EPA-R4-73-016c
March 1973
Environmental Monitoring Series
A Coupled Two-Dimensional  Diffusion
and Chemistry Model for Turbulent
and Inhomogeneously Mixed
Reaction Systems
                          Office of Research and Monitoring
                          U.S. Environmental Protection Agency
                          Washington, D.C. 20460

-------
                                            EPA-R4-73-016c

      A Coupled  Two-Dimensional

    Diffusion  and  Chemistry Model

for  Turbulent and  Inhomogeneously

         Mixed Reaction Systems
                           by

             Glenn R. Hilst, Coleman duP. Donaldson,
          Milton Teske, Ross Contiliano, and Johnny Freiberg

         Aeronautical Research Associates of Princeton, Inc.
                     50 Washington Road
                  Princeton, New Jersey 08540
                   Contract No. 68-02-0014
                  Program Element No. A-11009
              EPA Project Officer:  Kenneth L. Calder

                    Meteorology Laboratory
              National Environmental Research Center
            Research Triangle Park, North Carolina 27711
                       Prepared for

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

                        March 1973

-------
This report has been reviewed by the Environmental Protection Agency



and approved for publication.  Approval does not signify that the



contents necessarily reflect the views and policies of the Agency,



nor does mention of trade names or commercial products constitute



endorsement or recommendation for use.
                                    11

-------
                  TABLE OP CONTENTS
    Nomenclature
1.  Introduction
2.  The Problem
3.  An Evaluation of the Effects of
    Inhomogeneous Mixing
4.  Closure of the Chemical Sub-Model
5.  The Construction of a Two-Dimensional Coupled
    Diffusion/Chemistry Model for a Binary
    Reaction System
6.  Some Calculations of the Interactions of
    Turbulent Diffusion and Chemistry
7.  Conclusions and Recommendations
    Appendices
A.  Effect of Inhomogeneous Mixing on Atmospheric
    Photochemical Reactions
B.  Chemical Reactions in Inhomogeneous Mixtures:  The Effect
    of the Scale of Turbulent Mixing
                           iii

-------
                        NOMENCLATURE
A.         moment term defined by Equation 4.5
B,         ratio of mean concentrations
           ratio of reaction rates

C,         concentration of ith chemical species
f          moment term defined by Equation 4.7
F          horizontal flux defined by Equation 6.3
g          acceleration due to gravity
k          reaction rate constant
M          value of  -=-   when  C ' CA = 0
                     77 TT          a  B
n.         frequency distribution of ith species
N          ^n±
p          pressure
q          square root of twice the turbulent kinetic energy
r          correlation coefficient defined by Equation 3.15
S          wind shear
t          time
t.         intermittency factor
T          absolute temperature
T          adiabatic temperature
u          velocity.
x          axial coordinate
                              IV

-------
z          vertical distance



X          micro-scale length



A-,         length scale



AO         length scale



A,,         length scale



v          kinematic viscosity



a          standard deviation
 Z




Superscripts



—         mean component



 1          fluctuating component






Subscripts



chem       reaction rate due to chemistry



I          reaction rate neglecting third-order correlations

           (Table 1)                   • .   •



s          steady state value (see e.g. Equation 3.12)



a, &,-y,f)    chemical species



o          initial value



1,2        species.designation
                              v

-------
                                                             1-1
                      1.  INTRODUCTION

      This section of the A.R.A.P. final report on Contract'
EPA 68-02-0014 covers the work on modeling of chemical
reactions in turbulent and inhomogeneously mixed binary
reaction systems performed during the period. September 1972
through January 1973.  The primary intent of the EPA-supported
portion of this work has been the assessment of the combined
effects of turbulent diffusion and inhornogeneous chemistry on
the dispersion and chemical alteration of reactive pollutants
and natural constituents of the lower atmosphere.  Parallel
programs aimed at similar assessments in the lower strato-
sphere,, particularly as they pertain to • the impact of proposed
SST exhaust emissions on the natural environment, have been
supported (under Contracts NAS1-11433 and NAS1-11873) by
NASA/Langley and, by transfer of  funds, by the DOT/CIAP
program.  Credit for the support  of the basic technological .
developments common to these problems is therefore shared by
EPA, NASA,  and DOT.  The applications of this technology
reported here are restricted to the EPA orientation, however,
and have been supported solely "by that Agency.
      At the time this work was undertaken in mid-1972, a
general assessment of the potential importance of inhomo-
geneous mixing in chemical reaction rates, and the basic
approach to modeling these effects via second-order closure
of the chemical kinetic equations had been developed by an
in-house program at A.R.A.P.   These earlier developments
were reported in two papers by Donaldson and Hilst, one
entitled "Effect of Inhomogeneous Mixing on Atmospheric
Photochemical Reactions, " which was published in ENVIRONMENTAL
SCIENCE AND TECHNOLOGY, Volume 6, September 1972, and a second
entitled "Chemical Reactions in Inhornogeneous Mixtures:  The  '
Effect of the Scale of Turbulent Mixing, " which appeared in

-------
                                                             1-2

the Proceedings of the 1972 Heat Transfer and Fluid Mechanics
Institute. Stanford University Press, June 1972.  Since the
work reported here stems directly from these earlier considera-
tions, reprints of these papers are appended to this report.
The reader is urged to read these appendices (A and B) first,
if he is not already familiar with their contents.
      In addition to this preliminary work on chemical kinetics
modeling, several years of effort at A.R.A.P. have been devoted
to invariant modeling (second-order closure) of the structure
of turbulence and turbulent diffusion in boundary layer shear
flows under various conditions of hydrostatic stability and
surface roughness.  This facet of model development has also
been sponsored by several agencies, including work under the
EPA Contract EPA 68-02-0014, and is reported extensively in
Volumes I  and  II of this final report.  In the present report
on coupled diffusion and chemistry models, familiarity with
the material in Volumes I and II will be assumed.

-------
                                                             2-1
                       2.  THE PROBLEM

      The objective of  this program, and therefore the primary
problem which we have  addressed, has been to fabricate a use-
ful coupled model which, can simulate the combined effects of
turbulent diffusion and chemical depletion on the concentra-
tion patterns of reactive chemical species emanating from
common or separate sources.  However., against the background
of assessment and modular model development described in the
previous section, it was evident at the initiation of the
present program that two major problems had to be solved first,
      1.  More rigorous analyses were required in order to
determine the magnitude of the effects of inhomogeneous mixing
on chemical reaction rates,  the conditions under which these
effects could be.realized, arid an evaluation of the likelihood
that these conditions actually occur in atmospheric pollutant
situations.   For example,  if it could be shown that these
effects were either always insignificant, or constituted only
a transient perturbation of the chemical kinetic rates, a
coupled diffusion/chemistry model could be readily constructed
using the conventional mean-value chemical kinetic equations.
      2.  Given that the results of the above analyses were
not totally negative, i.e.,  negative in the sense that no
important real-world situations could be found in which
concentration fluctuations played a significant role, it
was recognized that the second immediate problem was the
development of a useful closure scheme for the third-order
correlations inherent in the complete chemical kinetic
equations.  It was also recognized that under some circum-
stances the third-order correlations could be neglected.
However, this assumption restricts strongly the range of
joint frequency distributions of reacta.nt concentrations

-------
                                                             2-2
which can be considered, and denies any semblance of generality
in the chemical sub-model.  A more appropriate,  although
approximate, closure scheme was required.
      With these considerations in mind, first efforts were
devoted to these two problems.  By mid-November 1972 both, had
been resolved, and attention was focussed on the assembly of
the first coupled diffusion/chemistry model.  The results of
the earlier work on the analyses of the magnitude and signifi-
cance of inhomogeneous mixing on chemical reactions and the
development of a closure scheme at the level of third-order
correlations of concentration fluctuations have been assembled
as a technical paper which was presented at the llth Aerospace
Sciences Conference of the AIAA in'Washington, D. C.,
January 10, 1973.   A slightly modified version of this paper
is included as Section 3 and 4 of the present report.  The
major results discussed there are:
      1.  There are indeed real-world atmospheric pollution
problems in which neglect of the fluctuations of concentrations
of reacting species introduces significant errors.  These
effects associate primarily with multiple source situations,
many of which are very common.in the urban pollution arena.
      2.  An approximate closure scheme, for the chemical
sub-model which conforms to the principles of invariant
modeling and which accounts for the effects of inhomogeneous
mixing over a wide range of conditions  (concentration
variance-to-mean-squared ratios up to 100) has been developed.
This sub-model predicts reaction rates to within a factor of
two of the exact chemical kinetic solutions for situations
where the mean-value chemical kinetic approximation Incurs
errors of a factor of 100.  On the other hand, the chemical
kinetic sub-model recovers the mean-value approximation when
the concentration fluctuations are indeed insignificant, in
 G. R. Hllst,  "Solutions of the Chemical Kinetic Equations
 for Initially Inhomogeneous Mixtures, " AIAA Paper No. 73-101,
 January 1973, Washington, D. C.

-------
                                                             2-3
chemical reaction rates.  This second-order closure model may
therefore be considered as a generalized (but still approxi-
mate) solution of the chemical kinetics equations.

-------
                                                           3-1
              3.   AN EVALUATION OF  THE EFFECTS
                    OF INHOMOGENEOUS MIXING

 The  Basic  Chemical  Kinetic  Equations for
 Inhomogeneous Mixtures
       Following  Donaldson and  Hilst  (Appendix  A)  vie. assume  an
 isothermal,  irreversible, two-body reaction  between chemical
 species   a  and   p   to  form -y  and  6  .

                         a  + p -•• 7 + 6                    (3-1)
 Further,  we  assume  that the reaction rate for  any joint  values
 of the concentrations of the reacting chemical species are
 correctly  specified by

and
where  C.   denotes  the  concentration  of the  ith  chemical
"species  (expressed  as a mass  fraction),  and   k,   and   kp
are  the  reaction  rate constants.
      Equations  (3.2) and  (3.3)  specify the  local instantaneous
rate of  change of the concentration of  the reactants.   In order
to determine  the  average rate of change,  we  assume the. local
history  of  the joint values of  Ca and  Cg   at  a fixed' location
comprises a stationary  time series and  each  may  be dissected into
.its mean and  fluctuating components

-------
                                                           3-2
and
                                                         (3.6)
and, by definition, C ' = CA = 0 .   Under these assumptions the


chemical kinetic equations for the average rates of change of


the concentrations of  a  and  p  at that location are readily


shown to be
and
                                                          (3.7)
                                                          (3.8)
      In order to solve Equations (3. 7) and (3.8) we need a predic-


tion equation for  C 'C£ .   This is readily derived  (Appendix A)
as
           Q P
                                       Ca
                                                          (3.9)
which introduces four new terms.  C1  . CA  . C'CA  , and C' CA
                                   a    p    a p        ap
The prediction equations for  C'   and  CA   are
TE—
a  = - 2k,  (CAC|
         Ivf3a
and
C|2C')
     x
                                                         (3.10)
                                                         v -^ •   '

-------
                                                            3-3
and they do not introduce any more new terms.  In order to
close Equations (3.7) through (3.11), and thereby achieve a
chemical sub-model for reactions in inhomogeneous mixtures,
we require prediction equations for the third-order correla-
tions  C^2CU  and  C^CA2  .
      Before proceeding to the closure problem, however, it
is instructive to examine more closely the limits of the
effects of concentration  fluctuations on chemical reaction
rates and the conditions  under which these effects become
significant.  This examination may be made in two steps;
1. when are the fluctuations of concentration negligible
(i.e., when are the reaction rates predicted satisfactorily
by the mean values of concentration alone?) and 2. when may
the third-order correlations be neglected?  For the latter
cases, Equations (3.7) through (3.11) comprise the closed
set discussed by Donaldson and Hilst (Appendix A).

The Limits of Errors in Reaction Rate Predictions
if Concentration Fluctuations are Neglected.
      Since the neglect of concentration fluctuations  in
determining reaction rates is equivalent to the assumption
that the local values of  C   and  CQ  are constant in time,
                           a        p
C'1 = CA = 0  and the reaction rates predicted under this
assumption are simply
and
                     / V7T \
                           _  ^ 7T 7T                      (3.13)

-------
                                                            3-4
where the subscript  s  denotes the steady state assumption.
Then the ratio of reaction rates predicted from the inclus-
ion of concentration fluctuations to those predicted neglect
ing these terms are
                                                         (3.14)
and an identical equation for the relative rates of. depletion
of the  (3  species.
      The limits on Equation (3.14) are readily determined
from Equations (3.7) or (3.8) and elementary statistics.
First, we note from Equation (3.7) that for irreversible
reactions,  ScT/dt < 0  and therefore
                           -°L  > -i                      (ci)
                           Ca°P
Further, from elementary statistics we note that

                           ^Cl
                      -1 <  a P .-   .. < + i                (C2)
                           fC'2C'2)^
                           •  a  p '
and therefore
                                                          (C3)
Substituting conditions (Cl) and (C3) into Equation  (3.14),
we establish the limits

-------
                                                           3-5
                                                          (04)
Conditions (C4) set the maximum errors in the prediction of
reaction rates which the neglect of concentration fluctuations
can produce.   These limits are set by the individual variance-
to-mean-squared ratios of  C   and  Cg  and are therefore
f\
C,
functions of the marginal frequency distributions of  C   and
      The limits established by conditions (C4) are shown
graphically in Figure 1 and we note immediately that the
limiting errors in reaction predictions occasioned by neglect
                                             p |2 p ,2
                                                  A
                                             — § -- 73— I  < 0.5
                                              ~   7
but increase to highly significant values as this ratio exceeds
1.0.  The potential for order-of -magnitude errors in the
prediction of the reaction rate exists whenever the product
of the variance-to-mean-squared ratios greatly exceeds 1.0.
      The actual error depends, of course , on  C 'CA/C CA .
                                                Co p  ~ Ct p
This actual error may be examined by forming the ratio
                                                         ,.
                                                         (3
where  r  is the ordinary correlation coefficient and in this
usage expresses the ratio of the actual error in reaction rate
predictions to the maximum possible error for any given Joint

-------
                                                     3-6
                                                  r-1.0,   /
Figure 1.  The limits of errors  of prediction of chemical
          reaction rates incurred by the neglect of
          concentration fluctuations.  (See text for
          explanation of terms.)

-------
                                                         3-7
distribution of  C   and  CM .  Selected values of  r  are
                  a        p
also graphed in Figure 1.  In the limit of  r = 0  no error
in reaction rate predictions is occasioned by the neglect of  .
concentration fluctuations.  This is, of course, the situa-
tion when  C   and  CB  are randomly distributed and  c'c« = 0
However, it is clear from Figure 1 that even modest values of
r  produce significant errors.in the reaction rate prediction
     In   8  \
when  -~ —|— \ > 1 , particularly when  r < 0  .
     V r   r   I
     \Ca  CP  /
      We shall return to this analysis, and identify joint
distributions of  C   and  CM  for which the fluctuations of
                   a        p
concentration must be included later.  For the purpose of
model development, we now turn attention to the importance of
the third-order correlations in the chemical kinetics equations
The Role of the Third-Order Correlation Terms
      Returning1 to Equations (3.9) to  (3.11), it is evident
that the primary role of the third-order correlations is to
be found in their control of the rate  of change of  CT'C^ ,
both directly and through the rates of change of the variances,
C '   and  CM  .  The effects of the third-order correlations
 a         p
on the reaction rates will therefore appear primarily as a
time-integrated effect on  Cf^CM  and any cumulative error in
                    2             2
the estimates of  C ' CM  and  C'CM   will produce a cumulative
          	     a  p        a p
error in  C 'CM .
           a P
      We may deduce immediately from Equations  (3.9) through
(3.11) that if     	
                   C '2CM « CRC |2 + C C 'CM
                    a  p     p a     a a p
and
(C5)

-------
                                                           3-6
their effect on  C 'C£  is negligible and vie may close the
model. equations by setting  C^CA = C&CQ2 = ° •  To illustrate
that conditions (C5) are met under any given circumstance, we
must evaluate the joint distributions of  Ca  and  Cg  from
which these moments are derived since there are now no limit-
ing conditions • on their marginal distributions.  In other
words,  we must turn attention to the distribution functions
from which the means and moments have been derived if we are
to determine the importance of third- or higher-order correla-
tions in chemical reaction rates.  Ideally, we would examine
simultaneous experimental measurements of  C .   and  Cgi  to
make this assessment; unfortunately, very few such data exist.
However, so long as we assume that the basic chemical kinetic
equations are correct (and our whole theory is based on this
assumption) we can proceed by solving these basic equations
for various initial distributions of  C .  and  Cg.  , determin-
ing in the process the time histories of all of the relevant
moments of these distributions.

The Moment Generating Model           ,                      ...
      Under the assumption that only chemical reactions are
operative in changing the concentrations of  a  and  £ , the
chemical kinetic equations can be written as total derivatives
and integrated directly as a function of reaction time.

                        dC ,
                        dC
                             - -k2caicf3i
and

-------
                                                           3-9
K2
pi kn di
i
k2 Ca.l(Q)
ir f f rn
K! cB1ioj
expf- (CB,(0) - T-£ C ,(0))k,tl
^ pi K^ ai i j
while

       r  d- ^ -
       '"fti \ ^ ) ~
                                                         (3.19)

Equations (3.18) and (3.19) specify the j.oint values of  C .
and  Cg.  at -time  t ,  given their initial values and the
reaction rate constants.  They may be used to specify the
frequency distribution of  (C ., Cg. )  at any time  t ., given
their initial frequency distribution,  n.(0)  since, in the
absence of mixing,  n.   is conserved as  C ,  and  Cg.   change
value due to chemical reaction.   Equations (3.18) and (3.19)
provide the information necessary to calculate all of the
relevant moments of  n.(C . , Cg.)  and their rates of change.
We may introduce any arbitrary initial distribution  ^(C^,  Cgj
subject only to the constraints
                        0 < C  . <
                          —  ai —

                                                          (06)
and
                        0 < C  .
                          —  ai

      In order to illustrate the general behavior of the
distribution of  C   and  Cg  and the associated first-,  second-,
and third-order moments, we have chosen a simple distribution
of points along the line  C  = 1 - Cg  and weighted each  point
equally  (HI = 1/N) .  The time history of  n1(C ±, Cg.)  is
shown in Figure 2 and the moments of these distributions  are
plotted in Figure 3.  Note particularly the distortion of the
originally linear distribution of  (C , Cg)  and the associated
decrease (from zero) of the third-order moments.  The relative

-------
                                                         3-10
                                            RUN NO.  1

A<

.2<


0<
h N
> \
x
<<
\ •
N
^
V
\
> W2
^
tco
J 1^
) .2
v^^

yv
.4
x°
\
XX
^^>
^
"*"*""'-C
s±
.6


>>
>>v
^

_^\_
.8
                                                             1.0
Figure 2.  Example of the time history of the joint frequency
           distribution of  (cm> CPi^  given the initial
           distribution shown for  t0  and  k,  = kp = 1.
           Each point was weighted equally  (n^ = i)  for
           calculation of the moments of these distribu-
           tions (shown in Figure 3).

-------
   10'
   10°
   10
    -2
   IO'3
            3-11




       RUN NO. 1


       — &/-\ == -L
        \
         \
            V
                                         — o
                                                /2 -p/2
     C/2 -
       ~
           s

        I
                                    8
10
                        Kt
Figure 3.  .Time history  of  the  first-,  second-,  and third-
           order moments of  ^(CtfL,  C8i)   for t'ne distribu
           tions shown in Figure  2.

-------
                                                            3-12
                 Q             p   __  __
magnitudes of  C '  CJ,  and  CRC '   + C C 'CU  are also plotted in
                a  p        pa     a a p
                                       2             2
Figure 3 and, as can be seen there,  C1 CA  and  C 'CA
completely dominate the initial  change of  C ' C A .   However,
this latter effect is too short-lived to be significant in
the prediction of the time history of mean concentrations.
This fact is shown in Figure 4,  where the predictions of  C"
                                                           OX
and  C"6  as a function of time,  first, neglecting the fluctua-
tions completely,  and then neglecting only the third-order
moments, are compared with the exact solution.  The latter
assumption produces an error of  approximately 10 per cent at
kt = 10 while the  total neglect  of the fluctuations produces
an error of 300 per cent at that time.
      As a further example, and  one which illustrates the
importance of the.third-order correlations, we have constructed
the distribution functions which simulate the case of intermit-
tent sources.  For physical perspective, imagine a free-way,
oriented across the wind and on  which the automobile traffic
ranges from a steady, bumper-to-bumper stream to only an
occasional vehicle.  We assume that each vehicle emits approx-
imately the same amount of pollutants per unit time, but.that
the ratio of the  a  and  p1  species emitted is slightly
variable from one  vehicle to another.  Now we ask,  "What is
the average reaction rate for these exhaust materials
immediately downwind from the roadway as a function of the
intermittency of the traffic?"
      We simulate  this situation by the frequency distribution
for  (C , Cft)  shown in Figure 5.  The variability of  C   and
       ct   p                                       -     ct
CQ  due to variable exhaust -emissions is portrayed as a
circularly symmetric distribution and we take  C  = CR = 0
when there is no traffic upwind  of our observation line.
(Small background  concentrations have been assumed in another
calculation but produce no significant effect.)  We assume
further that the pairs of nonzero values of  (C ,  .Cft)  occur
                                               ct   p

-------
                                                            3-13
      IOV
 QQ.
10
  •»
  e
10
10"
     10"
                  O =  Exact Solution

                  Q =  Solution neglecting 3rd-order moments
                        Solution neglecting concentration
                        fluctuations
                                                 8
                                 Kt
                                                     10
   Figure 4.  Comparison of the predictions of  C   and  UQ

              under assumptions listed, with exact values
              from the moment generating model using the
             .distributions shown in Figure 3.

-------
                                                     3-14
             •7r
             .6
             .5
         t
         03.   ,
         O   -3
             .2
              .1
             OO
                  7O
               03
.1
                                        .5
.6   .7
Figure 5.  Joint distributions of   (Ca, Co)   chosen  to
           simulate chemical reaction rates  immediately
           downwind of a roadway on which  traffic  is
           variably intermittent.

-------
                                                         3-15
with equal frequency and we measure the intermittency of the
traffic as the fraction of time there is a vehicle upwind of
the observation line,  t. .  The moment-generating model has
been used to determine the chemical reaction rates as a
function of  kt  for  t± = 1.0, 0.5, 0.33, 0.2, 0.1  (t± = 1.0
corresponds to a steady, bumper-to-bumper stream of  traffic).
The values of  An =[-2	i- \   r = 	£_E	   and
                        C ' C '
                         a  P -  at kt = I
                    C~eC ' 2 + C" C 'C '
                     pa     a a p
and for .each value of  t..  are shown along with  the observed
ratios of  cKT /dt  to the reaction rates predicted assuming
steady values of  C   and  Cg ,  (BcT/dt)  , and  to the  rates
predicted neglecting the third-order correlations,  (d(T /dt)..
in Table 1.  These results, which are now firmly grounded  in
reality, show clearly that the potential errors  in the  predic
tion of reaction rates neglecting the concentration fluctua-
tions can be realized.  Although the effects of  concentration
fluctuations are negligible for a steady stream  of traffic,
t. = 1 , chemical kinetics based on this assumption under-
estimate the initial depletion rate of  C~   and  (To   by a
factor of 9 when  t. = 0.1.'
      The effect of neglecting the third-order moments  is  not
evident at  kt = 1 , however, since we started with the
correct value of  (T'C^  and the integral effect  of neglecting
  ip         	   Ct P
C ' CA  and  C 'CA   is not yet large at  kt = 1 .  Their
integral effect on the predictions of  C"   and   d(T/dt   can
be seen, however, in the time histories of these quantities

-------
                                                          3-16
when the third-order correlations are neglected.  Values of
B! = C'Q/C^J  and  B2 = (dU^/dt )/(dC"a/^t )I  for each value of
t.  and  kt = 1, 5, 10, 16, and 20 are listed in Table 2.
[(    ) indicates  "C ,-  or  (dcT/dt)...  has wronS sign.]
      An inspection of this table shows that the first effect
of neglecting the third-order correlations, while retaining
the first- and second-order moments, is to produce significant
errors in the reaction rate.  This is, of course, the first
integral effect on  C ' C ' ..  The second effect is on the mean
                     a p
values predicted for the concentrations, an error which
depends upon the time integral of  C'CA .  We also note that
not only can the.errors be large (for example, the reaction
rate is over-predicted by a factor of 100 when  t. = 0.1
and  kt = 20)  but we get the totally erroneous results of
positive values of  <3C" /dt  and negative values of  C"   .'
      Prom these results we conclude that a generally useful
chemical kinetic model must include the representation of
the moments of .the concentration fluctuations through the
third-order.

Summary
      The various considerations and examples of the effects
of concentration fluctuations on chemical reaction rates
discussed above may be summarized as follows:
      1.  The effects of concentration fluctuations can be
significant, to the point of dominating chemical reaction
rates.   The situations under which they are significant are
characterized by joint distributions of the reactant concentra-
tions which are skewed toward large values of these concentra-
tions,  since this is the condition under which the variance-to-
mean-squared ratios can be large with respect to one.

-------
                                                      3-17
      2.  The chemical reaction rate is significantly
accelerated when the concentration fluctuations are
positively correlated and is depressed when these fluctua-
tions are negatively correlated.  Combining .these two
requirements for significant.effects of concentration
fluctuations, we may sketch the general character of
the joint distributions of  C   and  Cg  which require
the inclusion of concentration fluctuations in chemical
kinetic models.  These are shown in Figure 6.  The first,
when  C 'CA > 0  is essentially the distribution of concen-
trations used for the intermittent traffic situation in
the last section.  The second, when  C "C'j < 0 , may be
                                      CM P
quickly identified with the situation when the chemical
reaction is retarded or stopped by local depletion of one
of the reactants.  This latter case depends strongly upon
the relative rates of chemical reaction and of local
diffusive mixing.  Further study of this situation depends
upon an appropriate coupling of the chemical and diffusion
equations.
      3.  For strongly skewed distribution of  C   and  CA
                                               • ct        p
the case when concentration fluctuations become dominant in
the determination of the chemical reaction rate,  the third-
order correlations of these distributions mus_t be included
in the generalized chemical kinetic model.  The neglect of
these terms leads to the highly erroneous result that
either the reactants are produced rather than depleted, or
the mean concentrations go to negative values.

-------
TABLE  1
•f-
ri
1.0
0.50
0.33
0.20
O.10
A
Al
0.19
1.40
2.7
5.0
11.0


-0.41
0.6l
0.67
0.74
0.75
A
Ag.
-0.04
0.02
0.50
1.48
4.00
(d^/dt)'
(^cc/at)s
0.92
1.86
2.80 .
4.70
9.25
( So ydc )
(5c(/at)I
1.0
1.0
1.0
1.0
1.0
  TABLE 2
•1-
ti
1.0
0.50
0.33
0.20
0.10
•kt = l kt = 5 kt = 10 kt = 16 kt = 20
Bl
1.0
1.0
1.0
1.0.
1.0
B2
1.01
1.00
1.00
1.00
1.00
Bl
1.0
1.05
1.21
1.67
2.31
B2
1.33
l.6o
1.17
0.54
0.40
Bl
1.0
0.92
1.24
7.20
(-1.42)
B2
1.25
2.60
1.50
0.30
0.41
Bl
1.0
0.83
1.08
(-1.33)
(-0.30)
B2
1.40
(-6.80)
(-1.50)
0.09
0.03
Bi
1.0
0.77
0.96
( -o . 50 )
(-0.14)
B2
1.60
(-2.35)
(-0.50)
0.02
0.01
                                                      (JO

                                                      M
                                                      OO

-------
   I
   QQ.
  O
                                                       3-19
   t
   QQ.
   O
                                m
Figure 6.  General types of joint frequency distributions of
           Ca  and  Co  for which concentration fluctuations
           are significant in determining chemical reaction
           rates.

-------
                                                     4-1
        4.  CLOSURE OF THE CHEMICAL SUB-MODEL

      Having demonstrated the need for a more general
chemical kinetic model for reactions in inhomogeneous
mixtures, we may proceed along either of two fronts.
The more general of these is to write down prediction
equations for the thirdTorder correlations and seek
closure by suitable assumptions regarding the fourth-
order moments.  If no such assumptions exist at the
fourth order, we may proceed to higher-order moments
until they are found.  This line of inquiry has been
pursued to the level of the fifth-order moments in the
present study but has been set aside, primarily because
the proliferation of simultaneous partial differential
equations for the higher-order moments poses staggering
computer requirements.  This latter feature of higher-
order closure becomes prohibitive when one considers
the requirement for an equivalent level of closure of
the diffusion equations with which this chemical sub-
model is to be coupled.
      With these facts in mind, we have turned attention
to the development of approximate closure schemes in
which maximum possible information regarding the third-
order correlations is sought from the first- and second-
order moments.  As is noted later, such a closure scheme
cannot be exact^.  We begin by reviewing the definitions
Of  *H ' CcfCp  and  CaCe2'
      First, by definition,
            :i°£ = If ni
-------
                                                           .4-2
where  n.  is the frequency of occurrence of the joint values
of  (C ,, Cft, )  and  N = Z n. .  Expanding Equation  (4.1). .and
      cjt j-   p -i_            T  JL
making use of the definitions of  C"^  and  CL , i.e.,

C . =   Z n  C,( j = a,p) , one. obtains
  .
          1  ,±
           a P
                         n. C .Cft.
                          i  ai pi
                                                          (4.2)
We note in particular from Equation (4.2) that, since
                               n
                                i
                                                          (C7)
C'CA/C Cg = -1  only when the lower limit of conditions  (C?)'
is satisfied.  This can occur only when any nonzero values of
                                       g^ , and vice versa.
           C 'C/c c  = -1 , which is the condition for the
C .   are coupled with zero values of
                 "
Then, when
                  c  =
termination of the chemical reaction, all joint moments of
nt(C
     ., Cg.)  about the origin must be zero.
      Now consider the definition of  C' CA

                             ~ Ca)2(°pi  - Cp)
                          n. C2.C,
                           i  ai
C '2
-i a
+ ^a + '
*""
C 'C '
-i Ct P
C" € R
a P
                                                          (4.4)

-------
                                                            4-3
using Equation (4.1) and the definitions of  C^  and  Cg .
It is immediately evident from Equation (4.4) that our
                            2
problem of approximating  C ' C^  by first- and second-order
moments reduces to finding a suitable approximation for the
first term on the right-hand side of Equation (4.4) in terms
of these moments.  For convenience we denote this term by
A. .  i.e . .
 ior      '
                 A,  = — — Z n. C.CR.                 (4.5)
                  ia            X  ai Pl
In seeking this approximation we may. note the following
conditions which  A.   must satisfy
      1.  A= 0               when C/C(   -1
      2.  A1Q= 1               when C/C  = 0
                    C'     C'C,
      3.  A   = 1 + -- + 2--  when C'2CA = 0
           ia       cd     c ^        a  p
                     a      a P

      4.  Ala 2. 0 at all times.                          (C8)

Condition 4 in (C8) operates primarily to constrain the joint
values which  C ' /C    and  C 'CA/C Cft  may assume.  For example,
               ct   ct         ct p  ct p
when conditions 3 and 4 in (C8) are applied jointly, we derive
directly the further condition that

-------
                                                           4-4
                       c •
                        a
                       C2
                        a
              C 'C '
            2  -SL&
              C
> - 1
or
                     a P
                         > - i
                           " 2
                                a
               1  x
                                                          (C9)
                                    2             2
      Finally, we must note  that  C' C'  and  C'C '   cannot be
                                   a  p        a p
specified exactly from first- and second-order moments alone.
Consider, for example, the two distributions of  C  .  and  Cg.
shown in Figure 7.  All of the first- and second-order moments
are  identical for these two  distributions, but the third-order
correlations are of opposite sign.  Clearly, we are looking for
useful approximations, not exact relationships, and any such
approximations must be delimited as to their regions of applica-
bility.  The criteria for usefulness of these approximations
must be based on the accuracy with which the chemical model
within which they are incorporated predicts the chemical
reaction rates.  However, in this development stage we shall
be primarily interested in the accuracy of specification of
the  third-order correlations.
Approximations of  A.   From First-
and Second-Order Moments"
      Utilizing conditions  (C8) and  (C9), it is not difficult
to formulate approximate expressions for  A.   and  A.ft  from
 a
     CV
and C '2 and C'2
     a       p
        la
 For  example, an  early
approximation which has been tested is

-------
                                                         4-5
     1.0
      .8
                           .4
.6
.8
1.0
Figure 7.  An example of two joint distributions of  Ca  and
           CR  for which all first- and second-order moments
           are equal but the third-order moments are not
           equal.

-------
                                                           4-6
             A
              ia ~
                        CaCp
                    a
                                        (4.6)
      Since  C 'C' = 0  when  C '  = 0  and  C 'C '/C CA > - 1 ,
              cxp             ct             a. p  a p —
it is evident from Inspection of Equation (4.6) that this
approximation satisfies conditions 1, 2, and 4 of (C8), but
satisfies condition 3 only if  C^CA - 0  when  C^2CA = 0 .
According to condition (C9), the latter result is admitted
but is not required, i.e.,  C 'CA   may be nonzero when
  p
                             ex p
      The development of an improved approximation of  A.   ,
improved in the sense of more appropriate realization of
condition 3 in (C8), proceeds as follows:  Let
          A
           ia
                       2
1 +

                      a
+ 2
f (C ,C0;C 'C ',C |C') (4.7
 ax  a'  P' a $' a  P  ; v
where the function  f   satisfies the conditions
               a
                 - 0 when
              f  = 1 when C' CA = 0
               a           a  p
                                                          '.CIO)
A simple function which satisfies conditions  (CIO) is
                                 + 1
              f  =
               a
       M + 1
                                        (4.8)
where  M  is defined as the value of  C 'CA/C; CQ  when  C^ CA
We may note immediately that  M  as defined must satisfy
condition (C9), i.e.,   •
                                              0

-------
                      M > -
                               1 +
                                    a
      As a further measure of  M , however, we may also

note that jsince the ordinary correlation coefficient,

                  must lie between i 1 (condition (C3))
          72^
                       (Oil)
             C '  C'
             _o	P_
              o   o
             7T^  7T^

             Ca  °P
                          a
                         C
                          oTP
p i  PI
_«_ _JL
772  ^2
Ca  Cp
and we expect  M  to be related to both
                                           p~ —p
                                         C' /C   and
                                          a '  a
                                                         (C12)
This expectation is reinforced by the fact that, as defined,
                                                      .2
M  must have the same value for both
                                      C'2CA  and  C'C •'
                                       a  P        a P
            p —p                                 p —p
although  C ' /C   is not necessarily equal to  CA /Co
      Some values of  M  versus
                                  —o-
                                  p i  pi'
                                    a  °P
                                  Ca  CP
                                               as determined
from the moment generating model, are shown in Figure 8.

(Each of these solutions was derived from a log-normal or


                                                      Cpi

                                                   CR .)  As
                                           "^0        P
can be seen from Figure 8, within the range tested,   M  is
a composite of log-normal distributions of  C  .  and

which were skewed toward large values of  C    and
a relatively well-behaved function of
     r'  PI
     Gq  ,°p

     —2  —2
      a   P
                                                    For our
present stage of approximation, however, we have chosen a

dichotomous relationship for  M , namely,
                M = 0 when
                             C '  CA
                              a   P
                                  P
                                       < 1
                                                         (C13)
                M = 1 when

                                       >

-------
  M
      2.5
      2.0
       1.5
       1.0
        .5
         10
           -i
Figure 8.  Values of   M   vs
 C'2 C'2
  a   P


 C"2
-a
                                            C/
                               10'
as determined by solution of the basic
•*=•

CD
           chemical  kinetics  equations  for various  initial log-normal frequency
           distributions  of   C    and   C
                               a

-------
                                                            4-9
      Using Equations (4.4),  (4.7),  and (4.8),  our  approximate
predictor equations for
                      and   C 'C/4    are
                           a p
  T^     v  ft
Ca CP = M + 1
and
                       + 2

                                        o
                                      a P
    2    o B
°icfT - M¥T
                         1 +
and'-we specify  M  by conditions (C13) .
                                                           (4.9)
                                                (4.10)
      As a first, but severe test of this approximation,
                                                      2
comparison of the predicted and observed values of  CM CM  for
the most extreme value of

                                CA
                             a
                             shown in Figure 8 (Run
L24) is shown in Figure 9, and the predictions of
from this model are compared with those predicted by 1) neglect-
ing the fluctuations completely and 2) neglecting only the third-
order correlations, in Figure 10.                  .
The Approximate Chemical"Sub-Model
For Inhomogeneous Mixtures,
      Employing .-Equations (4.9) and (4.10) and conditions (CIS),
the chemical .sub-model for two-body reactions in inhomogeneous
mixtures is
                                                           (4.12)
      - It,
> c«c '2
-Pa
                                      Ca
                                                 (4.13)

-------
              4-10
RUN  L24
^^ ^^ ^^

(

700



600


QQ. 500
10
M O >
'^ 4°°
b
csj H
o 30°
200
100
0

-100
C|2C'
_o^ a p
" r} — Exact Solution
v "r ~r
J W O r\
a p-
t r — P
\ C C ' G •'
\i Q y i > B
jk T O ^
c"2 ~ ~,
I U a fi
\ C C,1
\ U I' ,
— - - - - -L
\ r r
\ °a°P
- \ L J
\

\ ^
\\

\V
- \v
\ ^^^
°^^_^ ^""'^--^^ 	 ^

"°--— o 	 0
1 1 1 1 1 1 1

0 10 20 30 40 50 60 7(
Kt
Figure 9. Approximation of the third -order correlation for an
<-> pi
C 
-------
           16
     00
     O

     X
      a
     10

      I
Figure 10.

14


12


10

8


4

2
i
\ \
P \ RUN L24
\ ^
]\ \ ,r-Mark I
\\ \ /

v ^<
\\ \
V /-Mark'M. \
\\ / \
y x
• \
\ v
V
\X ^-Mom. Gen. Model
Uniformly \ ^-o^ \
Mixed — v ^^ ^V
\ \ ^^^
l""°""a °^ 	 ^^~ 	 ~":=s^-
Q! 	 i i i i i i
0 20 40 60 80 100 120
Kt
Comparison of predicted vs actual chemical reaction
for Run L24 under the following assumptions:
1. Neglect of concentration fluctuations.
(uniformly mixed)
2. Neglect of third-order correlation (Mark I)
3. Inclusion of approximate estimates of third-
order correlations (Mark VII)











»
^^


•••=

140

rates







-------
                                                            4-12
               = - 2k,
               C^C'2 +
                P a
(4.14)
C ''
 a
          a P
             M =
                                       - M
                                                           (4.15)
                                                           (4.16)
                                                           (4.17)
             M = 1 when
and the sub-model is closed at the level of the first- and
second order moments.
      The only rigorous test of this approximate model
available to us now is a comparison with the exact solu-
tions of the chemical kinetic equations, as a function
of reaction time  kt , and for various initial distribu-
tions of  (C ., CQ.) .  The much more realistic case of
          v  ai   pi
coupled chemical depletion and dilution by turbulent
mixing must await the coupling of the chemical and
diffusion sub-models.  However, if the local diffusive
mixing is very, very slow compared with chemical deple-
tion, this sub-model must  "track" the. chemical depletion
correctly.  The following tests of the chemical sub-model
                                                           (C14)

-------
                                                         4-13
are therefore restricted to this circumstance so far as any
degree of reality.is concerned.
      Since our primary concern in the coupled chemistry-
diffusion models will be accuracy in the prediction of the
local mean reaction rates, we are particularly concerned
with this facet of the chemistry sub-model.  The comparison
of interest is between the local depletion rates of the
reacting chemical species as measured by  dC~ /dt .    Prom
a variety of initial distributions for  n.(C ., Cg.) , we
have chosen four which exhibit varying degrees of the effects
of inhomogeneous mixing.  Their initial distributions are
shown in Figure 11 and the comparisons of reaction rates as
a function of reaction time are shown in Figures 12 to 15>
(including the initial reaction rates predicted .when the
concentration fluctuations are neglected).
      These comparisons, although by no means exhaustive,
show that the approximate chemical sub-model developed here
captures a very large fraction of the effects of inhomogeneous
mixing on chemical reaction rates.  Over a very wide range of
reaction rates, this model predicts the exact rate to within
a factor of two, while the neglect of the fluctuation terms
in the chemical kinetic equations incurs errors of up to a
factor of 100.  On this basis it seems safe to proceed to
the coupling of this chemical sub-model and the invariant
(second-order closure) diffusion sub-model.

-------
                                                     4-14
O
        Run  No.  2
        n.  =.  1
                  01
 oa.
o
         Run  No.  308
                Oi
 CO.
o
Run No.  380




       1
           X>    Oi
              o
O
                                           Run  "A "
                 O

              O    i
            0     10
                                            v     1000
                                          O         100
Figure 11.   Schematic  of  four distributions used to test

            the  approximate  chemical model.

-------
io-
                                                  4-15





IO-2

<_
'O
a
10



io-3










in-4
1 RUN NO. 2
?
• O Mom. Gen. Model
\j A Mark VII
\ (dCg/dt) at kt = 1 is 2.0 x
\
1 V
IN
- \\
V\
"A
v^
\
\
V
\
\
\
\
\
\
\
\
\
\
\
1 1 1 1 I X 1 1
                                                              10
                                                                -1
             10
                 20
30
40
50
60     70
                               Kt
Figure 12.
        Comparison of reaction rates predicted by the
        chemical kinetic model (Mark VII) with exact
        solution for initial distribution of  (Ca,
        shown in Figure 11.

-------
                                                        4-16
iu •
10-2
I0~3
Ti
a
10
t>
io-4
I0'5
(
& RUN NO. 308
6
\\ O Mom<) Gen. Model
\Q
- JA A Mark VII
\
^j (^c/^t)B at kt . 1 is 6.2 x io"^
\
- X
\X
X
^^
\^xx^
Vv
A
\
\
1 1 1 1 1 1
D 10 20 30 40 50 60
Figure 13,
                     Kt

Comparison of reaction rates Predicted  by the
chemical kinetic model (Mark VII)  ^thr®xacj
solution for initial distribution of  ^Ca, op

shown in Figure 11.

-------
     10
  r2
                                 RUN NO.  380
                                                              4-17
10
     I0~3
     io-4
     1C
      ' 5
                                 O  Mom.  Gen. Model
                                 A  Mark VII
                                \\
                                      \
                                        \
                                                  \
                                                     \
                                                          \
10
        6
                  10
                       20
                                  I
30


Kt
40
50
•^-1
   60
        Figure  14.
               Comparison of reaction rates predicted by the
               chemical  kinetic model (Mark. VII) with exact
               solution  for  initial distribution shown in
               Figure 11.

-------
    io-9
                                        4-18
                                      RUN NO. A

                                    O   Mom. Gen. Model

                                    A   Mark VII.
                                               at kt =. 1 is 1.0 x 10
    I0~6-
  a
10
                                                                    -7
              100
       200    300
400
500    600
700
                                Kt
  Figure 15.
Comparison of reaction  rates  predicted by the
chemical kinetic  model  (Mark  VII) with exact
solution for initial  distribution shown in
Figure 11.

-------
                                                          5-1
        5.   THE  CONSTRUCTION OF A TWO-DIMENSIONAL
             COUPLED  DIFFUSION/CHEMISTRY MODEL
                FOR  A  BINARY REACTION SYSTEM

       The  development of  the  closure scheme for the chemical
 sub-model  described in  the previous section, coupled with
 the  models for  prediction of  turbulence structure, fluxes,
 and  turbulent diffusion of matter  described in Volume  I of
 this report, provides the necessary modules for a coupled
 diffusion/chemistry model, the objective of this program..
 In particular,  these  developments  make possible coupled
 models which permit examination in detail  of the processes
 of generation and decay of the fluctuating components, as
 well as the mean values,  of turbulent diffusion, and inhomo-
 geneous reactions of  a  binary or two-body  reactive system.
       As a starting point for these coupled models we  have
 chosen the relatively simple, but  realistic, situation of
 two  reactive but otherwise passive pollutants emanating
 from either common  or separate cross-wind  line sources.
 This choice reduces the diffusion  calculations to two
 dimensions and  permits  the decoupling of the diffusion/
 chemistry  model from  the  turbulence model, since there is
 no feedback into the  turbulence field due  to either pollutant
 density or exo- or  endothermic reactions.  The decoupled
 turbulence model is used  to generate the field of turbulent
 motions and fluxes, which are then used, along with source
 specifications  and  reaction rate constants, as input  to  the
 coupled diffusion/chemistry model. As can be seen from
 the  full derivation of  the two-dimensional model presented
 in the next section,  this system,  involving as it does
 nine simultaneous partial differential equations, is  already
.rather complicated.  However, one  of the primary reasons for
 starting at this basic  level  of complexity is to permit  close

-------
                                                          5-2
examination of the Interactions between turbulent diffusion and
chemical reactions in the simplest realistic mode in which they
could occur.  When these interactions are understood,.extension
of the models to three-dimensional configurations, nonpassive
pollutants, and three-body reactions can be undertaken with a
much better appreciation of the complex nonlinear system within
which they will operate.
      Due to the limited time and funds for this project, only
a few test calculations of the combined effects of turbulent
diffusion and chemistry have been possible.  Some basic calcula-
tions, which begin to define the effects of turbulence vis a vis
chemical reaction rates, and two sets of calculations for the
NO  - 0-  patterns to be expected in a multiple freeway situa-
  ^    ->
tion are presented and discussed in Section 6.  Despite their
limited number, these examples already point up sharply the
effects of inhomogeneous mixing (produced by the turbulence field)
and the effects of diffusion-limited conditions on fast reactions
characteristic of photochemical chains.

Derivation of the Modeled Equations for the Mixing^of
Two Chemically^ Reacting Materials Emanating
from Cross-Wind Line Sources
      For an atmospheric shear layer in which the Schmidt number
is equal to one and the adiabatic density is constant, we may
follow Donaldson (Vol. I) and write the equation governing the
diffusion and chemistry of a reacting species,  a, ,  as
                                    o
               Af         ^f       r\ C*
                 '"' = - u_. -^ + v. —T& - k,C Co           (5-1)
               'dF"  ' ~  j dx .    o^2      a a
                            J      OXj

where  k   is the reaction rate of the  a  species  with a
second species  p .   A similar equation may also be written
for  CB , but this will not be done until the modeling is
completed.

-------
                                                           5-3
      We may express our variables in terms of their mean
and fluctuating parts as
                        C  = C"  + C'
                         a    a    a
                        u. = u. + u!               .     (5.2)

Substituting these expressions into Equation (5.1) and averag
ing, we obtain the mean local rate of change of concentration
in terms of convection, molecular and turbulent diffusion and
chemical reaction
                    = _.u.     _     (IC~0 + v  — -;
                dt       j d5cj   °^j   J a     ° ^x2
                                                   J
                                                        (5'3)
In deriving Equation (5.3) we have used the continuity equation

                            chl
                                - 0
                             x
                              J
                            du !
Prom Equation (5-3) we see that we now require expressions
for  u'C '  and  C 'C ' .  We first subtract Equation (5.3)
      J Ot        CX p
from Equation (5.1) to obtain an equation for the fluctua-
tion  C ' ,

-------
                                                          5-4
       dC '         dC '       dC       dC '    x               d2C '
         * = - u.    ° - u. :r-2 - u! :r-2 + !— (u^T) + v     2
       dt        j  dxT    j d3cT    j dxT   dx7 v j a'
 Following Donaldson (Vol.  I), we may write the expression for
 u.1   as
       du.1    [    du.1      du.       du^      du^
                            ,    uJ^ruJ^J
                          d2u,'
                ST-  +v/,l^-|£i                      (5.6)
                        ° dxj
 Multiplying Equation (5.6) by  C^  and  (5-5)  by   u^  ,  adding,
 and averaging, we obtain an expression  for   u   ')
           '         wu. v-/    ______   uu rv
           _a = -1J. ^-i-2 - 'uTTTT  . -r^  -  ujc.V  .
                  J dx^.      i j
                 ,                 	     ,  v32^7c7     du '  dCT
                 S	 /  ... ,n . \   K  TTTsrr  ,.  J     i  a   0   i   __a
                 BT:
                   j
                               .
Equation (5.7) introduces  various  third -order correlations,
pressure correlations and  dissipation terms that have been
modeled previously.  We must, however,  obtain an expression
for  C "T '  .  Returning to  Volume I ,  we  may write the equa-
tion governing the temperature  fluctuation  T1  as

-------
                                                           5-5
dT     .  dT '     .  dTr1      d2T'
       dT'     f-  dT'     ,  dT      ,  dT '     .  6T H
       -vr—  =  - <  U .  -r	 +  U   -s	 +  U J  -T	 -  U  -T	 \ +
       dt      \  j  dx,    j  oxT    j  dx.     j  dx. J
                                                         (5.8)
We then  multiply  this  equation  by   C'  ,  Equation (5-5)  by  T'
                                     \JL
add them together and  average to obtain the  equation governing
the rate of  change  of   C 'T
                      -  fc   CpC'T1  + C  C'T'  + T'C'C'     (5.9)
                                                         v->^'/
The triple correlation  u !C 'T '  and the dissipation  term have
already been modeled.  The correlation  T 'C 'CA  will be
                                           a  p.
discussed below.
      Returning to Equation  (5.3)> ™e see  that we  must determine
the governing equation for   C 'C~I  .  If we  multiply Equation (5.5)
by  CA  and add to it the equation obtained by multiplying  C '
by the fluctuation equation  for   CA - a  replaced  by  p   and  (3
by  a  in (5.5) - then the average of that expression gives the
equation for  C 'CM

-------
                                                           5-6
                                             dC ' dCA
                                            2 ^- -a   P
                                             5x, dx,
                                                         (5.10)
We see from Equation (5.10) that we must model  the  third-order
            •      2         2
correlations-.-  C..'CA   and  C ' CA  , but also  that we  require
               ct p         ct  p
                   2         2"
expressions for  € '   and  CA  .  These two equations  are
identical except for the transposition of   ct  and   P  .   The
  2
C1   equation is found by multiplying  (5.5)  by  C'   and
 Ct                        "                        CX
averaging to obtain
                     - 2k  fCpC |2  + C C 'C '   +  C'2C'  >    (5.11)
                         a|Pa     aap      a  PJ    v
    -  Now, since we are dealing with an atmospheric  shear layer,
and our initial attention will be directed  to cross-wind  line
sources of  a  and  P  , we expect that the  only  derivative of
importance is the one normal to the mean flow  ~  in the   x
cartesian direction.  Thus, only  x . = x^ = z  will  be  important
                                   J    j  —
in the equations.  Also, we may set  t = x/u  without loss of
generality. The modeling for the third-order correlations,
pressure correlations, and dissipation terms is  prescribed by
Donaldson in Volume I as

-------
                                                         5-7
                               oj3
             u'C'T' = - A
                             dC 'T'
                               a  .
             n '
             ^  dx,      A,   K." a
            du.1 dC '   u'C'
              i   a    i a
where  q = (u 'u ' + v >v ' + w 'WT)?  and  X, A, , A2 and A_  are
length scales.  Also,  A0  =  (A0  A0 )5  where  a  and   P
                        ^B     da  ^P
correspond to the species in  question.  When Equations  (5.12)
are substituted into Equations (5-3),  (5.7), (5-9),  (5.10),
and (5.11), we obtain our final equation set governing  the
simultaneous diffusion and chemical reaction of   a  and  p
                dC       d C    aC 'w '
              —   a         a     a
                                                         (5.13)
                                                         (5.14)

-------
                                                5-8
u
    g
        =  -  w'w '
                x-                UU '


                {(2A2   + A  )  q^

                L    a     Ja



                             d2C 'w '"
             -

             A
                  c 'w '  +  v
                  -  2v
              l
               a
           - k,
  C 'w '
   a


0  X2
    a
                        a p1 1


                       — )


                           (5.15)
             A^ C6W '  + vo
                               'P
                       C^w '


                         ~2~
      i ip i
 U
   ~5x
-r	  -  O 'W '  -3—
dz      a   oz
                 {.  a

                 ^2c-nf7
            +  v.
                         -  2v
            C 'T1
             a
                                 a
                                               (5.16)
               , < C CAT1  +  CftC'T' + T'C 'C'
               llap       pa        ap
                                               (5.17)

-------
               d2C~^
            v      °L£ -  2v
                               C 'C
   dC
                               A
           •f  V
                 d2C'
                    a
               o  , 2
                 dz
                            C1



                             X
                                a
    ac-*


  '•3^= - 2^
    ^-/.A.           i
                            /       ^
                  -5-c-  + -3- < A0 q  -T-S
                  dz    dz   2^   dz
                               P
                 d";
                            o
                                                       5-9
u
dCAT


~5x
         =-• - w 'T
           + v
                                         -P
                        - 2 v
                               cpT'
                             o
                r_
              •V2
                          C0C'.
                                                  (5.18)
u
    a
                 dCg


                 3F~

dCjC


"d~z
'CA >»
g P  I
               1   a P
           -  k.J'C C*1
               c'i  a a
                      p
                           ''a  + Ca CP
    (5.19)
                                                    (5.20)

-------
                                                          5-10
We have written
k  = k,
 a    1
and
            kQ =
             p
                                          Since we have
decoupled the diffusion/chemistry model from the back-
ground turbulence model, a solution of Equations (5.13) -
(5.21) requires knowledge of the initial distributions of
C"   and
and
         Cfg  and of the flow parameters
                        u
                  T, T ,
                                WW
                                                         q,
    w'T * .   The macro scale lengths  A  and micro lengths
X  must also be known as functions of the background turbu-
lence or the plume characteristics.
    Finally, the coupled diffusion/chemistry model is
closed by modeling the third-order chemistry correlations
as described in Section 4.

                   	
                   CaC

                                             ^L& - Ml   (5.22)
                       - M
                                     (5.23)
where
                   M = 1 when
           C '
            a
                                  "
                  >
M = 0 when
                                  p
                                     < 1
                              (5.24)
A generalization of (5.22) and (5.23) gives the appropriate
modeling for  T'C'CL
    T'C
                            T C
                         - M
                                     (5.25)
With the properly modeled equations, we can now proceed to a
discussion of some of the results of computer solutions of
these equations.

-------
                                                           6-1
       6.  SOME CALCULATIONS OF THE INTERACTIONS
           OP TURBULENT DIFFUSION AND CHEMISTRY

      As must be evident from the derivation and recapitula-
tion of the closed diffusion/chemistry equations (Section 5),
it is virtually impossible to trace the effects of variations
of any one variable through this simulation system.  This
being the case, the validation of the model must.involve
multiple iterations with a systematic, step-by-step variation
of each of the input variables, and comparison of these model
predictions with observed values of concentration patterns,
chemical depletion rates, turbulent flux divergences, and the
like.  Neither time nor resources have permitted this valida-
tion of the model, of course, and the calculations presented
here must be regarded as suggestive rather than authoritative
as to how real chemically reactive and turbulent systems may
operate.  The verification of these predictions must be
deferred, but the results which are presented here argue
strongly that at least in some circumstances the interactions
of turbulent diffusion and chemical reactions are highly
significant and, if verified from observations, models of
this type may also improve predictions of air quality in
the lower atmosphere significantly.
An Illustrative Calculation
      In view of novelty of simultaneous consideration of the
turbulent diffusion of reactive chemical species and their
reactivity in inhomogeneous mixtures, it appears highly
desirable to examine in detail the individual processes by
which turbulent diffusion and chemical reactions produce
observed patterns of reactant concentrations and reactant
depletion in a simple but realistic system.  To this end we
have chosen to calculate the combined processes of diffusion
and chemistry for the case of a plane jet of reactant  a

-------
                                                           6-2
released continuously and isokinetLcally into a uniform environ-
ment of reactant  p .  The environment of  £  is characterized
by a uniform transport speed  TT = io m/sec  and an isotropic
and homogeneous field of turbulence characterized by the verti-
                               2      22
cal intensity of turbulence  w '  = 1 m /sec  .  The plane Jet
of the  a  species is oriented across the mean field of flow
and the initial vertical distribution of the concentration of
a., C" ,  is taken as gaussian with a central value of one and
    ct
standard deviation  a  = 0.4 m  .  In view of the requirements
                     z
that the mass fractions of  a  and  £  equal one, this  "Jet"
of the  a  species displaces the ambient  (3  species at the
source in such a way that the initial distribution of the
concentration of the  (3  species is the complementary gaussian,
Cg0 = 1 - C" 0 .   This geometry of the initial distributions
of  Cf   and  Cg  is shown in Figure 16.  Note that no initial
fluctuations of  C   and  C0  are introduced at the source.
                  a        P
Finally, we take  k, = kp = 1.0 .
      In keeping with the constraints imposed in the construc-
tion of the model, we assume the reaction of  a  with  P
proceeds isothermally.  For our present purposes we shall also
assume that this reaction is irreversible, even though this
assumption is not mandatory?   With these input conditions the
model calculates the redistribution and the chemical depletion
of  a  and  (3  as a function of travel distance or time after
emission.
      As a first partial view of the coupled effects of diffusion
and chemistry .in this flow reactor, we may compare the predicted
axial concentrations of the  a  species as a function of distance
from the source and under the following conditions:
*
 The effects of reversibility of reactions and catalytic
 cycles may be accommodated to a certain extent by approp-
 riate choices of the reaction rate constants,  k]_  and  k2-
 Similarly, three body reactions of the type  ~^-±^oPeP^,
 may also be simulated, if  dCM/dt « 0 , by taking  fc = kj_CM
 where  M  is the third body.

-------
       t
            0
                        1.0
                                             Plume axis
                      Ci/C,
Figure 16.
Source configuration for a plane jet of pure
a  species  injected isokinetically into an
environment of  pure  |3  species.

-------
                                                          6-4
      1.   a  does not react with  £  (diffusion only)
      2.   Diffusion and chemistry occur,  but the
          chemical reaction rates are calculated on
          the basis of the local mean values of
          concentration only.  (We have termed this
          "homogeneous chemistry" since  C "Cl  is
                                          U» }~J
          neglected.)
      3.   Diffusion and inhomogeneous chemistry are
          operative.  (The full model described in
          Section 5.)

This comparison is shown in Figure 17 and it is immediately
evident that the neglect of  C 'CA  leads to a significant
                              C£ p
over prediction of the rate of decrease of the axial concentra-
tion of the  a  species.  'For example, the ratio of the
predicted concentrations at  x = 40 m is two, and, as is clear
from Figure 17, this ratio is increasing with  x .  This  .
result reflects primarily the effect of  C'CA  on the. chemical
reaction rate.  However, such a simple portrayal of the results
of coupled chemistry and diffusion does not portray the balance
of the diffusive and chemical processes at work.  In -order to
gain this insight we must examine in detail the'balance of
turbulent diffusion and chemical reactions going on across the
plume.
      In order to examine this balance we have neglected the
molecular diffusion terms and plotted each of the rates which
determine  dcT/dt  and  dCL/dt  as a function, of distance from
the plume centerline at  x. = 37 m or t = 3,7 sec.  The profiles
of mean concentrations of  a  'and  p  are shown.in Figure 18
and the diffusion and chemical reaction rates"predicted by the
model are shown in Figure 19.  In order to discuss and interpret
these results,, we recall the balance equations for  dCf/dt  and

-------
                                              Diffusion-
                                                no chem.
  X
  o
  E
  o

loe
  X
  o
  E

10°
                                                     6-5
Inhomogeneous
   chem. -+-
     diffusion
                                                 Homogeneous
                                                    chem. +
                                                     diffusion
                            x (m)
  Figure 17.  Comparison of the axial concentrations  of the  a
             species as a function of travel distance from the
             plane jet as estimated for 1) diffusion only,
             2^  inhomogeneous chemistry plus diffusion, and
             3)  homogeneous chemistry and diffusion.  (See
             text for details.)

-------
    fi,                                     t=3.7  sec
    oi
E

-------
                                Height above plume  Cj_  (m)
                                                             2
                                                                  Above this height the chemical reaction is
                                                                  essentially "diffusion" limited, i.e., the total
                                                                  reaction rate is determined by the rate at
                                                                  which the a  species is supplied by turbulent
                                                                  diffusion.
                                                                  Below this height the chemical reaction rate
                                                                  is increasingly "mixedness" limited, i.e., C'aCjj
                                                                  becomes significant.(Compare curves 182)
-6
-4
           Figure 19.
     -2             0             2            4
Rates for  terms indicated  (ppp/sec) (xlO2)
8
         Calculated  values of. local  rates of change  of the concentrations  of
         a  and   P   due  to diffusive flux divergence and  to inhomogeneous
         chemical reactions.

-------
                                                          6-8

and
                 c>C
                         -,
                         Iz
where we have neglected  v  —3—  and  v  —**-  as small in
comparison with  ^— C 'w '  and  -r— C^w ' .  Each of the retained
terms is plotted as a function of height above (or below) the
plume centerline in Figure 19.  (Recall that  d(T/dt  and
        are the local rates of change of the mean concentra-
tion of  a  and  £  due to both turbulent diffusion and
                    ^  _       ^  _
chemical reaction;  -r— C 'w '  and  -r— C^w '  are the local diver-
gences of the turbulent flux of the  a  and  £  species^.
-k(C~ Cfg)  is the average local chemical reaction rate due  to
the local mean concentrations of  a  and  P  ;  -k CfMjJ   is
the average local chemical reaction rate due to correlated
fluctuations of the concentrations of  a  and  p .)
      A patient inspection of Figure 19 reveals the following
facts regarding the diffusion and chemistry  processes through
the plume :
      1.  The diffusion of the  a  species is removing  a
from the plume core from the center line to  z.= 1.25 m  and
is causing an accumulation of  a  from 1.25  to 5 m (Curve  4).
      2.  The diffusion of the  p  species into the plume  is
operating to remove  p  from the height zone  2 to 8 m  and
accumulate  p  in the height zone  0 to 2 m  (Curve 6).

-------
                                                           6-9
      3.  The chemical reaction is depleting both the  a  and
P  species between the plume center line and  z = 6 m  (Curve l),
the upper height being the limit of  a  penetration into the
P  environment at this time.
      4.  The chemical reaction diminishes the rate of increase
of the concentration of  P  below  z = 1.5 m and accelerates
the decrease of concentration of  P  from 1.5 to 5 m (Curve 5).
      5.  Below 2.75 ni the chemical reaction accelerates the
depletion of  a  in the plume core, but above 2.75 m the
reaction rate very nearly balances the diffusive transfer rate
for  a , i.e., above  z = 2.75 m  the chemical reaction is
diffusion limited (Curve 3).
      6.  Below  z = 5 m  the diffusive mixing of  P  with
a  becomes increasingly inhomogeneous so that at the plume
center line the chemical reaction rate is proceeding at only
60 per cent of the rate computed on the basis of mean values
of  P  and  a  concentration at that height.  (Comparison of
Curves 1 and 2)
      Prom these detailed comparisons we can immediately deduce
that the  a plume is not only being rapidly depleted by reaction
with  P  but that it is also growing in vertical width only
slowly because of the balance between diffusion and reaction
rates in the outer limits of the plume.  On the other hand, the
P  deficit in the initial plume is being filled in with only
minor interference from the local chemical reaction with  a .
Further, the depletion of the  a  species in the core of the
plume is significantly slower than would have been expected
from mean-value chemical kinetics.  In this region the chemistry
is "mixedness " limited while above this region it is clearly
"diffusion" limited.

-------
                                                          6-10
      This simple example is intended only to illustrate the
balance of diffusive and chemical processes (and, incidentally,
points to a laboratory experiment which may verify these
predictions).  However, the example does illustrate the model's
power to simulate and portray complex chemistry/diffusion
processes.
The Sensitivity of Chemical Reactions to Turbulent
Diffusion Rates
      The example discussed above is, of course, only one
particular case and cannot provide any real insight into the
sensitivity of the combined diffusion and chemical depletion
of reactive species to variations in the input variables.
Extensive sensitivity analyses have not been possible, but we
have done a partial analysis of the effect of turbulence.
intensity on the chemical depletion rate of the  a  species
for the flow reactor described above.
      Returning to Equation (6.1) we note that since  U  is
constant
                n oo  dc       -\   r> °°          dF
where  F   is the horizontal flux of the  a  species at
        ct
distance  x .  Then
                 /•> 0° SC 'W '         .-> oo
                        - dz -
and since the first term on the right of Equation  (6.4) is zero,
the rate of change of the flux of the  a  species  is determined
by the total reaction rate over the height of the  plume at any
distance  x .  In order to compare the model 's .prediction of
this rate against a limiting condition, we may note that as

-------
                                                           6-11
C 'CA ' goes to zero and  Cfi  tends to be uniformly distributed
through the  a  plume (a condition which can only be approached
asymptotically), the basic chemical reaction goes over to a
first-order reaction and for  U - constant
where  C"D~  is the environmental concentration of the  (3  species.
        pO
We can rewrite Equation (6.. 4) as

                                                          < 6 •
and compare Equations (6.5J and (6.6) from the model calculations
                           2
using various values of  w '
      This comparison is shown for the flow reactor problem and
at  x = 40 m  in Figure 20.  As can be seen from Figure 20, the
relative chemical depletion rate of the  a  species is quite
sensitive to the intensity of turbulence for small values of
w '  but becomes quite insensitive to this input parameter when
  p
w '  becomes large ^  We may also note that even with the vigorous
                 2      22
turbulence of  w '  = 3 m /sec  , the chemical depletion rate
has only achieved about 60 per cent of the limiting, first -order
reaction rate at  x = 40 m or kt = 4 sec,

A Simulation of  NO - 0~,  Reactions and Diffusion Downwind
of a Multiple Freeway System
      The preceding calculations illustrate the basic capabilities
of the A.R.A.P. coupled invariant diffusion/chemistry model and,
of course, provide only a minimal excursion into the possible
interactions of diffusion and chemistry processes in a turbulent
flow reactor.  The stage is set for in-depth analyses of this
kind, but neither time nor resources have permitted the multiple

-------
                                                        6-12
       1.0
       .8
 'P    -6
-La  -4
-Limit of first-order  reaction
                                            x=40m
                                            — — -O


            I
           I
           I
                                    2             3

                               w/2 (m2/sec2)
     Figure  20.  The partial dependence of the chemical depletion *
                of the  a  species on the intensity of turbulence..

-------
                                                          6-13
calculations which are required.  However, as an exercise of
the coupled dimensional model we have constructed an initial
simulation of the diffusion and chemical reaction of  NO
(emitted from vehicles travelling on surface roads or free-
ways) with the ambient  0_ .   In order to examine the effec'ts
of multiple sources of  NO ,  we imagine four parallel freeways,
oriented across the mean wind and separated by a distance of
300 m.  Steady traffic is assumed on each freeway so that
each represents a continuous line source of  NO .  The geometry
of the freeways and the assumed initial source distribution of
NO  at each freeway are shown in Figure 21.  Shown there also
are the mean wind profiles and the potential temperature lapse
rate chosen for this calculation (a neutral lapse rate and a
realistic boundary layer wind profile).  Just upwind of the
first freeway (Si)we assume that  0~  is uniformly distributed
in the vertical at a concentration of 10 ppm.  At the first
freeway we displace  0.,  with  NO  so that the initial street
level value of  0-  immediately downwind of (Si) is zero.
      For the purposes of this calculation we assume that  NO
reacts with  0-  irreversibly so that the only sources of  NO
are the freeways  (NO)  and the reservoir of  0_  above the
NO  plumes.  With this assumption (which, of course, denies
the multiple reactions and catalytic cycles which enter into
the photochemical problems associated with  NO  and  0,)  we
                                                      •3      c;
choose the value of the reaction rate constant as  k = 5 x 10^
(l/ppm-sec).  The calculation proceeds from the first upwind
freeway to a distance of 1200 m (300 m downwind from the
fourth and final freeway.  From these calculations we can
recover the predicted vertical profiles of the mean concentra-
tions of  NO  and  0~  and of the diffusion and chemical rates
at any distance downwind.
      Before looking at some of the details of these calcula-
tions, we may see the general result by examining the mean

-------
   20
3

O
0)


O
.0

O
 *
    10
I10'
1 8
o 6
UJ
§ 4
m
°<
^f
%
SOURCE CONFIGURATION
? OF NO AT S,,S2,S3,S4
X
^\
. . . . \
) 2 4 6 8 R)
?^CONC. OF NO (PPM)
' \\
-9
U
j
;
/
) 600 900 1200 0 _2 4 6 8
83 84 u (m/sec)
                                                                                       I
                           downwind

                    first freeway  (m)
    Figure 21
                The  source geometries and the boundary layer meteorological conditions

                chosen  for simulation of a multiple freeway problem.
                                                                                        i
                                                                                        M
                                                                                        -pr

-------
                                                          6-15

values of the concentration of  NO  and  0^  at 1.625 m above
street level and as a function of distance downwind from the
first freeway.  These values are plotted in Figure 22 and we
pee immediately the gradual accumulation of  NO  and the
depletion of  0   in the surface layer as the air moves
across multiple freeways.
      A closer examination of Figure 22 also reveals the
balance between diffusive mixing and chemical reactions as
reflected in the surface layer concentrations of the  0-, .
Note that the residual  0,,  entering each freeway is drastic-
ally depleted by reaction with the fresh  NO  during the
first  50 m of travel, and then the  0~  concentration
increases due to diffusive mixing from above at a rate
which exceeds the chemical destruction rate.  Recalling the
illustrative calculation discussed in the previous sections,
we see here the suggestion of a fine balance between diffusion
of the reactants into each other and the chemical depletion
rate.  In this case the chemistry is restricted by the supply
rate of  0,,  .to the  NO  plume.
      With this general result in hand, we may turn attention
to the details by examining the predicted vertical profiles
of  NO  and  0_. concentration and of their reaction rates
at selected positions in the array of freeways.  For this
purpose we have elected to plot these profiles at  x = 400
and  1200 m ; these are shown in Figures 23 and 24 (Note the
difference in height scales when comparing these two figures).
      Examination of these figures shows immediately the
surface layer depletion of  0,,  and the inability of diffusive
mixing to maintain a uniform distribution of  0_  with height.
Even at 1200 m the surface layer concentration of  0-  is
less than 1/10th its value outside the  NO  plume.  And we
may note in passing that this diffusion limitation can be

-------
 to
 CVJ
 (0
  •
 II
 Kl


 0
10
                                             in
                                             CVJ
                                             c
                                              M
                                             10
                          300
600
900
1200
                                      x, meters downwind
           Figure 22.  Surface  layer concentrations of NO  and  0-z predicted for
                      multiple freeway simulation.

-------
                                                          6-17
readily checked by vertical profile measurements of the
concentrations of suitably chosen reactants.
      Of considerably greater interest, however, is the
portrayal of the chemical reaction rates with height at
400 and 1200 m.  These are plotted on the right-hand side
of Figures 23 and 24, and for comparison we have also
plotted the rates determined by the local mean values of
concentration alone.  As is more than evident, thes-e .•
reactions are strongly  "mixedness" limited; the total
reaction rate over the plume height is only 19 per cent
of the mean value rates at 400 m and 13 per cent at 1200 m.
This result points to the  "folding" nature of turbulent
mixing', a process in which discreet volumes of each reactant
are folded into one another, but are not intimately mixed.
and therefore react chemically at a much slower rate than
their local average values of concentration suggest.  The
calculation also points to the maintenance and even enhance-
ment of this  "mixedness " limitation when we are dealing with
multiple sources.  Its true role in a complex photochemical
system needs much further study, of course.  But the portrayal
of a diffusion/chemistry interaction which may alter the
chemical reaction rates by a factor of five or more can
hardly be ignored.1
      This demonstration of the potential significance of
the correlations of concentration fluctuations in the
determination of chemical reaction rates also points up the
necessity for second-order closure models if these effects
are to be simulated.  The possibility that important multiple-
source urban air pollution problems exist in which this effect
will be significant commends second-order closure models,
either for the direct simulation of these situations or for
the development of correction factors which can be applied
to first-order closure models.

-------
    10
    8
-  6

 N
01  yi
«   4
      -\
   0
     0
         \
            X
          /

                   \
/
                                           8
It   ,	
J   /(CaC)8-f-C/aC/)9)dz

 l\         p— —          =^'
                                                                                  19
                                                               CaC/9 ,  A

                                                           Units = ppm/sec
                                                                                      o\
                                                                                      i
                                                                                      t->
                                                                                      CO
        Figure 23.  Vertical profiles of the concentration of  NO  and  0
                    and of their chemical reaction rate at

                    from the  first freeway.
                                                                     -,

                                                        x = 400 m downwind

-------
                                8   0
                                                              = 0.13
                                            flC^ 4- C^C^ ,
                                                CaC^f  A
                                           Units = ppm /sec
                                                                    VO
Figure 24.  Same as Figure 23 except at x = 1200 m.

-------
                                                          7-1
           7.  CONCLUSIONS AND RECOMMENDATIONS

      Given the objectives of this program to construct a
coupled invariant diffusion and chemistry model and to
exercise such a model sufficiently to show its applicability
and advantages in real-world situations, we can only conclude
that this effort has been more than successful.  This work
has demonstrated the feasibility of incorporating the
stochastic nature of turbulent diffusion and chemistry in
dynamic models, and has provided the first working version
of such a model.. Most importantly, perhaps, even this first
and relatively simple version has revealed interactive effects
between turbulent diffusion and chemical reactions which
could not possibly be revealed by mean-value or first-order
closure models.  It seems clear to us that, on the one hand,
the atmospheric chemists must reexamine their traditional
assumptions of a well-mixed system in quasi-equilibrium-,
and on the other the atmospheric dynamicists must extend
their considerations beyond the classical calculations of
average values of pollutant concentrations.  The further
exercise and development of these second-order closure models
can provide the tools necessary for the joint evaluation of
turbulent, multi-source flow reactors, such as the atmospheric
boundary layer in urban areas.
      We make this latter recommendation with a full awareness
of the complexities of the chemistry of urban air pollution
and the proliferation of the invariant model equations." as
multiple or chain reactions are introduced.  However, 'this
Increasing complexity need not deter the development and use
of these concepts, since they may first be used to analyze
critical turbulent reactions, then to define areas where
simpler models are quite adequate, and finally to provide
simulation capabilities for those situations where first-
order closure models are demonstrably inadequate.

-------
                                                         7-2
      As a desirable prelude to further development of these
models, such as their extension to coupled three-dimensional
systems, the present model should be subjected to a rigorous
sensitivity analysis wherein the input variables of the
turbulence field, the initial plume geometries, and the
chemical reaction coefficients are systematically varied., and
the outputs of reaction and diffusion rates and the concentra-
tion distributions are tested for sensitivity to these, input
variations.  Second, critical experiments, first in .the
                                          •
laboratory and then in the atmosphere, should be designed
and conducted to verify and validate not only the basix; model
output such as mean values of concentration, but also the
processes internal to the model's workings.  In an atmospheric
experiment, a minimum measurement program would require an
array of towers oriented downwind from cross-wind line sources
and equipped to measure the simultaneous means and fluctuations
of at least two reacting chemical species (coming from well
defined sources) and the turbulent flux ©f these materials,
all as a function of height.  The conduct of such an experi-
ment within the broader measurements program of the EPA/RAPS
appears particularly desirable.

-------
                  APPENDIX  A







EFFECT OF INHOMOGENEOUS MIXING ON ATMOSPHERIC



           PHOTOCHEMICAL REACTIONS

-------
                     Reprinted from
ENVIRONMENTAL
Science & Technology
Vol. 6, September 1972, Pages 812-816
Copyright 1972 by the American Chemical
Society and reprinted by permission of
the copyright owner
Effect of Inhomogeneous Mixing on Atmospheric Photochemical Reactions


Coleman duP. Donaldson and Glenn R. Hilst1
Aeronautical Research Associates of Princeton, Inc., Princeton, N.J. 08540
• The conventional assumption of local uniform mixing of
reactive chemical species is reexamined  by derivation of the
chemical reaction equations to include  the effect of locally
inhomogeneous mixtures on the reaction rates. Preliminary
solutions of a simplified version of these equations show that
inhomogeneities in reactant concentration  generally tend to
slow the reaction rate. Estimates of the relative roles of local
diffusive mixing  and chemical reactions in inhomogeneous
mixtures show that there  are several relatively fast photo-
chemical reactions which may be limited by local diffusive
mixing. In these  cases,  the reaction proceeds much  more
slowly than would be predicted if the reactants were uniformly
mixed.
                             The importance of chemical reactions in  the atmosphere
                           has been increasingly recognized in the problems of air pollu-
                           tion. These are probably most acute in dealing with photo-
                           chemical smog formation (Worley, 1971). We have, therefore,
                           drawn our examples from photochemistry, but we have not
                           attempted to go beyond an  examination of the possible im-
                           portance of inhomogeneous mixing in these processes.

                           Basic Chemistry Model
                             We assume a bimolecular  reaction
+ 0
                              + S
                                                                               (1)
                           where a, /3, y, and 5 denote chemical species and that the re-
                           action of a with /3 to form y and 5 is stoichiometric and is
                           governed by equations of the form
    In developing  either  mathematical simulation models or
     laboratory chambers for the study of chemical reactions
in the atmosphere,  it has been generally assumed  that the
reacting materials are uniformly mixed. However, observations
of the time history of concentrations of trace materials show
quite clearly that uniformly mixed materials are the exception
rather than the rule in both air and water (Nickola et al., 1970
Singer etal., 1963; CsanadyandMurthy, 1971). Local fluctua-
tions of concentration are particularly significant during the
early stages of atmospheric mixing, immediately following dis-
charge of trace materials into the atmosphere, and when there
are multiple point sources of pollutants. Our purpose here is to
make a preliminary estimate of the importance of these fluctua-
tions on atmospheric chemical reaction  rates and determine,
at least approximately, the relative roles of reaction rates and
diffusive mixing in the control of atmospheric chemical re-
actions.
   1 To whom correspondence should be addressed.

 812  Environmental Science & Technology
                                                                               (2)
                                                      ri  vm
                                             —  = -K[a]{p]
                           [/] denotes the molar concentration  of the rth  chemical
                           species and K is the reaction rate constant in units of (sec
                           mol/cm3)"1. It is convenient  to transform the concentration
                           terms in Equations 2 to dimensionless mass fractions, C,, by
                                                    =  Mt[i]
                                                    (3)
                           where p0 is the density of the mixture (g/cm3) and  Mt is the
                           molecular  weight  of the ith chemical species.  Then the de-
                           pletion rates for the a and /3 species may be written

                           and
                                                                                       Ma
                                                                               (4)
                                                                                                                 (5)

-------
where Ka = KpJM^ and has dimensions (sec-ppm)"1 when C0
and C0 are expressed in parts per million (ppm) by wt.
  We may now examine the relative contributions of the means
and fluctuations of Ca and Q to the chemical reaction rate by
assuming the time history of these quantities at a fixed point
constitutes a stationary time series and that
                      =  C
                               cfl'
                                                      (6)
where the overbar indicates a time average and the prime
indicates the instantaneous fluctuation about  the  average.
Noting that Ca' = Ce' = Oand that
                  ac,    &c,
                  —  = — H
                          dr     dr
we obtain directly from Equations 4-6
and
                                                      (7)
                                                      (8)
                                                      (9)
where we have suppressed the dependence of Ka on the tem-
perature and pressure. (This analysis can be extended  to in-
clude the fluctuations of Ka owing to significant fluctuations of
temperature and pressure. For our present purposes, we shall
assume an isothermal reaction at ambient pressure.)
  The role of concentration fluctuations in chemical reactions
is immediately evident from Equations 8 or 9. The second-
order correlation in the joint fluctuations of Ca and Cp  either
enhances the reaction rate (when the correlation is positive) or
suppresses the reaction when Ca 'C$' is negative. Only when
these fluctuations either do not  exist or are uncorrelated  is the
average  reaction rate governed by the average concentrations.
As a simple example of the importance of this correlation
term, imagine that the materials a and /3  pass the point of
observation  at different times—i.e., they are never in contact
with each other. Values of Ca and C0 would be observed, but
it is readily seen that CaCp =  — Ca'Cp' in this case, a  result
which correctly predicts no chemical reaction.
  If we  assume no diffusive mixing of the reacting materials,
we may  model the chemical reactions by noting that
                                                     (10)
                                                              and

  "\y^ /   "\ f*
  oCt    oC
    - -  =     —
   d/     dr


V  - r ' —"-'
    — t-a

                                                                                                                   (12)
                                                              Performing the necessary operations and time-averaging, we
                                                              get, repeating Equations 8 and 9,
                                                                                                                   (13)


                                                                                                   a  ,
                                                                                              cac»
     CCa'C
                                                                                                           'Q")
                                                                                                                   (14)
                                                                                                                   (15)
                                                                                                                   (16)
                                                                                                                   (17)
                                                              Equations 13-17 provide a closed set, except for the third-order
                                                              correlation terms €„'€/* and C3'Ca'J.
                                                                The appearance of the third-order correlations complicates
                                                              the modeling problem very considerably since the statistical
                                                              description now requires consideration of the  distribution
                                                              functions for Ca and Cft. In  an independent study, O'Brien
                                                              (1971) has proceeded from Equations 13-17 by assuming the
                                                              form of these distribution functions. Another approach, which
                                                              we are pursuing, is to model the  third-order correlations in
                                                              terms of the second-order correlations. However, for our
                                                              present purpose of determining whether or not the effects of
                                                              inhomogeneous  mixtures on chemical  reactions  may  be
                                                              significant, we  may  neglect  the third-order  correlations  by
                                                              assuming C0' and C^' are symmetrically distributed about Ca
                                                              and Cf,, respectively. This assumption is, of course, untenable
                                                              for more general cases but it does permit a solution of Equa-
                                                              tions 13-17 by numerical techniques.
                                                                For an initial test of the significance of inhomogeneities in
                                                              chemical reactions, we assume a  reaction box in which the
                                                              initial concentration distributions of a and ft are arbitrarily
                                                              specified by ^, Q,, C^, C/*, and €„'€/,'. As a further con-
                                                              straint which isolates the chemical reaction process, we assume
                                                              there is no mixing in the reaction vessel and no wall effects.
                                                                As a  reference case, let us assume  a completely uniform
 10
 ^.
 10
                                      UNIFORMLY
                                      MIXED CASE
                                         10°

                                         Kl
                                                                                  Figure 1.  Chemical depletion of randomly
                                                                                  mixed reactants (Ca'Cp'  = 0) for various
                                                                                  initial degrees of inhomogeneity, as measured
                                                                                  by C-VC.'
                                                                                    Volume 6, Number 9, September 1972  813

-------
Figure 2.  Chemical  depletion  of initially
inhomogeneously  mixed  reactants (C'2/C02
= 0.40) for various degrees of initial correla-
tion between Ca' and C0', as measured by
                   Cg'Cf,'
                                                                                                                0.0
                                                                                                                  -1.0
                                                                                 10°

                                                                                 Kl
mixture of a and /?—i.e^  no Jluctuations in concentration,
Ma ^ Mp, and initially Ca = Cp = C0. The predicted values
of Ca are shown in Figure 1 as a function of time normalized
by the reaction rate constant.  As can be seen,  the reaction
proceeds to exhaustion of the reacting materials.
  Now let us assume that a and |3 are initially inhomogene-
ously mixed but that there is no initial correlation between
Ca' and cy—i.e., initially Ca'Cy ss 0. As a measure of these
fluctuations, we take GyVO* = 0.2, 0.4, 0.6, 0.8, and 1.0.
The results of these calculations are also shown  in Figure 1,
and  it  is  immediately  evident  that  any  inhomogeneities
operate to suppress the chemical reaction rate and to stop it
completely before  the  reacting  materials are exhausted.
Mathematically, the model predicts that, in the absence of mix-
ing,  initial inhomogeneities  operate  to produce  values  of
Ca'Cp' which eventually become equal to —CaCp and  the re-
action ceases. Physically, the local reactions have everywhere
proceeded to exhaustion of one of the reactants, leaving a
residue of the other reactant and products at that  site.
  It is of special interest to note, from Equation 15, that the
suppression of  the reaction rate by C~'Cy depends only on
one of the  reactants being nonuniformly distributed initially.
A negative rate of change of Ca'Cy can be generated by non-
zero values of either Ca'2 or Cy*, since the terms C>Ca'2 and
CQCy2 are positive definite.  The  presence of concentration
inhomopeneities in one of the reactants generates inhomogene-
ities in the other.
  The effect of an initial correlation between Ca' and Cy may
now be examined by assigning initial nonzero values to Ca'Cy,
CV^jind  Cy2. To illustrate  this effect, we have chosen
c7~2/c02 =  cyvc,2  =   o.4  and c7cy/(c72c7')1/2  =
+ 1.0, +0.5, 0.0, -0.5,  and -1.0 where Ca'Cy/(Ca'2Cy2)"2
= Ra#, the  ordinary correlation coefficient.  The  resulting
predictions of CjC0 are shown in Figure 2 and are again com-
pared with the uniformly mixed case. As might have been ex-
pected,  initial  positive  correlation accelerated  the reaction
rate, but only when  this initial  positive correlation was per-
fect did the reaction go to exhaustion of the reacting materials.
In this case, although there were concentration  fluctuations,
stoichiometrically equal amounts  of a  and /3 were initially
placed in each  local volume. In all other cases,  the reaction
was  again  halted when one of the reactants was exhausted
locally, leaving a residue of the other reactants and products
of the reaction.
  The combined effects of initial inhomogeneities  and correla-
tions between the fluctuations are summarized in Figure 3 by
plotting the depletion of C during the first normalized time
step as a function of Raiff and C7"2/^2. The effect  of the mag-
Figure 3. Joint effect of initial correlation and inhomogeneity on de-
pletion of reacting materials at Kt = 1.0
nitude of the fluctuations, as measured by C'2, reverses as one
goes from large positive toward small positive and negative
values of Ca'Cf'.
  These results point toward an important role for fluctua-
tions of concentration in controlling chemical reaction rates.
For example,  if two reacting  materials are  discharged  si-
multaneously from a point source, during their initial mixing
with the atmosphere their concentration fluctuations should
be large and positively correlated. We would then expect,  on
the basis of this effect, that their reaction rate would be con-
siderably faster than if they were uniformly mixed  from the
start.  The emission of hydrocarbon and NO* from auto ex-
hausts is a case in point. Discharge of SO2 and particulate
matter from power plant stacks is another.
  On the other hand, if reacting materials are  randomly
mixed or if positive fluctuations in one are associated with
negative  fluctuations in  the other, the reaction should  be
suppressed,  compared to the uniformly mixed case.  Both of
these cases could be important, but their true importance de-
pends critically on the rate at which atmospheric  diffusion
tends to mix chemical species and, hence,  to diminish these
fluctuations, as compared with  the rate of chemical reaction
produced by the concentration fluctuations.

Estimates of Local Mixing Rates in the Atmosphere
  The only way in which the correlation Ca'Cy can be elim-
inated, if it  exists, in given  flow situations is by the process
of molecular diffusion. To estimate the rate at which this can
occur, we may write the expressions for the contribution of
814  Environmental Science & Technology

-------
molecular diffusion to the time rate of change of Cn' and
They are
=  Da
  n  ;X.r_«
                \ d/ JAM       dyjdyi
                                                    (18)
                                                    (19)
Multiplying Equations 18 and 19 by Cp' and Ca', respectively,
adding, and time-averaging gives
     /
     V
-2 D
                                (20)
(We have assumed Da =^ £>0 =  Z), consistent with the assump-
tion Ma ~ M0. See O'Brien (1971) for a discussion of this
assumption.) The first term on the right-hand side of Equation
20 is nondissipative—i.e., it measures the transfer of the Ca'Cp'
correlation by gradients in the value of this correlation within
the  field. The second term is dissipative—i.e., it measures the
local diminution of Ca'Cp' by the action of molecular diffu-
sion. The appropriate expression for this term is
          2 D Cg'CV
              X2
                                                    (21)
In this expression, the dissipative scale length X must be chosen
as it is chosen for the calculation of other turbulent correla-
tions when performing calculations of the structure of turbu-
lence (Donaldson, 1969).
  For such calculations, 1/X2 is given approximately by
                            °-05 p°q
                      X2
where p0 is the atmospheric density, q1 = V2 + e'2 + tv'2, n0
is the molecular viscosity of air, and A is a length scale related
to the integral scale of the atmosphere, and is of the order of
1000 cm in the earth's boundary layer.
  If we choose typical values of the parameters involved in
evaluating the magnitude of the expression for (dCa'Cy/d/)dift
given in Equation 20, we have

                A =  1000cm
                p0 =  10-' g/cm8
                Ho =  1.7 X 10-
14. C3H6 + HO2 =
      CH3O + CH3CHO           3.4 X  lO"2     10-'
15. C2H3O+ M =
      CH3 + CO + M            1.7X10-'  2 X 10~2
                                         and the reaction rate will be suppressed. In this case, the re-
                                         action is controlled by the rate of species mixing and will de-
                                         pend on parameters other than Ka and C0Q.
                                           We may estimate which two-body reactions will proceed as
                                         though C'acy ~ 0—i.e., in the usual manner, and which will
                                         be modified by having values of |C<,'Cy| of the same order as
                                         \CaCe\ by forming the ratio
           /dCa'Cy\
      N =  \   df  Jdift =   2D   =  3.4X1

           \   d?   /Ch
                                                                                              (22)
                                         When N » 1.0, Ca'C/,' will tend to zero and the reaction will
                                         be controlled by the reaction rate constant and the mean con-
                                         centrations; when N  and the reaction will proceed at a rate deter-
                                         mined largely by the rate at which one reactant can be mixed
                                         with another and will depend on the scale of the patches of
                                         unmixed reactants.
                                           Typical values of Ka for various reactions which enter  into
                                         the photochemical chains are listed in Table I along with esti-
                                         mates of N. For these two-body reactions, only the first and
                                         second are sufficiently slow for conventional kinetic models
                                         to apply. The propylene reactions with O3,02, and HO2 (num-
                                         bers 12, 13, and 14 in Table I) tend to represent a transition
                                         stage between diffusive mixing control and chemical reaction
                                         control of the reaction rate. The remaining reactions are all
                                         clearly diffusion-limited  in inhomogeneous  mixtures  and
                                         should proceed at a rate which is much slower than conven-
                                         tional chemical kinetics would predict.

                                         Conclusion
                                           These results indicate there are clearly reactions in the ph>; to-
                                         chemical chain which will be suppressed by the inability of the
                                         atmosphere to mix the reacting materials rapidly enough to
                                         prevent serious local depletion of one of the reacting materials.
                                         In these cases, conventional models of the reaction will tend to
                                                                                   Volume 6, Number 9, September 1972  815

-------
seriously overestimate the reaetion rate, and, therefore, the
production rate of the chemical species which enters into the
next reaction in the photochemical chain. On the other hand,
the enhancement  of  the reaction rate for two  materials
emanating from a common source and, therefore, occupying
the same volume of the atmosphere, during the initial period
of incomplete mixing,  may  also represent a significant de-
parture from conventional simulation models.
  It is hoped that this brief and necessarily incomplete dis-
cussion will serve to demonstrate the importance of turbulent
fluctuations  of concentrations in atmospheric chemical re-
actions. Consideration of these effects in refining simulation
models of these reactions appears to be important.

Nomenclature
Ci   =   mass fraction of /'th chemical species, ppm
Di   =   molecular diffusion  coefficient   for  /th  chemical
  species, cm2/sec
K  =   chemical reaction rate constant, cm3/sec-mol
K«, K/t   —   chemical reaction rate constants, 1/ppm-sec
MI  =  molecular weight of /th chemical  species, g/mol
N  =  nondimensional ratio of characteristic  times
q2   =  ur*  + vi* + W7*, cm2/sec2
^a,/s   =  ordinary second-order correlation coefficient
/   =  time, sec
u',v',w'   =  orthogonal  components of turbulent fluid
   motion, cm/sec
yt   =  length along /th direction of a cartesian coordinate
   system, cm
[ ],      -   averaged quantity
'   -   departure from the average of the primed quantity

GRI-UK LIUTKRS

a, #, 7, 8   =  chemical species
A  =   dissipation scale length, cm
A   =   macroscale of atmospheric turbulence, cm
Hi,  =  dynamic viscosity for air, g/cm-sec
p,,  =  fluid density, g/cm3
Literature Cited
Csanady,  G. T.,  Murthy, C.  R., J. Pliys. Oceanogr.,  1,  I,
   17-24 (1971).
Donaldson, C. duP., J. AIAA, 7, 2, 271-8 (1969).
Nickola, P. W., Ramsdell, J.  V., Jr., Ludwick, J. D.,  "De-
   tailed Time Histories  of Concentrations  Resulting  from
   Puff and Short-Period Releases of an  Inert Radioactive
   Gas: A  Volume of Atmospheric  Diffusion Data," BNWL-
   1272 uc-53  (available from Clearinghouse,  NUS,  U.S.
   Dept. of Commerce),  1970.
O'Brien, E. E., Phys. Fluids, 14, 7, 1326-9. (1971).
Singer, I.  A., Kazuhiko,  I., del Campo, R. G., J.  Air Pollut.
   Contr. Ass., 13, 1, 40-2 (1963).
Worley, F. L., Jr., "Report on Mathematical  Modeling of
   Photochemical  Smog," Proceedings of the  Second Meeting
   of  the  Expert Panel  on  Modeling, No. 5,  NATO/CCMS
   Pilot Project on Air Pollution, Paris, July 26-7, 1971.
Received for review  November 26, 1971. Accepted May  II,
1972.

-------
                  APPENDIX  B  '







CHEMICAL REACTIONS IN INHOMOGENEOUS MJXTUKFiS:




 THE EFFECT OF THE SCALE OF TURBULENT MIXING

-------
                                 15


 CHEMICAL REACTIONS IN  1NHOMOGENEOUS  MIXTURES:   THE EFFECT OF THE
                   SCALE  OF  TURBULENT MIXING

          Coleman duP.  Donaldson*  and Glenn R.  Hilst**
                             ABSTRACT

     Recent studies by O'Brien  [1]  and  the  authors  of this paper [2]
have provided a theoretical  framework  for the  assessment of
chemical reaction races when the  readmit!.;  are embedded in a
turbulent fluid and are inhomof.eneour..! y  mixed.  The results of
ther.e studies, which are  reviewed here,  point  towards a profound
effect on chemical production and depletion rate:-, when the charac-
teristic reaction time, as measured hy  the  product  of the chemical
kinetic reaction rate constants and the  average and fluctuating
concentrations of the reactions,  is short compared  with the
characteristic molecular  diffusion  time.   The  latter time is
measured by the ratio of  the molecular  diffusion coefficient and
the square of the dissipation scale length, and is, therefore,
dependent upon the scale  of  the turbulent motions.   Both the fact
of inhomogeneous mixtures and this  dependence  upon  turbulent
scales of motion pose significant problems  when extending labora-
tory results to other scales of motion,  such as the free atmosphere.

     These theoretical results, which are partially substantiated
by observations, point towards the  need  for simultaneous measure-
ments of turbulence and chemical  reaction rates over a range of
turbulence scales and reaction rate constants.  If  substantiated
by such new experimental  measurements,  tiie  theoretical results
point towards a clear requirement for- joint consideration of the
chemical reactions and the scale  of turbulence in such diverse but
critical problems as the  design of  large  combustion apparatus and
the calculation of photochemical  reactions  in  the atmosphere.


                          INTRODUCTION

     Although the effects of inhomogeneous  mixing of reacting
chemical species on the reaction  rate,  as measured  by either the
depletion of reacting species or  the production of  new species, have
been recognized for at least ten  years  [3], methods for accounting
for this effect have only recently  emerged  [1,2].  Neither of
these methods are as yet  fully developed, but  they  are sufficiently
advanced that we may make some preliminary  estimates of the
situations under which the effects  of inhomogeneous mixing will be
pronounced or perhaps even completely dominate the  reaction.

     In particular, we find  for .the case  of an irreversible two-
body reaction at constant temperature that  the following limita-
tions are imposed by inhomogeneous  mixing of either or both of the
reacting species:
* President, Aeronautical  Research  Associates of Princeton, Inc.
  50 Washington Road,  Princeton,  New Jersey 085*10     (A.R.A.P.)

**Vice President  for Environmental  Research,  A.R.A.P.
                 Reprinted from PROCEEDINGS OF THE 1972 HEAT

                 TRANSFER AND FLUID MECHANICS INSTITUTE,

                 Raymond  B. Landis and Gary J  Hoffmann. fHitorl.

                 Stanford University Press, 1972.  © 1972 by th»

                 Board of Trustees of the Leland Stanford Junior

                 University.

-------
254   Donaldson and Hilst: Chemical Reactions and Turbulent Mixing


   • If the chemical reaction rate Is slow compared with the
     molecular diffusion rate, no effect is noticed, and the
     reaction proceeds according to conventional chemical kinetics.

   • If the chemical reaction rate is fast compared with the
     molecular diffusion rate, the reaction rate is limited by
     the diffusive mixing rate, tending, in the limit of very
     slow diffusive mixing, to zero before the reactants are
     exhausted.

     A large number of reactions in combustion processes and photo-
chemical smog formation fall within this latter category.  It is,
therefore, of considerable interest to investigate further just
how much the reaction rate is curtailed by inhomogeneous mixing
under such circumstances.  In the following pages, we derive the
basic equations for prediction of the joint effects of chemical
reactions and molecular diffusion, examine the effects of the
dissipation scale length of the turbulent motions, and identify,
on a preliminary basis, the two-body reactions inherent in photo-
chemical smog formation for which inhomogeneous mixing is a
limiting condition.


             MODELING OF CHEMICALLY REACTING FLOWS

     For most computations of chemically reacting turbulent flows,
it has been customary for engineers to proceed with the calculation
according to the following scheme.  First, the engineer develops by
some method (mixing length, eddy diffusivity, or other method)
equations for the tirne-averaged or mean values of the concentrations
of the reacting species of interest (say, species  a and P ) at
each point in the turbulent flow under consideration.  He also
obtains an equation for the mean value of the temperature that is
expected at each point in this flow.  It is then customary, if the
equations that generally govern the reaction between  a  and  P , are


                            BF* = -klCaCp                       , p p                    .    ( ?)
                            Dt     K2 a P                    .

to assume that valid equations for the time rates of change of the
mean values of the mass fractions of  a  and  P are

                            DC\
                                  ~ 1  n  A                       \ 3 J
                            	
                            Dt
                            n+-     ^o^nr-'a
                            IJt      £ Ct p


In these equations,  C   and  CR  are the time-averaged mass

fractions of the two species and  k-  and  k0  are the reaction

rates  k,  and  k?  evaluated at the mean temperature T , i.e.,
                                                                CO

-------
             Donaldson and Hilst: Chemical Reactions and Turbulent Mixing   255
             and  k2 = k2^  '

     Although equations such as (3) and  (^) are used extensively at
the present time, it is not difficult to show that they are
incorrect when reaction rates are fast and the scale of the turbu-
lence is large.  This may be done by considering the proper forms
of Eqs. (1) and (2) when they are averaged.  The well-known results
are*
and
     DC
     DT
     DC
To demonstrate the character of these equations, let us discuss
them under the assumption that  k'= k' = 0  .  Equations  (5) and
(6) then reduce to                                     .
                       DC
                         a _  57 (r P"
and
                       Dt
                       DC

                       DF
                                         o:Cp)
                                                                 (8)
It is clear from these equations that, if one wishes to calculate
the reaction of a  with  P , it will bep necessary to have an
equation for the second-order correlation  CCi  unless one can
           C'Ci «
show that

conditions required for
for the particular flow in question.  The

             can be derived in the follow-
                         C 'C '
                          er p
                              « (J CQ
                                  ct p
ing way.  First, by following the method used by Reynolds for the
derivation of the equation for the turbulent stress tensor, one
finds the following equation for the substantive derivative of the
correlation  C'Cl :**
              a p
                          - u°
                     -chem
                                                                 (9)
*  For a discussion of these equations that is related to the
   present treatment, reference should be made to O'Brien [1] which
   was published after this work on the modeling of chemically
   reacting turbulent flows was started.
** For the purposes of this illustrative discussion, the flow is
   treated as incompressible.
T  The  notation  is  that of general tensor analyses.
   contravariant form of the metric tensor  gmn .
                                                      mn
                                                          is the

-------
256   Donaldson and Hilst: Chemical Reactions and Turbulent Mixing
where the term (DC 'CA/Dt )cnem
                               is the contribution of chemical

kinetics alone to the substantive derivative of
                                                 C'C
                                                  a
                                                         This
expression can be found from Eqs . (1) and (2), and is
      DCaCP
          K
      Dt
                           2
                          '
                                            2
                                            '
It is instructive to discuss the behavior of the correlation
for the case of turbulent reactions in the absence of any
appreciably -large gradients.  In this case, Eq. (9) becomes
                                chem
                                                               (11)
The second term on the right of Eq .  (11)  is the destruction of the
correlation  C^CX  by the action of molecular diffusion.  In line
with our previous work [5]5 we will model this term by means of a
diffusion scale length   X   so that Eq. (11) becomes
                          Dt
                                chem
                                          C'C'
                                           a P
                                                               (12)
The diffusion term in this equation is such that  C'C'
approach zero with a characteristic time that is
                                                       tends to
                           diff
What is the overall effect of the first term on the right-hand side
of Eq. (12)?  The effect is difficult to see from an inspection of
Eq .  (11) , but we may derive an expression for what this term _
accomplishes from Eqs. (7) and (8).  First, multiply (7) by  Co and
(8)  by  C/a  and then add the resulting equations.  The result is
                      chem
This equation can be interpreted by saying that the effect of chem
istry alone is to drive  CaCft  to the negative of  C^CL   (or  C/CA
to the negative of  cCo) with a characteristic time
     Equation (14) s_tates that the reaction between a and  p  will
always stop, i.e., C.,C  + C'C'  will become zero, short of the
                    \Jf p    Up
exhaustion of  a or  P unless  a and P are perfectly mixed wherever
they occur in the turbulent flow under consideration.  The physical

-------
             Donaldson and Hilst: Chemical Reactions and Turbulent Mixing  257


reason for this is that, in the absence of diffusion, if  a and p
are not perfectly mixed to start with, the final state of the gas
in any volume element will be  a  and products, P  and products,
a  alone, or  P alone, but never any region containing both  a and
P .  It is easy to see that, no matter what the values taken on by
C   and Cg  are as a function of time, if   Ca  is never nonzero
when  Co  is nonzero and vice versa_ so that no reaction is possible,
it is mathematically true that  C CQ  +   C'C' = 0 .  . Thus, Eqs. (7)
                                 dp       Cl p
and (8) state that no reactions are possible as required by the
physics of the problem.
     An actual example may make the meaning of  C^CJ  more clear.
Consider that the flow of material by a given point is such that
alternate blobs of  a and  p pass the point.  Let us suppose that
half the time the flow is all  a and half the time it is all  p  .
The resulting concentrations are sketched in Figure 1.  If this
pattern keeps repeating, the average values of  CQ  and  C_  are

obviously  C~a= 1/2  and  C~6= 1/2 .  Whenever the flow is all  a ,
C  =+1/2  and  C  =-1/2.  Whenever the flow is all  P , C  =-1/2
and  C' = +1/2 .   We find then that the average value of  C'C'
      p  _                 _   __                   a p
must be  C'C' =-1/4 .   Since  C'C' = C C0  , no reaction is possible
          a p                  a p    a p  '                ^
according to Eqs. (7)  and (8). and obviously no reaction should
occur.
                   THE EFFECT OF SCALE LENGTH

     We may now return to Eq. (12).  If, in this equation, the
scale  X is small enough and the reaction rates are slow enough,
the second term on the right-hand side of the equation will be
dominant and the flow will be such that  C'C^  is always almost
zero. This means that molecular diffusion is always fast enough
to keep the two species well mixed.  On the other hand, if the'
reaction rates are very fast and X  is very large, the first term
on the right-hand side of Eq. (12) will be dominant and  C'C'  will

tend to be approximately equal to -C CR and the two species will be

poorly mixed.  The rate of removal from the flow of a  and p by
reaction will then not be governed by reaction rates but will be
limited by molecular diffusion.  To put these, notions into quanti-
tative form, let us consider the ratio of the two characteristic
times

               N = ;


and a contact index
                   C C_ + C'C'
               I =  a P _  a P
                      C CQ
                       a P
We note that if  N  is much smaller than one, diffusion will be
very rapid and the two species  a and  P will be in intimate

-------
258  Donaldson and Hilst: Chemical Reactions and Turbulent Mixing
contact with each other.  In this case  C'C'/C CL. will be small
                                         u. p  up
and the contact index will approach one.  If,  on the other hand,
N  is much larger than one, mixing will be poor and  C'C'   will
          	 	                                         ct p
approach -C Cg .   The contact index will then  approach zero.
In this case The reaction will be diffusion-limited.

     In many laboratory flows, the dissipative or diffusive scale
of turbulence is very small and  N  is, indeed, small so that the
neglect of  C'CA  in the kinetic equations is  permissible.  On the
other hand, if the laboratory experiment is just increased in size,
holding all other parameters such as velocity, temperature, etc.,
constant, one soon finds that the character of the flow changes.
This may be seen by examining the expression for the dimensionless
quantity  N  in more detail.

     Let us assume the diffusive scale of a turbulent flow is of
the order of the dissipative scale so that we  may relate  X to the
integral scale length of the turbulence  A,  by (cf. Ref. [5])

                                     bpqA1/u)                   (18)

where  a  and  b  are constants and  pqA,/|i  is the turbulent
Reynolds number.   Substituting this expression into Eq.  (16) gives


                          AiN% * kA)                      U9)
For relatively high Reynolds numbers, this expression becomes
If an experiment is performed in the laboratory and a value of  N
for this experiment is determined or estimated and is found to be
small compared to one, then we know that the diffusive mixing of
the flow is such that the species a and (3  are in contact.   The
reaction rate of these species is then chemically controlled.  Now
if the apparatus is just scaled up in size, all other things being
equal,  N  will increase linearly with size since the scale  A
increases linearly with the size of the apparatus.  When the scale
has been increased sufficiently, so that  N  is no longer very
small compared to one, the nature of the flow in the device must
change, for the species  a and  p will no  longer be in intimate
contact, at equivalent positions in the apparatus.

     The turbulent atmospheric boundary layer is- a good example of
a flow in which it is essential to keep track of the correlation
Cg-Ci  if one is to be able to make sense of the reaction of species
which are introduced into the flow.  To demonstrate this, we list
in the table some of the second-order reactions responsible for the
production of photochemical smog.  We have also listed in this
table the reaction rate recommended for each reaction [4]  and an
estimate of the number  N  for each reaction if it occurs in the .
atmospheric boundary layer where a typical value for  X  is  10
centimeters.  It is interesting to note that it is, in general,

-------
             Donaldson and Hilst: Chemical Reactions and Turbulent Mixing   259


those reactions listed in the table for which N  is greater than
one that investigators have found to proceed more slowly than pre-
dicted by formulas such as Eqs. (3) and (4) when the reaction rate
determined from laboratory experiments is used.  This difficulty
has led some investigators to search for other chemical reactions
that might be considered which would explain this discrepancy.

                         •CONCLUSION

     It certainly appears unwise to follow this course until such
time as one has at least developed a viable scheme for properly
computing turbulent reacting flows.  It is the authors' opinion
that an acceptable method of computing such flows can be developed
through the use of second-order correlation equations such as
Eqs. (9) and (10).  Methods of modeling the third-order correlations
that appear in these equations can be found that are similar to
those used to study the generation of turbulence and turbulent
transport [5].  The development of a viable method for computing
chemically reacting turbulent flows according to such a scheme is
under active development by the authors.  It is important to note
in this connection that it is essential in developing this general
method to consider fluctuations in density and in the reaction rate
constants when the chemical rate equations are considered.


                          NOMENCLATURE

a, b      = constants

C., CJ    = concentration of subscript species, expressed as a
            mass fraction
    &    = molecular diffusion coefficient
    I     = contact index (Eq. (17))

k, , k«    = chemical reaction rate constants
    N     = dimensionless ratio of characteristic times for mole-
            cular diffusion and chemical reaction

    q     = rms value of turbulent kinetic energy

    A,    = integral scale length of turbulence
    X     = dissipative scale length

    M.     = viscosity
    p     = density of the fluid
    T ;    = characteristic time


      ':                     REFERENCES

1.  O'Brien, Edward E.   Turbulent Mixing of Two Rapidly Reacting
    Chemical Species, Physics of Fluids, 1971, 11(7), 1326-1331
2.  Donaldson, Coleman duP. and Hilst, Glenn R.  The Effect of
    Inhomogeneous Mixing on Atmospheric Photochemical Reactions.
    Submitted to Environmental Science and Technology, 1972.
3.  Toor, H.L.  Mass Transfer in Dilute Turbulent and Nonturbulent
    Systems with Rapid Irreversible Reactions and Equal D.lf fusivlty.
    J.Amer.Inst. Chem.  Eng., 1962, 8, 70-78.

-------
260   Donaldson and Hilst: Chemical Reactions and Turbulent Mixing


*).  Worley, Frank W. "Report on Mathematical Modeling of  Photo-
    chemical  Smog," paper presented at  Panel on Modeling,  NATA/CCMS
    Pilot Project on Air Pollution, Paris, July 1971.

5.  Donaldson, Coleman duP. and Rosenbaum, Harold.   "Calculation  of
    Turbulent Shear Plows Through Closure of the Reynolds  Equations
    by Invariant Modeling," presented at NASA Symposium on
    Compressible Turbulent Boundary Layers, Hampton,  Virginia,  •
    December  1968 and published in NASA SP-216, pp.  231-253.
Some Second-Order Reactions Responsible for Photochemical Smog


       Reaction                k (ppm-sec)~             N

 0  +  NO = N02 + 02 *            8.3 x 10"^            0.25 *

 N02 + 03 = NO, + 02             1.7 x 10~5         5.0 x 10~3

 N03 + NO = 2N02                    4.8             1.1 x 103

 NO +  H02 = N02 + OH             1.7 x 10"1           50.0

 OH +  03 = H02 + 02                 1.7           .  5.0 x 102

 OH +  CO = H + C02               5.0 X 10~2         1.5 y 102

 CH 02 + NO = CH30 + N02            1.7             5-0 x 102

 C2H302 + NO = C2H30 + N02          1.7             5-0 x 102

 C2H402 + NO = CH3CHO + N02         1.7             5.0 x 102 • .

 CH30 + 02 = HCHO + H02             1.7             5.0 x 102

 C0H,-  + 0 = CH, + C.H..O          6.0 x lO'1         1.8 x 102
  jo          J    ^ J

 C3Hg  + 03 = HCHO + C2H1|02       8.3 x 10~3            2.5

 C..H,  + 0_ = CH00 + C0H00        1.7 x 10~2            5.0
  3o     2     3     ^3

 C3Hg + H02 = CH 0 + CH3CHO      3.4 x 10~2           10.0

 C2H30 + M = CH3 -I- CO + M        1.7 X 10'1           . 50.0
*Those reactions for which  N  is  small compared  to one  are  those
 which can be treated using mean quantities in  the basic equations
 of chemical change, i.e., correlations in fluctuating quantities
 may be neglected.

-------
          DonaJdson and Hilst: Chemical Reactions and Turbulent Mixing   261
ca\ ii - •• g/3u '
r'-4. '
Ca"1" 2
c' — i.
P 2
/ s+l] / =+l
«i-|j «j9-i
^'^
«i-i
- - I


Figure  1.   Simple problem  illustrating

            reactions are possible
a P
= -C CD when no
    a P

-------