Saws Day 3 Master

―Statistical Considerations Involving
Particle Size Analysis‖

or ―Big numbers and little numbers‖ or ―Heterogeneity is the mother of variability‖
Alan F Rawle Ph.D

Applications Manager
Malvern Instruments Inc., Westborough, MA 01581
USP Workshop on Particle Size: Particle Detection & Measurement
December 8 - 10, 2010
USP Meetings Center, Rockville, Maryland
Synopsis
3 areas I‘ll be considering
 Averages: Volume/mass & number
 Statistics of sampling – fundamental sampling
error (FSE), Poisson, Student‘s T-Test
 Statistical Process Control/PAT
But we‘ll start with a brief introduction
In the material I‘ll provide an Appendix with

references and some additional basic
material, but will not burden you with this
during the talk
What do I need to think about and define before
I even attempt a particle size measurement?
What is the AQL (acceptable quality level) of the organization?
Or alternatively, what is the required or desired standard error
(SE) in the measurement?
This relates to any specification that is or will be set.
This is defined by the end user (perhaps a hint in ISO/USP
guidelines)
Is a bulk size (‗as is‘/with agglomerates) required or is a dispersed
(primary) size desired? This is the answer to the question ―What
is the purpose of taking the measurement? The answer governs
the energy that will be required in the measurement and is related
to the end use of the material and the fitness of purpose thereof.
This is defined by the end user
What is the top end (largest size) and the polydispersity
(width/spread) of the particle size distribution and the density of
the material?
This is set by the material and its mode of manufacture
What is the mass of sample that you utilize in your particle size
experiment?
This is set by the accessory utilized or purchased by the user and
the amount of sample added. You‘re in control!
Heywood
―However, it must be realised that particle size analysis is

not an objective in itself but is a means to an end, the end
being the correlation of powder properties with some
process of manufacture, usage or preparation‖
H Heywood Proc. 1st Particle Size Anal. Conf. September
1966 p 355 – 359, Heffer
This comment is in the
summing-up plenary lecture
Professor Harold Heywood

(1905 - 1971)
Specification
―Fit for purpose‖
―Just good enough‖
What happens to the numbers?
What are they used for?
What are the consequences of:
 ―Failure?‖
 ―Out-of-specification?‖
 ―Reject?‖ (verb)
We have to get both the size (x) AND quantity (y) axes
correct… a 2-D problem for a 3-D particle
And we may want control with a single number (1-D)
Particle counting and particle size analysis
or particle size distribution
USP <788> Particle Counting
USP <429> Particle Size Distribution
(USP <776> Optical Microscopy)
How many? Particle Counter
 Concentration
 Contamination
 Cleanliness
 Naturally leads to numbers
How much? Ensemble-based statistics
 > x mm
 < y mm
 Naturally leads to volumes/masses
Number means
D[1, 0] – Microscopy – the simplest

number mean (number-length): 1-D
D[2, 0] – Image analysis/obscuration –
projected or circular equivalent area
(number-area): 2-D
D[3, 0] – Electrical sensing zone – volume of
individual particle (number-volume): 3-D
Deal with the numbers of particles

Cope well with a linear vertical (quantity) axis
Moment (strictly Moment-Ratio) means
See Appendix for ISO 9276-2 equivalents
D[3, 2] – Surface area moment mean

(Sauter mean diameter) – relates directly to a SSA
measurement (6/D[3, 2] = SSA)
D[4, 3] – Volume/mass moment mean (De Brouckere
mean)
Numbers of particles irrelevant (after all it‘s a

distribution!)
1g of SiO2 r ~ 2.5 g/cm3 contains ~ 7.6 X 108 particles if they are 10 mm in size
Cope well with a logarithmic horizontal (size) axis
Reflects the value of the particles (in situ nugget effect)
Number and volume
- objects in space
Size Number % by % by % by
(cm) of number mass $/€/¥
Objects
10 - 1000 7000 0.2 99.96 99.96
1 - 10 17500 0.5 0.03 0.03
0.1 - 1 3500000 99.3 0.01 0.01
Total 3524500 100.0 100.0 100.0
Adapted from New Scientist 13th October 1991
Transformations theoretically possible but not advised

Number/visualization and volume/mass
How many? How much?
Highest resolution Low resolution
Every particle seen, counted or imaged Ensemble distribution
Size and shape information Size only but statistically large numbers of particles
Limited or no shape information
Small numbers can be seen Direct value - $ - distribution
Hints as to nature of particle No chemical information
Can be slow - replicate information difficult or impossible Rapid - replicates usually trivial
Lots of information Averaged information
Measured directly Often derived (terminal velocity, angular dependence
of light scattering
Direct visualization Graphs and tables - statistics
The unusual can be seen The unusual may be difficult
Static/2-D Dynamic - changes in size due to dispersion can be
monitored
D[x, 0] D[p, q] Typically (p -1) = q

Number means Volume/mass moment means
Responds well to linear y-axis Responds well to logarithmic x-axis
Weights as d Weights as d^3
Dust? Smaller particles much less significant

Lower limit (2 or 3mm) for visible Limits can be low (sub-nm for DLS or SAXS)
light - mathematical artefact
The harder you look the more you see
Small numbers of large particles pose a problem
2-D representation (z-axis?) Volume/mass

Wide distributions provide challenges
Need statistically valid numbers of particles
Small amounts of material used Can use grams or more of material
Difficult to carry out on-line Good for quality control and on-line applicability
USP <776> Optical Microscopy
―For irregularly shaped particles,

characterization of particle size must include
information on particle shape‖
3R‘s & 3M‘s in particle sizing
Allow the variability (‗error‘) balance to be assigned
Accuracy – do I hit the hole?
Precision – how close are my shots (after Pitard and Deming)?
Measures of precision:
 Repeatability – same group of particles measured consecutively in a set
of measurements: intra-assay precision
Machine + Method (changes in material during measurement: time)
 Reproducibility – another sample; more samples: batch heterogeneity
Machine + Method + Material
 Robustness – how a key variable affects the end result (Club selection?
Imaginary part of RI)
3 R‘s sorted out in method development
But bulk versus primary/dispersed?
Dry powder – PST - Pressure-Size Titration – DP of
venturi
 No plateau
 Compare with dispersed wet result
 Sample heterogeneity - reproducibility
Wet – BDAS – before, during and after sonication
(wetting, separation, stabilization)
 Plateau possible
 Microscopy to assist
 Repeatability/intra-assay precision (single sample)
Calculation of mean and RSD/CV for 10 or more
results
Effect of time
Stability, stability, stability
Heywood again – myth of sensitivity
A number counter isn‘t necessarily more sensitive to the fine material –
otherwise a 100 nm latex would be measured just fine on a visual light
image analyzer or microscope
Bottom-end imposes a mathematical as well as a physical barrier – the
harder we look then the more we see:
Dr. H Heywood: Dr. Bronowski asked whether anyone could comment upon
automatic star counting methods; although I cannot say what methods were used for
this purpose there was an interesting similarity between star and particle counting
results when the limit of visibility was approached. Early records of microscopical
counts, when extended to the extreme limit of resolution, showed a false maximum
frequency just greater than the lower limit, whereas it is now known that the
numerical frequency for normal dusts increases continuously as the particle size
decreases. Star counts made by unaided vision about 300 years ago showed a
maximum frequency at about 4th magnitude, the limit being around 6th magnitude.
With the aid of telescopes it was easily shown that the numerical frequency of stars
increased as the magnitude increased, i.e. brightness decreased. The important
feature of this comparison was that false effects might be observed if any counting
process was worked to the extreme limit of its capabilities.
The Physics of Particle Size Analysis Conference University of Nottingham

6 – 9 April 1954 British Journal of Applied Physics Supplement No. 3 Page S-177
―Your decisions are only as good as your
samples‖ Francis F Pitard
Are we sampling or taking a specimen?
Are we gambling?
―A sample is representative when the mean square of the
total Sampling Error is not larger than the sum of the
acceptable accuracy and acceptable precision‖
‗Batch homogeneity‘ is an oxymoron for a distribution
‗Batch (or lot) heterogeneity‘ is correct
USP <429>: on a heterogeneity basis 3 s = 15% (using
Student‘s T-Test if applicable) is actually tough to hit, if
care in sampling in not taken (segregation, in situ nugget
effect, fundamental sampling error) Gy used 1 s as 16% as
his basis for sampling error.
Learn from the geologists – 50 assays minimum taking 5kg
samples ground to < 150 mm and 30 g for a fire assay
(inadequate amount – Pitard)
―Sampling is nothing but representative mass reduction‖
Pierre Gy
And how do you take your sample?
With a scoop? Like a grocer…
Reliability of selected sampling methods using a 60:40 sand mixture
Sampling technique Standard deviation
Cone and quartering 6.81

Scoop sampling 5.14
Table sampling 2.09
Chute slitting 1.01
Spinning riffling 0.146
Random variation 0.075
Taken from Table 1.5 page 38, T.Allen Particle Size Measurement 5th Edition Volume 1
Chapman and Hall 1997 ISBN 0 412 72950 4
Note: +/- 3 s for scoop is ~ +/- 16%!

Standard Error (Gy: Fundamental Sampling Error, FSE)
A population has a true mean and standard deviation. When we're sampling we don't
know what those 'true' values actually are. The standard error is the estimate of the
true standard deviation based on the actual samples we take - it's the measured as
opposed to the true/real - unknown - s.d.
Is inversely proportional to the square root of the

number of particles
S.E. a 1/√n or n a 1/s2
For 1% S.E. then:
n = 1/(0.01)2 = 10000
10000 particles total will be needed to specify the

mean to 1% S.E.
If the x90, for example, is required to 1% S.E. then
10000 particles need to be in the x90+ size band – and
this defines the minimum mass of sample for
statistical validity
―I wish to detect a small amount of
agglomeration in my system‖
Assume: 1% of agglomeration at 2 mm (2000

mm) in an aqueous pharmaceutical slurry
(20% solids by mass). Assuming a particle
density of the solid of 1.5 g/cm3, we can
calculate the minimum mass of sample to
have a 99% confidence limit in seeing this
agglomeration
In cgs units:
~ 6300 g
―I wish to detect a small amount of
agglomeration in my system‖
This table indicates the minimum mass of material

needed to specify the x99 point to a standard error of
1% when the x99 point is at d mm and the density of
material is 1.5 g/cm3
d (mm) d/10000 d^3 Density (g/cm3) 10000*100*(p/6) Minimum Mass (g)
1 0.0001 1E-12 1.5 523598.7756 0.000001
10 0.001 1E-09 1.5 523598.7756 0.000785
100 0.01 0.000001 1.5 523598.7756 0.79
200 0.02 0.000008 1.5 523598.7756 6.28
500 0.05 0.000125 1.5 523598.7756 98.2
1000 0.1 0.001 1.5 523598.7756 785.4
1500 0.15 0.003375 1.5 523598.7756 2651
2000 0.2 0.008 1.5 523598.7756 6283
5000 0.5 0.125 1.5 523598.7756 98175
10000 1 1 1.5 523598.7756 785398
So you‘re taking only 20mg……….
Because it‘s all you can spare

Because it‘s valuable (but you could recover
it in a wet measurement)
x99 point required
Can we keep the material in suspension?
d (mm) r (g/cm3) Mass used (g) Mass of 1 particle (g) No. particles Calculated SE (%)
1 1.5 0.02 7.85398E-13 254647909 0.01
20 1.5 0.02 6.28319E-09 31831 0.56
50 1.5 0.02 9.81748E-08 2037 2.22
100 1.5 0.02 7.85398E-07 255 6.27
200 1.5 0.02 6.28319E-06 32 17.72
500 1.5 0.02 9.81748E-05 2 70.06
1000 1.5 0.02 0.000785398 0 198.17
1500 1.5 0.02 0.002650719 0 364.05
2000 1.5 0.02 0.006283185 0 560.50
Poisson statistics
Number of particles in 10 bottles:
0, 0, 0, 0, 0, 0, 1, 0, 0, 0
Probability of occurrence of 0, 1, 2, 3, 4, 5, 6 particles?
Average number of particles ()
Formula is:
P (x = r) = (r/r!)e
 = [(9 X 0) + (1 X 1)]/10 = 0.1
n e^-0.1 0.1^n n! P
0 0.905 1 1 0.9050
1 0.905 0.1 1 0.0905
2 0.905 0.01 2 0.0045
3 0.905 0.001 6 0.0002
4 0.905 0.0001 24 0.0000
5 0.905 0.00001 120 0.0000
6 0.905 0.000001 720 0.0000
Mode is below the average

Tendency to underestimate with small numbers – agglomerates in powder?
Outliers? These are actually vitally important and not ‗out and out liars‘
Student‘s t-Test (USP <429>…..)
When numbers are small (3 i.e. N = 2) the
measured standard deviation can exceed the
given value:
Two Sided Number
70% 90% 95% 99%
of estimates (N-1)
1 1.96 6.31 12.71 63.7

2 1.39 2.92 4.30 9.93
3 1.25 2.35 3.18 5.84
4 1.19 2.13 2.78 4.60
5 1.16 2.02 2.57 4.03
6 1.13 1.94 2.45 3.71
7 1.12 1.90 2.37 3.50
8 1.11 1.86 2.31 3.36
9 1.10 1.83 2.26 3.25
10 1.09 1.81 2.23 3.17
15 1.07 1.75 2.13 2.95
20 1.06 1.73 2.09 2.85
25 1.06 1.71 2.06 2.79
Is this being assessed?

http://en.wikipedia.org/wiki/Student's_t-distribution
The number ‗3‘
―Before the inherent variability of the test animals

was appreciated, assays were sometimes carried out
on as few as 3 rabbits: as one pharmacologist put it,
those were the happy days.‖
E C Fieller ‗The biological standardization of insulin‘
Suppl., J. Royal Stat. Soc., Vol VII, 1940-41 page 3
Quoted in Deming ―Some Theory of Sampling‖ p 76
(Incidentally a review of Walter Shewhart‘s ―Statistical
Method from the Viewpoint of Quality Control‖ book
appears in the same volume)
With a Gaussian distribution then +/- 1 s is 68% of the
population – 1 in 3 must lie outside….
Throw away the 2 outliers and you are left with a
perfect result
x100 – when the numbers are big
OK for screens – 100% passing
Largest particle in the system

Consider the 1 g of SiO2 r ~ 2.5 g/cm3 and all particles
10 mm
760 million or so…. (Say ~ 0.8 parts/billion)
Our system will need to be able to discriminate to this
level…
OK, if we could count all these particles (typically a
liquid particle counter will deal with up to 50000
particle/mL…)
―Quotation of an x100 value by laser diffraction is
specifically deprecated by this International Standard‖
ISO 13320: 2009 Section 7, page 20, ―Reporting of
results‖
http://www.iso.org/iso/iso_catalogue/catalogue_tc/catalogue_detail.htm?csnumber=44929
Standard deviations cannot be added
Only variances can…
So:
 Estimated sampling error 1 s = 16%
 Estimated dispersion error 1 s = 10%
 Estimated RI robustness factor 1 s = 12%
 Estimated analytical error 1 s = 0.5%
 Estimated maximum instrumental error 1 s = 0.5%
Wrong: Total Error = 16 + 10 + 12 + 0.5 + 0.5 = 39%
Correct: Total variance s2 = (0.16)2 + (0.1)2 + (0.12)2 + (0.05)2
+ (0.05)2 = 0.0550
Or total s = 23.5%
(Note: the largest error controls the game as expected)
Statistical Quality Control
Is the system in ‗equilibrium‘?
Are the variations we observe a statistical and natural result of the
process? If so, we cannot do better without a substantial change – or
‗breakthrough‘ as Juran referred to it. This is the role of management.
Deming: Out of the Crisis Juran: Managerial Breakthrough

Statistical Quality Control
Cannot inspect quality into a product

100%/200%/300% inspection?
Pass-fail on a single sample of a powder?
Steady-state operation
Mill variations
100% = No particles Transmission Mill Current Dv(50)

100 40
90
Mill Current [Amps]
80
Transmission [%]
30
70
Dv(50) [µm]
60
50 20
40
Manual sampling
30
10
20
10 Set Point
0 0
0% = Lots of particles 0 50 100 150 200 250 300 350 400
Dt Time from Start (s)

Steady-state operation
Mill variations - after smoothing the hopper feed
80
70 Transmission Dv(50)
Transmission (%)
60
Particle Size (µm)
50
40
30
20
10
0
0 100 200 300 400 500 600
Time from Start (s)

Non-steady-state operation
Screen mill
Feed rate Feed rate Screen

Incident Incident Breaks
Statistical Process Control
When the numbers are big
On-line: 1 measurement every second
Thus 3600 measurements/hour
Imagine a 6s (+/- 3s) regime
+/- 3 s takes in 99.7% of the population
3 in 1000 measurements will be outside this (above or
below)
Thus approximately 10 measurements/hour will be
‗out of specification‘ in such a system designed to
build quality in (cannot inspect it in!)
How will you handle this? Variogram (Pitard) a
correlation form of the Cusum (cumulative sum)
method of analyzing trend data
Summary
For heterogeneous systems (i.e. those with

a particle size distribution) then statistical
information is needed
Some areas where statistical understanding
is vital include:
 Calculation and understanding averages
 Number and volume/mass
 Display of information (discontinuous/continuous
distributions)
 Sampling
 Statistical Process Control
Thank you!
alan.rawle@malvern.com
+ 1 508 768 6434
Questions?
Appendix
References
Basic Principles
T Allen Particle Size Measurement Volume 1 Powder sampling and particle
size measurement Fifth Edition Chapman and Hall (1997)
R D Cadle Particle Size Determination Interscience New York (1955)
G Herdan Small Particle Statistics Second Revised Edition Butterworths
London (1960)
J M DallaValle Micromeritics The Technology of Fine Particles Second
Edition Pitman New York (1948)
Clyde Orr, Jr. Particulate Technology The Macmillan Company New York
(1966)
Clyde Orr, Jr., J M Dallavalle Fine Particle Measurement Size, Surface and
Pore Volume The Macmillan Company New York (1959)
H Heywood The Scope of Particle Size Analysis and Standardization‖ in
Symposium on Particle Size Analysis Supplement to Transactions,
Institution of Chemical Engineers, Vol. 25, 14 – 24 (1947)
R R Irani, C F Callis Particle Size: Measurement, Interpretation and
Application John Wiley & Sons Inc., New York (1963)
Appendix
References
Basic Principles
H E Rose The Measurement of Particle Size in Very Fine Powders London
Constable & Company (1953)
American Society for Testing Materials Symposium on New Methods for
Particle Size Determination in the Subsieve Range Washington Spring
Meeting March 4, 1941 ASTM, Philadelphia, U.S.A.
A Jillavenkatesa, A J Dapkunas, L-S H Lum Particle Size Characterization
NIST Practice Guide Special Publication 960-1 U S Government Printing
Office Washington (2001)
M J Groves Particulate Matter Sources and Resources for Healthcare
Manufacturers Interpharm Press Buffalo Grove, Illinois (1993)
H G Schroeder, P P DeLuca ―Theoretical Aspects of Particulate Monitoring
by Microscopic and Instrumental Methods‖ Journal of the Parenteral Drug
Association Vol. 34, No. 3 183 – 191 (May - June 1980)
S J Borchert et al Particulate Matter in Parenteral Products: A Review
Journal of Parenteral Science & Technology Vol. 40, No. 5 212 – 241,
(September – October 1986)
Appendix
References
Basic Principles
H G Merkus Particle Size Measurements Fundamentals, Practice, Quality
Springer (2009)
Clive Washington Particle size analysis in pharmaceutics and other
industries: theory and practice Ellis Horwood (1992)
A F Rawle Instrument Qualification and Performance Verification for
Particle Size Instruments Chapter 12 in Practical Approaches to Method
Validation and Instrument Qualification Ed. C C Chan, H Lam and X M
Zhang 255 – 298 Wiley (2010)
A F Rawle Analytical Tools for Suspension Characterization Chapter 6 in
Pharmaceutical Suspensions: From Formulation Development to
Manufacturing A K Kulshreshtha et al (Eds.) Springer Science + Business
Media, 177 – 230, DOI 10.1007/978-1-4419-1087-5_6 AAPS (2010)
A F Rawle Particle Morphology and Characterization in Preformulation
Chapter 3.2 in Preformulation in Solid Dosage Form Development M C
Adeyeye, H G Brittain (Eds.) Volume 178, 145 – 184, Informa Healthcare
(2008)
A F Rawle The Basic Principles of Particle Size Analysis Malvern
Application Note MRK-034 (1993) Downloadable:
http://www.nbtc.cornell.edu/facilities/downloads/Basic%20principles%20of%20particle%20size%20analysis.pdf
ISO 9276-2:2001 Representation of results of particle size analysis -- Part
2: Calculation of average particle sizes/diameters and moments from
particle size distributions
Appendix
References
Basic Principles
R A Mugele, H D Evans ―Droplet Size Distribution in Sprays‖ Ind. Eng. Chem. 43(6),
1317 – 1324 (June 1951)
M Alderliesten: 5 part series in Part. Part. Syst. Charact. (from 1990 – 2005) [Part 6 in
press]
M Alderliesten Mean Particle Diameters Part I: Evaluation of Definition Systems Part.
Part. Syst. Charact., 7 (1990) 233-241
M Alderliesten Mean Particle Diameters Part II: Standardization of Nomenclature Part.
Part. Syst. Charact., 8 (1991) 237-241
M Alderliesten Mean Particle Diameters Part III: An Empirical Evaluation of
Integration and Summation Methods for Estimating Mean Particle Diameters from
Histogram Data Part. Part. Syst. Charact., 19 (2002) 373-386
M Alderliesten Mean Particle Diameters Part IV: Empirical Selection of the Proper
Type of Mean Particle Diameter Describing a Product or Material Property Part. Part.
Syst. Charact. 21 (2004) 179 – 196
M Alderliesten Mean Particle Diameters Part V: Theoretical Derivation of the Proper
Type of Mean Particle Diameter describing a Product or Process Property Part. Part.
Syst. Charact. 22 (2005) 233 – 245
ASTM E 2578 – 07 Standard Practice for Calculation of Mean Sizes/Diameters and
Standard Deviations of Particle Size Distributions
The ISO terminology for particle size
ISO TC 24/SC 4 N rev1 Date: 2010-10-17 ISO/CD 9276-2 (Michael

Stintz + Maarten Alderliesten)
Moment notation (ISO)
Systematic code Terminology
x 1,0 arithmetic mean size
x 2,0 arithmetic mean surface size
x 3,0 arithmetic mean volume size
x 1,1 length weighted mean size
x 1,2 surface weighted mean size, Sauter mean diameter
x 1,3 volume weighted mean size

The ISO terminology for particle size
(in progress)
Moment-Ratio notation
Systematic code Terminology
D  3,0 harmonic mean volume size
D  2,1 diameter-weighted harmonic mean volume size
D 1,2 surface-weighted harmonic mean volume size
D  2,0 harmonic mean surface size
D 1,1 diameter-weighted harmonic mean surface size
D 1,0 harmonic mean size
D 0,0 geometric mean size
D1,1 diameter-weighted geometric mean size
D 2,2 surface-weighted geometric mean size
D 3,3 volume-weighted geometric mean size
D1,0 arithmetic mean size
D 2,1 diameter-weighted mean size
D 3,2 surface-weighted mean size, Sauter mean diameter
D 4,3 volume-weighted mean size
D 2,0 mean surface size
D 3,1 diameter-weighted mean surface size
D 4,2 surface-weighted mean surface size
D 5,3 volume-weighted mean surface size
D 3,0 mean volume size
D 4,1 diameter-weighted mean volume size
D 5,2 surface-weighted mean volume size
D 6,3 volume-weighted mean volume size

Appendix
Basics – Internet material
A F Rawle Webinars:
15th Mar 2007 Size aspects of particle size analysis - the math
http://tinyurl.com/23hspao
28th Oct 2008 Raising the standard (2) – Logical method development for
particle size analysis in the pharmaceutical industry
http://tinyurl.com/l8xth4
22nd Apr 2008 Raising the standard (1) - The need and use for
standardization and standards in the pharmaceutical regulatory
environment
http://tinyurl.com/24goplh
A little dated (webinar) now:

Basic Principles of Particle Size Analysis
http://tinyurl.com/2cvpach
A F Rawle The Basic Principles of Particle Size Analysis Malvern Application Note
MRK-034 (1993) Downloadable:
http://tinyurl.com/2c4unfn
Continuous and discontinuous distributions
Source of confusion
 Contrast a screen with light scattering
The number of size classes in a histogram
affects the vertical axis - a density
distribution is required in the vertical axis
A frequency curve is a continuous
distribution
 Can pass through any point in a histogram bar
 An infinitely small increment in size means an
infinitely small quantity in the vertical axis
An undersize plot avoids the confusion in
horizontal or vertical axes
Reconciliation
The horizontal and vertical axes present a

mathematical challenge with any particle
sizing technique:
 What do we mean by size
 What do we mean by quantity?
A linear vertical axis can‘t deal with small

numbers of particles
A horizontal linear axis can‘t deal with the
ranges typically encountered in particle size
analysis
Continuous/discontinuous distributions
Particle size technologists specify (ISO 9276-1:1998 Representation of
results of particle size analysis -- Part 1: Graphical representation):
Logarithmic size (x) axis, linear quantity (Qx) y-axis
Preference is for a frequency or cumulative (under or oversize)plot
because of possible histogram issues. Vertical axis then should be a
density distribution.
These are exactly the same:
K Sommer 40 Years of Presentation Particle Size Distributions – Yet Still

Incorrect Part. Part. Syst. Charact. 18, 22 – 25 (2001)
As is this….
• So, is everything really an average of 1000mm on an instrument with a

linear scale from 0.02 – 2000 mm?
• Of course not, but it can be seen that the width of the histogram size
class affects any generated average (s) or points of the distribution (if they
are interpolated)
• This becomes very important when the size of the distribution
approaches that of a class width (typically when the class width is 1/10th
or more of the distribution width)
Need to get the axes right!
Same number of horizontal size bands

Same way of representing the vertical axis
 Continuous distribution
Correct understanding of the meaning of the
axes:
 Understand the nature of the measurement
 Understand the lower end mathematical
‗brick-wall‘
Apply appropriate transformation(s)
 Understand any assumptions and applicability
Density distribution
(For example % per mm)
Appendix
References
Sampling
R Richards Ore Dressing 4 volumes (1909) Set back sampling practice by 50 years
(Ms = kd2 for ‗convenience‘ - not the correct Ms = kd3). See next slide.
Pierre Gy Sampling of Particulate Materials: Theory and Practice Second Edition
Elsevier Scientific Pub Co. (1982)
Pierre Gy Sampling for Analytical Purposes Wiley (1998)
Readable chapter on theory and practice of sampling: M G Whateley, B C Scott
Evaluation Techniques Chapter 10 in Ed. C J Moon M G K Whateley, A M Evans
Introduction to Mineral Exploration Second Edition 199 – 252 Blackwell (2006)
F F Pitard Pierre Gy‘s Sampling Theory and Sampling Practice CRC Press Boca
Raton (1993)
K Sommer Sampling of Powder and Bulk Materials Springer-Verlag Berlin (1986)
C H Murphy Handbook of Particle Sampling and Analysis Methods Verlag Chemie
International (1984)
Chapter 1 in Allen Particle Size Measurement referenced earlier
A F Rawle Sampling for Particle Size Analysis Malvern Instruments Application Note
MRK-456-01 (2001) From:
http://www.malvern.com/common/downloads/mrk456-01.pdf
A F Rawle Webinar:
June 3rd 2010 Sampling for particle size analysis - estimation of standard error
https://www.brainshark.com/brainshark/vu/view.asp?pi=654544412&text=M021507
ISO 14488:2007 Particulate materials -- Sampling and sample splitting for the
determination of particulate properties http://www.iso.org 124CHF
Appendix
References
Sampling
H Masuda, K Gotoh Study on the sample size required for

the estimation of mean particle diameter Advanced Powder
Technol. Vol 10(2) 159 – 173 (1999)
M W Wedd Procedure for predicting a minimum volume or
mass of sample to provide a given size parameter precision
Part. Part. Syst. Charact. 18(3) 109 – 113 (2001)
A F Rawle Webinar:
June 3rd 2010 Sampling for particle size analysis -
estimation of standard error
https://www.brainshark.com/brainshark/vu/view.asp?pi=65454441
2&text=M021507
ISO 14488:2007 Particulate materials -- Sampling and
sample splitting for the determination of particulate
properties http://www.iso.org 124CHF
Calculation of minimum mass
Totally different approaches: Gy via s and Rawle via numbers of particles
Pitard/Gy:
MS = 18 r f d3/σ2
(where d refers to 95% passing)
f = shape factor (= p/6 for spheres, 1 for cubes)
Rawle:
MS = 20 r (p/6) d3/σ2
(for x95; spheres considered)
r is density in g/cm3, d is size in cm, s is Fundamental

Sampling Error, FSE, (Gy/Pitard)
―Standard Error‖ Rawle – A E Annels Mineral Deposit
Evaluation: a Practical Approach Chapman and Hall,
London (1991). This is the term used in ―Mathematics and
Statistics for Chemists‖ C J Brookes, I G Betteley, A M
Loxston John Wiley & Sons (1966) page 266 and
―Statistical Methods‖ Harold Lucas Butterworths (1970)
page 77; these were my college and school statistical texts
Appendix - Sampling
Above: Henry A Vezin‘s calculations in Richards‘ Ore Dressing (Volume II page 850)
Interestingly Reference ‗14‘ in Richards is to H O Hofman Metallurgy of Lead‖ (1899)
where on page 45 (not 42 – the start of the chapter), Vezin is named with an 1866 comment
but no actual text reference is given….. (Hofman was a MIT Professor like Richards)
Appendix – Sampling
Rawle-Vezin comparison
X99 1% SE (99% confidence limits) density = 2.6 g/cm3

Needs 10000 particles and this represents 1% of the total mass of the system
d (mm) d/10000 d^3 Density (g/cm3) 10000*100*(PI/6) Minimum Mass (g) Vezin (1865/1866) Vezin/Rawle
1 0.0001 1E-12 2.6 523598.7756 0.000001
10 0.001 1E-09 2.6 523598.7756 0.001361
100 0.01 0.000001 2.6 523598.7756 1.36
125 0.0125 1.9531E-06 2.6 523598.7756 2.66 2.92 1.10
200 0.02 0.000008 2.6 523598.7756 10.89
250 0.025 1.5625E-05 2.6 523598.7756 21.27 23.3 1.10
500 0.05 0.000125 2.6 523598.7756 170.2 186.7 1.10
1000 0.1 0.001 2.6 523598.7756 1361.4 1493 1.10
1500 0.15 0.003375 2.6 523598.7756 4595
2000 0.2 0.008 2.6 523598.7756 10891 11950 1.10
4000 0.4 0.064 2.6 523598.7756 87127 95570 1.10
5000 0.5 0.125 2.6 523598.7756 170170
8000 0.8 0.512 2.6 523598.7756 697015 764600 1.10
10000 1 1 2.6 523598.7756 1361357
Vezin referenced in Richards' Ore Dressing

Volume II Page 850
Appendix
Sampling – Pierre Gy
Appendix - Sampling
New notations (Pitard/Esbensen)
Old and new notations
Pitard/Esbensen
Errors Old Term New term

Heterogeneity Fluctuation Error CE HFE
Quality Fluctuation Error QE QFE
Fundamental Sampling Error FE FSE
Grouping and Segregation Error GE GSE
Increment Weighting Error WE IWE
Increment Delimitation Error DE IDE
Increment Extraction Error EE IEE
Increment Preparation Errors PE IPE
Appendix
References
Statistical Process Control/Management
W A Shewhart ―Statistical Method from the Viewpoint of Quality Control‖
Dover Publications New York (1986) [Originally published in 1939]
W Edwards Deming ―Some Theory of Sampling‖ Dover Publications New
York (1966) [Originally published in 1950]
W Edwards Deming ―Elementary Principles of the Statistical Control of
Quality - a Series of Lectures‖ Nippon Kagaku Gijutsu Remmei, Tokyo,
Japan (Second Printing: June 1952) This is a nice rare book!
W Edwards Deming ―Out of the Crisis‖ 19th Printing Massachusetts
Institute of Technology, Cambridge, MA (August 1992)
J Seddon ―The case against ISO 9000‖, Second Edition, Oak Tree Press,
(November 2000)
J Butman ―Juran: a lifetime of influence‖ John Wiley and Sons., Inc (1997)
J M Juran ―Quality Control Handbook‖ McGraw-Hill Second Edition (1962)
J M Juran ―Managerial Breakthrough‖ McGraw-Hill New York (1964)
USP Workshop on Particle Size:
Particle Detection & Measurement
Analytical Measurement Options
Light Obscuration
Neal Zupec
Research Scientist/Baxter Healthcare Corporation
December 10, 2010
Baxter Healthcare Corporation

Overview
• light obscuration and <788>
• instrument considerations
• practical vs philisophical application

<788> Particulate Matter in Injections
First Supplement to the USP 33-NF 28 Reissue

Official October 1, 2010 to February, 2011
―Particulate matter in injections and parenteral infusions consists

of extraneous mobile undissolved particles, other than gas
bubbles, unintentionally present in the solutions‖
Instrument Requirements
• use a suitable apparatus based on the principle of light blockage
that allows for an automatic determination of the size of particles
and the number of particles according to size
• the apparatus is calibrated using dispersions of spherical

particles of known sizes between 10 µm and 25 µm. These
standard particles are dispersed in particle-free water
• care must be taken to avoid aggregation of particles during

dispersion
• system Suitability can be verified by using the USP Particle

Count RS.

Test Environment and Materials
• the test is carried out under conditions limiting particulate matter,
preferably in a laminar flow cabinet
• very carefully wash the glassware and filtration equipment used,

except for the membrane filters, with a warm detergent solution,
and rinse with abundant amounts of water to remove all traces of
detergent
• immediately before use, rinse the equipment from top to bottom,

outside and then inside, with particle-free water.

Environment Suitability (Background)
• determine the particulate matter in 5 samples of particle-free

water, each of 5 mL, according to the method described below
• if the number of particles of 10 µm or greater size exceeds 25 for

the combined 25 mL, the precautions taken for the test are not
sufficient
• the preparatory steps must be repeated until the environment,

glassware, and water are suitable for the test.

Method
• mix the contents of the sample by slowly inverting the container
20 times successively
• if necessary, cautiously remove the sealing closure
• clean the outer surfaces of the container opening using a jet of

particle-free water and remove the closure, avoiding any
contamination of the contents
• eliminate gas bubbles by appropriate measures such as

allowing to stand for 2 minutes or sonicating.

Method
• for large-volume parenterals, single units are tested
• small-volume parenterals having a volume of 25 mL or more may

be tested individually
• for small-volume parenterals less than 25 mL in volume, the

contents of 10 or more units are combined in a cleaned container to
obtain a volume of not less than 25 mL; the test solution may be
prepared by mixing the contents of a suitable number of vials and
diluting to 25 mL with particle-free water or with an appropriate
particle-free solvent when particle-free water is not suitable.

Method
• powders for parenteral use are reconstituted with particle-free

water or with an appropriate particle-free solvent when particle-
free water is not suitable.

Method
• The number of test specimens must be adequate to provide a

statistically sound assessment. For large-volume parenterals or
for small-volume parenterals having a volume of 25 mL or more,
fewer than 10 units may be tested, using an appropriate sampling
plan.

Method
• Remove four portions, not less than 5 mL each, and count the
number of particles equal to or greater than 10 µm and 25 µm.
Disregard the result obtained for the first portion, and calculate
the mean number of particles for the preparation to be examined.

Data Evaluation
Reporting: average of 3 aliquots not to exceed particle limit
≥10 µm ≥25 µm
<788> Test 1.A. Large-Volume 25 per mL 3 per mL

Injections (more than 100mL)
<788> Test 1.B. Small-Volume

6000 per container 600 per container
Injections (less than and equal to
100mL)
<789> Ophthalmic Solutions 50 per mL 5 per mL

Light Obscuration Test Elements
10 unit pool Single units
(<25mL) (>25mL)
Calibration Validation
Resolution Sampling Plans

Light Obscuration
Volume accuracy Assay Blanks
Count accuracy Training

(PCRS)
Limits

Sampler Options
• vacuum or syringe sampler- general use
• pressure sampler – used with viscous or volatile fluids, degases

samples containing entrained air
• stir bar – can be used with either sampler type (as long as it is
not a source of particulate matter itself)

Liquid Counter Options
• numerous methods and electronic instruments exist for the
enumeration and sizing of particles in liquids that measure
• light scattering
• light extinction
• or a combination of light scattering and light extinction
• resistance modulation

Light Obscuration Sensors - HIAC
HRLD Series Measurement Concentration Limit Flow Rates Standard Calibrated
Model Number Range (μm) (<10% coincidence) (mL/min) Flow Rate (mL/min)
HRLD-100 4 μm-100 μm 10,000 20-100 60
HRLD-100HC 4 μm-100 μm 18,000 10-50 20
HRLD-150 1.2 min.-150 18,000 10-50 10 or 25
HRLD-150JA 1.2 min.-150 18,000 10-50 10 or 25
HRLD-400 2.0 min.-400 10,000 20-100 60
HRLD-400HC 2.0 min.-400 18,000 10-50 20
HRLD-600JS 2.0 min.-600 6,000 30-200 100
Hach Company PO Box 389 Loveland, CO 80539

Principle of Light Obscuration Detection
 light is supplied by a laser diode source
 particles are constrained to pass through a defined view volume

located between the source and detector
 variation in light intensity is converted by the photodiode

detector to a voltage signal

Laser diode
View volume Photodiode

detector

Laser diode
View volume Photodiode

detector

Full
Pulse Height Analyzer
Scale
Electrical noise and debris
Voltage level for calibration particle median diameter

Counts
Voltage level for median diameter plus one STD. deviation
Voltage level for doublet particles median diameter
0 100 200 300 400 500 600 700 800 900 1000
Channel No.
Sources of Erroneous Count Data
• immiscibles due to oils such as silicone (from closures) or

plasticizers from flexible containers
• air or gas bubbles
• particle size and shape
• particle overconcentration
• refractive indice differences

Data
• demonstrates compliance
• used to determine manufacturing deviations
• used to detect formulation issues

Ideal
• in development, once the acceptable level of particulate matter

has been obtained and the particle population make-up has been
determined to be heterogeneous, the light obscuration method
becomes a monitoring tool over the life of the product

Considerations
Light ―obscuration‖ counters
detect differences that exist
within and relative to a uniform
fluid medium in motion

Considerations

Particle Contamination- Definitions
• a particle is any object having definitive physical boundaries in

any direction
• contamination is the unintentional presence of a particle or

particles in some place

Particle Contamination- Considerations
• particulate matter for release-testing samples is generally
heterogeneous in nature (i.e. randomly sourced)
- process and/or closure system related
• particulate matter for stability samples is generally

heterogeneous in nature with the potential to additionally contain
―stability‖ derived particles
- precipitates from solution/storage and solution/container
interactions
- degradents of the container system

Thank you for your attention
Questions?

Particle Detection &
Measurement
Analytical Measurement Options
Membrane Microscopy
Scott Aldrich
Outline
• Method Scope
• Equipment & Lab
• Operator Training
• Important Steps
• Blank Control
– USP 788 blank requirement
– Lab control blanks
• Counting and Sizing: full and partial
– Quantitative practices
– Additional information
• Calculations
• Limits
Method Scope
• Particulate matter in the product is isolated and counted:

– Visible to subvisible sizes of essentially solid particulate matter from the
product liquid will be retained on a membrane filter
– Counts of retained particles are conducted on the flat, dry membrane
• Full drying important
• Flatness ensured by tape or grease
• Membrane quality is key
• Apply the method for:
– USP 788 compliance
• Large-volume injections
• Small-volume injections
– In Development
• Specialty products, devices
• Low-volumes
Method Scope
• Membrane microscopic method is used when

– Articles fail LO due to - oils/air? - real particles?
– Articles cannot be tested meaningfully by light obscuration
• Viscous liquid
• Emulsions
• Dilution changes stability
• Coloration
• Insufficient volume available
– In some instances, light obscuration results are confusing
• Very high counts <25µm
• Few counts >25µm
• High variability among containers
• Individual monographs may specify use of only the microscopic
assay or may exempt articles from membrane microscopic analysis
Limited Samples/Sample Volumes
• Full container sampling of

– Single LVIs
– Single SVIs 25 to 100 mL
– Pooling of ten SVIs <25mL; however a minimum volume is
unnecessary
• No dilution or partial sampling needed
• Particle counts reported as
– Per container
– Per mL
• Is the best approach for low volume containers, and when
partial volumes are available – Counts are
– Direct yield from the package
– Or normalized per mL
Equipment & Lab
• The Test Environment - Zoning and Hoods

– Use Zoning: proceed from Common to Cleanest Area
– Use of Wet Hood/Dry Hood recommended
• Filtration equipment in the Wet Hood
– Glassware
– Filtration apparatus
– Drying/staging racks
– Collection Drain
• Microscope in the Dry Hood
– Lens systems provide 100  10x
– USP graticule calibrated to  2%
– Reliable X-Y stage
– Illumination
• Oblique 10-20
• Coaxial brightfield epi-illumination
Wet Hood Operations
Filtration Apparatus
• Use filter barrel suitable for the volume to be tested

– 25mm membranes offer a small area for full counting, having a minimum
EFA diameter of about 16mm
– 47mm membranes may be necessary, but consider the increased counting
area in the 36 mm diameter
– The filtration system is best determined by the requirements of the product
formulation, volume and package systems to be tested
• Large or small volume
• Water-like or viscous product
• Special sampling needs
• The filtration barrel may be plastic, glass, or stainless steel
• Use a flat filter support
– sintered glass is the flattest base
– Poor product flow ? Use a stainless steel screen base
• Get the apparatus really clean……..use it exclusively for particle isolation
work
Filtration and Membrane Apparatus
• Filtration facilitated with:

– Good vacuum source
– Filtration gangs (3x, 6x) with valves
– Pressurized dispenser (N2) filled with filtered water, capable of
delivering the rinse at 10 to 40 psi, preferably through a 0.22µm final
filter
• Test membrane filters
– ideally mixed esters of cellulose (depth), 25mm, non-gridded, dark
(black or dark gray) to provide contrast
– may be 47mm, may be gridded, may be alternate composition
• Membrane surface is the key feature
– ―Smooth porosity‖
– composed of a material compatible with the product
– ideally 0.45µm porosity
Product Filtration
• Use filtered water to pre-rinse both sides of the membrane.
• Center the rinsed membrane on the base, then place the cleaned filtration
barrel on top of the membrane/filtration base.
• Ensure that a distinct impression of the barrel flange is pressed onto the
membrane – it defines the EFA.
• Depending upon the dosage form being tested, prepare the sample, invert
to mix and then add to the wetted membrane in the waiting apparatus.
• After the last addition of product solution, the walls of the barrel are rinsed,
stopping when the volume falls below about one-fourth of the fill level.
Maintain the vacuum until all of the liquid in the barrel has passed through.
• Remove the filtration barrel from the base while maintaining vacuum, then
turn the vacuum off and remove the membrane with smooth forceps
• Place the membrane onto a prepared, labeled holder
– Double-sided tape on a microscope slide
– Grease in a plastic holder, etc.
• Protect the sample in a Petri dish or in a commercial slide holder
• Allow the membrane to air-dry in the HEPA enclosure with the cover ajar.
Blank Control
• USP 788 and 789 blank requirement is 50mL water,

limits of NMT 20 ≥10µm and 5 ≥25µm
– Lab control blanks recommended
• Represent total rinse volume+
• Use the same membranes and filtration apparatus
– Each Sample?
– Each Filtration Gang?
– Each Product Type?
– Each Operator?
Dry Hood Operations
Compound Binocular Microscope
• Ideally located in a separate HEPA

laminar flow hood
• Eyepieces:
– 10× magnification
– 1 eyepiece must accept and focus an eyepiece
graticule, better if both are focusable
• Mechanical stage to hold and traverse
entire filtration area of the 25 or 47-mm
membrane filter EFA
• Correct interpupillary distance choice
• Objective:
– 10× nominal magnification
– Planachromat or better quality
– Minimum numerical aperture: 0.25
– Compatible with episcopic illuminator
attachment
• Lens combination yields magnification of
100 ± 10x
Illuminators
• Two illuminators are required:
– External illuminator, adjust it first
• Incident oblique illumination
angle: 10° to 20°
• Is the primary illumination
– Episcopic bright-field illuminator
internal to the microscope
• Supplemental to the oblique
illumination
• Readily illuminates flat, thin flakes
– Both provide a bright, even
source of illumination
– Blue daylight filters provide:
• Truer colors
• Decrease operator fatigue
The Epi-illumination Source
• First – Adjust the oblique illumination. The method has used

oblique illumination to provide particle surface relief and to
enhance the shadow from larger particles since 1975
• Second - carefully adjust the epi-illumination so as not to
diminish the effect of the oblique incident light
• Use a known high-count sample to optimize illumination
• In 1995 - addition of coaxial/perpendicular incident
illumination (epi-illumination)
– Does not detract from standard oblique illumination effects
– Provides an additional resolution of thin, reflective particles
• The drawback for two illumination sources is diminished
resolution IF they are poorly adjusted/aligned, resulting in a
negative count bias
Microscope Alignment
• Microscope optical alignment and illumination are critical for

the success of this method
– Is it 10µm or 25µm ?...easy
– Tougher - differentiating a 10µm particle from the background
• Compromised by:
– Low Objective N.A.
– Unmatched ocular lenses
– Intermediate lenses not matching the microscope system
– Poor maintenance
– Poor optical alignment
• Operator fatigue results when not addressing these factors
Test Apparatus – USP Graticule
• Method-specific
design
• Linear scale
graduated in 10µm
increments
• Certified at
installation using a
NIST calibrated stage
micrometer
• Standardizes the
source of calibration
across manufacturing
sites
• Facilitates partial-
counting
Microscope Alignment
The operator examines the membrane preparation, locates a typical

array of particles,
•Adjust external, incident illumination at an oblique angle so that:
– An even ellipse of reflected light is visible on the membrane
– Illumination is evident through the eyepiece field of view
– Even, full illumination with good shadowing result
• Adjust the internal episcopic bright

field illuminator to yield an even
illumination at high setting on the
transformer control
• Then decrease the illumination until
one observes the evident shadow from
larger particles
• High reflectivity of flat, glassy particles
and distinct shadows of more equant
particles is evident when adjustment is
optimal
Important Steps
• Glassware restriction – use only for <788>, <789> operations

• Membrane quality and choice
– Most aqueous products – cellulosic esters
• Color-contrast, plain, 0.45µm, 25mm diameter
• Water-rinsed
– Certain non-aqueous, high concentration or emulsion products
• Color-contrast, plain, 1.0µm, 25-47mm diameter
• Water-rinsed/solvent-rinsed
• Polycarbonate, metal, polysulfone, etc.
• Membrane flatness – affix with
– Grease
– Tape
• Prepared sample hold
– storage in HEPA
– proximity
– age
Sizing and Tabulation
• The graticule facilitates estimation of

particle size through comparison to
open and filled circular areas with
10µm and 25µm diameters, or to the
linear scale in 10µm divisions
• Not necessary to directly
superimpose the circle or scale over
the particle; estimate the particle size
in the field of view by "mental"
comparison to circles
• The operator uses the linear
scale to:
– Calibrate
– Assist longest chord assessment
– Evaluate larger threshold sizes
Sweeping the Membrane under the GFOV
Microscopic Particle Count Test History
• Method in use since 1975

• Revised (Improved Microscopic Assay) method used by
several participating companies in 1992 collaborative study
on expired parenteral lots
– Counts via light obscuration and membrane methods were
requested by Pharmaceutical Manufacturer's Association —
Quality Control Particulate Sub-Committee
– Most participant products performed well in both methods
• Counts via light obscuration and membrane methods were
well within proposed limits
• The HIMA report demonstrated usefulness of method and
general acceptability
• Made official January 1995
Operator Training
• Success of microscope method is highly reliant upon operator

comfort and training
• An experienced microscopist tabulates particles as they move
across the field of view
• Provide a comfortable environment with little traffic, dimmed room
lighting, and no interruptions
– Interrupting the count is not acceptable: when interrupted or
delayed, microscopist should start anew
– Individuals have become ill at the microscope
• from the strain of poor optics and
• the disorienting effects of membrane movement
• A comfortable work environment and supportive management result
in operator satisfaction and improved counting precision
Operator Training
• To count or not to count ?

– “In performing the Microscopic Particle Count Test, do not attempt to size
or enumerate amorphous, semi-liquid, or otherwise morphologically
indistinct materials that have the appearance of a stain or discoloration on
the membrane filter."
• Counting performance with standards
– Bi-modal preparations of 15µm and 40µm
• Variance Expectation
– Individual  5%
– Intralab  10%
– Interlab  20%
• What about natural standards ?
– HIMA data
• NMT  100%
– Observation
• NMT  50%
Counting Anomalies
• If the retained material appears to have three dimensions, it

is a particle, however…
– “In performing the Microscopic Particle Count Test, do not attempt to size
or enumerate amorphous, semi-liquid, or otherwise morphologically
indistinct materials that have the appearance of a stain or discoloration on
the membrane filter." Why Not?
• Investigate materials that:
– Show little or no surface relief
– Present gelatinous or film-like appearance
– May indicate system/condition out of control
• If persistent, the material may be indicative of:
– Product change
– Instability
Blank as System Suitability
• Consistency of blank count is important factor

– A day of use 50mL blank count assures the environment and
operations are acceptable. Failure to consistently pass the
blank test, or variation in the count, or a blank count trending up
may indicate problems with:
• Method equipment preparation
• Lab environment
• Operator training
• Operator fatigue/lack of concentration
• Operator:
– Attains better method control by pairing blank and sample
membranes – one blank/one sample
– Runs blank membrane just before using the filtration apparatus
for the sample
Blank as System Suitability
• 50mL water - Blank Limit is

– NMT 20 particles ≥ 10µm
- and -
– NMT 5 particles ≥ 25µm
• <1788> Recommendation for Chapter <789>

– Blank Limit
• NMT 10 particles ≥ 10µm
• - and -
• NMT 1 particle ≥ 25µm
• - and -
• No particles ≥ 50µm
– Why?
Counting and Sizing: Full or Partial
• Quantitative practices
– Full count for all samples is preferred
– Partial count for high level of retained particles - above 1000
• Additional information may be sought
– Size thresholds
– Particle type assessment
Sizing and Tabulation
• Directly or mentally compare circle diameter or linear scale to

the particle to indicate appropriate size category
– Is the size equivalent circular diameter? Or longest chord?
– Intention of HIMA scientists to correlate MM count to LO
• Equivalent circular diameter assessment will best correlate particle size by
microscopy to light obscuration
– What are we after?
• There is a bias of size by LO when particles are aspherical
• Consider a 70µm needle:
– Needle of 70μm x 1μm x 1μm has
• Equiv. Cir. Dia. of 55μm2 and Eq. Dia. of 8μm
• Thus some may not count it (<10μm)
Total Count Procedure
• The microscopic test is flexible in that it can count from very few to
many particles over the entire membrane
• The operator counts all particles on the membrane surface or
effective filtration area (EFA). The graticule field of view (GFOV),
defined by the large circle of graticule, may be ignored and the
vertical crosshair used as the counting plane.
• Scan the entire membrane in a path that adjoins but does not
overlap the first scan path
• Repeat the procedure, moving the membrane under the
objective until all particles on the membrane are counted
• Tabulate:
– Total number of particles that are ≥10µm - 25µm = P10-25
– And the number that are ≥ 25µm or larger = P25
Total Count <788>
• Large Volume Injections

– Calculate particles per mL for each threshold counted
– Report total ≥10µm (P≥10-15µm + P≥25µm)  Volume
– Report total ≥25µm (P≥25µm)  Volume
• Small Volume Injections
– Calculate particles per container for each threshold
counted
– Report total ≥10µm (P≥10-15µm + P≥25µm)  # Containers
– Report total ≥25µm (P≥25µm)  # Containers
Partial Count Procedure
Performing a Partial Count of Particles on a Membrane

• When many particles are present, particles in random portions of
the membrane surface may be counted. The result is normalized to
the whole sample.
• Analyst must ensure an even distribution of particles present on the
membrane
– Assessed by 50x exam and/or qualitative scanning at 100x to
assess content of particulate clumps: few should be present
• Count 10µm and larger particles in one GFOV at the edge of the
filtration area, and one in the center of the membrane
• The number of 10µm or larger particles should not differ more than
2x between them
Reject the membranes failing these criteria

Partial Count Procedure
• Partial count is allowed; however…

– Undefined in <788>
– Defined in <1788>
• If 1000 or less particles are present, the membrane should
be fully counted
• If the count difference between a GFOV at center and one at
the edge is <2X, then…
• Count 20 GFOV for 25mm (16mm EFA) membranes
• Count 100 GFOV for 47mm (37mm EFA) membranes
Partial Count Diagram - Random
Count 2 GFOV, center + edge.

Vary NMT 2x
Count at least 20 GFOV in the EFA

Partial Count
• Large Volume Injections

– Calculate particles per mL by the formula:
P size x Total Area/Area Counted x Volume

Particles ≥10 µm = (P10-25 + P25) AT/APVmL
• Small Volume Injections
– For pooled <25mL units or single units ≥ 25-100mL of a
small-volume injection, calculate the number of particles per
unit by the formula:
P size x Total Area/Area Counted x No. Units
Particles ≥10 µm = (P10-25 + P25 )AT/Apn
Particle Limits
The injection meets the requirements of the test if the average number of particles
present in the units tested does not exceed the appropriate value listed in Table 3
Table 3: Microscopic Method Particle Count, Parenteral

Product Type ≥ 10 µm ≥ 25 µm
A. Large-Volume Injections 12 2 per mL
B. Small-Volume Injections 3000 300 per container
The ophthalmic solution meets the requirements of the test if the average
number of particles present in the units tested does not exceed the
appropriate value listed in Table 4
Table 4: Microscopic Method Particle Count, Ophthalmic

Size ≥ 10 µm ≥ 25 µm ≥ 50 µm
Limit 50 per mL 5 per mL 2 per mL
Microscope Attributes
• Value of Microscopic Method:

– Almost any volume may be used
– We observe the actual particle in a dry state
– Particles may be retained for study
– Common liquid artifacts not an issue
– An important evaluation for products in development
• Count in additional size thresholds
• Almost any volume or device may be tested
• Point source particles are evident and retained
Using DLS and SPOS to assess the quality
and safety of pharmaceutical injectables
in the nanometer to submicron size range
David F. Nicoli, PhD

Vice-President, R&D
Particle Sizing Systems, LLC
Santa Barbara, California
Port Richey, Florida

Particle Detection & Measurement
Rockville, MD December 8-10, 2010
Outline of Presentation
 Dynamic Light Scattering (DLS) – How it works

 DLS applied to protein aggregation
 Single-Particle Optical Sizing (SPOS) – How it works
 Expansion of SPOS: Light Scattering (LS) added to
Light Extinction (LE) for counting/sizing, 0.5-400 um
 SPOS applied to injectable fat emulsions, USP <729>
 NEW SPOS Techniques: Focused Extinction (“FX”) and
Focused Scattering (“FX-Nano”) – How they work
 FX-Nano applied to proteins and other nano particles,
permitting counting/sizing down to 0.2-um and below
Dynamic Light Scattering (DLS): obtaining the
Diffusion Coefficient, D, from the fluctuations
(“noise”) in the Scattered Light Intensity, IS
 Laser light (wavelength ) is focused into a sample

cell containing particles suspended in liquid
 The particles scatter light (in all directions) because

their refractive index differs from that of the liquid
 For very small particles (diameter d << ) – the

Rayleigh region – the scattered intensity IS is approx.
independent of scattering angleS (after correction
for the effective scattering volume,  1/sinS)
 IS = f(nP ,nS) (MW)2 I0 or,
 IS = g(nP ,nS) V2 I0 i.e., IS = g(nP ,nS) d6 I0

 Consider only TWO particles in suspension that are

illuminated by the incident laser light beam
 The illuminated particles will re-radiate (“scatter”)

light in all directions, having essentially the same
wavelength  (“quasi-elastic” light scattering, QELS)
 The scattered waves “interfere” everywhere in

space, and in particular at the point of detection
 The net detected intensity IS will depend on the

relative phases of the two waves, which depends on
the difference in optical path lengths, from the laser
wave front to each particle and then to the detector
 The scattered light intensity produced by two particles

depends on the relative phases of the scattered waves
 Intensity IS varies from a MAXIMUM value, for total

constructive interference (optical path lengths differ
by N), to ZERO, for total destructive interference
(path lengths differ by (N+1/2) (N = integer)
 The suspended particles constantly undergo random

Brownian motion, or diffusion, caused by unbalanced
collisions of the surrounding water molecules
 This random-walk diffusion causes the difference in

optical path lengths, and therefore IS , to fluctuate
 The fluctuations in IS are random over short times but

carry very useful information over longer times
 DLS is based on the mutual diffusivity, or Brownian

motion, of particles suspended in a liquid
 In the case of only two particles, the scattered light

intensity fluctuates between a maximum value and
zero, as the particles diffuse in the liquid
 The useful diffusivity, D, is buried in the “noisy” IS

signal and can be obtained by the mathematical
operation of Autocorrelation:
C(Dt) = < IS(t)  IS(t -Dt) >
where IS measured at the current time t is multiplied

by IS measured at the earlier time, t -Dt, for a given Dt
and < > denotes an ensemble (i.e., running) average
 C(Dt) is computed continuously for each of many

different values of Dt, ranging from “small” to “large”,
relative to the characteristic time t of the fluctuations
in IS … where t depends on the diffusivity, D
 Small particles: large D, requiring small values of Dt

Dynamic Light Scattering (DLS): in the case of
uniform particles, C(Dt) is a simple exponential
function, from which D is easily obtained
 Autocorrelation of the “random” noise in IS yields a

simple result for C(Dt) in the case of uniform particles:
a single decaying exponential function:
C(Dt) = A exp(-Dt/t) + B (where B = <IS(t)>2 )
where t is the characteristic time of the fluctuations,

related to D by 1/t = 2DK2
and K, the “scattering wave vector”, is given by

K = (4pn/) sin(/2)
where n = refractive index of the suspending liquid
 What remains: obtaining the hydrodynamic radius, RH ,

from D (in turn, obtained from t)
uniform particles, C(Dt) is a simple exponential
function, from which D is easily obtained
 Particles of uniform size (0.09-um here) yield a nearly

perfect decaying exponential function for C(Dt)
Dynamic Light Scattering (DLS): the particle
size (hydrodynamic radius RH ) is obtained from
the diffusivity, D, using the Stokes-Einstein eq’n
 The desired quantity RH is connected to the measured

quantity D by the Stokes-Einstein relation,
D = kT/6pRH
where k = Boltzmann’s constant

T = temperature (0K)
 = viscosity of the liquid (cP)
 DLS is an “absolute” technique, in the sense that the

measured diffusivity D depends only on the particle
size, and not its physical properties (e.g., refractive
index) – provided the particles are of uniform size
Dynamic Light Scattering (DLS): for single-peak
(e.g., gaussian or log-normal) PSDs, the simple
method of cumulants offers adequate analysis
 It all depends on the complexity of the PSD …

 Most “real” samples of interest contain a distribution
of particle sizes, rather than mostly a single size
 Hence, the intensity autocorrelation function, C(Dt), is
no longer a single decaying exponential function
 Instead, an algorithm is needed to “invert” C(Dt), in
order to obtain (approximately) the desired PSD
 The simplest algorithm: the method of cumulants
CR(Dt)  ln[C(Dt)-B] “reduced” autocorrel. function
CR(Dt)  a0 + a1Dt + a2(Dt)2 + … (least-squares fit)
i.e., a polynomial (quadratic) fit to the log of C(Dt)
(e.g., gaussian or log-normal) PSDs, the simple
method of cumulants offers adequate analysis
 The particle size distribution (PSD) corresponding to

the previous C(Dt) is nearly uniform: diameter  93-nm
PSDs (e.g., gaussian or log-normal in shape),
the simple method of cumulants is effective
 Example: micelles of CTACl (0.05M) + NaCl (0.1M)

 The resulting Intensity-wt “Gaussian” PSD yields:
mean diam 4.9 nm (D = 9.6 10-7 cm2/s), CV = 20%
more complex PSDs (e.g., multimodals) a more
sophisticated “inversion” algorithm is needed
 The cumulants algorithm fails for “polydisperse” PSDs
 In general, C(Dt) contains many decaying exponential

functions, each representing a different diffusivity:
C(t) = A[ I g(DI) exp(-DIK2Dt)]2 + B
where sum I is over all diffusivities, DI and g(DI) is
the intensity-wt distribution of diffusivities, from
which the PSD (distribution of RI) is easily obtained
 In general, g(DI) is strongly dependent on RI :

g(DI) = f(RI)  VI2  GI2  f(RI)  RI6  GI2(RI)
where GI2(RI) is the intra-particle Mie “form-factor”
 For very small particles, RI << and GI2  1 (Rayleigh)

Dynamic Light Scattering (DLS): a good (high-
resolution) inversion algorithm is needed to
analyze “difficult” PSDs (e.g., multimodals)
 Example: PSL bimodal, 100-nm + 170-nm

 Many DLS instruments are unable to resolve such a

closely-spaced bimodal, yielding only a gaussian PSD
 Example: Protein BSA – 0.5% (5 mg/mL)

 Peak #2: agglomerate “tail” (56% of total intensity)
 Example: Protein BSA – 0.5% (5 mg/mL)

 Peak #2: agglomerate “tail” (7% of volume, or mass)
 Example: Protein BSA – 0.05% (0.5 mg/mL)

 Peak #2: agglomerate “tail” (59% of total intensity)
 Example: Protein BSA – 0.05% (0.5 mg/mL)

 Peak #2: agglomerate tail (only 3% of total volume)
 Ex: Protein “X” (M.W. 50K) – 0.7 mg/mL, 4-deg C

 Peak #1: native protein; #2: agglomerates (83% int.)

 Agglomerates represent only 0.5% of total volume

 Agglomerates now account for nearly 99% of total int.

 Agglomerates now account for almost 5% of volume
 Ex: Protein “X” (M.W. 50K) – 4-deg C vs 37-deg C

 The simple gaussian PSDs show only the crude trend

 Clear: the large shift in intensity by the agglomerates

 Clear: the growth in the amount/size of agglomerates
 Ex: Drug-carrying micelles + lactose (high concen.)

 The simple 2-parameter “Gaussian” (cumulants) PSD
yields a very poor fit to C(Dt) (huge chi-squared value)

 C(Dt) shape indicates a high polydispersity, caused by
the much smaller 2nd component (lactose molecules)

 The PSD reveals the lactose (“A”, 0.8nm), micelles
(“B”,7.1 nm) and misc. “contamination” (“C”, 100nm)
A
B
C

 The lactose (MW 342, est. size 1.1 nm) provides (est.)
43% of IS , corresponding to >99% of the total volume
B C
Single-Particle Optical Sizing (SPOS), Part I:
Conventional Light Extinction (LE) is effective
for counting/sizing particles above  2-um
 Light Extinction (LE): light refraction (incl. scattering)

 Momentary reduction in the transmitted light flux
 Particle momentarily “blocks” some incident light

 Fraction of light blocked  2 x (particle area)/aw
 Typical LE Sensor Response for a light-extinction

sensor (PSS Model LE400)
 Corresponding Calibration Curve for the LE sensor

 Note the drop in the response below 2-um
SPOS(LE): ideal for measuring the large-droplet
“tails” of injectable fat emulsions (USP <729>)
 Fat emulsion “LD” (20% w/v)
(Auto-inject 2 mL into 30 mL, auto-dilute 50)

SPOS(LE): ideal for measuring the large-droplet
“tails” of injectable fat emulsions (USP <729>)
 Fat emulsion “LD” (20%): 1.63 x 105/mL > 5 um

PFAT5 = volume fraction > 5 um = 0.017% (pass)
USP<729>
Single-Particle Optical Sizing (SPOS), Part II:
Light scattering (LS) provides added sensitivity,
needed to count/size particles down to 0.5-um
 Light scattering (LS)

 Momentary flash of scattered light occurs when a
particle passes through the same “view volume”
 Problem: the light scattering response has a very

limited dynamic range, due to strong size dependence
 Light Extinction (LE)

 Advantage #1: large upper size limit, and therefore
large dynamic size range
 Advantage #2:  independent of refractive index
 Disadvantage: limited sensitivity (minimum size)
 Light Scattering (LS)

 Advantage: high sensitivity (lower size limit)
 Disadvantage #1: limited upper size limit, and
therefore relatively small dynamic size range
 Disadvantage #2: dependent on refractive index
Single-Particle Optical Sizing (SPOS), Part III:
Combine LE and LS (“LE+LS”, Pat.) to obtain
the best of both worlds (sensitivity, size range)
 Separate detectors (same view volume) for LE and LS

 The separate signal pulses produced by the LE and

LS detectors are added to form a composite pulse
 The resulting LE+LS Response (calibration curve)

has a large dynamic range:  0.5 um to 400 um
Combine LE and LS (“LE+LS”, Pat.) to obtain the
best of both worlds (sensitivity, size range)
 High sensitivity and large dynamic size range

 Highest possible resolution (single-particle) -- e.g.,
0.7+0.8+1.3+2+5+10+15+20+50+100+200 um
“Expanded” SPOS (LE+LS): ideal for measuring
the large-droplet tail in injectable fat emulsions
 Injectable fat emulsion: typical PSD “tail” (> 0.56 um)

“Expanded” SPOS (LE+LS): ideal for measuring
the large-droplet tail in injectable fat emulsions
 Same injectable fat emulsion tail,

exp. 100X: many particles > 2 um
Question: How can background contamination
counts – especially problematic at the lowest
size limit (0.5-um) – be effectively reduced?
 Background contaminants increase greatly with

decreasing size – a general fact of SPOS life
 Example: 10% sample concentration (r = 1) and
sensor coincidence limit of 10,000/mL
 10-um: 1.91 x 108/mL … must dilute  20,000X
 2-um: 2.39 x 1010/mL … must dilute  2,400,000x
 0.5-um: 1.53 x 1012/mL … must dilute  150,000,000x
 Solution: Make the “view volume” smaller, and also
avoid looking at all the sample suspension
 Result: “Focused Extinction” (“FX”)
Focused Extinction (FX) vs conventional LE:
higher sensitivity (and working concentration)
 Drastic reduction in the size of the “view volume”
 Much lower minimum

particle size limit (much
higher sensitivity)
 Much higher particle

coincidence limit (sample
concentration)
Focused Extinction (FX) vs conventional LE:
higher sensitivity (and working concentration)
 Pulse height depends on particle size and trajectory
 The higher the fraction of

incident light “blocked”, the
larger the pulse height, DVLE
 Particles of a given size

yield a wide range of pulse
heights, depending on their
path through the light beam
 The resulting pulse height

distribution (PHD) must be
deconvoluted to obtain the
particle size distribution (PSD)
Question: how, in practice, can particles as
small as 0.2-um be counted/sized easily?
Not by conventional SPOS …
 The numbers are against us traditional SPOS …

 Example: 10% sample concentration (r = 1) and
sensor coincidence limit of 10,000/mL
 0.5-um: 1.53 x 1012/mL … must dilute  150,000,000x
 0.2-um: 2.39 x 1013/mL … must dilute  2,400,000,000x
Such large dilutions are impractical (background counts)
 It gets worse: we haven’t yet considered the needs of
Lord Rayleigh (a law … not a choice!)
Question: how, in practice, can particles as
small as 0.2-um be counted/sized easily?
Not by conventional SPOS …
 Rayleigh’s Law: for small particles (<<  ) the

scattering intensity falls as the 6th power of the size.
 1st issue: insufficient light (very small signal …)
 2nd issue: narrow size range (amplifier saturation)
 Going from 1- to 0.5-um, the intensity drops 64X

 Going from 0.5- to 0.2-um, the intensity drops 244X !
 Conclusion #1: We can’t achieve 0.2- to 1-um using

just one amplifier gain range (unlike light-extinction)
 Conclusion #2: We need very high laser intensity
 SOLUTION: New SPOS technology (“FX-Nano”)

FX-Nano: focused scattering is used to extend
the lower size limit of the focused-extinction
(FX) sensor, from 0.56-um to below 0.2-um
 Where there’s extinction, scattering is closeby ...

 Same beam, similar “view volume” as used for the FX
FX-Nano: provides particle size analysis down to
0.2-um, with single-particle resolution and high
counting statistics, in a relatively short time
 0.2-um PSL (High-gain Scattering)

FX-Nano: permits particle size analysis down to
 3-modal: 0.2-, 0.24-, 0.30-um (High-gain Scattering)

 0.3-, 0.34-, 0.4-, 0.5-, 0.6-um (Low-gain Scattering)

 4-modal: 0.7-, 1-, 1.36-, 2-um (Focused Extinction)

FX-Nano: provides the entire PSD for submicron
injectable fat emulsions, with unprecedented
(single-particle) resolution and accuracy
 Emulsion “LD”, 1K dilution; high+low-gain scattering

 Emulsion “LM”, 1K dilution; high+low-gain scattering

 Emulsion “LN”, 1K dilution; high+low-gain scattering

 Compare the three PSDs (high+low-gain scattering)

 Same comparison, but with 10X expansion

 Same comparison, but with another 10X expansion

Interesting: compare the submicron “tails”
obtained from FX-Nano with the > 2-um tails
obtained using conventional LE technology
 Fat emulsion “LD”: result from standard LE (> 2 um)

Interesting: compare the submicron “tails”
obtained from FX-Nano with the > 2-um tails
obtained using conventional LE technology
 Same trend found for PSDs obtained by LE (> 2 um)
Emulsion LD:
1.01 x 106/mL
Emulsion LM:
4.47 x 106/mL
Emulsion LN:
7.37 x 106/mL
FX-Nano: powerful tool for analyzing submicron
agglomeration of proteins and other biologicals
in injectable pharmaceutical formulations
 Vaccine #2, agglomerate “tail” above 0.2um (200nm)

 Vaccine #3, similar “tail” > 0.2um, but different shape

 Vaccines #2, #3, #4: different large-particle tails

 Vaccines #2, #3, #4: different large-particle tails, X10

 Vaccines #2, #3, #4: different large-particle tails, X100

 Vaccines #2, #3, #4: cumulative PSD, log scale

Image Analysis for Particle
Size, Shape and other Valuable Information
Kevin Powers
Particle Detection and Measurement
December 10, 2010
Always Look at your Particles under a Microscope!
No Exceptions!
• Particle Count
• Particle Size Distribution
• Particle Shape
• Other information
With Respect to USP 788/789
Light Obscuration – does not capture images (flow microscopy)

Count/Size
Refractive index
Droplets
Aggregation (you don’t know)
Microscopic Examination - Membrane Method
can capture images
Emulsions/droplets/fragile aggregates
Count Size/shape
Aggregation is apparent
Additonal information
Static Image Capture
Static Image Capture (can be manual or automated)

• More flexibility to adjust lighting and optics
• Sample can be saved for reanalyis
• Sample can be mounted for additional testing (SEM/EDS)
• Sampling/dispersion is a big issue
• Shape bias?
INTERNATIONAL ISO
STANDARD 13322-1
Particle size analysis — Image analysis methods —
Part 1:
Static image analysis methods
Analyse granulométrique — Méthodes par analyse d'images —
Partie 1: Méthodes par analyse d'images statiques
Dynamic Image Capture
Dynamic Image Capture

• Generally optimized adjustment lighting and optics
• Analysis Parameters designed by experts
• In motion – in focus?
• Images are saved for further inspection/analysis
• Sampling/dispersion can still be an issue
• Shape bias?
INTERNATIONAL ISO
STANDARD 13322-2
Particle size analysis — Image analysis methods —
Part 2:
Dynamic methods
Analyse granulométrique — Méthodes par analyse d'images —
Partie 2: Méthodes par analyse d'images dynamiques
Issues in Microscopy and Image Analysis
Image analysis has many of the same issues as other
Particle sizing methods such as sampling and dispersion
• Sampling methods
• Dispersion
• Image Capture and Resolution (diffraction limit and camera optics)
• Depth of field and focus
• Calibration
• Number of particles sampled/Statistical Analysis
• Software/Man-machine interface National Bureau of
Standards estimates one
must count 10,000
individual pictures
for statistically valid
results!
A L Dragoo et al “A critical assessment of
requirements for ceramic powder characterisation”
Advances in Ceramics Vol 21 Ceramic Powder
Science (1987). The American Ceramic Society Inc.
Training
Operator/training Instrumentation/software
• The instrument you use is only as good as the sample preparation

• The image you capture is only as good as the optics/instrument
• The information you collect is only as good as the image you capture
• The software is only as good as the information you analyze
• These are all tools that are only as good as the analyst using them
Resolution
Resolution is a function of both the microscope and the Image capture device. The
microscope’s resolving power is dependent on its optics and physical principles.
The image can also be captured or digitized at varying resolutions.
• Resolving power of Microscope = Diffraction limit defined by
d 0.6 
no sin 
• In general optical size and shape information is difficult to obtain below 3-5mm
• Electron microscopes can resolve much smaller particles with greater
depth of field but suffer from lower contrast levels STED
*See NIST Practice Guide SP960-1
Eva Rittweger, Kyu Young Han, Scott E. Irvine, Christian Eggeling, Stefan W. Hell (2009). "STED microscopy reveals crystal
colour centres with nanometric Resolution.". Nature Photonics 3: 144–147.
Magnification 2- 10 mm?
Recommended Magnifications for selected size ranges
(ISO 13322-1 Static Image Analysis)
Sample size range Resolution (mm) Field size (mm) Total Magnification
0.1 – 1.0 0.01 10 x 10 20Kx
0.3 – 3.0 0.03 30 x30 8Kx
1 –10 0.1 100 x 100 2Kx
3 – 30 0.3 300 x 300 800x
Optical Microscope 500x

How many particles should I Count?
This depends on the particulate system, sampling efficiency and the

degree of accuracy required. Masuda and Iinoya (1971) developed
A simulation which estimates the number of particles required for a given
error in a perfect log-normal particle distribution (see extract).
 GSD n*(MMD) n* (MVD)

0.05 1.10 585 131
1.20 2939 528
1.30 8526 1247
1.40 19026 2363
1.50 36007 3956
1.60 60811 6092
Where  = relative error, GSD= Geometric Standard Deviation,

And n*= number of particles (MMD=mass median diameter,
MVD = mean volume diameter)
Calibration
Calibration of the image should be accomplished with a traceable

calibration standard such as a certified stage micrometer or image
reference graticule. A new calibration should be accomplished with
each new magnification.
Depth of field and focus

All particles in field of view should be in focus and magnification Selected to give
the appropriate depth of field. Particulate systems with large size ranges become
difficult to size optically.
Image Analysis
Can be used to derive

• Size information
• Shape information
• Morphology
• Crystallinity
• Color
• Degree of agglomeration Number Histogram for GCP 3-30
2000
1800
Population 10293
120.00% Morphology
100.00%
1600
Comparison of Laser Diffraction Data and Optical Analysis
1400 Frequency for GCP 3-30 Carbon (10196 particles)
80.00%
Cumulative % 14
1200
Frequency
1000 60.00% 12
800
10
Differential Volume %
40.00%
600
8
400 Optical GCP 3-30 carbon
20.00%
Instrument A
200
6
0 .00%
4
2
8
4
2
1
9
2
6
1
3
6
7
3
3
8
2
6
6
9
07
92
55
00
4
04
11
17
28
45
72
14
83
65
42
.8
.8
.0
.9
.4
1.
4.
9.
3.
6.
12
20
0.
2.
11
18
30
47
76
12
19
30
49
78
0.
0.
0.
0.
0.
0.
1.
1.
4.
7.
Diameter (microns)
2
0
1 10 100
Shape Size Diameter (microns)

Measurements
A wide variety of measurements are possible using automated

image analysis software. Primary measurements include:
• The Area of each object

• Maximum dimension (maximum Feret diameter)
• Minimum dimension (minimum Feret diameter)
• Area equivalent diameter (derived)
• Shape factors  (derived)
X MAX
4  Area 
XA  X MIN
p
USP 788
The projected area of each particle can be converted to the
area equivalent circular diameter, xAi.
4 Ai X = 10µm
x Ai 
p A=78.54 µm2
Is 100x enough?
Image Analysis Software – See like a computer
Image analysis software can differ in capabilities and may vary in the way
operations are defined. Operators have a wide variety of image
manipulation, sorting, and measurement options which can aid or hinder
statistical reliability.
• Border definition
• Threshold levels Contrast!
• Area definition (holes – euler number)
• Filters and image processing operations
• Splitting agglomerates or touching particles
• Size and length Measurements
• Discarding agglomerates
• Roundness and shape factors
• Data Presentation and Statistics
Area and aggregates
IMAGE PRO PLUS

VER 6.2
Raw particle Count/Size data sorted by aspect ratio to

identify agglomerates
Split aggregates are always smaller
Same data as in previous slide with agglomerates split by limited

watershed filter and sorted by roundness
Comparison Image Analysis and Light Scattering
Number Histogram for GCP 3-30
Population 10293
2000 120.00%
1800
100.00%
Optical Image Analysis
1600
1400 Frequency
on Spherical Reference
80.00%
1200
Cumulative %
Powder 3-30 microns
Frequency
1000
800
60.00%
vs Laser Diffraction
40.00%
600
400
20.00% Comparison of Laser Diffraction Data and Optical Analysis
200 for GCP 3-30 Carbon (10196 particles)
14
0 .00%
4
2
8
4
2
1
9
2
6
1
3
6
7
3
3
8
2
6
6
9
07
92
55
00
04
11
17
28
45
72
14
83
65
42
.8
.8
.0
.9
.4
1.
4.
9.
3.
6.
12
20
0.
2.
11
18
30
47
76
12
19
30
49
78
12
0.
0.
0.
0.
0.
0.
1.
1.
4.
7.
Diameter (microns)
10
Differential Volume %
Number distribution is derived 8
Optical GCP 3-30 carbon
from equivalent area diameter and 6

Instrument A
converted to equivalent volume 4
distribution 2
0
1 10 100
Diameter (microns)
Comparison Image Analysis and Light
Scattering (1-10micron spheres)
MBP 1-10 Glass Powder
(4038count)
12
Representative Example 10 Instrumet A

Instrument B
Optical Analysis
Relative Volume %
8
Laser Diffraction PSD vs

Optical Image Analysis for 6
MBP 1-10 micron Glass

4
0
0.1 1 10 100 1000
Diameter (microns)
MMD GSD
Instrument A 4.73 1.38
Instrument B 4.77 1.42
Optical 4.46 1.46
Shape
INTERNATIONAL ISO
STANDARD 9276-6
Representation of results of particle size analysis —
Part 6:
Descriptive and quantitative representation of particle shape and
morphology
Représentation de données obtenues par analyse granulométrique —
Partie 6: Description et représentation quantitative de la forme et de la
morphologie des particules
Pop Quiz EMA 5008 - What is Spherical?
A Little Help?
More Help
54 93 42 77 81 7 58 36 39 105 61 46 59
What percentage of these
particles are spherical?
11 68 62 111 41 74 8 82 38 89 53 116 12
102 76 72 26 3 22 6 71 44 32 47 37 73 94
33 60 114 80 29 40 21 63 70 66 24 3 9 These Particles are arranged

in increasing order of Roundness.
56 90 99 1 45 13 69 24 27 14 23 34 20 67
Where would you draw the
line for a ―spherical‖ particle?
18 30 98 17 16 110 115 85 75 113 31 25 57 117
2 19 55 96 84 107 87 35 15 91 50 108 43 88 86 10 5 48 109

Even More Help – but Decision Time!
Obj.# Roundness Aspect Radius Ratio Axis (major) Axis (minor) Obj.# Roundness Aspect Radius Ratio Axis (major) Axis (minor)
54 3.115319 2.100244 2.284157 124.7807 59.41248 83 1.120581 1.156105 1.327115 59.86917 51.78526
93 2.886467 2.441599 28.8902 239.5288 98.10325 22 1.120206 1.010085 1.082241 81.35082 80.53861
42 2.282457 1.272543 1.877138 85.54147 67.22089 6 1.120192 1.060499 1.233071 82.1179 77.43324
77 2.042687 2.267449 7.431973 141.7795 62.52817 71 1.119002 1.029821 1.10955 57.59219 55.92448
81 1.960888 2.287875 5.868821 133.8645 58.51038 44 1.118028 1.052749 1.140191 66.48984 63.15833
7 1.762736 2.080468 3.665487 161.2341 77.49896 32 1.117717 1.17961 1.354659 60.3385 51.15122
58 1.524376 1.203358 1.430892 80.40601 66.81803 47 1.113584 1.058597 1.120571 65.58897 61.95838
Appendix 1 – Quantification of Results by Image Analysis
36 1.520728 1.127447 1.453993 64.42514 57.14248 37 1.108 1.244848 1.559374 60.88822 48.91218
39 1.472157 1.368183 2.128397 114.2203 83.48321 73 1.107493 1.169396 1.225367 96.77805 82.759
105 1.439647 1.910014 2.355201 69.82136 36.55541
In this project
94
the figure of
1.106367 1.036575
merit is the sphericity
1.080746
of the
77.71529
particles as
74.97311
61 1.417583 1.36837 2.142832 106.004 77.46737 determined
33 by image analysis.
1.10537 1.063748 There are several
1.233824 ways
79.04694to measure
74.30983the
46 1.374841 1.448694 2.149516 111.4455 76.92822 roundness
60 of a 21.1032
dimensional image. 1.223357
1.133718 One of the 80.04663
most popular70.60544
is take
59 1.329038 1.684423 1.992413 128.5577 76.3215 the perimeter
114 of the object squared
1.098743 1.017071 and divide
1.067636 it by the
75.324654p times the
74.06034
11 1.303918 1.348388 1.611476 74.1125 54.96377 Area. For
80 a perfect
1.098064circle this ratio will
1.063904 equal one69.80113
1.141651 and the ratio
65.60849
68 1.272464 1.015281 1.124969 76.46819 75.3173 increases
29 as the two dimensional
1.097646 1.065271 image becomes 77.78089
1.124362 less circular.73.01509
Two
62 1.25648 1.113632 1.224073 61.67119 55.37841 other pertinent
40 measures of
1.095572 1.036168shape are aspect
1.079312ratio and radius
93.63419 ratio both
90.36581
111 1.25355 1.298196 1.552185 98.55019 75.91319 of which21 will 1.095439
equal one for a perfect circle.
1.062765 1.157252Image Pro Plus image
74.66385 70.25433
41 1.249221 1.109601 1.37046 70.76276 63.77316 analysis63software provides
1.094615 a variety of
1.011602 automated60.28266
1.105511 measurements of
59.59129
74 1.213959 1.021676 1.215329 75.80112 74.19292 which a70selected few are1.036232
1.093878 defined below.1.091978 62.72553 60.53235
8 1.210175 1.513895 1.853862 89.87395 59.36606 Roundness
66 1.091402 1.133501 2
Perimeter /4pArea
1.372003 55.40286 48.87764
82 1.193472 1.028305 1.115213 68.44372 66.55976 Radius24 ratio 1.090975 1.00905 1.11609
ratio between max 56.75483
radius and min56.24579
radius
38 1.18814 1.175315 1.436836 99.98438 85.07026 Max radius3 1.090608 maximum distance between object’s
1.067525 1.150758 88.19406 82.61546
89 1.177157 1.00498 1.068111 66.50791 66.17836 9 1.08827 1.031955 centroid1.084828
and outline92.40705 89.54562
53 1.162123 1.173624 1.408054 81.65877 69.57833 56
Min radius 1.086873 minimum distance 62.04723
1.089144 1.166333 between object’s56.9688
116 1.160043 1.068724 1.13158 68.85111 64.42365 90 1.086778 1.050651 centroid1.101041
and outline81.41013 77.4854
12 1.148352 1.434366 1.48953 121.4706 84.68594 99
Perimeter 1.086586 1.009525 1.112539 56.04374
length of objects outline 55.51497
102 1.131074 1.014715 1.136911 88.29057 87.01022 1
Aspect ratio 1.083822 1.04687 1.114407
ratio between major74.67649 71.33312
axis and minor axis
76 1.129261 1.071567 1.125137 74.11531 69.16534 45 1.083756 1.037442 1.153889 53.12645 51.20908
of ellipse equivalent to object
58.077 13 1.082524 1.039362 1.112905 68.6231 66.02428
72 1.12653 1.002854 1.094935 58.2436 82 69 1.080605 1.005295 1.080238 73.6284 73.2406
Other Shapes
Single Crystal Gold Platelets

Particle Analysis on 200nm ANOPORE™ Filter
Passes USP 788?

3-D Shape Analysis (Stereo microscopy and Tomography)
Topology
3D Shape information
from SEM Analysis
Alicona MeX – surface metrology
What is Other Information? - Example
Spray Dried Kaolin and the case of the

Phantom fines
Complex and Dynamic Systems in the context of USP788
How many particles/aggregates 10μm >25 μm ?
USP
788
Water Formulations
PBS
Drug Bloodstream
pH Isotonic Buffers Blood
Culture media Salts Cells
Surfactants Proteins/Opsonins
Rheology Modifiers Filtration
Shear!
How do Particles Behave?
Image Analysis for Complex systems
Primary Particles Coated, 15min, Liver Coated, 24h, Liver
50nm 200nm 1um
Images by Kerry Siebein

Conclusions
• Image Analysis is a crucial part of particle characterization in

general and for drug delivery in particular.
• The same issues with other Particle Size Analysis methods also
apply to sizing by Image Analysis - plus a few more!
• Image analysis is the only quantitative method for shape

determination - but it has limitations!
Imaging Techniques
“The Gold Standard – or are they?”
From hence, ye Beauties, undeceiv'd,

Know, one false step is ne'er retriev'd,
And be with caution bold.
Not all that tempts your wand'ring eyes
And heedless hearts is lawful prize,
Nor all, that glisters, gold.
Thomas Gray – ―Ode On The Death Of A Favourite Cat Drowned In A Tub Of
Goldfishes‖
Particle Detection and Measurement
December 8-10, 2010; USP Headquarters
USP Council of Experts
Mario Sindaco, M.S., MBA

USP Director, Compendial Affairs
USP Governing and Advisory Bodies
Council of the
Convention
Convention
Membership
Board of Council
Trustees & of Experts &
Expert Panels
Board Expert
Committees USP Staff Committees
Advisory Stakeholder
Bodies and Forums &
Groups Project
Teams
USP Council of Experts – General Information
• Serves as the scientific

decision-making body for the
U.S. Pharmacopeia.
• Is composed of 20 Expert Council
Committees in areas related to of Experts &
Expert Panels
Expert
USP compendia.
Committees
• The Chairs of each Expert
Committee together serve as
the Executive Committee of the
Council of Experts.
• The Chairs of the Expert
Stakeholder
Committees are elected by the Forums &
USP Convention. Project
• Dr. Roger L. Williams serves as Teams
the Chair of the Council of
Experts’ Executive Committee.
2010–2015 Expert Committees
2010-2015 Council of Experts - Demographics
Professional Background
2% 1%1%1%
4%
8% Industry - 175
Academe - 62
Consultant - 36
Government - 25
11% 53% Practitioner - 15
Other - 7
Association - 5
Pharmacopeias - 5
Retired - 2
19%
USP Council of Experts – Expert Committees
• Expert Committees address specific standards-setting areas within

USP.
• Expert Committees are responsible for developing and revising

standards for medicines and foods that appear in the USP
compendia (USP-NF, FCC, DS, P2).
• Expert Committee members develop and approve monographs

and general chapters, as well as related USP Reference
Standards.
USP Council of Experts – Expert Panels
• Expert Panels are formed to provide additional expertise on a

particular compendial topic, thereby supplementing Expert
Committee expertise.
• Expert Panels are advisory to one ore more Expert Committees

but are not decision-making bodies themselves. They have a
specific charge and are dissolved at the conclusion of their work.
• Examples:
• Unfractionated Heparin Expert Panel
• Medicare Model Guidelines Expert Panel
• Metal Impurities Expert Panel
Requirements to serve as expert volunteers
 USP invites qualified candidates:

• Pharmaceutical scientists
• Academicians
• Regulatory professionals
• Healthcare professionals
• And others
• From the areas of:
• Nomenclature • Reference Standards
• Small Molecules • Compounding
• Biologics and Biotechnology • Food Ingredients
• Excipients • Dietary Supplements
• General Chapters
Requirements to serve as expert volunteers
 Volunteers generally spend 3.5 hours per week on USP

work.
 Volunteers travel 1-2 times per year to USP Headquarters

for their Committee’s or Panel’s face-to-face meeting.
Expert Committees and Panels also meet by teleconference
and Web regularly.
 Candidates should have advanced degrees and should be

actively working in related scientific and/or regulatory
disciplines, or manufacturing quality assurance.
How You Can Be Involved
Go to: www.usp.org/goto/nominate and submit an application.

Saws Day 3 Master

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Saws Day 3 Master

Uploaded by

Copyright:

Available Formats

―Statistical Considerations Involving

Particle Size Analysis‖

Alan F Rawle Ph.D

In the material I‘ll provide an Appendix with

―However, it must be realised that particle size analysis is

Professor Harold Heywood

D[1, 0] – Microscopy – the simplest

Deal with the numbers of particles

D[3, 2] – Surface area moment mean

Numbers of particles irrelevant (after all it‘s a

Adapted from New Scientist 13th October 1991

Transformations theoretically possible but not advised

D[x, 0] D[p, q] Typically (p -1) = q

Dust? Smaller particles much less significant

2-D representation (z-axis?) Volume/mass

―For irregularly shaped particles,

The Physics of Particle Size Analysis Conference University of Nottingham

Reliability of selected sampling methods using a 60:40 sand mixture

Sampling technique Standard deviation

Cone and quartering 6.81

Note: +/- 3 s for scoop is ~ +/- 16%!

Is inversely proportional to the square root of the

10000 particles total will be needed to specify the

Assume: 1% of agglomeration at 2 mm (2000

This table indicates the minimum mass of material

Because it‘s all you can spare

Mode is below the average

1 1.96 6.31 12.71 63.7

Is this being assessed?

―Before the inherent variability of the test animals

Largest particle in the system

Deming: Out of the Crisis Juran: Managerial Breakthrough

Cannot inspect quality into a product

100% = No particles Transmission Mill Current Dv(50)

Dt Time from Start (s)

Time from Start (s)

Feed rate Feed rate Screen

For heterogeneous systems (i.e. those with

ISO TC 24/SC 4 N rev1 Date: 2010-10-17 ISO/CD 9276-2 (Michael

x 1,0 arithmetic mean size

x 2,0 arithmetic mean surface size

x 3,0 arithmetic mean volume size

x 1,1 length weighted mean size

x 1,2 surface weighted mean size, Sauter mean diameter

x 1,3 volume weighted mean size

D  3,0 harmonic mean volume size

D  2,1 diameter-weighted harmonic mean volume size

D 1,2 surface-weighted harmonic mean volume size

D  2,0 harmonic mean surface size

D 1,1 diameter-weighted harmonic mean surface size

D 1,0 harmonic mean size

D 0,0 geometric mean size

D1,1 diameter-weighted geometric mean size

D 2,2 surface-weighted geometric mean size

D 3,3 volume-weighted geometric mean size

D1,0 arithmetic mean size

D 2,1 diameter-weighted mean size

D 3,2 surface-weighted mean size, Sauter mean diameter

D 4,3 volume-weighted mean size

D 2,0 mean surface size

D 3,1 diameter-weighted mean surface size

D 4,2 surface-weighted mean surface size

D 5,3 volume-weighted mean surface size