EPA-R2-73-219
June 1973 Environmental Protection Technology Series
Feasibility Study of In-Situ
Source Monitoring
of Particulate Composition
by Raman or Fluorescence Scatter
Office of Research and Monitoring
U.S. Environmental Protection Agency
Washington, D.C. 20460
-------
EPA-R2-73-219
Feasibility Study of In-Situ
Source Monitoring
of Particulate Composition
by Raman or Fluorescence Scatter
by
M. L. Wright
Stanford Research Institute
333 Ravenswood Avenue
Menlo Park, California 94025
Contract No. 68-02-0594
Program Element No. 1A1010
EPA Project Officer: John S. Nader
Chemistry and Physics Laboratory
National Environmental Research Center
Research Triangle Park, North Carolina 27711
Prepared for
OFFICE OF RESEARCH AND MONITORING
'o.S. ENVIRONMENTAL PROTECTION AGENCY
WASHINGTON, D.C. 20460
June 1973
-------
This report has been reviewed by the Environmental Protection Agency and
approved for publication. Approval does not signify that the contents
necessarily reflect the views and policies of the Agency, nor does
mention of trade names or commercial products constitute endorsement
or recommendation for use.
11
-------
ABSTRACT
The purpose of this project was to assess the feasibility of in-
stack monitoring of an air-suspended particulate stream by fluorescence
or Raman optical interactions. The study explored the feasibility of
two approaches: quantitatively monitoring a prescribed constituent, and
monitoring the relative concentrations of several constituents simulta-
neously. Fluorescence-monitoring systems were found suitable for the
second.
The method of approach was to assess the magnitude of the Raman and
fluorescence interaction, and then calculate the detectability of that
material for a typical in-stack system. Thirty-four materials were
investigated on the project; thirteen materials had significant fluo-
rescent responses and twenty-two materials had measurable Raman responses.
When these responses were used to calculate in-stack detectability, all
thirteen materials could be detected by fluorescence systems (although
few could be uniquely identified), and fifteen of the twenty-two Raman-
active materials could be detected by a Raman system.
The use of a laboratory Raman instrument to analyze conventionally
sampled particulates was considered. The primary advantage of this
instrument appears to be the capability for measuring ions—for example,
sulfate.
Finally, a few crude experiments were made to detect the fluorescent
response of a particulate material suspended in a liquid (rather than
air). The-c- measurements showed substantial interference from fluores-
cence by the liquid medium; nevertheless, a component of the particulate
iii
-------
fluorescence was detectable. This experimental result partially verifies
the calculated feasibility of detection by fluorescence.
It is concluded that both fluorescence and Raman in-stack monitoring
systems can yield useful information about the quantity and composition
of a particulate stream. Recommendations are made for additional efforts
toward achieving an operational in-stack monitoring system.
IV
-------
CONTENTS
ABSTRACT iii
LIST OF ILLUSTRATIONS vii
LIST OF TABLES ix
I INTRODUCTION 1
II MATERIALS 3
III FLUORESCENT RESPONSE MEASUREMENTS 7
A. General 7
B. Measurement Results and Discussion 11
IV RAMAN RESPONSE MEASUREMENTS 21
A. General 21
B. Measurement Results and Discussion 27
V IN-STACK MONITORING SYSTEMS 41
A. General 41
B. Fluorescence-Monitoring Systems 42
C. Raman Monitoring Systems 48
1. General 48
2. Spectrometer-Type System 48
3. Filter-Type System 59
4. Performance Summary for Raman Systems 65
VI LABORATORY MEASUREMENT CAPABILITIES 71
A. General 71
B. Measurable Material Properties 71
C. Instrumental Considerations ; . 72
VII AEROSOL MEASUREMENTS 75
-------
VIII CONCLUSIONS 77
IX RECOMMENDATIONS 79
Appendix A MEASURED FLUORESCENT RESPONSE SPECTRA 81
Appendix B MEASURED RAMAN SPECTRA 91
REFERENCES 110
vi
-------
ILLUSTRATIONS
1 Schematic Diagram of a Fluorescence Measuring Instrument . 8
2 Optical Attenuation of Air 12
3 Typical Fluorescence Measurement Curve (for CaSO ) .... 13
4 Fluorescence Response Shift with Varying Excitation
Wavelength—for Phosphate Rock Feed Material 15
5 Fluorescence Response Shift with a Single Constituent
(for CaF ) 16
£
6 Minimal Fluorescence Response Shift — for Super Phosphate
Storage Product 17
7 Excitation and Fluorescence Response Wavelengths 19
8 Schematic Diagram of Raman Measurement Instrument 22
9 Raman Response Curve for (NH ) SO 28
10 Raman Response Curve for CaSO 29
11 Wavenumber Shift-to-vVavelength Shift Conversion Chart ... 30
12 Raman Spectral Response Summary 36
A-l Relative Response for Baird-Atomic Spectrofluorimeter
Source and Detector 83
A-2 Fluorescent Response of A1F 84
*J
A-3 Fluorescent Response of CuSO 84
A-4 Fluorescent Response of Cryolite 85
A-5 Fluorescent Response of Al (SO ) 85
2 43
A-6 Fluorescent Response of EPA Raw Alumina 86
A-7 Fluorescent Response of HgSO 86
A-8 Fluorescent Response of EPA Zinc Smelter Feed Material . . 87
A-9 Fluorescent Response of EPA Coal—Source, NBS 87
A-10 Fluorescent Response of EPA Phosphate Rock Feed Material . 88
A-ll Fluorescent Response of EPA copper Smelter Feed Material . 88
A-12 Fluorescent Response of EPA Fly Ash 89
vii
-------
A-13 Fluorescent Response of EPA Lead Smelter Feed Material . 89
A-14 Fluorescent Response of Particulate A1F in Water .... 90
O
B-l Raman Response of HgSO 92
B-2 Raman Response of PbSO 93
B-3 Raman Response of CdSO 94
B-4 Raman Response of Al (SO ) 95
2 4 3
B-5 Raman Response of Al (SO ) (6471 A) 96
2 4 3
B-6 Raman Response of HgCl 97
£
B-7 Raman Response of CdCl 98
^
B-8 Raman Response of Cud 99
£t
B-9 Raman Response of PbO 100
B-10 Raman Response of CdS 101
B-ll Raman Response of CaF 102
^
B-12 Raman Response of A1F 103
o
B-13 Raman Response of EPA Phosphate Rock Feed Material . . . 104
B-14 Raman Response of EPA Zinc Smelter Feed Material .... 105
B-15 Raman Response of EPA Triple Super-Phosphate Storage
Product (6471 $) 106
B-16 Raman Response of EPA Coal—Source, NBS (6471 A) .... 107
B-17 Raman Response of EPA Coal—Source, NBS 108
B-18 Raman Response of Napthalene 109
viii
-------
TABLES
1 Materials Analyzed for Optical-Monitoring Potential 5
2 Relative Fluorescence Intensities 14
3 Relative Raman Response Intensities 31
4 Comparison of Various Measurements of the Raman Cross
Section of Benzene (992*cm"1 Line) 39
5 Fluorescent-Reference-Material Measurement 44
6 Relative Raman Cross Sections for Gases 56
7 Raman In-Stack Monitoring-Instrument Performance Summary . . 66
8 Raman Shifts of Molecular Ions 73
IX
-------
I INTRODUCTION
Analytical techniques currently used for determining the composi-
tion of aerosols require some type of sampling. This sampling pro-
cedure may change the characteristics of the aerosol in such a way that
the measured sample properties are not representative of the aerosc
properties in the stack. An in-situ technique, particularly one uti-
lizing optical methods, would avoid disturbing the aerosol in the ack
and would give an indication of the true characteristics of the ae~osol
in the stack.
The two most promising optical-material interactions are the well-
known Raman and fluorescence scattering properties of materials. Raman
scattering from materials is often proposed as a method of analysis
because of the relatively narrow and distinct Raman peaks obtained in
the spectra of many materials. However, this specificity advantage is
offset by the weak nature of the Raman interaction. In many cases too
few photons are scattered to permit measurements to be made using Raman
scatter. Fluorescence, on the other hand, offers a much stronger
optical interaction than the Raman scatter. This advantage is partially
offset by the broad nature of the spectral excitation and resrjnse
characteristics of most fluorescent materials, particularly in the
solid state. Discrimination between several materials is much more
difficult with fluorescence than it is with Raman scatter, due to the
broad, diffuse nature of these fluorescence spectral responses. Thus,
both effects present some difficulties, which in some cases will pre-
vent realistic field measurements.
-------
This project was concerned with the analysis of the feasibility
of in-situ stack monitoring using either Raman or fluorescence scatter-
ing properties. The analysis considered monitoring from two points of
view: (1) the feasibility of quantitatively monitoring a prescribed
molecular constituent, and (2) the feasibility of qualitatively monitor-
ing the relative concentrations of several molecular constituents simul-
taneously. The relatively strong but nonspecific nature of the
fluorescent response makes it suitable for quantitatively monitoring
a prescribed molecular constituent. Qualitative monitoring of several
constituents is best accomplished through use of the Raman effect.
Both methods were found to be feasible for several materials.
The general approach taken in this project was, first, to determine
by experimental measurements whether a specific material had a signifi-
cant Raman or fluorescence response. If a significant response was
found, an estimate of the detectability of that material was made for
typical monitoring system configurations.
The key to the successful application of either monitoring approach
is the existence of a sufficiently large optical interaction. It should
be noted that not all materials possess a significant Raman or Fluo-
rescent response characteristic. This is particularly true for glassy
materials such as fly ash. This material was of particular interest to
both EPA and SRI; however, measurements of the fluorescent and Raman
spectra of fly ash were made, and no Raman or fluorescent response was
found for any of the fly-ash samples. Thus, fly ash appears to be a
distinctly unpromising material for in-stack analysis by Raman or in-
situ fluorescence techniques. Other materials were found to be much
more promising and are discussed in later sections of the report.
-------
II MATERIALS
The contract specifies that ten materials would be selected jointly
by SRI and EPA personnel for optical characterization. The procedure
finally adopted was for EPA to specify broad guidelines detailing the
rationale for particulate monitoring and indicating specific industrial
processes whose pollutants were believed to be of greatest concern at
the present time. Within these guidelines, SRI was directed to select
the specific materials that would be examined. The materials selected
were primarily simple chemical compounds that were easily prepared for
optical analysis; thus it was possible to examine more than the ten
materials specified in the contract, and a total of 34 materials were
examined. This total does not include 14 samples of stack emission
products that were obtained well after the technical work stopped and
thus were not completely evaluated. The latter samples were supplied
through the courtesy of Dr. Milton Feldstein of the Bay Area Air
Pollution Control District (BAAPCD).
The EPA guideline stated that the particulates of primary interest
in this project should be those with chemical-related health effects,
and that other particulate material should be of only secondary interest.
The materials of interest were grouped into several categories, again
by EPA direction. These groupings are (1) fly ash from coal-burning
power plants; (2) calcium and ammonium sulphates from heating sources;
(3) lead, copper, cadmium, and mercury compounds from smelters, incin-
erators, alloy plants, and steel mills; (4) fluorides from aluminum
reduction plants or phosphorous plants; (5) other particulates of gen-
eral interest, including sulphates, chlorides, oxides, and sulphides of
-------
lead, cadmium, zinc, mercury, copper, nickel, chromium, vanadium, arsenic,
and beryllium; and (6) a representative organic material.
The specific materials that were examined on this project are listed
in Table 1.
The organic material originally selected was the particulate emis-
sion from an asphalt batch plant. Difficulty in obtaining such a sample
resulted in the substitution of the NBS coal sample for the originally
chosen asphalt batch plant sample. Both of these materials probably
contain a complex mixture of many organic compounds, and the Raman and
fluorescence measurements were not expected to yield constituent data
on either material. Time and funds did not permit the examination of
simpler organic pollutant materials on this project, although they offer
significant potential for either Raman or fluorescence monitoring.
-------
Table 1
MATERIALS ANALYZED FOR OPTICAL-MONITORING POTENTIAL
Materials supplied by SRI
Calcium fluoride (CaF )
£i
Mercury sulfate (HgSO )
Aluminum fluoride (AlF )
Aluminum sulfate [Al (SO ) • 18H 0]
jL T *J £
Cupric sulfate (CuSO • 5H O)
Calcium sulfate (CaSO • 2H 0)
Cryolite (Na AlF)
3 6
Water (H 0)
Cupric chloride (CuCl • 2H 0)
2 2
Lead chloride (PbCl )
Mercuric oxide (HgO)
Cupric oxide (CuO)
Cuprous oxide (Cu 0)
Mercuric sulfide (HgS)
Sulfuric acid (H SO )
Lead sulfate (PbSO )
4
Lead sulfide (PbS)
Lead oxide, mono (PbO)
Cadmium sulfate (SCdSO • 8H 0)
4 £
Cadmium oxide (CdO)
Aluminum oxide (alumina) (Al O )
Ammonium sulfate [(NH ) SO ]
4 2 ^
Fly ash (4 samples)
Materials supplied by EPA
Triple superphosphate storage product
Phosphate rock feed material
Raw alumina
Coal (source, NBS)
Zinc smelter feed material
Copper smelter feed material
Lead smelter feed material
Fly ash
-------
Ill FLUORESCENCE RESPONSE MEASUREMENTS
A. General
The two main requirements of the fluorescence measurement program
were, first, to be able to detect fluorescent response amplitudes down
to a level comparable to that of the Raman response peak of water, and,
second, to detect fluorescent responses over the entire wavelength
range of 200 to 700 nm. These requirements led to the selection of two
separate fluorescence measurement systems for complete wavelength cover-
age. One of these instruments was modified to obtain a substantial
improvement in performance and was also calibrated to provide more
quantitative information for comparison of the response levels of
various materials.
The spectrofluorimeter on which the majority of the sample measure-
ments were made was a Fluorispec, Model SF-1, made by Baird Atomic, Inc.
This instrument covers the wavelength range 220 to 700 nm. It was
chosen primarily because it utilizes double monochrometers for both
the source and detector, thus giving superior scattered-light rejection
for fluorescence measurements. A diagram of this instrument system is
shown in Fig. 1. Most routine laboratory fluorescence measurements are
made on transparent or semitransparent solutions in which the scattering
of the source (or exciting) light is relatively low. Solid samples were
used on this project, many of which are white powders reflecting large
amounts of source light scattered to the detector. Thus, it is particu-
larly important, when measuring either solid samples or aerosols, to
achieve a high degree of scattered-light rejection.
-------
SOURCE DOUBLE
MONOCHROMETER
/
./
XENON
LAMP
\
DETECTOR DOUBLE
MONOCHROMETER
//
SAMPLE
HOLDER
\
V
N
\
PMT
CURRENT
METER
PHOTOMULTIPLIER
DETECTOR
SA-2039-2
FIGURE 1 SCHEMATIC DIAGRAM OF A FLUORESCENCE MEASURING INSTRUMENT
-------
Several modifications of this instrument were made to improve the
sensitivity, accuracy, and repeatability of the fluorescence measurements
First, the exciter lamp in the source unit was replaced in order to
improve the ultraviolet-light output. This new lamp increased the UV
light ^utput by a factor of approximately five. The UV output of these
lamps degrades faster than does the visible performance, however, so
this degree of improvement was not maintained over the entire measure-
ment program. Second, a large portion of the original instrument elec-
tronics was bypassed in order to increase the stability and repeatability
of the measurements and to lower the effective noise level below that
provided by the original instrument. These instrument modifications
included replacement of the photomultiplier power supply with a well
regulated Power Designs commercial power supply, and the direct monitor-
ing of the photomultiplier current by a sensitive Hewlett-Packard 425A
current meter. Joine portions of the passive photomultiplier circuitry
have been retained as in the original instrument, but all of the active
portions of the photomultiplier circuitry have been replaced. In addi-
tion, the original 1P21 has been replaced by a 1P28 photomultiplier
to 1'uj ther improve the UV performance of the instrument. The effect
of these modifications has been monitored by observing the Raman peak
oi water and has resulted in a clearly enhanced signal-to-noise ratio
(SNRJ or the Raman response. In addition, the amplitude response was
stabilized and is available as an absolute current level that can in
turn be related to an absolute light level at the detector. These modi-
ficationb allow a quantitative comparison of response levels to be made
over the f .ill dynamic range of the photomultiplier.
Both the source and the detector portions of the Baird-Atomic
spectrofluorimeter were calibrated in order to allow accurate amplitude
correction to be made on the measured fluorescent responses. These
-------
relative-amplitude calibrations can be partially converted to absolute
1 2*
calibrations through the use of materials with known quantum yields. '
This type of absolute calibration is still of limited utility in the
present program, however, because such measurements are ordinarily done
in liquid samples with more easily defined geometries. The use of solid
samples, as in the present program, does not offer the possibility of
easily controlled geometries; this factor will contribute substantially
to errors in determining absolute cross sections for the materials, even
though the system has been calibrated with materials of known quantum
yields.
The second measurement system consists of an Aminco-Bowman Spectro-
photofluorimeter. This machine covers the wavelength range 200 to 800 nm,
although it was utilized only for measurements in the source range 200
to 220 nm because of the superior performance of the Baird-Atomic
instrument at other wavelengths. Modifications of this machine were
also planned. However, the scattered-light response of this single-
monochromator instrument was so large that the only modification made
was a replacement of the photomultipiier tube to enhance the UV perform-
ance of the instrument. The performance of this instrument, even with
this scattered-light limitation, was still judged adequate to determine
the presence of any responses that would be useful for identification in
in-stack measurement systems.
No measurements were made at wavelengths shorter than 200 nm because
of the abrupt increase in the attenuation of air at these shorter wave-
lengths. This attenuation will affect both the measurements and the
operation of the final system. The attenuation curve for "standard
*References are listed at the end of the report.
10
-------
atmosphere" air is shown in Fig. 2. For special applications, opera-
tion is possible in the "notches" of the rising curve, but in general
200 nm is considered to be the transmission limit.
The procedure used for the measurement of each material involved
a manual scan of wavelengths to locate the positions of any significant
fluorescent responses. When definite response-peak locations could be
determined manui ~.ly, only those regions of greatest significance (i.e..
peaks within a iactor of 10 of the largest peak) were run quantitatively
on the system. Materials without an easily discerned maximum response
level were examined at a variety of wavelengths for low-level fluores-
cence responses. A typical example of an actual response run is shown
in Fig. 3 for calcium sulfate (CaSo ). This figure shows the two types
of response curves often used to characterize fluorescent materials.
The curve on the left side of the figure shows the response of the
material to a varying source (or excitation) wavelength with the detector
fixed in wavelength at the maximum fluorescent response position (440 nm
in this example). The sharp rise at the long-wavelength end of the
excitation curve is caused by the response to scattered source light as
the source and detector wavelengths become close together. The right-
hand curve shows the response of the material for a varying detector
wavelength with the excitation wavelength at the peak of the excitation
response curve (369 nm in this example). This pair of curves for each
material indicates the maximum amplitude response chat would be avail-
able for optimum source and detector wavelengths.
B. Measurement Results and Discussion
The results of the measurements ipade on. the Baird-Atomic system
are shown in Table 2. This table shows the wavelength location of the
excitation and ilvor^r-- - :;: ks and indicates the normalized relative
amplitude of the £ic.c.••O5c_uc«; peak for each material. The response level
11
-------
10a
to
10°
.E
to
2 10°
D
Z ,
ai in"'
10
,-2
10
,-3
10
,-4
10
,-5
170 175 180 185 190 195 200
I I I I
205 210 215 220
WAVELENGTH — nm
225 230 235 240 245 250 255
SA-2039-3
FIGURE 2 OPTICAL ATTENUATION OF AIR
-------
0.3
U!
W
O 0.2
CL
-------
Table 2
RELATIVE FLUORESCENCE INTENSITIES
1
Material
CaFo
2
EPA phosphate
rock feed sample
HgS04
EPA triple
phosphate sample
EPA raw alumina sample
A1F
3
A12(S04)3
CuSO
4
CaSO
4
Cryolite
-Water < Raman)
EPA Zn sample
EPA coal sample
CuCl
2
PbCl
2
HgO
CuO
Cu 0
2
HgS
H SO
2 4
PbSO
4
PbS
PbO
CdO
CdSO
4
Al O
2 3
(NH4)2S04
EPA fly ash sample
EPA copper sample
EPA lead sample
SRI fly ash (4 samples)
Peak Wavelength
(nm)
Excitation
372
392
309
364
347 "
372
364
341
369
375
340
349
365
380
None
None
None
None
None
None
None
None
None
None
None
None
None
None
None
None
None
Fluorescence
452
470
564
447
Normalized Peak
Signal Level
(jiA)
8.23
4.43
1.51
1.31
!
420
452
430
388
440
458
384
467
443
438
Peak
Photon Count
(photons/s)
6
2.96 X 10
1.6 X 10
5
5.44 X 10
4.72 X 10
5
1.18 4.25 X 10
0.345
0.32
0.120
0.111
0.109
0.063
0.038
0.032
0.0103
5
1.24 X 10
5
1.15 X 10
4
4.32 X 10
4
4.0 X 10
4
3.92 X 10
4
1.36 X 10
4
1.15 X 10
3
3.7 X 10
14
-------
The two mercury compounds showed slightly larger responses than the two
cadmium compounds; however, all four materials had relatively weak
responses and were not much larger than the stray-light response of
the instrument. Although no quantitative calibrations were made between
the two instruments, an estimate of the relative response would be
comparable to aluminum sulfate in Table 2.
It should be noted that the shape and peak location of the excita-
tion and fluorescence curves may-shift with operation at other than
these two optimal wavelengths shown above. This shift in peak location
and shape of the response curve is illustrated in Fig. 4 for the EPA
phosphate rock feed sample. This figure shows the change in fluorescence
response curve as the excitation wavelength is shifted from 325 nm to
475 nm. Note that the amplitude of the maximum value of each fluorescent
curve is different, as is the wavelength at which this maximum occurs.
Note also that the width of the curve varies with the excitation wave-
length and becomes narrower as the excitation wavelength becomes longer.
10
HI
to
i*
lil
O
Z
01
O
V)
111
£E
O
I
EX 475
EX 325
200
300
400 500
WAVELENGTH — nm
600
700
SA-2039-5
FIGURE 4 FLUORESCENCE RESPONSE SHIFT WITH VARYING EXCITATION WAVELENGTH
FOR PHOSPHATE ROCK FEED MATERIAL
15
-------
Most of the materials investigated in this project have a changing
response characteristic similar to that shown in Fig. 4. This changing
response curve is often associated with the presence of more than one
constituent in the sample. This could indeed be the case for the
phosphate rock feed material; however, it is also possible for this
effect to occur in single constituent materials as well. An example
of this effect in a single material is shown in Fig. 5 for reagent-
grade CaF . Some materials do not show a pronounced shift in the
£t
fluorescence response curve with varying excitation. For these mater-
ials the peak location and general shape of the fluorescence curve
remains approximately the same with varying excitation wavelengths.
The overall fluorescence curve moves up and down in amplitude, uniformly
with varying excitation wavelengths. No material examined on this
project was completely free of fluorescence curve shift. An example
of a material with minimum curve-shape change is shown in Fig. 6 for
the EPA triple super-phosphate storage product.
u.
200
300
400 500
WAVELENGTH — nm
600
700
SA-2039-6
FIGURE 5 FLUORESCENCE RESPONSE SHIFT WITH A SINGLE CONSTITUENT
(for CaF2)
16
-------
200
300
400 500
WAVELENGTH — nm
600 700
SA-2039-7
FIGURE 6 MINIMAL FLUORESCENCE RESPONSE SHIFT — FOR SUPER-PH^PHATE
STORAGE PRODUCT
This characteristic of changing emission-peak location will compli-
cate simultaneous quantitative measurements of several materials by
fluorescence techniques. Consider, for example, a monitoring system
in which several detectors, each sensitive to a different wavelength,
examine sequentially the fluorescence response of a mix of materials
excited by different exciting wavelengths. If each individual material
produces a fixed ratio of outputs in the detector array for each excita-
tion wavelength, then that ratio matrix can be considered characteristic
for that material. If, on the other hand, the ratios for each detector
differ with different exciting wavelengths, then the characteristic
matrix for each material is two-dimensional, thus significantly compli-
cating the tauk of determining the relative composition of the unknown
materials.
The fluorescence response curves for most of the materials examined
in this project show rather broad, smooth curves with little fine
17
-------
structure. Also, most of the peak locations in wavelength are approxi-
mately the same for the materials examined. A summary chart showing
the peak location (letter E or F) and half-amplitude width (solid dots)
for each of the materials as a function of wavelength is shown in Fig. 7.
It is evident from this figure that considerable response overlap
exists between materials and would make the simultaneous identification
of many of these materials relatively difficult. Note that CaF and
^
A1F have virtually identical wavelength responses. A few materials,
o
such as HgSO and CuCl , are separated more widely from the others and
4 £*
could be distinguished on the basis of simple fluorescence measurements.
It is evident from these data that the fluorescence measurement tech-
nique is not sufficiently selective to be a general chemical analysis
method, but it can be useful in specific situations where the list of
interfering materials is known and where the spectra are easily separable.
The measured response curves for materials with significant fluo-
rescence are shown in Appendix A. These are uncorrected curves, taken
directly from the recorder traces. The calibration curve for both the
source and detector of the Baird-Atomic Spectrofluorimeter is also given
in Appendix A, so that any portion of these curves could be corrected
for instrument response if desired.
18
-------
COAL-
EPA Zn
Ca F,
Al F,
AI203
Cu CU
Ca SO.
Cu SO,
PHOSPHATE ROCK
TRIPLE PHOSPHATE
Na3 Al
HgSO.
E.
E-
-E
_F
F-
•E-
-E-
--F-
(SO4)
,_
F
I
200
300
400 500
WAVELENGTH — nm
600
SA-2039-8
FIGURE 7 EXCITATION AND FLUORESCENCE RESPONSE WAVELENGTHS
-------
IV RAMAN RESPONSE MEASUREMENTS
A. General
The Raman measurements were made on a Spex Ramalog Raman spectro-
meter. This instrument consists of a Coherent Radiation Model 52
argon-krypton laser source and a Spex Model 1401 double monochromator
as the detection optical filter. The optical detector is an ITT FW130
o
photomultiplier that is thermoelectrically cooled to -30 C. A schematic
diagram of this instrument is shown in Fig. 8. Two types of detection
electronic systems are used--a dc system and a photon-counting system.
The dc system monitors the photocurrent from the photomultiplier, with
an adjustable electronic averaging time, and presents this information
as a function of wavenumber on a strip-chart recorder. The photon-
counting system digitally accumulates the photon count for a given time
interval and again presents this information as a function of wave-
number on a strip-chart recorder. The photon-counting system provides
better performance at extremely low light levels and is the preferred
type for use with extremely weak signal levels. Both systems are adequate
in sensitivity for determining Raman responses that will be usable in in-
stack monitoring applications. The reason for this is that reasonably
large Raman responses must be present in order to be usable with practical
field instrumentation. It is evident that difficulties will be encountered
in field measurements if, in the laboratory, many hours of integration
are necessary to detect a material with a laboratory instrument whose
optical geometry can be optimized for maximum signal on a fixed solid
sample.
21
-------
ARGON-KRYPTON
LASER
BEAM
SPLITTER
to
to
SAMPLE
HOLDER
OPTICAL 1
FILTER .
DC CURRENT
METER AND
RECORDER
COOLED
PHOTOMULTIPLIER( f
DETECTOR
POWER
MONITOR
< /
SPEX 1401 DOUBLE
MONOCHROMETER
T
OR
1
PHOTON COUNTER
AND
RECORDER
SA-2039-9
FIGURE 8 SCHEMATIC DIAGRAM OF RAMAN MEASUREMENT INSTRUMENT
-------
The most important information needed in evaluating the feasibility
of remote monitoring of various species by laser Raman spectroscopy is
the intensity of the Raman scattered light, which is given by
total
I . I Ma
P op
where I is the incident laser light intensity, I is the Raman scattered
o p
intensity, M is the amount of material present, and a is the total
P
molar Raman scattering cross section. This Raman cross section is the
pertinent material property for determining the detectability of a
given material and is similar to the quantum efficiency associated with
the fluorescence process.
Before laser excitation was available, the stability, intensity,
and other characteristics of the sources were among the major impedi-
ments to the measurements of absolute cross sections in Raman scatter-
ing, which is very weak. Since the advent of the laser, and with various
improvements and refinements in analysis and detection techniques, the
problem of absolute-cross-section measurements has become somewhat more
tractable. However, such measurements have been made only in a small
3-5
number of cases due to the considerable difficulties still present.
A great deal of thought has been given to this problem, which needs
resolution before many areas of application can proceed to operational
systems.
The sample properties and instrumental factors that affect the
observed intensity include the following:
(1) Refractive index. This will affect:
(a) The intensity of the exciting radiation reaching
the interior of the sample.
23
-------
(b) The degree of convergence (relative intensities
of perpendicular and parallel polarized components
of exciting light as they appear to the scattering
molecule).
(c) The fraction of the Raman radiation that can enter
and pass through the monochromator.
(2) Molecular environment (caused by changes in sample composition)
(3) Fluorescence.
(4) Color.
(5) Intensity of the Raman source.
(6) Variations in the relative intensities of the parallel and
perpendicular polarized components of the exciting radiation.
(7) Properties of the sample cell, if used, and the sampling
geometry.
(8) Spectral sensitivity of the spectrophotometer.
(9) Polarization sensitivity of the spectrophotometer.
(10) Background emission. This may arise from scattering of fluo-
rescence at the cell walls, from scattering of the continuum
in the spectrum of the exciting radiation, or from stray light
inside the monochromator.
(11) Instrument sensitivity (relation between the intensity of the
radiation entering the monochromator and the recorder response)
The following methods are used to eliminate or correct for some of the
above factors:
(1) Use of an internal standard, which eliminates the effects of
variations in refractive index, as well as such instrumental
factors as the intensity of the source, cell properties, and
instrument sensitivity. This will not compensate for the
effect of color except in gray samples.
(2) Use of a polarized laser beam, which essentially eliminates
the effect of convergence and the partial polarization of
the exciting light.
(3) Measurement of band intensities above a linear baseline drawn
between preselected, fixed spectral positions on each side of
the band, which largely corrects for background emission and
24
-------
for a limited amount of sample fluorescence. (The intensity
of the fluorescence normally changes slowly with the frequency
with no sharp bands. However, too much fluorescence will
unduly influence the signal-to-noise ratio for the Raman
spectrum.)
(4) Use of a polarization scrambler placed at the entrance to the
monochrnpi^^r- *•-• • eliminates the polarization sensitivity
of the s^cc—.... *er.
(5) Periodic checks for changes in spectral sensitivity, which
may be made by substituting a light source of reproducible
relative spectral emissivity for the sample cell and record-
ing the intensity over the spectral region of interest. Any
changes in the observed ratios of the intensities can be used
to correct the observed sample intensities.
Because of the attendant difficulties mentioned above in the
measurement of Raman intensities, determination of absolute Raman cross
sections is a major undertaking. For the purposes of the present feasi-
bility study, it was deemed adequate to make relative-intensity measure-
ments of the Raman lines for the various solid samples. The cross-
section standard was a material (CdS) of known cross section.
Experimentation with several sampling techniques led to the adoption
of the following sampling method. First, the chemicals were packed under
pressure into the end of a 1/8-inch-diameter stainless steel rod into
which a depression had been machined. The powder was given a relatively
flat SU;face by pressing down on another flat surface. The stainless
steel rod was then mounted in a holder and arranged so that the surface
o
of the sample made an angle of approximately 45 to the vertical laser
beam, which was incident from below the sample. The holder was capable
of sufficient three-dimensional motion such that 'ie laser focal spot
could always be imagined centrally on the spectrurrter slit. If the
particle size of the sample was so coarse that sample fell out of
the end of the rod when turned upside down, the sample was ground in a
mortar to a finer particle size. No attempt was made to determine the
particle size of the samples.
25
-------
The spectra were initially recorded at a spectral slit width of
10 cm and at a scan rate of 1000 cm per minute. If interesting
features appeared, the particular spectral ranges were reexamined at
higher resolution. All the white samples were examined with 4880-A
excitation. The colored samples, including green cupric chloride,
absorbed too much energy from the focused laser beam and were altered,
as evidenced by discoloration at the position of the focal spot. Some
of the hydrated samples appeared to lose their water of hydration at
high power densities. Such samples were run with 6471-A excitation or
at reduced power (or both), as dictated by sample integrity and adequate
SNR.
The ideal conditions and techniques that are desirable for Raman
cross-section measurements were noted earlier. During the present in-
vestigations, several compromises had to be made, based on the available
time, the large number of materials, and the expected eventual use of
the data. For example, the spectral and polarization sensitivity of
the spectrometer was not calibrated. Also, while every effort was made
to achieve identical scattering geometries with the different samples,
the effects of particle size, refractive index, color, and absorptivity
were ignored. Because the effects of particle size were ignored in
these measurements, no attempts were made at maintaining uniformity of
particle size. All sample runs were performed after the laser power
had stabilized. The laser power output, however, was monitored only
occasionally, and small variations may still have occurred from run to
run. Finally, because of the large variation in color, absorption,
particle size, and uniformity, no attempt was made to use an internal
reference standard. The net effect of all these compromises was believed
not to lead to large errors in the measured results. Based on the
26
-------
experience with repeated measurements of different samples and pro-
cedures, it is estimated that the results obtained are accurate to
approximately ± 50% for the white samples.
B. Measurement Results and Discussion
Typical Raman response curves are shown in Figs. 9 and 10 for
(NH ) SO and CaSO . Figure 9 is an example of a Raman curve with
4^4 4
relatively low background and no fluorescence interference. A clean,
strong Raman line appears at about 900 cm , and smaller lines are
shown. Figure 10 shows the effect of fluorescence interference, appear-
ing as a broad, smooth curve over the entire chart. The Raman lines
extend above this interference.
The Raman response cr.rves are plotted as a function of wavenumber
shift. A conversion to wavelength shift can be made by reference to
Fig. 11, which gives wavelength shift versus wavenumber shift at several
laser excitation wavelengths.
A summary of the Raman measurements is given in Table 3, listing
the material, the wavenumber shift for all significant lines, and the
measured signal and background levels. Also given in Table 3 is the
cross section for each line, which will be discussed in a later section.
For the sake of uniformity, the relative intensities in Table 3 are
presented as counts per second normalized to a 10 cm spectral slit
width with 4880-A excitation at 350 mW. This normalization removes the
effects of different power levels, excitation wavelengths, and detection
systems. In order to use this normalization procedure, linearity of
the various devices was assumed. Also, the throughput of the system
was assumed to be equal at 4880 A and 6471 A, and the scattered intensity
-4
was assumed to have a \ dependence. This latter assumption is valid
for the Raman intensity and the scattered background. However, it is
not valid if the background was due to fluorescence or to Mie (particle)
27
-------
CO
IU
f-
Z
LU
I I Fill
•0.3 x 10~6 A-
T I T I
' ' L
4800 4400 4000 3600 3200 2800 2400 2000 1600 1200 800 400 0
WAVENUMBERS — cm
-1
SA-2039-10
FIGURE 9 RAMAN RESPONSE CURVE FOR (NHd) SO
<* *
28
-------
I-
LLI
to
CO
4800 4400 4000 3600 3200 2800 2400 2000 1600 1200 800
I x 10~6 A
WAVENUMBERS — cm
SA-2039-11
FIGURE 10 RAMAN RESPONSE CURVE FOR CaSO,
-------
10
103
WAVELENGTH SHIFT—A
SA-2039-12
FIGURE 11 WAVENUMBER SHIFT-TO-WAVELENGTH SHIFT CONVERSION CHART
30
-------
Table 3
RELATIVE RAMAN RESPONSE INTENSITIES
Material
PbO
PbCl
2
HgS04
HgCl
2
RA
PbSO
4
(NH4)2S04
CuCl
2
Excitation
Wavelength
6471
4880
4880
4880
4880
6471
4880
4880
4880
Wavelength
Shift
60
69
86
138
279
370
60
86
156
93
121
230
300
410
495
580)
>
588)
660
987
1043
1125
1180
1342
1670
70
121
312
380
420
750
1250
720
1022
1052
443
608
640
978
1065
1165
450
613 )
623)
976
1090
1420
1650
3150
64
109
215
236 )
\
249 )
407
700
Peak Photon
Count Rate
6
7.8 X 10,
7
1:3 x 10
7
5.4 X 10
7
1.4 X 10
6
1.9 X 10
5
8.4 X 10
7
1.5 X 10
5
5.0 X 10,
7
1.5 X 10
6
2.1 X 10
6
2.8 X 10_
6
8.8 X 10,.
6
9.6 X 10_
6
1.5 X 10,.
6
2.3 X 10
5
2.5 X 10
4
6.3 X 10
7
1.4 X 10
6
7.9 X 10,
7
2.4 X 10
5
6.7 X 10
3
4.2 X 10
3
2.9 X 10,,
6
1.9 X 10.
o
5.4 X 10.
6
8.1 X 10
5
9.2 X 10,,
5
1.8 X 10,
4
7.5 X 10.
6
3.1 X 10.
5
1.7 X 10
5
2.9 X 10
5
2.3 X 10
1.5 X 105
6
1.1 X 10,,
4
4.6 X 10,
4
3.1 X 10
4
2.5 X 10_
4
9.6 X 10
6
1.1 X 10
5
3.4 X 10
5
1.6 X 10,
4
4.2 X 10
Background Photon
Count Rate
6
4.3 X 10
6
4.3 X 10
g
3.0 X 10
6
1.7 X 10
6
1.0 X 10
6
7.9 X 10
6
4.1 X 10
6
1.8 X 10,,
6
1.5 X 10
6
1.3 X 10
6
1.3 X 10°
6
1.3 X 10
6
1.3 X 10
6
1.3 X 10,.
6
1.4 X 10
6
1.5 X 10,.
6
1.7 X 10
6
5.4 X 10,.
6
2.3 X 10
5
2.1 X 10
5
1.3 X 10
4
6.7 X 10
4
5.0 X 10
6
5.4 X 10
4.9 X 10e
6
4.9 X 10
5
1.3 X 10
4
4.2 X 10
4
3.3 X 10^
4
1.3 X 10
4
1.3 X 10
4
1.3 X 10
4
4.6 X 10
1.9 X 104
3
6.3 X 10,
3
6.3 X 10
3
6.3 X 10
3
6.3 X 10
0 ?
5
3.2 X 10
5
2.5 X 10
5
1.1 X 10
4
4.9 X 10
Effective Raman
Cross Section
797
1330
5520
1430
194
174
3110
100
3000
420
560
1760
1920
300
460
50
12.6
3050
1720
5230
146
196
38.4
16
662
36.3
61.9
75
48.9
359
15
10.1
8.15
31.3
322
99.6
46.9
12.3
Fluorescence
Response
Maximum
Response
Location
2400
Maximum
Photon
Count Rate
5
1.9 X 10
31
-------
Table 3 (continued)
Material
CdCl2
CdSO
HgO
HgS
CaSO
CdS
Excitation
Wavelength
4880
4880
6471
6471
4880
6471
Wavelength
Shift
80
85
115
158
218
320
1585
3470
160
185
255
280
315
415
450
497
6031
615)
658V
670)
835
925
1000
10501
1063)
1100 I
1118)
11681
1173)
1554
2700-3500
25
35
65
130
328
550
250
275
341
425
490
620
670
1015
1150
212
305
347 1
365 )
563
599
Peak Photon
Count Rate
7.5 X 10^
8.4 X 10
8.4 X 10
1.8 X 10
7.1 X 104
2.5 X 10
9.6 X 10
6.3 X 104
1.9 X 104
7.1 X 105
9.6 X 104
5.4 X 104
5.9 X 104
2.5 X 10
6.3 X 10
7.5 X 10
3.4 X 10."
3.2 X 10
7.1 X 10
2.1 X 104
3.8 X 10
3.3 X 10
2.3 X 10
2.5 X 10
1.7 X 10
1.2 X 104
7.0 X 10
4
6.2 X 10
1.9 X 104
S.9 X 10
Background Photon
Count Rate
8.4 X 10g
3.1 X 10
1.2 X 10
5.4 X 10
7.9 X 10
6.7 X 104
5.4 X 10
2.9 X 10
2.6 X 10
1.6 X 10
1.8 X 104
1.6 X 10
4
1.6 X 10
1.3 X 104
1.6 X 104
7.5 X 10
3.4 X 104
3.0 X 10
1.7 X 10
2.1 X 10
1.8 X 10
1.5 X 10
1.5 X 10
1.6 X 10
1.8 X 10
5
2.6 X 10
i.S X 10
1.3 X 105
6.2 X 104
5.9 X 10
Effective Raman
Cross Section
225
25.2
2.52
53.9
25.7
9.06
34.8
22.8
6.88
257
34.8
19.6
21.4
9.06
53.7
6.39
44.7
4.2
9.33
5.32
9.63
8.36
5.83
63.4
4.31
15.7
9.16
8.63
8.11
2.49
7.72
Fluorescence
Response
Maximum
Response
Location
2400
Maximum
Photon
Count Rate
2.4 X 105
32
-------
Table 3 (concluded)
Material
A12(S04)
o
Phosphate
Rock Feed
A12°3
CaF2
"V1F6
Lead
Concentrate
NBS Coal
PbS
CdO
CU2°
CuS
A1F3
CuO
Zinc
Concentrate
Copper
Concentrate
Super
Phosphate
B47
Fly Ash
Excitation
Wavelength
4880
6471
6471
4880
4880
6471
4880
6471
4880
6471
4880
6471
6471
6471
6471
6471
6471
4880
6471
6471
4880
6471
6471
6471
Wavelength
Shift
480
620
1000-1200
480
620
1000-1200
945
580
378
415
585
643
750
None
322
1040
550
950
1350
1600
None
None
None
None
None
None
None
None
None
None
None
None
None
Peak Photon
Count Rate
2.5 X 104
1.9 X 10
3.8 X 10
7.1 X 10
4.5 X 10
1.4 X 10
1.7 X 104
4.2 X 10
8.4 X 103
1.3 X 10
1.3 X 104
7.8 X 10
8.8 X 10
3.8 X 102
375
750
Background Photon
Count Rate
2.0 X 104
1.9 X 10
4.2 x 10
6.6 X 104
5.3 x 10
3.7 X 10
7.1 X 104
7.5 X 10
1.6 X 10
1.4 X 10
1.1 X 106
3.1 X 104
3.3 X 10
4.2 X 102
9.4 X 103
1.0 X 10
Effective Raman
Cross Section
13.8
10.5
20.9
1.01
1.56
1.39
2.46
2.78
Fluorescence
Response
Maximum
Response
Location
2000
300
r.ooo
800
2200
2100
1400
500
2700
>6000
400
1100
Maximum
Photon
Count Rate
5.0 X 10
6
2.5 X 10
1.2 X 10
1.3 X 105
5.0 X 10
1.1 X 10
1.2 X 104
4.8 X 106
6.7 X 10
5.4 X 104
4.0 X 10
5.4 X 103
33
-------
scatter. The values of relative intensities are estimated to be accurate
to ± 50% in the case of the white samples. In the case of the colored
samples, the values are less reliable due to absorption effects.
Both the background and line intensities were determined at the
position of the peaks. The background was determined by drawing a
smooth line through the base of the peaks. The line intensity was then
measured as the peak height above this background. It should be noted
that the spectrometer slit of 10 cm does not significantly influence
the relative intensity of the background and broad Raman lines (broader
than 20 cm , say). The relative intensity of the narrower lines and
the background will, however, depend on the slit width. These narrower
lines are usually located at small wavenumber shifts from the exciting
line and would not normally be used in field systems because of the
stray-light-rejection problem. For this reason, the narrow lines have
not been evaluated, although they could be, from the measured data.
It was not feasible to examine the region close to the exciting
line (0 to 150 cm ) under a uniform set of conditions for all the
samples, since the variation in scattered intensity in this region was
very large for different samples. While several lines appear in this
region, the available time and the eventual use of the data led to the
decision not to examine them in detail. Scattered source light in this
region was believed to be too sensitive a function of particle size.
The level of the background from scattered exciting radiation within
100 cm~ (or a spectral separation of 24 A at 4880 A) of the (unshifted)
rayleigh line was felt to be too high to permit use of this region in
any nondispersive system. The need for discrimination between the
several substances that have lines in this narrow region is a further
nontrivial problem.
34
-------
The absolute values of the listed wavenumber shifts are estimated
_]_
to be within ±5 cm . The linewidths shown are believed to be accurate
±20%, without deconvolution of the instrument functions.
Some of the spectra (see Fig. 10, for example) show lines at 2900,
3700, and 4150 cm . These have been shown to be grating ghosts and
not characteristic of the scatterer. The spectral region above 2000 cm
o
was not evaluated for responses with 6471-A excitation because it had
too many artifacts caused by unfiltered plasma lines or grating ghosts.
During the data analysis some of the very weakest lines were ignored
when it was felt that their value relative to other lines for the same
substance was minimal.
The measured lines are also shown in "spectral" form in Fig. 12
in order to illustrate the spectral distribution of the lines observed
for the present materials. Note that there is a relatively uniform
distribution of lines both in wavenumber offset and amplitude. Fig.
12 indicates amplitude in measured counts per second. Note that this
count-per-second scale corresponds to the laboratory experimental meas-
urement values and not to the in-stack monitoring-system values given
in Table 2 for the fluorescence measurements. The relative magnitudes
of these count rates would be correct for the in-stack monitoring system,
however.
Some general remarks about the observed spectra are in order. It
can be seen from Fig. 12 that the white solids were the most efficient
scatterers. The colored solids either did not have a developed spectrum,
or their intensities were low. It was also observed that the substances
involving heavier elements had stronger lines. For example, the strongest
lines in the figure are due to mercury and lead compounds. Even in the
_T
region 900 to 1200 cm where the Raman lines are due to the sulfate
35
-------
38S
368
304
277
WAVELENGTH SHIFT —A
224 188
122
7X6 48.1
23.9
1300
10
z
8
Q
tu
X
I
1"
i »
|
si
5,
»
8
p
"V
X
°r,
<
f
c»
2?
130
13
§
I
ft
'o
fc
UJ
M
i
6
0.13
aois
1700 1600 1SOO 14OO
1200 1100
1000 800 800 700
FREQUENCY SHIFT, WAVENUMBERS—cm'1
800 600
400 300 200 100
FIGURE 12 RAMAN SPECTRAL RESPONSE
SUMMARY
36
-------
ion, the lead and mercury sulfates have the highest count rates. The
spectra of alumina and other aluminum compounds (fluorides, sulphate)
were among the weakest even though they were all white.
The chosen materials included oxides, sulfides, halides, and sul-
fates. Among these only the sulfates have a vibrating sub-unit—the
SO molecular Ion On examining the spectra, it is noted that with
4
very few exceptions, only the sulfates have Raman lines with shifts
greater than 700 cm . Note that the sulfate lines are grouped in three
regions: 400 to 500 cm~ , 600 to 700 cm" , and 950 to 1200 cm . It
is thus seen that even though the vibrational sub-unit SO has frequencies
4
which are, to first order, characteristic of the SO unit, the actual
4
frequencies are dependent on the specific compound in which they are
present.
The point to be made most strongly is that the Raman spectral
shifts or intensities cannot be accurately predicted by existing theory
and are capable of wide variation. Thus, a general method for positive
identification cannot depend on one specific spectral feature but must
necessarily depend on several. Also, in the samples measured there are
about 50 lines in the range 100 to 700 cm (about 150 A). This close
spacing means that good resolution will be needed to discriminate between
the several lines in any general identification scheme.
The spectra generated during the course of this program agree well
with the published spectra that are available. The spectral shifts
agree to within 5 cm . Since relative-intensity data are usually not
available, no comparisons of amplitude can be made. Further, since the
published spectra are for oriented single crystals, a direct comparison
with powder samples may not be completely accurate.
To convert the intensities to a cross-section value for the purposes
of feasibility analysis, the published value of the cross section of
37
-------
— 1 6
the 207 cm line of CdS powder was used. This reference is not clear
about the details of the excitation frequency, the units used, and
whether the cross section is for the entire line or the peak value. It
is inferred from the body of the paper that the quoted Raman cross sec-
-1 -28 2 n
tion for the 207 cm line of CdS powder (0.7 x 10 cm at 293 K) is
the total cross section (4rr sr) per molecule, measured with 6328-A
-4
excitation. Assuming \ dependence on wavelength and that the entire
line was measured, the cross section at 4880 A is calculated as
-29 2 -1 -1
1.57 x 10 cm sr molecule
Only one value for the CdS cross section is available. In other
materials a wide spread of measured cross-section values is often re-
ported. For example, the cross-section value quoted in the same paper
for the 992 cm line of liquid benzene leads to a calculated cross-
o -29 2 -1 -1
section value at 4880 A of 4.5 x 10 cm molecule sr . Table 4
shows the range of values of the Benzene cross section as measured by
various workers. It is seen that there is a spread of more than an
order of magnitude in the values, with the value of Ref. 6 being the
highest reported value. While comparable results were not available
for CdS powder, it is conceivable that the techniques of Ref. 6 may have
-29 2
systematically given high values. Thus, the value of 1.57 x 10 cm
molecule sr is used with reservations about its accuracy. Since
the linewidth measured during this investigation (with a 10 cm
spectrometer slit) is 20 cm , the measured peak intensity corresponds
5
to the entire line. Thus, the measured rate of 1.2 x 10 counts per
second corresponds to the above cross section with 350 mW of 4880-A
excitation.
38
-------
Table 4
COMPARISON OF VARIOUS MEASUREMENTS OF THE
RAMAN CROSS SECTION OF BENZENE (992 cm"1 LINE)
Excitation
X
6328
4880
4880
4880
4880
6943
Reported
Cross Section
p-i -i
(cm sr •*• molecule )
-29
1.59 x 10
-29
3.25 x 10
-29
2.42 x 10
-30
5.48 x 10
-30
3.0 x 10
-31
4.5 x 10
Cross Section
at 4880 A
2 _i _i
(cm sr molecule )
-29
4.5 x 10
-29
3.25 x 10
-29
2.42 x 10
-30
5.48 x 10
-30
3.0 x 10
-30
1.84 x 10
Reference
6
3
4
7
8
9
A measurement-system constant, K , can be calculated from a knowl-
o
edge of the measured count rate and the published cross section as
N M
K =-^- .
o pa
5 -1 -3
Using N = 1.2 x 10 counts s , p = 4.82 g cm , M = 144.476, and
r
— 29 2 —1—1
a = 1.57 x 10~ cm molecule sr for CdS, we calculate K = 2.29 x
o
35 -1 -1
10 counts cm molecule sr s molecule . The cross section of any
other material can then be calculated as
a = N M/K p .
r o
39
-------
It should be recalled that the measurements were made /ith a slit width
of 10 cm ; thus the above relation gives the cross section for the
entire line for lines with halfwidths smaller than 10 cm , but for
-1
lines with larger widths it gives the cross section per 10 cm at the
peak.
The calculated cross sections for the materials evaluated on this
contract are shown in Table 3. Iu te again that the measured count rate
is more representative of the variation in sign«j. strength with material
than is the cross section.
40
-------
V IN-STACK MONITORING SYSTEMS
A. General
In considering the application of optical in-stack monitoring
systems, two types of monitoring functions were postulated. The first
function is the assessment of the relative concentration of a number of
constituents without a knowledge of the absolute concentration of any
single component. Because absolute numbers are not required, this type
can also be considered as a qualitative measurement. Second is the
quantitative measurement of the amount of a single constituent present
in the stack particle stream. Ideally, of course, one would hope that
a quantitative measure of each of the constituents in the particulate
stream would be measurable. The extent to which this is possible can be
determined by extrapolating the results of the quantitative and quali-
tative approaches taken in this project.
In addition to these two types of applications discussed above,
both Raman and fluorescence material interactions were considered as
sensi;.;., mechanisms. As the project progressed, it seemed that there was
a match between the application and the sensing mechanism. In particular,
the fluorescence response seemed more appropriate for the quantitative
sensing of one kind of material in the particulate stream, and the Raman
response seemed more appropriate for the relative analysis of several
constituents on a qualitative basis. This comes about because the fluo-
rescence i .- = pjnse is relatively large and can thus yield relatively good
quantitative information, but the spectral characteristic of fluorescence
is such that it would be difficult to separate a variety of const^. -nts.
In contrast, the Raman response shows relatively distinct spectral loa
tures for each material; however, the low level of the response would
41
-------
make the collection of quantitative information rather difficult. For
this reason, the systems analyzed in this section will be divided into
fluorescent systems, with emphasis on quantitative measurements, and
Raman systems, with primary emphasis on separation of spectral informa-
tion and less emphasis on amplitude-calibration techniques.
The in-stack fluorescence monitoring system can provide a high
level of performance relative to the experimental laboratory measure-
ments because of the large fluorescence response. This means that all
the materials that have observable fluorescence in the laboratory measure-
ments can be detected by an in-stack system, although a few materials
may be marginal with inexpensive monitoring systems. Thirteen materials
were observed to have significant fluorescence responses, and these can
be detected by the in-stack system. Twenty-one materials did not have
observable fluorescence and would not be detectable by the in-stack
system.
Two Raman systems have been considered. The first uses a spectro-
meter and is similar to the laboratory measuring instrument. The other
uses an optical filter and may provide for a limited analysis capability
at a lowered cost. Generally speaking, for these systems, a lower per-
formance level would be obtained in the in-stack system than was present
in the laboratory measuring instruments. Twenty-two materials had mea-
surable Raman spectra and twelve materials did not. Of the twenty-two
materials that had observable Raman spectra, approximately fifteen mate-
rials would be detectable in the in-stack system, four materials would
be marginal, and three materials would not be detectable in the in-stack
system, even though they had observable Raman spectra in the laboratory.
B. Fluorescence-Monitoring Systems
In order to make reasonable estimates of the in-stack detectability
of the fluorescent materials measured in the laboratory, some common
42
-------
signal reference level must be used to relate these laboratory amplitudes
to calculated in-stack-monitoring-system signal levels. The most appro-
priate approach was believed to be the use of a diffuse white reflector
to relate calculated and measured signal levels. The reflector chosen
was a freshly-prepared, diffuse, white magnesium oxide surface. This
reference material was measured on the Baird-Atomic spectrofluorimeter
and served as a reference amplitude level to which material fluorescent
responses could be related. The assumed unity diffuse reflection coef-
ficient could then be used as a reference response signal level for the
in-stack monitoring system as well. If the effects of scattering, ab-
sorption, obscuration, etc., are neglected for the particle stream, it
is thus possible to obtain estimates of the in-stack fluorescence-moni-
toring system performance for each of the materials measured by the
laboratory instrument.
The measured reference signal lev«l for the diffuse white.reflector
is shown in Table 5. This reference measurement was made at a peak wave-
length of 455 nm and after normalization resulted in a relatively large
peak signal level compared to the existing fluorescent-material responses.
The peak photon count for the diffuse reflector is shown in the sixth
column and represents the calculated in-stack' monitoring-sys tern reference
signal level (i.e., the signal level that would be observed with an in-
stack monitoring system viewing a white reflector rather than a particu-
late stream).
Also included in Table 5 are the responses for three highly fluo-
rescent scintillation dye materials: bis-MSB [p-bis (0-methylstyryl)
benzene], POPOP [2, 2'-p-phenylenebis (5-phenyloxazole)J, and PPO [2, 5
diphenyloxazolej. These highly fluorescent materials have a quantum
efficiency of approximately 0.5. A crude check on this quantum effi-
ciency, and thus on the assumed reference level, was made by calculating
the normalized integrated signal level in addition to the peak signal
lavel. The integrated signal level is a calculated value of the instrument
43
-------
Table 5
FLUORESCENT-REFERENCE-MATERIAL MEASUREMENT
Material
Diffuse
white
reflector
BIS (MSB)
POPOP
POP
Peak Wavelength (nmX
Excitation
455
402;425
424
365
Fluorescence
—
480
460
395
Normalized Peak
Signal Level (ua)
34,700
1,340
1,880
2,411
Normalized
Integrated
Signal Level (ua)
34,700
17,900
21,400
23,900
Peak
Photon Count
(pnotons/s)
10
1.25 x 10
8
4.83 x 10
8
6.77 x 10
8
8.69 x 10
Integrated
Photon Count
(photons/s)
~\ f\
10
1.25 x 10
9
6.45 x 10
9
7.71 x 10
9
8.61 x 10
-------
response, assuming that all the light emitted over the entire fluores-
cence response curve was measured by the instrument, while the peak
signal level corresponds to the optical signal measured just through
the narrow bandwidth of the instrument itself. These two levels are
substantially different because these dye materials were measured at the
maximum resolution of 2 nm by the Baird-Atomic instrument, and the width
of the fluorescent response is significantly larger than 2 nm* Note
from Table 5 that the integrated signal levels for the three highly
fluorescent materials represent a substantial fraction (~ 0.5) of the
integrated response for the diffuse white reflector. This check of the
integrated signal level for the three highly fluorescent materials
serves as an additional verification for the use of the diffuse white
reflector as a fluorescence amplitude reference.
The response of the in-stack monitoring system to a diffuse white
particulate material must be calculated in order to derive estimates
of detectability for the measured materials in a typical in-stack moni-
toring system. This response can then be related to the material mea-
surements through the white-reflector measurement. This method assumes
that the relative particulate fluorescence properties are the same as
the relative bulk fluorescence properties and that the relationship
between the fluorescence response and the diffuse white-reflector re-
sponse is identical for both powders and bulk reflecting surfaces. This
assumption neglects the effects of self-absorption, Mie and other par-
ticulai.3 scattering effects, and some geometrical optical effects.
Assume a fluorescence-monitoring system that illuminates a volume
of particulate material, suspended in an airstream, with an optical
signal whose wavelength is assumed to be at the peak of the desired
material excitation response curve. The light from the fluorescing
particulate materials is collected by a collecting lens, and, after
optical filtering, is directed to an appropriate optical detector. It
45
-------
is also assumed that che function of this fluorescence-monitoring system
is to obtain quantitative data on a known constituent or constituents.
Thus, the predominant signal-processing requirements have to do with
amplitude measurements in a fixed channel (or channels) of optical
filtering, with no intercomparison of multichannel amplitude information
being required for constituent analysis.
The schematic diagram of this in-stack fluorescence monitoring
system would be virtually identical to the schematic diagram for the
laboratory fluorescence-measurement systems as shown in Fig. 1. The
most significant difference in this schematic diagram would be that the
common optical volume would be ;within the stack rather than within some
sample chamber, as was the case for laboratory measurements. Other
differences will exist in physical configuration and in component
values; however, these difierences in numerical value do not change the
schematic diagram for the resulting instrument.
If we assume a common optical volume of diameter D and length
Li
•L , the number of particles illuminated in this common volume is
TTD -f-V
L
n =
where v is the particulate density in particles per cubic centimeter.
The equivalent particle area is
A =
P 4
where D is the mean particle diameter. The fraction of light inter-
cepted by the total particulate area in the common illuminated volume
is equal to
46
-------
2
If we assume an illumination power of P watts and a collection ef-
ficiency co, an optical efficiency yj, and quantum efficiency Q,
the detector count in photons per seconds is equal to
20 P
N = 0.2 x 10
F 4
With u> = 0. l/4n, -p = 0.1, and Q = 0.21, the detector count is
10
1.25 x 10 photons/s. This is shown in Table 5 as the reference peak
photon-count level for the diffuse white reflector. Also given in
Table 5 are the peak photon counts for the three highly fluorescent
dye materials. An integrated photon-count level is also shown for the
fluorescent material to indicate the photon count that would be avail-
able assuming that all of the fluorescent response light could be col-
lected and utilized by the optical detector. It is not usually feasible
to use the full response width in practice because of the necessary re-
jection of the excitation light by the optical filter. Thus a response
closer to the peak photon count would be observed in practical in-stack
monitoring systems. Fortunately, the peak photon-count level is suffi-
ciently high that it provides a useful signal level for all of the
materials investigated. This peak photon-count level is given in Table
2 for the chemical materials investigated on this project. Reasonably
high count rates would be obtainable for all of the fluorescent materials
investigated in this project. A count rate is not given for the Raman
response of water because of the significant difference in geometry for
the liquid measurement. The larger common volume of the liquid samples
would yield a higher response level than would be achieved with the
smaller common volume that is obtained with the solid samples. The
47
-------
position of the entry for water in Table 2 does, however, give an in-
dication of the signal levels of Raman responses relative to the re-
sponses of slightly fluorescent materials.
The lower limit for response measurements is set by the dark count
of the detector. The dark count for a typical photomultiplier detector
(i.e., Type 1P28) is approximately 8000 counts/s. This dark count sets
a minimum level on the number of photons/s that can be detected by the
system. This dark count can be reduced by using a low-noise tube such
as the EMI 6094S, with a dark count of 660 counts/s. If necessary,
these count rates can be reduced by at least a factor of 10 by cooling
and tube selection. Thus, it is possible to detect even the low-fluo-
rescence materials with feasible systems.
C. Raman Monitoring Systems
1. General
Two types of Raman measurement instruments will be discussed
for in-stack monitoring applications. The first of these is an instru-
ment utilizing a spectrometer that scans the entire spectral region
of interest. The second is a lower-cost filter-type instrument that
looks only at specific wavelengths, and will be considered for more
limited applications of the Raman monitoring technique.
The spectrometer-type instrument is similar to the instrument
used for the laboratory measurements. The major difference, as for
the fluorescence system, is in the common optical volume, which is in-
side the stack rather than located at the sample-holder position. A
schematic diagram for such a system would be similar to that shown in
Fig. 8 for the laboratory system.
2. Spectrometer-Type System
For this instrument, assume a laser beam of power P watts,
diameter w cm, wavelength \ cm, and negligible divergence passes
48
-------
through an aerosol containing the particles of interest. These particles
are assumed to be spherical, with an average diameter d cm, and to have
-3 -3 -1
a number density n cm , density p g cm , molecular weight M g mole ,
2 -1 -1
and Raman cross section a cm molecule sr
o
Assume that the Raman light is collected at 90 by a lens
system and analyzed by a spectrometer of f-number F, and dispersion
D cm/cm . If the needed resolution is v cm , then the slit width to
be used is Dv cm. Let the slit height be H cm. The acceptance angle
2
of the spectrometer is (1/F). The corresponding solid angle is rr/(4F ).
2
The entendue of the system is then rrDvH/(4F ), which should remain con-
stant in the optimum optical system.
The following typical values are used in the discussions:
-3
P = 1 watt p = 4.82 g cm
-4 -1
\ = 0.488 x 10 cm M = 144.46 g mole
-29 2 -1 -1
w = 0.1 cm a = 1.57 x 10 cm molecule sr
-4
d = 1 x 10 cm F = 6.8
6-3 -3 -1
n = 10 cm D = 4.5 x 10 cm/cm
7] = 0.268 v = 10 cm
Q = 0.15 H = 1 cm .
The mass density of the aerosol is then
3 -6 -3
mrd p/6 = 2.52 x J.O g cm
The number of particles per unit length of the laser beam is
2 3-1
rnrw /4 = 7.85 x 10 cm
The total geometrical cross-sectional area of these particles is
Q o «c\ O — 1
n(rrw /4) (TO /4) = 6.17 x 10 cm cm
49
-------
The fraction of the laser-beam cross section occupied by the particles
is
2 -3 -1 -1
rrnd /4 = 7.85 x 10 cm = 0.79% cm
This is a small fraction of the laser beam and therefore represents
only a small loss or attenuation of light in the common optical volume.
The actual fraction of light scattered is determined by an effective
cross section, which may be .different from the geometrical cross section.
The theory of scattering by small particles indicates that the scattering
efficiency, q, of the particles can vary from 2 to 6 depending on the
size and refractive index (i.e., 2 to 6 times the energy intercepted by
the geometrical cross section of the particle is scattered). This scat-
tered light appears as attenuation for the transmitted beam. Thus, the
2 -1
attenuation coefficient due to scattering is a = nqrd /4 cm . The
scattered radiation may still be effective in causing an observed Raman
signal because of multiple scatter. This has the effect of increasing
the beam diameter or decreasing the photon flux density. However, cal-
culations of these multiple-scattering events is a difficult undertaking.
Because of this difficulty and the fact that the attenuation in the par-
ticulate common optical volume is a small percentage under the assumed
conditions, it will be neglected in the following treatment.
If the laser beam of length £ is imaged to fill the slit with
a collection angle of Qsr, then
w/n = nDvH/(4F )
where wm = Dv; and H = £m where m is the magnification of the sys-
tem. Thus,
50
-------
2
_ TTDVH _1_ _ TT m
2 w£ ~ 4 F
4F
2
TT Dv -3
= - — = 3.44 x 10 sr .
4 wF
The photon flux density is then
4P \ 20 -2 -1
- ; — = 3.13 x 10 photons cm s
2 he
TTW
2
2 TTW Hw 3
= 0.017 cm
The sample volume is
4 Dv
The number of particles in the sample volume is
nn!LJL= 1>74 x 104 .
4 Dv
Neglecting attenuation, the number of scattering molecules in the sample
volume is
3 3
TTW H TTd p 14
n ---- A = 1.84 x 10 molecules
4 Dv 6 M
From this, the number of Raman scattered photons is
_
~
3 3
X nrrw H nd pAq
r ~ 2 he 4Dv 6 M
TTW
3
PXA rnrd pa wH
he 6 M Dv
5 -1 -1
= 9.03 x 10 photons s sr
With a spectrometer transmission of ji = 0.268 and a detector
quantum efficiency Q = 0.15, the number of counts for CdS is
51
-------
PXA nnd pa wH
N = -1- QnO
r he 6 M Dv '"
3 9
_ PX.A nrrd £O wH TT /Dv\
he 6 M Dv 4 \wF7
3
IL ^A — ^rcd pg DvHT|Q
4 he w 6 M 2
F
= 125 count s
The above equation may be rewritten as
- l 16 x 1042 - - £2
_ i.ib x 10
P ,, -1
= K — LJHS counts s
1 w
42 -1 -1
where K = 1.16 x 10 molecule mole J
3 -7
L = nnd /6 = 5.24 x 10
- pa/M = 5.24 x 10 mole molecule cm sr (for CdS)
2 ~5 2
S = DvH-y]Q/F = 3.91 x 10 cm sr
for the chosen spectrometer system.
is determined by material characteristics, S by spectro-
meter and detector characteristics, and L is the volume fraction of the
material in the aerosol. Thus, given a particular spectrometer system,
particle type, and concentration, the only parameters that can be changed
to increase the number of counts are the laser power and the beam diameter.
52
-------
Increases in laser power are limited by technology and expense. The
beam diameter w cannot be reduced indefinitely because the beam diver-
gence will increase and lead to a lower limit on w. More importantly,
the total number of particles in the sample volume will also decrease
as the cube of the beam diameter w. The decreasing sample volume is
compensated for by a higher flux, and a larger collection angle Q, both
2
of which increase as w .
Up to this point, the analysis has considered only the optical
effects in the common volume illuminated by the laser and viewed by the
spectrometer. In particular, the attenuation in this region has been
neglected. Although the attenuation in the common volume is small, the
attenuation in the optical path to and from the common volume may be
significant and is calculated as follows:
Assume that the laser beam travels a distance u through the
o
aerosol to the common sample volume, and the Raman scattered photons
travel a distance u through the aerosol to reach the detector optics.
Since the sample length is •£, it is readily shown that, if the light
attenuation of the aerosol is taken into account, the above expression
must be multiplied by an attenuation factor A given by
~a(Uo * V
A =5
[1 - e^]
2 -1
where en = nqird /4 cm
For a typical system, u = u =30 cm, and assuming q = 3,
then
A = 0.237
and for CdS,
— 1
N = 29.6 counts s
r
53
-------
The count rate is thus reduced from 125 counts/s to 29.6 counts/s; it
is evident that the attenuation in the path is significant under the
assumed conditions.
Implicit in the above calculation of count rate is the assump-
tion that the particle density is constant. Actually, however, there
will be fluctuations in the number of particles in a given volume, and
this can appear as noise in the detected optical signal. For example,
for the values used above, the sample volume contains on the average
4 4 1/2
1.74 x 10 particles. The rms fluctuation in this value is (1.74 x 10 )
= 132 or 0.76%. If the beam diameter were reduced by a factor b under
stationary conditions (particles are stationary) , the percentage fluc-
tuation will increase by b *b. However, when the aerosol is moving at
a flow rate of v cm s , the actual aerosol volume swept through the
3 -1
sampling volume will be wv£ cm s , if the laser beam is normal to
the direction of the flow. Thus, the number of particles in the volume
will be nwv£ s . For v = 100 cm s , the number of particles
7
will be 2.2 x 10 . The calculated number density will then be in error
by 0.02%. Thus, for reasonable in-stack effluent velocities, noise due
to this source is not expected to be a problem. Optimization of the beam
diameter, however, will be dependent on the particle density and the in-
stack velocity if one wishes the photon noise to limit the accuracy of
measurement.
The SNR characteristics may be evaluated using the data pre-
sented in the previous section on measured Raman and background ratios.
There are, however, additional sources of background in the in-stack
configuration. Contained in the laser-illuminated sample volume, along
with the particles of interest, are molecules of the gas and possibly
other constituents that cause Raman and fluorescent scatter. Estimation
of the fluorescent background intensity cannot be made other than by
actual measurement. The Raman scatter from the gas molecules may be
o
estiir" '- - '. as follows. At atmospheric pressure and 300 K, the number
54
-------
19 -3
density of any gas molecule is n = 2.45 x 10 molecules cm . The
3 2 17
number in the sample volume is (TTW £n /4) = 4.16 x 10 molecules.
a
3
Thus, the number of air molecules is 2.26 x 10 larger than the number
of molecules of interest and could perhaps cause a large signal that
would mask the return from the desired species.
-1
Fortunately, most gases have Raman shifts higher than 1000 cm ,
which is the approximate upper limit needed for the solid materials of
interest. Table 6 is a list of the vibrational Raman shifts of these
gases as well as their measured cross sections relative to nitrogen.
-31 2
Nitrogen, the reference gas, has a cross section of 3.3 ± 1.1 x 10 cm
s molecule . Also, the gas cross sections are, on the average, an
order of magnitude smaller than those of the solids, and will partially
compensate for the larger concentration. The pure rotational Raman
shifts of most of the gases occur in the region below 100 cm and have
higher cross sections. Again, this spectral region will not be useful,
-1
because of scatter from particles. Thus, except for the 519 cm line
of SO , no large interference from gases is expected in the range 100
2 -1
to 1000 cm where most of the lines of interest occur. However, it
should be remembered that these gas molecules will contribute to rayleigh
scattering and, with low particle concentrations, to the background level
in the spectrometer.
Now that a relatively interference-free spectral region is
assured, the detection limits for the spectrometer system can be esti-
mated. The detection limit will be presented as an integration time to
detect a given standard condition. This method was chosen to simplify
the data presentation for both long- and short-integration-time systems.
One important condition on the detection process is the expected
dilution of the active material by a neutral substance. In an aerosol,
this dilution can be accomplished in two ways; first, the active mate-
rial may be only a part of the particle composition, and, second, the
55
-------
Table 6
RELATIVE RAMAN CROSS SECTIONS FOR GASES
Gas
N2
0
2
H (sum)
H2 (Q(l))
CO
NO
CO^)
CO (2V )
2 2
N2°(V
N2°(V3)
SO (v )
2 1
S°2(V2)
H2S(V
NH3(V
ND^)
CH4(V1)
C2H6(V3)
WV
WV
Wave number
Shift (cm'1)
2331
1556
4161
4161
2145
1877
1388
1286
1285
2224
1151
519
2611
3334
2420
2914
993
3062
992
Relative
Cross Section
1.0
1.3
2.4
1.6
1.0
0.27
1.4
0.89
2. 2
0.51
5.2 i
0. 12
6.4
5.0
3.0
6.0
1.6
7.0
9.1
56
-------
solid-material content in the aerosol sample volume may be changed by
varying the particle size, shape, or concentration. It is convenient
to define a dilution factor x that includes both of these effects in
a single factor.
In order to estimate the fractional dilution at which these
compounds are detectable, the following assumptions are made: (1) the
measured experimental values for the Raman lines and background are for
the sample at theoretical density; (2) the intensity of the Raman lines
and the intensity of the background are proportional to each other and
the incident power; (3) the dilution of the sample is accomplished by
mixing with a neutral material that does not contribute to the back-
ground. This last assumption is true only for changes in particle size
or concentration. It is not possible to change the solid-material dilu-
tion ratio without changing the ratio of Raman photon counts to back-
ground photon counts. Also, the assumption is true for aerosols only
if the particulate scatter is the predominant background source and the
detailed Mie scatter properties are neglected.
For consistency with the measured data, assume a Raman line of
-1 -1
width v cm with a peak count rate of N counts s when this rate
r r
is measured with a 10 cm slit. Let the background counts at the posi-
-1 -1
tion rf the line be N counts s for a 10 cm slit. Assume also that
b
the detection system uses a bandwidth of Vr cm"1. Then the Raman
counts received will be Nr(vr/10) counts s"1 and the background counts
will bu N (v /10) counts s~1, if vr > 10 (otherwise, v =10).
L r r
The total number of counts over a counting period T will then be
(N + N, ) ,10. If the dilution factor is x, then the number of
r b r
counts is (N + N )v xT/10.
r b r
Neglecting other noise sources, except Poisson statistical
noise (photo:: noise) , the noise component is then
1/2
Noise = [(N + N )v xT/10]
r b r
57
-------
and the ratio of the Raman signal to the noise xn signal is
n N v xT/10 /N v xT\
Raman r r i - ~ \
where
Noise r -i 1/2
[(N + N )v xT/10]
r b r
Z = (N H- N )/N
r b r
If one desires an accuracy of a% in the result, this may be
interpreted as a desired SNR of (100/a). Thus,
2 5
Txa = 10 Z/N v s .
r r
This expression gives the time, in seconds, to obtain enough counts to
attain an accuracy of a% when the material of interest is at a dilution
of x relative to experimental conditions. With v = 10 cm and
r
N calculated for other materials from
r
P -1
N = K — LS/ffA counts s
r 1 w
then
Txa2 = 104Z/K
If we use the values previously assumed, we obtain a figure
of merit:"
2 —29
Txa = 4.196 x 10
—31 2
For CdS,J{ = 5.24 x 10~ , Z = 3.17, A (0.1) - 0.866, and Txa = 293 s,
Thus, for CdS, to detect a mass loading of 10% of the standard
o _3
conditions [i.e., (mrd 6/6) x 0.1 = 0.252 \j.g cm ] with an accuracy of
58
-------
1%, the counting time required is T = 2930 s, or 48.9 minutes. A
summary of the performance of the spectrometer instrument in terms of
this "figure of merit," Txa , is given in Section V-C-4.
It is important to recall the premises on which these calcula-
tions are based. First, the aerosol opacity is assumed small in the
common volume so that attenuation of the laser beam as well as the Raman
scatter is small. The number density and velocity of the aerosol is
such that the fluctuation in the number of laser-illuminated particles
has a value more precise than the attempted accuracy a, relative to
the sample volume and the time of measurement. Also, the Raman-to-
background ratio is assumed to be the same as that measured in the
powers in the experimental program. The assumed cross section for CdS is
a reference value and determines all the other cross-section values.
Sample-to-sample changes in the sampling geometry during the experiments
with powders are neglected.
3. Filter-Type System
With the development and general availability of high-quality
dichroic and interference filters, the need for expensive dispersive
optics is always under scrutiny, especially when performance is expected
only over a small wavelength range. The design and analysis of a system
based on such interference filters is given in this section.
Before proceeding with the analysis, however, some of the
characteristics of interference filters will be reviewed. Interference
filters are usually made up of thin dielectric films and can be made to
achieve a narrow bandpass, typically 10 to 20 A in the visible region
of the spectrum. Transmission efficiencies on the order of 0.5 in the
-4
passband and attenuation on the order of 10 in the stopband can be
achieved with these filters. In general, the peak transmission decreases
as the desired passband narrows, and the cost is inversely related to
the width of the passband.
59
-------
The passband of an interference filter is designed for normal
o
(90 ) incidence of the optical beam. At all other angles of incidence,
the passband position shifts to higher frequencies as the square of the
angle of incidence. The bandpass thus widens with increasing angle of
incidence. The bandpass position also changes with the temperature of
the filter. Typical temperature coefficients are about +0.2 A/ K. A
practical problem in the fabrication of large-area filters is the uni-
formity of the passband over various portions of the filter.
As we have noted earlier, the desired Raman lines of solids
-1 O
occur within a wavenumber-shift range of 0 to 1000 cm . With 4880-A
excitation, this range translates to roughly 250 A. Typical Raman line
widths are about 10 cm or about 2.5 A. A 20-A passband therefore means
a wavenumber spread of 80 cm and may include the lines of several
species. Even if an interference filter of 2.5 A bandwidth were obtained
at some expense and loss of transmission, it is a difficult undertaking
to maintain the passband at a given absolute spectral position, due to
the effects of temperature and angle of incidence. Thus, it should be
noted that the isolation of the Raman spectral lines of a single species
for general analysis purposes is not practical for present interference
filters because of the limited resolution capabilities. However; if the
species of interest are known and a general analysis shows that inter-
ference from closely spaced lines is not a problem, then one might con-
sider a nondispersive technique based on interference filters for
estimation of the particular species.
The off-peak rejection capability of interference filters is
important because the scatter from air molecules and particles produces
a very large signal at the laser wavelength that must be rejected by
the filter. It will be recalled that the number of air molecules in a
given sample volume is about 2000 times higher than the number of mole-
cules of the species of interest. Rayleigh scattering cross sections
60
-------
3
are typically 10 times higher than Raman cross sections, and Mie scat-
tering cross sections can be several orders of magnitude higher than
rayleigh. It is thus seen that for each Raman photon of interest, there
6 8
are typically 10 to 10 other photons present, in addition to other
Raman photons that are not of interest. These additional pnotons must
be rejected if the desired Raman signal is to be detected.
A primary advantage of the spectrometer-type instrument is
the large rejection of this scattered light that can be obtained with
double monochrometors. The interference filter, on the other hand,
faces a formidable task in sorting out the Raman photons. Elaborate
interferometric techniques have been used for reducing this high back-
ground of scattered laser light, but they add to the expense and com-
plexity, and, in particular^ detract from the ruggedness of the device.
Recently, the narrow absorption lines of molecular iodine vapor have been
used to suppress the 5145-A line of the argon ion laser with an attenua-
8
tion of 10 , with a corresponding attenuation of only about 6 for the
10
desired wavelengths. This is one technique for attenuating the ray-
leign line. The following analysis assumes that such a technique is
used.
The typical attenuation capability of a spectrometer of the
-8 -10
type used in the experiment is on the order of 10 to 10 for the
rayleigh wavelength, even as close as a few wavenumbers (e.g., 10 cm )
away. Thus, the background levels measured during the experiments would
be realistic values to use in the following analysis, assuming that a
-8 -10
comparable suppression of 10 to 10 is achieved for the rayleigh
wavelength.
Let the filter have a bandwidth of v cm and transmittance
at normal incidence. In order to transmit radiation from an extended
source, the filter must pass light at a finite angle 0.
61
-------
The transmission peak shifts spectrally as
v = — 1 + = v +6
m 2 cos 9 o
vvhere
26 = (v /cos 6) - v
o o
The effective bandwidth of the filter will be broader, being a convolu-
tion of v and 6, and the peak transmission 7] is also reduced. It
U
has been shown that
222
v = v + 166
6
and that
v / 46
v / 46 \
Tle=7]-arctan(v)
If we define the angle 9 such as to make the effective bandwidth of
the filter */2 v, then 46 = v and T) = rrTl/4. If we define R = v /v,
9 o
then
v = 46 = 2v - 1
o cos 9
1 - cos 9 =
2R + 1
The acceptance solid angle in the medium of the filter is given by
2TT
ou = 2rr(l - cos 9) =
2R + 1
*
If the refractive index of the filter is n , then the acceptance angle
in air will be
3
since R ~ 10 .
*2 _
n 2rr *^ .
co = « n TT/R
2R + 1
62
-------
To perform a system analysis, we assume, as before, a laser
beam of dimaeter w cm and a filter whose useful linear dimension is
A cm. From constancy of entendue we have
A (ju = w Q
or
0 = cu(A/w) = (ir/R)(n*A/w)2 .
Following the earlier derivation, we calculate the number of Raman counts
to be
V * 2
XA P TT 3 pa r (n A)
N = TT nd —
r he w 6 M 10 R
v
If the optical efficiency of the imaging system is 83.3% as before, and
the peak transmission of the filter at normal incidence is 80%, then
7| = 0.523. For a typical 20-$ filter at 5000 A*, R = 250 and n =1.5.
A = 5 cm is a practical size for filters. Assume Q = 0.15. From the
above equation, as before, the smaller the beam size, the larger the
Raman signal. The same comments regarding practicable Q values and
particle densities are in order. Assume w = 1 cm, and Q = 0.18 sr
(or a 15-cm-diameter lens at 30 cm). For these conditions, for CdS,
the number of Raman counts will be
N = 5633 x 0.236 counts s
r
= 1335 counts s
Again, the measured signal-to-background ratio, Z, will be
used to estimate detection limits for this system. Note, however, that
this signal-to-background ratio was measured with a spectral width of
63
-------
10 cm. If the effective filter bandwidth is */2 v cm , the signal-
to-background ratio will be different. If we assume the background per
cm is constant, then the background counts will be Jz N v/10. If
-1 b
the full width of the Raman lines is less than 10 cm , The Raman
counts will stay the same even under a broader passband. If the Raman
-1
line width is greater than 10 cm , then the Raman counts received Under
the wider passband will be N v /10, or J2 N v/10, whichever is
r r r
smaller. Hence, the signal-to-background parameter now becomes
It follows that
N + *2 N V/10
r b
- for v < 10
N r
N v + /v/2 N v
Z1 = r r for J% V > v > 10
N v v r
r r
N + N
r b /—
for v > V2 v .
N r
r
for v < 10
r
S N v XT
Raman = ( -LJ_ for v > v > 10
Noise \ 10Z1
r- 1/2
N V2 VxT \
— 1 for v > J2 v
10Z1 / r
64
-------
and
104Z'
for v < 10
N r
r
2 10 Z' 10 i-
Txa = for V2 v > v > 10
N v r
r r
10 Z' 10 ^ ,_
j=— for v > V2
N J2 v r
r v
Using the appropriate values in the expression
N = K - LS;
r 2 w
we get
-31
9.3 x 10 (Z'/A) for v < 10
r
2 —31
Txa = 9.3 x 10 (Z'/A)(10/v ) for ^2 v > v > 10
r r
9.3 x 10~31 (Z'/A)(10A/2 V) for v > ^2 v .
r
As before, the value of attenuation, A, depends on the aerosol con-
centration x. A summary of the performance levels of this filter
instrument is given at the end of this section.
4. Performance Summary for Raman Systems
2
The normalized detection time, Ta , is given in Table 7 for
each Raman-active material and for both Raman systems. The concentration
6
parameter; x, of the particulate stream is relative to 10 particles/
3
cm and is varied from 2 to 0.001 to demonstrate a variety of typical
concentrations. The attenuation of the aerosol is accounted for in these
fi
calculations and results in a maximum detectability at about 10 par-
3
tides/cm . Larger concentrations produce high optical-attenuation
65
-------
Table 7
RAMAN IN-STACK MONITORING INSTRUMENT PERFORMANCE SUMMARY
Material
HgO
HgS
CaSo
4
CdS
Al (SO )
2 4
3
Phosphate
Rock Feed
CaF^
2
Na A1F
3 6
Lead
Concentrate
NBS Coal
Line
328
550
250
275
341
425
490
620
670
1015
1150
212
305
347
365
563
599
480
620
1000-
1200
580
945
322
550
1040
950
1350
1600
2
Normalized Detection Time (Ta )
Spectrometer Instrument
x = 2
139.2
1254
276.7
5.2 X 10
3
1.50 X 10
4
4.48 X 10
4
1.34 X 10
4
1.44 X 10
4
2.80 X 10
561
4
5.85 X 10
3
2.26 X 10
3
3.84 X 10
3
4.04 X 10
3
4.27 X 10
4
1.92 X 10
3
2.90 X 10
3
5.15 X 10
3
9.0 X 10
4
2.71 X 10
5
3.84 X 10
4
2.60 X 10
5
5.65 X 10
5
3.74 X 10
4
5.45 X 10
5
4.61 X 10
6
5.95 X 10
6
1.63 X 10
X = 1
66.0
595
131.2
2.45 X 10
708
4
2.12 X 10
6347
6815
4
1.33 X 10
266
4
2.76 X 10
3
1.07 X 10
3
1.82 X 10
3
1.92 X 10
3
2.03 X 10^
3
9.10 X 10
3
1.37 X 10
2920
4269
4
1.29 X 10
5
1.82 X 10
4
1.23 X 10
5
2.67 X 10
5
1.77 X 10
4
2.59 X 10
5
2.19 X 10 .
6
2.82 X 10
5
7.75 X 10
X = 0.1
181
1630
359
6720
1937
4
5.81 X 10
4
1.74 X 10
4
1.87 X 10
4
3.63 X 10
728
4
7.57 X 10
2929
4983
5249
5543
4
2.49 X 10
3762
7990
4
1.17 X 10
4
3.52 X 10
5
4.98 X 10
4
3.38 X 10
5
7.31 X 10
5
4.86 X 10
4
7.08 X 10
5
5.98 X 10
6
7.71 X 10
6
2.12 X 10
x = 0.01
1590
4
1.43 X 10
3150
4
5.90 X 10
4
1.70 X 10
5
5.11 X 10
5
1.53 X 10
5
1.64 X 10
5
3.19 X 10
6390
5
6.65 X 10
4
2.57 X 10
4
4.38 X 10
4
4.61 X 10
4
4.87 X 10
5
2.19 X 10
4
3.30 X 10
4
7.02 X 10
5
1.03 X 10
5
3.09 X 10
6
4.38 X 10
2.97 X 10
6
6.42 X 10
g
4.26 X 10
5
6.22 X 10
6
5.25 X 10
7
6.78 X 10
7
1.86 X 10
x = 0.001
1.57 X 10
1.41 X 10
4
3.11 X 10
5.83 X 10
5
1.70 X 10
6
5.04 X 10
6
1.51 X 10
6
1.62 X 10
6
3.15 X 10
4
6.31 X 10
6
6.55 X 10
5
2.54 X 10
5
4.32 X 10 )
5 (
4.55 X 10 >
5 I
4.81 X 10 J
6
2.16 X 10
5
3.26 X 10
5
6.93 X 10
6
1.01 X 10
6
3.05 X 10
7
-1.32 X 10_
6
2.93 X 10
7
6.34 X 10
7
4.21 X 10
6
6.13 X 10
7
5.18 X 10
8
6.69 X 10
8
1.84 y 10
Filter Instrument
x = 2
3.88
53.8
11.9
124
50
4098
2842
1351
2757
62.5
288
403
33.3
329
183
62
95.5
67.5
4
2.30 X 10
1373
4
3.50 X 10
4
2.30 X 10
744
3
8.3 X 10
4
1.49 X 10
3
6.25 X 10
X = 1
1.84
25.5
5.63
58.9
23.7
2181
1348
641
1308
29.6
137
191.1
15.8
156.1
86.8
29.3
45.3
31.9
4
1.09 X 10
651
4
1.66 X 10
4
1.09 X 10
353
3943
7037
2974
x = 0.1
5.03
69.9
15.4
161
64.8
5970
3690
1750
3580
81.1
374
523
43.2
427
237
80.3
124
87. 4
4
2.99 X 10
1780
4
4.54 y 10
4
2.98 X 10
965
4
1.08 X 10
4
1.93 X 10
81-10
x = 0.01
44.1
614
135
1420
569
4
5.24 X 10
4
3.24 X 10
4
1.54 X 10
4
3.14 X 10
712
3290
4590
380
3750
2090
705
1090
768
5
2.62 X 10
4
1.56 X 10
5
3.99 X 10
5
2.62 X 10
8480
4
9.48 X 10
5
1.69 X 10
4
7.15 X 10
x = 0.001
436
6060
1340
4
1.40 X 10
5620
5
5.17 X 10
5
3.20 X 10
5
1.52 X 10
5
3.10 y 10
7030
4
3.25 X 10
4
4.53 X 10
3750
4
3.70 X 10
4
2.06 X 10
6960
4
1.07 X 10
7580
6
2.59 y 10
1.54 X 10
6
3.93 y 10
6
2.58 X 10
4
8.37 X 10
5
9.35 X 10
6
1.67 X 10
5
7. or; x 10
Minimum Detectable
Material Concentration
(mg/m3)
25.359
357.127
57.299
683.348
241.506
33,828.182
20,908.019
9,942.166
20,287.603
110.434
2,124.925
4,806.547
95.578
3,926.227
596.102
99.743
175.435
108.657
79,962.4
11,939.34
110,643.648
66,254.56
5,364.188
54,204.604
22,908.127
-------
Table 7 (continued)
Material
A12°3
PbO
HgS04
PbCl
2
HgCl
2
RA
PbS(>
4
Line
378
415
69
86
138
279
370
410
495
5801
588)
660
987
1043
1125
1180
1342
1670
86
156
70
121
312
380
750
1250
720
1022
1052
443
608
640
978
1065
1165
o
Normalized Detection Time (Ta~)
Spectrometer Instrument
x = 2
2,04 X 10
1.37 X 10
17.0
8.75
1.67
6.85
68.7
787
6.25
65.8
44.7
11.1
10.1
106.4
59.8
2394
3.80 X 10
1059
7.25
8.45
13.98
3.60
152.4
3.45 X 10
5.4 X 10
173
30.2
16.9
106
586
1642
27.7
542
308
x = 1
4
9.68 X 10
6.48 X 10
8.07
4.15
0.793
3.25
32.6
373
2.97
31.3
21.2
5.29
4.80
50.5
28.4
1136
1 . 80 X 104
502
3.44
4.01
6.63
1,70
72.27
1.64 X 10
2.55 X 10
82.0
14.3
8.04
50.3
278
779
13.1
258
146
X = 0.1
2.65 X 10^
1.77 X 10
22.1
11.4
2.17
8.89
89.2
1020
8.14
85.6
58.0
14.5
13.1
138
77.6
3110
4.93 X 10
1375
9.42
11.0
18.1
4.66
198
5
4.48 X 10
6.98 X 10
224
39.2
22.0
138
760
2130
36
703
400
X = 0.01
2,33 X 10
1.56 X 10
194
99.8
19.1
78.1
783
8970
71.5
752
510
127
115
1210
682
2.73 X 10
4.33 X 10
1.21 X 104
82.7
96.5
159
41.0
1740
3.93 X 10
6.13 X 10
1970
344
193
1210
6680
1.87 X 104
316
6170
3510
x = 0.001
2.30 X 10
1.53 X 10
1910
985
188
771
7730
8.85 X 104
706
7420
5030
1250
1140
1.20 X 104
6730
2.69 X 10^
4.27 X 10
1.19 X 105
817
952
1570
404
1.71 X 10
3.88 X 10
6.05 X 10
1.95 X 10
3400
1910
4
1.19 X 10
6.59 X 104
1.85 X 105
3120
6.0 X 10
3.47 X 104
Filter Instrument
x = 2
4
4.89 X 10
1.93 X 10
1.76
0.69
0.057
0.114
1.99
158
0.270
7.2
4.23
0.58
0.152
13.7
1.24
233.8
9.2 X 10
243
0.0473
0.73
0.75
0.087
9.05
2.06 X 10
3.22 X 10
1.48
3.96
1.84
5.35
38.3
151
0.64
3.39
4.10
x = 1
2.32 X 104
911.6
0.834
0.328
0.0271
0.0542
0.941
75.0
0.128
3.43
2.01
0.273
0.0722
6.47
0.587
111
4370
115
0.0225
0.3-14
0.355
0.0412
4.29
9763
1.53 X 104
0.703
1.88
0.870
2.54
18.2
71.6
0.304
1.61
1.94
x = O.i
6.34 X 104
2495
2.28
0.897
7.42 X 10
0.148
2.57
205
0.350
9.38
5.49
0.747
0.198
17.70
1.61
303
1.20 X 10
315
6.15 X 10
0.942
0.972
0.113
11.7
2.67 X 10
4.18 X 10
1.92
5.13
2.38
6.95
49.7
196
0.831
4.40
5.32
x = 0.01
5.57 X 105
2.19 X 10
20.0
7.88
0.652
1.30
22.6
1800
3.07
82.3
48.2
6.56
1.74
156
14.1
2660
1.05 X 10
2760
0.54
8.28
8.53
0.989
103
5
2.35 X 10
3.67 X 10
16.9
45.1
20.9
61.0
•137
1720
7.30
38.6
46.7
x = 0.001
5.50 X 106
2.16 X 10
198
77.8
6.43
12.8
223
4
1.78 X 10
30.3
813
476
64.7
17.1
1540
139
2.63 X 10
1.04 X 10
4
2.73 X 10
5.33
81.7
84.2
9.76
1020
6
2.32 X 10
3.62 X 10
167
445
20(5
602
•1310
1.70 X 10
72
381
461
Minimum Detectable
Material Concentration
(mg/m^)
170,195.2
16,718.744
9.856
3.873
0.320
0.637
11.101
695.007
1.027
27.563
16.138
2.193
0.58
52.888
4.712
3,763.211
148,155.236
3,525.21
0.163
2.329
2.400
0.278
29.361
179,053.42
112,240.8
3.063
8.161
3.778
19.558
141.972
636.765
2.339
12.378
14.977
-------
Table 7 (concluded)
Material
(NH ) SO
4 4
2
CuCl 2H O
CdCl
2
CdSO
4
Line
450
613 •
623)
976
1090
1420
1650
3150
215
2361
}
249 J
407
700
218
320
1585
3470
415
450
497
603 I
615 >
658)
>
670 /
1000
1050 J
1063 )
1100 1
1118 /
1168)
1173 >
1554 /
2900-
3500
2
Normalized Detection Time (Ta )
Spectrometer Instrument
x = 2
446
643
78
2114
3319
4283
895
100
437
902
4411
242
4780
4
2.47 X 10
489
2550
4
1.26 X 10
1392
1982
4
1.06 X 10
123
1039
2053
1843
3
5.2 X 10
x = 1
212
305
37.1
1002
1574
2031
422
47.6
207
428
2092
114.6
3
2.26 X 10
4
1.17 X 10
232.1
1207
5970
660
940
5055
58.4
493
974
874
2466
X = 0.1
579
834
101
2740
4310
5560
1160
130
566
1170
5730
313.7
6197
4
3.21 X 10
635
3300
4
1.63 X 10
1810
2570
4
1.38 X 10
160
1350
2660
2390
6750
x = 0.01
5090
7320
891
4
2.41 X 10
4
3.78 X 10
4
4.88 X 10,
4
1.02 X 10
1140
4980
4
1.03 X 10
4
5.03 X 10
2755
4
5.44 X 10_
5
2.82 X 10
5578
2.90 X 104
5
1.44 X 10
4
1.59 X 10
4
2.26 X 10
5
1.22 X 10
1400
4
1.18 X 10
4
2.34 X 10
4
2.10 X 10
4
5.93 X 10
X = 0.001
5.02 X 10
4
7.23 X 10
3
8.80 X 10
5
2.38 X 10
5
3.73 X 10
5
4.82 X 10
5
1.00 X 10
4
1.13 X 10
4
4.91 X 10
5
1.02 X 10
5
4.96 X 10
4
2.72 X 10
5
5.37 X 10,,
6
2.78 X 10
5.51 X 10
2.86 X 105
6
1.42 X 10
5
1.57 X 10
5
2.23 X 10
5
1.20 X 10
4
1.39 X 10
5
1.17 X 10
5
2.31 X 10
5
2.07 X 10
5
5.85 X 10
Filter Instrument
x = 2
26.9
13.7
1.83
9.6
24.1
32.5
1.75
7.4
9.65
48.8
48.5
34.6
339
3
3.87 X 10
2.52
363
613
146
54
436
3.35
11.6
46.9
40.7
10.2
X = 1
12.8
6.50
0.87
4.55
11.4
15.4
0.828
3.51
4.58
23.2
23.0
16.4
160.9
1836
1.19
172
291
68.9
25.7
207
1.59
5.51
22.3
19.3
4.83
31 = 0.1
34.9
17.8
2.38
12.5
31.2
42.1
2.26
9.60
12.5
63.4
63.0
44.3
440
5026
3.27
471
795
189
70.4
566
4.35
15.1
61.0
52.3
13.2
x = 0.01
306
156
20.9
109
274
370
19.9
84.3
110
557
553
394
3867
4
4.41 X 10
28.7
4130
6980
1660
618
4970
38.2
132
535
464
116
x = 0.001
3020
1540
206
1080
2710
3650
196
832
1090
5500
5460
3890
4
3.82 X 10
5
4.36 X 10
283
4.08 X 10*
4
6.89 X 10
4
1.64 X 10
6100
4
4.90 X 10
377
1310
5280
4580
1150
Minimum Detectable
Material Concentration
o
(mg/ra )
28.381
14.469
1.911
10.109
25.413
34.317
1.818
10.376
13.718
69.464
68.966
83.615
3,414.62
38,963.592
6.006
4,227.003
7,151.5
464.479
151.877
5,087.15
9.265
32.44
131.479
114.031
28.508
00
-------
values that increase detection time; smaller concentrations produce lower
Raman signals that also increase detection times. The accuracy, a, is
2
in percent and the normalized detection time, Ta , can be interpreted
directly as a time with an accuracy of 1%. At this accuracy level, the
detection times vary from about 0.03 s for PbO, a clearly detectable
4
material, to 1.66 x 10 s (4.6 hours) for CaF , a clearly unreasonable
£
material to monitor.
69
-------
VI LABORATORY MEASUREMENT CAPABILITIES
A. General
The Raman effect may prove useful for particulate monitoring even
as a laboratory tool in which collected samples are analyzed. For
example, the sulfate content of a collected sample of particulate may
be determined using a laboratory Raman instrument, because of tne dis-
tinctive response of the sulfate ion. In general, the performance level
that could be expected from a laboratory Raman instrument is identical
to the performance that was discussed in Section IV on Raman measure-
ments. About 50% of the materials analyzed on this project had mea-
surable Raman responses; it appears likely, then, that such a laboratory
instrument could provide useful information about particulate composition.
B. Measurable Material Properties
It is potentially possible to monitor sources of the sulfates,
chlorates, carbonates, nitrates, phosphates, fluorides, chlorides, oxides,
and sulfides of various metals such as lead, mercury, cadmium, zinc,
copper, nickel, iron, chromium, vanadium, calcium, aluminum, and beryl-
lium. More complex substances such as ammonium compounds and various
organic materials are of interest and can also yield Raman spectra.
The Raman spectra of solids may be classified, for the present pur-
poses, into two categories. In the first, there are vibrating molecular
subunits such as amonium, sulfate, phosphate, or carbonate ions, or water
of hydration. In the second category, the entire solid is one vibrating
unit. In the first case, the presence of different anions (i.e., calcium
carbonate vs. sodium carbonate) does not perturb the Raman shifts and the
71
-------
latter maintain their character, to first order, even though crystal
structures and lattice spacings may change. In the second case, however,
the crystal structure determines the Raman spectrum and thus there is no
guarantee that if cadmium sulfide has a Raman spectrum, lear1 sulfide will
also have one. Even if these two sulfides do have spectra, the spectra
need have no relationship to each other except, of course, if they have
similar crystal structures. Table 8 lists the typical Raman shifts of
some molecular ions.
It will be recalled that, as a general rule, the heavier elements
such as lead and mercury were observed to have higher scattering cross
sections than lighter elements such as aluminum. These heavy metal com-
pounds will thus be more easily detected by both laboratory and in-situ
instruments.
C. Instrumental Considerations
For these laboratory-type measurements, the use of a conventional
spectrometer, in which a slit is swept over the desired spectral region,
satisfies the basic measurements requirements. However, if the time
taken for completing a sweep (typically 15 to 30 minutes under favorable
SNR conditions) is unacceptable, additional instrument sophistication
may be necessary. The time taken to scan a spectrum depends on the
acceptable SNR. In order to maintain the same SNR with a decrease in
overall measurement time, it is necessary to increase the exposure time
per spectral interval. Another reason for decreasing the measurement
time is that changes in sample composition may occur during a scan
period. Thus it is desirable to look simultaneously at as much of the
entire spectrum as is possible.
There are two different techniques available for reducing the mea-
surement time. The first technique uses an optical multichannel analyzer
in which a special-purpose vidicon tube replaces the moving slit arid
detector. The resolution elements of the tube divide the entire spectrum
72
-------
Table 8
RAMAN SHIFTS OF MOLECULAR IONS
Ion
SO
4
CIO
4
PO
4
NH
4
CO
3
NO
3
SO
3
Raman Shifts
981
935
935
3033
1063
1050
1069
451
462
363
1685
652
1104
1102
1082
3134
1415
1390
1330
613
628
515
1397
680
7^.0
532
into separate channels, and photon counts are accumulated simultaneously
on all channels. Each resolution element is in essence a separate exit-
slit/detector combination. Thus, SNR considerations are identical to
those in a conventional system. The advantage of the system is a
reduction of scan time by a factor of N if there are N spectral
resolution elements.
In the second technique, called multiplex spectroscopy, specially
coded rrjipks (combinations of slits at different spectral positions) are
success? '"--1" placed at the exit plane of the spectrometer. The light
passed by the masks is recombined and is incident on a single conven-
tional phototube. Since a large number of slits (say, n) are open at
any one time, the SNR is improved by a factor of n (for the case where
the count rate through each slit is the same) for the same counting
time. The output of the detector is now, however, coded by the position
73
-------
of the slits on the masks. A decoding procedure (usually done by a com-
puter) is necessary to develop a spectrum. The increased SNR is achieved
at the expense of additional complexity in signal processing.
The advantages of these two techniques may be quantified as follows.
.First, assume that the dispersing instrument is the same for each tech-
nique and the resolution limitations are not due to this device. Further,
the detectors are assumed to have the same characteristics in all cases.
Dark-current noise is neglected and Poisson statistics are assumed (i.e.,
the SNR with P counts is
If there are N resolution elements in the spectrum and T is
the integration time needed to obtain a given SNR, S, in the weakest
spectral element, then for a conventional spectrometer the time needed
to generate a spectrum is NT. For the optical multichannel analyzer,
the time needed to achieve an SNR of S in the channel with the lowest
count rate is T. In the multiplex case where n (< N) slits are open
at any one time, the count rate is thus at least n times as large and
an SNR of S can be achieved in a maximum period of T//n. However,
in order to cover the entire N spectral-resolution elements, we need
to repeat the measurement N times. Thus the total time needed to
achieve S over the entire spectrum is NT/v/n.
It is thus seen that a significant advantage in either time or SNR
(or compromise combinations) may be achieved with these techniques, at
the expense of additional complexity in equipment. It should be noted,
however, that the state of the art in these techniques is such that a
scanning spectrometer has superior resolution when a large spectral band
has to be examined.
The additional speed advantage of these two advanced techniques
would probably only be required in a large central laboratory facility
in which a large number of samples from a wide region would be processed.
It does not appear that the more sophisticated techniques would be re-
quired for regional air pollution areas.
74
-------
VII AEROSOL MEASUREMENTS
An attempt was made to observe the fluorescent spectra of the more
highly fluorescent materials in particulate form even though this was
not called for in the contract work statement. The initial measurement
attempt was oriented toward detecting a fine particulate stream in air,
utilizing a closed system in which the particulate stream was pumped
through the spectrofluorimeter. Several particulate pumping systems
were considered, and the best readily available system was tried in the
laboratory. Unfortunately, contamination of the particulate air stream
and clumping of the particulates was sufficiently severe to prevent
meaningful measurements of this type to be made.
The second attempt intended to utilize a liquid particulate sus-
pension medium, rather than air; to relieve the problems encountered in
a flowing air system. Because of the pump-contamination difficulties
encountered in the first system, it was decided to abandon flowing sys-
tems and to use relatively thick viscous liquids in order to suspend the
particulate matter for a sufficient length of time to make meaningful
measurements. Several of the more fluorescent materials were found to
be relatively insoluble in alcohols. Several alcohols were examined for
fluorescent interferences. These ranged from methyl alcohol, which
would still require a pumping system to maintain suspension, to tert-
butyl alcohol, which is a solid at room temperature. It was found that
certain batches of glycerol were low enough in fluorescence to permit
some measurements to be made, and at the same time were sufficiently
viscous to permit suspension of the particulate material for the length
of time required in the measurements. Even though the best available
75
-------
glycerol had relatively low fluorescence, it was still sufficiently high
to mask the extremely small fluorescent return from the particulate mate-
rial at low concentrations. Thus it was not possible to obtain
interference-free fluorescent spectra for any of the desired materials
at concentration levels approximating those that may be found in stacks.
This masking effect from the glycerol fluorescence was not expected,
based on the fluorescence amplitude of the pure glycerol samples.
However, it was noticed that the introduction of particulate material
into the glycerol enhanced the amplitude of the measured glycerol fluo-
rescence by a substantial factor. This resulted in an increasing masking
effect as the particulate concentration was increased in an attempt to
overcome the glycerol fluorescence.
In both the air and glycerol suspension experiments, however, a
component of the fluorescence response of the particulate matter was
detected in the presence of the severe interferences mentioned above.
Thus, even though these crude initial attempts did not demonstrate a
clear measurement capability, both experiments indicate a strong potential
for making such measurements if the interferences can be reduced or
eliminated. Thus, from both the calculations and the initial experiments,
it appears that additional work in this area would result in a demon-
strated particulate-measurement capability.
76
-------
VI11 CONCLUSIONS
The results of the measurements and analysis accomplished on this
project indicate that both fluorescence and Raman in-stack monitoring
systems can yield useful information about the quantity and composition
of a particulate stream. This conclusion is also supported by the results
of a few initial measurements of the fluorescence of particulates at con-
centrations comparable to those expected in smoke stacks. These conclu-
sions are, however, based on the measurements of materials that are input
substances to various industrial processes. The final determination of
feasibility for various industrial-process-monitoring applications will
depend on the existence of significant optical interactions with the
effluents of these processes. Initial measurements made on fourteen
effluent samples provided by the Bay Area Air Pollution Control District
indicate observable fluorescence in at least a few of the samples. Thus,
the monitoring feasibility for effluent materials also appears encouraging
at the present time.
Fluorescent systems are characterized by relatively large optical-
response signals over broad spectral regions. These characteristics
make the analysis of mixed constituents by fluorescence a difficult task,
but would allow a quantitative measurement of a single or known mix of
fluorescent components to be made. Thirteen of the thirty-tour materials
examined on this project had fluorescent responses, and all of these are
expected to be observable in the in-stack particulate stream.
Raman measurements are characterized by a relatively small optical
signal from the material, but this signal is concentrated in a very
77
-------
narrow, specific spectral region. Raman systems are suited to measure-
ments of the relative concentrations of a variety of constituents, and,
although it may be possible to quantitatively determine the concentration
of each of these, this task is made more difficult by the low response
level of the Raman interaction. Twenty-two of the thirty-four materials
investigated on this project have measureable Raman responses. Of the
twenty-two materials with observable spectra, approximately fifteen
appear detectable by an in-stack instrument, four appear marginal, and
three appear unlikely to be detected by such an instrument. Quantitative
detection of these materials by Raman systems is less feasible than for
fluorescence systems because of the low level of the Raman response and
the more critical nature of Raman monitoring instruments. Thus, the
number of materials in each Raman detectability category should be viewed
as a rough estimate and could vary from a low of one or two, to a high
of as much as all twenty-two materials. As for fluorescence, the fea-
sibility assessment will depend on the strength of the Raman interaction
of effluent materials. No measurements of the Raman interactions of the
BAAPCD samples have been made. It is expected, however, that at least
a few of these effluent materials will have Raman spectra comparable to
those measured in the present project, and would thus be detectable.
78
-------
IX RECOMMENDATIONS
The calculations and initial experiments performed on this project
indicate that both Raman and fluorescence systems are capable of detect-
ing materials in particulate form in smoke stacks. This result was more
encouraging than had been expected initially in the project, and as a
result of this potential detection capability two tasks are recommended
as steps toward a prototype in-stack monitoring system.
The first task would be to measure the Raman and fluorescence char-
acteristics of the fourteen samples obtained from the BAAPCD and to
assess their detection feasibility using the methods developed on this
project. These samples do not represent the full range of industrial
effluents for which monitoring would be useful. It is recommended that
at least one effluent sample from each industrial process of potential
monitoring interest be included in this measurement program.
The next task would be to show the experimental feasibility of
detecting an appropriate particulate stream of an actual effluent mate-
rial under laboratory conditions. This experiment would provide direct
measurement information that would require fewer assumptions in deter-
mining in-stack feasibility than is the case with the present, material-
measurement techniques. Particularly for fluorescence monitoring, this
experiment would provide results that would allow much more accurate
prediction of the capabilities of an actual stack-monitoring system.
With minor modifications, the laboratory instrument used in this experi-
ment could conceivably be employed for an actual in-stack measurement.
79
-------
Appendix A
MEASURED FLUORESCENT RESPONSE SPECTRA
This appendix contains the results of the measurements of materials
that have significant fluorescence responses as obtained on the Baird-
Atomic Spectrofluorimeter. The curves shown are taken directly from the
instrument and are not corrected for variations in wavelength response.
They can and should be corrected for the source and detector character-
istics shown in Fig. A-l if quantitative use is to be made of the curves
in this appendix.
The procedure for correcting the measured curves is as follows.
First, determine the excitation wavelength and the relative response at
that wavelength as determined from Fig. A-l. Next, determine from Fig. A-l
the wavelength of the fluorescent response and its relative response.
The two relative-response numbers are multiplied together and the fluo-
rescence response amplitude in the measured curve is divided by this
factor. This procedure will result in a uniform quantitative response
level for all wavelengths from 220 to 700 nm. It should be noted that
this procedure does not provide an absolute response level but is de-
signed only to make the response at different wavelengths uniform,
taking into account the variation in source and detector performance in
the instrument.
Several features of the curves in this appendix are worth noting.
First, in Fig. A-7 for HgSO , note the wide variation in wavelength
between the excitation and fluorescence response curves. This was the
widest separation between these curves for any material investigated on
this project. In Fig. A-9, for the EPA coal sample, note the particularly
wide excitation and fluorescence response curves for that material. This
81
-------
is to be expected, based on the probable composition consisting of a
large number of organic components. Each of these components is ex-
pected to fluoresce at a slightly different wavelength, thus yielding
a broader curve than would be observed for single constituents. Figures
A-ll, A-12, and A-13 for EPA copper, EPA fly ash, and EPA lead, respec-
tively, are included to show examples of materials that do not have
significant fluorescence responses. In these three figures, the smooth
baselines represent the scattered-light response of the instrument,
and it is evident that no fluorescent responses are visible. Figure A-14
shows the response of suspended particulate A1F in water. The water has
O
been thickened in viscosity by the addition of methyl cellulose, which
is slightly fluorescent, but less so than the A1F . The excitation and
O
fluorescence responses shown in this curve are a composite between the
low concentration of A1F in particulate form and the fluorescence of
O
the methyl cellulose. Note the decrease in amplitude between the two
excitation spectra labeled Run 1 and Run 2. This decrease in amplitude
was the result of particulate settling during the time required to make
this measurement. The two large peaks around 375 nm are the Raman
response of the water. Note that the Raman response amplitude is changed
by the amount of particulate material suspended in the water. This
effect was noticed for other materials as well. The level of fluores-
cence that is measured here is clearly lower than the Raman response for
the water. Although there is some enhancement in this response due to
the suspended particulate material, this response level would be typical
for an aerosol measurement.
82
-------
UJ
w
z
O
a.
w
ai
cc
'Jj
cc
0.001
0.01
200 300 400 500
WAVELENGTH — nm
600 700
SA-2039-13
FIGURE A-1 RELATIVE RESPONSE FOR BAIRD-ATOMIC SPECTROFLUORIMETER
SOURCE AND DETECTOR
83
-------
200
u.
200
300
400 500
WAVELENGTH — nm
600
700
SA-2039-14
FIGURE A-2 FLUORESCENT RESPONSE OF AIF3
300
400 500
WAVELENGTH — nm
600
700
SA-2039-15
FIGURE A-3 FLUORESCENT RESPONSE OF CuSO4
84
-------
200
300
400 500
WAVELENGTH — nm
600
700
SA-2039-16
FIGURE A-4 FLUORESCENT RESPONSE OF CRYOLITE
200
300
400 500
WAVELENGTH — nm
600
700
SA-2039-17
FIGURE A-5 FLUORESCENT RESPONSE OF AI2(SO4),
85
-------
Ol
(O
§2
a.
c/>
ai
oc
o
z
LLJ
UJ
tr.
o
FL 405
200
300
400 500
WAVELENGTH — nm
600
700
SA-2039-18
FIGURE A-6 FLUORESCENT RESPONSE OF EPA RAW ALUMINA
1.0
0.8
III
to
I 0.6
w
ui
-------
0.1
0.08
CO
£ 0.06
co
ID
tr
H 0.04
UJ
LU
tr
§ 0.02
200
300
400 500
WAVELENGTH — nm
600
700
SA-2039-20
FIGURE A-8 FLUORESCENT RESPONSE OF EPA ZINC SMELTER FEED MATERIAL
0.08
300
400 500
WAVELENGTH — nm
600 700
SA-2039-21
FIGURE A-9 FLUORESCENT RESPONSE OF EPA COAL — SOURCE, NBS
87
-------
300
400 500
WAVELENGTH — nm
600
700
SA-2039-22
FIGURE A-10 FLUORESCENT RESPONSE OF EPA PHOSPHATE ROCK FEED MATERIAL
30
oc
25
I 20
15
til
o
DC
O
10
200
300
400 500
WAVELENGTH — nm
600
700
SA-2039-23
FIGURE A-11 FLUORESCENT RESPONSE OF EPA COPPER SMELTER FEED MATERIAL
88
-------
25
w ?0
I
CO
HI
CC 15
m
O
2
LU
o 10
CO
tu
OC
O
3 5
LL
o
200
EX 250
300
400 500
WAVELENGTH —nm
700
SA-2039-24
FIGURE A-12 FLUORESCENT RESPONSE OF EPA i=LY ASH
25
20
• t
a.
S 15h-
LU
O
10
o
CO
Ul
CC
§ 5
200
EX 300
EX 250
300
EX 350
I
400 500
WAVELENGTH — nm
600 700
SA-2039-25
FIGURE A-1J FLUORESCENT RESPONSE OF EPA LEAD SMELTER FEED MATERIAL
89
-------
300
400 500
WAVELENGTH — nm
600 700
SA-2039-26
FIGURE A-14 FLUORESCENT RESPONSE OF PARTICULATE AlFg IN WATER
90
-------
Appendix B
MEASURED RAMAN SPECTRA
This appendix contains the measured spectra of the materials in-
vestigated on this project. The curves are reproduced from the actual
instrument traces.
91
-------
2000
1600
1200
WAVEN UMBERS — cm
800
-1
400
SA-2039-27
FIGURE B-1 RAMAN RESPONSE OF HgSO4
92
-------
2000
1600
1200
WAVENUMPERS — cm
800
-1
400
SA-2039-28
FIGURE B-2 RAMAN RESPONSE OF PbSO4
93
-------
rO.3 x 10"6 A-
z
UJ
UJ
>
UJ
CC
2000
1600
1200
WAVENUMBERS — cm
800
-1
400
SA-2039-29
FIGURE B-3 RAMAN RESPONSE OF CdSO4
94
-------
4800
4400
4000 3600
3200
2800
2400 2000
1600
1200
WAVENUMBERS — cm
-1
1 x 10~6 A
800
400
SA-2039-30
FIGURE B-4 RAMAN RESPONSE OF AI2(S04)
-------
0.3 x 10"6 A
2000
1600
1200
WAVBNUMBERS — cm
800
-1
400
SA-2039-31
FIGURE B-5 RAMAN RESPONSE OF AI2(SO4) (6471 A)
o
96
-------
CO
IU
H
Ul
>
01
tr.
-6 x 10~6 A-
2000
1600
1200
WAVENUMBERS — cm
800
-1
400
SA-2039-41
FIGURE B-6 RAMAN RESPONSE OF HgCI2
97
-------
00
4800
4400
4000
3600
0.1 x 10"6 A
I
I
3200 2800
WAVENUMBERS — cm
2400 2000
-1
1600
1200
800
400
SA-2039-32
FIGURE B-7 RAMAN RESPONSE OF CdCI,
-------
RELATIVE INTENSITY
VD
C
30
m
CD
oo
>
2
>
30
m
co
-a
O
•z.
CO
m
o
c
O
NJ
m
z
c
CD
m
33
M
O
CO
(0
I
u
u
-------
LU
I-
UJ
>
<
UJ
oc
5 x 10~6 A-
T
2000
1600
1200
WAVENUMBERS —cm
800
-1
400
SA-2039-34
FIGURE B-9 RAMAN RESPONSE OF PbO
100
-------
4800
4400 4000
3600 3200 2800 2400 2000
WAVENUMBERS —cm"1
1600
1200
800
400
SA-2039-35
FIGURE B-10 RAMAN RESPONSE OF CdS
-------
4800 4400
4000
3600
3200
2800
2400
2000
1600
1200
800
400
WAVENUMBERS — cm
-1
SA-2039-36
FIGURE B-11 RAMAN RESPONSE OF CaF2
-------
,0.21 x 10 6 A-
o
u>
WAVEN'IMBERS —cm"
SA-2039-37
FIGURE B-12 RAMAN RESPONSE OF AIF3
-------
4800
4400 4000 3600 320C
2800
WAVENUMBERS — cm
2400 2000
-1
1600
1200
800
400
SA-2039-38
FIGURE B-13 RAMAN RESPONSE OF EPA PHOSPHATE ROCK FEED MATERIAL
-------
0.3 x 10~6 A
6000 5600 5200 4800 4400
4000 3600 3200 2800 2400 2000 1600 1200 800 400 0
WAVENUMBERS — cm~1
SA-2039-39
FIGURE B-14 RAMAN RESPONSE OF EPA ZINC SMELTER FEED MATERIAL
-------
•0.07 x 10~6 A-
oo
z
LU
I-
LiJ
>
LU
DC
2000
1600
1200
WAVENUMBERS — cm
800
-1
400
SA-2039-40
FIGURE B-15 RAMAN RESPONSE OF EPA TRIPLE SUPER-PHOSPHATE STORAGE
PRODUCT (6471 A)
106
-------
-5x10° CPS
01
I-
01
3
01
QC
2000
1600
1200
WAVENUMBERS —cm
800
400
-1
SA-2039-42
FIGURE B-16 RAMAN RESPONSE OF EPA COAL — SOURCE, NBS (6471 A)
-------
o
00
tn
uj
I-
IU
QC
-1 x 10° CPS-
2000
1600
1200
800
WAVENUMBERS — cm
-1
400
SA-2039-43
FIGURE B-17 RAMAN RESPONSE OF EPA COAL — SOURCE, NBS
-------
•1 x 10~6 A-
co
ui
I-
ui
LU
DC
L
2000
1600
1200
WAVENUMBERS — cm
800
-1
400
SA-2039-44
FIGURE B-18 RAMAN RESPONSE OF NAPTHALENE
109
-------
REFERENCES
1. W. H. Melhuish, J. Opt. Soc. Am,„ /ol. 54, p. 183 (1964).
2. E. H. Gilmore et al., J. Chem. Phys., Vol. 20, p. 829 (1952).
3. Y. Kato et al., J. Opt. Soc. Am., Vol. 61, p. 347 (1971).
4. J. G. Skinner et al., J. Opt. Soc. Am., Vol. 58, p. 113 (1968;.
5. F. J. McClung et al., J. Opt. Soc. Am., Vol. 54, p. 641 (1964).
6. V. S. Gorelik et al., in Light Scattering Spectra of Solids, G. B.
Wright, ed. (Springer-Verlag, New York, N. Y., 1969).
7. W. F. Murphy et al., Appl. Spectrosc., Vol. 23, p. 211 (1969).
8. W. R. Fenner et al., J. Opt. Soc. Am., Vol. 63, p. 73 (1973).
9. B. P. Stoicheff, "Stimulated Raman Emission and Absorption Spectro-
scopy," Semi Annual Report No. 7, Department of Physics, University
of Toronto, Toronto, Canada (December 1968).
10. G. E. Devlin et al., Appl. Phys. Letters, Vol. 19, p. 138 (1971).
11. C. R. Pidgeon et al., J. Opt. Soc. Am., Vol. 54, p. 1459 (1964).
110
-------
BIBLIOGRAPHIC DATA '• ReP°« No- 2-
SHEET EPA-R2-73-219
4. Title and Subtitle
Feasibility Study of In-Situ Source Monitoring of Particulate
Composition by Raman or Fluorescence Scatter
7. Author(s)
M. L. Wright and K. S. Krishnan
9. Performing Organization Name and Address
Stanford Research Institute
Menlo Park, California 94025
12. Sponsoring Organization Name and Address
Environmental Protection Agency
Research Triangle Park
North Carolina 27711
3. Recipient's Accession N ->.
5- Report Date
June 1973 (D/A and D/I)
6.
8. Performing Organization Rept.
No.
10. Project/Task/Worlc Unit No.
SRI Project 2039
11. Contract /Grant No.
68-02-0594
13. Type of Report & Period '
Covered
FINAL June 72 thru Apr '
14.
15. Supplementary Notes
16. Abstracts
The purpose of this project was to assess the feasibility of in-stack monitoring
of an air-suspended particulate stream by fluorescence or Raman optical interactions.
The study explored the feasibility of two approaches: quantitatively monitoring a
prescribed constituent, and monitoring the relative concentrations of several consti-
tuents simultaneously. Fluorescence-monitoring systems were found suitable for the
second.
The method of approach was to assess the magnitude of the Raman and fluorescence
interaction, and then calculate the detectability of that material for a typical in-
stack system. Thirty-four materials were investigated on the project; thirteen mate-
rials had significant fluorescent responses and twenty-two materials had measurable
(continued on reverse side)
17. Key Words and Document Analysis. 17o. Descriptors
Source monitoring
Particulate
Raman
Fluorescence
Stack monitoring
Monitoring systems
Aerosols
Spectra
17b, Identifiers/Open-Ended Terms
17e. COSATI Field/Group
18. Availability Statement
19.. Security Class (This
Report)
UNCL.
20. Security Class
Page
UNCLASSIFIED
21. No. of Pages
228
22. Price
FORM NTIS-SS (MKV. 1-72)
USCOMM-DC 149B2-P7Z
111
-------
(Abstract-concluded)
Raman responses. When these responses were used to calculate in-stack
detectability, all thirteen materials could be detected by fluorescence
systems (although few could be uniquely identified), and fifteen of the
twenty-two Raman-active materials could be detected by a Raman system.
The use of a laboratory Raman instrument to analyze conventionally
sampled particulates was considered. The primary advantage of this
instrument appears to be the capability for measuring ions—for example,
sulfate.
Finally, a few crude experiments were made to detect the fluorescent
response of a particulate material suspended in a liquid (rather than
air). These measurements showed substantial interference from fluores-
cence by the liquid medium; nevertheless, a component of the particulate
fluorescence was detectable. This experimental result partially verifies
the calculated feasibility of detection by fluorescence.
It is concluded that both fluorescence and Raman in-stack monitoring
systems can yield useful information about the quantity and composition
of a particulate stream. Recommendations are made for additional efforts
toward achieving an operational in-stack monitoring system.
G. !-. u. 1973 - 747-788 /3I3. REGION NO. «
------- |