Aerosol measurement


Material Information

Aerosol measurement
Physical Description:
xxiv, 716 p. : ill. ; 24 cm.
Lundgren, Dale A
Aerosol Measurement Workshop, 1976
University Presses of Florida
Place of Publication:


Subjects / Keywords:
Aerosols -- Measurement   ( lcsh )
Particle size determination   ( lcsh )
bibliography   ( marcgt )
non-fiction   ( marcgt )
conference publication   ( marcgt )


Includes bibliographies and indexes.
General Note:
"A University of Florida book."
Statement of Responsibility:
editors Dale A. Lundgren ... et al..

Record Information

Source Institution:
University of Florida
Rights Management:
Copyright by the Board of Regents of the State of Florida. This work is licensed under a modified Creative Commons Attribution-Noncommercial-No Derivative Works 3.0 Unported License. To view a copy of this license, visit You are free to electronically copy, distribute, and transmit this work if you attribute authorship. However, all printing rights are reserved by the University Press of Florida ( Please contact UPF for information about how to obtain copies of the work for print distribution. You must attribute the work in the manner specified by the author or licensor (but not in any way that suggests that they endorse you or your use of the work). For any reuse or distribution, you must make clear to others the license terms of this work. Any of the above conditions can be waived if you get permission from the University Press of Florida. Nothing in this license impairs or restricts the author's moral rights.
Resource Identifier:
oclc - 04056708
lccn - 78015424
isbn - 0813006031
lcc - TD884.5 .A34 1976
ddc - 628.5/3
System ID:

This item is only available as the following downloads:

Full Text




Dale A. Lundgren
Environmental Engineering Sciences
University of Florida
Gainesville, Florida

Franklin S. Harris, Jr.
Physics and Geophysical Sciences
Old Dominion University
Norfolk, Virginia

William H. Marlow
Brookhaven National Laboratory
Associated Universities
Upton, New York

Morton Lippmann
Institute of Environmental Medicine
New York University Medical Center
New York, New York

William E. Clark
Environmental Engineering Department
California Polytechnic State University
San Luis Obispo, California

Michael D. Durham
Environmental Engineering Sciences
University of Florida
Gainesville, Florida

A University of Florida Book

University Presses of Florida
Gainesville / 1979


Library of Congress Cataloging in Publication Data

Aerosol Measurement Workshop, University of Florida,
Aerosol measurement.

"A University of Florida book."
Includes bibliographies and index.
1. Aerosols-Measurement-Addresses, essays, lectures.
2. Particle size determination-Addresses, essays, lec-
tures. I. Lundgren, Dale A. II. Title.
TD884.5.A34 1976 628.5'3 78-15424
ISBN 0-8130-0603-1

The University Presses of Florida is the
scholarly psablishing agency for the
State University System of Florida.




Preface ......................... ............ ............ ix
Introduction...................... ........................ xi
Contributors ..................... ....................... xxi

Part I: Inertial Classification

Aerosol Centrifuges. Marvin I. Tillery ............................ 3
Comments on Centrifuges. Owen R. Moss ........................ 24
Lovelace Aerosol Particle Separator Design Modifications.
D. Talley, 0. G. Raabe, and J. A. Mewhinney ................. 29
The Goetz Aerosol Spectrometer. Hermann E. Gerber .............. 36
-Analysis of the Cyclone as a Size Selective Aerosol Sampler.
J. M Beeckmans ......................................... 56
Use of Cyclones for Size-Selective Aerosol Sampling. Morton
Lippm ann .................. ....... .. .... ............. 66
Cyclone Discussion. Howard E. Ayer and John M.
H ochstrasser ............. ........ .... ................. 70
Aerosol Measurement by Cyclone Collectors. S. L. Soo ............. 80
Inertial Impactors. Virgil A. Marple and Klaus Willeke .............. 90
Comments on Inertial Impactor Calibration and Use. Ronald H.
K nuth ...................... ... .... .................. 108
Some Comments on the Selection and Use of an Impactor. A. Kishan
Rao ................................................ 117
Measurement of Particulate Aerosol Mass Concentration Using a
Piezoelectric Crystal Microbalance. David C. Woods ............ 119
Comments on Surface Coatings for Lundgren-Type Impactors.
Thomas A Cahill ..................... .................. 131

Design and Use of the Mercer-Style Impactor for Characterization of
Aerosol Aerodynamic Size Distributions. Otto G. Raabe ......... 135

Part II: Light Scattering Particle Counters

Single Particle Optical Counters: Principles and Field Use. Kenneth
T. Whitby and Klaus Willeke ................................ 145
Royco Instruments Particle Counters: Capabilities and Limitations.
A. Lieberman ......................................... 183
Synoptic Analysis of Aerosol Particles. James M. Lepper, Jr. ......... 194
The Aerodynamic Size Calibration of Optical Particle Counters by
Inertial Impactors. Virgil A. Marple ......................... 207
Design Considerations of the TechEcology Particle Counters. Louis
J. Petralli, Jr. ........................................... 216
An Evaluation of the Climet 208 and Royco 220 Light-Scattering
Optical Particle Counters. William E. Clark and Edward L.
Avol ....... ..... .. .......... ....... ........ 219
Aerosol Growth Measurements Using the Differential II and Climet
208 Light Scattering Spectrometers. I. N. Tang, H. R.
Munkelwitz, and J. G. Davis ............................... 231
Aerosol Field Measurements Using Light Scattering Photometers.
Richard W. Storey ........................................ 241
Computer Graphics and Analysis of Atmospheric Aerosol Size
Distributions. Ronald J. Sentell ............................ 260
Holography and Image Analysis. Alexander E. Martens ............. 269
Single Particle Light Scattering Spectrometers. Robert G.
Knollenberg ........................ .................. 271
Performance of the PMS Axially Scattering Spectrometer Probe.
Richard K. Jeck .......................................... 294
Aerosol Size and Shape Determination Using a Laser Light
Scattering Spectrometer. J. Allen, R. B. Husar, and E. S.
Macias ................................................ 312
Counting Efficiencies of Three Single Particle Aerosol Counters.
David Rimberg ........................................ 321

Part III: Electrical Aerosol Analyzer

Electrical Aerosol Analyzer: History, Principle, and Data Reduction.
Benjamin Y. H. Liu, David Y. H. Pui, and Abde Kapadia ....... 341
Electrical Aerosol Analyzer: Calibration and Performance. David Y.
H. Pui and Benjamin Y. H. Liu ............................. 384
Electrical Aerosol Analyzer: Operation, Maintenance, and
Application. Gilmore J. Sem................ ............. 400



The Diffusion Charging Mobility Analysis Hypothesis Revisited.
W. H. Marlow ............................................ 433
The Reduction of Data from the Electrical Aerosol Analyzer.
L. Willard Richards ....................................... 438
A Comparison of the Simultaneous Response of Four Electrical
Aerosol Analyzers. William E. Clark and Edward L.
Avol ............. ..................................... 451
The Role of the Electrical Aerosol Analyzer in Photochemical
Aerosol Research. W. C. Kocmond, J. Y. Yang, and D. B.
K ittelson ............................... ................ 458
Measurement of Ultrafine Particle Size Distributions in Industrial
Flue Gases Using an Electrical Aerosol Analyzer. James W.
Ragland .............. ............... ...... ....... ... 473
Comparison of Electron Microscopy and the Electrical Aerosol Size
Analyzer for Determination of Size Distribution of a
Submicronic Salt Aerosol. R. F. Phalen, A. T. Ho, and J. L.
Kenoyer ...................................... 480
Comments on Aerosol Measurement. Peter J. Groblicki ............. 488
Electrical Aerosol Analyzer Constants. K. T. Whitby and B. K.
Cantrell .............. ..... .. .... .. ................ 492

Part IV: Condensation Nuclei Counter and Diffusion Battery

Aerosol Detection by Condensation Nucleus Counting Techniques.
Austin W. Hogan ....................... ............... 497
Fine Particle Measurement. T. A. Rich ........................... 515
Calibration of Aitken Nuclei Counters Deviations from Classical
Picture of Condensational Growth of Nuclei. Josef Podzimek
and James L. Kassner, Jr. ................................ 527
Automatic Analysis of Submicron Aerosols. David Sinclair,
Richard J. Countess, Benjamin Y. H. Liu, and David Y. H.
Pui .............. ............ ...................... 544
Instrumentation for Real-Time Measurement of Condensation
Nuclei. Ernest A. DeMetre ................................ 564
General Electric Condensation Nuclei Counters. J. Brooks
H aberl ..................... ............................. 568
Droplet Growth Kinetics and Design Principles for Condensation
Nuclei Counters. Paul E. Wagner ........................... 574
NCAR Airborne Condensation Nucleus Counter. G. Langer .......... 585
Tropospheric Airborne Condensation Nuclei Sampling: Meteorology
Research, Inc., Air Pollution Mapping Programs. David S.
Ensor and J. A. Anderson .................................. 590
Experiences in the Determination of Particle Size Distributions in
Industrial Flue Gas Emissions Using Diffusional Methods.
Joseph D. McCain ............. ... ................. 600



The Diffusion Processor. William Marlow ......................... 609
Calculation of Aerosol Size Distribution from Diffusion Battery
Measurements-A Computer Program for the Graphical
"Stripping" Method. David Sinclair, David A. Christy, and
Kenneth Snyder ................. ...................... 615
Part V: Other Methods for Aerosol Measurement

Remote Sensing of Atmospheric Aerosols. Alex E. S. Green ......... 635
Aerosol Particle Size Measurements Using the Electrical Resistance
Principle. S. Kinsman .................................... 650
Aerosol Monitors Based on the Contact Electricity Principle. Walter
John, Georg P. Reischl, and William Devor ................... 655
Performance Evaluation of a New Piezoelectric Aerosol Sensor.
Gilmore J. Sem and Major Peter S. Daley .................... 672
New Instruments in Light Scattering Techniques. Franklin S.
Harris, Jr. ............................................. 687
Particulate Mass Monitoring Techniques Applied to Emission
Sources. John S. Nader ..................... ............. 690

Subject Index ................. ............................ 703
Author Index .................... .... ....................... 711




In this volume we present a complete text on the basic principles, mechanical
operation, calibration, data gathering procedures, data interpretation, instrument
limitations, and specific applications of aerosol measurement instrumentation.
Information contained herein is not found in normal technical papers or manufac-
turers' literature. It should enable the engineer and scientist to better understand
the application and use of aerosol measurement devices. All answers are not
provided because knowledge is incomplete and differences of opinion exist
among the leading experts. However, different viewpoints are presented by
various writers. Manufacturers point out instrument advantages while users point
out shortcomings and limitations. Several articles by users of these instruments
present performance data unobtainable elsewhere. Summary articles, complete
with extensive literature review lists, enable the reader to become better aware of
the extensive published literature.
This book is a result of the interest, support, and hard work of many individu-
als actively involved in the field of aerosol measurement. The Aerosol Measure-
ment Workshop held at the University of Florida, March 24-26, 1976, could not
have been held nor this book published without them. I wish each person and his
or her contribution could be individually acknowledged.
A group of about twenty persons, while attending a 1975 Gordon Research
Conference on Atmospheric Aerosols, discussed, supported, suggested funding
methods, and helped formulate the workshop which led to this book. Three of
these individuals, Drs. Morton Lippmann, William Clark, and William Marlow,
have assisted from the formulation of the workshop idea through the organization
and review of three sections of this volume. A fourth section was organized and
reviewed by Dr. Franklin Harris. All sections of this book were assembled,
edited, and coordinated with the able assistance of Ms. Ann Root and Ms. Kathy

x Preface
In addition to the more than fifty persons who authored or co-authored the
articles contained herein, many others assisted and contributed through active
workshop audience participation which resulted in the discussions, questions,
and answers also found in this book.
I have relied heavily on my graduate students to assist me with everything
from the initial workshop on through to the final publication of this volume. One
student in particular, Dr. Michael Durham, was involved in all aspects of this
task, including the book's indexing, during his four year "apprenticeship," and
his name deservedly appears as a co-editor.
Special thanks go to Al Breslin and Bob Beadle of the Energy Research and
Development Administration, now the U.S. Department of Energy, who helped
obtain funding; to Edwin Stokely and John Kane in the Office of Public Affairs,
who administered the funding; to Parker Reist and Lawrence Kornreich from the
Triangle Universities Consortium on Air Pollution (TUCAP) for providing funds
from the Environmental Protection Agency (EPA); to Major Peter Daley for his
assistance in obtaining funds from the U.S. Air Force Office of Scientific Re-
search; and to Ronald Vance, who administered these funds which were used to
support the Aerosol Measurement Workshop and to help in the preparation of this
book. Thanks go to Richard Palmer, Division of Continuing Education, Univer-
sity of Florida, for his help in presenting the workshop and administering these
Receipts from the sale of this royalty-free book will be used to pay all costs of
publication, including typesetting, printing, and binding. Final publication was
delayed, however, because all costs had to be covered in advance of sale.
Therefore, special thanks are extended to Thermo-Systems, Inc. (now TSI), St.
Paul, Minnesota, for providing the financial backing required to guarantee this
book becoming a reality.


Fine particle measurement has been an important discipline to many areas of
science and industry, and the topic of many excellent books. Articles on the
techniques and significance of particle size distribution measurements date back
to about 1900. Industries as diverse as paint pigment manufacturing and flour
milling have been affected by particle size distribution since their beginnings.
Particle size distribution can be measured in a powder or an aerosol. Mea-
surement in an aerosol differs from that in a powder in several ways and for
several reasons. Powders consist of relatively large particles (normally 1-1000
/xm diameter) with large quantities of material available for the particle size
measurement (1 gram or greater quantities). Aerosols, however, involve much
smaller particles covering a wider size range (0.001-100 /m diameter) with
much smaller quantities of particulate mass available for an analysis (often 1
milligram or less).
Measurement of the particle size distribution of an aerosol by instrumental
techniques is the topic of this book. It is intended as a reference text for those
who use or anticipate using commercially available instrumentation to study,
measure, monitor, or analyze particle size distribution or size fractionated sam-
ples of an aerosol. Four specific techniques are covered: inertial classification
(including centrifuges, cyclones and inertial impactors), light-scattering particle
counter (single-particle optical counters), electrical aerosol analyzer (electrical
mobility analyzer), and condensation nuclei counter-diffusion battery (combined
for size analysis). These four techniques are essentially the researcher's only
methods for studying the particle size-dependent properties of an aerosol such as
formation, condensation/evaporation, or coagulation. As such, they form the
basis for discussion.
Sieving, sedimentation, elutriation, simple air classification, and some air

permeability and surface area techniques used for size measurement of powders
are not directly applicable to aerosol measurement and as such are excluded from
this volume. These methods of size classification are discussed in books by R. D.
Cadle (1955), C. Orr and J. M. DallaValle (1959), G. Herdan (1960), L. Silver-
man, C. E. Billings, and M. W. First (1971), T. T. Mercer (1973), R. Dennis
(1976), and others (see attached bibliography).
Microscopy, both light and electron, is equally useful in the measurement of
powders and aerosols and is very basic to the calibration of other measurement
techniques and to the description of particle shape. No measurement instrument
is more important, useful or versatile than the light microscope. Similar com-
ments are applicable to the relatively complicated electron microscope and the
scanning electron microscope (SEM). Although it is not discussed as a subject in
this text, it is assumed that the reader has a basic knowledge of microscopy and
understands its uses and limitations. Several of the books referred to above
include discussions of microscopy and there are several specific texts on the
subject. These works also treat the techniques and problems of aerosol sampling
and sample preparation for viewing. (General aerosol literature sources are re-
viewed later.)


A particle size characteristic of primary importance in aerosol studies is the
particle aerodynamic diameter-the diameter of a unit density spherical particle
which has the same steady state velocity in a gravitational or centrifugal force
field as the particle of interest. This diameter allows determination of residence
time, potential inhalation hazard, and particle inertial properties such as losses in
a pipe bend or impaction against an obstacle in a flow stream. Aerodynamic
diameter is a measurement parameter describing how a particle will act, not how
it will look (except for a non-volatile, solid, spherical particle of unit density).
Because most particles of interest are of irregular shape or of unknown density, it
is often difficult to relate this diameter to optical appearance. For example,
during the collection of gas-borne particles onto a solid surface suitable for
subsequent microscopic viewing, liquid particles may coalesce, volatile particles
may evaporate, particles may agglomerate, or agglomerated particles may break
Inertial classification devices separate particles from a gas stream but do not,
by themselves, measure aerosol concentration or size distribution. This means
that gas-borne particles are fractionated, or removed from a gas stream, in an
aerodynamic size-dependent manner for subsequent analysis of the known size
particulate fraction. Size-fractionated samples containing many particles are ob-
tained in this manner, with the analysis being performed on the gross sample and
normally not on single particles.
Aerodynamic diameter classification devices include gravitational sedimenta-



tion chambers, air elutriators, aerosol centrifuges, cyclones, and inertial impac-
tors. Sedimentation chambers and air elutriators are frequently used for size-
fractionation of aerosolized powders or dusts but not often for ambient concentra-
tion aerosol sampling.
Aerosol centrifuges are mechanical devices which involve a rapidly spinning
body through which an aerosol sample is flowing at a constant controlled rate.
They are capable of size-separating small particles because a very high cen-
trifugal force can be applied to the particles of interest. High rotational speed
centrifuges work best in the general size range from about 0.1 to 1.0 /m
(aerodynamic) diameter. Classification is also accomplished with high flowrate
air centrifuges used by various industries to coarse fractionate powders in the
5-50 ,m size range. These devices are not normally used for the study of aerosol
concentrations. Aerosol centrifuges as a class of air samplers are discussed in
detail in Part I, beginning with the article by M. Tillery. Discussion of the much
simpler elutriators can be found in other sources (Mercer, 1973; Silverman,
Billings and First, 1971).
Cyclones are similar to centrifuges in that they depend upon centrifugal force
for particle separation. Cyclones, however, are simpler mechanical devices
which rely upon a rapidly spinning air stream inside a fixed housing for the
separation force, as opposed to the mechanically complicated rapidly spinning
body of a centrifuge. Particle size separation is not as sharp in an air sampling
cyclone as it is in a centrifuge. An optimum size range for cyclones is from 3 to
30 /m (aerodynamic) diameter. Cyclones are especially useful as pre-collectors
for other aerosol size measurement devices. J. M. Beeckmans' article introduces
a discussion of cyclone design and calibration.
Inertial impactors are the third category of aerodynamic diameter classification
devices treated in Part I in a series of articles starting with a report by V. Marple
and K. Willeke. Several impaction stages connected in series are normally re-
ferred to as cascade impactors. These impactors are intermediate between cy-
clones and centrifuges in both mechanical complexity and particle size classifica-
tion, being optimally suited to the 0.5-5 pm (aerodynamic) diameter size range.
Through the use of very small dimensions and very high gas velocities, the lower
size limit can be decreased below 0.1 gm with reduced pressure; conversely,
through the use of very large dimensions, low velocities and high gas flowrates,
the upper size limit can be increased to perhaps 50 gm. If it were not for
operational problems, cascade impactors could be ideal classification devices
because theoretically they can cover a very broad particle size range.
Aerosol centrifuges, cyclones, and inertial impactors are all devices which
collect (or remove) gas-borne particles in a size-dependent manner. Samples of
particulate matter are then available for subsequent chemical or physical
analyses. In general, these devices are not aerosol monitors and are not intended
for in situ particle size distribution measurement. D. Woods discusses one tech-
nique of automating the readout of these devices, and the uses and limitations of
inertial classification devices are treated in about a dozen other articles.




Single particle optical counters provide a very different or supplementary method
of aerosol measurement. These particle size distribution monitors do not collect
an aerosol sample, except after the measurement. The basic measurement param-
eter is the amount of light scattered by a particle. This particle is then assigned an
"optical" size which is equal to the actual size of a spherical calibration particle
which scatters an equal amount of light (as measured by the instrument photosen-
sitive detector).
Optical counters have several definite advantages. Even though some are
mechanically complicated, they can be very reliable. Electronics are such that
these devices can count and optically size gas-borne particles extremely rapidly
and with a fair degree of reproducibility.
Although optical counters have been primarily used in clean environments,
they can be adapted to fairly dirty environments. Again, there is an optimum
particle size range for these devices; most operate best for particles between 0.5
and 5.0 Am diameter. Other models or modifications work well over either a
slightly lower or higher size range with no theoretical upper limit, and a lower
limit of less than 0.1 /m for the most sophisticated unit. This class of counters
fulfills a very important function in the aerosol measurement field for the
monitoring of number concentration in a preselected size interval or intervals.
The operating principle, theoretical response, calibration, limitations, and use
of single particle optical counters are discussed in a comprehensive article by K.
T. Whitby and K. Willeke. Counters manufactured by three different companies
are treated in separate articles by A. Lieberman, J. Lepper, Jr., and L. Petralli,
Jr. Details of calibration, instrument comparison, field use, and data interpreta-
tion are presented in other articles. More novel, but commercially available,
optical devices are described by R. Knollenberg and discussed by R. Jeck. In all,
the pros and cons of commercially available light-scattering particle counters are
discussed in detail.


The electrical aerosol analyzer is a device especially suited to the measurement of
particle size distribution below the range of the optical counter. Size separation is
obtained because the electric mobility of a gas-borne particle in an electric field is
a function of particle size. This device collects the aerosol during measurement,
but it is not a true aerosol sampler. Rather, it is a rapid response (several
minutes), aerosol size distribution monitor with good size measurement capabil-
ity over the general 0.01-0.5 Am (electrical) diameter size range. The electrical
aerosol analyzer determines the electric mobility of an unknown particle in a
known electric field and then assigns a diameter to the unknown particle equal to
the diameter of a known size calibration particle having an equal electric



Although the first instrument to measure the mean size of an aerosol electri-
cally was built over 50 years ago, it was not until about 1973 that a truly
successful commercial electrical aerosol monitor was available. This device, its
development, calibration, and use are discussed in detail in Part III.
The history and principle of the electrical aerosol analyzer is first discussed by
B. Y. H. Liu, D. Y. H. Pui, and A. Kapadia. The calibration and performance of
the instrument are detailed by D. Y. H. Pui and B. Y. H. Liu. Instrument opera-
tion and maintenance, a summary of applications, and a complete bibliography
are included in G. Sem's article. Additional articles cover theory, data reduction,
instrument comparison, and applications to several types of aerosol measurement
situations. Comments and discussion of this measurement technique made at the
aerosol measurement workshop are also included.


As the name implies, a condensation nuclei counter (CNC) is a device to count
the number of nuclei which serve as condensation sites for a supersaturated
vapor, normally water. The counting can be done manually, as it was in early
devices and still is in some units, or automatically by measuring the light extinc-
tion caused by the liquid droplets formed in the condensation process. These
devices, discussed in detail in Part IV by A. Hogan, T. Rich, and others, are
capable of sensing or counting particles greater than about 0.005 am (diffu-
sional) diameter. This lower detectable particle size is somewhat uncertain but is
definitely within the 0.001-0.01 /m diameter range, depending upon the compo-
sition of the particles and the instrument operating conditions (such as degree of
supersaturation). CNCs are capable of particle size distribution measurements to
a limited extent. If several fractions of a stable aerosol are sensed at different
supersaturations and a count recorded for each sample, then the differences in
number will indicate the number of particles in a calculated size interval. Opera-
tion of a CNC will be clarified after reading P. Wagner's article on droplet
growth theory, J. Podzimek and J. Kassner, Jr.'s, discussion of basic calibra-
tion and E. DeMetre's and J. B. Haberl's descriptions of commercially avail-
able instruments.
A main advantage of these devices is their ability to rapidly sense or detect
smaller particles at lower concentrations than can be detected by any other device
or method. The devices also serve as sensing or counting mechanisms for other
methods of particle size classification, namely, the diffusion battery, which
preferentially removes the very smallest particles first because of a higher diffu-
sion coefficient.
A diffusion battery is a set of circular or rectangular cross section tubes
through which an aerosol stream flows. Several of these cells are normally
mounted so that an aerosol stream can be deflected through various combinations
of these cells to produce the differential removal as a function of particle size. A



CNC is used at the output of a diffusion battery to provide a readout or measure-
ment of the particle number concentration. In this combination a particle size
distribution device is achieved which works well over the 0.01-0.1 /m diameter
range even for low aerosol concentrations. This range can be extended almost
fivefold in both directions. Above 0.3 /m other techniques work far better.
Below 0.01 m the CNC is the only device capable of detecting low concentra-
tion. Therefore, this separation technique has definite applications down to a
theoretical lower limit of perhaps 0.002 /m diameter. Design and operation of
diffusion batteries are discussed by Sinclair and Countess.


Although this book is specifically concerned with four aerosol measurement
methods (based upon commercially available instrumentation) it was also con-
sidered desirable to acquaint the reader with several related papers presented in a
general session at the Aerosol Measurement Workshop and submitted for publi-
Several remote aerosol sensing techniques are discussed in an article by A.
Green. J. Allen, R. Husar, and E. Macias present the possibility of both size and
shape determination by laser light scattering. One related size distribution mea-
surement technique based upon the electrical resistance principle is summarized
in an article by S. Kinsman. Aerosol sensing based upon contact electricity is
discussed by W. John, G. Reischl, and W. Devon. Performance of a new
piezoelectric aerosol sensor is the subject of G. Sem and P. Daley's article.
Ideally, it would be desirable to summarize in this one volume all information
related to or necessary for a complete understanding of aerosol measurement.
Because that was not possible, nor even reasonable to attempt, the following
discussion of published literature was prepared

Many books have been published that specifically relate to the field of aerosol
measurement. Several of these discuss the related problems of aerosol sampling
(a representative sample is necessary in order to obtain a meaningful measure-
ment), aerosol generation (necessary for proper instrument calibration), particle
statistics (including presentation of data and calculation of measurement error),
aerosol physics (physical laws governing the motion and interaction of gas-borne
particles), and aerosol chemistry (formation, reaction, and composition of parti-
To help make the reader aware of books pertaining to the fine particle and
aerosol fields, and acquaint him or her with the general content, the following list
was prepared; it includes a short discussion of the topics covered, particularly
those that relate to aerosol measurement. In addition, a list of journals is ap-
pended in which articles related to aerosol measurement would most likely be
published. These lists are not complete, but we hope they include the majority



of aerosol works published in English in the open literature. Bibliographies
accompanying the articles in this book are quite complete for most of the mea-
surement techniques discussed, and the reader should check these lists carefully.


Allen, T., 1968: Particle Size Measurement. Chapman and Hall Ltd., London, 248 pp.
Topics include: Particle sampling, particle size distributions, sieving, microscopy, sedimentation,
centrifugal methods, Coulter counter use, light scattering, permeametry and gas adsorption.
1975: Particle Size Measurement. 2d ed., John Wiley and Sons, New York, 454 pp.
All aspects of particle size measurement with emphasis on powders.
American Conference of Governmental Industrial Hygienists, 1972: Air Sampling Instruments. 4th
ed. P.O. Box 1937, Cincinnati, Ohio 45201, 587 pp.
A collection of articles dealing with many aspects of gaseous and aerosol sampling equipment and
methods. Areas of concern include industrial hygiene, ambient and source sampling, biological
aerosols, instrument calibration, flow systems, particle collectors and direct reading devices.
American Society for Testing Materials, 1959: Symposium on Particle Size Measurement. ASTM,
Philadelphia, 303 pp.
Describes particle size measurement by means of sieving, gravitational and centrifugal sedimenta-
tion, light scattering, microscope counting, and using a Coulter counter.
Butcher, S. S. and R. J. Charlson, 1972: An Introduction to Air Chemistry. Academic Press, New
York, 241 pp.
Included in this introductory text are discussions of sampling and analysis of aerosols and chemi-
cal and physical properties of atmospheric aerosols.
Cadle, R. D., 1955: Particle Size Determination. Interscience Publishers, Inc., New York, 303 pp.
Principal topics covered are particle statistics, sampling, optical and electron microscopy, sieve
analysis, sedimentation, elutriation and surface area measurements (powders). Light scattering is
also considered.
,__ 1965: Particle Size. Reinhold Publishing Corporation, New York, 390 pp.
Text devoted to the practical importance of particle size distributions. General introductory dis-
cussion on particle behavior leads to application of basic aerosol principles to air pollution
problems and particle deposition in lungs and on the skin. Also covered is the importance of size
distribution in the paint, cement, powder, and paper and pulp industries.
,__ 1975: The Measurement of Airborne Particles. Wiley-Interscience, New York, 342 pp.
Comprehensive coverage of the collection and analysis of aerosols. Topics include: Particle
statistics, filtration theory, sedimentation, thermal and electrical precipitation, optical and electron
microscopy, centrifugal classifiers, multistage impactors, optical measurement of aerosols, con-
densation nuclei counters, acoustic counters and electric aerosol analyzers.
DallaValle, J. M., 1943: Micromeritics. Pitman Publishing Corporation, New York, 428 pp.
Basic text on the behavior and characteristics of small particles. Includes discussion on particle
size measurement, size distributions, dynamics of small particles, characteristics of packing, and
physical and chemical properties of aerosols.
Dautrebande, L., 1962: Microaerosols. Academic Press, New York, 366 pp.
A summary of the author's and his co-workers' experiences in the generation and measurement of
aerosols used to study their physiological and therapeutic effects after deposition in the respiratory
Davies, C. N., Ed., 1961: Inhaled Particles and Vapours. Pergamon Press, New York, 495 pp.
Proceedings of an international symposium organized by the British Occupational Hygiene Soci-
ety 3/29-4/1/60. It contains four papers on the size-selective sampling of airborne dust.
,__ 1964: Recent Advances in Aerosol Research. Pergamon Press, The Macmillan Company,
New York, 80 pp.
A bibliographical review of aerosol research performed between 1957 and 1962.
,__ 1966: Aerosol Science. Academic Press, New York, 468 pp.
A detailed text composed of chapters written by the leading authorities in the field. Chapters cover
aerosol generation, coagulation, electrical behavior, filtration, optical measurement, and adhesion
and deposition of particles.



- Ed., 1967: Inhaled Particles and Vapours II. Pergamon Press, New York.
Proceedings of an international symposium organized by the British Occupational Hygiene Soci-
ety 9/28-10/1/65. It contains six papers on dust exposure evaluations and instruments.
,__ 1973: Air Filtration. Academic Press, New York, 171 pp.
A complete analysis of the removal of aerosols by fibrous and membrane filters including descrip-
tions of filtration theories and experimental results.
Dennis, R., Ed., 1976: Handbook on Aerosols. National Technical Information Service, TID-26608,
Virginia, 142 pp.
Subjects discussed are aerosol generation, aerosol dynamics, optical properties, aerosol sampling
and particle size measurement. Extensive reference lists are included for each subject as well as a
bibliography of aerosol books and periodicals containing aerosol articles.
Drinker, P. and T. Hatch, 1954: Industrial Dust. McGraw-Hill Book Company, Inc., New York,
401 pp.
Comprehensive text covering both the engineering and medical aspects of the dynamics, charac-
teristics and control of industrial aerosols.
Friedlander, S. K., 1977: Smoke, Dust and Haze: Fundamentals of Aerosol Behavior. John Wiley
and Sons, New York, 317 pp.
Written as text for a course in particulate pollution. Topics include size distribution, transport,
deposition, optical properties, measurement techniques, aerosol behavior, and gas-to-particle
Fuchs, N. A., 1959: Evaporation and Droplet Growth in Gaseous Media. Pergamon Press, New
York, 72 pp.
A complete survey devoted to the kinetics of evaporation and growth of droplets of pure liquids in
stationary and nonstationary media.
,__ 1964: The Mechanics of Aerosols. Pergamon Press, New York, 408 pp.
This classic work reviews almost all the important papers on the subject "Mechanics of Aerosols"
up through 1960. Topics covered include rectilinear, curvilinear, and Brownian motion of parti-
cles; diffusion and coagulation of aerosols; and the classification, dispersion, and turbulent motion
of aerosols.
Green, H. L. and W. R. Lane, 1964: Particulate Clouds: Dusts, Smokes and Mists. D. Van Nostrand
Company, Inc., New York, 471 pp.
A comprehensive coverage of the formation, behavior and effects of both man-made and natural
particulate clouds. Descriptions of aerosol sampling, measurement, industrial collection and the
dispersion and effects of particles and clouds in the atmosphere are included.
Hatch, T. F. and P. Gross, 1964: Pulmonary Deposition and Retention of Inhaled Aerosols.
Academic Press, New York, 192 pp.
A study of aerosol deposition from a viewpoint of both aerosol mechanics and human and animal
Herdan, G., 1960: Small Particle Statistics. Butterworth and Company Ltd., London, 418 pp.
A complete treatment of particle statistics. This work also considers sample preparation and sizing
by microscope, gravitational and centrifugal sedimentation (in liquids), surface area measurement
methods and general information on powders. Some aerosol measurement discussion.
Hidy, G. M., Ed., 1972: Aerosols and Atmospheric Chemistry. Academic Press, New York, 348 pp.
Contains numerous articles presenting a wide variety of studies on the physical chemistry of
aerosols and their relationship to atmospheric chemistry. Included is a detailed discussion of the
Pasadena Smog Study.
__ and J. R. Brock, 1970: The Dynamics of Aerocolloidal Systems. Pergamon Press, New York,
379 pp.
A comprehensive text providing an extensive theoretical analysis of the dynamics of single and
collective aerosols.
___ and 1971: Topics in Current Aerosol Research. Pergamon Press, New York, 157
Contains two articles-the first is an analysis by N. A. Fuchs and A. G. Sutugin on the formation
and physical properties of high-dispersed aerosols with particle size below 0.1 /m. The second is
an article by S. L. Soo on the transport in multicomponent gas systems.
Intersociety Committee, 1972: Methods of Air Sampling and Analysis. American Public Health
Association, Washington, D.C., 480 pp.
Describes the EPA recommended methods for the sampling of gases and aerosols.
Irani, R. R. and C. F. Callis, 1963: Particle Size: Measurement, Interpretation and Application.
Wiley, New York, 165 pp.



Detailed descriptions of the proper use of microscopy, sieving, light scattering, turbidity,
sedimentation and Coulter counters in the analysis of particle size distributions.
Junge, C. E., 1963: Air Chemistry and Radioactivity. Academic Press, New York, 382 pp.
Text dealing with the constituents and chemical processes in the lower 50 km of the atmosphere
with an emphasis on basic and large scale phenomena. Analysis of the physical properties,
chemical composition and distribution of aerosols in the troposphere is presented along with a
similar analysis of the gaseous constituents and precipitation.
Kerker, M., Ed., 1963: Electromagnetic Scattering. Pergamon Press, The Macmillan Company,
New York, 592 pp.
A detailed treatment of electromagnetic scattering with extensive references. Basic treatment is for
spheres, but chapters also on stratified spheres and infinite cylinders, particle sizes and applica-
tions to polydisperse systems. Applications are primarily to physical chemistry problems in liquids
rather than to the atmosphere and air pollution.
1969: The Scattering of Light and Other Electromagnetic Radiation. Academic Press, New
York, 666 pp.
Summary of the theory with applications to light scattering.
Liu, B. Y. H., Ed., 1976: Fine Particles. Aerosol Generation, Measurement, Sampling and
Analysis. Academic Press, New York, 837 pp.
Proceedings of the 1975 Symposium on Fine Particles. Contains numerous articles on aerosol
generation, aerosol sampling, aerosol measurement and analysis.
Lundgren, D. A., M. Lippmann, F. S. Harris, Jr., W. E. Clark, W. H. Marlow, and M. D. Durham,
Eds., 1979: Aerosol Measurement. University of Florida Press.
Detailed coverage of four specific aerosol measurement techniques: inertial classification (cen-
trifuges, cyclones, and impactors), light scattering particle counters, electrical aerosol analyzers,
and condensation nuclei counters together with diffusion batteries. Numerous articles on each
topic cover basic principle, mechanical operation, calibration, data gathering procedures, data
interpretation, instrument limitations and specific application.
McCartney, E. J., 1976: Optics of the Atmosphere. Scattering by Molecules and Particles. John
Wiley and Sons, New York, 408 pp.
Good summary of scattering processes and effect on light propagation.
Mednikov, E. P., 1965: Acoustic Coagulation and Precipitation of Aerosols. Consultants Bureau,
New York, 180 pp. (Translation from Russian by C. V. Larrick.)
This text treats the problem of artificial coagulation of an aerosol through use of high-intensity
sonic and ultrasonic vibrations.
Mercer, T. T., 1973: Aerosol Technology in Hazard Evaluation. Academic Press, New York, 394
Comprehensive text on aerosol generation, sampling and measurement. General discussion of
aerosol properties, particle statistics, aerosol sampling for size and concentration measurement,
instruments for aerodynamic, optical, electrical and other size measurements, respirable samplers
and production of test aerosols.
Orr, C., Ed., 1977: Filtration, Principles and Practices, Part I. Marcel Dekker, Inc., New York.
Articles on aerosol filtration and other topics.
Orr, C., Jr. and J. M. DallaValle, 1959: Fine Particle Measurement. The Macmillan Co., New
York, 353 pp.
Topics included are microscopy, particle statistics, sieving, sedimentation, inertial classification,
light scattering, surface area measurement by flow resistance, gas adsorption and liquid-phase
sorption, and pore size measurement (powders).
Shaw, D. T., Ed., 1978: Fundamentals of Aerosol Science. Wiley, New York, 372 pp.
Topics covered include aerosol impactors, Brownian coagulation, droplet evaporation, aerosol
filtration theory, sampling of fibrous aerosol particles and electrostatics in aerosol filtration.
Silverman, L., C. E. Billings and M. W. First, 1971: Particle Size Analysis in Industrial Hygiene.
Academic Press, New York, 317 pp.
Complete coverage of particle sampling and size analysis for both powders and aerosols. Review
of sampling instruments, size analysis by optical, electron and scanning electron microscopy.
Discussion of all common particle sizing apparatus (both manual and automatic), particle statistics
and application areas.
Spumy, K., 1966: Aerosols: Physical Chemistry and Applications. Gordon and Breach, Science
Publishers, New York, 943 pp.
The proceedings of an international conference in Czechoslovakia in 1962. Comprised of 943
pages-98 papers in English, 35 in German, and 7 in French. The subjects covered on aerosols are



physical chemistry, radioactivity, meteorology and astronomy, industry, agriculture, and biologi-
cal effects.
Twomey, S., 1977: Atmospheric Aerosols. Elsevier Scientific Publishing Co., New York, 302 pp.
This text covers aspects of aerosol physics important in the formation, evolution and removal of
particulate material in the atmosphere. Chapters are devoted to the influence of atmospheric
particles on weather, atmospheric optics and radioactive transfer, atmospheric electricity and
atmospheric energetic and climate.
van de Hulst, H. C., 1957: Light Scattering by Small Particles. Wiley, New York, 470 pp.
The most important and detailed book on single scattering by homogeneous spheres. A masterly
presentation including mathematics, physical insight, critical discussion of literature and applica-
tions to chemistry, meteorology and astronomy.
Walton, W. H., Ed., 1971: Inhaled Particles III. Volumes I & II. Unwin Brothers Ltd., Surrey,
England, 1090 pp.
Proceedings of an international symposium organized by the British Occupational Hygiene Soci-
ety 9/14-9/23/70. Volume II contains seven papers on dust sampling and standards.


American Industrial Hygiene Association Journal
Annals of Occupational Hygiene
Applied Optics
Atmospheric Environment
British Journal of Applied Physics
Environmental Science and Technology
Health Physics
Journal of Aerosol Science
Journal of the Air Pollution Control Association
Journal of Applied Meteorology
Journal of Colloid and Interface Science
Journal of Geophysical Research
Journal of the Optical Society of America
Journal of Scientific Instruments
Powder Technology
Review of Scientific Instruments
Staub-Reinhalt Luft (English Translation)




ALLEN, J., AeroChem Research Laboratories, Inc., P.O. Box 12, Princeton,
N.J. 08540
ANDERSON, J. A., Meteorology Research, Inc., 464 Woodbury Road, Altadena,
California 91001
AVOL, E. L., Rancho Los Amigos Hospital, 7601 E. Imperial Highway, Dow-
ney, California 90242
AYER, H. E., Kettering Laboratory, University of Cincinnati Medical Center,
3223 Eden Avenue, Cincinnati, Ohio 45267
BEECKMANS, J. M., Faculty of Engineering Science, The University of Western
Ontario, London, Ontario, N6A 5B9, Canada
CAHILL, T. A., Physics Department, University of California, Davis, California
CANTRELL, B. K., Particle Technology Laboratory, Mechanical Engineering
Department, University of Minnesota, Minneapolis, Minnesota 55455
CHRISTY, D. A., Health and Safety Laboratory, U.S. Energy Research and
Development Administration, New York, New York 10014
CLARK, W. E., Environmental Engineering Department, California Polytechnic
State University, San Luis Obispo, California 93407
COUNTESS, R. J., Environmental Measurement Laboratory, U.S. Department of
Energy, New York, New York 10014
DALEY, P. S., Major, Air Force Civil Engineering Center, Tyndall AFB, Florida
DAVIS, J. G., Brookhaven National Laboratory, Department of Applied Science,
Upton, New York 11973
DEMETRE, E. A., Energy Systems Programs Department, General Electric
Company, 1 River Road, Schenectady, New York 12345

DEVOR, W., Air and Industrial Hygiene Laboratory, California State Department
of Health, 2151 Berkeley Way, Berkeley, California 94704
ENSOR, D. S., Meteorology Research, Inc., 464 Woodbury Road, Altadena,
California 91001
GERBER, H. E., Naval Research Laboratory, Washington, D.C. 20375
GREEN, A. E. S., Department of Physics and Astronomy, University of Florida,
Gainesville, Florida 32611
GROBLICKI, P. J., GM Research Laboratories, Environmental Science Depart-
ment, GM Technical Center, Warren, Michigan 48090
HABERL, J. B., General Electric Company, 100 Plastics Avenue, Pittsfield,
Massachusetts 01201
HARRIS, JR., F. S., Rockville, Utah 84763
Ho, A. T., Air Pollution Health Effects Lab., Department of Community and
Environmental Medicine, University of California, Irvine, California 92717
HOCHSTRASSER, J. M., Industrial Hygiene, Tenneco Chemicals, Inc., Park 80,
Plaza West, 1, Saddle Brook, New Jersey 07663
HOGAN, A. W., Atmospheric Sciences Research Center, State University of
New York at Albany, 1400 Washington Avenue, Albany, New York 12222
HUSAR, R. B., Air Pollution Research Lab., Washington University, St. Louis,
Missouri 65830
JECK, R. K., Naval Research Laboratory, Washington, D.C. 20375
JOHN, W., Air and Industrial Hygiene Laboratory, California State Department
of Health, 2151 Berkeley Way, Berkeley, California 94704
KAPADIA, A., Particle Technology Laboratory, Mechanical Engineering De-
partment, University of Minnesota, Minneapolis, Minnesota 55455
KASSNER, JR., J. L., Graduate Center for Cloud Physics Research, University of
Missouri-Rolla, Rolla, Missouri 65401
KENOYER, J. L., Air Pollution Health Effects Laboratory, Department of Com-
munity and Environmental Medicine, University of California, Irvine,
California 92717
KINSMAN, S., Coulter Electronics, Inc., 590 W. 20th Street, Hialeah, Florida
KITTELSON, D. B., Mechanical Engineering Department, University of Min-
nesota, Minneapolis, Minnesota 55455
KNOLLENBERG, R. G., Particle Measuring Systems, Inc., 1855 S. 57th Court,
Boulder, Colorado 80301
KNUTH, R. H., Environmental Measurement Laboratory, Department of
Energy, 376 Hudson Street, New York, New York 10014
KOCMOND, W. C., Desert Research Institute, University of Nevada System,
P.O. Box 60220, Reno, Nevada 89507
LANGER, G., National Center for Atmospheric Research, Boulder, Colorado
LEPPER, JR., J. M., Climet Instruments Company, 1320 W. Colton Avenue,
P.O. Box 151, Redlands, California 92373



Contributors xxiii
LIEBERMAN, A., Royco Instruments, Inc., 141 Jefferson Drive, Menlo Park,
California 94025
LIPPMANN, M., Institute of Environmental Medicine, New York University
Medical Center, 550 First Avenue, New York, New York 10016
Liu, B. Y. H., Mechanical Engineering Department, University of Minnesota,
Minneapolis, Minnesota 55455
MCCAIN, J. D., Southern Research Institute, 2000 Ninth Avenue S., Bir-
mingham, Alabama 35205
MACIAS, E. S., Department of Chemistry, Washington University, St. Louis,
Missouri 65830
MARLOW, W., Brookhaven National Laboratory, Department of Energy and
Environment, Associated Universities, Upton, New York 11973
MARPLE, V. A., Mechanical Engineering Department, University of Minnesota,
Minneapolis, Minnesota 55455
MARTENS, A. E., Bausch & Lomb, 820 Linden Avenue, Rochester, New York
MEWHINNEY, J. A., Inhalation Toxicology Research Institute, Lovelace Founda-
tion, P.O. Box 5890, Albuquerque, New Mexico 87115
Moss, 0. R., Battelle PNW-Biology Department, Battelle Boulevard, P.O. Box
999, Richland, Washington 99352
MUNKELWITZ, H. R., Brookhaven National Laboratory, Department of Applied
Science, Upton, New York 11973
NADER, J. S., Environmental Sciences Research Laboratory, U.S. Environmen-
tal Protection Agency, Research Triangle Park, North Carolina 27711
PETRALLI, JR., L. J., Met One, 154 San Lazaro Avenue, Sunnyvale, California
PHALEN, R. F., Air Pollution Health Effects Laboratory, Department of Com-
munity and Environmental Medicine, University of California, Irvine,
California 92717
PODZIMEK, J., Graduate Center for Cloud Physics Research, University of
Missouri-Rolla, Rolla, Missouri 65401
Pui, D. Y. H., Mechanical Engineering Department, University of Minnesota,
Minneapolis, Minnesota 55455
RAABE, 0. G., Radiobiology Laboratory, University of California, Davis,
California 95616
RAGLAND, J. W., Southern Research Institute, 2000 Ninth Avenue S., Birming-
ham, Alabama 35205
RAO, A. K., Monsanto Company, 800 N. Lindbergh Boulevard, St. Louis,
Missouri 63166
REISCHL, G. P., Air and Industrial Hygiene Laboratory, California State De-
partment of Health, 2151 Berkeley Way, Berkeley, California 94704
RICH, T. A., Deceased
RICHARDS, L. W., Air Monitoring Center, Rockwell International, 2421 W.
Hillcrest Drive, Newbury Park, California 91320

RIMBERG, D., North American Pemco, Inc., P.O. Box 655, Bardonia, New
York 10954
SEM, G. J., TSI, Inc., P.O. Box 3394, St. Paul, Minnesota 55165
SENTELL, R. J., Tennecomp Systems Inc., P.O. Box J, Oak Ridge, Tennessee
SINCLAIR, D., Environmental Measurement Laboratory, U.S. Department of
Energy, 376 Hudson Street, New York, New York 10014
SNYDER, K., Drexel University, Philadelphia, Pennsylvania 19014
Soo, S. L., Department of Mechanical and Industrial Engineering, University of
Illinois at Urbana-Champaign, Urbana, Illinois 61801
STOREY, R. W., NASA Langley Research Center, Hampton, Virginia 23665
TALLEY, D., Los Alamos Scientific Laboratory, P.O. Box 1663, Los Alamos,
New Mexico 87545
TANG, I. N., Brookhaven National Laboratory, Department of Applied Science,
Upton, New York 11973
TILLERY, M. I., Los Alamos Scientific Laboratory, University of California, Los
Alamos, New Mexico 87545
WAGNER, P. E., Institut Fo Experimentalphysik, Der Universitat Wien, A-1090
Wien, Strudlhofgasse 4, Vienna, Austria
WHITBY, K. T., Particle Technology Laboratory, Mechanical Engineering De-
partment, University of Minnesota, Minneapolis, Minnesota 55455
WILLEKE, K., Department of Environmental Health, University of Cincinnati,
3223 Eden Avenue, Cincinnati, Ohio 45267
WOODS, D. C., Langley Research Center, Hampton, Virginia 23665
YANG, J. Y., Union Carbide, Linde Division, 61 E. Park Drive, Tonowanda,
New York 14150




Inertial Classification



Los Alamos Scientific Laboratory, University of California, Los Alamos, New Mexico

The characteristic of primary interest in determination of airborne residence time
and potential inhalation hazard of airborne particles is gravitational or terminal
settling velocity. This velocity is the result of gravitational force and drag exerted
on the moving particle by air. Interaction between a moving particle and air also
defines the particle's dynamic or inertial properties. Thus, particle characteristics
that determine sedimentation velocity can, in most cases, also be used to define
the particle's inertial properties. Sedimentation velocity of spherical particles of
known density can be calculated if the particle diameter is known. However, in
many cases of interest, particles are irregular in shape and of unknown density.
In this case, the sedimentation velocity or a characteristic related to sedimenta-
tion velocity must be measured. A hypothetical diameter can then be used to
define and compare inertial properties. The aerodynamic diameter or the diame-
ter of a unit density spherical particle with the same sedimentation velocity as the
particle of interest is the most frequently used hypothetical diameter (Mercer,
Aerosols are usually encountered as a distribution of different size particles.
The distribution of the parameters of interest with respect to some particle charac-
teristic must be used to describe the behavior and potential effects of the parti-
cles. Distribution of number, mass, surface area, radioactivity, or other property
of interest with respect to aerodynamic diameter is often used to describe the
range and magnitude of inertial properties. Sampling instruments that separate
particles according to their inertial properties permit direct determination of these
Aerosols are relatively stable suspensions of airborne particles so gravitational
Work performed at the Los Alamos Scientific Laboratory under the auspices of the U.S. Energy
Research and Development Administration. Contract No. 7405-ENG-36.

Inertial Classification

sedimentation velocities are low. Accurate determination of these velocities re-
quires long times with stable atmospheres, accurate measurement of small dis-
tances, or use of an increased deposition force. Centrifugal force is often used to
increase the deposition rate and provide accurate measurements in reasonable
times. Centrifugal sampling instruments normally determine deposition velocity
by determining penetration distance of particles passing through a duct that is
perpendicular or nearly perpendicular to the deposition force. If the inlet of the
entire duct is filled with dusty air, a mixed deposit occurs with a continual

Figure 1. Top view of cylindrical duct aerosol spectrometer.

decrease in the deposit of large particles along the duct. Instruments of this type
are called "semidispersive aerosol centrifuges." Centrifugal size classification
instruments in which the dusty air enters the duct in a thin layer covered by a
much thicker layer of clean air are called "aerosol spectrometers." Spectrome-
ters provide a continual separation and grading of particles according to deposi-
tion velocity and are currently the most widely used aerosol centrifuges. A
variety of centrifugal aerosol spectrometers have been built emphasizing specific
aspects of the operating characteristics. Operating principles of all these instru-
ments are similar. Thus, performance characteristics and operational limitation
can be illustrated by considering a simple instrument.
Figure 1 shows the top view of a cylindrical duct aerosol spectrometer. Several


Aerosol Centrifuges 5
instruments of this type have been built (Hochrainer, 1971; Matteson, Boscoe
and Preining, 1974; Anderson, 1976; Tillery, 1974). In most of these instruments
the depth of the duct (h) is greater than the width in the radial direction (R2 -
Ri). The elutriating layer of clean air is introduced into the deposition duct
through a laminating device. Air is drawn down the duct by applying a vacuum
behind a filter at the end of the duct. Sampling rate is regulated by drawing more
air from the duct than is introduced through the clean air inlet. The deficit enters
the duct as a thin layer of dusty air next to the inner wall. Flow in the duct is
laminar so particles tend to remain next to the inner wall. Flow through the
curving duct, and rotation of the duct results in a centrifugal force field and a
radial acceleration of the particles. Instantaneous radial velocity can be repre-
sented by ignoring acceleration and equating centrifugal force to the resistance
given by Stokes law:

dR pCD2 27r2V2R
dt = 9r (1)

where R = radial position, t = time, p = particle density, C = Cunningham'Slip
Correction, D = particle diameter, v = rotational rate in revolutions per unit
time, and qr = viscosity of duct gas.
Penetration distance (f), or distance particles will travel down the duct from
inlet to deposition, can be determined for particles of a given size by assuming a
velocity profile across the duct given by

T= f(R, b), (2)

where b = vertical position in the duct, and t = distance from the sample inlet.
For tall narrow ducts of height h, penetration distance at the center line (b = h/2)
can be determined by assuming that the velocity profile is parabolic across the
radial dimension with a maximum at the center (R2 + R 1)/2, or

(d 6U -R-R (R R1)2 (2a)
R, +R, Rt2 R (R2 R1)29

where U = Q/h(R2 R1) = average velocity in the duct, and Q = volumetric
flow rate.
Maximum penetration distance (VD) for particles of diameter D results from
particles entering the duct at R1. Thus, tD is determined by solving Eqs. (1) and
(2a) for dt and integrating across the velocity profile from R, to R2, i.e.,

9r- K U ,R2 + R1 2R1 R2 R
D r2pCD2k l(R2 Ri) (R2 Ri)2 lh

Inertial Classification

K = ratio of center line velocity to average velocity (1.5 for tall narrow ducts).
For a given instrument this can be reduced to

D = "7pCD22) (3a)

where I is the instrument geometrical constant given by

Kj R2 + RI 2RR2 RI
I = +KR2 + R + 2RIR2 n (n ). (3b)
h (R2 R )2 (R2 R1)3" '2) (3b)

Figure 2 is a plot of penetration distance against QI/pCD2v2 and experimental
data from three different cylindrical instruments. Good agreement is obtained for
the instruments having rectangular deposition ducts (Hochrainer, 1971; Ander-
son, 1976; Tillery, 1974). In the square duct, instrument penetration is somewhat
greater than predicted (Matteson, Boscoe and Preining, 1974). Eq. (3a) illus-
trates some of the operational limitations. The total flow rate Q must be limited to
prevent particles of interest from penetrating the deposition duct. Small particle
collection requires long ducts because penetration distance is inversely propor-
tional to particle diameter squared.
Eq. (3a) indicates that high flow rates and small particle sizes can be compen-
sated for if sufficient rotational speed is used. However, in addition to mechani-
cal difficulties involved with high rotational rates, secondary flows develop as a
result of Coriolis forces and inertia in the rotating curved duct (St6ber and
Flachsbart, 1969b). These secondary flows cause localized displacement of the
dusty air layer resulting in distorted deposition patterns and disagreement be-
tween deposition points and calibration curves. Secondary flows develop as
vortices, with opposite directions of rotation in the top and the bottom of the
duct. Depending on the relationship between direction of flow in the duct and
rotation of the duct, the direction of secondary flow in the center of the channel
will be toward the deposition wall or toward the center of rotation. When the
period of the vortex flow is approximately equal to or greater than the time
required for the sample to pass through the deposition duct, the deposition
patterns on the outer wall will not be seriously disturbed. However, the deposi-
tion rate in the center of the channel will be altered depending on the direction of
the vortex. This results in deviations from predicted penetration distance. This
effect is most important for small particles collected at the end of long ducts.
When the period of the vortex flow becomes significantly less than transit time
through the duct, marked distortions can occur in the deposits, making it difficult
or impossible to relate deposited and sampled concentrations as a function of
particle size. The development of secondary flows that distort deposition patterns
limits the range of permissible flow rates and rotational speeds.
Deposition patterns on the filter located at the end of the duct (Figure 3)
illustrate the location of secondary vortices at the top and bottom of the duct
(Tillery, 1974). In this case, the secondary flows did not seriously disrupt deposi-

I0 0 O I Il I I I' I II 1 I I I I '-

10 *

10 -

_-- --*

0 .1 I 1.
3 4 5 0
10 10 2 10 10
Ql /CD i
Figure 2. Penetration distance as a function of Ql/pCD2v2.

Figure 3. Filter deposit from concentric instrument.

Inertial Classification

tion patterns on the outer wall. A discussion of secondary flow effects in aerosol
centrifuges and illustrations of distorted deposition patterns are given in several
papers (St6ber and Flachsbart, 1969b; Stober, 1976; St6ber, Flachsbart and
Boose, 1972; Stober and Boose, 1973).
Assumptions used in derivation of Eq. (3), and failure to consider the effect of
secondary flows, limit the accuracy of these analytical relationships. However,
Eq. (3) defines parameters that regulate performance of these instruments and is
adequate for design purposes. An empirical calibration should be used for deter-
mination of aerosol size distributions. Many instruments in common use have
deposition duct shapes that prohibit mathematical analysis (Stober and
Flachsbart, 1969b; Moss, Ettinger and Coulter, 1972; Kotrappa and Light,
1972a); these instruments must be calibrated experimentally. Calibration is often
accomplished by sampling an aerosol produced by nebulizing an aqueous sus-
pension of monodisperse polystyrene latex particles. Evaporation of the droplets
leaves aerosol particles consisting of multiples of the fundamental particle size.


Figure 4. Deposition of polystyrene latex particles in a spinning, spiral duct spectrometer.

These particles will be collected in discrete bands in a spectrometer. Samples are
normally collected on metal or paper foils lining the outer wall of the deposition
duct. Figure 4 shows a foil with a collection of polystyrene latex particles. The
aerodynamic diameters of aggregates of uniform sized spheres as a function of
the number and configuration of particles in the aggregate have been determined
by Stbber, Berner and Blaschke (1969). These values can be used with measure-
ment of penetration distance on the foil to determine aerodynamic diameter (DA)
as a function of penetration distance. A calibration curve determined in this
manner is shown in Figure 5 (Moss, Ettinger and Coulter, 1972).
The resolution of aerosol spectrometers is normally defined in terms of the
ratio of the range of particle diameters found at a location to the diameter
expected at the location (8D/D), and is a function of the deposit width for
monodisperse particles. The deposit width for monodisperse particles results
from particles having different radial positions at the duct inlet. Particles entering
the duct at the inner wall (R = R1) will penetrate the farthest as they have the
greatest distance to travel to the deposition surface. Particles of the same size

Aerosol Centrifuges 9
entering the deposition duct at R = R, + A, where A is the thickness of the dusty
air layer, will penetrate a shorter distance. An expression for the resolution can
be obtained by differentiating Eq. (3) with respect to R1 and D

8DA 'D 2R2 (R2 Ri) + R2(R2 + R1) In (R1/R2) 5R,
DA 2fD R22 R12 + 2RiR2 In (R1/R2) (R2 Ri) (4)

An estimate of the ratio of width of deposit to penetration distance for particles of
the same size can be obtained by determining penetration distance for particles


J 1.0- 1000 RPM 15.4 1/min

2 3000 RPM, 15.4 I/min

0 D

0 .0 1 1 I I 1i I I I I I 1 1 1 I I I I 1I I I1111
Qo 0 18 36 54 72 90 108 126 144 162
Figure 5. Calibration curve for spinning, spiral duct spectrometer.

entering the duct at R, and at Ri + A in the same manner that Eq. (3) was
determined, giving

S2R2A A2 + 2RR2ln R+A)
(D R22 R,2 + 2RiR2 In (R,/R2)

The size of A depends on the ratio of the sampling rate S to the total flow rate
Q. A reasonable approximation for the velocity profile given in Eq. (2a) over the
range 0.02
A = 0.612 (R2 R1). (6)

Using this approximation of A in Eq. 5 for the instruments described by Hoch-
rainer (1971) and Tillery (1974) indicates that

-D when 0.02 < 0.2. (7)
1e Q Q

Thus, assuming 5R is equal to the dusty air layer thickness, spectrometer resolu-
tion is a function of the ratio of sampling rate to total flow rate, i.e.,

8DA -S 12R2(R2 RI) + R2 (R2 + R1) In (R1/R2) S1/2
DA 2Q+ 0.612 R22 R12 + 2RiR2 In (R,/R2) IJ (8)

A plot of resolution for two different sizes of spectrometers is given in Figure 6.
Resolution improves as R2 R, decreases. This effect is most important for
smaller instruments. The circles on the curve for R2 = 7.0 cm are two relative
standard deviations for particles found on a one-eighth inch diameter electron
microscope grid located in the center of the collection foil (Tillery, 1974). The
experimental resolution is better than predicted. This would be expected, since in
the laminar flow velocity profile, fewer particles enter next to the inner wall, so
the thickness of the dusty air layer (A) is effectively smaller than predicted.
The radial position of the particles at the end of the duct depends on deposition
rate, resulting in size separation on the filter located at the end of the duct (Figure
1). This separation is not amplified by air flow normal to the deposition direction.
Separation will thus be less with more overlap. An estimate of the width of the
deposit can be made by calculating the radial position for particles entering the
duct at the extremes of the dusty air layer (R1, R1 + A). A plot of the fraction of
the duct width in which particles are found as a function of particle size is given
in Figure 7. Both curves are for an instrument having R2 equal to 7 cm, operated
at 3000 RPM, with a total flow rate of 5 epm and a sampling rate of 0.5 1pm.
With these operating conditions, particles having VpCD = 0.5 pm will be
deposited on both the filter and the outer wall of the duct. The curve indicates
poor separation of larger particles that are more than halfway across the duct.
Particles larger than 0.25 m in diameter will be deposited across more than 30%
of the filter with the operating conditions described above. Half of the deposit of
0.3 ,im particles will overlap the deposit of 0.4 /m particles. The separation
of small particles found close to the inner wall will be better. The thickness of the
dusty air layer for an S/Q ratio of 0.10 is about 19% of the duct width. The filter
deposit thickness for equal size small particles tends to be less than this value
because particles entering the duct at R, have a lower velocity down the duct,
thus remaining in the centrifugal force field for a longer period of time than
particles entering at Ri + A.
Penetration distance in aerosol spectrometers having ducts of more compli-
cated geometry cannot be determined analytically. However, the parameters that
affect performance of these instruments can be determined from the equations for
the cylindrical instrument. The form of I in Eq. (3a) will vary with deposition


Inertial Classification

Aerosol Centrifuges

duct geometry. The penetration distance for all spectrometers of this type will be
directly proportional to flow rate through the duct and inversely proportional to
the square of the rotational rate and particle diameter. The resolution of spec-
trometers is dependent on the ratio of the sampling rate to the total flow rate.
Resolution also depends on the width of the deposition duct. However, this effect
is not very important for instruments having R2 greater than 5.0 cm.
The first centrifugal aerosol spectrometer was built in 1950 by Sawyer and
Walton (1950). A schematic view of this instrument called the Conifuge is shown
in Figure 8. The instrument consists of a narrow annulus between two right

0.35 ,, ,

0.25 (R2 Ri) 2.0cm

(R 2- R1) 1.Ocm

0.15 (R2 RI) = 0.5cm

0.05 R2= 3.0cm -

< | 00 1 I I I l I I

0.25 (R2 RI) = 2.0,cm
(R2 R 1 1.Ocm

0.15 ,(R2 Ri) 0.5 cm

S* R2 = 7.0cm
0.05 I -

0.02 0.06 0.10 0.12 0.20
Figure 6. Resolution for spinning, concentric duct spectrometer.


Inertial Classification

cones. The cones, in a sealed chamber, are rotated by an electric motor. The
rotating cones function as a centrifugal pump pulling air in the top and out
through jets located around the cone base. Clean air recirculates through the
sealed chamber and the top of the cones. When air is drawn from the sealed
chamber, an equal amount of dusty air comes down the aerosol inlet, spreads out
in a thin layer over the inner cone, and starts down the annulus covered by a
much thicker layer of clean air. In the annulus, the particles are subjected to a
centrifugal force accelerating the particles toward the outer wall as they pass
down the channel. All particles have essentially the same distance to travel to the
outer wall so they are deposited in a continuously graded spectrum according to
radial velocity. Velocity down the channel decreases with distance due to the
increase in cross-sectional area as centrifugal force increased due to greater

1.0 1 T I I I I I I I

0.4 (R2- R1) 1.0 cm

(R2 R ) = 2.0 cm


0.1 I I I I II I I 1 1 11i
0.01 0.02 0.03 0.05 0.10 0.2 0.3 0.5 0.7 1.0

Figure 7. Filter deposit width as a function of particle size for spinning, concentric duct spectrometer.

radius. This tends to compress the spectrum with respect to penetration distance.
Operational limits of this type of instrument result from the requirement of
laminar flow in the annulus and the necessity of a thin layer of dusty air, thus
limiting sampling rates. A thin layer of dusty air is required to provide reasonable
resolution or separation. This is illustrated by the particle trajectories shown in
Figure 8. Equal size particles will deposit over a length At depending on where
they enter the annulus. The thinner the dusty air layer, the better the resolution
and the lower the sample rate.
An equation similar to Eq. (3) relating penetration distance to particle size,
Conifuge design, and operating conditions has been developed by St6ber and
Zessack (1966). This equation is

4( D + -a3) sin0 + 3w (D2 a2) COS20 = 322


where fa = distance from the top of cone to sample inlet, GD = maximum
penetration distance for particles of diameter D, and 0 = half angle of the cone
(Figure 8). Several simplifying assumptions were necessary in deriving the equa-
tion. However, Mercer (1973) has shown that the equation does a reasonable job
of predicting the performance of several Conifuges.
The width of the deposit for equal size particles is dependent on the width of
the dusty air layer as shown in Figure 8. The width of the dusty air layer is
indicated as equal to the separation of the sample inlet funnel and the inner cone.
Dusty air layer thickness is actually determined by the fraction of the annulus


Figure 8. Schematic view of the Conifuge.

required to carry the sample flow rate. This is given in Eq. (6) as a function of the
ratio of sample flow rate to total flow rate for a parabolic profile across the short
dimension of a rectangular duct. The ratio of the deposit width, for uniform sized
particles, to maximum penetration distance (AVD/GD) can be determined by cal-
culating penetration distance for particles entering the annulus at the extremes of
the dusty air layer. Calculations carried out assuming fa = 0 indicate a slight
dependence on particle size with (AD/GD) decreasing slightly with increasing
penetration distance and decreasing particle size as shown in Table I. The most
important parameters in determining (AD/ D) are the half angle of the cones (0),
and the ratio of the sample flow rate to total flow through the annulus as shown in
Figure 9. The lines in Figure 9 are the average values for the particle size range

Aerosol Centrifuges


TABLE I. Ratio of Deposit Width to Penetration Distance
for 45* Conifuge Operated at 4000 RPM
with a Total Flow Rate of 13 cc/sec.


(,Tm) (cm) S/Q=0.01 0.05 0.10 0.15 0.20

0.1 14.75 0.030 0.070 0.102 0.129 0.152
0.2 9.23 0.030 0.070 0.103 0.130 0.153
0.3 7.00 0.030 0.071 0.103 0.130 0.154
0.4 5.75 0.031 0.071 0.104 0.131 0.155
0.5 4.93 0.031 0.071 0.104 0.131 0.155
0.6 4.35 0.031 0.072 0.105 0.132 0.156
0.7 3.91 0.031 0.072 0.105 0.132 0.157
0.8 3.56 0.031 0.072 0.106 0.133 0.157
0.9 3.28 0.031 0.073 0.106 0.133 0.158
1.0 3.05 0.031 0.073 0.106 0.132 0.158
2.0 1.86 0.032 0.075 0.109 0.137 0.162
3.0 1.39 0.033 0.076 0.111 0.140 0.165
4.0 1.12 0.033 0.077 0.113 0.142 0.168
5.0 0.95 0.034 0.078 0.115 0.144 0.170




0.02 0.06 0.10 0.14 0.18


Figure 9. Length of deposit for equal size particles in two Conifuges.


given in Table I for two different "Conifuges" (450, 300). The relative standard
deviation of (AG/t2D) over the range of diameters for each value of S/Q was less
than 5%. Resolution defined as AD/D at any penetration distance is proportional
to (AtD//D)- Some decrease in resolution results from broadening of the nonrotat-
ing, dusty air layer as it joins the rotating clean air layer at the lower edge of the
inlet tube. Instruments have been designed that rotate both air flows prior to
entering the deposition annulus (Hochrainer and Brown, 1969; Berner and
Reichalt, 1969). These instruments have improved resolution, but they also have
significant inlet losses.
Stober and Flachsbart (1969a) designed a Conifuge having a truncated conical
annulus. This design markedly increased the area of the sample inlet without
increasing the thickness of the dusty air layer in the annulus. This instrument was
designed to operate over a wide range of flow rates and rotational speeds and
revealed the limitations of Conifuge design. At high rotational speeds and high

TABLE II. Conifuge Characteristics

Particle Size
Conifuge 0 L(cm) W RPM S(tpm) Q(tpm) Range (p/m) (S/Q)"2
Sawyer and Walton, 1950 300 6.0 0.58 3,000 0.025 0.800 0.48-21.0 0.18
Keith and Derrick, 1960 450 12.5 1.00 8,000 0.300 3.240 0.09-5.0 0.03
Tillery, 1967 450 12.5 1.00 4,000 0.085 0.780 0.07-3.0 0.33
Hauck and Schedling, 1968 450 12.5 1.00 3,000 0.674 6.740 0.50-7.0 0.31
Hauck and Schedling, 1968 450 12.5 1.00 5,000 0.224 2.240 0.10-2.00 0.31
Hochrainer and Brown, 1969 30' 6.0 0.57 10,000 0.026 1.180 0.12-2.00 0.15
Stober and Flachsbart, 1969b 200 19.1 1.00 1,500 0.500 8.300 0.5-10.0 0.24
Stober and Flachsbart, 1969b 200 19.1 1.00 6,000 1.000 9.700 0.085-3.00 0.32
Stober and Flachsbart, 1969b 200 19.1 1.00 3,000 1.200 10.000 0.22-6.00 0.35
Stober and Flachsbart, 1969b 200 19.1 1.00 6,000 0.100 3.200 0.04-1.50 0.18
Hochrainer and Brown, 1969 90 4.0 0.50 10,000 0.012 0.576 0.55-2.00 0.14

flow rates, Coriolis forces caused secondary flows that disrupted the deposition
patterns, and caused mixing that led to diffuse deposition patterns. Limiting
conditions for regular deposits over the entire deposition surface were achieved at
total duct flow of about nine Cpm at 6000 RPM.
The characteristics of a number of Conifuges are given in Table II. The
sampling rate for more than half of the instruments is less than 0.1 (pm. The
instrument of Stober and Flachsbart, having a sampling capability of around a
liter per minute, is large and practically limited to laboratory use. The 900
instruments (Hochrainer and Brown, 1969; Berner and Reichalt, 1969) consist of
an annulus between two right cylinders with flow from top to bottom similar to
flow through Conifuges. A rigorous determination of penetration distance in 90
instruments has been made by Stober and Zessack (1966) and is given by

97)Q (In R2 In R1)
S27T3pCD2(R22 R2)v2 (10)

Aerosol Centrifuges


Inertial Classification

where 1D is the maximum penetration distance measured down the side of the
outer cylinder, and R1 and R2 are the respective radii of the inner and outer wall.
Conifuge samples are normally collected on foils lining the outer cone or on
slides located in slots cut into the outer cone. The foil or slide can be cut into
strips and analyzed to determine the mass or particle number with respect to the
diameter difference across the strip given by the calibration curve of the instru-





Figure 10. Blowup of spinning, spiral duct spectrometer.

ment. The use of slides or portions of the foil requires appropriate area correc-
tions. Determination of size distribution requires corrections for inlet losses and
for small particles that penetrate the annulus.
The low sampling rate of the Conifuge and the instabilities encountered at high
rotational speeds and flow rates indicated the need for a different design. In 1969
Stober and Flachsbart (1969b) designed a centrifugal spectrometer with a spiral


duct similar to a semidispersive centrifuge designed by Kast (1961). A blow-up
view of a similar instrument is shown in Figure 10 (Moss, Ettinger and Coulter,
1972). The instrument consisted of a long (-~ 180cm), rectangular duct (3.3 cm
deep by 1.0 cm wide except at the start of the duct where the width is 1.7 cm) cut
into a flat cylinder. Air flow through the instrument is controlled externally
through ducts separated by rotating seals in the drive shaft of the instrument and
in the base of the instrument. Filtered air is introduced into the deposition duct
through a laminator consisting of several parallel foils. Suction is applied at the
end of the duct. The suction flow rate is greater than the laminating air flow with
the difference being drawn down the aerosol inlet, on the axis of rotation. The
center inlet directs the dusty air down the deposition duct as a thin layer next to
the inner wall. Particles passing down the spinning duct are subjected to an
increasing centrifugal force accelerating the particles toward the outer wall where
they are deposited on a collection foil. The particles are separated and deposited
on the outer wall in a continuously graded spectrum according to deposition rate
or aerodynamic diameter.
The complicated geometry of the deposition duct limits analytical determina-
tion of particle trajectories or penetration distances. However, the functional
relationship of the parameters regulating performance is similar to the cylindrical
duct instrument. Penetration distance is inversely proportional to the particle
diameter squared and rotational rate squared, and is directly proportional to total
flow rate. Resolution is dependent on the ratio of the thickness of the dusty air
layer to the thickness of the clean air layer, or the ratio of the two flow rates.
Total flow through the instrument and rotational rate are limited by the develop-
ment of secondary flows that disrupt deposition patterns. Instruments of this
design operate satisfactorily at 3000 RPM with total flows as high as 19 fpm.
The particle size range deposited on the collection foil covers almost two orders
of magnitude (0.09Aim-5p/m) under these conditions (Stober and Flachsbart,
1969b). At 6000 RPM with a total flow of 10 tpm the secondary flows markedly
disrupted deposition patterns (Stober and Flachsbart, 1969b). Theoretical con-
siderations of secondary flow development in rotating curved ducts indicate that
tall, narrow ducts will tend to suppress secondary flows (Stober, 1976). Instru-
ments have been built with ducts as deep as 6 cm in an effort to extend the
operational range. However, only marginal improvement has been noted with the
deep duct instruments.
The physical size of long duct instruments and the large deposition area make
these instruments somewhat impractical for portable use. Several small, single
turn, spiral duct spectrometers have been built for particle separation and to
provide reasonably portable sampling instruments. The first short duct instrument
was designed by Kotrappa and Light (1972a). This instrument, called the
Lovelace Aerosol Particle Separator, has essentially a single turn, spiral duct
with approximately a one to three expansion in the radial direction over the
length of the duct. The duct was expanded to utilize the Conifuge effect of
decreasing velocity to enhance collection of small particles (Kotrappa and Light,
1972b). However, expansion in the radial direction increased the distance to

Aerosol Centrifuges


Inertial Classification

deposition site, and decreased the average centrifugal force in the duct compared
to a constant width duct with the same outer radius. These effects compromise
the enhanced deposition due to increased time in the centrifugal force field.
Stober (1976) has also built several short duct instruments utilizing constant
width, narrow ducts or ducts of decreasing width to suppress secondary flows.
The constant duct width instrument is shown in Figure 11 with top and center
inlet removed. The characteristics and performance of spiral duct spectrometers
are listed in Table III. The particle size collection range for the short duct
instruments is for particles collected on the deposition foil. Smaller particles are
collected on the filter at the end of the duct.


Figure 11. Short, spinning, spiral duct spectrometer.


TABLE III. Characteristics of Spiral Duct Aerosol Spectrometers

Deposition Duct Dimensions (cm)

Maximum Total Duct Sample Particle Size
Reference Length Width Depth Radius (cm) RPM Flow ({pm) Flow (epm) Collection Range (A.m)

Stober and Flachsbart, 1969a 180 1 3.3 12 3000 10.0 0.5 0.09-5.0
Stober and Flachsbart, 1969a 180 1 3.3 12 6000 5.0 0.3 0.18-4.0
Stober and Flachsbart, 1969a 180 1 3.3 12 3000 19.0 1.6 0.09-5.0
Moss et al., 1972 178 1 5.08 12 3000 15.4 0.44 0.10-4.0
Kotrappa and Light, 1972 45 1.5-3.9a 3.9 8 3000 5.0 0.40 0.6-3.5
Kotrappa and Light, 1972 45 1.5-3.9a 3.9 8 3600 5.0 0.40 0.48-3.5
Kotrappa and Light, 1972 45 1.5-3.9a 3.9 8 4500 5.0 0.40 0.42-3.5
St6ber, 1976 55 1.67 3.3 8 3000 5.0 0.4-3.5
Stober, 1976 55 1.67 6.0 8 3000 5.0 0.32-2.0
St6ber, 1976 55 1.6-0.8b 3.4 8 5000 5.0 0.22-2.0
St6ber, 1976 55 1.6-0.8b 3.4 8 3000 10.0 0.48-3.5

a. Expands over length from 1.5 to 3.9 cm.
b. Contracts over length from 1.6 to 0.8 cm.

Inertial Classification

In all rotating, spiral duct samplers a portion of the inlet rotates so that inlet
loss corrections are necessary to determine size distributions. Losses can be
calculated for well-defined rotating geometries (Mercer, 1973). In most cases the
inlet geometry is not simple and is complicated by interfaces between fixed and
rotating components so collection efficiencies must be determined experimen-
tally. Correction factors relating center line deposition concentrations to airborne
concentrations have been determined by Kops et al. (1974) for the long, spiral
duct of St6ber. Center line deposition concentration for particles smaller than 0.2
/m was found to be enhanced by the secondary flows. Center line deposition
concentration for particles larger than 0.4 gm was found to be low because of
inlet losses. These losses are quite high for particles larger than 2 /m in diameter
(St6ber, Flachsbart and Boose, 1972; Kops, Hermans and Van De Vate, 1974).
Similar corrections are required if the entire height of the foil deposit is used to
determine size distributions, as inlet losses result in lower concentrations on the
lower half of the foil for large particles (DA > 0.4/gm) (Ferron and Bierhuizen,
1974). These losses are most significant for large particles, and so are important
when determining mass distributions.
Centrifugal aerosol spectrometers have also been built in which the particle
flow is in the radial direction (Burson, Keng and Orr, 1967/8; Redkin, 1970). In
these instruments a thin stream of particles is injected into a rotating chamber of
still air or a chamber in which the air is also flowing radially at a low velocity.
The particles are deviated from the radial direction due to the interaction of
Coriolis forces and drag. Thus, small particles deviate only slightly and deposit
on the outer wall close to the intersection of the radial line from the inlet to the
outer wall. Large particles deviate significantly and deposit on the outer wall at
some distance from the intersection of the radial line with displacement being
opposite the direction of rotation.
Several types of semidispersive aerosol centrifuges have been built. The most
frequently encountered device of this type is the Goetz Aerosol "Spectrometer"
(Goetz, Stevenson and Preining, 1960). The instrument shown in Figure 12
consists of two helical ducts of 2.5 turns on the outside of a 30 right cone. The
radius of the cone is 3.8 cm at the base. The cone forming the outer wall of the
ducts is removable and lined with a foil or paper for collection and subsequent
analysis of the deposit. The cones are rotated by a variable speed motor at rates
up to 24,000 RPM. The impeller action of the rotating ducts draws the sample
down the inlet and through the ducts. Flow rate through each duct is independ-
ently controlled by jets located at the end of the duct. The farther the particles
travel down the duct the greater is the centrifugal force and the deposition
velocity. The deposit at any distance down the channel contains a mixture of all
sizes of particles having a maximum penetration distance equal to or greater than
the distance. Maximum penetration distance for a given sized particle is the
distance traveled by a particle, of the given size, entering the deposition duct next
to the inner wall. The increase in deposition velocity as particles travel down the
duct results in a variation in deposited concentration of a given size particle out to
the maximum penetration distance. A theoretical study indicates an increase of





Figure 12. Goetz Aerosol "Spectrometer."


Inertial Classification

deposited concentration with penetration distance up to near maximum, then
decreasing to zero at the maximum (Stober and Zessack, 1964). An experimental
study indicates the theory is approximately correct with cutoffs at maximum
penetration distance not as sharp as indicated by theory (Baust, 1967). Turbulent
flow conditions exist for a short distance below the inlet to the ducts. Theoretical
treatment of deposition in the turbulent region would be difficult or impossible
and has not been done. Experimental determinations of deposition in this region
have been made (Baust, 1967; Raabe, 1967). Gerber has found that under limited
operating conditions with a modified instrument, center line deposition concen-
trations are constant (Gerber, 1971; Stober and Boose, 1973; Hochrainer, 1972).
However, marked variations in deposition concentration were found across the
channel requiring correction factors to determine particle size distributions from
center line deposits (Gerber, 1971). Complicated deposition patterns have also
been noted with a spiral duct, semidispersive aerosol centrifuge (Ruger, Maiwald
and Feddersen, 1968).
The principal advantage of centrifugal aerosol spectrometers is the collection
of the sample as a continuously graded spectrum in terms of aerodynamic or
inertial properties. This separation simplifies the determination of particle size
distributions with respect to aerodynamic diameter for most parameters of inter-
est. The resolution, or separation of particles according to aerodynamic charac-
teristics achieved by these instruments is not attainable by any other means. The
principal disadvantages are relatively low sampling rates, collection of the sam-
ple over a large surface, and delay between the time of sampling and determina-
tion of results.
The only advantage of the semidispersive centrifuges is the high sampling rate
(up to 125 tpm for Kast instrument [1961]) relative to the centrifugal spec-
trometers. The variations in deposit concentration, poorly defined maximum
penetration distance, and particle losses in the turbulent flow region make deter-
mination of particle size distributions very difficult and susceptible to errors.
At the present time few of these instruments are commercially available.
However, plans for almost all of the instruments are available and fabrication can
be carried out by most well-equipped machine shops.


Anderson, P. L., 1976: Development of a centrifuge for source sampling to determine particle size
distribution. Environ. Sci. Technol., 10, 145.
Baust, E., 1967: Use of the Goetz Aerosol Spectrometer to measure the size spectra of polydispersed
aerosol. Staub-Reinhalt. Luft, 27, 16.
Berner, A. and H. Reichalt, 1969: On the inlet slit systems in conifuges-part I: The ROSL system.
Staub-Reinhalt. Luft, 29, 5.
Burson, III, J. H., E. Y. H. Keng and C. Orr, Jr., 1967/8: Particle dynamics in centrifugal fields.
Powder Technol., 1, 305.


Aerosol Centrifuges

Ferron, G. A. and H. W. J. Bierhuizen, 1974: The measurement of polydisperse aerosols with the
spiral centrifuge. Aerosol in Naturwissenschaft, Medizen und Technik, Jahreskongress der
Gesellschaft fuir Aerosolforschung e.v.
Gerber, H. E., 1971: On the performance of the Goetz Aerosol Spectrometer. Atmos. Environ., 5,
Goetz, A., H. J. R. Stevenson and 0. Preining, 1960: The design and performance of the aerosol
spectrometer. J. Air Pollution Control Assoc., 10, 378.
Hauck, H. and J. A. Schedling, 1968: A modified conifuge. Staub-Reinhalt. Luft, 28, 18.
Hochrainer, D., 1971: A new centrifuge to measure the aerodynamic diameter of aerosol particles in
the submicron range. J. Colloid and Interface Sci., 36, 191.
,__ 1972: On the reliability of measurements with the Goetz Aerosol Centrifuge, Discussion.
Atmos. Environ., 6, 699.
and P. M. Brown, 1969: Sizing of aerosol particles by centrifugation. Environ. Sci.
Technol., 3, 830.
Kast, W., 1961: Neues staubmessgerat zur schnellbestimmung der staubkonzentration und der
kornverteilung. Staub-Reinhalt. Luft, 21, 215.
Keith, C. H. and J. C. Derrick, 1960: Measurement of the particle size distribution and concentration
of cigarette smoke by the conifuge. J. Colloid and Interface Sci., 15, 340.
Kops, J., H. Hermans and J. E. Van De Vate, 1974: Calibration of a Stober Centrifugal Aerosol
Spectrometer. Aerosol Sci., 5, 379.
Kotrappa, P. and M. E. Light, 1972a: Design and performance of the Lovelace Aerosol Particle
Separator. Rev. Sci. Instrum., 43, 1106.
S and 1972b: Centrifuge Separator. U.S. Patent No. 3, 698, 626.
Matteson, M. J., G. F. Boscoe, and 0. Preining, 1974: Design, theory and calibration of a field type
aerosol spectrometer. Aerosol Sci., 5, 71.
Mercer, T. T., 1973: Aerosol Technology in Hazard Evaluation. New York, Academic Press.
Moss, 0. R., H. J. Ettinger and J. R. Coulter, 1972: Aerosol density measurements using a modified
spiral centrifuge aerosol spectrometer. Environ. Sci. Technol., 6, 614.
Raabe, 0. G., 1967: Calibration and use of the Goetz Aerosol Spectrometer. Assessment of Airborne
Radioactivity, Vienna, International Atomic Energy Agency, 417.
Redkin, J. N., 1970: Properties of atmospheric aerosol measured with a centrifugal spectrometer. J.
Geophysical Research, 75, 3563.
Ruger, G., E. Maiwald and C. Fedderson, 1968: A centrifugal separator and its properties. Staub-
Reinhalt. Luft, 28, 9.
Sawyer, K. F. and W. H. Walton, 1950: The "Conifuge": A size sampling device for airborne
particles. J. Sci. Instrum., 27, 272.
Stober, W., 1976: Design, performance and applications of spiral duct aerosol centrifuges. Fine
Particles, Aerosol Generation, Measurement, Sampling, and Analysis. New York, Academic
A. Berner and R. Blaschke, 1969: The aerodynamic diameter of aggregates of uniform
spheres. J. Colloid and Interface Sci., 29, 710.
__ and C. Boose, 1973: Developing flow and particle deposition in horizontal elutriators and
semidispersive aerosol centrifuges. Atmos. Environ., 7, 119.
and H. Flachsbart, 1969a: Aerosol spectrometry with a ring slit conifuge. Environ. Sci.
Technol., 3, 641.
_ and 1969b: Size separating precipitation of aerosols in a spinning spiral duct. Envi-
ron. Sci. Technol., 3, 1280.
and C. Boose, 1972: Distribution analyses of the aerodynamic size and the mass of
aerosol particles by means of the spiral centrifuge in comparison to other aerosol precipitators. J.
Colloid and Interface Sci., 39, 109.
__ and U. Zessack, 1964: On the theory of a conical aerosol centrifuge. Staub-Reinhalt. Luft,
24, 295.
S and 1966: The measurement of aerosol particle size spectra by means of centrifugal
separators. Z. Biol. Aerosol forschung, 13, 263.
Tillery, M. I., 1967: Design and calibration of a modified conifuge. Assessment of Airborne Radioac-
tivity, Vienna, International Atomic Energy Agency, 405.
1974: A concentric aerosol spectrometer. J. Amer. Industrial Hygiene Assoc., 35, 62.



OWEN R. Moss
Department of Radiation Biology and Biophysics, University of Rochester, Rochester, N.Y.

These comments deal with the calibration and use of the spinning spiral aerosol
spectrometer. I will first present some work of Dr. Ted Martonen (1975/76) and
then I will discuss some practical points I have encountered while operating these
centrifuge devices.
The overview paper on centrifuges by Tillery shows a drawing of the cen-
trifuge built at Los Alamos. A particle enters at the center of rotation, crosses the
channel, and deposits on the collection foil lining the outer wall. Martonen has
recently solved the equation for the motion of a particle under the influence of the
centrifugal and Coriolis forces in the channel. He assumed that there is no
secondary flow. A preliminary comparison of the theory with experimental cali-
bration points shows that the two curves diverge greatly for particles less than
0.8j/m aerodynamic diameter (Dae). Particles with Dae of 0.25 /m are predicted
to pass through the entire length of the channel. Instead, they deposit about half-
way down the channel for the operating conditions used in the comparison. The
instrument works much better than one would expect.
The centrifuge devices are used to sample a variety of aerosols. Air may be
only a small component of the sampled gas. Martonen investigated whether the
calibration of the instrument changes if the aerosol carrier gas is different from
the laminating air in the channel. He used carbon dioxide, air, and helium as
aerosol carrier gases. These three gases have specific gravities of approximately
1.4, 1.0 and 0.12, respectively. Figure 1 shows the results of a part of this work.
Under identical operating conditions, the more dense gas tends to stream across
the channel. This is very much like the results seen with cigarette smoke (Hinds,
1975; Martonen, 1976; St6ber, 1976).
This paper is based on work performed under contract with the U. S. Energy Research and
Development Administration at the University of Rochester Biomedical and Environmental Research
Project and has been assigned Report No. UR-3490-933.

Comments on Centrifuges

The aerosol carrier gas can be cleaned and used for the laminating air flow in
the channel. Martonen found that a laminating flow of carbon dioxide or nitrogen
required that the instrument be run at lower rotational speeds in order to maintain
distinct deposition patterns like those obtained during the initial calibration with
dry air. It is not clear whether the conditions he used in his experiments were a

1111111 I I1 II i Ii III I

LN b
j I

emissions might be close to what he was thinking of.

The instrument appears to keep its calibration under the various sampling con-
p \ \\

infor the haerosonnel carris 1.04. No change is seen in the desampling condositions wpattere identihcalibr. Clean
tion sampling condition were 3000 rpm and 8sol7 .pm.
good model fused the Stberaerosols the centrifuge might Rochbe used to relate the aerodynamic


Inertial Classification

diameter to the projected area diameter of the particles. This type of measurement
is made by gluing electron microscope (EM) grids to the collection foil. The loca-
tion of the grid provides a measure of the aerodynamic diameter of the particles
that land on it. A scanning electron microscope was used to observe the sample.
The grid-bars contained more particles on their surface than the carbon-film-
substrate, grid squares. This was very noticeable for the samples containing
particles with Dae greater than 5/pm. The particles were very irregular in shape.
For Dae around 2 um and less, there was little difference in the count density.
The size distributions measured on the grid bars and grid squares were essentially
the same. The grids are mechanically deformed in pulling them off the foil. The
carbon film substrate would vibrate more than the adjacent metal grid bars.
Particles would then be lost preferentially. This tendency to lose larger particles
should be kept in mind when counting. Care should be taken in handling the foil
or EM grid sample.
Obtaining a representative sample of a dust composed of particles larger than
3 gm Dae is a problem with the centrifuge units. Two particle sinks located at
the radius of the aerosol inlet tube from the center of rotation must be considered.
One sink is the point where the walls of the stationary inlet tube match up with
the rotating sides of the hole in the centrifuge head. The particle loss that occurs
at this point has been reduced by machining and polishing (Kotrappa, 1971/1972).
Another method that I have tried is to insert the stationary tube down into the rotor
until its end opens into the spiral channel. The aerosol particles now pass from a
stationary walled tube to one with a moving slit, the inlet into the channel. The
loss should be less with this arrangement than that predicted for rotating tubes
by Tillery (1976).
The other particle sink is the metal sheath used to form the aerosol inlet slit.
If the slit is moved away from the center of rotation, the deposition patterns from
the calibration aerosol become very sharp. The particle loss in such an inlet also
increases (Stober and Flachsbart, 1969). An inlet should allow the larger particles
(at least up to 10 pm Dae) to enter the channel. It should also allow calibration-
aerosol deposition patterns of minimum overlap. The deposition patterns should
be centered down the middle width of the sampling foil.
Figure 2 shows sketches of calibration aerosol deposition patterns for different
aerosol inlet configurations. The first inlet is the extended sheath inlet used for
high resolution studies. I have not drawn in the faint lines that tend to extend back
from the deposition patterns near the top and bottom of the foil. Particles with
D ae greater than 4 /.m are deposited inside this type of inlet and not on the col-
lection foil. The second inlet configuration is an open one. The singlet deposition
patterns become smeared.
If a slit is to be used, and one must be if the deposit is to be centered down the
collection foil, then it should be located no more than one inlet-tube-radius from
the center of rotation. Inlet 3 in Figure 2 shows one such configuration. Slits were
constructed at the closest point to the outer wall. None of the slits designed for
this position gave good deposition patterns. The aerosol appeared to have an
early involvement in the development of the secondary flow. This particular inlet


Comments on Centrifuges

showed that the top aerosol stream had a larger velocity component towards the
outer wall than the lower aerosol stream.
Another alternative is to place a slit at the base of a rounded step function. The
example is given as inlet 4 in Figure 2. Surprisingly, the slit works fairly well. It
can be shaped so that deposition occurs down the middle width of the collection
foil. Particles as large as 10 /m Dae appear to get in with very little loss. The
deposition pattern at the center end of the foil begins to tail off visibly where the
12 to 15 Am Dae particles would be expected to land (5 to 7 cm). The 10 /m Dae
particles then must be getting into the channel. Actually, the open inlet, number


I~ o( 1). 5 5 S0 9300
Siem) 5 is O
E)at(cp-1 J (^s ,81 ,36


a I I



Figure 2. Sketches of deposition patterns obtained with different aerosol inlets. "1" is the
distance down the foil. The small dotted lines indicate where the deposition pattern repeated due
to doublets or triplets. The centrifuge was the original Stober unit at the University of Rochester.
The basic sampling conditions were 1500 rpm and 10 fpm.

2 in Figure 2, works fairly well with a resolution on the foil of 5 to 10 percent.
I would like to make one last comment on sampling polydisperse aerosols. In
general, the sample is collected on a foil which is then cut and stored for
chemical and physical analysis. I found that in cutting the foil, some of the
material was lost on the scissors. When dealing with heavy deposits of dust
particles having Dae greater than 5 gm the loss can be more than 10%. One solu-
tion is to use foil of overlapping metal pieces. The pieces fit into a backing foil
containing slots. With a little practice, selected pieces can be removed or re-
placed without disturbing the others. Short aluminum strips have also been used
with success (Ferron and Bierhuizen, 1976).


ot w--c-

28 Inertial Classification


Ferron, G. A. and H. W. J. Bierhuizen, 1976: The measurement of polydisperse aerosols with the
spiral centrifuge. J. Aerosol Sci., 7, 5-11.
Hinds, W. C., 1975: Measurement of the aerodynamic size distribution of tobacco smoke. Paper #63.
Presented at Amer. Industrial Hygiene Assoc. Conf., Minneapolis.
Kotrappa, P., 1971/2: Personal communication. Lovelace Foundation for Medical Education and
Research, Albuquerque.
__ and M. E. Light, 1972: Design and performance of the Lovelace Aerosol Particle Separator.
Rev. Sci. Instrum., 43, 1106-1112.
Martonen, T., 1975/6: Personal communication. University of Rochester School of Medicine and
Dentistry, Rochester.
Raabe, 0., 1976: Design and use of the Mercer-style impactor for characterization of aerosol
aerodynamic size distributions. Presented at Aerosol Measurement Workshop, Gainesville, Fl.,
March 24-26.
Stober, W., 1976: Personal communication. Institute fur Aerobiologie, Grafschaft, Schmallenberg/
Sauerland, Germany.
S and H. Flachsbart, 1969: Size-separating precipitation of aerosols in a spinning spiral duct.
Environ. Sci. Technol., 3, 1280-1296.
Tillery, M., 1976: Aerosol centrifuges. Presented at Aerosol Measurement Workshop, Gainesville,
Florida, March 24-26.


Inhalation Toxicology Research Institute, Lovelace Foundation, Albuquerque, N. Mex.

Four years of experience employing the Lovelace Aerosol Particle Separator
(LAPS) for preparation of monodisperse particles and characterization of
aerosols have suggested several aspects of the engineering design which could be
changed to improve performance and reliability. Specifically, the original LAPS
units used three bearings on the drive shaft with the center bearing serving as a
seal between the inlet for LAPS air and the exhaust. In some units the center
bearing malfunctioned as a seal, leaking as much as 1.5 cc per sec of air between
inlet and outlet at a pressure difference of only 0.3 psig. Also, the bearings were
occasionally observed to generate a foreign aerosol which contaminated the
collected aerosol. This occurred when the bearing lubricant became aerosolized
and was deposited on the collection foil. Because of this loss of lubricant from
the bearings and possible slight misalignment of the three bearings with respect
to the LAPS shaft, premature bearing failure occurred in several LAPS units.
A complicating factor in use of the LAPS units at the Inhalation Toxicology
Research Institute (ITRI) is that those used to prepare monodisperse particles are
normally in glove boxes and may be highly contaminated with various long-
lived radionuclides such as 238Pu. Since the configuration of the bottom bearing
housing prevented repair inside a glove box, extensive and costly decontamina-
tion was required before the units could be removed to a clean work area for
Since a fairly major design change was required to solve the bearing and seal
problem it was decided to take this opportunity to incorporate other changes in
design to improve the LAPS and its reliability. A brief description of each of the
design modifications follows. The original design is represented in Figure 1. The
Work performed for the Division of Biomedical and Environmental Research, U.S. Energy Re-
search and Development Administration under Contract No. E(29-2)-1013.



Figure 1. Exploded view of the original Lovelace Aerosol Particle Separator. Approximate scale:
diameter of disk equals 17.8 cm.

Lovelace Aerosol Particle Separator

modified design is represented in Figure 2. A detailed view of the drive shaft,
magnetic seals and bearing housing is shown in Figure 3.


Magnetic Seals
Compact axial shaft seals were obtained from the Magnetic Seal Corporation,
West Barrington, Rhode Island. Three of these seals were used in the drive shaft
bearing housing so that the laminar air inlet and the LAPS exhaust outlet were
isolated from each other and from the ambient atmosphere and bearing lubricant.
Vented seal carriers were designed for effecting this separation. Another seal was
used on the aerosol inlet shaft to prevent bearing lubricant contamination and
aerosol dilution with ambient air.

Channel Profile
The inside wall of the spiral channel was redesigned for the final 1800 so that the
undesirable concave pouch at this end of the channel in the original design was
eliminated and the channel made more narrow to provide an improved aspect
ratio (Stober, 1976). This allowed a square channel end measuring 3.175 cm on a
side. The channel was shortened about 0.5 cm to allow for a larger filter holder,
but the effective length of foil did not change due to the use of the 0.5 cm of the
filter holder in front of the filter.

Channel Filters
A filter was added upstream of the laminating air to remove possible contami-
nants from the drive shaft seals. The spiral channel outlet filter was made to
occupy the entire area of the channel outlet eliminating the constriction at the end
of the original spiral.


Filter Holders
An injection-molded polystyrene filter holder was designed to support both the
inlet and the outlet filter. The holders are inexpensive, can be loaded and stored
for use as needed, and, in the case of the outlet filters, allow for convenient and
safe storage of collected aerosol particles. The filter holders were designed to seal
against the channel around their perimeter eliminating leakage that was some-
times observed in the original LAPS design.







Figure 2. Exploded view of the modified Lovelace Aerosol Particle Separator. Approxi-
mate scale: diameter of disk equals 17.8 cm.



Lovelace Aerosol Particle Separator

Laminator-Aerosol Inlet
The two piece clean air laminator and aerosol inlet were combined into a one
piece cylindrical unit. The one piece design can be fabricated more easily with a
higher degree of precision. Laminators and aerosol inlets no longer need to be
supplied in matched pairs for a particular LAPS.

Drive Shaft Bearing Housing
This support base was redesigned so that assembly and disassembly could be
accomplished with relative ease inside a glove box with a minimum of inexpen-
sive special tools required. Should bearings or seals require replacement after a
LAPS has been contaminated during use, expensive decontamination is not

Drive Shaft and Aerosol Inlet Shaft
These shafts were made of tool steel and brass, respectively, in the original
LAPS. The modification uses 304 stainless steel for both. Rusting and corrosion
as a result of decontamination procedures are thus eliminated.

Alignment of Spiral Rotor and Lid
Alignment of these mating parts in the original LAPS design was accomplished
using the protruding ends of the aerosol inlet and laminator as a positioning key.
This caused wear and resulting inexactness of fit between parts whose alignment
was relatively critical. The modified design has dowel pins for alignment.


The original LAPS was made of 6061T6 aluminum which, while being relatively
easy to machine, is soft and easily damaged. The modified LAPS is made of
7075T6 aluminum which is easier to machine and much less likely to suffer
damage as a result of routine handling.

Top and Bottom Disks
The bottom disk was made thicker for greater stiffness and better sealing against
the spiral rotor and for more contact with the drive shaft taper. The top disk was
thickened for stiffness and for more metal support of the countersunk cap screws
used in assembly.


Inertial Classification

Bolt Pattern

The modified LAPS has a new bolt pattern including three centrally located
screws for more uniform clamping and a more secure seal of the top disk to the
spiral rotor. The threads in the rotor are reinforced with stainless steel "Keen-
sert" thread inserts.

Dynamic Balance

Dynamic balance of the original LAPS was accomplished by attaching appro-
priate steel counter weights to the outside of the top and bottom disks each being
secured with four or five number 10-32 flathead screws which penetrated approx-



Figure 3. Cutaway section view of the driveshaft bearing housing of the modified Lovelace Aerosol
Particle Separator.

imately 0.3 cm into the disk. These weights represented a potential hazard since
some had been observed to loosen during LAPS operation. In the modified
design appropriate quantities of metal were removed from heavy sections of the
rotor to accomplish the counterweighting by metal removal rather than metal
addition. In all, more than 200 g of metal were removed.

Center Bearing

Eliminating the center bearing from the drive shaft and thus eliminating any
preloading of the bearings because of slight misalignment of the shaft and bear-
ing bores should eliminate premature bearing failure. The use of three magnetic
seals in place of the center bearing should protect the two remaining bearings


Lovelace Aerosol Particle Separator 35

from damage by cleaning solvents which might accidentally enter the bearing
housing during decontamination. Such damage has occasionally occurred in
LAPS units of the original design.
Acknowledgements-The authors acknowledge the help of Sandia Corpora-
tion, in particular John Lewis and his associates for drawings, structural calcula-
tions and configurations. Also, Mr. W. (Butch) Yaple of Continental Machine
Company for his help on the design of the laminator-inlet slit.

Kotrappa, P. and M. E. Light, 1972: Design and performance of the Lovelace Aerosol Particle
Separator. Rev. Sci. Instrum., 43, 1106-1112.
Raabe, 0. G., H. A. Boyd, G. M. Kanapilly, C. J. Wilkinson and G. J. Newton, 1975: Development
and use of a system for routine production of monodisperse particles of 23Pu02 and evaluation of
gamma emitting labels. Health Phys., 28, 655-667.
Stober, W., 1976: Design, performance and applications of spiral duct aerosol centrifuges. Fine
Particles, B. Y. H. Liu, Ed., New York, Academic Press, pp. 351-397.


Naval Research Laboratory, Washington, D.C.

A major advertised advantage of the Goetz aerosol spectrometer (Preining,
Stevenson and Goetz, 1959; Goetz et al., 1960; Goetz and Kallai, 1962) over its
immediate predecessor, the original conifuge of Sawyer and Walton (1950), was
that its sampling rate was two orders of magnitude larger. This advantage made the
Goetz centrifuge popular when it first appeared commercially in the early 1960's.
In the years that followed, however, confidence in the instrument declined when
many papers in the literature gave inconsistent performance evaluations of the
instrument. Most of the studies (St6ber and Zessack, 1964; Baust, 1967; Raabe,
1967; Stober and Boose, 1973; Hochrainer, 1972) reported difficulties in using
the instrument, although the originators and others (Gerber, 1971; Horvath,
1973; Gerber, 1974) reported acceptable performance. The present purpose is to
review those evaluations and to describe those factors which must be considered
when applying the instrument.


The Goetz centrifuge is shown in the cutaway sketch of Figure 1. Two indepen-
dent channels with a constant cross section of a parallelogram spiral down the
outside of the conical rotor. During operation the rotor is spun rapidly, so that the
airscrew action of the helical channels draws the aerosol sample into the instru-
ment. The aerosol enters the centrifuge through the inlet tube and is deflected into
the channels by a baffle at the bottom of the stationary tube. The particles upon
entering the channels experience the centrifugal field and move radially across
the channel flow until they hit the outside wall. This wall consists of a metal foil

Goetz Aerosol Spectrometer

or other thin flexible material held tightly against the channel walls by a conical
cup. After a sample is deposited, the foil is removed for analysis. A flattened
collecting foil with a deposit of polydisperse latex particles is shown in Figure 2.
The flow rate F ([pm) through the channels depends on the rpm R of the
centrifuge rotor and on the size 0 (cm) of the interchangeable orifice at the end of
each channel. Goetz et al. (1960) gave

F = 2.75 x 10-1 RO cm2, (1)

which is an approximation, since the flow rate depends to some extent on condi-
tions external to the instrument. Flow-rate measurement must be made with a
low-resistance meter, e.g., soap film meter (Barr, 1934), since the instrument is
quite sensitive to pressure changes at the inlet and the outlet.
The geometry of the deposit in Figure 2 shows similarities to that of a horizon-
tal elutriator, which is to be expected, since the operation of the instruments is
similar in principle, with the force of gravity in the elutriator being replaced by
the centrifugal force in the centrifuge and with the aerosol being introduced over
the entire cross section of the channel in both devices. In the elutriator
monodisperse particles form a deposit of uniform concentration with a length eD
along the channel proportional to the inverse square of the aerodynamic diameter


Figure 1. Cutaway sketch of the Goetz Aerosol Spectrometer. In assembled form the vertical axes (1)
and the horizontal dashed lines (2) each coincide.


D of the particles. Under ideal operating conditions the concentration of those
particles would also be constant on the centrifuge deposit, although 'D is related
to D in a more complicated manner due to the involved geometry of the cen-
trifuge spirals. St6ber and Zessack (1964) in a theoretical study of the perfor-
mance of the instrument showed that along the midline of the channel deposit

D 2.29 "- F/A D2 R2 + (fa + X)3/212/3 X, (2)

where A = particle slip factor, T = viscosity of air, X = 16.5 cm, and
fa 6 cm is defined as the distance along the deposit side of the channel
from the beginning of the channel to the point opposite the channel entrance.
The simplest technique of measuring [D along the flattened channel deposits
which describe Archimedean spirals is with accessory equipment (Goetz and
Kallai, 1962). It can also be obtained by measuring the length r of the line
perpendicular to the top edge of the foil and extending to the deposit end, which
is formed by the arc of radius r + fo (Figure 2). Goetz and Kallai (1962) gave the
relationship between eD and r as

D r(2 ) 1/ a In [r+ (22 + r2)1/2 r + Fo, (3)
2 a ro

where fo = 7.0 cm, ro = 7.4 cm, and a = 1.77 cm/rad is the pitch of the deposit
spirals. An approximate expression,

eD 3 3.0 r'-355, (4)

accurate for tD > 4.0 cm, can also be used.
Under the assumption that the deposit concentration of monodisperse particles
on the centrifuge deposit is uniform and ends abruptly at each tD, a deposit of
polydisperse particles is evaluated in the same way as an elutriator deposit. The
discrete size distribution of the particles is given by

AD A= f AD F t

The number C of particles per unit area is counted at different eD values near the
midline of the deposit, and the resulting curve of C vs. eD is differentiated to
yield AC/AfD. The ratio AtD/AD is obtained from calibration data which relates
D of monodisperse particles to their fD value (Figure 3). The last ratio is a
weighting factor needed to give the size distribution an absolute value; it includes


Inertial Classification

Goetz Aerosol Spectrometer

the area AD on the deposit covered by particles of size D, the flow rate through
the channels, and the time t during which the aerosol sample is collected.
The cumulative size distribution of polydisperse particles is simply

N( M tIc AD AC, (6)
M MFt c=o

where N( than D and M is their total number.


The purpose of almost all experimental evaluations of the Goetz aerosol cen-
trifuge was to establish the geometry of the centrifuge deposit with monodisperse
latex particles and thus verify the accuracy of Eq. (5) for sizing polydisperse
aerosol. All the evaluations showed a more or less sharp dropoff in the latex
particle concentration curves, which is the first requirement for an elutriatorlike
deposit. The values of fD and D corresponding to the dropoffs are shown in the
calibration curves provided by the builders of the centrifuge (Figure 3). These
curves, apparently for an early centrifuge model, differ somewhat from those

0 5

Figure 2. Flattened collecting foil with a deposit of polydisperse latex particles. The deposit length of
the monodisperse fraction is tD.


10 -


0 R 0
SZ: x xl0x 3 amil

< I 6 60
2 12 60
3 18 60
4 12 40
5 24 40
6 18 30
7 24 30
8 24 20
10 1 I 1 1 1 1 111 I I 1
5 10 20 30 40
Figure 3. Deposit length for particles of a given aerodynamic diameter as a function of centrifuge rpm
R and the size 0 of the flow-limiting orifice, according to Goetz, Stevenson and Preining (1960).

Goetz Aerosol Spectrometer

corresponding to the commercial instrument. The curves for the commercial
instrument have proved to be reproducible.
Evidence for the second requirement, that of a uniform deposit concentration
for monodisperse particles, was given by Preining, Stevenson, and Goetz (1959).
Their two concentration curves measured along the midline of the deposit for
two values of 0, and one value each of R and D (Figure 4) did indeed show
the required uniform deposit concentration as well as the sharp dropoff. How-
ever, the originators of the centrifuge neglected to give concentration curves
for other possible combinations of D, R, and 0 (D/R/O), thus leaving the im-
pression that the instrument operated ideally for at least all D/R/O described by
the curves in Figure 3. Subsequent evaluations (Stober and Zessack, 1964;
Baust, 1967; Raabe, 1967; Stober and Boose, 1973; Gerber, 1971) of the






0 0IO 20 30
Figure 4. Concentration of latex particles 0.56 /m in diameter along the midline of the foil deposit for
R = 12,000, 0 = 40 (A), and 0 = 60 (o), from Preining, Stevenson and Goetz (1959).

centrifuge tested other D/R/O and found concentration curves which usually
were non-uniform and sometimes showed broad cutoffs. The reputation of the
centrifuge for erratic behavior came about because investigators usually tested
their own choice of D/R/O and found shapes of concentration curves that had
not been seen previously.
The last (Gerber, 1971) of the series of evaluations found a consistent pattern
in all of the previous calibration work. All latex-particle concentration curves,
measured near the midline of the deposit, had three distinct regions as shown in
the example in Figure 5. The curves increased over the distance fa, had a nearly
constant slope S (with both positive and negative values) from ea to fb, and
dropped off with an average slope of S2 past 4b. It was discovered that for any
given operating condition of the centrifuge, i.e., any given values of R and 0,
the shape of the concentration curves for various values of D was nearly invariant


Inertial Classification

when the length tb ta was normalized (Figure 6). The previous calibration data
as a whole reflected that behavior (Figure 7). Some concentration curves showed
the required S = 0, although most showed S < 0, which behavior was also
found by Baust (1967).
Figure 8 shows a wide range in the slopes S2 of the dropoff region; however an
overall pattern in the data is not as evident as for S1, perhaps due to the greater
sensitivity of the cutoff to the technique of latex particle generation. As a general
rule large values of 1/Si are found when R> 20000 rpm and when flow rates
greater than 5 fpm are used. Although the operating conditions exceeding those
limits still give a Reynold's number reflecting laminar flow in the channels, it
appears that those limits mark a transition to turbulence in the channel entrances.
A series of calibration runs was repeated with the baffle on the inlet tube
removed (Gerber, 1971). Those concentration curves consistently showed values
of S1 > 0 (Figure 7). Velocity-profile measurements near a channel entrance in
the centrifuge (point P, Figure 1) run with and without the baffle at the same
values of R and 0 showed major differences, suggesting that the inlet geometry






0 5 10 15
Figure 5. Typical latex particle concentration curve (D = 0.357 /m, R = 18,000, 0 = 30 mil).


1 .0

0.8 -


0.4 A 0.2341m 28.
Z O. 357 29. SEE
2 0. 2 A 0.500 30. TABLE I
0 0.714 31.
W 1.0; .........

o0 0.8-

j 0.6

-" 0.4 0.264 Jm 56.
C. ............0365 57.
0.L 0.2- 0.557 58.
0.796 59.

> 0

W 1.0 b



0.4 a- =0.1 pm 8.
b- = 0.188 9.
O. 2 c = 0.365 10.
d =0.630 II.

0 .2 .4 .6 .8 1.0 1.2
Figure 6. Experimental latex particle concentration curves (Gerber, 1971, upper curve; Horvath,
1973, middle curve) and theoretical predictions by St6ber and Zessack (1964), lower curve. All
curves were normalized over the distance b ea and they reflect the same operating conditions of
the centrifuge. The instrument used by Horvath (1973) differed slightly from the Goetz centrifuge in
baffle geometry and channel dimensions.


Inertial Classification

influences the slope of the concentration curves by causing strongly developing
velocity profiles in the first few centimeters of the channels.
Stbber and Zessack (1964) calculated the concentration of monodisperse parti-
cles along the channel deposit for a given set of operating conditions under the
assumptions that the particles were uniformly distributed across the channel at
point (a, that laminar, fully developed flow existed beyond ea, and that no
velocity gradients existed between the inner and outer channel walls. Their





0 1I

-0 I





20 30 40 50 60 70 80
Figure 7. Normalized slope S of latex-particle concentration curves as a function of the operating
conditions of the centrifuge. The stock instrument is indicated by dashed lines, and the instrument
with baffle removed is indicated by solid lines. See Table I for the key.

theoretical results do not compare favorably with the pattern demonstrated by the
experimental concentration curves of others (Figures 6 and 7), or with their own
experimental curves included in Figure 7. Stober and Boose (1973) qualitatively
attributed the disagreement to the effect of the developing velocity profile as well
as secondary circulation in the channel induced by the Coriolis force.
Figures 7 and 8 show that operating conditions exist which give the required
elutriatorlike deposit at least along the center line of the deposit. Horvath (1973)
pointed out, however, that the instrument operating with nonuniform concentra-
tion curves could still give the correct size distribution of aerosols if recurrence
correction formulas are applied to Eq. (5).


R = 12000 rpm

0 40
0 41 6000


18000 0 2 42 12000
D4649 %q51 22 33 4-,440O 5 18000 3 36
0 47 X 9 5 24 A6 6 240000 35
0 48 I o X. 14

S126 -

(D17 [3B30-"- 32o0
M 28
7 13 XA 10 o.

XI 5
A 19

5 -1-1-11 I -I

Goetz Aerosol Spectrometer

TABLE I. Key for Figures 7 and 8. Each D(/zm)/R(rpm x 10-3)/O(mil)
Represents a Calibration of the Goetz Centrifuge with Latex
Particles. For Some Tests D Is Expressed as a
Function of t-,)


Preining et al. (1959)

* Preining (1962)
3. 0.557/10/40

Stober and Zessack (1964)
A Experimental
4. 0.365/6/60
5. 0.365/18/30
6. 0.365/18/60
7. 0.365/24/20



C Raabe (1967)
16. 1D(60cm)/24/60
17. 1 D(32cm)/24/20

A Baust (1967)
18. 0.163/22/20
19. 0.365/22/20
20. 0.796/22/20
21. 1.305/22/20
22. 0.163/22/40
23. 0.211/22/40
24. 0.365/22/40
25. 0.796/22/40

Gerber (1971)
E[ centrifuge A
26. 0.357/12/30
27. 1.090/12/60
28. 0.234/18/30
29. 0.357/18/30
30. 0.500/18/30
31. 0.714/18/30
32. 0.357/22/40

o centrifuge A,
baffle removed
33. 0.357/18/30

O centrifuge B,
baffle removed
34. 0.357/6/80
35. 0.714/6/80
36. 1.090/6/80


Deposit Area

The accuracy of Eq. (5) also depends on the values of AD, which are only
approximately equivalent to the area eD x bo available for particle deposition on
the collecting surface (bo is the distance between the upper and lower channel
walls). Goetz and coworkers (Preining, Stevenson and Goetz, 1959; Goetz and
Kallai, 1962; Preining, 1962) realized early that deposits of monodisperse parti-
cles were not of width bo but instead gradually narrowed to a fraction of that
value for long deposits. Further they found that the deposit concentrations away
from the midline of the deposit were nonuniform to some degree (e.g., Figure 9),
even though the midline concentration was constant. They measured the areal
extent of the latex particle deposits with a planimeter and divided that area by the
maximum possible coverage, tD x bo. The resulting ratio Ki is applied to Eq. (5)
to correct the assumption that AD = eD x bo. Similar measurements (Gerber,
1971) which in addition took into account the deposit inhomogeneities away




Horvath (1973)
56. 0.264/18/30
57. 0.365/18/30
58. 0.557/18/30
59. 0.796/18/30

46 Inertial Classification

from the deposit midline gave values of K1 which agreed well with the earlier
values (Figure 10).

Entrance Losses

Baust (1967) determined particle losses in the centrifuge inlet by measuring the
activity of radioactively tagged particles on the centrifuge deposit as well as on a
parallel reference filter. His results are given in Figure 11 along with the results
of recent work based on the counts of latex particles. No obvious explanation was

14 1

0 10 20 30 40 50

60 70 80 90

Figure 8. Slope S2 of the cutoff region for latex particle concentration curves. S2 = AC/A', where AC
= 1.0.

found for the significant differences seen between the data of Baust and the other
data. The recent work shows that the losses depend on the operating conditions of
the centrifuge. Also, the losses are small for particles with sizes that correspond
to ID > 6 cm, which is the location on the deposit about opposite the channel
entrance. The large loss of particles with smaller ID values suggests that those
particles are lost in the short connection between the inlet tube and the channels.

Pressure Sensitivity

The reduction of atmospheric pressure from the sea-level value shortens the
values of Do and requires that a correction be applied to the calibration curves



2 A6

0 -

A 23
A 24

6 17

A 21
A 19 A 25
A 20
4 0 55 1 27
A 18 056 054 O 39
AS 5
A 7
1 26 0 45
2 28 0 52 0 44
130 02 0 43
846 56 31 w 53 52 I 38
57 40 2949 32
5859w 33 37
0 I I s 1 ,jtI



Goetz Aerosol Spectrometer

(Figure 3) and to the value of F in Eq. (5) (Gerber, 1964). Two factors contribute
to shorten the deposits: The particle slip correction becomes more pronounced at
lower pressures, and the flow rate through the channels decreases. The correction
factor K2 applied to values of D in Figure 3 is given with an accuracy of 5% by
the empirical relationship

D2 exp -76)C2
K2 ,() (7)

where P is atmospheric pressure in cm Hg, D is given in micrometers, C1 = 3.0
and C2 = 0.95 for operating conditions with 0 0.75 mm, and C1 = 2.0 and C2
= 0.98 for 0 > 0.75 mm. After the calibration curves are corrected, new values
of AfD/AD must be calculated. The value of F which in addition affects the
absolute value of the size distribution through Eq. (5) was found to require a
factor of 0.8 at P = 30 cm Hg.

Resolution Limit
The ability of the centrifuge to resolve fine structure in a polydisperse size
distribution depends on the sharpness of the deposit dropoff for particles of each
size in that distribution. A measure of the sharpness is the distance on the deposit
over which the dropoffs of monodisperse latex particles are smeared. Such mea-
surements were made by fitting cumulative normal curves to the dropoff regions
of latex particle concentration curves (Gerber, 1971). Subtracting the known
value of the variance in the size of the latex particles from the variance of the
normal curves gave the dispersion in the dropoff inherent to the instrument. The
resolution limit of the centrifuge is defined as the relative standard deviation
(o/D) of that dispersion.
The resolution limits shown in Figure 12 for several operating conditions of
the centrifuge show that exceptionally precise particle sizing is possible. Similar
resolutions should exist for other operating conditions as long as R : 20000 rpm
and F : 5 epm, since under those limits the average slope S, in the dropoff
region was found to be steep (Figure 8). The precise resolution suggests that the
flow throughout the channels is laminar for those operating conditions, since
otherwise the sharp border between aerosol and particle free air in the channels
which creates the fine resolution would not exist.

Statistical Sampling Error
The particle counts per unit area along the deposit which form the basis of the
derived particle-size distribution introduce a statistical sampling error governed
by the Poisson distribution. Some have suggested that the magnitude of that error
was sufficient to make use of the centrifuge impractical (Hochrainer, 1972;
Horvath, 1973). The relative statistical sampling error was calculated for a range
of particle-size distributions to place that suggestion in proper perspective. The


Inertial Classification

size distributions were described by the Junge distribution

dN/d log D = C/Dy, (8)

for which -2 < -y < 3, thus encompassing most possibilities. In the evaluation of
those hypothetical distributions on the centrifuge deposit, eleven equal areas




0 5 10 15 20 25 30 35 40

Figure 9. Deposit concentration of monodisperse latex particles for the centrifuge with the baffle
removed (R = 18,000, 0 = 30 mil). Separation between solid concentration isolines is 0.2, and that
between solid and dashed lines is 0.1. The outermost line is 0.0, and areas labeled (+) have values
larger than 1.0.



a, 0.9-

z 0.7

- 0.6 -

' 0.5
o 0

S0.4 -


0. -

5010 20 30 40

Figure 10. Deposit area correction factor K, vs. 'D to correct for the assumption
that the deposit width is a constant 1.15 cm.

0.1 1.0
Figure 11. Loss of particles in the inlet of the centrifuge.

Inertial Classification

were counted to give a ten-point size distribution with the points separated by
equal geometric intervals over one decade of particle size. The total count was
25000 particles. The average relative statistical sampling error was calculated for
discrete and cumulative size distributions as well as for the total number of
particles (Figure 13). As expected, the error is small for all cumulative distri-
butions, although it is also reasonable for discrete distributions with /f < 2.
However, for size distributions with a predominance of small particles (e.g., in
the atmosphere 3 3), a substantial error is found. By reducing the number of




z 0.01

0 .0011 1 x I I I III I I I III
0.01 0.1 1.0
Figure 12. Resolution limit oT/D of the Goetz centrifuge. The contribution from the Brownian
diffusion of the particles is /3.

data points in the distribution or increasing the particle counts per point, the
errors shown in Figure 13 can be significantly reduced.


In retrospect the lack of sufficient calibrations of the Goetz centrifuge by its
builders was the principal reason the instrument developed a poor reputation. The
few data they showed indicated that the centrifuge operated ideally in that it gave
elutriatorlike deposits for monodisperse particles. Most of the subsequent calibra-


Goetz Aerosol Spectrometer

tions showed deposits of other shapes. When all calibrations were compared,
however, they showed a consistent pattern, with the same deviations from the
ideal elutriator behavior. By taking those deviations into account, as has been
described here, the instrument can be successfully applied to measure the size
distribution of aerosol particles.
The weakness of the instrument lies in the large statistical sampling error
which is unavoidable when trying to measure the discrete size distribution of
aerosols with a much greater number of small than large particles.
The strengths of the instrument are the relatively high sampling rate, accurate




W 0.1

>r ~ s

-2 -I 0 I 2 3

Figure 13. Average statistical sampling error for ten-point size distribution described by the Junge
distribution, dN/d log D = C/D7, over the particle-diameter interval 0.1 to 1.0 /im and for a total
particle count of 25,000 particles. The dashed lines give the error when the counting areas on the
centrifuge deposit are optimized.


Inertial Classification

measurement of the cumulative size distribution, and ability to measure particles
with diameters as small as a few hundred angstroms (Gerber et al., 1970; Gerber,
1976). Operation of the Goetz centrifuge at much higher rpm values than those of
other aerosol centrifuges makes the measurement of such particles possible and a
unique capability.


Barr, G., 1934: Two designs of flow meter, and a method of calibration. J. Sci. Instrum., 11,
Baust, E., 1967: Use of the Goetz Aerosol Spectrometer to measure the size spectra of polydisperse
aerosols. Staub-Reinhalt. Luft, 27, 16-23.
Gerber, H. E., 1964: Pressure calibration of the Goetz Aerosol Spectrometer. Ft. Monmouth, N.J.,
U.S. Army Electronics Command Technical Report No. 2456.
____, 1971: On the performance of the Goetz Aerosol Spectrometer. Atmos. Environ., 5, 1009-
,__ 1974: Discussions: Developing flow and particle deposition in horizontal elutriators and
semi-dispersive aerosol centrifuges. Atmos. Environ., 8, 1344-1346.
_ 1976: Relationship of size and activity for AgI smoke particles. J. Atmos. Sci., 33, 667-677.
__ U. Katz, C. I. Davis, and L. 0. Grant, 1970: Some size distribution measurements of AgI
nuclei with an aerosol spectrometer. J. Atmos. Sci., 27, 1060-1067.
Goetz, A., H. J. R. Stevenson and 0. Preining, 1960: The design and performance of the aerosol
spectrometer. J. Air Pollution Control Assoc., 10, 378-383.
- and T. Kallai, 1962: Instrumentation for determining size- and mass-distribution of submi-
cron aerosols. J. Air Pollution Control Assoc., 12, 470-486.
Hochrainer, D., 1972: Discussions: On the reliability of measurements with the Goetz Aerosol
Centrifuge. Atmos. Environ., 6, 699.
Horvath, H,, 1973: Discussions: On the performance of the Goetz Aerosol Spectrometer. Atmos.
Environ., 7, 1003-1011.
Preining, 0., 1962: Das Goetzsche Aerosolspektrometer, problem bei betrieb und auswertung.
Staub-Reinhalt. Luft, 22, 129-133.
,__ H. J. R. Stevenson and A. Goetz, 1959: The analysis of aerosol spectra. 52nd Annual Air
Pollution Association Meeting, Los Angeles. Preprint 59-42, 29 pp.
Raabe, 0. G., 1967: Calibration and use of the Goetz Aerosol Spectrometer. Assessment of Airborne
Radioactivity. Vienna, International Atomic Energy Agency.
Sawyer, K. F. and W. H. Walton, 1950: The 'Conifuge'-a size-separating sampling device for
airborne particles. J. Sci. Instrum., 27, 16-23.
Stober, W. and U. Zessack, 1964: On the theory of a conical aerosol centrifuge. Staub-Reinhalt.
Luft, 24, 295-305.
- and C. Boose, 1973: Developing flow and particle deposition in horizontal elutriators and
semi-dispersive aerosol centrifuges. Atmos. Environ., 7, 119-130.


Goetz Aerosol Spectrometer


Q: I'd like to know the physical size of the last instrument described by Dr.
Stober: While the prototype design followed the conventional rotor design of
several years ago, the new rotor model which we are working on now has a
diameter of about 33 centimeters and the geometry of the duct will be somewhat
different. There are only 1V turns on the spiral duct, and the width of the duct
has been reduced to something like 7 mm, instead of the usual depth of 35 mm.
Q: What about the humidity response and the difficulties with using this type
of instrument?
Stober: One has to maintain constant relative humidity at least to a reasonable
degree. The shift of the frequency in this case, where the basic frequency is at
10 megacycles, is about 1 cycle per percent relative humidity and we can adjust
to that. The sampling rate is up to 10% of the total flow. If we maintain a clean
air flow at 0% humidity and we introduce aerosol at 100% humidity then we have
only 10% humidity in the air mixture. Within 0 and 10% we can adjust for
the differences caused by changes in humidity. We can also provide the possibility
of just using humid air at a certain level and maintaining that humidity within
the duct. Some particles, particularly hydrophilic ones, grow and shrink with
humidity; this is something one has to take into consideration.
Q: What can you do to prevent water condensation when you are sampling a
hot humid air stream?
Stober: We have used this instrument for sampling diesel exhaust which has a
lot of humidity. We use a silica gel column with a hollow center core so the
diesel exhaust aerosol goes through that column and the water diffuses to the
surrounding silica gel wall. That works very well and we have no dewpoint
problems whatsoever. Of course, again, particles might shrink and swell with the
different humidity levels but I think that is something one has to keep in mind and
can live with.
Q: I have a question for Otto Raabe. He spoke about the importance of
discharging particles. If we discharge an aerosol, we change its deposition prop-
erties. Shouldn't we really be sampling the aerosol in its natural charged state?
Raabe: There are two times when you have to worry about charge. First, if you
are calibrating with monodispersed polystyrene latex particles you have to dis-
charge them, or else you are going to get some anomalous deposit positions
because of the charge and you won't be able to accurately define the physical size
and density of those particles. The second case is in field sampling, for example,
plutonium aerosols in glove boxes. We have found a high electrostatic charge
which did lead to anomalous results with respect to the deposition both in cascade
impactors and in the spiral disk centrifuge if we didn't first pass the aerosol
through a Krypton 85 discharger. Now I say we must discharge because there is
no way to characterize this complicated phenomenon that relates both to electro-
static charge and aerodynamic diameter. They should be characterized separately.


Inertial Classification

Study the charge in one set of experiments. Study the actual aerodynamic prop-
erties of a particle without charge in another set.
Q: What do you mean by discharging?
Raabe: Discharging here in the broadest meaning means that we have reduced
the charge to the Boltzmann equilibrium. This is not actually completely dis-
charging, but reducing the charge to such a relatively small amount that the
behaviour of the particles is not markedly affected by their charge. We are
talking about reducing charges from hundreds or thousands of charges per parti-
cle to Boltzmann equilibrium which involves some uncharged, some with posi-
tive, some with negative; but small charges do exist.
Stober: Maybe I should add some remarks to that. I think from our experience
with the spiral centrifuge the only difference we obtained in the deposit of
charged or uncharged particles is that the arrangement of the particles deposited
close to the foil is different when the particles are charged than when they are
uncharged. The reason is that a particle which is settling in a location and cannot
get rid of its charge throws up a field and influences an approaching particle
which is settling in the same location. The electron micrographs are different
depending on whether particles were charged or not, or whether they settled on
an insulating foil which we use very often, such as Formvar film, or on a metal
foil. The location, as far as my experience goes, is not changed, only the picture
of the deposit.
Moss: I'd like to make a comment about the question of charge. Again, this
comes back to the basic thing we'll be talking about in this Workshop and that is
what are we trying to measure? With the centrifuge devices we're measuring a
very specific thing, terminal settling velocity. And why are we measuring that?
Because it is one of the mechanisms by which the lung takes out particles from
the air. So the question on charge, in terms of these instruments, reduces to how
long does a charge remain on a particle when it's in 100% humid atmosphere
such as you have when you are inhaling it into your lungs? In that sense you want
to discharge your particle when you are looking at this particular property.
Tillery: I think the important thing here is that latex aerosols that are generated
from generators have very high charges, extremely high charges, and such
charge levels will not normally be encountered in sampling conditions. The
enhanced deposition is a result of image forces which require fairly high electric
charges. So the type of aerosol you're normally going to encounter isn't going to
be a problem, and it's not going to be a problem, I think, in relating the deposi-
tion in these instruments to lung deposition.
Q: If you are sampling a natural unknown aerosol, what is the accuracy with
which you can determine size distribution and how does this relate to optical size
Raabe: That's an exceptionally difficult question. Size distribution measured
with a centrifuge depends on the density and shape of the particle. Optical
counter sizes depend on the refractive index of the particles, while centrifuges do
not. That is the crux of the problem, and I don't believe there is an answer yet.
St6ber: I think the problem really is whether you can relate mass distribution


Goetz Aerosol Spectrometer 55
measurement in terms of aerodynamic diameter to number distribution measure-
ment. You can do both analyses with the instrument. You can put Formvar film
along the sampling foil, take it off, and weigh your deposits. Immediately
you get a mass distribution in terms of aerodynamic diameter. This is not neces-
sarily related to a number distribution because if you sample an unknown aerosol
having particles with all kinds of shape factors, then you are at a loss. There is no
direct relationship to the number distribution and vice versa. If you look at
electron micrographs and count your particles in certain locations then you come
up with a number distribution but you cannot transfer that into a mass distribution
without knowing shape factors. Of course you can determine shape factors from
electron micrographs, but it's a messy procedure. This is a basic problem which
we cannot get away from. It's always worthwhile to state what you are referring
to; a primary mass distribution measurement, or a primary number distribution
measurement. The other kind of distribution cannot directly be derived from the
first measurement.


Faculty of Engineering Science, The University of Western Ontario, London, Ontario, Canada

The reverse flow cyclone has long been used as a cheap and reliable device for
removing relatively coarse particles from air streams. Although all cyclones will
collect particles in all sizes with at least some measure of efficiency, cyclones are
not generally considered to be efficient devices for removing particles smaller
than 2.0 to 3.0 u/m. Recent experimental studies on cyclone efficiency have,
however, shown that quite high collection efficiencies can be obtained with
submicron particles in some circumstances (Chan and Lippmann, 1975;
Blachman and Lippmann, 1974; Knuth, 1969; Lippmann and Chan, 1974;
Lippmann and Kydonieus, 1970). A wide range of cut sizes (also called effective
cut-off aerodynamic diameters (ECAD), defined as the aerodynamic diameter of
particles having a collection efficiency of 0.5 at the stated operating conditions)
may be obtained by suitable selection of cyclone geometry, cyclone diameter and
inlet velocity, which raises the possibility of using calibrated cyclones for deter-
mining size distributions in heterodisperse aerosols. A multistage sampler with
six identical cyclones in parallel, all operating at a different flow rate, was
developed by Lippmann and Kydonieus (1970) for this purpose. A multistage
sampler based on cyclones is inherently able to collect much larger samples than
a cascade impactor, and may be capable of yielding much more accurate informa-
tion than the impactor, particularly in determining particle size distributions by
mass weight. The design of efficient cyclone samplers, and the extrapolation of
data to conditions outside of the calibration range requires some understanding of
cyclone theory and cyclone operation, and it is our purpose to review existing
data and theories to assist in this objective.

The Cyclone as a Size Selective Sampler


Dimensional similarity between two geometrically similar cyclones, or between
two operating conditions in a given cyclone, requires, as a minimum, similarity
in fluid flow and in particle trajectories. These conditions will be satisfied by the
equality of the cyclone Reynolds numbers (Re) and of the Stokes numbers (Stk).
There may, however, be other mechanisms or forces at play beyond those scaled
by equating the Reynolds numbers and Stokes numbers. These include electrostat-
ic forces, turbulent velocity, fluctuation, induced inertial impingement of particles
on the cyclone walls, reentrainment of deposited particles, and elastic recoil of
particles on deposition. Moreover, in some cases the particle Reynolds number
based on centrifugal drift velocity may be outside the Stokes law range, in which
case the inertial forces will not be correctly scaled by the Stokes number. All of
these complications should be borne in mind in evaluating cyclone data, and in
extrapolating such data outside the calibration interval.


Early cyclone theories, such as those of Rosin, Rammler and Intelman (1932),
and Barth (1956), considered only inertial forces and resulted in models which
predict capture efficiencies of unity for particles slightly larger than the cut size.
This conclusion, which is contrary to observation, resulted from neglect of the
role of turbulent diffusion. Nevertheless, although the older theories failed to
predict a reasonable shape for the efficiency curve, they did predict the cut size of
most industrial cyclones reasonably well. The cut size may be expressed in terms
of Stk and of the cyclone geometrical parameters in these theories.
Subsequent theoretical developments have included effects attributable to tur-
bulent diffusion. Leith and Licht (1972) assumed complete mixing of particles in
a horizontal plane, and a deposition velocity at the wall which was a function of
the axial distance coordinate. This velocity was taken as equal to the centrifugal
drift velocity which a hypothetical particle initially released near the center of the
cyclone would have attained by the time it had descended to the level of the
horizontal plane in question. The model is physically and conceptually unsound;
nevertheless its predictions are in good agreement with limited data on industrial
cyclones obtained using heterogeneous test dusts. Such data are inherently sus-
pect, however, because it is not clear to what extent the dust had been deagglom-
erated prior to reaching the cyclone. The agreement between the theory and the
data of Lippmann and his collaborators, who used dilute monodisperse aerosols,
is generally poor.
Turbulent diffusion was quantitatively taken into account in the theories of
Beeckmans (1972, 1973). The results were obtained as solutions of differential
equations, whose parameters could be expressed in terms of the Stokes number
and of the cyclone geometrical parameters. It was assumed that the turbulent
diffusivity is uniform throughout the cyclone, and proportional to the Reynolds


Inertial Classification

number. The expression for turbulent diffusivity was obtained by a momentum
balance on the spinning gas, and it is based on the assumption that the tangential
gas velocity is inversely proportional to r'5. Data on gas velocity distribution in a
cyclone are scarce, and it is generally considered that the exponent varies be-
tween 0.5 and 0.7. No data are available on the effect of Re on the exponent, but
if the exponent were to depend on Re the parameters of the differential equation
would depend on Re as well as on Stk. Beeckmans' theories are not particularly
successful in correlating experimental cyclone efficiency data.


Data obtained with redispersed powder cannot be considered reliable, and will
not be considered further. The data of Lippmann and collaborators were obtained
with dilute monodisperse aerosols, and will be considered here, together with
limited, unpublished data from our laboratories obtained with monodisperse,
uranine-methylene blue aerosol.
The least restrictive assumption for correlating the data for a given cyclone is
that the efficiency is a function of the particle size over the domain of Re and Stk,
as well as of the electrical parameters, the elastic constants of the particles, and
of the cyclone walls, etc. Such a broad generalization clearly has little utilitarian
value in our present state of knowledge of cyclone operation.
The somewhat more restrictive assumption that efficiency may be correlated
over the domain of Re and Stk cannot be verified for Lippmann's data, because
the only source of variation for Re is the inlet velocity Vi. Thus, although the data
cover a certain area of the domain of Re and Stk, there are no data in which both
Re and Stk were respectively equal in two independent experiments. A reliable
test of the assumption in question would require experiments in which cyclone
diameter D, gas viscosity /, gas density p, or particle density pp were varied, as
well as particle size and inlet velocity.
While Lippmann's data do not furnish a proof of the validity of the assumption
= = -(Stk, Re), it would seem logical to seek an empirical correlation in terms
of these variables. A linear relationship is clearly unsuitable, but a variety of
other functions have been found which correlate the data more or less well.
Functions which were tested include the following:

In Nr = ao + a1 In (Re) + a2 In (Stk) (1)

In NT = ao + a1 In(Re Stk) + a2 In2 (Re Stk) (2)

,q = F (X) (3a)


X = ln[(Stk'" Reb)/(Stko-' Reb)50o]/ln o



The Cyclone as a Size Selective Sampler

TABLE I. Ranges of the Experimental Variables

D VI dae
Cyclone (mm) (m/sec) ('/m) Re x 10-3 Stk x 103 n

Unico 25.4 1.82-2.39 2.3-5.1 3.6-4.7 1.4-7.9 30
Aerotec 3/4 19 1.94-2.47 2.2-6.6 2.9-3.7 1.8-21 30
Aerotec 2 55.2 6.0-8.6 2.7-6.6 26-37 3.0-25 20
Q < 5 Cpm 10 3.7-20.8 1.07-9.27 2.9-16 6.3-230 47
Q > 5 tpm 10 24.3-133 0.24-2.0 19-100 1.9-79 56
Beeckmans 76-152 6.1-21.3 0.71-5.6 54-150 0.054-16 21

NOTE: For all cyclones, it was assumed that the kinematic viscosity had a value corresponding to air
at 200C, 1 atm (v = 1.51 x 10-5 m2 sec-1).

NT is the cyclone efficiency expressed in transfer units, F is the normalized
cumulative log probability function, (Stko.5 Reb)50 is the value of the quantity
in brackets at -q = 0.5, o-, is the geometric standard deviation, and b is an
arbitrary constant. The cyclones used by Lippmann and collaborators are de-
scribed in the literature. Our own experiments were performed with two geo-
metrically similar cyclones with diameters equal to 3 inches and 6 inches. Details
of these experiments will be published elsewhere. Experimental variables and
their ranges for all cyclones considered are given in Table I. Results of the
regression analyses are shown in Tables II and III. Note that the data for the
Dorr-Oliver 10 mm cyclone were split into two groups, one for which the flow Q
was less than or equal to 5 tpm, and one for which Q exceeded this value.
As can be seen in Table II, this procedure resulted in a considerable reduction
in the standard error of estimate e. The last column in these tables may be

TABLE II. Linear Regression Analyses on the Function
fn NT = ao + a1 n Re + a2 en Stk

Cyclone ao a, a2 Ea e4

Unico 2.90 1.093 1.179 0.098 1.10
Aerotec 3/4 6.52 1.58 1.29 0.153 1.16
Aerotec 2 0.205 0.526 1.202 0.133 1.14
all data 6.81 1.09 1.10 0.705 2.02
Q < 5 fpm 3.03 0.96 1.90 0.387 1.47
Q > 5 (pm -11.65 1.36 0.67 0.246 1.28
Beeckmans 3.86 0.584 0.516 0.195 1.22

a. E equals standard estimate of error. (e = i,(ln NTi Y)2/(n 1), Y
being the value of In NT calculated from the regression equation using
appropriate values of Re and Stk corresponding to the i'th observation,
and NTI being the transfer unit efficiency corresponding to the i'th experi-


Inertial Classification

interpreted as follows. Let zi equal the ratio of the observed NT to that calculated
using the regression equation, and suppose that whenever zi < 1 it is replaced
by its reciprocal, so that for all zi, zi > 1. It may be shown that geometric
mean value of zi equals

= e EVV(n-1)/n

Thus for n > 1, the last column in the tables approximates the geometric mean
value of zi.
The data could not be fitted by linear regression to a function of the form

TABLE III. Linear Regression Analyses on the Function
fn NT = a) + a, en Re Stk + fn2 Re Stk

Cyclone ao a, a2 E e'

Unico 240 4.0 1.51 -0.063 0.096 1.10
Aerotec 3/4 5.92 2.53 -0.206 0.10 1.11
Aerotec2 12.6 3.36 -0.195 0.122 1.13
Q < 5 epm -12.9 2.82 -0.11 0.611 1.84
Q > 5 tpm 6.8 1.41 -0.052 0.44 1.55
Beeckmans 3.2 0.59 0.0073 0.196 1.21

TABLE IV. Non-linear Regression Analyses Based on the Function S2

Cyclone b (Stko.5 Reb)50 og E(%)

Unico 240 0.50 4.0 1.64 2.58
Aerotec 3/4 0.725 26.5 1.53 1.95
Aerotec 2 0.225 0.83 1.56 2.11
Q -< 5 tpm 0.225 1.51 1.30 5.84
Q > 5 (pm 1.05 6058 2.13 3.79
Beeckmans 0.5 13.7 3.15 5.83

TABLE V. Non-linear Regression Analyses Based on the Function S1

Cyclone b (Stko05 Reb)50o "g eE

Unico 240 0.475 3.4 1.68 0.103 1.11
Aerotec 3/4 0.50 4.24 1.55 0.107 1.11
Aerotec 2 0.475 11.0 1.57 0.117 1.12
Q < 5 tpm 0.25 1.88 1.46 0.442 1.56
Q > 5 tpm 1.1 10,400 2.36 0.191 1.21
Beeckmans 0.60 43.0 3.39 0.200 1.22


The Cyclone as a Size Selective Sampler

represented in Eqs. (3a) and (3b), but instead a non-linear minimization routine
was used to find the minimum value of sums of squares based on two functions of
the dependent variable -7, on the domain of the three disposable parameters
contained in the equations. The functions minimized were

S, = I ln (NTi/Yi) (4)


S, = 2( i)2 (5)

Yi equals the value of NT calculated using the fitting function with the values of
Stk and Re corresponding to the i'th data point, and ,7i is the corresponding
calculated value of the efficiency. Results of these analyses are given in Tables
IV and V. Figures 1 and 2 show plots of some of the data using probability and
logarithmic coordinates.


A comparison of Tables IV and V shows that changing the minimization criterion
has an appreciable effect on the parameters, even when the form of the function
remains unchanged. This is particularly significant with respect to the parameter
b, which determines the relative effects of the Reynolds number and of the
Stokes number. Note that b changed appreciably for both AEROTEC cyclones as
the minimization function changed. Changes in (Stk 5 Reb),o are less significant,
since they reflect primarily changes in b. The ratio b/0.5 is compared in Table VI
with a,/a2, the ratio of regression coefficients relating to the functions ln(Re) and
ln(Stk) in Table II. Agreement between columns 2 and 3 is within 20% for all
cyclones, and with the exception of the AEROTEC 2 cyclone there is equally
good agreement between columns 1 and 3. We conclude that data based on
monodisperse aerosol studies indicate that cyclone efficiency cannot be corre-
lated in terms of the Stokes number alone, and that the Reynolds number must
also be included in the correlation. Furthermore, for a given cyclone geometry
this type of correlation may only be valid for fixed values of cyclone diameter,
gas viscosity and gas density. Changes in the proportions of the cyclone will, of
course, have a profound effect on the parameters of the correlation.


The ability of the cyclone to sample large volumes without saturation effects has
been alluded to earlier, and this is an important advantage relative to impactors.
On the other hand, as can be seen in Tables IV and V, the cyclone efficiency


Inertial Classification

curve does not have a particularly sharp cut-off, as evidenced by relatively large
values of org. Mercer (1974) has compared values of o-g found by plotting impac-
tor efficiency curves on log probability coordinates against Stk o.5. In order to
compare these data with cyclone data, we must reduce both sets of data to a
common basis. This presents no problem in comparing impactor data with cy-
clone data obtained by varying particle size at constant flowrate, because in this
case Re is constant and consequently, in Eq. (3b) the normalized variable be-
comes ln(Stk/Stkso)0.5. The slope of an efficiency plot on log probability paper,








Dorr Oliver Cyclone
Q> 5 Ipm

0.5 I.1
Stk05 Re*
Figure 1. Efficiency of the 10 mm Dorr-Oliver cyclone for flows in excess of 5 [pm, vs. Stk 'Re' '.
Line shows correlation using cumulative log-normal distribution function (Table V).




The Cyclone as a Size Selective Sampler

which determines o-g, is the same whether the abscissa is Stk0'5 or (Stk/Stkso')05,
hence Mercer's calculated o-0's for impactors may be compared directly with
oj,'s for cyclones operated at constant flow.
If the velocity of the cyclone is changed at constant particle size, then both Re
and Stk will be affected. It is possible, however, to plot efficiency vs. Stk at a
single particle size, and to determine o-a for such a plot. If the efficiency is in fact
a function of both Re and Stk, a different curve will be obtained for each particle
size. It is easy to show that the following relationship exists between o'1), the
dispersion parameter for the argument Stk'.5, and o-g2), the dispersion param-
eter for the argument StkosReb:

0 +(1) -- (2))l/(1 + b)


Eq. (6) shows that o-" 0. Thus for the cyclones studied,
since b is always positive, the variance of a plot of efficiency vs. Stk'.5 for

Unico 240 Cyclone
o Re,,< 4176
* Re> 4176




Re o93 Stk1-179

Figure 2. Efficiency of the Unico 240 cyclone in transfer units, vs. Rel'0aStk1'179. Line shows linear
regression equation (Table II) on the arguments en Re and [n Stk.


Inertial Classification

experiments at constant flowrate and variable particle size is invariably smaller
than a similar plot obtained by varying flowrate at constant particle size. The fact
that Mercer found the same phenomenon with impactor data suggests that these
might be better correlated as a function of the form Stko5Reb rather than as a
function of Stk only.
The data correlated in Tables IV and V were obtained over a range of flows
and particle sizes. The slope of an efficiency curve plotted on logarithmic and
probability coordinates using as argument the function StkO5Reb must equal the
slope of a similar plot of efficiency versus Stko'5 at constant inlet velocity. This
follows from the fact that if Re is constant the abscissas of two curves are merely
displaced by a constant value on a logarithmic scale. Thus the o-,'s in Tables IV
and V may be directly compared with o-0's for impactor data obtained at constant
flow. Mercer calculated values for these for round jets in the range 1.10 to 1.54,
with a mean value of 1.27. For cyclones, the lowest value found for o-g was 1.3

TABLE VI. Comparison of (b/0.5) with Ratio of Coefficients a,/a2
in Regression Analysis on Variables [n(Re) and [n(Stk)

b/0.5 b/0.5 a /a2
Cyclone (Table V) (Table IV) (Table II)

Unico 240 0.95 1.0 0.93
Aerotec 3/4 1.0 1.45 1.22
Aerotec 2 0.95 0.45 0.438
Q -< 5 epm 0.5 0.45 0.505
Q > 5 tpm 2.2 2.1 2.03
Beeckmans 1.2 1.0 1.13

for the 10 mm Dorr-Oliver cyclone in the low flow range (Table IV). Several
cyclones had o-g's in the range 1.5-1.7, and some cyclones had o- 's considerably
less sharp than this. In general, therefore, the cyclones had cut-off charac-
teristics appreciably higher than most impactors. The same conclusions may
be reached by considering the slope of the plot of -q versus Stk "/(St ")5o
For both round and rectangular impactors at constant inlet velocity the value
of this slope is approximately 1.6 (Mercer, 1965). It may be shown that
for a device whose efficiency may be correlated by a function of the form
StkoSReb, the slope s of efficiency vs. Stko05/(Stko-5)50 at -T = 0.5 is given by the
following expressions:
a. Re constant, Stk varied by changing particle size

s = 1/VV2 In org. (7)

b. Particle size constant, inlet velocity varied

s = (2b + 1)/V2--ln org.


The Cyclone as a Size Selective Sampler

Slopes computed for the cyclones, using Eq. (7), ranged between 0.33 and 1.05.
These values are less than the corresponding impactor slopes, in agreement with
the previous conclusion based on o-, values for cyclones and impactors.
The more diffuse cut-off characteristics of cyclones relative to impactors
means that there will inevitably be a greater loss of definition in the reconstruc-
tion of the particle-size distribution of the aerosol under test from cyclone selec-
tive sampler data relative to cascaded impactor data. It is difficult to generalize
quantitatively about the relative errors in the computed particle size distributions
obtained by the two methods. It seems likely that the error obtained using a given
set of cyclones or impactors is dependent on the ratio of the dispersion param-
eters of the particle population and of the size-selective samplers, as well as on
how well the ECAD's of the set of samplers is distributed over the effective range
of particle sizes in the aerosol. In any case, the dispersion parameter (o-g) of the
aerosol population should be considerably larger than that of any single sampler
if a reasonable estimate is to be obtained for the population parameters of the
aerosol. Thus if the particle sizes of the aerosol cover a broad spectrum it is likely
that a set of size-selective cyclones would give almost as good an estimate of
particle population as an impactor cascade. On the other hand, should the aerosol
contain a relatively narrow range of particle sizes, the error in determining the
aerosol particle size distribution obtained using cyclones would probably be
appreciably larger than the corresponding error in using impactors.


Barth, K., 1956: Brennst.-Wirme-Kraft, 18, 1.
Beeckmans, J. M., 1972: A steady-state model of reverse-flow cyclone. J. Aerosol Sci., 3, 491-500.
1973: A two-dimensional turbulent diffusion model of the reverse flow. J. Aerosol Sci., 4,
Blachman, M. W. and M. Lippmann, 1974: Performance characteristics of the multicyclone aerosol
sampler. J. Amer. Industrial Hygiene Assoc., 35, 311-326.
Chan, T. L. and M. Lippmann, 1975: Presented at the 25th Canadian Chemical Engineering Confer-
ence, Montreal.
Knuth, R. H., 1969: Recalibration of size-selective samplers. J. Amer. Industrial Hygiene Assoc.,
30, 379-385.
Leith, D. and W. Licht, 1972: A.I.Ch.E. Symp. Ser. 68, No. 126, 196-206.
Lippmann, M. and T. L. Chan, 1974: Calibration of dual-inlet cyclones for respirablee" mass
sampling. J. Amer. Industrial Hygiene Assoc., 35, 189-200.
and A. Kydonieus, 1970: A multi-stage aerosol sampler for extended sampling intervals. J.
Amer. Industrial Hygiene Assoc., 31, 730-737.
Mercer, T. T., 1974: Presented at a meeting of Gesellschaft fuir Aerosol Forschung, Bad Soden.
Rosin, P., E. Rammler and W. Intelman, 1932: Z. Ver. Dt. Ing., 76, 433.



Institute of Environmental Medicine, New York University Medical Center, N.Y., N.Y.

The penetrating analysis presented by Dr. Beeckmans in his review paper in
these Proceedings has, at long last, provided a scientifically sound basis for
characterizing the performance of small cyclones in air sampling applications.
Chan and Lippmann (1976) have shown that previously proposed predictive rela-
tions for cyclone efficiency (Rosin et al., 1932; Lapple, 1951; Barth, 1956; Leith
and Licht, 1972; Beeckmans, 1973) are inconsistent with experimental observa-
tions in tests using low concentrations of monodisperse test aerosols. Using a
hyperbolic tangent relation for cyclone performance first developed by Blachman
and Lippmann (1974), they also described the empirical constants of this equation
for four commercially available, small cyclones which have been used in air
sampling, and for several Stairmand-type cyclones. It should be noted that the
previous predictive theories were developed from test data on, and have primar-
ily been applied to, cyclones used for air cleaning applications.
The discrepancies between the predictive theories and experimental perfor-
mance could be due to any or all of the following differences between the
conditions within cyclones used in air cleaning as opposed to cyclones used in air
sampling: 1) In air cleaning, the concentrations are much higher than in the
ambient aerosols which are drawn into sampling cyclones, and may lead to
particle agglomeration within the cyclone; 2) In air sampling, the particles usu-
ally remain on the cyclone wall where they were initially deposited, while in air
cleaning, deposited particles are continually scoured from the initial deposition
sites and migrate to the collecting hopper; 3) The flow patterns in air cleaning
cyclones are always turbulent, while in the smaller cyclones used in air sampling,
the flows may be completely or partially laminar. The presentation of Ayer and
Hochstrasser in these Proceedings clearly demonstrates transitions in pressure-

Cyclones for Size-Selective Sampling

flow relationships in cyclones which imply that laminar flow conditions can exist
within small cyclones.


By far the greatest usage of small cyclones in aerosol sampling has been as the
first stage of a two-stage sampler. In most cases, they have been operated at
flowrates which were believed to permit cyclone penetrations equivalent to the
penetration of inhaled dust to the non-ciliated deep lung spaces in humans. Such
applications have been called respirablee" dust sampling, and their rationale has
previously been discussed (Lippmann, 1970, 1976; AIHA Aerosol Technology
Committee, 1970).
In respirablee" dust sampling, it is essential that the cyclone cut-off charac-
teristics be known and constant, and that each cyclone is operated at a flowrate
which produces the desired cut-off. In the absence of a usable theory, it has been
necessary to design and calibrate air sampling cyclones empirically. Unfortu-
nately, dependence on an empirical calibration can lead to controversy and/or
errors. The original calibration of the Dorr-Oliver 10 mm nylon cyclone
(Lippmann and Harris, 1962) failed to include an appropriate aerodynamic shape
factor. With Kotrappa's (1971) shape factor correction, the recommended sam-
pling rate was reduced to 1.8 [pm, which is in good agreement with most of the
more recent empirical calibrations (Ettinger et al., 1970; Lippmann and
Kydonieus, 1970; Caplan et al., 1977; Seltzer et al., 1971) which recommend
either 1.7 or 1.8 [pm. However, MESA, which specifies acceptable samplers
and their operating characteristics in mine dust evaluations, still insists that the
10 mm cyclone be operated at 2.0 (pm on the basis of its own calibration (Tomb
and Raymond, 1969).
There are conflicting calibration data for several of the 12.5 to 50 mm diameter
stainless steel cyclones which have also been used for size-selective sampling.
Some of the differences can be attributed to the same causes as for the 10 mm
cyclone, i.e., differences in calibration techniques, data interpretation, and fail-
ure to use appropriate correction factors. Another factor is the dimensional var-
iations between nominally identical samplers which were documented by
Lippmann and Chan (1974). The steel cyclones are individually assembled, and
the quality control exercised by the manufacturers has not always been adequate.
The 10 mm nylon cyclones, which are injection molded, have not been observed
or reported to vary significantly from one to another.
Finally, empirical calibrations can only be valid if the field applications are
made without significant changes in the sampling train configurations. The paper
by Ayer and Hochstrasser in these symposium proceedings demonstrates that the
cut-off characteristics of a given cyclone can be significantly changed by altering
the coupling between the cyclone and back-up filter and thereby altering the flow
pattern within the outlet pipe of the cyclone. These observations appear to ex-
plain many of the current discrepancies between different laboratories' calibra-


Inertial Classification


Size-mass distribution analyses of aerosol constituents have been made using
cascade impactor samplers for more than thirty years, beginning with the original
May (1945) impactor and the Laskin (1950) Univ. of Rochester modification.
Analyses have also been performed with aerosol centrifuges which deposit parti-
cles on a collection foil according to their aerodynamic size. However, both
types of sample collectors have significant limitations when used for such appli-
cations. Solid particles tend to bounce off or be reentrained from impaction
surfaces which lack an adhesive layer or whose adhesive becomes saturated.
Unfortunately, saturation usually is reached either before the investigator is
aware of it, or before the accumulated sample is large enough for the subsequent
analysis. Applications of centrifuges are limited by their low sampling rates, and
large physical size and cost.
The limitations of the other inertial classifiers for compositional analysis by
particle size fraction have led to the development of samplers using cyclones in
either parallel or series arrays. The first development of a portable sampler using
a parallel array of cyclone-filter series samplers was described by Lippmann and
Kydonieus (1970). Each cyclone had a different cut-size and the overall size-
mass distribution of the total aerosol or any of its chemical constituents could be
determined from analyses of the collection on the filter following each cyclone
and a parallel filter sampler operated without a cyclone pre-collector. The per-
formance characteristics of an improved version of this sampler were described
by Blachman and Lippmann (1974). A much larger version of this sampler for
fixed station ambient air pollution sampling was described by Bernstein et al.
(1976). A parallel multicyclone sampling train for stack sampling applications
was described by Chang (1974).
The major advantage in using cyclones instead of impactors for such applica-
tions is that their performance is not significantly affected by the amount of
sample collected (Blachman and Lippmann, 1974). They share with impactors
the advantage of relatively low fabrication cost.
Acknowledgements.-These investigations were supported by research con-
tracts HSM-099-71-48 and CPE-70-NEG-128 from the National Institute for
Occupational Safety and Health, by RP 117-1 from the Electric Power Research
Institute, and is part of a center program supported by Grant ES 00260 from the
National Institute of Environmental Health Sciences.


AIHA Aerosol Technology Committee, 1970: Guide for respirable mass sampling. J. Amer. Indus-
trial Hygiene Assoc., 31, 133-137.
Barth, W., 1956: Calculation and design of cyclones on the basis of recent tests. Brennstoff-Warme-
Kraft, 8, 1-9.
Beeckmans, J. M., 1973: A two-dimensional turbulent diffusion model of the reverse flow cyclone.
J. Aerosol Sci., 4, 329-336.


Cyclones for Size-Selective Sampling

Bernstein, D., M. T. Kleinman, T. J. Kneip, T. L. Chan and M. Lippmann, 1976: Development of a
high-volume sampler for the determination of particle size distributions in ambient air. J. Air
Pollution Control Assoc., 26, 1069-1072.
Blachman, M. W. and M. Lippmann, 1974: Performance characteristics of the multicyclone aerosol
sampler. J. Amer. Industrial Hygiene Assoc., 35, 311-316.
Caplan, K. J., L. J. Doemeny and S. D. Sorenson, 1977: Performance characteristics of the 10 mm cy-
clone respirable mass sampler, Part I-monodisperse studies. J. Amer. Industrial Hygiene Assoc.
38, 83-95.
Chan, T. L. and M. Lippmann, 1976: Particle collection efficiencies of air sampling cyclones.
Environ. Sci. Technol., 11, 372-382.
Chang, H. C., 1974: A parallel multicyclone size-selective particulate sampling train. J. Amer.
Industrial Hygiene Assoc., 35, 538-545.
Ettinger, H. J., J. E. Partridge and G. W. Royer, 1970: Calibration of two-stage air samplers. J.
Amer. Industrial Hygiene Assoc., 31, 537-545.
Kotrappa, P., 1971: Revision of Lippmann-Harris calibration for two-stage sampler using shape
factor. Health Phys., 20, 350-351.
Lapple, C. E., 1963: Dust and mist collection. Chemical Engineers Handbook, J. H. Perry, Ed. New
York, McGraw-Hill.
Laskin, S., 1950: The modified cascade impactor. UR-129, Rochester, New York, The University of
Rochester Atomic Energy Product.
Leith, D. and W. Licht, 1972: The collection efficiency of cyclone type particle collectors-A new
theoretical approach. AIChE Symposium Series: Air Pollution and its Control, 68, 196-206.
Lippmann, M., 1970: "Respirable" dust sampling. J. Amer. Industrial Hygiene Assoc., 31, 138-
Lippmann, M., 1976: Size-selective sampling for inhalation hazard evaluations. Fine Particles, B.
Y. H. Liu, Ed., New York, Academic Press, 287-310.
Lippmann, M. and W. B. Harris, 1962: Size-selective samplers for estimating respirablee" dust
concentrations. Health Phys., 8, 155-163.
Lippmann, M. and A. Kydonieus, 1970: A multi-stage aerosol sampler for extended sampling
intervals. J. Amer. Industrial Hygiene Assoc., 31, 730-738.
Lippmann, M. and T. L. Chan, 1974: Calibration of dual-inlet cyclones for respirablee" mass
sampling. J. Amer. Industrial Hygiene Assoc., 35, 189-200.
May, K. R., 1945: The cascade impactor, an instrument for sampling coarse aerosols. J. Sci.
Instrum., 22, 187-195.
Rosin, P., E. Rammler and W. Intelman, 1932: Principles and limits of cyclone dust removal. Z.
Ver. dt. Ing., 76, 433-437.
Seltzer, D. F., W. Bernaski and J. R. Lynch, 1971: Evaluation of size selective presamplers. II.
Efficiency of the 10 mm nylon cyclone. J. Amer. Industrial Hygiene Assoc., 32, 441-446.
Tomb, T. F. and L. D. Raymond, 1969: Evaluation of collecting characteristics of horizontal
elutriator and 10 mm nylon cyclone gravimetric dust samplers. Presented at the 1969 Annual
Meeting of the American Industrial Hygiene Association, Denver, May.



Kettering Laboratory, University of Cincinnati, Cincinnati, Ohio

Our interest in cyclones has been related to the two-stage sampling of dust to
provide a respirablee" fraction for correlation with health effects. The first
portion of our comments, therefore, will discuss the use of small cyclones for this
purpose, and the interpretation of results which are obtained from the cyclone
sampling. The balance of the talk will discuss results of laboratory experiments
with miniature cyclones used for two-stage sampling and give some of our
theories as to the performance of the cyclones.
The small cyclones are routinely used to sample for dust which may produce
silicosis or coal workers' pneumoconiosis. Because this is a routine procedure,
the main requirements are convenience and low cost with accuracy. It is these
factors of convenience, cost and availability which have led to the widespread
use of cyclones, in particular the 10-mm Dorr-Oliver cyclone, in North America.
Our comments are directed toward the accuracy with which we represent the
relative health hazards by these measurements. In addition to the 10-mm cy-
clone, we have used the Health and Safety Laboratory (HASL) 1/2"-cyclone
(Unico 18) and the Aerotec 2 most often, but we have also used the Aerotec 3/4"
and the HASL 1" (Unico 240). Based primarily on results of comparative "res-
pirable" mass and impinger sampling in the Vermont granite sheds, a threshold
limit value for quartz based upon two-stage sampling was recommended (Ayer,
1969). A subsequent trip to the granite sheds, this time a pilot study for granite
cutters operating their equipment without exhaust ventilation, produced the aver-
age data of Table I (Ayer et al., 1973). From this table it may be noted that for
practical purposes the 10-tpm elutriator made in the NIOSH shops, the Isleworth
sampler, the HASL 1/2" cyclone at 10-4pm, and the 10-mm nylon cyclone at
1.7-fpm were equivalent in average performance.
Lippmann and Chan (1974) have presented a calibration of dual inlet cyclones

Cyclone Discussion

including the Aerotec 3/4" and 2" models, and the HASL 1/2" and 1" models.
Others attempting to use these cyclones at the flow rates recommended by
Lippmann and Chan have encountered inconsistencies, however. Yablonsky et al.
(1975) noted that the concentration of a clay dust cloud passing the Unico 240
(HASL 1") was 1.3 times the concentration passing the Aerotec 3/4" when both
were operated at the flow rates recommended by Lippmann and Chan. Thompson
et al. (1976) in using the Unico 240 and the Aerotec %3" at the recommended flow
rates, noted that the Unico 240 passed 1.24 times as much dust as the Aerotec
/4". Using an Andersen Impactor, Yablonsky determined that the Aerotec 3/4"
was passing essentially that fraction of the total dust which would be expected for
a cyclone with characteristics which corresponded to the retention criteria
specified by the Threshold Limit Values Committee of the ACGIH (1970).
Thompson et al. (1976) used a Coulter counter to determine the efficiency of the
cyclone for various size fractions and determined that the Aerotec 3/4" operating
at its recommended flow rate performed very close to the criteria specified in the
Threshold Limit Values. Thus, independent analyses by differing methods con-
firmed the calibration of Lippmann and Chan for the Aerotec 3/4", but produced

TABLE I. Average Respirable Dust Concentrations


Horizontal Elutriator Samplers
Isleworth (MRE) sampler (2.5-4pm) 10.7
10-4pm elutriator 11.6
Hexhlet (50-fpm) 14.4
Cyclone Samplers
/2-inch steel cyclone at 10-4pm 10.9
10-mm nylon cyclone at 1.7-4pm 10.7

rather different results for the Unico 240. We believe that data obtained in a study
of the performance of miniature cyclones (Hochstrasser, 1976) can possibly
explain some of these discrepancies.
The study was performed using the "high efficiency" Stairmand design
(1951). Stairmand's design was varied by using two shorter cone lengths, three
different outlet inside diameters, and three different outlet outside diameters
giving a total of 15 configurations of cone and outlet diameters. The basic body
diameter was 19 mm. For the cyclones with the shorter cones, the pressure drop
across the cyclone varied regularly with the flow as shown in Figure 1. However,
with the long cone configuration, it was found that there were two definite
pressure-flow regimens which could be defined by whether or not the flow in the
outlet pipe was laminar or turbulent, i.e., whether the Reynolds number was
less than 2,000 or greater than 4,000. When the Reynolds number was between
2,000 and 4,000 either the laminar or the turbulent flow pressure drop could
exist, with the turbulent condition being the more stable. This is illustrated in
Figure 2. The prediction of cyclone efficiency for a Stairmand design cyclone by
Leith and Licht (1972), Lapple (1950), and Barth (1956) are shown in Figure 3
for one of the higher flow rates. The predicted efficiency according to Blachman








4 5 o 7 8 9 10

Q-Flow Through Cyclone (m /sec x 10 ')

Figure 1. Pressure drop vs. flow for cyclones nos. 2, 3, 5, 6, 8, and 9.






1.5 2 2.5 3

















F 4 F ~ 4 F -I----,




-... ----.. Legend --
_. __.

S- ---- Stable Flow
S----Unstable Flow
-.- ... + .. -- TF Line for Turbulent ---
L F@ Outlet Pipe Flow
_ r LF Line for Laminar
- Outlet Pipe Flow --_
_--:-- --A--77

--i-- F -

1 1.5 2 2.5 3

4 5 6 7 8 9 10


E7T V. F



















Q-Flow Through Cyclone (m3/sec x 10 4)
Figure 2. Pressure drop vs. flow for cyclones nos. 1, 4 and 7.

T7-i 7_- 1 -

Cyclone 7 -

Cyclone #4

Cyclone #1


74 Inertial Classification

and Lippmann (1974) is also shown, along with the experimental efficiency data.
It may be seen that the observed efficiency is predicted reasonably well by either
the Barth equation or that of Blachman and Lippmann. In this case there ap-
peared to be little difference in efficiency for large particles whether the outlet

- F--.

" ; ~ ("^ '" -: ; I

--F -:-. F. .^- :-- :: .-7 -


:_ I ._ :. / i L L
i, __!___/_-_______ I I

_,. ; .. / I-
/ ..... K-- -7.. -: i -:: -

I- "v I : -

-1 -1+




.., 1-.

I 7r
/ : 7 I
.. .. .. .. .. .. r .I . r T -

_I/_: ,: .. ,-.[: : I _-t ...
II ) :" :1 I: : I .. l I ~ l : l I I I ..t '

..: .-- --. -r -. -.--- -- : -- .. .. I-; -' : ": .
i1/- -fL : -- "____-.-__--_::l: F-: : ; :: -

.. .. .. . .. 2 T .. : -- --- -- -- .- --_ ;- ., :. : ._ --r: .

_- __-"-_ __- :_. ..- I---: I -:. ::: .::. I -:_-_ .;. "
F--T--- .-F ...i -:-:::: K -;





0 1 2 3 4 5 6 7
dp (MMAD)-Particle Diameter (,A)
Figure 3. Cyclone efficiency vs. particle size for cyclone no. 1 at Q = 3.32 x 10-4 n3/sec.

Cyclone Discussion

pipe flow was laminar or turbulent. At the lowest flow for cyclone configuration
No. 1, (illustrated in Figure 4) only the Blachman and Lippmann prediction
reasonably matched the observed efficiencies. It may be noted at this lowest flow
the 50% efficiency point was roughly 3.5 /m aerodynamic diameter, the 50%
efficiency specified in the TLV criteria. It is theorized that when the outlet pipe


90 .---"
80 _1_

.. ---7 ... .




SA DATA (Laminar O.P. FLOW)
- y- -::t-----.---. -----LAPPLE

5 6 7

Figure 4. Cyclone efficiency vs. particle size for cyclone no. 1 at Q = 1.66 x 10-4 m3/sec.

0 1 2 3 4
dp (MMAD)-Particle Diameter (u)


Inertial Classification

flow is turbulent, the cyclone is operating in the conventional manner as illus-
trated in Figure 5. However, when the outlet pipe flow is laminar, the flow
pattern in the cyclone is different; the outlet vortex does not form in the normal
manner, but rather the flow pattern is as shown in Figure 6. The existence of the
"zone of stagnation" shown in this diagram is confirmed by both a ring of
particles collecting at the plane of transition with none below this point, and by

I|Edd y

*- Main Vortex

Vortex Core

Figure 5. Eddy and vortex flows in a cyclone collector.

the fact that there is no further change of pressure inside the cyclone below the
plane of transition. The cyclones calibrated by Lippmann and Chan came rea-
sonably close to the ACGIH retention criteria only when the cyclones were
operating with the outlet pipe in the laminar flow condition.
It thus appears that the operation for respirablee" dust sampling cyclones is
dependent upon the laminar flow condition existing in the outlet pipe and the
short circuited flow condition. If this is so, then it is imperative that the outlet


Full Text
xml version 1.0 encoding UTF-8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd
INGEST IEID E9XB5N1GI_35PHFG INGEST_TIME 2013-07-25T21:53:57Z PACKAGE AA00010114_00001