EPA-650/2-74-094




DECEMBER 1974
Environmental  Protection Technology  Seri
                                                             es

-------
                                        EPA-650/2-74-094
   INFRARED  GAS  FILTER  CORRELATION

INSTRUMENT  FOR IN-SITU  MEASUREMENT

         OF GASEOUS POLLUTANTS
                        by

               D. E. Burch and D. A. Gryvnak

                 Philco-Ford Corporation
                 Newport Beach . CA 92663
                 Contract No. 68-02-0575
                Program Element No. 1AA010
                   ROAPNo. 26AAP
              Project Officer: William F. Herget

              Chemistry and Physics Laboratory
            National Environmental Research Center
          Research Triangle Park , North Carolina 27711
                    Prepared for

           OFFICE OF RESEARCH AND DEVELOPMENT
            ENVIRONMENTAL PROTECTION AGENCY
                WASHINGTON, D.C. 20460

                    December 1974

-------
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
An infrared analyzer employing gas cell correlation techniques has been
designed and constructed to measure the concentrations of carbon monoxide,
nitric oxide, sulfur dioxide, hydrogen chloride, and hydrogen fluoride
in the effluent of stationary sources.  An infrared beam is directed across
the stack to a retroreflector and back so that the instantaneous average
concentration is measured continuously without disturbing the constituents
of the effluent.  A small, removable, fixed-position grating monochromator acts
as a unique optical filter that passes narrow spectral intervals that are
centered at wavelengths where the gas to be detected will absorb.  One grating
monochromator is used for CO and NO, another for S02 and HC1, and a
third for HF.  The useful ranges of concentration times path length, in atm
cm, over which each gas can be measured are:  0.005 to 0.4 to NO;
0.0013 to 0.15 for CO; 0.001 to 4.0 for S02; 0.0003 to 0.2 for HCl and
0.0001 to 0.02 for HF.  The discrimination against other gases in the
effluent is excellent.

This report was submitted in fulfillment of contract number 68-02-0575 by
Philco-Ford Corporation, Aeronutronic Division, under the sponsorship of
the Environmental Protection Agency.
                                    iii

-------
                                  CONTENTS


                                                                      PAGE

ABSTRACT	     ill

ILLUSTRATIONS  	      V

LIST OF TABLES	     vii

ACKNOWLEDGEMENT  .	   viii

SECTION

   1      INTRODUCTION AND SUMMARY	      1

          Background 	      1
          Brief Description of Instrument   	      2
          Definitions, Symbols and Units  	      3

   2      SPECTROSCOPIC PRINCIPLES 	      5

          Correlation in Simple Spectral Models   	      5
          Energy and Voltage Relationships 	     11
          Design Factors 	     18
          Two-Cell  System	     20
          Multiple  Correlation Cell Systems   	     24

   3      INSTRUMENT DESIGN AND PERFORMANCE   	     25

          Optical Layout of Basic  Instrument  	     25
          Instrument  Employing Grating Assembly   	     29
          Spectral  Properties  of the Gases 	     35
         Laboratory Tests  and Calibration Data   	     40
          Field Tests	     47

   4     REFERENCES	     53
                                    LV

-------
                              ILLUSTRATIONS
FIGURE                                                             PAGE

   1     Simplified optical diagram of a basic gas-cell
         correlation spectrometer.                                    6

   2     Plots of transmittance for a model gas illustrating
         the spectroscopic principles of detection.                   8

   3     Plots of transmittance illustrating interference by a
         gas species other than the one being measured.              9

   4     Plots of transmittance illustrating the effect of an
         absorbing gas of species y whose spectrum is not
         correlated with that of species x.                         10

   5     Simplified diagram of a gas-cell correlation spectro-
         meter that employs high-frequency chopping and low-
         frequency modulation.                                      12

   6     Calculated spectral plots of transmittance for a
         model spectrum similar to a portion of a CO band.          14
   7     Plots of calculated TT • T    and T tt « T     versus
             i   ,    .     i „ , L    sam      att    sam              , ..
         sample absorber thickness.                                 16

   8     Plot of calculated A/2 versus sample absorber thickness.   17

   9     Calculated spectral plots of transmittance for a model
         spectrum illustrating the principle of detection of
         the two-cell system.                                       21

   10     Calculated spectral plots of transmittance demon-
         strating the principles of discrimination against
         continuum absorption by the two-cell and the  one-cell
         attenuator systems.                                        22

   11     Optical diagram of a gas-cell correlation instrument
         for across-the-stack measurements.                         26
                                    v

-------
                            ILLUSTRATIONS (continued)


FIGURE                                                                 PAGE

  12        Optical diagram of a retroreflector  assembly used
            to redirect the radiant energy beam back across
            the stack to the gas-cell correlation instrument.            28

  13        Optical diagram of a gas-cell correlation instru-
            ment incorporating a grating assembly.                       30

  14        Optical diagram of a retroreflector  and grid
            assembly.                                                    oo

  15        The grating assembly for S02 and  HCl.                        33

  16        The power  supply unit above  the electronics unit.            34

  17        Photographs  of  the optical assembly.                         35

  18        Spectral curves  of transmittance  of H,0  and of H00 + NO
            between 1862 and 1928 cm'1.                                  37

  19        Spectral curves  of transmittance between 2465 and
            2525 cm'1.                                                   39

 20        Curves  of V1 versus normalized absorber thickness
            for CO  and NO.                                              4!

 21         Curves  of V1 vs normalized absorber thicknesss for
            S02 and HCl.                                                42

 22         Curve of V1 vs normalized absorber thickness for HF.         43

 23        The optics assembly and the retroreflector  mounted
           on the smokestack of the Duke Power Co. at
           Charlotte,  N. C.                                            48

 24        Plots of NO concentration vs  time  for 3 different
           days at the Duke Power Co., Charlotte, N. C.                49

 25        Plot of CO  concentration vs time of day for
           25 June 1974 at  Duke Power Co., Charlotte, N.  C.             50

 26         Plots  of S02  concentration vs time of  day for
           3 different days  at Duke Power Co.  In  Charlotte,  N. C.       52
                                    vi

-------
                               LIST OF TABLES
TABLE                                                                  PAGE

  1       Comparison of the sensitivity and discrimination
          of two-cell and one cell-attenuator system.                   24

  2       Instrument parameters and performance.                        46
                                      vii

-------
                              ACKNOWLEDGEMENT
The authors would like to express their gratitude to two of their co-workers,
Mr. Francis J. Gates and Mr. John D. Pembrook, for their cooperation and
assistance.  Mr. Gates' ingenuity and skills in setting up test apparatus  and
in assembling small components were invaluable.  Mr. Pembrook is responsible
for designing and testing the electronic components.
                                   viii

-------
                                 SECTION 1

                           INTRODUCTION AND SUMMARY
BACKGROUND

     Most commercially available instruments for measuring the concentrations
of gaseous pollutants in the effluent of stationary sources require that a
sample be drawn from a stack or duct through a probe into a separate container
for analysis.  The analysis may be made continuously on-line or at a later
time, possibly in a laboratory.  Such sampling techniques have two major
drawbacks:  (1) concentration changes may occur along the sampling  train; and
(2) concentration gradients may exist in the flow field so that several
samples may need to be taken from different positions in order to determine
the average concentration.  By employing in-situ absorption spectroscopy, the
first difficulty can be eliminated and the second one greatly reduced.  In
absorption spectroscopy radiant energy from a thermal source is transmitted
through the effluent from one side of the duct to the other side.   By compar-
ing the amount of transmitted energy at characteristic wavelengths where a
particular gas species absorbs strongly to that at wavelengths where it does
not absorb, or absorbs weakly, it is possible to determine the instantaneous
average concentration across the flow field.  The effluent flow is  undisturbed,
and the measurement along the path across the duct adequately represents the
average unless there are unusually large gradients in the concentration.
The required wavelength selection for absorption spectroscopy may be achieved
in a variety of ways.  Narrow band interference filters provide a simple,
relatively inexpensive method, but in most cases they do not provide the high
spectral  resolution required for good specificity, or discrimination against
other gas species that may  absorb in the same spectral interval.  Dispersing
instruments employing a prism or grating and a scanning mechanism provide
better spectral resolution, but are more complex and, in turn, more expensive
and more  subject to misalignment.

     A class of instruments employing gas-cell correlation spectroscopy  to
provide good sensitivity and specificity at relatively low cost have been
developed for a variety of  applications.  Gas-cell correlation instruments
of various types have been  discussed in References 1-5.  In one  form,  the
energy beam  transversing the sample  gas is alternately directed  through
either a  correlation cell or an attenuator.  The correlation cell,  also

-------
 frequently called a gas filter, contains the gas species to be measured, and
 the attenuator has nearly constant transmittance at all wavelengths of interest.
 An interference filter restricts the spectral bandpass to a region of absorp-
 tion by the gas species being measured.  The gas in the correlation cell is
 essentially opaque at the wavelengths of maximum absorption for the gas species
 being measured.  Therefore, as the correlation cell and the attenuator are
 alternated in and out of the beam, the energy of wavelengths of strong absorp-
 tion by the gas are modulated.  In the following section, some very simple
 models of absorption spectra are used to describe the spectroscopic principles
 of detection and discrimination of gas-cell correlation spectrometers.  The
 illustrations used are helpful in understanding many of the factors involved
 in designing a gas-cell correlation instrument for a particular application.

 BRIEF DESCRIPTION OF INSTRUMENT

      The primary purpose of the work reported here has been to perform the
 required research and to design and construct a gas-cell correlation instru-
 ment for across-the-stack monitoring of stationary source emissions.  Any of
 the five gases, NO, CO, S02, HF,  or HC1, can be measured over the concentra-
 tion ranges ordinarily of interest for stationary sources.   SO  was originally
 included, but subsequent research indicated that this gas coula probably be
 measured better by methods other than gas-cell correlation spectroscopy.

      Conversion of the instrument from one  gas species to another is relatively
 simple and involves changing a correlation  cell and a bandpass filter.  Chang-
 ing the radiant energy source and/or the detector may also be required, de-
 pending on the gas species involved.   The instrument has been designed for
 versatility,  it is necessarily more complex than an instrument designed to
 use the same  principle of operation for a single gas species on a particular
 type of stack.  In some cases, the stability, the signal-to-noise ratio and
 the ease of operation have been compromised in order to retain the versatility.
 The instrument is  intended primarily as a tool for the Environmental Protection
 Agency to evaluate gas-cell correlation methods for a variety of across-the-
 stack measurements.   By testing it under various conditions with different
 instrument parameters,  valuable information can be obtained for the design
 of simpler,  single-purpose gas-cell correlation instruments and to determine
 the factors  that  limit their performance.

     By combining gas-cell correlation techniques along with dispersion
techniques to carefully control the spectral  bandpass,  it has  been possible
to greatly improve the discrimination and the stability over that  attainable
with either technique used separately.  A small, permanently aligned assembly
employing a grating as the dispersing element is included for  each of  the  5
gases with the spectral bandpass selected to  give optimum discrimination,
stability, and signal-to-noise ratio.  Each assembly uses a multiple-slit
grid in the focal plane of the dispersed  radiant energy to pass  narrow spectral
intervals that contain strong absorption by the gas.  Discrimination and
stability are improved by rejecting much  of the energy at wavelengths  where
the gas does not absorb strongly.  The improved discrimination against 1^0
is particularly important when measuring  NO.  The instrument can be used
with a grating assembly or without  it, as in  the more  conventional manner.

-------
     Stack diameters between approximately 1.5 m and 10 m can be accounted
for with only minor adjustments of a few optical components.   In normal
operation, the stack is "double-passed" by using a retroreflector  on the
side opposite the radiant energy source to reflect the energy back through
the stack to the remainder of the instrument.   However, by moving the source
and some associated optics to the opposite side, it is possible to "single-
pass" the stack.

     The instrument has been tested in the laboratory with synthetic samples
of effluent to provide calibration data at various temperatures and to measure
the interference by other gases typically in stack effluent.   Field tests
have also been performed on the stack of a coal-burning power plant while the
gas concentrations of interest were simultaneously being measured by other
instruments.  The data obtained on the stack and in the laboratory indicate
that gas-cell correlation instruments can be built with adequate sensitivity
and discrimination to monitor any of the 5 gases listed above.  The ultimate
performance of any instrument of this type for across-the-stack measurements
will probably be limited by the uncertainty in determining and maintaining
the "zero-balance" that corresponds to no absorbing gas in the stack.  This
is a much more serious problem for across-the-stack instruments than for
instruments in which the sample area can be evacuated or flushed with a non-
absorbing gas.

 DEFINITIONS,  SYMBOLS AND UNITS

 T       transmittance at a particular wavenumber (or wavelength) that would
         be measured with a spectrometer having infinite resolving power.
         Subscripts_indicate the transmittance of different components or gas
         samples.  T indicates the average transmittance over a specified
         interval.

 T       transmittance of the sample gas being analyzed.
  S cUu
 T       transmittance of the gas species x to be measured in the sample gas.
  x      This  is equivalent to T    if only species x is present in the sample
           --                    sam
         cell.

 T       transmittance of an interfering gas of species y in  the sample gas
  y      that  may produce a false indication of species x.

 T       transmittance of the spectral bandpass filter.

 T       transmittance of the gas of species x at low pressure in a correla-
         tion  cell.

 T       transmittance of an attenuator, assumed to be constant over the
  a      spectral bandpass of interest.

 T       transmittance of the gas of species x at high pressure in a correla-
  H      tion  cell.

-------
 v       wavenumber of radiant energy (cm  ).

 v       wavenumber of the center of an absorption line.  Other subscripts or
         superscripts indicate particular values.  (cm  ) .

 ^       wavelength of radiant energy expressed in micrometers (|j,m) .

 f       carrier frequency (high) (Hz) .

 f       alternator frequency (low) (Hz).
  a

 p       partial pressure of a particular gas species, usually species x (atm) .

 P       total pressure of a gas mixture (atm).

 ^       geometrical path length of the  radiant energy beam through a gas
         sample (cm) .
            j  concentration of a gas species, usually expressed in parts per
         million (ppm) or percent.

 u,  w    pC-  (atm cm), absorber thickness of a particular gas species,  u
         refers to gas in the sample cell and w to gas species x in a correla-
         tion  cell.  One atm cm is  equivalent to 10  ppm meters.

 k       absorption coefficient at  wavenumber v of an absorption line (atm cm)  ,

 S       Pkdv,  intensity of an absorption line (atm   cm"  cm  ) .

 V       voltage component of amplified  detector signal at carrier frequency f
  c                                                                            c
 V       voltage component of amplified  detector signal at alternating
         frequency f .
                    cL

 V'       V  /V  ,  normalized voltage.   Normalization is made so that V' = 1
         when there is 100% modulation of the beam at frequency f  .  See
         Equation (5) and  related text.                          a

 D.R.     discrimination  ratio,  ratio of  the  concentration of an interfering
         gas species  to  the  concentration of the species  being measured that
         produces  the same reading.   This ratio may be positive or negative.

 F        lim V1   Pembrook  Factor  (atm cm)
         u-»0 u

 E        chopped  radiant energy.  Equations  (1)  and (2)  and  related text
         define  E   and E     and explain  the  subscripts.
                 LI      att

M        constant relating radiant energy to transmittances.   See  Equations (1)
         and (2).

A and Z  are defined  and explained by Equations  (3) and  (4)  and related text.

-------
                                 SECTION 2

                           SPECTROSCOPIC PRINCIPLES
CORRELATION IN SIMPLE SPECTRAL MODELS

        Gas-cell correlation spectrometers of the type described here depend
for their sensitivity on the correlation between the structure in the spectrum
of the gas species x to be measured in the sample cell and that of the gas in
a correlation cell.  The desired correlation is achieved by filling the
correlation cell with the same gas species x.  Generally, the spectral band-
pass of the instrument includes several absorption lines so that the gas in
the correlation cell has large  fluctuations  in the transmittance at different
wavenumbers.  Figures 1-4 are based on a simple instrument and a greatly
simplified model spectrum to illustrate the spectroscopic principles of
detection and discrimination in such a way that the reader can easily follow
the mathematics involved.

        An optical diagram of a simple gas-cell correlation spectrometer is
illustrated in Figure 1.  Radiant energy from a source passes  through either
a correlation cell or an attenuator, then through the sample cell and a band-
pass filter to the detector.  The sample cell may be a stack or a closed cell
with windows, or it may simply be an open space such as  an atmospheric path.
The correlation cell is filled with gas species x to be  measured.  The attenu-
ator and the correlation cell are alternately moved into the beam at the
alternator frequency f  .  An ac amplifier measures the component of  the
detector signal at frequency f  that results from a difference between the
amounts of radiant energy on tne detector during the two halves of the alterna-
tor cycle.

        For simplicity, we assume that the transmittance of the bandpass
filter is unity from wavenumber v  to v  +1 cm"  , and is  zero elsewhere.   A
transmission spectrum of the gas in the correlation cell is illustrated  in
the upper-left corner of Figure 2.  The upper-right portion shows the  trans-
mittance spectrum of the attenuator over  the same spectral region.   Note_that
~T  , the average transmittance of the correlation cell, is  made equal to  T    ,
the average transmittance of the attenuator.  When there is no absorbing gas
in the sample cell, the amplifier output  is  zero because the  amount  of radiant
energy incident on the  detector is the same  during both  halves of alternator

-------
       vSOURCE  CORRELATION
                CELL
           BAND-PASS
            FILTER
            X1
        ATTENUATOR
SAMPLE CELL
                                             DETECTOR
                                           AMPLIFIER
                                             &
                                           METER
FIG. 1.   Simplified optical diagram of a basic  gas-cell
correlation spectrometer.

-------
cycle.  A sample of gas of species x in the sample cell  is  represented  by  the
middle panel of Figure 2.  Because the same gas species  exists  in both  the
correlation cell and the sample cell, the absorption occurs over  the  same
spectral interval in both cells.  The difference in the  transmit tance results
from different amounts of gas in the two cells.

        The transmittance of the correlation cell and the sample  in series is
given by TL . Tsam.  In this case, the sample does not affect  the transmittance
because the correlation cell absorbs all of the energy at wavenumbers where
the sample absorbs.  During the correlation-cell half of the alternator cycle,
the energy on the detector is proportional to J TL . Tgam dv,  which is
numerically equivalent to TL . Tsam in the 1 cm"1 interval.  The  lower  right-
hand portion of Figure 2 illustrates the spectrum of the energy during  the
attenuator -ha If of the cycle.  Note that the two halves  are no longer balanced
with the sample cell.  TL . Tsam = 0.50 >Tatt  • T     =  0.375.   The differ-
ence is measured by the a.c. amplifier and meter and is  related to the  concen-
tration of the gas in the sample.  Because Tatt is a constant equal to
                                 Tsam'  ^en there is a P°sitive correlation
between the spectral structures of the gases in correlation cell L and the
sample cell, TL  . Tsam > T£ . T^^, as in the example above.  If TL .  Tsam <
TL  * ^sam' tne correlation is negative.  The two terms are equal when there
is  no correlation.

        Figures  3 and 4 are based on the same correlation cell and attenuator
as  the previous  example and illustrate the interference by two different
absorbing gas species other than species x in the sample.  The gas species
illustrated in the upper panel of Figure 3 absorbs, but at different wave-
length numbers than does species x.  The correlation between the spectra of
the sample and species x is obviously negative with the resulting misbalance
opposite to that when only species x is in the sample.  TL  . Tsam = 0.40 <
Tatt • 'I'sam = 0«45.  If the interfering gas is alone in the sample cell,
as  indicated by  Figure 3, the instrument will indicate a negative amount of
species x.  When both species x and the interfering gas are present, the
indicated concentration of species x is too low.

        The interfering gas of species y illustrated in Figure 4 has two
absorption  lines, one inside the absorption line of species x and one outside.
No  correlation exists between the  spectra of this gas and species x, and
f,  . TJT = 0.4 =  Tatt  • Ty»  The interfering gas decreases,  by the same
fraction, the energy  incident on it during both halves of the cycle, and  the
system  remains balanced.  The lower panels of Figure 4 correspond to a  sample
consisting  of species y, whose transmittance is plotted  in  the upper panel,
and species x, represented by the  middle panel of Figure  2.  We  note  that
T — . Tsam  - Tatt . Tsam = 0.4 - 0.3 =  0.1.  This difference is  less than
tne 0.125 produced by the same amount  of species x, but with no  species y
present  (Figure  2) .   Although the  interfering gas y by itself does not
produce a misbalance, it reduces the misbalance produced  by species x to  0.8
its original  value.   This factor is,  of course, equal  to  TT.

-------
J.
J
H
0
i ,
g
to
H"
0

1,
1
i
co
H


~
—

—
—

—
••-
D

i 1


I 1


I I
^







I
>~ V
i
4J
0



1_
B
(0
u

I
(cm"1)

1 | 1 1




n r
1 1 1 1
0 1
FIG. 2.  Plots of transmittance for a model gas illustrating

the spectroscopic principles of detection.  The average trans-

mittances are as follows:


Upper panels, T  = 0.50 and f    = 0.50.  Middle panel,  T    = 0.75.
               •k             ate                          sam
Lower pane Is, T
                    sam
0.50 and f
                                    att
                                           sam
0.375.

-------
e
cd
CO
H
•
o


to
H
S
n

1

—
—
0
1 i


i i
V
L
|

L

j
'Va

i
01
4-
n
H
0
(cm )
1
cd
01
H
•
4J
4->
cd
H
o

1 I I 1 .
FIG. 3.  Plots of transmittance illustrating interference by a

gas species other than the one being measured.
Upper panels, T    =0.9.  Lower panels, T
               sam                        L
                                               sam
                                             0.40 and
"att
sam
              0.45.

-------
                  o
      I   I    I
                         J	I
                                      4J
                                      «B
                                            I   I   I
                                      I
                                      (0
                                V-V  (cm" )  .


FIG. 4.  Plots of transmittance illustrating the effect of an absorbing
gas of species y whose spectrum is not correlated with that of species x.
The upper panel represents gas species y alcme in the sample cell.  The
combined transmittances of species y with the correlation cell and with
the attenuator are shown in the two middle panels.  The corresponding two
plots in the two lower panels are for a sample consisting of both species
x and species y.
Upper panels, T  = 0.80.  Middle panels, T_
               y
Lower panels, T
T    =0.40 and T
 sam             att
 Tv = 0.40 and T
= 0.30.
                                          att
                                                T  = 0.40.
                                    10

-------
     A similar reduction in the misbalance due to a sample of species x would
be produced by continuum absorption or by dirt on the windows of the sample
cell.  The reduction in the misbalance and in the resulting detector signal
can be accounted for by processing the detector signal in such a way that
the output is proportional to (TL . Tgam - Tatt . T8am)/(TL . Tsam + l'att . T   ).
For example, from Figure 2, we see that this ratio for species x, (0.50 - 0.375)/
(0.50 + 0.375) = 1/7, is equal to the corresponding ratio for the same quantity
of species x plus the uncorrelated species y represented by Figure 4:  (0.40 -
0.30)7(0.40 + 0.30) = 1/7.  The instrument described in the following section
measures a quantity proportional to this ratio.  In most real situations, the
spectral band includes several absorption lines of the gas being measured and
may include several lines of one or more interfering gases.  If the positions
and intensities of the lines of the interfering gas are completely uncorrelated
with those of species x, the interfering gas will behave as the one illustrated
in Figure 4 and not produce a false indication.  In practice, the spectrum of
a gas absorbing within the spectral bandpass will usually have some correlation
either positive or negative, with that of species x and consequently will in-
terfere with the measurement.


ENERGY AND VOLTAGE RELATIONSHIPS

     Figure 5 illustrates a gas-cell correlation spectrometer that is more
complex than the basic one shown in Figure 1.  It is similar in operating
principle to the one built by us for across-the-stack measurements and can be
used with two correlation cells or with one correlation cell and an attenuator.
The alternator consists of correlation cell L and either an attenuator or
correlation cell H along with the optical and mechanical components that alter-
nately direct the beam through them.  The one-cell attenuator system employs
correlation cell L and an attenuator.  In the  two-cell system, the attenuator
is replaced by correlation cell H, which contains the gas species to be
measured at high pressure.  In the alternator, the beam may remain fixed
while the correlation cells  (or attenuator)  move as indicated, or the correla-
tion cells may remain fixed while the beam is alternately directed through
them by moving mirrors.  In either case, the beam passes through the sample
cell during both halves of the attenuator cycle.

     Before passing through the alternator, the energy beam is chopped at a
relatively high frequency f , called the carrier frequency.  The difference
in the energy during the two halves of the alternator cycle modulates the
amplitude of the carrier signal.  Employing the high frequency chopping in
addition to the low frequency modulation makes it possible to measure a
quantity proportional to V /V , where these voltages are proportional to the
components of the detector signal at frequencies f  and f  , respectively.
The ratio V /V  can be related experimentally  to tne gas concentration and
is relatively independent of the source brightness, detector sensitivity,
dirt on windows or mirrors, etc.
                                      11

-------
                     CORRELATION CELL  L
SOURCE
         \
      CHOPPER
  (Frequency f )
                                       SAMPLE CELL
                     ATTENUATOR
                        OR
      	T —' CORRELATION CELL H
    ALTERNATOR I
   (frequency f  )
               cl

              Ref
         I	Ref
o
          OUTPUT
       (Proportional
           to
         Va/V
               SYNCHRONOUS      I
               DEMODULATOR
                                CONTROL
                                                              BANDPASS FILTER
J [
1


1
1








                                                                    D
                                                                         DETECTOR
                                                                  PREAMPLIFIER
                                                               IUNED FILTER (f )
                                                                  AGC AMPLIFIER
                                                                        &
                                                                  SYNCHRONOUS
                                                                  DEMODULATOR
 FIG. 5.  Simplified diagram of a gas-cell correlation spectrometer that employs  high-
 frequency chopping and low-frequency modulation.

-------
     Figure 6 contains several plots of calculated transmittance for a model
spectrum used to illustrate the spectroscopic principles of the one cell-
attenuator system illustrated in Figure 5.  The curves correspond closely  to
those of Figure 2, except that the rectangular-shaped spectral lines of
Figure 2 are replaced by two absorption lines with spacings,  intensities,  and
half-widths typical of a portion of the fundamental CO band.   The curves of
Figure 6 are quite realistic, but the effects of correlation are less obvious
than in Figure 2.  For simplicity, we assume that the bandpass filter passes
only the spectral region between the centers of the absorption lines (T  = 1
for 0 s v - v  54 cm'1, and T  = 0 for all other v).
             3                I
     The radiant energy chopped at frequency f  incident on the detector
during the half-cycle that the correlation eel? L is in the beam is given by

     E  = M pT, T_ T    dv.                                             (1)
      L     .)  f  L  satn

The corresponding energy during the attenuator-half of  the cycle is

     E    = M PT. T  ^ T    dv.                                         (2)
      att     .1  f  att  sam

Although T- is constant in the present example, it is left under the integral
sign so that the equations will be valid  for all  situations.  The  constant M
depends on  source intensity and size, aperture sizes, window  transmittance,
mirror reflectivities, etc.  For simplicity, we assume  that the  spectral
emissivity  of the source and sensitivity  of the detector are  both  constant
over the spectral interval of  interest.   The remainder  of  the discussions
on spectroscopic principles is based on the condition that E  =  E    with
no sample in the sample cell.   (T    =1.)
                                 S cLUl

     The curve in the upper panel of Figure 6  labeled (T    - TL)  represents
the spectral response of the system at frequency  f  .  Sample  absorption where
(T    - T ) >0 produces a positive V  .  On  the other hand, a  gas in the sample
ceftSnth a strong absorption  line wftere  (T    -  TL) 0 would have a negative correlation with the  gas  in
the correlation cell and would  produce a  negative V  .   If  the spectrum of  the
sample gas  is completely uncorrelated with  the spectrum of the  gas in  the
correlation cell, the absorption producing  negative V   balances  that  producing
positive V  .

     The spectrum shown in  the  center  panel  of Figure 6 corresponds to 0.01 atm
cm of CO at 1 atm pressure  in  the sample  cell.  During  the half of the cycle
when correlation  cell L is  in  the beam, the  transmittance  of  the beam that has
passed through the sample  is  given  by  TT  .  T    .   This  quantity is indicated
in the lower panel along with  the corresponding  curve that applies during the
attenuator-half of the  alternator cycle.  The  sample  absorbs  only  a very  small
portion of  the radiant  energy  that  has passed  through correlation  cell L
because little energy remains  at  the wavenumbers  where  the sample  absorbs.
During the  attenuator half  of  the cycle,  the  sample  absorbs a larger portion
of the energy with the  result  that  E  >E    .   A  detector voltage component V
                                    LI   att                                 a
proportional  to  (E   - E    )  appears  as a  modulation  of  the carrier voltage V .


                                       13

-------
   o


   1


   i
   CO
   •ZL
   <
   cm
       1.0
4  0.5
       •  0
              sam
       1.0
       0.5
        0


\

—
i
1
\
\
<>•-•'<
s'
n
i
f- TL.
1
!
1
	 — • 	 ] — — — • - . i_ .__
' v
X Tatf rsam \^
i
T \
Tsam
i
\
,


(

i
i
i
\J
            0
                   1           2           3

                 - v   WAVENUMBER (cm-1)
                     a
FIG.  6.   Calculated spectral  plots of transmittance for a model spectrum

similar  to a portion of a CO  band.  The curves  illustrate the principle of

detection of the one cell-attenuator system.  The  plot of TL is based  on a

correlation cell containing 0.2 atm cm of CO  at 1  atm pressure.
                                 14

-------
     The curves in Figure 7 illustrate the dependence of Eatt an<^ EL on sample
absorber thickness for the spectral model illustrated in Figure 6.  Each pair
of curves correspond to a different absorber thickness w of species x in the
correlation cell; the pressure is maintained at 1 atm.  Because If = 1 over
the region of interest, 0 ^ v - va ^ 4 cm'1, TL . T8am = EL/4M, and Tatt. Tsam
Ea*-«-/4M.  As the sample absorber thickness u approaches zero, 1'sam apPr°aches
unity, and TL . Tsam and Tatt • 'I'sam approach T^

     We define two useful quantities.


              TL - Tatt> Tsam dV = (EL ' Eatt)/M'
    E  -J*  VTL + Tatt > Tsam dV = ( EL + Eatt)/M'

For the spectral model of Figure 6, A is proportional to V  and to the
difference between values plotted in Figure 7 for given values of u and w.
Similarly, Z is proportional to the sum of the same values and to the average
value of V .  Neither A nor £ is easy to measure directly; however, the ratio
can be measured with an instrument of the kind illustrated in Figure 5.  In
order to calibrate the instrument, the radiant energy beam through the attenu-
ator is temporarily blocked so that T    and, in turn, E   , equal zero.  Under
this condition A/Z = 1, and the amplifier gain settings are adjusted so that
the output voltages V  and V  are equal.  We define
                     £L      C

     V = V /V                                                          (5)
           a  c
with the gain settings adjusted as just described.  If the gain settings
remain fixed, it follows that V1 = V /V  = A/I! when the attenuator beam is
unblocked and the voltages are measured.

     An automatic gain control circuit that maintains V  constant can  further
simplify the measurement.  When this is used, V" is proportional to V  , which
is measured directly.  Thus, only one voltage, V , need be read in order  to
determine A/2 .

     Curves based on calculated values of V1 = A/Z are shown  in Figure 8  for
5 values of w,  including the 2 represented in Figure  7.  Recall that  the
calculated curves are based on a portion of a model spectrum  similar  to a CO
spectrum and  that the spectral interval corresponds to the spacing between
two adjacent  lines.  Because the curves of Figure 8 are based  on ratios of
integrated transmittance, it follows that the same curves would be obtained
if the spectral interval of the model spectrum were widened to include
several identical lines with the same spacings.  Such a model  is similar  to
the 8 or 10 strongest lines in either the P branch or the R branch of  the
fundamental CO  band.  We have employed a gas-cell correlation  spectrometer
with a spectral bandpass that included approximately  10 of the strongest  lines
in the R branch of the CO band.  Each experimental curve we obtained was  very
similar to the  one for  the corresponding value of w in Figure  8.  The  differ-
ences between  the experimental and the calculated curves could be explained
                                      15

-------
   1.0
LU

o
    0.5
CO
^

<
oe:
     0
         __   w = 0.1 atm cm
w = 2 atm
                                                               TL' Tsam
                                      1
        0.001
               0.01                  0.1                    1


        SAMPLE ABSORBER THICKNESS     u (atm cm)
          10
   FIG.  7.  Plots of calculated T
                    T   and T
                     sam      att
                                                T    versus sample absorber thickness.
                                                Sain
The
   curves are based on the spectral model similar to CO that is  illustrated in Fig. 6.

-------
       0.001
            0.001
          0.01                0.1                   1
SAMPLE ABSORBER THICKNESS    u (atm cm)
FIG. 8.   Plot of calculated A/£ versus sample absorber thickness.  Each curve  is identified by w, the
absorber thickness of gas  species x in correlation cell L.  T  for each value  of w is:   2 atm cm, 0.461;
1 atm cm,  0.600; 0.5 atm cm, 0.711; 0.2 atm cm, 0.817; and O.I atm cm, 0.875.

-------
 by the difference between the model spectrum and the real CO spectrum and  by
 the shape of the transmission curve of the bandpass filter used in the in-
 strument.  A "rectangular" bandpass in which T  = 0 or 1 has been assumed  for
 the calculations.

      An experimental curve of V  vs u can be obtained without adjusting the
 amplifier gains by the procedure described above.  The shape of the curve  on
 a logarithmic scale is the same regardless of changes in the gain setting,
 which shift the curve vertically.  However, adjusting the gains as outlined
 makes it possible to relate V  directly toA/S.   For small values of u, the
 curves of Figure 8 have unityaslope, indicating  that A/E is proportional to u.
 Furthermore, for a given value of u, A/Z varies  by a factor of only about  1.5
 for variations of a factor of 20 for w.

      If w were further reduced to values much less than those indicated, T
 would no longer approximate zero near the line centers, and A/Z would reduce
 more rapidly.  In many situations,  the  major factor limiting the accuracy  of
 a gas-cell correlation instrument is "zero-drift", the slow variation in the
 signal output when u = 0.   An important instrumental parameter is given by

        = lira — = lim  -&-  (atm cm)"1.                                 (6)
          u-»0  U    ir»0

 For lack of an unambiguous and descriptive name,  we have named F the
 Pembrook Factor after one  of our colleagues,  John Pembrook,  who has done con-
 siderable work in developing methods to evaluate  instrument performance.  For
 example, from Figure 8 we  see that A /T  s-  0.001 when u = 0.001 atm cm and
 w = 1 atm cm.  Thus, F = 1.00 (atm cm)" ,  and a  zero-drift corresponding to
 a misbalance of A/Z » 0.001 produces an error equivalent to 0.001 atm cm of
 the gas being measured.   From Equations (3) and  (4),  we see  that this corres-
 ponds to a zero-drift such that 0.998 as  much chopped radiant energy is trans-
 mitted through the  alternator during one-half of  its  cycle as during the
 other half.   [U-  0.998)7(1 + 0.998)  ~ 0.001D.  It follows that the spectral
 bandpass and correlation cell parameters  should be chosen to produce a large
 value of F if the  sample cannot be  removed so that zero drift can be checked
 and  accounted for  regularly.

      The  slopes of  the  curves of Figure 8  decrease with increasing u.   For
 values  of u  more  than 1  or 2  times w, the  dependence  of V"  on u is so slight
 that  it  establishes  a practical upper limit of u  for  an instrument.

 DESIGN  FACTORS

      From Figures 6,  7,  and  8,  we can establish some  general  "rules of thumb"
 and  factors  to be considered  in determining the optimum spectral bandpass,
 sample  cell  length,  and  correlation  cell  parameters  for a particular applica-
 tion.  The spectral  bandpass  should be  chosen to  include  strong absorption
 lines of  gas  species  x to  be  measured and  a minimum of absorption by other
 gas species  that may  be  present  in the  sample gas.  Interference by other
 gas species  can generally  be  decreased  by  employing a selected  narrow  bandpass.
However,  a lower practical  limit  on  the width of  the  spectral  bandpass  may be
determined by  filters  that  are  available or by detector  noise.   If the  band-
pass  is very narrow,  the signal  output, which is  proportional  to A,  (Equation (3))
 for small u may be less  than  the  detector  noise.

                                      18

-------
     The absorber thickness w in the correlation cell should be great enough
that the gas is essentially opaque at the wavenumber of strong absorption
near the line centers.  The curves of Figure 8 show that by employing a
relatively large w, a larger Pembrook Factor F can be achieved along with an
extension of the dynamic range to larger values of u.  Two factors establish
a practical upper limit on w.  If the correlation cell is opaque over too wide
of a spectral interval near each line center, the "sharp" structure is lost
and discrimination_against other gases may suffer.  Furthermore, as w increases
to the point that T  is small, the signal output resulting from a small u
decreases, and detector noise may become a problem.  The decrease in signal
output with increasing w can be observed in Figure 7 in which A is proportional
to the separation between a pair of curves.  For a given u less than approxi-
mately 0.1 atm cm, A is larger for w = 0.1 atm cm than for w = 2 atm cm.  If
w were increased further, A would continue to decrease, but the ratioA/£
would continue to increase slightly as indicated in Figure 8 because 2 would
also decrease.

     The optimum combination of sensitivity and discrimination are generally
achieved when the pressure of the gas in the low pressure correlation cell is
equal to or somewhat less than the pressure in the sample cell.  This ensures
that the widths of the pressure-broadened lines are approximately the same
in both cells.  We have shown experimentally and analytically  that the per-
formance is seldom affected significantly when the correlation cell pressure
varies from the sample pressure down to less than half of its  value.  If  the
spectral bandpass contains only spectral lines of approximately equal intensity,
the dependence on pressure is less than if the lines cover a wide range of
intensities.  It is understood that as the pressure is varied, the absorber
thickness is adjusted to maintain near opacity near the centers of the strong
lines.  The desired combination of total pressure and absorber thickness  is
achieved by varying the length of the correlation cell or by adding a non-
absorbing gas such as N_ to collision broaden the lines.

     For a  fixed sample cell  length, the range of concentrations  that can be
measured is limited on the low end by noise or by uncertainty  in  the correct
zero reading and on the upper end by the "saturation" indicated by the curves
of Figure 8.  If it is practical, the sample cell length  is  selected to  be
consistent with the range of  concentrations  to be measured.  Two  or more
sample cell lengths may be required  to cover the entire  range  of  concentrations.
In some applications, such as across-the-stack measurements  of stationary
source pollutants, the optical length through the sample  gas may  be difficult
to change.  Double-passing the energy beam  through  a  stack  to  increase the
length may  be practical but  limits on window sizes  and  on the  mechanical
stability of the optical platforms make  it  difficult  to  pass  the  beam across
the stack more  than twice.   If one pass  of  the  stack  produces  too much
absorption, it may be necessary  to select a  different spectral bandpass  in
which the absorption  is weaker.   Another  solution  is  to  use  an optical  arrange-
ment that passes the  radiant  energy  beam across  only  part of the  stack.   This
method does not provide  the  averaging of a  path  that  extends across  the  entire
stack.
                                      19

-------
      Narrowing  the  absorption  lines by  decreasing  the  sample pressure to less
  than 1  atmosphere can  greatly  improve the  discrimination  and sensitivity for
  some gases.  This is particularly  true  for SO-,  whose  absorption lines  are  so
  closely spaced  that they overlap and smooth out  much of the spectral structure
  when the  sample  is  near 1 atm.  At pressures less  than about 0.1 atm, the
  spectral  structure  of  SCL is greatly increased so  that the performance of a
  gas-cell  correlation spectrometer  is improved.   The optimum sample cell length
  and correlation  cell parameters for any application obviously depend on the
  sample  pressure.  The  half-width of a collision-broadened absorption line is
  proportional to  the total gas  pressure  and inversely proportional to the
  square  root of  the  temperature.   Therefore,  the  spectral  lines  of  a hot gas
  in a  smokestack are narrowed by the  higher temperature although the pressure
  is approximately atmospheric.

  TWO-CELL  SYSTEM

      The  two-cell system is illustrated schematically  by Figure 5 when a high-
  pressure  correlation cell H is  used in place of  the neutral density attenuator.
  Figure  9  illustrates the spectroscopic principles of this  system with the same
  low-pressure sample in correlation cell L as in  the case illustrated in
  Figure  6.  The model is based  on correlation cell H filled to 5 atm with CO,
  the gas to be detected.  The absorber thickness of the CO in this cell,  0.108
  atm cm, was adjusted so that fTfTRdv =  jTfTLdv.   Because  of the pressure
 broadening of the absorption lines, the  gas in cell H absorbs less  near the
 centers of the lines,  but  more  in the  wings of the lines,  than does the  low-
 pressure gas in cell L.  The curve  labeled  (T   -  T ) represents  the spectral
 response of the system at  frequency f  .   By comparing Figure 9  with Figure  6,
 we see  that the response |TR -  TL  | ,  of  the two-cell system is  less than
 ' Tatt  " TLr°f the one'ceH attenuator  system in the region midway between the
 lines.   Thus,  the two-cell  system would  discriminate better against an  absorp-
 tion line of an interfering  gas in  this  region.   The two-cell  system is  only
 about  3/5 as sensitive  as  the one  cell-attenuator system because of the
 smaller value  of (TR -  TL) near the centers of  the  lines.

     In matching fT^dv with fTfTRdv , the  pressure of  the gas  in cell H could
 have been made higher with a corresponding  shorter  path to maintain the
 balance  between  the  two beams.   This change would further  broaden the lines,
 increasing TR near the  line  centers and  decreasing  it  in the wings.   As  a
 result,  the spectral response would be greater near the centers  of  the lines.
 However,  this would  be  accompanied  by an increased  negative  response  in  the
 wings of the lines,  causing  the  system to respond more  to  absorption  by  other
 gases with absorption lines  between the  CO  lines.  The  optimum pressure  in
 the cells  depends on the spectra of the  gas species  to  be  measured  and any
 interfering gases in the sample.  In many cases,  the decrease in response of
 the two-cell system  is  more  than compensated for  by  an  increased discrimination
 over the one cell-attenuator system.

     Figure 10 illustrates model spectra for which  the  discrimination by the
 two-cell system is far  superior to that by  the one  cell-attenuator system.
The curves labeled T  and T  in the upper panel correspond to the transmit-
 tances  of  low and high pressure  correlation cells, respectively.  Both have


                                     20

-------
      1.0
o

I
00
      0


     -0.2
                  v-.v    WAVENUMBER(cnrl)
                      3
 FIG. 9.   Calculated spectral plots of transmittance for a model spectrum

 illustrating the principle of detection of the two-cell system.
                                21

-------
     1.0
     0.5
       0
          0
                    v - v
2345

   WAVENUMBER (cm"1)
FIG.  10.  Calculated spectral plots of  transmittance demonstrating the
principles of discrimination against continuum absorption by the two—
cell  and the one-cell attenuator systems.
                              22

-------
same average transmittance, which is represented by the curve labeled T tt«
Thus the two curves T^ and T^ correspond to the two-cell system,  and the
curves TL and Tatt correspond to the one-cell attenuator system.   Both of the
absorption lines have a half-width 01 of 0.05 cm   for the low pressure sample,
and 0.5 cm~l for the high pressure sample.  The intensities of the lines at
(v-va) = 1, and 3 cm"^- are 4 atm'^-cm'-'-cm"-'- and 1 atm~lcm"i'-cm~-'-,  respectively.
The absorber thicknesses of gas x in the low and high pressure cells are 0.5
and 0.24074 atm cm, respectively.  It is assumed that Tf = 1 for 0 < (v-va)
< 6 cm~l and Tf = 0 for all other v.

     Curves A, B and C in the lower panel represent three simple types of
continuum absorption by interfering gases in the sample.  Table 1 summarizes
the relative sensitivities of the two systems to the three types of continuum
and to a small amount of gas species x at the same pressure as the gas in
correlation cell L.  The equivalent error corresponds to the amount of gas
species x that produces the same A/£ as that produced by the interfering con-
tinuum.  Continuum A has constant transmittance and produces no error, whereas
Continua B and C produce signals in the one-cell attenuator system that cor-
respond to a sample containing approximately 0.108 atm cm of the gas x.  The
equivalent error for the two-cell system is only about 2% as much.  Sample B
produces an error that corresponds to a positive amount of gas x because the
general slope of the transmittance curve is similar to that of gas x; i.e.,
the transmittance increases with increasing wavenumber for both gas x and
continuum B.  As expected, the opposite is true for continuum C, whose  trans-
mittance curve slopes in the opposite direction.

     A small amount of gas species x produces a smaller signal in  the two-
cell system than in the one-cell attenuator system; however, the increase in
discrimination may be more important than the  loss in sensitivity.  Similar
results would be obtained  if the continua were replaced by spectra with the
same general slopes, but consisting of many  lines.  Thus,  curves B or  C may
be representative of the spectra near the edge of  an absorption band  of an
interfering gas.  The model spectrum of gas x  in the correlation cells  is
also typical of the edge of many absorption bands  in which the  lines  get pro-
gressively weaker with increasing distance  from the band  center.   As  an
example, interference by the wing of a  C02 band when measuring  CO  is  similar
to the model illustrated in Figure  10.  In  such a  case,  the  discrimination
by the two-cell system can be much better  than  that by  one-cell attenuator
system.
                                       23

-------
                                    TABLE 1

                COMPARISON OF THE SENSITIVITY AND DISCRIMINATION
                   OF TWO-CELL AND ONE-CELL ATTENUATOR SYSTEM
 0.001 atm cm
 of gas

 Continuum A

 Continuum B

 Continuum C
                          Two-Cell
0.0002066


0           0

0.0004778   +0.00231

-0.0004778  -0.00231
                                     One-Cell Attenuator
Sample



A
Z
Equivalent
Error
(atm cm)

A
z
Equivalent
Error
(atm cm)
0.0003203


0

0.03474

-0.03474
0

+0.1084

-0.1084
 MULTIPLE CORRELATION CELL SYSTEMS

      The discussion above has  been  restricted  to  relatively  simple  instruments
 for which the  processing of the  detector  signal is  not  excessively  complicated
 By  using more  than  two  correlation  cells  with  different amounts  of  gas  either  '
 at  different pressures  or at the same  pressure, it  may  be  possible  to obtain
 more information  about  the  sample absorption and  to better account  for  inter-
 fering  gases,  particulate matter, and  variations  in sample temperature  or
 pressure.  In  general a  measurement with  a correlation  cell  containing  a
 small amount of gas  at  low  pressure provides information about the  absorption
 near the  centers  of  the  lines.  Similarly, a correlation cell with  more gas
 absorbs over a wide  interval near the  line centers  and  provides  information
 about absorption  in  the wings of the lines.  An alternate  method is to  use a
 single correlation cell and vary the gas  pressure in it.   Processing the
detector signal for an instrument with more than two correlation cells, or with
one  of varying pressure,  is necessarily more complicated than for two cells or
for  one cell and an attenuator.  The added complication is usually not re-
quired for absorption measurements, but it may be justified  for  passive
measurements in which the hot gas being measured serves as the radiant  energy
source.                                                                    &y
                                      24

-------
                                 SECTION 3

                      INSTRUMENT DESIGN AND PERFORMANCE
OPTICAL LAYOUT OF BASIC INSTRUMENT

     Figure  11 shows  the optical  diagram of  the instrument that was built
 and tested for  EPA.           It employs the principles described above and
can be used to compare different methods of  gas-cell correlation spectroscopy
under a variety of situations.  Different gas species can be measured by in-
terchanging the gas correlation cells, a bandpass filter, and possibly the
source and detector.  The instrument is equipped  with correlation cells and
filters to measure NO, CO, SO,, HC1,  or  HF.  All of the optical components
shown in Figure 11 are mounted on a baseplate approximately 47 cm x 53 cm.
Except for a preamplifier, all of the electronic  components are mounted
separate from the optics.

     In Figure 11, mirrors indicated by an N are  in the entrance section that
directs the beam of radiant energy into the  stack.  A  tungsten filament bulb
in  a quartz envelope  serves as the radiant energy source for wavelengths less
than approximately 3.6 ^m.  Either a Nernst  glower or an electrically heated
nichrome wire covered with a ceramic material serves as a source for longer
wavelengths.  Two sources can be  kept  in place. Mirror Nl is removed or put
in  a place as shown,  depending on the  source being used.  In normal operation,
bypass mirrors Bl, B2, and B3 are not  in place.  Mirror N2 forms an enlarged
image of the  source that overfills aperture  Apl; a further enlarged image  of
the source is formed  on mirror N5, which directs the beam through  a CaF2
window into  the  stack.  A field lens  near aperture Apl  images  mirror N_ on
mirror N  .   A retroreflector  described in the  following paragraph, placed
on  the opposite  side  of the stack directs the beam back through  the stack
and forms  another image of the source  on mirror  XI.  Mirror XI images  mirror
C3  of the  retroreflector  onto aperture Ap2 just ahead  of  the  alternator.   A
rotating mirrored  chopper Al, with  two blades and  two  open sectors, alter-
nately directs  the  beam through one  of two  paths.  One  alternate path  contains
correlation  cell CC, and the  other  contains an  attenuator.   As  discussed
 previously,  the  attenuator may be replaced  by a  high-pressure  correlation cell.
Mirror A5  directs  the beam upward through  a bandpass  filter  to the detector.
Either of  two liquid-nitrogen cooled detectors is  used:  PbS for \^ 4 u.m, and
pbSe for ^  between  4  and  6 u-m.  The  bandpass filter  passes  a spectral band that
 contains  absorption by the gas being measured.
                                       25

-------
  •Window

           X2

  From Stack
  To  Stack
 10 Hz
Mirrored
Chopper
                                435 Hz .
                     Source u   Chopper
              r

             ^Alternate Source
FIG. 11.  Optical diagram of a gas-cell correlation instrument for across-the-stack
measurements.
                                 26

-------
     Figure 12 shows an optical diagram of the retroreflector  assembly,  which
is a modified version of a "cat's eye" reflector.   Mirror Cl forms an image of
mirror N5 on the small spherical mirror C2,   which  is adjusted to correspond
to the distance between Cl and N5.  Mirror C2 images Cl near C3.   Mirrors Cl
and C3 are cut from a single spherical mirror and  mounted so that their
centers of curvature are at the same height but separated slightly in an
azimuthal direction in order to image N5 on XI.  Provided the relative posi-
tions of the three mirrors Cl, C2, and C3 are rigidly fixed, the  position of
the image formed near mirror XI relative to the effective source  near N5 is
insensitive to small angular or linear displacements of the retroreflector
assembly.

     The radii of curvature of the mirrors are chosen to be optimum  for a stack
approximately 3 m in diameter.  For this size of stack, mirrors N2, N3, Cl,
and C3,  aperture  Ap2, and the detector are  conjugate to each other.  The
source is conjugate to Apl, N5, C2, and XI.

     The instrument is very sensitive  to differences  in  the energy passing
through  the  two paths of  the  alternator section.   Therefore,  after the alter-
nator has been balanced with  no absorbing gas  in the  beam,  it  is  very  important
that any changes  in aperture  or angular field  of view affect  both paths  of  the
alternator  in  the same way.   This  is  best accomplished by  ensuring that  the
limits of  the  beam  are well defined outside  of  the alternator and that the
alternator  transmit  to  the detector all of  the  beam entering  it.   The  image
of C3  at Ap2  is  larger  than the  aperture, and the  image  of Ap2 at the  detector
is smaller  than  the  sensitive element.  Therefore, Ap2  forms  the  effective
aperture as long  as  any misalignment  of the  optical components preceding it is
sufficiently small  that Ap2 remains completely filled.   The windows  and  mirrors
in the  alternator section are oversized so  that a  mask in front  of mirror XI
limits  the  angular  field-of-view  of  the detector.   The optical components
preceding  mirror XI  are designed  so  that  the opening in the mask on  mirror XI
is completely filled.

      In order to balance  the  alternator,  the bypass mirrors Bl,  B2,  and B3
 are easily positioned  as  indicated in Figure 11 to bypass the stack and provide
 an optical path with no absorbing gas.  Bypass mirror B3 replaces XI, which
has a different radius of curvature,  and  sits in the same position;  the same
mask is used for XI and B3.   When the bypass optics are in place, Ap2 is con-
 jugate to mirrors N2 and  N3  as it is  when the beam is traversing the stack.
 Similarly, B3 is conjugate to the source and Apl as is XI when it is in place.
Thus,  the instrument has  the  same field-of-view while it is being balanced as
 it has during a measurement.   This method of bypassing the stack in order  to
 balance the alternator matches the aperture and field-of-view about as well as
 is possible without passing the beam through the  smokestack.  Errors due to
 non-uniformities in the transmittance of stack windows or  in the reflectivity
 of mirrors in the entrance section are minimized, but they cannot be eliminated
 completely.  The best balance could probably be achieved by placing a pipe
 across the stack with the radiant energy beam passing through the pipe.  The
 absorption could be eliminated by flushing  the pipe with a non-absorbing gas.
 However, this method of balancing is frequently not  practical because of space
 limitations or because of possible interference with the operation  of the  plant
                                       27

-------
NJ
c»
                                          C3
                               FIG.  12.  Optical diagram of a  retroreflector assembly used
                               to redirect the radiant energy beam back across the stack to
                               the gas-cell correlation instrument.

-------
being monitored.  It seems probable that the performance of gas-cell  correla-
tion instruments for stack measurements will ultimately be limited  by the  un-
certainty in the balance corresponding to no absorption in the stack.

     The radiant energy from the source incident upon the detector  is chopped
at 435 Hz with a 10 Hz modulation that is related to the concentration of the
gas being measured.  The detector signal is amplified and processed in such a
way that the 435 Hz signal serves as a carrier for the 10 Hz signal as indi-
cated by Figure 5.  In this way, the system is insensitive to 10 Hz variations
in the radiant energy on the detector unless the energy is first chopped at
435 Hz.  Thus, emission by hot gases in the stack, or by any optical compon-
ents that follow the 435 Hz chopper, does not influence the instrument output.

INSTRUMENT EMPLOYING GRATING ASSEMBLY

     The stability and discrimination of an instrument of the type  shown in
Figure 11 could frequently be improved if the spectral bandpass could be
controlled more accurately that can be done with an interference filter.
Figure 13 shows another version of the instrument described above with a
grating assembly to provide the desired bandpass.  Either CO or NO can be
studied with the grating assembly shown by blocking the beam at the  appropriate
retro-reflector Rl or R2.  The bandpass filter near the detector eliminates
overlapping orders of shorter wavelengths that are passed by  the grating
assembly.  It is also necessary to change the correlation cell, and  possibly
the attenuator, when changing from one gas  to another.  A second grating
assembly is built  for SO- or HC1, and a third one serves  for HF.  Each
grating assembly is constructed in its own box, and they can be interchanged
quickly with little or no realignment of optical components required.  The
box for each assembly is mounted to the main base plate on a  3-point system
that will not transfer stresses from  the base plate to  the box.  All of the
parts  in the grating assembly are bonded in place with  epoxy  cement  to avoid
misalignment.

     Mirrors Tl, T2, and T3 are bonded to a block that  can be removed or  put
back in place easily.  This transfer  optics assembly  directs  the beam into  an
entrance slit SI that is below  the optic axis of  the  alternator.  The beams
of the selected bandpass exits  from the grating assembly  through slit S2  at
the height of the  optics axis of the  alternator.  Mirror  T3 directs  the beam
into the alternator.  When the  transfer optics  are  installed, it is  necessary
to tilt mirror  XI  to direct the beam  to  the lower level of mirrors Tl, T2,  and
slit SI.  Otherwise, no  adjustments are required when the  transfer optics
assembly is installed or removed.

     Mirror T2  images mirror XI, and  thus  the source, on the  entrance slit  SI.
For a  stack diameter of  3 m, mirror C3  of  the  retroref lector is imaged  on
mirror T2.  As  the diameter changes with  the  resulting change in distance
between mirrors C2 and XI, the  conjugate  to mirror  C3 moves  slightly from
mirror T2.  A mask on the grating  limits  the  field-of-view of the  detector.
                                      29

-------
                                                                  Detector
   Window

           X2

   From Stacl^
                                                    10 Hz.
                                                   Mirrored
                                                   CHopper
                                435 Hz
                     Source "  Chopper

             ^Alternate Source
>'IG. 13.  Optical diagram of a gas-cell correlation instrument incorporating a
grating assembly.
                                    30

-------
     The grating assembly shown in Figure 13 disperses  the radiant  energy  and
directs that of wavelengths near 5.25 p-m toward the retroreflector  Rl.
Energy from a portion of the spectral band selected for the measurement  of NO
is redirected back to the grating, which undisperses it and directs it to
mirror Gl and the exit slit S2.  Similarly, wavelengths near 4.6  y,m fall on
retroreflector  R2, which returns a portion selected to measure CO.  Only
one gas is measured at a time; the retroref lector  not  being used is covered
with a mask.

     A previous paper by one of us(°/ describes the method by which the  radia-
tion is dispersed and a selected portion is undispersed and imaged.  The
original paper described the method for a prism as a dispersing element;
however, the obvious application to a grating and to systems with a "tailored"
bandpass were pointed out.  Decker(7) has also described an application  of the
method to Hadamard spectroscopy.

     A view of  the retroref lector and grid  assembly such as the ones indicated
by Rl or R2 in Figure 13 is shown in Figure 14.  A converging beam of dis-
persed radiant energy from mirror G2 (Figure 13) is incident upon mirror G4
whose reflecting plane is 45° from the vertical.  The horizontal grid is in
the focal plane where monochromatic energy forms an image of the entrance
slit Si.  Openings in the grid, such as the two shown in Figure 14, pass the
selected spectral intervals to mirror G5.  Mirrors G4 and G5 are normal to
each other  so that each ray of the beam returns to mirror G2 parallel to,  but
displaced vertically downward from, the ray before it struck mirror G4.   The
returning beam is completely undispersed by the grating, and the image formed
at the exit slit S2 is undispersed and contains only those wavelengths passed
by the grid in the retroref lector and grid assembly.  Because an  image of
slit Si is  formed at slit S2, only one of these slits is required  to provide
the desired spectral resolution.  Slit S2 is intentionally made somewhat
wider  than  slit SI; the effective part of S2 is only that portion  conjugate
to the narrower SI.  Slit S2 reduces scattered energy reaching the detector;
having it oversize does not degrade the resolution and makes the positioning
of the slits much  less critical.

     A photograph of the grating assembly for SOo and HCl  is shown in
Figure 15.  A single mirror, G2,  reflects the dispersed  energy to  both  retro-
ref lector and grid assemblies.  Two separate mirrors cannot be used  as  G2
and G3 are  for the NO and CO instrument shown in Figure  13.  At mirror  G2
of the grating assembly for S02 and HCl,  the beams  of dispersed  energy  for
the two wavelengths of interest,  4.0 and  3.6 u-m, overlap each  other.  The
two beams near 5.2 and 4.6 (j-m used to measure NO and CO,  respectively,  are
separated enough that two separate mirrors, G2 and  G3,  are  used  to focus
the beams on the retroreflector   and grid assemblies.   By rotating G3 rela-
tive to G2,  the two retroref lector  and grid assemblies  can be placed close
to each other as shown in Figure  13.  The grating  assembly for HF  contains
a single  retroreflector and grid assembly.

     The power  supply unit and  the electronics unit  shown in Figure 16  are
designed for rack mounting and  are connected to  the  optics  assembly by
approximately  15 meters of electrical cables.  Power supplies  and  appropriate
                                      31

-------
OJ
ro
                                                                            G?
                                                                               GRID
                   FIG. 14.  Optical diagram of a retroreflector and grid assembly.

-------
              Slits
FIG. 15.  The grating assembly for S02 and HCl.  The
          two retroreflector  and grid assemblies are
          labeled according to the gas they are used
          to measure.
                           J3

-------
                    |   .
                    «^PP
                                                u.
              t ••*•    t  *
              ..%  *
                                               Q

FIG. 16  The power supply unit  above  the  electronics unit.

-------
switches are provided for three different radiation sources:   a Nernst  glower,
a nichrome-heated ceramic glower, and a tungsten-iodide bulb.  Also  included
in the power supply unit are amperite ballast bulbs for the Nernst glower
and a switch so that either 1, 2, or 3 of the ballasts  can be  connected in
parallel to maintain a current of approximately 0.27,  0.54, or 0.80  amps.
Motors for the two choppers and a heater to start the Nernst  glower  are also
controlled from the power supply unit.

     The amplified detector signal from the preamplifier on the optics
assembly is fed into the electronics unit shown in Figure 16.  All of  the
signal processing indicated by the electronics diagram of Figure  5 takes
place in the electronics unit.  Included on the front panel are gain con-
trols and adjustments for the phases of the reference signals  used in the
synchronous demodulators.  An electronic balance control, not  shown  in
Figure 5, on the front panel of the electronics unit makes it  possible to
properly account for a possible small misbalance between the  two  beams  in
the alternator section when the stack is being by-passed to determine  a
"zero" reading.

     By setting a switch to the appropriate position, the panel meter can be
made to read either of five different voltages:  signal input, carrier level,
signal output, carrier reference, or modulation reference. The  signal input
represents the ac input at all frequencies and makes it possible  to  measure
unwanted "pickup" when the chopped beam is blocked from the detector.   The
carrier level represents the signal at the output of the carrier-frequency
demodulator.  After the instrument has been properly calibrated,  the absorber
thickness of the gas being measured can be determined from the signal output,
which can be read from the panel meter or from a recorder connected  to an
output jack.  Nine different ranges exist for the signal output.   The panel
meter is also used when adjusting the levels of the carrier reference and
the modulation reference signals.  Time constants of 0.3, 1,  and  3 seconds
are available.

     Two photographs of the optics assembly, with and without the cover,
appear in Figure 17.  The chopped radiant energy leaves the optics assembly
before traversing the stack and returns from the stack through the open port
in the front of the cover.  The bracket directly beneath the  opening in the
cover fastens to a flange on the port of the stack.  The detectors are filled
with liquid nitrogen through an opening in the top of the cover;  a cylin-
drical cap covers the opening and protects the detector.  A small fan mounted
above the source housing and visible in the lower photograph  of Figure 17
circulates air and carries away much of the heat from the source.  A pre-
amplifier and a box for the bias battery are mounted on the baseplate near
the front of the instrument and adjacent to the grating assembly.

SPECTRAL PROPERTIES OF THE GASES

     Figure 18 shows two experimental curves of transmittance in the region
of the Q-branch and the R-branch of the fundamental NO band.   The absorption
represented by the upper curve is due to approximately  1.5% H~0 in 8 m of
                                     35

-------
                                               'JflHhb.

                                        ,<^^^l^^^
FIG. 17.  Photographs of the optical assembly.  The  top picture
          shows the assembly with the cover  installed.

                               36

-------
          1920
1910
1900
1890
I860
1870
                                  WAVENUMBER  (cm-1)
FIC. 18.  Spectral curves of transmittance of H20 and of HO + NO between 1862 and 1928 cm
width = 0.35 cm~l.  The transmittance scales of
                                                                 '1
             "the two curves differ by 0.20.
                                               Spectral  slit-

-------
 room air.   The lower curve,  which is  displaced downward  for  clarity, repre-
 sents the  same amount of  H20 and a sample  of  0.03  atm of NO  in  a  10 cm
 sample cell.   The need to reduce interference by H20  is  apparent.  Absorption
 by H20 is  even stronger in the  lower-wavenumber P-branch of  the NO band.  The
 cross-hatch areas correspond to the 6 narrow  spectral intervals passed by
 the retroreflector and grid  assembly  for NO.   Interference by H?0 is reduced
 by choosing the intervals with  a minimum of H^O absorption.  Each interval
 contains a pair of NO lines.

      Blocking  the spectral intervals  where there is little or no NO absorp-
 tion decreases  £ with essentially no  decrease  in A. (See Equations 3 and 4.)
 Thus,  the  Pembrook Factor F  is  increased with  a resulting increase in the
 stability  of the zero balance.   The blocked energy, if allowed  to pass,
 would  provide  no information about  NO concentration,  but it would add to
 any misbalance  of the alternator.   Information on  the positions, intensities,
 and haIf-widths  of the NO lines  can be obtained from  two articles by Abels
 and Shaw.(B'y;

     Two transmittance  curves of  the  S02 band  employed are shown in Figure 19.
 The structure in Curve A  is due  to  "clusters"  of many closely-spaced lines
 that are not resolved.  When  the  total pressure of the sample is greater than
 about  1 atm, the  lines  are broadened  enough that they overlap considerably
 and much of the  structure  in  a spectrum of 1 atm sample that could be observed
 with a very narrow spectral slitwidth is not greatly different from Curve A.
 Further smoothing  of  the  spectrum with increased pressure is illustrated by
 Curve B of Figure  19.  At  the high pressure, the widths of the S02 lines are
 comparable to the  spacings between  the clusters of lines.

      The relatively  "shallow" structure in the spectrum  of S02  near 1 atm
 contributes little to the sensitivity of a gas  cell correlation instrument
 for thisgas when the  sample  is  near 1 atm, as  in a smokestack.  The greatest
 sensitivity can be achieved  by making use of  the coarse, deep structure.  In
 order  to do this,  the grid for  S02  has been made with only two  openings that
 pass two spectral intervals,  each about 4.7 cm"1 wide.   One  is  centered in
 the region of strong  absorption  near  2513 cm'1, and the  other is centered
 outside the region of significant absorption near 2538 cm"1.

     Enough S02  is contained  in  the correlation cell  that it is essentially
 opaque in  the 2513 cm"1 interval while it is nearly transparent in the other
 interval.  It follows  from the previous discussions on the spectroscopic
 principles that  the transmittance of  the attenuator is approximately 50%
 over both  spectral intervals.  The  amount of radiant  energy passing through
 the  correlation  cell  is only  very slightly dependent  on  the amount of SOo
 in  the sampling  region.  When the attenuator is in the beam, the SO- in the
 sampling section absorbs  in one of  the two spectral intervals passed to the
detector.   Approximately  twice the  contrast in  the two beams of the alter-
nator could be achieved if the alternator transmitted only in one of the two
intervals during one-half of each cycle and only in the other interval during
the other half.  Because of the relatively wide spectral intervals involved,
4.7 cm"1,  it would not be difficult to accomplish this by modulating the
energy with a  chopper in the plane of the dispersed energy in the grating
                                     38

-------
   1.0
o
CO
     0
                                                  6
  1
1
                               2500  WAVENUMBER(cm'l)    2470
                                                         —,
T
                                4.0       WAVELENGTH (Mm)      4.5


   FIG. 19.  Spectral curves  of transmittance between 2465  and 2525 cm"1.  The
   total pressures of the two dilute, room-temperature, mixtures of SO, in N2
   are 1.00 atm for Sample A  and 15.0 atm for Sample B.   u  = 0.78 atm cm for
   Sample A; u for Sample B was reduced  a few percent to  produce nearly the same
   average transmittance as Sample A. The spectral slitwidth ^ 0.35 cm"l.
                                     39

-------
 assembly.  In this case a correlation cell would not be required.  Alternating
 between two similar intervals could also be done with narrow -bandpass filters.
 It follows from the above discussion that gas -cell correlation methods do
 not provide the advantages in terms of effectively high spectral resolution
 for S02 that they do for other gases such as HCl, HF, NO, CO, or C0? whose
 spectra consist of distinct, well-separated lines.

      Two bands of S02 centered near 1150 cm   and 1360 cm~l are stronger
 and would provide higher sensitivity than the one used.  However, H20 inter-
 ference is much more troublesome and detectors are less sensitive in the
 regions of the two lower -wavenumber bands.  Data on the absorption by these
 bands are given in Reference 10.

      Data on the absorption properties of CO,  HCl, and HF are available in
 a number of reports and papers.   Line parameters used in the design of the
 present instrument are from Benedict, et al(H)  for CO; Benedict, et al(!2)
 for HCl; Meredith and Kent(-'U-)  and Herget , et  al(14)  for HF.

 LABORATORY TESTS AND CALIBRATION DATA

      Figures  20, 21,  and 22  show plots of V1  = V /V  for each of the 5 gases
 with the appropriate  grating assembly in place as  indicated by Figure 13.
 These curves  can be used for calibration to  convert the instrument  output  to
 the partial pressure  p of  the gas species being  measured.   The ordinate for
 each of the figures corresponds  to the ratio A/E plotted in Figure  8, and
 the abscissa  is  the absorber thickness of the gas  normalized to room tempera-
 ture,  296  K.  JThe  curves for CO  and NO in Figure 20 are based  on samples at
 296 K,  and u  = pi,  where p is the partial pressure  of  the  absorbing gas  in
 atm and I  is  the geometrical path length  in cm.  All  of the  samples used to
 obtain the data  contained the absorbing  gas mixed with  N2  at 1 atm  total
 pressure.

     Samples  of  CO  + N2  and  NO + N2  were  also investigated at  approximately
 420 K  (279°F) which is  typical of smokestack exhaust.   We  found  that  a single
 curve  can  be used if the value of u  is normalized  to 296 K by
              „/„,.„      \ -          (cm) 296            U6
              u(atm cm296) -             -  , or p
where 6 is the stack temperature.  By normalizing in this manner, the system-
atic error introduced by using a single curve is not more than approximately
5% over the temperature range investigated.  The validity of this simple ex-
pression for normalizing indicates that the absorption is nearly independent
of temperature changes provided the number of molecules in unit cross -sectional
area of the beam remains constant.

     The S02 data obtained at the two different temperatures cannot be repre-
sented by a single curve.   Thus, accounting for the stack temperature is
somewhat more complicated for S02 than for CO or NO.  Concentrations of SO,
could be determined graphically from the two curves of Figure 21 by interpolation
                                     40

-------
0.000!
0.00!
0.01           O.I
    u (atm cm
10
    FIG. 20.  Curves of V1 versus normalized absorber thickness for CO and NO.  The
             correlation cell parameters are:  p = 0.461 atm, L = 2 cm for
             CO; and p = 0.414 atm, L = 2 .cm for NO.

-------
to
                  0.00!
0.0!              0.!
    u  (atm  cm296)
            FIG. 21.   Curves of V1  vs normalized absorber thickness for S02 and HCl.  The results
                      for the elevated temperature 422 K for SO,  are shown as a broken curve.  The
                      correlation cell parameters are:  p = 0.539 atm, L = 1 cm for HCl; and p = 2 atm,
                      L = 5 cm for  S02»

-------
U)
              O.I
             0.01
            0.001
           0.0001
                0.00001    0.0001      0.001        0.01
                                       u ( atm crr,296)
O.I
1-0
                  FIG. 22.  Curve of  V1 vs normalized absorber thickness for HF.  The
                           solid portion is based  on laboratory data, the dashed portion on
                           calculations.  The correlation cell'parameters are p = 0.164
                          'atm,  L =  1 cm.

-------
  for temperatures between 296  K and 422  K.   The  temperature  can  also be
  accounted for empirically by  use of the follwoing  expression:
    u(atm cm . ) 6      g
p =	
                        2%
                                                          _|-
 The  uncorrected  value of u1  is determined from V1 and the solid S00 curve of
 Figure  21.                                                        '

       No  calibration data were obtained for HCl or HF at other than room
 temperature.  For elevated temperatures less than approximately 420 K, it
 is probably safe to account  for temperature in the simple manner used for CO
 and  NO.

       Table 2 summarizes the important parameters and the performance of the
 instrument for the 5 gases of interest when the appropriate grating assembly
 is being  employed.  The second column lists the spectral intervals passed by
 the  grid  in the retroreflector  and grid assembly.  As indicated in the dis-
 cussion of Eq. (6), the Pembrook Factor given in the third column relates
 the  absorber thickness of the gas in the sample to the resulting percentage
 misbalance in the two alternator beams.

       A Nernst glower source heated by 0.8  amp  of alternating current  was
 used in measuring the absorber thickness required to produce a signal  equiva-
 lent to the peak-to-peak noise arising from the detector.  A 3-second  time
 constant was  employed.   The laboratory optical  path  simulated a double-passed
 stack 3 m in  diameter with  5 cm diameter windows.  Both  the  PbSe  and PbS  de-
 tectors have  2 mm x 2 mm sensitive  elements  cooled with  liquid nitrogen.

       Water vapor produces  the most serious  interference when measuring NO
 The  discrimination ratio (10,000:1)  is based on a  3 m diameter stack contain-
 ing  5% H20.  Because  of  a non-linear relationship  between  the H00  absorber
 thickness  and  the  instrument  response, the discrimination  ratio depends on
 the amount of  H20.  Unrealistically high discrimination  ratios  can be  observed
 with  high  concentrations of  the interfering gas.  A meaningful  discrimination
 ratio should be measured with approximately the same  absorber thickness of the
 interfering gas as will be encountered during a measurement.

      Several measurements of H-O interference were made while measuring NO
with  a different instrument employing various bandpasses that  included several
of the H20 lines seen in Figure 18.  Typically, the H?0  interference was
 10-100 times as great as that observed with the grating assembly with a band-
pass  of 6 narrow spectral intervals.  By employing the grating assemblies to
carefully select the bandpasses, it has been possible to reduce the inter-
ference by other gases to a negligible amount while measuring CO, SO ,  HCl
or HF.                                                              2     '
                                      44

-------
     Hydrogen fluoride is highly reactive,  and  it  adsorbs  and  desorbs  on  and
off of the walls of the gas-handling system and correlation cell.   Therefore,
concentrations of the dilute HF mixture were difficult  to  determine accurately.
Samples of HF concentration between 0.1% and 1% were  prepared  in a  1 cm sample
cell in the following manner to provide calibration sample absorber thick-
nesses between approximately 0.001 and 0.01 atm 0*1295.   A.  stainless steel
cylinder was used to make a 1% HF mixture in N2.  The cylinder was  precondi-
tioned for adsorption by introducing pure HF into  the cylinder at a slightly
higher pressure than would be used in the mixture  and allowing the  gas to
react with the walls for approximately one hour.  The HF was pumped out and
the desired 1% HF in N2 mixture made.

     The cylinder was attached to the sample cell  with  a short piece of
tubing to reduce adsorption in the lines.  In order to  obtain  samples  with
concentrations below 1%, the sample cell was filled with the 1% mixture to
a predetermined pressure, always less than 1 atm.   The  total pressure  in  the
sample cell was then increased to 1 atm by adding  N2«  This method  gave con-
sistent results for concentrations as low as approximately 0.1%.  At lower
concentrations, apparent variations in the amount  of adsorption and desorp-
tion made the results unreliable.  A shorter sample cell was not  available
to enable us to make samples of smaller absorber thickness while  maintaining
an accurately measurable concentration.

     In order to extend the calibration curve to smaller values of  absorber
thickness, we employed published values for the HF absorption line  para-
meters to calculate the expected instrument response.  The calculations also
involved the spectral bandpass, the amount of HF in the correlation cell,
and use of the Ladenberg-Reiche functionC15) for absorption by a spectral line.
For very small absorber thicknesses, V1 is proportional to u,  as  is indicated
by the broken line of unity slope in Fig. 22.  The calculated values agreed
with the experimental values for u between approximately 0.001 and 0.005 atm
cm296.

     We have used this calculation method previously for other gases whose
line parameters are well-known.  Generally, the shape of the curve can be
calculated quite accurately for values  of u up  to where the slope  of a cali-
bration curve such as the one in Fig. 22 has decreased to  about 0.5.  The
calculated curve of V1 vs u may require a slight shift up  or down  to match
an experimental curve, but the shapes of the two curves on log-log scales
are nearly identical.  Because of our previous  success in  calculating  cali-
bration curves, and because of the  agreement between the calculated and
measured values for several values  of u, we believe  that both  the  solid  and
dashed portions of the calibration  curve in Fig. 22  are accurate to less than
* 10%.

     When measuring HF,  the discrimination  against H20  is  50,000 :  1.  Most  of
the interference by HoO  is due to  the  longest wavelength  interval  which  is
passed by the grid silt  nearest  the grating.   If more discrimination  against
HoO is needed,  the grid  slit  that  passes the  longest wavelength could be
blocked and  only the  other  two grid slits  used. New HF calibration data
would have  to be obtained  for  the  assembly  used in this fashion.


                                     45

-------
                                                           TABLE 2
                                              INSTRUMENT PARAMETERS AND  PERFORMANCE


Gas to be
Measured Spectral Band Pass
NO 6 intervals, each centered


Pembrook
Factor j
(atm cm)
0.81

Detector and "~
Noise-Equivalent
Absorber Thickness
(atm cm)_ Remarks on Interfering Gases
PbSe HO D.R. = 10,000:1 with H-0 at 125°C.
0.003 „
           on a pair of NO lines near
           1900 cm"1
                                                                                5  ppm  NO.  The  6  pairs  of NO  lines  in
                                                                                the  spectral bandpass are relatively
                                                                                free of HjO absorption.  Inclusion  of
                                                                                other  lines would increase  the  inter-
                                                                                ference by HjO.
 CO        11 intervals,  each centered
           on a CO line
           and 2200 cm"
                                               1.85
on a CO line between 2150
           -1
                                                       PbSe
                                                       0.0008
                  C00
                         Interference  too  small  to measure.
                         D.R. > 200,000:1  with O>2 at  125°C.
                         107. CO2 produces  error  corresponding
                         to  less than  0,5  ppm CO
                         D.R. S- . 50,000:1 with H.O at 125°C.
                         57.  H20 produces error corresponding
                         to  -1 ppm CO.
SO- 2 intervals, near 0.088
2538 and 2513 cm"1.
HC1 4 intervals, each includes 0.7
corresponding lines of 35_
L>
PbS
0.011
PbS
0.003
No significant interference by other gases
in stacks of coal-burning plants in this
spectral region.
Intervals in the spectral bandpass are
essentially free of interfering absorption.
          and 38  isotopes.  Inter-
                C
          vals are between 2750 and
          2850  cm"1.
                                                                            absorbs near the adjacent lines.
                                                                                                              SO
                                                                        might interfere with one of the HC1 lines.
HF
          3 intervals,   each passes a
          line between  3995  and
          4080 cm
                 -1
                                    4.2
PbS
0.0001
                                                                        D.R. = 50,000:1 with H20 at 150"C.
                                                                        Improved discrimination can be obtained by
                                                                        blocking the long wavelength grid slit.
                                                             46

-------
     No data have been obtained with the instrument  shown schematically  in
Figure 11 without the grating assembly.   However,  from other  data  obtained  in
our laboratory with instruments employing narrow bandpass filters  in the
place of a grating assembly, we can conclude that  the simpler filter instru-
ment provides a higher ratio of signal to detector noise at the expense  of
poorer discrimination and a lower Pembrook Factor.

FIELD TESTS

     Field tests of the instrument measuring NO, CO, and S02  were  performed
on a smokestack of a coal-burning power plant in Charlotte, N.C.  Photographs
of the optics assembly and retroreflector mounted  on the stack are shown in
Figure 23; the cover of the optics assembly has been removed.  The side  of
the optics assembly next to the stack is connected by a hinge to a flange
welded on the stack.  Two adjustable legs support the back side of the optics
assembly; the legs can be adapted to a variety of stacks.

     Two ports with 10 cm openings are welded permanently on opposite sides of
the stack at approximately  1.5 meters above the level of the flat roof.   The
effluent temperature was approximately 422°K (300°F).  The two-way optical
path length through the effluent was approximately  580 cm.  A small building
approximately 6 meters from the stack housed the power supply unit and  the
electronics unit.

     A CaFo window is mounted on a shutter to prevent stack effluent from
escaping into the instrument.  Compressed air blown  over the inside surface
of the window prevents the  accumulation of dirt on  the window.  The air then
passes into the stack through a cylindrical tubing  that  extends approximately
30 cm into the stack.  Because of the air flow  through the tubing, very little
effluent reaches the window.  Although the required  air  flow was  greater than
expected, the windows remained clean enough for operation  for  several hours.
The window is mounted on a  sliding shutter that can be closed  easily to pro-
tect the window when the instrument is not in use.   A similar window arrange-
ment with plumbing for compressed air is mounted  on the  retroreflector  assem-
bly on the side of the stack opposite the optics  assembly.

     Smoothed plots of the  NO  concentration measured at  four different  times
are shown in Figure 24.  During three of the four periods, measurements were
compared with valves obtained with a DuPont Model 460-1  analyzer.   We note
that the data obtained with the two instruments agree quite well.   The  DuPont
instrument extracts a portion  of the effluent from  the  stack and  employs the
uv to measure the N0£ concentration.  The NO is then oxidized  to  N02, and
another NO? measurement  is  made.  From  the difference between  the N0£ concen-
trations, the original NO  concentration  in  the  effluent  can  be determined.
The N02 concentration in the  effluent before it leaves  the stack  is  generally
much  lower than  that of  the NO.

      Similar smoothed  plots of the CO concentration are shown in  Figure 25.
The  relatively low concentration varying between  5  ppm and 20 ppm indicates
 that  the  combustion was  nearly complete.  Simultaneous measurements made with
 an extractive NDIR CO  analyser resulted in average values near 9.5  ppm  during
                                     47

-------
FIG. 23.  The optics assembly and the retro-
          reflector mounted on the smokestack
          of the Duke Power Co. at Charlotte,
          N. C.
                       48

-------
   200
§
£ 100
i
      945
                         23 July 1974
                                                                                     T
                                                                         23 July 1974
                      950
                                       955
                                                                    1610
                           Tla*
                                                                                     1615
   200
   100
                         26 July 1974
                  1535
                                                                             27 July 1974
                                  1540

                                  T1JM
                                                    1545
940
                 945
     FIG.   24.  Plots  of NO concentration vs  time for 3  different days at the  Duke Power
                 Co.,  Charlotte, N.C.   The peak-to-peak noise was 14  ppro.  The  symbol Q)
                 represents the data obtained  with a DuPont Model 460-1 analyzer.
                                           49

-------
                                          20
                                     c
                                     o
                                     o
                                     B
                                     O
                                     U
                                          10
                                            1000
                                                                   1010



                                                                   Time
                                                                                        1020
      20
8
  a
  o
jjt
     '10
  c
  o
  u
           1440
                                 1450
1500
1510
                                                                                                            1520
                                                                  Time
  FIG. 25  Plot  of CO concentration vs time of day for 25 June 1974 at Duke Power Co.

           The peak-to-peak noise was 2.8 ppui.
                                                                                     Charlotte, N. C.

-------
the 3 day period of 25, 26 and 27 June 1974.   Several months earlier we  had
employed a small grating spectrometer to scan transmission spectra of the
same stack.  From these spectra we estimated  the average CO concentration  to
be 15 ppm.

     Data on the S0~ concentration are shown  in Figure 26 for periods during
three different days.  Agreement with the data by the DuPont instrument  is
good.  Effluent was extracted from the stack  by the DuPont instrument and  uv
spectroscopic methods were employed to determine the S0~ concentration.
During the beginning of the period of measurement on 26 June 1974, both
instruments indicated essentially the same concentration.  After about 9:10,
the DuPont instrument indicated a nearly constant concentration while the
Philco-Ford instrument indicated that the SO,, concentration was decreasing
slowly.  No satisfactory explanation for the  difference is given.

     The high level of noise observed while measuring S0« was due to the
PbSe detector being used.  Ordinarily a liquid-nitrogen cooled PbS detector
is used for 802? however, the dewar for this  detector had lost its vacuum
and could not be used.  The detectivity of the PbS detector is only about
one-tenth that of the PbSe detector at 4 |im.

     The noise-equivalent concentration of NO and CO were both approximately
twice as high as would be expected on the basis of the  laboratory data repre-
sented in Table 2.  For  example, the noise-equivalent absorber thickness of
0.003  (atm 0^295) of NO  corresponds to approximately 7  ppm  for the 5.80 meter
two-way path across  the  stack at 422 K.  Any one of several  factors may have
attributed to  the  lower  signal-to-noise ratios  observed in  the field:  dirt
on the windows, turbulence and vibration,  or improper alignment.   Exposure
of the detector to  sunlight during the alignment also probably increased  the
detector noise  for  several hours after the exposure.  This  latter  factor  seems
quite  probable  since  the detector did appear to behave  abnormally for a while
after  it had been exposed  to  the sun.

     No field  tests have been made for HCl or  HF  to  date,  although such tests
are  planned by EPA personnel  in  the  near  future.

     In general, the instrument  performed  quite satisfactorily during the
field  tests.   The  sensitivity,  detectivity,  and discrimination against  inter-
ference by other gases are more  than adequate  for  most  applications. As
indicated  previously,  the  largest  contribution to  the measurement errors
probably  comes from uncertainty in the  zero  setting.   A few minor problems
were encountered,  but these  can be remedied  easily.   Among these were the
accumulation  of dirt on the windows,  binding of the window shutters, and
difficulties  in changing the  detector,  or refilling it with liquid nitrogen,
without  exposing it to the sunlight.

     It  is  anticipated that  additional tests performed with this versatile
instrument will provide information that is  helpful in the design of simpler
gas-cell  instruments.
                                     51

-------
         1000
to
       c
       o
4J
c
0)
o
C!
O
u

CM
O
                             25 June 1974 ;
                       -1530-
                                Time
T701
                                              T7TO   10TKJ
                                                                                 27 June  1974
                                                             TOTD-
—roio-
 Time
                                                                                                       1530
         1000
                      T
                       e
                                                                           o
                                                       26 June  1974
0!	L_
   900
                              910
       920
                                                     930

                                                       Time
                                                                         940
                                                   950
                                                                                           1000
         FIG.  26.   Plots  of S02 concentration vs time of day for 3 different days at  Duke  Power  Co.  in Charlotte,  N.C.
                    The  time during zero adjustment and checking is represented by the dashed  lined in the  lower  figure.
                    The  peak-to-peak noise was 120 ppm.  The symbol  ffi represents data obtained  with the DuPont
                    Model  460-1 analyzer.

-------
                                SECTION 4

                                REFERENCES


 1.   Burch, D. E., and J. D. Pembrook.  "Instrument to Monitor CH ,  CO,  and
     CO   in Auto Exhaust."  Prepared by Philco-Ford Corp. for EPA under
     Contract No. 68-02-0587.  EPA Report No. 650/2-73-030, Oct.  1973.

 2.   Ludwig, C. B., Bartle, and M. Griggs.  "Study of Air Pollutant  Detection
     by  Remote Sensors."  Prepared for National Aeronautics and Space
     Administration under Contract NAS12-630 by Convair Division of  General
     Dynamics.  Report No. GDC-DBE68-011, Dec. 1968.

 3.   Bartle, E. R., S. Kaye, and E. A. Meckstrath.  "An In-Situ Monitor  for
     HC1 and HF,"  AIAA Paper No. 71-1049, presented at the Joint Conference
     on  Sensing of Environmental Pollutants, Palo Alto, Calif., Nov. 1971.

 4.   Goody, R.  "Cross-Correlating Spectrometer," J. Opt. Soc. Am. 58,   900
     (1968).

 5.   Advances in Environmental Science and Technology.  John Wiley and  Sons,
     Publishers, 1971.  Edited by J. N. Pitts and R. Metcalf.  Discussion on
     gas cells appears in a section "Spectroscopic Methods for Air Pollution
     Measurements," by P. L. Hanst.

 6.   Burch, D. E.  "Adjustable Bandpass Filter Employing a Prism," Appl.
     Opt. 8,  649  (1969).

 7.   Decker,  J. A., Jr.  "Experimental Realization of  the Multiplex Advantage
     with a Hadamard-Transform Spectrometer," Appl. Opt. 10. 510  (1971).

 8.   Abels, L. L., and J. H. Shaw.  "Widths  and  Strengths  of Vibration-
     Rotation Lines in the Fundamental Band  of Nitric  Oxide,"  J. Mol.
     Spectrosc. .20, 11  (1966).

 9.   Shaw, J. H.  "Nitric  Oxide  Fundamental," J.  Chem. Phys.  24,  399 (1956).

10.   Burch, D. E., J. D. Pembrook,  and D.  A.  Gryvnak.   "Absorption  and
     Emission by  SO. between  1050 and  1400 cm"1  (9.5  - 7.1 M-m ,"  prepared by
     Philco-Ford  Corp.  for EPA under  Contract No. 68-02-0013.  Philco-Ford
     Report No. U-4947,  (ASTIA PB 203523), July  1971.

11.   Benedict, W.S., R.  Herman,  G.  E.  Moore, and S.  Silverman.  "The Strengths,
     Widths,  and  Shapes  of Lines in the  Vibration-Rotation Bands  of CO."
     Astrophys. J.,  135.  227  (1962).
                                    53

-------
                           REFERENCES (Cont'd.)


12.  Benedict, W. S., R. Herman, G. E. Moore, and S. Silverman.  "The
     Strengths, Widths, and Shapes of Infrared Lines II.  The HC1
     Fundamental."  Can. J. Phys. 34, 850 (;956).

13.  Meredith, R. E., and N.  F. Kent.  "Line Strength Calculations for the
     °~*"1' 0—*2, 0->-3, and l-*-2 Vibration-Rotation Bands of Hydrogen
     Fluoride."  Prepared for the Advanced Res. Proj. Agency under Contract
     No. SD-91 by The University of Michigan.  Report No.  4613-125-T
     April 1966.                                                     '

14.  Herget,  W. F.,  W.  E.  Deeds,  N.  M. Gailer,  R. J. Lowell,  and  A. H.
     Nielsen.   "Infrared  Spectrum of Hydrogen Fluoride:  Line Positions and
     Line  Shapes.  Part  II.   Treatment of Data  and Results."  J.  Opt.  Soc.


15.  Goody, R. M., "Atmospheric Radiation",  1964, Oxford University Press.
                                   54

-------
 BIBLIOGRAPHIC DATA
 SHEET
                  1. Report No.
                    EPA-650/2-74-094
                                                                 3. Recipient's Accession No.
                                                                    5. Report Date
                                                                      December 1974
Title and Subtitle
Infrared  Gas Filter Correlation Instrument for in  situ
Measurements of Gaseous Pollutants
                                                                    6.
 . Author(s)
  D. E. Burch and D.  A.  Gryvnak
                                                                  I. Performing Organization Kept.
                                                                    No.   U6121
  Performing Organization Name and Address
  Philco-Ford Corporation, Aeronutronic Division
  Ford Road
  Newport Beach, California 92663
                                                                  10. Project/Task/Work Unit No.
                                                                    1AA010/26AAP/14
                                                                  11. Contract /Grant No.

                                                                    EPA 68-02-0575
 "2. Sponsoring Organization Name and Address
  Chemistry and  Physics Laboratory
  National Environmental Research Center
  Environmental  Protection Agency
  Research Triangle Park, N. C. 27711
 ^^Supplementary Notes
                                                                  13. Type of Report Si Period
                                                                     Covered  Final Report
                                                                     7/72  to 6/74
                                                                  14.
             infrared analyzer employing gas  cell  correlation techniques has been designed!
 n   constructed to measure the concentrations  of  carbon monoxide,  nitric oxide, sulfur    I
 ioxide, hydrogen chloride,  and hydrogen  fluoride in the effluent  of stationary sources.
 n  infrared beam is directed across the stack  to  a retroreflector  and back so that  the
 "nstantaneous average  concentration is measured continuously without disturbing the con-
 tituents of the effluent.  A small, removable, fixed-position  grating monochromator  acts
as  a unique optical filter that passes narrow  spectral intervals  that are centered  at
 ,avelengths where the  gas to be detected  will  absorb.  One grating monochromator  is used
 -.or CO and NO, another for S02 and HC1, and  a  third for HF.  The  useful ranges of concen-
     ion times path length, in atm cm, over which each gas can be  measured are:  0.005 to
 .   to NO; 0.0013 to 0.15 for CO; 0.001 to 4.0 for S02; 0.0003  to 0.2 for HC1 and 0.0001
 -o  0.02 for HF.  The discrimination against  other gases in the  effluent is excellent.
      report was submitted in fulfillment  of  contract number  68-02-0575 by Philco-Ford
(-.orp.» Aeronutronic Div., under the sponsorship of the Environmental Protection Agency.
    Key Words and Document Analysis.  17o. Descriptors

   Exhaust Gases
   NO,  CO, S02,  HC1 and HF
   Infrared  Analysis
   Correlation Techniques
 •|7b. Identifiers/Open-Ended Terms

   Gaseous Pollutants
   Gas Filter Correlation
   Effluent Concentration
   Stationary Source
   C- COSATI Field/Group
, Availability Statement

 Release Unlimited
                                                        19. Security Class (This
                                                          Report)
                                                               :LASSIFII
                                                          20. Security Class (This
                                                             Page
                                                          	UNCLASSIFIED
                                                                                21. No. of Pages
                                                                              22. Price
     NTIS-38 110-70)
                                                                                USCOMM-DC 40328-P71

-------
   INSTRUCTIONS FOR COMPLETING FORM  NTIS-35 (10-70) (Bibliographic Data Sheet based on COSATI
  Guidelines to Format Standards for Scientific and Technical Reports Prepared by or for the Federal Government,
  PB-180 600).

   1.  Report Number.  Each tepott shall carry a unique alphanumeric designation. Select one of the  following types:  (a)  alpha-
      numeric designation provided by the sponsoring agency,  e.g., FAA-RD-68-09; or, if none has been assigned, (b) alphanu-
      meric designation  established by  the performing organization e.g., FASEB-NS-87;  or, if none has been established, (c)
      alphanumeric designation derived from contract or grant number, e.g., PH-43-64-932-4.

   2-  Leave blank.

  3.  Recipient's Accession Number. . Reserved for use by each report recipient.

  4.  Title and Subtitle.  Title  should indicate  clearly and briefly the  subject coverage of the report, and be  displayed promi-
      nently.  Set subtitle, if used, in smaller type or  otherwise subordinate it to main title. When  a report  is prepared in more
      than one volume, repeat the primary title, add volume number and include subtitle for the specific volume.

   5.  Report Date. Each report shall carry a date  indicating at least month and year. Indicate the basis on which it was selected
      (e.g., date of issue, date of approval, date 01 preparation.


  6.  Performing Organization Code. Leave blank.
      Translation of ...  Presented at conference of ... To be published in ...  Supersedes . . .       Supplements . . .

  16.  Abstroct.   Include a brief  (200 words or less) factual summary  of the most significant information contained in the report.
      If the report contains a significant bibliography or literature survey, mention it here.

  17.  Key Words  and Document Analysis,  (a).  Descriptors.  Select from the Thesaurus of Engineering and Scientific Terms the
      proper authorized terms that identify the major concept of the  research and are sufficiently specific and precise  to be used
      as index entries for cataloging.
      (b).  Identifiers and Open-Ended Terms.  Use identifiers  for project names, code names, equipment designators, etc.  Use
      open-ended terms written in descriptor form for those subjects for which no descriptor exists.
      (c).  COSATI Field/Group.  Field and Group assignments are to be taken from the.1965 COSATI  Subject Category  List.
      Since the majority of documents are multidisciplinary in nature, the primary Field/Group assignmem(s) will be the specific
      discipline,  area of human endeavor, or type of physical object.  The application(s) will be cross-referenced with secondary
      Field/Group assignments that will follow the primary posting(s).

  18.  Distribution Statement.  Denote reusability to the public or limitation for reasons other than  security for  example "Re-
      lease unlimited".  Cite any availability to the public, with address and price.

  19 & 20.  Security Classification.   Do not submit classified reports to the National Technical Information Service.

  21.  Number of Po9e».  Insert the total number of pages,  including this one and unnumbered pages, but excluding distribution
      list, if any.

  22.   Price.  Insert the price set by the National  Technical  information Service or the Government Printing  Off ice, if  known.
rORMNT,..38(,0.70,	USCOMM-OC 4032.-P7

-------