EPA-R2-73-207

January 1973            Environmental Protection Technology Series
Development of Improved Systems
for Obtaining Time Integrated
Measurements of SC>2, NC>2,  NOX,
and Other  Pollutants
                            Office of Research and Monitoring
                            U.S. Environmental Protection Agency
                            Washington. D.C. 20460

-------
                                            EPA-R2-73-207

Development of Improved Systems

   for  Obtaining Time Integrated

Measurements  of  SC>2>  NC>2,  N(3X>

          and  Other  Pollutants
                         by

                   Victor R. Huebner

                   Instra-Tech, Inc.
              1223 South State College Boulevard
                Fullerton, California 92631
                 Contract No. 68-02-0318
                Program Element No. 1A1010
           EPA Project Officer:  Andrew E. O'Keeffe

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

              OFFICE OF RESEARCH AND MDNITORING
            U. S. ENVIRONMENTAL PROTECTION AGENCY
                 WASHINGTON, B.C. 20460

                     January 1973

-------
This report has been reviewed by the Environmental Protection Agency
and approved for publication.  Approval does not signify that the
contents necessarily reflect the views and policies of the Agency,
nor does mention of trade names or commercial products constitute
endorsement or recommendation for use.
                                   ii

-------
                        TABLE OP CONTENTS
Section                       Title
 1             Introduction and Summary                          ].
 2             Overall Objectives                                2
 3             Basic Experimental Approach                       3
 3.1           General Background                                3
 3.2           Specific Characteristics of the Meloy Analyzer    3
 3.3           Secondary Standards                               5
 4             Direct Photographic Film Detection                6
 4.1           Basic Problems                                    6
 4.2           Basic Film Sensitivity Considerations             7
 4.3           Comparison of Different Films                    10
 4.4           Laboratory Tests of Film Sensitivity             11
 4.5           Reaction Chamber Tests                           18
.4.6           Gas Permeation Studies                           31
 4.7           Direct Air Injection                             33
 5             Alternative Detectors                            34
 6             Photon Counting                                  37
 6.1           General Background                               37
 6.2           Electronic Considerations                        39
 6.2.1          General                                          39
 6.2.2          Power Supply                                     4i
 6.2.3          Photomultiplier Tube                             4l
 6.2.4          Amplifier Circuit                                44
 6.2.5          Digital Division                                 46
 6.3           Results                                          tip
 7             Suggestions for Future Work                      53

-------
                       LIST OF TABLES
                            Title
             Relative Light Emission from the Meloy
               Reaction Chamber                               4
2            Comparison of Film Sensitivity                  12
3            Polaroid 3000 Sensitivity                       13
1»            Tri-X Sensitivity                               17
5            Diffusion Rate Parameters for a 1.5mm
               I.D. by 10mm Long Tube                        20
6            Relative Intensity Obtained from
               Different Reaction Chambers                   24
7            Relative Reflectivity of Various Tubes          27
8            Effect of Diffusing Ethylene into the
               Reaction Chamber                              31
9            Response of Various Photodetectors              35
                                                             11

-------
                        LIST OP FIGURES
Figure                        Title                            Pagi

  1            Typical Response Curve for Photographic
                 Film                                            9
  2            Chemllumlnescent Camera                          16
  3            Chemllumlnescence Reaction Chanber               2.1
  4            Effect of Gas Velocity on Signal                 23
  5            Effect of Reaction Chamber Length                29
  6            Gas Permeation Tube Test                         32
  7            Basic Components of the Chemllumlnescence
                 Photon Counter                                 40
  8            PMT, Preamp, and Comparator Circuit              43
  9            Photon Counter/Readout Circuit                   47
 10            Count Rate vs Current at Various
                 Threshold Levels                               50
 11            Threshold vs Counts at Various Light
                 Levels                                         52
 12            Correlation of Count Frequency with
                :Display Reading                                54
 13            PMT Current vs Meter Reading                     55
 14           .Ozone Concentration vs LED Matrix Readout        56
 15            Appendix A - Digital Division Circuitry          59
 16            Appendix B - Photon Counter Description          60
                                                                111

-------
 1.  Introduction and Summary
      This program was primarily directed toward Investigating new
 and improved methods of detecting and recording the weak light
 emission resulting from the chemiluminescent reaction of air pollutants
 with specific gaseous reactants.  Major emphasis was placed on
 attempting to use photographic film as both a detector and integrator
 for the light emission.   The intent was to expose film to the
 chemiluminescent reaction,  develop the film, and relate film density
 to accumulated air pollutant.   The major problem was that it requires
 almost  a million times  as much light to produce a detectable response
 on a one Inch square film plate as Is required  for a photomultiplier
 tube.   This  disadvantage  can be minimized  by concentrating  the  reaction
 over a  very  small  surface area.   Consequently reaction chambers
 having  extremely small frontal areas were  developed.   Chambers  that
 were ,5mm diameter  provided  1400 times  as  much  light  Intensity.
 Unfortunately,  this  still wasn't  enough  to  overcome  photographic
 film's  poor  sensitivity.
     Although  the photographic  film  approach was unsuccessful, several
 avenues  were explored that may be  of  value  in the  future.  One of
 the  more  important findings was  that  the chemiluminescence reaction
 is primarily limited by physical mixing efficiency of the reactants.
 Long, narrow reaction chambers provided much better mixing efficiency.
 With excessively narrow tubes,  however, optical reflectance losses
 counteracted any Improved mixing efficiency.  Nevertheless, moderately
 small diameter reaction chambers should yield better signals and make
 it possible to use small diameter detectors.  The other major finding
of interest was that permeation tubes could be effectively used to

-------
 meter in the reactant gas.  This may make it possible to eliminate
 the need for auxiliary reactant gas cylinders.
      Photon counting was then explored as a means of increasing the
 reliability and reducing the cost of chemiluminescent detectors.
 This approach was quite successful.  A system was developed that
 amplified each photoelectron,  discriminated against noise pulses,
 accumulated all the photon counts,  and displayed the sum of the
 photon counts on a 5 x 7 light emitting diode display.   The photon
 counting system was packaged in a housing and was shipped to EPA
 laboratories for further tests.

 2.   Overall objectives
      Chemiluminescence techniques for  monitoring air pollutants  such
 as  S02,  N02,  NOx,  ozone,  etc.  are extremely  sensitive and  reliable.
 Their major advantage  is  that  they  require minimal  mechanical components,
 and  no wet  chemicals.   Although  Chemiluminescence methods  are basically
 simple and  reliable, they  impose severe sensitivity  requirements upon
 the  optical detector system  required to monitor  light emission.
 Photomultipller  tubes  coupled  to high impedance  analog amplifiers
 are  utilized almost exclusively in  commercial Chemiluminescence
 instruments.  These Instruments generally perform quite well under
 controlled laboratory conditions where the temperature is held constant
 and  technicians are available to zero and calibrate the  instruments.
 One of the purposes of this program was to develop methods of detecting
 light emitted by Chemiluminescence which are more stable and reliable
than those presently used.  The other prime objective of this program
was to develop a simplified method of time-integrating and recording

-------
 the  outputs of  chemilumlnescence instruments.  Data reduction
 through  time Integration la a practical necessity for any large
 scale monitoring program.  For time Integration to be effective,
 instrument stability must be extremely good.
     Present instruments must either have periodic manual calibration
 or have  very expensive self-correcting circuitry if long term
 stability is required.  Photomultlpller tube gain changes caused
 by temperature  variations or power supply changes are excessively
 high.  Even the best amplifiers are somewhat marginal in respect
 to zero  drift for this application.  Consequently primary effort
 was  placed on elimination of the PMT system by utilizing photographic
 film as  the detector as well as recorder.  When this approach proved
 to be unsuccessful, digital photon counting methods were explored.

 3.   Basic Experimental Approach
	3*1  General background
     This program was oriented primarily towards the "readout"
portion of air pollution monitors rather than towards the chemical
transducer components.  Consequently, a commercial ozone analyzer,
the Meloy Laboratories, Inc. OA320, was purchased for this program.
This analyzer is typical of the current state-of-the-art in pollution
analyzers based on the chemlluminescence principle.
     3»2  Specific Characteristics of the Meloy Analyzer
     The Meloy ozone analyzer has a 19mm diameter by approximately
20mm deep chamber with concentric 6.3 and 3.2mmO.D. tubes terminating
at the front glass wall to carry ethylene and air, respectively.
This IR essentially identical to the Nederbragt design.   Recommended

-------
flow rates are 1 liter per minute for air and 28ml/min. for ethylene.
A UV lamp Is used to provide ozone for calibration purposes, and
gave a reading of 0.22ppm ozone under normal operating conditions.
     When a Kodak Wratten 2B filter (400aim long pass) was Inserted
between the reaction chamber and the PMT, the ozone reading decreased
from 0.22ppm to O.l4ppm.  When a Wratten 12 filter (510nm long pass)
was used, the ozone reading decreased to 0.02ppm.  These results
Indicated that 36% of the light emission Is below 400nm, 555? Is
between 4oOnm and 510nm, and only 9% of the light emission Is greater
than 510nm.  These results agree with those of Hodgeson, et al1.
     In another series of tests, masks with various size center holes
were placed over the reaction chamber.  These results are shown In
table 1. The data Indicate that light emission probably is evenly
                               TABLE 1
        Relative Light Emission from the Meloy Reaction Chamber
Diameter hole
   In mask         Hole Area       PMT Reading        Intensity/mm2
    6.0mm            28mm              .Olppm             3.6
   10.5              86                .08                9.3.
   13.5             145                .14                9.7
   19  (no mask)    284                .22                7.7
distributed throughout the reaction cell.  The lower value for the
smallest hole probably Is due to a lower volume of gas since this
space Is occupied by the inlet tubes.   The decline for the largest
area probably is due to a restricted  angle of view because of the
chamber walls.   Thus,  it appears that  light Is emanating uniformly
1. Hodgeson,  J.A.,  Martin,  B.E.,  and Baumgardner,  R.  E.  Laboratory
   Evaluation of Alternate  Chemlluminescent Approaches for the Detection
   of Atmospheric Ozone.  ACS Meeting,  Chicago,  Sept. 1970.

-------
 from all  portions  of the  chamber.   This  would  be  the  case  If  Incomplete
 or Just barely  complete mixing  of  the  ethylene and  oxygen  were
 occurring.
      The  response  of the  entire  system was  found  to be  9x10-9 amperes
 per ppm ozone.   If we assume reasonable  values for  typical  PMT's,
 this corresponds to  approximately  IxlO-10 lumens, or  llxlO-11  lumens
 per centimeter.  As  wll]  be shown  In table  2 this is  about  100 fold
 less than the sensitivity threshold level for  photographic  film.
 The response was inversely proportional  to  air flow rate (when the
 ozone generator was  used), and was moderately  sensitive to  ethylene
 flow rate.  As will  be seen later, this  is  symptomatic of incomplete
 reaction  between ozone and ethylene.

      3.3  Secondary  standards
      Several types of secondary light  sources  were  used instead of
 the  Meloy Chemiluminescence monitor in order to provide greater
 versatility and convenience.  A Monsanto MV-1  light emitting diode
 (LED) was used in most cases.   By appropriately adjusting the voltage
 and  current-limiting resistors, highly reproducible light levels
 could be  easily achieved.   The LED Ijght output was calibrated by
measuring its output with the  Meloy analyzer PMT,  and recording the
output in "equivalent parts-per-mlllion ozone".  In this manner,  the
relative  sensitivity of experimental detection modes could  be easily
determined.   The primary disadvantage of the LED  standard was that
light emission occurred at 6lOnm rather than at the ^30nm peak
wavelength of the ozone chemiluminescent  reaction.
     The other secondary standard consisted  of a  24v.  tungsten

-------
filament lamp driven by a 2lJvdc power supply through a series of
current limiting resistors.  This lamp was optically coupled to the
detector under test (or to the Meloy PMT for calibration) through a
60 mm long by 6.3 mm D. quartz fiber optic tube, a Schott BG 18
infra-red filter, and a Kodak Wratten 36 filter.  This combination
produced light with maximum intensity at 420nm. and a half-band width
of ^nm.  Although the tungsten lamp's output Is less controllable than
a LED1s, it's better spectral properties was needed for some tests.
In all cases, light output was cross-correlated with the response of
the Meloy PMT system (indicated in ppm ozone), so that meaningful
comparisons could be made.

4.  Direct Photographic Film Detection
     4.1  Basic Problems
     Photographic film was utilized in an attempt to eliminate
the present photomultiplj.er tube detection system because it has
a long history of usage in low light level application.  Photographic
film also appeared to be highly advantageous for its simplicity in
integrating and recording the relative light intensity.  Unfortunately,
photographic film also has two major disadvantages - reciprocity failure
and a dependance upon light intensity rather than total quantity of
light.  Photographic film has a sensitivity threshold below which
an image will not be formed regardless of exposure time.   Consequently,
low light levels will either have a very non-linear relationship to
film density or will go undetected.  The mode of sensitizing -
through light intensity rather than light quantity further compounds
this problem.  Whereas PMT detectors linearly respond to  total light

-------
with no regard for light distribution, film response Is based upon a

light Intensity per unit area relationship.  The only feasible method

of Improving this situation Is to concentrate the available light of

reaction into a smaller area.  However, since the chemlluminescence

reaction is a gaseous reaction, the mixing problem becomes more

severe as reaction tube area is decreased.

     4.2  Basic Film Sensitivity Considerations1

     Essentially all film manufactured today is based on the silver

bromide reaction to light.  The primary factor affecting film

sensitivity is the concentration of sulflde specks at the surface

of the silver crystals.  These specks originate from high molecular

weight compounds in the gelatin used as the film base (synthetic

sulfur compounds give poor sensitization) and congregate at imperfections

in the silver bromide lattice.  These sulflde specks require a

significant number of photon-generated electrons before they are

capable of promoting silver bromide sensitization through an electron

transfer process.   This requirement for high concentrations of photon-

generated electrons arises because of a competing electron capture

process within the sulfide specks.  Once this threshold level has been

reached, the silver bromide conversion can proceed at a rate governed

by light intensity.  Larger silver bromide crystals hold more sulfide

specks and absorb  more light than smaller crystals, thereby having

better sensitivity.

     Sensitivity characterists of photographic film can be described

by a density vs.  light exposure curve similar to the one shown in


1.  Extracted from "Photographic  Film" by P.  Glatkides,  Fountain Press,
    London,  1958.

-------
  figure 1.  All film exhibits a nominal density called "fog" with no
  light exposure. As the light Is Increased, the density begins increasing
  at the threshold point.  The curve then assumes a curved toe, a linear
  section,  and a curved shoulder.  Very high light intensities create
  a density decrease called "solarizatlon".   Various systems are used
  to designate film speed.   Perhaps  the poorest  system Is  the ASA system,
  which has no scientific basis,  Is  only valid  for high light levels,
  and cannot  be determined  with  any  reasonable degree  of precision.   For
  our purposes,  the  threshold  level  Is  of paramount  importance  and re-
  ferences  to  "sensitivity  improvement"  will  refer to  a  decrease  In the
  quantity  of  light  required to reach the threshold  level.   Consequently,
  many  of the observations made may appear to be contradictory to
  popularly accepted notions.
       It is generally believed that film sensitivity can be improved
 by different developing procedures.  Unfortunately, high speed
 developing techniques only Improve contrast -  they have no effect on
 the film threshold level.   Since the ASA ratings  are  based on contrast
 ratios, darkroom manipulations  will Improve the ASA value.  They can't
 make an invisible  latent image  appear  if the film threshold hasn't
 been reached.
      The addition  of  sensitization  chemicals to unexposed  film  can
 greatly enhance  sensitivity to long  wavelength  (red)  light.  They
 cannot  improve the  basic film sensitivity at its maximum sensitivity
 wavelength (350nm.)  Thus,  they  serve only to extend the wavelencth
 response characteristics and can do nothing  to improve  film response
 to the ozone chemllumlnescence reaction, which already occurs near the
maximum wavelength sensitivity.

-------
                               Figure 1
             Typical response curve for photographic  film
                                                         solarization
Optical
Density
1
                             Quantity of light

-------
      The most  important  sensitization techniques  have  consisted  of
 pre- or post-exposure  of the  film  to  light,  mercury  vapor,  ozone,  sulfur
 dioxide, and various organic  chemicals.   These methods  generally are
 optimized by exposing  the film  to  a level Just below its  minimum
 threshold level,  so that the  image of interest can easily trigger  the
 exposure.  These  methods require very careful control,  and  sensitization
 beyond  a factor of 3 improvement generally is considered  to be too
 tricky  and unreliable.
      4.3  Comparison of  Different  Films
      Most of the  faster  types of film are compared in table 2.   Column
 1  lists  the type  of film, while its ASA value is listed in column  2.
 The  third column  lists the basic threshold of the film.   These values,
 expressed In meter-candle-seconds, represent the lowest light level
 that  gives a barely perceptible increase  in density  over  normal  film
 "fog".   The threshold values were  estimated in most  cases from manu-
 facturer's published curves.  The  next column lists  the approximate
 ratios of  sensitivity at  400 nm. versus that at 500  nm.  This is
 important because film threshold values are determined at $00 nm,
 whereas  the ozone chemilumlnescense emission is near 400 nm.  The
 columns  listing the sensitivity ratios for longer exposure times
 (slOOsec/     slOOsec/         reflect the increase in density with
      /slsec,       /slOOsec.)
 increased exposure time.  Values of 100 would represent perfect linearity,
while lower values are  the result  of  reciprocity  failure.   The conversion
to lumens per era2  puts  film sensitivity on a  radiometric basis.   These
values represent  the  lowest concentration of  400  nm.  light that  is
capable-of creating a  film Image..
     Type laO is  an astronomy  emulsion that is  coated on glass slides.
                                                                   10

-------
It Is characterized by a relatively slow speed jut very little


"reciprocity failure".  Thus,  for very long exposure tines, it is more


sensitive than some of the "fast" film.  Tri-X is a high speed'film


that is widely used by photographers.  It is more sensitive than laO


for exposure times less than 100 seconds, but its poor reciprocity


characteristics make it less sensitive than laO at longer exposure


times.  Royal Pan ^1*11 has approximately the same sensitivity as Tri-X.


Polaroid 3000 film is definitely faster for exposure times less than


100 seconds, but has very poor time-integration properties with longer


exposure times.  Polascope 410 is not significantly better than



Polaroid 3000 film.


     In summary it can be concluded that Polaroid 3000 film is


significantly more sensitive than laO, but has approximately the


.some sensitivity as do the other films listed in table 1 J f the exposure


time it> approximately 100 seconds.  For very long exposure times, the


excellent reciprocity characteristics of laO make it superior to other


films.   However, since photomultiplier tubes typically have sensitivities

           ~\ li              ">
of l-5x lO"-1  lumens per cm"",  even the best flirts have approximately


10,000 times poorer sensitivity than do photomultipller tubes.


	4.4 Laboratory tests of film sensitivity


     In addition to a theoretical assessment of the film sensitivity


problem, a large number of film tests were made using an MV-1 LED to


simulate chemilumlnescence.   The standard procedure followed in all


cases was to prepare the test  set-up, attach the Meloy PMT and


associated electronics, and then operate the LED at various current


levels in order to obtain a plot of input current vs. "equivalent ppm


ozone".  In all cases, the LED current was less than one-tenth as great


as the current, required to be  barely perceptible  to a trained observer





                                                                   11

-------
                                       TABLE  2

                            COMPARISON OF FILM SENSPr
Film

laO

Tri-X Pan

Royal Pan

Polaroid 3000
ASA value   Threshold (s)
                                                        slOOsec/   slOOsec/
                    25

                   320

                   400

                 3,000

Polascope lJ10   10,000
            .01 meter candle
                  secor.ds
            .001

            .001

            .0003

            .0001
x500nm
1.0
1.0
0.6
1.0
1.0
s Isec
100
15
30
10
5
slOOsec
70
5
10
2
2
lOOsea
1x10-8
7x10-9
6x10-9
3x10-9
2x10-9
' "ij
1x10-10
1x10-9
6x10-10
1x10-9
1x10-9

-------
after 5 minutes of acclimation in a totally dark room.  In other words,

the light intensities were exceedingly low.

     After the calibration curve for a particular test set-up had been

prepared, the PMT was removed and photographic film was placed in the

same relative orientation as the PMT cathode had been.  Various current

levels were than applied to the LED (one level for each film negative)

until it was determined that no image had been recorded.  The lowest

current level applied to the LED that had resulted in an image

(usually Just barely perceptible) was then recorded as the minimum

film threshold value.  By referring to the current vs. equivalent

ozone chart, this minimum threshold value could be reported in terms

of equivalent ppm ozone".

     The test procedure previously described was applied to Polaroid

3000 film. Results of these tests are shown in table 3« The exposure

areas were estimated by measuring the developed negative's images.  The

LED current used is only of academic Interest since the actual light

reaching the film was greatly influenced by the test set-up. The

                                TABLE 3

                       Polaroid 3000 Sensitivity
LED test set-up

One inch from film

One inch from film
Exposure
  time

  1 min

   60
Through 1.6mm x 45cm
 filter optic           5

Through 1.6mm x 30cm
 tube                   3

Through 0.8mm x 30cm
 tube                   3
Exposure
  Area
      ?
3,0cm.

3.0


 .070


 .018


 .005
  LED
Current

7xlO~5A

3xlO~5
Equiv.
Ozone

70ppm

30
                                                             Ozone/
  cm'
                      1.6x10-5   2
5x10'
                                  .1
                      3x10
                          -5
23ppm

10


28


55


80
                                                                   13

-------
equivalent ozone values were determined by comparing the LED current


required to produce a barely perceptible Image with the PMT output


response for that current. The ozone per cm2 values are simple


arithmetic derivations based on exposure area and equivalent ozone


threshold values.


     The first test was made with the uncolllmated LED one inch from


the film.  This resulted in a dense exposure in the center which diminished


in intensity towards the edges.  This non-uniformity would have a tendency


towards lower equivalent ozone values.  As may be seen, the 60 minute


exposure only resulted in a factor of two sensitivity improvement over


the one minute exposure time.  This is the result of Polaroid film's


poor reciprocity characteristics.  When the LED light was transmitted


through a fiber optic bundle, the total sensitivity was much enhanced


(due to the smaller area over which the light was spread) but the


sensitivity on a per unit area basis was approximately the same.  Where


a 1.6 mm or 0.8 mm teflon tube was used to transmit the light, their


smaller exposure area created a better gross sensitivity value, but


the sensitivity per unit area wasn't appreciably affected (within


experimental error).  It thereby appeared that Polaroid 3000 film has


an apparent sensitivity threshold of between 23 and 80 ppm ozone when

        <•>
a one cm  area is utilized.  This sensitivity is only improved by a


factor of two when  very long exposure times are used.  This estimate


is biased strongly in favor of the film, since the film sensitivity


is essentially the same at ^30 nm, (ozone chemilumlnescence) or 610 nm.


(LED), whereas the Meloy PMT has only 5% as much sensitivity to 610 nm.


radiation as it does to 430 nm radiation.  Thus, the true equivalent


sensitivity of Polaroid film is In the neighborhood of 1000 ppm. ozone/cm2.

-------
                                                         p
Since.PMT systems can easily detect  lOppb  ozone,  a one  cm  Polaroid

system Is approximately 100,000 times less sensitive than a PMT.   This

value agrees quite well with the factor of 10,000 calculated for  the

best film in a theoretical system.   Thus,  'the film system must  overcome

a 10,000 - 100,000 fold disadvantage before it can compete  with a PMT

system.  One method, which Is feasible only for astronomy emulsions

such as Kodak Ta-0, is to Integrate  the available light over exceedingly

long time intervals.  This may provide up  to a factor of 100 improvement

under Ideal conditions.  The other  technique is to reduce the frontal

area of the reaction chamber so that the chemiluminescence  is concentrated

in a smaller area, thereby Increasing the  intensity per unit area.

Simple gas dynamic considerations dictate  that the optimal  chamber

configuration would be a long narrow tube  with infinitely good  internal

reflectivity.  These types of chambers will be discussed In section 4.5.

     In order to accomodate the astronomy  type film plates  and  to hold

the long tubular types of reaction chambers that appeared to be needed,

a special camera chamber was designed and  built.  This is shown in

figure 2.  The rubber gasket was sandwiched between the upper and

lower halves to provide a light-tight, gas-tight seal.  The lower

chamber contains an aluminum film holder that can be positioned by

means of a rod sealed by 0-ring bushings.   The upper section contains

a light-emitting dJ.ode holder, a holder for chemiluminescence reaction

chambers, and a vacuum line.

     All initial tests of this camera chamber were made with 16mm.

Trl-X film.  The film was held on the film plate by double sided

Scotch tape.  In a typical test, the film would be positioned in place,

it would be exposed under prescribed conditions, and then moved about
                                                                   15

-------
                      1.25cm D.   hole to accept
                          reaction chamber
                                         To vacuum
                                      Rubber gasket
                                         Actuating rod
                              Felt

                           Film holder
Position Indicator
          Figure 2  Chemiluminescent Camera
                                                    16

-------
1/JJ inch to a new position where different exposure conditions were

used.  The film was moved between each exposure until all tests were

completed or untJ.l the entire film strip was used.  Development by

Mlcrodol-X (according to the manufacturer's recommended procedures)

always immediately followed the last exposure.

     The MV-1 LED produced an image l.lmm. in diameter for very low

light levels.  Higher light levels resulted in a central dense spot

about l.lmm.  in diameter, surrounded by a 5mm. diameter spot having

about one-third the density of the central spot.  The values shown in

table lJ indicates some of the relative densities obtained under varying

conditions.  The equivalent ozone values were obtained as before by

                               TABLE 4

                       Trl-X Sensitivity

LED CURRENT    Equlv. Ozone    Exposure time    Relative Density

                                                        0
                                                        1
                                                        3
                                                        6
                                                        0
                                                        3
                                                        5

determining the Meloy PMT response to varying currents through the

LED. The relative density values were subjective appraisals on a 0-10

scale where 0 Indicated no perceptible image.  It may be seen that a

light exposure equivalent to 0.18 ppm ozone produced no image, even

for 60 minute exposure times.  Increasing the exposure time from 20

minutes to 60 minutes resulted in a marked increase in density for the

same exposure levels, and a decrease in the exposure level required to

produce a given density.  This indicates that Tri-X has reasonably

good reciprocity characteristics.  This is in marked contrast to prior
1 uA
2
H
8
1
2
3
0.18 ppm
0.7
2.6
9.5
0.18
0.7
1.6
20 min.
20
20
20
60
60
60
                                                                   17

-------
results with Polaroid 3000 film where a 60 fold increase in exposure



time only resulted in a 2-fold improvement in sensitivity.



     The results shown in table 4 Indicate that the minimum threshold



level for Tri-X film is about 0.3 ppro ozone for a 1.1 mm. spot,  or



approximately 30 ppm/cm2.  This is essentially the same as  was found



for Polaroid 3000 film.  Since this film also has a "flat"  spectral



response characteristic, it's true sensitivity also is 20 times  poorer



(due to the poor response of the PMT to the light wavelengths emitted



by the LED). This "true" sensitivity then becomes 600 ppm.  ozone/cm2.



Clearly then, even with reasonable good photographic, films, the  entire


                                                          -M      -R
chemiluminescence reaction must ocnur within an area of 10    to  10 2



cm before photographic film can be comparable to PMT's.



     4.5 Reaction Chember Tests



     It was previously shown that the chemiluminesoenoe must be



concentrated in a very small area if adequate sensitivity is to  be



achieved with photographic film.  Theory dictates that it is impossible



to concentrate light by simple optical lenses or mirrors.  Some  preliminary



tests with lenses, mirrors, light guides, verified that normal optical



concentration techniques were useless.  One possible optical concentration



technique would be to collect light emitted within a glass  sphere with



fiber optics arranged in "porcupine" fashion.  The fiber optics  could



then be bent and arranged in a single bundle, thereby creating a rough



collination of the heretofore diffuse light.  This collimated light



could then be concentrated by conventional optical methods.  The



mechanical problems associated with this concept would be exceptionally



severe and it did not appear warranted to follow this path.
                                                                   18

-------
     It was shown that the chemilumlnescense reaction must be concentrated


                              -i|      -52
in an area of approximately 10   to 10 J cm .   This is equivalent to a



circle 0.04 to 0.11 mm. in diameter.  Since this small size precludes



a spherical reaction chember, a ]ong tubular-shaped chamber was



dictated.  Although the fact that the Meloy chamber exhibited marginal



mixing behavior even though it had a very large size was discouraging,



we proceeded with the deve]opment of smaller reaction chambers.  The



first chamber had a 0.76mm O.D. teflon tube with a 3.2mm  O.D. x 1.52mm



I.D. teflon tube that opened 15 mm back of the viewing port. This



pair of tubes was inserted into a tube that could be evacuated and



had a //O cover slip as its window.  All Joints were sealed with



silicone rubber.  This chamber was positioned  against the PMT.  Various



air flows and ethylene flow rates were tested.  The ethylene flow



rate had only slight influence on a signal.  The signal response was



essentially linear with the reciprocal of the  air flow rate.  Using



the inner tube for ethylene and the outer tube for air gave about



50$ greater response than the reverse case.



     When this reaction chamber was operated under identical conditions



to those used for the Heloy chamber, a reading of . OOUppro ozone was


                                                          2          P
obtained.  With the smaller area of the new chamber (l.Smnr vs. 28^1111^),



this represented an intensity per unit area improvement of 3.0.



     In viev; of the relatively poor improvement shown by the new



chamber, some basic diffusion parameters were  calculated using Pick's



principles.  These are shown In table 5,  The  first column lists various



flow rates, while the second column indicates  the degree of mixing



occurring between ehtylene and air.  For example, a flow rate of 105ml/



minute would mix 105? of the gases.  It Is obvious that the normal flow
                                                                   19

-------
rate of one liter per minute is responsible for less than \% mixing

efficiency.

                                TABLE 5

Diffusion Rate Parameters for a 1.5mm I.D. by 10mm Long Tube

           Plow Rate                    Mixing
           Requirement                Homogeneity

                7.4ml/min                 99?
               15                         90
               42                         50
              105                         10
              280                          1

     Since it appeared that a variety of tests v/ould be needed before

a small chamber that provided adequate mixing could be obtained, a

more versatile chamber was built.  This is shown in figure 3.

     In operation, the smaller assembly was inserted through the rubber

grommet until it touched the cover slip.  The large housing could be

mated with either the PMT or the camera.  The 1.45jrmO.D. tube was

used for ethylene.  This tub* was considerably longer than Its outer

teDescoping tube so that its end position could be varied without

disassembling the chamber. Although ethylene flow rates were varied

with all chambers tested, no large differences were ever noted.

Consequently, a standard ethylene flow rate of 28ml./min. was used

for all tests.

     In the first test,  the ethylene tube was positioned 4.0 cm. back

of the outlet.  When a one liter per minute air flow rate was used

with the ozone generator on, a reading of 0.078ppm ozone was obtained

from the PMT. After a mask with a 4mm D. hole was  used to block out

all but the tube Itself, a reading of only 0.0l8ppm was Indicated.

This Indicated that the  bulk of the reaction was occuring after the
                                                                   ?0

-------
                                 Figure 3

                     Chemiluminescence reaction chamber
#0 cover 3Jip~~-^|J

           ^1
1.3cm D. opening.
                       1.5cm O.D.
                           .10cm
                                   1.5mm  O.D.
                                   teflon
                                                .2mm D. tube to vacuum
                                                    I.D. rubber gromrcet
                                                       3mm  O.D.  x 1.5mm
                                                        I.D.  Tfeflon
                             6.3mm O.D. x 3.2mm  I.D.
                                Teflon tube

-------
gases had left the reaction tube.



     The relationship between air flov/ and ozone reading for this



3.3mm I.D. Teflon tube configuration is shown in figure 4.



     It may be seen that a very linear relationship was obtained



between the reciprocal of the air flow and output signal (as with



the Meloy chamber).  A 3.Omm  I.D. aluminum tube inserted into the



chamber did not change the values, thereby indicating that it was



not sufficiently reflective to exhibit light-piping effects, or



sufficiently reactive to Increase ozone consumption.  Results of the



previous  1.5mm  I.D. teflon tube chamber also are shown in figure [l.



Since this chamber had more light attenuation as well as a greater



gas velocity, its light output would be expected to be less.



     The other two "orifice" curves in figure ^ are of greatest



interest.  In these cases, a 3.0min  O.D. by 3.0cm. long by either



1.5mm I.D. or 0.76nunI.D. was inserted in the throat of the chamber.



The ethylene tube v/as moved Just back of the insert.  Both of these



curves exhibited increasing signals with decreasing flows at high



flow rates, but then reached a plateau at the lower flow rates.  This



probably indicated complete consumption (due to complete mixing)



of the available ozone.  The smaller orifice had a. lower signal probably



because of optical attenuation.  It also required a lower flow rate



before the plateau was reached, presumably because of the higher



velocity characteristics of small tubes.



     When data are plotted against the reciprocal of the air flow



rate as in this case, it is really analogous to plotting the data



against ozone concentration, since the ozone concentration decreases



as the air flow increases.  Although it may appear that a straight line
                                                                   22

-------
                              Figure 4

                  Effect of gas velocity on signal
           .20
           .15-
EQU!valent
   Ozone
   (ppm)
           .1C.
^"^. 5mm
                                                        5mm prlflce
                                                      (X*76mn  orifice
                                                        3.3mm Teflon tube

                                                        3mm  alum,  tube

                                                        1.5mm Teflon tube
                                 1.0       1.5       2.0

                               1/Alr Flo* (liters)

-------
function should occur with no plateau,  this  is  not  the case  if

efficient mixing occurs.   The correct model  for this  type  of test  is

to assume that there are  three gas flows - ethylene,  pure  ozone, and  air.

Since the ozone generator produces a constant  flow  of ozone  regardless

of air flow, the air merely serves as a diluent. Consequently,  under

complete mixing conditions, one would expect a constant response

regardless of air flow rate.

     The preceding data are tabulated as a function of chamber area

in table 6.  In all cases, the values obtained at an air flow rate

                                TABLE 6

                       Relative Intensity Obtained
                       from different reaction chambers

Type of                    Ozone    Reading/     Relative
Chamber          Diameter  Reading   Diameter2   Intensity

Original           1.95mm   .?2ppm    o,06          1.0
Teflon tube        1.5      .004      0.17          2.8
Teflon tube        3.3      .017      0.16          2.6
Aluminum tube      3.0      .017      0.19          3.1
3cm long orifice   1.5      .20       8.6         113
3cm long orifice   0.76     .05       8.6         113

of one liter per minute were used.  It  may be seen  that the  straight

tubes provided an intensity increase of about 3 regardless of

diameter or material.  The orifice tubes which provided a  turbulence-

producing edge, had 11? times as great  a light intensity as  the

original tube.  The 1.5mm inch orifice  chamber had  essentially the

same total quantity of light as did the original chamber.   In view

of the large signal as well as the plateau formation (fig. 4), we  can

safely assume that essentially complete reaction is occurring.  The

0.76mm inch orifice tube did not show the expected 4 fold Increase  in

intensity.  Since the air flow curve shape indicated complete mixing,

this lack of additional improvement simply was due  to loss of available
                                                                   24

-------
light due to the small hole.  In spite of the 100 fold .intensity
improvement;, it was still not possible to obtain a photographic
image.
     If one assumes that the average light-producing collision
occurs in the middle of the tube (1.5cm back of the viewing port),
the viewing angle even with perfectly straight chamber walls is only
6° and 3° for the 1.5mm  and 0.76mm chambers, respectively.  This
corresponds to a total light gathering efficiency of only  .025$ for
the l.5mm chanber, and .006? for.the 0.76mm chamber.
     Other chamber lengths and ethylene tube placements resulted in
either the same intensity or decreased intensity.  An interesting
observation was that a significant ozone response v/as found for as
long as three hours after the ethylene was turned off.  With the
smaller orifice chamber,  this response was essentla]ly the same as
when the ethylene v/as on.  Extensive flushing of all ethylene tubes
with air Immediately after ethylene shut-off did not alter this
situation.  It thus appears that ethylene probably is being absorbed
in the tubing, and exceedingly small quantities of ethylene are
required if a well designed chamber is used.
     The next step was to explore the possibility of utilizing internal
reflectance in order to increase the light gathering efficiency.  It
was hoped that reaction chambers with good internal reflectivity properties
would permit the use of longer tubes (for better mixing)  and smaller
dlataeter tubes (for better light concentration).
     A series of tests were made to determine the reflectivity properties
of various tubes.   A controlled diffused light source was oriented
at one end of the  tubes,  while the Meloy Labs PMT was used at the
                                                                   ?5

-------
 other end to determine relative light Intensity values.  Twelve Inch
 tubes were tested in all cases.  Results of these tests are shown In
 table 7.  The first column lists the type of tube tested, the second
 column Indicates the Inside diameter, the third column shows the
 amount of light passed by the tubes  (In equivalent ppm ozone), and the
 last column relates the average light Intensity per unit area. It
 may be seen that the glass tubes gave the best results, with the small
 bore capillary providing an intensity gain of 3^ fold.  If we assume
 a critical angle of 30° the Increase in light Intensity should have
 been a factor of ^0,000 gain.  The relatively low gain could be
 accounted for if the glass only reflected an average of 96% of the
 light with each bounde.  Although a factor of 3*» gain can be achieved,
 much higher gains are possible if the average reflectivity could be
 Increased above 96%.  However, this may not be practical for a field
 Instrument.
     The calculated efficiency for the small bore tube Is about 0.1^.
 If we take a rather simplistic view and assume that the average
 chemiluminescent collision occurs in the midpoint of the tube, the
 overall efficiency should be 3.2%.   A  3QCm   long by 0.5 mm reaction
 chamber should then exhibit an Improvement of .0327.052 or 12.8 fold
 over a one cm^ chamber.
     In order to evaluate actual chamber performance, various  tests
were made with small bore capillary tubing.  The 6.3mm O.D.  x 0.5 nun  I.D.
capillary tube was mounted in a 13mm vacuum tube equipped with a. #00
cover slip and a suitable adaptor for mounting on the photomultiplier
tube.  Air and ethylene were mixed  right at  the entrance to  the tube.
Ethylene flow rates  of  0.25,  0.5,  1.0,  1.5,  and 2.0 liters  per minute
                                                                   26

-------
                          TABLE  7
    Relative reflectivity of various  tubes
 Type of
  Tube
 None-holes spaced
     30cm apart
Inside
Di ameter
1.8mm,
Light
Intensity
1.35
                                                      Relative
                                                      Intensl ty
                                                      mm..
1.0
 Teflon

6.3mm O.D. glass

1.5mm  O.D. pi

 Stainless
.75
.50
1.0
1.1
.74
3.80
4.30
.87
3.0
3^
10
1.6
                                                                 27

-------
were tested.  Three different tube lengths - 23,  30,  and 53cm
tested.  Results of these tests are shown In figure 5.
     In all cases,  the lower ethylene flow rate showed a moderate
decrease In sensitivity.  The 23cm chamber had very good response but
the plateau was not as flat as previously experienced with larger
diameter chambers.   This  Indicates that complete  mixing Is not occurring
In the tube, and longer tubes are Indicated. The  30cm chamber still
didn't have a flat  plateau, and had considerably  less response than
did the 23cm chamber.The  lower response is due to greater reflectance
losses.  These losses are further indicated in the 53cm chamber. Here,
only the very high  flow rates gave a signal.  At  normal air flows,
the reaction apparently occurred so far down the  tube that its light
was completely absorbed before reaching the PMT.
  Both the 23 and 30cm reaction chambers showed better response than
would have been Indicated by the internal reflectance tests.  Whereas
the reflectance efficiency for the 30cm had been calculated at 3.2?,
the actual response was roughly 50% as great as the original Meloy
chamber.  This discrepancy probably is due to a better mixing efficiency
in the capillary tube than the Meloy chamber, since the Meloy chamber
never has shown any plateau characteristics on a  response vs. velocity
plot.  Other possibilities are an underestimation of the reflectivity
(a 1% error in the  estimated reflectance per bounce would have a
large effect on total efficiency) or to the bulk  of the reaction
occurring forward of the  tube's midpoint.  In any case, the test
results looked very encouraging.  If we compare the response per unit
area of the 23cm tube  (0.5mm  I.D.) with the original Meloy chamber
(19mm I.n.), we find a net decrease in area of 1,400 fold coupled with

-------
    .35
    .30
    .25
 I
n
c
o
a
>/)
0)
o
M
o
    .20
    .15
    .05
                                 Figure  5


                    Effect of reaction chamber  length
              X53CIH long
                                                      long
                                                      long
                                            28 ml./min.  ethylene

                                             7 ml./min.  ethylene
                                    3         «         5


                               I/Air flow in litera/min.
                                                                     29.

-------
only a slight decrease in total response.  Thus, the brillance was
Increased 1,000 foil1.  In comparison to a one cm4- area cell, brilliance
was increased approximately 250 fold.  Since previous calculations had
indicated that the brilliance of a hypothetical 1cm  area chamber must
be increased by between 10  and lO-5 before photographic film v:onld be
comparable to a PMT, we were still approximately two orders of magnitude
away from our goal.  Since the use of even smaller diameter reaction
tubes would begin to pose serious pressure drop, mixing, and plumbing
problems, we decided to terminate any further work on smaller chambers
and concentrate on the photographic film aspects.
     Various capillary tube reaction chambers including the 23cm
0.5mm I.D. tube, a 7.6 by O'.lcm I.D.tube, and others were used in
conjunction with various photographic films.  These films included
Polaroid 3000, Tri-X, and Ia-0.  Exposure times up to 4 hours were
used, and all developing was by manufacturer's recommended procedures.
Air flow rates and ethylene flow rates were varied over wide ranges.
The Meloy ozone generator (0.3ppm ozone) was used In all testa, and
proper operation was confirmed by checking light emission with the
Meloy PMT prior to each series of photographic tests.  In no instance
did we obtain any perceptible images on the films.  Even optical coupling
of the capillary reaction tube to the surface of the film (by using
appropriate quartz lenses) did not yield any images.  These results
merely confirmed the theoretical calculations which stated that several
orders of magnitude more sensitivity was still needed.  In view of the
increasing difficulty of obtaining additional significant sensitivity
improvements, this approach was terminated.  Perhaps the future will
                                                                   30

-------
yield filmj with better sensitivities (some of the newer solid state

imace storage plates already look promising).  In that event, the

present v/ork on improved mixing efficiency and smaller viewing areas

should prove valuable.

     *l.6 Gas Permeation Studies

     Even though the photographic film recording approach had not

proved fruitful, the use of small capillary tube reaction chambers

had opened up the possibility of using gas permation rather than

direct injection of ethylene into air.

     One of the major logistic disadvantages of chemiluminescent o^one

a.nalyzers is the necessity to supply ethylene.  If this could be

replaced by a reactive liquid whose vapors diffused into 'the chamber,

a much simpler system could be developed.  Consequently, the feasibility

of using a diffusion membrane was tested.

     Figure 6 describes the test setup used.  The air flowed past the

silicone rubber tube, through the capillary tube, and to the vacuum

port.  The ethylene was allowed to flow at 15ml. per minute through

the reservoir surrounding the diffusion tube.  Response dropped

severely when ethylene flow was stopped, presumably due to rapid

depletion.  The diffusion device was tested against direct injection

of ethylene into the end of the capillary tube.  The results are

shown in table 8.

                               Table 8

     Effect of Diffusing Ethylene into the Reaction Chamber
       (Responses given in equivalent pprn, ozone)

Air flow        Ethylene @        Ethylene @        Ethylene
 rate           28ml. /min.        7ml. /min.        diffused in
 2.01pm            .110              .095
 1.5               .1^5              .120             .090
 1.0               .190              .165             .120
 0.5               .275              .2^5             .175
 0.25              .280              .265             .170
                                                                   31

-------
                          Figure 6



                  Gas permeation tube test
to ethylene
vent
                           1mm. I.D. x 30cm glass tube




                                            #00 cover slip
                                                 PMT
            5mm O.D. silicon rubber wlth it3cin exposed

-------
     In all cases, the usual pattern of increasing response with



decreasing air flow, finally reaching a plateau, was found.  The response



decreased about 10/5 as the ethyl ene flow was reduced, from 28ml./min.



to Trcl./min.  With the diffusion device, response dropped another



30%.  Nonetheless, the response obtained by diffusing ethylene through



silicone rubber was quite respectable.  It opens the possibility of



diffusing liquid reaotants, thereby eliminating the gas bottle



requirement.



	4.7  Direct Air Injection



     Film manufacturers have a problem of film fogging due to the



chepiiluminescence reaction of ozone in the atmosphere with gelatin.



In order to overcome this problem, they use sealed shipping containers



and add antioxidants to the film.  It was attempted to use this



reaction ar, the basis of an ozone detector.



    One end of a 3«2mm Teflon tube was connected to the ozone



generator and the other end was positioned about 3.2mm from the film



plate.  Vacuum was applied to the camera case to create a flow rate



of one liter per minute.  No reaction v/ith Kodak Tri-X 16mm. film



could be found after one hour exposures.



     In other tests an air pump was connected to the ozone generator



inlet, and positive pressure was used to direct a one liter per



minute air flow onto the film.  Again, no exposure could be seen



after development



     It thus appears that either the 0.3 ppm ozone concentration



cannot overcome the antioxidant, or the one hour exposure time is much



too short.  In any case, this does not appear to be a fruitful approach.
                                                                   33

-------
5.0  Alternative Detectors
     Even though the direct film exposure concept proved to be
unfeasible, a drastic need still existed for a reliable, low cost,
long term recording monitor for air pollutants.  Although the
chemiluminescence reaction Itself Is simple and reliable, the extremely
low light emission level has made it necessary to use photomultipller
tubes as sensors.  The gain as well as zero level of PMT's is markedly
aJtered by changes in temperature and voltage.  Consequently, a portable
instrument would require a thermostatted oven (very high power consumption),
and a very well regulated power supply (expensive).  In an effort to
overcome these disadvantages, various other photodetectors were tested.
     In most cases, the test setup consisted of an 2^v. power supply
with 1.5 Kohm, 1.2 Kohm, 1.0 Kohm and 820 ohms current llmiters
(positions 1,2,3,11), driving a 2l
-------
                                         TABLE 9
                         Response of Various Photodetectors
       Detector
       r:elov  PUT
       reloy  PI.T
       RCA  1P28  PMT
       ?CA  1F28  PI-T
       CL902L
       PD 1900-Sl
       PD 1900-S10
       DT 1737
       CLR  2060
       CLR  2180
    Test Condition
O^one amplifier, 1200v.
Electrometer, 1200 v.
Ozone amplifier, 800v.
Electrometer, 800v.
Electrometer, 15v.
Electrometer, 15v.
Electrometer, 15v
1 r-eg x 1000 (lmv=lpa.)
Electrometer, Iv.
Electrometer, Iv.
 Dark    Lamp 1  Lamp 2   Lamp 3   Lamp 1
 Onpm.    0.3    5.5     Overran^e
 OuA      .01    .13     1.38      3.18
 Oppm
 1.37nc
 2.00na
25.Oca. 25.2
I2.0pa. 12.0
41.7pa
.01
1.13
2.01
5.2
2.0


1.85
6.53
.086
1.85
2.04
25.2
12.0
4fi 7
TO . (
1.87
6.67
.25
2.75
2.05
28.0
43.0
h Q 7
T y . /
2.09
6.77
1.10
8.72
2.25
29.0
11.3
P Q n
O j ,U
2.31
21.1
OJ

-------
     The first test with the Meloy PMT showed the correlation between



our simulated light conditions and the chemiluinlnescent light emission.



The actual current output of the Meloy PMT was directly related to the



ozone reading.  The 1P28 PMT had considerably less response than the



Meloy PMT.  This would be expected in view of the lower operating



potential.  The Clairex CLQ02L cadmium sulfide photoresistor tracked



the simulated light conditions a.lmost as well as the 1P28.  It had a



higher relative dark current and its response was exceedingly slow.



     The next three detectors are vacuum photodiodes manufactured by



the Tung-Sol division of V/arner Electric Corporation.  The PD 1900-S4



appeared to follow the light source somewhat, but the signal to dark



current ratio was very poor.  Also, the current increase equivalent



to 5•5ppm ozone was only 0.2 picoamperes, which makes it difficult to



design suitable amplifier circuits.  The PD 1900 with S10 response



characteristics gave even poorer response, presumable due to its



greater sensitivity to red light, which J.s a noise and dark current



producer in this application.  The DT 1737A is a PD1900-S10 photodiode



coupled to a  PET amplifier.  It is advertised as being equivalent to



photomultiplier tubes.  Our tests indicated that it did indeed follow



our simulated signals, but that its noise and drift characteristics



were unacceptable.



     The CLR 2060 and CLR 2180 detectors are silicon photodarlingtons



sold by Clairex.  They differ only in that the CLR 2180 has a. lens



to collect light.  Even though these detectors have maximum sensitivity



to 900nm light and only about 5% as much sensitivity to l»00nm light,



both detectors were able to measure light intensities equivalent to



     ozone or greater.  In going from darkness to lamp ^, CLR 2060
                                                                   36

-------
Increased by 0.46na.  while CLR 2180 .increased by l4.6na., thereby
showing the importance of good light collecting ability.
     The preceding results cannot be strictly related to detectability
of cheip.ilurnlnescence  because of their disproportionate sizes.  Since
these tests were made with a small diameter (3.2mm)light source, all
detectors collected a high proportion of the available light.  With
a large diameter (over 2.5cm)    reaction chamber as used in the Meloy
ozone analyzer, only  the Meloy PMT (which has an end-on configuration)
will collect the light efficiently.  The 1P28 PMT, which has a side
window configuration  collects less than one percent of the light.  No
signals due to the ozone reaction were obtained with the 3.P28.  The
vacuum photodiodes would collect about 20? of the light but their
responsivity characteristics are poor.  The solid state detectors collect
very small  proportions (less than 0.15?) of the light and have poor
responsivity characteristics.  It was concluded that only end-on
photomultjp]ier tubes are capable of monitoring the ozone-ethylene
chemllumlnescent reactions at this time.

6. Photon Counting
     6.1  General Background
     All light originates as discrete energy particles termed "photons".
The eye, most detectors, and most amplifiers are too slow to respond
to individual photons.  Consequently, the output of most photodetectors
appears to be an analog signal, the magnitude of which is proportional
to light intensity.  With a sufficiently fast response system (risetimes
less than one microsecond), the detector output appears as a number of
sharp spikes, each of which represents a single photon.  The height of
                                                                   37

-------
 the spikes  remains  constant  regardless  of  light  intensity.   Increased
 light  only  results  in Increased  numbers of spikes.
      A recent  article by  Zatzick (1)  reviews  the  advantages  and
 general state  of  the  art  of  photon  counting.   Photon  counting  is
 universally agreed  to be  the  ultimate method  for  obtaining the best
 reliability and signal-to-noise  ratios  in  low light level applications.
 In  photon counting  only the  number  of photons are counted.   Electronic
 noise  is eliminated by threshold  circuits  and dc  drift is eliminated
 by  using ao  coupled amplifiers.   With conventional analog systems,
 not  only the number of photons affect the  signal, but also their peak
 height, the  frequency of  cosmic  rays, electronic,  noise, and  electronic
 drift.  The apparent  photon peak  height  is governed by the gain of the
 photomultlpller tube  and amplifier.   Although  amplifier gain Is stable,
 photomultiplier gain  Is greatly  influenced by  power supply voltage and
 is moderately Influenced by temperature.  This means that a very well
 regulated power supply must be used,  and a. controlled temperature oven
 must be provided for the PMT If it is used in  the field.   Cosmic rays
 strike the photomultiplier tube at a rate of approximately one/cm2/sec.
 Since they are orders of magnitude larger than photon signals,  they
 can have an appreciable effect on the baseline at low light  levels.
 Electronic noise is  another source of baseline drift.   Although it  is
 simple to remove this source of error by threshold limiting  with a
photon counting circuit,  it becomes  an appreciable signal in  analog
circuits.   Electronic drift of both  the  PMT and amplifiers is an
exceedingly  difficult problem to  solve with standard  analog  circuits.

1.   ZatzJck, M.R.  Electro-Optical Systems Design,  June 1972,  pp.  20-27
                                                                   38

-------
Photon counting circuits always are AC coupled so as to eliminate this
problem. Thus, the photon counting technique eliminates many error
sources which limit the sensitivity and reliability of conventional
systems.
     6.2 Electronic Considerations
         6.2.1 General
     Although photon counting has nany theoretical advantages over
standard analog integration procedures, the electronic circuitry must
be relatively sophisticated to avoid introducing nev/ problems.  Fig.
7 illustrates the basic components required to monitor chemiluminescence
reactions by photon counting techniques.  A photomultiplier and its
power supply are needed Just as in an analog system.  The pre-amplifier
serves  to lower the impedance of the signals.  The comparator discriminates
between the photoelectron signals and noise, and amplifies the photoelectron
signals to a saturation level voltage.  The divider network reduces
the total counts to manageable levels, while the readout system permits
a number representative of the pollutant level to be recorded.  Perhaps
the most difficult area is in obtaining adequate risetime speeds  from
the PMT amplifier.  If the response speed is too slow, the photoelectron
signals will appear as broad peaks instead of sharp spikes.  This
results in a poorer signal-to-noise ratio and in peak overlap at  higher
light levels.  The peak overlap causes doublet and triplet photoelectron
peaks to appear, which results in an exponential power of 2 or 3
relationship between  counts and light  intensity, instead of the normal
linear  relationship.  The other major problem area if?  in handling the
massive number of  photoelectron counts.  Since approximately 1000 photons
riust be counted each  second, a 24 hour accumulation would result  in
                                                                   39

-------
  Co
  K
  o
  o

 •o
  o
  ct
High
Voltage
Power
Supply



PMT



Pre-amn




Corcparato



Divider



Readout

 O
 3
 H-
 D
OQ
 C

 0>

—J
 o
 0)
 3
 O
 ill

•a
 D'
 O
 ct
 O
 3

 o
 o
ct
ft

-------
approximately 100 million counts.   This is handled most easily Ly using



a "brute force" approach of digital division.



          6.2.2. Power Supply



     The high voltage power supply can he relatively poorly regulated



since amplitude changes of the photoelectron signals do not have a



direct influence on the number of counts.  The current capability of



this power supply need be only one-tenth as great as required for



analog systems, because the dividing chain can be composed of much



higher resistance values.  This is possible because each photoelectron



places a constant current load upon the PMT, as opposed to the variable



load imposed when the PMT is operated into an integrating current amplifier.



     Almost all of the lower cost high voltage power supplies operate



as dc-dc converters.  The Input d.c. power is chopped at approximately



10KII-7. transformed to a higher voltage, and then rectified to obtain a
    fj *


high voltage dc output.  The chopper frequency appears at the output



as a ripple, which ultimately becomes noise.  ?ince a photon counting



system can discriminate against this ripple, it can tolerate the vise of



these low cost power supplies.  All studies on photon counting were



made with a Mi]. Electronics type VL15 power supply.  This unit transforms



input voltage to a voltage 100 times greater.  It was normally operated



with a 12vdc input and a 1200 vdc output.



          6.2.3.  Photomultlplier tube



     Most of the requirements for a photomultiplier tube used for



photon counting are  similar to those for normal analog usage.  In



order to obtain best sensitivity and reliability, it should have



minimal response to  longer wavelengths,  since longer wavelength response



Increases the thermionic noise problem.  It should also have the minimum

-------
possible cathode area,  since excess cathode area increases thermionic



and cosmic ray noise.  DC leakage is unimportant for photon counting



purposes, so dark current values are not especially meaningful.   A



special requirement is  that the PMT must have fast response characteristics.



     The EMI 9524S PMT  was chosen for this application.  This  30mm.



end-on tube has an S-ll spectral response.  It was primarily chosen



for its overall reasonable characteristics, rugged construction,  and



low cost.  It has a somewhat slow rise-time response of 18 nanoseconds,



but this is adequate if only low light levels are measured.  Even



though this is a moderately small diameter tube much smaller tubes are



available at higher costs.  Alternatively, a magnetic focusing lens



could be used to reduce the effective cathode diameter.  This tube was



used as is, however, since it fit the existing Meloy reaction chamber



and permitted us to proceed immediately to the photon counting problem.



     The dynode chain configuration used with the PMT is shown in



figure 8.  This is a relatively standard configuration except that



the chain resistance is ten-fold greater than normally used, and



capacitors are used on the last three stages to provide a more constant



gain regardless of the  photon rate.  Higher resistance is possible with



photon counting circuits because less current is required at the  low



light levels generally  encountered.  It should be noted that the  plate



is tied to ground through a 22K ohm resistor.  This permits the output



to be monitored either  by the photon counting amplifier or by a



conventional dc analog  amplifier.  This permitted us a high degree of



flexibility in our experimental tests.  A major disadvantage of this



configuration is that the PMT shield must be tied to the high voltage



supply.  A much safer and lower noise configuration v/ould be to tie

-------
J>  •'*
                                 £N\1  V24S
       3WI8LD
   —/loov-
                                            I    I
                                                    T
                                                 00 f
                                               I  -OO/
                                                                                                            0
                                                                                                        7
-------
the cathode and shield to ground, and tie the load resistor to the



positive high voltage output.  This would necessitate an ac coupled



circuit, which would preclude any do current measurements.  Nonetheless,



if only photon counting were to be used, it would be recommended that



the cathode and shield be grounded.



	6.2.^.  Amplifier Circuit



     Also shown in figure 8 is the final amplifier circuit delivered



in the breadboard, instrument.  The 22 Konm load resistor was found to



be the largest possible resistance that would not unduly broaden the



photoelectron signals when unshielded wire was used between the plate of



the cathode and the preamp]ifier.  With the 22Kohm resistor, it v/as



not possible to use shielded cables without broadening the signals due



to the extra capacitance.  This did not prove troublesome since shielded



cables are not needed for photon counting purposes.  The Harris HA-2000



preamplifier is a newly introduced FET Input, unity gain amplifier which



is capable of following 100 MHZ signals.  This preamplifier was very



satisfactory and was used in all experimental configurations.  A slow-



response amplifier was used at the output o.C 1;he preamplifier r;o that



EPA personnel could conveniently monitor dc current vs. photon counts.



Very little work was done with this amplifier since v/e preferred to



disconnect the PMT output lead and measure current directly whenever



a comparison test was made.  The auxiliary slow amplifier is completely



functional and does reflect the true dc state of the chemiluniinescent



reaction.  The output of this amplifier is quite rippled and drifts



womewhat since this is typical of analog systems.  The Meloy amplifier



output appears smoother because it has approximately a 10 second time



constant.

-------
      The  output of the preamplifier also  IP  fed  through a coupling

 capacitor to an PET amplifier  stage.  A diode was used between these

 stages  to prevent negative peak overshoot, which would unnecessarily

 increase  peak doubling effects.  This amplifier  stage presented the

 greatest  difficulties in achieving adequate  speed and stability.  Single

 transistors, as used here, are not thermally stable unless additional

 compensation circuitry IP used.  Integrated  circuit amplifiers would

 be the  ideal choice if available.  Various amplifiers, including the

 LM101 with feedforward compensation and the  Harris HA-2520, were tested

 in various configurations.  In all cases, a  linear re] ationRhip

 between current and count rate could not be  achieved.  Rest re
                                                                 i
were obtained when the IIA2520 was direct.! y coupled to the KA2000

preamplifier in a standard two amplifier feed-back arrangement.  Under

these conditions, a power slope factor of 1.1 was achieved under low

light level conditions.  Even though the HA2520 has a very impressive

slew rate of 120v//us, the high gain demanded of this amplifier slows

it to a point where response time is greater than one microsecond.

This causes peak broadening, which results in some photon doublets

that tend to degrade linearity (note:  if only doublets were present,

the photon count would be proportional to the square of the current).

     In view of the poor thermal characteristics of the amplifier

section, its output is capacltively coupled to the comparator.  This

comparator (a Falrchild uA710) is used with open loop gain so that

any peaks exceeding a preset  threshold value become amplified to

the comparator saturation level.  The comparator used in this circuit

performed quite well.  However,  a far superior comparator has recently

become available - the Advanced  Micro Devices Am685.   This comparator

-------
 has  a  2ns resetime Instead of the 40ns risetime of the yuA?10.  Even
 more important, it has excellent stability (10uV/°C. drift) which permits
 i.t to  be used directly with very small signals.  This makes it possible
 to couple it directly to the preamplifier, or to use an amplifier with
 low gain, thereby minimizing rise time.  Thus, this comparator makes
 possible an amplification system made up entirely of stable temperature-
 compensated integrated circuits. Unfortunately, this comparator came
 on the market too late to be incorporated in our existing circuitry.
 This comparator uses ECL logic outputs instead of TTL as does the older
 710 comparator. While this does not represent a difficult interface,
 it would have entailed making an entire new circuit board.  Since the
 photon counting technique could be easily shown to be far superior to
 analog systems, the new comparator was not incorporated into the system.
     The output of the comparator is available on an output jack
 labeled "counter".  This makes it possible to use any standard digital
 counter to monitor the rate of photon production.  The signal then
 proceeds to the "digital division" portion of the circuit, where the
 photon pulses are counted.
	6.2.5  Digital Division
     Figure 9 is a functional diagram of the decade division circuitry,
while the complete circuit is shown in appendix A.  The pulsed output
of the comparator is  fed to the array of decade dividers.   F.ac'n stage
performs a 10-fold division of the number of counts.   Thus,  a single
pulse appearing at the output of the third decade divider  would
represent ]000 counts at the input.   A rotary switch  is used to select
which decade divider  output (the extent of division)  rshall be permitted
to accumulate.   The Dast decade divider outputs are decoded  to binary

-------
                                P )  Voltmeter
                                                                             4 )Counter
| 1200vdc


FH? I
i

1 i.'A 2005
!. Preamp

—S


AC
Amp.


Conpar
at or

-!4 1 10
< i •
Y
TJ
1 !
 o
 ct
 O
 3

 O
 O
to
• P.
O
c
ct
O
e
H-
    ^
    0)
         1 10
j. 10
Start Timer
v  {Tim
*LL
               2?imer
                27
                  Sequence
                    y A
                      	j
    I  \-
    Start Cnunl-.Jnc
                               Enter  liemoj'v
                               Transfer I-iernory
                               to  Display



                               Advance F.ov/
r 10
                                            i-p-i r 1Q i
                                                      X1
                                           Gate!
                                              f  10
r 10



^^^
Over-
range
                                                                      Binary
                                                                     ! 7 Bit
                                                        I
                                                                       Res*
                                                        Tin
                                       i Fulse j
                                         7s-   r
                                              	_^ Memory
                            I Row H "
                            JAdvance
                                                      Display

-------
 form by a 7 bit binary counter.  This counter provides a 7 bit parallel



 output that is related to pollutant concentration.  This provides a



 resolution  of one part in 127.  Greater resolution is possible simply



 by using larger counters.  The 7 bit accuracy was chosen since a 5x7



 LED matrix was chosen as the readout device.  Naturally, once the



 pulses are in digital form, any desired resolution can be obtained



 simply by using binary converters and displays with greater capacity.



     The counting time is regulated by an RC network attached to a



 digital timer (SE 55).  A rotary switch changes the timing period in



 decade- steps from 0.1 minute to 100 minutes.  At the end of each



 timing period, a number of events occur.  These are:



     1.  Replace the previous counts in the memory with the accumulated



 counts from the last time period.  These counts will be stoker! in the



 memory during the entire next timing period.



     2.  Advance the LED readout by one row and display the new counts



 during the next courting period.  This results in a 7 bit nunber in one



 row of the LED display.  This row is advanced during the start of each



 new timing period.  In addition, the circuit shown in Appendix A also



 incorporates a "Blip" capability in which the LED matrix is dark during



 most of the timing period.   The binary coded value only appears as a



 0.1 second blip at the end  of the timing period.  If this is coupled to



 a photographic recording system, a very low power system can result.



     3.  Reset all dividers to zero.  This readies them for the next



 timing period.



     *J.  Reset  the timer.



     This logic  system thereby results in a system in which both the



amount of count  division and  counting  time can be independently varied.

-------
 This  d'.-gree  of  flexibility  probably  wouldn't  be  required  in  a  field
 instrument,  but  it  is  helpful  for  R  &  D  purposes.  The  f.  x 7 LED matrix
 displays  a binary coded  number  that  is proportional  to  pollutant
 concentration on one of  its  7  bit  rows.  This  display row is advanced
 one row at the  end  of  each  timing  period in order to display the new
 reading.  This  system  has functioned very well with  no  failures at
 any time.
      6.3  Results
      All  data were  obtained  either by  using the  Meloy ohemilumlnescence
 chamber or by using a  light  bulb simuDator having a  filtered output of
 410 nm.   Photon,  counts were  plotted against "ppm ozone" or against
 "analog output current"  whose values were obtained by applying exactly
 the same  experimental  conditions while obtaining the analog or digital
 data.
      When analog output  currents were compared to photon  counts, an
 extremely non-linear relationship was obtained in most cases (in early
 experiments).  Reasonably linear relationships could be obtained in
 many  cases if the count  rate was raised to an exponential power
 between 1.5 and 3.  Close examination of the photon shape on a
 Tektronix oscilloscope revealed that peak widths in excess of one
 microsecond were causing doubling-up of the photon signals.   With
 the high speed, circuitry shown in figure 8,  however,  good linearity
 was achieved in practically all cases.
     A plot of photon counts vs. analog current is shown for various
 threshold level settings in figure  10.   The  count rate  is displayed
on the Y axis,  while the equivalent analog  current is shown  on the
X axis.  The different  corves are labeled according  to  the threshold

-------
""tgure
                      Count rate vs. Current at various threshold levels
•s
o
o
0)
tn

fn
V
A

at
^>
o
o
           -    - -'^. '•
                                                       l^jfl§»i
                                                     BS^Sl :
                                           ^       . -.«j:-n

                                           •   •  •  • 'JtfM
                                            -  !'> i ' i:
                                           •;.;- •: ::"
               ;-«   1
                                                        :  li^^
                 UlU
                           6   8 10       20    30  10   60  80 100    200


                           Current in Nanoamperes                  50

-------
voltage applied to the comparator.  Zero volts corresponds to essentially



no threshold offset, whereby all counts are recorded.  This curve has



a slope less than unity because many of Its counts are derived from



electronic noise, which remains constant regardless of light level,



thereby yielding a fairly horizontal slope.  The 0.8v. threshold



provided a slope greater than unity.  This is presumably due to counting



many "doublet" photon signals composed of two photon signals appearing



at the same time, thereby yielding a signal twice as intense.  All the



curves obtained with threshold voltages between 0.06 and 0.6 volts



are quite linear, thereby confirming the good high speed capabilities



of the electronic circuit.  From these data, It may be Inferred that



most of the noise pulses are smaller than 0.06 volts, and that most



of the photon signals have amplitudes between 0.06 and 0.6v.



     The data of figure 10 is reformated in figure 11 to show the



relationship between threshold voltage and count rate for various light



levels. As the threshold level is reduced from 0.06v. towards zero, the



dark count rate increases very rapidly.  This is due to counting



electronic no3se pulses near the zero threshold level.  The region



between 0.5 and 0.6 volts also is characterized by a steep slope.  This



is presumably due to a transition from counting only single photon



signals to counting doublets.  The region of least slope lies between



0.3 and 0.5 volts.   This is presumably the peak height of the average



photon signal.  If the photomultlplier tube provided precisely the



same gain for all photons., this region would be perfectly horizontal.



If this were the case, the photon count would be completely oblivious



to small changes In PMT gain.  As it is,  changes in PMT gain will affect



the count rate,  but much less severely than If It  were operated  in the
                                                                   51

-------
10 OK
 10K
•d
c
o
o
a;
CO
  IK  •
0)
a

w
-p
t} _
      2..
 100
  10
                             Tv
                           r • j-\
                        ! ,    •
                         t:   h'  rr.m
                                Threshold (Volts)



                            Figure 11



                  Threshold vs. Counts at Various  Light Levels

-------
 analog mode.  lietter  stability would he possible with gallium phosphide



 PMT  cathodes which generate 30 electrons per photon.  These PMT's are



 more expensive  than the simple PMT used for these tests.



      A plot of  count  rate vs. LED matrix readout is shown in figure 12.



 As would be expected  with a digital system, a perfect correlation exists



 A plot of PMT output  current vs. ozone reading is sh^wn in figure 13.



 This also shows a good linear relationship, as would be expected.  The



 relationship between  count rate and ozone concentration is shown in



 figure 14.  A threshold setting of 0.2v. was used to obtain these data.



 It may be seen  that very good linearity was obtained.



      In summary, the photon counter exhibited very good linearity over



 a wide dynamic  range.  The threshold voltage setting for the comparator



 could be set between 0.06 and 0.6 volts and still achieve linearity.



 Lower thresholds settings tended to Include electronic noise,  while



 higher thresholds Included primarily those photons that happened to



 appear at the same time as another photon.  The primary known  weaknesses



 of the present system - a cheap Pi'T and less than optimum analog



 amplification probably were responsible for a dependency of count rate



 upon threshold voltage setting.



 7.   Suggestions for future work



     The feasibility of applying photon counting principles  to



 chemllumlnescence detectors has  now been shown.   Sensitivity is  now



 limited only by thermionic noise  - amplifier drift,  cosmic rays,  power



 supply drift,  etc.  create little or no  interference.   It is  suggested



that  the already good  performance of the present  photon counter  can be



Improved at  least 10 fold by  utilizing  a small  area  detector to  decrease



the  thermionic  noise contribution.   This will be  especially  important
                                                                   53

-------
                                                      3 4567891
100K

                                           I -I I = __JI__L__~ ,-lLl-
                                          ' i I   ~ li.  'i .
 100 iL.
                          Matrix  Readout
                      Time Base  = 6 seconds
    Figure 32 Correlation  of  Count Frequency with  Display Reading

-------
   P.O
   1.5
0)
n
o
N
O


D.
P.
1,0
    .5
                10
                                  ?0
                             Current  Output (na)
                 Figure 13 PfTT  Current  vs.  Meter Reading

-------
   2.0
0)

o

o
tr
                      Counts/sec/lOQ/O
               Figure lii  3?;one  Concentration vs. LED
                           Matrix  Readout

-------
 when  longer wavelength reactions such as NO  - ozone are monitored.
                                           X

      As a further aid to thermionic noise reduction, some of the


 newer liquid crystal films such as N-(p-ethoxybenzylldene) -p-n-


 butylaniline should be tested as a chopper.  This will function to


 alternately sense the background alone vs. the background plus signal


 without the complexity of mechanical choppers. This type of light


 chopper could be easily incorporated into photon counters simply by


 using up-down counters.  The "on" cycles would count up, while the


 "off" cycles would count down.  This then, would represent the ultimate


 in high reliability signal detection.


      On the commercial aspects, the present transistor amplifier


 should definitely be replaced by an integrated system employing the


 new AM685 voltage comparator.  This will provide faster response


 times, much greater reliability and stability, and a greater linear


 dynamic range.  Also, the "brute-force" chain of decade dividers


 should be replaced by an integrated counter such as the Mostek MK5009P.


 This  single chip device can divide by as much as 3.6x10^, it's division


 is completely programmable, and it uses considerably less power than


 does  a comparable number of TTL decade dividers.


      The readout device can be any number of digital output  devices.


 This  includes numeral displays,  printers, a computer,  teletype lines,


 etc.  In the device delivered,  a 5 x  7 LED matrix was  used with the


 idea  that a permanent record  could be made on photographic film.   A


more  convenient readout would be a magnetic card such  as used with


 the Hewlitt-Packard or Monroe desk-type computers.   With this type of


 recording card, over a month's  operation could be stored on  one card.


The card could then be entered  into  a computer to be collated with other
                                                                  57

-------
data.  In any case,  once the data Is  in digital  form,  many  readout



options are available,
                                                                 r
                                                                  8

-------
        APPENDIX A




Digital Division Circuitry
                                         59

-------
 PAGE NOT
AVAILABLE
DIGITALLY

-------
        APPENDIX n




Photon Counter Description
                                        60

-------
C.
          S   3
        'x  A
                          07
BUP
                          orr
                           10
                        100
                            MIMTES
                   *>—-o  o
                            3 V—©
                   LF

-------
                                 APPENDIX B

                      Description of the Photon.Counter


1.  Electrical connector - llOv., 50-60 HZ.

2.  Photomultlplier tube - nates directly with Meloy reaction chamber.

3.  Preamp output - 0 - O.lv.  photoelectron signals  prior to amplification.

4.  Counter output - 0 5v. digital signals that can  be connected to any
    digital counter.

5.  Meter output - 0 - 10v.t damped analog output.   This signal corresponds
    to the conventional meter output used In most commercial chemiluminescence
    monitors.

6.  On - off switch - this unit has essentially no warm-up time.

7.  Indicator light - Operates from the main power supply to confirm that
    the unit Is on.

8.  Time adjust - sets the cycle rate to 0.1, 1, 10, or 100 minutes.

9.  Divider - Divides the number of photon counts   by exponent Indicated
    (100 to 10 million).  This should be set so that the LED readout (12)
    provides a reasonable reading,

10. Overrange Indicator - Lights and stays on whenever the capacity of the
    LED readout (12) Is exceeded (over 127 units).

11. Reset - This button should be pushed to turn the overrange Indicator off.

12. LED readout - Contains 7 horizontal lamps and 5  vertical rows.   The
    7 horizontal lamps Indicate a binary number between 1 and 127 which Is
    proportional to ozone concentration.  Each row normally indicates the
    accumulated photon counts for the previous cycle.  After the current
    cycle time Is completed, the lighted row will advance upward one row
    and display the latest reading.

13. Blip control - Inhibits the LED readout from constant indication.  In
    the "Blip11 position, the LED readout will display the accumulated counts
    for 0.1 second at the end of each cycle.

14. Threshold control - A screwdriver adjust potentiometer that permits the
    noise discrimination threshold to be Increased or decreased.

-------