N A S A CONTRACTOR .

"
i
I-
NA
REPORT

LOAN COPY: RETURN TO

AFWL (WLIL-2)
KIRTLAND AFB, N MEX

UNIFIEDANALYSISOF
T U R B U L E N T J E T MIXING

by Joseph A. Schetx

Prepared by
JOHNS HOPKINS UNIVERSITY
Silver Spring, Md.
and
UNIVERSITY OF MARYLAND
College Park, Md.

for

NATIONAL
AERONAUTICS
AND
SPACE
ADMINISTRATION WASHINGTON, D. C. JULY 1969

I-.
 NASA CR-1382
TECH LIBRARY KAFB, NM

*
UNIFIED ANALYSIS OF TURBULENT J E T MIXING

By Joseph A . Schetz

Distribution of this report is provided in the interest of

informationexchange.Responsibilityfor the contents
resides in the author or organization that prepared it.

Funded by NASA, Contract No. R-76 and prepared under Naval Ordnance
Systems Command Contract No. NOW-62-0604-c by Johns Hopkins University,
Silver Spring, Md. and University of Maryland, College Park, Md.
I

for

NATIONAL AERONAUTICS AND SPACE ADMINISTRATION

I".."
. ." . ,., ~ _ _ _ _.. _ .

For s a l e by theClearinghouseforFedercrlScientificandTechnica!Infc:rnotior
-
S p r i n g f i e l d ,V i r g i n i a 2 2 1 5 1 C F S T l price $3.00
 SYMBOLS

a Initial jet radius or half-height

Widths of the mixing regions

Drag coefficient

d Diameter

g = P + -P a = g/ge(nondimensional)
N

Total pressure; g
2
Constants

Characteristic length

P Pressure

S 3
Transformed axial coordinate; = S/$" (nondimensional)
a
T Temperature
N

U Axial velocity; u = u/u (nondimensional)

e
V Normal velocity

XYY Axial and normal coordinates

Mass fraction of species

"i"

Eddy viscosity, defined by Equ.

(1)

Displacement thicknesses

PT = PC Turbulent viscosity

D Dens i ty

Shear
-
t Stream function; $ = $/$a (nondimensional)

Subscripts

a Conditions at y = a, x = 0

C Conditions along jet centerline

e Free stream conditions

J Initial jet conditions

2
 INTRODUCTION

The general flow field of interest is simply an inner stream

of fluid injected parallel atomoving outer stream, and turbulent mixing
between these streamsis the subject of the present investigation. This
is, of course, one of the oldest viscous flow problems that remains
essentially unsolved, and the continuing interest in this general problem
stems from thewide variety of practical devices whose intelligent design
depends upon an understanding of such processes. Jet ejectors and combus-
tion chambers are, perhaps, two of the most common examples.

In an area where s o much work has been reported, it is fortunate,

indeed, that several authors have endeavored to review previous studies.
References (1) through (7) are the most usefulof these as background
information for what follows. Other work that has appeared more recently
or that which is of particular interest here is also listed in the Refer-
ences. In Table I, the principal experimental studies are cited along
with a very cryptic description of their main characteristics. Wake ex-
periments also bear upon the present problem a general
in way, but, as we
shall see later, a direct quantitative applicationof such results must
await further developments in our basic understanding of turbulent shear
flows.

The history of the analytical treatments of this problem is

also very lengthy;here, however,it is instructive to consider wake
problems concurrently. The classical treatment (Refs. 2-6) of free mixing
problems is based ona view of the flow field where there are two main
regions o f the flow: 1) the transitional or developing region and2) the
similarity regions, where suitably scaled profiles are self-preserving
with axial distance. In the jet case, an initial region containing the
potential core precedes the transitional region. Mixing in this initial
region is normally taken as that for the well known half-jet problem
(two unbounded parallel streams of different velocity initially separated
by an infinitesimally thin plate). When the inner boundary of this
mixing region intersects the axis of the jet, the initial region is deemed
ended, and the next stage in the calculation must begin. At this point,
however, a very crude assumption was commonly made: the transitional
region was neglected and the similarity region was 'assumed to begin im-
mediately. For wake problems, of course, one does not normally treat an
initial region (the near wake) directly but the same assumption of
neglecting the transitional region and taking similar profiles immediately,
now at the "initial" station, has been commonly made.

This general procedure is open to criticism on two counts.

First,
the flow in the initial region is not well represented
by the idealized
half-jet problem since real initial profiles are far from ideal. The
boundary layers on the inside and outside of the splitter plate and the
thickness of the splitter plate itself serve to produce large distortions
of the initial profiles which are known to produce large effects on the
development of the near field. This manifests itself in the generally
poor prediction of the length of the initial region by the classical
theories (Ref. 4 ) . Second, it is well known that the similarity region

3
does not begin near the end of the initial region. Townsend (Ref. 3)
indicates, for example, that the wake behind a circular cylinder does
not become "similar" until an axial distance greater than about100
diameters. These two criticisms are not merely academic; taken together
they result in rather poor agreement between the classical theory and
experiment. On the.basis of these observations, it seems clear that the
transitional or non-similar region of the flow field must be treated
directly. Thus, we can consider either analytic approximations that do
not employ a similarity assumption (Refs. 28, 29) or strictly numerical
procedures (Ref. 30). Both of these choices will be discussed in detail
in the present report.

To this point, our considerations of the analytical treatment

of turbulent free mixing problems has dealt primarily with the question
of solving the equations of motion. While being important, this is not
the area of greatest concern. The major difficulty associated with these
problems is the specification of some representation of turbulent trans-
port processes. Our understanding of turbulence at this time is not such
that one can treat the "turbulent" processes within
a shear flow directly.
Rather, the most that can be expected is some prediction of the "mean" flow
properties. To this end, we relate the shear to the mean flow variables
as following Boussinesq
7 = (pe)u
Y
where c is an "eddy viscosity". This is certainlya gross simplification
of turbulence and can be attacked on physical grounds (Ref. 3), however,
at the present time, it represents the only fruitful avenue of approach
to turbulent shear flows. The eddy viscosity models that have been pro-
posed for the main mixing region of free mixing flows are summarized in
Table 11. The initial region in the jet problem requires special con-
sideration, and we shall discuss this point shortly. The earlier "mixing
length" theories given as the first four cases in the table have been
generally discredited (Ref. 3 4 ) , and most workers in the field now favor
some application of the concept of Prandtl's third model given as the
fifth item in Table 11.

We digress hereto discuss the flow in the initial region. A gen-

eral treatment of this part of the flow must be able to handle arbitrary
boundary layers on the inside and the outside of the splitter plate where
either or both may be laminar or turbulent. Further, the wake trailing
from the finite thickness splitter plate and its interaction with both
splitter plate boundary layers must be treated. This isa formidable
fluid flow problem in itself and will require further extensive study
before its solutioncan be confidently incorporated into a treatment of
the transitional and far flow fields. There are, however, many technolog-
ical applications of jet mixing (e.g. fuel injection) primarily concerned

(x- 102a) compared to the initial region(x-

with the development of the flow field over distances that are long
loa). Under these cir-
cumstances it is to be expected that the details of the initial region

4
will have a small effect on the region of interest, and some approximation
there can be tolerated. If an approximation is to be made, a convenient
one might as well be selected. In the greatest part of what follows,
therefore, the eddy viscosity in the initial region has been based on
the gross profile in the same manner as for the main region of the flow.
This modelis certainly not physically correct, but it will be shown later
that it does, in fact, produce predictions at least as accurate
as the
classical half-jet treatment. In a few cases, the experimental data to
be compared with theory do not extend far enough downstream of the initial
region to make this approximation adequate. In those cases,it is necessary
to start the analysis with an experimental "initial" profile just downstream
of the initial region. This is exactly the course of action followedin the
wake case where it is necessary to start with a profile at theof endthe
"near wake". The agreement between experiment and theoretical prediction
in the main region of the jet is always improved by starting the calculation
with an experimental "initial" profile at the end of the potential
core
region. This procedure hasnot been employed in general here as it is
felt to be artificial. A useful theory mustbe able to provide reasonable
predictions in the regionof interest starting with the true initial con-
ditions or some simplification thereof.

Returning now to Table11, it is important to note that the three

entries following Prandtl's third model
can be shown to be equivalent to
it and each other. Consider first the Clauser model+

E = k, u ''6 (2)
e
where k l b,-
.018 and the Prandtl model using
(x) as the width,b5,
2

c = .037 b
,(x) Iumax - u
2
I
min
For simple profile shapes such as a rectangular or triangular defect or
excess, one finds that the two expressions agree exactly in form and to
the extent of .036 versus .037 as the proportionality constant. Later,
specific numerical results will be presented for jet solutions with both
models and comparison with experiment made. This will only strengthen
the assertion made here thatEqu. (2) and ( 3 ) are equivalent. The "wake"
models of Schlicting and Hinze can be reduced toth'e same form as the
Clauser model. Nothing that

C d = 2 0 (4)
D
and taking a representative value of =C 1.20 for a circular cylinder
D
in high Reynold's number flow, these expressions become respectively

E = .044 ue 0 (5)
and

+ Here 6* must be interpreted to be based on the absolute value of

(1-
5
 I n t h et r e a t m e n t of wake problems, i t i s common t o n e g l e c t t h e f a c t o r
( u / u ) i nt h ed e f i n i t i o n of 8 s i n c e( u / u e ) 2 1. Thisdoes,however,
e
r e n d e r 8 = A* (A7k =
m
r-- 11"
U
I dy) SO t h a t ,t h e s ef o r m u l a sc a n as
J -03 ue
w e l l be written

E = .044 ue A*

Q = .027 ue A*

The C l a u s e r model w r i t t e n i n t h e s e terms i s

E = .018 ue A*

where we h a v e t a k e n t h e d i s p l a c e m e n t t h i c k n e s s a p p r o p r i a t e t o a "two-sided",
p l a n a rf r e em i x i n gp r o b l e mr a t h e rt h a nt h e" o n e - s i d e d "b o u n d a r yl a y e rc a s e
c o n s i d e r e d by C l a u s e r . A s s t a t e da b o v e , i t w i l l b e shown t h a t Equ. ( 2 a )
p r o v i d e sp r e d i c t i o n s i n good agreement wiAh j e t e x p e r i m e n t s . The q u e s t i o n
a r i s e s as t o why t h ec o n s t a n tf o rw a k e s( u
j
< 1) i s s o much l a r g e rt h a n
t h a tf o rj e t s( u j1 ) . Abramovich (Ref. 4 , pp. 211) n o t e st h i se f f e c t
and a t t r i b u t e s i t t o i n c r e a s e d t u r b u l e n c e c a u s e d by t h e s e p a r a t e d b a s e
flow i n t h e wake caseA The w r i t e r i s n o ta w a r e o fa n yc o n s t a n td e n s i t y
j e t e x p e r i m e n t sw i t hu j < 1 t h a tm i g h th e l pt oc l a r i f yt h em a t t e r .F o r
o u rp u r p o s e sh e r e ,h o w e v e r ,t h ei m p o r t a n tr e s u l t i s t h a tt h e s ef r e e mix-
i n ge d d yv i s c o s i t ym o d e l sa r e a l l e q u i v a l e n ti nf u n c t i o n a lf o r m .

The s i x models l i s t e d i n T a b l e I1 f o l l o w i n g t h e wakemodels are

a t t e m p t st oe x t e n dt h eb a s i cP r a n d t l model t o p r o b l e m si n v o l v i n gs i g n i f -
i c a n td e n s i t yv a r i a t i o n s . The Ting-Libby Model r e s u l t s fromanattempt
t oa p p l yt r a n s f o r m a t i o nt h e o r yt ot u r b u l e n tf r e em i x i n g ; i t hasbeen shown
t ob eu n r e l i a b l e( R e f .1 8 ) .F e r r i l ss u g g e s t i o no fu t i l i z i n g a mass flow
d i f f e r e n c et or e p l a c et h ev e l o c i t yd i f f e r e n c ei nt h eP r a n d t l model has
p r o v i d e dp r e d i c t i o n so fu n r e l i a b l ea c c u r a c yf o rt h ea x i - s y m m e t r i cc a s e
(Refs. 2 2 , 2 3 ) . However, when t h e m a s sf l o wd i f f e r e n c e was a p p l i e dt o
t h ep l a n a rc a s e , a good p r e d i c t i o n( R e f . 29) was a c h i e v e df o rt h e one
e x p e r i m e n tt h a te x i s t si nt h el i t e r a t u r e( R e f . 21). The A l p i n i e r i model
i s c o n t r a r y i n form t o a n y o t h e r modeland m u s t beviewed as e s s e n t i a l l y
e m p i r i c a l ,q u a l i t a t i v e l ya s w e l l a sq u a n t i t a t i v e l y . The s i m p l e Zakkay
axi-symmetric model g i v e n i n T a b l e I1 i s based on t h ep r e s u m p t i o nt h a t
t h ea s y m p t o t i cd e c a yo ft h ec e n t e r l i n ev e l o c i t ya n dj e tf l u i dc o n c e n t r a -
t i o nb e h a v ea st h ei n v e r s e of t h es q u a r eo ft h es t r e a m w i s ed i s t a n c e ,
i . e .x - nw h e r e n = 2 , i nt h ea x i - s y m m e t r i cc a s e . I t i s t r u et h a tt h e s e
q u a n t i t i e s dobehaveas "x" t o some n e g a t i v ee x p o n e n t ,n ,b u tt h ev a l u e
of I 1 n I t i s a t l e a s t a f u n c t i o n o f t h e j e t . t o f r e e stream mass f l u x r a t i o ,
pjuj/peue as c a nb es e e ni nF i g . No. 1. Zakkay a s s e r t st h a ta ne x t e n d e d
formof h i s model c a nb eu s e df o rc a s e sw i t h n # 2.0. ExaminationofRef. 23,
h o w e v e r ,r e v e a l st h a t i t i s n e c e s s a r yt o know t h ed e c a ye x p o n e n tf o rb o t ht h e
c e n t e r l i n e c o n c e n t r a t i o n and t h e v e l o c i t y a p r i o r i b e f o r e t h e e x t e n d e d model
can b e s p e c i f i e d . This , i n e s s e n c e , r e q u i r e s k n o w l e d g e of t h e b e h a v i o r of t h e
solution before the solution can be obtained. Thus, thereno is
unified
picture of the treatment of flows with significant density variations on
the basis of existing eddy viscosity models, particularly for the axi-
symmetric configuration.

Since itwas possible to demonstrate some unity of the models for

planar, constant density cases, it is instructive to examine new means for
A s these models are all
extending these models to the compressible case.
equivalent, we may start with any one. Rather than the usual procedure of
starting with the Prandtl model, start here with the Clauser It model.
is
simple, at least formally, to extend this to varying density, i.e.

co
G = .018
ueA* = ,018
ue dY

and to show that for simple profiles, sucha as

triangular velocity defect,
that this expression is equivalent to the planar mass flow difference model
given in Table11. Recalling that this model produced predictions in good
agreement with experiment, it may be stated that the Clauser model can be
viewed as an adequate representation of planar, free mixing flows with or
3 6 the state-
without strong density variations. With the results of Ref.
ment can be broadened to include the effects of strong
streamise pressure
gradients.

In view of allthis, one may ask what the difficulty is with

the axi-symmetric case. It is the purpose of this paper to propose an
answer to that question. First, some new resultsfor the planar case will
be presented to substantiate some of the assertions made above. Then, the
axi-symmetric case will be treated in detail. All of the work is limited
to two-dimensional (either planar or axi-symmetric), steady, fully turbulent
flows without chemical reaction. Pressure gradient and three-dimensional
effects are left toa later report.

ANALY s IS

The turbulent laws are taken here to be identical to the correspond-

ing laminar lawswhen expressed in terms of mean fLow quantities and tur-
bulent transport coefficients. Further, the boundary layer form of the
conservation laws is assumed to apply.

Planar Flows. In this section, some new results for planar flows
will be developed and discussed. The primary aim of this
work is to substan-
tiate some of the assertions made in the Introduction; for this purpose
con-
stant pressure cases only will be considered.

It has been stated that it is necessary to treat the non-similar

development of the flow field in the transitional region. The choice aof
method for solving the flow equations thus reduces to either
a strictly
numerical procedure oran analytic approximation. Here, an analytic approx-
imation of the linearized type (Ref.37) will be employed. This type of

7
approach has been shown to provide very good approximations for free
mixing problems (Ref. 38). Specifically, the physical normal coordinate,
y, is replaced bya stream function,S(x,y),

Sy = U(X,Y) ; -$x = V(X,Y) (8)

and the total pressure,g(x,$) E P(x) + pu2(x,$)/2, is introduced as the

dependent variable. With this, the Mass Continuity and Momentum Equ's. be-
come

gx = su(xy$)gglli
This is approximated as

gx = y(x)g (10)
$$
where uc(x) is the velocity along the centerline of the jet.
Introducing
a new streanwise coordinate
rx
S s(x')uC(x')dx'
JO

and using the boundary conditions

g(0,llr) = g j = P + pu?/2 ;
J
O < JI < $a

-
-
= P + puz/2 ; $ > $a
e
'

lim$ "t 03 g(S,JI) = g e

the solution becomes

where g = g/ge, 3 = S I $ : , $ = $/$,. The solution foru(S,JI) can be found

by simple algebraic manipulation. It remains then to invert theS(x) and
$(y,x) transformations to obtain the solution in the physical plane.

It is at this point, inverting theS(x) transformation, that a

model for the eddy viscosity must be specified. It has been asserted that
the Clauser model representsa unified and adequate statement;it was used
here. Some results of the analysis are compared with the experimental data
presented in Ref. (14) in Fig. No. 2 on the basis of the centerline velocity
decay for u./ue= 2.0. Also shown is the Classical Theory (Ref. 4 ) predic-
J
tion for each case. The first point of importance to note is the poor pre-
diction of the initial region length by the Classical Theory even though
some attempt at including the effects of initial non-uniformity
was made in

8
this case. It is true that the description of this region by the heuristic
approximation on the eddy viscosity proposed herein is not truly adequate
either, but it is certainly at least as accurate as that of Classical Theory.
Secondly, the present theory does seem to predict the behavior of the flow
in the non-similar transitional region more accurately. That is, the slope
of the velocity decayin the region25 5 x/a 5 75 is more accurately pre-
dicted by the present solution. Lastly, the present and the Classical
results become the same far downstream in the similarity region.

This has been accomplished using the generalized Clauser model for
the eddy viscosity. To the author's knowledge this is the first time that
this model has ever been applied to anything but a boundary layerItflow.
has been previously asserted here that this modelis equivalent to.the Prandtl
model commonly employed for free mixing problems, and it is proposed now
to justify this assertion. This can be expeditiously accomplished by
taking the solution developed here and calculating the eddy viscosity
based on the Prandtl model from these profiles. The results of asuch cal-
culation are given in Fig.No. 3 . Certainly, the two models are in very
close quantitative and qualitative agreement. Also shown in the figure is
the eddy viscosity in the initial region as predicted by the half-jet or
free jet boundary. Taking this in conjunction with Fig. No. 2 one can con-
clude that the true eddy viscosity in the initial region lies somewhere
between this prediction and the results of the heuristicapproximation for
this region employed here.

Axi-Symmetric Flows. Starting with the now substantiated conclusion

that the Prandtl model, the Clauser model and the wake models are all equiv-
alent in the planar case, one may choose to seek
a corresponding axi-symmetric
model using any of these three as the starting point. Schlicting (Ref.5)
simply used the planar form of the Prandtl in model
the constant density
case and Ferri (Ref.18) then employed his mass flow difference concept to
this. It will be shown here that neither step is adequate. Rather, an
axi-symmetric equivalent of the planar models will
be sought starting with
the generalized form of Clauser's planarmodel. Rewrite this as

where pT(x) is the turbulent viscosity. This can be read to

say,
"The turbulent viscosity is proportional
to the mass flow defect (or
excess) per unit width in the mixing region". One can carry this state-
ment over into axi-symmetric flow by defining
a new displacement thick-
ness, &+cy, as

This gives
pT(x) k2 ( p u rrS:+)/L
=
e e r
The proportionality constant, k2, will have to be determined aby
comparison between theory and experiment for one case as is done with
all eddy viscosity models.It will of course be a function of the
choice of the characteristic length, L; here we take L = a, the initial
jet radius. With this, the model becomes

9
 The question of solving the equations of motion remains. The
writer was fortunate to be able to apply somework of a previous
colleague, S. L. Zeiberg, in this regard. The finite-difference pro-
gram for treating hypersonic wake problems developed at the General
Applied Science Laboratories (Ref.30) was used directly, only minor
alterations were made to include the new eddy viscosity model. The
cooperation of the Special Projects Office of U.the S. Navy in agree-
ing to make the program available is gratefully acknowledgea.

In order to make the one-.time determination of the constant

in Equ.(l7) and to begin the comparison of theory and experiment, the
constant density experiments of Forstall and Shapiro (Ref. 1) were
employed. The particular case chosen was for u./u= 2.0 and the results
J e
are shown in Fig. No.4 . The value of b n determined is 0.018. One can
observe that the unified model gives excellent qualitative as well as
quantitative agreement with the data; the asymptotic decay predicted by
the Prandtl model isat variance with the data. Note that again, the
Classical theory providesa poor prediction. Returning to the question
of the asymptotic decayrate, it is interesting to put the data and the
numerical results for the two eddy viscosity models on log paper since
the asymptotic behavior should bea straight line (i.e. x-n). This has
been done in Fig. No. 5 where it can be clearly seen.that the unified
model predicts a decay rate in much better agreement with experiment
than the Prandtl Model. The axial variation of the eddy viscosity as
predicted by the two models is given in Fig. No.6.

At this point, comparisons between theory and experimentcan be

extended to covera wider range of parameters. The effects of density
variations are now included, limited, for the moment,to those due to
temperature variations. Ferri's extension of Prandtl's model is em-
ployed along with the unified model, and the experimental data of Landis
and Shapiro (Ref. 12) is used as the standard against which the adequacy
of the theories are measured. A constant Prandtl number of 0.75 in
accordance with previous (Refs.1, 12) suggestions is used throughout.
Results for heated jets (T./T = 1.19) with u./ue = 2.0 are given in
J e J
Fig. No. 7 and those for (T./T = 1.31) with u./u = 4.0 are given in
J e J e
Figs. No. 8 and 9. In these cases also, the solution including the
temperature field based on the unified model is generally in better
agreement with the data than that based on Prandtl's model or Ferri's
extension thereof. This conclusion is reached on the basis of the
slope of the centerline velocity and temperature decays. The limited
downstream extent of the data does allow some variety of opinion on
this point especially for the case in Fig.No. 9.

10
 The last variable that must be considered in the general
treatment of this flow problem is jet and free stream composition.
Many of the practical app1ications.involve the injection of one fluid
into a moving stream of another, and a useful eddy viscosity model
must be able to handle such cases. The casea of jet of hydrogen
injected into an air stream provides a stringent test of the theory
since there isa very large density gradient across the mixing zone.
This also represents a situation of current technological interest.
Some tests of Alpinieri (Ref. 22) and Zakkay and Krause (Ref. 20) will
be used to make the comparisons of theory and experiment, and the Ferri
model and the unified modelwill both be used with a constant turbulent
Schmidt number of0.75 (Refs. 1, 12). Consider first the data of Zakkay
and Krause (Ref. 20). Note that their external air stream was supersonic
(M 1.6), but it will be seen that this does not prevent these results
from fitting into the general picture we that
are attempting to construct.
Using the data for the case with J
u./ue
= 1.14 and p.u./peue= 0.14, a
J J
comparison between theory and experiment is given in No. Fig.10, on the
basis of the centerline decay of hydrogen concentration.It is clear that
both models underestimate the rate of decay in the far flow field. The
unified model is, however, in better quantitative and qualitative agree-
ment with the data.

It is interestingto consider a case with an even lower value of

pjuj/peue, and for this purpose Alpinieri's results for u./u = 0.67 and
J e
p.u./p u = 0.04 are employed. These experiments were made with a large
J J
jet (2$ ziameter) which reduces initial boundary layer effectsbut limited
the non-dimensional axial length of the data to =x/a 20. In this short
distance, it can be expected that the heuristic approximation of the eddy
viscosity in the initial region employed previously will introduce serious
error. The calculations were therefore begun using the measured profiles
at x/a = 5.0 as "initial" conditions. The development of the flow field
from x/a= 5.0 to x/a = 20 as predicted by the theory was then compared
with the data as shown in Fig. No. 11. Both models give results in reason-
able agreement with the data although the far downstreamofrate decay is
again under-estimated. Interestingly, the results for the two models are
in very close agreement. Apparently, thetwo models tend to become the
.
same in the limit of very small values of the parameter up.u./p
J J e e
One further aspect of the specification of an eddy viscosity model
was considered. The Prandtl model and all its descendants, including that
introduced here, are based upon the difference of some quantity across the
mixing region. When the particular difference upon which a given model is
based becomes zero, that model will predict a vanishing eddy viscosity.
This, of course, is notin agreement with common experience, and one must
simply accept this limitation upon the applicability of the given model.
The actual extent of that limitationis, however, very important.In the
present case, if the radius weighted integral of the mass flow difference
becomes zero, the limitation has certainly been reached. The case of uniform
initial mass flow (i.e. p.u./peue = 1.0) will produce this result,s o that
J J

11
the extent of the limitation of this modelcan be discussed in terms
of the initial mass flow ratio. caseA with p - u./peue = 0.63 (Ref. 26)
J J
was considered, and a comparison between theory and experiment is pre-
sented in Fig. No. 12. Clearly, the present model is still applicable,
and the limiting valueof p.u./p u must lie closer to unity.
J J e e
An additional case of this type, i.e. pjuj/peue near unity, with
hydrogen injection and Ju./uenot near unitywas studied. The experimental
data of Chriss (Ref. 27) for his case 1A (p.u./peue = 0.56, uj/ue = 6.3)
were compared with analytical predictions Juslng
.J
both the unified model and
the Ferri model. The calculation was started with experimental profiles
at x/a = 5.9. Results for centerline hydrogen concentration and velocity
decay are givenin Figs. No. 13 and 14, respectively. Here, as has been
generally true, the unified model produces a superior prediction.

DISCUSSION

Analysis of the turbulent mixing of a in jeta co-flowing external

stream has been considered in detail with particular emphasis placed on
the form for a model for the eddy viscosity. A new interpretation was
placed on Clauserls low speed, planar model for ‘bake”the region of a
turbulent boundary layer that permitted the derivation of a model for
cases with varying density and/or an axi-symmetric configuration. Exten-
sive comparisons between experiment and calculations using the unified
model as well as previous suggestions were made. Cases considered
included planar and axi-symmetric flows, heated and unheated jets, sim-
ilar and foreign gas injection, subsonic and supersonic external flows
and a range of the mass flux ratio, p.u./peuey from 0.04 to 4 . 0 . Good
J J
quantitative agreement between theory and experiment was obtained over
this range using the unified model for the eddy viscosity; the poorest
agreement was at the lowest values of p.u./peue with hydrogen injection.
J J
Most importantly, the axial rate of decay of the centerline values of
velocity, temperature and jet concentration in the axi-symmetric con-
figuration as predicted by the unified model in were
better agreement
with experiment than previously suggested models. In this regard, it
is interesting to note that the unified model and Ferri’s modeltend
to give identical predictions in the limit as the mass flux ratio be-
comes very small. In summary, it may be stated that the present results
in conjunction with those of(Ref. 361, demonstrate that the unified model
provides a unified description of the eddy viscosity in free mixing
flows including planar and axi-symmetric flows, varying temperature,
varying composition and axial pressure gradients.

The formal statement of the unified model permits easy extension

to truly three dimensional
cases, and some studiesin this area are planned.
Also, further work with, both experimental and analytical, strong axial
pressure gradients is needed.

12
 REFERENCES

1. A. H.: Momentum and Mass Transfer in

Forstall, W., Jr., Shapiro,
Vol. 72, pp. 339-408
Coaxial Gas Jets. Journal of Applied Mechanics,
(1950).
2. Pai, S. I.: Fluid Dynamics of Jets. Chapter V, VI and V I I , D. Van Nostrand
Co., Inc., New York(1954).

3. Townsend, A. A.: The Structure of Turbulent Shear Flow. Chapter8, Cambridge

Univ. Press (1956).

4. Abramovich, G . N.: The Theory of Turbulent Jets. MIT Press, Cambridge,

Mass. (1963) .
5. Schlichting, H.: Boundary Layer Theory. Fourth Edition, Chapter XXIII,
, New York(1962).
McGraw-Hill Book Co.

6.

7. Halleen, R. M.: A Literature Review on Subsonic Free Turbulent Shear Flow.

Rept. MD-11, Mechanical Engrg. Dept., Stanford University
(1964).

8. Ledgett, L. A.: Mixing of Fluid Streams. ASME Summer Meeting, Berkeley,

Calif. (1934).

9. Viktorin, K.: Investigation of Turbulent Mixing Processes. English Trans-

lation, NACA TM-1096(1941).

10. Pabst, 0 . : Die Ausbreitung Heisser Grasstrahlen in Bewegter Luft,

I Teil-
Versuche in Kerngebeit. Deutsche Luftfahrtforschung, Aug.(1944).

11. Ferguson, C. K.: Mixing of Parallel Flowing Streams in

a Pressure Gradient.
Heat Transfer and Fluid Mechanics Institute (1949).

12. Landis, F. , and Shapiro, A. H.: The Turbulent Mixing of Co-axial Gas Jets.
Heat Transfer and Fluid Mechanics Institute (Preprints and Papers) Stanford
University Press (1951).

13. Helmbold, H. B., Luessen, G., and Heinrich,A. M.: An Experimental Com-
parison of Constant Pressure and Constant Diameter Jet Pumps. Univ. of
Wichita, Engineering Rept. No. 147(1954).

14. Weinstein, A. S., Osterle, J. F., and Forstall, W.: Momentum Diffusion
from a Slot Jet into a Moving Secondary. Journal of Applied Mechanics,
pp. 437-443, Sept. (1956).

15. Curtet, R.: Confined Jets and Recirculation Phenomena with Cold Air.
Combustion and Flame,2, pp. 383-411 (1958).

13
 _. .. ___ . _.. .

16. Mikhail, S.: Mixing of Coaxial Streams insidea Closed Conduit.

J. of Mechanical Engrg. Sci., Vol.2, No. 1, pp. 59-68 (1960).

17. Maczynski, J. F.: A Round Jet in an Ambient Coaxial Stream. J. of Fluid

Mechanics, 13, pp. 597-608 (1962).

18. Ferri, A., Libby, P. A., and Zakkay, V.: Theoretical and Experimental
Investigation of Supersonic Combustion. Third ICAS Conference, Stockholm
(1962) .
19. Becker, H. A., Hottel, H. C., and Williams, G. C.: Mixing and Flow in
Ducted Turbulent Jets. Ninth Symposiumon Combustion, Academic Press,
pp. 7-21 (1962).

20. Zakkay, V., and Krause, E.: Mixing Problems with Chemical Reaction. 21st
Meeting AGARD Combustion and Propulsion Panel, London(1963) also in
Supersonic Flow, Chemical Processes and Radiative Transfer, Pergamon Press.
(1964).
21. Marquardt Corp., TMC Report PR297-24, Second Quart. Prog. Rept. AF Contract
NO. AF 33(657)-8431 (1963).

22. Alpinieri,L. J.,: Turbulent Mixing of Coaxial Jets. AIAA Journal,pp. 1560-
1567 (1964)*

23. Zakkay, V., Krause, E. and Woo, S. D. L.: Turbulent Transport Processes
for Axisymmetric Heterogeneous Mixing. AIAA Journal, pp. 1939-1947 (1964).

24. Barchilon, M. and Curtet, R.: Some Details of the Structure of an Axi-
of Basic Engrg., pp. 777-787,
symmetric Confined Jet with Back Flow. Journal
Dec. (1964).
25. Bradbury, L. J. S. and Riley, J.: The Spread of a Turbulent Plane Jet into
a Parallel Moving Airstream. Journal of Fluid Mechanics, V o l . 27, Pt. 2,
pp. 381-394 (1967).

26. Henry, J. R.: Private Communication.

27. Chriss, D. E.: Experimental Study of Turbulent Mixing of Subsonic Axi-

symmetric Gas Streams. AEDC-TR-68-133,August (1968).

28. Libby, P.A.: A Theoretical Analysis of the Turbulent Mixingof Reactive

Gases with Applicationto the Supersonic Combustion of Hydrogen. ARS Journal,
pp. 388-395, March (1962).

29. Schetz, J.A.: Supersonic Diffusion Flames. In Supersonic Flow, Chemical

Processes and Radiative Transfer, Pergamon Press
(1964).

30. Zeiberg, S. L. and Bleich, G. D.: Finite Difference Calculationof Hyper-

sonic Wakes. AIAA Journal, pp. 1396-1402, August (1964).
31. Clauser, F. H.: The Turbulent Boundary Layer. in Advances in Applied
Mechanics, Vol. IV, Pergamon Press(1956).

32. Ting, L. and Libby, P. A.: Remarks on the Eddy Viscosity in Compressible
Mixing Flows. Journal Aerospace Sci., pp. 797-798, October (1960).

33. Bloom, M. H. and Steiger, M. H.: Diffusion and Chemical Reaction in

Free Mixing. IAS Annual Meeting, PaperNo. 63-67 (1963).

34. Townsend, A.A.: The Eddy Viscosity in Turbulent Shear Flow. The Philoso-
phical Magazine, pp. 890-906, September(1960).

35. Keagy, W. R. and Weller, A. E.: A Study of Freely Expanding Inhomogeneous

Jets. Heat Transfer and Fluid Mechanics Inst., ASME, New York(1949).

36. Schetz, J. A. and Jannone, J.: Planar Free Turbulent Mixing with an
Axial Pressure Gradient. Journal of Basic Engrg., December (1967).

37. Schetz, J. A. and Jannone, J.: A Study of Linearized Approximations to

the Boundary Layer Equations. Journal of Applied Mechanics, pp. 757-764,
December (1965).

38. Zeiberg, S. L. and Kaplan, B.: Approximate Analysis of Free Mixing Flow.
ATAA Journal, pp. 552-554, March(1965).
 TABLE I
SUMMARY OF EXPERIMENTS FOR
A JET IN A CO-FLOWING STREAM

Axi-
Variable
Foreign
Pressure
Gas
Planar
symmetric
No.
Author
Ref.
Year
Temperature
Injection
Gradient

Ledget t 1934 8 X

Viktorin 1941 9 X

Pabst 1944 10 X X

Ferguson 1949 11 X

Forstall & Shapiro 1950 1 X He

Landis & Shapiro 1951 12 X X He,C02
Helmbold, et a1 1954 13 X X

Weinstein, et a1 1955 14 X

Curtet 1958 15 X X
Mikhai1 1960 16 X
Maczynski 1962 17 X
Ferri, et a1 1962 18 X
H2
Becker, et a1 1962 19 Aerosol X

Zhestov, et a1 1963 4(pp.28,344) x X

Yakovlevskiy & 1963 4(PP. 274) X

Pechenkin
Borodachev 1963 4(PP. 29) X

Zakkay & Krause 1963 20 X

H2
Marquardt Corp. 1963 21 X
H2
Alpinieri 1964 22 X H2 3CO2
Zakkay, et a1 1964 23 X H2 ,He,C02,A
Barchilon & Curtet 1964 24 X X

Bradbury & Riley 1967 25 X

Torrence & Eggers 1967 26 X X

Chriss 1968 27 X X
H2
 TABLE II EDDY VISCOSITY MODELS FOR MAIN MIXING REGION OF JETS AND WAKES

s (pp.48S) ~ x ~ x

REWIRES TWO MIXING LENGTHS 1

lNTR0DUCED"VELDCITY DIFFERENCE"
1 CONCEPT; WITH b TAKEN AS bh, K~
5 (pp. 481) X 0.037 IN PLANAR JETS AND
I( 0.25 IN AXCSYMMETRIC JETS.

SCHLICH -INC
I
1942 5

I x ~ 0.0222 y e cDa WAKE OF A CYLINDER OF ARBITRARY

CROSS SECTION

Il x 1l
\PPLIED TO "WAKE"-LIKE OUTER
CLAUSER 19S6 31 REGION OF A BOUNDARY LAYER,
0.016 < K < 0.018

HlNZE 1959 6 I I 0.016 ye d

I WAKE OF A CIRCULAR CYLINDER
I
Lo IS THE CONSTANT DENSITY EDDY
TINO-LIBBY 1960 32 VISCOSITY AND pc IS THE CENTER.
LINE DENSITY

lx l
z 0 IS THE CONSTANT DENSITY EDDY
TING-LIBBY 1960 32 p2, = PC 2, VISCOSITY AND ps IS THE CENTER-
LINE DENSITY
I
~
EXTENDED PRANDTL'S THIRD MODEL
FERRI,ET AL 1962 18 l x TO VARIABLE DENSITY, INTRODUCED
"MASS FLOW DIFFERENCE"CONCEP1

ATTEMPT TO EXTEND PRANDTL'S

BLOOM 6 STEIGER 1963 33 X THIRD MODEL TO VARIABLE DENSITY,
6' IS TRANSFORMED WAKE RADIUS

SIMPLE APPLICATION OF "MASS FLOW

SCHETZ 1963 28 X
DIFFERENCE"T0 PLANAR FLOWS
I

ALPlNlERl 1964 22

PRESUMES THAT CENTERLINE

ZAKKAY, ET AL 1964 23 0.011 by, uc VELOCITY A D CONCENTRATION
DECAY AS x'
"TURBULENT VISCOSITY PRO.
SCHETZ 1968 PRESENT X PORTIONAL TO MASS FLOW UNIFIED MODEL
DEFECT (OREXCESS) IN THE
MIXING REGION"
 H2-AIR, REF 27 * *

0 H,-AIR, REF 20
0 H,-AIR, REF 22
- 0 A-AIR, REF 23
C0,-AIR,REF 22
He-AIR,REF 1
- B He- AIR, REF 12
AH=,CO,, N,-AIR,REF. 35

0.01 0.02 0.05 0.10 0.20 0.50 1.0 2.0 5.0 10.0

Fig. 1 EXPERIMENTALDETERMINATION OFASYMP-

TOTIC CONCENTRATION DECAY EXPONENT
1.oo

0.75

0.50

0.25 --
0 50 100 150 200

Fig. 2 THEORY AND EXPERIMENT FOR PLANAR, CONSTANT

DENSITY JET MIXING, Ui/U, = 2.0
 0.05

0.04
I

0.03
0

15-

0.02 I

I
I

I I

- X,/a
-
- - I
PRANDTL MODEL
CLAUSER
MODEL
I

0- FREE JET BOUNDARY

0.01 w

0
25 50 75 100
X /a

Fig. 3 EDDYVISCOSITYMODELS INPLANARJETMIXING

20
 1.oo

0.75

Q Q
3 3
u .- 0.50
3 3

0.25

Fig. 4 THEORYANDEXPERIMENT FOR AXISYMMETRIC,

CONSTANT DENSITY JET MIXING, Ui/U, = 2.0

21
 1 .oo

0.75

0.50 L

e
3
I
.I

3
0.25 t
- 0 0 EXPERIMENT (REF (1))
-THEORY, PRANDTL MODEL
--THEORY, NEN MODEL

Fig. 5 THEORY AND EXPERIMENT FOR ASYMPTOTICDECAY

OF CENTERLINE VELOCITY FOR AXISYMMETRIC,
CONSTANT DENSITY JET MIXING, Ui/U, = 2.0
 I

0.015 1

l 0)
w
"", "-
O.QO5
/
I

-
- - BRANDTLMODEL

,/'
-- NEW MODEL
FREE JET BOUNDARY

n
"0 25 50 75 100 125
X /a

Fig. 6 EDDYVISCOSITYMODELS IN AXISYMMETRIC JET

Me XING
 1 .oo

0.75 -

+I? 0.50
2 3-

I I "
0.25
-1

-- EXPERIMENT(REF (12))
THEORY, FERRI MODEL
--THEORY, NEW MODEL

0
25 50 75 1 00
X /a
Fig.7THEORY AND EXPERIMENT FOR HEATED, AXISYMMETRIC
AIR J ET; Ui/Ue=2.0, Ti/Te= 1.19

24
 1.00

0.75

Q,
--
0

--
EXPERIMENT(REF (12))
THEORY, FERRI MODEL
THEORY, NEW MODEL
3 -
.-I 0.50
3

O - 10101
25 50 75 1 00
X /a
Fig. 8 THEORY AND EXPERIMENT FOR HEATEDAXISYMMETRIC
= 4.0, Ti/T, = 1.31
AIR JET; Ui /U,

25
 1 .oo
Ui/lJ, = 4.0
Ti/T, = 1.31

\'\\
*
e
0.75
--
--
0 EXPERIMENT (REF (12))
THEORY, FERRI MODEL
THEORY, NEW MODEL

e o
7 7 0.50
k" G-

-
*i
0.25

-"
I

\ ""
+-
--- -
Fig. 9 THEORY AND EXPERIMENT FOR TEMPERATURE
FIELD OF HEATED AXISYMMETRIC AIR JET;
Ui/Ue = 4.0, Ti/Te = 1.31

26
 0.7!

I
I

I
0.5( i
-1
u p e= .14

i
W
-
.z
J
K
W
I-
z
W
U
h
cy
I
*
I
Y

--
--
0 EXPERIMENT(REF
FERRI MODEL
NEW MODEL
(20))

0.2!

P 0

25 50 75 1
X/a

Fig. 10 THEORY AND EXPERIMENT FOR AXISYMMETRIC

HYDROGEN JET MIXING IN AIR,
~iUi/~cUc=0.140

27
I
U*/U = 0.67
I =
-
J
0.20
Q!
W
I I
-- --
I-
Z
W 0 EXPERIMENT(REF (22))
I u FERRI MODEL
n
(v NEW MODEL
I
I Lu
co = 0.10 W
I
)I
Y I I

0
I I
5 10 15 20 25.

Fig. 11 THEORY AND EXPERIMENT FOR AXISYMMETRIC

HYDROGEN JET MIXING IN AIR,
I PeUe = 0,040
Pi U*/
1 .oo

- DATA(REF. 40)
THEORY

X /a

Fig. 12 THEORY AND EXPERIMENT FOR AIR JET MIXINGIN

I
A SUPERSONIC AIR SYSTEM,PjUj
-
PeUe
= 0.63
 1.00

0.75

d
>-” 0.50

0.25

10 20 30 40

F i g . 13 THEORY AND EXPERIMENT FOR CONCENTRATION

DECAY WITH HYDROGEN JET IN AIR,piUi
”
- 0.56
PeUe

30
1

1 .oo

I UNIFIED MODEL
I MODEL
0.75

hr

1 -u, 0.50
1 -ui

0.25

0
0
L1
10 20

X /a
L
30 40 50

Fig. 14 THEORY AND EXPERIMENT FOR VELOCITYVARIATION

WITH HYDROGEN JET IN AIR,piUi
-= 0.56
P eUe

31