particle size

Upload: bharat-bajaj

Post on 12-Oct-2015

80 views

Category:

Documents


0 download

DESCRIPTION

Fluid mechanics

TRANSCRIPT

  • This article was downloaded by: [Indian Institute of Technology - Delhi]On: 25 July 2013, At: 08:58Publisher: Taylor & FrancisInforma Ltd Registered in England and Wales Registered Number: 1072954 Registered office: MortimerHouse, 37-41 Mortimer Street, London W1T 3JH, UK

    Journal of Dispersion Science and TechnologyPublication details, including instructions for authors and subscription information:http://www.tandfonline.com/loi/ldis20

    Particle Size Distribution Measurement fromMillimeters to Nanometers and from Rods toPlateletsP. BowenPublished online: 05 Feb 2007.

    To cite this article: P. Bowen (2002) Particle Size Distribution Measurement from Millimeters to Nanometers and fromRods to Platelets, Journal of Dispersion Science and Technology, 23:5, 631-662, DOI: 10.1081/DIS-120015368

    To link to this article: http://dx.doi.org/10.1081/DIS-120015368

    PLEASE SCROLL DOWN FOR ARTICLE

    Taylor & Francis makes every effort to ensure the accuracy of all the information (the Content) containedin the publications on our platform. However, Taylor & Francis, our agents, and our licensors make norepresentations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose ofthe Content. Any opinions and views expressed in this publication are the opinions and views of the authors,and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be reliedupon and should be independently verified with primary sources of information. Taylor and Francis shallnot be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and otherliabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to orarising out of the use of the Content.

    This article may be used for research, teaching, and private study purposes. Any substantial or systematicreproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in anyform to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://www.tandfonline.com/page/terms-and-conditions

  • Particle Size Distribution Measurementfrom Millimeters to Nanometers andfrom Rods to Platelets

    P. Bowen

    Powder Technology Laboratory, Materials Department, Swiss FederalInstitute of Technology Lausanne (EPFL), 1015 Lausanne, SwitzerlandE-mail: [email protected]

    ABSTRACT

    There are many instruments for particle size distribution (PSD) measurement, each

    using a particular physical phenomenon to define the size (e.g., sedimentation, laser

    diffraction). Particle size distribution measurement although important, most people

    want it done as quickly as possible. This paper will compare rapid methods [ laser

    diffraction and photon correlation spectroscopy (PCS)] with sedimentation

    techniques and image analysis.

    Nanometer powders (primary particles 1050 nm) have been studied using,

    photocentrifuge, PCS, and x-ray disc centrifuge. The median diameters are very

    consistent for all instruments for narrow size distributions at around 20 nm but

    a divergence of results for the 50 nm range when distributions are broader.

    Comparison with image analysis for a spherical silica (50 nm) illustrates the

    accuracy possible in this domain.

    In the 0.1 to 5 micron range examples showing how the width and size range of

    the particle size distribution of typical commercial aluminas and calcites, can

    influence greatly the reported values from different instruments will be presented.

    The resolution and accuracy of PSD measurement for laser diffraction has been

    investigated using spherical glass beads ranging from 70 to 400 mm, sieved into fivefractions. Particle size distributions measured by image analysis are compared with

    the laser diffraction results.

    The shape also has a profound effect on the interpretation of data provided by

    commercial instruments which ordinarily assume a spherical shape. Results for

    plate and rod like particles will be presented and the most appropriate method for

    the shapes suggested.

    631DOI: 10.1081/DIS-120015368 0193-2691 (Print); 1532-2270 (Online)Copyright # 2002 by Marcel Dekker, Inc. www.dekker.com

    JOURNAL OF DISPERSION SCIENCE AND TECHNOLOGY

    Vol. 23, No. 5, pp. 631662, 2002

    2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

    MARCEL DEKKER, INC. 270 MADISON AVENUE NEW YORK, NY 10016

    Dow

    nloa

    ded

    by [I

    ndian

    Insti

    tute o

    f Tec

    hnolo

    gy -

    Delhi

    ] at 0

    8:58 2

    5 July

    2013

  • INTRODUCTION

    The aim of the current paper is not to be an exhaustive

    review of all methods but to share the experience gained

    in our laboratory over the past 1015 years which may

    then help the reader in their choice of particle sizing

    method for particles in the nm to mm range. Where other

    work in the literature is relevant it has been cited in an

    effort support our general findings and conclusions. We

    shall also try and point out in which domains we have

    found the methods used in our labs to be particularly well

    adapted and where one must take great care in simply

    interpreting the data provided by various commercial

    instruments. We shall not restrict ourselves to particles

    with spherical shapes but also approach the issue of

    anisotropic particlesrods and platelets in particular.

    We shall however restrict ourselves to inorganic particles,

    and not include comparisons based on latex spheres, as

    the world of organic based particulates bringing often

    their own particular problems or simplifications quite

    different to inorganic powders. There will be a short but

    concise first section on the general definitions of

    diameters and distributions which will allow people

    new to the field to follow the more detailed discussions

    without having to refer to basic texts for an introduction.

    Why measure particle size distributions? In industrial

    as well as everyday life materials are often found as

    powdersfrom coffee granules, washing powder, house-

    hold dust, drugs and cosmetics, to ceramic and metallic

    powders. The powder properties are often influenced by

    the particle size; raising questions such asdoes the

    powder flow or how quickly does it dissolve? (sugar is a

    good household example). In one of the best monographs

    on particle size measurement,[1] Allen mentions a survey

    made at Du Pont by Davies and Broughton on 3000

    products where they found that 80% involved a powder at

    some stage of the manufacture. This highlights the

    importance of powder technology in the modern indus-

    trial world. The study of powder technologya field in

    its own rightis heavily dependent on our capability to

    measure the powders particle size and its distribution.

    The particle size distribution of a powder is a key

    characteristic that influences its properties, handling and

    domain of application. Some examples are; the covering

    power and color quality of cosmetics and paints,[2] its

    packing behavior,[3] reactivity or sinterability of ceramic

    and metallic compacts.[46] If we take the example of

    advanced ceramic materials, in order to increase reacti-

    vity and minimize the final grain size, which consequently

    influences the properties of the final products,[7] the

    ceramic powders have more and more particles in the

    sub-micron range. The particle size and distributions will

    have an influence on which processing routes may be

    suitableslip casting or dry pressingand how the

    powder compacts during this process. On firing, the

    shrinkage, final density and properties are all influenced

    by the raw materials particle size and distribution. So,

    the characterization of powder size is clearly important

    but particle size and distribution (PSD) measurement on

    its own is not enough and the results must be coupled

    with other characterization techniques (such as

    microscopy, x-ray powder diffraction, surface area

    measurement, and chemical composition analysis) to

    correctly interpret and use the measured PSD. Often a

    PSD measurement is made with the aim of relating the

    PSD to a particular property or behavior of a powder and

    when choosing a method for the PSD measurement the

    application should always be borne in mind.

    Different methods for PSD measurement often have

    limitations and when these are ignored, correlation with

    a particular property of interest and the conclusions

    drawn can be erroneous. The aim of this overview is to

    familiarize the readers with some of the methods

    currently available along with their limitations and to

    what degree we can use these methods for absolute or

    comparative particle size measurement. Particle shape is

    one of the factors that often limits the use of certain

    instrumentsalthough comparison of measurements

    from instruments using different principles can also

    provide information on particle shape. After a brief

    introduction to some key definitions for particle

    diameters, shapes, and distributions, a brief overview of

    current methods will be given before moving onto more

    detailed discussions of some specific instruments and

    comparison between the instruments for particles from

    nanometers to millimeters.

    Diameters, Distributions and Shape

    All too often when one reads the powder character-

    ization section of scientific papers one finds the average

    particle size was 2 micronswith no reference to the

    method of measurement or distribution base. Figure 1

    shows an example of the same distribution plotted as a

    number or volume based distribution. The marked differ-

    ence shows how important the precise definition of the

    average particle size can become.

    632 Bowen

    2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

    MARCEL DEKKER, INC. 270 MADISON AVENUE NEW YORK, NY 10016

    Dow

    nloa

    ded

    by [I

    ndian

    Insti

    tute o

    f Tec

    hnolo

    gy -

    Delhi

    ] at 0

    8:58 2

    5 July

    2013

  • Particle Size

    There are many different ways of defining a particle

    sizefor a sphere we have an unambiguous definition

    of diameter and likewise the edge length of a cube.

    Most particles we come across in everyday work and

    life are unfortunately, rarely spherical or cubic and more

    than one dimension may be necessary to describe the

    particle. Table 1 lists a number of particle diameter

    definitions and some of these are illustrated in Fig. 2.

    The most widely used diameters are the equivalent

    spherical diameters and particularly the equivalent

    volume spherical diameter, dv, which is the diameter

    of the sphere which has the same volume as the

    particle. A cube of length 1 mm has an equivalentspherical diameter equal to 1.24 mm. Each measuringtechnique measures a certain diameter and for a sphe-

    rical shape these sometimes give the same result. For

    irregularly shaped particles there will however be an

    influence on the assigned size and this should always be

    taken into account when choosing the method of

    analysis (again bearing in mind the application of

    your PSD measurement). The sieve diameter for exam-

    ple is the minimum size that can pass through the

    square aperture and for irregularly shaped particles will

    not necessarily be the same as the Stokes diameter

    measured by a sedimentation technique. Irregular parti-

    cles can have the same equivalent spherical diameter

    but vastly different shapes. For regularly shaped parti-

    cles we can also have a distribution in shapes or aspect

    ratios (e.g., length to diameter ratio for a cylinder).[8]

    Microscopy is a very popular method for shape analysis

    as one can identify individual particles and quantify a

    shape coefficient. A microscopic examination of a

    powder should always be carried out even if the

    shape and size are not going to be analyzed quantita-

    tively by image analysis. An image of the particles

    contained in your powder will very quickly tell you if

    Figure 1. Example of the same PSD represented either as a

    number or volume distribution.

    Table 1

    Different Definitions of Particle Diameters[1]

    Diameter Symbol Definition

    Stokes diameter dst Diameter of free falling sphere which would fall at the same rate as the particle

    in a given fluid

    Sieve diameter dT Minimum square aperture through which the particle will pass

    Volume diameter dv Diameter of the sphere with the same volume as the particle

    Surface diameter ds Diameter of the sphere that has the same surface area as the particle

    Projected area diameter dA Diameter of the circle which has the same area as the projected area of the particle

    Ferets diameter dF Distance between two parallel tangents which touch the outline of the particle

    projection

    Average Feret diameter dFav Average Feret diameter from diameters measured over all angles between 0

    and 180

    Maximum Feret diameter dFmax Maximum distance between two parallel tangents which touch the outline

    of the particle projection

    Minimum Feret diameter dFmin Minimum distance between two parallel tangents which touch the outline

    of the particle projection

    PSD Measurement from Millimeters to Nanometers 633

    2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

    MARCEL DEKKER, INC. 270 MADISON AVENUE NEW YORK, NY 10016

    Dow

    nloa

    ded

    by [I

    ndian

    Insti

    tute o

    f Tec

    hnolo

    gy -

    Delhi

    ] at 0

    8:58 2

    5 July

    2013

  • the powder is homogeneous in shape e.g., disc-like or

    rod-like or spherical. Irregularly shaped particles such

    as the ground calcite shown in Fig. 2(b) is typical of a

    ground ceramic powder and is often encountered in the

    ceramic industry. With the improvement in the power of

    computers over the last 10 years, image analysis

    programs have become very powerful and quantitative

    data can be obtained from microscopic observations.[8]

    The limitations often being linked with sample

    preparationoverlapping particlesand the 2D repre-

    sentation of our 3D particle.

    Shape Factors

    There are many ways of describing shape depending

    on the regularity of the shape itself.[8] Here we shall

    describe two simple concepts, a particles sphericity and

    an aspect ratio. For anisotropic particles with a relatively

    regular morphology such as a rod or plate one can define

    the aspect ratio as the ratio of the major to the minor axis.

    For irregular particles, where a simple definition of major

    and minor axes is not so clear the ratio between the

    Maximum and Minimum Feret diameters (dFmax=dFmin)gives a very good indication of the elongation of a

    particle.

    Aspect ratio Major axisMinor axis

    ordFmax

    dFmin1

    The shape most often assumed in particle size analysis is

    the sphere and one can define how close a particle

    approaches a sphere by the use of a sphericity shape

    factor. This can be defined as

    cw Surface area of a sphere having the

    same volume as the particle

    Surface area of the particle2

    Equivalent Spherical Diameter

    Most commercial instruments assume a spherical

    shape for data analysis and the result is given as an

    equivalent spherical diameter (ESD). The ESD is the

    diameter of the sphere that would give the same result as

    that observed for the real powder under observation. The

    various physical phenomena used in particle size

    measurement often show functional variations with

    particle shape. Therefore, the ESD measured by sedi-

    mentation will not be equivalent to that measured by laser

    diffraction for non-spherical particles. The ESD which is

    reported by most automatic PSD instruments currently

    available will also vary as the aspect ratio of the particles

    vary. Jennings and Parslow[9] have computed the varia-

    tion of the ESD with the aspect ratio for oblate and

    prolate spheroids for the different physical phenomena

    used in several commercially available instruments. They

    showed from the theoretical calculations that the ESD is

    always smaller than the real major dimension of the

    anisodiametric particle. There have been many attempts

    to determine shape distributions mainly by comparison of

    different techniques and when judiciously chosen an

    Figure 2. (a) Examples of some diameters for a non-spherical particle (b) typical irregular particlea ground calcite.

    634 Bowen

    2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

    MARCEL DEKKER, INC. 270 MADISON AVENUE NEW YORK, NY 10016

    Dow

    nloa

    ded

    by [I

    ndian

    Insti

    tute o

    f Tec

    hnolo

    gy -

    Delhi

    ] at 0

    8:58 2

    5 July

    2013

  • estimation of the aspect ratio and the particle minor

    dimension can be made.[10,11]

    Distributions and Average Diameters

    The powders we deal with everyday are rarely single-

    sized or monodispersed and are part of a distribution of

    sizes which occur with different frequencies and we have

    to choose how to represent this distribution in a concise

    and accurate manner. The distribution can be represented

    as either a cumulative distribution [Fig. 3(a)] or as a

    density or frequency distribution [Fig. 3(b)]each

    having its use depending on the information we want

    from a graphical representation. The base to which the

    distribution is normalized is of great importance and

    certain measurement techniques give us directly a

    number (microscopy) or weight (sieving) based distribu-

    tion. The transformation between different distributions

    is not always easy and care should be taken to verify with

    what accuracy this can in fact be done.[1]

    The average of a distribution is a measure of the central

    tendency which is not greatly affected by a few extreme

    values in the tails of the distributions. There are several

    central tendencies which can represent a distribution and

    some examples of these, the mode, median, and mean are

    shown in Fig. 3(b). The mode is the value which occurs

    most frequently in the distribution and will be the peak in

    a frequency curve. The median is the middle value of a

    distribution where the total frequency of values above and

    below it are equal (i.e., the same total number or total

    volume of particles below the median as above it)from

    a cumulative distribution curve it can be read directly

    [Fig. 3(a)]. The mean value has to be calculated and is the

    point about which the moments of the distribution are

    equalseveral examples of different mean diameters are

    listed in Table 2. For a symmetrical distribution, such as a

    normal distribution, the mode, mean, and median coin-

    cide but for skewed distributions they differ [Fig. 3(b)].

    The manner in which a continuous size distribution is

    sampled to provide a frequency distribution has an

    important consequences on the accuracy and form of

    the frequency distribution.[1] For broad size distributions

    a geometric progression should be used e.g., x,2

    px, 2x,

    2p

    2 x, where x is the smallest size interval, du. For

    narrower size distributions arithmetic progressions can be

    used. The underlying rule is that the resolution of each

    class interval should be similar, where the resolution is

    defined as the width of the interval divided by the mean

    class size. The cumulative distribution, FN(x), for a

    number based distribution, is defined in Eq. (3) and can

    be represented unambiguously on linear and logarithmic

    axes for the particle size.

    FN x numer of particles < du

    Total number of particles3

    where du is the upper particle size of the size interval.

    The frequency or density distribution, fN, is defined as

    the derivative of the cumulative distribution Eq. (4),

    which usually discretized into different size classes, di,

    as discussed above. The frequency

    fN dFNddN

    FN d2 FN d1d2 d1

    DFNd2 d1

    4Figure 3. Examples of (a) a cumulative distribution and(b) a frequency distribution.

    PSD Measurement from Millimeters to Nanometers 635

    2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

    MARCEL DEKKER, INC. 270 MADISON AVENUE NEW YORK, NY 10016

    Dow

    nloa

    ded

    by [I

    ndian

    Insti

    tute o

    f Tec

    hnolo

    gy -

    Delhi

    ] at 0

    8:58 2

    5 July

    2013

  • distribution however can give ambiguous representations

    when using a log scale and for comparisons should be

    plotted as size (linear scale) vs. the fraction of particles in

    the interval, DFN , as recently re-iterated by Sommer.[12]

    Many instrument manufacturers make the mistake of

    providing frequency distributions as relative % frequen-

    cies which can be very misleading when comparing

    distributions which do not have the same size classes.

    The easiest way to correctly represent and compare PSDs

    is using the cumulative representation.

    The average of a distribution indicates the central

    tendency; the dispersion or width of the distribution is

    described by its standard deviation, s the root meansquared deviation Eq. (5). Each standard deviation is

    calculated with respect to the central tendency used in the

    formula of Eq. (5); in this example, dw ,the mean weight

    diameter.

    sw fidi dw2

    W

    s5

    where, fi is the frequency particles (as a weight) of that

    diameter and W total weight for all of the diameter

    intervals.

    Distributions can often be represented by mathematical

    expressions which describe the whole of the distribution

    with the central tendency and the standard deviation.

    For a detailed discussion of the normal, log-normal, and

    RosinRammler distributions in particle size measure-

    ment the reader is referred elsewhere.[1] If the particle size

    distribution under investigation does follow a type of

    distribution, comparison and transformation from number

    to volume distributions are more easily calculated.

    Having seen the sometimes frightening number of

    possible particle sizes and distributions the question of

    which diameter should one use is often asked. As a general

    rule, most powder suppliers quote an equivalent sphere

    median volume diameter. When dealing with powder

    packing the volume distribution is a convenient base.

    When we are dealing with agglomerating systems or

    milling processes where population balances are often

    used, the number distribution is usual. As mentioned in

    the introduction the PSD measurement is usually made

    with the aim of relating it to a particular property, either of

    the powder, process or product. In this case the best base

    must be chosen as a function of the correlation sought after.

    For example, if you are dealing with a process sensitive to

    fine particles, such as the flowability of flour the number

    distribution is more sensitive. If it is the strength of

    sintered ceramics then it is a small agglomerate population

    that determines the critical flaw size and a volume based

    distribution will be the most adequate.

    CURRENT METHODSA BRIEF

    OVERVIEW

    There are a vast number of methods and suppliers of

    equipment for particle size analysis and a few are listed

    in Table 3. Other types of machines and analysis ranges

    can be found in the literature.[13,14] Many groups have

    made evaluation studies on the various commercial

    instruments available,[1,1519] only a very brief descrip-

    tion of some popular methods will be presented here.

    Microscopy

    Microscopy has obvious advantages in that you actu-

    ally get information on the morphology of the particles at

    the same time as a size distribution. The problems arise

    when you consider the sampling and statistics. Sample

    preparation is very important and overlapping particles

    and sampling errors even on well prepared slides (e.g.

    size segregation) are often problematic. These limitations

    can be minimized by analyzing the whole area prepared

    and not just a central area, as well as numerous slides to

    Table 2

    Some Examples of Different Mean Diametersand Their Mathematical Representation[1]

    Diameter Definition

    Number-length dnl ePn

    i1 diNiPni1 Ni

    Number-surface dns Pn

    i1 di2NiPn

    i1 Ni

    2

    s

    Number-volume dnv Pn

    i1 di3NiPn

    i1 Ni

    3

    s

    Length-surface dls Pn

    i1 di2NiPn

    i1 diNi

    Surface-volume dsv Pn

    i1 di3NiPn

    i1 di2Ni

    Volume-moment

    (weight-moment)

    dv Pn

    i1 di4NiPn

    i1 di3Ni

    Specific surface area dBET 6SBETr (mm)

    636 Bowen

    2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

    MARCEL DEKKER, INC. 270 MADISON AVENUE NEW YORK, NY 10016

    Dow

    nloa

    ded

    by [I

    ndian

    Insti

    tute o

    f Tec

    hnolo

    gy -

    Delhi

    ] at 0

    8:58 2

    5 July

    2013

  • minimize variations between specimens. With the possi-

    bility of automating some of the analysis with the use of

    image analysis packages the number of particles counted

    can be put into the thousands and the statistics become

    quite acceptable [for a sampling error of

  • using an analytical ultracentrifuge (AUC).[20] The main

    assumption made here is the spherical shape of the particle

    in the Stokes law calculation. The detection method can

    also lead to limitations. Some instruments use light absor-

    bance to detect the sedimentation of the particles and when

    the size is similar to that of the wavelength of the light used,

    scattering phenomena become important. The use of x-rays

    alleviates this problem but one must have a material that

    adsorbs enough x-rays at a dilution which avoids hindered

    settling problems. So for materials with a low x-ray

    absorption such as, organic polymers the use of x-rays

    are not an ideal case. By using multiple speeds, centrifuga-

    tion can access the whole colloidal range, and with the

    AUC even in one experiment.

    Single Particle Counters

    Single particle counters such as the electrical sensing

    zone technique,[1,15] is a well established technique and

    works well down to about 0.5 mm. It is one of the fewtechniques that measures a mass and number distribution

    directly. The particles are suspended in an electrolyte and

    pass through an aperture between two electrodes. The

    voltage pulse produced as the particles pass through this

    aperture is proportional to the volume of the particle. One

    of the draw backs with this method is that different sized

    apertures must be used to cover the whole range of sizes

    (albeit a very large size range 0.5 to 1000 mm). Care mustalso be taken to count enough particles in the largest size

    category. The use of a mass balance is imperative to

    check what quantity of material is outside the size range

    of the aperture in use and how accurate a calibration has

    been carried out. Also, care must be taken to ensure the

    nominal values quoted for the sampling volumes are

    correct.[15] More recently optical detectors have been

    introduced where the passage of particles in a photo-

    sensitive zone give rise to pulses which are proportional

    to the particle diameter.

    A method often used for dry powders, of particular

    interest to the pharmaceutical industry is the aerosol time

    of flight particle sizer which is another variation of a

    single particle counter. Particles are fluidized and pass

    through a detection zone via a nozzle, carried by

    compressed air which accelerates the particles. The

    time of flight between two laser beams is measured and

    the size calculated. When using spherical particles the

    results are directly comparable with other methods[21] but

    care must be taken when dealing with bi-modal or

    very broad size distributions.[17] For non-spherical or

    irregularly shaped powders, the difficulty in relating the

    aerodynamic diameter to other equivalent spherical

    diameters has limited the use of such instruments. Irre-

    gularly shaped particles can however align in the aerosol

    flow conditions and measured PSDs therefore depend on

    particle shapecomparison with other methods can lead

    to information on the shape of the particle.[21]

    Gas Permeability

    The measurement of particle size using gas perme-

    ability is frequently used for metallic powders and is

    often quoted FSSSFischer sub-sieve sizerthe type of

    instrument used. The particle size is deduced from a

    measurement of the permeability of a gas through a

    compact of the powder under investigation.[1] This

    method does not give a distribution but the surface-

    weight mean diameter and is not very suitable for sub-

    micron powders where high resistance to the flow occurs

    and the theoretical approach used is no longer valid.

    Light Scattering

    Light scattering particle size analyzers fall into two

    main categories-the low angle laser light scattering

    (LALLS) or laser diffraction group which measure from

    3500 to around 0.05 mm and the dynamic light scattering(DLS) which measures from about 500 nm to around 2 nm.

    Laser diffraction analyzers can be used with suspen-

    sions and dry powders, where more recently, better

    sampling systems have helped improve the dry powder

    measurement.[17] Both diffraction and light scattering

    theory are needed to cover the range of particles sizes

    claimed by these instruments as discussed in more detail

    below. The main limitations arise when non-spherical

    particles need to be analyzed and when the optical proper-

    ties necessary for the models are not available. These

    methods are very quick and the results reproducible. More

    recent instruments have added supplementary detector

    arrays, light polarization and or different wavelengths to

    try and access with more accuracy, the sub-100 nm range.

    The dynamic light scattering (DLS) method is based on

    the light scattered by particles (colloidal) which are under-

    going random thermal motions. The light scattered fluc-

    tuates with time and can be related to the particle diffusion

    coefficient in a given suspending medium and conse-

    quently a particle diameter. For sub 200 nm narrow size

    distribution particles, DLS is a very good, quick, and

    accurate technique. The main problems arise once the

    638 Bowen

    2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

    MARCEL DEKKER, INC. 270 MADISON AVENUE NEW YORK, NY 10016

    Dow

    nloa

    ded

    by [I

    ndian

    Insti

    tute o

    f Tec

    hnolo

    gy -

    Delhi

    ] at 0

    8:58 2

    5 July

    2013

  • powder under investigation does not have a narrow or

    single mode distributionthe method gives an idea of the

    dispersion or polydispersity of the distribution but does not

    give a particle size distribution only a mean diameter and a

    polydispersity. The technique will be discussed in more

    detail below. Borkovec and Sticher have shown the utility

    of using both techniques in seriesLALLS and DLSto

    cover the large size range often found in natural materials

    such as soils.[22]

    Gas Adsorption

    Nitrogen adsorption can be used to measure the

    specific surface area of a powder. If the powder is

    assumed to be monodispersed and spherical, an average

    particle diameter having the same surface in m2=g as thepowder under investigation can be calculated. This is

    normally written as dBET after the adsorption model used

    to describe the nitrogen adsorption isotherm (Brunauer,

    Emmett, and Teller).[23] When compared with the median

    diameter, dv50, from other methods the dv50=dBET ratiocan be defined as the agglomeration factor FAG indicative

    of the state of agglomeration of the powder under

    investigation.[24]

    For all the methods described, particle size standards

    can be used to get a more reliable quantitative number from

    the analysis but, beware that your material to be analyzed is

    similar to that of your standard used to calibrate the

    instrument otherwise very erroneous results can ensue.

    SPECIFIC METHODS

    In this section we shall describe in more detail several

    different methods used for PSD measurement and the

    commercial instruments used in our laboratory to cover

    the nm to mm range. We shall describe briefly their

    physical basis and experimental set up. Then we shall

    look at their performance and highlight certain limita-

    tions pertaining to the particular method before moving

    on to the final section where we shall compare results for

    a series of different powders. The instruments used were,

    laser diffraction in general a Malvern Mastersizer S

    (MMS) [and some preliminary results with Micromeritics

    Saturn DigiSizer (MSD)], cuvette photocentrifuge a

    Horiba CAPA-700 (PC), Brookhaven X-ray Disc Centri-

    fuge BI-XDC (XDC), and for photon correlation spectro-

    scopy a Malvern ZetaSizer 4 (MZS4). The dispersion

    concentrations used for PSD measurement were around

    0.1 mg=mL for DLS, 0.5 mg=mL for PC, MMS and

    MSD, and 100 mg=mL for the XDC. The percentagevolume of solids for the XDC was less than the 2% at

    which hindered settling is expected to occur.[25] All

    samples underwent an ultrasonic treatment (150 W,

    10 min) prior to measurement.

    SedimentationCentrifugation

    Cuvette Photocentrifuge

    A typical cuvette photocentrifuge used in our labora-

    tory is the Horiba CAPA-700. It uses the principle of

    liquid-phase sedimentation and light absorption to

    measure a particle size distribution. A homogeneous

    suspension of particles is prepared and as the particles

    settle either by gravitation or centrifugation, the change

    in concentration of the suspension is monitored by the

    change in intensity of a light beam (Fig. 4). The sedi-

    mentation velocity is calculated using Stokes law Eq. (6)

    for sedimentation by gravity and Eq. (7) for centrifugal

    sedimentation.

    d

    18Z0Hr r0g

    s6

    d 18Z0 ln x1=x2 r r0o2t

    s7

    where d is the particle diameter, Z0 the viscosity of thedispersion medium, H the sedimentation distance, r thesample density, r0 the dispersing liquid density, t thesedimentation time, x1 the distance between the center of

    rotation and the sedimentation plane, x2 the distance

    between the center of rotation and the measuring plane,

    g the acceleration due to gravity, and o the rotationalangular velocity. The change in the optical transmission

    during a certain time interval corresponds to a particular

    size range. The relationship between the change in light

    intensity, absorbance, the size and number of particles,

    from the Beer-Lamberts law, is given by Eq. (8),

    log I0 log Ii AXni1

    KdiNidi2 8

    where I0 is the intensity of the initial light beam, Ii is the

    light intensity after passing through the sample, A is an

    optical coefficient for the cell, di is the particle diameter,

    Ni the number of particles, and K(di) the light extinction

    efficiency or coefficient which is a complex function of

    PSD Measurement from Millimeters to Nanometers 639

    2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

    MARCEL DEKKER, INC. 270 MADISON AVENUE NEW YORK, NY 10016

    Dow

    nloa

    ded

    by [I

    ndian

    Insti

    tute o

    f Tec

    hnolo

    gy -

    Delhi

    ] at 0

    8:58 2

    5 July

    2013

  • size and the relative refractive index of the suspending

    medium and particle.

    When the particle diameter is comparable to the light

    wavelength, light scattering becomes important and a

    correction must be made using the Mie theory.[26] One of

    the problems encountered in our laboratory when trying

    to use the light scattering correction mode with the

    Horiba CAPA-700 was the lack of resolution allowed

    for in the light scattering extinction coefficients in the

    small microprocessor which controls the apparatus.[27,28]

    The fitting of a function which oscillates and varies

    steeply with particle size using a polynomial function

    was very unsatisfactory. Consequently a computer

    program was written which takes the raw absorbance

    data and corrects for light scattering using an extinction

    coefficient calculated using a modified Mie theory.[27,28]

    An example of the function K(di) for a calcite powder

    with spherical particle morphology is shown in Fig. 5.

    For a 50=50 wt.% mixture of two spherical silicapowders with nominal diameters of 0.25 mm and 0.6 mmthe differences with and without the extinction coefficient

    correction is striking, as shown in Fig. 6. The small

    Figure 4. Schematic representation of cuvette photocentrifuge (a) sample cell and detection geometry and (b) typical absorption

    curves, illustrating the need for two plateaus as a good data criterion as discussed later in the text.

    Figure 5. Calculated extinction coefficient K(di) vs. diameter

    for a sphere (calcite in water).

    Figure 6. Particle size distributions for a 50=50 wt.%. sphe-rical silica powder bi-modal mixture with and without the light

    scattering correction.

    640 Bowen

    2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

    MARCEL DEKKER, INC. 270 MADISON AVENUE NEW YORK, NY 10016

    Dow

    nloa

    ded

    by [I

    ndian

    Insti

    tute o

    f Tec

    hnolo

    gy -

    Delhi

    ] at 0

    8:58 2

    5 July

    2013

  • particle fraction is grossly underestimated if the correc-

    tion is not used, indicating that our 50=50 mixture was infact 70=30 wt.% mixture of the 0.6=0.25 mm powders.The data corrected for light scattering indicates a

    54.4=45.6 mixture. These results show the improvementin accuracy using the light scattering correction for this

    ideal case of a mixture of two powders both with narrow

    size distributions and spherical shape [Fig. 7(a)].

    The PSD results for a commercial alumina powder

    [Fig. 7(b)] are shown in Fig. 8. A large discrepancy is seen

    between the uncorrected Horiba CAPA-700 data and the

    suppliers x-ray sedimentometer data. The corrected data

    shows a very good correspondence with the suppliers x-

    ray sedimentometer data and gives a more complete PSD

    analysis below the 0.2 mm size range. An important aspectof this result is the fact that the Horiba CAPA-700 data

    had to be collected in two size ranges to give the data

    quality necessary to apply the light scattering correction.

    If the good data criterion is not satisfied, i.e., a plateau

    of absorption at beginning and zero absorption at end

    [Fig. 4(b)],[28] then the light scattering correction on poor

    data [Fig. 4(b)] gives an erroneously fine distribution, as

    for example, in one single measurement on the alumina

    powder shown in Fig. 9. This is because the size of last

    data point in the incomplete data is underestimated at 0.01

    because of the simple averaging method used by the

    Figure 7. Powders used to illustrate importance of the light scattering correction with the cuvette photocentrifuge (a) TEM of

    0.25 mm model silica spheres (b) Commercial alumina powder BY2.

    Figure 8. Particle size distributions for a commercial alumina

    powder showing the suppliers data, the Horiba photocentrifuge

    uncorrected data and the corrected data (light scattering and

    radial dilution).

    Figure 9. Example of the a commercial alumina powder

    where the data collected did not fit the Good Data Criterion

    for light scattering correction to be correctly applied when

    using a photocentrifuge.

    PSD Measurement from Millimeters to Nanometers 641

    2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

    MARCEL DEKKER, INC. 270 MADISON AVENUE NEW YORK, NY 10016

    Dow

    nloa

    ded

    by [I

    ndian

    Insti

    tute o

    f Tec

    hnolo

    gy -

    Delhi

    ] at 0

    8:58 2

    5 July

    2013

  • software.[28] The percentage of particles smaller than

    0.1 mm increases from 0.4% for the uncorrected data to22% for the corrected data (and is 11% for the data which

    fits the good data criterion Fig. 9). This inconvenience, a

    straight forward lack of versatility of data collection with

    the Horiba CAPA-700, undermines the otherwise quick

    and reliable data collection. For finer powders in the sub

    100 nm range the light scattering correction and good data

    criterion becomes even more critical as for a 40 nm

    diameter alumina particle the extinction coefficient is

    only 2.75105. For the gamma alumina powdersdiscussed later in the comparison of instruments section;

    diameters can be overestimated by up to a factor of ten.[29]

    The upper range of the Horiba is 300 mm and althoughLALLS is better suited to this size range, even with dense

    powders such as atomized steel powders reproducible and

    reliable results have been collected up to 150 mm.The PSD data for a European Community standard

    quartz powder, BCR 66, is shown in Fig. 10. When

    Horiba CAPA-700 data is corrected for light scattering a

    much finer PSD than the standard data is seen. The two

    major assumptions made in the light scattering correction

    are the quartz refractive index and the spherical shape of

    the quartz particles. The real part of the refractive index

    was verified using the liquid immersion method to be

    1.55 0.015 and the imaginary part was estimated tobe much less than 0.1.[28] The particle shape observed by

    scanning electron microscopy is very irregular and

    platelet-like, rod-like and equiaxed particles were

    observed [Fig. 11(a)]. If the extinction coefficients for

    an infinite cylinder are used instead of those for a sphere,

    the quartz PSD is similar to the standard data distribution

    (Fig. 10). This result suggests that for such very irregu-

    larly shaped particles the simple spherical particle light

    scattering correction is not applicable and the shape has

    to be taken into account. In the case of the standard

    quartz powder studied, the mixture of morphologies is

    too complex to attempt a correction and the use of the

    infinite cylinder extinction coefficients are used only to

    illustrate the significant effects that shape can have on the

    PSD when correcting for light scattering. In Fig. 10 the

    standard data and Horiba uncorrected data show a very

    good correspondence suggesting an averaging out of

    the effects of light scattering. Some striking examples of

    such averaging out of some of the fluctuations in the

    light scattering coefficients for irregularly shaped quartz

    particles can be found in the monograph of Bohren and

    Huffman[26] again supporting the above interpretation

    with respect to particle shape and the light scattering

    correction.

    Figure 10. Particle size distributions for the BCR 66 quartz

    powder showing the standard data, the Horiba uncorrected data

    and the corrected data.

    Figure 11. SEM micrograph of irregular and anisotropic powers (a) the BCR 66 standard ground quartz powder illustrating the

    irregular shape of the particles (b) glass fibers of constant diameter (c) copper oxalate rods of constant aspect ratio.

    642 Bowen

    2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

    MARCEL DEKKER, INC. 270 MADISON AVENUE NEW YORK, NY 10016

    Dow

    nloa

    ded

    by [I

    ndian

    Insti

    tute o

    f Tec

    hnolo

    gy -

    Delhi

    ] at 0

    8:58 2

    5 July

    2013

  • In light of this sensitivity to shape and the importance

    of anisotropic particles in practice further work has been

    carried out on the PSD analysis of anisotropic particles

    using the photocentrifuge.[3032] For cylindrical particles,

    model glass fibers and precipitated copper oxalate

    rods,[31,32] [Fig. 11(b) and (c)] the CAPA-700 has been

    found to give a very good correlation with image analysis

    as illustrated in Fig. 12. To compare the equivalent Stokes

    spherical diameter that the photocentrifuge provides, with

    the image analysis data the anisotropic particle must have

    some regularity in its shape distribution.

    One possibility is when the diameter of the rods is

    constant [glass fibers, Fig. 11(b)] then a fiber length (or

    spherical volume) can be calculated by equating the

    spherical and cylindrical volumes. The fiber lengths are

    consequently well predicted from the photocentifuge

    ESD.[31] This approach assumes that the drag coeffi-

    cients for the sphere and cylinder are the same and until

    now no satisfactory hydrodynamic explanation as to

    why this is the case has been found. This is discussed in

    more detail elsewhere[31] and further investigations are

    in progress.

    A second approach is if the particles have a reasonably

    constant aspect ratio then an equivalent spherical volume

    can be computed from the cylinder volume for

    comparison purposes [Fig. 11(c) copper oxalate]. A

    very good correlation was also found[32] this time for

    cylinder lengths in the micron range rather than the

    600 mm of the glass fiber example discussed above. If apowder does not fit either of the above cases (constant

    diameter, length or aspect ratio) then correlation between

    image analysis datacylinder length and the Stokes

    diameter cannot be made. For platelet like particles no

    correlation was found[31] even when light scattering was

    taken into account for thin (0.5 mm) platelets.The main conclusion to draw from this section is that,

    in order to make accurate sub-micron particle size

    measurements using a cuvet photocentrifuge by correct-

    ing for light scattering, care must be taken to characterise

    the powder under observation with respect to particle

    shape. Particular care must also be taken with respect to

    data quality and light scattering characteristics such as

    the effects that impurities can have on the imaginary part

    (absorption) of the refractive index. The cuvette photo-

    centrifuge has been shown to give useful information on

    cylindrically shaped particles, if there is some constant

    factor in the geometry. If the particle shape is not

    homogeneous or, cannot be approximated by a simple

    geometric shape such as a sphere or cylinder then an

    alternative method for accurate particle size measurement

    should be sought. For such cases the x-ray centrifuge

    could be a suitable alternative where the mass of the

    sample in the x-ray beam is measured directly and no

    scattering phenomena are encountered as discussed in the

    next paragraph. One advantage of the cuvette photocen-

    trifuge is measurements can be made with just a few mg

    of samplex-ray sedimentometers often need up to 100

    times more material.

    X-ray Sedimentation

    X-ray sedimentometers have been used for many years

    and are particularly well adapted for inorganic powder

    where the method of particle detectionabsorption of

    x-rays is sufficiently high to allow detection. The most

    celebrated instrument is the Micromeretics

    SedigraphTM[33]which uses gravitation to separate

    particles and measures an undersize mass PSD using

    Stokes law and the x-ray absorption. For particles finer

    than around 0.25 to 0.5 mm the gravitational force is notsufficient and is about the lower limit for the instrument.

    Earlier instruments were also relatively slow for sub-

    micron particle size measurementalthough by introdu-

    cing a relative movement between sample cell and

    detector this has been significantly improved in recent

    years. The main advantage of such systems is that x-ray

    absorbance is simpler than the use of light for particle

    detection as seen above for the photocentrifuge,

    where light extinction corrections can become veryFigure 12. Comparison of equivalent spherical diameter for

    glass fibers from image analysis and photocentrifuge.

    PSD Measurement from Millimeters to Nanometers 643

    2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

    MARCEL DEKKER, INC. 270 MADISON AVENUE NEW YORK, NY 10016

    Dow

    nloa

    ded

    by [I

    ndian

    Insti

    tute o

    f Tec

    hnolo

    gy -

    Delhi

    ] at 0

    8:58 2

    5 July

    2013

  • complicated. The main limitation of this instrument is

    that one must have a material that absorbs enough x-rays

    at a concentration which avoids hindered settling

    problems (less than 2% volume). So for organic polymers

    x-rays are not an ideal case and for inorganic materials a

    couple of grams of material is sometimes needed whereas

    with the light based systems often 5 to 50 mg is enough

    for several measurements. The advantage with the couple

    of grams quantity is that it is normally a representative

    sample of the distribution. Whereas with the 2050 mg

    care must be taken and repetitions performed to ensure a

    representative result especially when the powder has a

    wide distribution of sizes. This can be achieved by using

    sampling procedures such as spin-riffling to generate

    sub-samples which are completely consumed.[1] The

    gravitation based sedimentometer is widely used and

    very reliable but to speed up analysis and extend the

    finer end of the range an x-ray disc centrifuge has

    become available.

    The x-ray sedimentometer used in our laboratory is

    the Brookhaven X-ray Disc Cenftrifuge (XDC). This

    instrument was designed to analyze inorganic powders

    quickly, accurately, and reproducibly using centrifugation

    and x-rays for the particle detection. The sample cell is a

    hollow x-ray transparent disc, into which a suspension

    is injected and then spun to centrifuge the

    particles (Fig. 13). The principle is then as described

    above for the photocentrifuge using Eqs. (2) and (3) to

    follow the sedimentation of the particles and the change

    in x-ray intensity. The spherical assumption is again a

    limitation for the hydrodynamic aspects of sedimentation

    but no scattering phenomena are encountered within the

    range of 100 to 0.05 (0.01) mm. Results from Allenswork on the BCR 66 quartz powder (discussed in the

    previous section) show very good agreement with the

    standard data.[34]

    To illustrate the capabilities of the XDC for fine

    particle analysis a commercial boehmite was dispersed

    in nitric acid and its PSD measured assuming its theore-

    tical density of 3.01 g=cm3. The specific surface area ofthis powder was measured to be 214 m2=g which gives adBET of 9.3 nm (assuming spherical primary particles).

    The XDC gave a dv50 of 16 nm suggesting a certain

    amount of agglomeration which is probable when look-

    ing at the TEM photographs [Fig. 15(a)]. The micrograph

    also shows the primary particles to be somewhat elon-

    gated which can also modify the dBET although for

    platelets often attributed to boehmite particles it would

    decrease, increasing the apparent degree of agglomera-

    tion. The data was collected in just under 2 hours using

    Brookhavens recently upgraded 10,000 rpm disc and is

    presented in Fig. 14. Comparison with other instruments

    Figure 13. Schematic representation of an x-ray disc centrifuge and cuvette photocentrifugeillustrating the different geometries.

    Figure 14. Illustration of fine particle analysis

  • gave very consistent results [photon correlation spectro-

    scopy (PCS) and photocentrifuge] and is discussed in

    more detail elsewhere[29] and later in the final section of

    this paper.

    To what degree the median diameter of 16 nm is

    accurate is difficult to assess with these boehmite particles

    because of their non-sphericity and difficulty in preparing

    TEM specimens without touching particles [Fig. 15(a)].

    Consequently a spherical silica was used which simplifies

    the image analysis considerably. When using the standard

    density of 2.2 for a precipitated silica a large discrepancy

    between the XDC and image analysis was observed(Fig.

    14). The silica powders were supplied as a dispersion and

    diluted in 0.0025 M HCl for the analysis. A sample of this

    dispersion was freeze dried and the porosity measured

    using nitrogen adsorptiondesorption. If the density of the

    silica is recalculated taking into account this porosity and

    the thickness of the electrical double layer (assumed to

    move with the particle) then a figure of 1.52 is found for

    this effective hydrodynamic density. Re-analyzing the

    XDC data with this shows a very good correspondence

    between the XDC and image analysis data.[26] These

    results imply that in order to have accurate PSD data in

    this

  • alumina at both 2000 rpm and 600 rpm and the data

    renormalized or merged as the manufacturer describes

    it. This was carried out for four pairs of data and four of

    the five percentiles have standard deviations less than 5%.

    One of the disadvantages with the CAPA-700 photo-

    centrifuge was its limited run time capacity which led to

    collection of data in two or three separate runs to cover

    broad size distributions. The XDC does not have this

    limitation, but if we are dealing with a relatively broad

    size distribution (e.g., 5 mm to 50 nm) then in order tocollect sufficient data points at the large size end of the

    distribution excessively long runs, >8 hours, have to beperformed. One problem with running samples for

    8 hours apart from the analysis time is that there is an

    evaporation of the dispersion liquid which renders base-

    line identification difficult. To remedy this the XDC

    already has the tool for merging data collected at

    different speeds to allow more rapid data collection for

    such samples. An example for the commercial alumina

    studied above for runs at 600, 2000, and 3000 rpm

    in Fig. 17 shows the type of data that is typically

    produced. The median diameters are really in very

    good agreement but the tails differ greatlythe high

    particle size being poorly sampled at the higher speeds

    only four and three points before 0.5 mm for the runs at2000 and 3000 rpm respectively. The overlap with the

    600 rpm run just below 0.3 mm is really very goodindeed. In fact the 600 rpm run only collects data to

    0.23 mm with 38% of the distribution being finer thanthis value. The 2000 rpm data finds a 0% value at

    0.12 mm. The data plotted in Fig. 16 for the dv5 anddv15 of runs at 600 rpm is the result of an extrapolation

    and both show values lower than the 0% value found

    for the 2000 rpm run, showing the extrapolation to be in

    significant error. The raw data is available to check on

    this possible over extraction of data and it should

    always be consulted to avoid gross errors. This over

    extraction of data has been corrected in more recent

    software. The overlap of two measurements at different

    speeds is not always as smooth as found in this example

    Figure 16. Repeated PSD analysis with the XDCdifferent sample preparations showing good reproducibility from run to run.

    Figure 17. Example of same alumina dispersion run at

    different speeds using the XDC.

    646 Bowen

    2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

    MARCEL DEKKER, INC. 270 MADISON AVENUE NEW YORK, NY 10016

    Dow

    nloa

    ded

    by [I

    ndian

    Insti

    tute o

    f Tec

    hnolo

    gy -

    Delhi

    ] at 0

    8:58 2

    5 July

    2013

  • and care must be taken with respect to the cross over

    point used.

    In conclusion the XDC has very good fine particle

    capabilities (4 mm theoptical properties of the material and medium do not play

    an important role and the resulting diffraction pattern can

    be described by diffraction theory.[35] For a powder

    suspended in air or a liquid this produces a series of

    rings, the position and intensity of these diffraction rings

    being dependent on the particle size Fig. 18. When the

    particles are

  • shape. If the particle size distribution to be analyzed is

    known to follow a particular distribution e.g., log-normal

    or Rosin-Rammler the initial guess can use this informa-

    tion to simplify the data analysis. There are a vast number

    of LALLS instrument manufacturers and each has their

    preferred algorithm for data reduction which can lead

    to differences of up to 20% when using the same

    dispersion.[37]

    There have been many studies using this type of

    instrument with various particles[15,36,37] and very good

    results are found for particles >10mm. For sphericalpolystyrene particles[36] the results are very accurate and

    very reproducible. For inorganic particles both spray dried

    powders[37] and glass beads have been investigated[38]

    using various laser diffraction instruments and compared

    with image analysis. Figure 19 illustrates an example of a

    narrow sieved fraction (160200mm) for glass beads wherelaser diffraction and image analysis show good agreement

    (5%) for the dv50 with slightly more variation (10%) at the

    tails (dv10 and dv90) of the distribution.[38] The results with a

    standard quartz powder in the 1090mm range are a littleless accurate (1015% variation for median) as we move

    away from the perfect spherical shape assumed in all data

    reduction algorithms but still very reproducible.[15] The

    method is also very quick, the analysis takes only a few

    minutes so a high throughput can be achieved and auto-

    matic samplers are also available.

    When we move below 10 mm towards 1 mm then lightscattering becomes more important and the Mie theory has

    to be applied. Further limitations of the instrument arise

    when particles are non-spherical and when the relative

    refractive index between solvent and powder is small or

    optical constants of the powder unknown. Another limita-

    tion is found when trying to use the method for on-line

    analysis and suspensions are too concentrated leading to

    multiple scattering problems.[36] When the refractive

    indices of the medium and particle are similar (ratio of

    real parts of the RI< 1.1), even for particles >5 mm, thenagain the Mie theory has to be used and the optical

    constants input into the analysis routine if not assigned

    diameters can be as much as 25% in error.[36]

    When we have spherical particles and a narrow size

    distribution (e.g. 0.5 to 2.0 mm) the results compared toimage analysis can be very good even in the micron

    range. For example a spherical silica of nominal diameter

    1 mm gave a median diameter of 1.04 mm from imageanalysis[39] and 1.032 mm (standard deviation 0.0083from five repetitions) when using a MMS with a refrac-

    tive index of 1.412 (real) and 0.00 (imaginary) for

    amorphous silica. When there is a significant fraction

    of sub-micron particles present in the powder to be

    analyzed, the optical model chosen in the set-up becomes

    critical. The Mie theory requires knowledge of the

    refractive indices of the particles and suspending

    medium. A systematic study of laser diffraction instru-

    ments with the BCR 66 standard quartz powder covering

    the 3 to 0.3 mm range by Allen and Davies[15] stillshowed limitations in this range even when using the

    Mie theory correction. This study was carried out some

    years ago and current algorithms may have improved but

    this effect, to a major extent, may be due to the irregular

    shape of these particles as seen in the photocentrifuge

    study described earlier.

    There has been much discussion about the complex

    part the refractive index of powders which corresponds to

    the absorption of light where often there is no informa-

    tion available and the value has to be estimated. One

    method of arriving at a value is to adjust the value until

    the best fit between the calculated and measured data is

    found; though naturally care must be taken assure the

    result for the imaginary part is reasonable for the material

    under investigation. In the ISO 13320-1 Particle Size

    Analysis by Laser Diffraction Methods the conclusion

    was that small values of the imaginary part of the

    refractive index (about 0.010.1) are applied to cope

    with the surface roughness of the particles. The authors

    personal opinion is that this correction has more to do

    with shape than surface roughness as illustrated for the

    Figure 19. Example of PSD comparison between image

    analysis and laser diffraction (MMS) for sieved spherical

    glass beads.

    648 Bowen

    2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

    MARCEL DEKKER, INC. 270 MADISON AVENUE NEW YORK, NY 10016

    Dow

    nloa

    ded

    by [I

    ndian

    Insti

    tute o

    f Tec

    hnolo

    gy -

    Delhi

    ] at 0

    8:58 2

    5 July

    2013

  • light scattering correction with the photocentrifuge

    previously discussed. The motivation behind this

    approach may be understood from the results shown

    in Fig. 20. Here the standard quartz powder BCR 66

    data is compared to the laser diffraction PSDs using

    absorption coefficients of 0.1 and 0.001. The results for

    the 0.1 coefficient are closer to the standard data (for all

    percentiles >40%) even though the coefficient for such arelatively pure quartz should be closer to 0.001. As

    discussed above the reason for this discrepancy is

    thought to be related to the non-spherical particle shape

    of the BCR 66 and this effect of particle shape will now

    be discussed in more detail. Before we do that, one last

    point is that the BCR 66 was standardized by a sedi-

    mentation method where perhaps the very irregular shape

    has less influence. One of the favored ways of standar-

    dizing a powder is to use image analysisseeing is

    believing as they saybut using image analysis of an

    irregular shaped particle to then produce an ESD for

    comparison with diffraction or sedimentation is in itself

    a difficult task.[8] This point for irregular particle

    shapes will be further discussed in the comparison of

    instruments section.

    When the particles deviate significantly from a sphere

    then all the spherical assumptions in the optical models

    are obviously at fault. Azzopardi[36] quotes an example

    from the work of Swithbank et al. where for cylinders

    with lengths more than three times the diameter, the

    diffraction method gave a particle size (volume

    diameter) which was 12% smaller than the cylinder

    diameter and not at all sensitive to the cylinder length.

    Other studies have also shown that with aspect ratios

    below three then the ESD from laser diffraction reason-

    ably represents the real volume of the particle.[40] So

    although there is the possibility of detecting changes in

    shape using such methods[41] it is a detection and not a

    PSD measurement. Gabas et al.[42] studied some model

    non-spherical particles, namely cubes, flat rectangular

    plates, and cylinders using laser diffraction. The parti-

    cles were monodispersed with dimensions between 200

    and 1000 mm and they saw that the minimum andmaximum dimensions were detected with a more or

    less continuous size distribution between. This leads to

    the conclusion that an aspect ratio could be calculated

    reliably for these regular monosized anisotropic parti-

    cles, but information on the size distribution was not

    possible. Naito et al. have also looked at the effect of

    particle shape on the response from five different types

    of instruments[43] laser diffraction included, and also

    concluded that an aspect ratio can be extracted for rod

    like particles and orientation effects in the cell play a

    significant role. More recently Matsuyama et al.[44] have

    derived quantitative diffraction patternESD relation-

    ships for laser diffraction illustrating the above effects.

    They show that for aligned monosized ellipsoids a bi-

    modal distribution will result from the diffraction data. If

    the ellipsoids have a size distribution and are randomly

    oriented the resulting distribution becomes monomodal.

    The fact that the diffraction pattern for an ellipsoid is no

    longer symmetric in the x-y plane [Fig. 18 (c)] means

    that in order to derive shape information from the

    diffraction pattern data has to be collected in this

    plane and not just the y direction as for most

    instruments. This has recently been developed with

    the Micromeritics Saturn Digisizer but although shape

    analysis is not yet available the potential for regular

    shape investigation is becoming a reality.

    Recent work at LTP has looked at the use of some

    model anisotropic particles with regular cylindrical or

    platelet morphologies,[32] as well as some more realistic

    natural powders namely a commercial mica and some

    precipitated copper oxalates rods.[32] Several methods as

    well as laser diffraction data were compared with image

    analysis to help interpret the results. The glass fibers

    [regular cylinders, 10 mm diameter and lengths from10600 mm, Fig. 11(b)] showed no correspondencebetween laser diffraction and image analysis data in

    agreement with the work of Gabas et al.[42] but contrary

    to the photocentrifuge discussed above.[32]

    Figure 20. Standard quartz powder BCR 66 data and the

    measured laser diffraction PSDs using absorption coefficients

    of 0.1 and 0.001.

    PSD Measurement from Millimeters to Nanometers 649

    2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

    MARCEL DEKKER, INC. 270 MADISON AVENUE NEW YORK, NY 10016

    Dow

    nloa

    ded

    by [I

    ndian

    Insti

    tute o

    f Tec

    hnolo

    gy -

    Delhi

    ] at 0

    8:58 2

    5 July

    2013

  • For the platelet morphology laser diffraction has

    already been shown by Baudet et al.[45], with clay

    samples and an estimated thickness, to give good

    median diameter (d50) correlations. To compare data

    collected, contrary to the Baudet et al. method which

    needs a calibration sample, we chose to calculate rota-

    tional volume diameters (Dvrot) from the image analysis

    data (simply the spherical volume of a disc rotated 360

    through the diametral axis). The orientation of the plate-

    lets as they pass through the laser diffraction sample cell

    was treated as being perfectly oriented perpendicular to

    the laser beam (i.e., plane of disc normal to the incident

    beam) or randomly oriented. The final assumption made

    to allow a comparison of the Dvrot and Dv laser diffraction

    data is that the diffraction pattern of a disc approximates

    that of a spherefrom the results of Baudet et al.[45] this

    assumption seems to be reasonable. The platelets studied

    were alumina single crystals (Atochem, France) which

    had a very regular morphology and a reasonably constant

    thickness of 0.6 mm (Fig. 21(a)). Figure 22 shows thePSD results and we see a reasonable correlation between

    the oriented image analysis Dvrot median (6.8 mm) and thelaser diffraction median (7.7 mm). This was by far thebest correlation between the image analysis and any of

    the four PSD instruments studied.[31] The tails of the

    distributions show a large divergencethis was also the

    case with the study by Baudet et al.[45] The discrepancy

    can be attributed both to the laser diffraction deconvolu-

    tion algorithm and to some degree for the coarse fraction

    the image analysis. As agglomerated platelets were

    eliminated in the image analysisand some degree of

    agglomeration seems very likely from the micrograph

    shown in Fig. 21(a). So care should be taken when using

    this approach, if the tails of the distribution are of

    particular importance in an application but the laser

    diffraction median is representative of the real platelet

    median.

    Further work on mica particles with aspect ratios

    between 10 and 20 has confirmed this correlation

    between laser diffraction and image analysis.[32] The

    image analysis was carried out on several different size

    ranges (010, 10250, 250400, and >400 mm) to mini-mize sampling errors.[32] The distribution is very broad

    ranging from 10 mm to 1000 mm. The correspondencebetween image analysis and laser diffraction is reason-

    ably good (Fig. 23). The discrepancies seen at the higher

    particle sizes have been attributed to the sampling

    Figure 21. Micrographs of (a) the model alumina platelets

    and (b) the commercial mica.

    Figure 22. Particle size distributions of alumina platelets

    from laser diffraction and image analysis as a function of the

    volume ESD.

    Figure 23. Comparison of image analysis and laser diffraction

    PSDs for a commercial mica.

    650 Bowen

    2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

    MARCEL DEKKER, INC. 270 MADISON AVENUE NEW YORK, NY 10016

    Dow

    nloa

    ded

    by [I

    ndian

    Insti

    tute o

    f Tec

    hnolo

    gy -

    Delhi

    ] at 0

    8:58 2

    5 July

    2013

  • problems of image analysis despite the precautions of

    analyzing four fractions. Even though 8000 particles in

    total were counted by image analysis we sample less than

    1 mg and to have a statistically representative sample for

    the higher end of the distribution 2 g is needed[32]

    (the laser diffraction was collected using 12 mg of

    mica). To actually back calculate a disc particle size

    distribution, rather than the rotational equivalent

    compared in Fig. 23, an assumption of constant thickness

    would be necessary this will depend on the production

    route of the powder but is surprisingly often a reasonable

    assumption.

    In conclusion we have seen that despite the complica-

    tions of the data reduction in laser diffraction methods

    when a certain number of the basic assumptions are met,

    very reliable data can be collected. The method is

    particularly well suited to particle size measurement in

    the 101000 mm range where comparison with imageanalysis gives dv50s within 5%. When dealing with

    narrow size distributions of a spherical silica even at

    around 1 mm dv50s better than 5% can be obtained. Thetails of the distributions seem to be in less agreement

    with image analysisbut then image analysis is also

    prone to sampling errors if the distribution is particularly

    broad or there is a population of agglomerates difficult to

    interpret in the image analysis. For very irregular shaped

    particles such as ground quartz there is also the problem

    of transforming the shape into an ESD for comparison

    with laser diffraction. These aspects for a sub-micron

    alumina will be discussed further in the comparison of

    instruments section.

    For anisotropic particles such as discs and plates a

    very good correspondance can be found for discs

    assuming alignment in the measuring cell and that the

    data is put on the same equivalent volume basis. Many

    authors have pointed out the possibility of gleaning an

    average aspect ratio from rod like particles,40,42,43,45 but

    information on the length or size distribution does not

    seem possible. The non-symmetry of the diffraction

    pattern for such ellipsoid particles does allow for further

    development with new 2D detection systems which

    could help improve this aspect. In general, when the

    dispersions used are themselves reproducible, the laser

    diffraction method gives very reliable results. Much

    effort has been made very recently by the various

    manufacturers to improve the sensitivity and accuracy

    in the sub-micron and even sub-100 nm range by using

    extra light sources, detector arrays, polarized light and

    different wavelength light.[46,47] This along with more

    rigorous improved algorithms and inversion techniques

    due to improved computing power are helping push this

    speedy and reliable method to its limits.

    Photon Correlation Spectroscopy

    The DLS method or PCS or quasi-elastic light scatter-

    ing (QELS) as it was first termed is another method

    which depends on the interaction of light with parti-

    cles.[48] The earliest work on light scattering was carried

    out by Rayleigh[49] and the assumptions used in his

    approach limits the application to particles much smaller

    than the wavelength of the incident radiation. Rayleigh

    showed that the scattered intensity I (flux per unit area),

    I I016p4R6n2 1n2 22

    r2l410

    Where I0 is the incident intensity, n n1=n0 the relativerefractive index of particle of refractive index n1 and

    suspending medium n0, R is particle radius, l thewavelength of light in the medium, and r the distance

    between the scattering particle and the detector. The Mie

    theory,[26] as already mentioned, gives a complete solu-

    tion for spherical particles of any size and refractive

    index. The light scattered by colloidal particles in suspen-

    sion, which undergo Brownian movement due to thermal

    agitation, will fluctuate with time and can be related to the

    diffusion coefficient, Dt of the particle.[48] The diffusion

    coefficient can be related to a hydrodynamic diameter dhand for spherical morphology is given by the Stokes

    Einstein equation

    Dt kBT

    3pZdh11

    where T is the absolute temperature, Z is the suspendingliquid viscosity, and kB Boltzmans constant. This

    hydrodynamic diameter is very similar to the geometric

    diameter in most cases. The main exception being very

    small (

  • the cumulant method. For a Rayleigh scatterer the results are

    independent of the angle at which you make measurements

    but once out of this region (n17n0)R=l 1 more completetheories have to be used.[48,50]

    The PCS method is particularly well suited to the

    measurement of narrow particle size distributions in the

    range 1500 nm and very accurately as for the examples

    of latex standards frequently cited and Weiner even

    quotes an example where the standard was corrected

    after some PCS measurements.[48] Photon correlation

    spectroscopy is capable of measuring accurately down

    to 1 or 2 nm as illustrated by the example on tetra-

    propylammonium bromide[48] but not with a low inten-

    sity laser and the preclusion of dust is imperative. The

    dependence of the scattered intensity for Rayleigh

    scattering was shown in Eq. (10) to be proportional to

    the sixth power of the diameter. A particle only twice

    the size will give 64 times the intensity so any dust

    contamination will bias the results significantly. One

    method often used to remove dust for very fine colloids

    (

  • Zetasizer (MZS) manual) where several PSDs are shown

    all of which fit the PCS data with similar accuracy. One

    can derive acceptable particle size distributions from PCS

    measurements but more complicated algorithms for the

    data analysis must be used and more accurate and multi-

    angle data collected.[53] However, as this method is

    extremely quick, rapid analysis can be carried out and

    if the resulting distribution is narrow and reproduceable,

    the data is probably reliable. If a certain polydispersity is

    registered then another method, such as microscopy or

    sedimentation, should be used to discover which of the

    various solutions (e.g. Fig. 24) gives the best agreement

    with the other methods.

    The upper size limit of the PCS method has been stated

    as being around 500 nm. Wiener[48] has very clearly

    described the reasons behind this which are related to

    the number of particles per unit volume, sedimentation,

    slower diffusion for larger particles. First sedimentation, if

    the particles are too dense with respect to the suspending

    liquid they may sediment during the course of the data

    acquisition. Also the diffusion slows down as particles get

    bigger and experiments take longer to collect the same

    quality data as collected for smaller particles. Also one

    must be careful to make sure that there are enough

    particles per unit volume otherwise small fluctuations in

    the baseline can have drastic effects on the data quality.

    The difficulty in assuring that the volume fraction of

    particles is sufficient is that at higher concentrations we

    meet the onset of particleparticle interactions (around

    102) and this is often difficult to achieve in the 0.5 to1 mm range where methods such as centrifugal sedimenta-tion are better suited.

    So in summing up the PCS section, we can say that it

    is very useful for rapid measurements of narrow size

    distributions between 1 and 500 nm. Most of the results

    on sub-100 nm particles collected at LTP over recent

    years, have used PCS as a starting point or a back up

    screening method. Most of the examples are best

    described and discussed in comparison with other meth-

    ods such as sedimentation and microscopy (linked with

    image analysis) which leads on to the following and final

    section of this paper.

    COMPARISON BETWEEN

    INSTRUMENTS

    In this section we shall look at a series of powders

    from 20 nm boehmites to 400 micron glass spheres and

    101000 mm mica flakes and compare results from eitherdifferent methods or different instruments that use the

    same basic principles. Before doing this one or two

    statements about which methods and why different

    methods should produce more reliable results for certain

    powders than others. The main difference is between the

    ensemble techniques; where we have a diffraction pattern

    or light scattering signals that come from all of the

    sample and the separating techniques, such as sedimenta-

    tion or even microscopy, where the contribution from

    different size ranges is separated to a certain degree. The

    separation techniques are generally believed to give more

    reliable representation of the size and distribution widths

    whereas the ensemble techniques are very rapid. So the

    separation techniques are slower and are to be preferred if

    an accurate size and distribution are needed whereas the

    ensemble techniques are well adapted to rapid screening

    or comparison between samples for say quality control

    techniques. Bearing in mind these points we shall now

    look at the various powders from nm to mm.

    Colloidal Boehmite (1040 nm)

    A commercial boehmite, Disperal HP 14=2 (suppliedas a dry powder, Condea, Germany), has been analyzed

    using, PCS, XDC, a photocentrifuge (CAPA), and gas

    adsorption. The boehmite was dispersed in a 2% acetic

    acid solution and treated for 10 min using an ultrasonic

    probe (150 W, 20 kHz, 50 mL volumes). The samples

    were diluted as necessary for the photocentrifuge and

    PCS measurements. For the sedimentation methods the

    theoretical density 3.01 g=cm3 was used because the verylow agglomeration factor after dispersion (see Ref. [29]

    for more details). The specific surface area from nitrogen

    adsorption was 214 m2=g which gives a spherical dBET of9.3 nm [Fig. 26(a) shows a somewhat elongated primary

    particle shape]. The PSDs for the three other methods

    are shown in Fig. 25 and show very good agreement with

    each other showing a median diameter of 16 nm (1 nm,average from 23, PSDs for each instrument). The

    photocentrifuge data is corrected for light scattering

    after re-normalization and averaging of 69 separate

    measurements (of between 1 to 3 hrs)the whole data

    collection and data processing taking about 18 hrs i.e.,

    nearly two working days. The XDC took less than 2 hrs

    and has a much greater data resolutionthe photcentri-

    fuge having only three data points in this interval. That

    the photocentrifuge gives such a good correlation with

    the other two methods when taking into account the lack

    PSD Measurement from Millimeters to Nanometers 653

    2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

    MARCEL DEKKER, INC. 270 MADISON AVENUE NEW YORK, NY 10016

    Dow

    nloa

    ded

    by [I

    ndian

    Insti

    tute o

    f Tec

    hnolo

    gy -

    Delhi

    ] at 0

    8:58 2

    5 July

    2013

  • of resolution and data correctionis quite remarkable.

    The PCS with powder dispersion included took less than

    20 minillustrating its great benefits when dealing with

    narrow size distributions in the nm range. To what degree

    the dv50 of 16 nm is accurate has been discussed to some

    degree in the XDC section and also by Staiger et al.[29]

    Commercial Gamma Alumina 30200 nm

    The particle size distribution results for a commercial

    gamma alumina powder Aluminiumoxid C (Deg C,

    99.6% Al2O3, 2.8% alpha, Degussa, Switzerland) for

    PCS, XDC, and the CAPA are shown in Fig. 27. Here

    we see a much greater variation in the PSDs measured by

    the various instrumentsvarious characteristic values are

    given in Table 4. The specific surface was measured to be

    92 m2=g leading to a spherical dBET of 19.2 nm indicatinga certain amount of agglomeration when compared to the

    other instruments. For these sedimentation measurements

    hydrodynamic densities were calculated using porosity

    data from nitrogen adsorption measurements. These were

    made on loose or freeze dried powders to avoid inter-

    particle porosity that might be introduced from packed

    powders.[29] For comparison of results the XDC has been

    used as the yardstick. This choice was made as the x-ray

    absorption is directly proportional to mass, no models for

    light scattering needed, as for the two other methods.

    Which method gives the most accurate absolute measure-

    ment is not always easy to discern without image

    analysiswhich is difficult to perform on such fine

    powders.

    For the photocentrifuge, bearing in mind the assump-

    tion of spherical particle shape and a refractive index

    estimated, assuming the porosity in the agglomerate was

    filled with waterone can see that the light scattering

    correction for Horiba data is really quite good. The

    correction seems to slightly overcorrect, showing smaller

    values for mean and median diameters but are within

    20% of the XDC data. The results from PCS however

    seem to significantly underestimate the sub-100 nm range

    and consequently the fine tail of the distribution. This

    gives a mean diameter of 134 nm and a median diameter

    of 123 nm. This is a typical example of the sensitivity of

    light scattering to a low population of agglomerates

    discussed above. This can be seen in the XDC standard

    Figure 25. Comparison of PCS, XDC and photocentrifuge

    (CAPA) PSDs measured for a commercial boehmite.

    Figure 26. TEM images of (a) the commercial boehmite and (b) gamma alumina investigated.

    654 Bowen

    2002 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

    MARCEL DEKKER, INC. 270 MADISON AVENUE NEW YORK, NY 10016

    Dow

    nloa

    de