United States
Environmental Protection
Agency
Municipal Environmental Research EPA 600 279127
Laboratory August 1979
Cincinnati OH 45268
Research and Development
Optical Detection of
Fiber Particles in
Water
-------
RESEARCH REPORTING SERIES
Research reports of the Office of Research and Development. U.S. Environmental
Protection Agency, have been grouped into nine series. These nine broad cate-
gories were established to facilitate further development and application of en-
vironmental technology. Elimination of traditional grouping was consciously
planned to foster technology transfer and a maximum interface in related fields.
The nine series are:
1 Environmental Health Effects Research
2. Environmental Protection Technology
3. Ecological Research
4. Environmental Monitoring
5. Socioeconomic Environmental Studies
6. Scientific and Technical Assessment Reports (STAR)
7. Interagency Energy-Environment Research and Development
8. "Special" Reports
9. Miscellaneous Reports
This report has been assigned to the ENVIRONMENTAL PROTECTION TECH-
NOLOGY series. This series describes research performed to develop and dem-
onstrate instrumentation, equipment, and methodology to repair or prevent en-
vironmental degradation from point and non-point sources of pollution. This work
provides the new or improved technology required for the control and treatment
of pollution sources to meet environmental quality standards.
This document is available to the public through the National Technical Informa-
tion Service, Springfield, Virginia 22161.
-------
EPA-600/2-79-127
August 1979
OPTICAL DETECTION OF FIBER PARTICLES IN WATER
by
S. R. Diehl, D. T. Smith, M. Sydor
Department of Physics
University of Minnesota, Duluth
Duluth, Minnesota 55812
Grant No. R804361-02-0
Project Officer
Gary S. Logsdon
Drinking Water Research Division
Municipal Environmental Research Laboratory
Cincinnati, Ohio 45268
MUNICIPAL ENVIRONMENTAL RESEARCH LABORATORY
OFFICE OF RESEARCH AND DEVELOPMENT
U.S. ENVIRONMENTAL PROTECTION AGENCY
CINCINNATI, OHIO 45268
-------
DISCLAIMER
This report has been reviewed by the Municipal Environmental Research
Laboratory, U.S. Environmental Protection Agency, and approved for publica-
tion. Approval does not signify that the contents necessarily reflect the
views and policies of the U.S. Environmental Protection Agency, nor does
mention of trade names or commercial products constitute endorsement or
recommendation for use.
ii
-------
FOREWORD
The Environmental Protection Agency was created because of increasing
public and government concern about the dangers of pollution to the health
and welfare of the American people. Noxious air, foul water, and spoiled
land are tragic testimony to the deterioration of our natural environment.
The complexity of that environment and the interplay between its components
require a concentrated and integrated attack on the problem.
Research and development is that necessary first step in problem
solution and it involves defining the problem, measuring its impact, and
searching for solutions. The Municipal Environmental Research Laboratory
develops new and improved technology and systems for the prevention, treat-
ment, and management of wastewater and solid and hazardous waste pollutant
discharges from municipal and community sources, for the preservation and
treatment of public drinking water supplies, and to minimize the adverse
economic, social, health, and aesthetic effects of pollution. This publica-
tion is one of the products of that research; a most vital communications
link between the researcher and the user community.
Water filtration for asbestiform fiber removal has been studied at
Duluth, Minnesota and Seattle, Washington, and a 30 million gallon per day
plant has been built and is operating at Duluth. This report describes
research to develop a rapid means of detecting fibers in water so that the
quality of potable water can be monitored as it is actually being produced,
in contrast to days or weeks of delay necessary for electron microscope
analysis.
Francis T. Mayo, Director
Municipal Environmental Research
Laboratory
111
-------
ABSTRACT
Light scattering by individual particulates is used in a multiple-
detector system to categorize the composition of suspended solids in terms of
broad particulate categories. The scattering signatures of red clay and
taconite tailings, the two primary particulate contaminants in western Lake
Superior, along with two types of asbestiform fibers, amphibole, and chryso-
tile, were studied in detail. A method was developed to predict the
concentration of asbestiform fibers in filtration plant samples for which
electron microscope analysis was done concurrently. Fiber levels as low as
5 x 10 fibers/liter were optically detectable. The method offers a fast
and inexpensive means for measuring, either on a continuous basis or as
discrete samples, the fiber levels of filtration plant output. Further
calibration of the instrument could enable analysis for other specific
particulate contaminants as well.
This report was submitted in fulfillment of Grant No. R804361-02-0 by
the University of Minnesota, Duluth Department of Physics under the sponsor-
ship of the U.S. Environmental Protection Agency. This report covers the
period March 8, 1976 to September 7, 1978, and work was completed as of
May 16, 1978.
IV
-------
CONTENTS
Foreword ±±±
Abstract iv
Figures vi
Tables vii
Abbreviations and Symbols viii
Acknowledgment ix
1. Introduction 1
2. Conclusion 3
3. Recommendations 4
4. Apparatus 6
5. Analysis Method 7
6. Results 9
Individual particle types 9
Predicting concentrations 12
7. Application to Samples Containing Fibers 14
8. Refinements to the Apparatus 21
References 48
Appendix 50
-------
FIGURES
Number Pagt
1 Size distribution of fibers found in Duluth drinking water .... 24
2 Block diagram of apparatus 25
3 Number of events at the ± 45° detector pair versus concentra-
tion for each particle type 26
4 Region assignments for the ± 45° detectors 27
5 Region assignments for the ± 90° detectors 28
6 Region assignments for the + 45° and + 135° detectors 29
7 Region assignments for the + 45° and - 135° detectors 30
8 Percent of total counts versus counting region at ± 45° for
each particle type 31
9 Percent of total counts versus counting region at ± 90° for
each particle type 32
10 Percent of total counts versus counting region at + 45° ,
- 135° for each particle type 33
11 Percent of total counts versus counting region at + 45°,
+ 135° for each particle type 35
12 Predicted versus measured total fibers for Duluth water
samples and ± 45° scattering angles 37
13 Total counts at ± 45° versus EM total fiber concentrations .... 38
14 Predicted versus measured total fibers for Duluth water and
± 90° scattering angles 39
15 Predicted versus measured total fibers for Duluth water and
+ 45°, - 135° scattering angles 40
16 Predicted versus measured total fibers for Duluth water and
+ 45°, + 135° scattering angles 41
17 Predicted versus measured total fibers for Seattle raw water ... 42
18 Predicted versus measured total fibers for Seattle finished
water 43
19 Comparison of predicted total fibers and EM measurements for
data collected using the modified apparatus 44
20 Prediction of amphibole fibers versus the EM measurements for
the modified apparatus 45
21 Particle counter data versus EM amphibole fiber counts 46
22 Particle counts versus EM total fiber counts 47
VI
-------
TABLES
Number Page
1 Coefficients of for the ± 45° fit to Duluth filtered water .... 16
2 Coefficients for ± 90° fit 17
3 Coefficients for the + 45°, - 135° fit 18
4 Coefficients for the + 45°, + 135° fit 18
5 Coefficients for the fit to Seattle raw water 20
6 Coefficients for the fit to Seattle filter water 20
vii
-------
LIST OF ABBREVIATIONS
ADC analog to digital converter
A/D analog to digital
C coincident
EM electron microscope
Hz hertz
mm millimeter
N non-coincident
SAD selected area diffraction
sec second
ym micrometer
Mg microgram
Ug/J, microgram per liter
°C degree centrigrade
viii
-------
ACKNOWLEDGMENTS
We would like to offer our sincere thanks to the personnel at the
Duluth Filter plant for their cooperation. We are also especially grateful
to the people of the Lake Superior Basin Studies Center who were involved
with the collection of the electron microscope data. Their dedication and
assistance were greatly appreciated.
IX
-------
SECTION 1
INTRODUCTION
Over the past several years the presence of fibrous asbestiform parti-
culates has been observed in a number of municipal water supplies throughout
the U.S. The possible health hazard which these fibers present has spurred
a great deal of interest in the problems of detection and removal of the
submicroscopic particulates in water. While the removal of amphibole-
asbestos has reached a high state of the art in the Duluth, Minnesota
filtration plant, advanced detection techniques have been slow in developing
and investigators have been forced to rely on tedious and time consuming
methods.
Electron microscopy and x-ray diffraction techniques are the most widely
used in the detection of asbestiform fibers. Both techniques require expen-
sive equipment, tedious sample preparation, and have other severe limitations
X-ray diffraction has a very low detection sensitivity (^ 10 ug)l»2 and
requires a relatively large amount of asbestiform material, thus necessita-
ting filtration of a very large volume sample, (over 50 liters has been
needed for some clean samples). Filtration of large volume samples creates
problems since the filters often become completely plugged .with suspended
material, such as the aluminum hydroxide present in filter plant effluent,
making the collection of enough asbestos material for analysis impossible.
o
Electron microscopy has good sensitivity (0.1 pg has been claimed) but
is both costly and time consuming requiring from several hours to two days
for one sample analysis. Electron microscope results are also highly depend-
ent upon the method of sample preparation and interpretation by the observer
doing the actual counting. Thus repeatability is poor and the error for a
particular low-concentration sample can be very high.
The purpose of this paper is to introduce a new technique for the
detection of submicroscopic fibers which is fast, inexpensive, and has demon-
strated high repeatability. The method employs single-particle scattering
in which a focused laser beam passing through a liquid sample illuminates
single particles at the focal point of a system of detectors. The signal
received at the detectors is recorded sequentially for each particle drift-
ing through the focal point. At the end of the test the data are analyzed
for particle type and particle concentration. Thus far, the primary purpose
has been to detect rod-like particles and the data has been correlated only
to fibrous particles, but other particle symmetries can be discerned with
proper calibration procedures.
-------
The scattering of electromagnetic waves by dielectric objects has
received much attention over the years. The fundamental theory was given by
Lord Rayleigh in 1871-5 and has been extended by many others since then. The
theory is straight-forward for the cases in which the scattering bodies are
very small (Rayleigh Scattering), or very large (geometrical optics)
compared to the wavelength of the incident radiation. However, the mathe-
matics for the so-called "resonance region," in which particle dimensions are
of the same magnitude as the wavelength, is complex. In this region the
complete wave nature of the incident radiation must be considered. While
early solutions to the scattering problem were found for spheres in three
dimensions, (Mie theory)6»7 ancj infinite circular cylinders >>° in two
dimensions, only recently have solutions been obtained for arbitrarily shaped
bodies including finite circular cylinders.
Work by Barber and Yeh^ as well as that by Birkhoff et al. has shown
that the scattered radiation from rod-like particles is dependent upon the
angle of incidence to the radiation field. Mie theory demonstrates that the
radiation pattern of spherical particles depends only upon particle size and
index of refraction and has complete symmetry about the optical axis.
Observing the symmetry of the scattered light of single particles in a
monochromatic beam should thus allow for the discrimination of needle-like
particles from a mass of other particles in a sample.
The problem is complicated by the fact that the actual radiation
patterns for both spheres and cylinders show many maxima and minima. The
books by Van de Hulst" and Born and WolfH give radiation patterns for
spheres in the resonance region. Patterns for cylinders in various orienta-
tions can be found in the works by Barber and Yeh , Farone and Kerker^ , and
Kerker, Cooke, Farone, and Jacobsen . Particles of other than spherical or
cylindrical symmetry (e.g. ellipsoids) also show characteristic multiple-lobe
radiation patterns".
Asbestos fibers occurring in the environment are commonly of a size
which places them within the resonance region of visible light (Figure 1).
Thus the detection of these particles cannot be carried out by commonly
available instruments using volume reflectance measurements based upon Mie
or Rayleigh scattering. Furthermore, volume scattering has been shown to
be of limited usefulness for the detection of particles which lack charac-
-l / -i C
teristics that produce a volume interference effect14'-1- . Identification
should be possible, however, by observations of the symmetry of the
scattered light from individual particles. However, the problem is compli-
cated by the fact that the fibers exist in a medium containing many other
particles of varying shapes and sizes. It is thus necessary to find a
scattering signal unique to needle-like particles amid a great deal of back-
ground "noise" due to other suspended particles.
To this end an instrument has been designed using multiple detectors to
monitor the symmetry of the scattered light due to single particles in a
laser beam. Comparison of the scattering signal to electron microscope
fiber counts for each sample allowed the extraction of the fiber signature
from the background noise.
-------
SECTION 2
CONCLUSION
The multiple-detector scattering method has shown itself to be of
considerable value for the measurement of fiber concentrations in water. If
calibrated with a sufficient amount of EM data, the method actually yields
results which are more accurate than a single standard EM fiber analysis.
Furthermore, it seems likely that any particle type whose size is on the
same order of magnitude as the wavelength of the incident beam could be
distinguished from a background of other particulates. They need only to
differ in shape or the index of refraction. However, routine calibration
checks should be made and the calibration criteria should be updated as
additional data becomes available. Checks on irregularities of filtered
product could also be built in.
Once completely programmed, the microcomputer will make the apparatus
portable and extremely easy to operate. After answering a few questions
from the keyboard about such things as the desired test duration, the
operator must only wait for the concentrations to be displayed. If needed,
the apparatus could be modified to function as an on-line monitoring device
with the most recent fiber concentrations continually displayed. In fact,
the increased rate at which the particles would be carried through the beam
in a flow-through system would mean an increase in the accuracy of the meas-
urements for a given test duration.
-------
SECTION 3
RECOMMENDATIONS
An optical method for the detection of fibers in water has been
constructed to provide a capability for checking the effectiveness of filtra-
tion processes for removing fibers. The method differs from a turbidity
meter in that it provides not only the information on total turbidity as does
an ordinary turbidity meter or particle counter, but also provides informa-
tion on the actual fiber count. Thus, even if the total turbidity remains
relatively steady but the fiber removal process is not working properly, the
instrument would sound a warning within several minutes. This feature is
quite essential in monitoring plants where most of the turbidity is due to
particulates other than the fibers.
The instrument works on the principle of optical categorization of
turbidity samples particle by particle. In this method particulates are
identified by their light scattering envelope. A particle illuminated by a
beam of light scatters photons in various directions with intensities which
depend on particle size, its shape, its relative index of refraction, and
its orientation in the beam. The scattered light envelope is thus character-
istic of the physical properties of the particle. In this apparatus
individual particles drifting through a tiny volume of space are viewed
simultaneously by several detectors. The set of signals from the detectors
are fed into a microprocessor-computer which automatically identifies the
set of simultaneous pulses as those belonging to a scattering envelope
characteristic of a certain species of particulates. Thus, by counting the
abundance of various species of particulates we can determine the composition
of the suspended solids on a statistical basis. In the case of monitoring a
filtration plant output the instrument can be programmed to record only the
concentrations of fibers and the total particulate count. The system is
calibrated for detection of a specific particulate category, by using
electron microscopy data for that species in conjunction with scattering
information. The instrument remains in calibration provided no new major
species of particulates is introduced into the background and the optical
surfaces remain aligned and clean. Thus, one would anticipate, for mainten-
ance purposes, a monthly check on the optics and a semiannual check on
calibration.
The present system is a bench type instrument but the method should be
modified and applied to an on-line operation where it could greatly improve
the monitoring of the fiber removal process. The basic components and rough
costs for a system are: light source ($2,000), circuitry and detectors
($1,000), machining of chamber and detector mountings ($3,000), and
-------
microprocessor and programs ($4,000). The cost of calibration would be
around $8,000 plus the cost of electron microscope data.
The system can be constructed to require minimal operator skill to read
out fiber count and perform routine tests on the operation of the system.
The entire method can be reduced to simple push button operation. Actually,
for reasonably skilled operators, the instrument is flexible enough to
provide a facility for testing of plant operation efficiency and for data
compilation and recording.
-------
SECTION 4
APPARATUS
The schematic for the apparatus is shown in Figure 2. A laser beam
which is focused to a diameter of . 1 mm passes through a water sample
containing suspended particles. Six photodiode detectors are mounted around
the sample at the angles ± 45°, ± 90°, and ± 135° to the incident beam. As
individual particles drift through the beam, the resulting scattered light
produces, at each detector output, signals which appear as a series of pulses
of various sizes and shapes with an average width of about .2 - .3 seconds.
A lens in conjunction with a narrow aperature increases the light
gathering ability of each photodiode and yet limits the length of the beam
viewed by the detectors to 1 mm. This, together with the narrow beam
diameter, ensures, for low concentration samples, that only single-particle
events are viewed by the detectors at any time. In addition, since each lens
subtends an angle of about 10°, all the light scattered in a cone toward the
lens is detected by the photodiode producing an integrated signal which
smooths out the narrow angular fluctuation of the scattered light.
An analog to digital converter was designed to simultaneously sample any
two of the six detectors at a 15 Hz rate (by integrating for 1/15 second),
and transmit the data by telephone directly to the university computer where
it was stored for future analysis. A higher sampling rate was not possible
due to the fixed transmission rate of the computer phone link. To observe
adequately the characteristics of the scattering signal, a two channel
digital storage device was used to sample and display a 5 or 10 second time
segment of the two detector output.
-------
SECTION 5
ANALYSIS METHOD
In considering the scattered signal by examining simultaneous pulse
heights for the six detectors many sampling schemes were possible. Three
pairs of viewing angles were initially chosen: ± 45°, ± 90°, and + 45° with
- 135°. A fourth combination at + 45° and 4 135° was added later. The
polarization plane of the laser beam was always held perpendicular (vertical)
to the plane of the detectors. The parallel case was also examined but was
not considered in detail.
To measure accurately the concentrations of unknown samples, the number
of events seen per unit of time for any given sample should remain constant
and should be proportional to the concentration of particulates. It was
found, however, that weak convection currents caused by the difference in
temperature between the sample and its surroundings were able to influence
the rate at which particles drift through the beam. In fact, the event rate
for a typical sample chilled to near 0°C prior to testing, slowly dropped
over an 8 hour period to about one-fourth of its starting value. After 24
hours with the sample near equilibrium, particles would cease to travel
directly through the beam and would on occasion appear to rotate or move
repeatedly in and out of the beam.
The following test procedure was thus implemented to improve repeat-
ability and to obtain a high event rate:
1) The sample was first refrigerated at near 0°C for at least 12 hours.
2) One hour prior to testing, the sample was vigorously agitated to
thoroughly resuspend the particles and then again refrigerated to allow
bubbles to dissipate.
3) The sample was removed from refrigeration one-half hour prior to testing.
4) Ten minutes prior to testing, the sample was gently agitated to resuspend
any settled particles and placed in the scattering chamber.
5) Using the A/D converter, ten mintues of detector data was sent to the
computer for each angle pair.
Once data files were created for a sample, all the advantages and
options of the computer were available for data analysis. A program was
written which first searched each digitized detector output for a repetitive
minimum, i.e. baseline, to use for the detector zeros. The program then
-------
searched the data for peaks. Once a peak was located, its height and the
level of the signal in the accompanying detector were stored in memory.
However, only events which exceeded a predefined cutoff set sufficiently
above the noise level were stored. If the accompanying detector also
produced a peak, i.e. if the peaks occurred simultaneously, the zero-
corrected peak heights were stored together and labeled as a "coincident"
pair. Otherwise, the event for a given detector was labeled as non-coinci-
dent. In order to reduce the uncertainty between the measured and actual
peak heights due to the finite sampling rate, the stored value for the peak
height was averaged over the adjacent digital sampling times.
-------
SECTION 6
RESULTS
INDIVIDUAL PARTICLE TYPES
Files for the peak heights and the corresponding coincidence informa-
tion were created from the original digitized data for a variety of different
particle types. Amosite, red clay, and taconite tailings in two size ranges,
< 2 ym and 2-5 ym, were studied. Information on spherical scatterers was
provided by data from .6 and 1.1 ym uniform latex spheres. Canadian chryso-
tile with an average fiber length of 2 - 3 ym was also investigated. Four
concentrations of each sample type were tested as a cheek on linearity and
our dilution techniques, and each sample was run at least twice as a check
on repeatability.
A plot of the total number of events at the ± 45° detector pair versus
concentration can be seen in Figure 3 for the smaller particle size range.
Each sample counting rate was corrected for background by also testing the
water used for sample dilution. The counting rate from background water was
subtracted from the counting rates for the samples. The plotted points in
Figure 3 appear quite linear, and the small deviations from the straight-line
fits could easily arise from any number of factors including sample contami-
nation or sample dilution errors which are always possible at such low
concentrations. For the < 2 ym range at higher concentrations, the maximum
spread in the total number of events changed less than 5% after repeated
testing with the ± 45° detector pair. The spread was less than 10% for the
other three detector pairs. The lower gain and the improved zero stability
of the ± 45° detector configuration may account for this difference. Fluctu-
ations in the number of events were a little more noticeable for the larger
particle sizes. In such cases changes in particle settling rates become
noticeable even with small deviations in the test procedure.
Distribution plots for both the coincident and non-coincident events
can be found in the appendix for each of the particle types. Only the ± 45°
and the + 45° with - 135° detector pairs are shown in detail. The results
for the ± 90° and + 45° with + 135° pairs are similar. The ± 45° data were
categorized by sorting the events into a 21 x 26 element array on a plot
representing maximum pulse height along one axis and pulse ratio along the
other. The pulse ratios for coincident events were obtained by dividing the
smaller of the two peaks by the larger to give ratios always less than one.
For the non-coincident events the ratio value was taken as the level of the
detector at which no pulse occurred divided by the peak height found at the
opposite detector. Because of the symmetry of the ± 45° detectors, no
-------
information is lost by plotting ratios in this fashion, but instead, the
number of events of a given ratio are in effect doubled. It was hoped that
the pulse ratios would tend to be independent of particle size but dependent
on particle type.
For the distribution plots with the detectors at + 45° and - 135°, the
events were sorted into an array either by peak height versus peak height
in the case of coincident events or by peak height versus the level of the
opposite detector for the non-coincident events. Because of the lack of
symmetry of the + 45° and - 135° detector pair with respect to the beam, the
non-coincident events were categorized separately for the + 45° from those
which occurred at the - 135° detector.
All the samples of each particle type were combined to reduce statist-
ical fluctuations between adjacent array elements. The distributions were
background corrected by smoothing both the sample array and the background
array and then subtracting the corresponding array elements. (The smoothing
was done in a manner which conserved the total number of events and allowed
a change of at most four events per array element.) To compare between
particle types, the arrays were normalized by dividing the number of events
in each element of an array by the total number of events. Thus each
contour line on the pulse height versus pulse ratio array represents a line
of equal event probability given in percent. To avoid the loss of data,
whenever a large peak went off scale, the event was summed into a location at
the array boundaries. This accounts for the concentration of contour lines
seen near the boundaries of the arrays, particularly for the runs represent-
ing the larger particle types. For the ± 45° non-coincident case, events
with ratios larger than 1.0 were summed into the edge array locations. The
contour lines are also artificially compacted near the lower boundaries
because of the cutoff level.
The distribution plots for the .6 ym latex spheres reveal a number of
peculiarities of the equipment and the method. Due to spherical symmetry of
the latex particles, light should scatter symmetrically about the beam in
the plane of the detectors. Thus one would expect for the ± 45° detector
pair that all of the events would fall in the coincident plot and have
ratios of exactly one. Obviously, this is not exactly the case. Although
most of the events did bunch up on the coincident distribution plot near the
1.0 ratio boundary, about 25% of the events fell in the non-coincident plot.
One explanation for this is that the detectors are slightly out of align-
ment, and consequently, the lengths of the beam viewed by each detector do
not completely overlap. An occasional particle could be seen by one
detector and not the other, which would primarily account for the events with
low ratios.
The remainder of the non-coincident events, however, result from the
interaction of the low A/D converter resolution with the cutoff level. Due
to the ± 1 bit out of 7 bit (1.5% of full scale) accuracy of the A/D
converter, it is possible for a pulse to just exceed the cutoff level of one
detector but not the other. Such an event would thus be defined as non-
coincident. This is especially noticeable for + 45° and - 135 detectors
where the light scattered by the .6 ym latex spheres seldom if ever exceeded
10
-------
the - 135° detector cutoff and most of the events were labeled as non-
coincident.
The apparent concentration of events on the ± 45° plots at a few parti-
cular ratios is also caused by the finite step size of the A/D converter'.
At low peak heights the ratios begin to take on noticeably discrete values
due to the division of one small integer value by another. It should be
noted here that the Gaussian-shaped beam intensity and slow sampling rate
also produced a significant spread in the measured peak heights, despite the
uniformity of the latex spheres. But since such deviations from the mean
peak height are the same at both detectors, the ratios should be unaffected.
To make quantitative measurements of the amounts of various types of
suspended solids found in an unknown sample, it is necessary that each
particle type must have its own scattering signature. Aside from the more
obvious differences due to particle size, however, the pulse height and pulse
ratio distribution plots for the various particle types, excluding the latex
spheres, all have a somewhat similar pattern. This is not surprising when
one considers the asymmetric nature of such particles. For example, under
the electron microscope, many red clay particles appear as eliptical plate-
lets which could at least in some orientations scatter light in a fashion
similar to the fibers found in the amosite and chrysotile samples. The wide
variation in particle size and shape found even in just one general particle
type also tends to wash out any characteristic features of the scattered
light.
However, significant differences do exist. The ± 45° distribution plots
for < 2 um red clay have nearly twice the number of coincident events with
ratios near 1.0 than does any of the other particle types. This simply
suggests that the smaller red clay particles tend to be mojre spherically
symmetric. Comparing the red clay (< 2 um) and tailings (< 2 ym) contour
plots to the plots for amosite (< 2 ym) and chrysotile for the ± 45° non-
coincident distributions, one observes a shift in amosite and chrysotile
toward low pulse-height ratios. The samples of larger particle size (2-5
um) also showed pronounced differences; amosite produced more high-intensity
off-scale coincident events than did red clay even though red clay generated
more events with intermediate peak heights. Some of the contour plots also
revealed a few pronounced maxima. Chrysotile, for instance, has a build-up
of non-coincident events at the ratios of .2 and .5 on the ± 45° detector
pair plots.
Similar differences exist for the + 45° and - 135° detector pair. Red
clay again had the highest percentage of coincident events with small peak
heights. On the - 135° non-coincident plot, chrysotile produced a maximum
in the number of events which is over four times higher than the same region
for other event distributions. The association of fibrous material, spheres,
and block particles with certain regions on the pulse height and pulse ratio
plots provide a means of categorizing the signals in unknown samples into
fibrous and nonfibrous particle types, and a means for predicting particle
type concentrations.
11
-------
PREDICTING CONCENTRATIONS
Since the event distributions showed significant differences for
particle types, it should be possible to make quantitative predictions about
the concentration of an individual particle type in a sample. Consider
subdividing the event distribution plots of an unknown sample into n smaller
regions. In any given region the event rate, i.e. the number of events per
unit time, should be proportional to the concentrations of the various
particle types. Thus, for all of the n regions one has a system of equations
of the form:
Nl -
bl C2
(1)
N =
a C + b C + c C0 +
n 1 n 2 n 3
where N , N . . . , N
1 <- n
C-. , C-, C , . .
and a , b , c , . . .
= the event rate in each region,
= the unknown concentrations of various types of
suspended solids,
= the coefficients of proportionality.
Provided the subdivision of the pulse height-pulse ratio distribution
incorporates a sufficient number of regions to identify all major particu-
late types and provided the resulting system of n simultaneous equations is
linearly independent, one may solve for each of the unknown concentrations.
For example, the equation for C is given by
Cl " V Nl + V N2 + V N3
V Nn
(2)
where a ', b-i'> c-i"> ^i '> are functions of the original a's, b's, c's,
d's, . . . etc. So each unknown concentration can simply be expressed as a
linear sum of the event rates of the n regions.
In practice it is desirable to keep the number of unknowns to a minimum,
i.e. group the categories of particles which are not of interest into broader
regions. Unfortunately, it may be necessary to subdivide a classification of
a specific category of particles if the constituents vary independently from
sample to sample. For example, particles labeled as "fibers" may have to be
grouped by type or perhaps even length. To obtain total fibers, however, one
just sums the individual equations which again results in a linear sum of the
n region event rates, but with different coefficients.
12
-------
Solving for the necessary coefficients by experimenting with each
individual particle type which is present in an unknown sample would usually
be impractical, especially if one is only interested in the level of some
specific contaminant. In fact, it may not be possible to isolate or identify
some of the background particulates. One approach is to test samples whiph
contain various known amounts of the particle type that one desires to
identify and an independently varying amount of other background particulates.
Linear-regression analysis can then be used to find the set of coefficients
which gives a least-square fit to the measured concentrations of the known
particle type. The number and the manner of subdivision of the pulse height
regions chosen can be altered until a satisfactory fit is obtained. It
should be pointed out, however, that if one uses as many variables, i.e.
regions, as there are test samples, the linear regression will produce an
exact fit independent of the actual number of particle types. Therefore to
obtain a statically valid fit, the number of test samples should be at least
several times larger than the final number of variables.
13
-------
SECTION 7
APPLICATION TO SAMPLES CONTAINING FIBERS
During a six month period, over 60 Duluth and 30 Seattle water samples
were tested for which electron microscope (EM) data was available. All of
the Duluth samples were filter effluent from either the main plant which
supplies Duluth tap water or from an adjoining pilot plant operation. The
Seattle samples consisted of both pilot plant,filter effluent and raw intake
water. Since the samples contained measured concentrations of fibers,
including both amphibole and chrysotile asbestos, they presented an ideal
opportunity to directly test and improve the scattering method for the detec-
tion of fibers in water.
Prior to analyzing the filtration plant data, the pulse height distribu-
tion plots for each detector pair were subdivided into 12 regions in a manner
which accented the differences seen between the various test samples of the
known particle types: amosite, chrysotile, red clay, tailings, and spheres.
This yielded pulse height ratio regions which would differentiate at least
the tested particle types and possibly others. The choices for each of the
four detector pairs are depicted in Figures 4-7. Figures 8-11 show the
percentage of event types found in each region for the particle types tested
and for the Duluth and Seattle samples. The unique signatures for various
particulates is apparent at all four of the detector pairs.
Since the EM analysis for the Duluth and Seattle samples was performed
by independent laboratories using different techniques, the two sets of data
were handled separately. Otherwise, sources of error which vary from one
EM analysis to the other would make correlation of the optical data to EM
data difficult. If the data for Duluth and Seattle samples were to be
combined, the presence of different background particulates and different
types and sizes of fibers would also necessitate a large number of fitting
variables, complicating the interpretation of the results.
The primary fiber of interest in the Duluth water is amphibole asbestos,
although chrysotile fibers are also present in smaller amounts. The Duluth
EM data which we received included the counting of amphibole and chrysotile
fibers that were positively identified using selected-area diffraction (SAD).
Other fibers were also counted provided they had the correct morphology,
such as nearly parallel sides, even though no SAD patterns were observed or
the patterns were ambiguous. (The usual 3:1 aspect ratio was used to define
the cutoff between fibers and non-fibers.) The total number of fibers that
were recorded per sample was usually less than 20, of which only a few were
positively identified as amphibole or chrysotile. As pointed out by
14
-------
Leineweber^, the statistical uncertainty for such low numbers of fibers is
extremely high, and the resulting spread in the data makes it difficult to
evaluate the quality of the mathematical fits. To improve the situation, the
lower concentration samples were grouped by their number of observed fibers
and averaged. This tends to smear out any sample to sample variations which
might occur. Even with sample averaging, the number of positively-identified
amphibole or chrysotile fibers was too low to attempt a good correlation
with the scattering data, thus only the total number of recorded fibers was
used in the calibration procedure.
Least-square fits were made for each detector pair to determine which
of the pairs would be most suitable for fiber identification work. Trial
fits were first computed using all of the coincident and non-coincident
regions, (as many as were permitted by either computer limitations or the
number of samples). Regions of low significance were then eliminated,
regions which showed similar coefficients were combined, and the subsequent
fit was recomputed. This iterative process was continued until a satisfac-
tory fit was achieved based on as few region combinations as necessary.
The ± 45° detector pair produced the best attainable least-square fit
using only six variables. A plot of the fiber concentrations calculated
using the least-squares fit versus the actual EM fiber data can be seen in
Figure 12 for the Duluth samples. With a standard deviation of only .42
million fibers/liter, the quality of the fit is quite good, and suggests that
the optical scattering method, once calibrated with sufficient EM data, is
actually more accurate than the EM fiber analysis. When the number of pulse
height-pulse ratio region combinations is reduced still further, the quality
of the fit deteriorates somewhat. For instance, a four variable fit was
obtained with a standard deviation of .61 million fibers/liter.
Of the 60 samples tested, only 10 percent were omitted, usually for
obvious reasons such as excessive filter flocculent or fiber clumping.
Seven of the 21 data points represent sample averages. A constant was added
to the fitting equation to offset error which might shift the data. Elec-
tronic noise which could inadvertently be counted as an event is one such
source of error. A random loss of fibers to container walls or in the EM
analysis is also a possibility, as well as sample contamination.
Shown in Table 1 are the final region combinations and significance
values used to produce the fit. The coefficient values correspond to the
a', b% and c', etc. of equation (2) where the N^s are the summed event rates
of the listed region combinations. The significance values, defined as the
absolute value of the product of the coefficient and the average event rate
for each particular region combination yield a measure of the relative
importance of each variable. As regions were selected to produce the best
fit to the measured fiber concentrations, it became clear that only a few
regions played an important role in particle identification and many of the
peripheral regions supplied only small corrections to the fit. Thus the
final selection of the peripheral regions was influenced by the statistical
errors associated with the particular set of data used in the calibration.
15
-------
TABLE 1. Coefficents for the ± 45° fit to Duluth filtered water.
#
1
2
3
4
5
6
C
COEFF.
-.136 x 10"1
-.254
i
.453 x 10
.595 x 10"1
-.157
.950 x 10"1
.279
SIG.
1.336
17.257
3.937
7.020
1.176
11.819
REGION COMBINATIONS
C-l, C-3, N-3, N-4, N-7
C-2, C-12, N-ll
C-7, C-10, N-2
C-4, C-5, C-9, N-9, N-10
C-ll
C-6, C-8, N-8
The significance values and signs of the coefficients used for identifi-
cation of particle types in filtration samples vary in comparison to the
coefficients found for the samples of known particle types. For example, the
coincident regions 6 and 8 associated with the most significant positive
coefficient for filtration plant samples are also regions of high red clay
(< 2 ym) probability (see Figure 8), a background particle type which the fit
for fibrous particles should suppress. The main types of background particu-
lates in the Duluth filter effluent, however, are biological material, mostly
diatom fragments and bacteria, and aluminum hydroxide globules from the
filtering process. Thus it is not clear just how the test particle types
such as red clay and tailings actually influence the selection regions for
fibers in filtration samples, since the red clay and tailings may constitute
an insignificant fraction of the background of non-fibrous material in the
samples. Furthermore, the test samples for known particle types had differ-
ent particle size distributions than those of the Duluth filtration plant
samples. For example, about 25 percent of the amphibole fibers found in the
amosite samples were greater than 2 ym even though the samples were sized by
centrifuge to less than 2 ym. In contrast, most of the fibers in the Duluth
water were less than 1 ym.
Another factor which makes the interpretation of the coefficients
difficult, has to do with the mathematical nature of the fitting equation.
The condition necessary for a valid fit is not that one particle type be
more probable than another in a given region, but more specifically, that
changes in the probability differences (and ratios) exist from region to
region.
16
-------
One caution concerning the reliability of the fitting equation should be
considered. If from sample to sample the concentration of a particular back-
ground particle type remains directly proportional to the fiber concentra-
tion, the coefficients will adjust to give a least-square fit to both types
of particulates weighted toward the particle type which is more numerous.
Then if the ratio of fibers to this background particulate should ever
change, the equation would predict incorrect fiber levels. In view of the
fluctuating levels of the various suspended solids found at the filter plant
intake, it is unlikely that this will be a problem provided enough samples
are taken over a long enough period to allow for sufficient variability.
A plot of the total number of events during each 10 minute test period
versus the EM fiber concentrations for a series of filtration samples is
shown in Figure 13. Each point in Figure 13 was averaged identically to the
manner used in establishing the original calibration for fiber counts in
filtration samples. The large dispersion of the points in Figure 13 indi-
cates, as suspected, that the ratio of fibers to the total number of
suspended particles is not constant, supporting the fact that the calibration
actually represents the fiber count. This result also suggests that a
straightforward particle count alone would be of limited usefulness for
detecting the actual fiber concentrations. An electron microscope estimate
on all particulate constituents of filtration plant samples would be helpful
in interpretation of the pulse height-pulse ratio categories found for the
filtered water samples.
Least-square fits to the EM "total" fiber concentrations are shown in
Figures 14 - 16 for each of the other three detector pairs. The respective
coefficients for equation (2) are found in Tables 2-4. For the ± 90° and
TABLE 2. Coefficients for ± 90° fit.
#
1
2
3
4
C
COEFF.
_i
.763 x 10
.144
.162
i
.242 x 10
-.714
SIG.
2.77
12.59
13.17
.69
REGION COMBINATIONS
C-ll, N-2, N-3, N-5, N-12
C-l, C-2, C-5, C-7, C-8, C-10
C-4, C-6, C-12, N-7
N-9
17
-------
TABLE 3. Coefficients for the + 45°,- 135° fit.
#
1
2
3
4
5
6
C
COEFF.
-1
.267 x 10
-1
-.545 x 10
-.237
.343
.134
-.290
.415
SIG.
3.45
1.35
9.94
14.12
2.56
5.98
REGION COMBINATIONS
C-3, 1-4, 2-2, 2-5, 2-9, 2-12
C-8, 1-10, 2-6
C-5, C-6, C-7, C-12
C-9, 2-8, 2-11
C-10, 1-2 i
1-1
TABLE 4. Coefficients for the + 45°, + 135° fit.
#
1
2
3
4
5
6
C
COEFF.
-1
.843 x 10
_i
-.679 x 10
_i
-.500 x 10
_i
.939 x 10
_i
.612 x 10
-.352
.996
SIG.
.77
1.66
3.47
1.56
7.91
2.73
REGION COMBINATIONS
C-5
C-8
1-3, 1-5, 1-8, 1-12
1-9, 1-10
1-1, 1-4, 1-6, 2-5, 2-6, 2-8
2-11
2-1
18
-------
+ 45° with - 135° fits, the EM data and sample averaging were identical to
that used for the ± 45° case. However, a smaller amount of data was avail-
able for the + 45° and + 135° detector pair. The ± 90° fit was obtained
with only four variables since little improvement was found by using more.
This together with the point dispersion of Figure 14 indicates poor resblu-
tion of the ± 90° detector pair. Six variables were needed to produce
reasonably good fits for the other two detector pairs.
The standard deviation of these cases is about double that of the ± 45°
case, reflecting the fact that the repeatability was the best for the ± 45°
detector pair. The increased zero stability due to the lower gain settings
is the main difference between the ± 45° case and the other three detector
pairs. Both zero drift and the ± 1 bit accuracy of the A/D converter can
substantially influence the number of small peaks which exceed the cutoff
level. Another source of error which has also been mentioned earlier is due
to the statistical fluctuations in the number of fibers measured using EM
techniques. Even after averaging the samples, each data point usually
represented fewer than 50 fibers. Assuming that the EM data fluctuations are
random and can be described by the usual Poisson statistics, a best-care
error of about ± 15% is implied. A similar source of error results from the
statistical uncertainty associated with the number of counts in each pulse-
height region combination. Although for a typical sample as many as 1000
events may have been detected, once these are split into their respective
regions and inserted into the fitting equation, the possible error can become
appreciable.
The EM data for the Seattle samples consisted of counts of amphibole
and chrysotile fibers with no quantitative information on the other back-
ground particulates. Because the number of samples which contained
significant amount of both types of fibers were low, the two fiber counts
were summed to yield total fiber concentrations. The ± 45° detector pair
again produced the best results. The Seattle samples included both raw and
finished water with greatly differing fiber concentrations and background
material. Separate comparisons of EM and predicted concentrations from
optical data are shown in Figures 17 and 18 for the raw and filtered samples
respectively. Six variables were used for each calibration fit with the
coefficients found in Tables 5 and 6.
The finished water samples, Figure 18, yield a greater dispersion of
the points. This can be easily accounted for by the statistical uncertainty
associated with the EM data. Due to an insufficient number of samples,
averaging was not practical, and some of the points represent as few as 5 to
10 fibers. Furthermore, especially for the finished water, a fair percen-
tage of the fibers were often less than .5 um in length and it is not known
whether such small fibers would produce scattering peaks which would exceed
the cutoff level. It should also be mentioned that the average "sample fiber
concentration was quite low compared with that of the raw water, or even the
previously discussed Duluth samples. Thus, only a small percentage of the
scattering events were produced by fibers.
19
-------
TABLE 5. Coefficients for the fit to Seattle raw water.
#
1
2
3
4
5
6
C
COEFF.
_i
.882 x 10
i
.213 x 10
.112
.153
-.201
_i
-.526 x 10
.430 x 10"1
SIG.
55.49
18.63
155.73
76.23
114.18
72.27
REGION COMBINATIONS
C-2, N-l
C-4, C-10, N-5
C-6, N-8
C-12
C-8
N-4, N-6
TABLE 6. Coefficients for the fit to Seattle filtered water.
r
T
2
3
4
5
6
C
COEFF.
-.263 x 103
-.825 x 10~3
.618 x 10"4
.143 x 10~2
_2
-.572 x 10
.646 x 10~3
_i
.374 x 10
SIG.
.05
.06
.01
.42
.41
.19
REGION COMBINATIONS
C-l, N-3, N-7, N-10
C-5, N-5, N-12
C-3, C-10
C-2, N-l, N-4, N-6, N-8
N-9
C-6, C-8, C-12, N-2
The raw water samples, for which as many as 200 fibers were counted,
produced a much better fit, Figure 17. To ensure that no more than one
particle was usually in the beam, many of the raw samples were diluted prior
to testing.
20
-------
SECTION 8
REFINEMENTS TO THE APPARATUS
In the past six months much time has been spent incorporating a micro-
computer (Processor Technology, Sol-20) into the scattering equipment. A
number of distinct advantages are realized. The microcomputer has its own
seven-channel analog to digital converter with twice the resolution as the
one previously used. The increased accuracy plays special importance in the
die-termination of the baselines and in the measurement of small peak heights
where small errors can influence the number of events which exceed the cutoff
level. In addition, the microcomputer allows us to sample any number of the
six detectors at any desired rate up to about 200 Hz, which reduces the
uncertainty in the peak-height measurements inherent in the previous slow 15
Hz rate. Operator interaction is also greatly reduced by the microcomputer.
In the past it was necessary to secure the large university computer by phone
and continually reopen the data file every two minutes as the test
progressed. Thus test durations of over 10 to 20 minutes were impractical.
Using the microprocessor, the only limitation is due to an overflow of the
microcomputer memory which for low concentration samples can accommodate
runs of many hours. Finally, and perhaps the most important feature, the
microcomputer will eventually make the apparatus portable, and more useful
as a research tool or an on-line monitoring device.
After interfacing the microcomputer to the detectors, a machine language
program was written and debugged to do the following: 1) simultaneously
sample three detectors at 60 Hz rate, 2) find the baseline (zero) of each
detector, 3) search for peaks at each detector and determine if they are
coincident with peaks found at the other two detectors, and 4) store the
peak heights and coincidence information in the microcomputer memory. The
program divided the events into seven categories: three non-coincident, one
for each detector, and four coincident, one for each of the three combina-
tions of two detectors, and one indicating complete overall coincidence.
Since the complexity of the program and data analysis increases rapidly with
the number of detectors that are included, the program was written to handle
only three detectors simultaneously.
The sampling rate of 60 Hz, which is conveniently derived from the power
line, was chosen so that the A/D converter would sample in phase with the 60
Hz noise present on the detector outputs, thus further reducing sampling
errors. The cutoff level has also been lowered somewhat to improve the small
peak-height sensitivity. After the microcomputer memory is full or the
desired test duration is reached, the peak-height information is sent by
phone to the university computer. Although this greatly facilitates the
21
-------
research, the use of another computer will be unnecessary in the fiber detec-
tion system once a satisfactory calibration of the instrument with the EM
data is achieved.
Samples were tested at two different combinations of three detectors:
1) ± 45° with + 135° and 2) ± 135° with + 45°. So far the data has only
been analyzed for the first of these cases, chosen because of the promise
shown earlier by the ± 45 fits and by the pronounced signatures seen at the
+ 135° detector for the known particle types. Programs have been written to
further subdivide each coincidence type into three categories according to
peak-heights in the + 135° detector. A frequency distribution plot of the
- 45° versus the + 45° pulse heights was then made for each group.
When a new series of Duluth samples with EM data became available,
including both raw and finished water, the samples were tested using the new
three-detector method. Many of the samples were tested repeatedly, usually
at least a few days later, as a check on reliability. The results confirm a
phenomenon first pointed out by Tom Biele at the Duluth Filter Plant - the
number of particles in the finished water samples increases with time. For
a few of the pilot plant effluent samples, for instance, the total number of
events for a given test duration increased by nearly a factor of ten over a
two week interval. The continuing precipitation of the aluminum hydroxide
which is added in the filtration process is believed to be responsible. The
effect was not considered in the original testing of Duluth samples because
they were scanned optically a number of days after their filtration date at
which time the samples had stabilized.
Before applying the fitting programs to the new data, the frequency
distributions, three for each of the seven categories of coincidence events,
were subdivided into a total of 84 regions in a manner which accented the
differences observed for the known particle types samples. Next, a program
was written which, after finding the number of events per region for each
sample, computed the relative independence of the 84 regions. Finally, the
number of regions was reduced to a total of 24 by combining regions which
displayed little independence.
Shown in Figure 19 is a least-square fit to the total fiber concentra-
tions of the finished water samples. The term "total fibers" refers to the
previous discussion concerning the Duluth EM data. Six variables consisting
of various combinations of the 24 regions were used. Since the average
number of EM observed fibers per sample was around two to three times higher
than that of the previous Duluth water, the samples were not combined and
averaged. About 30% of the sample points, however, do represent less than
20 fibers, which may account for much of the point dispersion. Thirty
minute optical scans were usually employed.
A seven variable fit for determination of amphibole fibers concentra-
tion is shown in Figure 20. The quality of the fit is as good as one can
expect considering the small number of observed fibers. In fact, 16 of the
23 points represent less than 6 fibers per sample.
22
-------
Since it was not possible to test each sample immediately after it was
drawn, many of the samples still contained an abundance of aluminum hydroxide
particles at the time they were tested. The fact that a satisfactory fit was
obtained suggests that the scattering method is able to distinguish fibers
among a large amount of the precipitate.
For comparison, plots of both the amphibole and total fiber concentra-
tions versus total particle concentrations are shown in Figures 21 and 22 for
the two respective sets of data. The total particle counts were obtained by
the personnel at the Duluth Filter Plant using a Hiac, Model PC 320, particle
counter with a minimum particle-size detection limit of 1.0 um (average
particle dimension). The plots point out the great uncertainty encountered
by using a single-detector particle counter for fiber level measurements.
For example, in Figure 22 two of the points (represented by triangles) which
were plotted at nearly the same fiber concentrations differed in their total
particle count by over a factor of 100. In contrast, the fiber concentra-
tions of these same two samples were predicted accurately using the three-
detector scattering fit.
23
-------
.1 .2 .3
DIAMETER
23 4
LENGTH (/im)
Figure 1. Size distribution of fibers found in
Duluth drinking water. (Taken from
electron microscope data on 50 samples
drawn in 1977.)
24
-------
DETECTORS
LASER
LENS
POLARIZATION
ROTATOR
SCOPE
AMPLIFIER!<***!*
OFFSET »
TRANSIENT
STORE
GAIN *
AMPLIFIER
AMPL
1 '
A r\f»
, 4
1 A r\
GAIN
TO
COMPUTER
PHONE
MODEM
, » * .
SERIAL DATA
TRANSMITTER
Figure 2. Block diagram of apparatus.
25
-------
20
18
16
14
w.
2 12
x
co 10
z
3
o
0 8
2 -
/
P
V TAILINGS «2um)
A RED CLAY (<2 urn)
O AMOSITE (<2um)
D CHRYSOTILE
I I
01 23456789 10
CONCENTRATION ( ug/l x 10)
Figure 3. Number of events at the ± 45° detector pair
versus concentration for each particle type.
26
-------
zuu
1-
X
UJ
X 100-
x:
UJ
0.
50-
n-
1
2
3
5
6
4
f
8
9 II
10 12
.2 .4 .6 .8
DETECTOR RATIO
COINCIDENT
1.0
\£.V
90-
r
X
o
iU 60-
X
v
S ^0-
i
2
3
4
5
7
8
9
10
II
12
NON-COINCIDENT
.2 .4 .6 .8
DETECTOR RATIO
IJO
Figure 4. Region assignments for the ± 45°
detectors.
27
-------
X
CD
:*:
ui
80
60-
40
192
5
6
7
8
36
10
* 4
It
1
12
.2 .4 .6 .8
DETECTOR RATIO
COINCIDENT
1.0
«HJ-
>-
X
0
UJ
x 20-
UJ
1
C
~~"^
3
T
1
i
5
4
6
7
8
) .2 .4 .6
9
10
II
12
J .8 1.0
NON-COINCIDENT
DETECTOR RATIO
Figure 5. Region assignments for the ± 90°
detectors.
28
-------
12.0
9.0 J
6.0-
3.0-
10
II
12
COINCIDENT
20 40 60 80 100 120
UJ
X
UJ
Q.
-
5
**
r»-
2
y|
J H
o
-6J7p
J'
1
"fl"
>
3
r
i
i
L- 1
i i->
ii
0
10
15 20 25
PEAK HEIGHT + 45
NON-COINCIDENT
AT 45°
20 40 60 80 100 120
6.0'
45-
r»-
i a 12
2
4
7
3
5
8
6
9
10
II
J
NON-COINCIDENT
AT 135°
30
Figure 6. Region assignments for the + 45°
and + 135° detectors.
29
-------
Ifc.W
9.0
6.0
3.0-
o
in
m
1 "
H
X
O
X
UJ R_
a.
0.
0
C rt
o.O
4.5-
3.0-
1.5-
rt -
0
4
g
18
4
5
8
3
r-li
8
2
3
*
^
2
2
6
9
5|
r
0
1
0
1
*L
~6~
1
9
1
10
l_
10
40
5
i
1 O
IO
1
1
40
7
7
5
l
i
||
6(
1
60
h
J
r
5
2
4
12
1 i
80 100 12
2
NO. IO
a \d
8O 100 12
12
0
h
0
N
COINCIDENT
NON-COINCIDENT
AT 45°
NON-COINCIDENT
AT -I35°
5 10 15 20 25
PEAK HEIGHT + 45
30
Figure 7. Region assignments for the + ^
and - 135° detectors.
30
-------
COINCIDENT
NON-COINCIDENT
8
6
4
2
8
6
4
2
27 35
18
14
10
6
2
18
14
10
6
2
18
14
10
6
2
18
14
10
6
2
RED CLAY
<2um
2-5urn
_-' N
AMOSITE
<2um
2-5um
DULUTH
SEATTLE
LATEX
TAILINGS
CHRYSOTILE
123456789 10 II 12 123456789 10 IM2
REGION
Figure 8. Percent of total counts versus counting
region at r 45° for each particle type.
31
-------
COINCIDENT
NON-COINCIDENT
123456789 10 II 12
is
14
10
6
2
18
14
10
6
2
18
14
10
6
2
18
14
10
6
2
j
RED CLAY
< 2 um
AMOSITE
<2um
2-5 um
DULUTH
SEATTLE
i \ LAJEX
: TAKINGS
! ! CHRYSOTILE
I \
123456789 10 II 12
REGION
Figure 9. Percent of total counts versus counting
region at ± 90° for each particle type.
32
-------
6-5
O
U
O
E-t
8
6
4
2
8
6
4
2
8
6
4
2
8
6
4
COINCIDENT
RED CLAY
<2um
2-5 urn
NON-COINCIDENT+45'
AMOSITE
< 2um
2-5 um
DULUTH
LATEX
TAILINGS
CHRYSOTILE
r~f
,^::' v-
I 23456789 (Oil 12
18
14
10
6
2
I 23456789 (Oil 12
REGION
Figure 10. Percent of total counts versus counting
region at + 45°, - 135° for each particle
type. (continued on following page)
33
-------
NON-COINCIDENT AT-I35C
8
6
4
2
8
6
4
o
0
8
H e
o 6
H
4
2
8
6
4
2
-V
S\
I 23456789 1011 12
REGION
Figure 10 continued.
34
-------
COINCIDENT
NON-COINCIDENT AT+45
CO
E-i
§
O
8
6
4
2
8
6
4
2
RED CLAY
<2um
2-5um/\
AMOSITE
<2um
2-5um
4
DULUTH
SEATTLE
^*Ix
I I I
8
6
4
2
LATEX
TAILINGS
CHRYSOTILE
A
18
14
10
6
2
18
14
IO
6
2
18
14
10
6
2
18
14
10
r^ix
i i i i. i V i~*i i\
I 23456789 10II 12 123456789 1011 12
REGION
Figure 11. Percent of total counts versus counting
region at + 45°, + 135° for each particle
type. (continued on following page)
35
-------
NON-COINCIDENT AT+ 135*
O
O
H
O e
H 6
4
2
8
6
4
2
H-«-
_i i
I \
i I
i \
\ ' !
\ /x.' \
\ / »i
r >"** * V
*' _.xx>^x- A
I 2 3 4 5 6 7 8 9 10 II 12
REGION
Figure 11 continued.
36
-------
64
to
O
x
£48
tn
o
32
UJ
s
3
_J
O
|6
XX
- X
16
MEASURED
32
(fibers
48
'liter x I05)
64
Figure 12. Predicted versus measured total fibers
for Duluth water samples and ± 45°
scattering angles.
37
-------
20
DULUTH
45°
15
x
~ 10
CO
o
o
CONCENTRATION (total fibers/liter x I06 )
Figure 13. Total counts at ± 45° versus EM total
fiber concentrations.
38
-------
64
o
O
x
a>
= 48
32
CO
Ul
3
0 ic
UI 10
O
<
O
I I I I
16
48
64
MEASURED VALUES (fibers /liter x I05 )
Figure 14. Predicted versus measured total fibers
for Duluth water and ± 90° scattering
angles.
39
-------
I
I
16 32 48 64
MEASURED VALUES (fibers/liter x 1C?)
Figure 15. Predicted versus measured total fibers
for Duluth water and + 45°, - 135°
scattering angles.
40
-------
^ 64
10
o
x
_~ 48
-------
12
20
28
MEASURED VALUES (fibers/liter x I06)
Figure 17. Predicted versus measured total
fibers for Seattle raw water.
42
-------
o 28
X
to
UJ
12
o
o
4 12 20 28
MEASURED VALUES (fibers/liter x IQ4)
Figure 18. Predicted versus measured total
fibers for Seattle finished water.
43
-------
16
32
48
64
MEASURED VALUES (fibers/liter x I04)
Figure 19. Comparison of predicted total fibers and
EM measurements for data collected using
the modified apparatus.
44
-------
18
O
x
CO
Ul
O
£
14
10
XX
10
14
18
MEASURED VALUES (fibers /liter x I04)
Figure 20. Prediction of amphibole fibers versus the
EM measurements for the modified apparatus.
45
-------
40
10
2 30
cc
UJ
E
o" 20
CO
UJ
_l
o
K-
cr
<
CL
10
492
J260
o
0 8
o o
8
_ o
.04
.08
.12
.16
AMPHIBOLE FIBERS/LITER x
Figure 21. Particle counter data versus EM amphibole
fiber counts. (same data as used in amphibole
fit of Figure 20)
46
-------
40
D 30
o
X
OC.
Ul
w
>20
LU
1
PARTICl
10
492 t 26ot 2IIT
0
O
_. 0
o 8
00 ° °
o
o
o
o
_ o
o
o °
0 0 0
V
1 1 t 1 1 1
1 23456
TOTAL FIBERS/LITER x 10
Figure 22. Particle counts versus EM total fiber counts.
(same data used in 6 variable fit of
Figure 19)
47
-------
REFERENCES
1. Crable, J. V. Determination of Chrysotile, Amosite and Crocidolite by
X-ray Diffraction. Am. Ind. Hygiene Assoc. J., 27:293, 1966.
2. Rickards, A. L. Estimation of Trace Amounts of Chrysotile Asbestos by
X-ray Diffraction. Anal. Chem., 44:1072, 1972.
3. Rickards, A. L. Estimation of Submicrogram Quantities of Chrysotile
Asbestos by Electron Microscopy. Anal. Chem., 45:809, 1973.
4. Leineweber, J. P. Statistics and the Significance of Asbestos Fiber
Analysis. Presented at NBS Workshop, Guithersburg, MD, July 18 - 20,
1977.
5. Rayleigh, Lord. On the Scattering of Light by Small Particles. Phil.
Mag., 41, 447-454, 1871.
6. Van de Hulst, H. C. Light Scattering by Small Particles. Wiley Press,
New York, 1957.
7. Kerker, M. The Scattering of Light. Academic Press, New York, 1969.
8. Wait, J. R. Electromagnetic Radiation From Cylindrical Structures.
Pergamon Press, New York, 1959.
9. Barber, P. and C. Yeh. Scattering of Electromagnetic Waves by Arbitrar-
ily Shaped Dielectric Bodies. Applied Optics, 14(22):2864-2872, 1975.
10. Birkhoff, R. D., J. C. Ashley, H. H. Hubbel Jr., and L. C. Emerson.
Light Scattering from Micron-Size Fibers. J. Opt. Soc. Am., 67(4):564-
569, 1977.
11. Born, M. and E. Wolf. Principles of Optics. Pergamon Press, New York,
1964.
12. Farone, W. A. and M. Kerker. Light Scattering From Long Submicron Glass
Cylinders at Normal Incidence. J. Opt. Soc. Am., 56(4):481-487, 1966.
13. Kerker, M., D. Cooke, W. A. Farone, and R. A. Jacobsen. Electromagnetic
Scattering From an Infinite Circular Cylinder at Oblique Incidence. J.
Opt. Soc. Am., 56(4) .-487-491, 1966.
48
-------
14. Gibbs, R. J. Light Scattering From Particles of Different Shapes. J.
Geophys. Research, 83(11):501, 1978.
15. Diehl, S. R. and M. Sydor. The Feasibility of Using Optical Methods for
the Detection of Asbestos and Red Clay Particles in Lake Superior Water.
Mimeo, U.M.D. Dept. of Physics, Duluth, MN, 1975. 42 pp.
49
-------
APPENDIX
CONTOUR PLOTS OF VARIOUS PARTICLE TYPES
The following are contour plots of the pulse height versus ratio or
pulse height versus pulse height arrays for the ± 45° and + 45°, - 135°
detector pairs respectively. The plots were made for each particle species
tested by smoothing and normalizing the pulse height arrays and then connec-
ting equal event rates to give lines of equal event probability in percent.
These figures can thus be viewed as 3-dimensional frequency plots showing the
differences in scattering signatures of the various particle species.
50
-------
0.00
8
ftl
O
in
e>
O
0.00
.20 .40 .60 .80
DETECTOR RATIO
RED CLAY
<2
o
o
CM
X
O
Iu
UJ
0.
.1
CHRYSOTILE
< 2
.1
0.00
.40 .60
DETECTOR RATIO
.80
1.00
.20
.40 .60
DETECTOR RATIO
1.00
TACONITE TAILINGS
< 2
0.00
.40 .60
DETECTOR RATIO
LOO
Al. Contour plots of coincident events at the ± 45° detectors for the different particle
types investigated.
-------
Ul
N>
0.00
1-
X
(9
UJ
UJ
a.
x
(9
UJ
X
til
a.
o
10
.40 .60
DETECTOR RATIO
80
1.00 0.00
RED CLAY
2-5 /am
o
o
CM
O
t- «>
X
2
UJ
1 o
UJ
a.
0.00
.20
.40 .60
DETECTOR RATIO
1.00
DULUTH FILTER EFFLUENT
.20
.20 .40 .60 .80 1.00
DETECTOR RATIO
LATEX SPHERES
0.6
.40 .60 .80
DETECTOR RATIO
1.00
Al.continued
-------
Ul
UJ
A2,
CHRYSOTILE
< 2 /tm
0.00
.40 .60
DETECTOR RATIO
.40 .60
DETECTOR RATIO
.80
il.OO
8-
o
UJ
* O
-------
o
N
O
I.OO
LATEX SPHERES
0.6
0.00 .20 .40 .60
DETECTOR RATIO
.80 il.OO 0.00
A2. continued
.20
.40 .60 .80
DETECTOR RATIO
bl.OO
-------
A3,
40 60 80 100 120 0 20 40 60 80 100
PEAK HEIGHT+45 DEG. PEAK HEIGHT*45 DEG.
Contour plots of coincident events recorded at the + 45°, - 135° detectors for
different particle types investigated.
120
AMOSITE
< 2/im
CHRYSOTILE
2
60 80
HEIGHT+45 OEG.
40 60 80
PEAK HEIGHT4-45 OEG.
TACONITE TAILINGS
2ttm
RED CLAY
< 2
the
-------
Ln
O
U
Q Q
*><"
1C
u
X
AMOSITE
2-5
(9
IU
in
7
_
Hi
x
*
<
a.
20 40 60 80
PEAK HEIGHT-I- 45 DEG.
100
120
o
en
in
o
o (0
UJ
LATEX SPHERES
0.6/xm
I
I
I
20 40 60 80 100
PEAK HEIGHT+45 DEG.
120
(M
o q
to
I
£ o
-------
o
Q
CVJ
(9
<£
UJ
UI
0.
§.
AMOSITE
< 2/tm
I
20 40 60 80
PEAK HEIGHT+ 45 DEG.
100
20
40 60 80 100
PEAK HEIGHT4-45 DEG.
120
TACONITE TAILINGS
< 2/xm
60 80 100
HEIGHT+45 DEG.
120
20
40
PEAK
60 80 100
HEIGHT + 45 DEG.
120
A4. Contour plots of events non-coincident at 45° for the + 45°, - 135° detectors.
-------
00
o
UJ
o o
to
in
10
T
(9
UJ
X
u
a.
in
8
AMOSITE
2-5
o
20
40
60
80
100
PEAK HEIGHT + 45 DEG.
LATEX SPHERES
0.6
60 80
HEIGHT* 45 DEG.
DULUTH FILTER EFFLUENT
60 80
HEIGHT4-45 DEG.
100 120
A4. continued
40 60 60 100
PEAK HEIGHT + 45 DEG.
-------
o
111
Q
m m.
ro *
I
t-
X O
V) Q
UJ K>
< o
uj «
Q.
8
AMOSITE
< 2
I
1
0.0 5.0 10.0 15.0 20.0
PEAK-HEIGHT* 45 DEG.
O 10
10
s
10
U
X
Id
0.
-vr
RED CLAY
< 2
I
25.0 30.0
0.0
A5.
5.0 10.0 15.0 20.0 25.0
PEAK HEIGHT 4-45 DE6.
30.0
CHRYSOTILE
< 2/im
15.0 20.0
HEIGHT+45 OEG
30.0
m i
jo
I
t- o
x o
2 «
UJ
X
< in
TACONITE TAILINGS
< 2
I
I
J_
'0.0 5.0 10.0 15.0 20.0 25-0
PEAK HEIGHT+-45 DEG.
30.0
Contour plots of events non-coincident at - 135° for the + 45°, - 135° detector
pair.
-------
AMOSITE
.1 a
_L
I
O
O
C9
LJ
0 O
in ">
10 t
Ig
in
5.0
10.0 15.0 20.0 25.0
PEAKHEIGHT445 DEG.
30.0
o
q
d
DULUTH FILTER EFFLUENT
I
0.0 5.0 10.0 15.0 20.0
PEAK HEIGHT*45 DEG.
25.0 30.0
ON
O
o
o
<0
(9
So
LATEX SPHERES
0.6/xm
j_
I
I
0.0 5.0 10.0 15.0 20.0
PEAK HEIGHT»45 DEG.
25.0 30.0
RED CLAY
2-5/im
0.0
10.0 15.0 20.0 25.0
PEAK HEIGHT*45 DEG.
30.0
A5. continued
-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before completing)
1. REPORT NO.
EPA-600/2-79-127
2.
3. RECIPIENT'S ACCESSION NO.
4. TITLE AND SUBTITLE
Optical Detection of Fiber Particles in Water
5. REPORT DATE
August 1979 (Issuing Date)
6. PERFORMING ORGANIZATION CODE
7. AUTHOR(S)
S.R. Diehl, D.T. Smith, M. Sydor
8. PERFORMING ORGANIZATION REPORT NO.
9. PERFORMING ORGANIZATION NAME AND ADDRESS
Department of Physics
University of Minnesota, Duluth
Duluth, Minnesota 55812
10. PROGRAM ELEMENT NO.
1CC614 SOS 1 Task 07
11. CONTRACT/GRANT NO.
R804361
12. SPONSORING AGENCY NAME AND ADDRESS
Municipal Environmental Research LaboratoryCin., OH
Office of Research & Development
U.S. Environmental Protection Agency
Cincinnati, Ohio 45268
13. TYPE OF REPORT AND PERIOD COVERED
Final 3/76 - Q/78
14. SPONSORING AGENCY CODE
EPA/600/14
15. SUPPLEMENTARY NOTES
Project Officer: Gary S. Logsdon (513) 684-7345
16. ABSTRACT
Light scattering by individual particulates is used in a multiple-detector
system to categorize the composition of suspended solids in terms of broad
particulate categories. The scattering signatures of red clay and taconite
tailings, the two primary particulate contaminants in western Lake Superior,
along with two types of asbestiform fibers, amphibole, and chrysotile, were
studied in detail. A method was developed to predict the concentration of
asbestiform fibers in filtration plant samples for which electron microscope
analysis was done concurrently. Fiber levels as low as 5 x 10 fibers/liter were
optically detectable. The method offers a fast and inexpensive means for
measuring, either on a continuous basis or as discrete samples, the fiber levels
of filtration plant output. Further calibration of the instrument could enable
analysis for other specific particulate contaminants as well.
17.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b.IDENTIFIERS/OPEN ENDED TERMS C. COSATI Field/Group
amphiboles, asbestos, clays, colloids,
detectors, fibers, light (visible
radiation), optical detection,
particle shape, water
light scattering,
taconite tailings,
red clay, Lake Superior,
chrysotile particle
detector
20 F
18. DISTRIBUTION STATEMENT
RELEASE TO PUBLIC
19. SECURITY CLASS (ThisReport)
UNCLASSIFIED
21. NO. OF PAGES
71
20. SECURITY CLASS (This page)
UNCLASSIFIED
22. PRICE
EPA Form 2220-1 (9-73)
61
US GOVERNMENT PBINTUG OFFICE; 1979 -657-060/5447
------- |