Advances in Heat Transfer 2 (1965)

Contributors to Volume 2
D. R. BARTZ
PAUL M. CHUNG
F. M. DEVIENNE
A. B. METZNER
E. M. SPARROW
Advances in
HEAT
TRANSFER
Edited by
James P. Hartnett Thomas F. Irvine, Jr.

Department of Mechanical State Uniuersity of New YorA
Enginemng nt Stony Brook
University of Delaware Stony Brook, Long Island
Newark, Delaware Nau York
Volume 2
ACADEMICPRESS New York London

COPYRIGHT
0 1965, BY ACADEMIC PRESSINC.
ALL RIGHTS RESERVED.
NO PART OF THIS BOOK MAY BE REPRODUCED I N ANY FORM,
BY PHOTOSTAT, MICROFILM, OR ANY OTHER MEANS, WITHOUT
WRITTEN PERMISSION FROM THE PUBLISHERS.
ACADEMIC PRESS INC.

111 Fifth Avenue, New York, New York 10003
United Kingdom Edition published by

ACADEMIC PRESS INC. (LONDON) LTD.
Berkeley Square House, London W.l
OF CONGRESS
LIBRARY CATALOG
CARDNUMBER:63-22329
PRINTED I N THE UNITED STATES OF AMERICA

LIST OF CONTRIBUTORS
D. R. BARTZ, Jet Propulsion Laboratory, California Institute of Tech-

nology, Pasadena, California
PAUL M. CHUNG, Aerospace Corporation, San Bernardino, California
F. M. DEVIENNE, Laboratoire Miditerranken de Recherches Thermo-

dynamiques, Nice, France
A. B. METZNER, University of Delaware, Newark, Delaware
E. M. SPARROW, Heat Transfer Laboratory, Department of Mechanical

Engineering, University of Minnesota, Minneapolis, Minnesota
PREFACE
Since the appearance last year of Volume 1 of Advances in Heat

Transfer, research in this special field continues unabated, primarily
associated with the atomic energy industry, and the aerodynamics and
astronautics efforts. Development of new instrumentation and refine-
ment of. high speed computers continues to improve our experimental
and analytical capacities, and accordingly we are able to attack new
and more complex problems in a much more definitive fashion. T h e
results of these research efforts are normally published as individual
articles in national and international journals. I t is understandable that
such journal articles, because of space limitations, assume that the
reader be well aware of the existing state of knowledge, and so present
in an abbreviated and concise manner the new piece of information.
It is extremely difficult for a nonspecialist to make engineering use of
individual papers appearing in such a journal. It is clear from time to
time-as a given area in heat transfer evolves to a definitive state-that
a review article or a monograph which starts from widely understood
principles and develops the topic in a logical fashion would be of value
to the engineering and scientific community. It is our continued hope
that Advances in Heat Transfer will fulfiIl this function.
April, 1965 JAMES P. HARTNETT

THOMAS F. IRVINE,
JR.
[vii]
Turbulent Boundary-Layer Heat Transfer
from Rapidly Accelerating Flow of
Rocket Combustion Cases
and of Heated Air1
.
D. R BARTZ
Jet Propulsion Laboratory. California Institute of Technology
Pasadena. California
I . Introduction . . . . . . . . . . . . . . . . . . .... 2
A. Origin of the Problem . . . . . . . . . . . . .... 2
B. An Approach to the Solution . . . . . . . . . . . .. 4
.
C Background of Analyses of the Problem . . . . . . . .. 5
.
I1 Analyses . . . . . . . . . . . . . . . . . . . . . . . . 8
A . Integral Momentum and Energy Equation Solution . . . . 8
.
B Closed-Form Approximation . . . . . . . . . . .... 32
C . Transport Properties . . . . . . . . . . . . . . . . . 36
D . Variable Properties . . . . . . . . . . . . . . . . . . 39
E . Driving Potential . . . . . . . . . . . . . . . . . . . 44
111. Air Experiments . . . . . . . . . . . . . . . . . . . . 52
.
A Purpose . . . . . . . . . . . . . . . . . . . . . . . 52
B. Literature . . . . . . . . . . . . . . . . . . . . . . 52
C . Experimental Techniques . . . . . . . . . . . . . . . 53
D . Experimental Results . . . . . . . . . . . . . . . . . 57
E . Effect of Acceleration on Turbulence . . . . . . . . . . 72
IV. Rocket Thrust-Chamber Measurements . . . . . . . . . . . 74
A. Literature . . . . . . . . . . . . . . . . . . . . . . 14
B. Experimental Techniques . . . . . . . . . . . . . . . 75
C. Combustion-Chamber Heat Flux . . . . . . . . . . . . 19
D . Nozzle Heat Flux . . . . . . . . . . . . . . . . . . 87
This Chapter is contributed jointly by the author and the Jet Propulsion Laboratory.
California Institute of Technology. where portions of the work were done under Depart-
ment of the Army Contract No . DA-04-495-0rd 18 and National Aeronautics and
Space Administration Contract Nos . NASw-6 and NAS 7.100 .
r11
D. R. BARTZ
V. Concluding Remarks . . . . . . . . . . . . . . . . . . 95
Appendix A. Coles' Skin-Friction Coefficient and von Kdrman
Form of the Reynolds Analogy. . . . . . . . . . . . . . . 98
Appendix B. Boundary-Layer Shape Parameter Evaluation . . . 99
Appendix C. Boundary-Layer Thicknesses and Integral Equations
for Thick Boundary Layers................. 101
Nomenclature . . . . . . . . . . . . . . . . . . . . . 102
References . . . . . . . . . . . . . . . . . . . . . . . 105
I. Introduction
A. ORIGIN
OF THE PROBLEM
Ever since the development of rocket engines for practical applica-
tion, there has been a recognized need to predict the heat transfer from
the combustion gases to the walls of both the combustion chamber and
the nozzle. Since, in the early days, these walls were generally constructed
of materials with negligible strength above about 1500F and had to
contain gases at pressures of a few hundred pounds per square inch
and temperatures of 4000-500OoF, the consequence of underdesigned
wall-protection provisions was a serious local wall failure and, frequently,
a blown-up engine; the consequence of grossly overdesigned wall-
protection provisions was excessive pressure drop and weight, or demands
of shifts in the engine operating mixture ratio toward lower performance.
As a result, the prediction of heat transfer with sufficient accuracy to
avoid failures and sacrifices in weight or performance became increasingly
important. More recently, chamber pressures of large booster engines
have reached the 1000-lb/in.2 level and there %reindications that in the
foreseeable future these pressures may be doubled. I n addition, the use
of more energetic propellants has driven combustion gas temperatures
of these engines up near 8000F. New trends in smaller engines for
upper stages and for spacecraft have also increased the demand for
knowledge of the heat transfer and boundary-layer growth. In the
interests of simplicity, or in the absence of sufficient or suitable propellant
coolant, use has been made of ablating walls or refractory metal walls
cooled by radiation. Although there is no general acceptable theory of
the ablation process for heterogeneous materials, it has been shown to
be related to the heat transfer ( I , 2). T h e need for detailed knowl-
edge of the momentum losses or boundary-layer growth in the
supersonic part of nozzles has been increased by the trend toward very
large expansion ratios, which result in performance gains for space
operation. At some area ratio, depending on the design of the wall, the
gains in performance obtained by a still higher area ratio are offset by the
121
HEATTRANSFER
FROM RAPIDLY
ACCELERATING
FLOWS
added weight of the wall. Hence, it becomes important to ascertain the

real performance by assessing the friction losses. Other recent develop-
ments such as generating shocks to provide thrust-vector control also
require knowledge of boundary-layer development. Thus, the need for
knowledge of both heat transfer and boundary-layer development in
rocket thrust chambers (used commonly to refer collectively to both
combustion chamber and nozzle) has expanded with the years during
which rocket engine development has evolved. Fortunately, the knowl-
edge available has expanded significantly as well. However, as will
become evident from this chapter, the problem is not solved.
One might ask in what way heat transfer and boundary-layer develop-
ment in rocket thrust chambers are so special that the problem is still
not solved after more than a decade of concentrated analysis and experi-
ment. For most flow fields too complex to permit exact solutions, the
practice has long been to create a model by making typical, and usually
permissible, assumptions such as inviscid core flow, laminar boundary
layer or turbulent boundary layer with specified eddy difhsivity, one-
dimensional, constant properties, steady flow, etc. T h e deviations of
the real flow from the model are accounted for by small correction
factors determined by correlating data for the real flow against predictions
of the model. T h e problem with the rocket thrust-chamber flows is that
the real flow is characterized by numerous deviations from flow des-
cribable by a simple model, and the deviations are not necessarily small.
T h e most significant deviation or complexity is that the free stream
flow cannot in general be successfully described in terms of steady,
average, one-dimensional flow variables. In particular, the flow in the
combustion region (and this may include a substantial portion of the
nozzle, depending on propellants and configurations) is frequently
characterized by severe large-scale secondary flows, nonlinear oscilla-
tions, and variable total temperature. Each is characteristic of particular
propellants, propellant injectors, operating conditions, and combustion-
chamber configurations. As yet, our knowledge of combustion is insuffi-
cient either to predict or to control this behavior. Since heat-transfer
predictions cannot proceed beyond our ability to describe the fluid
dynamics and energy states, it should begin to be evident why prediction
of heat transfer and boundary-layer development in rocket thrust
chambers is difficult. T h e next most significant complexity is that the
free stream, and hence the boundary-layer flows, are rapidly accelerating;
thus, it becomes impossible to neglect axial pressure-gradient terms in
the momentum and energy equations. Because of this, it is no longer
possible to express the momentum and energy equations in similar form,
and hence to derive useful analogies between momentum and energy
131
D. R. BARTZ
transport except in an heuristic fashion. Another significant complexity

is the possible occurrence of chemical reaction in the free stream which
leads to an axially varying total temperature or enthalpy, and chemical
reaction in the boundary layer due to the recombination of dissociated
chemical species. Since most of the chemical reactions in question are
exothermic, they can play a pronounced role in modifying the driving
potential for heat transfer. A further significant complexity is the fact
that for most rocket flows of interest the boundary layer (and probably
the free stream as well) is very likely to be turbulent. One arrives at this
conclusion by noting the generally very high Reynolds numbers due to
very high mass flow rates per unit area, not compensated by unusually
high viscosity or small linear dimensions. Furthermore, from the usual
free-stream turbulence due to combustion, one would expect boundary-
layer transition to occur at unusually low Reynolds numbers, out-
weighing the stabilizing effects of cooling and acceleration. There are
still other complexities such as extreme property variations across the
boundary layer, an uncertain flow origin, separation due to high ambient
pressure leaking up the subsonic portion of the boundary layer in
the divergent part of the nozzle, etc.
B. AN APPROACH
TO THE SOLUTION
I n the face of a problem with as many complicating elements as
described, one must adopt a pragmatic approach such as restricting
the analysis to that part of the problem that can be handled (or
almost handled with plausible assumptions) and then fully recognize
the limitations of the result. One can only hope to remove the
currently necessary restrictions by deeper specific knowledge of the
phenomena involved, to be gained by carefully instrumented and
controlled experimentation coupled with thorough analysis. Among
the currently necessary restrictions in this authors opinion are that
(1) the flow to be considered be beyond the region of severe secondary
flows due to combustion; (2) the ultimately attainable combustion
temperature be established in the free stream, and (3) the engine
be operating without significant combustion pressure oscillations.
(Research on the effect of pressure and velocity fluctuations (see 3, 4, 5 )
may some day make it possible to avoid the latter restriction.) A further
necessary restriction is that chemical recombination in the boundary
layer proceed according to local chemical equilibrium conditions or the
close equivalent, a diffusion-controlled chemically frozen boundary layer
with a catalytic wall. [Theory and experiment of the type described by
Rosner (6) could make this restriction unnecessary if certain assumptions
r41
HEATTRANSFER
FROM RAPIDLY FLOWS
ACCELERATING
are made. However, critical experiments, such as reaction rate measure-

ments, that would determine which of such assumptions are plausible
have not been made in the rocket engine environment.] Additional
restrictions are that the boundary layer be fully turbulent, have some
specified thickness at the starting point of the analysis (such as that
appropriate to growth in a pipe entrance region of approximately the
combustion-chamber length when starting at the nozzle entrance), and
that the region of unseparated flow be of primary interest. Thus, it is
evident that, in effect, one has either had to ignore the combustion zone
or to assume its heat transfer to be no higher than that predicted at the
starting point of the analysis. This may or may not be a good assumption,
as will be evident later. With these restrictions, it is possible, by making
a number of plausible assumptions, to predict both boundary-layer
development and heat transfer. T h e plausible assumptions, and
methods for making such predictions are discussed in Section I1
of this chapter. T o establish the validity of these methods of
prediction, it is necessary to compare the predictions with the results
of carefully controlled experiments in which existence of the restricted
conditions assumed is assured. This is done to the limit of the availability
of such experimental results in Section 111. Finally, one must compare
the predictions with results from measurements of real rocket thrust-
chamber flows under a wide variety of conditions to determine to what
extent the real flows deviate from that assumed in the model. Such
comparisons are made, also to the limit of availability of experimental
results, in Section IV.
C. BACKGROUND
OF ANALYSES
OF THE PROBLEM
Before proceeding with a current version of the analysis of the problem

in Section I, B, it might be of value to trace the stages of evolution of
earlier analyses of this problem. Initially, for want of better information
on turbulent boundary layers in nozzles, the classic turbulent pipe-flow
heat-transfer correlation equations of McAdams and of Colburn (7)
were applied by considering the nozzle flow to be a series of fully
developed turbulent pipe flows. Each point in the nozzle was assumed
to have been preceded by a very long pipe of the local diameter of
interest. Because this approach seemed to work (although there
was a very limited amount of local heat-flux data with which
to compare it), there was a tendency to lose sight of the fact that
the flow was by no means fully developed in the sense of boundary
layers extending to the flow axis of symmetry. Not satisfied with
the apparent incompatibility of the actual flow regime with that
~51
D. R. BARTZ
which served as the basis for the analytical prediction, several

workers attempted to solve the nozzle heat-transfer problem from a
boundary-layer viewpoint making use of the integral momentum and
energy equations (8, 9). T h e essential difference between the nozzle
problem and most of the turbulent boundary-layer analytical treatments
then published was the necessity for retaining the pressure gradient
terms in the equations of motion. The way was already partially paved
since the momentum transfer problem had been solved by approximate
methods under the impetus of computing boundary-layer corrections
to the contours of supersonic wind-tunnel nozzles (10, 11, 22). Experi-
mental results were found to agree quite satisfactorily with predicted
boundary-layer thicknesses.
The new extension achieved in refs. (8) and (9) was the handling
of the heat-transfer, as well as the momentum-transfer problem.
Numerical results from the approximate solutions obtained were
found to agree reasonably well with limited experimental data then
available and with predictions made on the pipe-flow basis except in
nozzle-entrance regions. Here, the possibility of extremely thin boundary
layers was shown to result in correspondingly high local heat fluxes,
as should be expected. Since the boundary-layer approach was considered
physically valid, it was expected that this method of analysis would
serve the purpose of making reasonably accurate predictions of con-
vective heat transfer in supersonic nozzles. It was anticipated that the
analyzis would be improved as further basic knowledge of the skin
friction and heat transfer in accelerating turbulent boundary layers was
obtained.
I n that era, which preceded the wide availability of high-speed com-
puters and the practice of sharing programs among organizations, it
soon became evident that a method of analysis requiring the solution
of a pair of differential equations with coefficients varying in accordance
with each particular nozzle contour could not be used very widely.
Consequently, a closed-form equation which could be hand-computed
and which closely approximated the results of the boundary-layer anal-
ysis for a particular typical nozzle configuration and typical initial
boundary-layer conditions was sought and found. It was clearly evident
from the boundary-layer heat-transfer calculations (9)that the dominant
parameter in the variation of the local heat-transfer coefficient was
the local mass flux raised to the eight-tenths power. This suggested
the possibility of again utilizing the dimensionless parameter
approach employing Reynolds number, Prandtl number, and Nusselt
number. Such an approach, however, raised the question of the charac-
teristic length dimension to be used. A review of the boundary-layer
[61
HEATTRANSFER FLOWS
FROM RAPIDLY ACCELERATING
development in a nozzle (9, Fig. 3) showed that the local boundary-layer

thickness varied in a systematic relationship with the local diameter,
suggesting that the local diameter be used as the characteristic length.
When the diameter was so employed, the dimensionless equation looked
identical in form to the McAdams and Colburn pipeflow equations
with a proportionality constant to be determined, thus accounting for
the early success of such equations when applied to nozzle flow. T h e
proportionality constant was determined by fitting the closed-form
equation to the boundary-layer heat-transfer calculations at the throat
for a particular case, estimated to be reasonably typical of then current
rocket nozzles. Some additional minor modifications resulting from
variable properties considerations, and effects of throat radius of curva-
ture were deduced from the boundary-layer results and applied to the
closed-form equation, giving the result published in ref. (13).
This closed-form equation served its purpose quite satisfactorily until,
with time, several changes occurred. First, with the increasing availability
of high-speed computers, the compromises inherent in such an equation
were no longer necessary; an exact solution, to the extent permitted by
knowledge of the turbulent boundary layer, could be computed almost
as readily as the closed-form equation. Second, nozzles of interest were
no longer restricted to simple conical convergent-divergent nozzles
in which local flow conditions were easily expressible in terms of local
area ratio. So-called bell nozzles resulted in regions of severely turned
flow, in which the mass flux near the wall was considerably different
from that predicted by one-dimensional calculations. Annular-throat
nozzles of the plug type also raised the question of the applicable local
diameter to be used in the closed-form equation. Third, the closed-form
equation provided only heat-transfer coefficients, whereas the increased
precision required of rocket nozzle design made it desirable to know
such boundary-layer parameters as the displacement and the momentum
thicknesses. These thicknesses permit computation of nozzle performance
corrections and provide nozzle-contour corrections for calculations of
the free-stream flow. Thus, it appeared desirable to reformulate the
turbulent boundary-layer heat-transfer equations in a form suitable to
accommodate all of these new requirements, to eliminate compromises
originally made to ease computational difficulties, and to program the
result for digital-computer solution (14). At the same time, the analysis
in ref. (9) was reexamined in the light of new information and altered
where it seemed advisable.
T h e Blasius skin-friction formula employed in ref. (9) was replaced
by Coles correlation (15), which better fits the data at high Reynolds
and Mach numbers. Momentum thickness was made the characteristic
171
D. R. BARTZ
dimension in computing the skin-friction coefficient since this thickness

is a more fundamental property of the boundary layer than the velocity
thickness employed in ref. (9), and is the dimension employed in
Coles correlation. Energy thickness with a correction for differing
momentum thickness was made the characteristic dimension in com-
puting Stanton number, rather than velocity thickness with a correction
for temperature thickness. Mach number at the edge of the boundary
layer was made an optional parameter to be prescribed in place of area
ratio, facilitating application to nozzles of the bell and plug type. For
convenience, axial distance, rather than distance along the wall, was made
the position variable. Adiabatic recovery temperature, instead of stagna-
tion temperature, was made the driving potential in computing heat
flux, improving accuracy at high Mach numbers, and provision was
made for optionally employing enthalpy (rather than temperature)
driving potential in cases in which chemical reaction or variable specific
heat must be considered. T h e momentum and static-temperature-distri-
bution equations of ref. ( 9 ) were corrected to apply more accurately
to unequal momentum and energy thickness. Finally, a simultaneous,
iterative solution of the momentum and energy equations was formulated,
rather than stopping at the first approximation as in ref. (9).
Unfortunately, the intervening years have shed no fundamental light
on the most important postulate of ref. (9), that the skin friction and
heat flux at any point in a nozzle are the same as they would be on a
flat plate at the same free-stream conditions and boundary-layer thick-
nesses. Furthermore, as discussed later, the question of a variable-proper-
ties correction for severely cooled boundary layers has become clouded
rather than clarified. Although little more basic insight into these
questions has been achieved, a considerable amount of supersonic-nozzle
and rocket-thrust-chamber data has been obtained with which some
degree of gross comparison can be made. The need for additional
specific experimental results will become apparent from the compari-
sons.
11. Analyses
A. INTEGRAL
MOMENTUM SOLUTION
AND ENERGYEQUATION
T h e only method of approach amenable to analysis known to this

author for the solution of the turbulent boundary-layer development
and local heat transfer in rapidly accelerating flows is the simultaneous
solution of the integral forms of the boundary-layer momentum and
PI
HEATTRANSFER FLOWS
FROM RAPIDLYACCELERATING
energy equations. As mentioned in the Introduction, Sibulkin (8) and

Bartz (9) describe some of the earliest attempts at such solutions. Others
have been discussed in the literature but have differed only in minor
detail or, in the interest of devising a simpler method, have necessarily
involved considerably more arbitrary assumptions and heuristic argu-
ments. I t is beyond the scope of this chapter to critically review and
compare detailed results from these analyses since the differences in
most cases are small compared with the differences between the pre-
dictions and experimental results. The purposes of this chapter are
served satisfactorily by the derivations adapted from Elliott et al. (14)
and presented in this section, which illustrate the essential features of
the analyses. I t is important to recall that one necessarily proceeds with
analyses of the turbulent boundary layer primarily through plausible
assumptions and intuitive arguments. Fortunately, errors and uncertain-
ties introduced by approximations made in determining the development
of the boundary layers are reduced considerably by the fact that the
boundary-layer thicknesses enter into the heat-transfer coefficient to
about the power. T h e most direct effect on the heat-transfer
coefficient is encountered in the skin-friction coefficient and Stanton-
number correlations adopted. Unfortunately, the correlation equations
must be based on experiments, the results of which do not always
agree.
T h e integral momentum and integral energy equation of the turbulent
boundary layer are usually derived either ( 1 ) from integration of the
Prandtl boundary-layer equations, with certain questionable assumptions
made about the turbulent fluctuation correlation terms; or (2) from the
control-volume viewpoint, in which these turbulent fluctuation terms
are ignored. T h e derivation presented here, although related to the
second approach, differs by starting with the displacement, momentum,
and energy boundary-layer thicknesses as basic definitions of the respec-
tive deficiencies in mass, momentum, and energy resulting from friction
and heat transfer. This derivation is based on comparison of the real flow,
with a hypothetical adiabatic potential flow extending all the way to
the wall of a slightly different nozzle and having the same wall static-
pressure distribution and total mass flux as the real flow. I n the following
treatment, the nomenclature employed for the real-flow and potential-
flow nozzles will be introduced first, followed by the definitions of the
displacement, momentum, and energy thicknesses, a discussion of the
assumptions employed, the derivation of the integral momentum and
energy equations, the presentation of the skin-friction and heat-transfer
correlation equations adopted, and the derivation of relations for the
boundary-layer shape parameters linking the various thicknesses.
[91
D. R. BARTZ
I. Nomenclature
a. Real Flow. Figure 1 presents the nomenclature for the real nozzle
flow. T h e stagnation conditions of the gas flowing through the nozzle are:
temperature T o , pressure p o , specific heat cp , specific heat ratio y ,
Prandtl number Pr, and viscosity po . At a given station, the distance
along the nozzle axis is z, the distance along the wall is x, the radius of
the wall from the axis is 7 , the wall temperature is T, , the wall shear
stress that retards the fluid motion is T ~ and , the heat flux to the wall
is qw . At a distance y from the wall, the time-mean values of the tur-
bulently fluctuating density, stagnation temperature, and x-component
of the velocity are p , io , and ti, respectively. T h e velocity ti varies from
/
nth STREAMLINE
FIG. I .
- T,. rw.qw
Nomenclature for real flow.

HEATTRANSFER FLOWS
zero at the wall to the free-stream value U at distance S from the wall;
S is the velocity thickness of the boundary layer. T h e stagnation tem-
perature io varies from T , at the wall to the free-stream value To at dis-
tance d from the wall; d is the temperature thickness of the boundary layer.
There is a streamline of the flow, the nth streamline, which, for a finite
distance upstream and downstream of station x , lies just beyond S and A .
Thus, all boundary-layer effects are confined to a wall layer defined as
containing the flow between the nth streamline and the wall. At station z,
the nth streamline lies a distance 6', from the wall, and the gas flowing
at that point has density p , pressure p , Mach number M , static tempera-
ture T , and viscosity p. Although s,.' is greater than both S and d , the
separation of these three points is assumed small enough to make the
difference in free-stream properties between them negligible.
T h e fluxes of mass, momentum, and total enthalpy between the nth
streamline and the wall for the real flow of fig. 1 are m, (lb m/sec),
A$,, (ft Ib m/sec2), and f i r (Btuisec), respectively.
b. Potential Flow. Figure 2 presents the nomenclature required for
describing the potential-flow nozzle, in which the conditions at the nth
streamline (conditions U , p, and To)extend all the way to the wall. T h e
nth streamline in the potential-flow nozzle is at identically the same
location with respect to the nozzle axis as in the real-flow nozzle, but
the wall must, in general, be at a different distance S,' from the nth
streamline in order to satisfy the requirement that the mass flux "ip
between that streamline and the wall to remain equal to m, . T h e momen-
tum flux and enthalpy flux of mp in the potential-flow nozzle are
and fip , respectively. Under the assumption that boundary-layer effects
are confined to a small distance from the wall, relative to I , the wall
radius r, of the potential-ff ow nozzle is approximately equal to 1. Results
of the derivation of the integral momentum and energy equations for
cases in which the boundary-layer thicknesses S and d are not small
with respect to r are given in Appendix C.
2. Dejnitions
a . Deficiency Thicknesses. Under the assumption that S,' is small
compared with r, it is seen from Fig. 2 that the fluxes of mass, momen-
tum, and enthalpy (referenced to the wall temperature T,) in the poten-
tial-flow nozzle are
mp = 2wpU6,' (1)
hil, = 2nrpU%gl (2)
H,, ( -
= 2 ~ p U c , ,To T,)S,,' (3)
1111
D. R. BARTZ
2
ro.po, cp, y , Pr, CONSTANT
..
n t h STREAMLINE
FIG. 2. Nomenclature for adiabatic potential flow.
For 6,' also small compared with Y , it is seen from Fig. 1 that the fluxes
of mass, momentum, and enthalpy in the real-flow case are approximately
HEATTRANSFER FLOWS
Equations (4)-(6) are approximate in that a product of mean values

is not, in general, equal to the mean value of the product; the cross-
correlation terms must be considered. For example, the product pa in
Eq. (4)is not necessarily equal to the time-mean flow density pU which
would have to appear in Eq. (4)to make the equation exact. However,
it can be argued that the cross-correlation terms substantially cancel out
when the integration is performed over the boundary layer (16,p. 1090).
Since 8,' has been selected such that kT= k, , Eqs. (1) and (4)can
be equated to yield the following expression for the difference in the
wall positions between the nozzles:
6,' - 6', = 1;' (1 -

pu
--)PU dy (7)
T h e integral above is customarily defined as the displacement thick-

ness 6*. Thus, the physical significance of the displacement thickness
is that 8" is the distance the wall must be moved inward or outward for
adiabatic potential flow as compared with the position of the wall for a
realflow having the same massflux. That is, the physical definition of the
displacement thickness is
6* = 6', - 6', (8)
while the integral definition is

6* = J:' (1 - --)pu dy
(9)
PU
Because of the approximate nature of Eq. (7) (resulting from the

approximation in Eq. 4),Eqs. (8) and (9) do not define exactly the same
quantity. T h e question of which definition to adopt as fundamental
will be discussed later.
Subtracting Eq. ( 5 ) from Eq. (2) yields, with the aid of Eq. (7), the
deficiency of momentum flux in the real flow as compared with the poten-
tial flow:
$l, - $lr = 27rrpuz J

%# pu
-(1 -
u
u)dy
0 PU
T h e integral above is customarily defined as the momentum thickness 8.

Thus, the physical significance of the momentum thickness is that 8
is the thickness of potential flow which has a momentum flux equal to that
by which the momentum flux of the potential flow exceeds t h e momentum
l-131
D. R. BARTZ
jlux of the realflow for the same massflux. Hence, the physical definition
of the momentum thickness is
and the integral definition is

8 = r ' - - (pa1 - v ) d y a
0 PU
Subtracting Eq. (6) from Eq. (3) yields, with the aid of Eq. (7), the
deficiency of enthalpy flux in the real flow as compared with the potential
flow:
&- H , = 2i7rpUc,(T0 - T W ) 1 6,'
0
pa
-(1 -
PU
tb - Tw
To - Tw
) dY 113)
T h e integral above is customarily defined as the energy thickness 4.

Thus, the physical significance of the energy thickness is that 4 is the
thickness of potentialflow which has an enthalpy flux equal to that by which
the enthalpy flux of the potential flow exceeds the enthalpy flux of the real
$ow for the same massflux. Hence, the physical definition of the energy
thickness is
H D- H , = 2i7rpU,,(To - Tw)+ 114)
and the integral definition is
b. Coejicients. T h e skin-friction coefficient C, is defined as the ratio

of the wall shear stress to the dynamic pressure of the flow at the edge
of the boundary layers. Thus,
(16)
T h e stanton number C, is defined as the ratio of the wall heat flux

to the enthalpy flux of the flow at the edge of the boundary layers based
on the difference between adiabatic and actual wall temperature. Thus,
HEATTRANSFER FLOWS
T h e adiabatic wall temperature Taw is the wall temperature for zero

heat flux and is related to Mach number by
where R is the adiabatic recovery factor.
3. Assumptions
T h e following assumptions are made in the analysis:
(1) T h e flow is axisymmetric and steady without tangential compo-
nents of velocity.
(2) T h e boundary layer is confined to a distance from the wall which
is small compared with the distance from the axis of symmetry.
(3) T h e only forces acting on the gas are those due to pressure gra-
dients and to skin friction at the wall.
(4) T h e only changes in total enthalpy in the flow direction are those
due to heat flux through the wall.
( 5 ) T h e flow immediately outside the boundary layer is reversible
and adiabatic and parallel to the wall.
(6) Static pressure is constant through the boundary layer perpendi-
cular to the wall.
(7) T h e gas is perfect; however, the restriction that specific heats be
constant can be removed in computing the driving potential for heat
flux.
(8) T h e gas has a constant Prandtl number, a viscosity which varies as
a power of the temperature, and a constant adiabatic recovery factor.
(9) T h e skin-friction coefficient is the same as for constant-pressure
constant-wall-temperature flow on a flat plate at the same free-stream
conditions, wall temperature, and momentum thickness.
(10) T h e Stanton number is the same as for constant-pressure con-
stant-wall-temperature flow on a flat plate at the same free-stream
conditions, wall temperature, energy thickness, and momentum thick-
ness.
(1 I) T h e Stanton number for unequal momentum and energy thick-
nesses is that for equal thicknesses multiplied by (+lo)., where n is a
small interaction exponent.
~ 5 1
D. R. BARTZ
(12) Heat transfer affects the skin-friction coefficient in either one of

two ways: (a) there is no effect, and C, is the same as for adiabatic flow,
or (b) C, is the same as for adiabatic incompressible flow at a density
and viscosity evaluated at the arithmetic mean between the actual
wall temperature and the free-stream static temperature.
(13) T h e Stanton number for equal momentum and energy thicknesses
is related to the skin-friction coefficient by von KhrmPns form of
Reynolds analogy.
(14) Any chemical reactions in the boundary layer affect only the
driving potential for heat flux.
(15) T h e boundary-layer shape parameters OjS, d/S, and S*/8 are
those for 1/7-power profiles of velocity and of the difference between
stagnation and wall temperature. Such profiles are typical of turbulent
boundary layers on flat plates.
(16) Heat transfer by thermal radiation is negligible compared with
convection.
(17) There is no significant net mass transfer from wall to gas or gas
to wall.
Assumptions 1-4 define the situation to which the analysis applies.
Assumption 3 excludes, for example, magnetohydrodynamic forces, and
Assumption 4 excludes combustion effects (except for a possible direct
effect on heat flux as allowed by Assumption 14). Assumptions 1 and 2
have already been employed in defining S*, 8, and 4. Assumptions 5
and 6 are good approximations if the flow has no strong shocks. Assump-
tions 7 and 8 introduce little error for most gases.
Assumptions 9 and 10 are the ones which most affect the results, and
they are also the most uncertain. That skin friction and heat flux have
the flat-plate dependence on local conditions is certainly valid asymptoti-
cally for gradual nozzle contours (drjdz ---f 0 and dT,/dz --+ 0), but the
extent of departure in practical nozzles remains unexplored by experi-
ment except for the limited data of Ludwieg and Tillman ( 1 7 ) which
tend to suppost the present assumption. I n the absence of a correlation
of heat transfer obtained with large differences between momentum and
energy thicknesses, Assumption 1 1 was selected based upon intuitive
reasoning and was found to agree with a small sample of data to be
discussed.
Assumption 12 has two options, either of which can be selected.
Assumption 12a is based on recent experiments (18-24, which as dis-
cussed in Section 11, D, showed no measurable effect of heat transfer
on skin-friction coefficient or Stanton number for cooled boundary layers
[I61
HEATTRANSFER FLOWS
compared with values for adiabatic walls. Assumption 12b is the widely
used procedure of evaluating properties at some reference temperature.
T h e one selected being the arithmetic mean between the wall and the
static temperature at the edge of the boundary layer, gives only slightly
different values of C, from those determined by the reference-tempera-
ture method (22) out to Mach numbers of interest for most nozzle flows.
However, the background of data on which the film or reference
methods were established for the cooled turbulent case is sketchy, con-
sisting mainly of data with negligible temperature differences (i.e.,
T,/T,, only slightly less than unity) on average data over long pipe
lengths in which uncertain axial property variations clouded the picture
considerably, and data which generally scattered to the same extent as
the magnitude of the variable properties corrections. These data, as well
as newer data which conflict with the reference temperature method,
are discussed in Section 11, D. Thus, one is faced with the quandary of
having doubt thrown upon the accepted variable properties correction
by the new data and yet not being fully convinced that there should be
no correction on the basis of the limited data cited. For this reason, the
analysis presented here retains the option of either treatment of the
variable properties question.
Assumption 13 is well substantiated by flat-plate and pipe-flow
experiments, as will be shown later. Assumption 14 represents the com-
putationally convenient viewpoint that the effect of chemical recombina-
tion can be accounted for by employing a generalized recovery enthalpy
(rather than temperature) driving potential, leaving the heat-transfer
coefficient unaltered. Assumption 15agrees roughly with observed velocity
and temperature profiles on flat plates and wind-tunnel nozzles. T h e only
effect of Assumption 15 on the other parameters computed, however, is
through the ratio 6 * / O which is relativelyinsensitiveto theprofilesassumed,
and which, in turn, has only a secondary effect on momentum thickness
and skin friction, and little or no effect on energy thickness and heat Aux.
Assumptions 16 and 17 are, of course, only statements of the limit to
the scope of the analysis.
4. Integral Equations
T h e usual approach to the derivation of the integral momentum and

energy equations for a turbulent boundary layer (9, 16) is to start with
the boundary-layer differential equations and introduce an approximation
by eliminating the fluctuating cross-correlation terms through arguments
that they substantially cancel out when integrated across the boundary
~ 7 1
D. R. BARTZ
layer (26, p. 1090). Through this process, one arrives at integrals of

time-averaged variables such as
1'' (1 - --)pa dy
0 PU
which are then defined as exactly equal to new variables 6*, 8, and +.
In this case, the definitions given by Eqs. (9), (12), and (15) are con-
sidered fundamental. However, the variables 6*, 8, and + are then
related only approximately to the physical mass, momentum, and energy
defects, and the resulting momentum and energy equations become
approximations of uncertain accuracy when written in terms of these
integrally defined variables.
An alternate derivation of the integral momentum and energy equa-
tions, which will be presented here, adopts at the outset the physical
definitions of 6*, 8, and # given, respectively, by Eqs. (8), (1 I), and (14).
I t will be seen that this derivation leads directly to the integral momentum
and energy equations without further approximation and without con-
sideration of the internal structure of the boundary layer. T h e resulting
equations are identical in appearance with those derived from the
differential equations, differing only in the definitions associated with
6*, 8, and 4. However, the uncertainty in the integral expressions for
6*, 8, and 4 due to the turbulent fluctuation terms may still affect the
results, to a minor extent, through the use of the integral expressions in
the shape parameters introduced later, and in evaluating Reynolds
numbers in most of the available skin-friction data.
a. Momentum Equation. For the potential flow along the wall, mp
the streamwise gradient of momentum flux ap
is, by Assumption 3,
balanced only by the pressure gradient acting over the flow area
27cySP', where the latter, from the physical definition of the displacement
thickness, Eq. (8), is equal to 27rr(6,' - a*). Thus, employing the physical
definition of the momentum thickness from Eq. ( l l ) , the momentum-
flux gradient is
For the real wall flow mr, the streamwise gradient of momentum flux
a, is balanced by both the wall shear force and the pressure gradient
acting over the area 2 7 ~ 8 ~Thus,
'.
d M. ,
- = -2.rrrrW - 2 ~ ~ 6dP
-,'
dx dx
HEATTRANSFER
FROM RAPIDLY FLOWS
ACCELERATING
Subtracting Eq. (20) from Eq. (19), and noting that by Assumptions
5 and 7 dpldx = -pU dUldx, the following relation is obtained:
dxd (rpU28)= 17, - rpU6*

dU
-
dx
Equation (21) is the integral momentum equation for thin axisym-
metric boundary layers. I t can be put i n . a more convenient form by
differentiating, introducing the definition of the skin-friction coefficient,
Eq. (16), and rearranging to give
de -
- -
c, +----
dx 2 (22)
Under Assumptions 5 and 7, the expressions involving p and U can

be written in terms of the Mach number M , as follows:
U dx M ( l ++MY) dx
pU dx M ( l + q M 2 ) dx
Substituting these expressions into Eq. (22) and transforming the

independent variable to x by noting that dxjdx = [I (d~/dx)~]~/~, +
the final form of the integral momentum equation is obtained:
d0 C dr f 2 - M 2 +(8*/8)
-d=z - L [ l2+ ( = ) ]
-"[ M(1 + TY - M
1 Z )
b. Energy Equation. For the wall flow mp without heat transfer, the
enthalpy flux, by Assumption 4, remains constant. Thus, employing
the physical definition of the energy thickness from Eq. (14), the stream-
wise gradient of the enthalpy flux flp is
d
+
-& [fir 2mp Uc,(To - T&1 =0 (26)
For the real wall flow f i r , the streamwise gradient of the enthalpy
flux f i r is exactly equal to minus the rate at which energy is transferred
to the wall. Thus,
d .
- H,. = -2mqw (27)
dx
1191
D. R. BARTZ
Subtracting Eq. (27) from Eq. (26) yields
Equation (28) is the integral energy equation for thin axisymmetric

boundary layers. It can be put in a more convenient form by differenti-
ating, introducing the definition of the Stanton number (Eq. 17), and
rearranging to give
Substituting Eqs. (23) and (24) and transforming the independent

variable to z yields the final form of the integral energy equation
Note that the effect of variable surface temperature is accounted for in

the development of the energy thickness boundary layer. This is the
general relationship upon which the usual variable surface-temperature
correction to heat transfer on a flat plate is based.
5 . Skin-Friction Coeficient
a. Diabatic Skin-Friction Coeficient. I n accordance with Assump-
tion 9, the skin-friction coefficient in a nozzle is taken to be the same as
that on a flat plate at the same conditions at the edge of the boundary
layers p, U , p, T o ,and M , the same wall temperature T , , and the same
momentum thickness 8. Unfortunately, even this drastic assumption
does not permit a completely reliable evaluation of C , , since only the
adiabatic skin-friction coefficient C l o , obtained when T, = T a w , is
known accurately. T h e relationship between C, and C,, for severely
cooled turbulent boundary layers, when gas properties vary greatly
between the free stream and the wall, is sufficiently uncertain that both
relationships discussed earlier are included in the analysis as alternatives.
T h e first relationship, Assumption 12a, is that of computing the value
of C,by assuming it to be exactly equal to that for an adiabatic wall, i.e.,
HEATTRANSFER
FROM RAPIDLY FLOWS
ACCELERATING
T h e correlation of adiabatic turbulent boundary-layer skin-friction

coefficients developed by Coles (15) was found to correlate accurately
the trends and magnitudes of nearly all of the reliably measured data
from high-speed flow over flat plates thus far reported in the literature.
In particular, it fits the data at high Reynolds and Mach numbers better
than previously used correlations. Consequently, in updating the analysis
of ref. (9), the Coles correlation (the details of which are presented in
Appendix A) was adopted by Elliott et al. (14) and was used in obtaining
the new boundary-layer calculation results presented in this chapter.
Unfortunately, since the Coles correlation is not explicit in the momen-
tum thickness Reynolds number Ro , the important functional relation-
ships of the analysis become difficult to recognize. Therefore, in order
to illustrate the essential features of the problem in the clearest manner,
the Blasius equation, coupled with variable properties corrections, will
be used in the remainder of this Section. (Comparing the Blasius equa-
tion with the tabulated values of low-speed skin-friction coefficients c,
used in the Coles correlation [Table A.1 and Fig. 31, the maximum
deviation is 5 yo between Ro of 400 and Ro of 15,000.) T h e Blasius equa-
tion, expressed in terms of R, , is
0.0256
Cf = __
(R*)t
where cfis the low-speed value of skin friction, Ro equals pUO/p, and
p and p are properties evaluated at the local static temperature T a t the
edge of the boundary layers. T h e relation between Cf,and cfin the
Coles correlation, derived in Appendix A, is
and m is the exponent of the viscosity relationship adopted, p

For our purposes, Eq. (33) can be approximated by
-
where T , is a sublayer temperature specified in Eq. (A.3) of Appendix A,
T".
with less than about 10% error over the range of R Band M of interest
for m = 0.6. Combining Eqs. (31), (32), and (34),
D. R. BARTZ
0.010
a009
0.008
0.007
0.006
0.005
0.004
IZ 0.003
0.002
aooi
I00 10' 10: 10
F, Rg
FIG.3. Adiabatic skin-friction coefficient for low-speed flow.
T h e second relationship, Assumption 12b, is that of accounting for

the effect that property variation may have on skin friction, either as a
result of compressibility, or of Tw # T a w ,or both, by evaluating the
properties p and p at a temperature which is the arithmetic mean between
T and T , . This is the same relationship employed in ref. (9) and
elsewhere. T h e reasoning behind such a correction and a discussion
of its validity are presented in Section 11, D.
C, 1
Or, again making use of Eq. (32) and assuming m = 0.6,

-0.6
(37)
6. Stanton Number
For flow with substantial pressure gradients, it is no longer possible
to derive an appropriate analogy between the Stanton number C ,
P I
HEATTRANSFER
FROM RAPIDLY FLOWS
ACCELERATING
and C,/2 in a straightforward manner because the similarity between

the momentum and energy equations is destroyed by the presence of
the pressure-gradient terms. Nevertheless, in order to proceed, it was
found necessary to adopt some form of Reynolds analogy, modified
according to arguments in this section. By Assumption 10, the Stanton
number C , in a nozzIe is taken to be the same as that on a flat plate at
the same free-stream conditions p, U , p, To , M , the same wall tempera-
ture T, , and the same local energy and momentum thicknesses and +
e.
T h e most appropriate Pran.dt1-number correction to the Reynolds
analogy is believed to be the von KBrmBn form (23, p. 225), which was
derived by consideration of the respective thermal resistances in laminar
sublayer, a buffer layer, and a turbulent outer region.
A Prandtl-number correction of this form has been utilized in all of the

new boundary-layer calculations from which results are presented in
this chapter. Again, however, as in previous section, the purposes of
illustration of the most significant relationships of the problem can be
better served by replacing Eq. (38) with the simpler but less widely
valid Colburn form of Reynolds analogy:
It is important to note at this point that relationships such as Eqs. (38)

and (39) have been established by experimental correlation and analysis
of Aows in which the ratio of energy thickness to momentum thickness,
410, is essentially constant in the streamwise direction at a value close to
unity, being dependent only upon a small fractional power of the Prandtl
number. I n the nozzle flow situation, analyses of the type presented by
Sibulkin (8)and Bartz (9) have predicted that the ratio +/e may increase to
values as high as 5 in the throat region because of the differences in the
integral momentum and energy equations resulting from the combined
presence of the term
-e -
[ M (1
1
+
+ @*P)
M2) "1
D. R. BARTZ
in the momentum equation and the term
in the energy equation. Thus, it was apparent in the analyses (8, 9)

that some account must be taken of 4/!3 # 1. In those references, the
Reynolds analogy was modified by multiplication by a factor (4/6)-117,
which is fairly close to (4/!3)-l17for most conditions of interest. T h e
exponent was arrived at by intuitive arguments linked to the power-
relation velocity and stagnation temperature distributions assumed to
exist in the boundary layer. Again following such reasoning in this
development but broadening it slightly by allowing the exponent to be
a parameter to be chosen, the result is
Substituting into Eq. (40) the expression for C,/2 from Eqs. (35) and (37),
which can be made equivalent in form by expression in terms of a
temperature Tr e f, [where in Eq. (35) Tref= T,, , and in Eq. (37)
Tref= ( T , + T ) / 2 ] ,the result is
or
Noting that C, increases sharply as the downstream distance approaches

zero from the start of a thermal boundary layer (i.e., location of start of
heating or cooling downstream of the flow origin), it might be argued
that the upper limit for n must be a. On the other hand, it might be
argued intuitively that n should probably not fall below zero since a
growing momentum thickness would be predicted to increase C,, con-
trary to the usual behavior. (The value of n corresponding to the (4/6)-117
factor of refs. (8)and ( 9 )is 3/28, or about 0.1 .)
One might consider rearranging Eq. (42) to
This form suggests that a Reynolds-number-like parameter R, = p U $ / p

is perhaps equally relevant to the correlation of C, as Ro is to the correla-
~ 4 1
HEATTRANSFER
FROM RAPIDLY FLOWS
ACCELERATING
tion of C, . Such a suggestion has been made previously by Seban and

Chan (24) and Kutateladze and Leontev (25) and used to advantage in
their analyses. I n addition, an equation of the form of Eq. (43), derived
in Appendix A (Eq. (A.8)) using the Coles and von Khrmhn relations,
has been utilized to correlate one set of data for a flat-plate flow (19)
and another for tube-entrance flow (20).T h e RE values with which the
data of Reynolds et al. (19) were correlated were multiplied by +/x,
where +/x was computed from the low-speed, constant-surface-tempera-
ture, flat-plate energy equation, d+/dx = C,. Since these data were
obtained with TWITobetween 1 and 1.05, i.e., with a low temperature
difference across the boundary layer, the variable properties factor
differed only negligibly from unity irrespective of the choice of T r e l .
Wolf (20) presented values of the heat flux in successive separately
cooled sections downstream of the abrupt start of cooling of a
fully developed adiabatic pipe flow. These values have been used to
obtain 4 and the values of differentiated to determine local C,. I n
#J
these tests, with both air and COz , severe wall cooling was employed,
yielding values of TWITobetween 0.3 and 0.7. In computing R , , p and
p were evaluated at the local free-stream temperature, which is equivalent
to assuming TTef= Taw for this low-speed flow. T h e data from these
two references plotted as CIL(+/O)-o.l versus R, are compared with
Eq. (A.8) in Fig. 4 and found to agree quite satisfactorily. Such agree-
ment between the da.ta of Reynolds et al. (19) and Eq. (A.8) should,
of course, be expected, since both TWIToand $18 are near unity. It is
significant, however, that most of the data of Wolf (20) also agree with
the equation to within * l o % even though they were obtained with
a wide range of values of both Tw/T,, and +/8. A limited sample of data
such as this cannot, of course, be used to make a sensitive determination
of the most appropriate value of n, the interaction parameter. Conse-
quently, a value of n of 0.1 was arbitrarily selected and found to improve
the correlation somewhat compared with no correction at all (i.e.. n of 0).
T h e data of Wolf (20) also suggest that values of C, are insensitive to
variations of TWITofor cooling since, by taking T r e f= Taw= T , no
correction for property variation has been made and no systematic
deviation of the data with TWITois noted. Agreement with this trend was
also exhibited by data obtained from several other experimental in-
vestigations discussed in Section 11, D.
7 . Solution of Integral Equations

With the specification of C, by Eq. (43) (or its more accurate counter-
part, Eq. (A.8)) and C, by Eqs. (35) or (37) (or their more accurate
P51
D. R. BARTZ
P u9 / P
FIG. 4. Comparison of modified Stanton number correlation (Eq. (A-8)) with data
from low-speed flow.
counterparts, Eqs. (33), (A.4), (AS), and Table A.l), only the local
Mach number at the edge of the boundary layer and the boundary-layer
shape parameter 6*/0 need be determined in order to proceed with the
solution of the integral momentum and integral energy equations for 0
and 4, respectively. T h e local M distribution is, of course, a
function of the nozzle configuration and may be taken as that for
one-dimensional reversible adiabatic flow, or that for two-dimensional
flows resulting from method-of-characteristics solutions where
necessary (see Section 111, D). I n order to compute the value
of S*/0 as well as some of the auxiliary shape parameters 0/S, $/A,
and A/S to which S*/d is related, it is necessary to specify some
velocity and stagnation-temperature distribution through the boundary
layer. For this purpose, +-power distributions of zi in terms of y/S and
to - T, in terms of y / A have been adopted. These distributions obvi-
ously give grossly inaccurate values of zi and (to- T,) and their deriva-
tives very near the wall. Fortunately, these distributions are utilized only
in integration across the boundary layer in computing the shape para-
meters. Hence, the errors near the wall have negligible over-all effect.
[261
HEATTRANSFER
FLOWS
I n fact, the shape parameters are quite insensitive to the arbitrary

4
specified exponent over a range from about to but are quite sensitive
to the local value of M and T,/T, for which the value of the shape
parameter is being computed. Thus, the use of low-speed adiabatic
values of the shape parameters rather than the correct values obtained
from distributions and integrals as presented in Appendix B can lead
to significant errors near the nozzle throat. It is evident from the shape-
parameter integrals that the shape parameters depend upon the ratio
4/0 and that, consequently, the integral momentum and energy equa-
tions (25) and (30) must be solved iteratively, except when the equations
have been uncoupled by assuming n = 0, and the expression for ch
given by Eq. (43) is utilized. Hence, except for this special case, it
becomes almost essential to program the equations for a digital com-
puter since no simple analytic solution can be obtained without con-
siderable compromise. Such a program is described in detail in Elliott
et al. (14). I n ref. (9) (written before the availability of a high-speed
digital computer), the solution of the integral momentum and energy
equations was not obtained by iteration but was terminated with a
first approximation, which was made by assuming a reasonable value
of 6 / A to be a constant 1.O only for the purpose of evaluating the bound-
ary-layer shape parameters. Although it was argued at the time that such
an assumption would have little effect, it was found that it does have a
large effect on 6*/0 and even a signification effect on hg , as will be demon-
<
strated. I n particular, in the throat region of a highly cooled (TWITo 1)
nozzle of typical rocket-nozzle shape, +/O, and hence A / 6 , may approach
values as high as 5 and, as a consequence, 6*/0 becomes negative. This
was predicted and readily explained by Reshotko in his discussion of
ref. (9). Because of the extreme cooling, the temperature falls faster
than does the velocity across the boundary layer. T h e result is that the
local mass flux per unit area in the cooled boundary layer exceeds
that in the free stream, giving rise to the negative displacement
thickness.
Finally, the local heat flux gw is computed from the defining equation
for c h , (Eq. (17)), where ch has been determined as a function of 0
and 4 from an equation such as Eq. (43) (or its more accurate counter-
part, Eq. (A.8)). T o make this calculation, choices must be exercised as
to (a) the method of properties evaluation-either p and p at film tem-
perature or C, for an adiabatic wall by way of the Coles correlation; and
(b) the value of n, the interaction parameter, between 0 and 4.
For the special case in which n is assumed to be equal to zero and
Eq. (43) is utilized to specify the functional relationship for c h , the
integral energy equation can be analytically solved separately from the
1271
D. R. BARTZ
integral momentum equation. T h e form of Eq. (30) under these condi-

tions is
where
where it has been assumed pUrZ = constant, and
T h e solution of Eq. (44) is
(45)
It must be remembered, however, that Eq. (45) is based upon three
approximations: (1) the Blasius low-speed skin-friction coefficient;
(2) the Colburn form of Reynolds analogy for Prandtl-number correction,
and (3) the approximation that ( Tu,/T)0.6M (Tuw/T)( Ts/T,,)o.6 for
rn = 0.6 when basing the calculation on C, for adiabatic wall. For a
film-temperature properties calculation, only the first two approxima-
tions pertain. T h e degree of approximation compared with what are be-
lieved to be the most valid values of g ( ~i.e.,
) , Eq. (A.8), varies with both
Rdand M. Under an extreme set of conditions, M = 4, Rd = Re m lo4,the
deviation of the value Q(z) resulting from these approximations is about
+35%. At lower M and Rd , the deviation is considerably less. If, how-
ever, Q(z)were taken to be 35 % high throughout the nozzle, this would
have about a - 6% effect upon C, through the value of +& in Eq. (A.8)
for the case when c $ ~ is zero.
Another interesting limiting case occurs when ( + o / ~ o ) is large with
respect to a parameter specified below. This parameter can be derived
by evaluating the integral in Eq. (45)for the special case of flow in a
contracting nozzle of average contraction angle 31, a constant wall
temperature T , , low speed flow such that (Taw- T , ) M ( T o - Tw)
and Tref M T,, M T (i.e., out to low supersonic Mach numbers).
P I
HEATTRANSFER FLOWS
Under these restrictions, the n =0 solution of the energy equation,

Eq. (43,assumes the form
It is clear that when (40/~o) is sufficiently large (the magnitude primarily

depending upon nozzle design and RD*),that the second term in the
brackets becomes small with respect to I , and 4 becomes proportional
to Y . (In the event that the second term is as large as 4,the neglect of
the term induces an error of - 38 yo in 4 and an error of +8 yo in C, .)
When 4 is proportional to Y , and is substituted for in Eq. (43) by its
equivalent ( c $ ~ / Y ~ ) Ythe result is
0 .O 1 28( Trei/ T)-Oa6 - 0.0152( 7'ref/ T)-o.6
c,,= - -
Pr2/9(pU ~ / p ) f ( + ~ / r ~ ) 'Pr2/3RDt(q5,/ro)+
(45b)
In the case of interest, that of flow approaching the nozzle through a

constant area duct of diameter 2r0 and length 1, the value of ( + o / ~ o ) *
from Eqs. (29) and (43) is given by (0.44) (Pr)-2~15(RD)-1~20(l/~o)1~5(~o/
which when substituted into Eq. (45b) yields
By mere algebraic rearrangement Eq. (45c) can be expressed as
which bears a close resemblance to the closed-form approximation

described in the following section.
8. Sample Results
I n order to demonstrate the qualitative behavior of the boundary
layer and heat transfer in a typical small nozzle operating under typical
rocket conditions, a sample calculation has been made for the nozzle
depicted in Fig. 5 for the conditions listed. T h e results obtained by
selecting the same nozzle configuration and operating conditions of a
similar calculation (9) also show the effect of making an "iterative
~ 9 1
D. R. BARTZ
c = 1.80
FLOW - --
r 8 2.50
-I- r 1.85
FIG. 5. Nozzle contour and flow conditions for sample calculation. All dimensions in
inches. Flow conditions:
p o = 300 psia ctr = 0.567 Btu/lb OR y = 1.2
To = 4500R Pr = 0.83 p N T0.66
po = 1.3 x lb sec/ft2
simultaneous solution of the integral momentum and energy equations

(as outlined in Section 11, A, 1-7) rather than the first-approximation
solution as in ref. (9). T h e same three pairs of inlet boundary layer
thicknesses 6 and d have been selected. I n Fig. 6, the values of 6 pre-
sented show that a thick inlet boundary layer shrinks rapidly as a result
of the subsonic acceleration, reaching a minimum just ahead of the
throat, whereas a thin inlet boundary layer initially grows rapidly and
then, too, shrinks to about the same value as the flow approaches the
throat. It is also evident that the iterative simultaneous solution predicts
considerably thicker boundary layers throughout the nozzle than first-
approximation solution, although the qualitative behavior is essentially
unchanged. I n Fig. 7, a similar comparison of values of d is presented.
Here, the changes between the first-approximation and the iterative
simultaneous solution are somewhat smaller, as one might suspect,
since the most significant change that occurs is the value of 6*/8 becoming
negative in the momentum equation. T h e behavior of 6* as determined
by the iterative simultaneous solution is illustrated in Fig. 8. As stated
and explained in Section 11, A, 7, the values of 6* are negative over most
of the length of the nozzle. Finally, the local h, distributions for the
three assumed entrance conditions are presented in Fig. 9 and compared
~301
HEATTRANSFER FLOWS
.-6
Mi
rn
rn
w
z
0.20
r
+
a 0.16
w
ziJ0.12
t
a
a
0 0.08
z
3
0
m ao4
A X I A L DISTANCE RATIO Z/L
FIG. 6. Velocity boundary-layer thicknesses for nozzle and conditions of Fig. 5.
FIG. 7. Temperature boundary-layer thicknesses for nozzle and conditions of Fig. 5.

1311
D. R. BARTZ
.-c
I-
oo
AXIAL DISTANCE RATIO z/L
FIG. 8. Displacement thickness for nozzle and conditions of Fig. 5.
with those ,of the first-approximation solution. The thicker boundary

layers 6 determined for the iterative simultaneous solution effect approxi-
mately a 15% reduction in the maximum values of h, . It should be
noted that in order to permit comparison with the h, distribution ( 9 ) ,
the film-temperature property evaluation option (Assumption 12b) was
utilized. It is seen that the initial value of 6 or 6 has little effect on the
h, distribution over most of the nozzle when the initial temperature
boundary-layer thickness is small.
B. CLOSED-FORM
APPROXIMATION
From results such as those discussed in Section A, it was evident
that the variation of local mass flow rate per unit area at the edge of the
boundary layer is still the dominant variable affecting the heat transfer
distribution even in an accelerating flow. This suggested that by selecting
some linear dimensional variable that varied in even a rough approxima-
tion to the variation of the boundary layer, a closed-form Nusselt-
~321
HEATTRANSFER FLOWS
FIG. 9. Heat-transfer coefficient for nozzle and conditions of Fig. 5.
Reynolds type correlation equation could be found that might approx-

imate the results of the boundary-layer analysis reasonably well. Such an
equation was developed (13) by selecting the local diameter at the station
of the nozzle of interest as the linear dimension. This became an obvious
choice when boundary-layer distribution results such as those shown
in Fig. 6 where obtained. Only near the nozzle entrance for an initially
thin boundary layer did this selection appear to be qualitatively in-
appropriate. I t was later shown, as described in Section 11, A, 7, that
the exact proportionality of the energy thickness 4 to the local
diameter is a direct result of a n = 0 (see Eqs. 4-43) solution of the
energy equation for constant wall temperature and the limiting con-
ditions of thick entering thermal boundary layer and/or high Reynolds
number.
[331
D. R. BARTZ
Starting with the assumption that the local heat-transfer coefficient

is principally dependent on local mass flow rate per unit area,
T h e exponent as established by the results of the boundary-layer

solutions is 0.8 because of the direct influence of the skin-friction
correlation adopted. If Eq. (46) is nondimensionalized and is multiplied
by a function of Pr as suggested by Eq. (38), one obtains the familiar
Nusselt-Reynolds type equation
Nu = C(RD)o.ePro.a (47)
where Reynolds number is based on the local diameter D, which was

assumed to be the characterizing linear dimension. For a range of
Prandtl numbers near unity it can be shown that the von Khrmin-
Prandtl correction (Eq. (38)) can be reasonably approximated by PrO.4
for low values of RB. Thus, the closed-form approximation is complete
except for an arbitrary constant C. For the purposes of this equation
the value of C might be selected so as to force exact agreement of the
closed-form equation and results of a boundary-layer solution at one
particular point in the nozzle for a particular set of operating conditions.
C was evaluated to be 0.026 (13)by forcing agreement at the throat with
first-approximation results for the conditions listed and the nozzle
configuration of Fig. 5. For other nozzle configurations and conditions,
this constant would vary but not very drastically, especially if a factor
(D*/tc)O.l,as suggested by nozzle similarity considerations (9), were
multiplied into the equation.
T h e possible influence of variable properties considerations are
readily demonstrated by the closed-form approximation. If it is assumed
that both cp and Pr are constant over the boundary layer (as they are
to a close approximation for a wide temperature interval), the properties
whose variations must be accounted for are only p and p. If these are
evaluated at some reference temperature to be specified then Eq. (47) can
be used to determine h, , as follows:
If, for the moment, the reference condition is assumed to be at tempera-

ture halfway between the wall temperature and free-stream static tem-
P41
HEATTRANSFER FLOWS
perature, the variable properties factor can readily be expressed in terms

of TWIToand M :
where it has been assumed that p N Tm.(Arguments about the appro-

priateness of variable properties corrections in general, and this reference
condition in particular, for severely cooled boundary layers are presented
in Section 11, D.) Finally, by assuming that the local mass flux is related
to that at the throat by the local area ratio (ix., one-dimensional flow),
the convenient form of the equation for h, can be obtained
where the throat mass flux per unit area has been related to the rocket
performance parameters characteristic velocity c * and chamber pressure
Po *
T h e success of this equation in fitting results from turbulent boundary-
layer calculations over the whole nozzle when agreement is forced at
the throat by selection of the constant C is illustrated in Fig. 10. It is
evident that the agreement can be made to be excellent except near the
entrance for a case in which the entrance boundary layer is thin (see
Fig. 9). T h e weakness of the closed-form equation, of course, is in the
uncertainty of the value of the constant C for the flow of interest, since
it is sensitive to the inlet boundary-layer conditions to the extent of
about 10% at the throat. T h e value of C as determined from the iterative
simultaneous solution of the turbulent boundary-layer equations for
typically thick inlet boundary-layer conditions (i.e., case 2a of Fig. 9)
is 0.0225, as compared with the value of 0.026 obtained from the first-
approximation calculation and used in ref. (1.3). T h e worth of the
closed-form equation (Eq. (50)) lies in its simplicity, permitting the
determination of the approximate h, distribution by a rapid slide-rule
calculation. An approximate evaluation of the influence of inlet boundary-
layer thickness on the heat transfer coefficients can be made by noting
the extra factors (l/r0)1/5and ( Y , J Y ) ~ / ~ O appearing in Eq. (45d) but absent
from Eq. . .(47).
. The same objective of slide rule simplicity can be accom-
P I
D. R. BARTZ
t6
w
L
A X I A L DISTANCE RATIO z/L
FIG. 10. Heat-transfer coefficient for nozzle and conditions of Fig. 5 ; closed-form
approximation comparison.
plished with less uncertainty due to entrance conditions by making use

of Eq. (45d) to achieve an expression comparable to Eq. (50)
where u in Eq. (45d) was taken to be (Tref/T)-0.6 which is slightly

different than Eq. (49). T h e reason for the difference is discussed in
Section 11, D. T h e limitations necessary to arrive at Eq. (45d) must of
course be observed in applying Eq. (50a), i.e., +o/io or lir, large in the
sense of the criteria of Eq. (45a).
C. TRANSPORT
PROPERTIES
It is evident from the analyses of Section B that the problem under
discussion involves all three of the transport properties usually found
in heat-transfer problems, p, A, and Pr. If it is accepted that the closed-
~361
HEATTRANSFER
FROM RAPIDLY FLOWS
ACCELERATING
form approximation is a reasonable over-all approximation to the results

of the detailed boundary-layer analyses, then some idea of the impact
of the transport properties on the problem can readily be obtained.
In Eq. (48),it is seen that the transport properties have been combined
in such a way as to eliminate the direct use of A, in the interests of
convenience, leaving only the ratio of p0.2/Pr0.6.Since in a theoretical
calculation of high temperature viscosity and thermal conductivity, the
same force constants and functional relationships are utilized for both
p and A, it makes sense to cancel these factors out to the maximum
extent possible in order to reduce the influence of uncertainties in the
values. T h e final result, po.2/Pro.6,is a parameter which is readily
determined with fairly low uncertainty because of the low power ex-
ponent on viscosity compared with its range of uncertainty and the
asymptotically constant characteristic of Pr at high temperatures.
A search of the literature for reliable, consistent values of the viscosity
and Prandtl number (or thermal conductivity) reveals only a negligible
amount of such data for a limited number of species above about 2000R
for viscosity and 1000R for thermal conductivity and Prandtl number.
T h e reason for this lack of information is that such measurements are
very difficult to make. Surveys of the availability of data on high-tempera-
ture viscosity and Prandtl number as of 1958 and 1960 are reported in
refs. (26) and (27), respectively. Undoubtedly the most complete com-
pilation of experimental transport properties for gases is contained in
the NBS tables (28). As evidenced by the total absence of new experi-
mental values of high-temperature transport properties of gases in
the Thermodynamic and Transport Properties Symposia of 1959
(29) and 1962 (30), the situation does not appear to have changed
appreciably.
Thus, continuing to face this dearth of experimental data at high
temperatures, one is forced to turn to theoretical calculations. For the
most part, such theoretical calculations are essentially extrapolation
formulas derived from statistical mechanics models and based on two
empirically determined constants. These constants, usually related to a
collision diameter and to an attraction energy, are generally derived
from room-temperature measurements of one of the transport proper-
ties. T h e results of an ambitious set of calculations such as these for
200 gases, covering the range 100-500OoK, are presented by Svehla (31).
Andrussow (32) reports simple correlations of the results of detailed
statistical mechanical calculations of the transport properties of 34 gases
(many of which were polar). He correlates the results by use of a series
expansion of the exponent of the temperature dependence of the trans-
port properties. In addition, he proposes some rather simple means for
[371
D. R. BARTZ
predicting the transport properties of mixtures of gases. Brokaw (33)

also presents a simple method for mixture calculations which utilizes
alignment charts. Both of these techniques, as well as the earlier proposed
technique of Buddenberg and Wilke (34),avoid the complexities required
by the statistical mechanical methods of Hirschfelder et al. (35). Unfor-
tunately, there are very few experimental data at high temperatures that
would test the validity of any of these methods. A moderate tempera-
ture test of the method of Buddenberg and Wilke (34) indicated agree-
ment within 10% for nitrogen-stream mixtures to 1200K (36).
It was shown (26) that, consistent with the level of potential error
introduced by approximations of Eq. (48),it is possible to make very
simple, and hence rapid, approximations to the values of Pr and p.
T h e approximation to the value of Pr results from the use of the same
collision integrals for both viscosity and thermal conductivity. T h e
equation for Pr on this basis is
Pr = Y
1.94~- 0.74
where the ratio of mean-free paths for diffusion and viscosity was taken
to be 1.2 for a smooth sphere model as opposed to 1.0, which is the Euken
approximation used in ref. (13). T h e higher value of the ratio was
found to agree better with available data (26). Svehla has presented a
similar equation with slightly different constants:
Pr = Y
1.77~- 0.45
T h e latter agrees with the NBS air data (28)somewhat better in the range
500-2000"R, whereas from 2000 to 4000"R they both differ by about
2 yo from the Chapman-Cowling high-temperature prediction; i.e.,
Pr = 0.715.
Figure 2 of ref. (27) presented a plot of the one-fifth power of viscosity
versus temperature of a large number of gases frequently present in
rocket exhaust nozzles. Over the temperature range from loo0 to
8000"R, the maximum deviation of p1I5 of any of these gases (except
hydrogen) from that of air was only 8q4. T h e value for hydrogen was
consistently about 15% lower than that for air. Thus, it was concluded
that the value of p1I5of air could be taken as a reasonably accurate value
for most combustion gas mixtures for use in calculations based on the
closed-form approximation. For that purpose, a power-relation viscosity
equation for air based on the NBS data was presented (13):
p = (46.6 x lb/in. sec
10-10)(,lt)~(TT"R)o.6
[381
HEATTRANSFER FLOWS
T h e calculated high-temperature values for air (32) are best fitted by

p = (33.8 x 10-10)(A)t(T0R)0.66 lb/in. sec
over the range from 1000 to 9000R.At most, the two equations differ
by only 10% at the high-temperature end. Since there are no experi-
mental data above 2000R, the choice between these equations is arbi-
trary.
D. VARIABLE
PROPERTIES
By virtue of the fact that most rocket thrust-chamber walls are neces-
sarily cooled severely with respect to the free-stream temperature and
there is a region of supersonic flow (both conditions causing variations
in the local static temperatures in the boundary layer), the question
immediately arises as to how to adapt constant properties (i.e,, T, T)
correlations of skin friction and heat transfer to such problems. One
method that has gained wide acceptance due principally to reports by
Eckert (22) is the use of a reference temperature, at which the properties
are evaluated in the constant properties correlations in order to adapt
to calculations of heat transfer and skin friction for a boundary layer with
large temperature variation. This procedure is widely accepted for
adjusting for both the property variation due to cooling or heating and
that due to compressibility. A mathematically equivalent procedure is to
compute the constant properties T~ or gWvalues by using the free-stream
temperature properties and then to multiply the Q, or (fw by a function
of the ratio of properties at the reference temperature to those at the
free-stream static temperature.
I n most cases, both c,, and Pr have been justifiably assumed not to vary
significantly with the temperature variation across the boundary layer,
giving the convenient result
For a laminar boundary layer, it can be shown that

D. R. BARTZ
whereas, for a turbulent boundary layer, skin friction or Stanton number

correlation based on boundary layer thickness, the corresponding
expression is
or the alternative relationship suggested by Coles (Eq. (33)). In ref. (9),

u was modified slightly to permit viscosity evaluation at the stagnation
temperature at the edge of the boundary layer and recognition of the
dependence of 6 on (pU/p)-l/5, which alters the exponents of Eq. (57) to
Eckert has shown (22) that the most satisfactorily definition of the
reference state, accounting for both heating and cooling and compress-
ibility effects at high speed, is
T* =0 4T + T,) + 0.22 Pri( To - T ) (59)

which, for low-speed flow, reduces to
*
T l o w speed == TO, = 0.5(T + T,) 160)
This is also sometimes referred to as the film temperature. Although

Eq. (60) was adopted initially (9, 13) in order to be consistent with
skin-friction and heat-tansfer correlations then in use, its continued use
is probably questionable in view of the analytical and experimental
results that have since led to widespread use of Eq. (59). Interestingly,
however, it will be shown in Section E that a variable properties
correction based upon the arithmetic mean temperature (Eq. 60) seems
to show the best agreement with one set of data at high Mach numbers
for heated-air experiments. Although this result may not be general and
the differences are not large, the use of Eq. (60) has been retained in
calculations made here in order to maintain some degree of consistency
with earlier published calculations. At a Mach number of 4 (T,/T,, of i),
the highest M for which data are available, the value of the variable
properties correction factor based upon T,,,, is 20/, above that based
on T*.
While the T* or T,,,, treatment of the variable properties problem has
gained wide acceptance, it is worthwhile to re-examine the premises,
data, and theoretical solutions upon which it is based. As far as the
author can determine, the reference-temperature concept was initiated
[401
HEATTRANSFER
FROM RAPIDLY FLOWS
ACCELERATING
by Rubesin and Johnson (37). They found that all of the C, results
of Crocco (38) (numerically solved from the laminar boundary-layer
equations for a flat plate for Pr = 0.725, M = 0-5, T w / T = 2-2,
viscosity-temperature exponents m = 0.5, 0.75, 1.00, 1.25) could be
uniformly reduced to values within 1 % of the constant properties value
(c,dK = 0.664) by using a factor u (as in Eq. (56)) and a reference
temperature T having constants only slightly different from those of
Eq. (59). Significantly, the maximum value of u (corresponding to
M = 0, TWIT= 2, m = 0.75) was only 1.07, whereas the corresponding
value of 5 for a turbulent boundary layer would be 1.357, indicating the
greater sensitivity of the turbulent boundary layer to such a correction.
Young and Janssen (39) later solved the flat-plate boundary-layer equa-
tions numerically out to higher Mach numbers and with cooler walls,
using a Sutherland viscosity law rather than a simple power relation.
By use of the reference temperature T* (Eq. (59)), Eckert was able to
correlate all of the Young and Janssen C, data to the constant properties
value to within 2.6%. When he tried the same thing on the C, data,
deviations running from - 14 yo to $9 yo were encountered. Later
laminar boundary-layer calculation results by Van Driest (44,in which
an enthalpy driving potential was utilized in the C , definition, were
found by Eckert to be correlatable with the constant properties value
within a few percent using Eqs. (56) and (59). When the Young and
Janssen C, results were recorrelated on this basis, all were converged
to the constant properties value within 2.6 yo except for one set of results
calculated for T of 100R.For these data, the correlation was made poorer
by the T* correction, differing by from 14 to 20%. T h e difficulty was
attributed to peculiar transport properties variations at the low tempera-
ture.
T h e way in which the T,,, variable properties correction was carried
over to the turbulent boundary-layer case is of interest. In the concluding
remarks of Young and Janssen (39), the statement is made: More
important, there are indications that this procedure ( T properties
evaluation) can be applied approximately to the turbulent layer. That is,
use the incompressible relations for the turbulent heat-transfer and
skin-friction coefficients and the following relation for T:
T
-= 1
T
+ 0.032M2 + 0.58 (-T-T w - 1) for M < 5.6 (61)
T h e indications referred to are not discussed in the paper nor in any

of the references of the paper. Eckert (22, 41) recommended the refer-
ence-temperature variable properties correction for turbulent boundary
layers based on the Young and Janssen statement, some limited skin-
[411
D. R. BARTZ
friction data obtained with a cooled wall, and reference to the NACA
investigation of turbulent air flowing in tubes (42),in which some success
in correlating the heat-transfer results with large temperature differences
had been achieved using the film-temperature correction. I t should be
noted that in Humble et al. (42),although the film-temperature correction
was apparently quite satisfactory in collapsing the data to the constant
properties line, the majority of the data were obtained with heat
addition. T h e performance of the film-temperature correction was quite
inconclusively indifferent with respect to the liniited data obtained with
heat extraction. Furthermore, all of the data of Humble et al. (42) were
correlated on the basis of property values averaged over the length of the
tube. This includes the obviously hazardous procedure of taking the nu-
merical average of a density changing by as much as a factor of ten from
inlet to outlet. Thus, these data (particularly those obtained with heat
extraction) cannot be looked upon as a conclusive demonstration of the
validity of the film-temperature correction for turbulent flow. Deissler
(43) and Van Driest (44) almost simultaneously published solutions
of the turbulent boundary layer for flow in tubes and over flat plates,
respectively, in which uniform distributions of heat flux and shear stress
across the boundary layer were assumed and a mixing length theory
employed, together with an assumption of a turbulence Prandtl number
of unity. Deissler, using a two-zone integration of the boundary-layer
equations, obtained a result qualitatively different with respect to the
variable properties question from that obtained by Van Driest, who
used the same eddy transport relation all the way across the boundary
layer. It was found by Deissler that his low-speed (incompressible)
skin-friction and heat-transfer coefficient results with heat transfer could
be reduced to the constant properties result within a percent or so by
employing the film temperature Tamof Eq. (60) for property evaluation.
This success encouraged the use of the film temperature in correlating
the results of Humble et al. (42). Van Driests results could not be so
correlated inasmuch as the influence of wall cooling according to his
analysis was quite sensitive to the Reynolds number and M . At some
Reynolds number for each M (about 2 x lo6 for M = lo), the curves
for different T,/ T crossed, indicating a qualitatively different effect of
wall cooling on either side of this Reynolds number. Going back even
earlier than the analyses that led to the reference-temperature method,
one encounters the Colburn equation for flow in smooth tubes, which
is based on film-temperature evaluation of properties (45). While it is
true that the Colburn film-temperature is equivalent to Tan, of Eq. (60)
when T is replaced by T, , it is significant that the Colburn equation is
based on a density evaluated at the bulk temperature rather than the
[421
HEATTRANSFER FLOWS
ACCELERATING
FROM RAPIDLY
film temperature. Hence, the equivalent expression for u from the

Colburn equation is
which predicts that for gases with Pr insensitive to temperature, the heat
transfer is lower than the constant properties value in the case of wall
cooling. T h e Colburn equation was used to correlate data for both gases
and liquids. Its success was not critically tested by the available data,
which generally had an experimental scatter of f40%, masking the
effects of properties variation.
Later, more definitive experimental results appeared which tend to
disagree with the reference-temperature concept for predicting heat
transfer and skin friction with cooled walls. Lobb et al. (18) reported
results of the influence of heat transfer on skin friction at M = 5.0 and
6.8 and showed that for TWITuwas low as 0.5 the skin-friction coefficient
values were essentially equal to or slightly lower than those for an
adiabatic wall. T h e reference-temperature correction (Eqs. (57) and(59))
would have predicted values 31 and 35 % higher than the adiabatic wall
skin-friction coefficients for M = 5.0 and 6.8, respectively. Zellnik and
Churchill (21) have reported local heat-transfer data from the inlet
region of circular tubes with air entering at temperatures from about
500 to 2000F and the wall maintained at about 100Fby water cooling.
It was found that data obtained with TWITbdown to about 0.25 agreed
with the accepted constant properties values and showed no discernible
influence of the ratio TWITb, within the experimental scatter of f15%
when properties were evaluated at the bulk temperature. Similar results
were also reported by Wolf (20)with both air and carbon dioxide trans-
ferring heat to cooled entrance regions of pipes over a range of TWITb
from 0.3 to 0.7. Wolfs local data, reproduced in Fig. 4 of this chapter,
could also be correlated with no effect of TWITb,within the experimental
scatter of & l o % , when properties were evaluated at the bulk tempera-
ture. While the data of Zellnik and Churchill (21) and Wolf (20) were
obtained principally with turbulent flow, Kays and Nicoll (46) have
reported length-mean heat-transfer data in circular tubes with low-speed
laminar flow over a range of TWIT,,,from 0.85 down to 0.55, where T ,
is the log mean bulk temperature. These data were found to be equal to
95% of the constant properties value and were also independent of the
ratio TWIT,,,, when the properties were evaluated at T,,, . It must be
noted, however, that the magnitude of the predicted variable properties
correction is only slightly more than the 5 % scatter.
While these data are probably insufficient to settle the question of
1431
D. R. BARTZ
the effect of property variation on heat transfer and skin friction defini-
tively, they certainly cast suspicion on the validity of the reference-
temperature methods as applied to severely cooled walls. I n the authors
opinion, there probably should be no correction made to the constant
properties values of c, and cj based on free-stream properties except
for that due to compressibility effects. Note that the arguments presented
here have been based upon experimental data obtained with negligible
flow acceleration. Laminar flow solutions for accelerated flows have
indicated that the effect of wall cooling is to lower predicted skin friction
coefficients below the constant property values. On the other hand,
Stanton number predictions may be altered either above or below
constant property values, depending upon the viscosity-temperature
relation selected. Thus, for accelerated laminar flow, cold walls further
disturb the equality between C , and C j / 2predicted by Reynolds analogy.
Lacking the capability of making similar calculations for the turbulent
boundary layer, there is no alternative but to resort to intuitive argu-
ments such as presented in Section 11, A and then to check the results
against the best data available. For turbulent boundary layers, a method
for relating the value of C,n for an adiabatic wall at high speeds to the
constant properties value c, has been suggested by Coles (1.5) and is
presented in Appendix A of this chapter. Calculations have also been
made (and are presented here), however, which are based on the arith-
metic mean temperature as a reference temperature. This has been done
to illustrate the possible magnitude of the variable properties correction
and to relate the results with previously calculated and correlated results.
T h e fact that the data in some regions and under some conditions best
fit the predictions based upon the arithmetic mean temperature may or
may not be significant. Other factors at work, such as transition, free-
stream turbulence, acceleration, etc., as will be discussed, may account
for this behavior. Additional experiments in which boundary-layer
details are determined by probing will be needed to sort out these
effects.
E. DRIVING
POTENTIAL
1. Without Chemical Reaction
T o this point, attention has been focused primarily upon the heat-
transfer coefficient -that part of the Newton cooling equation which
is dependent upon the fluid motion. T o complete the calculation of the
heat flux, this coefficient must, of course, be multiplied by a driving
potential which represents the difference of the energy levels between
1441
HEATTRANSFER FLOWS
free stream and wall. T h e object of this separation of variables is to

arrive at a coefficient which is independent of both the difference and
the absolute levels of the energy in the free stream and wall. In the case
of low-speed flow with small temperature difference between free stream
and wall, this ideal separation of variables is possible, the driving poten-
tial in this case being simply the temperature difference
qw = K,(T - T,) (63)

As the speed of the flow increases to the point at which there is an
appreciable difference between free-stream stagnation and free-stream
static temperatures, it becomes evident that in order to maintain the
separation of variables, one must modify the low-speed correlations
of the heat-transfer coefficient as in Eq. ( 5 5 ) and also replace T with
a recovery temperature T a w ,the temperature the wall would attain
if insulated:
I t should be noted that

U2
Taw=T+R- (65)
2%
where R , the recovery factor, has been shown by both theory and
experiment (including flow with acceleration) to be equal to Pr* for
turbulent boundary layers ( 4 7 , 4 8 ) .
As the temperature difference and velocity increase to the magnitudes
of interest to a cooled supersonic nozzle flow, the question of the appro-
priate value of cp in both Taw (Eq. (65)) and h, (Eqs (17) and (48))
becomes important. Van Driest (44), Eckert (22), and others have shown
that the enthalpy difference is the fundamental measure of the energy
level difference, or driving potential, and can, in general, be substituted
for the product of c, and temperature difference; thus,
-
qw = W ( L w - Iw) (66)
where hi is h, with cp eliminated (e.g., Eqs. (17) and (48)):
2. With Chemical Reaction

I n recent years, the question of the effect of chemical reaction on
the heat transfer has received considerable attention. I n particular,
for both high-speed stagnation point flow and flow of high-energy
[451
D. R. BARTZ
rocket-engine combustion gases, an appreciable mole fraction of dis-

sociated species such as N , 0, H, OH, F, etc., is present in the free
stream. Upon being brought to rest against a cooled wall, the natural
tendency toward chemical equilibrium results in very energetic exo-
thermic recombination reactions. If the local conditions are such that
these reactions are very slow with respect to species residence times
near the wall, the flow is said to be chemically frozen, and no effect on
heat transfer is to be expected, provided the wall is noncatalytic to the
recombination reaction. If, at the other extreme, the conditions are such
that the reactions are so very fast that the flow is able to maintain
local chemical equilibrium everywhere within the boundary layer, then an
enhancement in the heat transfer is to be expected as a result of the
energy released by the recombination. Two other intermediate condi-
tions are also of interest: (a) a boundary layer which is chemicallyfrozen
but is in contact with a catalytic wall of such activity that all dissociated
species diffusing to the wall are immediately recombined; and (b) a
boundary layer with intermediate arbitrary recombination rates both
within the flow and at the wall. Because of the extensive treatment of
these effects on heat transfer by Denison and Dooley (49), Lees (50),
Rosner (6), and others, it will suffice here to give only the results as
they apply to the rocket nozzle case and to discuss the possible magni-
tudes of the effects.
Before proceeding to the general result that will apply for the cases
described above, it is important to establish the influence of chemical
reaction on heat transfer in the special case in which the parameter
pD,,E,/X is unity. As discussed by Lees (50), when a gas contains more
than one chemical species, heat is trasported not only by heat conduction
but also by diffusion currents carrying both thermal and chemical en-
thalpy. I n two-dimensional or axially symmetric flows of boundary-layer
type, the rate of heat transport across stream lines in laminar flow is
given by
where ii is the local static enthalpy:
and where 1: is the heat of formation of the ith species. In the special
case of a binary mixture, Ficks law states that
HEATTRANSFER FLOWS
Thus, Eq. (68) becomes
T h e complete static enthalpy, which includes both the thermal and

chemical enthalpies of the mixture, is defined by i = ZCKiii,so that
di = Fp dt + C i, dk, , where Fp = Z k , c p , . Thus, substituting in Eq.
(71), the expression for q becomes
where the first two terms are due to conduction and the third to diffusion.
From Eq. (72), it is clear that when the parameter pDl,Cp/X is unity, the
heat transfer is given by
from which Lees draws the conclusions that the net heat flux is then
independent of the mechanismof heat transfer and of thechemical reaction
rates in the mixture for a given enthalpy gradient. T h e parameter
pDl,Cp/h has been designated as the W. K. Lewis number and given
the symbol Lef . Although this result has been derived for a binary
mixture in a laminar boundary layer, both Lees (50) and Rosner (6)
imply that the same result can in principle be derived for a multi-
component reacting gas mixture by using the generalized conservation
equations as presented by Hirschfelder et al. (35). However, to make
the problem more tractable, Lees (50) considers a gas mixture con-
sisting of two groups of species, each with about the same atomic
or molecular weight and about the same mutual collision cross
sections. He asserts that these can be replaced by an effective
binary mixture, in which each group acts like a single component so
far as diffusion is concerned. T h e enthalpy of each individual species
must be carefully distinguished in calculating the energy transport,
but there is only one effective diffusion coefficient, and Ficks law
is applicable. I t would appear that this approximation is equally
applicable to the chemical recombination within a rocket nozzle. As
for the applicability of the result to the turbulent boundary layer,
Lees (50) presents a series of plausible arguments and assumptions, as
rigorously convincing as any on the turbulent boundary layer, which
suggest that as long as some form of Reynolds analogy is still applicable,
results for the reacting laminar boundary layer are applicable to the
turbulent boundary layer. T h e molecular W. K. Lewis number
appears in the turbulent reacting boundary-layer equations just as the
P71
D. R. BARTZ
molecular Prandtl number appears in the turbulent nonreactive bound-

ary-layer equations because of the calculation of the total thermal
resistance as the sum of that in a laminar sublayer, a buffer layer, and
a turbulent core for both the conductive and diffusive energy flux, similar
to that made by von Khrrnhn in the nonreactive case.
For the special case of a frozen laminar boundary layer and a catalytic
wall, as described above, Lees (50) and Rosner ( 6 ) have obtained the
result
where 1, is the free-stream value of i, as given by Eq. (69), dIkinetic

is
U 2 / 2 and Alrllem is Ci (Ki- Kiw)lio.When Lef is unity, Eq. (74)
reduces to
q =W U U - 1,) (75)
as one should expect from Eq. (73). Note, however, that Eq. (75) is not
identical to Eq. (66) except in the special case in which dl,,,,,, is zero.
I n order to generalize the driving potential for a reacting multi-
component laminar boundary layer for a W. K. Lewis number
other than unity, Rosner (6) has made use of the energy equation as
written by Lees (50) in the following form:
where Pr, is the frozen Prandtl number of the rnulticomponent mix-

ture, Dim is the diffusion coefficient of the ith species with respect to
the mixture, is (pDi,llFP,f)/h,,and i, is given by Eq. (69) and con-
tains chemical enthalpy. By making the approximation [see derivation
of Eq. (33), Lees (SO)]
2 i, 3
aY
ZjjO) 3
aY
(77)
which neglects the differences in thermal enthalpy between components

compared with the heats of formation (usually highly accurate for mix-
tures of interest), Rosner ( 6 ) has rewritten Eq. (76) as follows:
1 + (Pr, - 1 ) a i k 1net I c / ay
az,/ay
FROM RAPIDLYACCELERATING FLOWS
HEATTRANSFER
where i ki nct i c = u2/2 and ichern,i

= Iio)ki.Noting the formal similarity
between the general energy equation (as written in the form of Eq. (78))
and Eq. (74) for its special case, Rosner (6) has proposed a generalized
driving potential of the form
4 = h i ( z , , R - 11,to) (79)
where the subscript R is for recovery, f for frozen, and the value of
I! = I, - dI,cinetic- dIchem = ZiKiJ: F ~ dT , ~is computed for the
different equilibrium compositions existing, respectively, at the edge
of the boundary layer and at the wall. T h e relation between Eqs. (77),
(76), and (74) becomes evident by expanding I f , R i.e.,
;
4 + ( R - 1)
= hi[111 dlklnetlc +2 a
(Rc,i - 1) dfchem.i - lf.w] (80)
or, equivalently,
where it is noted that u = 0 at the wall so that &kinetic = I k i n e t i c ,

and where Rc,i is a chemical recovery factor dependent upon
It is suggested that the dependence of R, upon Lef will take a slightly
different form from one flow to another just as, for example, R for a
laminar boundary layer is equal to Prt and for a turbulent boundary
layer it is equal to Pri. I t was already shown in Eq. (74) that for a frozen
laminar boundary layer, R, is equal to Leg. Fay and Riddell (51) who
carried out numerical calculations of heat transfer to a stagnation point
in air with dissociation, found that their results were correlated well
with R, given by for equilibrium and by for a frozen
boundary layer with a catalytic wall. For turbulent boundary layers,
Lees (50) argues that R, by Reynolds analogy is dependent upon Le
to the negative of the same power that h, is dependent upon Pr. Hence,
from Eq. (48), it is concluded that R, is equal to whereas from
Eq. (A.8) it would appear that R, is also Ro-dependent.
As for the effects on the heat transfer to be anticipated when recom-
bination rates are finite at the wall or in the boundary layer (or both),
Rosner (6) has presented qualitative arguments which suggest that in
no case will the finite rates result in heat transfer either below the
frozen case with the noncatalytic wall or above the complete equilibrium
case. He notes that the heat-transfer rate becomes insensitive to the gas-
phase reaction rates when the surface recombination rates become very
large (i.e., for a catalytic wall). Analogously, as one would expect, the
[@I
D. R. BARTZ
heat-transfer rate becomes insensitive to surface recombination rates when

the gas-phase recombination rate becomes large since no atomic species
reach the wall. As mentioned in the previous paragraph, the heat transfer
in these two limits is nearly equal. For a low recombination rate, the
heat transfer approaches the chemically frozen case as the catalytic
activity of the wall approaches zero. For a noncatalytic wall, the heat
transfer approaches that of full chemical equilibrium as the chamber
pressure, and, hence, homogeneous reaction rates increase. Unfortun-
ately, since few reaction rate data are available against which to compare
results, it is not possible at present to make calculations with confidence
of the effect on heat transfer of finite rates. Thus, it is recommended
that the equilibrium limit (Eq. (80) or (81)), with R, equal to Leg.6, be
utilized in computing rocket thrust-chamber heat flux. This is probably
well justified by the high pressures prevalent in the high heat-flux
regions of most rocket thrust chambers. Of course, it remains to be
demonstrated with definitive local data whether or not the accounting
for chemical reaction suggested here actually applies for accelerating
turbulent boundary layers with varibale properties and streamwise
varying species concentrations.
T h e maximum influence on heat transfer of the chemical recovery will
probably be encountered with one of the most energetic chemical rocket
propellant combinations such as H, + 0, at its maximum performance
mixture ratio of 5 ; or in the case of the nuclear rocket, with hydrogen
as the working fluid, both with nozzle walls at maximum temperature
(say, about 1500K for H, regenerative cooling). Under these conditions,
at 10 atm, L I I ~ ~ ~ ~-, /I,,J +
( I , , for H , 0, would be about 15%,
whereas for the nuclear rocket it would be only about 7%. T h e value
of Le, for H atom diffusion through equilibrium-dissociated hydrogen
is given by Rosner (6) as 1.3 at these conditions. Thus, the heat-transfer
enhancement due to chemical reaction for the nuclear rocket case from
Eq. (81),with R, = Ley.6, amounts to only about 9%, of which only
1 *yo is due to Le, > 1. T h e value of Le, for H atom diffusioti'through
water vapor, the major combustion gas constituent for the H, + 0,
case, has been suggested by Rosner (6) as being conservatively as high
as 4.0. T h e principal reason for the high Le, is the large imbalance in
molecular weight between the H atoms and the water vapor. Even with
this large value of Let for H atoms, the net enhancement of the heat
transfer over that of a frozen boundary layer with a noncatalytic wall
is not as great as one might suspect because of the 15% dIchem/(Io - I,.,,),
only 8 % is due to H atom recombinations, the other 7 % being the
result of reactions forming H,O. T h e energy release from the latter
processes while comparable to that of the H atom, will be limited by
POI
HEATTRANSFER
FLOWS
the diffusion of the much heavier species 0 and OH, which, diffusing
through H,O vapor, will probably have values of Let near unity, as
will the H,O vapor itself. Thus, virtually all of the 12% enhancement
in heat flux due to Le,,, # 1 results from the H atom reaction according
to Rosner ( 5 2 ~ ) .T h e total enhancement in heat flux over that for a
completely frozen boundary layer with noncatalytic wall is 32 yo. It
must also be remembered that the 15% dIchem/(IO - I,,,) resulted from
100% of theoretical performance, with a wall at 1500OK, at 10 atm
pressure. This fraction will fall rapidly with reduced combustion effi-
ciency (it is down to 7% at 95% combustion efficiency) and with
reduced wall temperatures and increased pressures.
It is interesting to note that when Le, is unity, or when the enhance-
ment due to R, + 1 is to be neglected in a calculation, it is unnecessary
to determine dIchem . For such a situation, the value of I, can be calcul-
ated by a method described by Welsh and Witte (52), in which it is
unnecessary to solve the simultaneous species equilibrium equations.
Hence, a calculation can be made with a slide-rule in a few minutes.
T h e method is based upon the assumption that the species at equilibrium
at the wall condition are readily known from the initial reactants and
that the wall is sufficiently cool to preclude the existence of any dis-
sociated species in equilibrium there. (This situation pertains for cooled
walls for relatively simple reactants such as H, + F, and H, + 0,.
For a system such as N,O, + N,H,, for example, some assumption
about the presence or absence of NH, must be made but introduces
only small differences.) T h e method is possible because the total energy
available to the system is exactly equal to the heat of reaction, which is
known from the heat of formation of the reactants and the products. T h e
equation for I , - I f , wfrom Welsh and Witte (52),converted from volume
to mass units and translated to the nomenclature of this chapter, is
+ i
Kj /rn cDi dT -
Tref
ITW
cp, dt
Tref
(82)
where subscript j refers to reactant species and i to product species.

Since heats of formation lir are commonly tabulated at 298K, this is a
convenient temperature to take as Trer.If the species KiWare known
because of the assumption of no dissociated species at the wall, then they
are also the same at Trefsince generally Trpr< T , . T h e third term of
Eq. (82) represents a difference in enthalpy between the reactant consti-
tuents at Tin,and T,,, . If Tinj is sufficiently close to Tref(as it usually
[5 11
D. R. BARTZ
is), no appreciable dissociation or phase change occurs between these

temperatures, permitting the use of a simple expression for the sensible
enthalpy change such as that used in Eq. (82). The fourth term represents
the change of enthalpy of the product constituents between Trefand T,, .
Again, if no dissociation or phase change occurs between these tem-
peratures, a simple expression for the sensible enthalpy change such as
appears in Eq. (82) is permissible. If dissociation or phase change does
occur, the final term must be modified accordingly, degrading the sim-
plicity of the equation. While Eq. (82) was derived for of theoret-
ical combustion performance, it is shown in Welsh and Witte (52) that
by multiplying the two bracketed terms by the ratio of the square of the
actual c * to the loo?(, (or ideal theoretical) c * , Eq. (82) can be adapted
for c * lower than 1 0 0 ~ o .
111. Air Experiments
A. PURPOSE
At a very early stage in the investigation of the problem of heat transfer
in rocket nozzles, it was realized that it would be essential to obtain some
heat-transfer data in nozzles operating under controlled conditions for
which the flow would be describable. This need suggested experiments
with heated air flowing through cooled supersonic nozzles for the
purpose of eliminating combustion effects such as large secondary
flows, nonlinear oscillations, variable total temperature, recombination,
excessive free-stream turbulence, and uncertain transport properties.
With hot-air experiments, it is still possible to retain the essentials of the
problem analytically modeled in Section 11; i.e., rapidly accelerating
turbulent boundary layers, variable properties, and variable flow origin.
B. LITERATURE
Even with the general recognition of the necessity of such experi-
ments for many years, relatively few significant results of such experi-
ments are cited in the literature. T h e reason for this paucity of data is
that a rather large air supply and a method of heating the air are required
and that measuring local heat flux from a hot gas to a cold wall in a
region in which the heat flux varies rapidly in the flow direction is
experimentally difficult. T h e earliest experiments of this type reported
were those of Saunders and Calder (53), followed later by Ragsdale and
Smith,s (54). Unfortunately, the nozzles used in both of these investiga-
tions were built with convergent and divergent half-angles of about
r521
HEATTRANSFER
FLOWS
I deg, so that there was negligible acceleration to the flows. Baron and
Durgin (55) obtained experimental data in a two dimensional nozzle
at a stagnation pressure range of 6 to 30 psi at 570"R. They succeeded
in correlating very low heat-flux data from low pressure-gradient air
flow with turbulent boundary-layer flat-plate correlation equations to
h25 yo after considerable manipulation of the "effective flow origin"
and of effects of surface-temperature variation. Their raw or "uncor-
rected" data tended to fall 50 to 70% below a Colburn analogy correla-
tion for a flat plate, i.e., St = 0.0374(R,)-1/5 for supersonic regions of
the nozzle, but showed fair agreement in the subsonic region. It was not
until still later that investigations were made for the specific purpose
described in Section 111, A by Massier (56), Kolozsi (57), Fortini and
Ehlers (58), and Back et al. (59). It is the latter of these references that
will form the principal basis for discussion in the remainder of this
section, although brief comments will be made about the results of
Kolozsi and of Fortini and Ehlers. Kolozsi obtained data with a 78-deg
half-angle convergent-divergent nozzle at 1200"R at stagnation pressures
of 226 and 370 psia. Fortini and Ehlers obtained data with a nozzle having
a 30-deg half-angle convergence and a Rao-design divergent section
at 1600"R and 300 psia stagnation pressure. T h e data of Back et al.
were obtained with a 30-deg half-angle convergent, 15-deg half-angle
divergent nozzle over a range of stagnation temperatures from 1000
to 2000"R, and stagnation pressures of 30 to 250 psia. T h e wide range of
operating conditions of Back et al. (59) makes the results of particular
interest.
C. EXPERIMENTAL
TECHNIQUES
The flow system and instrumentation locations utilized in the investi-
gation reported in Back et al. (59) are depicted in Fig. 11. T h e source
of the heated air was a wind-tunnel compressor system, followed by a
turbojet combustor can in which very lean mixtures of methanol and
air were burned. T h e mole fraction of methanol required to reach even
2000"R was small enough that the resultant gas has thermodynamic
and transport properties only slightly different from air. T h e fluid-
dynamic effects of the combustion were adequately damped out by the
low-velocity calming section and the system of baffles and screens
followed by an aerodynamically contoured exit from the calming section.
T o prevent the unwanted growth of thick thermal boundary layers, the
calming section and contoured exit were lined with inconel, and the
system was designed so that the liner very nearly reached the stagnation
temperature. By varying the length of the instrumented cooled duct
P31
D. R. BARTZ
BAFFLES AND
TEMPERATURE PROBES
IN NOZZLE WALL
WATER FLOW RATE AND

-C +o~[~~~$~~~,p TEMPERATURE RISE
ORIFICE EXIT
STATIC PRESSURE
FIG. 1 1 . Flow and instrumentation diagram; heated-air investigation of Back et al. (59).
All dimensions in inches.
downstream of the settling chamber, the boundary-layer conditions

at the nozzle inlet could be varied. Free-stream temperature was
measured by two shielded thermocouples just upstream of the nozzle
inlet. Traverses were made of boundary layers about l a in. upstream
of the nozzle entrance to determine the temperature and velocity profiles
and boundary-layer thicknesses. Probe tip details are shown in Fig. 12.
T h e nozzle utilized for all of the measurements of heat flux had an
internal contour closely similar to that depicted in Fig. 5, only with a
slightly larger throat radius of 0.902 in. T h e static-pressure distribution
0042 D
0.040D Izq & ,-- MgO INSULATION
J M E L THERMOCOUPLE
L0.006-D ASPIRATION HOLES ( 4 PLACES)

(b)
FIG. 12. Tip details of traversing boundary-layer probes; heated-air investigation of
Back et al. (59). (a) Stagnation-pressure probe; (b) stagnation-temperature probe. All
dimensions in inches.
WI
HEATTRANSFER
FROM RAPIDLY FLOWS
ACCELERATING
along the wall of the nozzle was measured with thirty-two static-pressure
holes of 0.040-in. diam, spaced both circumferentially and axially in
the nozzle wall.
Steady-state wall temperatures and heat fluxes were determined from
the output of thermocouples imbedded in cylindrical plugs, a typical one
of which is shown in Fig. 13. Three thermocouples were formed along
ALUMEL WIRES CHROMEL WIRES
FIG. 13. Thermocouple plug diagram; heated-air investigation of Back et af. (59).
All dimensions in inches.
the length of each plug, which, when instrumented, was pressed into
a hole drilled through the nozzle wall. By making the plugs from the
same billet from which the nozzle was machined and using a force fit,
the thermal disturbance to the wall was minimized. T h e material selected
for the nozzle and plugs was type 502 stainless steel because of the known
insensitivity to temperature of its thermal conductivity. Measurements
of the thermal conductivity for a sample from the billet used for the
nozzle were obtained from the National Bureau of Standards. T h e
locations of the thermocouples were determined to an accuracy of less
than 1 yo of the inter-thermocouple spacing by a Kelvin bridge electrical
technique described in detail by Back et al. (59). Twenty-two plugs were
used to obtain the heat-flux distribution, with spacing made both
axially and circumferentially. I t can be shown that if the isotherms of
P I
D. R. BARTZ
the nozzle wall are essentially parallel (as they were experimentally
shown to be in Fig. 14), the local heat flux can be determined from the
local temperature gradient measured normal to the nozzle inner surface
without the necessity of having to make any correction for axial con-
duction.
40
&* 3.0
1
0
t-
2 2.0
v)
2
0
Q
LT
I .o
0
0 0.1 0.2 03 0.4 0.5 0.6 0.7 0.8 0.9 1.0 I I
FIG. 14. Nozzle wall isotherms; heated-air investigation of Back et al. (59).
I n the investigation reported in Fortini and Ehlers (58), the problem

of axial conduction was fairly well circumvented by the use of thermally
insulated, ane-dimensional heat-conduction plugs. Both of these thermo-
couple plug techniques have been made to yield what are believed to be
reliable and accurate data; however, not without considerable effort.
As a consequence of these evaluations of the effort required, Back et al.
(59) are using calorimetrically cooled nozzles in continuing their air
experiments. These are made of axially short sections, which permit
measurement of the less desirable circumferentially averaged and axially
semilocal values of heat flux. Nozzles of this type have been used
extensively in obtaining measurements under rocket thrust-chamber
conditions. An example of such a nozzle will be shown in connection
with Section IV.
From this discussion, it is evident that it is much more difficult to
measure local heat-transfer coefficients from a hot gas to a cool wall than
to measure the wall temperature for a controlled local input of heat
from a hot wall to a cool gas. T h e extra effort is apparently well justified
for the purposes of the experiments under consideration here in view of
the suspected different influence of variable properties between heating
and cooling.
P I
HEATTRANSFER
FLOWS
D. EXPERIMENTAL
RESULTS
1. Static Pressure Distributions
Measured wall static pressures from tests made at stagnation pressures
ranging from 45 to 150 psia are presented in Fig. 15. Deviations from
the one-dimensional-flow predicted wall static pressures are clearly
evident and are obviously beyond the spread of the data. The deviations
undoubtedly result from significant radial velocity components caused
0
5
K
W
K
2
v)
W
lY
n
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
AXIAL DISTANCE RATIO Z/L
FIG. 15. Ratio of static to stagnation pressure along the nozzle (of Back et ul., 59).
by the taper and curvature of the nozzle. I n Fig. 16, the distribution
of local mass flux per unit area at the edge of the boundary layer pU
computed from the wall static-pressure data and normalized by the mass
flux per unit area predicted from one-dimensional flow p1U , is presented
for 75-psia stagnation pressure. Since from the closed-form approxima-
tion equation (Eq. (48)), 11, is proportional to (PU)O.~,
deviations of the
real flow pU from the one-dimensional values of up to 150/, result in
errors of over 10% in h, if the one-dimensional plU1 values are used.
Deviations considerably greater than this in severely turned flows in the
1571
D. R. BARTZ
THROAT z/L = 0.603 C

07 I I 1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
AXIAL DISTANCE RATIO Z/f
FIG. 16. Ratio of local to one-dimensional mass flux along the nozzle (of Back et al., 59)
divergent portions of contoured nozzles have been predicted by method-

of-characteristics flow analyses and have been observed by Fortini and
Ehlers (58). I n the transonic region of a nozzle, the prediction of the
real flow is considerably less certain than for the fully supersonic region.
Nevertheless, Oswatitsch and Rothstein (60) did consider two-dimen-
sional flow in a convergent-divergent nozzle. The wall boundary layer
was neglected, and it was required that the fluid velocity at the wall be
exactly parallel to it. Their result of the ratio of the mass flux per unit
area at the nozzle wall to that for one-dimensional flow is given by
where
T h e predicted ratio pU/plUl from Eq. (83) is in fair agreement with the
data presented in Fig. 16. T h e position of the sonic line is predicted to be
somewhat upstream of the geometrical throat. Note the predicted dis-
continuity of the mass flux at the intersection of the conical sections of
(581
HEATTRANSFER FLOWS
the nozzle with the circular arc throat region. From results such as these,
one concludes that if requirements of precision of the calculation of
nozzle heat flux are sufficient to justify a boundary-layer (rather than a
closed-form) calculation, then the deviations from one-dimensional flow
should be determined and fed into the calculation.
2. Boundary Layers at Nozzle Inlet

In order to determine the nature and thicknesses of the velocity and
temperature boundary layers entering the nozzle, the boundary layers
were probed just upstream of the nozzle inlet. Typical results of such
measurements made with an approach length of about 3& diameters are
presented in Fig. 17 for 1500"R temperature and a range of stagnation
pressures. A 1/7-power constant properties profile (as assumed in Section
11) for both temperature and velocity is found to agree satisfactorily with
the data. T h e thicknesses 6*, 8, and r$ were calculated using the thick
boundary-layer relationships of Appendix C and are about 5 % less
than those predicted assuming the thin boundary-layer relations of
Eqs. (9), (12), and (15).
3 . Heat Transfer Results

In Fig. 18, a composite presentation is made of raw heat-transfer
coefficient data covering the full range of stagnation temperatures and
pressures investigated. T h e majority of the tests were duplicated, with
the results found to be reproducible to within 1 2 % . Some evidence of
circumferential variations of greater extent than this can be seen in the
figure by observing symbols that are similarly tagged. Although circum-
ferential nonuniformities were not evident from probing the free stream
ahead of the nozzle, it is possible that they could have gone undetected
in the boundary layer. T h e data from the several tests presented con-
sistently show a maximum heat flux upstream of the geometrical throat
very close to the point of maximum mass flux per unit area, as indicated
in Fig. 16. T h e apparently strange behavior of the data near the nozzle
exit for the lower stagnation pressures is due to separation.
I n order to compare these results with distributions predicted from
the turbulent boundary-layer calculation of heat flux presented in
Section 11, A, the distributions from two tests have been selected as
being typical of most of the remainder of the data. T h e data of which
these results are not typical will be discussed separately. T h e two tests
selected were both made at a stagnation temperature of about 1515"R,
one at a stagnation pressure of 75 psia and the other at the highest
W1
D. R. BARTZ
1.0
0.8
0.6
0.4
1.0
0.8
0.6
I
0.4
FIG. 17. Boundary-layer profiles 1.25 in. upstream of nozzle inlet with 18-in. cooled
approach length (59).
stagnation pressure investigated, 254 psia. T h e heat-transfer coefficient

distributions from the two tests are presented in Figs. 19 and 20,
respectively. T h e four solid lines on the figures are the iterative simul-
taneous turbulent boundary-layer solution results (Eq. (A.8)) obtained
using the experimentally measured boundary-layer thicknesses at the
nozzle inlet for the initial conditions and the experimentally determined
pU distribution (see Fig. 16). In four separate calculations, use was made
of different combinations of the properties evaluation temperature and
the interaction exponent n (Eq. (A.8)). T h e properties alternatives
P I
HEATTRANSFER
FROM RAPIDLY-ACCELERATING FLOWS
0.043
0.046
0.031
AXIAL DISTANCE RATIO z/f
FIG. 18. Heat-transfer coefficient versus axial distance ratio with 18-in. cooled
approach length (59).
D. R. BARTZ
0 0.1 0.2 0.3 0.4 0.S 0.6 0.7 0.8 0 9 1.0

AXIAL DISTANCE RATIO Z/L
FIG. 19. Comparison of measured and predicted heat-transfer coefficients for heated air
at p o = 75 psia (14, 59).
FIG. 20. Comparison of measured and predicted heat-transfer coefficients for heated
air at p o = 254 psia (14, 59).
C621
HEATTRANSFER FLOWS
selected were (a) density and viscosity evaluated at a reference tempera-

ture equal to arithmetic mean between wall and free-stream static (film)
temperature (Eq. (60)); and (b) properties evaluated in accordance with
Coles' method of computing C,for an adiabatic wall (Eqs. (A.l)-(A.3)).
T h e two alternatives for the boundary-layer interaction exponent n
were (a) 0.1, which is about the value used in ref. (9), where some
physical justification for its selection was presented; and (b) 0, i.e., the
momentum and energy equations essentially decoupled. T h e lowest of
the four curves, curve D, is seen to be approximately 30% below
curve A in the throat region of both figures. Curve A, based on n of 0.1
and properties at Tan,, it should be pointed out, is based on the same
assumptions as those behind the results presented in ref. (9) (and
in Figs. 9 and lo), to which the closed-form equation (Eq. (50)) was
fitted. It is seen, however, that values from Eq. (50) lie above curve A,
since curve A resulted from an iterative simultaneous solution, as dis-
cussed in Section 11. T h e different relationship between curve A values
and Eq. (50) values of h, in Figs. 19 and 20 is due to differing entrance
boundary-layer thicknesses for the two sets of calculations. T h e dis-
similarity of the shapes of the curves as shown in Fig. 10 is due to the
use of experimental p U values in the boundary-layer calculations and
the use of plU1 in Eq. (50). (More closely similar results could be ob-
tained by using experimental pU values in Eq. (48).) Significant (but
not conclusive) is the agreement in the throat region between the data
and curve D, exhibit in both Figs. 19 and 20, I n the contraction and
expansion sections .of th'e nozzle, the data were perhaps in equally good
agreement with curve C. I t is evident that unless a large quantity of such
data is compared with predictions covering a wide range of operating condi-
tions, differences of this order probably cannot be resolved in this manner.
T h e effect of varying nozzle-inlet boundary-layer thicknesses on the
heat transfer is shown in Fig. 21 for a stagnation temperature of 1500"R
over a range of stagnation pressures from 75 to 200 psia. With no cooled
approach length (i.e., nozzle connected directly to contoured discharge
nozzle of calming section), the inlet velocity boundary layer was deter-
mined by probing to be about 5 % of the nozzle-inlet radius. T h e heat-
transfer coefficients for tests made under this condition were consistently
higher over most of the nozzle than those of tests in which an 18-in.
cooled approach section was used. For these tests, it was determined by
probing that the inlet boundary-layer thickness was up to about 25%
of the nozzle-inlet radius. T h e trends of higher heat-transfer coefficients
in the contraction and throat regions are consistent with predictions of
the type shown in Fig. 9. T h e predicted smaller, but still persistent,
differences in the divergent region are also observed. It is clear that in
~631
D. R. BARTZ
-
IA
0
N.
.-E
Y
0
55
P
2
b
.2:
I-
z
w
uLL
LA
W
0
V
a
W
IA
v)
2
a
a
cI
t
W
I
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
FIG. 21. Comparison of measured and predicted heat-transfer coefficients for various
boundary-layer thicknesses at nozzle inlet for heated air (59).
real flows of interest, it is unlikely that near-zero inlet boundary-layer

thicknesses will be encountered, nor were they experimentally observed
even with thc zero length inlet duct. Rather, thicker boundary layers,
such as those observed in the tests of Fig. 18, will be encountered.
As mentioned earlier in this Section, some of the data did not behave
in a qualitatively similar manner to those of Figs. 19 and 20. Data from
tests of this type are presented in Fig. 22 and are compared with a
type-D prediction curve; i.e., n = 0, properties evaluated in accordance
r641
HEATTRANSFER
FLOWS
3.5
TEST No. 266
NOMINAL CONDITIONS
3.0 - p0 = 44.8 psia ~
To 1503OR
.P T, = 630 - 68093
c 25 -
2
1
I
I 0
0 I
FIG. 22. Comparison of measured and predicted heat-transfer coefficients for heated
air at po = 44.8 psia (59).
with Coles' prediction of C, for an adiabatic wall. Note that although

near the nozzle inlet, the data agree fairly well with the prediction curve,
a short way into the nozzle, they begin to fall considerably below the
prediction. Beyond the throat but before the separation point, the data
rise closer to the prediction curve. Data exhibiting this behavior relative
to predictions (type-D curve) were obtained only at the lower stagnation
pressures where it was reproducibly observed. In order to get a better
picture of this behavior and of the variation of the heat-transfer coeffi-
cient at intermediate pressures, the heat-transfer coefficients at a single
axial station in the nozzle were correlated versus stagnation pressure for
virtually all the tests made. These plots, it was found, could be made
more general by (a) nondimensionalizing the heat-transfer coefficient
by dividing by pUc, and multiplying by thereby forming the
modulus ChPro.6;and (b) nondimensionalizing the stagnation pressure
by converting it to its proportional equivalent pU, multiplying by the
local diameter2 D of the axial station of interest, and dividing by the
viscosity at the edge of the boundary layer p, forming the modulus R , .
* Although a local boundary-layer thickness, 6 , 8, or would have been preferable,

it was not measured directly. Hence, the more convenient linear dimension D was used
and justified by the arguments of Section 11, B.
~651
D. R. BARTZ
When this was done, increases in R, at a given station at constant

stagnation temperature were in direct proportion to increases in stagna-
tion pressure. T h e data from virtually every test made with an 18-in.
cooled approach duct are presented in this fashion in Fig. 23 for selected
0.8
- 3 -I0 0.6
as
3 I U O ? m I 1
lo-: I I
a9 I
I I I l l I
I I l l 1 864 1 P
I I I 1 I 1 111
la3
0.8
06 I i l l z / L =0.469 0.6
1 d L ' I I II
1 i I i I I 1 I I I I J]
0.2 OA 06 a8 106 2 4 6 0.2 a4 06 ae 106 2 4 6
puD puD
P P
FIG.23. Correlation of CnPro.6versus RD ar various subsonic and supersonic area
ratios, with 18-in. cooled approach length (59).
Prediction from Eq. (50), Trer Tam; T,/To as noted
=
- - - Prediction from Eq. (48), Taw Trer =
- - Prediction from Eq. (A. 8) for To = 15WR, n = 0, C, for adiabatic wall:
0 :To=1030"R, y=1.380; 0 :T 0 = 1 5 W R , y=1.345; A : To=2000"R,
y = 1.328
[661
HEATTRANSFER
FROM RAPIDLY FLOWS
ACCELERATING
axial stations identified by their respective area ratios A / A , and axial

position z/L. When these nondimensional coordinates are used, it is
simple to draw a curve corresponding to the closed-form approximation
equation (Eq. (48)). This was done in Fig. 23, where u' was computed
for film temperature properties evaluation for an appropriate range of
Tw/Toand also properties evaluated at T a w .T h e dashed lines represent
results from turbulent boundary-layer calculations (n = 0, properties
appropriate for adiabatic wall Ct). I n general, at the higher R , , the data
tend to vary with R , in about the same qualitative manner (i.e,, slope)
as these predictions. I n making quantitative comparisons between the
data and the predictions in this higher R, (or parallel) region, cognizance
should be taken of the typical spread of data presented, as in Fig. 18,
believed due to some circumferential nonuniformities. T h e most uniform
agreement with the data in the higher R, region was found to be the
prediction based upon Eq. (48),with Tret = Taw. Except at A / A , = I ,
where such a prediction was about 30% high, agreement to within
about 15% was observed at all A / A , for which data were obtained.
T h e most surprising trends exhibited in Fig. 23 occur at the lower
R , (i.e., lower stagnation pressures), where significant but correlatable
departures from the predictions were observed. This is the R, region
typified by data such as those presented in Fig. 22. I t should also be
noted that at the highest subsonic area ratio, i.e., near the nozzle inlet,
the slope of the data generally follows the predictions for the whole
range of R,. T h e low R, departure of the data persists through the
throat and into the supersonic region. It could actually continue out to
the nozzle exit. However, it was not possible to operate the nozzle at low
staganation pressures without separation at the highest supersonic-area-
ratio stations.
One possible explanation for this behavior of the heat-transfer results
is that because of either the extreme acceleration or the combination of
acceleration and cooling, the entering turbulent boundary layer ex-
perienced a reverse transition back toward (but probably not all the way
to) a laminar condition. At some point downstream in the nozzle, there is
perhaps again a forward transition back to fully turbulent conditions.
For the present, this must stand as merely a hypothesis, which, when
checked against the currently available data, was not found to be violated.
T h e hypothesis must now be checked by making detailed boundary-
layer surveys in the maximum acceleration region. Because boundary
layers are thinning out so rapidly in this region (at least according to
the turbulent boundary-layer calculations), such measurements are dif-
ficult to obtain experimentally. Other investigators have observed
unexpected trends accompanying the acceleration of turbulent boundary
~ 7 1
D. R. BARTZ
layers. The proposed hypothesis may well explain some of the anomalous
variation of heat transfer with stagnation pressure in rocket thrust
chambers reported in Welsh and Witte (52).These data will be discussed
in Section IV. I n Sergienko and Gretsov (61), a turbulent boundary
layer at the entrance of a supersonic nozzle was found to undergo tran-
sition to a nearly laminar one at the nozzle exit when the stagnation
pressure was 4.3 psia. When the pressure was increased to 14.2 psia,
a turbulent boundary layer was found at the nozzle exit. No boundary-
layer or local heat-transfer measurements were made within the nozzle.
I n Back (62), it was reported that heat transfer trends of the type under
discussion here were also observed at lower pressure-gradient subsonic
flow conditions. I n that investigation (62), there was departure from
fully turbulent flow throughout the acceleration region, as indicated by
the linearity of the measured velocity profiles near the wall. In Section
111, E, a derivation is made of a parameter relating to the level of eddy
transport. It is shown that the severe acceleration does influence this
parameter in a manner consistent with the hypothesis of a reverse
transition. I n Fig. 24, the variation of the R8 with axial position (according
to predictions from turbulent boundary-layer calculations) is presented
for a few of the tests of Fig. 18. Note especially that even for the lower
stagnation pressures, the minimum Ra is considerably above values at
which forward transitions are customarily observed for lower pressure-
gradient flows such as over flat plates and in pipes.
T h e heat-transfer coefficient distributions obtained by Kolozsi (57)
for two different staganation conditions are illustrated in Fig. 25,
as is the nozzle contour used. T h e wall temperatures were main-
tained nearly constant throughout the nozzle, such that the ratio
T J T , remained between 0.50 and 0.55 at all stations. Thus, the
variable properties correction based upon the arithmetic mean tem-
perature (i.e., in Eq. (50)) was a maximum of +20% at the nozzle
inlet and about f 1 0 yo at the throat. Note that the throat heat-
transfer coefficient for the higher stagnation-pressure test was about
65% of that predicted by the closed-form approximation (Eq. (50)),
with the variable properties correction noted. This comparative result is
quite close to that obtained by Back et al. (59),illustrated in Fig. 20 for
approximately similar stagnation conditions. The data of Kolozsi have
also been compared with type B and D boundary-layer predictions (see
Fig. 19 for designations) in Fig. 25. T h e data appear to fit the two types
of predictions without much preference in the subsonic region. I n the
throat and supersonic region, the data compare preferentially and quite
reasonably with curve D. Note the rapid decrease beyond the throat of
both the data and the predictions. Because of the small divergence
[681
HEATTRANSFER
FROM RAPIDLY.ACCELERATING FLOWS
AXIAL DISTANCE RATIO r/L
FIG. 24. Predicted momentum-thickness Reynolds numbers along nozzle (59).
angle, a relatively larger growth of the thermal boundary layer results

compared with a wider-angle expansion region and a predicted more
rapid decline of the heat-transfer coefficient. T h e comparatively high
value of the heat-transfer coefficients measured in the entrance region is
readily explained by the fact that a fine screen was placed at the entrance
of the nozzle to disrupt the existing boundary layer in the nozzle approach
passage, thus creating a new boundary layer beginning at the entrance
to the first segment of the nozzle (57). T h e boundary-layer calculation
was made with necessarily finite (but arbitrary) entrance values of 4
and 0 of 0.001 and 0.005 in., respectively. T h e resulting predicted
behavior of h, follows that in the nozzle-inlet region reasonably well
but would be better if thinner boundary layers had been assumed for
the initial condition.
T h e data of Fortini and Ehlers (58) were restricted to essentially one
nominal set of operating conditions, 300 psi stagnation pressure and
1600"R stagnation temperature. Of the tests reported, three were made
~ 9 1
D. R. BARTZ
AXIAL DISTANCE FROM NOZZLE INLET, rn
Comparison of measured and predicted heat-transfer coefficients for air

(data from Kolozsi, 57).
at the nominal conditions, with no flow obstruction upstream of the

nozzle. T h e average of these data is presented in Fig. 26 where they are
compared with predictions of the closed form equation (Eq. (48)), and
to results of turbulent boundary-layer calculations as derived in Section
11, A, with n = 0. Since rather large deviations between the measured
mass-flux and the one-dimensional-flow distribution were observed,
the experimental distribution was used in making both the closed-form
approximation and boundary-layer heat-transfer predictions. (It was
found that the measured mass-flux distribution agreed very well with
predictions from axially symmetric method-of-characteristics calcula-
tions in the supersonic region.) As is evident, the agreement between
the data and one of the turbulent boundary-layer predictions (n = 0,
film properties curve B) is quite reasonable out to very high area ratios.
Unfortunately, the few data obtained in the throat region were somewhat
more scattered than elsewhere in the nozzle. Some difficulty was en-
countered with spurious readings from two thermocouple plugs. The
arbitrarily drawn average curve presented in the original Reference 58
is omitted because of its uncertainty in the throat region. One could
conclude, however, that the most probable value of Fortini and Ehlers'
throat heat-transfer coefficient lies between 70 and 85% of the closed-
form equation (Eq. 50) prediction. In comparison, the data of Back et al.
(59) at the throat for their closest test conditions (254 psi, 1500"R) were
about 70 yo of the comparable closed-form equation prediction (Fig. 20).
~701
HEATTRANSFER
FLOws
d
.-
L - 2
5
Y 14
5
&
c 12
2
10
&
5
W B
9
b.
w 6
8
0:
k l 4
ln
2
a
F 2
2
w o
r) 2 4 6 8 10 12 14 16 18
A X I A L DISTANCE FROM NOZZLE INLET I , in
FIG. 26. Comparison of measured and predicted heat-transfer coefficients for air
(data from 58).
Over most of the expansion region, the data of Fortini and Ehlers
closely follow the boundary-layer prediction based upon film-tempera-
ture properties evaluation and n = 0, curve B, whereas the data of
Fig. 20 tend to fall 5-15 yo below the comparable prediction, consistent
with the throat data comparison. In the contraction region, both the
limited data of Fig. 26 and the average of the data of Figs. 19 and 20
follow the closely similar curve B and curve D predictions satisfactorily
without much preference.
Some data were also obtained by Fortini and Ehlers (58) with signifi-
cant flow disturbance upstream of and within the nozzle contraction
region. These data shed some light on the deviations from the steady,
uniform-flow heat transfer which may result from combustion-initiated
large-scale turbulence or secondary flows to be expected in a rocket
nozzle, or from reactor-core-initiated flow disturbances in a nuclear
rocket nozzle. In their tests, a simulated reactor core (a plug with many
holes) mounted just upstream of the nozzle inlet resulted in about 25%
higher throat heat flux than was obtained in the tests with undisturbed
approach flow. A V-gutter turbulence generator mounted right in the
D. R. BARTZ
contraction region of the nozzle (up to an area ratio of 10) resulted in

about 557; higher throat heat flux.
ON TURBULENCE
E. EFFECTOF ACCELERATION
In order to gain some insight into the mechanism responsible for the
reduction of heat transfer below that anticipated for a fully turbulent
boundary layer observed at low stagnation pressures, Back et al. (59)
have considered the boundary-layer turbulence energy equation (e.g.,
Hinze, 63). Using the conventional notation of Hinze (63), Back wrote
the equation for the convection of turbulent kinetic energy by the mean
flow as
where the first term on the right-hand side of the equation represents
the production of turbulent kinetic energy by the working of the mean
velocity gradients against the Reynolds stresses. T h e second term
represents work done by the turbulence against the fluctuation pressure
gradients, and the third term the convection of turbulent kinetic energy
by the turbulence itself. Finally, the last term represents the transfer
of energy by turbulent viscous stresses. For a two-dimensional flow with
a pressure gradient, the first term can be resolved into its significant
parts, representing the production (or decay) of convected turbulent
kinetic energy.
T h e remaining terms on the right-hand side of Eq. (84) depend on the

turbulence produced. T h e first term in Eq. (85) is always positive and
leads to a production of turbulent kinetic energy. For more usual flow
with negligible streamwise flow acceleration, it is clear that the second
term of Eq. (85) can be neglected. However, for flow with acceleration,
the second term leads to a decay of turbulence, provided t i 2 > vI2.
Thus, a measure of the importance of flow acceleration in reducing the
net production of turbulent kinetic energy is given by the ratio of the
two terms; i.e.,
HEATTRANSFER FLOWS
T o establish a variation of x in the streamwise direction requires a

knowledge of the turbulent quantities across the boundary layer. In the
absence of direct turbulence measurements in an accelerating flow,
Back et al. (59) have adapted flat-plate measurements of Klebanofi (64)
at a momentum-thickness Reynolds number of about 8 x lo3, which
yields an average value of (u"- v'*)/(-u'v') M 1.8. This ratio is
believed not to vary appreciably across most of the boundary layer.
T h e velocity gradient au/ay was taken from the law of the wall to be
(2.5/30)(~,/pv). T h e streamwise velocity gradient was approximated
as being equal to the free-stream value dU/dx, finally giving
22v dUldx
x = --- TudP
(87)
with the proportionality constant obviously being arbitrary. T h e variation

of the parameter x along the nozzle, the contour of which is shown in
Fig. 14, was computed and is presented in Fig. 27 for one stagnation
' 6 I ' O
N
N
AXIAL DISTANCE RATIO r / L
FIG. 27. Predicted effect of flow acceleration in reducing net production of turbulent
kinetic energy at different stagnation pressures (59).
P I
D. R. BARTZ
temperature and a range of stagnation pressures. With decreasing

stagnation pressures, the ,increasingvalues of x characterize the predicted
reduction in net production of turbulent kinetic energy. Although the
development of this paramater and the arguments behind it are somewhat
less than rigorously adaptable to the problem at hand, the variation
of x along the nozzle does display a trend of being largest in the conver-
gent section before diminishing through the throat and divergent section,
which is consistent with the observations of reduced heat transfer at low
stagnation pressures. Detailed turbulence measurements in strongly
accelerated flows would be of considerable value in assessing the
validity of the proposed reverse transition mechanism and its influences.
IV. Rocket Thrust-Chamber Measurements

A. LITERATURE
As discussed in the Introduction of this chapter, the rocket thrust
chamber has not been a very satisfactory research apparatus for investi-
gating the rocket thrust-chamber heat-transfer problem. The numerous
competing, nonrepetitive, complicating factors encountered therein
which influence heat transfer make it extremely difficult to clarify the
picture sufficiently to permit one to obtain results of general validity.
Nevertheless, the final test of the utility of analytical and more basic
experimental investigations must be the comparison of predictions
resulting from these investigations with actual rocket thrust-chamber
heat fluxes. Unfortunately, the only meaningful comparisons are those
with local heat-flux data or, at least, with data which are circumferentially
averaged over a short axial length. Such data are obtained only from
thrust chambers specially instrumented for this purpose, as described
in Section B. Most engine development programs have not included such
tests. Consequently, the limited data available are from applied research
investigations at thrust levels small compared with those of the engines
now under development for boosters. I n 1958, Rose (65) made semilocal
calorimetric measurements in a small thrust chamber operating with
nitric acid-ammonia propellants and summarized results of similar data
available from other investigations up to that date. In 1960, Neu (66),
reported experimental data from liquid oxygen-heptane rocket thrust-
chamber tests at small thrust. This particular propellant combination,
widely in use for large booster engines, has the unique characteristic of
causing deposits of carbon to be rapidly laid down on the thrust-chamber
walls. Although this adds still another significant unpredictable com-
plicating factor, the result is a dramatic reduction in heat flux, serving to
C741
HEATTRANSFER
FLOWS
relieve the engine cooling problem. The question of the prediction of

the cause and effect of this deposition has been discussed by Sellers
(67). More recently, local rocket thrust-chamber data useful for com-
parisons with predictions have been obtained over a wide range of
conditions with nitrogen tetroxide-hydrazine propellants by Welsh and
Witte (52) and by Witte and Harper (68).Powell et al. (69,70)has gathered
local heat-flux data with the higher-energy propellants chlorine tri-
fluoride-hydrazine and liquid oxygen-liquid hydrogen at a thrust level
of about 5000 lb. Rupe and Jaivin (71) have obtained data at a thrust
level of about 20,000 lb and have shown dramatic local variations in
chamber heat fluxes due to injection mass distributions and the influence
of unstable combustion.
TECHNIQUES
B. EXPERIMENTAL
T h e operating rocket thrust chamber as a heat-transfer device has
three unique characteristics: (1) very rapid establishment of steady flow,
(2) very high heat fluxes; and (3) very sharp axial (and sometimes
circumferential) gradient of heat flux. T h e third characteristic
establishes a requirement for very localized measurements, whereas
the first two characteristics suggest methods of making such
measurements. T h e extremely high heat flux establishes the
requirement for considerable cooling if the thrust chamber is to
operate for any appreciable length of time (i.e,, more than a
few seconds). Because the fluxes are generally so high, it is possible to
divide the thrust chamber into numerous axially short segment lengths
which are individually cooled. An example of a thrust chamber designed
for such steady-state calorimetric measurements (68, 69), is illustrated
in Fig. 28. By regulating the flow of each of the passages, 'it is possible
to equalize the temperature rise in each passage so as to minimize
passage-to-passage heat transfer by coolant convection to the barrier
walls. T h e error introduced by axial conduction in the wall between
the gas and the coolant can be made acceptably low by thinning the
wall out to about loo/, of the passage axial length, as illustrated in Fig. 28.
Although the number of passages shown in the figure are sufficient to
permit determination of axial variations in heat flux with adequate reso-
lution, the principal drawback of this technique is that it cannot resolve
circumferential variations. Such variations can become significant with
certain types of propellant injectors and certain operating conditions to be
discussed in Section C. T h e most frequently used coolant for calorimetric
measurements such as these is water, generally metered by turbine meters
and temperature-measured with differentially wired thermocouples.
[751
D. R. BARTZ
a
W
N
0
x
0
'-d- TEST-STAND MOUNTING BRACKET
FIG.28. Sectional, water-cooled thrust-chamber assembly for steady-state calori-

metric measurements of semilocal heat flux (68).
T h e very rapid establishment of full flow conditions in rocket thrust

chambers (i.e., typically much less than I sec) and the currently available
capability of fully measuring all necessary performance parameters in
a few seconds of steady flow operation have led to the practice of building
most thrust chambers of heavy uncooled metal walls for initial develop-
mental testing of the combustion characteristics. This practice naturally
suggested a transient heating measurement of local heat flux, in which
time-temperature histories of local areas of the wall are recorded during
the short test duration. In principle, it is then possible to feed this time-
temperature history into a computer programmed to solve the transient
heat-conduction equations for the configuration of interest to determine
what unique heat-flux input was required to produce the experimentally
recorded history. Analytical investigations of this problem (72) have
shown that such a computation can be made to yield adequately
accurate heat-flux results by solving the one-dimensional (radial) heat-
flux equation in cylindrical coordinates with two experimentally measured
time-temperature boundary conditions, using appropriately tempera-
ture-variable thermal properties of the wall. T h e analysis also demon-
strated the necessity of getting one of the thermocouple junctions as
close to the combustion surface of the wall as possible and further
dictated the necessary temperature sampling rate. An example of one
configuration of the thermocouple plug utilized by Powel and Price (72)
~761
HEATTRANSFER FLOWS
is illustrated in Fig. 29. T h e question of neglecting the axial variation of

heat flux in solving the transient conduction equation was investigated
analytically and experimentally by comparing measured heat flux distri-
butions in identically contoured calorimetric and transient plug-instru-
mented nozzles operating under closely similar combustion conditions.
T SURFACE THERMOCOUPLE
THERMOCOUPLE PLUG
/
PLATING 0005 THICK (MAXI
INSULATED THERMOCOUPLE WIRE

APPROX 0,0080SWAGED IN PLUG
FIG.29. Surface thermocouple plug assembly for transient measurement of local

heat flux (72).
This comparison, in which agreement within about the & 10% circum-
ferential variations of the local data was demonstrated, is reported in
Powell and Price (72), as are further details of the transient conduction
program, the plug construction techniques, and the data-sampling
requirements. T h e most significant advantages of this method of heat-
flux determination are the possibility of making pointwise (i.e., truly
local) measurements and the adaptability to inexpensive thrust chambers
and short test duration.
Although a comprehensive discussion of propellant injection and
combustion principles is beyond the scope of this chapter, several
representative injector types are illustrated in Figs. 28 and 30. A brief
qualitative description of their characteristics will be of use as back-
ground to the discussion of heat transfer in the combustion chamber.
T h e injector illustrated in Fig. 30a is known as an enzian or splash-plate
injector. By directing the propellant jets to impinge behind the splash
plate, a tightly confined initial combustion is established, with a strong
[771
D. R. BARTZ
.
\
SPLASU PLATE OXIDIZER FUEL ORIFICE
MANIFOLD I OXIMZER ORIFICE
ITYFKAL OUAORAN
I /
FUEL
MANIFOLD \
FUEL-OXIDIZER TURUST
IMPINGEMENT CUAMBER ID
1OCATlON
(a) (b)
FIG. 30. Propellant injectors (52): (a) enzian type, (b) showerhead type.
recirculation of hot gases behind the plate. This recirculation action

probably tends to obliterate the discrete nature of the combustion gas
flow that would otherwise result from the use of a limited number of
pairs of propellant jets. It also results in a rather short flow length, in
which the first 90% of the final total temperature is achieved. As the
partially combusted gases Aow through the aperture of the splash plate,
a separated region is created which, because of the vortex shedding, is
characterized by a strong reverse flow near the wall. Figure 30b shows a
showerhead-type injector comprised of a large number of axially directed
propellant jets. Because there is no forced mixing due to impingement
or recirculation (both of which are provided in the enzian injector),
reaction results only by jet breakup, followed by radial diffusion of the
droplets and evaporation and diffusion of the propellant due to heat
release from the initial small amount of random mixing and combustion.
Because of the nature of the process, it is almost mandatory that the
number of jets be kept very large in order to keep the mixing scale small
with respect to chamber dimensions. T h e dependence on the relatively
slow diffusional process causes the approach to the final total tempera-
[781
HEATTRANSFER FLOWS
ture to be rather gradual. The type of injector most prevalent in very

large thrust-chamber injectors also utilizes a very large number of jets
and usually has some impingement of adjacent propellant jets provided
by drilling the jets at an angle to the axis. However, there is no organized
recirculation of part of the hot gases such as results from the splash
plate of the enzian injector. T h e energy release patterns of this type of
injector are probably about midway between the enzian and showerhead
injectors. T h e injector illustrated in Fig. 28 typifies the use of impinging
jets in small thrust injectors, where, because of lower limits on the size
of practically drillable jet orifices, it becomes necessary to use only a
limited number. This results in spatially nonuniform mass-injection distri-
butions. T h e nonuniformities have been shown by Rupe and Jaivin (71) to
result in substantial circumferential gradients in heat flux, which, sur-
prisingly, persist all the way through the nozzle. A wide variety of other
injectors is in use, each with its own peculiar flow and energy-release
patterns. As will be shown, these peculiarities of the various injectors
directly influence the heat-flux distribution.
HEATFLUX
C. COMBUSTION-CHAMBER
Because of the general lack of specific quantitative knowledge of the
heat-release patterns of injectors, the usual practice in design analyses
of the cooling of rocket thrust chambers is to assume that the combustion
chamber heat flux is constant at the value calculated for the nozzle
entrance. If the closed-form approximation (Eq. (50)) is used, this is
equivalent to assuming a boundary-layer thickness at the nozzle entrance
equivalent to that which would be expected at the end of a constant-area
duct about equal in length to the nozzle (see Fig. 10, case 2a). An
example of a measured chamber heat-flux distribution which agrees
closely with such a prediction is shown in Fig. 31, where the enthalpy
driving potential specified by Eq. (82) has been employed. Since driving
potentials are not directly measured or known, all results from rocket
thrust-chamber tests are presented as direct heat-flux rather than heat-
transfer coefficients as were used for the air results. T h e propellants
used in the tests of Fig. 31 were nitrogen tetroxide-hydrazine. T h e
contraction-area ratio was 4 : 1, which is typical for thrust chambers
of a few thousand pounds. (Thrust chambers designed to operate at
smaller thrust levels generally tend to have higher contraction ratios
because of the difficulty in building a very small-diameter injector. On
the other hand, thrust chambers designed to operate at larger thrust
levels generally tend to have lower contraction ratios in order to reduce
weight.) Note that the heat flux in the segment of the chamber adjacent
1791
D. R. BARTZ
.-It
-,-
k- -INNER SURFACE
cn
2
NOMINAL CONDITIONS
P O = 144 psi0
1 I
cc = 5038
- m r = 1.01
L* = 23.7in
cC = 4:l
8-PAIR ENZIAN INJECTOR
00 u-
-1
i NOTE: ONE T E S ~INDICATED
CONSISTENTLY HIGHER 4
2 4
AXIAL DISTANCE I, in.

8 10
I
FIG. 3 I . Comparison of predicted values with experimental heat-flux measurements

made with 4 : I contraction-area-ratio nozzle at low chamber pressure (52).
HEATTRANSFER FLOWS
to the injector is essentially up to the full value of the heat flux in the
remainder of the chamber and that at the nozzle-entrance region. This
is undoubtedly the result of the very rapid initial rise in total tempera-
ture and the strong reverse flow pattern, both characteristic of the enzian
injector.
Unfortunately, agreement between predictions and measured heat
fluxes in combustion chambers such as that illustrated in Fig. 31 is not
typical. In particular, in a thrust chamber with a large contraction ratio
(like the one for which data are presented in Fig. 32 for the same propel-
4 I I
INNER SURFACE
I
I NOZZLE
THROAT
2.
Y I
PLANE OF
INJECTOR
0
AXIAL DISTANCE 1, in
FIG. 32. Comparison of predicted values with experimental heat-flux measurements

made with 8 : 1 contraction-area-ratio nozzle at low chamber pressure (52).
lants and injector as in Fig. 31), it is observed that the measured heat
flux in the chamber is nearly 100% above the prediction. While there is
no known quantitative relationship, qualitatively, this pattern of high
heat flux (relative to prediction) is generally observed for high contrac-
tion-ratio chambers for a variety of propellants and injector types. This
is probably due to the fact that local combustion-dominated flows (such
as the recirculation for the enzian injector) can have velocities far in
[811
D. R. BARTZ
excess of the cross-sectionally averaged velocity which is utilized in

making the predictions.
T h e conclusion should not be hastily drawn that the contraction ratio
is sufficient to correlate heat flux in a combustion chamber, inasmuch as
data obtained with the same propellants and injector but with a thrust
chamber having a contraction ratio of only 1.64 : 1 (presented in Fig. 33)
also show dramatic differences between predictions and experimental
observations. I n this case, the average flow velocities are so high that the
residence time of the reacting propellants is considerably reduced. Thus,
local effects due to chemical reaction (i.e., combustion) appear to persist
all the way to the throat. There is no good detailed mechanistic argument
readily available that would quantitatively predict or explain this in-
fluence on heat flux. Elevated chamber and contraction-region (even
throat-region) heat fluxes have been frequently (but not necessarily
typically) observed with low contraction ratio thrust chambers.
Figure 34 shows a chamber heat-flux distribution measured with an
enzian ejector (see Fig. 30a) compared with that of a showerhead injector
(see Fig. 30b) operating at similar conditions with the same propellants
and thrust chamber. Note the gradualriseinheat flux from the injector
to the nozzle entrance, where the values reach a level quite close to that
predicted by the closed-form approximation (Eq. (50)). This pattern
is quite consistent with the characteristics of a showerhead injector
(described in Section IV, B), i.e., a very gradual heat-release pattern and
a very small scale of flow nonuniformities. It is probable that no signi-
ficant organized combustion-dominated flows departing from the average
flow are present. I n view of the high fraction of theoretical performance
( c * ) and low chamber heat flux exhibited by this injector, it is fair to ask
why other types such as the enzian should be used. One reason is that
the number of injection orifices tends to run into the many thousands
when the injector is scaled up to large thrusts. Thus, the cost and
reproducibility of flow from jet to jet tend to become unacceptable. But
more serious is the fact that the gradual heat-release pattern which pro-
motes low heat fluxes also makes large operating thrust chambers more
susceptible to severe combustion pressure oscillations. The severe oscil-
lations are believed to be due to periodic disturbances moving through
regions of mixed- but only partially reacted-propellant gases, with
sudden energy release occurring as the disturbance passes. This wave
process is reinforced and becomes steeper to the point at which sharp-
fronted waves with pressure ratios greater than 10 : 1 have been observed,
causing some combustion research investigators to believe that continu-
ously spinning detonation waves are established. Such waves are very
damaging, principally because they tend to increase wall heat fluxes
WI
HEATTRANSFER
FLOWS
8
I
c INNER SURFACE7
I NOZZLE
rumni
NOZZLE EXIT
CHAMBER CENTERLINE
- i - L -
r
NOMINAL CONDITIONS
po = 123 psia
C. i 5 2 5 0 f t l n c
mr = 1.02
6 - L" = 16.9sin
cc = 1.64:l
8- PAIR ENZIAN INJECTOR
3
-7
s
I L,
1 ; ;
" :
II
1 7
D
I )
01.
r+ I
-
f 4
a
m.
ANALYTICAL,
METHOD OF
- - - - - .-- - - --
I
2 TESTS
I
I
I ' tz;\
NOTE. ONE TEST INDICATED

CONSISTENTLY HIGHER q
I
0 i
FIG.33. Comparison of predicted values with experimental heat-flux measurements

made with 1.64 : 1 contraction-area-ratio nozzle at low chamber pressure (52).
~831
D. R. BARTZ
NOMINAL CONDITIONS
SYMBOL - ----___
po,psio = 199 207
c?ft/sec = 5300 5600
mr = 1.02 1.03
l * i n = 17 17
6c = 1.64 164
INJECTOR FIG 30(0) FIG. 30 (b)
r\VG.OF 2TESTS 3 TESTS
c/-E~.(W)'FOR
', Po' 207psi
C- = 560Dpri
- A - - - - I
I
I
,
I
-J
0 4 B
AXIAL DISTANCE FROM INJECTOR. in
FIG. 34. Comparison of experimental heat-flux measurements obtained using enzian

and showerhead injectors ( 52, 73).
sharply beyond normal engine cooling capabilities, with spectacular

failures resulting.
Examples of the range of influences of nonlinear combustion oscilla-
tions on chamber heat fluxes are illustrated in Fig. 35a, in which the
ratio of observed heat flux to that predicted by Eq. (50) is plotted
versus distance from the injector face, multiplied by the contraction
ratio. (By use of this modification of the length coordinate, the abscissa
values become proportional to the average residence time of the propel-
lant within the trust chamber.) It is significant to note that present
understanding of the causes and effects of nonlinear combustion pressure
oscillations is insufficient to predict whether or not they will occur and,
if so, what their influence will be on temporal or spatial distributions of
~ 4 1
6.0
-5 5.0
-
4.0
U
w 30
d
\2.0
2
I.o
0
3.0
2.5
-
0 2.0
*2s 1.5
2 1.0
0.5 I mI I T I I I I I
0 0
0
(%/%) Zlnj.ctor 1 in.
FIG. 35. Distributions of experimentally measured heat flux in combustion chambers

operating at thrust levels from 1000 to 20,000 lb with several liquid propellants.
Fig. Sym. Propellants Ref. Injector A J A , F/1000 P o , mr c*/c$t TWITo

Ib psia
a 0 RFNA-An., FA 71 52, Dublet 2.03 20.0 287 3.3 "0.92 0.31-0.49
a 0 RFNA-An., FA 71 52, Dublet 2.03 20.0 31 1 2.8 60.96 0.19-0.46
a 0 N,O,-UDMH 71 47, Dublet 2.03 20.0 300 1.4 '0.99 0.22-0.38
h o N,O,-N,H, 68 8, Enzian 2.50 2.2 207 0.98 *0.93 0.19
b A N,O,-N,H, 68 8, Enzian 2.50 1.0 100 0.99 "0.93 0.19
h 0 N,O,-N,H, 52 8, Enzian 8.00 4.0 126 1.0 ".89 0.22
b a N,O,-N,H, 52 8, Enzian 4.00 4.0 144 1.0 "0.86 0.19
b B N,O,-N,H, 73 180, Showerhead 1.64 3.0 207 1.0 "0.96 0.16
b o SFNA-An., FA 74 8, Dublet 8.00 1.0 316 2.8 '0.89 0.19
c o CI F,-N,H, 69 14, Splash Cup 1.65 5.0 298 1.9 ".93 0.15
C A C1 F,-NzH4 69 14, Splash Cup 1.65 5.0 298 2.2 "0.94 0.15
c o 0,-HEPTANE 66 9, Parallel Sheet 6.15 1.8 265 ,2.3 "0.84 0.15
c o 0,-HEPTANE 66 9, Triplet 6.15 1.8 314 2.3 '0.98 0.14
~~ ~
t u, unstable; s, stable.
D. R. BARTZ
heat flux. Some basic experiments have been made to determine the
influence of oscillating pressures on heat transfer (3-5).Unfortunately,
such experiments were made with isentropic waves of amplitude com-
parable to the mean pressure and not with nonisentropic waves of
amplitude many times the mean pressure such as those believed to have
been encountered by Rupe and Jaivin (71).
I n addition, Fig. 35a, illustrates some of the circumferential variations
in heat flux (indicated by range covered by arrows) that result from the
mass-flux pattern established by the injector. By taking point measure-
ments with transient temperature plugs in a fairly large thrust engine
(20.000 lb) with a relatively coarse injector pattern, Rupe and Jaivin (71)
were able to resolve experimentally heat-flux variations of nearly 10 : 1
(near the injector), depending on local position with respect to an
individual injector-orifice pair. On some scale, relative to the injector-
element scale, variations such as these probably occur within all thrust
chambers.
Figure 35 also shows data obtained with two high-energy propellant
combinations-chlorine trifluoride-hydrazine and liquid oxygen-liquid
hydrogen. T h e special feature of combustion of these propellants is
the fact that a substantial fraction of the energy released by combustion
can become tied up in dissociating species such as hydrogen molecules
to hydrogen atoms and water into hydrogen and hydroxyl radicals (see
Section 11, E). T h e limit of the effect of the recombination of these
species in the cool,.bouridary layer can be accounted for approximately
by assuming equilibrium recombination and adopting the enthalpy
driving potential concept, the details of which are specified in Rosner (6)
and discussed in Section 11, E. This has been done in calculating the
predictions to which the experimental data of Figs. 31-35 have been
compared. I t was found, however, that the percentage increases due to
the possibility of equilibrium recombination are too small in these tests
to be verified by comparisons with chamber heat-flux data because of
the wide variety and magnitude of competing unpredictable influences
of the combustion. In fact, the principal point to be established by the data
and the discussion of this section is that the reliable prediction of heat flux
in a combustion chamber is not possible because of the inadequate status
of quantitative knowledge of rocket-engine combustion phenomena. Until
sufficient quantitative knowledge is gained, attempts at refined chamber
heat-transfer calculations are not justified. T h e limit of useful analysis
is probably calculations from some simple correlation equation such as
Eq. (50), modified by qualitative and intuitive arguments about the
combustion flow as characterized by the injector, propellants, and
operating conditions of interest.
P I
HEATTRANSFER FLOWS
D. NOZZLEHEATFLUX
Because of the severe acceleration that occurs in the nozzle, and the
increasing distance from the injector, it might be argued qualitatively
that the direct influences of combustion such as secondary flows and
heat-release distribution should begin to decay as the acceleration com-
mences at the nozzle entrance, and should be largely decayed by the point
of maximum acceleration, the throat. This qualitative expectation is
supported to some extent by the experimental nozzle heat-flux distribu-
tions from Welsh and Witte (52), presented in Figs. 31-33 of this chapter.
Even though there are considerable qualitative differences in the heat-
flux distributions in the contraction region and the region just ahead of
the geometrical throat, the distributions beyond the throat all compare
quantitatively quite well with the prediction based upon the closed-form
approximation equation (Eq. (50)). I n the expansion region [out to the
limited area ratio utilized in the tests of Welsh and Witte (52) about 3.51,
the experimentally measured heat fluxes all fall within a band from 80
to 100% of the value predicted by Eq. (50) at corresponding area ratios
and most within an even tighter band, i.e., 85-90% of the prediction.
Although plots such as those of Figs 31-33 are useful for conveying
an idea of the distributions and numerical values of the heat flux, they
are not very satisfactory for generalizing the results or for ready compari-
sons with predictions over a range of conditions. For this purpose, it is
appropriate to turn back to the dimensionless parameter presentation
such as that used to correlate the data from the air experiments. I n
particular, the parameter C, Pro.s is plotted versus R, for the reasons
described in Section 111, D, 3. As in the case of the air data, these para-
meters are computed on the basis of evaluating the properties at the
free stream static temperature. I n order to reduce the experimental
heat-flux values to heat-transfer coefficients, computations of the driving
potential were made on the enthalpy, difference basis (Io - I,,,) des-
cribed in Section 11, E. Account was taken of actual reductions in per-
formance below the theoretical value, but is was assumed that the com-
bustion process was completed by the time the flow entered the nozzle.
For comparison with these data, presented in Figs. 36-39, lines are drawn
corresponding to (1) the predictions from Eq. (50) (based on arithmetic
mean temperature evaluation of properties) for a value of Tw/Totypi-
fying the data of the particular figure; and (2) the predictions from Eq.
(48), with Trefselected as the adiabatic wall temperature T a w .
Attention is first called to the part of each figure corresponding to the
highest subsonic area ratio for which data are reported for each of the
four different contraction-area-ratio nozzles. They have in common the
~ 7 1
D. R. BARTZ
-
3
4 2
pt c- f
10-3
1 15
." 10
FIG. 36. Dimensionless parameter correlation of heat-transfer coefficients in 8 : 1

contraction-area-ratio nozzle operating with Na04-N,H4propellants (68).
A J A , = 8 nozzle, L* = 63 in. p o = 77 - 292 psia
NZO, - NZH,; = 0.98 - 1.03 Eq. (50):--, Trer = Tarn,Tw/TD= 0.18
C*/C; = 0.89 - 0.96 Eq. (48): - -, Trer = Taw
fact that the slopes of the curves drawn through the experimental data
of all four of these nozzles is quite close to that of the prediction equation;
i.e., C,, PrO.6 REO.~.
N For all but the 2.5 : 1 nozzle (Fig. 38), the data
exceed the predicted values. T h e experimental values for the highest
subsonic-area-ratio plot for the 8 : 1 area-ratio nozzle lie considerably
P I
HEATTRANSFER
FROM RAPIDLY
ACCELERATING
FLOWS
SUBSONIC SUPERSONIC
10-3
0.3 0.4 0.6 0.8 106 2 3 0.3 0.4 0.6 0.8 I06 2 3
k y Y
FIG.37. Dimensionless parameter correlation of heat-transfer coefficients in 4 : 1
contraction-area-ratio nozzle operating with N,O,-N,H, propellants (68).
A J A , = 4, L* = 24in. po= 99 - 267psia
N z 0 4 - N,H4; mr = 0.96 - 1.02 Eq. (50): -, Trer = Tam, T,/To = 0.20
c*/c,*, = 0.85 - 0.89 Eq. (48): - -, Tref = T,,,
above the predicted line for reasons discussed in Section IV, C. Notice,
however, that as the flow proceeds to lower subsonic area ratios, the lower
R , (Lea, lower total pressure) data begin to drop off considerably with
respect to the prediction lines. As the R , increases, value of C , Pro.s
rise abruptly relative to the prediction line. At still larger R, , they begin
to fall off gradually with increasing R , in most cases reaching what
~ 9 1
D. R. BARTZ
tq
h
3
10
FIG. 38. Dimensionless parameter correlation of heat-transfer coefficients in 2.5 : 1

contraction-area-ratio nozzle operating with N204-N2Hdpropellants (68).
A J A , = 2.5 nozzle, L* = 35.8 in. Po= 10&301 psia
N2Oa-NZH4; = 0.99 Eq. (50); -, Tref = Tam, T w / T ,= 0.20
c * / c i = 0.93-0.94 Eq. (48): - -, Trer = Tow
appears to be the start of a -0.2-power dependence region. Note that

there is evidence of this behavior out to even the highest supersonic-area
ratios. This behavior is significantly similar to that exhibited by the data
from the hot-air tests in Fig. 23. Perhaps this results from the same
mechanism proposed to explain the behavior of the air data; i.e., at
lower R , (lower stagnation pressures) turbulent boundary layers entering
[901
HEATTRANSFER FLOWS
xh
i:
10-
I I I I I I
IO- I
0.6 0.8 10 2 3 5 2 3 5
-
P
P
-
PUO
P
FIG.39. Dimensionless parameter correlation of heat-transfer coefficients in 1.64 : 1

contraction-area-ratio nozzle operating with N,O,-N,H, propellants (68).
A i / A , = 1.64 nozzle, L* = 17 in. p o = 97-246 psia
Nz02-N2HII 0.98 Eq. (50): -, Tkr = T,, , Tw/To= 0.28
c*/c& = 0.91-0.93 Eq. (48): --, Trer == Taw
the nozzle are worked upon by the severe acceleration, decaying the
turbulence near the wall by some turbulence decay mechanism (such as
discussed in Section 111, E), resulting in a reduction in the wall gradients
of the velocity and temperature back toward those characteristic of
transitional fldw. At higher R, , i.e., higher stagnation pressure, the
~911
D. R. BARTZ
turbulence decay mechanism is not as strong (as shown in Fig. 27). Thus,
at higher RD , the acceleration does not substantially reduce the wall
gradients and, hence, the heat flux. It is granted that this argument and
proposed mechanism are speculative, being based on limited data, all
obtained with the same propellants and injector. Injector performance
and flow characteristics are known to be typically quite flow-rate
sensitive and hence, sensitive to stagnation pressure. T h e principal
reason for suggesting the mechanism at all is the remarkable qualitative
similarity between these rocket-nozzle results and the air results pre-
sented in Fig. 23, and the data obtained by Back (62) with accelerated
low-speed wind-tunnel flow. Unfortunately, there are no data available
from other rocket tests which cover a sufficient range of operating
conditions such that a useful comparison could be made.
I n order to illustrate how the heat-transfer coefficient varies at higher
area ratios (and Mach numbers), typical data of Witte and Harper (68)
are plotted in Fig. 40 and are compared with boundary-layer predictions
(Eq. (A.8)) as well as the closed-form approximations (Eq. (50) with
Tret= Tamand Eq. (48) with Trer= Tau,). For each of the boundary-
layer heat-transfer predictions, the initial boundary-layer thicknesses
at the nozzle inlet were selected so as to yield a heat flux equal to the
average of the experimental data in the entrance region. The test selected
at 300 psi stagnation pressure is at sufficiently high RD to preclude
behavior of the type discussed in the previous paragraph. For a conical
expansion section such as that used in the test from which data these
were obtained, the deviations of p U from p1 U,are negligible except very
near the throat. T o facilitate the comparisons, the data have also been
normalized by dividing by the local h,-value predicted by Eq. (50) in
the lower part of the figure, as have the other predictions. Except for a
few passages believed to have had erratic thermocouples, the data follow
a more or less smooth decline from about 25% above the type-C
predictions just beyond the throat to very close agreement with the
type-C or -D predictions at high area ratios. This behavior is closely
similar to results of other high area-ratio data from rocket tests at lower
stagnation pressures presented in Witte and Harper (68). I t is also
reasonably similar to that of the air data of Figs. 19 and 20 over the
limited range of supersonic-area ratios at which data were obtained.
T h e convergence of the high Mach-number data toward the type-C or
-D boundary-layer prediction suggested the possible value of a closed-
form approximation to this prediction. This was found to be readily
possible over the expansion region of the nozzle (where the boundary-
layer growth has become regular) by adopting Tawas Trefin evaluating
hg from Eq. (48) (see Section 11). As is evident from the bottom portion
r921
HEATTRANSFER
FLOWS
E
NOZZLE
l I
c THROAT 4
z0 INNER
In
2 -
PLANE OF
INJECTOR
I
0
2 NOZZLE ENTRANCE-
I
2 4 1
a
3
CHAMBER - / NOZZLE EXIT-
3 C E N-T-
E R L-I N
- E
-.0
-, -I -
20 24 28 32 36
AXIAL DISTANCE x , in.
I .s
I.o
0. s
0
0 4 0 12 16 20 24 28 32 36
I AXIAL DISTANCE FROM INJECTOR c. in
FIG.40. Comparison of measured and predicted heat fluxes for N,0p-N2H, at

pa = 301 (68).
D. R. BARTZ
of Fig. 40, the agreement with the type-C and -D boundary-layer

h,-prediction over this portion of the nozzle is excellent.
I n the contraction and throat regions, the air data and the rocket data
behave quite differently, the air data being equal to or below the type-C
boundary-layer prediction and the rocket data exceeding the prediction
by as much as 40%. If one were forced to speculate about a mechanism
that would account for this situation, one might postulate that in the
rocket nozzle flow there is superimposed upon the normal convective
processes an added process, such as free-stream turbulence, which
originates in the combustion chamber and which decays in the stream-
wise direction. It is well known that even for rocket engines operating
without organized combustion pressure oscillations, a so-called com-
bustion noise is frequently present having pressure oscillation amplitudes
of several percent of the mean pressure without regular or dominant
frequencies. (These free-stream flow fluctuations are probably related
to free-stream turbulence phenomena,) Combustion noise of this
character was usually encountered in tests in which the rocket data
discussed in this section were obtained. Although these noise-like
pressure oscillations are almost always monitored only by transducers
back near the injector, it is very likely that the magnitude of the oscilla-
tions decays in the flow direction. Although neither the status of theory
nor basic experiments on the influence of free-stream turbulence on
heat-transfer (75-77) are sufficiently well advanced to make an authori-
tative prediction of the influence of the free-stream turbulence, the 40 yo
elevation of the rocket data above the air data is not inconsistent with
influences of free-stream turbulence observed by others (76). Because
of the large, irregular influences on heat transfer in the contraction
region believed possible due to combustion-dominated secondary flows,
one cannot readily differentiate between such effects and those of the
free-stream turbulence in this region.
T h e rocket data of Fig. 40 clearly show an asymptotic preference
toward predictions made on the basis of evaluating properties at the
adiabatic wall temperature, supporting the arguments and evidence
presented in Section 11, D. Unfortunately, the air data of Back et al. (59),
which also tend to favor the same prediction basis, were restricted to low
area ratios, where other factors at work make conclusive comparison diffi-
cult. T h e air data presented in Fig. 26 (from Fortini and Ehlers, 58) were
obtained out to high expansion ratios and show a decided preference for
predictions based on the film-temperature property evaluation. There
is no explanation obvious to this author for the distinct difference
between these particular air data and the rocket data at high area ratios.
Thus, the variable properties question is not resolved by these data.
WI
HEATTRANSFER
FROM RAPIDLYACCELERATING FLOWS
Powell et al. (69, 70) has measured the distribution of local heat transfer
in rocket engines operating with propellants of sufficient energy content
to make considerations of the effects of recombination on heat transfer
(see Section 11, E) potentially important. Unfortunately, the fraction of
theoretical performance achieved in these investigations thus far
(90-95% c*) has not been sufficient to cause predicted chemical recom-
bination enhancement of heat transfer to exceed about lo%, which is
not readily discernible within the experimental scatter and the other
influences on heat transfer. Consequently, the measured heat-transfer
distributions of Powell et al. (69, 70) are quite similar to those of
Figs. 31-33 with respect to predicted values.
Although this chapter has been restricted to the problems of heat
transfer from the hot gases to the cooled walls of nozzles and chambers,
the reader should also be aware of the interesting and often design-
limiting problems associated with the wall cooling. I n the particular case
of cooling with hydrogen, the two problems become closely coupled and
must be solved simultaneously. Correlations of the heat-transfer coeffi-
cients of hydrogen as they apply to the rocket-engine cooling problem
have been discussed by Benser and Graham (78). I n the case of cooling
with subcooled liquids, it is frequently advantageous to make use of
nucleate boiling. As discussed (79), this regime of operation is character-
ized by a fixed and easily determined wall temperature, thus decoupling
the problems to a large extent.
V. Concluding Remarks
As should be clear to the reader from the evidence presented in this
chapter, the problem of heat transfer from hot gases to the cooled walls of
nozzles is not yet solved in the sense of producing a theory or an
empirical correlation universally accurate for all the flows of interest.
Nevertheless, the analyses and experiments that have been discussed
have accomplished two things: (1) they have suggested the specific
basic physical processes that must be understood quantitatively before
the problem can be solved; and (2) they have shown how far existing
methods of analysis may be in error under the limited conditions thus
far investigated thereby establishing some basis for corrections to these
analyses that will permit an acceptable prediction of the heat transfer
for most requirements. T h e delineation of the processes that must be
undqrstood to solve the problem and the specification of procedures
recommended for nozzle heat-transfer predictions at this time conclude
this chapter.
P51
D. R. BARTZ
Of basic importance to all turbulent boundary-layer flows are two

problems brought into sharp focus by the cooled-nozzle heat-transfer
problem. First, the question of the influence of variable properties
on heat transfer to severely cooled walls must be settled by definitive
experiments in which local conditions, including velocity and tempera-
ture profiles, are measured over a range of velocities extending into the
supersonic region and over a range of stagnation pressures. Second, the
influence on heat transfer of both free-stream turbulence and gross
disturbances to the boundary layer by secondary flows must be investi-
gated and understood to such an extent that their effects can be predicted
with reasonable assurance from known or specified initial conditions.
T h e presence of a strongly favorable pressure gradient, of the magni-
tude peculiar to nozzle flows, adds another potentially significant, yet
largely uninvestigated, basic dimension to boundary-layer flow. In
particular, direct shear measurements are required to determine whether
or not it is reasonable to assume that for the same local momentum-
thickness Reynolds number, the skin friction of an accelerating turbulent
boundary layer is the same as without a pressure gradient. T h e effect
of strongly favorable pressure gradients on turbulence production and
decay within the boundary layer must be understood and somehow
correlated to make the occurrence of both forward and reverse transition
predictable. Some means of correlating skin friction in the transition
region must also be found. For flows with strongly favorable pressure
gradients, where, as indicated by predictions, the thermal boundary
layer can become substantially thicker than the velocity boundary layer,
it is necessary to find either (1) some modified Reynolds analogy by
gaining new insight into the turbulent form of the boundary-layer
equations with pressure-gradient terms retained, or (2) some general
correlation of the Stanton number in terms of the local energy thickness
irrespective of past history of the boundary layer. Finally, fluid mechanics
and thermodynamics of the rocket combustion process must become
quantitatively understood and made both predictable and describable
on an instantaneous, local basis in order to make significant improvement
in the prediction of the heat flux in the combustion chamber and
contraction region.
As for recommended procedures for making predictions of rocket-
nozzle heat transfer, it is probably worth the effort to employ one of the
iterative boundary-layer solutions of the type described in Section 11, A
for radically new nozzle configurations or peculiar entrance or operating
conditions for which such solutions are not available. For such calcula-
tions, one should attempt to account for deviations from one-dimen-
sional flow by method-of-characteristics calculations in the supersonic
[961
HEATTRANSFER
FLOWS
region and use of Eq. (83) in the transonic region. For further predictions
for only slightly different configurations or conditions, especially when
an answer is required in a hurry, the closed-form approximation
equations can be used with reasonable precision by adjusting the constant
C to fit the previously obtained boundary-layer results. T h e question
remains, however, as to which boundary-layer calculation method and
which closed-form approximation should be used and how these com-
puted results might relate to the heat fluxes to be anticipated in the
rocket nozzle. Since data presented and discussed in this chapter failed
to substantiate conclusively either of the variable properties correction
methods described, the choice on this critical question remains arbitrary.
The author tends to favor using the adiabatic wall temperature Tawas
the reference temperature in the closed-form approximation equation
(Eq. (48)), or the closely equivalent procedure of utilizing the Coles
(25) C,for an adiabatic wall in the boundary-layer calculation (Eq. (A.8)).
T h e reason for this selection is threefold: (1) the pipe and flat-plate data
discussed in Section 11, D suggesting its basic validity; (2) the agreement
at high velocities with the rocket data of Fig. 40;and (3) the way this
assumption fits in with the hypothesis of decaying free-stream turbulence
(Section IV, D). I n order to relate predictions on this basis to expected
heat fluxes in a rocket nozzle, it is probably necessary to start by multi-
plying the values in the transonic region by a factor of about 1.3 to 1.4
and to decrease this factor gradually to unity at stations at which the
Mach number is about 4. I n the chamber and the contraction regions,
the values require multiplication by some factor arrived at intuitively,
based upon arguments about the nature of the combustion influences,
as discussed in Section IV, C. A factor of 1.3 to 1.8 is probably a reason-
able starting range around which adjustments can be made as the com-
bustion flow situation may demand. For combustion gases in which a
substantial fraction of the total energy is tied up in dissociation, account
should be taken of the possible recombination enhancement of heat
transfer by adjustments to the driving potential, as discussed in Section
11, E.
ACKNOWLEDGMENT
The author is grateful to P. F. Massier, L. H. Back and D. E. Rosner for their detailed
review of the manuscript and their constructive suggestions. In addition, it is a pleasure to
acknowledge the considerable contributions made to the rocket heat-transfer problem by
the authors colleagues past and present, including L. H. Back, W. J. Colahan, D. G.
Elliott, G . W. Elverum, H. L. Gier, E. Y. Harper, D. T. Harrje, G. I. Jaivin, P. F.
Massier, M. B. Noel, W. B. Powell, T. W. Price, R. W. Rowley, J. H. Rupe, W. E. Welsh,
and A. B. Witte.
1971
D. R. BARTZ
Appendix A
COLES' SKIN-FRICTION AND VON KARMANFORM

COEFFICIENT OF THE
REYNOLDSANALOGY
Coles (1.5) has shown that most of the carefully measured skin-friction
coefficient data from adiabatic flow over flat plates over a wide range of
Re and M can be correlated to within a few percent by a single curve
of ef versus Cf&. These low-speed values of c, and R, are related to
the actual Cf, and Re values by
and
where subscript aw refers to the adiabatic wall or recovery temperature

and subscript s refers to a sublayer temperature given by
T h e values of c, versus &!,I?! utilized by Coles are plotted in Fig. 3,

where they are compared with the Blasius equation. They are also
tabulated in Table A. 1, For values of ,& above 64.8, the extrapolation
,Re G C188 CI
2.51 0.00590 16.36 0.00290
3.10 0.00524 23.2 0.00269
3.97 0.00464 29.6 0.00255
4.88 0.00426 35.9 0.00246
5.73 0.00398 41.8 0.00238
7.41 0.00363 53.6 0.00227
8.94 0.00340 64.8 0.00219
12.75 0.00308
a From Coles ( I 5).
curve given by
(-$)* = 2.441n [ cf C,Ri

---)
(3.781 - 25.104 + 7.68 (A.4)
(2rfY
HEATTRANSFER FLOWS
can be used. Values of C,for ,R, below 2.51 can be computed from
0.009896
-- ( c f & ) 0 . 5 6 2
I n this region, the Ro are below those normally associated with tur-
bulent flow. Equation (AS) is used simply to get an order of magnitude
Utilizing a power relation for viscosity, p -

value of C, with which to start a thin-entrance-condition calculation.
T", Eqs. (A.l) and (A.2)
can be put into a more convenient form given, respectively, by Eq. (33)
and by
T h e form of the Reynolds analogy adopted to obtain the calculation

results presented in this chapter is based on Assumption 13, that when
4 = 8, C, is related to C,/2 by the von KBrmBn form of the Reynolds
analogy given by Eq. (38). For 4 = 8, Eq. (38) can be rewritten directly
as
where R , is the parameter pU+/p By the intuitive reasoning and the

comparison with data given in Section 11, A.6, Eq. (A.7) is modified for
cases in which d, # 8 by multiplying (A.7) by (+/8)n,giving
Ch =
(cf(Rm(4ie).
(A4
1-5(7) CAR,) [I-Prfln--
5Pr 6+ 1 l
where the notation C,(R,) denotes a value of C, determined either from
Eq. (33) (for adiabatic wall C,) or from Eq. (36) (for film properties C,),
C, having been evaluated from Table A.l, Eqs. (A.4) and ( A . 9 , with
RQreplacing RB.
Appendix B
BOUNDARY-LAYER
SHAPEPARAMETER
EVALUATION
I n order to calculate the boundary-layer shape parameters 6*/8,
O j8 , + / A , etc., it is necessary to specify velocity distributions over the
P I
D. R. BARTZ
velocity boundary-layer thickness 6 and temperature distributions over

the temperature boundary-layer thickness A . T h e distributions adopted,
in accordance with Assumption 15, are
U = y < 6,
and
By rearrangement of Eq. (B.2), making use of the isentropic relation-

ship between static and stagnation temperature and the velocity distri-
bution of Eq. (B.l), the local density distribution is given by
i = =P= = a
P
where
I n evaluating the shape parameters, cognizance must be taken of two

special cases in which Eq. (B.3) must be modified.
b
C a s e I : when S G A , S<y<A, i/T=a I+-s-c) (B.4)
( 1 ;
Case II: when S ,: A , A <y < 6, i / T = a(1 + b - cs2) (B.5)
Note that the Crocco temperature distribution is a special case of Eq.
(B.3), in which 5 = 1, the situation for flow over a flat plate, with Pr = 1.
From the integral definition of 0 for a thin boundary layer given by
Eq. (12), it can be shown that
e
= a Il
7
for Case I ,
0 7
- = - (I,
6 a
+ I,) for Case II (B.6)
where the definite integrals I are given by

HEATTRANSFER
FLOWS
Similarly, from Eq. (1 5 ) ,
1S = 757 ( [ I ; + 131) for Case I , a_6 = 3-1,' for CaseII (B.7)

a a
where
From Eqs. (B.6) and (B.7),
for Case II
(B.8)
Finally, from Eq. (9),
s* 7
- = 5 7 - - ( I , + 13) for Case I ,
6*
-= 1 - - (I6 + I,)
7
for Case II
S a S a
(B.9)
where
I ?! -- S:W/a.- & ds,
's
13 = I' +
1 1
S6
(b/& -c
ds
Appendix C
BOUNDARY-LAYER
THICKNESSES
AND INTEGRAL
EQUATIONS
FOR THICK
BOUNDARYLAYERS
If now the restriction is removed that the boundary-layer thicknesses
S and d must be small with respect to Y, it can be shown that the dis-
placement thickness 6 * has the same physical meaning as it did for a
thin boundary layer (Eq. (8)) when S* is related to a new integral given by
--
6*(1 --) s* cos a: = j y ( l - ) P U( 1 y cos a
- 7 ) d y (C.1)
2r PU
where N is the angle the wall makes with respect to the centerline at the
local station z. Making use of this thick boundary-layer definition of S*, it
Poll
D. R. BARTZ
can be shown that the deficiency of momentum of the real flow near the
wall compared with the same mass flux of potential flow is
ni, - ni, = 2 n p ~ z y(1e - 2Y
(C.2)
O P
and similarly with the enthalpy flux of the real flow near the wall,
H , - ilT= 2npUc,(To - T,)Y(b ( 1 -

2r
--
= 2npucy(TQ- TJY jBr
3( 1 --io) - T w
PU -
o f To - T,,
y cos a
a ( 1 --)dy
Y
(C.3)
From these equations, it can be shown that the integral momentum and
energy equations are identical to those for the thin boundary layer
+
(Eqs. (25) and (30)) if 8,6*, and are replaced by 8, S*, and +, where
S* = 6*
( 1 ----)6* cos
2Y
a , e=e(i--), e cos
2Y
+
+ = + ( I - - - - ) cos L11
2Y
(C.4)
NOMENCLATURE
A Local cross-sectional flow area C,, Adiabatic skin-friction coefficient,

A, Cross-sectional flow area of com- i.e., for T, = To,
bustion chambers C,(Rd) Skin-friction coefficient based on
A, Throat cross-sectional flow area, R b , Eq. (A.8)
(7/4)D* Stanton number, Eq. (17)
a Temperature-ratio parameter, Eq. Temperature-ratio parameter, Eq.
(B.3) (B.3)
a1 Speed of sound based on one- Specific heat of gas
dimensional flow calculation of T Specific heat per unit mass of
a, Speed of sound for stagnation mixture C kit,(
conditions Characteristic velocity, p o g A , /w
b Temperature-ratio parameter, Eq. Local diameter
(B.3) Nozzle-throat diameter
C Coefficient in closed-form approxi- Coefficient of mass diffusion of
mation, Eq. (47) species i into gas mixture
C, Skin-friction coefficient, Eq. (1 6) Coefficient of mass diffusion of
(?, Low-speed adiabatic skin-friction species 1 into species 2
coefficient Thrust
[lo21
HEATTRANSFER
FLOWS
g Gravitational constant Exponent of temperature depend-

H Enthalpy flux of wall layer ence of viscosity
h* Heat-transfer coefficient based on Mass flux of wall layer, Eqs. (1)
temperature driving potential and (4)
hi Heat-transfer coefficient based Mixture ratio, oxidizer to fuel by
on enthalpy driving potential, weight
h*lc, Boundary-layer interaction expo-
I Static enthalpy per unit mass at nent
edge of boundary layer Variable coefficient in Eq. (44)
I, Frozen static enthalpy per unit Prandtl number, pc,/A.
mass at edge of boundary layer, Static pressure
X Ki JTcni dT Stagnation pressure
dIChemChemica1enthalpy per unit mass, Variable coefficient in Eq. (44)
I:(Kj - Kiw)Iio Local heat-transfer rate across
Enthalpy of formation per unit streamlines within boundary layer
mass at standard conditions Heat flux to wall
&kinetic Portion of total enthalpy per unit Turbulent kinetic energy, Eq. (84)
mass represented by kinetic Adiabatic recovery factor for kinetic
energy, $UB energy, Eq. (18)
Total or stagnation enthalpy per Chemical recovery factor
unit mass at edge of boundary Reynolds number based on D,
layer PUDlP
Enthalpy per unit mass of gas Reynolds number, p U x f p
mixture at Tw Reynolds number based on mo-
... I , Definite integrals in boundary- mentum thickness, pUB/p
layer shape parameter expres- Low-speed Reynolds number based
sions, Eqs. (B.6)-)(B.9) on momentum thickness,Eq.(A.2)
Complete local static enthalpy per Parameter based on energy thick-
unit mass of mixture of gases ness, PU+/p
within boundary layer, C k,ii Radius of wall from axis of sym-
Local static enthalpy per unit mass metry
of ith species within boundary Radius of curvature of nozzle throat
layer Throat radius
Mass fraction of ith species at edge Dummy variable in boundary-
of boundary layer layer shape parameter integrals,
Mass fraction of ith species at wall Eq. (B.3)
Mass-flow rate exponent, Eq. (46) Static temperature at edge of
Local mass fraction of ith species boundary layer
within boundary layer Adiabatic wall temperature, Eq.( 18)
W. K. Lewis number pD,&,/A Mixed mean or bulk temperature
Characteristic length (combustion Injection temperature of reactants, j
volume/d, ) Log-mean bulk temperature (46)
Nozzle length along axis Stagnation temperature
Combustion chamber or approach Thermodynamic reference tempe-
duct length rature, 298K
Mach number at edge of boundary Sublayer temperature in Coles
layer transformation, Eq. (A.3)
Momentum flux of wall layer, T W Wall temperature
Eqs. (2) and ( 5 ) t Local static temperature within
Molecular weight of gas mixture boundary layer
D. R. BARTZ
Local time-mean stagnation tem- h Thermal conductivity
perature in boundary layer P Viscosity at edge of boundary layer
Velocity at edge of velocity bound- PO Viscosity of gas at stagnation con-
ary layer ditions
Local time-mean x-component of Pa Viscosity of gas evaluated at T,
velocity in boundary layer V Kinematic viscosity
D i m e n e l e s s velocity parameter, P Gas density at edge of boundary
ti/ d\/.PlP layer
Components of velocity parallel and P Local time-mean density in bound-
normal to surface, respectively ary layer
Local scalar diffusion velocity of 0 Variable properties correction fac-
ith species within boundary layer tor, Eqs. (49) and (58)
Dummy variable in boundary-layer 0 General variable properties cor-
shape parameter integrals, Eq. rection factor, Eq. (55)
(B.7) Retarding wall shear stress
W Weight rate of flow per unit time Energy thickness of boundary
through nozzle layer, Eqs. (l4), (1 5)
X Distance along wall Thick boundary-layer energy thick-
Y Distance from wall along normal ness parameter, Eq. (C.4)
Y+ Dimensionless distance from wall, X Turbulence decay parameter, Eq.
(Y GGG (86)
z Distance along axis of symmetry
Ly Angle between wall and axis at
SUBSCXIPTS
station z
Y Ratio of specific heat at constant
pressure to that at constant volume am Arithmetic mean
A Thickness of temperature boundary aw Adiabatic wall
layer b Property evaluated at mixed mean
6 Thickness of velocity boundary or bulk temperature T6
layer f Frozen, evaluated based on
6 Thickness of layer containing all absence of chemical change
wall effects 1 ith species (products)
6* Displacement thickness of bound- j j t h species (reactants)
ary layer, Eqs. (8) and (9) 0 Entrance or initial condition
Thick boundary-layer displace- P Potential flow
ment thickness parameter Eq. R Recovery condition
(C.4) r Real flow
Contraction ratio A J A , ref Property evaluated at Trer = T*,
Boundary-layer thickness ratio T a m 9 or T a w
(A/6*)/ ref Thermodynamic reference state
8 Momentum thickness of boundary W Wall
layer, Eqs. (11) and (12) 0 Stagnation
e Thick boundary-layer momentum- 1 One-dimensional flow value
thickness parameter, Eq. (C.4) ( ) Fluctuating quantity
HEATTRANSFER
FLOWS
REFERENCES
1. R. W. Rowley, An experimental investigation of uncooled thrust chamber materials
for use in storable liquid propellant rocket engines. Tech. Rept. No. 32-561, Jet
Propulsion Laboratory, Pasadena, California, January 20, 1964.
2. E. P. Bartlett, Thermal protection of rocket-motor structures. Aerospace Eng. 22 ( I ) ,
86-99 (1963).
3. R. C. Saunders and D. T. Harrje, Heat transfer in oscillating flow. Progress Rept.
No. 3, Princeton Univ., Aeronautical Eng. Dept., Princeton, New Jersey, October 1,
1960 through September 30, 1962.
4. D. T. Harrje, Effects of oscillating flow on heat transfer in a tube. Progress Rept.
No. 20-362, Jet Propulsion Laboratory, Pasadena, California, August 27,
1958.
5. C. E. Feiler and E. B. Yeager, Effect of large-amplitude oscillations on heat transfer.
NASA T R R-142, 1962.
6. D. E. Rosner, Convective heat transfer with chemical reaction. Aerochem Research
Laboratories, ARL 99, Part I, August, 1961.
7. W. H. McAdams, Heat Transmission, 3rd ed. McGraw-Hill, New York,
1954.
8. M. Sibulkin, Heat transfer to an incompressible turbulent boundary layer and
estimation of heat transfer coefficients at supersonic nozzle throats. J . Aeron. Sci.
23 (2), 162-172 (1956).
9. D. R. Bartz, An approximate solution of compressible turbulent boundary-layer
development and convective heat transfer in convergent-divergent nozzles. Trans.
A S M E , 77 (S), 1235-1245 (1955).
10. M. Tucker, Approximate calculation of turbulent boundary-layer development in
compressible flow. NACA T N 2337, April 1951.
11. G. E. Gomf, Supersonic nozzle design for viscous fluids. Thesis in Aeronau-
tical Engineering, California Institute of Technology, Pasadena, California,
1949.
12. N. Tetervin, Approximate formulas for the computation of turbulent boundary-
layer momentum thicknesses in compressible flows. NACA Wartime Rept. No. L119,
March 1946.
13. D. R. Bartz, A simple equation for rapid estimation of rocket nozzle convective
heat transfer coefficients. Jet Propulsion 27 (I), 49-51 (1957).
14. D. G. Elliott, D. R. Bartz, and S. Silver, Calculation of turbulent boundary-layer
growth and heat transfer in axi-symmetric nozzles. Tech. Rept. No. 32-387, Jet
Propulsion Laboratory, Pasadena, California, February 15, 1963.
15. D. E. Coles, The turbulent boundary layer in a compressible fluid. Rept. No. P-2417,
Rand Corporation, Santa Monica, California, August 22, 1961.
16. A. H. Shapiro, The Dynamics and Thermodynamics of Compressible Fluid Flow,
Vol. 11. Ronald Press, New York, 1954.
17. H. Ludwieg and W: Tillman, Investigation of the wall-shearing stress in turbulent
boundary layers. NACA T M 1285, May 1950.
18. R. K. Lobb, E. M. Winkler, and J. Persh, Experimental investigation of turbulent
boundary layers in hypersonic flow. J. Aeron. Sci. 22 (l), 1-10 (1955).
19. W. C. Reynolds, W. M. Kays, and S. J. Kline, Heat transfer in the turbulent incom-
pressible boundary layer with constant wall temperature. NASA Mem. 12-1-58W,
December 1958.
D. R. BARTZ
20. H. Wolf, The experimental and analytical determination of the heat transfer charac-
teristics of air and carbon dioxide in the thermal entrance region of a smooth tube
with large temperature differences between the gas and the tube wall. Ph.D. thesis,
Purdue Univ., Lafayette, Indiana, March 1958.
21. H. E. Zellnik and S. W. Churchill, Convective heat transfer from high-temperature
air inside a tube. 3. Am. Znst. Chem. Engrr. 4 (I), 37-42 (1958).
22. E. R. G. Eckert, Engineering relations for heat transfer and friction in high-velocity
laminar and turbulent boundary layer flow over surfaces with constant pressure
and temperature. Trans. ASME, 78, 1273-1283 (1956).
23. E. R. G. Eckert and R. M. Drake, Jr., Heat and Mass Transfer. McGraw-Hill,
New York, 1959.
24. R. A. Seban and H. W. Chan, Heat transfer to boundary layers with pressure gradients.
University of California, Institute of Engineering Research, Berkeley, California.
Contract No. AF33(616)-348, WADC T R 57-11], ASTIA AD 118075, July
1957.
25. S. S. Kutateladze and A. I. Leontev, Drag law in a turbulent flow of a compressible
gas and the method of calculating friction and heat exchange. Akad. Nauk Belorusrk.
SSR, Minsk, USSR pp. 1-23 (1961).
26. D.R. Bartz, in Transport Properties in Gases, pp. 105-1 18. Northwestern Univ.
Press, Evanston, Illinois, 1958.
27. D. R. Bartz, Prediction of thermal properties. 3. Heat Transfer C82, (3), 162-166
( 1 960).
28. J. Hilsenrath, C. W. Beckett, W. S. Benedict, L. Fano, H. J. Hoge, J. F. Masi,
R. L. Nutall, Y. S. Touloukian, and H. W. Woolley, Tables of thermal properties
of gases. Natl. Bur. Std. (U.S.), Circ. 564 (1955).
29. Thermodynamic and Transport Properties of Gases, Liquids and Solids. McGraw-
Hill, New York, 1959.
30. J. F. Masi and D. H. Tsai, eds., Progress in International Research on Thermo-
dynamic and Transport Properties. Academic Press, New York, 1962.
31. R. A. Svehla, Estimated viscosities and thermal conductivities of gases at high tempe-
ratures. NASA T R R-132, 1962.
32. L. Andrussow, in Progress in International Research on Thermodynamic and
Transport Properties (J. F. Masi and D. H. Tsai, eds.), pp. 279-287. Academic
Press, New York, 1962.
33. R. S. Brokaw, Alignment charts for transport properties. Viscosity, thermal con-
ductivity, and diffusion coefficients for nonpolar gases and gas mixtures at low
density. NASA T R R-81, 1961.
34. J. W. Budenberg and C. R. Wilke, Calculation of gas-mixture viscosities. Znd. Eng.
Chem. 41, 1345-1347 (1949).
35. J. 0. Hirschfelder, C. F. Curtiss, and R. B. Bird, Molecular Theory of Gases and
Liquids. Wiley, New York, 1954.
36. C. F. Bonilla, S. E. Greer, and E. A. Taikeff, in Thermodynamic and Transport
Properties of Gases, Liquids and Solids, pp. 346-349. McGraw-Hill, New York, 1959.
37. M. W. Rubesin and H. A. Johnson, A critical review of skin-friction and heat-
transfer solutions of the laminar boundary layer of a flat plate. Trans. A S M E ,
71, 383-388 (1949).
38. L. Crocco, The laminar boundary layer in gases. APL/NAA/CF-1038, North
American Aviation Inc., Los Angeles, California, July 15, 1948. (Translation.)
39. G . B. W. Young and E. Janssen, The compressible boundary layer. J. Aeron. Sci.
19 (4), 229-236, 288 (1952).
[I061
HEATTRANSFER
FLOWS
40. E. R. Van Driest, The laminar boundary layer with variable fluid properties. North
American Aviation Inc., Rept. No. AL 1866, 1954.
41. E. R. G. Eckert, Survey on heat transfer at high speeds. WADC T R 54-70, April
1954.
42. L. V. Humble, W. H. Lowdermilk, and L. G. Desmon, Measurements of average
heat-transfer and friction coefficients for subsonic flow of air in smooth tubes at
high surface and fluid temperatures. NACA Rept. 1020, 1951.
43. R. G. Deissler, Analytical investigation of turbulent flow in smooth tubes with
heat transfer with variable fluid properties for Prandtl number of 1. NACA T N
2242, December 1950.
44. E. R. Van Driest, Turbulent boundary layer in compressible fluids. J. Aeron. Sci.
18 (3), 145 (1951).
45. A. P. Colburn, A method of correlating forced convection heat transfer data and a
comparison with fluid friction. Trans. Am . Inst. Chem. Eng. 29, 174-210
(1933).
46. W. M. Kays and W. B. Nicoll, Laminar flow heat transfer to a gas with large tempera-
ture differences. J. Heat Transfer, COO (4), November 1963.
47. H. B. Squire, Heat transfer calculations for aero-foils. Great Britain Aeronautical
Research Council, R & M 1986, November 1942.
48. R. A. Seban and D. L. Doughty, Heat transfer to turbulent boundary layers with
variable free-stream velocity. Tram. A S M E , 78, 217 (1956).
49. M. R. Denison and D. A. Dooley, Combustion in the laminar boundary layer of
chemically active sublimators. Publ. No. U-110, Aeronutronic Systems, Inc.,
Glendale, California, September 23, 1957.
50. L. Lees, Convective heat transfer with mass addition and chemical reactions. Combust.
Propulsion, A G A R D Colloq. 3rd Palermo, Sicily, 1958, p. OO( 1958). Macmillan
(Pergamon), New York; also in Recent Advances in Heat and Mass Transfer
(J. P. Hartnett, ed.), p ~ 161-207.
. McGraw-Hill, New York, 1961.
51. J. A. Fay and F. R. Riddell, Theory of stagnation point heat transfer in dissociated
Air. J. Aeron. Sci. 25 (2), 73 (1958); also in Recent Advances in Heat and Mass
Transfer (J. P. Hartnett, ed.), pp. 115-141. McGraw-Hill, New York, 1961.
51a. D. E. Rosner, private communication, 1963.
52. W. E. Welsh, Jr., and A. B. Witte, A comparison of analytical and experimental
local heat fluxes in liquid-propellant rocket thrust chambers. J. Heat Transfer
CS4, ( I ) , 19-28 (1962).
53. 0. A. Saunders and P. H. Calder, Some experiments on the heat transfer from a
gas flowing through a convergent-divergent nozzle. Proc. Heat Transfer Fluid Mech.
Znst. 1951 (1951). Stanford Univ. Press, Stanford, California.
54. W. C. Ragsdale and J. M. Smith, Heat transfer in nozzles. Chem. Eng. Sci. 11,
242-51 (1960).
55. J. R. Baron and F. H. Durgin, An experimental investigation of heat transfer at
the boundaries of supersonic nozzles. WADC T R 54-541, Naval Supersonic laboratory,
Massachusetts Institute of Technology, December 1954.
56. P. F. Massier, Convective heat transfer in nozzles. Combined Bimonthly
Summary No. 63, Jet Propulsion Laboratory, Pasadena, California, February 15,
1958 (28-30).
57. J. J . Kolozsi, An investigation of heat transfer through the turbulent boundary
layer in an axially symmetric, convergent-divergent nozzle. Masters thesis, Dept.
of Aeronautical and Astronautical Engineering, Ohio State University, Columbus,
Ohio, 1958.
D. R. BARTZ
58. A. Fortini and R. C. Ehlers, Comparison of experimental to predicted heat transfer

in a bell-shaped nozzle with upstream flow disturbances. NASA T N D-1743,
August 1963.
59. L. H. Back, P. F. Massier, and H. L. Gier, Convective heat transfer in a convergent-
divergent nozzle. Intern. J. Heat Mass Transfer 7 ( 3 , 549-568 (1964).
60. K. Oswatitsch and W. Rothstein, Flow pattern in a converging-diverging nozzle.
NACA TM-1215, March 1949.
61. A. A. Sergienko and V. K. Gretsov, Transition from a turbulent into a laminar
boundary layer. Soviet Phys. Doklady (English Transl.) 4 (2), 275-276 (1959).
62. L. H. Back, Heat transfer to turbulent boundary layers with a variable free-stream
velocity. Ph.D. Thesis, University of California, Berkeley, California, June 1962.
63. J. 0. Hinze, Turbulence p. 62. McGraw-Hill, New York, 1959.
64. P. S. Klebanoff, Characteristics of turbulence in a boundary layer with zero pressure
gradient. NACA TN-3178, 1954.
65. R. K. Rose, Experimental determination of the heat flux distribution in a rocket
nozzle. M.S. Thesis, Purdue Univ., Lafayette, Indiana, January 1958.
66. R. F. Neu, Comparison of localized heat transfer rates in a liquid-oxygen-heptane
rocket engine employing several injection methods and oxidant-fuel ratios. NASA
T N D-286, June 1960.
67. J. P. Sellers, Jr., Effect of carbon deposition on heat transfer in a LOX/RP-1 thrust
chamber. ARS (Am. Rocket SOC.) J. 31 (9,662-663 (1961).
68. A. B. Witte and E. Y. Harper, Experimental investigation of heat transfer rates in
rocket thrust chambers. AZAA J. 1 (2), 443-451 (1963).
69. W. B. Powell, J. P. Irving, and M. E. Guenther, Chlorine trifluoride-hydrazine
liquid propellant evaluation and rocket motor development. Tech. Rept. No. 32-305,
Jet Propulsion Laboratory, Pasadena, California, May 15, 1963.
70. W. B. Powell, Liquid oxygen-liquid hydrogen systems. Space Programs Summary
No. 37-25, Vol. IV, JetPropulsion Laboratory,Pasadena, California, February 29,1964.
71. J. H. Rupe and G. I. Jaivin, Effects of mass flux distribution and resonant combustion
on local heat transfer rates in a liquid propellant rocket engine. Tech. Rept. No. 32-
648, Jet Propulsion Laboratory, Pasadena, California, September 23, 1964.
72. W. B. Powell and T. W. Price, A method for the determination of Iocal heat flux
from transient temperature measurements. ZSA (Znstr. SOC. Am.) Trans. 3 (3) (1964).
73. A. B. Witte, Heat transfer in rocket motors. Research Summary No. 36-5, Vol. 11,
pp. 49-5 I. Jet Propulsion Laboratory, Pasadena, California, November 1, 1960.
74. W. E. Welsh, Jr., Review of results of an early rocket engine film-cooling investigation
at the Jet Propulsion Laboratory. Tech. Rept. No. 32-58, Jet Propulsion Laboratory,
Pasadena, California, March 13, 1961.
75. R. A. Seban, The influence of free stream turbulence on the local heat transfer from
cylinders. J. Heat Transfer. C82, 101 (1960).
76. J. Kestin, P. F. Meader, and H. E. Wang, Influence of turbulence on the transfer
of heat from plates with and without a pressure gradient. Intern. J. Heat Mass
Transfer 3, 133-154 (1961).
77. S. P. Sutera, P. F. Meader, and J. Kestin, On the sensitivity of heat transfer in the
stagnation-point boundary layer to free-stream vorticity. J. Fluid Mech. 16 (4),
497-520 (1963).
78. W. A. Benser and R. W. Graham, Hydrogen convective cooling of rocket nozzles.
Am. SOC.Mech. Engrs. Paper 62-AV-22 (1962).
79. D. R. Bartz, Factors which influence the suitability of liquid propellants as rocket
motor regenerative coolants. Jet Propulsion 28, 46-53 (1958).
PO81
Chemically Reacting
Nonequilibrium Boundary Layers
.
PAUL M CHUNG
Aerospace Corporation. Sun Bernardino. California
.
I Introduction . . . . . . . . . . . . . . . . . . . . . . 110
I1. Governing Equations . . . . . . . . . . . . . . . . . . . 112
.
A Diffusion Velocity . . . . . . . . . . . . . . . . . . 114
.
B HeatFlux . . . . . . . . . . . . . . . . . . . . . . 115
C . Boundary Layer Equations . . . . . . . . . . . . . . 116
.
D Boundary Conditions . . . . . . . . . . . . . . . . . 118
E. Surface Heat Transfer . . . . . . . . . . . . . . . . 120
111. Chemical Kinetics . . . . . . . . . . . . . . . . . . . . 122
A . Homogeneous Chemical Reaction . . . . . . . . . . . 123
B . Heterogeneous Chemical Reaction . . . . . . . . . . . 135
IV. Boundary Layers with Surface Reactions . . . . . . . . . . 138
A . Governing Equations and Boundary Conditions . . . . . . 138
B. Transformation to 6-S Coordinates . . . . . . . . . . . 139
C . Momentum and Energy Equations . . . . . . . . . . . 141
D . Damkohler Number for Surface Reactions . . . . . . . . 143
E . Self-similar Solutions of Species Conservation Equation . . 144
.
F Surface Reaction and Heat Transfer From High Temperature
Dissociated Air and Chemical Diagnostics . . . . . . . . 161
G . Nonsimilar Solutions of Species Conservation Equation . . 1 64
V . Boundary Layers with Gas Phase Reactions . . . . . . . . 192
A . Governing Equations and Heat Transfer . . . . . . . . 192
B. Transformation of Governing Equations . . . . . . . . . 192
C . Damkohler Number for Gas Phase Reaction . . . . . . . 194
.
D Self-similar Solutions at Stagnation Region . . . . . . . 195
E . Thermal Ignition Problems . . . . . . . . . . . . . . 213
F. Nonequilibrium Flows of High Energy Gases with Dissocia-
tion and Recombination . . . . . . . . . . . . . . . . 218
G . Near-Equilibrium Flows . . . . . . . . . . . . . . . 251
V I . Concluding Remarks . . . . . . . . . . . . . . . . . . 266
Nomenclature . . . . . . . . . . . . . . . . . . . . . 267
References . . . . . . . . . . . . . . . . . . . . . . . 268
[lo91
PAULM. CHUNG
I. Introduction
T h e chemical reactions associated with the interaction of gases in

motion and solid or liquid surfaces have interested engineers for ages.
Classically, the subject has received much attention in connection with
the combustion of solid and liquid fuels. T h e subject has also been of
great interest to the chemical industries for the operation of their
catalytic reactors.
T h e real surge of interest, however, in the chemically-reacting flows
adjacent to solid or liquid surfaces, has occurred during the past decade
or so with the advent of the hypervelocity vehicles in the aerospace
technology. Aside from the problem of combustion in the propulsion
systems, the study of heat transfer to the exterior or the rocket nozzle
of such vehicles necessitates the understanding of chemical interactions
between the gases and the solid boundaries. Most recently, the problem
of chemically reacting flows past solid boundaries, as well as in the wakes
and trails of solid objects, has received additional attention in connection
with the ballistic re-entry bodies. This is due to the belief that much
of the characteristics of the re-entry bodies can be foretold by observing
the chemical structure (including the ionization profile) of the flows by
means of an optical receiver or a radar.
I n order to study the problems of the hypervelocity vehicles men-
tioned herein, various high temperature experimental facilities are being
built at many laboratories, T h e high temperature gas is usually generated
by an electric arc or a shock. One of the methods of determining the
chemical properties of the generated gases (chemical diagnostics) at
the test section is to employ some sort of a differential catalytic gage.
T h e development of such gages for the various experimental facilities
has generated a series of problems which also requires the knowledge
of the chemical interaction between the gases and the solid surfaces.
Most of the problems of chemical reactions described above can be
analyzed within the framework of a boundary layer theory. I t is therefore
natural that the progress of analytical treatments of the problems has
followed closely that of the boundary layer theory which took place
during the past decade.
There are already in existence several monographs and books ( 1-6)
which treat the subject of chemically reacting boundary layers. I n these,
however, the treatments are rather confined to the cases in which the
chemical reaction, whether it occurs in the gas phase or on the surface,
is considered to take place at an infinitely fast rate so that the chemical
rate-kinetics is no longer a factor. For this extreme limit, the solution
of the problem can usually be obtained in terms of the existing solutions
PlOI
CHEMICALLY
REACTING
NONEQUILIBRIUM LAYERS
BOUNDARY
of the conventional boundary layer equations, and the problem is more

or less reduced to re-evaluating algebraically the proper energy driving
potential. Though this extreme limit (equilibrium limit) constitutes an
important regime of the chemically reacting flows, most of the problems
of reacting boundary layers described herein do not fall into this category.
When the chemical reaction is considered to occur at an infinitely fast
rate, the governing equations are essentially the same as those for the
chemically inert boundary layers, and the chemical reactions are governed
by the characteristics of the boundary layer alone. T h e most interesting
features of the chemically-reacting boundary layers, which are due to
the coupling of the boundary layer characteristics and the finite-rate
chemical reaction, are completely lost in the equilibrium limit. It is only
in the nonequilibrium cases where the chemical reaction takes place
with a finite rate, that the true behavior of a chemically reacting boundary
layer becomes manifest.
A few nonequilibrium problems were included in the monographs
(2, 4, 6). They are the conductivity cell (2), simple surface reactions (4),
and the linearized Rayleigh problem (6).
T h e present monograph is devoted to the nonequilibrium, chemically
reacting boundary layers. We shall more or less take up where Rosner
( 4 ) and Moore (6) left off. Since the previous works (1-4) discussed
fully the equilibrium cases, including the energy-driving potential and
the variation of the transport properties, etc., we shall mention them
here only to aid the reader in understanding the general nonequilibrium
problem.
This monograph is not intended as a collection of engineering data
or an exhaustive review of all the pertinent published w0rk.l I t is rather
the intention of the author to present some of the basic characteristics of
the nonequilibrium chemically reacting boundary layers and to present
the methods of solution of the physical problems described earlier.
Only the reacting laminar boundary layers are included herein, for very
few published works exist for turbulent nonequilibrium boundary
layers. T h e basic concept presented, however, should also be applicable
to the turbulent cases.
We shall begin with the formulation of the governing equations and
a brief discussion of the chemical kinetics. We shall then follow with
the various cases of nonequilibrium surface reactions assuming that the
gas phase reaction is frozen. T h e nonequilibrium gas phase reaction will
then be considered for surfaces either nonreacting or in equilibrium.
Finally, an approximate treatment of simultaneous nonequilibrium
See Ref. (7) for a complete bibliography.

[1111
PAULM. CHUNG
surface and gas phase reactions will be considered. Each of the above
three cases will begin with the analysis of the stagnation region where
a self-similar transformation of the governing equations exists. T h e
analysis is then extended to the nonsimilar cases. It is presumed here
that the reader has a basic knowledge of boundary layer theory.
11. Governing Equations
T h e steady state governing equations for the chemically reacting,

two-dimensional and axisymmetric laminar boundary layers will be
formulated. We begin the formulation by writing the boundary layer
equations in the following general form:
Continuity
a(p2lY')
+--=o
apvr'
ax aY
Momentum
p u au a
i -+v*)
ax ay
= dp au
Energy
Conservation of ith species

a
( ac.
p u1
ax
+ v ac') aY (pC,6,) + wi
ay = - - (4)
Equation of State
p = ( z y -ci
)pRT
(5)
i Mi
where the enthalpy, h, is defined to include the chemical energy as
h =zCihi
and
hi = 1 T
0
cpi dT + h: (7)
T h e variables x., <y,. and r , and u and v are the distances and the velocitv
components defined in Fig. 1. T h e exponent E is 0 or 1 depending o h
PI21
REACTINGNONEQUILIBRIUM
CHEMICALLY BOUNDARYLAYERS
Y
4 BOUNDARY LAYER EDGE, e
W a (FLAT PLATE)
-:.::,: .:..:..,.."
x
I - L I
-
b (SUPERSONIC WEDGE OR CONE)
\
FIG. 1. Coordinate system employed for entire chapter.
whether the boundary layer is two dimensional or axisymmetric. p,p, and

p are the density, viscosity, and the pressure of the gas mixture, respec-
tively. C , , d , , c p j , and hio denote the molecular weight, mass
fraction, diffusion velocity, constant pressure specific heat, and the heat
of formulation of the ith species, respectively.
T h e quantities 6, q, and W irepresent the diffusion velocity, heat
flux, and the mass rate of formation of the component i per unit
volume, respectively. These three quantities distinguish the present
energy and species conservation equations from those of the chemically
inert boundary layers. I n order that Eqs. (3) and (4) be useful, the
quantities 6,q, and W imust first be specified. We shall develop the
~ 3 1
PAULM. CHUNG
expressions for 6 and q adaptable to the present purpose in the following.

T h e discussion of W iwill be deferred to a later section.
A. DIFFUSION
VELOCITY
A rigorous development of the expression for the diffusion velocity Gi
can be found in the literature (8, 9). Here, we shall derive the simplified
expressions which will be used in the present analysis.
Following the lead of most of the published papers on the chemically
reacting boundary layers, we first neglect the thermal diffusion in com-
parison with the concentration diffusion. Then for a multicomponent
boundary layer wherein ap/ay = 0, the following equation is obtained
from the kinetic theory (8, 9):
where i # j, and i = 1, 2, 3, ..., n. Dij is the binary diffusion coefficient

between the species i and j .
For a binary mixture of components 1 and 2, Eq. (8) reduces to the
familiar Fick's relationship
C16,= -D12 VC, (9)
As it is seen from Eq. (8), a rigorous treatment of diffusion velocities for
a multicomponent boundary layer adds considerable complexity to the
species conservation equations. A nonequilibrium chemically reacting
boundary layer is a quite formidable problem and a simplification of the
expression for the diffusion velocity is greatly to be desired. We shall now
see that the simple Fick's form of the diffusion velocity can be deduced
for a multicomponent system when certain special conditions are satisfied.
First, consider that we are interested only in the diffusion of a particu-
lar component i among the n-component system. Such is often the case
with the combustion of premixed fuels. Then if for the diffusion of a
particular component i, all the Dij and fij, where i # j , are equal
then Eq. (8) can be reduced to the following form:
CiGi = -Di VCi (W2
Notice here that Mimay be different from Mj. Equation (10) can be
written for each component simultaneously if all M j are equal including
.
J = 1.
.
* Since D,, are equal for allj, we will simply denote it by Di

~1141
CHEMICALLY
REACTING
BOUNDARY
Next, consider that the sum of all the species of our prime interest
constitutes a trace in the n-component system. For this case, we begin
with the following expression given by Curtiss and Hirschfelder (10)
for a particular component, i, present as a trace in an n-component
system :
where the summations include all j . Now if the sum of the diffusing
species of our main concern constitutes a trace, and if all D, and &Ij
are, respectively, equal where j does not include the diffusing species,
then Eq. (1 1) gives
Ci8i FZ -Di VCi (12)
simultaneously for each component i of our prime interest.
Now we have seen that it is possible for certain cases of multicom-
ponent systems to express the diffusion velocities of the pertinent species
by the simple Ficks form. These possibilities have been explored herein
because a chemically reacting boundary layer is seldom a true binary
system, and often it is desired to consider several diffusing species
simultaneously.
Often, the multicomponent boundary layer does not exactly satisfy
all the special conditions set forth above for the reduction of the diffusion
velocities to the simpler Ficks form. A considerable amount of approxi-
mation, however, usually accompanies the solution of reacting boundary
layer problems because of the inherent uncertainties in the chemical
rate processes and transport properties, etc. I n view of these approxima-
tions, it is a rather common practice to employ the simpler Ficks form
of the diffusion velocities, instead of the more rigorous forms, if the
conditions described herein can be roughly satisfied. T o facilitate the
analysis, the present monograph will be limited to the discussion of the
boundary layers wherein the diffusion velocities can be reduced to the
Ficks form.
B. HEATFLUX
With the diffusion velocities specified, we are now ready to formulate
the heat flux q. I n a multicomponent system, the heat flux is due to the
thermal conduction and the energy carried by the interdiffusing species.
Thus,
PAULM. CHUNG
where h is the thermal conductivity. Substituting the Fick's form of G i ,

we obtain
C. BOUNDARYLAYEREQUATIONS
Now we may, with the aid of the preceding discussions of the diffusion
velocity and the heat flux, write the explicit forms of the boundary layer
equations to be subsequently analyzed. The continuity and the momen-
tum equations, Eqs. (1) and (2), are unchanged. T h e energy and the
species conservation equations become
Energy
+
Now, basically, the problem is to determine the ( 5 i) number of
variables, p, u, v, h, T,and Ci's, by the use of the same number of
equations, Eqs. (1); (2), (15), (16), (5), and (6). There are as many
conservation equations, Eq. (16), as the number of species. However,
the number of the necessary species conservation equation is reduced
by one, and usually some of the conservation equations are substantially
simplified by the use of the following relationships:
pi 1 i
=
2 Ci& 0
a
=
zwi=o
i
There are two alternate forms of the energy equation whichat times
are more useful than Eq. (15). T h e first of these is the energy equation
which is expressed mainly in terms of the total enthalpy, h , , rather than h,
where
U2
h,=h+-
2
CHEMICALLY
REACTING
BOUNDARY
This energy equation is derived by first multiplying the momentum

Eq. (2) by u and combining it with the energy Eq. (15). A little mani-
pulation of the combined equation with the aid of the continuity equa-
tion and the equation
+
dh = c,,dT z h , d C i
i
(21)
which is obtained by differentiating Eq. ( 6 ) , gives the desired equation
where the Prandtl number, Pr, is defined by
and
T h e Lewis number for the component i, L e i , is defined by

pD.c
Le.1 -- 1 Pr
9= -
h sc,
Another alternate form of the energy equation is that mainly based

on the temperature (static temperature) T. This equation is derived
from Eq. (15) with the aid of Eq. (21) and the species conservation
Eq. (16) as
Equations (15), (22), and (27) are the three alternate forms of the energy
equation applicable to a chemically reacting boundary layer.
When analyzing a problem of surface reaction with the frozen gas
phase reactions, it is convenient to use the frozen total enthalpy h,,
defined by
htj = 2 Cihif +
i
112
(28)
[1171
PAULM. CHUNG
where
hi, = lo
T
cpi dT
For the special cases of frozen gas phase reactions, a frozen total energy
equation can be readily derived in a manner similar to Eq. (22). T h e
resulting equation is identical to Eq. (22) except h , and h, are replaced
by h,, and h i , , respectively.
D. BOUNDARYCONDITIONS
T h e boundary conditions for the continuity and the momentum
I
equations are the well-known ones and are, at y = 0,

u=o (30)
v = v, (31)
and, at y = co,
li = u, (32)
where the subscripts w and e refer to the wall and the edge of boundary
layer, respectively.
Consider now the boundary conditions for the energy equations (1 5 ) ,
(22), and (27). T h e enthalpies and the temperature are considered to be
known at the boundary layer edge. Thus, at y = co,
ht = hte = he + (ue)2/2 (34)

T = T, (35)
We write at the wall with the aid of Eqs. (6) and (7),
2 Ci, + 2 Ci,Jzio
Tw
h = h, = cpi dT
i 0 i
h, = h,, = h,
T = T,
Equations (36) and (37) show that the wall boundary conditions for
Eqs. (15) and (22) cannot be specified unless the wall mass fractions,
C,, are known. Usually, C,, are not known a priori and therefore the
P181
wall boundary conditions of Eqs. (15) and (22) are coupled to the solu-
tions of the species conservation equations. The wall temperature is
considered to be known. Though the wall boundary condition for Eq.
(27) is known a priori, the use of Eq. (27) instead of Eq. (15) or (22)
does not necessarily expedite the solution because Eq. (27) itself is
more strongly coupled to the species conservation equation than the
other two energy equations.
Finally, consider now the boundary condition for the species conser-
vation Eq. (16). We have, at y = co,
Cl = ca, (39)
At the wall (gas-solid interface), the component i is transported from the
gas to the solid by diffusion at the rate (pDi aCi/8y), (see Fig. 2a). At
FIG.2. Mass and energy balance at the interface.

~191
PAULM. CHUNC
the same time, the component i is transported away from the interface
by the normal current at the rate (pv),Ci, in the gas, and toward the
interface at the rate (pv),(Ciu,)- in the solid (may be a porous solid).
(pv),( Ct,,)- is that of the solid if the solid wall is ablating whereas it is
that of the injected fluid when a fluid is injected through a porous wall.
T h e net rate of production of the component i by the surface reaction,
Jiw , is then given by
aci
Jiw = -PwDiw (F)
3w
+ ~ w v w [ C i w- (Ciu,)-I (40)
(C,,,)- is zero when the particular ith component is not included among
the injected or ablated gas mixture, whereas it is 1 when the injected
or ablating substance is entirely comprised of the component i.
T h e surface reaction rate Ji,, together with the gas phase reaction
rate W iwill be discussed in a subsequent section.
E. SURFACE
HEATTRANSFER
Before leaving the present section, we shall develop a few general
expressions for surface heat transfer.
At the surface of a solid, energy is transported from the boundary
layer to the solid by conduction and diffusion at the rate ( A aT/ay +
pEDihi a C i / a y ) , (see Fig. 2b). At the same time energy is transported
away from the interface by the normal current at the rate (pv), ECiwhi,
in the gas, and toward the interface at the rate (pv),E(Ci,hi,)- in the
solid or in the fluid through a porous wall. T h e subscript (-) denotes
the interior of the solid where the normal gradients are zero. T h e net
rate of convective heat transfer, qw , to the solid interior is given by
Equation (41) can be transformed into the following with the aid of
Eq. (21):
CHEMICALLY
REACTING
NONEQUILIBRIUM
BOUNDARY
LAYERS
Another form of qw is also derived from Eq. (41) as
where h, > 0 is the latent heat of vaporization at the surface. Equations

(41), (42), and (43), are now respectively rewritten with the aid of Eqs.
(40) and (44), as
-(pv>w
i
hifw[Ciw - ( c t w ) - l - 2
i
i ~ (pv)wht
~ i w h- (47)
T h e overall effect of the gas phase and surface chemical reactions on the
surface heat transfer can now be seen from the governing equations and
the heat transfer expressions derived above.
For most gases of our interest, Pr and Lei respectively are close to
unity. It is seen therefore that the effects of the last two terms of Eq.
(22) on the solution of the equation will be rather small. In any case,
we should be able to discuss the behavior of the energy equation without
loss of generality by assigning Pr = Lei = 1. When chemical reactions
take place, we may employ Eqs. (22) and (46) to obtain heat transfer,
and the equations become, for Pr = Lei = 1,
PAULM. CHUNG
and
When, on the other hand, chemical reaction takes place neither in the
gas phase nor on the surface, we may employ the frozen total energy
equation, which becomes identical in form to Eq. (22a), and Eq. (47) as
and
T h e preceding four equations show that the energy equations and the
heat transfer expressions are respectively identical between the chemically
reactive and inert boundary layers except h , is used for the former case and
h , for the latter case. We may therefore state in general that the total
enthalpy difference, which includes the difference of chemical energy,
replaces the frozen total enthalpy difference as the driving potential for
heat transfer when the boundary layer becomes chemically reactive.
In particular, when h, = pwvw = 0, the preceding four equations give
= I +
Ci C,,hio - Ci C i d i o
htfe - htfw
Eq. (48) shows that for this case the heat transfer is increased by
Cj Ci,hio - Ci C . h."
IW ' 1
due to the chemical reaction whether the reaction takes place in the gas
phase or on the surface.
It should be noted here that Eq. (48) is only qualitative because
CCiWh,"is not known a priori except when the gas is in chemical
equilibrium at the surface.
111. Chemical Kinetics
T h e chemical kinetics which provides the general method of deter-

mining the mechanism of reaction is a very complex and extensive sub-
[ 1221
CHEMICALLY
REACTING
BOUNDARY
ject. I t is not intended here, therefore, to give the subject a comprehen-

sive treatment. We shall only consider those portions of the chemical
kinetics necessary to continue our study of the reacting boundary layers.
The reader is referred to the literature (9, 21-14) for a more comprehen-
sive study of the subject.
A. HOMOGENEOUS
CHEMICAL
REACTION
Chemical reactions are of two types: homogeneous reactions and
heterogeneous reactions. T h e former occurs in a homogeneous phase
<<
such as in an all gaseous system. Heterogeneous reactions, on the other

hand, are characterized by their preferential occurrence at an interface
such as the surface of a solid or a liquid.
Here, we consider the homogeneous reactions. T h e heterogeneous
reactions will be considered subsequently.
1. Law of Mass Action

Consider the following stoichiometric equation,
2
j=1
ajMj+ 5 bjMj
j=1
(49)
where aj and bj are the stoichiometric coefficients for the reactants and
reaction products, respectively. M j represents the chemical component j .
We shall parenthesize as (M,) to designate the mole concentration,
e.g., moles per cubic centimeter, of the component j . According to the
(<
law of mass action, the rate of production of a particular component,
i, is proportional to the products of the concentrations of the reacting
chemical species, each concentration being raised to a power equal to
the corresponding stoichiometric coefficient. Thus,
T h e quantity (bi - ai) is included in the equation because the com-

ponent i may also appear in the reaction product. T h e quantity k, is
called specific reaction-rate coefficient for the reaction of Eq. (49), and
is defined by Eq. (50). A,, therefore, is essentially a proportionality
constant defined through the law of mass action. T h e units of k, vary
depending on the particular kinetic equation, such as Eq. (50). T h e
sum Zp, is called the order of the reaction.
11231
PAULM. CHUNG
I n general, chemical reactions can proceed in both forward and back-

ward directions. Thus, the more general stoichiometric equation is
where k, and k, represent the forward and the backward rate coefficients
respectively corresponding to the arrows shown in the equation. T h e
law of mass action then gives for the net production rate of the compo-
nent i,
4_
dr
) --(bi - ai) [k,
M i_ fi
j=1
(Mi)'> - k b fi
j=1
(l%fj)bj]
Now, consider that the above reaction is allowed to take place un-
disturbed with an infinite amount of time available. Then, the state of
chemical equilibrium will be achieved by the system wherein no further
change in composition takes place. T h e true equilibrium state from the
kinetic viewpoint is approached asymptotically, and the achievement
of the same requires an infinite amount of time. However, the practical
asymptotic equilibrium state is acquired in a finite amount of time which
is proportional to reciprocals of the reaction rates. In a true equilibrium
state
and Eq. (52) gives
i54)
where subscript E denotes the equilibrium state. Equation (54) now

defines the equilibrium constant K, based on the concentrations. Now,
either of the specific rate coefficients, k, and k, , can be eliminated from
Eq. (52) with the aid of Eq. (54). Equation (52) may be written as
or
I n order that Eqs. (55) be quantitatively useful in determining a reaction

rate, the equilibrium constant and one of the specific rate coefficients
c 1241
CHEMICALLY
REACTINGNONEQUILIBRIUM LAYERS
BOUNDARY
must be known. We therefore turn to the equilibrium constant and the

specific rate coefficients in the following.
2. Equilibrium Constant
Let us start with the Boltzmann law of statistical mechanics which
states that the probability of a particle being in a state of energy ei , and
statistical weight g, , called the degeneracy, is proportional to gie-ci/kT.
T h e probability that the particle will exist in any one or another of its
possible states, given by a series of values of the index i, will be propor-
tional to
This sum is called the partition function for the particle.

Without any elaboration, one can see from the above definition that
the relative concentrations of the chemical species in an equilibrium
system are related to their partition functions. In fact, it can be rigorously
derived from statistical mechanics that the partition functions and the
equilibrium concentrations of the chemical species, in the reacting
system of Eq. (51), are related by the equation
where Qc is the partition function for the standard states of unit concen-
tration. Thus
P
O c = - Q HT
where R is the universal gas constant. From Eqs. (54) and (57), the
equilibrium constant is then also defined in terms of the partition func-
tion as
T h e products and ratios of the partition functions as written in Eqs.

(57) and (59) are based on the supposition that the same zero of energy
is used in expressing the energy levels ei of each chemical species.
~1
PAULM. CHUNG
However, it is more convenient to take the lowest energy level of each

species as the zero for that species. I n this case we define
C airj"
n n
A E O = bjrjo - (60)s
3=1 3=1
where ei0 is the energy of formation of species j , per particle, at the

absolute zero of temperature. T h e numerator of Eqs. (57) and (59),
Q2,
IIy=l then must be replaced by II%, Q2 exp( -Ae"/KT) since
exp( --E,"/KT)will factor out of each term of the partition-function
summation, Eq. (56). Now Eqs. (57) and (59) are rewritten as
where AE" = N A Aec and N A is Avogadro's number.

T h e equilibrium constant of a system is often based on the partial
pressures of the species, pi rather than the concentrations (illj). T h u s
we define
From the relationship between p j and ( M j ) ,it is seen that

K , = K,(RT)"I-"~~ (63)
T h e partition functions defined in general by Eq. (56) can be calculated
accurately from statistical mechanics, and the reader is referred to any
standard text for a detailed treatment of the subject. It is sufficient for
the present purpose only to note that the partition functions, therefore
the equilibrium constant, can be computed accurately from the
statistical mechanics.
3 . Spec$c Reaction Rate Coeficient

T h e analytical approaches of obtaining the rate coefficients have been
progressing mainly along the lines of collision theory and the theory
of absolute reaction rates. I n spite of much analytical work, it is perhaps
Aro thus defined, multiplied by the Avogradro number, (AE"), is then the heat of
reaction for the chemical reaction expressed by Eq. (49). Positive quantity of AE" in
this case means that the process is endothermic.
[I261
CHEMICALLY
REACTING
BOUNDARY
justifiable to state that the theoretical analysis has not yet advanced to
the point where a significant contribution can be made to the solution
of the present-day complicated high-temperature reactions, Experimen-
tally, however, considerable strides have been made in recent years in
measuring the high-temperature reaction rates for certain cases, such
as the chemical reactions taking place in a high-temperature air [for
instances, see Refs. (IS), (16)]. Though the theoretical approaches
alone seldom produce satisfactory numerical values of specific reaction
rates, the theoretical analyses are invaluable in understanding the nature
of the reaction and interpreting the experimental data.
T h e rigorous analytical treatment of the specific reaction rate coeffi-
cient is beyond the scope of the present endeavor. T h e following simple
analysis is intended to show the basic meaning of the quantities com-
prising a rate coefficient, T h e reader will find it helpful in adapting the
reation rate data presented by physicists and chemists to the boundary
layer problems.
For a given chemical reaction, the specific rate coefficient, k, or kb ,
is independent of the concentration and depends only on the tempera-
ture. I n general, K ( K , or k b ) is given by an expression of the form
k = Z ( T )exp [-E,/(RT)] (64)
where E, is the activation energy for the reaction. Arrhenius first sug-
gested this form, with 2 considered as constant, as an imperical formula
of correlating experimental data. Let us investigate, at least qualitatively,
the theoretical explanation for the particular form. We shall investigate
it from the viewpoint of the theory of absolute reaction rates since it
suits our purpose better than the collision theory. We consider the
reaction of Eq. (49) proceeding with the specific rate coefficient K , .
According to the theory, the collisions among the reactants, under
favorable conditions, lead to the formation of a transitory chemical
species called the activated complex. T h e activated complex is then
transformed to the reaction products at a certain rate. Let us designate
the activated complex by X, and its concentration by (X+). Then the
+ +
reaction of Eq. (49), according to the law of absolute reaction rates,
is actually a two-step reaction expressed by
2 a,Mj 2bjMj
n n
+ X+ -+
j-1 + j-1
Now we consider the potential energy of the reacting system as a function

of the reaction path (reaction coordinate). Depending on the choice of
~271
PAULM. CHUNG
the reaction coordinate, there will be different amounts of energy

required for the reaction described by Eq. (65). T h e activated complex
is located at the point of highest energy on the most favorable reaction
path and the activation energy E, per mole is required to form X,
+
(see Fig. 3a). It is usually assumed that the reaction between the reactants
and the activated complex takes place much faster than that between
the activated complex and the products. T h e reactants and the activated
REACTION COORDINATE -
- ----
-
Y
REACTION COORDINATE
FIG.3. Activation energy and heat of reaction.

CHEMICALLY
BOUNDARYLAYERS
complex are therefore in chemical equilibrium, and the reaction between

the complex and the products is the rate determining reaction of the
overall reaction Eq. (49). Equation (65), therefore, is rewritten as
2ajMj
j=l
X+
+ j=i
(X,) may be obtained by utilizing the equilibrium relation of Eq. (61) as

+
T h e rate of formation of the products is then obtained by multiplying

(X,) by the frequency of crossing of the energy barrier (maximum energy
+
corresponding to the activated complex). T h e major work of the theory
is comprised of evaluating the partition function of the activated com-
plex Qc+ , the activation energy E, , and the frequency of crossing of the
+
energy barrier. Without going any further with the theory of absolute
reation rates, it is now clear that
A comparison of Eqs. (64) and (68) shows one of the theoretical expla-
nations for the function Z( T) and the exponential function exp( -Ea/RT).
Now, turning to the more general reaction of Eq. (51), the specific rate
coefficient for the backward reaction, k,, is obtained from Eqs. (68)
and (54) as
k, = k,/Kc (69)
4. Exponential Functions exp( -- AE"/RT) and exp( -Ea/RT)

We have seen in the preceding sections that the exponential functions
exp( -AE"/RT) and exp(-Ea/RT) appear rather consistently in the
expressions for the reaction rates. Though the two functions look similar,
it is quite important that one remembers the basic differences between
the two. Since the temperature dependency of reaction rates is predomi-
nantly governed by the exponential functions we shall give our attention
to these functions herein.
~ 9 1
PAULM. CHUNG
As it can be seen from Fig. 3a, the heat of reaction AE" and the activa-
tion energy E,, are different quantities. T h e accurate values of heat of
reaction can be obtained theoretically or experimentally. T h e accurate
activation energies, on the other hand, are seldom available. We shall
illustrate the manner in which exp( --d E " / R T ) and exp( -E J R T ) enter
into the reaction rates by the following two examples.
First, consider a typical combustion of hydrocarbon fuel. Actual
combustion process of a hydrocarbon is a rather complicated chain
process. However, it is a common engineering practice to consider the
process as a one-step bi-molecular reaction expressed by
where M , , M , , and Mp represent the fuel, the oxidizer, and the com-
bustion product, respectively. T h e energy-reaction coordinate relation-
ship is sketched in Fig. 3b. Applying the law of mass action, Eq. (52),
to the stoichiometric relationship of Eq. (70), we obtain for the com-
bustion rate of the fuel
Now according to the preceding section,
It has been found, mostly experimentally, that for most hydrocarbons,
5000K 2 4 5 40,000"K (73)
T h e equilibrium constant K, is, according to Section 2,
= Z E ( T )exp[--dE"/(RT)] (74)
where, for the present definition of K,

AE" = E," - (Ef" + E,") < 0 (75)
T h e magnitudes of AE'jR for hydrocarbons are often within the same
range as those of E J R shown by the relationship Eq. (73). Thus, K,,
[1301
CHEMICALLY
REACTING BOUNDARYLAYERS
NONEQUILIBRIUM
is usually very large for the combustion temperatures of 3000 or 4000" K.

T h e backward reaction of Eq. (70) is therefore usually negligible and
the combustion rate becomes
It is now seen that the exponential term which is important in a com-

bustion problem is that based on the activation energy, exp(-Ea/RT).
Equation (76) may not be sufficient for the reaction of a combustible
gas injected into a hypersonic boundary layer. Often, especially near
the stagnation region, the temperature of the combustion zone may
rise to much above 4000 OK when a vehicle is flying at hypersonic speed.
For such cases the equilibrium constant K, can become sufficiently
small that the reverse reaction may not be negligible. In practice, of
course, decomposition reaction of Mp other than that given by Eq. (70)
would also set in when the temperature is extremely high.
Now we turn to the dissociation-recombination processes of diatomic
gases. Some of the most important ones are those which occur in the
high energy air boundary layers associated with the aerospace technol-
ogy. Most of the high temperature dissociation-recombination processes
of gases such as oxygen and nitrogen are known to take place as (15-17).
A, + A , + X*A, + x
kf
kb (77)
where A , and A , designate the atomic and the molecular species,

respectively. X denotes the gas particles which act as a catalyst. I n a
binary mixture of atoms and molecules, X may be either the atoms or
the molecules. The recombination therefore, is a third-order reaction
whereas the dissociation is a second-order reaction. Here we write
Eq. (77) such that the forward reaction leads to a lower energy state as
was the case with the preceding problem of combustion (See Fig. 3b,c).
Since k, and k, represent the specific rate coefficients for recombination
and dissociation, respectively, we let
k, = k R , k, = k~
From the law of mass action, we write for the production rate of the
atoms
- - - -2kR(X)[(A,)2 - K,(A,)]
d(A1) (78)*
dr
Many authors include the factor 2, appearing in the law of mass action of Eq. (78),
in kR for convenience. The KR thus defined is then equal to (2kR)of the present definition.
PAULM. CHUNG
where
K ( A )' = Z E ( T )exp [--dE"/(RT)]
- _
kD=
- kR [2
(A,) IE
(79)
Notice here that the equilibrium constant is defined as k,/kf rather than
k,/k, so that AE" would be positive. Now as before
KR = Z R ( T )exp [--E,/(RT)I (80)
It is known, however, that the three-body recombination processes of
the present type require negligible activation energies. Therefore, Eq.
(80) becomes
k ~ T( ) = ZR(T ) (81)
I n fact, it has been quite well established experimentally that for the
recombinations of oxygen and nitrogen atoms
Much of the recombination rates have been obtained through shock

tube experiments in which the dissociation rate is measured behind a
strong shock (see Table 11). Let us therefore consider the form of k , .
From Eqs. (79) and (82)
kD = ZE(T)ZR(T )exp [ -AE"/(RT)] (83)

Thus the exponential term which appears in the dissociation is that
based on the heat of reaction rather than E, .
5. Dissociation- Recombination of Diatomic Gases

More work has been done on the boundary layers with dissociation
and recombination reactions than with any other chemical reactions
because of its importance in the hypersonic flows. We shall therefore
investigate the properties of dissociating oxygens and nitrogens a little
further.
First, let us consider the equilibrium relationships. I n Eq. (79) the
equilibrium constant K, was defined in terms of the equilibrium mole
concentration of atoms and molecules. It is more convenient in engineer-
~321
CHEMICALLY
REACTING
BOUNDARY
ing applications to employ the mass fractions, C , rather than the mole
concentrations. We let the subscripts 1 and 2 represent the atoms
and molecules, respectively. Then
c, + c, = 1 (84)
Eq. (79) can now be rewritten as
and
T h e partition function Qc for the various species of air have been

<
calculated elsewhere [for instance, see Ref. (14) for TE 15,000 OK].
Lighthill (18) defined the quality Q;J4Qc, in Eq. (86) as the charac-
teristic dissociation density, pD , and showed that it varies very slowly
with respect to temperature for T E ,< 7000 OK. He thus suggested to
assume pD as constant and the equilibrium relationship be given by
T h e gas whose equilibrium relationship is given by the above equation

is commonly called a "Lighthill gas" when the contribution of the
vibrational state to the specific heat is assumed to be one-half that of
the fully excited state. T h e values of pD are tabulated in Ref. (18) for
oxygen and nitrogen for various temperatures and are reproduced here
in Table I.
TABLE I
CHARACTERISTIC
DISSOCIATION
DENSITY
(18)
p~ (gm/cm3) when T (OK) is:

1000 2000 3000 4000 5000 6000 7000
Oxygen 145 170 166 156 144 133 123

Nitrogen 113 135 136 133 128 123 118
As was mentioned previously, the predominant temperature depend-

ence of an equilibrium constant is usually due to the exponential
function. An equilibrium relationship even simpler than Eq. (87) can
be deduced from Eq. (86) with a slight added approximation. As it can
[I331
PAULM. CHUNG
be seen from Table I, p D , though slowly, does vary with temperature.

It can be shown that the parenthesized quantity in the right-hand side
of Eq. (86) may be considered a constant for T 7000 O K without2
losing much accuracies. Thus we let
=-
1 exp (16.2) exp (- --)ARTE o =
4p
16.2 - --)ARTE o (88)
Equation (88) is found to be satisfactory for oxygen and nitrogen for

engineering purposes (19).
T h e quantity BEo/? is defined as the characteristic dissociation
temperature, T, , and is equal to 61,000 and 116,000 OK, respectively,
for oxygen and nitrogen.
I t is seen from Eq. (88) that practically no dissociation takes place
in air at temperatures below about 2000 O K . T h e oxygen begins to
dissociate at about 2000 OK. T h e nitrogen then begins to dissociate at
temperatures between 3500 and 4500 OK depending on the pressure.
Let us now turn to the specific reaction rate coefficients. As it can be
seen from Refs. (15-1 7), refined rate coefficients are continuously being
measured. Some of the more recent values for the major reactions
occurring in pure air (17) are given in Table 2. The reactions involving
TABLE I1
MAJORNEUTRAL
G A S REACTION
RATESFOR AIR BOUNDARY ( I 7)
LAYERS"
A" X
Reaction (eV) Catalyst
(1) 0 2.25 x loz0T - 3 / n

op+ x e 0 + 0 + x 5.1 0 2 0.8 x 1020'PsJ2
kR NZ 6.19 x 1015 ~ - 1 1 2
N , NO 3.02 x 1015 T - ~ / z
-
(2) N 2.36 x lozLP 3 i 2
N, + X s N +N +X 9.8 Nz 2.76 x 10lET-l12
kit o,,o,NO 1.09 x lo1*T-'/*
~
(31 NO 2 x loz1PS/*
NO +X 2 N f0 +X 6.5
kR 0, N, 0 2 N2
, 1.02 x loz0T-3/2
)
(4) cm3
0 + N, s NO + N 3.3 1.62 x loLs
(mole sec
~
kR
~ ~ ~~
Note: T is in "K.
NO are not important in the analysis of heat transfer because the

equilibrium concentration of NO and thus the energy associated with
the NO reactions are negligible. I t is, however, very important for the
study of radar observables. Reaction 4 of Table 11, for instance, may
produce N along with NO at the temperatures much below the dissocia-
tion temperatures of N, . N thus produced can quite readily react with
0 to form NO+ and an electron. I n fact, for slender bodies, the process
mentioned above is one of the leading processes of forming electrons
which determine the radar cross sections.
B. HETEROGENEOUS
CHEMICAL
REACTION
T h e heterogeneous reactions of the present interest are mainly those
which occur at the gas-solid interface. Often the detailed theoretical
treatment of the surface chemical kinetics is even more complex than
that of the homogeneous ones. Here, we shall only consider the simple
phenomenological description of the reaction laws pertinent to the
present study.
There are basically two types of surface reactions that are of interest.
They are those wherein the surface acts as a catalyst for the gas reactions
and those in which the surface participates in the chemical reaction
respectively. T h e catalytic surface recombination of atoms is an important
example of the former category whereas the surface combustion is that
of the latter.
Consider the following overall surface reaction:
T h e wall may either act only as a catalyst or react with A and B. KW1 and
K,, are the specific rate coefficients for the forward and backward
surface reactions respectively. Phenomenologically, the rate of production
of the reactant per unit area JA,w may be expressed by
J A . ~= --Kw,(A)nl + -Kw#P (90)
where n, and n2 are the orders of the forward and backward reactions,
respectively. T h e units of and K,s depend on the units employed
for the concentrations of the reactants and products.
If a sufficient amount of time is allowed for the reaction5 the surface
For a steady-state but nonstatic system, this implies that the reaction takes place at
a rate much greater than the rate at which the gaseous reactants are diffused to or from
the surface.
P351
PAULM. CHUNG
condition will approach an equilibrium state wherein there is no further
variation in the concentration of gaseous species. For equilibrium then
Eq. (90) gives
Eq. (90) is now written as
I n a boundary layer analysis the surface reaction enters the problem

through the wall-boundary condition. The reaction rates such as given
by Eq. (92) are equated to the appropriate diffusion fluxes of the species
at the wall. T h e solution of the boundary layer equations with these
wall conditions gives the wall concentrations of reactants and products.6
Let us now consider a few particular surface reactions of our interest.
First, consider the catalytic recombinations of dissociated diatomic
gases. T h e surface reactions associated with such recombinations are
usually known to be first-order reactions. Thus if we let A and B of
Eq. (89) represent atoms A, , and molecules A , , respectively, the rate
equation (92) becomes
where k, = k,, . The equilibrium ratio (A1)/(A,) is extremely small

for oxygen and nitrogen atoms recombining on a surface whose tem-
perature is below about 2000 OK [see Eq. (88)]. The second term of
Eq. (93) therefore, can usually be neglected for the recombination of
oxygen and nitrogen atoms. T h e specific reaction rate coefficient k,
depends on the surface material as well as on the gaseous reactant.
It is also a quite strong function of the surface temperature. T h e values
of k, have been obtained experimentally for recombination of oxygen
and nitrogen atoms on various surfaces. Some typical values given in
Ref. (20) are reproduced here in Fig. 4.
For most of the surface reactions prominent in the catalytic reactors
of chemical industries, the equilibrium ratio, (A):l/(B)P, is usually
also of negligible magnitude. T h e reaction rate of Eq. (92) can there be
commonly written as
JA.ro = --kw(A)" (94)
More detailed analysis will be given in the subsequent Section IV.

~361
7mF-- ,=lcm/sec -k, = 10
k, = 10'
4F
Licr
I o-' Io -~ 10-2 10'1

CATALYTIC EFFICIENCY Y'
FIG.4. Specific rate coefficients for catalytic recombination of 0 and N.

[y = d 2 r M J R T k,]: (Ref. 20)
Finally, consider the surface combustion. T h e gaseous reactant, A,

in this case is the oxygen. As was with the gas phase combustion dis-
cussed previously, the backward (endothermic) reaction is negligible,
especially at the surface temperatures which are usually much lower than
the temperatures of the gaseous flame zones. Since in most of the experi-
mental work the rate of variation of solid fuel is measured rather than
that of the gaseous oxygen, we express the consumption rate of fuel,
from Eq. (92), as
-J,,,o = kw(A)n (95)
T h e specific rate coefficient K, is expressible in the following Arrhenius

form
ku = KO exp [--E,,/(WI (96)
where KOis a constant or a function of temperature, usually referred to

as the frequency factor, and E,, is the activation energy for the surface
reaction. For the combustion of carbons, Scala (22) has shown that
[I371
PAULM. CHUNG
much of the reported kinetic data can be bracketed by the following

kinetic constants:
fast reaction
KO= 6.73 x lo* lb/ft2 atm sec
E,, = 44 K. cal/mole
iislow reaction
KO= 4.47 x lo4 lb/ft2 atm sec

Eaw = 42.3 Kcal/mole
with which partial pressure of oxygen is to be used in Eq. (95) instead
of mole concentration (A). Various investigators (see for instance, Ref. 22)
have found reaction orders, n, between zero and one.
IV. Boundary Layers with Surface Reactions
T h e surface chemical reactions in boundary layer flows will now be

considered. It will be assumed that no chemical reaction takes place in the
gas phase. T h e boundary layers with gas phase reactions will be con-
sidered in a later section.
A. GOVERNING
EQUATIONS
AND BOUNDARY
CONDITIONS
T h e governing equations are those derived in Section I1 with Wi = 0
since the gas phase reaction is considered to be frozen. Specifically,
the equations of immediate concern are the continuity equation ( l ) ,
momentum equation (2), energy equation (22), and the species eonser-
vation equation (16). T h e boundary conditions are given by Eqs.
(30)-(40). With the aid of Eq. (92) we can now rewrite the boundary
condition (40) at y = 0 in a more specific manner:
T h e specific rate coefficient A, is considered known (See Section 111)

as a function of x. The subscripts i and j in the above equation designate
the reactant and the product, respectively.
With W, = 0, the governing equations and the boundary conditions,
except the boundary condition (97), are essentially the same as those
~381
CHEMICALLY
REACTING
NONEQUILIBRIUM
BOUNDARY
LAYERS
for the classical boundary layers which have been fully analyzed else-
where. Basically then, the only new analytical problem at hand is to
solve the species conservation equation with the wall boundary condition
given by Eq. (97).
Following the customary boundary layer practice, we shall first
investigate the possibility of obtaining a self-similar solution. For this
purpose we transform the governing equations and the boundary
conditions in the following manner.
TO 4-f COORDINATES
B. TRANSFORMATION
Following Lees (23) and Fay and Riddell (24) we define
and
P = J X pl.peuer2c
dx (99)
0
T h e continuity equation (1) is automatically satisfied by introducing the

stream function 4,which is defined by the usual relations
and
Define a nondimensional stream function f as
We then have
j; = Y
ue
and
pv= --
If
1
[(dWi+ -1
dzf)P, + d Z j 4 (104)
Also we define
PAULM. CHUNG
Now Eqs. (2), (22), and (16) are respectively transformed from (x, y )
coordinates to (4, f ) coordinates for constant Pr and Sc as:
Momentum
Total energy
where C, is considered constant. For the frozen boundary layers of

present interest, it is more convenient to employ an energy equation
based on the frozen total enthalpy h,, rather than Eq. (107). As can be
seen from the sentences following Eq. (29), the frozen total energy
equation takes exactly the same form as Eq. (107) and is
Frozen total energy
where
T h e appropriate boundary conditions found among Eqs. (30)-(40) and

the boundary condition (97) are transformed as follows:
at7j=O
ji, = 0
c 1401
CHEMICALLY
REACTING
NONEQUILIBRIUM
BOUNDARYLAYERS
H = Hw(f)
and at i j -+ co
f+)= 1
H = l
mi = 1
H, = 1
where
and
In the above, X = x/L where L is a characteristic length of the body

(see Fig. 1)
C. MOMENTUM
AND ENERGY
EQUATIONS
I t is seen that the momentum equation and the frozen total energy
equation (109) are coupled to the species conservation equation only
through I, p,/p and the fourth term of Eq. (109). For most engineering
' H,(cu) = 1 because C , , is considered to be constant.

~411
PAULM. CHUNG
purposes the dependence of I and pJp on Cican be neglected in com-
parison to their dependence on temperature. Moreover, as we shall see
subsequently, 1 may be often assumed as constant, and the entire third
term of the momentum equation is negligible for many cases of our
interest. T h e fourth term of Eq. (109) vanishes when either Lei = 1
or all cpi are equal. Again for many engineering problems, Le, is close
to unity while cpiare similar. This term therefore can safely be neglected
for the present purpose. Finally, consistent with the above approxima-
tions, the boundary condition (1 15) may be written as
( 1 15a)
where cpw is the specific heat of the gas mixture at the wall and is con-
sidered to be known.
Now we see that the momentum and the frozen total energy equa-
tions together with their boundary conditions are those of chemically
inert boundary layers and, for all practical purposes, are decoupled
from the species conservation equation. T h e flows for which the partial
differential equations for momentum and frozen total energy can be
reduced to ordinary differential equations with 4 as the sole independent
variable (self-similar flows) have been explored fully elsewhere. We shall
only mention here that in general, the types of flows amenable to self-
similar analysis are the flows associated with flat plate, subsonic wedges
(Falkner-Skan), superso~nicwedges and cones, and the hypersonic blunt
bodies. It is further noted here that the momentum equation (106)
for these flows, except the Falkner-Skan flows, become reduced to the
following Blasius equation:
(lf) +jj = 0 ( 122)
where the denotes total differentiation with respect to
(I) 4. T h e self-
similar energy equation is from Eq. (109)
Equation (123) is applicable to the Falkner-Skan flows also, since for

this subsonic case, the third term is negligible. T h e self-similar nature
of the above flows are not destroyed by surface mass transfer, (pwziw),
only if the mass transfer is distributed along the surface in such a manner
that the right-hand side of Eq. (1 1 1) is constant. For this case Eq. (1 1 1)
becomes
(Illa)
CHEMICALLY
REACTING
BOUNDARY
In order not to unduly complicate the analysis, we shall in the fol-

lowing, confine the study of the species conservation equation to the
flows wherein self-similar solutions of the momentum and frozen energy
equations exist.
D. DAMKOHLER FOR SURFACE

NUMBER REACTION
Before considering solutions of the species conservation equation, we
shall investigate the physical meaning of the function pi(s) defined by
Eq. (120). It is because this function plays an important role in the sub-
sequent analysis of boundary layers with surface reactions.
Consider that a uniform diffusion potential. for component i,
(Ci, - C,,), is applied across the boundary layer along a body. I t is well
known for such case that the right-hand side of Eq. (108) vanishes and
mican be solved as function of i j only. T h e rate of diffusion of component
i at the surface Jiw , is then given by
For a given (C, - CiW)and I,, , mi,,,is constant. T h e rest of the term
in the right-hand side if Eq. (124) therefore represents the characteristic
diffusion rate of the component i for unit surface area and for unit
concentration difference across the boundary layer. Now returnin
Eq. (120), it is seen that the denominator of the equation defining i ( s )PtO
is exactly this characteristic diffusion rate. T h e numerator of Eq. (120),
kw(pwCie)nf,on the other hand, represents the characteristic surface
reaction rate for unit surface area. The function ti,therefore, denotes
the ratio of characteristic surface reaction rate to characteristic surface
diffusion rate for the ith species. Also, one may take the reciprocals of
the respective characteristic rates and, after multiplying each of them
by the product of density and boundary layer thickness, consider ti
as the ratio of characteristic diffusion time (Tdiff) t o characteristic surface
reaction time ( T J . Thus we may write
ti($)= (---characteristic surface reaction rate
characteristic surface diffusion rate 1
= (2y)i (125)
Such characteristic time ratio is called surface Damkohler number

for the component i.
[I431
PAULM. CHUNG
Now we are in a position to clearly define the two extreme cases of

frozen and equilibrium surface reactions.
Let us first consider the case in which ti --t 0. It is seen that the
right-hand side of Eq. (1 14) is zero and therefore, the effect of surface
reaction on the species boundary layer is practically nil. Such case is
defined as the flow with frozen surface reaction. It should be noted
here that the frozen surface reaction does not necessarily imply that
R, --+ 0. It only implies that the reaction rate is much smaller than the
diffusion rate though the absolute value of the reaction rate itself may
be quite large.
Next, consider the other extreme case of f i -+ 03. The left-hand side
of Eq. (1 14) is finite. Therefore, the quantity in the bracket at the right-
hand side of the equation must approach zero as f, -+ co. This means
that (m,),---+ (mi), . This extreme limit wherein the surface concentra-
tions of the species approach those corresponding to the equilibrium
state, is deffined as the flow with equilibrium surface reactions.
Again one should note that the equilibrium limit does not necessarily
imply that K , --t 00, but rather that the reaction rate is much greater
than the diffusion rate. Consider for the equilibrium limit, that either
(mi& is invariant with respect to x or it is, as in most cases of our
interest, negligibly small. For these cases, the wall-boundary condition
(1 14) can be replaced by the a priori known value (mi), = (m&, , and
the solution of the conservation equation can be obtained in the con-
ventional self-similar manner.
T h e full boundary condition (1 14) must be used for the general cases
of finiteti, and the analysis of the species conservation equation for
finite values of ti constitutes most of the remainder of the section.
When ti is finite, miW # (mi), and the gas layer at the surface is in a
state of chemical nonequilibrium. These general cases are therefore
defined as the flows with nonequilibrium surface reactions.
E. SELF-SIMILAR
SOLUTIONS
OF SPECIES
CONSERVATION
EQUATION
Now we turn to the species conservation equation (108) with the
boundary conditions (1 14) and (1 18). Within the same degree of approxi-
mations made in the preceding section, the right-hand side of Eq. (1 14)
can be written as
Now for the flows mentioned in the preceding section C, wherein self-
similar solutions are available for the momentum and the frozen total
energy equations, the stream function .f is a function of only, and Eq.

(108) and the wall-boundary condition Eq. (1 14) become, respectively,
and
Now let us consider the boundary value problem comprising Eq. ( 1 26)
and the two boundary conditions (1 18) and (127). A study of Eq. (126)
and the boundary conditions shows that mi becomes a function only
ti
of 4 if and [(mi)Lfi/(mj)2],are respectively constant. T h e equilibrium
ratio [ ( m i ) ~ ~ / ( r n is
j ) constant
~], when the pressure and the wall tempera-
ture are constant along the surface. Even when the pressure and T,
vary, this term usually does not present difficulty, because for most
engineering problems, it is of such small order of magnitude that the
last term of Eq. (127) can be neglected. Next consider the function pi .
A study of Eq. (120) shows that ti is finite and constant for a given
value of K, only at the stagnation region of blunt bodies where u, N x
and p and pw are constant. Therefore, an exact self-similar solution of
the boundary value problem exists only at the stagnation region when
ti is finite.
There are few special cases which approximately satisfy the self-
similar criteria. For these cases, the "locally similar" solution should
produce sufficiently accurate results. T h e streamwise variation of
the Damkohler number given by Eq. (120) is due to the quantity
(p,/p,o)ni/F(X). Suppose now that the inviscid flow around a certain body
shape is such that the quantity (p,/p,,)ni/F(X) vary very slowly with
respect to x. A satisfactory locally similar solution then could be obtained
for this body with the accuracy of the solution depending on the degree
to which (p,/p,)"',F(X) approximates a constant. T h e hypersonic flow,
for instance, around blunt bodies comprised of a spherical nose and a
conical after-body is found to satisfy the above approximate criterion
when ni = 1.
T h e numerator and the denominator of the function (p,/p,,)ni/F(X)
represent the variation of the surface reaction rate and the surface
diffusion rate, respectively. T h e fact, therefore, that this function
varies slowly with respect to x implies that the surface reaction and
diffusion rates vary in a similar manner. T h e resulting surface concentra-
tion mi, thus varies very slowly along the surface and, for this case,
[I451
PAULM. CHUNG
it is well known from the conventional boundary layer analysis that a

locally similar solution is satisfactory.
In the following, we shall study two self-similar cases and a locally
similar case. For the former cases we will study the surface catalytic
reactions with and without surface mass transfer, and a surface com-
bustion problem all at the stagnation region. For the latter case, we shall
study the first-order catalytic surface reaction around the hypersonic
blunt body.
1. Surface Catalytic Reactions with and without Mass Transfer at

the Stagnation Region
T h e problem of surface catalytic reaction at the stagnation region was
analyzed by Scala (25) and Goulard (20) without mass transfer and by
Chung (26) with mass transfer. Scala obtained exact numerical solutions
for a selected flow condition by the use of a digital computer. We shall
herein rather follow the analyses of Goulard (20) and Chung (26) which,
though less accurate than the analysis of Scala, gives more analytic and
general treatment of the problem.
For simplicity, we shall confine the analysis to that of the boundary
layers where p,/p, << 1. I t is now well known that (see Ref. 23) for such
boundary layers, Eq. (122) is applicable as the momentum equation and
also the third term of the energy equation (123) is negligible. Moreover,
for most engineering purposes, one may consider I to be constant and
(see Section V, D , le)
For constant 1, we redefine the similarity variable and the stream function
as
and
It is considered that the boundary layer comprises a binary mixture

of a reactant and a product. We let C and C, represent the reactant and
the product, respectively. Thus for the surface recombination of atoms,
the former represents the atoms and the latter the molecules. I n the case
of a dissociated but frozen air boundary layer, the atoms and molecules
~461
CHEMICALLY
BOUNDARY
LAYERS
of oxygen and nitrogen may respectively be grouped together to com-

pose a binary mixture since the atomic weight and the diffusion coeffi-
cient of oxygen and nitrogen are similar. T h e momentum equation (122),
frozen total energy equation ( 1 23), and the species conservation equation
(126) then become, respectively,
f+f f = 0 (130)
1
-H f
Pr
+ fH, =0 (131)
1
- m
sc
+fm =0 (132)
where the prime () here denotes the total differentiation with respect
to 7. m 2is obtained by the difference ( l/Cze)(l - Gem). At the stagnation
region where all properties along the boundary layer edge remain
constant and u, = /3x, the boundary conditions given in Eqs. (1 11) to
( 1 19) become:
atq=O
f=f - - (P4W
w -
d(1 +E ) ( P e d L P
f=O
H f = Hfw = (CIlwTtu)/hfe
and at q = co
f = 1
Hf = 1
m = l
In the above, surface Damkohler number for constant 1 is defined more

conveniently as
which becomes at the stagnation region

PAULM. CHUNC
I t is also assumed in Eq. (136) that the gas injected into the boundary
layer from the wall does not contain the reactant, thus (mu,)- = 0.
Solutions of Eqs. (130) and (131) are available elsewhere (see for
instance, Refs. 1 and 23).
Integrating Eq. (132), we obtain
When the boundary conditions (1 36) and (139) are applied to Eq. (142),
there results
where
T h e function G(Sc, fw) is given in Ref. ( I ) and is reproduced here, with

some modifications, in Fig. 5. G(Sc, f,) is equal to the well-known
value 0.47 S Cwhen
~ f, = 0.
m, which gives the surface reaction rate, J w , is now obtained by
solving the algebraic equation (143) with the aid of Fig. 5. As was ex-
plained in Section 111, the equilibrium ratio [mEn/(m,E)n2],is usually
negligible. Also, many catalytic reactions are first order or between
8 and first-order reactions. For mEn/(mZE)n2= 0, Eq. (143) can be solved
as :
for n = 1
and for n = a
. r. sc f,,, 1
Some of the typical values of m, given by the above equations are

plotted in Fig. 6 . It is seen that m, for all n and fE approaches the equi-
~481
CHEMICALLY BOUNDARY
LAYERS
2.4 - I 1 I I I I I
0.8 - -
0.4 -
0 0.25 0.50 0.75 1 .o

-vT ,f
FIG. 5. Function G(Sc,f,); Ref. ( I )
0.8 - -
0.6 - -
s
E
0.4 - -
0.2 - -
0
10-2 1 10 Id
to
FIG,6. Variation of rn, with respect to lodefined by Eq. (141).
[I491
PAULM. CHUNG
librium value, which is zero, as 5, -+ 00. In the other extreme of CO+O,

the m, assumes the value l / ( l - (Scf,)/G) for all n. For the inter-
mediate finite values of to, m, is a function of both f, and n. T h e
transition from the frozen limit (5, + 0) to the equilibrium limit
(to+ a)is seen to take place faster as n is decreased.
Let us now study the surface heat transfer. For convenience we shall
consider that h, = 0 for it does not alter the basic nature of the present
analysis. We assume that the specific heats of the reactant and the
product are equal. Then the heat transfer relation of Eq. (47) becomes
where Ah" = h," - h," and Ah" is positive heat of recombination.8 T h e

first term on the right-hand side of Eq. (147) represents the heat transfer
due to the convection of thermal and kinetic energies, whereas the second
term represents that due to the surface reaction and the subsequent
release of the heat of reaction Ah". We shall denote the former by qc and
the latter by qd . qc is obtained from the solution of Eq. (131) as (see Refs.
1,231
1
+
4 c = pr d(1 4PCCLR)OB~( h f , - ~ t , w ) G ( P r J w ) (148)
where m, is given by the solution of the species equation such as Eqs.

(145) and (146).
Let us elaborate further on the heat transfer expressions for n = 1
because many of the important catalytic reactions such as surface
recombination of air atoms are first order and also the manipulation
is simplest when n = 1. Substituting Eq. (145) for m, in Eq. (149),
Eq. (149) gives, after some rearrangement,
50
Combining Eqs. (148) and (1 50), we obtain for the total heat transfer
=jqd(l + 4(PCCLP)"B~G ( P r , f w ) ( h , f P - h , w )
1
9w
8 For the surface recombination of dissociated air, Ah" is the averaged value for oxygen
and nitrogen based on their respective concentrations at the boundary layer edge.
r I 501
CHEMICALLY
REACTING
BOUNDARY
T h e second term in the curly bracket is qa/qc.It is zero when the surface
reaction is frozen (5, -+0) and it is equal to the maximum value,
L e G(Sc
A- fw)
G(Pr, fw)
(-'-
~ ~ - ) when co -+ co for all f,.
htfe htfto
When Le = 1, the
maximum contribution of the surface reaction to heat transfer becomes
C,dh"/(h,,,- htt,) which is in accordance with that predicted earlier
by Eq. (48).
T h e manner in which qc is reduced by the surface mass transfer is
well known. Let us investigate the simultaneous effect of f w and $,
on qd . I t can readily be deduced from Eq. (1 50) that for n = 1 and for
all Sc,
where the subscript E denotes the equilibrium surface condition reached

as co -+ co. I t can be readily shown with the aid of Fig. 5 that the
quantity [G(Sc, f,) - S c f J varies very little with f, when Sc is order
one. Therefore the ratio (qd/qdE)given by Eq. (152) is quite insensitive
to surface mass transfer for the entire range of l o . Hence, for n = I ,
Noticing the fact that = (qc)fw/(qc)fw=o

(4iE)f,J(~dE)f,=0 , we write from
Eq. (153) for n = 1,
(qd)f,
__- (qdE)fw (qC)f,
---
% ( 154)
(qd)f,-O (qdE)f,.=O (qc)f,-O
Now we see that the relative effect of surface mass transfer on reducing
heat transfer is fairly constant for all Damkohler numbers.
Such, however, is not the case when n = Q. We can qualitatively
verify this with the aid of Eqs. (146) and (150). Since the quantity
[G(Sc, f,) - Scf,] is rather insensitive to f, , it can be shown by
rearranging Eq. (146) that, for n = 8 and finite values of , qd varies
as
at the most. q d E , on the other hand, is independent of n because

(mE),-+0 for all n. Hence, it can be shown with the aid of Eq. (150) that
PAULM. CHUNC
Eqs. (1 55) and (1 56) give
gd- 1
(157)
QdE dG(Sc,lfUT
Now as it is seen in Fig. 5, G decreases with mass transfer. Therefore,
and
(qc)f,
--->
(qd)f,,, (qdE)f,,
____ N -- (158)
(Qd)f,.=O ( 4 d E ) f -0 (Qc)f,. -0
Thus, the mass transfer is not as effective in reducing qd as in reducing

qc when n < 1 and 5, is finite.
2. Surface Combustion of Graphite at Stagnation Region

Surface combustion, instead of gas phase combustion, prevails for
a graphite when the heat transfer rate is such that the surface temper-
ature is below about 5000-8000R depending on the particular grade
of graphite. T h e surface combustion therefore is prevalent in many
engineering applications including the ablative heat shield for space
vehicles. We shall study the surface combustion rate of graphite in
stagnation boundary layer flows.
Perhaps the surface burning of carbon is one of the most familiar
combustion processes of all time. Yet, three fundamental problems
associated with it have not been well understood: (a) the heterogeneous
kinetics of combustion at low surface temperatures; (b) the relative
proportions of CO and CO, produced by the combustion at the surface
at any surface temperature; and (c) the interaction of (a) and (b) with
the boundary layer which supplies the oxygen.
Scala (21) gave a rather complete survey of literature pertaining to
the combustion of graphite along with the numerical solutions of the
stagnation boundary layer equations for the diffusion-limited combustion
regime. By diffusion-limited regime we mean the regime wherein the
Damkohler number is very large so that the surface is in equilibrium, and
thus the combustion rate is governed solely by the rate at which the
oxygen is diffused to the surface. Scala also considered the regime of
nonequilibrium surface reactions by the so-called series-resistance
method. This method, however, is not correct in general as was shown
by Welsh and Chung (27) and others.
[I 521
CHEMICALLY
NONEQUILIBRIUM
Welsh and Chung (28) analyzed the coupling of the three phenomena,
(a), (b), and (c), in a self-consistent manner for the stagnation region.
Particularly, a new concept was advanced which connected the propor-
tions of CO and CO, products to the nonequilibrium surface reaction.
T h e governing equations and the method of analysis were essentially
the same as those employed in the preceding section for the catalytic
surface reactions. We shall herein give the main steps of the analysis
given in Ref. (28).
T h e main difference between the catalytic reaction problem and the
present combustion problem can be explained from the surface boundary
condition (136). Eq. (136) is applicable to the present problem when
m is considered to denote the normalized mass fraction of the gaseous
reactant, the oxygen. In the case of catalytic reaction, f, represented the
mass transfer rate of the coolant injected at the surface at an arbitrary
rate. I n the present case, f, represents the combustion rate of carbon
which is being sought.
In addition, as we shall see, the parameter 5 of Eq. (136) becomes a
function of m. This, however, presents no problem for we can always
redefine the Damkohler number such that it is free of m.
T h e surface reaction of present interest is described by the stoichio-
metric relation
y,O* + Yccs-Y~coi+ rsCO (159)
where y denotes the number of moles and C, designates the carbon in

solid state. For a given supply rate of oxygen, the combustion rate of
carbon varies by up to factor of two depending on the proportion of
CO, and CO produced, y z / y 3, and this proportion is not known a priori.
In Section 111, the surface combustion rate of carbon was derived in
terms of the partial pressure of the oxygen at the surface. Changing the
partial pressure to the mass fraction of oxygen, the mass rate of carbon
combustion, J c [see Eqs. (95) and (96)], becomes
where subscript 1 denotes the gaseous reactant, oxygen. T h e equation

of state, Eq. ( 5 ) , was also used in deriving Eq. (160). T h e mass rate of
oxygen consumption, J1, is related to Jc from Eq. ( 1 59) as
PAULM. CHUNG
Now in the following we shall solve for the oxygen consumption rate J1,
and the ratio y l / y r . which determines the ratio y z / y 3 .The carbon com-
bustion rate J r will then be obtained from Eq. (161).
T h e species conservation Eq. (132) for oxygen becomes Eq. (142)
upon integration. I t can be seen in Fig. 5 that for the relatively small
values off, of our interest (f, 5 - 0.2) the integral G(Sc,,f,) can be
approximated by
G(Sc,fw) = dl + A,f, (162)
where for Sc = 0.51 (Le = 1.4 when Pr = 0.72)
A, = 0.372, A, = 0.266 ( I 63)s
Hence, upon applying the boundary condition ml(co) = 1, Eq. (142)
becomes with the aid of Eq. (162)
Now the wall boundary condition (1 36) becomes

Eqs. (160) and (161)
where anticipating that ylA?l/yr.A?cwill be a function of m,, the Dam-

kohler number is redefined as
-
- KO exp [--Ea,"/(RTw)l [P(AIIC;l1)wln(C1PY-l
I1 66)
du + 4 ( P e c L e ) o P
As was mentioned earlier in this section, fw represents the combustion
rate of carbon, and from Eqs. (133) and (160),
fw = -LaC1e(m,w)n ( 167)
Now we have three algebraic equations, Eqs. (164), (169, and (167)
with four unknowns, m,, , mi,, f, , and ylii?JyC&Zc. Eliminating
mi, and fw from the three equations we obtain
These values of Sc and Pr are suggested for carbon combustion in Ref. (1).
c 1 541
CHEMICALLY
NONEQUILIBRIUM
We shall now express the ratio y l f i l / y C f i cin terms of m,, so that

Eq. (168) may be finally solved for m,, . For this purpose, we first
postulate that the carbon surface is an ideal catalyst in promoting the
following chemical equilibrium of the gas at the interface:
2CO,~2CO+O, ( 169)
T h e equilibrium constant K p for the above reaction is derived as
K, =
(Pco)2(Po,)
__I_
(Pco,)2
= exp
68.224 x lo3
(20.926 - --_____
T 1 (170)
where T is in OK. Then the ratio of CO-to-CO, partial pressures is
obtained from the above equilibrium relationship as
Now from Eqs. (159), (170), and (171), the ratio y l M l / y c f i cbecomes
It is seen in the above equation that as 5, --+ 03, and hence m,, -+ 0,
T h e stoichiometric relationship Eq. (159) becomes for this limit

O2+2C-t2C0 (174)
showing that CO is the sole product of combustion. Now Eq. (168) can
be solved for mlw with the aid of Eq. (172). T h e combustion rate of
carbon is then obtained from Eq. (160).
T h e maximum diffusion-limited combustion rate, (Jc)d,I , wherein
-+ 0, may be obtained rather directly from Eq. (164) with the aid
m,,,,
PAULM. CHUNG
Now let us turn to the Damkohler number, con,

for the surface com-
bustion of graphite defined by Eq. (166). It is seen that the Damkohler
number is dependent most strongly on the surface temperature through
the exponential function. A few typical combustion rates computed from
the solutions presented herein are shown in Fig. 7. T h e combustion
1.2- I I 1 I I I I I
CD
w
c 0
2 0.8-
Z
0
c
Y)
0 DATA OF REF 28 6 30 ( A )
-
0 DATA OF REF 29 f B )
THEORETICAL RESULTS
(n =1/2. "FAST" REACTION)
A 0.107
B 0.019 0.0097
I I I
1200 1600 Zoo0 2400 2800 3200 3600 4000 MOO 4800
SURFACE TEMPERATURE, RANKINE
FIG. 7. Surface combustion of graphite.
rates are shown normalized by the diffusion limited rate of Eq. (1 7 5 ) and
are given as the function of the surface temperature. Also shown are
the experimental values obtained by Diaconis et al. (29) and Welsh (30)
from their respective arc tunnel tests.
T h e results of the theory show the following interesting phenomena.
With the initial rise in the surface temperature and hence the Dam-
kohler number, the surface reaction enters into the nonequilibrium
regime. This regime is seen to be traversed rather rapidly, and the
temperature rise of about 400"R is seen to bring the abscissa to the
first plateau with J J ( J c ) d , I of &. During this plateau, m,, is sufficiently
small compared to one, and the combustion is practically diffusion-
limited with CO, as the predominant combustion product. With the
continuous increase in surface temperature, a point is reached wherein
the CO begins to appear among the products. T h e carbon combustion
[I 561
CHEMICALLY
REACTING
BOUNDARY
rate increases with the ratio of CO to CO, until the maximum combustion
rate is reached wherein CO is the sole product. Summarizing, there
seems to be two transitions occurring in carbon combustions when c,,,
is varied. T h e first is that due to the variation of surface chemical
condition from the frozen to the equilibrium regime; and the second
is that due to the variation of the equilibrium combustion product from
that comprised predominantly of CO, , to that comprised predominantly
of CO. T h e two transitions may occur at two distinctive Damkohler
number regimes or may occur simultaneously depending on the property
values such as KOand E,, .
T h e agreement between the theory and the experimental results
(see Fig. 7) is seen to be satisfactory as a whole. Moreover, the trend of
the experimental data seems to support the general phenomena discussed
above though no definite conclusion can be drawn.
I n the analysis presented, the oxygen at the surface is considered to
be in molecular form and I is assumed to be constant. T h e oxygen should
predominantly be in the molecular form since the surface temperature
is below about 2000 OK and the carbon surface is quite catalytic for the
atomic recombination. Also the constant value of I as given by Eq. (128)
should be satisfactory for the carbon combustion problem (see Ref. I).
In closing, the graphite combustion through the transition regions is
encountered at the stagnation region of a blunt space vehicle during a
lifting re-entry wherein the heat transfer rate is substantially lower than
that during a ballistic re-entry, but the peak heat transfer period is
much elongated.
3. Locally Similar Solution for Catalytic Reaction around a

Hypersonic Blunt Body
I t was explained in the early part of the present Section E that a
locally similar analysis would produce a satisfactory solution of the
species conservation equation along a body if the Damkohler number
varied very slowly. We shall now show that the catalytic recombination
of dissociated diatomic gases around hypersonic blunt bodies comprising
a spherical nose and a conical after-body (see Fig. Ic) can be analyzed
by the locally similar method. T h e following analysis is essentially that
given by Chung and Anderson (31).
For constant I, the governing equations are given by Eqs. (1 30)-( 132),
for the highly cooled boundary layers around the blunt bodies. Equation
(132) is applicable here because the locally similar solution of species
conservation equation is anticipated. Limiting ourselves to the cases
with no mass transfer, the momentum and the energy equations have been
[I571
PAULM. CHUNG
solved by Lees (23) with appropriate boundary conditions. We shall

presently then consider the locally similar solution of Eq. (132).
T h e boundary conditions are
m(a) = 1
and
mw = 5m,
where n = 1 for the catalytic recombination of atoms. T h e Damkohler

number given by Eq. (140) is rewritten here for n == 1 as
In the preceding analyses of stagnation region, we have assumed that

pw is known a priori. Strictly speaking, however, pw is a function of the
unknown m, as well as the pressure and surface temperature in a com-
pressible flow [see the equation of state ( 5 ) ] . Usually, however, as seen
from Eq. (9,pw is a strong function of p and T , , and the accurate
account of its dependency on C , is not important. Moreover, k, which
appears with p,, is usually not known accurately to warrant such
detailed consideration of pw . Nevertheless, the dependency of pw on
C, should be taken into account in the analysis of a surface reaction
wlien it can be done without excessive added effort. With this in mind,
the Damkohler number is redefined with the aid of Eq. (9,which for
diatomic gases is p = (1 + , that it will be free of un-
C ) ( R / f i 2 ) T so
knowns as
(179)O
and the boundary condition (177) becomes
where &, = C(1 +

Gem,). I, denotes the constant position of &, , and
we shall call it a reference Damkohler number.
Now it is seen that 56 would vary very slowly along a body when
E(P/Po)/~(x)l does.
lo In deriving this equation it is assumed that pape/(p&, c11 pJp, .
r 1581
CHEMICALLY
BOUNDARYLAYERS
It can readily be shown by solving the frozen energy equation (131)

or the total energy equation of the similar form with equilibrium surface
condition, that the frozen or the equilibrium heat transferll varies along
the surface as
(2%)
qwo frozen or equ11.
-F(X)
Therefore, we may say that if the available pressure and equilibrium or

frozen heat transfer distributions for a particular blunt body are found
to be similar, then the local similarity concept should be applicable in
the analysis of the surface reaction. It can be seen from the equilibrium
heat transfer calculated in Ref. (23) that the hypersonic blunt body,
with spherical nose and conical after-body, satisfies the above criteria
for the locally similar analysis.
Here, a locally similar solution means that Eq. (1 32) is solved locally
with the boundary condition (180) applied as though it were not a
function of x,
however, with &, being calculated at the point in
question. A self-similar solution can in this way be obtained locally for
each position along the body.
Integration of Eq. (132) locally with the boundary conditions (176)
and (180) gives
T h e heat transfer due to surface recombination of atoms qd , is ob-

tained by
T h e heat transfer around the spherical nose with conical after-body

shown in _Fig._ l c _ calculated by Eq. (181) and (182) for
was ~
rb Sc-I 417.03 (u,/LP) = 5 and C, = 1. T h e gas phase reaction was
assumed to be frozen. T h e result is plotted in Fig. 8 as a fraction of
the heat transfer to a perfectly catalytic wall. T h e following inviscid
l1 See the equilibrium heat transfer expression given in Ref. (23).

P591
PAULM. CHUNG
I
0
r
+---0
11VM 3 I l A l V l V 3 0 1 d33SNVkIl lV3H 4 0 NOll3VkI3
CHEMICALLY
REACTING
BOUNDARY
flow properties employed in Ref. (23) are used in the present computa-
tion:
_ -- cos2x
Po
ue = px
2 = cosx,
I for
for
0 < X <Xe
X >, Xe
Po
u, = pxe
and
T h e result from a more accurate integral method, to be discussed later,

is also shown in Fig. 8. T h e agreement between the results of the two
methods are seen to be excellent.
F. SURFACE AND HEATTRANSFER

REACTION FROM
HIGHTEMPERATUREDISSOCIATED
AIR AND CHEMICAL
DIAGNOSTICS
We have thus far developed the general heat transfer relations in
Section 11, E , and obtained in the preceding Section E, the heat transfer
formulas for the cases wherein the self-similar solutions exist. This
may therefore be a good place to consider a little more quantitatively
the effect of surface chemical reaction on heat transfer from high tem-
perature air because of its importance in current aerospace applications.
Let us specifically consider the catalytic surface recombination of
atoms supplied from the inviscid region. We have seen in the preceding
Section E that for Le near unity, the maximum contribution of sur-
face recombination to heat transfer, (qn/q,.)max, is of the order of
C,d h / ( h , ,- htjw)and occurs as 5 3 00. Actually this is true whether
the chemical reaction is taking place on the surface or in the gas phase,
and this basic fact is not limited to the stagnation region12 (see Section
11, E ) . Let us consider that the air at the boundary layer edge is dissoci-
ated to near its equilibrium value. Such dissociation in practice may
occur due to the strong shock preceding a blunt body in a hypersonic
flight, or by an electric arc in a stagnation chamber of an arc tunnel,
or in many other ways. T h e value of C,dh/(h,,,- htfw)increases with
the equilibrium temperature as the degree of dissociation, C , , increases.
l2 This is assuming that there is no unusual variation of surface condition along x

such as a discontinuous variation of surface catalycity.
[I611
PAULM. CHUNC
T h e dissociation is complete (C, = 1) at the temperatures of 7000-

10,OOO OK when the pressure is about to 10 atm, and there the
maximum value of C,dh"/(h,,,- h,,,) is about 2 to 3. Because of this
great effect of atom recombination on high temperature heat transfer,
the surface recombination becomes very important in calculating heat
transfer to hypersonic blunt bodies and interpreting the laboratory
heat transfer results when gas phase reaction is nearly frozen.
T h e Damkohler numbers for the hypersonic blunt body considered
in Section E, 3 are computed in Ref. (31) for typical flight conditions.
These values together with the typical m, distributions obtained by the
integral method (31) are reproduced here in Figs. 9a and 9b. Since the
effect of surface reaction on heat transfer is zero when m, = 1 and
maximum when m, = (m,), = 0, Figs. 9a and 9b give this effect as
-------2---
b5
0.4 -
O ' I
0.2
U
1 2 3 4
X
FIG. 9a. Variation of m, along spherical nose and conical after-body.

CHEMICALLY BOUNDARY
LAYERS
100. 1
10.
L'u
I1
' 0
v)
hi
0.
0.0
15 00 200, OOO 250, OOO
ALTITUDE, f t
FIG.9b. Variation of Damkohler number with flight conditions.
the function of typical flight conditions and surface materials. One

may, of course, calculate q d from m, with the aid of Eq. (182). T h e
flight regime wherein the gas phase recombination is frozen is found
by studying the gas phase reaction which is to be considered later.
We shall only mention here that it was concluded in Ref. (31) that for
the blunt bodies of nose radius of the order of 1 ft, the gas phase recom-
bination will be frozen at altitudes above about 200,000 ft near the
stagnation region. T h e gas phase freezing will occur at much lower
altitudes along the conical after-body depending on the nose angle.
~631
PAULM. CHUNC
One of the most important current applications of the theories of

surface recombination arises from the fact that the heat transfer measure-
ments on the surfaces of known catalycities, k,, can be interpreted with
the theories to determine the free stream atom concentration in high
temperature laboratory facilities. Such determination is a part of the
facility diagnostic work, and the detailed discussion of the techniques
and some of the actual measurements are given by Rosner (5, 32),
Hartunian (33),and Winkler and Griffin (34). Most accurate diagnostics
are possible when the gas phase recombination is frozen. For this case
the basic principles of diagnostics are as follows. Consider that a pair
of noncatalytic and highly catalytic surfaces (double-surface gage) is
located in an experimental facility in such a way that the flow configura-
tions over the surfaces are identical. The measured heat transfer to the
noncatalytic surfaces gives q, , whereas that to the highly catalytic sur-
face gives qw = qc f qd . T h e difference of the two gives qd from which
C,may be obtained by the use of the present theories.
As was mentioned previously, the gas phase reaction first of all must
be frozen for accurate diagnostics. Then the accuracy of the diagnostic
depends, among other things, on finding the proper combinations of k ,
and flow conditions such that the ratio qd/qw becomes sufficiently large.
I t can be seen from the gas phase air chemistry discussed in Section I11
that the gas phase recombination rate decreases with pressure. It is
therefore often advantageous to locate the double-surface gages on
positions other than the stagnation regions of a blunt body, such as on
a flat plate, in order to freeze the gas phase reaction. Also, as we shall
see later, the sensitivity of the gages represented by qd/qw can be in-
creased substantially by locating the catalytic surface of the gages down-
stream of a noncatalytic surface.
T h e surface reactions along bodies other than the ones covered in the
present section, and the effects of variable k, on qd will constitute the
remainder of this chapter.
G. NONSIMILAR
SOLUTIONS
OF SPECIESCONSERVATION
EQUATION
I n Section 11, we developed the general relationship for surface heat
transfer. We have then shown in Section IV, C that the momentum and
the frozen energy equation needed for qc , can be decoupled from the
species conservation equation when the gas phase reaction is frozen.
Furthermore, the momentum and frozen energy equations decoupled
thus are the same as those for chemically inert boundary layers and the
solutions of these equations are available elsewhere. T h e only analytical
problem of our concern has been then to solve the species conservation
[I 641
equation for the surface distribution of reactants from which the reaction
+
rates and qa may be found. The final formulation of qw = qc qa has
been illustrated in Section IV, E, 1 in connection with the analysis of
the stagnation region.
We shall therefore in the present Section G be concerned only with
solutions of the species conservation equations. As was mentioned in
Section C, we shall limit our analysis to the flows for which the momen-
tum equation has a self-similar solution -flat plate, supersonic wedges
and cones, hypersonic blunt bodies, and Falkner-Skan flows. We shall
analyze mainly the first order reaction for it is the most prevalent one.
I t will also be assumed that p , is known a priori as the function of p
and T,: , and the surface mass transfer is negligible.
It will be also assumed that Sc and 1 are constant across the boundary
layer.
1. Exact Series Solution

The general governing Eq. (126) can be rewritten here for constant
Sc and 1 as
1
i&m4, +fm, = 25f,mc ( 184)
where m is the reactant mass fraction normalized to the boundary-layer

edge condition, and 5 is the nondimensionalized i defined by
T h e variables q and f have been defined by Eq. (129). The boundary

conditions become for fw = 0 and (m,), -+ 0:
Atq-0
m, = ( m
and at 7 = GO
m = l
T h e Damkohler number is given by Eq. (140) for the general body shape
and becomes for the first-order reaction
We shall now discuss the exact series solution of the boundary value
problem comprised of Eqs. ( 1 84),(186) and (187). T h e solutions about
~1651
PAULM. CHUNG
to be given here are based on the work of Chambre and Acrivos (35)
and Inger (36).
a. GeneralSolution in Terms of 7 and i. Let us assume that the
Damkohler number, <, is expressible as a continuous function of 5.
I n particular, let us assume that it is expandable into the following power
series.
where CY is an arbitrarily chosen positive number. We notice here that

it is immaterial whether the Damkohler number varies in the above
manner due to the body shape and flow or due to a continuous variation
of K, along the surface. Noticing the fact that Eq. (184) is linear we
construct the solution in the form
where ans are arbitrary constants. T h e substitution of Eq. (190) into

Eq. (184) and the separation of variables result in the following equation
for each n:
where yn is the eigenvalue. Equation (191) yields
a solution of
with Yn(7)
1
- Y; +fYn - 2ynJYn= 0 (193)
sc
We obtain the solution of Eq. (193) to satisfy the boundary conditions

Yn(0)= 1, Yn(.o)= 0 (194)
T h e solution (190) is then seen to satisfy the boundary condition (187).
Now, the wall-boundary condition (186) becomes, with the aid of
Eqs. (189)-( 194),
CHEMICALLY BOUNDARY
LAYERS
If we let yn = n / a , Eq. (195) gives the constants an's as
and the general solution of the boundary value problem is found as
I n principle, the above general solution applies within its region of

convergence to all body shapes with continuous distributions of k , when
the Damkohler number is expressible by the series (189). Let us now
adapt Eq. (197) to a few practical problems.
6 . Flat Plate, Supersonic Wedge, Supersonic Cone, and Falkner-Skan
Flows With Constant k , . These flows have in common the conditions
that (p,k,) and (pep.,) are, respectively, constant, and most of all, the
Damkohler number 5 can be expressed by a single term. Thus we may
let all b,,'s except, say, b, be equal to zero. T h e coefficients an's then
become
Qo = 0
a1 = b*/Yl'(O) (198)
...
an = b,"/II% Yi'(0)
When the Damkohler number is expressed in the form 5 = b l f l / d l

for each of the bodies concerned, Eqs. (197) and (198) give, after a little
manipulation,
where for a flat plate or a supersonic wedge

PAULM. CHUNC
for a supersonic cone
a = 6
and for Falkner-Skan flows of uJu, = BFXk,where 0 < k < 1,
(Y = 2(1 + k)/(l - k )
Equation (199) gives the distribution of reactant along the surfaces of
a flat plate, a supersonic wedge, a supersonic cone, and a subsonic wedge
as the functions of the Damkohler numbers given by Eqs. (200), (201),
and (202).
T h e values of Y,(O)s appearing in Eq. (199) are obtained by solving
Eq. (193) for the respective values of yn = n / a . T h e solutions of Eq.
(193) satisfying the boundary conditions (194) have been obtained
numerically by a number of investigators for different values of n and Sc
when OL = 2 .[See.for example, the work of Tifford and Chu (37) and
Chapman and Rubesin (38)].A few solutions for a corresponding to
the Falkner-Skan flow is also given in Ref. (37). For other values of a ,
the solutions of Eq. (193) are not readily available. This is one of the
reasons of the rather limited applicability of the exact method.
I t can be seen from the computed values of Y,(O) for 01 = 2 that
T h e series (199) therefore converges uniformly for all X , however,

actual calculations show that the convergence is rather slow.
Figure 10 shows the surface distribution of the reactant along a flat
plate or a supersonic wedge computed from Eq. (199).
c. General Body Shapes and Continuously Varying k, . It was seen
in the preceding sub-section that the general solution (197) reduces to a
rather simple form of Eq. (199), with the assured convergence, when the
Damkohler number can be expressed by a single term. For the body
[I681
CHEMICALLY
BOUNDARYLAYERS
1 I 1 I
=
0 1 2 3
z
1
(0.339) \rz- SC
FIG. 10. m, along flat plate and supersonic wedge.
shapes considered therein it is seen that the Damkohler number can still
varies as k, -
be represented by a single term if k, , though it may not be constant,
x k . T h e solution (199), therefore, gives the surface distri-
as x k . When k, -
bution of reactant along the bodies with constant k, or with k, varying
xk, of course, the Damkohler numbers given by
Eqs. (200)-(202) must be modified accordingly. Though it is seldom
that k, varies as x k in practice, the solutions obtained for such variations
of k, give a good insight into the general effect of the variable k, on the
surface reaction. Such solutions are discussed fully by Inger (36).
For other general body shapes and more realistic variations of k, ,
the Damkohler number cannot be expressed as a single term and it
remains as a series. Consequently, the full solution given by Eqs (196)
and (197) must be employed, and a considerable amount of numerical
work associated with the large number of the terms could become
inevitable in obtaining the values of m,. Furthermore, the general
convergence of the solution is no longer assured, and the region of
convergence must be investigated for individual cases. Nevertheless,
the series solution when it is obtained is still an exact solution within
~1691
PAULM. CHUNG
its region of convergence, and it can be quite useful in checking the

results of the approximate methods. For instance, it was shown in
Section E, 3, that the behavior of the species conservation equation
around a spherical nose in hypersonic flow is quite close to being self-
similar. The present series solution, therefore, should be able to yield
the surface mass fraction of reactant with a very few number of terms
for such flow conditions. The exact solutions were obtained in Ref. (36)
for the spherical nose up to the nose angle of 60" and they were compared
with the approximate solutions shown in Fig. 9a. The comparison
showed that the local similarity and the integral methods (31) give the
values of m,lm, within 3 yo of the exact values for this particular body
shape.
I n closing, one can in principle apply the present method to the
surface reactions other than first order also. Here again, however, the
numerical complexity usually becomes large, and most of all, the con-
vergence of the series solution becomes unassured. In fact, Chambre and
Acrivos (35) showed that for a second-oyder reaction over a flat plate
there exists an upper boundary on k, which limits the series solution.
We shall now consider several approximate methods of solution.
2. General Transformation of Boundary Layer Equations into

Incompressible Forlii
The subsequent approximate analyses begin with the transformation
of the general governing equations, Eqs. (l), (2), and (16) and the frozen
total energy equation, in the following manner. Though the present
interest does not directly involve the energy equation, we shall include
it in the transformation for completeness.
We first define a set of independent variables for constant I as
and
where 5 has been defined in Eq. (185).

Next, a set of dependent variables is defined in terms of the stream
function # as
T h e governing equations then become:

Continuity
+ - = o av
-au
as at
Momentum
Frozen total energy
Conservation of reactant
T h e last term of momentum equation (207), is zero for flat plate,

supersonic wedges, and supersonic cones. Also, as explained previously,
it is negligible for highly cooled blunt bodies wherein pJp, < 1.
Hence, it is seen that Eqs. (206)-(209) are in incompressible form for the
flows of the present interest because the density does not appear ex-
plicitly.
Boundary conditions for the transformed equations are as follows:
att=O
us0
PAULM. CHUNG
and at t = 00
U = l
ht, = htfe
c = c,
I n the following, we will be concerned with the approximate solutions
of Eq. (209) with the aid of Eqs. (206) and (207).
3. Approximate Solutions Based on Representation of Surface Reaction

by a Volterra Integral Equation
Fage and Falkner (39) made the now well-known suggestion that
in the solution of a forced convection problem one can replace the
velocity components of the flow in the appropriate field equation by
their asymptotic form close to the surface. Thus the streamwise velocity
profile, according to this suggestion, is replaced by u = (aujay), * y in
the analysis of an energy or a species conservation equation. Lighthill
(40) using this approximate velocity profile integrated the energy equa-
tion by Laplace transformation resulting in the well known Lighthills
heat transfer equation. Later, Rubesin and Inouye (41) developed the
same equation by separating the variables.
Application of the Lighthills integral relation to the problem of
surface reaction was investigated by Chambre and Acrivos (35),Rosner
(42), Chambre (43), and Acrivos and Chambre (44). T h e investigation
led to the formulation of a Volterra integral equation. The actual solution
of the integral equation, however, was obtained only for flat plate in the
first two references. Subsequently, Chung and associates (45) refor-
mulated the problem for more general compressible boundary layers
and obtained solutions for most of the body shapes of the present in-
terest. I n particular, the formulation (45) included both the discon-
tinuous and continuous variations of k, . T h e solutions for the variable
k, are given in Refs. (45, 46).
T h e following discussions are based on the above-mentioned works.
T h e only basic approximation involved in the following analysis is
that due to the simplified velocity profile. As we shall see later this
approximation produces almost negligible error in the surface distribu-
tion of the reactant.
a . Series Solutions for Constant or Continuously Varying k,, . It is

seen that the transformed species conservation equation (209) on the
s-t plane is identical in form to the energy equation analyzed by Lighthill
(do), and Rubesin and Inouye (41) for flat plate. Following the method
11721
REACTINGNONEQUILIBRIUM BOUNDARY
CHEMICALLY LAYERS
employed by Rubesin and Inouye (41) we derive the following general

equation from Eq. (209) for v,. = 0:
(t)am = SC~(I - mwo)

au
(a-)
ti' t 21'
We shall now illustrate the method of solution of surface reaction

problems based on the above equation by analyzing the surface reactions
along the flat plate, supersonic wedges, and the supersonic cones. For
these cases mu,o = 1 because the diffusion boundary layer thickness is
zero and therefore the Damkohler number is zero for all finite k, at
the leading edge. Also the last term of the transformed momentum
Eq. (207) vanishes. T h e present momentum equation on the s-t plane
is then identical in form to the momentum equation for incompressible
boundary layer over a flat plate. Hence we have
(GI B
= 0.332--
&OS
1
Now applying the surface boundary condition (213), for the first-order
reaction and v, = 0, to Eq. (217), we derive with the'aid of Eq. (218)
1 dm,(z)/dz
mu = - (0.339) dz S C__
~
where the value 0.339 is used in place of (0.332)'/12"(+). Equation

(219) is a nonlinear Volterra integral equation, and we desire to obtain
the solution of this equation.
We consider that, analogous to the previous exact analysis, the
Damkohler number is expressible by the series
'I)
((5) = z b n P / a (220)13
n=l
We then seek the solution m, also in a series form
+Z a n P a
m
mw(0 = 1 (22 1 )13

n=l
l3 Here the summation begins with n = I in Eq. (220), instead of n = 0, and m,(O)
is set equal to unity in Eq. (221) because we have limited the analysis to the cases wherein
((0) = 0.
[I731
PAULM. CHUNC
where cy is a positive quantity to be defined later. After substituting

Eqs. (220) and (221) into Eq. (219), and after a little manipulation we
transform Eq. (219) into
fi4 Sd 4n 2
ID
= - (0.339) a, ( E ) B (jr;,3) $ l a
n=l
where B denotes the Beta function which is related to the Gamma

function by
Now for a given set of bns, the coefficients ans can readily be obtained
from Eq. (222) by collecting the terms with like powers of 5. m, is then
obtained from Eq. (221).
T h e advantage of the present method compared to the exact series
method is that no numerical integration is involved, and all ans are
obtainable in terms of beta functions.
T h e surface distribution of the reactant can, in principle, be obtained
for all body shapes and continuous variations of K, by the use of Eq.
(217) and in a manner similar to the above illustration. However, it is
necessary to employ the appropriate expression for (aujat), instead of
Eq. (218) when the pressure gradient term is not negligible, and the
appropriate stagnation value of m,, when the body has a blunt nose.
I n practice, however, the convergence criteria of the resulting series,
such as Eq. (221), and the algebraic effort required when ((5) requires
a large number of terms, may limit the usefulness of the present method
as was the case with the preceding exact method.
Now returning to the flat plates, and supersonic wedges and cones,
simple analytical solutions can be obtained when K, varies in such a way
that the Damkohler number is expressible by a single term. We shall
consider these cases in the following.
P741
CHEMICALLY
REACTING
BOUNDARY
b. Solutions When 5 -
f n l a f o r Plates, Supersonic Wedges and Cones.
We may choose any one b, to be nonvanishing without loss of generality
because cy may be any positive quantity. We let
5 = bl(ll* and bn = 0 for n # 1
T h e following solution is then readily obtained from Eqs. (221) and

(222):
[
m, = 1
21
+ C(-l)
n=l
rr (x
i-1
I^
4 (0.339)b1d2 Sci
4i 2
9 3)
[nla
where
1
5 - 5
- (0.339) 42 Sch
cones when k, varies in such a manner than 5 -

T h e above solution is valid for the plates, and supersonic wedges and
t1Ia.T h e surface
distributions of reactant can be computed from the solution (223) by
the use of 5 and a given by Eqs. (200) and (201) when K , is constant.
T h e values of m, computed from Eq. (223) for flat plates and supersonic
wedges are plotted in Fig. 10 along with the results of the exact series
method. It is seen that for the region wherein the exact results are com-
puted, the present results agree with the exact results within one to two
percent. A similar close agreement was found for the cases of K ,
(m is any positive number) when the exact series results of Inger (36)
Xm -
were compared with the approximate results of Liu and Chung (46)
based on the present method.
Equation (223) and Fig. 10 show the variation of m, , and therefore
the variation of surface reaction rate as a function of the local Damkohler
number. I t is seen that the surface concentration of reactant relaxes
toward the equilibrium values (zero) as 5, is continuously increased.
For given flow and surface conditions, the local Damkohler number
increases and the relaxation occurs along the surface because the bound-
ary layer thickens and thus the diffusion time continuously increases
with x. For very small values of C,, the surface concentration of the
reactant is that corresponding to the boundary layer edge. T h e reaction
rate, therefore, is predominantly determined by the surface kinetics
El751
PAULM. CHUNG
represented by k, . This regime is called chemically or rate controlled

regime. I n the other extreme of 5,. + co the surface concentration of
the reactant approaches the equilibrium value as the surface reaction
consumes the reactants as fast as they are diffused to the surface. This
regime is called diffusion controlled regime.
c. Numerical Solutions for Discontinuously Varying k,,, . Thus far,
we have considered the cases wherein k , is either uniform or varies
continuously along the surface. In many practical cases, however, the
surface is made of different materials located adjacently and the specific
rate constant varies discontinuously.
T h e general equation (217) is applicable to the cases of discon-
tinuously varying k , also. Solutions of this equation for discontinuous
variations of k, have been obtained by Chung and associates (45) by
a series method, and a numerical method. The simple numerical method
employed by these authors seems most general and useful. We shall
herein describe this method and show a few of the solutions.
For purposes of illustrating the method, we consider the flows wherein
the last term of the transformed momentum Eq. (207) is negligible.
Furthermore, we consider the specific case wherein k, varies discon-
tinuously from zero to an arbitrary finite value at s1 (corresponding to XJ.
m, then is unity for 0 s < <
sl. Equation (217) is now transformed
for s 3 s1 with the aid of the wall boundary condition as
1
Cd(4 =
(0.339) .\/2 Sch
We now seek the solution of Eq. (225).

T h e method of solution essentially consists of subdividing the interval
of interest into small sections and approximating the function m,,(z)
by a function such that Eq. (225) can be integrated analytically for each
section. T h e segmentized solutions are then summed numerically,
+
step-by-step, beginning from = 1, so that a continuous solution of
m,(+) can be obtained. This method of solution has the merit in that it
is very general in nature, and the numerical work involved is not usually
excessive and can be performed by the use of a desk calculator.
r1761
CHEMICALLY
REACTING
BOUNDARY
T h e value of m, in the nth interval is found as follows. Let the sub-

division of the interval be represented by C $ o , d l , ..., 4kwhere +o = 1
(see Fig. 11). Then, we let the solution m , ( z ) for the kth segment
FIG. 1 I. Numerical solution of Volterra integral equation.
(#k--l \< x < C$,J be expressed by

mWsk= ak - b,zf
T h e particular exponent of z in the above equation is chosen so that Eq.

(225) will be integrable for each segment. Now for the nth segment,
Eq. (225) becomes, with the aid of Eq. (228),
Equation (229) has two unknowns: a, and b , . Another equation is

obtained from the fact that m, is continuous. For the nth segment, there-
[I771
PAULM. CHUNC
fore, m, at must be equal to that at obtained in the preceding

segment. Therefore we have from Eq. (228)
When Q, is eliminated from Eqs. (229) and (230), there results the
equation
+n+mw,n-l+
3
--
1
2bJ(+n: -
t d ( + n ) k51
+k')' - (+n3 - +!-I)*]
b, = (23 1)
- 4i-J + 2
1
+:(+nz (4Z,
__ - 4f-#
2 M+n)
Equation (231) can be used to calculate b , step-by-step starting from

the known value of mpu,l= 1. Once b, is obtained, a, is calculated from
Eq. (230) since mw,n-l is already known. m,,, is then obtained from
Eq. (228) as
m w , n = an - bn+n' (232)
T h e method discussed above was first tested by solving the problem

of surface reaction along a flat plate with an invariant k,, the exact
result of which is given in Fig. 10. A comparison of the numerical
solution with the analytical solution of Fig. 10 showed that the
present numerical method gives results which are within 3% of the
results of Fig. 10 when (bk - &.l) N 0.1. Higher accuracies, when
desired, may be obtained by using smaller intervals.
T h e method described above is very general and can be adapted
satisfactorily for most practical variations of Damkohler numbers
whether discontinuous or continuous and for most flow conditions of
practical interest. It can also be employed for surface reactions which
are of other than first order.
Figure 12 shows the flat plate results for several combinations of k,,'s
for 0 < <
x x1 and x 3 x1 . Figure 13 shows the distribution of atoms
along the surface following a noncatalytic surface on a sphere. T h e
external hypersonic flow properties given by Eq. (183) were used for
Fig. 13. We notice from Figs. 12 and 13 that the relaxation of surface
mass fraction of reactant following a discontinuous variation of k, ,
to those which would exist if the k, were uniform throughout, is quite
slow. Thus, when k, is varied from zero to a finite value for instance,
m, will stay above the corresponding value of m, for a uniform and
equal k, along a considerable length of the surface. Therefore, the sur-
face reaction rate and qd for a given reacting surface can be substantially
~781
CHEMICALLY
BOUNDARY
LAYERS
0.1 - -
1 m
I I
Fie. 12. Discontinuity of catalycity on flat plate and supersonic wedge.
increased by locating the surface on downstream of a nonreacting

surface.
I t was explained in Section F that the sensitivity of a double-surface
gage used for the high temperature facility diagnostics depends on the
heat transfer ratio q d / q m .I n view of the discussions given herein, it is
seen that this ratio, and therefore the sensitivity, can be substantially
enhanced by preceding the catalytic surface of the gage by a noncatalytic
surface. It was shown in Ref. (45) that the sensitivity of the gages can
readily be increased by such arrangements.
4. Approximate Solutions Based on Integral Methods

a. Pohlhausen Integral Method. T h e basic features of the Pohl-
hausen integral method are well known. We shall herein briefly describe
the method as it was applied to the problems of surface chemical reactions
by Chung and Anderson (31, 47, 48).
[I791
PAULM. CHUNG
0.9 - (P,O k, )sc23 -
- -
0.7
-- c c c - - c -
0.6 - X,/L = 0.6
E
3 - RESULT OF
NUMERICAL
0.5 METHOD EQS (230)
AND (231)
RESULT OF INTEGRAL
METHOD FOR
0.4 - UNIFORM k,
-
0.3 - -
-
0.2
___----
0. I I I I I I
0 0.2 0.4 0.6 0.8 1 .o 1 .z
X
T:
FIG. 13. Discontinuity of catalycity on sphere.
T h e analysis is simplified by assuming a priori that the diffusion

boundary layer has the same thickness as the momentum boundary
layer. This assumption is believed to be valid for Sc in the order one.
Again limiting the analysis to the cases wherein the pressure gradient
terms of the transformed momentum equation (207) is negligible, Eqs.
(207) and (209) are integrated across the boundary layer with the
[ 1801
aid of Eq. (206). T h e resulting integro-differential equations for

v, = 0 are
where
and the prime denotes differentiation with respect to s.

T h e boundary conditions become:
at Y = O
and at Y = 1
U=l
m = l
where
k,,,
T h e reason that the right-hand side of Eq. (238) is not a linear function
of mw for the first-order reaction was given in Section E, 3. T h e reference
[I811
PAULM. CHUNG
Damkohler number, rediffers only by a constant from r, defined in
Section E, 3. C is the mass fraction of atoms and m = CjC, .
We choose fourth and fifth degree polynomials in Y to represent the
profiles for U and m, respectively. By satisfying the usual boundary
conditions, including the boundary conditions (237)-(241), all the coeffi-
cients of the profiles are determined except one for m profile. One
coefficient in the nz profile is left to be determined as the solution of
Eq. (234).
T h e momentum equation is readily integrated by the use of the U
profile and there results
h = A'S = (34.05)s (242)
T h e species conservation equation (234) becomes, with the aid of the
profiles and Eq. (242),
1 - m , - (0.258)re
where ped(pe/+Jo is replaced by PIP,

T h e term (1 + Cemw) entered into Eqs. (238) and (243) as pw was
expressed in terms of p , T , , and C, through the equation of state. If
one considers that pW is known a priori, Eq. (243) becomes linear and
for flat plates and supersonic wedges it can be integrated in a closed
form. T h e solution is
m, =
1 - (1 +
0.4585,)-"3
(244)
1.0535,
T h e above closed form solution is seen in Fig. 10 to give very satisfactory
results.
For more general flow conditions, Eq. (243) must be integrated
numerically.
T h e equation is integrated in Ref. (31)by the use of a digital computer
for the higly cooled blunt bodies comprised of a spherical nose and a
conical after-body shown in Fig. lc. T h e inviscid flow properties used
around the body are those given by Eq. (183). T h e typical results are
given in Fig. 9a and they have already been discussed in Section F. We
only note here that the surface reaction practically does not freeze around
the nose in contrast to the gas phase reaction which, as will be seen later,
freezes quite drastically.
P821
CHEMICALLY BOUNDARY
LAYERS
b. Concentration-Integral Method for Various Orders of Reaction. T h e

analyses presented hitherto were mainly concerned with first-order
surface reactions. In principle, many of the approximate methods can
also be applied to the reactions other than of the first order. These non-
linear order reactions, however, usually require a considerable amount
of added numerical work. Rosner (49), however, presented an approxi-
mate integral method which gave closed form solutions for all orders
of reaction for flat plates and supersonic wedges. We shall herein
describe the method.
We first derive the following integro-differential equation by inte-
grating the species conservation equation (209) across the boundary
layer from t = 0 to t = CO:
where
-
9=
C-Cw
ce - c
w
, F, =I
m
0
U(l - 8 ) d t
T h e brief outline of the method is as follows. First, we assume a local

similarity. This then gives us the surface gradient (&/at), since the self-
similar solution of the species conservation equation (209) is available
elsewhere for many flow conditions. With this value of (at?jZt), sub-
stituted into Eq. (245), the equation is solved for the concentration
integral F3 . A new improved (adjat), is then obtained from F , . T h e
application of the boundary condition (213) to the expression of (a6jat),
thus obtained produces an integral equation which is solved for C,.
T h e essence of the method is therefore to employ the integro-differential
Eq. (245) and produce the first order improvement of (adiat), over the
local-similarity value.
Limiting ourselves again to the flows wherein the pressure gradient
term of Eq. (207) is negligible, the existing local similarity solution of
species conservation Eq. (209) [actually the full solution of Eq. (209)
for uniform C,] gives
d8
(%) = 0.47Sct (247)14
W
See Section E.
~831
PAULM. CHUNG
We now correlate the above local-similarity value of ( a d ( a t ) , to the

corresponding F , through Eq. (245). With the local similarity assump-
tion, d(C, - C,,)/ds is zero. Hence, Eq. (245) gives
Thus far we have dealt only with the locally similar relations. Now
we improve (adjat), from its locally similar value in the following manner.
We substitute Eq. (249) into Eq. (245) and solve the full equation,
without neglecting the second term, for F, . We then derive the improved
( & / a t ) , from Eq. (249) by the use of the F , thus obtained. T h e gradient
( d ; a t ) , is finally eliminated between Eqs. (249) and (213) and the
following integral equation is derived:
T h e above equation is a nonlinear Volterra integral equation. T h e solu-

tion of Eq. (250) for an arbitrary body shape is no less complicated than
that of the more exact Volterra equation studied in Section G, 3. How-
ever, for flat plates and supersonic wedges with uniform k, , the solu-
tion of Eq. (250) can be obtained rather simply by transforming the
equation into a differential equation as
-(2TL+l)
-5,dC, = mw ( 1 - mw)[2n - 2 ( n - 2)m,] dmw (251)
Integration of above equation gives, for flat plates and supersonic
wedges:
when n = t
5f2 = -
- 1-
- mw
+ I n -17- 3(1 - mw)
2 m W m W
when n = 1
and for all n # 0, 8, 1

where
=05c1ll(pli.)lu.l/x
1 Sc(p,"k,)C:-'
<f
2L
T h e Damkohler number, cf is related to those defined previously by
Equation (253) gives the surface distribution of reactant for n = 1

which is within about 2 yL of the exact values given in Fig. 10. The general
results are shown in Fig. 14. We notice one fact that the higher the
reaction order, the greater is the Damkohler number it takes for the
surface reaction to reach equilibrium state. This effect of reaction order
on the surface reaction is basically true for all body shapes.
5. Surface Reaction behind Strong Moving Shock and the Effect of

Vurying C,
Before closing the present chapter on surface chemical reactions,
we shall consider a slightly different problem compared to those we have
OL 1 I I I I
0 2 4 6 8 10 I2
[f
FIG.14. Effect of reaction order n on m, for flat plate and supersonic wedge (49).
PAULM. CHUNG
presented hitherto. It is the problem of surface chemical reaction behind

a strong shock moving along a solid boundary into a stationary dia-
tomic gas. This phenomenon is of interest in connection with shock tube
studies. T o measure, for instance, the dissociation level behind the
moving shock one may use a double-surface gage15 located on the shock
tube wall or on a thin plate inserted parallel to the flow. T h e theory
about to be presented gives the concentration of dissociated atoms along
the surface of uniform k, behind the moving shock.
In all the preceding analyses of the present chapter, we have considered
that the concentration of reactant is uniform along the boundary layer
edge. It is because the gas phase reaction along the boundary layer edge
is usually frozen when it is frozen within the boundary layer. This, how-
ever, is not always the case. Consider, for instance, the flow of a high
temperature dissociating air along a cold wall at a low, local Mach
number. Due to the cooling effect of the cold wall, the predominant
reaction within the boundary layer is recombination rather than dis-
sociation. It is likely, for such cases, that the dissociation along the
boundary layer edge will vary Ce but the gas phase recombination
within the boundary layer is negligible.16 It is nevertheless impossible
to analyze such a case by the use of a frozen species conservation
equation, because mathematically the species conservation equation
with frozen gas phase reaction cannot in general satisfy the boundary
condition of arbitrarily varying C, . This is obvious when one looks
at the species conservation equation (16). It is seen that at y = 00
where acejay = a2Ce/ay2= 0, peu, aCe/ax = W e . Therefore, aC,/ax
must vanish when W is neglected in the boundary layer equation. This is
consistent with the physical reasoning that C, must remain constant along
the boundary layer edge unless a gas phase chemical reaction takes place.
T h e present surface reaction problem behind a strong moving shock
is one special case which can be analyzed with the frozen species conser-
vation equation including the effect of arbitrarily varying Ce . T h e reason
for this exception is that for strong shocks which produce large amounts of
dissociation, the particle velocity behind the shock may for all practical
purposes be considered as equal to the shock velocity. Thus, for an
observer standing on the shock, u, = 0 and u, = -us where us is
the shock velocity. Hence, W for the boundary layer may be set at zero
for all finite aC,/ax. We have therefore an additional reason for dis-
cussing the present problem in that it is one case wherein we may see
the effect of varying C, on the surface reaction.
l5 See Section IV, F for chemical diagnostic principles.

l6 Inger (50), for instance, has investigated the existence of such a flow regime.
11861
CHEMICALLY
REACTING
BOUNDARY
In the following we shall outline a portion of the analysis given by

Chung (51).
It is considered that the gas dissociates in the inviscid flow behind
the shock of uniform velocity us and diffuses to the surface to recombine
catalytically. In the analysis the inviscid flow dissociation and the wall
recombination are considered to take place at arbitrarily finite rates.
On a laboratory-oriented system, the shock tube flow is a transient
one. We first fix an arbitrary origin x = 0 (see Fig. 15). The time that
e
' Y,"
////
ARB1TRARY ORlG IN
FIG. 15. Shock tube wall boundary layers.
the shock passes over the origin is defined as T = 0. T h e governing

equations for x < U ~ Tare then obtained by adding the terms a p l a 7 ,
p &/&, and aC,/& to the steady state continuity equation ( I ) , momen-
tum equation (2), and species conservation equation (16), respectively.
T h e governing equations are first transformed in the following manner.
T h e continuity equation is satisfied by defining the stream function as
We then define a set of dimensionless independent variables as

PAULM. CHUNG
It is seen that the present Damkohler number, [, is the same as 5 which

has been used hitherto, except the distance measured from the shock,
x, appears in the present case in place of the actual distance
U , ~ T-
along the wall which appeared in 5. A dimensionless stream function
is also defined as
T h e momentum and the species conservation equations now become,

respectively,
where the prime (') denotes the total differentiation with respect to ;i.
T h e boundary conditions are as follows:
atf.=O
f=f'=O (262)
ac
- = [c
a7
f' = 1 (264)
c = C,([) = Zb"%"
I
n=O
where it is considered that the distribution of atoms along the boundary

layer edge is expressible by the series given in Eq. (265).
Observing that Eq. (261) is linear, we construct the general solution as
2 [at,yn(.;i)+
oc
c(.i,5) =
n=O
bnwn(.i)~~n(O (266)
T h e substitution of the above equation into Eq. (261) separates the

variables, resulting with the following equations:
z,= 5" (267)
1
- Y,"
sc
+ (7 -f)Y,'- n(1 - f ' ) Y n =0 (268)
1
- W,"
sc
+ (7 - f)W,' - n( 1 - f ' ) Y n = 0 (269)
CHEMICALLY
NONEQUILIBRIUM
where the eigen values have been set equal to n and the limiting approx-
imation of the strong shock, u,/u, = 1, has been made.
We now obtain the solutions of Eqs. (268) and (269) to satisfy the
following boundary conditions, respectively:
Yn(0)= 1, YJ.0) =0
and
W,(O) = 0, W,(m) = 1
It is now seen that the solution (266) satisfies the boundary conditions
(262), (264), and (265). T h e application of the remaining boundary
condition (263) to the solution (266) produces the equation
Thus, the coefficients a, are determined in terms of Y,(O)and W,(O).

T h e values of Y,(O) and W,(O) are obtained by Chung (51) by inte-
grating Eqs. (268) and (269) numerically and are tabulated here in
Table I11 for Sc = 0.52.
T h e surface distribution of atoms now becomes
T o study the general effect of varying C, on the surface distribution

of atoms, let us consider the following arbitrary variation of C,:
C,(Q = 0.9998%- 0.6665p + 0.1481p (274)

Figure 16 shows the corresponding surface atom distribution calculated
from Eq. (273). T h e broken line in the figure shows the results obtained
by assuming dC,/d[ = 0 but using the local values of C, at each point.
I t is seen by comparing the broken line with the result of the correct
analysis that the effect of the variable C, on the surface atom distribution
cannot be neglected when C, varies rather greatly as in the present
example.
In order that the present theory be applied for a shock tube diagnostic
work the frozen energy equation must also be solved with the variable
h,,, = h,, - AhC,. As it is seen from the analogy between the species
I t was shown in Ref. (51) that the approximation of uJus = I is satisfactory for
the usual strong shocks for which u,/u, 21 0.9.
PAULM. CHUNG
TABLE I11
SOLUTIONSOF EQS. (268) AND (269) FOR UJU, = I AND Sc = 0.52 (51)
n Y"'(0) Wn'(0)
0 (I (1
1 -0.70037 0.28411
2 -0.94366 0.20924
3 - 1.1506 0.15831
4 -1.3325 0.12229
5 - 1.4960 0.09606
6 - 1.6456 0.07650
7 - 1.7840 0.06163
8 -1.9134 0.050 15
9 -2.0353 0.04118
10 -2.1508 0.03406
11 -2.2608 0.02837
12 -2.3661 0.02378
13 -2.4672 0.02004
14 -2.5645 0.01699
15 -2.6584 0.01448
16 -2.7494 0.01238
17 -2.8375 0.01065
18 -2.9232 0.00920
19 -3.0065 0.00795
20 -3.0877 0.00692
21 -3.1669
22 -3.2442
23 -3.3198
24 -3.3939
25 - 3.4664
26 -3.5374
27 -3.6072
28 -3.6757
29 -3.7429
30 -3.8090
I n Eq. (272), Wo'(0)and Y,,'(O)appear as W,,'(O)/Y,'(O) (= -I ) only. Th e individual

values of W,,'(O)and Y,,'(O)are therefore not needed.
conservation equation and the frozen total energy equation the latter
is amenable to an analysis similar to the one given herein. For details
the reader is referred to Ref. (51).
As to the shock tube conditions wherein the boundary layer gas
phase reaction would be frozen, Hartunian and Marrone (52) concluded
that the initial preshock pressures in the order of 0.1 mm H g or less will
ensure the frozen boundary layer when T , = 300 OK. Subsequently,
I1 901
0.5 1 .o
0.4 0.8
SAMPLE CeGIVEN BY EQUATION (274)
0.3 0.6
dC
s
U
C, BASED ON LOCAL ASSUMPTION OF
dc
e- = B
E
0.2 0.4
0.1 0.2
C ,CALCULATED BY EQUATION (273)
a I I 1 I I I I D
0.2 0.4 0.6 0.8 1 .o 1.2 1.4 5
. .
c
W FIG.16. C , along shock tube wall.
c
PAULM. CHUNG
Chung (51) suggested that the operating regime of the double-surface

gage could be substantially enhanced by preheating the gage surface
to 700 or 800 OK. It was also suggested that the value of 1 for the shock
tube flow be [ ( ~ w ~ w l ~ e ~ e ) o I ~ . ~ * -
I n closing our discussion on surface reactions, it is worthwhile
mentioning here that recently Freeman and Simpkins (95) applied the
Mellin transformation to Eq. ( I 6) and obtained a series solution for
flat plate which agreed closely with other solutions for flat plate discussed
herein.
V. Boundary Layers with Gas Phase Reactions
A. GOVERNING AND HEATTRANSFER

EQUATIONS
T h e governing equations were derived in Section 11. We first have
the continuity and the momentum equations (1) and (2), and the species
conservation equation (16). With them we may employ any one of the
energy equations (15), (22), and (27), and the frozen total energy equa-
tion. T h e surface heat transfer is obtained from the solution of the
equations by the use of Eqs. (45), (46), or (47) depending on the particular
form of the energy equation employed.
T h e expressions for the reaction rate term W iwere derived in Section
111.
I n the present section we shall consider the two classes of gas phase
reactions: the exothermic combustion of fuels; and the dissociation-
recombination processes occurring in high energy gases. Except in the
last subsection, we will consider that the surface reaction is either
frozen or it takes place at such a rate that the surface is in equilibrium.
B. TRANSFORMATION
OF GOVERNING
EQUATIONS
T h e governing equations will first be transformed to two useful
forms: one on the (ij-f) plane and another on the (t-s) plane.
1. Transformation to (ij-f) Coordinates

T h e independent variables ij and E have been defined in Eqs. (98)
and (99), respectively.
T h e transformed momentum and the total energy equations are those
derived in Section IV and are given by Eqs. (106) and (107), respectively.
r 1 921
For uniform Ci,, the species conservation equation (16) and the energy
equation (27) based on temperature are transformed respectively, as
= Z.f(fpif - fjmii) (275)
+ 2 (-e)f(j;O,
C
Cve
-
where F ( X ) has been defined in Section IV by Eq. (121), and
8=- T (277)
T,
When 1 is constant, it can be made to disappear from the governing
equations by replacing +j and f by 11 and f defined by Eq. (129). T h e
resulting equations in 11 and f are exactly the same as the governing
equations given above except 1 is missing.
T h e boundary conditions for Eq. (275) are the same as those developed
in Section I V and are given by Eq. (1 14) and (1 18). T h e right-hand side
of boundary condition (1 14) vanishes when the surface reaction is
frozen. T h e boundary conditions for energy equation (276) are the
known temperatures at the wall and at the boundary layer edge.
2. Transformation to ( t - s ) Coordinates
T h e independent variables t and s have been defined for constant
1 in Section IV and are given by Eqs. (204) and (203), respectively.
Defining U and V by Eq. (205) the continuity and the momentum
equations become Eqs. (206) and (207), respectively.
T h e species conservation equation (16) is transformed into the fol-
lowing form for uniform Cte:
PAULM. CHUNC
T h e only energy equation, on the ( t - s ) coordinates, of our interest, is

that based on total enthalpy, and this equation becomes
T h e boundary conditions for Eqs. (278) and (279) are given by Eqs.
(213) and (216), and Eqs. (1 13) and (117), respectively.
Before entering into the actual analysis of the equations, it is men-
tioned here that the gas phase reaction rate W ias was seen in Section 111,
is usually a quite complicated function of the dependent variables T ,
Ci, etc. Therefore, the governing equations such as Eq. (275) are quite
nonlinear and are not usually amenable to a completely analytical
approach. Nevertheless, with appropriate approximations, and with the
aid of digital computers, a fairly good general understanding of the non-
equilibrium boundary layer problems has been achieved.
Exact general methods of solution of the nonequilibrium problems
are yet to be developed because the handling of a set of coupled non-
linear partial differential equations, even numerically, is a very formi-
dable task. Only recently have the equations been integrated exactly by
a finite difference scheme for a few simple cases (53).
C. DAMKOHLER
NUMBER
FOR GAS PHASEREACTION
Consider the chemical reaction term (Wi/p).T h e dimension of this

term is a reciprocal of time and we may write
where wi is dimensionless and T . is a characteristic time for the gas

?
phase chemical reaction. T h e species conservation and energy equations
(275) and (276) now clearly show the fact that the chemical behavior
of a reacting boundary layer is largely governed by the parameter
Let us investigate the physical meaning of Cgi defined by Eq. (281).

T h e variables governing the characteristic diffusion rate of ith species
W I
CHEMICALLY LAYERS
BOUNDARY
is given by Eq. (124). T h e variables which govern the boundary layer

thickness 6 are derived from Eq. (98) as
I t can therefore be readily shown from Eqs. (124) and (282) that the
coefficient of in Eq. (281) denotes the characteristic diffusion time
of the ith species. It is now clear that
I n analogy to the surface Damkohler number, we shall call iuithe

gas phase Damkohler number.
As was the case with the surface reaction (Section IV, D), igi-+ 0
implies that the effect of the gas phase reaction on the boundary layer
flow is negligible. For lUi+ 0, the chemical reaction terms in the
governing equations vanish and the problem becomes that of the
chemically inert boundary layer. T h e boundary layer is then said to be
chemically frozen. T h e frozen boundary layer does not necessarily
imply that = 0. I t only implies that T~~ >> Tdiff .
I n the other extreme of igr--t m, wi must approach zero because
each of the terms in Eq. (275) is finite. This means that the forward and
the backward reaction rates comprising wi (see Section 111) must be
equal at each point in the boundary layer. Thus there must prevail a
state of local chemical equilibrium throughout the boundary layer.
Such boundary layer is commonly called an equilibrium boundary
layer. For an equilibrium boundary layer then the algebraic equation
wi = 0 may replace the differentia1 equation (275) and the problem is
greatly simplified.
It is seen that for the general nonequiiibrium cases of finite luithe
self-similar solutions of the governing equations discussed in Section
B, 1 can be obtained only if cgiis independent of a. For practical pur-
poses then, this limits the possibility of obtaining a self-similar solution
to the stagnation region of blunt bodies.
D. SELF-SIMILAR
SOLUTIONS REGION
AT STAGNATION
1. Nonequilibrium Boundary Layers of dissociated Air

a . Dissociation-Recombination Kinetics. T h e chemical kinetics of dis-
sociation and recombination was discussed in Section 111. Utilizing the
PAUL M. CHUNC
equations given in Section 111, A, 5 and the equation of state, we derive
the following equation for a pure diatomic gas:
p cz
- = - 2 ( 2 k ~ )(XT)[----
W
- (---)1-cc'
Ct
E (1 - C ) ] (284)
P 1CC
where C refers to the mass fraction of atoms and the equilibrium term
(Cz/(l - C2))is given by Eqs. (86)-(88). T h e first term in the bracket
represents the recombination, whereas the second term represents the
dissociation.
Now briefly consider the dissociation-recombination processes for
air in a boundary layer. T h e transport properties and k, are quite
similar between oxygen and nitrogen. Therefore, when the recombina-
tion is the predominant reaction one may combine all the atoms and
all the molecules respectively and consider the air as a binary mixture
of air molecules and air atoms. Then by employing a suitably averaged
value of heat of reaction one can usually obtain an acceptable solution
of the boundary layer problem. T h e error caused by such combining of
oxygen and nitrogen is much greater when the dissociation is the pre-
dominant reaction. It is because, as seen in Eqs. (86)-(88), the equi-
librium term (C2/(1 - C2))is basically an exponential function of the
heat of reaction, and the heat of reaction is quite different between
oxygen and nitrogen. Though much of the basic knowledge could still
be had by assuming a binary model of air for dissociation dominated
reaction one must be very careful in interpreting the results.
b. Formulution. We consider that either due to a strong shock
preceding a blunt body or, in the case of a laboratory work, due to an
electric arc or a shock upstream of a body, the air along the edge of the
boundary layer is dissociated to the equilibrium value. Thus, a substan-
tial portion of the total energy of the gas is associated with the atoms in a
chemical form.ls T h e atoms recombine as they diffuse through the cooled
stagnation boundary layer toward the wall. T h e degree of dissociation
at the boundary layer edge, C, , is known from the equilibrium relations.
T h e surface is considered to be either noncatalytic or perfectly catalytic
for the atom recombinations.
Following the works of Fay and Riddell (24), we will now study the
recombination occurring in the dissociated air-boundary layer by
analyzing the appropriate nonequilibrium boundary layer equations.
T h e heat transfer will be found from the solution of the equations. Similar
analysis was also carried out by Scala (25).
For the recombination-dominated stagnation boundary layer, we
la See Section IV, F.
r 1961
CHEMICALLY REACTING NONEQUILIBRIUM BOUNDARYLAYERS
consider that the boundary layer comprises a binary mixture of air

atoms and air molecules.
As was mentioned previously, the Damkohler number Cgi becomes
independent of s at the stagnation region, and the governing equations
can be simplified as functions of the variable +j only in a self-consistent
manner. T h e governing equations suitable for the present purposes
are the momentum equation (106), atom conservation (275), and the
energy equation (276), and these respectively become at the stagnation
region of two-dimensional or axisymetric bodies with the aid of Eq.
(284):
Momentum
(Zf) 1 ) [$- (fy]= 0

+ff + (Tc
Conservation of atoms
-1 -83.51[ 1 +C,m2 ---(1 ) (1 - C p ) ] = 0

c,m c, c2
1 - c2 (286)
Energy
where m = C/C,, and the Damkohler number defined by Eq. (281) is

written for the present problem as
k,, denotes the constant portion of the specific recombination rate

coefficient k, where
2kR = %
!T (289)
k,, is found for instance from Ref. (27)and here we let w = 1.5. When
k, is given in cm6/mole2sec, T , p , and R are in O K , atm, and
cm3 atm/mole OK, respectively. Aho is the heat of recombination per
[I971
PAULM. CHUNG
unit massl9 (h, - h,) and here it signifies the mass-averaged value
between oxygen and nitrogen.
T h e boundary conditions for Eqs. (285), (286), and (287) are as
follows:
atij=O
f=f'=O
. I
m' =0 (for noncatalytic surface)

m =0 (for perfectly catalytic surface)
f' =1
m=l
6=1
c. Solution. Fay and Riddell (24) integrated Eqs.,(285), (286), and

(287) by the use of a digital computer for the equilibrium boundary
layer-edge condition of h , = 10,500 Btu/lb and C, = 0.536, and for the
various Damkohler numbers. Solution was obtained for E = 1 and
T,,, = 300K only. T h e following property values were used for the
numerical integration:
Pr = 0.71
Le = 1.4
T h e viscosity was computed by Sutherland's formula as function of T

and C. T h e equilibrium atom mass fractions in the boundary layer
were approximated by the equation
C, = C, exp {const * [ 1 - (1 /6)]} (293)
Typical atom mass-fraction profiles are shown in Fig. 17.
lo Strictly h, - h, = ST
(cnl - cD2)dT +
Ah". The first term on the right-hand side
of the equation, however, is negligible compared to Ah".
I I I I I
x t
0 2 2
-
0
2 NOIl2W?1JS S V W W O l V
PAULM. CHUNG
d . Nonequilibrium Stagnation Heat Transfer. T h e surface heat trans-

fer is found by using the heat transfer expression (45) and the definitions
of 4 and s^ as
mzo' in the above equation is zero for noncatalytic surfaces. Defining the
Nusselt and Reynolds numbers as
we obtain from Eq. (294)
where mw' and 0,' are given by the solutions of Eqs. (286) and (287),
respectively. T h e computed values of Nu/.\/%-are shown in Fig. 18.
:I
0.5
0.2
o+-
I I I I I I I
----
I
P r = 0.71
1, = 30001:
I
1
0 I I I I I I I I 1
IO-~ IO-~ 10') I 10 lo2 lo3 10'
lRCga
FIG. 18. Nonequilibrium heat transfer at stagnation region (24).

PO01
CHEMICALLY
BOUNDARY
LAYERS
Consider the heat transfer shown in Fig. 18 for the perfectly catalytic
surface. It was shown in Section 11, E that for Le = 1 and Pr = 1
the total energy equation becomes the familiar form, Eq. (22a). At the
stagnation region where ue*fh,c+O, Eq. (22a) is true for all Pr. Hence,
Eq. (22a) is applicable for all tgr1
when Le = 1. Furthermore, the bound-
ary conditions h,, and h,,, are fixed when the surface is perfectly catalytic.
T h e heat transfer therefore should be independent of c,, except for the
effects of Le f 1 and varying fluid properties with [,,, . It is seen in
Fig. 18 that the heat transfer to perfectly catalytic surfaces is nearly
independent of l,,, showing that the effects of Le # 1 and property
variations are quite small.
T h e effect of chemical reactions on heat transfer was illustrated by
Eq. (48) for Le = 1 . Equation (48) can be rewritten as
I t is readily seen from Fig. 18 that the above relation is practically correct
even for Le = 1.4, and when the effect of the varying property values
due to the gas phase reaction is included.
Finally, it is seen in the governing Eqs. (285)-(287) that E appears
explicitly only in the momentum equation for a given value of c,, .
Lees (23) showed that the term in the momentum Eq. (285) which
includes E has a negligible effect on the solution of the equation for
highly cooled hypersonic boundary layers such as the present one.
Therefore, although Fay and Riddell integrated the equations for E = 1
only, the nondimensionalized results such as shown in Figs. 17 and 18,
are also applicable to two-dimensional bodies for all practical purposes.
e. Equilibrium Heat Transfer and Lees' Approximations. For the ex-
treme cases of --ti0 or the surface Damkohler number 5 + CO, the
following heat transfer relation was derived in Ref. (24) from the numer-
ical results for E = 1 :
where
a =
\0.52 when cqfl- cx)
(0.63 when 5,,-0 and 5 - 00
Lees (1, 23) suggested the following approximations in order to treat

analytically the governing equations for the boundary layers wherein
PJP,, < <
I and U e 2 / h , e 1:
PO11
PAULM. CHUNG
( I ) (pp) for each point in the boundary layer be replaced by I(pepLc)

where 1 is a constant.
(2) T h e third terms of Eqs. (106) and (107) will be neglected. With
the above approximations, the governing momentum and total energy
equations (106) and (107) become, respectively, self-similar for all s^
when , ,C .
+ 03 or 5 .+ 03. T h e equations then can be integrated readily
when Le = 1 with the aid of the existing Blasius functions. T h e heat
transfer is obtained ( I , 2.3) from the solution as
for L e =: 1 and c = 1 .
A comparison of Eqs. (298) and (299) shows that the simple analysis
of Lees gives the heat transfer values within lo?/, of the correct value
-
if we let 1 = ( ~ , , p ~ , / p ~ pin~ the
) : . ~final results of Eq. (299). Noticing the
fact that Eqs. (298) and (299) apply for both equilibrium and frozen
boundary layers when 5 co, this particular value of I is suggested
herein for all the nonequilibrium boundary layer solutions obtained with
Lees basic assumptions when pe/pu, < <<
1 and u,2/hle 1.
2. Combustion of Injected Premixed Fuel

Often the gaseous coolant injected into a boundary layer reacts
exothermically as its temperature is raised by the boundary layer. We
consider the simplest case wherein the chemical structure of the injected
gas is changed exothermically without the need of another gas such as
oxygen. T h e problem is then essentially that of the boundary layer
combustion of an injected premixed fuel. Such a problem was anlyzed
by Sutton (54, 55) and we shall herein follow his work.
T h e combustion problem will be analyzed for hypersonic stagnation
boundary layers with the following simplifying assumptions:
(1) Lees assumptions given in the preceding section are applicable
(2) T h e diffusion coefficient and molecular weight of each chemical
species are equal. Thus, we may employ the binary form of the species
conservation equation (see Section 11).
(3) Pr = Sc = 1.
(4) All cpi are equal.
(5) Chemical reactions do not take place on the surface.
(6) T h e reaction is first order proceeding only from the premixed
PO21
CHEMICALLY
BOUNDARYLAYERS
fuel to a product, and the reaction rate is given by a simple Arrhenius

form. Thus
W -- - -1
_
P
c exp [--E,,/(RT)I
T9b
(300)
where the C denotes the mass fraction of the reactant (premixed fuel)
and T~~ is constant.
(7) The effect of dissociation of oxygen and nitrogen is neglected.
With the above assumptions, the momentum equaiion (106), the
energy equations (107) and (276) and the species conseFvation equation
(275) for the reactant become, respectively,
f"'' +ff" = 0 (301)

H " +fH' = 0 (302)
e" +je' = -cgb (--I
I Ah"
c exp [-En/(RT,41 (303)
C" +fC' = L,,C exp [--E,/(RT,B)] (304)

where the prime (') denotes the total differentiation with respect to q
defined by Eq. (129), and
f' = o (307)
=-
1 (tpTw
hle
+ CwAh") (308)*O
e = ow (309)
Heat of formation of the reactant was considered to be equal to the heat of combustion,
Ah", whereas that of the rest of the gases is zero.
21 See Eq. (1 14) forfj = 4 = 0.
PAULM. CHUNC:
and at q = co
f' 1
H=l
0=1
c=o
Solution of the Blasius equation (301) is well known. We shall now
obtain 0 in terms of C and the known function f from Eqs. (301) and
(302). T h e system of Eqs. (301)-(304) will then be reduced to the
species conservation equation (304) and the expression for 0 in terms
of c.
T h e total energy equation (302) is identical in form to the momentum
equation (301). Hence,
We have
h = 2 C,hi = cvT+ dh"C (3 I 6)2'
Equations (315) and (316) now give
T h e substitution of Eq. (317) into Eq. (304) for 0 produces a nonlinear

equation in C only. T h e resulting equation was solved (54) by the use
of a digital computer for several values of &,, , Ah"ic,,T,, and E,/RT, ,
and for TWITe= 0.25 and fu, = -0.07071 1 , respectively.
a. Typical Solutions. T h e effect of increasing i& can be seen in
Fig. 19. For a value of 0.1 the combustion is slow, as indicated by the
proximity of the curve to the broken line, which represents the chemi-
cally inert case. As lgbincreases the local rate of consumption of injected
gas is increased which decreases the local concentration. For a value of
100 the injected gas reacts almost as soon as it enters the boundary
layer.
T h e effect of increasing the activation energy is seen from Figs. 19
and 20. As EJRT, is increased from 0.1 to 1 the combustion rate
decreases and therefore the concentration of reactant increases in the
boundary layer.
22 See the footnote for Eq. (308).

~2041
CHEMICALLY
REACTING LAYERS
BOUNDARY
NONEQUILIBRIUM
0.16 I I I
0.12
c 0.08
0.04
0 I I -
tl
FIG. 19. Mass fraction of premixed fuel (54).
6 . Heat Transfer f o r Large Values of a,( . I t was seen in Figs. 19 and

20 that as Cgb increases the predominant reaction zone moves toward
the wall. Taking advantage of this fact for large values of C u b , we shall
obtain the asymptotic behavior of surface heat transfer with respect to
{,$, by the following simple analysis. Since the important reaction zone
is adjacent to the wall when Cgb is large, we replace f (q) and f3(7) by fw
and Ow , respectively, in Eq. (304), which can then be integrated analyti-
cally to obtain, after satisfying the boundary conditions,
PAULM. CHUNG
E0 /RTe= 1
Ah"/cpTe=1
C
e, = 0.25
T h e above equation applies when &a, exp[-E,/(RT,,J] 9(fJ2. T h e

heat transfer now becomes with the aid of Eqs. ( 4 9 , (318) and (317)
Equation (319b) shows that as cobincreases for a given injection rate, a

greater portion of the combustion energy (-fit, Ah") returns to the wall.
PO61
CHEMICALLY
REACTING
BOUNDARY
I n the limit, the quantity within the braces becomes one showing that
all the combustion energy returns to the surface.
3. Reaction of Injected Combustible Gases

We have in the preceding section, studied the combustion of premixed
fuels injected into the boundary layer. We now consider the more general
case wherein a combustible gas alone is injected into a boundary layer
for the purpose of, say, mass transfer cooling. Combustion of the injected
gas now requires a continuous supply of oxygen from the inviscid region.
This problem has been studied only for the extreme limit of cob
-+ co
by several investigators [for instance, see Hartnett and Eckert (56)].
We shall herein formulate the general nonequilibrium problem and then
discuss the limiting solutions23.
First, we make all the simplifying assumptions made in the preceding
section except that Eq. (300) will be replaced by a more appropriate
reaction kinetics as follows. We consider that the combustion process can
be represented by
(fuel) + (oxygen) kf (product)

--t (320)
In accordance with the law of mass action (see Section 111), the consump-
tion rate of the fuel can be written as
where the subscripts 1 and 2 denote oxygen and fuel respectively. Again
employing the Arrhenius form we let
where rocis constant.

We may assign the heat of reaction to the heat of formation of either
of the reactants, oxygen or fuel, and let the heat of formation of rest of
the gas components be zero. I t is usually more convenient to assign the
heat of reaction to oxygen. We therefore define the total enthalpy as
h = r,T + C, Ah" (323)
where Ah" is the heat of combustion per unit mass of oxygen.
*3 Very recently, analogous nonequilibrium problems have been analyzed for certain
inviscid flow geometries. Some of these analyses will be discussed in Section V, G.

PAULM. CHUNG
T h e momentum equation is the same as Eq. (301). T h e remainder

of the governing equations become with the aid of Eqs. (321), (322),
and (323) as
H" + fH'= 0 (324)
I
c; +.fC,' = 5,c 3 c,c2 exp [-E,,I(RT,0)1 (325)24
C; + f ~ , '= c,, (-+)A 2 1

C,C, exp [-E,I(RT,~)I (326)
MI
c,"+fC,' = 0 (327)
c, + c, + c, + c, = 1 (328)
where the additional subscripts 3 and 4 denote the combustion product

and the inert nitrogen, respectively. T h e Damkohler number Cg,, is
defined as
I I
5Bc -- (1 )p + (329)
In addition to the governing equations, (301) and (324)-(328), Eq. (323)

gives the relationship between H and 8 as
Now we have, besides f,six unknown functions H , 0, C, , C2 , C, , and

C, , and the six equations, Eqs. (324)-(328), and (330). T h e equations
therefore can be solved for the unknown functions as soon as the bound-
ary conditions are specified.
T h e boundary conditions are as follows: [see Eqs. (40) and ( 1 14)].
atq=O
24 The density ratio p/p, is replaced by the temperature ratio I/O.

[2081
and at 71 = co
H = l
c,, = 0.22
(332)
c,, = 0
C,, = 0.78
I t is seen that Eqs. (324) and (327) can be integrated readily in closed
forms in terms off because of their similarity to the momentum equation.
Equations (325) and (326), on the other hand, require a numerical
integration. We may, however, introduce a simple equation similar to
Eq. (327) which can replace Eq. (326), thus leaving only Eq. (325)
to the numerical analysis. This is done by first subtracting Eq. (325)
from Eq. (326). If we let in the resulting equation
M
c, = (4)c, - c, (333)
M2
we obtain the equation
c;+jc; = o (334)
T h e boundary conditions for Eq. (334) become from the boundary

conditions for Eqs. (325) and (326) as:
atT=O
and at 77 = 03
c, = -Cle = -0.22 (336)

Now, the solution of the problem is reduced to integrating Ey. (325)
numerically with the aid of the closed form solutions of the rest of the
governing equations. Heat transfer is obtained from the solution and
Eq. (45)as
qw = d(1 + 4IS(PrCle)ol ( c J J 0 w ' (337a)
(337b)
T h e final numerical solution however, is not available in the literature.

Without obtaining the numerical results we may still discuss the
qualitative behavior of the equations with respect to I&. We may then
analyze the limiting cases of I & ~ --+ co.
~091
PAUL M. CHUNG
a . Development of Diffusion Flames. For smaller values of cur,
the combustion will take place slowly throughout the boundary
layer. T h e combustion rate will be distributed in accordance with
(t/B)C,C, exp[-E,,/(RT,B)], and the maximum rate will occur some-
where in the outer portion of the boundary layer where the temperature
is maximum (see Fig. 21).
As &,. is increased, most of the injected fuel burns before the boundary
layer edge is reached and, at the same time, most of the oxygen con-
vected into the boundary layer becomes depleted by reaction before they
reach the wall. Thus, the significant combustion zone becomes gradually
I
1
_---
1
FIG. 21. Development of diffusion flame.
CHEMICALLY BOUNDARY
LAYERS
narrowed as the combustion requires the coexistence of the fuel and the
oxygen. With lot increased to a sufficiently large value a condition is
reached wherein no fuel and oxygen which enter the boundary layer
from the wall and the boundary layer edge respectively survive across
the combustion zone. In the limit of (,: -+ 00, the reaction zone becomes
so thin that we may consider it a mathematical discontinuity in concentra-
tion and temperature gradients. At such a discontinuity (combustion
sheet) both C, and C, must vanish and at the same time the diffusion
rates of oxygen and fuel must be such that the stoichiometric relation-
ship of the combustion must be satisfied. Thus, as cU,.is increased con-
tinuously, the zone of maximum combustion moves from near the
boundary layer edge to the point where the above conditions on oxygen
and fuel supplies are satisfied and it finally becomes a combustion sheet
there. With the formation of a combustion sheet the reaction rate is now
completely governed by the rate at which the fuel and oxygen are sup-
plied to this sheet which signifies the development of a diffusion flame.
T h e diffusion-controlled combustion regime, 5,. 4 co,can be analyzed
rather simply. Basically following the analysis of Hartnett and Eckert for
flat plate (56), we shall analyze the present governing equations for the
stagnation region.
b. [,,+ co with Combustion Sheet off the Wall. We first consider
that the mass transfer rate (of fuel) is such that the combustion sheet
is located away from the wall. C,,,is then zero and Eq. (337b) immedi-
ately gives the surface heat transfer since
T h e reduction off Z, with respect to f, is well known. Equations (337b)

and (338) show that the surface heat transfer is reduced in the same
manner as f is reduced with respect to the mass transfer provided the
combustion sheet is lifted off the wall. One should remember here,
however, that the driving potential is ( h , - c,T,) which includes the
maximum chemical energy of Combustion.
Let us now obtain the location of the combustion sheet. With either
the fuel or oxygen missing from both sides of the combustion sheet the
reaction terms of Eqs. (325) and (326) may be neglected for these regions.
Equations (325) and (326) are then all similar to the momentum equa-
tion (301) and their solutions can readily be written in terms off as
PAULM. CHUNG
and
C2 = a 2 f ' + b, (340)
where a and b are constants of integration. Now in view of the discussion
of the preceding section a, we apply the following boundary conditions
to Eqs. (339) and (340) at the combustion sheet (sh):
Ci,sh = CP,sh = 0
1 1
-7- (C,')sh = - 7- (C*')m (341)
Ml M2
I n addition C, is made to satisfy the boundary condition at the boundary
layer edge given in Eq. (332), and C, that at the wall given by Eq. (331).
Equations (339) and (340) then give the result
For a given f; (sincefk is known for a given f,), fsh can be found from
the above equation, and ?p& (the position of combustion sheet) corre-
sponding to fi,, is then obtained from a table of Blasius functions.
We now consider the minimum value off, which will lift the com-
bustion sheet off the wall. At the onset of the lifting of the combustion
sheet we have f&= 0 and Eq. (342) becomes
- fw = (0.22) fi2
7 (343)
f:: MI
Therefore the combustion sheet is off the wall when
-fw > 0.22 3 2

7- f; (344)
Ml
c. tgc-+ co with Combustion Sheet on the Wall. When the mass
transfer rate (-f,) is less than that given by the right-hand side of
Eq. (344), the fuel will burn as soon as it enters the boundary layer.
Also, some excess oxygen will appear on the wall unreacted because not
enough fuel is being injected into the boundary layer to consume all
the oxygen diffusing to the wall. For this case we must satisfy at the wall,
M2 ac,
pwvw = 7(pD- - pvc,) (345)
mz, aY UI
T h e application of the above boundary condition together with the

boundary condition C,, = 0.22 to Eq. (339) gives
CHEMICALLY
NONEQUILIBRIUM
With the value of C,, found the heat transfer can be obtained from
Eq. (337b). It is seen in Eq. (346) that forf, = 0, C,, = C,, = 0.22,
indicating the absence of the chemical reaction. With an increase of fw ,
C,, decreases, and a study of Eq. (337b) shows an increase in heat
transfer. Rather quickly, however, the condition (344) is satisfied with
C , , = 0 and the combustion sheet is lifted off the surface, and the
heat transfer begins to decrease with increasing f, as was explained in
Subsection b.
I n closing, it is noted that when the stoichiometric coefficients are
not equal to unity, thus different from those of relationship (320), the
present analysis is still valid provided one interprets the molecular weight
ratio fi2/&Zl as the general fuel-oxygen stoichiometric mass ratio. Also,
the analysis of the diffusion-controlled regime for Pr and Sc differing
from one can be carried out in a manner similar to the present one by
the use of the functions G given in Section IV, E.
E. THERMAL
IGNITIONPROBLEMS
We now turn to the more general nonsimilar problems by departing
from the stagnation region. As it was with the surface reactions of
Section IV, we shall only consider the cases wherein a self-similar solu-
tion of the momentum equation exists.
Perhaps the earliest problem which necessitated the engineers to
tackle the nonsimilar, nonequilibrium boundary layer equations was
the problem of thermal ignition. Marble and Adamson (57) first studied
the problem of ignition under conditions of mixing between two gaseous
streams of a premixed fuel and a high temperature combustion product.
Only the case of equal stream velocities was considered. Dooley (58)
later re-examined the problem more exactly. T h e problem of mixing
layer ignition was subsequently extended to the wake of a flat plate by
Cheng and Kovitz (59) and Cheng and Chiu (60). They considered
the case wherein a cool premixed fuel enters the wake along one side
of a flat plate whereas a hot combustion product enters the wake along
the other side. Toong (61) presented an analysis, as well as some
experimental results, which studied the thermal ignition problem
associated with the boundary layer flow of a premixed fuel along a heated
flat plate.
T h e problem considered by Toong (61) is a little more straight-
forward than the others, yet possesses most of the main features of an
ignition problem. We shall therefore study this problem herein.
We consider a subsonic boundary layer flow of cool, premixed fuel
over a heated flat plate. Provided the conditions are favorable, the gas
~ 3 1
PAULM. CHUNC
will ignite adjacent to the wall somewhere downstream of the leading

edge. Once the gas is ignited, a flame front will be rapidly developed and
the front will eventually move into the main gas stream. Our problem
is then to determine the parameters and their magnitudes which govern
the ignition by analyzing the reacting boundary layer equations from the
leading edge to the point of incipient ignition.
1. Governing Equations for Flat Plate

First, we assume that the reaction is one of the second order of the
Arrhenius type such that
W = exp [--EaI(RT)I
-k ,Pp2C2 (347)
where C is the mass fraction of the premixed fuel and k, is the specific
rate coefficient considered as constant.
Then, assuming that 1 is constant and all c,,~ are constant and equal,
the momentum equation (106), the energy equation (276) and the
reactant conservation equation (275) respectively can be written as
f"' +ff" = 0 (348)
8qq + prf8q - 25gd.ff8Ipd = -'&d
1
- m2 exp
8 (349)
where the prime (') denotes the total differentiation with respect to 7
which becomes for the present case
and
8 = TIT,
m = C/Ce (352)
It is seen that
(353)
at q = O
f =f' = 0
8 = 0," (354)
m, = 0 (no surface reaction)
(355)
2. Solution of Equations
Solution of the momentum equation (348)is available elsewhere.
Expecting that the incipient ignition would occur at relatively small
values of Cgd we seek the solutions of Eqs. (349)and (350)in the following
form of series:
A substitution of the above series into Eqs. (349)and (350)and a col-

lection of the terms with like powers of Cgd result in the following set
of ordinary differential equations:
Im;
0; + Pr fe,'
+ Sc fm,l
(359)
PAULM. CHUNG
T h e boundary conditions on the above equations are as follows:
0,(0) = I , 4403) = 0
m,(O) = mo(m)= 0
and
e,(o) = e,(m) = o
m,(O) = m,(oo) = 0
where n = 1, 2, 3 , ....
Let us now briefly consider the series method which is being presently
employed for the solution of Eqs. (349) and (350). We recall that when
the series method was used to solve the governing equations for the
surface reactions in Section IV, it generated a set of equations which
were independent of each other. Moreover, all the equations were of
similar order of complexity. One could readily obtain the solution of as
many of the generated equations as were needed, usually by numerical
means, with the desired accuracy. Thus, a solution of the original
governing equations could usually be obtained by the series method
even for rather large values of the surface Damkohler number. T h e
reason for this is that the governing equation for surface reactions,
such as Eq. (184), is linear and homogeneous. I n contrast to this, the
governing equations for the gas phase reactions such as Eqs. (349) and
(350) are nonlinear and inhomogeneous. Hence, the equations generated
by the series method are inhomogeneous and the solutions of the succes-
sive equations depend on the solutions of the preceding equations
through their inhomogeneous terms. Moreover, as can be seen from
Eqs. (360), the complexity of the generated equations increases rapidly
with n. When n is greater than one or two, the effort required to solve
the equations is quite excessive and, at the same time, the accuracy of
the solution is reduced substantially as the errors in the solutions of the
preceding equations become accumulated in the inhomogeneous terms.
T h e series method therefore is useful only for small values of &,(, in
treating a nonsimilar gas phase reaction.
Now returning to the present analysis, Toong obtained solutions for
en and m, up to n = 5 numerically. T h e property values corresponding
to the stoichiometric ethanol-air mixture were used. The typical solu-
tions are shown in Fig. 22.
3. Incipient Ignition
I t is seen that the temperature profile at the leading edge corresponds
to that of a chemically inert boundary layer, for there the Damkohler
O L
" Pr = Sc = 0.65
8, = 3.9
I
.I
1 I I 1
1 2
rl
FIG. 22. Boundary layer ignition (61).

PAULM. CHUNG
number is zero. As the distance from the leading edge increases, the
temperature profiles are modified due to chemical reaction, with the
temperature gradient at the wall increasing from a negative value to a
positive one.
T h e distance downstream of the leading edge at which the first local
maximum in temperature forms in the gas phase is called the point of
incipient ignition. This initial bulge in temperature marks the result
of sufficient accumulation of heat due to chemical reaction to overcome
the local heat loss to the cooler outer portion of the boundary layer.
Subsequent to this initial bulge the reaction rate increases rapidly
and the maximum reaction zone moves from wall into the boundary
layer thus forming a laminar flame front.
T h e incipient ignition point (<,a), is marked by (6& = 0 for the
boundary layer flow. I n the case of a mixing layer, it is required that
6, = B,,,l = 0 at the ignition point.
For the solutions shown in Fig. 22, the incipient ignition occurs at
(Cod), = 0.67 x lo8. I t was found that the Damkohler number for
the incipient ignition is very sensitive to the wall temperature. Thus,
when 6, was slightly raised to 4.1 from 3.9 of Fig. 22 it was found that
decreased to 0.375 x lo8. Besides the Damkohler number and
the wall temperature, the thermal ignition is seen from the governing
equations to be governed by the parameters Pr, Sc, EJRT,, and
cp TJC, Ah.
Some experimental results were also given in Ref. (61) which showed
a good general agreement with the theory.
F. NONEQUILIBRIUM
FLOWSOF HIGHENERGY
GASESWITH
DISSOCIATION
AND RECOMBINATION
I n the remainder of the Section, we shall study the nonequilibrium

and nonsimilar boundary layers with dissociation and recombination.
T h e surface will be assumed to be either perfectly catalytic or non-
catalytic for the radical recombination. This problem is of great interest
in aerospace technology.
T h e present state of the art is that, with the basic behavior of the non-
equilibrium flows quite well understood, the job of obtaining the
elaborate digital computer programs is in progress which would solve
the enormously complicated governing equations for multicomponent
reacting boundary layers. We shall later show some of the results that
are being produced by such computer work. We shall, however, devote
most of the present section to more basic solutions of the problems so
that we may gain a rather fundamental understanding of nonequilibrium
[2 183
CHEMICALLY
NONEQUILIBRIUM
boundary layer flows without totally losing ourselves in a numerical

jungle.
1. Initial Relaxation near the Leading Edge of Flat Plates, and

Supersonic Wedges and Cones
As was seen in the thermal ignition problem of Section E, the series
method gives a satisfactory solution of the boundary layer problems
for small values of the Damkohler number. For a given flow condition,
a smaller Damkohler number implies a small x thus near the leading
edge. T h e method therefore has a rather limited application in obtaining
solutions of nonequilibrium boundary layer problems.
T h e method as used herein, however, produces relatively simple
general solutions, from which we can glean a considerable amount of
insight into the relaxation phenomena. Though the solution itself is
only for the near leading edge, the solution gives some important quali-
tative information concerning the behavior of the boundary layer as a
whole. I t is because the relaxation near the leading edge more or less
sets a pattern for the subsequent relaxations.
We shall herein depict portions of the analyses carried out by Inger
(62) and Rae (63).
a . Formulation. We shall formulate the problem for a partially
dissociated pure diatomic gas: T h e analysis is applicable also to air
provided one takes a proper caution in interpreting the results as was
explained in Section V, D, la.
First consider the production term of atoms WJp. T h e general
expression for this term was discussed in some length in Section 111, A, 5 .
For the nonequilibrium stagnation boundary layer analysis of Section
D, 1, Eq. (284) was employed for W J p . We rewrite Eq. (284) in the
following more specific form:
=
1
-2 (kRoF ) (RT)1
P C2
[i-i.l; - -
1
4P
exp (- RT ) (1
AE"
-- - C)] (363)
where
k,, is defined by Eq. (289).

Now we assume that Pr, Sc, and 1 are respectively constant and all
cui are constant and equal. T h e momentum equation is then the standard
~191
PAULM. CHUNG
Blasius equation whose solution is available elsewhere. T h e energy

equation (276) and the species conservation equation (275) for atoms
become, respectively, with the aid of Eq. (363) as
where C is the mass fraction of atoms, and Ah" and AE" are the heat of
recombination per unit mass and per unit mole, respectively. Also
m = CjC, . T h e Damkohler number defined in Eq. (281) becomes for
the present case
&,, = 7
4L kR0 P X
(R) (for flat plate and supersonic wedge) (367)
4L P
- (-) x (for supersonic cone)
kR0
l,, =zz -
3u, -
T,W+2
For units of the parameter of the above equations the readers are
referred to the sentence following Eq. (289).
e(o) = e w , e(4 = 1
m,(O) =0 (fork, = 0)
(369)25
m(0) = 0 (fork, 3 00)
m(a)= I
b. Solutions of the Equations by Perturbation. We let
2 S I tis actually meant here that the surface Damkohler number, 5, is either 0 or co.
However, in order that 5 may not be confused with 5, , we will let k, = 0 and K, --t co
represent 5 = 0 and 5 --fco respectively.
r2201
CHEMICALLY BOUNDARY
LAYERS
Only the first order perturbation functions 6, and m, are needed for our
present purpose. Substituting the series (370) into Eqs. (365 and 366)
and collecting the coefficients of one and c,,
, respectively, we obtain
-1 8 +f8,' - Z f ' 8 1 = -(
C Ah"
L
Pr CpTe
(372)
1
- m;
sc
+fm,' - 2f'm1 = Do
where

8,(0) = 6,"9 Urn) = 1
8,(0) = 8,(m) =0
and when k, = 0,
m,'(O) = 0, m,(co) = 1
(375)
m,'(O) = m1(03) = 0
whereas when k, -+ co
m,(O) = 0, mo(.o) = 1
(376)
m,(O) = ml(.o) =0
T h e zeroth perturbation equations (371) are the well-known governing

equations for chemically inert boundary layers and here they describe
the frozen boundary layers wherein Cge = 0. For all finite rates of the
chemical reaction then the boundary layer is always frozen at x = 0
as it is seen from the Damkohler numbers defined in Eqs. (367) and
(368). Physically, this is because the boundary layer thickness is zero at
x = 0 and therefore the diffusion time is also zero.
T h e solutions of Eqs. (371) are well known. We shall in the following,
analyze Eqs. (372) and find the first-order perturbations in atom concen-
tration and heat transfer caused by the chemical reactions.
c2211
PAULM. CHUNG
Noticing that Eqs. (372) are linear and also the homogeneous parts of
the equations are the same as Eq. (193) for yn = 1 already analyzed in
connection with the surface reactions, el(,) and ml(7) are readily ob-
tained by a formal integration of the equations as
whereas for k, + 00
In the above solutions Y,(Sc, 7) designates the solution of Eq. (193) for
yn = 1 but satisfying the boundary conditions Yl(Sc, 0) = 1 and
Y,'(Sc, 0) = 0. Also, the integrals Il(Pr, 7) and 12(Pr, 7) are defined by
T h e integrals Il(Sc, 7) and 12(Scl 7) are the same as those defined by

Eqs. (380) and (381) respectively except Pr is replaced by Sc.
T h e heat transfer becomes from Eqs. (45) and (370) as
T h e first-order perturbation in heat transfer is then obtained by the

use of Eqs. (377)-(379) as
CHEMICALLY
REACTING
BOUNDARY
Here again we see, from Eq. (384), the previously discussed fact that
the heat transfer is independent of the gas phase reaction when k, -+ co
and at the same time Le = 1.
We shall defer the discussion of the numerical results until after the
solutions have been obtained for a greater portion of the body than the
present region.
One of the important general informations concerning the reacting
boundary layers can be obtained from the solutions derived herein. It
is seen in Eq. (363) that we are actually considering two opposing
reactions occurring in the boundary layer. They are the recombination
and dissociation respectively represented by the first and the second
terms in the bracket of Eq. (363). Since one is exothermic reaction
whereas the other is endothermic, it is often helpful to find out quickly
which of the two opposing reactions is the predominant reaction for a
given flow condition. We shall herein obtain the boundary layer criteria
which govern the relative magnitudes of the dissociation and recombina-
tion reactions. T h e analysis is based on the near-leading edge solution;
however, the pattern set near the leading edge as to the predominancy
of one or the other reaction will be unchanged throughout the rest of
the boundary layer.
c. Boundary Layer Criteria for Predominancy of Either Dissociation
or Recombination. T h e overall information concerning the reactions
occurring in the present boundary layers can be obtained by studying
the net effect of the chemical reactions on the m,, when k,, = 0 and
q', when k, -+ co. ' T h e positive mlw for k, = 0 implies that the
chemical reaction in the boundary layer is more of dissociation than
recombination whereas the reverse is true when m,, < 0. When k, 00,
-+
the surface gradient mi, tells the story as to the predominance of dis-
sociation or recombination depending on whether mi, is positive or
negative. We obtain, from Eqs. (378) and (379),
mlw = --I,(Sc, co) (for k,u = 0) (385)
T h e integral I,(Sc, co) is positive and is given when Sc is assigned.

Hence, both the sign and magnitude of m,, and mi, depend directly
on - I,(Sc, 00).
Let us therefore consider the integral I,(Sc, to) defined by Eq. (380).
As is seen from the equation, I,(Sc, 00) essentially represents the non-
dimensional averaged value of @, weighed properly according to its
importance at each location across the boundary layer. Il depends on
12231
PAULM. CHUNG
C, rather strongly through @, . When the gas is in equilibrium along the

edge of the boundary layer it can be derived from Eq. (373) that, by
assuming 8 to be constant,
It is seen that (1 + C,)/C, - @, is rather insensitive to C, . Sub-

stituting Eq. (387) into Eq. (380), Inger (62) computed the function
(I + C,)/C, * I,(Sc, m) which is now quite insensitive to C, . T h e
results given in Ref. (62) are reproduced here in Figs. 23 and 24.
With the gases considered to be in chemical equilibrium along the
boundary layer edge, the observable reaction (net reaction between
dissociation and recombination) takes place within the boundary layer
as the equilibrium is being destroyed by heating and cooling of the
boundary layer by dissipation and cold wall, respectively. It is well
known that when the energy ratio (ue2/2)/cpTeis small the frozen
temperature profile, 8, decreases monotonically from the boundary
layer edge to the cold wall. Thus, the equilibrium of the gases in the
boundary layer, particularly those near a cold wall, is destroyed in such
a way that a tendency of chemical reaction is to predominantly recombine
[see Eq. (363)l. When (ue2/2)/cpT,is large, on the other hand, it is
known that the frozen temperature profile exhibits a temperature maxi-
mum within the boundary layer, I n his region where the temperatures
are much higher than T,, the equilibrium of the gases is destroyed in
such a way that the tendency to dissociate becomes predominant. Near
a cold wall, however, the tendency to recombine is prominent as was the
case with the small (ue2/2)/cpTe .
With this general background let us first consider the boundary
layers over perfectly catalytic walls. It is seen in Fig. 23 that the recom-
bination is never a predominant gas phase reaction when k, -+ 00. It
is because the surface, while forcing the gases in contact with it to become
in a state of equilibrium, creates near-equilibrium states among the gases
near the cold wall by drawing atoms away from them through diffusion.
Thus, the tendency to recombine in the gas phase near the wall is frus-
trated by the infinitely fast surface recombination. Therefore, very
little observable reaction takes place when (u,2/2)/cpTeis small. When
(u,2/2)/cpTeis large, however, the fast recombination at the surface has
no adverse effect on the dissociative reaction in the maximum tempera-
w41
CHEMICALLY
REACTING BOUNDARY
I I
lo-' 1 10 lo2
-
ue2
2Cp'e
FIG. 23. First-order perturbation integral for perfectly catalytic surface (62).
ture region of the boundary layer and the dissociation becomes the
predominant gas phase reaction. It is also seen in Fig. 23 that because
the dissociative reaction takes place in the gases sufficiently away from
the wall, the wall temperature has little effect on I,(Sc, a).Finally, the
dissociative reaction is seen to become prominent as (u,2/2)/c, T, becomes
greater than about 5.2s
Next, let us consider the boundary layers over noncatalytic surfaces.
26 Note that this particular value is only for dE"/(RT,)= 15.
~251
PAULM. CHUNG
sc= 0.5
(1) = 1.5
-= I5
RTe
c,<<1
-900
-1200
-
lo-' 1 10
c lo2
2cp'e
FIG. 24. First-order perturbation integral for noncatalytic surface (62).
Now the natural tendency to recombine in the cooled boundary layer

when (u,2/2)/c,Te is small is fully satisfied because the waIl no longer is
an atom sink. T h e recombination rate is proportional to 1 / T w - t 2 = l/T3.5
as can be seen from Eq. (387), and also most of the recombination
takes place near the wall where the temperature is minimum. Figure 24
shows therefore, the strong dependence of I,(Sc, co) on the wall tempera-
ture. Due to the rather large recombination activities, it is seen that the
dissociation predominates over the recombination only when (ue2/2),'c,T,
P261
CHEMICALLY
BOUNDARY
LAYERS
becomes greater than about 10 when 6 , .1. Also, it is seen that

due to the recombination activities near the wall, the effect of the wall
temperature on I,(Sc, co) persists even in the predominantly dissociative
regimes.
2. Dissociative Relaxation over Adiabatic, Noncataijtic Flat Plate

A complete nonsimilar solution of a nonequilibrium boundary layer
problem was obtained first by Chung and Anderson (64) for a flat plate
employing a modified Pohlhausen integral method. They considered
the predominantly dissociating boundary layers of oxygen over an
adiabatic flat plate. We shall briefly outline the method and show some
of the results.
a . Formulation for Pohlhausen Integral Method. It was shown in

Section I11 that Q defined by Eq. (364) may be considered as constant
without causing an undue error. It can be shown that we may let
Q = e16.8 for oxygen. With this value of Q chosen, Eq. (363) now gives
the production rate of oxygen atoms. For simplicity, we assume that 1
is constant and Pr = Sc = 1. We further assume that cpi are constant.
T h e wall is considered adiabatic and noncatalytic. Therefore
(aT/ay), = (aC/ay), = (ahJay), = 0 where C is the mass fraction of
atoms. T h e solution of the total energy equation (279) is then obtained
by inspection and is H = 1 for all s and t . T h e momentum equation
(207) and the conservation equation (278) for atoms are now integrated
across the boundary layer. It is assumed that the thickness of the diffusion
boundary layer is the same as that of the momentum boundary layer.
The integrated equation becomes, after some rearranging and with the
aid of Eq. (363),
Momentum equation
Conservation of atoms
(+) A', + F i x ,
1 1
15.8 - --]RT,e
AE"
(1 - C ) / dY (389)
PAULM. CHUNC
where
1
F, = UCdY
0
and F, was defined in Eq. (236). T h e (') here denotes the total differen-
tiation with respect to X.T h e Damkohler number is defined here as
where A' is constant to be found from the subsequent solution of the

momentum Eq. (388).
T h e total enthalpy defined by Eqs (20) and (6) becomes for the present
case, by letting h,O - h2O = Aho,
h, =c,~T + Cdh" + -2 111

(392)
since C(cp,- cD2)/cp2Q 1 where subscripts 1 and 2 denote the atoms

and molecules respectively. I t has been shown that h , = h , for the
entire flow field of the present interest. We can therefore rearrange
Eq. (392) and derive
0=1 + (*)M2 [l - (
31(=) -
Ah"
c (393)
where y is the ratio of the specific heats and M , is the free stream Mach
number.
b. Solution of the Equations. First employing a sixth degree poly-
nomial in Y to approximate the velocity profile the momentum equation
(388) is readily integrated to give
h=XX
(394)
A' = 36,0361985
Now let us consider the all-important species conservation equation

(389). T h e success of the method depends largely on finding an approp-
P281
CHEMICALLY
BOUNDARYLAYERS
riate profile for C. We are aware that the problem is nonsimilar due to
the chemical reaction term. T h e profile for C, therefore, should be able to
adjust its shape along X in such a manner that the main effect of the
chemical reactions, varying along X,can be correctly exhibited. It is
known a priori that the dissociation will be the predominant reaction.
Also, because of the exponential function appearing in the dissociation
term of Eq. (363), most of the dissociation will take place near the adia-
batic wall where the temperature is maximum. We shall therefore
choose a profile for C such that its general shape will be directly tied
to the local reaction rates along the wall.
With the above stated conditions in mind we approximate the atom
mass-fraction profile by the fifth degree profile
C ( X , Y ) = $,(X)Y" (395)
n=O
Five out of the six coefficients of the above polynomial are determined
by satisfying the following boundary conditions derived with the aid
of Eq. (278):
aty=O
ac
-- =0
%Y (396)
-I
. c,: 1 AE"
I 1 + c, --
4pexp
15.8 -
Y-1 M I Ah"
RT' +2 cp2TeC")
- (1 - C,)(1 (397)
It is seen that the C profile is now tied to the local reaction rates
along the surface through its second derivative at the wall, Eq. (397).
T h e boundary conditions (396)-(398) determine the five coefficients
of Eq. (395), b, through b, , in terms of the remaining coefficient bo(=C,J.
T h e profile of Eq. (395) together with the sixth degree U profile are
~291
PAULM. CHUNG
substituted into the conservation equation (389). T h e resulting differen-

tial equation is highly nonlinear, however, it is readily amenable to a
numerical integration.
T h e integration of Eq. (389) must begin at the leading edge. I t is
essential for a succesful integration that the function F4 and its deriv-
ative be well behaved at X = 0. T h e atom concentration is zero every-
where for X ,< 0; F4 is therefore zero for X = 0. T h e limiting value
of the derivative of F4 is obtained as follows. Evaluation of Eq. (389)
shows that
0
lim F4'-+6
x-0
Hence, an application of L'Hospitals' rule gives

AE"
(399)
Therefore F4' is seen to be well behaved at the leading edge. With the
value of F4' given by Eq. (399) and the boundary condition of F4 = 0
at X = 0, the numerical integration can commence in a regular manner.
T h e conservation equation (389) was integrated by the use of a digital
computer in Ref. (64).I n the numerical integration, the following values
were used:
AE"
-= 60,000"K
R
w = 2
y = 1.4
c. Typical Results. Figure 25 shows a few of the solutions obtained

by Chung and Anderson (64).T h e effects of the Mach number and the
other flow conditions on the boundary layer relaxation are seen from
the manner in which C, approaches the equilibrium value. I t is seen
that for the flow condition corresponding to the flight altitudes of 100,OOO
to 200,000 ft and M , < 20 the boundary layer cannot relax completely
within the length of about 10 ft. T h e relaxation will require a much
longer distance for nitrogen because d E o is much greater for nitrogen
than oxygen. Also, the boundary layer relaxations would be much
slower than those shown in Fig. 25 when the boundary layer is cooled
by the wall.
Recently, Blottner (53) obtained an exact solution of the present
problem by an elaborate finite difference method wherein a considerably
~2301
CHEMICALLY LAYERS
BOUNDARY
100 I I I I I 1 I
iB M,= 15
C, (AT EQUILIMIW)=0.527
-
C, (AT EQUILIBRIUM)- 0.366 -
--- RAES LOCAL SIMILMITY
0 2 4 6 8 10 12 14 16
x (feet)
FIG.25. Relaxation of oxygen along adiabatic wall.
greater amount of computer effort, compared to the integral method, was

expanded. T h e circles in Fig. 25 show Botners results. It is seen that
the results of the integral method agree with the exact results extremely
well for the adiabatic flat plate considered.
Rae (63) also obtained a solution to the problem by an approximate
local-similarity rule and the typical result is shown in Fig. 25.
T h e analyses of Blottner and Rae will be considered subsequently.
3 . Dissociative Relaxation over Cooled, Catalytic Flat Plate and

Supersonic Cones
a . Numerical Methods. Two numerical methods have been devel-
oped recently which successfully handled some of the nonsimilar, non-
equilibrium problems. They are the finite difference method as used
by Blottner (53) and the Dorodnitsyn integral method as applied by
Pallone et al. (65).
T h e finite difference method when it is successfully applied to a
nonsimilar boundary layer problem usually gives the most accurate
solution compared to other methods. It takes, however, unusual care
and manipulation to solve a problem by this method. Since the method
is a highly complicated and specialized numerical technique, we shall
[23 11
PAULM. CHUNG
not go into the method itself herein. We shall, however, consider some
of the results of the method subsequently.
T h e Dorodnitsyn integral method begins with dividing the boundary
layer into an arbitrary number of curvileaner strips in the streamwise
direction. The appropriate profiles, such as profiles for atom mass
fraction and temperature, are then approximated for each strip by
satisfying a certain number of boundary conditions at the dividing
lines of the strips.
Finally, the integro-differential equations derived for each strip by
integrating the original governing equations across the strip width are
solved simultaneously with the use of the profiles developed. T h e
Dorodnitsyn integral method therefore is basically the Pohlhausen
integral method applied to each of the prestriped portions of the bound-
ary layer simultaneously.
There are advantages of the Dorodnitsyn integral method- over the
Pohlhausen integral method when they are applied to nonequilibrium
boundary layer problems. It was mentioned in the preceding Section 2
that the success of the Pohlhausen integral method depended largely
on finding the profiles which can exhibit the major chemical phenomena
in the boundary layer and their variations along x. I n the analysis of the
reaction over an adiabatic flat plate, the most important reaction zone
was near the wall where the temperature was maximum. Thus applying
the proper boundary conditions at the wall, the profiles satisfactorily
exhibited the major chemical phenomena. T h e method therefore, gave
very accurate results. Consider now the problem of a highly cooled flat
plate. When (u,2/2)/cp,Teis large, one of the important reaction zones
is located somewhere inside the boundary layer where the temperature
maximum exists (see Section F, lc). T h e Pohlhausen method, by its
nature, satisfies the original governing equations only at the wall and
the boundary layer edge. I t is seen, therefore, that the Pohlhausen inte-
gral method cannot satisfactorily be applied to the nonequilibrium
boundary layers over highly cooled walls when the local Mach number
is high. For such cases it is readily seen that the method can be improved
by even merely dividing the boundary layer into two strips along the
maximum temperature zone, so that the proper boundary conditions
can be applied to the profiles at this important reaction zone. Thus, the
limitations of the Pohlhausen method naturally leads to the Dorodnitsyn
integral method. T h e latter method usually involves considerably more
computer effort than the former method.
b. Typical Results. T h e nonequilibrium boundary layers of oxygen

have been analyzed for flat plates and cones at supersonic speeds and
~321
CHEMICALLY
BOUNDARYLAYERS
for highly cooled perfectly catalytic surfaces both by the finite difference
method (53) and the Dorodnitsyn integral method (65). Since the first
method basically is the more accurate one, we shall herein discuss a few
of the results obtained by this method. T h e Dorodnitsyn method,
however, gave results which are in good general agreement with the
results of the finite difference method.
Figure 26 shows the atom mass-fraction profiles for the dissociating
5 I I I I I 1 I
4- -
-- -- -_- . -
\
\
\
_---
/ -
- - @
/
I 1
0.02 0.0) 0.06 0.08 0.10 0.12 0.14 0.16
oxygen over a flat plate. T h e same property values and flight conditions
as those used for the middle curve of Fig. 25 were employed except the
value of u, . As it is seen, a much higher value of u, was used for the
present case. It is seen from Figs. 25 and 26 that for x <
5 ft the peak
values of C in the boundary layer over the cooled catalytic wall are
about the same as those over the adiabatic noncatalytic wall in spite
of the fact that u, for the former caseismuch greater than that for the
latter one. T h e maximum and the minimum dissociations occur over the
12331
PAULM. CHUNG
noncatalytic adiabatic wall and the catalytic cooled wall respectively

for a given flow condition (see Section F, Ic). T h e above comparison
between Figs. 25 and 26 illustrates these extremes.
Figures 27 and 28 show the typical temperature and the atom mass-
fraction profiles over a 10" cone with a cooled, perfectly catalytic sur-
I
5 1 I I
1
ALTITUDE = 100, OOO FT
Urn= 21, OOO FTt'SEC
1
Pr = 0.7
L a = 1.4
e, = 1.81
I
I
4
1
6
FIG.27. Temperature profiles on perfectly catalytic 10" cone (53).
face. It is rather surprising to see that the oxygen has not relaxed com-
pletely to the equilibrium values even at the distance of 60 ft from the
cone tip. I t can be said from these and other results for flat plates and
cones, that it is very seldom that a boundary layer is in equilibrium
over a slender body.
4. Nonequilibrium Boundary Layer around Blunt Bodies

T h e nonequilibrium boundary layers around spherical nose and
conical after-body were analyzed by Chung and Anderson (19)employing
the Pohlhausen integral method. T h e surface was considered to be non-
P341
CHEMICALLY BOUNDARY
LAYERS
I I I 1 I
---- EQUILIBRIUM FLOW

ALTITUDE = 100. OW FT
uDD=21,OW~FT;Pc
I
4 Pr = 0.7
Le= 1.4
e, = 1.81
rl
0
0.02 0.04 0.06 0.08 0.10 0.12
C
FIG.28. Atom mass-fraction profiles of oxygen on perfectly catalytic 10" cone (53).
catalytic. Subsequently, Blottner (53) analyzed the nonequilibrium

boundary layers around hemisphere-cylinder combinations by the finite
difference method considering the wall to be perfectly catalytic. These
methods have been discussed in general in Section 3, a. We shall herein
briefly describe the analysis of Ref. (29).
a. Formulation f o r Pohlhausen Integral Method. We consider that
the boundary layer is comprised of oxygen atoms, nitrogen atoms, and
their molecules. Then, by considering the dissociation-recombination
reactions occurring between the atomic and molecular species of oxygen
and nitrogen, respectively, we write the production terms for the atoms
with the aid of Eq. (363) as
C," 1
-
WI
P = -2(2kR1)
P 2 [ 1 + c, + c3 - - K,,(0.22 - c,,]
(m) (400)
4p
1
w3 - - K,,(0.78 - C3)] (401)
P
PAULM. CHUNG
where the subscripts 1 and 3 denote oxygen atoms and nitrogen atoms
respectively. As in Sections 2 and 3, we shall use the following simplified
forms of the equilibrium constants:
K,, = exp (16.2 - e)

T
(
KD3= exp 16.2 -
T
T h e specific recombination rate coefficient k,, and k,, are considered

to be equal and
~ 3 x l0ls
2 k R , = 2kK3 = 2 k = (-)300
T
1.5 cm6
mole2 sec (403)
I n the following boundary layer we assume that l is constant, all

cpi are equal and constant, and the binary diffusion coefficients between
the species are all equal. It is further considered that the air dissociates
to the equilibrium state as it passes through the bow shock and reaches
the boundary layer edge near the stagnation region. T h e air is then
assumed to be chemically frozen at this state as it flows around the body.
Thus the atom mass fraction is uniform along the boundary layer edge.
A survey of the nonequilibrium inviscid flow analyses shows that the
reaction freezes very rapidly beyond the stagnation region and, therefore,
the present approximation of frozen boundary layer edge is usually
satisfactory.
As in Section 2, we assume that the energy and the diffusion boundary
layer thicknesses are the same as the momentum boundary layer thick-
ness. Now, the following integro-differential equations are derived by
integrating across the boundary layer the momentum equation (207),
total energy equation (279), and the diffusion equation (278) for oxygen
and nitrogen atoms:
Momentum
Total energy
Conservation of oxygen atoms
(%)= PePe
[F(X)I2
dF6
dX
AF
= &h
1
0
-
1
C1,64.5 1 [ + Cleml
(Clem,)
+ C3,m3
--
4P exp
[ 1-
6 1,000 (0.22 - Cleml)] dY
- Tee
Conservation of nitrogen atoms
where
8
y P
Y = fo,,dYIIokdy
1
A = - [uer6 -!
80
-
dy] s6
o Peo
1
F5 = U(l - H ) dY
0
F6 = / 1
U ( l - m,) dY
s
0
F, = U(1 - m3)dY
0
I n deriving Eqs. (404) and (405), the last terms of Eqs. (207) and (279)
have been neglected in accordance with the Lees approximation for the
highly cooled blunt body flows (see Section D, 1,e). T h e mass fractions
of oxygen and nitrogen molecuIes are found in terms of their respective
atom mass fractions as (0.22 - C,) and (0.78 - C,).T h e Damkohler
number is defined as
PAUL M. CHUNC
b. Solution of Equations. T h e momentum equation (404) is not

coupled to the rest of the equations and can be solved immediately.
A fourth degree polynomial in Y is assumed for U . T h e solution gives
1260
= As = __ S
37 (410)
Now we must solve the three coupled Eqs. (405)-(407).In order to

solve them we must first obtain the appropriate profiles.
T h e total enthalpy profile is quite regular at all reaction rates and
a cubic polynomial in Y is chosen to represent it. Hence
H = Hw + ( I - Hw)(3Y2- 2Y3) ;
+ (__ ) W
( Y - 212 + Y3) (411)
where Hw will be found in terms of the given T,,>,and C,, and CEU, to
be obtained from the solutions of the species conservation equations.
(aH/aY), will be obtained from the solution of the energy equation
(405).
As has been discussed in Sections F, 2 and F, 3, a, the success of a
Pohlhausen integral method greatly depends on finding the appropriate
atom mass fraction profile which can exhibit the main chemical charac-
teristics of the boundary layer. Each of the mass-fraction profiles for
m, = Cl/C,, and m3 = C,/C,, is expressed by a linear combination of
two functions. Since the same considerations apply in both cases, only
the derivation of the m1 profile will be explained here. We let
T h e function @, is chosen to represent the frozen profile for a given

value of mlw and satisfies the boundary conditions on the homogeneous
part of Eq. (278). T h e second function Q,(Y) acts as a perturbation
to the frozen profile due to the chemical reaction term of Eq. (278).
We express @, by a fifth degree polynomial and SZ, by a sixth degree
polynomial. These functions satisfy the following boundary conditions:
atY=l
CHEMICALLY LAYERS
BOUNDARY
T h e expression for m, becomes
m1(X y ) = m,, + (1 - m,,)(10Y3 - + 6Y5)

15Y4
+ &2( Y 2 - 4Y3 - 6Y4 - 4Y5 + Y6) (41 5 )
T h e function x1 is determined by satisfying Eq. (278) at the wall and is
T h e profile given by Eq. (415) satisfies the boundary conditions for

Eq. (278) plus the compatability conditions derived from it. T h e profile
approaches the frozen profile as the Damkohler number decreases to
zero. It is seen from Eqs. (406) and (407) that an expression for tempera-
ture is needed before the equations can be integrated. We derive from
the definition of total enthalpy [see Eq. (20)].
Now Eqs. (405)-(407) become ordinary first-order differential equations

with the aid of the profiles and Eqs. (410) and (417) derived above.
T h e equations are now amenable to a numerical integration.
T h e heat transfer is obtained from the solution as
4w = ($)ah W
(see Section D, 1, e).

where we may put 1 = [(p,pw/pepe)o]0.2
c. Numerical Results. T h e coupled nonlinear differential equations
(405)-(407) were solved in Ref. (19) by the use of a digital computer for
bodies comprising a spherical nose and a conical after-body. T h e inviscid
flow properties were approximated by Eqs. (183). Prandtl and Lewis
numbers were taken to be 0.72 and 1.4, respectively, and the averaged
value of cp for the boundary layer was employed.
Figure 29 shows the typical rn, and m3 profiles.
Heat transfer along the surface of the body is shown as a fraction of
the equilibrium heat transfer in Figs. 30 and 31. T h e figures indicate
12391
PAULM. CHUNG
1 .o
- I - -
0.8 -
c) 0.6 Alt = 150,000FT
s
E U 26, OOO Fl/SEC
= 0.22
4
Cle
CSe= 0.51
o'2
0
t 0.2 0.4
Y
0.6 0.8 1 .o
FIG. 29. Mass fraction profiles of oxygen and nitrogen atoms around cone with
spherical nose (k, = 0).
that the chemical reactions in the boundary layer freeze rapidly as the
air expands around the spherical nose and then recover toward equi-
librium as the boundary layer thickens along the conical after-body. It
is interesting to note that for the larger nose angle, 8 = 70", the bound-
ary layer is almost completely frozen at X = Xo for all the cases presented
here, and it will take an unreasonably long distance along the after-body
for the chemical state to recover to that at the stagnation region. Also,
a comparison of the present figures, Figs. 30 and 31, with fig. 8
shows that for the same blunt body the gas phase recombination freezes
much faster around the nose than the surface recombination. The
simultaneous gas phase and surface recombinations around blunt bodies
will be discussed at the end of the present section.
T h e influence of wall temperature and nose radius on the chemical
state of the boundary layer and heat transfer is also evident from the
figures.
T h e accuracy of the integral method was checked only at the stagna-
tion region where an exact self-similar solution is obtainable. After a
self-similar transformation analogous to that given in Section D, 1, the
governing Eqs. (278) and (279) were integrated exactly. T h e non-
~401
a
w
U
v
-
0
a
I- _ _ _ - ---- - -
0
0
0
m
0
LL
-2 _ _ _ - - ---- - -
0
0
U
F R O Z E N HEAT TRANSFER
-
LOCAL SIMIWIN, EQ ( 4 4 0 )
ALT = 150,000 F T ,1 = l 5 d K
U, = 26, 000 FT/SEC
I 1 1 I
O I 2 3 4
-
X
L
FIG.30. Heat transfer around cone with spherical nose (k, = 0).
PAULM. CHUNC
3 0.2 -
L
FROZEN HEAT TRANSFER -
0 1 2 3 4
FIG. 31. Heat transfer around cone with spherical nose (k, = 0).
equilibrium stagnation heat transfer obtained with the integral method

was found to be well within 10% of the result of the exact integration
for the cases shown in Figs. 31 and 32.
d. Limitations of the Method. T h e general limitations of the Pohl-
hausen integral method as applied to nonequilibrium boundary layers
have been discussed in Section 3, a. For the highly cooled boundary
layers of the present consideration there exists another practical limita-
tion which should be mentioned here.
As it is seen from Eqs. (413), (415), and (416), the effect of the most
important reaction zone near the wall for predominant recombination
on the atom mass-fraction profiles appears through (a2m/aY2),. T h e
magnitude of this second-order derivative increases rapidly as 8, is
decreased because of the factor 1/Owt2in Eq. (416). It was found that the
behavior of the present profile is not satisfactory when 8, < 3 while
w = 1.5 and at the same time is very large. T h e difficulty here when
8, is small for a large value of 5,a is quite similar to that one faces with
the velocity profile when the effect of the pressure gradient term becomes
too large (see Ref. 66).
12421
5. Approximate Closed Form Solutions

We have seen through the preceding sections that the governing
equations for chemically reacting nonequilibrium boundary layers are
of such nature that one must resort to some sort of numerical integration
for the final result. In the light of the numerical solutions obtained,
however, one begins to re-analyze some of the problems with a much
greater amount of approximations than that one originally dared to
make. S o p e of these analyses have resulted in closed form solutions. The
acceptability of such analyses depends on the comparison of the results
with the results of more accurate solutions. Since the numerical solutions
of the nonequilibrium boundary layer problems are still quite scarce,
the final judgment on some of these closed form solutions must be
deferred to a future date.
Herein we shall consider a few of the approximate closed form solutions
available. Also in line with the general theme of approximate closed form
analysis, we shall discuss the scaling laws.
a. Dissociative Relaxation over a Flat Plate. The first attempt to
obtain a closed form solution for the nonequilibrium flat plate boundary
layer goes back to the work of Jarre (67).Jarre approximated the reaction
term (w,/p) itself by a simple polynomial profile thus leading to a closed
form solution of the governing equations.
Rae (63)recently obtained a closed form solution based on the analysis
which emphasized the predominant dissociative reaction zone of the
boundary layer. As is seen in Fig. 25, the results of Ref. (63) approx-
imate the more accurate results quite well for an adiabatic and non-
catalytic wall. We shall first describe the analysis of Ref. (63) for a non-
catalytic surface.
We consider the dissociating boundary layer of a diatomic gas over a
flat plate, a supersonic wedge, or a supersonic cone. So let C, = 0
where C is the atom mass fraction. The analysis then begins with the
first-order perturbation solutions obtained in Section F, 1. Since C, = 0,
we will use C in the analysis instead of the normalized atom mass frac-
tion m. The solutions given in Section 1 are directly applicable to the
present case if we replace the perturbation function m, by C, and discard
C, whenever it appears in the solutions.
It is seen in the perturbed solutions, Eqs. (377)-(379), that the
expressions for the perturbed 8, and C, profiles can be simplified sub-
stantially if I , can be integrated in a closed form. It is because I , as
defined by Eq. (381) is strictly a universal function independent of the
chemical reaction. It is I , defined by Eq. (380) which incorporates the
effect of the chemical reaction into the first-order solutions. We shall
12431
PAULM. CHUNG
herein first integrate Ilapproximately in a closed form (up to I,) by the

Laplace method thus simplifying the solution. Then we shall show
that the solution for 0 and C constructed for small values of Cue by the
use of the zeroth and the first-order perturbations can, with certain
assumptions, predict 0 and C for large values of l,, also.
The quantitative description of the Laplace method (see Ref. 68,
p. 85) states that if in the integral
~ ( zis) bounded over the interval, a z < <

b, and there exists a point
z = a in this interval, at which ~ ( C =U )0 and G(a)> 0 then for
B > l the integral is asymptotic to
By parts the integral Ilcan be decomposed into the form
Both terms in the right-hand side of the above equation can be integrated
by the Laplace method. Hence the first-order perturbations, Eqs. (377)-
(379), can be expressed in terms of I , .
Now, let us consider a noncatalytic surface and find the surface mass
fraction of atoms for small values of Cue . Since C, is zero, we have from
Eq. (385)
c, = c,, = - I,(SC, 00) (422)
and upon the application of the Laplace method on Eq. (421) we obtain
* [I,(SC. 00) - I,@, 7*)1 (423)

where f*) denotes the point at which (l/d0) = 0 and (l/0,) > 0. The
Blasius momentum equation and the expression (373) for Qj0 , have been
used in deriving Eq. (423). Cue is defined by Eqs. (367) and (368).
P441
Equation (423) is, strictly speaking, applicable only when &

,, is small.
Rae proposed, however, that this simple solution be also applied to large
values of 6, by assuming that:
(1) Equation (423), which is rigorously correct only for 6, < 1,
is in fact, the correct form of the solution for all values of 5, , provided
only that the maximum of Bo , appearing in Eq. (423) be replaced by the
maximum of the local B at the point in question.
(2) The local maximum temperature and surface concentration bear
the same relation to each other at any value of Cge as they do for small
values of 5,.
We shall illustrate the proposed extention of Eq. (423)by applying it
to an adiabatic flat plate. So that we may compare the results to those
obtained by the Pohlhausen integral method in Section 2, we take all the
property values of the gas to be equal to those used in Section 2. For the
boundary layer over an adiabatic flat plate q* = 0. Thus applying the
above assumption 1 to Eq. (423), but bringing back the recombination
term of (wJp) missing in the first-order perturbation since C, = 0, we
obtain
Assumption 2 is not needed for the adiabatic case because 0, and C,

are exactly related by
we2 Ah
ew=i+---
2cvT, cvTecw
(425)
For a given C, Eqs. (424) and (425) can readily be solved for x.
Considering the approximations involved, Eqs. (424) and (425) give
surprisingly good results for the adiabatic flat plate (see Fig. 25). The
method should also produce equally good results when a wall is cooled
and K, -+ 03. When a wall is highly cooled while It, 0, however, the
--f
present method may not give as good results. It is because there exist
two important zones of reaction, in this case-one at the maximum
temperature region and another near the cooled wall. The Laplace
method emphasizes only one of the two important zones which is the
dissociation zone away from the wall and completely neglects the other
important zone near the wall which is the recombination zone.
It is seen, however, that if one integrates Il numerically instead of
resorting to the Laplace method and makes use of the extension proposed
by assumptions 1 and 2, then the simple method described herein will
12451
PAULM. CHUNG
be improved particularly for the highly cooled, noncatalytic surfaces.

In fact Il has already been integrated numerically by Inger (62) and the
results have been given in Figs. 23 and 24. A solution was obtained for
the highly cooled noncatalytic flat plate, by utilizing the assumptions
proposed by Rae and the values of IIgiven in Figs. 23 and 24, in Ref. (62).
Accuracy of the result cannot be checked because no other solution is
available for this case.
b. Similitude Parameters for Dissociating Flows over Slender Bodies.
The hypersonic flight of a slender body is of great interest to current
aerospace technology. The dissociative relaxation of gases in the bound-
ary layers over such bodies is important for one can compute the rate
of electron generation, found in the same manner as the atom production
rate, once the temperature and the species concentration profiles of the
boundary layer are obtained. The electron generation rate in the bound-
ary layer determines the electron concentration in the wakes which,
together with other excited species, constitute the observables.
Let us therefore consider the main parameters which govern the
chemical characteristics of nonequilibrium boundary layers over slender
bodies.
It is evident from the analyses of the boundary layers over flat plates,
supersonic wedges, and supersonic cones included in the present
Section F, that, when the predominant reaction is the dissociation, the
chemical state of the boundary layer is governed by the parameters T,,
-
T, ,and (u (l/p) for a given gas and a surface catalycity. Hence, consider-
ing that T, and T,,, are approximately fixed, when a slender body changes
its flight condition, the chemical state of the boundary layer is similar
along the points of constant cu * (1 / p ) given by
If we further fix u , , &,(l/p) becomes proportional to (px) and this is

sometimes called the Binary Scaling law.
c. Similitude Parameters for Flows around Blunt Bodies with Simul-
taneous Gas Phase and Surface Recombinations. Due to the strong
shock, the local Mach numbers are quitelow around blunt bodies.
The predominant reaction in the boundary layer therefore is the recom-
bination of the atoms produced by the shock.
The only existing accurate solutions of boundary layer equations
which consider the simultaneous gas phase and surface reactions,
[2461
CHEMICALLY
REACTING
BOUNDARY
both at finite rates, are the few particular numerical solutions obtained
by Scala (25) for the stagnation region of blunt bodies.
Recently, however, Chung and Liu (69),Inger (70),and Rosner (72),
obtained approximate but quite general solutions which predict the
chemical state of the highly cooled stagnation boundary layers with the
simultaneous gas phase and surface recombinations. Attempts are also
being made by Linan and DaRiva (72), Chung ( 7 3 , and Inger (74),
to extend the approximate analyses around blunt bodies.
The first analysis to correlate the numerical solutions and to obtain
closed form relations for nonequilibrium boundary layers around blunt
bodies was carried out by Goodwin and Chung (75).By using the Pohl-
hausen integral method described in Section 4, they obtained a series
of numerical solutions which covered a rather wide range of flow condi-
tions for a noncatalytic surface at 1500 OK. From the solutions a particular
Damkohler number was defined which correlated all the numerical
results within 2% or so at the stagnation region. Then a local similarity
rule based on the local Damkohler numbers was devised which predicted
the chemical state of the boundary layers around spherical nose and
conical after-body within about 10yo of the numerical results.
Subsequently, Inger (76) proposed that the stagnation correlation of
Goodwin and Chung be extended to other surface temperatures by
multiplying the Damkohler number by the ratio (15 0 0 / T , ) ~ +This ~.
temperature ratio correlated the solutions of Refs. (19) and (75) for
T, = 1500 OK and the solutions of Ref. (24) for T, = 300 O K quite
well. The results for T, = 10oO OK of Ref. (19), however, were not
correlated as well by the proposed temperature ratio.
As was mentioned at the beginning, these analyses are now extended
to include the nonequilibrium surface recombination. We shall follow
the theme set by Chung and Liu (69), and Chung (73) and derive a
simple relation which predicts the chemical state of the air boundary
layer around blunt bodies with simultaneous gas phase and surface
recombinations.
First, let us begin with the stagnation region. The correlating Dam-
kohler number defined by Goodwin and Chung when the surface tem-
perature effect proposed by Inger is included is
(427)
where Aho, C,and C, denote the mass averaged heat of recombination,
total mass fraction of atoms, and the mass fraction of nitrogen
12471
PAULM. CHUNG
atoms, respectively. T h e suggested units for the parameters are as

follows:
''0 - cm6 OK"
mole2 sec
/3 N l/sec
T-"K
p .- atm
cm3 atm
R = 82.06 -
mole "K
Since the predominant gas phase reaction zone on a highly cooled
blunt body is near the wall, passing to the limit, we replace the gas phase
reaction by an equivalent surface reaction. Noticing the fact, from the
expressions of (W/p),that the gas phase recombination term is propor-
tional to C2/(
1 + C) we let the equivalent surface reaction be represented
by
where k, is the equivalent surface reaction rate constant for the gas
phase reaction to be determined subsequently. Next, consider the actual
catalytic surface recombination. T h e rate of surface recombination is
given by Eq. (93) and is
-Jw = (PWk,)Ctu (430)
Now let us consider the gas phase and the surface reactions simultane-
ously taking place on the wall at the rates given by Eqs. (429) and
(430), respectively. T h e diffusion rate of atoms to the surface must then
be equal to the total rate of recombination of atoms by both the gas
phase recombination and the catalytic surface recombination. Hence
at the wall
Now the frozen species conservation equation and the above wall
boundary condition are transformed in terms of 7, and a self-similar
solution of the equation is obtained at the stagnation region in the manner
described in Section IV, E, 1. T h e solution gives the surface concentra-
tion of atoms as
CHEMICALLY
BOUNDARY
where
SC213 c,l
tun =
0.47 d(1+ )(Pepe)oSl k,
con is the equivalent surface Damkohler number for the gas phasereaction.
T h e remaining task for the stagnation region is to determine l g n .
A good correlation of m, for the stagnation region is available in terms of
I, defined by Eq. (427) when the wall is noncatalytic. T h e equivalent
surface Damkohler number is determined in terms of the gas phase
Damkohler number I, by matching the values of m, given by Eq. (432),
for r, = 0, to the correlated numerical results of m, given in Ref. (75).
Thus, we obtain
5gn = 21rg (434)
When all cpi are considered to be equal and constant, it can be shown
that
where the subscripts f and E denote the frozen gas phase and surface
reactions, and the equilibrium gas phase or surface reaction, respectively.
From Eqs. (432), (434), and (435), we derive
which gives the chemical state of the boundary layer at the stagnation
region in terms of the heat transfer ratio for any combination of actual
gas phase and surface Damkohler numbers. Equation (436) is plotted
in Fig. 32. Notice the close agreement between the values obtained by
Eq. (436) and those obtained from Ref. (75) for F, = 0.
Now we shall extend the closed form relation of Eq. (436) around a
blunt body in the following simple manner. We assume that the relation
given by Eq. (436) holds at each point of the body provided that the
gas phase and the surface Damkohler numbers respectively are modified
in accordance with the variations of the local Damkohler numbers along
the surface of the body. I n the case of the gas phase Damkohler number,
the variation along the surface is important because the predominant
reaction zone is near the wall.
P491
PAULM. CHUNG
rg
FIG. 32. Results of Eq. (436).
It is seen from Eq. (409) that the gas phase Damkohler number along
the surface varies as
(437)
from which we obtain
where subscript 0 denotes the stagnation point.

The variation of the surface Damkohler number is readily obtained
from Eq. (241) as
With the aid of Eqs. (438) and (439), Eq. (436) is modified for all X as
CHEMICALLY
BOUNDARY
Typical results obtained by Eq. (440) are compared with the results
obtained from the integral method for I', = 0 in Fig. 30. It is seen that
the results from Eq. (440) agree with those of the integral method very
well around the nose but the agreement is not as good along the after-
body.
Because the only existing solutions with which the present results can
be compared are for r,
= 0,27the accuracies of Eqs. (436) and (440)
cannot be checked for other finite values of r, . I n the light of the rea-
soning employed in the analysis, however, it is estimated that the present
solutions when r,# 0 should be as accurate as when r, = 0. T h e
numerical solutions with which the correlation of Eq. (427) was obtained
were based on w = 1.5. Within the same degrees of approximation
involved in the general analysis, however, the present closed form
relations should be applicable to the values of 1 & w & 2.
Finally, it is mentioned here that the present results and the results
of Refs (70) and (74) agree with each other quite closely.
FLOWS
G. NEAR-EQUILIBRIUM
We have seen in the present chapter that usually, only the full scale
numerical methods can give solutions of governing equations for the
complete range between frozen and equilibrium states, Often, however,
the most important chemical information could be had by studying only
the near-frozen limit, by a perturbation technique, as was the case with
the ignition problem of premixed reactants analyzed in Section E, 3.
Moreover, in the near-frozen regime, the perturbation technique gives
a much better insight into the behavior of initial relaxation than the
purely numerical results do (see for instance, Section F, 1). For analogous
reasons, the near-equilibrium solution is valuable in many of the reacting
boundary layer problems. Also, with both the near-frozen and near-
equilibrium solutions available, one can establish the Damkohler number
range through which a boundary layer undergoes the complete transition
from frozen to equilibrium states-usually a very important piece of
information.
More often than not, the limiting equilibrium flows create certain
singularities within the boundary layer. I n order to analyze the approach
to equilibrium state, therefore, one must perturb about the singularity
and the regular perturbation method such as was employed for the
near-frozen cases is no longer sufficient. As we shall see subsequently,
2 7 The results of Ref. (25) cannot be used for the comparison for it included, among
others, the thermal diffusion.

~511
PAULM. CHUNG
the singular perturbation technique, which is more commonly known
as the method of inner and outer expansions, is well suited for the
study of near-equilibrium regime when the limiting equilibrium state
leads to a singularity.
We shall, in the following, describe typical singularities which arise
in equilibrium boundary layer flows. We shall then analyze a particular,
singular perturbation problem associated with the combustion of
initially unmixed reactants.
1 . Typical Singularities Associated with Equilibrium Flows

One of the common singularities in equilibrium flows arises at the
gas-solid interface where the chemical kinetics governing the hetero-
geneous reaction is different from that controlling the homogeneous
gasphase reaction. These types of singularities can perhaps be best
described by the following sample case.
Let us reconsider the nonequilibrium flow over noncatalytic surface
with dissociation and recombination studed in section V, F. Out of
several choices we have,28 we choose the total energy equation, such as
Eq. (107), as the governing energy equation of the problem. Then, the
conservation equation for atomic species, Eq. (366) is the only equation
which contains the chemical reaction term explicitly. Let us, hence,
focus our attention to Eq. (366) with the boundary conditions m(co) = 1
and (amjaq), = 0 for noncatalytic surface.
For the large values of gas phase Damkohler number, which are of
our present interest, we define a small parameter c, as
6, = l/L(X) (44 1 )
Eq. (366) is then rewritten here as
Now;we expand the dependent variables in the following regular fashion,
28 See the sentences following Eq. (38).

~2521
CHEMICALLY
BOUNDARY
When the above series are substituted into Eq. (442) and other governing
equations, and when the terms with like powers of E , are collected,
there results a set of perturbed equations out of which we write for
Eq. (442)
-AEo/(RTeBo)] . C,mo) = 0 (444)
(445)LB
with the boundary conditions,
m6(0) = mi(0) = .** =0
(446)
mo(co) = 1, ml(co) = -.. = 0
In the limit of E , = 0, Eq. (444)with the help of the corresponding zeroth

order perturbed energy equation shows that the local gasphase equi-
librium will prevail everywhere throughout the flow field30-that is
everywhere except at the gas-solid interface. At the interface, the
heterogeneous chemical kinetics represented by the surface boundary
condition demands that mo(0) must vanish. Hence, as r ] -+ 0, m, must
change discontinuously from that determined by Eq. (444) to zero
creating a singularity in mO at the surface. T h e first layer of the gases
in contact with the surface, therefore, is not in equilibrium state and
does not satisfy Eq. (444). This layer, however, is infinitesimal in
thickness and, hence, the entire flow field may be said to be in equilibrium
for all practical purposes when E, = 0.
As E , becomes finite, though still much less than one, our physical
intuition tells us that the first effect of the nonvanishing E , is to increase
the thickness of the nonequilibrium gas layer adjacent to the wall.
Indeed, such is the case, and Hirschfelder (87) was the first to show
the existence of such a nonequilibrium sublayer through a numerical
integration of the governing equations for reacting static gas mixtures
contained between two parallel plates.
Since the nonequilibrium sublayer is created by the surface chemical
kinetics, only those equations which are capable of satisfying the surface
boundary condition can correctly describe the layer. As we see, Eqs. (444),
?@ w is assumed to be 2.
~ ~ the discussions in Section V, C.

3 0 A l see
PAULM. CHUNG
(445), and all the subsequent equations generated by the regular per-
turbation are algebraic and, hence, are incapable of describing the
nonequilibrium sublayer though they are applicable to the gases outside
of the sublayer. Therefore, the regular perturbation cannot be satis-
factorily used to study the near-equilibrium flows.
Having discussed the shortcomings of the regular perturbed equations,
now we may return to the original equation (442), and see, from a
slightly different viewpoint, why the regular perturbation technique
should fail. When one seeks the solution of Eq. (442) in the series
form of Eqs. (443), an implicit assumption is being made that the
coefficient of E , in Eq. (442) is of same order as the terms in the right
hand side of the equation for all 7. However, limca+o,,,+o m,, -+ 00 as
was discussed in the preceding paragraphs. Therefore, the order relation
implicitly assumed in the regular perturbation fails and hence the
method fails near the wall. T h e natural remedy of the difficulty is then
to postulate a different order relation near the wall consistent with the
limiting singularity and match this (inner) solution to the (outer)
solution obtained in the form of Eqs. (443). An actual application of the
inner and outer expansion technique to a singular, near-equilibrium
problem will be illustrated in the next section.
When AEOjR is very large compared to T , , as is the case with dis-
sociated air over a highly cooled wall, the regular perturbed equations
(444) and (445), show that m, and m, are vanishingly small within a
rather thick gas layer near the wall where the temperature is below
about 2000K. Under such special case wherein no surface reaction
takes place because of the absence of the reactant at the surface, the
surface boundary condition is trivially but automatically satisfied. No
singularity then exists for E , = 0 and, hence, the regular perturbation
describes the near-equilibrium state completely.
Let us now consider another singularity associated with equilibrium
flow. I n Section V, D, 3, we discussed the development of a diffusion
flame sheet. I n the combustion of initially unmixed reactants, fuel and
oxygen, the reaction zone becomes thinner as the Damkohler number
increases and, in the limit of 5, + co, the reaction zone becomes a
sheet. Across this sheet, discontinuities in the first derivatives of the
dependent variables exist and, hence, the flame sheet is a singular point
much like the gas-solid interface discussed previously. When &, is
reduced to a finite quantity, though it is still much greater than one,
the first order effect of the reduction appears as the broadening of the
flame sheet into a zone. Therefore, in order to study the approach to
equilibrium for the initially unmixed reactants, one must analyze this
flame zone broadening by -a singular perturbation technique. Also, as
[W
CHEMICALLY BOUNDARY
we shall see subsequently, there exists another cause of flame zone

broadening at extremely high temperatures which too can be analyzed
by the same technique. We shall study these problems in the following
section.
2. Diffusion-Flame Zone Broadening by the Method of

Singular Perturbation
We have discussed in the preceding sections the general nature of
singularities associated with equilibrium flows and suggested the singular
perturbation technique for the analysis of near-equilibrium regime. We
shall herein analyze the equilibrium and near-equilibrium regimes in the
combustion of initially unmixed reactants for a simple flow geometry.
Specifically, we consider the problems analyzed by Fendell (88), and
Fendell and Chung (89).
Let us consider two axisymmetric incompressible streams, fuel and
pure oxygen respectively, of uniform and equal velocities inpinging
against each other normally. Since the two free streams are of equal
velocity and irrotational, the vorticity is zero everywhere in the flow
field and, therefore, the flow is inviscid. We let the stagnation point of
the two streams be x = y = 0 where x is the radial distance measured
normal to the direction of the free stream and y , which is normal to x,
is taken to be positive toward the oxidant side. T h e well-known solution
of the flow near the stagnation point is then (See Ref. 90).
u = ax v = -2ay (447a)
where u and ZI are the x and y components of the velocity respectively.
T h e parameter a is the reciprocal of characteristic residence time of the
flow and it can be thought of as giving the asymptotic strength of the
streams. We consider the same one-step combustion process given by
Eq. (320) except we also include the possible reverse reaction with the
specific rate coefficient of k, . T h e reverse reaction becomes important
at temperatures above about 3000K for many hydrocarbons. Following
Fendell (88), we define the nondimensional variables as
I.' = (&3/&l)cl
F = (A,/A2)C2
T = (fi,/A,)(C,T/dh,O)
z = 1/2a/Dy (44713)
T h e subscripts I , 2, and 3 denote the oxidant, fuel, and the combustion
product respectively as before. Ah,'' is the heat of combustion per unit
W I
PAULM. CHUNG
mass of oxygen. T h e energy equation (1 5 ) , and the species conservation

equations (16), for oxidant and fuel respectively become with the aid
of Eqs. (447)
"i".)]
fi3
(448)
where
We shall assume the Damkohler number, Cgk, and the equilibrium

constant, K k ,to be constant in the present analysis. T h e fact that k,
and k, are functions of temperature does not alter the basic nature of
the analyses following. After adding the equations for Y and f', and
F and respectively in Eqs. (448), we rewrite the governing equations
as
+
d2(F p) + + =o
d(F If')
(452)
dz2 dz
T h e boundary conditions are as follows.
Forx-+oo - . .
Y = ll?f3/M1, F = 0, T = T, (454)
For z-+ -a
F =fi3/M2,
. . A
Y = 0, T = T-, (455)
Before we begin the actual analysis, let us briefly discuss the type
of solutions we wish to obtain in the light of the physical problem at
hand. When Kk = 0, the equilibrium limit (&,k --+ co) will lead to the
thin flame-sheet solution, similar to that obtained in Section V, D, 3,
wherein discontinuities in the first derivatives of the dependent variables
exist at the sheet. As {gk is reduced to a finite value, though still much
~561
greater than one, the first-order perturbation solution should show the
broadening of the flame-sheet to a thin flame zone and, at the same time,
erase the discontinuities and the singularities. This near equilibrium
solution when K, = 0 is then the first result we seek.
For the next case, consider that Cgk remains infinitely large but K ,
is increased to a finite quantity though it is still much less than one.
<
T h e nonvanishing K k 1 will, like the finite l g k ,broaden the flame-
sheet to a flame-zone, by preventing the complete disappearance of
both the oxidant and fuel in the combustion zone, and erase the dis-
continuities and singularities. However, the structure of the flame zone
thus produced will be quite different from that produced by finite l g k .
T h e perturbation solution obtained for Cgk -+00 and small but finite
Kk will tell us the manner in which the diffusion flame begins to deviate
from the thin flame-sheet at high temperatures.
T h e above two problems will be analyzed in the remainder of this
section by the method of singular perturbation (inner and outer expan-
sions). For a detailed discussion of the general method of singular
perturbation, the reader is referred to Refs. 91 and 92.
a . K k = 0, (l/Cgk) = cb <
1. We first define a small, inner region
(i) about the singularity, and the upstream ( u ) and downstream ( d )
OUTER (u)
\INNER (i)
- TO z =a
TO 2 s - m -
2-0
FIG.33. Sketch of first-order flame zone broadening.

C2571
PAULM, CHUNG
outer regions as shown in Fig. 33. We then let the dependent variables
in the outer regions be expanded into the following series forms.
yu(z, b b ) = yu.O(z) f EbYu.l(Z) + Eb2YU.Z(z) + '" (456a)

Fu(z,Eb) = Fu,O(z) +
ebFu.l(z) +
Eb2Fu,2(z) +"' (456b)
'u(2~ cb) = 'uu.O(z) + cb'u,l<z) + Eb2'U,Z(z) + '" (456c)
and similarly
.........
In formulating the inner expansion, we must keep in mind the following
requirements in the light of the discussions given hitherto in the present
Section G.
( 1) Since the singularity itself is associated with the inner region
going to zero, the independent variable, as well as the dependent
variables, should be able to stretch or contract with eb to show the
correct limiting behavior.
(2) T h e perturbed governing equations generated should retain the
second derivative, which has the singularity in the limit, and should
be well behaved so that the singularity can be erased when q, > 0.
(3) Besides the second derivative, the perturbed equation should
retain other physically important terms such as the chemical reaction
term in the present problem. T h e retention of the reaction term is
necessary because the first nonequilibrium effect should appear near
the singularity (in the inner region).
With these in mind, we expand about z* in the following manner
where (*) denotes the value for the limiting flame-sheet solution.
(4%)
where
(459)3'
31 The exponent of <b may be made different from n to give another degree of freedom
when needed.
CHEMICALLY REACTING BOUNDARY LAYERS
NONEQUILIBRIUM
T h e series, Eqs. (458), do not have zeroth order terms because the
inner region should disappear and thus the outer solution should become
valid all the way up to the singular point when q, = 0. T h e exponent n
will be determined subsequently.
We shall now generate appropriate perturbed equations for the outer
and inner regions and integrate them. T h e substitution of series (456)
and (457), into the governing equations (451)-(453), generates a set of
perturbed equations for the outer regions. T h e zeroth order perturbed
equations can be integrated readily and they become, after satisfying
the appropriate boundary conditions at z-+ co and z+ -CO for the
1
upstream and downstream solutions respectively,
+ il.,
Y ~ . ~ = A erfc(zid2) + [Fa + (fi3/fi1)] (460a)
Fu,o + Fu,o = -Cerfc(z/dj) + pa (460b)
1
Fu,o =0 (460c)
+ erf(z/dZ)]
o = E [ I i-
Y ~ ,p d , ~ + T-m (461a)
Fd,O + pd,o = D[I + erf(z/d?)] + + @13/fi2)] (461b)
Yd.0 =0 (46 1c )
where A , C, D, and E are the remaining constants of integration to be

determined by matching the solutions to the inner solution.
For the inner region, since the singular behavior of the problem
appears more explicitly through Eq. (453)32,we first employ this equation
to determine the exponent n. A substitution of Eqs. (458) into Eq. (453)
with the aid of Eq. (459) results in the following equation
= -k?-51f1 + E?-'(Ylfi + Y2fl) + *..I (462)

In view of the requirements set forth earlier for the inner expansion,
we would retain the chemical reaction term on the right hand side of
Eq. (462) rather than the convection terms (the terms with first deri-
vatives) in the first-order perturbation equations. Hence, we let
3 (463)
3* The functions ( Y + +
f)and ( F ?) are regular throughout the flow field. However,
such terms as 8Y/az2 and a2F/azz implicitly contained in Eqs. (451) and (452) have
singularities at z = z* in the limiting case.
[2591
PAULM. CHUNG
Eq. (462) then gives
With n = 9,the first order perturbation equations generated by sub-

stituting Eqs. (458) into Eqs. (451) and (452) can be integrated readily
with the result,
y1 + t, = +8
af (465a)
fi + t , = yf + 6 (465b)
Let us now consider the matching between the outer and the inner
regions. T h e matching should determine the constants of integration
for the inner and the outer solutions when they are known, and the
boundary conditions for yet unsolved equations such as Eq. (464). T h e
matching is crucial for the success of the method as was pointed out
in Ref. 92. T h e ultimate aim in the matching is that the outer expansion
as it is being brought into the inner region, and expressed in terms of
inner variables, should be identical to the inner expansion approaching
the outer region, and vice versa. If this is accomplished, it is seen that
there will be a region between the inner and outer regions wherein
both the inner and outer solutions are valid and hence the two solutions
over lap. There are some general rules which can be followed to accom-
plish this matching (see, for instance, Ref. 92). However, none of these
rules is always applicable, and one must use personal judgment and
reasoning in solving a particular problem.
Let us first match the function ( Y + f') between the upstream outer
region and the inner region. T h e outer and inner expansions for this
function become with the help of Eqs. (456a), (456c), (458a), (458c),
(460a), and (465a),
Now we wish to match such that, for each order, Eq. (466) as z -+ z*
[2601
CHEMICALLY BOUNDARY
LAYERS
will be identical to Eq. (467) as 2- 00. For x - + z * , each term of

Eq. (466) can be expanded with the result,
Y J z ,E + Tu(z,e b ) = {-A[]
~ ) - erf(z*/d/2)
- d/zi;;(z - z * ) exp(-(z*/d/Z>;) - ...I + Tm

+ Cfi3/fid> + Eb{[YU,l(Z*) + Tu,1(z*)l
+ (z - z * ) p
az
+ a f U J
I=Z*
+ (468)
Above equation is rewritten in terms of the inner variable as
Yzt(2,E + Fu(f,
~ ) E,,) = {-A[I - erf(z*/d/Z) - d2/aexp(-(z*/1/2)2)
. q;/32+ o ( ~ ; +
/ ~-1) + fa + (iiiyfi1)]
+ Eb{EYU,l(~*) + TU,,(~*)l+ ...I (469)
Now matching Eqs. (467) and (469) to O ( C ; / ~
gives
)
1
A=
erfc(z*/.\/2)
[ - f* + Tm (fi3/fi1)]
5 + (470)
(a2 + 8) = Ad/2/lrexp[-(z*/d/2)] .f (471)

Eq. (471) shows that p cannot be a constant unless B = 0. Of course,
fl could have been included in Eq. (470), instead of Eq. (471), by
assuming it to be O ( E ; ~ / ~
However,
). this violates Eq. (465a) where
each term is of O(1). Therefore, the only permissible value of is p = 0.
a is then obtained from Eq. (471) as
a = Ad/2/.rrexp[--(z*/~/2)~] (472)
With the aid of Eqs. (460b) and (460c), which define p* in the absence
of finite inner region, Eq. (470) becomes
A = [pa - f-m + (fiJfil)]/2 (473)
+
( Y , pd)for the downstream outer region is next matched to Eq. (467)
with the aid of Eqs. (472) and (473), in the same manner. T h e similar
matching is also carried out for the function ( F T).Through these +
matching processes, all the remaining constants of integration are
determined as
E=A
y = ~.\/2/.rrexp[-(z*/1/2)~1
C =D = [Fm- f P m - (k3/fi2)]/2 (474)
and 6 = 0 for the reasons analogous to that made B = 0.
P611
PAULM. CHUNG
Finally, the boundary conditions for Eq. (464) are obtained through
the following matching employing the yet unused outer solutions,
Eqs. (460c) and (461~). T h e outer expansion, Eq. (456b), gives for
z 3 z* in terms of the inner variable,
(475)
T h e inner expansion, Eq. (458b), becomes for 2 -+ co,
Since Fu,ois zero for all z >, z* according to Eq. (460c), F, = 0 up to

O(eb).Matching of Eqs. (475) and (476), therefore, gives
fi(2) = 0 as 0 -+ co
Eq. (465b) gives with the aid of the above equation,
T h e similar matching for the downstream results in,
tl+& as 2-+- w (477)

Eq. (464) is integrated with the above two boundary conditions in
Ref. 88 by the use of a digital computer. With tl(2)obtained, the solution
of the problem is now complete.
The near-equilibrium reactant concentration and temperature profiles
can be obtained by the use of Eqs. (460) and (461) for the outer regions,
and Eqs. (458) and (465) for the inner region, respectively. T h e quali-
tative sketch of the temperature profile is shown in Fig. 33.
Thus far, we have assumed that in the limit of eb = 0 the singular
perturbation will naturally give the thin flame-sheet solution hence, for
instance, will give the value of rf* at z* obtainable from the flame-sheet
solution. As we have seen in Section V, D, 2, the flame-sheet solution
was obtained independently of any limiting processes by assigning a
set of boundary conditions, Eqs. (341), at the flame sheet. These
boundary conditions are unique in that they are never employed at any
other Damkohler numbers, including the present case, than for co.
We shall now show that the p* obtained from the present analysis by
letting q,-+ 0 is indeed the same rf* obtainable from the thin flame-
sheet analysis such as that given in Section V, D, 2.
By employing Eqs. (451) through (453), and by employing the thin
w21
CHEMICALLY
BOUNDARYLAYERS
flame-sheet method similar to that given in Section V, D, 2, b, the

following expressions were obtained in Ref. 89 for z* and rf*.
T h e limiting expressions for z* and rf* as eb -+ 0 for the present analysis

can be derived readily by manipulating with Eqs. (460) and (461). One
finds that the resulting expressions are identical to those given by Eqs.
(478) and (479). Thus we have now shown that the present perturbation
approach has a unique limit given by the classical thin flame-sheet solu-
tion. I n closing it is pointed out here that Linan (93)also made a somewhat
similar study of the flame structures in the near-equilibrium regime.
<
b. luk -+ CO, and Kk = cC 1. This is a slightly simplified version
of the problem analyzed by Fendell and Chung (89). For .+ 00,
Eq. (453) can be replaced by,
T h e governing equations are then Eqs. (451), (452), and (480).

As we have done for the case considered in the preceding subsection a,
we expand the dependent variables for the two outer and inner regions
into the series given by Eqs. (456), (457), and (458) respectively except
q, is replaced by cC. Also, we define the independent variable for the
inner region given by Eq. (459) with n still to be determined. T h e zeroth
order perturbation equations and their solutions for the outer regions
are the same as those for the case considered in Subsection a. T h e
zeroth order outer solutions are hence given by Eqs. (460) and (461).
For the inner region, as was done previously, we first substitute the
inner expansions, Eqs. (458), into Eq. (480) and determine n. T h e sub-
stitution gives the following equation:
Ylfl + % Y Y l f i +Yzfd + ..'
Physically, as E , becomes finite though still much less than one, the
reverse reaction should first appear in the inner region and broaden
~ 3 1
PAULM. CHUNC
the flame-sheet to a flame zone. This means that, for finite E~ , Eq. (480)
for the inner region should contain a term from the parenthesized
4
portion of that equation. We, hence, let n = in Eq. (481). Then, the
first-order perturbed equation for Eq. (480) becomes
Ylfl = 1 (482)
A substitution of Eqs. (458), with n = $, into Eqs. (451) and (452)
generates a set of perturbed equations. T h e first-order equations are the
same as those for the subsection a and, hence, Eqs. (465) represent the
first order inner solutions for the functions ( Y +
p) and ( F p ) . +
We have now seen that both the zeroth order outer and the first
order inner solutions of functions ( Y + +
F ) and ( F F ) for the present
problem are respectively equal to those for the problem analyzed in
the preceding subsection. Therefore, Eqs. (460a), (465a), and (461a),
and Eqs. (460b), (465b), and (461b) are respectively matched exactly in
the manner described in Subsection a, and there results the constants
of integration given by Eqs. (472), (473), and (474). Also /3 = 6 = 0.
T h e only remaining problem is then to show that Yiand Fimatch to
the respective outer solutions without violating Eq. (482). With the use
of Eq. (460c), the outer expansion, Eq. (456b), becomes in terms of the
inner variable, as z+z*
F,(O,,) = cFu,l(z*) +O(p)
Matching this equation with Eq. (458b) or 2 4 co gives,
fl(2) -+0 as 2 3 co
Eq. (456a) becomes in terms of the inner variable, as z --+ z*
Y U ( k.,) = Y,,o(z*) + 1,~z2(aY,.o~az)z*

+ O(,)
where
(aYu,o/ax),* = ( A - C)d&xp[-(z*/dZ)2]
from Eqs. (460). Matching Eq. (484) with Eq. (458a) gives
YU.O(~*) =0
yl(a) -+ ( A - ~ ) d ~ e x p [ - ( z * / l / Z ) * ]. f
Now, when Eq. (486) is substituted into Eq. (482), there results, as
i-r co.
which is consistent with the matching requirement for F given by

Eq. (483). We have now shown that F and Y respectively match between
the upstream outer and the inner regions, and that the matching is
+ +
consistent with Eq. (482).33Since ( Y F ) and ( F T)have been already
respectively matched between the two regions, is also matched and
the matching between the regions ( u ) and (i)are complete. T h e matching
can be similarly completed between the regions ( d ) and (i).
The algebraic equations, Eqs. (465a), (465b), and (482) can be used
to solve for tl(2). T h e temperature in the inner region is then obtained
from Eq. (458c) to O(eC)as,
,I
1 i ( z , E,) = T* + $ { ( a + y ) ( z - z*) - [(ci - y ) ' ( ~ - z*)' + ~EJ'I'} (487)
At z = z*, this becomes
which shows the reduction of the temperature at z* where the flame-

sheet existed. When Eq. (487) is differentiated and set equal to zero,
we find that the position of the maximum temperature is no longer at
z* but is shifted to a new position zw given by
From the Eq. (489), it can be deduced that the maximum temperature
is displaced toward the oxidant source when [(l@JJ?J- (J?3/l@2) +
2( T, - PTm)]> 0 and toward the fuel when the expression is negative
as eC becomes finite. When it is zero, the position of the maximum
temperature remains at x*.
There are few other methods which have been devised to predict the
structure of the flame zone when the reverse reaction is not negligible.
One of these is the method advanced by Libby and Economos (94) in
which an ignition temperature is, a priori, chosen and assumption is
made that the gas is in equilibrium state in the flow region where the
temperature is greater than the ignition temperature whereas it is frozen
where the temperature is below the ignition temperature.
33 Strictly speaking, this is not a sufficient proof of matching for Eq. (482). A rigorous
proof is given in Ref. (89).

PAULM. CHUNG
VI. Concluding Remarks
I n the present discussion of the chemically reacting boundary layers

we have tried to present analyses which are rather general in nature such
that they may also be useful in studying other problems of nonequilib-
rium boundary layers than those considered herein. T h e emphasis was
placed on the basic behaviors and solutions of the boundary layer prob-
lems rather than on the particular engineering solutions. No attempt was
made either to cover all the existing solutions or to consider the detailed
aspects of the various engineering applications.
We have not included the analyses of Couette and Rayleigh problems
which preceded those of the steady-state boundary layer problems
because these have been included by Moore (6) in his discussion. For the
same reason, we have slighted the effect of variable (pp) and evaluations
of other thermodynamic and transport properties. We only mentioned in
Section V, D, 1, e that one may use 1 = [ ( p , ~ , , l p , p ~ ) ~ ]in
~ ~a~ boundary
layer solution obtained with the assumption of constant 1 if pe/pw < I,
ue2/2cpTe O( I), and if the boundary layer comprises gases whose proper-
ties are not much different from the properties of air. I n the case of
slender bodies wherein u,2/2cpT, >1, it seems that one may still employ
the above approximate value of I provided that (pepe)be replaced by
(p*p*), as was suggested by Clarke (77), where the asterisk (*) denotes
the maximum dissociation zone.
We have mentioned very little about the experimental verification of
the various analyses presented. Actually, some of the cited references
refer to the pertinent experimental results. T h e experimental counterpart
of most of the nonequilibrium theories, however, is not available to
this date. I n view of much experimental activities in various laboratories,
more experimental results should be forthcoming.
I n handling the diffusion phenomena only a slight deviation from the
assumption of binary mixture was allowed (Section V, F, 4). Rigorous
handling of the nonequilibrium reacting multicomponent boundary
layers requires substantially geater effort (mainly numerically) than that
already involved with the binary boundary layers. Solutions of the
equations for actual multicomponent boundary layers of air are presently
being attempted by the methods described in Section V, F, 3.
T h e general field of nonequilibrium viscous flows is presently ex-
panding into the wakes and viscous shock layers. Some of the typical
analyses on nonequilibrium wakes are given by Bloom and Steiger
(78, 79),Blottner (80) ,and Pallone (82),whereas those on viscous shock
layer are found in Cheng (82), Gibson (83), Scala (84), Stoddard (85),
12667
CHEMICALLY
BOUNDARYLAYERS
and Chung (86). T h e basic features of the nonequilibrium boundary

layers discussed however, are preserved through the reacting wakes
and shock layers.
NOMENCLATURE
(All the symbols have been defined in the text, only the most important ones are
listed here. Definitions in the text supersede those given here.)
C Mass fraction of reactant S Variable defined by Eq. (203)
C i Mass fraction of ith species S Variable defined by Eq. (99)
AE" Heat of reaction per mole T Absolute temperature
c, Constant pressure specific heat t Variable defined by Eq. (204)
E, Activation energy U ulu, , defined by Eq. (205)
F ( X ) Function defined by Eq. (121) U x-component of velocity
Dimensionless stream function de- V Variable defined by Eq. (205)
fined by Eq. (129) V y-component of velocity
Dimensionless stream function de- W Mass rate of production of reactant
fined by Eq. (102) per unit volume by gas phase
htlht, reaction
htilht,, X XlL
Enthalpy defined by Eq. (6) X Streamwise distance (see Fig. 1)
TotaI enthalpy defined by Eq. (20) Y Normal distance (see Fig. I )
Frozen total enthalpy defined by B du,/dx at stagnation point
Eq. (28)
Heat of reaction per unit mass
Surface production rate
Equilibrium constant for hypersonic flow
Forward, specific reaction-rate r Reference Damkohler number
coefficient 6 Boundary layer thickness
Backward, specific reaction-rate c
I = 0 for two dimensional body
coefficient = 1 for axisymetric body
Specific recombination rate coeffi- r Surface Damkohler number defi-
cient ned by Eq. (140)
Constant defined by Eq. (289) ri Surface Damkohler number of ith
Characteristic length (see Fig. I ) species defined by Eq. (120)
(PdI(PeP*) r, Gas phase Damkohler number
Molecular weight of gas mixture r,i Gas phase Damkohler number of
CICe ith species defined by Eq. (281)
CrlCie 7) Similarity variable defined by
nth perturbation of m Eq. (129)
reaction order, or li Similarity variable defined by
positive integer when appears with Eq. (98)
summation sign e TIT, or angle defined in Fig. I
Prandtl number of gas mixture h Thermal conductivity of gas mix-
Pressure ture
Total heat transfer to solid interior P Dynamic viscosity of gas mixture
Universal gas constant Y Kinematic viscosity of gas mixture
Schmidt number of gas mixture 5 Variable defined by Eq. (185)
PAULM. CHUNG
p Density of gas mixture e Boundary layer edge
4 Stream function defined by Eqs. 0 Stagnation point or leading edge
(100) and (101) w Wall
w Exponent defined by Eq. (289) 1 Reactant or atoms
2 Reaction product or molecules
Subscripts co Free stream
E Equilibrium
REFERENCES
1. L. Lees, Third AGARD Combustion and Propulsion Panel Colloquium, p. 451.

Pergamon, New York, 1959.
2. D. B. Spalding, Aeronautical Research Council, (Great Britain) 22776, R557,
C.F. 552; ASTIA AD 281702 (1961).
3. W. H. Dorrance, Viscous Hypersonic Flow. McGraw-Hill, New York, 1962.
4. D. E. Rosner, Aeronautical Research Laboratory Report 99, Part I, Wright-Patterson
Air Force Base, Ohio, 196).
5. D. E. Rosner, Aeronautical Research Laboratory Report 99, Part 11, Wright-Patterson
Air Force Base, Ohio, 1961.
6. F. K. Moore, in Theory of Laminar Flows (F. K. Moore, ed.), Vol. IV, p. 439.
Princeton University Press, Princeton, New Jersey, 1964.
7. D. E. Rosner, AIAA Preprint 63114 (1963).
8. J. 0. Hirschfelder, C. F. Curtiss, and R. B. Bird, Molecular Theory of Gases and
Liquids. Wiley, New York, 1954.
9. S.S. Penner, Chemical Reactions in Flow Systems. Butterworths, London, 1955.
10. C. F. Curtiss and J. 0. Hirschfelder, J . Chem. Phys. 17, 550 (1949).
11. A. A. Frost and R. G. Pearson, Kinetics and Mechanism. Wiley, New York, 1961.
12. D. B. Spalding, Some Fundamentals of Combustion. Butterworths, London, 1955.
13. A. H. Wilson, !Theirnodynamics and Statistical Mechanics. Cambridge Univ.
Press, London and New York, 1960.
14. C. F. Hansen, NASA TR-50 (1959).
15. K. Wray, J. D. Teare, B. Kivel, andP. Hammerling, AVCO Res. Lab. Res. Rept. 83
(1959).
16. J. C . Camm, B. Kivel, R. L. Taylor, and J. D. Teare, AVCO Res. Lab. Res. Rept.
93 (1959).
17. S. C. Lin and J. D. Teare, AVCO Res. Lab. Res. Rept. 115 (1962).
18. M. J. Lighthill, J . Fluid Mech. 2, 1 (1957).
19. P. M. Chung and A. D. Anderson, Proc. Heat Transfer Fluid Mech. Inst. 1 9 6 0 , ~ 150.
.
Stanford Univ. Press, Stanford, California.
20. R. J. Goulard, Jet Propulsion 28, 737 (1958).
21. S. M. Scala, IAS Paper No. 62-154 (1962).
22. P. L. Walker, F. Rusinko, and L. G. Austin, Advan. Catalysis Vol. 11, p. 133 (1959).
Academic Press, New York.
23. L. Lees, Jet Propulsion 26, 259 (1956).
24. J. A. Fay and F. R. Riddell, J . Aeron. Sci. 25, 73 (1958).
25. S.M. Scala, Proc. 3rd U . S. Nail. Congr. Appl. Mech. p. 799 (1958). Published by
ASME, New York.
26. P. M. Chung, NASA T N D-27 (1959).
27. W. E. Welsh, Jr. and P. M. Chung, Aerospace Corp.,TDR-I69(3230-12)TN-6 (1963).
P681
NONEQUILIBRIUM
REACTING
28. W. E. Welsh, Jr. and P. M. Chung, Proc. Heat Transfer Fluid Mech. Inst. 1963
p. 146 (1963). Stanford Univ. Press, Stanford, California.
29. N. S. Diaconis, P. D. Garsuch, and R. A. Sheridan, IAS Paper No. 62-155 (1962).
30. W. E. Welsh, Jr., Aerospace Corp., SSD-TDR-63-193 (1963).
31. P. M. Chung, and A. D. Anderson, NASA T N D-350 (1961).
32. D. E. Rosner, ARS J. ( A m . Rocket Soc.) 32, 1065 (1962).
33. R. A. Hartunian and W. P. Thompson, AIAA Preprint No. 63-464 (1963).
34. E. M. Winkler and R. N. Griffin, NASA T N D-1146 (1961).
35. P. L. Chambre and A. Acrivos, J . Appl. Phys. 27, 1322 (1956).
36. G. R. Inger, Intern. J . Heat Mass Transfer 6, 815 (1963).
37. A. N. Tifford and S . T. Chu, Proc. 2nd Midwestern Conf. Fluid Mech. Ohio State
Univ.,Colombus, Ohio, lg52 p. 363 (1952). Ohio State University Press, Colombus,
Ohio.
38. D. Chapman and M. W. Rubesin, J. Aeron. Sci. 16, 547 (1949).
39. A. Fage, and V. M. Falkner, Rept. Mem. Aeron. Res. Comm. London No. 1314,
(1931).
40. M. J. Lighthill, Proc. Roy. Sac. A202, 359 (1950).
41. M. W. Rubesin and M. Inouye, NACA TN-3969 (1957).
42. D. E. Rosner, J . Aerospace Sci. 26, 281 (1959).
43. P. L. Chambre, Appl. Sci. Res. A6, 97 (1956).
44. A. Acrivos and P. L. Chambre, Ind. Eng. Chem. 49, 1025 (1957).
45. P. M. Chung, S. W. Liu, and H. Mirels, Intern. J . Heat Mass Transfer 6, 193 (1963).
46. S. W. Liu and P. M. Chung, Aerospace Corp. SSD-TDR-63-345 (1963).
47. P. M. Chung and A. D. Anderson, A R S ( A m . Rocket Soc.) J . 30, 262 (1960).
48. P. M. Chung and A. D. Anderson, Advances in the Astronautical Sciences,
Vol. 8, p. 188. Plenum Press, New York, 1962.
49. D. E. Rosner, A . I. Ch. E. ( A m . Inst. Chem. Engrs.) J . 9, 321 (1963).
50. G. R. Inger, A I A A J . 1, 2057 (1963).
51. P. M. Chung, Phys. Fluids 6, 550 (1963).
52. R. A. Hartunian and P. V. Marrone, Cornell Aeron. Lab. Rept. AD-I 118-A-7 (1959).
53. F. G. Blottner, AIAA Preprint, No. 63-443 (1963).
54. G. W. Sutton, Symp. Combustion 7th London Oxford, 1958 p. 539 (1959). Academic
Press, New York.
55. G. W. Sutton, ARS Preprint 621-58 (1958).
56. J . P. Hartnett and E. R. G. Eckert, Heat Transfer Fluid Mech. Inst. 1958, Preprints
Papers p. 54 (1958). Stanford Univ. Press, Stanford, California.
57. F. E. Marble and T. C. Adamson, Jr., Jet Propulsion 24, 85 (1954).
58. D. A. Dooley, Ph. D. Thesis, Calif. Inst. Technical., Pasadena, California, 1956.
59. S . I. Cheng and A. A. Kovitz, Symp. Combustion 6th Yale Univ. 1956 p. 418 (1957).
Reinhold, New York.
60. S. I. Cheng and H. H. Chiu, Intern. J . Heat Mass Transfer I , 280 (1961).
61. T. Toong, Symp. Combustion 6th Yale Univ. 1956 p. 532(1957).Reinhold, New York,
62. G. R. Inger, Aerospace Corp. TDR-269(4230-10)-2 (1963).
63. W. J. Rae, IAS Paper 62-178 (1962).
64. P. M. Chung and A. D. Anderson, NASA T N D-140 (1960).
65. A. Pallone, J. Moore, and J. Erdos, AVCO, RAD TM-63-58 (1963).
66. P. A. Libby, Heat Transfer Fiuid Mech. Inst., 1958, Preprints Papers p. 216 (1958).
Stanford Univ. Press, Stanford, California.
67. G. Jarre, Torino Polytechnic Inst., Appl. Mech. Lab., T N - I 0 (1958); AFOSR
T N 38-944 (1958).
PAULM. CHUNG
68. G. Doetsch, Handbuch der Laplace Transformation, Vol. 2. Birkhauser, Basle.
1955.
69. P. M. Chung and S. W. Liu, AZAA 3. 1, 929 (1963).
70. G. R. Inger, AIAA J. 1, 1776 (1963).
71. D. E. Rosner, Aero Chem TP-54 (1962).
72. A. Linan and I. Da Riva, Chemical Nonequilibrium Effects in Hypersonic Aero-
dynamics, A paper presented at 3rd Intern. Congress in Aeronaut. Sciences, Aug.
27-31, 1962, Stockholm, Sweden.
73. P. M. Chung, unpublished.
74. G. R. Inger, unpublished.
75. G. Goodwin and P. M. Chung, Advances in Aeronautical Sciences (T. von Kirmin,
ed.), Vol. 4, p. 997.
Pergamon, New York, 1961.
76. G. R. Inger, ARS (Am. Rocket Soc.) J. 32, 1743 (1962).
77. J. F. Clarke, J. Fluid Mech. 4, 441 (1958).
78. M. H. Bloom and M. H. Steiger, IAS Paper No. 63-67 (1963).
79. M. H. Bloom and M. H. Steiger, Gen. Appl. Sci. Lab. T R 180, Westbury, L.I.,
New York.
80. F. G. Blottner, private communication, 1964.
81. A. Pallone, private communication, 1964.
82. H. K. Cheng, IAS Paper No. 63-92 (1963).
83. W. Gibson, IAS Paper No. 63-70 (1963).
84. S. M. Scala, private communication, 1964.
85. F. J. Stoddard, Proc. Heat Transfer Fluid Mech. Inst. 1963 p. 128 (1963). Stanford
Univ. Press, Stanford, California.
86. P. M. Chung, NASA TR-109 (1961).
87. J. 0. Hirschfelder, J. Chem. Phys. 26, 274 (1957).
88. F. E. Fendell, J . Fluid Mech. 21, 281 (1965).
89. F. E. Fendell and P. M. Chung, Aerospace Corp., TDR-4691S 5240-101-1, (1965).
90. H. Schlichting, Boundary Layer Theory. McGraw-Hill, New York, 1955.
91. S. Kaplun, 2. Angew. Math. Phys. 5 , 1 1 1 (1954).
92. M. Van Dyke, Perturbation Methods in Fluid Mechanics. Academic Press, New
York, 1964.
93. A. Linan, Ph. D. Thesis, California Institute Technology, Pasadena, California
(1963).
94. P. A. Libby and C. Economos, Znternl. J. Heat Mass Transfer 6 , 113 (1963).
95. N. C. Freeman and P. G. Simpkins, private communication, 1964.
Low Density Heat Transfer
. .
F M DEVIENNE
Laboratoire Mediterranien de Recherches Thermodynamiques.Nice. France
I . Introduction . . . . . ............... 212

I1. General Considerations . . . . . . . . . . . . . . . . . 214
A . Mechanism of Thermal Conduction . . . . . . . . . . 214
B. Apparent Decrease of the Heat Conduction Coefficient . . 214
C . Free-Molecule Conduction . . . . . . . . . . . . . . 215
I11. Heat Conduction of a Highly Rarefied Gas . . . . . . . . . . 276
A . Energy of a Gaseous Molecule . . . . . . . . . . . . . 216
B . Definition of the Accommodation Coefficient . . . . . . 211
C . Energy Transfer between a Wall and a Highly Rarefied Gas 219
D . Heat Conduction between Two Parallel Plates Immersed in a
Highly Rarefied Gas . . . . . . . . . . . . . . . . . 280
E . Heat Conduction between Two Coaxial Cylinders . . . . 282
F . Calculation of the Heat Exchanged between Any Two Sur-
faces Immersed in a Highly Rarefied Gas . . . . . . . . 284
G. Energy Received by.a Plane Plate Moving in a Highly Rarefied
Gas . . . . . . . . . . . . . . . . . . . . . . . . 284
H. Limit of Validity of the Preceding Formulae . . . . . . . 290
.
IV Heat Conduction in Rarefied Gases . . . . . . . . . . . . 291
A . Temperature Jump Regime . . . . . . . . . . . . . . 291
B. Heat Conduction between Two Parallel Plane Plates . . . 292
.
C Heat Conduction between Two Coaxial Cylinders . . . . 292
D . Temperature Jump Measurements . . . . . . . . . . . 293
.
F Theoretical Determination of the Temperature Jump in Rela-
tion to the Accommodation Coefficient and the Mean Free
Path of Molecules . . . . . . . . . . . . . . . . . . 296
.
V Intermediate Regime . . . . . . . . . . . . . . . . . . 300
.
A General Remarks . . . . . . . . . . . . . . . . . . . 300
B. Case of Two Parallel Plates . . . . . . . . . . . . . . 300
C . Case of Two Unspecified Surfaces . . . . . . . . . . . 303
VI . Experimental Determination of the Accommodation Coefficients 304
A. Different Methods of Measurement . . . . . . . . . . 304
B. Principle of the Method Employing the Determination of
the Heat Exchanged between Two Surfaces Heated to
Different Temperatures . . . . . . . . . . . . . . . . 304
~711
F. M. DEVIENNE
C. Contribution of the Different Modes of Heat Transfer . . . 307
D. Various Devices Employed . . . . . . . . . . . . . . 31 1
E. Measurements Obtained by Means of the Spectral Emission
by an Unequally Heated Gas . . . . . . . . . . . . . 317
F. Measurements Made by Means of Radiometric Apparatus
(Knudsens Method) . . . . . . . . . . . . . . . . . 317
G. Method Reported by Devienne . . . . , . . . . . . . . 319
H. General Results . . . . . . . . . . , . . . . . . . . 325
I. Different Factors Modifying the Values of Accommodation
Coefficients . . . . . . . . . . . . . . . . . . . . . 328
J. Influence of the Structure of the Surface and the Adsorbed
Layer . . . . . . . . . . . . . . . . . . . . . . . 329
I(. Influence of the Surface Temperature . . . . . . . . . . 329
L. Variation in Terms of the Energy Difference . . . . . . . 332
M. Influence of the Pressure. . , . . . . . . . . . . . . . 335
N. Accommodation Coefficients concerning Internal Energy. . 337
VII. Theoretical Calculation on Accommodation Coefficients . . . 339
A. General Observations . . . . . . . . . . . . . . . . . 339
B. Classical Theory Reported by Bade . . . . . . . . . . 339
C. Zener Theory . . . . . . . . . . . . . . . . . . . . 34 I
D. Theory of Jackson . . . . . . . . . . . . . . . . . . 343
E. Theory of Landau . . . . . . . . . . ... . . . . . 344
F. Theory of Devonshire . . . . . . . . . . . . . . . . 346
G. Theory of Zwanzig . . . . . . . . . . . . . . . . . . 349
H. Shortcomings of Present Theories . . . . . . . . . . . 350
VIII. Conclusion . . . . . . . . . . . . . . . . . . . . . . 350
List of Symbols . . . . . . . . . . . . . . . . . . . . . 352
References . . . . . . . . . . . . . . . . . . . . . . . 352
I. Introduction
T e n years ago research in the field of heat transfer associated with

rarefied gases involved a small number of specialists.
Recent problems encountered by long range missiles and satellites
have focused scientists attention on the problems of thermal conduction
in rarefied gases, and this topic has consequently attracted wide interest.
I n this monograph our purpose is to review the advances in heat
transfer in rarefied gases. We do not restrict our attention to research
that has been carried out during the last decade.
Rather, we thought it useful to present the whole field of thermal con-
duction in rarefied gases, stressing those results obtained recently and
noting those problems not yet solved.
I n this study, we have adopted a point of view similar to that usually
taken in rarefied gas dynamics. Thus, in the study of thermal conduction,
we distinguish four different regimes depending on the value of the ratio
~721
Low DENSITYHEATTRANSFER
of the mean free path L of the molecules and a characteristic dimension

d of the medium (for instance the distance between two parallel planes
in the case of the thermal conduction between two plates). This dimen-
sionless number is obviously the Knudsen number L l d which is so
important in rarefied gas dynamics.
These four regimes are (1) the regime of ordinary thermal conduction
in which we are not interested in the current study, L l d < 0.001;
( 2 ) the regime of free-molecule heat conduction, L / d 3 10; (3) the
regime of conduction corresponding to the temperature jump, 0.001 <
<
L / D 0.1; and (4) the one that will be called the intermediate regime
0.1 < L / d < 10.
T h e first section of this book deals with the presentation of phenomena
arising when the gas is truly rarefied. T h e second and third sections are
devoted to free-molecule conduction, i.e., conduction in highly rarefied
gases ( L / d >, lo), and conduction in rarefied gases in the regime
characterized by the temperature jump condition (0.001 < L / d < 0.1).
T h e fourth section shows how the connection between the two pre-
ceding regimes can be made by means of what we call the intermediate
regime (0.1 < L / d < lo). T h e fifth section is devoted to the experi-
mental determination of accommodation coefficients; their importance
is essential for all the regimes concerning the thermal conduction of
rarefied gases.
In the last section we will survey theoretical investigations that have
been carried out more or less recently with the object of predicting the
values of accommodation coefficients.
We will see that although the physics of thermal conductionin rare-
fied gases have been extensively studied and many results have been
obtained, further experimental and theoretical research is essential.
As a matter of fact, it appears that the available experimental and
theoretical results are not completely definitive, in the present state of
things.
We will see that the influence of certain parameters becomes pre-
ponderant when rarefaction increases. In particular the surface condition
of the solids which exchange heat with the highly rarefied gas is of
critical importance. Nevertheless, in all the research that has been
carried out so far, none specify this surface condition. We think that
nearly all the measurements must be taken up again, in order to try
and define the structure of the surface, by one means or another.
I n this study, we will generally assume that heat exchange occurs
between two solid surfaces at rest by means of conduction through a
rarefied gas. We have also calculated the exchange of energy in the case
of a single plate at rest or in motion, immersed in a rarefied gas.
WI
F. M. DEVIENNE
11. General Considerations
OF THERMAL
A. MECHANISM CONDUCTION
We know that heat conduction through a gas is the result of the
numerous collisions between gaseous molecules at every moment. T h e
molecules issuing from the hotter part of the gas carry greater energy
than the molecules issuing from the colder part. T h e collisions between
molecules have a tendency to produce an equalization of the energy
carried by the molecules. Inside the gas, a temperature gradient occurs;
it depends on the temperature and the nature of the solid surfaces
bounding the gas. I n particular, if we consider a layer of gas between
two parallel plane surfaces, the quantity of heat conducted per unit area
per second between these two plates at temperature T , and T , respec-
tively, the distance between them being d, is given by the equation
in which h is the thermal conductivity. (See Fig. 1 . j
'6 p2
FIG. 1
T h e experiment shows that this coefficient is constant at reasonable

values of the pressure. Generally speaking, we can assume it to be
constant when the mean free path of the molecules L is much smaller
than the distance between the plates d [ L / d ,< 0.0011.
B. APPARENTDECREASE
OF THE HEATCONDUCTION
COEFFICIENT
When the pressure decreases until the mean free path, though small,
compared to the distance between the two plates, is no longer negligible,
c2741
Low DENSITY
HEATTRANSFER
experiment shows that the quantity of heat conducted per unit area
per second between two identical parallel plates generally decreases
when the pressure is decreasing. Under these circumstances it can be
assumed in the case of two parallel plates that this quantity of heat is
given by
g is what we will call the temperature jump distance.

I n order to make everything clear, let us simply mention that the value
of g is generally equal to a few mean free paths. We will bring up this
subject again later. We may wonder why g has been called temperature
jump distance. I n fact, as far as heat conduction is concerned, it is as
if the temperature of the gas near the wall had not the value of the
temperature Tp of the wall, but a value Tgsuch as
where dTjdn stands for the temperature gradient normal to the wall in
the midst of the gas between the two parallel plates and away from the
wall.
Without insisting on the phenomenon, we immediately see that the
influence of the temperature jump is only to be noted when rarefaction
is such that the mean free path is not negligible when compared with the
distance between the two plates.
It appears that this influence does not absolutely depend on the
pressure, but on the ratio Lld, i.e., on the ratio of the mean free path
to the distance between the parallel plates. This ratio is nothing but the
Knudsen number Kn,a dimensionless number which is very important
in rarefied gas dynamics.
C. FREE-MOLECULE
CONDUCTION
If we inspect conditions more closely we note that the temperature
jump is due to the fact that near the wall, there are not only collisions
between gaseous molecules but impacts between the molecules of the
gas and molecules of the solid bounding it. As it will be seen further,
the energy is not completely exchanged between the molecules of the gas
and those of the solid. T h e result is that the layer of gaseous molecules
in the immediate vicinity of the solid surface has a temperature different
from the temperature of the surface,
W I
F. M. DEVIENNE
When rarefaction increases, the number of intermolecular collisions

between the gaseous molecules decreases as compared to the number
of impacts of gaseous molecules on the solid surface. We can assume
that intermolecular collisions may be neglected at very low pressure
and assume, for example, that in the case of two parallel plates, the
gaseous molecules strike one plate, then the other without ever colliding
with one another. This implies, of course, that the mean free path is
infinitely large compared to the distance between the plates, or generally
speaking the two surfaces bounding the gas. We then have the free-
molecule heat conduction. I t is easy to understand that, in this regime
of heat conduction, the conditions remaining unchanged, the heat
exchanged is proportional to the number of molecules striking either
surface and consequently proportional to the pressure of the rarefied
gas between the two plates.
However, there will always be a certain number of intermolecular
collisions, but these are more or less negligible when compared to the
number of impacts between the molecules and the solid surfaces. From
the above conditions, we may summarize by noting that for two surfaces
exchanging heat energy, it can be assumed that the heat conduction is
given by the classical formulae when the mean free path is small com-
pared to a characteristic dimension (for example, for two parallel plates
the distance between them is the characteristic dimension).
We can admit that the influence of rarefaction becomes significant
if the ratio L / d is larger than 0.001. For values of this ratio ranging
between 0.001 and 0.1 there is a regime corresponding to the tempera-
ture jump phenomenon. When L / d is larger than 10, free-molecule
heat conduction occurs, and finally for values of this ratio ranging
between 0.1 and 10, we consider we have an intermediate regime.
In. Heat Conduction of a Highly Rarefied Gas
A. ENERGY
OF A GASEOUS
MOLECULE
Before taking up the notion of the accommodation coefficient, we
think it necessary to recall a certain number of points specifying the
nature of the energy carried by a molecule of gas.
Depending on its structure, a molecule may possess several types
of energy. In general, whatever its atomicity may be, it possesses a kinetic
energy of translation which is, as we know, proportional to the absolute
temperature.
Besides the kinetic energy of translation, the molecule may possess
~761
a kinetic energy of rotation and an energy of vibration, when the molecule

contains two or more atoms. T h e value of these different sorts of energy
depends not only on the structure of the molecule but also on its
temperature.
We know that, in the case of diatomic molecules, the energy of rotation
is added to the energy of translation. T h e former energy is very small
at low temperature; in the vicinity of the standard room temperature,
it almost reaches a maximum value equal to kT which is of the energy
of translation of the molecule.
Polyatomic molecules also possess an energy of vibration. T h e varia-
tion of the specific heat of the polyatomic gases as a function of the
temperature shows that this energy has a noticeable value for a suffi-
ciently high temperature; the actual value of the temperature level at
which this contribution becomes important depends on the structure
of the molecule.
T o sum up, in a gas in thermodynamic equilibrium, only the energy
of translation is directly proportional to the absolute temperature; the
other types of energy of the gaseous molecule depend both on the struc-
ture of this molecule and on the temperature.
B. DEFINITION
OF THE ACCOMMODATION
COEFFICIENT
Consider a gaseous molecule having a certain incident energy Ei
and assume this molecule strikes a solid surface at temperature T p and
is reflected or re-emitted from it, carrying away an energy E, which
may be different from the energy E, that the molecule should have carried
if it had completely adjusted or accommodated its energy to the wall
temperature.
T h e accommodation coefficient is given by the equation
Ei - E,
a= (4)
Ei - E ,
This accommodation coefficient being thus determined, concerns the
total energy of the molecule. Accommodation coefficients corresponding
to energies of translation, rotation, and vibration of a polyatomic molecule
can be determined as well. I n the case of monatomic molecules, the
accommodation coefficient, whose definition is given above, corresponds
to the energy of translation since it is the only form of energy that such a
molecule can carry away.
I n case separate accommodation coefficients are determined for trans-
lational, rotational and vibrational energies it is not certain they are
equal. We shall further see that such is not the case.
12771
F. M. DEVIENNE
T h e definition of the accommodation coefficient we have just given,

may be, so to speak, an individual definition as far as it concerns one
molecule. We know however that the energies carried by the molecules
of a gas are not the same but that they are statistically distributed around
a certain value. Consequently, the notion of the individual energy of a
molecule must be replaced by the notion of mean energy, in a gas in
equilibrium.
I n some cases, it is convenient to replace the definition of the accom-
modation coefficient given by Eq. (4), by the following expression
Ti - T,
a=
Ti - T ,
This expression can be deduced from the preceding one in replacing the
energies of the molecules by the absolute temperature, which assumes
that energies are proportional to temperatures. We have seen that such
was the case for a monoatomic gas. O n the contrary, the expression is
not correct for a polyatomic gas. I n this definition it is assumed that the
flux of incident molecules and re-emitted or reflected molecules are in
equilibrium, and that the molecular speeds distribution is given by
Maxwell's law. Now, we must point out that in a highly rarefied gas
there can be no equilibrium because the collisions between molecules
are rare. T h e same phenomenon can be observed when a body is moving
in a highly rarefied atmosphere, the relative speed of the molecules in
relation to the wall they strike, no longer having a Maxwellian distribu-
tion since the translation speed of the body must be added to the speed
of the molecules. However, in many cases, for reasons of convenience,
we shall be led to replace the total energies of the molecules by the
absolute temperatures in the expression of the accommodation coeffi-
cient. T h e resulting formulae we will deduce can only be approximative
formulae. T h e expression in which the energies of the molecules appear
is the only rigorous one.
Without anticipating the results of our investigation, let us point out
that the accommodation coefficient depends on many parameters. In
particular it depends on the nature of the gas molecules and of the
molecular structure of the solid surface where the accommodation
occurs. I t also depends on the physical condition of the surface, i.e.,
its geometrical shape and cleanliness. Lastly, it depends on many varia-
bles, particularly on the difference of energy between the energy of the
impinging gas molecules and the energy carried away from surface by
the molecules if we assume that they are in equilibrium at T, . We shall
give further details about the influence of the parameters having an
influence on the values of accommodation coefficients.
~781
C. ENERGYTRANSFERBETWEEN A WALLAND A
HIGHLYRAREFIEDGAS
Consider a solid with a plane surface, at temperature T, and assume
this solid is immersed in a rarefied gas, at temperature Ta . Suppose that
the pressure is so low that we can neglect the dimension of the surface
compared to the mean free path of the molecules of the gas, but that the
gas extends to infinity so as to be in equilibrium at temperature TD. I n
order to make everything clear, suppose T, is higher than Tg.
T h e heat transfer rate across unit area per second is given by
Q == va(E,,, - E,) (6)
Where v gives the number of molecules striking the unit area per second,
a being the accommodation coefficient corresponding to the total energy
of the molecules.
I n the case of a monoatomic gas, the preceding relation becomes
-
Q = v a 2k(T, - T,) (7)
for the translational energy due to a stream of molecules striking a wall
is equal to the $ of the energy in the midst of the gas owing to the fact
that the fastest molecules that have a greater probability of impacts
on the solid surface carry a greater amount of energy. Finally, if the gas
is a diatomic gas whose temperature is close to standard room tempera-
ture, we can admit that the energy of the gas is equal to Ea = 3kT.
T h e expression is then equal to:
Q = va * 3K(T, - T,) (8)
T h e variation of the energy of vibration is neglected here.
Let us recall that v is equal to inv, where n being the number of
molecules per unit volume, va the mean arithmetical speed of the mole-
cules, it is equal to
va =J/8kT
(9)
k is the Boltzman constant and m the mass of the molecules.

T h e expression of the heat transfer between a wall and a rarefied gas
shows that if rarefaction is sufficient, this heat exchanged by conduction
is:
(1) Proportional to the pressure, i.e., to n,
(2) proportional to the accommodation coefficient, and
(3) proportional to the difference in temperature between the gas and
the wall.
12791
F. M. DEVIENNE
D. HEATCONDUCTION
BETWEEN T w o PARALLEL
PLATES
IMMERSED
IN A HIGHLYRAREFIED
GAS
Consider two parallel plates separated by a distance d. This distance
is regarded as small compared to the length of the mean free path L
of the molecules in the rarefied gas we are studying.
T h e respective temperature of the plates are Tl and T , , the mean
temperatures of the molecules issuing from the plates P , and P 2 , dif-
ferent in nature, T, and T2,and a, and a2 the accommodation coefficients
of the rarefied gas on the plates. We obtain the following formulae:
El - E2 == a,(E, - E2)
E2 - El == a2(E2- El)
El, E, , El,and E, figure the energies of the molecules corresponding

respectively to temperatures T, , T 2 , TI, and T2. Suppose T , (the
temperature of the first plate) is higher than T2 (the temperature of the
second plate). T h e energy carried by a molecule is equal to
AE = El -
If we assume the distribution of velocities of the molecules issued from

the two plates to be Maxwellian, the amount of energy carried across
unit area per second is:
I n this equation v gives the number of molecules striking the unit area
per second, hence v = $ 7/21, when we write that there is no accumulation
of molecules on a plate, we obtain the following formula:
~ I2 n1v - I n2 JVa2
0 1 - 2 (13)
T h e factor is 3 instead of $ because n, and n2 represent the number of
molecules per unit volume moving toward one side only, i.e., just half
the number of the molecules of a gas in which the velocity distribution
is Maxwellian. Besides, the formula can also be written
We define va in the following way:

Low DENSITY
HEATTRANSFER
Finally, by replacing the mean speed va' by its expression, we obtain

the following formula:
where temperature T' corresponds to velocity vat, whence:
T,' and T,' are the temperatures of the molecules of gas issuing respec-
tively from plates P, and P, .
When the gas is monatomic, the values of these temperatures is given
by the equations deduced from the preceding ones:
Let us calculate the density of an energetic stream due to the molecules

whose temperature varies from T,' to T,' and striking the two plates per
unit area per second. We must notice that the translational energy
carried by a stream of molecules equal to unit mass and issuing from a
gas at temperature T i , has a value 2 r T i , Y being the gas constant per
unit mass.
For a diatomic or polyatomic gas, the incident energy corresponding
to a mass of Ggms of gas at temperature Ti striking the unit area per
second is:
+
G(2rTi Ui) (17)
where U iis the internal energy of 1 gm of molecules at temperature Ti .
Let us recall that
T h e energy actually delivered or received by one of the walls is equal to
Q = G(C, + *y)(Tl' - T i ) (19)
C,,being the specific heat at constant volume for
c,=++-dUi
dT
F. M. DEVIENNE
since
c, + 31 2(y
=1 + l)C, with y = C,/C,,
Finally, the density of the energetic stream is equal to
In this expression, p' is the pressure of the Maxwellian gas having the
same density as the rarefied gas, but at mean temperature T'.If both
accommodation coefficients are the same, expression (20) can be written
as follows:
T h e two preceding expressions show that the quantity of heat trans-

ferred across unit area per second between two plates through a highly
rarefied gas is proportional to the pressure and independent of the dis-
tarice separating the plates.
If u2 = 1, Eq. (20) becomes
E. HEATCONDUCTION
BETWEEN T w o COAXIAL
CYLINDERS
Here a new feature enters in that one molecule will strike the outer
cylinder more often than it will strike the inner cylinder. T h e ratio of
thenumberof molecular impact on the two surfaces is proportional to the
radii of the cylinders. If rI is the radius of the inner cylinder and y 2
that of the outer cylinder, the ratio is r l / r 2 .
If a, and u2 are the accommodation coefficients of the molecules of the
rarefied gas on the cylinders 1 and 2, the equations can be written in
the form:
These equations can be deduced from Eqs. (10) by replacing El' by

the following expression,
HEATTRANSFER
Low DENSITY
that we obtain in writing that the stream of molecules striking the sur-
face of the outer cylinder is constituted on the one hand, by molecules
issuing from the inner cylinder, and on the other. hand, by molecules
having already struck the outer cylinder.
From Eq. (22), we obtain:
which are similar to Eq. (lo), provided a, is replaced by
Finally, from these equations we can obtain the heat conducted across
unit area per second from the inner cylinder:
with
If we assume the two accommodation coefficients are the same, we have

the resulting equation:
We notice that if yl = t 2 ,we obtain the formula (20) we had found in

the preceding paragraph.
For the particular case in which rl is supposed to be negligible in
regard to r 2 ,the Eq. (26) is then written in the form
Q = M T i - Tz) (28)
T h e latter equation shows that the quantity of heat exchanged between

the two coaxial cylinders-the diameter of the inner cylinder being
negligible in regard to the diameter of the outer cylinder- is proportional
to the accommodation coefficient of the molecules of the rarefied gas
on the surface of the inner cylinder. Moreover, this formula is employed
in calculating the accommodation coefficients.
T h e inner cylinder consists of a wire stretched along the axis of a
cylinder whose diameter is much longer than that of the wire.
~831
F. M. DEVIENNE
When we study the experimental determination of the accommodation

coefficient we shall go into every detail of the application of this method.
F. CALCULATION OF THE HEATEXCHANGED BETWEEN

ANY T w o SURFACESIMMERSEDIN A HIGHLYRAREFIED
GAS
We may notice that in the calculation of the heat exchanged between
two parallel plates, we have introduced no assumption whatever con-
cerning the parallelism of the plates. We have simply assumed unlike
the calculation we made in the precedent paragraph, that the molecules
striking one of the surfaces, would strike the other immediately. We see
that whatever the surfaces may be, this assumption being checked, the
result given in Section D is exact.
If, on the contrary, owing to the disposition of the surfaces one mole-
cule strikes one surface more often then the other (such is the case for
two concentric surfaces) similar formulae to the one established in the
preceding paragraph must be applied. We replace the ratio r J r z by
the ratio of the probabilities of impact of a molecule with surfaces 1 and 2.
I n any case, the different formulae point out that the heat exchanged
between any two surfaces immersed in a highly rarefied gas is independ-
ent of the distance separating the surfaces. I n addition, these amounts
of heat are proportional to the pressure of the rarefied gas provided the
mean free path is large in regard to the distance between the surfaces.
G. ENERGYRECEIVEDBY A PLANE
PLATEMOVINGI N A
HIGHLYRAREFIEDGAS
Though this case does not deal strictly with heat conduction, but rather
with rarefied gas dynamics, nevertheless we have found it interesting to
investigate this case on account of its importance. This will be a brief
presentation; further details will be found in reference (26).
Let us consider a plane plate moving in such a way that it remains
normal to the direction of the motion. We have seen that the number of
molecules striking this surface when it is at rest, immersed in a highly
rarefied gas per unit area per second is, for each side of the plate, equal to:
v = +zv,
When the plate is in motion, we understand that this number varies

particularly if the plate has a speed which is perpendicular to it, there
will be more molecules striking the front part of the plate, than molecules
striking the rear part of it. I n order to calculate the number of molecules
striking the plate, let us consider an element ds of the plane plate P.
~341
We assume that the speed of the plate P relative to the gas at rest is U.
Let us consider three axis with coordinates x, y , z in relation to the plate
with the x = axis perpendicular to the plate. Let us call I' the speed of a
molecule in relation to the plate, w being the speed of a molecule in
relation to the axes parallel to x-axis, y-axis, and z-axis but carried away
with the gas (Fig. 2). I t is obvious we have the following relation:
v=u+v (29)
FIG. 2.
We have:
T h e absolute speed of a molecule in relation with the surface is then

given by
+
v = dV,Z v,z V 2 + (31)
T h e number of molecules striking the surface element ds per second
with speeds ranging in the interval
(Vx,vx + dV,)(V,, v, + dV,)(VZ, + dV,) V z
are in a cylinder whose base surface is ds and whose height is V , . This

number of molecules is (we assume the regime flow to be steady):
f ( V ) n V , dS dVx dV, dV, = f ( ~ ) nV xdv, dv, dv, (32)

wheref(v) is the maxwellian distribution for the speed w , n being the
number of molecules per unit volume.
I n order to obtain the total number of molecules striking the front
side per second we only have to integratc the preceding expression in
~2851
F. M. DEVIENNE
making V , vary from 0 to f c o , Vv and V , from -co to +a.We

then obtain the equation:
V dS = n dS lirn
0
V,f( V )dVz
-m
dVv
-m
dV, (33)
with
v , being the most probable speed of the molecules at the temperature of

the gas.
We integrate this equation and obtain
expression which gives the number of molecules striking the unit area
of the front side of the plate, per second. It can be noted that if 8 is
the angle formed by the perpendicular to the surface and the direction
of the speed U, we have:
u, = u cos e
the preceding expression is written:
I n this expression the function @( Ux/vs) represents the probability

integral having Ux/vsas upper limit:
If we calculate the number of molecules striking the unit area of the

rear side of the plate, per second, we find the same expression in changing
the sign in front of U, which gives
or else:
Low DENSITY
HEATTRANSFER
I n order to calculate the energy received by the front and rear sides,
at first, we will assume the gas to be a monatomic gas. Let us give the
elements of this calculation effected for the first time by Stalder and
Jukoff (132, 133).
The total kinetic energy of the molecules striking the front side of the
plate per unit area per second is equal to
or
If we integrate, we have
If 0 is the angle between the perpendicular to the surface and the direction
of the speed U , we have:
(41)
Generally speaking, this expression can be written in the form
F. M. DEVIENNE
k being Boltzmann constant, Ei' the mean kinetic translational energy

of an incident molecule striking the front side of the plate, we can write
E.' = m- 7 +$'KT (43)

with
When the speed is null, we obtain $' = 2, whence Ei' = 2kT which is a
classical result of the kinetic theory of gases. On the contrary, when U
is very large, the preceding equation becomes
Ei'= m -u2 5
- +-kT
2 2
Figure 3 represents graphically the variations of $', we see that this

function rapidly increases from 2 to 2.5 and reaches fairly this value for
s, = uz/v,
= 2.
2.6
2.4
JI I'
2.2
2 .o
0 1.0 2.0 3.0 4.0 5.0
FIG.3
In the same way, let us calculate the translational energy El' of the
molecules striking per unit area and per second the rear side of the
plate:
whence
""q= Y" [m -
u2
2
+ +"AT1 (45)
with Y" given by the Eq. (36) or (37) with
T h e preceding equation can also be written in the following form:
Figure 4 represents the variation of +". It is a decreasing function

varying from 2 to - CO; this function can only be interpreted if molecules
strike the rear side, i.e., if U,/v,is inferior to 1.6. T h e formulae we have just
established allow the calculation of the energy of translation received by
the front and rear sides of the plate per unit area per second.
t 2.0
0
-4 0
-8 0
9''
-12 .o
-16.0
-20.0
-24.0
0 1 2 3 4 5
FIG.4.
F. M. DEVIENNE
When the gas is polyatomic, the energies of rotation and vibration

must be added to the translational energy. I n order to calculate the energy
yielded to the plate by a polyatomic gas, these different forms of energy
must be taken into account as a rule. However, there is ground for
noticing that the proportions of energy delivered on the surfaces depend
on the values of the accommodation coefficients corresponding to these
different energies.
I n the case of the diatomic gas we obtain-for the amounts of energy
received by the front and rear sides of a plane plate-the following
equations (in assuming the variations of the vibrational energy to be
negligible) :
Ei = n u2 + (16
2
+ 1)kT (49)
T h e amount of energy received by the front side per second is for a

monoatomic gas equal to
W = val[Ei - E,] = a,v[(m/2)U 2 + $kT- 2kTJ (51)
E, being the mean energy of a molecule if the coefficient of accommoda-

tion is equal to unity.
For a diatomic gas,
W = U ~ V [ E-~ E,]
= alv[(m/2) U2 + (4 + 1)kT - 3kTJ (52)
for the rear side we have, respectively,

W = a2v[E:(- E,] = a2v[(m/2)U 2 + 4kT - 2kTJ (53)
and
W = u~v[E;- E,] = ~2v[(m/2)U 2 + (111 + 1)kT - 3kTw] ( 54)
We notice, as we have seen before that the accommodation coefficients

of the front and rear sides a,, a, can be different, because the energy
of the impinging molecules are not the same.
H. LIMITOF VALIDITYOF THE PRECEDING

FORMULAE
I n establishing all the preceding formulae we have assumed the mean
free path to be infinitely large in regard to the dimensions of the solid
~901
bodies with which the gas would exchange amounts of energy. However,
collisions between molecules occur. Let us simply point out here that
if the ratio of the mean free path to the characteristic dimension, which
is Knudsen number is large, let us say superior to 10, the influence
of such collisions between molecules will be small, and the preceding
formulae are valid.
Whatever the value of the ratio of the mean free path L to the
characteristic dimension of the body may be (for instance the distance
separating two plates) and their influence is not always negligible. We
shall see further in a chapter dealing with the intermediate regime
where collisions begin to be important how corrections can be made in
order to take this factor into account.
IV. Heat Conduction in Rarefied Gases
A. TEMPERATURE
JUMP REGIME
We have seen that this temperature jump phenomen occurs when

the mean free path of the gaseous molecules in relation to the dimensions
of the bodies in contact with the rarefied gas is neither negligible nor
too large.
T h e study of the distribution of temperatures between two parallel
plates PIand P, respectively at temperatures TI and T , shows, as we
have seen that when the pressure sufficiently decreases, temperature
distribution between the two plates plotted against the distance is not
a straight line. Rather, it is represented by the solid curve in Fig. 5.
FIG. 5 .
F. M. DEVIENNE
As has been pointed out, it seems that near the wall, there takes place
a temperature jump given by
dT
T, - T,, = g -
dn
dTjdn being the temperature gradient normal to the wall in the midst
of the gas.
von Smoluchowski (122-126)was the first to show the existence of the
temperature jump. H e found that g the temperature jump distance,
is equal to 2.7 mean free paths for air, and 11 mean free paths for
hydrogen.
B. HEATCONDUCTION
BETWEEN Two PARALLEL
PLANEPLATES
We suppose the distance separating the plate is d. If the natures and
the structures of the two surfaces are identical the temperature jump
distances are the same. We may assume the apparent distance between
+
the plates to be equal to d 2g. T h e amounts of heat exchanged between
the plates per unit area per second are then given by the formula
Tl - T2
Q=A
+
d 2g (55)
When the two plane surfaces are different in nature or have a different
surface condition, the temperature jump coefficients of the gas in relation
to these surfaces are not the same; if we call them g, and g, the heat
conducted between the plane surfaces has the value
C. HEATCONDUCTION
BETWEEN Two COAXIAL
CYLINDERS
Suppose two coaxial cylinders whose respective radii are r , and r2
(yl being smaller than Y,). Let us first recall that the heat conducted
between the two coaxial cylinders when the pressure of the gas is
sufficiently high, is given by the formula:
Q stands for the heat conducted per unit area from the inner cylinder.
I n the case when we have a temperature jump, the heat exchanged
between the two coaxial cylinders can be written in the form
Low DENSITY
HEATTRANSFER
T h e expression log(r, + g 2 ) / ( r l- 8,) can be developed in series in

taking into account the fact that g , and g, are generally small in regard
to yl and Y , . This expression can then be written:
Finally, the heat transfered per unit area from the inner cylinder is given
by the formula:
This formula has been established in assuming g , to be small in regard

to Y , , for if g , is greater than or equal to Y , , the development in series
we have effected, limiting it to the first term would not be valid.
D. TEMPERATURE
JUMP MEASUREMENTS
We have already pointed out that von Smoluchowski (122-126) was the
first to determine the length of the temperature jump. His experiments
performed in a large range of pressures showed that this length g is
proportional to the length of the mean free path L ; this has been also
pointed out in the experiments conducted by Brush (13) at the same
period. Gherke (45) carried out the same experiments as Smoluchowski
with a more accurate apparatus, he confirmed the results obtained by
the latter; besides, he found out that the ratio g/L decreases at very low
pressures. Smoluchowski and Gherke deduced the temperature jump
from the measured heat conduction between the two surfaces at different
temperatures; on the contrary Lazareff (88) measured the temperature
distribution between two plates and deduced the value of g / L from
these measurements. Smoluchowski, Gherke, and Lazareff have used
Hydrogen and air. I n addition, Lazareff has operated with carbon dioxide.
T h e surfaces on which the accommodation took place were either clean
glass or metallic smooth surfaces.
T h e results obtained by Smoluchowski, Gherke, and Lazareff are
summarized in Table I. T h e value of L is the one deduced from the
expression given by Chapman, which is the following:
7 = 0.499nmv,L
where 7 is the coefficient of viscosity of the gas.

P931
F. M. DEVIENNE
TABLE I
~ ~ -~
Results obtained by
I
g/Lvalues obtained using Smoluchowski Gherke Lazareff
HP 11.22 9.20 12.40

Air 2.15 2.96 2.70
CO, - - 2.75
T h e agreement between the results obtained by the various experi-

menters is fairly good. Yet, if we take into account the theory that the
temperature jump distance depends on the accommodation coefficients
of the gas on the walls, and consequently the temperature jump should
depend on the nature of the surface. Such results are not in concordance
with the measurements reported by the preceding experimenters.
Later on, measurements of g/L have been take up again by Mandell
and West (95). They adopted Lazareff's method and found smaller
values. Such results can be explained by the existence of currents of
convection.
Weber (149-154) has determined with accuracy the temperature jumps
of many gases on platinum using the hot wire method, sometimes called
method of Schleiermacher, partly checked by measurements carried out
by means of double apparatus based on Goldschmidt method.
T h e radii of the outer cylinders are so small as to eliminate the
influence of the currents of convection. T h e measurements have been
effected with four different pieces of apparatus; the first two being
analogous to that ofschleiermacher, but improved, with a wire sufficiently
thick, while the other two had been built according to the method of
Goldschmidt.
Weber deduces the length of the temperatures jump g, on the wire from
the quantity of heat Qp given by the platinum wire at pressure p , and
from that measured at a very high pressure Qm .H e employs the following
formula:
where g, is the length of the temperature jump on the glass constituting

the walls of the apparatus yI and y2 are respectively the radius of the
wire and the radius of the outer glass cylinder. As Y /Y is small, he adopts
? ,
g, = g, in the term ( l/r2)(g,/gl) which is a corrective term.
12941
Low DENSITY
HEAT TRANSFER
From these measurements, he deduces that g p , i.e., g/L remains

constant in a great range of pressure. He obtains similar results with the
four apparatus, and the following average values of ratio g/L. On pla-
tinum, he obtains:
Argon 2.54
Helium 2.50
Neon 3.90
Hydrogen 11.70
Oxygen 2.74
Nitrogen 2.44
Air 2.85
Carbon dioxide 2.74
Methane 2.64
We notice that the values obtained by Weber with air, hydrogen and
carbonic anhydride are close to that obtained by Smoluchowski, Gherke,
and Lazareff.
It is worthy to note that there are few recent measurements.
Many experimenters who have endeavored to determine the values
of the accommodation coefficients have used the temperature jump
method.
However, the values of g they generally have given, are not expressed
in terms of the mean free path. I n this way, Thomas and Golike (142)
have compared experimentally the values of the accommodation coeffi-
cients of Helium, neon, and carbon dioxide by means of the temperature
jump method or by the low pressure method but unfortunately they have
not expressed the values of the temperature jumps in terms of mean
free paths. I n order to determine the temperature jump, we had to use
their reported values of accommodation coefficients and do some
additional calculations. T h e resulting value of g obtained in the case
of platinum is g = 1.436 for helium. It must be pointed out that this
value is much smaller than that obtained by Weber. T h e measurements
carried out by Thomas and Golike are certainly far more precise. Yet
even they give rise to important divergences in the values of the accom-
modation coefficients for taking into consideration only those experi-
ments which were casefully controlled, we still obtain an accommodation
coefficient of helium on platinum ranging between 0.146 and 0.196.
T h e value we have used corresponds to a = 0.149. If we adopt a = 0.196
we find a value of g which is about one third smaller.
I t can be noticed that the accuracy of the measurements of g as it
will be seen later in the measurement of accommodation coefficients,
is far from being satisfactory.
P951
F. M. DEVIENNE
F. THEORETICAL DETERMINATION
OF THE TEMPERATURE
JUMP
I N RELATIONTO THE ACCOMMODATIONCOEFFICIENT
AND THE
MEANFREEPATHOF MOLECULES
Several temperature jump theories have been proposed. T h e first
to be recalled is due to Smoluchowski. Kennard (68) gives an approxi-
mate theory which seems to be more satisfactory. This theory is based
on Maxwell's ideas and is analogous to the method of connecting the
slip coefficient with the fraction of the tangential momentum exchange.
We assume, as Maxwell did, that the stream of impinging molecules
is the same near the wall as it is in the midst of the gas. T h e temperature
gradient is assumed to be constant and the temperature close to the wall
is equal to T i .
These molecules will bring up to unit area of the wall in each second
the amount of energy of a Maxwellian stream issuing from a gas at
temperature Ti and the excess energy they carry as their contribution
to the conduction of heat is equal to
dT
+A-
dn
I n addition, the translational energy carried by a stream issuing from a

gas at temperature Ti is equal to 2rT, (this has been stated before).
Finally, we use the expression of that density of energetic stream as
calculated in Section 111, D:
but it must be pointed out that the energy exchanged between the wall
and the gas per unit area per second is equal to this density multiplied
by the accommodation coefficient of the molecules on the wall.
On the other hand, the energy actually delivered to the surface can
be equated to the total heat conducted across a parallel plane in the gas:
wi- w, = A - dT
dn
Hence we obtain
whence
2-a A
(Y + 1 ) G P
"7( 2 nT)t
~
Other theories connecting the accommodation coefficient with the tem-

perature jump have been established by Weber, and the one hand, by
Rating, Schafer, and Eucken on the other hand, and more recently by
Payne.
Weber starts from a relation given by Knudsen which connects the
heat conduction coefficient of the maxwellian gas with the coefficient
that he calls molecular heat conduction coefficient. He obtains the
following expression:
h = KepL (65)
where p stands for the pressure, L for the mean free path as it is given
by Chapmans formula.
From the relation (65) Weber obtains the expression of the tempera-
ture jump in taking the value 15/4 for K.
A e = g - dT =-.-
15 2 - a L -dT
dn 4 a dn
This expression which is valid for a monoatomic gas carrying only trans-
lational energy, is equally valid when the gas is polyatomic, provided the
value K = 2 is adopted.
I n this calculation of the temperature jump, Weber uses the values of
accommodation coefficients given by Knudsen (72-79),and by Keesom and
Schmidt (66, 6,7), and compares these values to those that he has found
in his experiments. I n the case of hydrogen the experimental value is
11.70L whereas the result of Webers calculation is 11.63L.
I n the case of air, the experimental value is 2.85 whereas the value
obtained by calculation is 2.81L. I n view of the excellent agreement of
the experimental results and theoretical calculation, Weber, in the
case of certain gases, such as methane and nitrous oxide, is confident
that he can deduce the average values of the accommodation coefficients
from the temperature jump measurements of these gases.
Finally, Weber uses formula (66) to work out the calculation of the
temperature distribution between two parallel plates brought up to
temperatures T, and T , respectively and separated by a distance d. If
we assume the accommodation coefficient a to be the same for both the
plates we obtain the following value of the temperature gradient in the
middle of the two plates:
_ - TI - T, .
dT - 1
(67)
dn 15 2 - a L
d 1 +T--
a d
W I
F. M. DEVIENNE
Weber compares the values observed by Lazareff with the values

calculated by means of this formula. Table I1 points out that, whatever
the pressure may be, the accordance between the calculated values and
observed values for air, on glass, is quite satisfactory. Schafer and
associates (119-121) attribute the apparent decrease in the heat
TABLE I1
Pressure Observed values Values calculated

(in mm Hg) of dT/dn from Eq. (67)
.__
P I = 760.0... 1.111 1.111

P z = 0.087... 0.850 0.830
P8= 0.065... 0.775 0.760
P4 = 0.019 ... 0.425 0.430
conduction of the gas at low pressure not only to the temperature

jump, but to the fact that the lower the pressure, the more difficult the
exchange of energy between the translational energy and the vibrational
energy.
If this be so, then a priori these must be a greater decrease in heat
transfer with decreasing pressure for polyatomic gases. I n this case, it
seems useful, after Schafer and his co-workers to introduce two tempera-
tures T , and T, respectively corresponding to the translational and
rotational energies, on the one hand and on the other hand to the
vibrational energy. They find the following relations between the tem-
peratures defined above:
cn aTn - h,,AT, - -(T,,

CV - T,)
v at VP
c* (T,, - T,)
cn aTa - h,AT, + -
---
v at VP
I n these expressions, C, represents the molecular heat capacity corre-

sponding to the translational energy and the energy of rotation, C, stands
for the one corresponding to the vibrational energy, a relaxation time,
and lastly V stands for the volume of a mole.
They arrive at the conclusion that the decrease in heat transfer with
decreasing pressure due to the fact that the energy of vibration has an
accommodation coefficient much smaller than that of the translational
energy, represents about 20 yo of the effect caused by the temperature
jump.
~981
These investigations show that the expressions previously given for

the temperature jump, are not complete in the case of polyatomic gases.
H. Payne has established from the kinetic theory of gases approximate
equations giving the temperature jump and the slip coefficient in the
case of an infinite plane solid surface of constant temperature. T,,,
bounding a gas whose states remains invariant throughout all planes
parallel to the surface.
I n order to obtain the temperature jump he makes no assumption
concerning the distribution of the molecules impinging on the surface,
but uses the mathematical theory of nonuniform gases. T h e value then
obtained for the temperature jump at the wall is as follows:
where
Besides
27rt *
8, = -12
f10
where no stands for the gas density near the wall, j , denotes a rather
complex expression which depends on the kinetic energy of the molecules
next to the wall.
I n the case when there is no slip but only a temperature jump, we have:
m av
g =- - v2fo+(v, z) dv
2 az
x being the direction normal to the wall.
Let us remark that the function of velocity distribution is given in the
six-dimensional space of velocities and positions by:
f(v, 4 =fO"l + +(v4 1

This theory is only valid when assuming $J to be very small in regard
to unity, which is the case when the temperature jump is small in regard
to the absolute temperature of the wall. I n the case of maxwellian
distribution the preceding expression becomes equal to
2-a
T , - Tw = -
2a 72To
which shows that with the help of the theory of Payne we can know the
limiting range on the temperature jump.
12991
F. M. DEVIENNE
V. Intermediate Regime
REMARKS
A. GENERAL
I n our investigation about the free-molecule regime, it has been
recognized that this regime was theoretically valid when there occured
no collisions between molecules or, better, that the collisions between
molecules could be totally neglected compared to the molecular impacts
on the wall; the accommodation coefficients of the energies of the mole-
cules striking the wall being consequently the only ones to be taken into
consideration.
In the same way, in the temperature jump regime the formulae that
we have established concerning the theoretical values of the temperature
jump are only valid if the rarefaction is not too great. As a matter of
fact, we assume the gas to be in equilibrium or at least, in pseudo-
equilibrium (since there is a temperature gradient); when the gas is
highly rarefied, the formulae are no longer valid, for the gas equilibrium
is not maxwellian.
It is necessary to replace the formulae giving the thermal conduction
between two surfaces, either in free-molecule regime or temperature
jump regime, by other formulae taking into account both the intermole-
cular collisions and the impacts on the walls.
B. CASEOF Two PARALLEL

PLATES
If L is the mean free path of the gas at pressure, p and d the distance
separating the plane surface, the number of probable molecular collisions
undergone by a molecule per second is equal to vJL,v, being the mean
speed of the rarefied gas molecules. If there are n molecules per unit
volume, there will be nd molecules for the volume included between
the unit area of the two surfaces.
Finally, the number of collisions between two molecules per unit area
is :
Va
vc = -nd (71)
L
This number must be compared with the number of impacts of the

molecules striking each surface per second. This number is
Low DENSITY
HEATTRANSFER
T h e ratio of the number of intermolecular collisions to the number of

impacts of the molecules on the wall is equal to
y c = 4 . zd
v
This ratio is inversely proportional to the mean free path that is to say,
proportional to the pressure. T h e ratio of the corrective term to the
heat conducted between the plates given by Eq. (20) is proportional to
the pressure, the corrective term itself being proportional to the number
of intermolecular collisions. T h e lower the pressure is, the more negligible
the corrective term in regard to the heat transfer rate thanks to the surface
impacts.
Furthermore, this result was to be expected. It is possible to calculate
this corrective term if we allow a few simplifying assumptions which are
logical, inasmuch as the pressure is supposed to be sufficiently low and
the mean free path large in regard to the distance separating the surfaces.
T h e first simplifying assumption is to admit that there occurs a
maximum of one collision during the trajectory of a molecule between
the two plates. T h e second assumption consists of admitting that the
accommodation coefficient of the molecules of a rarefied gas in regard
to the molecules of the same gas, has a determined value a.
Let us point that the number of molecular collisions required to
obtain in a gas the energetic equilibrium is far larger than one. As far as
translational energy is concerned, the number of collisions required is
10; in the case of rotational energy, it is about 1000. Besides a may
represent the average of the accommodation coefficients of the energies of
the rarefied gas molecules between them.
I n order to make this calculation, we assume the two streams of
molecules to carry energies corresponding T, and T, such as expressed
in formulae (16). We suppose that these gaseous molecules have the
same accommodation coefficient on the plates P, and P, . Let us consider
the molecules 1 and 2 which strike each other; they issue respectively
from plates P, and P, , After their impact they have energies E; and EL
corresponding to temperatures Ti and TL . We have
T 1 - T1 = a(T1 - Ti)= aa(T1- T;)

2 - T2 = a(T1 - T i ) = aa(T1 - T i )
T
If we write
dd = a(T, - T i ) = aa(T1 - T i ) (74)
r3011
F. M. DEVIENNE
we obtain
When the two molecules, after the mutual impact, go towards the same
direction, that is to say towards the same plates they transport energies
corresponding respectively to temperatures TL" cnd Ti'' after their
impact on the plates.
These temperatures are given by the equations
TI"- T''' = a(T; - T,)
Tr = T;' - u(T;' - T,) (76)
or, also, from the equation:

T;"= T,' -m(i -.) (77)
I n the same way,
'j''''' - T" = a(T - T")
2 1 2
T"'
1
= TI ' + AB(1 - U )
We see that the molecule 1 leaves plate P,at a temperature Ti'' inferior
to the temperature T,' it would have had if no molecular collision had
occured; the molecule 2 leaves plate PIat temperature Ti'' superior to
TI'. T h e energy transported between the two plates by means of mole-
cules 1 and 2 has increased (Fig. 6). T h e increase is proportional to the
difference in temperature:
2AB(1 - U )
r
I
D
I p2
FIG. 6.
Low DENSITY
HEATTRANSFER
When the molecules take the opposite direction after their mutual colli-
sion, we have
T;) - T; = U ( T , - T;) = a ( ~-, TJ +
ude
+
T i - Ti4) = u ( T ~- T,) = u(T, - T,) ad0 j (79)
Ti4)and Ti4corresponding respectively to the mean temperature that the

molecules have after striking plates P, and PI.Therefore, the energy
received by molecule I , which is proportional to Ti4)- TI is superior
to that received by molecule 2 when there occurs no molecular collision
between them. I n the same way, molecule 2 yields an energy superior
to that usually delivered by molecule 1. So, an increase in the energy
exchange between the plates must result from the molecular collisions.
When a molecule undergoes intermolecu!ar collisions between two
impacts on the plates, the temperatures T,!&, Ti;;, can be deduced.
We obtain the following formula:
T;,) = T,
T ; ; ~=
and for p collisions
) T,
- (1 - (1 -uye
+ (1 - u)de + (1 - u)(i - 1 (80)
) T, - (1 - u)m[i
T ; ~= + (1 - a) + (1 - ... + (1 - ay-11
which also includes formulae established for one molecular collisions

only. Accordingly, taking into account the probability of the number
of collisions in terms of the distance between two plates and the pressure,
we can deduce the corrective term that must be applied to the free-
molecule regime. It is obvious that, as soon as the number of collisions
increases appreciably we approach the slip regime. T h e correction
formula can a priori be only applied when there occur 2 or 3 intermole-
cular collision at a maximum, i.e., when the probability of collisions
does not exceed 3.
C. CASEOF Two UNSPECIFIED

SURFACES
T h e remarks we have made about free-molecule conduction are no
longer valid for the intermediate regime.
As a matter of fact, in our calculation of the heat transfer between two
plates, we have assumed the plates to be parallel and the distance between
PO31
F. M. DEVIENNE
them to be constant. Now, in the case of two unspecified surfaces, this

condition is only met in very particular cases and, anyway, even when
the two surfaces are parallel (case of two concentric cylinders or spheres),
if the calculation of the influence of one or several molecular collisions
upon the heat conduction is possible, this calculation cannot be carried
out in the same way as in the case of two parallel plates. Every time it
is necessary to specify the surface parameters allowing the calculation
of the distance covered by the molecules between two impacts on each
surface.
VI. Experimental Determination of the Accommodation Coefficients
A. DIFFERENT
METHODS
OF MEASUREMENT
(a) Determination of the heat exchanged between a wire heated by

passing electrical current through it and another surface, generally a
larger cylinder.
(b) Ornstein and Van Wycks method. This method is based on the
definition of the spectral emission of a rarefied gas unequally heated.
(c) Knudsens method using radiometric phenomena.
(d) Deviennes method using the rise in temperature caused by
radiation on a plate immersed in a highly rarefied gas or by the molecular
impact when the plate is in motion.
Above all, we shall be led to discuss the principle and realization of the
first method which has been the most widely used procedure for obtaining
accommodation coefficients.
T h e next two methods have been used to measure the accommodation
coefficients only under very particular conditions, the third one just
allowing the measurement of the accommodation coefficient of trans-
lational energy. T h e fourth method, quite recent, permits the measure-
ment of the coefficients under various conditions. I n particular it can be
employed when the first method fails to enable us to measure the accom-
modation coefficient. However, this method does not completely replace
the first one in the general case.
B. PRINCIPLEOF THE METHODEMPLOYING THE DETERMINATION

OF THE HEATEXCHANGED BETWEEN T w o SURFACES
HEATED TO DIFFERENTTEMPERATURES
T h e experimental device almost exclusively employed is that of the
heated filament, often called Schleiermachers method.
PO41
T h e principle of the method is as follows: a wire is stretched along the

axis of a cylinder immersed in a liquid maintained at constant tempera-
ture. T h e wire is heated by passing electrical current through it, and the
electrical resistance of the wire and the current flow through the wire
are measured. T h e measured resistance is a direct indication of the wire
temperature and the heat flow from the wire is found from the measured
current and resistance (i.e., q = PI?), thus, the electrical measurements
determine the total heat flow and the central wire temperature.
T h e temperature of the inner surface of the outer cylinder is also
measured usually by appropriately positioned small wire thermo-
couples. Usually the pressure of the gas in the apparatus is determined
by means of a MacLeod gage which is either directly connected to the
apparatus, or with a pipe device invented by Knudsen (this device allows
the measurement of the gas pressure, in a smaller volume, before its
entering the apparatus, hence a greater accuracy in the determination
of the pressure particularly when the pressure is low).
T h e heat exchanged between the wire and the concentric outer cylin-
der is caused not only by the heat conduction of the gas, but by the
radiation of the wire and also by heat flow along the wire itself and its
supports. It can also be produced by convection if the apparatus is large
enough and if the pressure is not very low.
We shall further see how the various investigators have actually
determined the heat exchange not only due to the heat conduction of
the gas. Let us point out, henceforth that this heat transfer rate (radiation
and conduction along the wire) can be easily measured by realizing total
vacuum in the apparatus, the wire being heated at the same temperature
as previously. T h e measurements carried out by the various experimen-
ters must be divided into two important groups, according to the value
of the pressure under which they have been performed.
When the pressure is low enough, that is when the molecules go from
one wall to the other without colliding with each other, the heat flow is
given by the following formula:
Usually, we can neglect rl in relation to r 2 . Thus we obtain
On the contrary, if the pressure is too high, free-molecule conduction

PO51
F. M. DEVIENNE
no longer is observed; we have the temperature jump regime. We have

previously seen that, in this case the formula we employ is as follows:
We have seen that, generally, g2/y2is very small in regard to gl/rl on

account of the small value of the radius rl of the wire in regard to radius
r2 of the cylinder. T h e formula can then be written as follows:
We put down
and we obtain
T h e value of g, can be deduced from the slope of the straight line repre-
senting the variation l/A plotted against the inverse of the pressure.
T h e preceding formula can be written in the following form:
with:
El
g11 = -
P
T h e values of accommodation coefficients can be deduced from such
relations as (64) between the accommodation coefficient and the tempera-
ture jump distance g.
There is ground for pointing out that these formulae are only an
approximation; consequently, the values of accommodation coefficients
so-obtained are approximate values only.
I n connection with this subject, it appears that many investigators
in measuring the accommodation coefficients of various gases may be
operating at sufficiently high pressures so as not to be in the free-mole-
cule conduction region. Many of these investigators give their results
without pointing out the specific formula they use to determine the
accommodation coefficients. On the other hand, some of them use
~3061
formula (83) which applies to free-molecule conduction, whereas,

considering the conditions of pressure at which they operate, it appears
that it is impossible to neglect the collisions between molecules. I n such
cases, it would have been more accurate to make a correction to take such
collisions into account.
When the measurements are actually carried out at very low pressures,
the energy due to thermal conduction may become very small in regard
to the total heat flow. In order to define more accurately the contribution
of the heat radiated by the filament and the heat flow transmitted
through the wire, it seems useful to make a preliminary calculation and
compare it with the molecule heat conduction.
C. CONTRIBUTION MODESOF HEATTRANSFER

OF THE DIFFERENT
First let us determine the energy loss from the wire by radiation; the
heat flow emitted by radiation from the wire to the inner surface of the
outer coaxial cylinder is given by the formula:
1
= Slo(T14- T t )
0,.
- + (-
1 1
-
S
I)$
1 2 2
where Sl stands for the surface of the wire, S, the inner surface of the
cylinder, (T the Stefan constant, and 4, and cg are the emissivities of the
wire and of the inner surface of the cylinder.
When the difference in temperature T,- T2 is small, the above
formula can be written under the form
At ordinary temperature we may notice that:
.
Q~ = 1.5 10-4sl
1
( T , - T,) cal/cm2//sec
-
1
1 + (z-
1 81
1) 3;,
If we denote by I the length of the filament or the length of the cylinder,

by d the filament diameter, and by D the cylinder diameter, the previous
equations can be written in the form
1
Or= 4uTm3nld 1 1 d ( T 1 - T2) (89)
-
1
+ (- 2
- 1)B
W I
F. M. DEVIENNE
with T, = (TI + T,)/2, and at ordinary temperature:

= 4.71 - 10-41d 1
(Tl - T,) cal sec-I
- + (-
1 1
- 1)-
1 2 D
At 27C the emissivities of tungsten and platinum are respectively
0.032 and 0.359; if the container is made of glass its emissivity is about
0.94. I n both cases, the term
is negligible relative to the term 1 / q .

Finally we can write that the heat flow exchanged by radiation between
the wire and the outer cylinder is given by the formula
and in the vicinity of ordinary temperature
d j r = 4.71 * - T,)
10-4~lZd(T1
T h e heat flow transmitted by conduction along the wire can be equated

d2
djC = T - A, grad T
4
I n this expression, the temperature gradient along the wire is unknown.

T h e temperature of the wire in its various points can be determined
theoretically. We set 0 as the difference in temperature between a point
of the wire and that at the wire extremities; the governing differential
equation giving this temperature is written in the form:
I n this equation A, represents the heat conduction coefficient of the wire,

A is the surface of its section, i is the current in amperes, po the resistivity
of the wire at room temperature, b the temperature coefficient of resisti-
vity, Y the radius of the wire, and q the heat loss per unit area and for a
difference in temperature equal to 1C. I t has been obtained by
equating the energy transferred from the wire, that is to say by radiation
or conduction.
~3081
Low DENSITY
HEATTRANSFER
T h e solution thus admitted is
1 being the length of the wire.

I n putting down
N = - -1 izp0
27rrq
M=- - bN (94)
4.2 X,A h,A
we deduced the mean excess of temperature of the wire in relation to
the outside, and also the temperature gradient at the extremities, which
defines the losses due to conduction:
This gradient is given by
Furthermore, in the case of purely molecular conduction the heat ex-

changed between a wire per unit area of this wire at a temperature T I ,
and on outer coaxial cylinder at a temperature T , is given by formula
(59). I n the present case, as rl/rz is very small this formula is written in
the form
= u , K ( T ~- TJ
For a surface of the wire equal to rrld the heat flow can be thus expressed.
= alTldK(T1 - T,) (97)
I n comparing these three quantities of heat, we see at first, that the
amounts of heat transmitted by radiation and by conduction through
the gas are proportional to the wire diameter; and besides it appears
that the heat transfer rate by conduction through the wire is proportional
to the square of the diameter. It is obvious that it is important to use
a very thin wire for the measurement of the accommodation coefficient
so that the latter quantity of energy should be small in regard to the
other two.
On the other hand, we notice that the ratio of the quantities of heat
exchanged by radiation and by conduction through the gas is equal to
F. M. DEVIENNE
It is to be noticed that this ratio is inverse to the pressure; besides it

does not depend on the length of the wire and its diameter; it is propor-
tional to the third power of the temperature at which we operate.
We must point out that the influence of the wire length is not negligible
when the losses by wire conduction and the distribution of temperatures
along the filament are concerned.
Let us indicate the respective values of heat exchanges by conduction
through the gas and by radiation, as numerical examples.
Let us consider the case the first experiments carried out by Roberts
(222-115) with helium. For these experiments he used a cylindrical
apparatus; the wire was a tungsten filament which was 0.07 mm in
diameter and 35 cm long. If the pressure is expressed in dynes per square
centimeter, the heat transferred to the gas from the wire per second for
a difference in temperature equal to 1C is given by the formula:
ag= 1.74 - 10-4parrld cal sec-l deg-1

dFT
I n this formula T represents the absolute temperature of the molecules
of the gas, m,their molecular mass and a the accommodation coefficient.
If we use the value obtained by Roberts for the accommodation coeffi-
cient of Helium on tungsten, 0.057, we have, in the case of helium:
Qg = 2.22 - lO-'p cal sec-l deg-' (99)
If we adopt the value 100 baryes for the pressure, this value is a little
lower than that used by Roberts, 0.1 mm mercury, we have:
a,, = 2.22 - 10-6 cal sec-1 deg-l (100)
T h e heat flow transferred from the filament by radiation, per second and
per degree, is
or = 1.5 . 10-4~,~1d (101)
i.e.,
@I = 3.69 * cal sec-l deg-'
el is equal to 0.032, the temperature is assumed to be equal to 300K.
We note that the heat transferred by radiation is about the sixth of the
energy transferred through the gas; then it is not negligible. This also
shows that these two quantities of heat are practically equal for a pressure
of about 20 baryes.
For lower pressures, the heat exchanged by thermal conduction is
therefore smaller than the sum of the other quantities of heat constituting
the corrective term.
[3101
I t must be noted that when the temperature of the filament is higher

than room temperature, the two quantities of heat are equal when the
pressure is higher than 20 baryes; on the other hand when the absolute
temperature decreases the accommodation coefficient varies little with
the temperature. This equalization exhibits that the heat exchanged by
thermal conduction trough the wire is smaller than that exchanged
by radiation, at least for such a wire as that used in Robert's first experi-
ments.
Many investigators have eliminated the influence of extremities in
using wires of various lengths.
D. VARIOUS DEVICESEMPLOYED
We shall indicate here the main devices employed since 1930. We
shall particularly apply ourselves to describing the differences that lie
between the device used by Roberts that will be taken as typical and
those used by various experimenters since that date :
1. Device Used by Roberts

T h e apparatus used by Roberts is mainly constituted by a tungsten
filament or a platinum filament in a glass tube immersed in a liquid, kept
at a fixed temperature by electric heating. T h e filament constitutes a
branch of a Wheatstone bridge. T h e thermoelectric effects are negligible;
the current running through the wire is not directly determined but the
total current running through the Wheatstone bridge is determined by
means of a potentiometric device.
I n the first experiments, the wire that had been used was 0.07 mm
in diameter and 35 cm long; it was stretched by means of a spring
constituted by a molybdenum wire with a diameter equal to 0.3 mm.
T h e cylinder containing the filament is connected on the one hand
with the diffusion pump and the MacLeod gage, on the other hand with
a reservoir containing the gas; liquid air traps are placed between the
cylinder containing the wire and the accompanying apparatus, so that
no impurity such as vapor of mercury can be adsorbed by the filament
and thereby modify the value of the accommodation coefficient. (Fig. 7.)
Robert's experiments were conducted in this way: first of all the
filament was flashed at a temperature markedly higher than 2O0O0C, in
order to outgas, then after interrupting the current, the filament was
connected with the Wheatstone bridge. A steady current then passed
through the filament so as to have in the filament a temperature 10-30"
higher than the temperature of the outer cylinder. T h e temperature of
[3111
F. M. DEVIENNE
Helium MacLeod
reservoir Pump gage
Molybdenum J(
spring
I
FIG. 7.
the filament is determined by means of the measurement of its resistance.

T h e resistance temperature coefficient of the filament on the one hand,
and its resistance at the bath temperature had been previously deter-
mined. T h e quantity of heat delivered by the filanient was then known.
Besides, the heat exchange by radiation and by conduction through the
wire was measured by obtaining high vacuum in the apparatus (Roberts
reached a pressure lower than mm mercury) and he heated the
filament at the same temperature; under these circumstance the total
heat flow is essentially the radiation exchange and the conduction along
the filament.
Measurements have been carried out at pressure ranging about 0.1 mm
mercury, the mean free path of helium being about 1.4 mm, that is
large in regard to the filament diameter which is only 0.07 mm. For
this reason Roberts applied the free conduction formula.
Roberts also checked that, in the range of pressures employed, the
heat exchanged was proportional to the pressure. I n other experiments
the purpose of which is to measure the variation of the accommodation
coefficient in terms of the temperature, Roberts uses a device analogous
to the previous one, the length of the filament being about half the other
length. He still measures the pressure by means of a MacLeod gage.
When the apparatus has a temperature different from ordinary tempera-
ture, he makes a correction (due to Knudsen) in order to take into account
the fact that the temperatures of the gage and the apparatus are different.
The device used by Roberts has been taken up by Raines (110) in
his measurements of the accommodation coefficient of helium on nickel.
[3 121
2. Device Used by Amdur and His Co- Workers

T h e measurements are at room temperature by means of the apparatus
represented in Fig. 8 mainly consisting of a polished platinum wire
between two brass plates.
T h e filament, which is 4.5 cm long, is disposed like a U in the middle,
parallel to the two plates; its diameter, 6.86 lop3mm, is measured from
the length and the resistance of the wire.
A wire analogous to the latter is placed between two strips of mica
in an ovoid tube made of platinum and connected with the apparatus.
This latter wire forms a branch of the Wheatstone bridge; its object is
to compensate the variation of temperature of the brass box. T h e other
two branches of the bridge are constituted by manganin resistances.
T h e whole apparatus is placed in a tank containing the gas whose accom-
modation coefficient is measured at low pressure; this low pressure is
obtained by the expansion of a small volume of gas the pressure of which
has been measured with all the accuracy required. Liquid air traps and
adsorbing apparatus have been placed in the device in order to purify the
gases employed by the experimenter.
FIG. 8.
3. Device Used by Grilly, Taylor, and Johnston

T h e apparatus they used consists of a steel tube whose diameter,
equal to 0.8 cm, is sufficiently small so as to avoid convection phenomena.
T h e axial wire is 0.075 mm in diameter.
T h e experimenters have taken special precautions for holding the
filament centered on the axis of the outer cylinder. A constant tension
is maintained on the wire through a 18 gm weight hanging on to a
PI31
F. M. DEVIENNE
piston in motion. T h e characteristic dimensions of the apparatus have

been measured with great care, in particular the radius of the wire and
its length.
T h e preceding apparatus is set up in a thermostat containing a suitable
liquid the temperature of which can be maintained between 120 and
400K. T h e investigators admit that when the thermostat is carefully
tuned, the variations of temperature in relation to the mean temperature
are about 0.003"C. T h e mean temperature may vary from 0.01" per
hour but these slow variations introduce no error in the measurements.
T h e resistance of the wire being measured by a millivoltmeter of
great precision, giving the potential difference between the extremities
of the wire; they simply had to divide this potential difference by the
intensity which was measured very accurately thanks to the potential
difference produced in a calibrated resistance of manganin. (See Fig. 9.)
2 I 3
FIG. 9.
T h e investigators operated at fairly high pressures; they were led to deter-

mine the correction of the temperature jump, they deduced the value of
the accommodation coefficient. T h e temperature jump was given by the
slope of the inverse of the apparent heat conduction of the gas, in
terms of the inverse of the pressure of the gas.
4. Device Used by Thomas and Brown

These experimenters, who have studied the accommodation coefficients
of nine different gases on platinum, have used a platinum filament in a
series of six different tubes. In particular, they describe the tube denoted
by B because they think that this design substantially improves the
technique of measurement.
This tube is conceived in such a way that the temperature along the
filament is uniform. Heat losses to the extremities are thus eliminated.
This result is obtained by raising two small blocks of platinum to the
same temperature as the filament. (See Fig. 10.)
T h e power spent by the filament and its resistance is obtained by the
classical method of measuring the potential difference at the extremities
of the filament and that occurring in standard resistance in series with the
filament. T h e temperature of the filament can also be deduced from its
W41
Sealed electrode!S
wire wound
1 Filament P t 121 100 mm 9
Tungsten (12110 m m 41
W = glass welding
FIG. 10.
value R, and its resistance coefficient. T h e thermostatic liquid which

was employed: kerosene was kept at 32"C, to f0.03"C.
T h e whole device, except the elements of the tube, was made of glass.
T h e usual precautions (liquid air traps) were taken in order to prevent
the vapor of mercury of the MacLeod gage from poluting the tube. T h e
other tubes varied as far as the diameter and length of the filaments were
concerned; the electric measurements were generally made by the same
method employed for tube B.
5 . Device Used by Oliver and Farber

Measurements were made of the accommodation coefficients of nitro-
gen, helium, and argon on platinum and tungsten, as a function of
the temperature of the filament on the one hand, and of the gas on
the other hand; these experimenters have used a fairly standard device
shown in Fig. 11. T h e originality of their device lies in the fact that
voltage is maintained constant by means of storage cells, chosen for
their ability to give essentially constant voltage under low-amperage
operating conditions.
13151
F. M. DEVIENNE
conductivity cell
FIG. 11.
6 . Device Used by Keesom and Schmidt

Keesom and Schmidt have measured the accommodation coefficient
of several gases, in particular that of helium on glass at 0C and at
much lower temperatures. T h e originality of their apparatus still based
on Schleiermacher device, lies in the fact that the wire on which the
accommodation occurs is positioned in a glass tube; specifically, inside
this thin tube a platinum wire 15p in diameter is welded. This platinum
wire serves to heat the glass tube and to measure its temperature. This
method can be extended to the measurement of the accommodation
coefficients of the same gases on various metals by producing a metal
film on the glass tube.
7. Other Devices
Some experimenters have used other devices based on methods which
are different from that of Schleiermacher. Accurate measurements of
heat transfer rate between two plates have been carried out, in particular
by Peck et al. (108).We must also mention the measurement of the accom-
[3 161
Low DENSITY
HEATTRANSFER
modation coefficients by mean of the measurement of the heat conduc-

tion through granulous bodies in dry air at very low pressure that we have
performed (20). It is obvious that this method can only allow the measure-
ment of the accommodation coefficient of a gas on the adsorbed layer
present on the solid body constituting the granular matter. I t is virtually
impossible to outgas by heating the powder.
E. MEASUREMENTS
OBTAINED
BY MEANSOF THE SPECTRAL
EMISSION
BY AN UNEQUALLY
HEATEDGAS
T h e experiment reported by Ornstein and Van Wyck (207) deserves
mention because they directly measured the velocities of molecules
after their impact on the wall. They passed an electric discharge through
a thin layer of helium at extremely low pressure (this pressure is not
given) between two coaxial glass tubes of which one was heated electri-
cally to 650K while the other was kept at 370K. They then observed the
shape of a spectral line emitted by the helium in a direction perpendi-
cular to the tubes. With this arrangement, one half of the observed spec-
tral lines came from the molecules that had struck the tube heated
electrically, while the other half came from those that had last struck the
cold tube, which had been roughened with 'CuO.
T h e half line from the colder molecules, when interpreted by the
Doppler-Fizeau theory corresponded to a Maxwellian distribution of
velocities at a temperature of 400"K, the other half, corresponding to a
temperature of 480"K, was not quite Maxwellian but exhibited a slight
deficit of low-speed molecules. From these experimental data, the experi-
menters calculated an accommodation coefficient of 0.32 for helium on
glass. Ornstein and Van Wyck recognize that their observations really
furnish information in regard to the energy of the molecules which are
present at a given moment in a given volume rather than information on
molecules which strike a surface in a fixed time.
F. MEASUREMENTS
MADEBY MEANSOF RADIOMETRIC
APPARATUS
(KNUDSEN'S
METHOD)
Knudsen has carried out measurements of accommodation coefficients
by using an apparatus based on radiometric forces. This device, shown
in Fig. 12, consists of a platinum strip P with one side bright and the
other side blackened. This strip is heated electrically so as to raise both
faces to the same temperature 8,; this temperature is determined by the
magnitude of the electric resistance of the strip. T h e temperature 8,,
of the tank is measured by means of a mercury thermometer. T h e values
PI71
F. M. DEVIENNE
FIG. 12.
i
of accommodation coefficients are deduced from the measurement of
the radiometric forces, determined by means of the torsion couple of
the wire holding the strip. From these observations, Knudsen measures
the difference between the accommodation coefficients on both sorts
of platinum: smooth and blackened. T h e experiments first performed
with helium enabled him to check the following formula he had
established theoretically:
p' - p" =1 4 - 60
(u' - a")
4P --
60
I n this formula p' and p" represent the pressures exerted by the gas on
both surfaces, p the gas pressure at the tank temperature, a' and a"
the accommodation coefficients on smooth platinum and blackened
platinum. Knudsen's device thus allows the determination of the dif-
ferences of two accommodation coefficients; yet it must be noted that in
the case of radiometric forces, the translatory energies alone have an
influence.
Consequently, the variation of the accommodation coefficients that
is obtained concerns only the energies of translation.
T h e determination of the accommodation coefficients of a gas in
relation on any conductive surface is possible in applying Knudsen's
method provided that the strip of platinum is replaced by a strip of the
body on which the accommodation coefficient of the rarefied gas is to
be measured; and the surface of this strip must be sufficiently rough on
[3 181
Low DENSITY
HEATTRANSFER
one side, so as to assume the accommodation coefficient to be nearly

equal to unity, as far as this surface is concerned.
T h e method invented by Knudsen is interesting because it allows the
determination of the accommodation coefficients dealing with the energy
of translation of the molecules. However, many important drawbacks
must be taken into account; it is not possible to raise the strip to a high
temperature. Besides the theory is only valid for very small variations
of temperature, which sets limits to its use.
G. METHOD
REPORTEDBY DEVIENNE
T h e method of the heated wire, though it is the most important one,
exhibits some deficiencies. First, it only allows the measurement of the
accommodation coefficient with a filament which makes the application
of this method difficult with certain metals. Besides with this method, it
is generally difficult to carry out measurements with nonconductive
bodies. Finally, the surface condition can hardly be varied and it is
difficult to study the structure of the surface by means of electronic
diffraction. Furthermore, the measurement of the accommodation
coefficient cannot be effected when the body is in motion in highly
rarefied atmosphere, That is why we have tried to introduce a new
method which can be applied in different ways as will be seen. T h e
principle of the method is to measure, on the one hand the ratio a /
and on the other hand the value of E (22).It must be added that these
two measurements can be made either successively or simultaneously.
Another characteristic of the method is that the rise in temperature is
measured between two analogous plates initially placed under the
same conditions, one undergoing afterwards a stream of energy, the
other one remaining under the initial conditions, practically unchanging.
1. Measurement of Accommodation Coe$cients under Usual Conditions

Two identical plates are located inside two compartments of a rectan-
gular copper box which is one to several centimeters thick. This box
is divided into two compartments by a copper plate of the same thickness.
Inside this rectangular box which is used as a thermal capacity two frames
are placed; they are connected with one another and two plates are fixed
in them, held in position by very thin nylon wires (Fig. 13).
T h e frames can be turned back during the experiment. Each compart-
ment is connected with the outside by an opening so as to have the same
pressure inside the apparatus and in the tank where the box is placed.
A thermocouple with a very thin wire whose junctions are fixed on
W I
F. M. DEVIENNE
the rear sides of the plates allows the measurement of the variation of
temperature between them. One of the plates is lit up by a light source
and received by this means an energy Wjsec. T h e difference in tempera-
ture between the two plates, 8, can be measured when we obtain a very
high vacuum, the pressure being less than lop6 mm of mercury. If we
assume 8 to range about a few degrees, we have
W = 8raT3O0 (103)
If cr denotes Stefans constant and T the temperature of the plate (the
two sides of the plate are assumed to be identical), we calculate the
accommodation coefficient of the plate if we measured the difference
in temperature between this plate and the other when a gas at a pressure
p is introduced.
T h e previous equation is replaced by the following one:
W +
= 8 ~ a T ~ 8 uKO (104)
K denotes the calorific capacity of the gas which strikes the unit area
of the plate per second.
When the gas is monatomic, we have: K = &.w,K. From Eqs. 103
and (104) we easily deduce the ratio a/.
- 0)
_a -- 16~aT~(O,
E nvaO
In order to have the absolute value of a, it is essential to know E.

We cannot rely upon the tables of constants which give average values
of E which do not always correspond to the experimental conditions.
At first the method applied to the measurement of E was to blacken the
rear side of the plate submitted to radiation. After vacuum, the rise in
temperature thus obtained gave
w = 4a(i + 4 7 7 3 4 (106)
From Eqs. (103) and (106) we obtain:
I t must be pointed out that, when we get small differences in tempera-

ture, ranging about a few tenths of degree, they are proportional to the
deviations observed on the galvanometer, which enables us to avoid
the calibration of the thermo couple. T h e measurement of c by means
of the curve giving the deviations in terms of the time is more reliable.
[32 11
F. M. DEVIENNE
At every moment, we can write that the energy W d t received during

a very short time dt is equal to the energy radiated by the plate R dt
plus the energy required to increase the difference of temperature
with the rectangular box by a value do.
We obtain the following differential equation:
W d t = RBdt + M,cdB
Whose solution is:
0 = #[I - e-tI7]
Mp being the mass of the plate, c its specific heat, 8' being the rise in
temperature given by Eq. (104).
We note that the time constant 7 is equal to
T h e mass and the nature of the plate being known, E is easily deduced
from Eq. (109).
Gilli (46) thus determined a certain number of accommodation
coefficients for helium, argon, nitrogen, oxygen, air, carbon dioxide and
methyl chloride.
This method has particularly allowed to point out that when we use
a nickelled copper plate, blackened with smoke-black, the accommoda-
tion coefficient for different gases generally ranges between 0.76 and
0.92; and for helium it is only 0.51. Therefore, it is not equal to unity,
which, a priori, we might have expected.
2. Measurement of the Accommodation Coeficients of the Molecules of a

Gas on a Plate in Motion
T h e method has been used by Devienne (23-25) to determine the
accommodation coefficients of various gases on various surfaces consti-
tuted by moving plates, at the extremity of a resolving arm, at speeds U
ranging from a few, up to 500 mjsec. T h e energy received by the plate in
motion, increases with the speed, as we have seen in Section 111, G.
We measure the difference in temperature of this plate, fixed on a small
insulating support at the extremity of the revolving arm, and of another
identical plate placed on an insulating support, fixed on the axis of the
revolving arm (Fig. 14).
T h e length of the latter support is large enough ranging between
10 and 20 cm so that the heat produced by friction inside the rotary
seal should not modify the temperature of the latter plate used as
~3221
Low DENSITY
HEATTRANSFER
eference weldlng
73 galvanometer
FIG. 14.
reference. From Eqs. (52) and (54) obtained in Section 111, G, we can
determine the energy theoretically received by the plate, either for a
monatomic gas or a polyatomic gas, on the assumption that the distribu-
tion of the gas is Maxwellian. This assumption is valid due to the fact
that the pressure is so low that the arm cannot give a motion to the gas,
and on the other hand that the volume of rotation of the revolving arm
is small in regard to the total volume of the vacuum tank in which this
rotation takes place.
In the present case, we can write:
w,= W' + W" (1 10)
hence:
with
s=-
U
VS
F. M. DEVIENNE
Therefore this method could a priori give only the ratio a / . i.e., the
relative value of accommodation coefficients. However, we simply have
to light the plate at rest either on the front side or on the rear side, by
means of a light source of constant energy, to obtain the value of E ,
in the same way as we did previously.
I n fact, the plate being supported by an insulator (generally Plexiglas)
we do not measure the factor of emissivity but its apparent value which
is, as a matter of fact, the one that figures in the previous equation.
From the measurement of a / we have been able to deduce the values of a .
3. Measurement of the Accommodation Coeficients of the Molecules of a

Molecular Beam
A third application of the method is that it allows the determination
of the accommodation of the molecules constituting a molecular beam
under various conditions.
I n particular, the method is now applied by Devienne and his co-wor-
kers to determine the accommodation coefficients of high energy mole-
cules (energy ranging from 100 to 3000 eV) obtained by charge exchange
(28). A plate is placed in the high speed molecular beam (it is held in an
insulating frame by 4 nylon wires). We measure the difference of tem-
perature between the latter and another identical plate hold in identical
frame, under the similar conditions of radiation, but placed outside the
high speed molecular beam.
T h e angle of incidence of the molecular beam of the first plate can be
varied by rotating the axis of the two frames by means of a small engine
operated by remote control. This allows, in particular, the study of the
variation of the accommodation coefficients in terms of the incidence
of the molecules. If v stands for the number of molecules, we have:
As the energy Ei of the molecules is very large compared with the

energy E, of these very molecules corresponding to the temperature
of the plate, we can write:
avEi = ~ E U T ~ ~ A S (1 12)
I n order to calculate E we can study the rise in temperature in terms
of the time, either by lighting the plate in vacuum, or by moving a screen
and receiving the molecular beam.
This method not only allows to measure the accommodation coeffi-
cient of high energy molecules on the plate and to syudy the influence
[3241
of the angle of incidence, but also to determine the influence of the gas
adsorbed by the surface of the plate.
As a matter of fact, in the previous device, it is easy to outgas by
means of an ion beam, for instance, and we can alternately outgas and
produce adsorption by sending a molecular beam either on one side or
on the other side of the plate which can be turned back. T h e molecules
of this molecular beam have an energy corresponding to the room
temperature and can be constituted by different gases.
T h e measurements which are being carried out at present enable us
to point out both the influence of a residual gas and a foreign gas. I t
must be noticed that the method which consists of measuring the ratio
a / and then E , so as to obtain the value of the accommodation coeffi-
cient, is a general one and that it particularly allows the measurement
of the accommodation coefficients when this measurement cannot be
made by means of the method of the heated wire.
I t also enables us to study the influence of such factors as the angle
of incidence of the molecular beam and the adsorbed gaseous layer,
influence that the method of the heated wire cannot exhibit.
H. GENERAL
RESULTS
T h e values of accommodation coefficients cannot easily be gathered.
As a matter of fact, the values obtained by various investigators for the
same accommodation coefficient are different from one another, even
when they operated under similar conditions. Different experimenters
indicating a probable error of about a few thousandths, sometimes obtain
results the values of which may vary in the proportion of one to two.
Obviously many results about accommodation coefficients must be
admitted cautiously some old data are unreliable owing to the use of
formulae that proved unsatisfactory in the calculation of the accommoda-
tion coefficient.
Accordingly, the results found by Sody and Berry (127), the first
results obtained by Knudsen and some results due to Dickins (30) must
be corrected in order to obtain values comparable with recent ones, as
Kennard (68) pointed out in his book, Kinetic Theory of Gases.
I n all the experiments connected with the measurements of accom-
modation coefficients, it would be necessary to give accurate details
about the state of surfaces. This was exhibited in the measurements of
the accommodation coefficient of helium on tungsten, reported by
Roberts.
When the wire has not been flashed at very high temperature, the
accommodation coefficient observed is, in fact, the accommodation
13251
F. M. DEVIENNE
coefficient of helium on the gas film adsorbed by the filament; its value
is then 0.19.
On the contrary, when the wire has been raised to a temperature
exceeding 2000"C, so as to outgas it, the value observed immediately
afterwards, is very low (0.05 for instance). As the experiment goes on
the accommodation coefficient increases and reaches 0.18.
Roberts ascribed this result to the fact that tungsten certainly roughens
after prolonged heating. This example points out the difficulty to obtain
the reproducibility of phenomena in the measurement of accommodation
coefficients.
Unfortunately, the accommodation coefficient does not depend upon
the nature of the gas and solid but upon the past history of the surface
struck by the gas molecules, i.e., upon its surface condition. It would
certainly be interesting to determine at a time, the accommodation
coefficients on a surface and the structure of this surface by means of
electron diffraction. However, we can affirm that the surface state greatly
accounts for the discrepancy between the results reported by various
investigators. A second cause of discrepancy lies in the fact that some
investigators used different types of formulae giving the heat conduction
of the gas, without questioning their validity when applied to the range
of pressures in which they operated.
I t must be noticed that the values of accommodation coefficients
we report, are those corresponding to the total energy of gas molecules.
We shall further see the attempts which have been made to determine
the various accommodation coefficient corresponding, in particular, to
the rotation and vibration energies.
1. Accommodation Coeficient of Rare Gases

T h e accommodation coefficients of rare gases particularly helium,
have formed the subjects of many determinations.
a. Helium. T h e value obtained for the accommodation coefficient
of helium on tungsten increases from 0.07 for helium on tungsten to
0.18 and even more, after prolonged heating of the tungsten, at a tem-
perature exceeding 2000C. I n presence of adsorbed helium, the accom-
modation coefficient varies from 0.19 to 0.55 for the two tungsten sur-
faces described above.
Recently, Oliver and Farber (206) have found an accommodation
coefficient of helium on tungsten close to 0.30. O n platinum, Soddy and
Berry found the value 0.50; Knudsen obtained the value 0.32 on bright
platinum and 0.91 on platinum blackened with platinum black; Amdur
(2, 2) obtained 0.403, whereas Thomas and Brown (242) found a value
13261
of about 0.2. Lastly, Mann (96-98) found 0.03 for platinum whose sur-
face had been purified by heating. Grilly et al. (57) reported 0.53. For
platinum, Farber and Oliver (106) found an accommodation coefficient
close to 0.2. Thomas and Golike (142) found values of the accommoda-
tion coefficient of helium on platinum ranging between 0. I55 and 0.253.
On a clean nickel surface Raines (110) found a value 0.072 whereas
Roberts reported 0.085. On glass, Keesom and Schmidt obtained a value
close to 0.38 at low temperature. On polished nickel Gilli (46) found an
accommodation coefficient equal to 0.35 whereas he obtained 0.51 on
nickel blackened with smoke-black.
6. Argon. For argon on platinum, Oliver and Farber found an
accommodation coefficient varying from 0.25 to 0.90 according to the
temperature of the wire. Gilli obtained a value of 0.86 for argon on
nickel blackened with smoke-black.
2. Diatomic Gases
a. Hydrogen. T h e accommodation coefficient of hydrogen on plati-
num has very different values, varying from 0.10 which was given by
Mann, to 0.40 given by Grilly and his co-workers. Soddy and Berry
found a value of 0.24; Knudsen, 0.32; Dickins, 0.34; Amdur, 0.312;
and Thomas, 0.27.
Let us recall that Knudsen found 0.74 for the accommodation coeffi-
cient of hydrogen on blackened platinum. Blodgett and Langmuir give
0.54 for the accommodation coefficient of hydrogen on clean tungsten
and 0.14 for a tungsten surface covered with a film of hydrogen. Lastly
Knudsen found the value 0.36 for the accommodation coefficient of
hydrogen on glass.
b. Oxygen. For oxygen, Soddy and Berry give 0.62, Dickins, 0.82,
Grilly and Taylor, 0.80; and finally, Mann gives values ranging between
0.42 and 0.55.
c. Nitrogen. Soddy and Berry found for nitrogen on platinum, an
accommodation coefficient equal to 0.68; Dickins gives the value as 0.81,
Amdur, 0.769, Thomas, 0.73, and Farber and Oliver obtained a result
of 0.55 on platinum and 0.53 on tungsten.
d. Carbon Monoxide. T h e accommodation coefficient of carbon
monoxide on platinum is indicated by Soddy and Berry as being equal
to 0.59, by Dickens, 0.80, and Amdur and Grilly give respectively
0.772 and 0.78.
e. Air. Gilli finds that the average accommodation coefficient of
dry air on nickel is equal to 0.75. Wiedmann and Trumpler report values
W I
F. M. DEVIENNE
for air on a number of engineering surfaces, in particular they obtained

values ranging between 0.87 and 0.95 for polished aluminum (see
Wiedmann, 156).
3. Triatomic Gases
I t is difficult to give all the values reported by the different authors.
However, let us give the coefficient of carbon dioxide on platinum.
Soddy and Berry give 0.552, Archer, 0.28, Dickins, 0.78, Grilly and
Taylor, 0.65, and Thomas and Golike, 0.78.
I. DIFFERENT
FACTORS
MODIFYING
THE VALUES
OF
ACCOMMODATION
COEFFICIENTS
I n the preceding paragraph we have reported values of accommodation
coefficients at the standard room temperature, and in spite of the dis-
crepancy between the values obtained by different investigators, it could
be observed that the nature of the gas molecule and that of the solid
surface played an important part.
We have also noticed that the structure and cleanliness of the surface
on which the accommodation took place, had to be taken into account.
A priori, there are many factors which have an influence on the values of
accommodation coefficients, which are as follows:
(a) nature of the surface,
(b) physical structure of the solid surface (cristalline or rough surface),
(c) presence of layers physically or chemically adsorbed,
(d) angle of incidence of the molecular beam or of the molecules on
the surface,
(e) temperature of the surface,
(f) shape of the surface,
(g) electrical state of the surface,
(h) degree of illumination of the surface,
(i) nature of the gas molecules,
(j) energetic state of these molecules,
(k) intensity of the molecular beam,
(1) background pressure, and
(m) energy difference Ei - E, .
T h e first factors, (a)-(h), are those which are related to the surface
on which the accommodation takes place; the other factors are related
[3281
respectively to the gas, (i)-(k), and simultaneously to the solid and the
gas, (1) and (m).
I n the following paragraphs, we shall study the experimental results
concerning the influence of some of these factors; the corresponding
studies of the other factors have not yet been carried out.
J. INFLUENCE
OF THE STRUCTURE
OF THE SURFACE
AND OF THE
LAYER
ADSORBED
We shall not develop this study in any detail. T h e only important
results are those obtained by Roberts; these results have already been
mentioned in connection with the values of the accommodation coefficient
of helium on tungsten. No systematic study has been undertaken on this
subject.
However, the results obtained by Roberts as well as these carried in
our Laboratory by Gilli have shown that when we increase the roughening
of the surface, the accommodation coefficient has a higher value.
K. INFLUENCE TEMPERATURE
OF THE SURFACE
In this connection, we must consider two distinct influences, which,

unfortunately many investigators have failed to do. On the one hand
we haveconsideredthevariation of the accommodation coefficient in terms
of the temperature.when the energy of the gas molecules nearly corre-
sponds to the temperature of the surface on which the accommodation
takes place; on the other hand, the variation of the accommodation
coefficients in terms of the energy difference E, - E, must be considered.
This exhibits again the complexity of the influence of different factors
on the values of accommodation coefficient. In particular, when we refere
to the variation in terms of temperature it is a priori necessary to specify
the difference of temperature that lies between the solid body and the
gas.
I n this paragraph, we shall only study the variation as a function
of the temperature; however, on account of the small number of experi-
mental data, it has not always been possible, unfortunately, to distinguish
the influence of the temperature from that of the energy difference.
The first accurate measurements of the variation of the accommodation
coefficient as a function of the temperature were made by Roberts on
helium and neon on tungsten. Up to the experiment reported by Roberts,
the accommodation coefficients were assumed to be equal to unity when
the absolute temperature approaches zero. T h e first experiments made
W91
F. M. DEVIENNE
by Roberts on the accommodation coefficient of helium on a clean

surface of tungsten have exhibited the following values:
T 79K 195K 295K
a 0.025 0.046 0.057
In the case of helium we remark that the accommodation coefficient

always seems to increase with the temperature. It must be added that
this result was equally reported by Raines who studied the variation
of accommodation coefficients of helium on clean nickel. This accom-
modation coefficient of helium on nickel is 0.048 at 90"K, 0.071 at 273"K,
and 0.077 at 369K. In the case of neon, Roberts finds that the accom-
modation coefficient slightly decreases when the temperature is raised:
0.08 at 79K and at 195"K, and 0.07 at 295K.
Rowley and Bonhoeffer (118)found a decrease of the accommodation
coefficient of hydrogen on platinum when the temperature increases.
At 109"K, the accommodation coefficient on a surface of clean platinum
is 0.27. When the surface of platinum is not clean but covered with
hydrogen, the accommodation coefficient is close to 0.5 at 120"K, and
close to 0.35 at a temperature ranging about 300K.
Grilly et al. (57) give the variation of the accommodation of nine
gases at temperatures varying between 80 and 380K. It must be
remarked that owing to the pressures under which they operate, an
important adsorption on the platinum wire certainly occurs which may
account for the high values reported. Figure 15 represents the variations
of the accommodation coefficients of different gases: air, hydrogen,
carbon dioxide, nitric oxide, carbon monoxide, methane, helium, oxygen,
and nitrogen peroxide.
We notice that in some cases, the accommodation coefficient increases
with temperature for instance the accommodation of helium on platinum
increases from 0.44 at 85"K, to 0.54 at 334K; the accommodation
coefficient of hydrogen decreases from 0.63 at 88"K, to 0.35 at 378K.
It must be remarked that the results obtained by Grilly et al. are not
very different from those reported by Rowley and Bonhoeffer for the
accommodation coefficient of hydrogen on a surface covered with the
molecules or atoms of this gas.
I n many cases, the accommodation coefficient exhibits a maximum
which is not veryimportant; such is the case for methane.Inher measure-
ment of the variation of the accommodation coefficient of helium on
nickel covered with methane, Raines finds a value varying from 0.42
at 90"K, to 0.36 at 273"K, and 0.343 at 369K.
W I
08
09
08
05 07 co
-
08 06
Q5
c
08 z
05 07 0
07 06 CH4 %
05 0
0
04 C
0
06 c
07 05 co, U
0
06 04 E
He 05 03 "
0
04 Q6 a
03 05
-02 0 4 H,
03
02
200 300 400 500 600
Temperature OK
FIG. 15.
In the present case, we notice that the variation of the accommodation

coefficient in terms of the temperature is reversed when the surface has
probably adsorbed the gas.
Archer observed the variation of the accommodation coefficient of
carbon dioxide on platinum in terms of temperature. He found that this
coefficient varies from 0.51 at a temperature of about 300"K, to 0.28 at a
temperature close to 600K; the intermediate values being 0.39 at 385"K,
and 0.33 at a temperature ranging about 300K. It must be pointed out
that the accommodation coefficient regularly decreases in terms of the
temperature. Therefore, it seems impossible to conclude in favor of a
systematic variation of the accommodation coefficient as a function of
the temperature; we remark that the assumption formerly made namely
that the accommodation coefficients necessarily approached unity when
the absolute temperature approached zero is false at least as far as helium
is concerned.
Keesom and Schmidt, studying the accommodation of helium, neon,
and hydrogen on glass at low temperatures (between 70 and 90K) find
that the accommodation coefficients decrease as the temperature increases
-for helium 0.383 at low temperature and 0.336 at standard room
temperature, for hydrogen 0.555 and 0.283, in the case of neon, 0.803 and
0.670, and for nitrogen 1 and 0.865 respectively.
They admit that at temperature lower than critical temperature,
adsorption occurs and the accommodation coefficient is equal to unity.
Above critical temperature, adsorption cannot be measured and the
W11
F. M. DEVIENNE
accommodation coefficient decreases with temperatures. Lastly, the

accommodation coefficient increases with the molecular mass of the
gas with which we are dealing. Most of these experimental results are,
as will be seen further, in contradiction with modern theories on the
accommodation coefficient; they all conclude that it increases with the
temperature.
More recently, Oliver and Farber have studied the variations of the
accommodation of helium, argon, and nitrogen in terms of the tempera-
ture of the filament on the one hand (this filament could be made of
platinum or tungsten) and on the temperature of the gas on the other
hand, These investigators have shown that the accommodation coeffi-
cients not only varies as a function of both the temperatures of the sur-
face and the gas, but also as a function of the difference between these
temperatures.
We shall see in the next paragraph the variation of the accommodation
coefficients in terms of the difference between the temperatures.
However, let us point out that, according to their results, the accom-
modation coefficient decreases as a function of the temperature of the
gas when the temperature of the wire is in the vicinity of this tempera-
ture. We must also indicate that the values obtained for the accommoda-
tion coefficient exhibit a maximum in terms of the difference of tem-
perature. Therefore, it seems that every time the influence of the
temperature of'the surface is to be discussed, it is necessary to specify
the difference lying between the latter temperature and the temperature
of the gas.
L. VARIATION
IN TERMS
OF THE ENERGY
DIFFERENCE
Very few investigations have been carried out in order to study the
variations of the accommodation coefficient in terms of the difference
in energy E, - E , , i.e., between the energy of the impinging gas
molecules and that corresponding to these very molecules at a tempera-
ture equal to the surface temperature.
Oliver and Farber (106) have studied the variations of the accommo-
dation coefficients of helium, argon, and nitrogen as a function of the
difference of temperature between the gas in the vicinity of the cold
wall and the tungsten filament,
T h e results relating to the temperature: 80K and 300K of the bulb
are represented by the curves of Figs. 16, 17, and 18. I n observing these
curves we notice that the accommodation coefficients vary markedly in
terms of the difference of temperature. I t seems, as it is assumed by
Oliver and Farber, that these accommodation values become very small
P321
Low DENSITY
HEATTRANSFER
0.6
A Bulb temperature = 300K

0.4
0.2
0
0 200 400 600 800 1000
A T (OK)
FIG. 16.
1.0
0.8
0.6
0.4
0.2
0
F. M. DEVIENNE
FIG. 18.
when the difference of temperature is small, about a few degrees.

Whatever the initial temperature of the gas may be, we note that these
values exhibit a maximum whose value is lower when the gas tempera-
ture is lower. On the other hand we remark that when we keep increasing
the temperature of the filament, the values of the accommodation
coefficients tend to be approximately constant. T h e results obtained by
these experimenters do point out what we said previously, namely that
it is necessary to specify the difference of temperature that lies between
the filament and the cold wall of the apparatus, in order to be able to
compare the values of the accommodation coefficients at a given tem-
perature of the filament.
Devienne (23-25) has studied the variation of the accommodation on
a nickel plate in terms of the energy of incident molecules by placing
the plate at the extremity of a revolving arm and by varying the speed
up to 500 mjsec.
These experiments have shown that all the accommodation coefficients
without exception decrease when speed increased. Very low values were
obtained for a high speed. At a pressure of about 0.25 p of mercury he
has found (for dry air) that a varied from 0.068, for a speed of 151.8 m/sec,
to 0.055 for 201.4 m/sec, reaching 0.042 for 401.1 mjsec (Le., for a value
U/v, close to unity). Figure 19 represents the variation of the ratio a/.
in terms of s = U/vyfor some gases. We remark that the higher the speed
W I
Low DENSITY
HEATTRANSFER
FIG.19. Nickel-plated copper plate: Air (1)p = 0 . 5 0 ~ ;(2)p = l p , Argon (4)p =

0 . 5 0 ~ (5)p
; ; = 0 . 5 0 ~ .Freon 12(8)p = 0 . 5 0 ~ .
= Ip. Carbon dioxide (6)p = 0 . 3 0 ~(7)p
Blachened-copper plate: Air (3)p = lp. Freon 12(9)p = 0 . 5 0 ~ .
is, viz., the higher the energy difference between the incident molecules
and the plate, the lower the accommodation coefficient is.
T h e values of a found in the course of these experiments are much
smaller than those found by many investigators. This is probably due
to the fact that our surfaces were outgassed for several days in very high
vacuum before the measurement. Moreover, experiments showed that
the accommodation coefficient decreased regularly during the time the
plate stayed in vacuum reaching a limiting value in about a hundred
hours under a low pressure close to mm of mercury.
M. INFLUENCE
OF THE PRESSURE
As it was pointed but before, the investigators operated at very different
pressures. Some of them like Amdur, Jones, and Pearlman systematically
studied the variation of the accommodation coefficients as a function
of the pressure. As a general rule, they explained the larger values they
found in the case of platinum covered with gases, by admitting that it
was necessary to reach a pressure o f & m m to get a saturation of the
platinum surface by the gas then employed. T h e measurements they
reported on different gases actually showed that the accommodation
coefficients vary markedly in terms of pressure.
I n the case of helium, for instance, they find that when the pressure
is close to l/lOO mm of mercury, the accommodation coefficient is close
to 0.296; on the contrary, when the pressure reaches a value slightly
exceeding & mm of mercury, the accommodation coefficient is close to
13351
F. M. DEVIENNE
0.400 and remains approximately constant at increased pressure. Amdur

and his co-workers find similar results for the accommodation coefficients
of other gas on platinum. In particular, for xenon, the accommodation
coefficient varies from 0.108 at the pressure of 0.007 mm of mercury
to 0.864 at a pressure of 0.4 mm.
In the case of krypton, the variation is also very large; from 0.232 to
0.840 when the pressure varies from 0.01 to 0.4 mm of mercury. I n all
these cases, as stated above, Amdur, Jones, and Pearlman find that from
a pressure exceeding 0.1 mm of mercury, the accommodation coefficient
remains constant. T h e results obtained for hydrogen at standard room
temperature exhibit the increase of the accommodation coefficient of
this gas in regard to platinum (0.239 for a pressure very close to 0.009 mm
of mercury; 0.312 for pressures higher than 0.1 mm). However these
values seem to be contradictory to those obtained by Rowley and
Bonhoeffer at much lower temperatures, ranging about 110K. At these
temperatures, it seems that the accommodation coefficient of hydrogen
on platinum slightly decreases when the pressure varies from 0.035 to
0.1 mm. As Rowley and Bonhoeffer did not operate at absolutely constant
temperatures, it is difficult to give an accurate value to the decrease of
the accommodation coefficient. From the table of the results obtained by
these experimenters the accommodation coefficient in the vicinity of
130K seems to vary from 0.348 for a pressure of 0.03 mm to 0.340 for
a pressure of 0.27 mm. T h e variation ranging about 1/40 has a very small
amplitude.
T h e results obtained by Amdur and Pearlman are logical since,
when the pressure of the gas increases, one or several adsorbed layers
are likely to be formed.
T h e mutual potential between the surface and the atoms of the sur-
rounding gas, is modified by the adsorbed layer. When the latter is
complete, the potential being modified in a constant way, the accommo-
dation coefficient must probably remain constant at a given temperature,
whatever the pressure may be if it exceeds a determined value.
However, Thomas and Brown find results which are in contradiction
with the preceding ones. On the one hand, when the pressure markedly
decreases, the accommodation coefficients of nine different gases, viz.,
krypton, argon, neon, helium, hydrogen, oxygen, carbon monoxide,
nitrogen, and mercury all approach a constant value. On the other hand
for most of them, the accommodation coefficients decreases when the
pressure increases; for instance, in the case of the krypton, they find
an accommodation coefficient slightly larger than 0.9 at a pressure
close to 0.01 mm, whereas Amdur and his co-workers has found an
accommodation coefficient of 0.232 for a similar pressure. When the
W I
pressure reaches 0.032 mm of mercury, the accommodation coefficient

decreases to a value equal to 0.64 whereas Amdur found 0.841, when the
pressure reaches 0.55 mm, the accommodation coefficient is equal to
0.500.
T h e measurements carried out by Devienne (23-25) by means of
the revolving arm have also exhibited an important influence of the
pressure. It seems, a priori, that the accommodation coefficient decreases
when the pressure increases, at least for pressure ranging between 0.1
and 2 p. This decrease is between 15 and 40% depending on the gas,
when the pressure varies from 0.25 to 1 p. of mercury.
In short there is undoubtedly an influence of the pressure on the
values of the accommodation coefficients but in the present state of the
experimental research it is diacult to establish laws, on account of the
numerous contradictions between results.
From the theoretical standpoint the influence of the variation of
pressure can be accounted for by the existence of an adsorbed layer which
is a priori thicker at higher pressures. It can also be explained by the fact
that the formulae we employ are more or less accurate and that their
application depends, in particular on the Knudsen number relating to
experimental conditions corresponding to measurements. In this con-
nection, let us remark that Thomas and Golike (242) have shown that
generally, a slight divergence existed between the values of the accommo-
dation coefficients obtained, on the one hand, from the temperature
jump and on the other hand, by the theory in free-molecule regime.
N. ACCOMMODATION CONCERNING INTERNAL

COEFFICIENTS ENERGY
It was previously remarked that if we consider a polyatomic gas, we

can define the accommodation coefficients of translational energy or
vibrational energy. The formulae given above, concerning heat transfer
are available when taking for a the mean value of the accommodation
coefficient corresponding to the various energies. T h e exchanges of
energy corresponding to the different types of energy of molecules cannot
be easily separated in order to measure a on the one hand, and a7 and
a, on the other. Yet we have seen that Knudsen, when studying the
radiometric forces at low pressures, measured the differences between
the accommodation coefficients of the translatory energy on hydrogen
on bright or blackened platinum. He finds that this difference is equal
to 0.415 whereas by heat conduction a difference equal to 0.420 is
obtained and corresponds, in this latter case, to the total energy of
hydrogen molecules.
W1
F. M. DEVIENNE
I n comparing the two values reported above it, it is shown that, in

the case of hydrogen, the accommodation coefficients of the energy
of translational and internal energy have very close values.
These conclusions can be explained by the fact thet in the vicinity of
standard room temperature conditions under which Knudsen operated,
the energy of oscillation of hydrogen molecules is small and consequently
it can be admitted that besides the energy of translation these molecules
only possess a rotational energy. This implies that the accommodation
coefficient for rotational energy must be nearly equal to that of trans-
lational energy, in the case of hydrogen; however, it is probable that at
elevated temperatures, the accommodation coefficient corresponding to
internal energy is weaker than the one corresponding to the energy of
translation. As a matter of fact, we know that at high temperature, the
energy of vibration has a substantial value and that, on the other hand,
a rather long time is required by the molecule to take a mean vibrational
energy corresponding to a given temperature. Therefore, it is easily
understood that, at high temperature there is a decrease of the accom-
modation coefficient due to the fact that the accommodation coefficient
corresponding to the energy of vibration is small.
This assumption is supported by the results obtained by Schafer
et al. (119) who admit that the accommodation coefficient for translatio-
nal energy and rotational energy in the case of molecules of carbon
dioxide, is practically equal to 1 , but that, owing to the small value
(0.13) of the accommodation coefficient in the case of the energy of
vibration, the accommodation coefficient has the value 0.8 1 normally
found. This shows that, owing to the difficulty of accommodation of
the energy of vibration, the exchanges of energy in rarefied carbondioxide,
have decreased by 20 yo.
These latter results are also confirmed by Duval and Niclause who,
in their study of the pyrolysis of acetaldehyde, showed that, at low
pressure the molecules of this body were not decomposed, after being
in contact with a platinum filament heated to a temperature of 1 3 W K ,
in a hot wire device, similar to those described above. This apparent
contradiction with the fact that acetaldehyde is highly decomposed above
750", can be explained in assuming that the accommodation coefficient
corresponding to the energy of vibration is much smaller that unity.
Duval and Niclause, adopting as limiting value of this accommo-
dation coefficient, the one corresponding to a similar value of the
accommodation coefficient for the energy of translation or else to a
value equal to unity for the latter coefficient. they obtain respectively
maximum values of a equal to 0.37 and 0.31 and minimum values equal
to 0.20 and 0.15 at temperatures of 10oO and 1300K. T h e result is that
13381
Low DENSITY
HEATTRANSFER
the molecules of acetaldehyde leave the filament with a mean internal

energy much smaller than that corresponding to the temperature of the
wire, hence the explanation of the fact that the molecules of acetaldehyde
are not decomposed.
VII. Theoretical Calculation on Accommodation Coefficients

A. GENERAL
OBSERVATIONS
I n this section, we shall study the different theories propounded to
explain the values obtained for the accommodation coefficients. T h e
preceding section showed that most of the experimental values are rather
contradictory; therefore, we cannot check the theories by calculating
the accommodation coefficient of a given gas on a particular surface.
Moreover, we shall see that apart from the solution of the problem from
the classical point of view (in spite of the simplifications effected) the
theoretical solution of the determination of accommodation coefficients,
is an extremely complicated problem. I n this way the theoretical works
carried out by Lennard, Jones, and Devonshire show that the calcula-
tion of the accommodation coefficient is only possible in very particular
cases and by assuming the metal constituting the wall, to have a parti-
cularly simple structure. Anyway, we have thought it useful to summarily
indicate the various theories on the accommodation coefficient, in this
study devoted to the heat conduction in rarefied gases.
I n fact, the importance of this accommodation coefficient is absolutely
fundamental, as it has been pointed out, in all the phenomena of heat
conduction in rarefied gases. We have also thought it useful to show
that if, theoretically after the kinetic theory of gases, the calculation of
heat conduction is easy in the case of a gas where the mean free path is
very small in regard to the dimensions of solid bodies, it is not so in
a rarefied gas; the difficulties are not only of experimental nature but
of theoretical nature.
THEORY
B. CLASSICAL REPORTEDBY BAULE
I n order to calculate the accommodation coefficient, B a d e (6) made

the simplifying assumption as follows.
T h e solid is supposed to belong to the cubic system, viz., the atoms or
molecules constituting it are located on the average on the corner of a
cubic lattice; it is understood that the atoms or molecules are animated
with oscillary movements due to thermal agitation; they have a mean
kinetic energy proportional to the temperature of the solid. H e also
P391
F. M. DEVIENNE
assumes the collisions between gaseous molecules and those of the solid
to be governed by the laws of elastic collisions. Accordingly, he deduces
that the kinetic energy of the molecules returning into the gaseous
phase after striking the molecules or atoms of the solid only once, can
easily be expressed in terms of the mean kinetic energy of the molecules
of the solid and that of incident molecules. He obtains the relation
where Ei , E, respectively represent the kinetic energy of incident mole-

cules and the mean energy of the molecules of the solid. This expression
can also be written in the form
or
When the incident molecule strikes successively two molecules of the

solid, we write:
E" = PE' + (1 - P)Es (1 16)
or
E" = P2Ei + (1 - P2)Es
Finally, if the gas molecule strikes n molecules of the solid its mean
kinetic energy, when this molecule is re-emitted into the gaseous phase,
has the following form:
E(") = P E i + (1 - P ) E S (118)
When comparing this result with the definition of the accommodation

coefficient, the expression of the latter is written as follows in Bade's
theory:
I n the particular case in which the molecules of the gas strike the
molecules of the solid only once, the accommodation coefficient has the
value
It is easy to remark that in this particular case, this expression has for max-
imum value 0.5 when the masses of the two sorts of molecules are equal.
It becomes much weaker if these molecular masses are very different.
I n order to obtain a value of the accommodation coefficient exceeding 0.5,
it is necessary that the gas molecules strike several times the solid mole-
cules before returning into the gaseous phase. If n is the number of
molecules striking the surface only once, a the number of the molecules
striking in twice, dfl)the number of the molecules striking the surface n
times before returning into the gaseous phase, the mean kinetic energy
of the molecules re-emitted into the gaseous phase, is equal to
It seems that for a given surface, platinum for instance, the accommoda-
tion coefficient must increase with the molecular mass of the gas in
which it is immersed. This result is roughly conform with those obtained
in the measurement of the accommodation coefficient of rare gases on a
platinum surface. These accommodation coefficients seem, in fact, to
increase with the mass of the gas atoms. It must be remembered that the
accommodation coefficient such as shown by Roberts considerably
depends on the roughening of the surface, which, as a matter of fact
Bade theory indicates; the average number of impacts of a gas molecule
on the solid surface obviously depends on the structure of the solid
surface. According to the modern conceptions of physics, this solution
can be but unsatisfactory and it is necessary to introduce wave mechanics
to obtain a solution effectively valid.
C. ZENERTHEORY
Zener (260) applies wave mechanics to the problem of the reflection
of atoms by crystals and to the determination of accommodation coeffi-
cients; he assumes that the atoms move perpendicularly to the surface
of the crystal which is assumed to exhibit cubic symmetry.
Among other simplifying assumptions, he admits that the interaction
energy between two atoms only depends on the distance I between
these atoms and that, consequently, it has the form V ( T ) he
; assumes that
we observe the same phenomenon for the mutual energy of two pairs
of atoms in crystal.
P411
F. M. DEVIENNE
Finally, in the case of light atoms, he finds that the factor of reflection
must be higher than 0.75 or equal to this value, which checks rather
roughly, it seems, the experiments of Johnson on the reflection of hydro-
gen beams by lithium fluoride. In order to determine the accommodation
coefficient after giving the same simplifying hypotheses, he assumes in
another paper that the mutual energy between the atoms of the gas and
the atoms of the surface has the following form:
V(Y)= ce-"d (123)
d is a constant and has the dimension of a length. H e thus obtains a
value expressed in the following formula:
where m, represents the mass of a molecule of the gas, m, the mass of an

atom of the solid, 0 being the characteristic temperature of the crystal
such as it is defined by Debye, and T the absolute temperature of the
crystal which is supposed to be close to that of the gaseous molecules.
T h e function x ( O / T )represents the effect of the quantification of the solid
on the value of a ; this function is written
its limiting values are:

x(0) = 1 x(00) = 0
T h e second function is defined by
with
if 0<3T
zo = T
zo = 3 if 0>3T
and
T h e ratio d/X can be written in the form:

2d . -1 _
- mv2
v 2 h
expression in which v is the speed associated to A. We deduce that dlX

is the ratio of the time during which the atom moves along a length 2d
to the period of oscillation of a linear oscillator whose energetic levels
are separated by a quantity equal to the kinetic energy of the atom.
When analysing the results obtained by Roberts on the variation of
the accommodation coefficient of helium on a clean surface of tungsten,
Zener shows, that when d = cm is adopted the theoretical represen-
tative curve then reproduces the experimental variations of a.
T h e theory propounded by Zener is very incomplete since, on the
one hand it only gives an expression of a as a function of the tempera-
ture, the accommodation coefficient being only calculated for a normal
incidence; on the other hand the formula contains a parameter d.
However, this theory is interesting for it represents a first stage in the
application of wave mechanics to the problems of the determination
of accommodation coefficients.
D. THEORY
OF JACKSON
Jackson alone (58),then successively with Mott (62) and Howarth (59,
60) established a theory of the accommodation coefficient which is
comparable with that established by Zener (259, 260).
I n a first paper he assumes that the potential energy of a gaseous atom
outside the surface of a solid is null and that it is constant on this surface.
Hence he deduces the calculation of the accommodation coefficient of
helium on tungsten and on nickel. He shows that this accommodation
coefficient first increases with the temperature; in particular, it is null at
absolute zero, then it exhibits a maximum at temperatures which are
different according to the values that are attributed to the molecular
field in the vicinity of a surface.
T h e theoretical calculation of the accommodation coefficient of helium
on tungsten gives for particular values of the constant, results which are
approximately the same as those obtained by Roberts at temperatures
79, 195, and 295K.
I n a second paper written in collaboration with Mott, Jackson assumes
that the mutual potential energy between a gaseous atom and the surface,
at a distance y from this surface, has the following form:
V(y)= ce-bY (129)
He then finds that a good agreement with the experimental results
obtained by Roberts is observed when the value adopted for the constant
b is equal to 9 * lo8cm. Finally, in a latest paper written in collaboration
with Howarth, Jackson suppresses the restriction that the atoms consti-
W I
F. M. DEVIENNE
tuting the solid oscillate with the same frequency. Jackson and Howarth
assume the solid to be a cubic crystal containing N atoms and having
consequently three N ordinary modes of vibration. When admitting
Debyes conceptions he obtain a mean expression of the accommodation
coefficient:
0 being the characteristic temperature of the solid, 0 being equal to

hv/K.
Finally, in applying wave mechanics and in adopting b = 4 . lo8 cm,
they obtain very satisfactory values for the accommodation coefficient of
helium on tungsten.
I t must be remarked that this theory is also incomplete since one is
obliged to choose a constant arbitrarily. This theory allows only a
prediction of the variation of the accommodation coefficients as a func-
tion of the temperature.
It must be noted that the results found by Zener and Jackson are
obtained with different values of the constants of the mutual potential,
since if we adopt the same data in Zeners theory, we must take
l / d = lo9 cm-l, whereas on the contrary, in the theory established by
Jackson and Howarth, we take b = 4 lo8 cm.
9
E. THEORY
OF LANDAU
Landau (82) reproaches the different authors who tried to establish

a theory of accommodation coefficient, with having used a certain number
of simplifying assumptions which constitute an obstacle to the general-
ization of the solution. He points out, in particular, the importance of
impacts on the solid in regard to its specific frequency. He admits that
in certain cases the surface of the solid can be replaced by a liquid
surface by applying the equations of hydrodynamics. He then obtains
the expression of the mutual energy of a gas molecule on the surface
whatever the direction of this molecule may be, in regard to the normal
at the impact point.
Landau distinguishes two cases according to whether the product of
the frequency of impact given by hi27 is small or large in regard to kT;
in other words, according to whether the last energy is weak or great in
regard to the energy of atoms or molecules constituting the gas. H e
remarks that at any temperature if the energy yielded is small in regard
to the energy of the atoms, the problem can be treated in a classical way.
W41
HEATTRANSFER
Low DENSITY
However, this expression of the potential energy of an atom in regard

to the surface depends on the expression adopted for the mutual potential
energy of the molecule in regard to the surface at rest. He thus obtains,
if 5 is the displacement of the liquid molecule in regard to the mean
surface, the following expression for the potential of the gas molecule
in regard to the surface:
dU
U(z - 5) = u - p -
dz
Landau then adopts for U an expression of the same form as the one
previously adopted by Zener on the one hand and Jackson and Howarth
on the other hand:
u= ( 132)
Finally, he obtains as a value of the accommodation coefficient of the

gas on the wall, the expression
a =-
7~
3pmt
(--)2d2c2
~kT
(133)
which shows that the accommodation coefficient increases as a function

of the absolute temperature according to the power 1.5 and in which d
represents the value of the constant in the expression (132) for mutual
potential energy between the gas and the surface at rest; c the sound
velocity; K the Boltzmann constant, h the constant of Planck. This
expression of a can also be written as follows:
a=-(--) 1 2nh2T 3
4pmt d2k@
where p stands for the mass of an atom of crystalline lattice and 0 the
characteristic temperature of the lattice in the case of a solid. This for-
mula is only valid when the frequency of impacts is low. If the product
of the frequency of collisions by h12.r is large in regard to kT, viz.,
if lost energy is large in regard to that of the molecule or the atom
impinging the surface, it can be considered that in the energy transfer
between the gas molecule and the surface of the solid, this latter behaves
as a rigid wall.
Landau obtains the following expression of the accommodation
coefficient:
F. M. DEVIENNE
It can be written in the form:

n~(kT)~
u = 430-
ph3c3
It must be remarked that the accommodation coefficient varies as the

third power of absolute temperature. T h e various results obtained by
Landau on the variation of the accommodation coefficient in terms of
temperature are not in concordance with those obtained by Roberts,
and also by Grilly, Taylor and Johnston.
I n the case of helium, Roberts actually obtains increase of the accom-
modation coefficient on tungsten in terms of temperature, but the
variation is certainly smaller than that represented by the 3/2 power
of absolute temperature; it is afovtiori much smaller than the one indi-
cated by the other formula of Landau, according to which the accommo-
dation coefficient increase proportionally to the third power of absolute
temperature; Landaus theory is interesting though unfortunately it
is not in agreement with experimental results.
F. THEORY
OF DEVONSHIRE
Let us recall, as we said, that Lennard Jones and some of his co-wor-
kers devoted a series of papers to the problems of energy transfer between
an atom of gas and a solid.
Devonshire (29) discusses the case of the transition between two free
states of an atom of gas and applies the result to the calculation of the
accommodation coefficient. He reproaches those who have studied this
theoretical problem before with neglecting the attractive field that
exists between a solid and a gas and with studying only the repulsive
field.
As in the case of the adsorption of an atom of gas by a surface, he
assumes the mutual potential between a gas atom and the surface to be
represented by a Morse potential whose expression is
u = De-2k(z-b-Z1 - 2De-k(z-b-Z) (137)
in which b stands for the distance between the surface atom of the
crystal and the minimum of the curve of potential energy, z the distance
separating a gas atom from the crystal surface, and Z the displacement
of the atom of crystal along the axis of x ; this axis is normal to the sur-
face.
This potential is attractive for relatively large distances to the surface,
and repulsive if these distances are very short. T h e formulae established
P461
Low DENSITY
HEATTRANSFER
by Devonshire are more general than those of preceding theoreticians

which they include as special cases. As we shall further see, they can be
applied to the experimental results obtained by Roberts for helium and
neon on tungsten Devonshire assumes, as the preceding others, that all
gas atoms move perpendicularly to the surface and that the solid can be
considered as constituted by harmonic oscillators that take or deliver
a quantum of energy. T h e latter assumption has been in particular
accounted for by Strachan who shows that the probability for a solid to
exchange a given amount of energy, is much greater for the exchange of
one quantum than for that of two quanta. If F is the total number of
atoms impinging on the surface per unit area per second, the number of
atoms having an energy ranging between E and E d E is equal to+
If, on the other hand, G(E, E +

hv)d(hv) figures the probability of
taking, at every atom impact, a quantum of energy of the solid in the
range of frequency v, v +
dv, the total energy taken from the solid is
given by
F
__
kT,
l: h2v dv
W
0
G(E + hv, E)e-ElkTzdE (139)
where v,,, is the maximum frequency of the solid. I n the same way the
energy yielded to the solid during the same time is given by the expres-
sion
F
KT 1; W
1
h2v dv r ( E + hv, E ) e - E + h v l k T z dE
0
+
in which T ( E hv, E ) d(Gv) represents the probability of energy E hv +
to deliver at every impact a quantum of energy to the solid, in the range
of frequency v, v +
dv.
From the above equations, it can be deduced that the mean energy
taken by a gas atom impinging on the surface is given by
E,'-E ---
2- kT,
1; h2v dv Jm [G(E,E
0
+ hv)
cElkTz
- T ( E + hv, E)e-E+hvlkTa]
dE (141)
But on the other hand, we have
r ( E + hv, E ) = G(E, E + hv)ehVlkTi ( 142)
W I
F. M. DEVIENNE
whence
E,' - E2 = T1-
k2T,3
T2 :J h(hv)2dv Jm
0
G(E, E + hv) dE (143)
if TI- T , is small, it can be written
whence finally
a=
E,'- E,
E, - Ez
=
h
k3T
sp aJ
( h ~dv) ~ G(E,E
0
+ hv)e-E/kr2dE (145)
Function G representing the probability of transition is determined in

terms of the mutual potential energy given by a function of Morse.
Finally, Devonshire obtains for a:
0 being the characteristic temperature defined by:

hv, = kO (147)
T h e calculation has been performed on the assumption that the gas
molecules moved along one dimension only. If we assume that the mole-
cules move along three dimensions and that besides, the momentum
exchange parallel to the surface is unchanged owing to the impacts on
crystal E, - E, must be replaced by 2k( Tl- T,) in the calculation of a,
which consequently diminishes by half the value obtained for the accom-
modation coefficient. On the other hand, if the variations of energy are
proportional to those occurring in the direction of z-axis, a remains
unchanged.
A priori, Devonshire admits that the value given for a by the preceding
expression is slightly larger than the one that should be obtained by
Landau and by Jackson and Howarth. He points out the shortcomings of
Landau's assumption and shows that he obtains the same results as
Jackson and Howarth except for a factor of 8 when he considers the
simple case of a repulsive field.
By applying these formulae in the case of helium and neon on tungsten,
he adopts the value 250K as the characteristic temperature of tungsten.
[W
T h e values K and D which best fit the experimental values found by

Roberts, are in the case of neon on tungsten:
K = 0.75 * los cm-'; D = 493 cal/mole.
For helium, the corresponding values are:
-
K = 1.154 lo8 cm-l; D = 60 cal/mole.
Raines checks Devonshire's theory by using his results on the variation
of the accommodation coefficient of helium on nickel with
K = 0.75 los cm-l; D = 428 cal/mole.
and a corrective factor 1.04.

T h e values found for the accommodation coefficient of helium on
nickel at different temperatures are in perfect concordance with the
experimental values.
So, a glance at Table I11 shows that the maximum divergence between
TABLE I11
Average experimental Values calculated

T (OK) values theoretically
369 0.077 0.077

273 0.071 0.070
195 0.060 0.061
90 0.048 0.047
the experimental values of Raines and the theoretical values ranges

about 2 yo.
It must be remarked that the accuracy in calculating the constants of
the surface field, is certainly small, as the theory of Devonshire applies
only to a one-dimensional model; theoretical values obtained are cer-
tainly larger by a factor ranging from 1 to 2 in relation to the values that
should be actually adopted.
T h e application of the theory of Devonshire is limited to the case
when the surface is clean and when no adsorption of the gas occurs. That
is why it seems that the theory should not be applied to very low tempera-
tures at which adsorption is important.
G.THEORYOF ZWANZIG
Zwanzig (161) reproaches Devonshire for introducing an approxima-
tion which implies that collisions, in which more than one vibrational
P491
F. M. DEVIENNE
mode of the solid lattice is excited, or a single phonon produced, are

improbable.
Zwanzig uses a single model that large amounts of energy can be
transferred from a gas to a solid. He uses a strictly classical model by
considering the interaction of gas molecules with spring-mass models
of crystals, where the interaction force-fields and the characteristic
crystal frequencies are derived from quantum-mechanical considera-
tions. We do not develop the calculations, we give the most important
result, i.e., that the efficiency transfer obtained from these calculations
is much higher than that which is predicted from the Devonshire theory.
H. SHORTCOMINGS THEORIES
OF PRESENT
As we have just seen the problem is far from being solved. T h e most
successful endeavor that was made was undubtably the theory given by
Devonshire. It seems a priori to give the outline of the variation of the
accommodation coefficient of a rare gas on a metal such as tungsten. At
least, it is in agreement with the experiments of Roberts which are
among the most particularly accurate at low temperatures.
However, the results obtained are in contradiction as it was already
pointed out, with those of Grilly, Taylor, and Johnston who find that
the accommodation coefficient of helium on platinum exhibit a maximum
when temperature increases.
I n summary, all the theories are incomplete due to the fact they
neither account for the variation of the accommodation coefficient in
terms of the angle of the incidence of the molecules, nor for the nature
and the amount of the gas adsorbed by the surface. It would be interes-
ting, in particular to determine the influence of the adsorption process
on the value of the accommodation coefficient. It is a very complex
problem which, it appears, may only be solved in very particular cases.
It is closely connected with the surface condition of the solid on which
accommodation occurs.
VIII. Conclusion
I n the course of this study on heat transfer in rarefied gases, we have

developed to a large extent the method of the measurements and the
theoretical calculation of accommodation coefficients.
When we calculate the heat transfer between one or several solid
surfaces and a rarefied gas, either the length of the temperature jump
or the value of the accommodation must be known.
POI
HEATTRANSFER
Low DENSITY
Finally, it is the value of the accommodation coefficient which rules

all the problems of charge exchange in rarefied atmosphere, inasmuch
as it is possible to deduce the value of the temperature jump from that
of the accommodation coefficient by applying the theoretical formulae
that we have previously given.
On account of the inaccuracy of the values given for the accommoda-
tion coefficients and the discrepancies observed between these values
and on account of the shortcomings of the theories concerning the accom-
modation coefficients, all the problems dealing with heat transfer are
far from being solved.
From the experimental standpoint, the first problem which is set is
to obtain accurate and reproducible determinations of the accom-
modation coefficients. T h e important divergences observed in the experi-
mental results do not, as a matter of fact, enable us at present, to say
a prior; what is the vafue of the accommodation coefficient of the gas
molecules on a given surface under determined conditions. We think that
it is necessary to use the technical means offered by modern sciences,
in order to specify with great care, the solid surface on which the
accommodation of the gas molecules occurs: electronic diffraction on
the one hand, and radioactive isotopes on the other, allowing the deter-
mination of the nature and number of adsorbed molecules on the surface.
Consequently, it seems that we could obtain values of accommodation
coefficients corresponding to conditions almost completely determined
and above all well known. So far, the values given by the different
experimenters perhaps correspond to particular macroscopic conditions
but ignore the surface conditions at the molecular scale.
Measurements of accommodation coefficients analogous to those we
have just defined will allow to determine the influence of the tempera-
ture, of the pressure and that of the energy corresponding the incident
molecules.
From the theoretical standpoint, it is obvious that when we obtain
values of the accommodation coefficients corresponding to well-defined
conditions at the molecular scale, it will be easier to establish a theory
giving the variations of the accommodation coefficients in terms of the
different parameters, we have just listed. It will then be possible, taking
into account the history of the surface which is more or less known
to us, to calculate theoretically with good accuracy the values of the
accommodation coefficients for a given gas and a given surface.
However, we do not think it possible to obtain, in a very near future,
by any theory, very accurate values of accommodation coefficients,
except when an important outgassing and very particular conditions
allow this. Yet we should be able to suppress the enormous contradictions
W I
F. M. DEVIENNE
between the data of different investigators in taking up again all the

measurements they have reported, and in taking into account the
importance of the surface conditions. This will allow us to solve in an
experimental and theoretical way, the totality of the problems dealing
with heat transfer in rarefied gases.
LIST OF SYMBOLS
d Characteristic distance, distance T, Impinging molecule temperature
of two plates T, Temperature of molecule after
d T / d n Temperature gradient normal to collision with a surface
to the wall T or T,Gas temperature
Impinging energy of a molecule T W Temperature of solid wall
Energy carried away from surface Va Mean arithmetical speed of mole-
by a molecule leaving the surface cules
Energy of a molecule leaving the V, Most probable speed
surface assuming the molecules Y Ration of specific heats
are in equilibrium at T, x Thermal conductivity
Temperature jump distance V Number of molecules impinging
Boltzmann constant the unit of surface per unit of
Mean free path time
Mass of a gaseous molecule V' Number of molecules that collide
Number of molecules per volume the unit area of the front side of
unit the surface per second
Amount of energy or amount of V" Number of molecules that collide
heat the unit area of the rear side of
Gas constant the surface per second
Absolute temperature 7) Viscosity coefficient
1
-J
"I)
L -
O(s) = erf (s) = error function e-"' dx
4; 0
REFERENCES
I. Amdur, M. C. Jones, and H. Pearlman, J. Chem. Phys. 12, 159 (1944).

2. I. Amdur, J. Chem. Phys. 14, 339 (1946).
3. C. T. Archer, Phil. Mug. [7] 19, 901 (1935).
4. C. T. Archer, Nature 138, 286 (1936).
5. C. T. Archer, Proc. Roy. SOC.A165, 474 (1938).
6. B. Baule, Ann. Physik [4] 44, 145 (1914).
7. 0. Beek, J. Chem. Phys. 4, 680 (1936).
W1
Low DENSITY
HEATTRANSFER
8. 0. Beek, J . Chem. Phys. 4, 743 (1936).

9. 0. Beek, J. Chem. Phys. 5 , 268 (1937).
10. E. Blankenstein, Phys. Rev. 22, 582 (1923).
1 1 . K. B. Blodgett and I. Langmuir, Phys. Rev. 40, 78 (1932).
12. J. G. M. Bremner, Proc. Roy. SOC.A201, 305 (1950).
13. C. F. Brush, Phil. Mug. [5] 45, 31 (1898).
14. J. Chariton and N. I. Semenov, 2. Physik 25, 287 (1924).
15. P. Clausing, Ann. Physik [ 5 ] 7 , 489 (1930).
16. P. Clausing, Ann. Physik [ 5 ] 7 , 521 (1930).
17. P. Clausing, Ann. Physik [5] 7 , 569 (1930).
18. J. 0. Cockroft, Proc. Roy. SOC.A119, 293 (1928).
19. J. H. de Boer, Z. Efektrochem. 44, 488 (1938).
20. F. M. Devienne, Thkse Doctorat, Fac. Sciences, Paris, 1948.
21. F. M. Devienne, Conduction thermique dans les gaz rarifies. Coefficient dacco-
modation. Gauthier-Villars, Paris, 1953.
22. F. M. Devienne, Compt. Rend. 243, 27 (1956).
23. F. M. Devienne, Research on the study of the conditions surrounding a body
moving at high speeds in the ionosphere. Tech. Rept. AF61 (514) 818 (1956).
24. F. M. Devienne, A. Edmond, and E. A. Brun, Etude expkrimentale de la Tempe-
rature darr0t en rigime molCculaire libre. Commun. 9th Congr. A p p f . Mech.,
Brussels, 1956 Vol. 11, p. 214.
25. F. M. Devienne, J . Aeron. Sci. 24, 403 (1957).
26. F. M. Devienne, Frottement et &changes thermiques dans les gaz rarCfiCs.
Gauthier-Villars, Paris, 1958.
27. F. M. Devienne (ed.), Rarefied Gas Dynamics. Macmillan (Pergamon), New
York, 1960.
28. F. M. Devienne, B. Crave, J. Souquet, and R. Cla.pier. in Rarefied Gas Dynamics
(J. A. Laurmann, ed.), p. 362. Academic Press, New York, 1963.
29. A. F. Devonshire, Proc. Roy. SOC.A158, 269 (1937).
30. B. G. Dickins, Proc. Roy. SOC.A143, 517 (1935).
31. S. Dushman, Scientific Foundations of Vacuum Technique. New York, 1949.
32. X. Duval and M. Niclause, Bull. SOC.Chim. France p. 428 (1950).
33. X. Duval and M. Niclause, J. Chim. Phys. 49, 51 (1952).
34. A. E. J. Eggleton, F. C. Tompkins, and D. W. B. Wanford, Proc. Roy. SOC.A213
266 (1952).
35. A. E. J. Eggleton and F. C . Tompkins, Trans. Faruduy SOC.48, 738 (1952).
36. P. Epstein, Phys. Rev. 23, 710 (1924).
37. A. Eucken, Nuturwissenschuften 25, 209 (1937).
38. A. Eucken and A. Bertram, 2. Physik. Chem. B31, 361 (1936).
39. A. Eucken and H. Krome, Z . Physik. Chem. B45, 175 (1940).
40. M. Fappo, Ann. Physik [5] 32, 392 (1938).
41. R. G. J. Fraser, Molecular Rays. Cambridge Univ. Press, London and New
York, 1939.
42. S. Frish, 2. Physik. 84, 443 (1933).
43. S. Frish and 0. Stern, Z . Physik 84, 430 (1933).
44. W. Gaede, Ann. Physik [4] 41, 331 (1913).
45. E. Gehrke, Ann. Physik [4] 2, 102 (1900).
46. M. Gilli, Compt. Rend. 17/10-60 (l9-).
47. H. S. Gregory, PYOC,Roy. SOC.A149, 35 (1935).
48. H. S. Gregory, Phil. Mug. [7] 22, 257 (1936).
F. M. DEVIENNE
49. H. S. Gregory and C. T. Archer, Proc. Roy. SOC.A110, 91 (1926).
50. H. S. Gregary and C . T. Archer, Phil. Mug. [7] I , 593 (1926).
51. H. S. Gregory and C. T. Archer, Proc. Roy. SOC.A121, 285 (1928).
52. H. S. Gregory and C . T. Archer, Phil. Mug. [7] 15, 301 (1933).
53. H. S. Gregory and E. H. Dock, Phil. Mug. [7] 25, 129 (1938).
54. H. S. Gregory and J. H. Marshall, Proc. Roy. SOC.A114, 354 (1927).
55. H. S. Gregory and J. H. Marshall, Proc. Roy. SOC.A118, 594 (1928).
56. H. S. Gregory and R. W. 3. Stephens, Nature 139, 28 (1937).
57. E. R. Grilly, W. J. Taylor, and H. L. Johnston, J. Chem. Phys. 14, 435 (1946).
58. J. M. Jackson, Proc. Cumbridge Phil. SOC.28, 136 (1932).
59. J. M. Jackson and A. Howarth, Proc. Roy. SOC.A142, 447 (1933).
60. J. M. Jackson and A. Howarth, Proc. Roy. SOC.A152, 315 (1935).
61. J. M. Jackson and N. F. Mott, Proc. Roy. SOC.A137, 703 (1932).
62. J. M. Jackson and H. Tyson, Munch. Mem. 81, 87 (1936-1937).
63. R. Jaeckel, Kleinste Drucke, ihre Messung und Erzeugung. Tech. Physik Ser.,
Vol. 9. Springer (Bergmann), Berlin, 1950.
64. H. L. Johnston and E. R. Grilly, J. Chem. Phys. 14, 233 (1946).
65. W. G. Kannuluik and E. H. Carman, p l o c . Phys. SOC.(London) B393, 701 (1952).
66. W. H. Keesom and G. Schmidt, Physicu 3, 590 (1936).
67. W. H. Keesom and G. Schmidt, Physicu 3, 1085 (1936).
68. E. H. Kennard, Kinetic Theory of Gases. New York, 1938.
69. C. Kenty and L. A. Turner, Phys. Rev. 32, 799 (1928).
70. G. B. Kistiakowski and F. Nazmi, J. Chem. Phys. 6, 18 (1938).
71. G. B. Kistiakowski, J. R. Lacher, and F. Stitt, J. Chem. Phys. 7, 289 (1939).
72. M. Knudsen, Ann. Physik [4] 32, 809 (1910).
73. M. Knudsen, Ann. PhyEik [4] 34, 651 (1911).
76. M. Knudsen, Ann. Physik [4] 48, 1 I13 (1915).
77. M. Knudsen, Ann. Physik [5] 6 , 129 (1930).
78. M. Knudsen, Kinetic Theory of Gases; Some Modern Aspects p. 46. Methuen,
London, 1934.
79. M. Knudsen, Kinetic Theory of Gases. London, 1946.
80. A. Kundt and E. Warburg, Phil. Mag. [4] 50, 53 (1875).
81. A. Kundt and E. Warburg, Poggendorfs Ann. 155, 337, 525 (1875).
82. H. G. Landau, Physik Z . Sowjetunion 8, 489 (1935).
83. I. Langmuir, J. A m . Chem. SOC.37, 417 (1915).
84. I. Langmuir, Phys. Rar. 8, 149 (1916).
85. I. Langmuir, J. A m . Chem. SOC.40, 1361 (1918).
86. I. Langmuir, Trum. Furuduy SOC.17, Pt. 111 (1921).
87. J. A. Laurmann (ed.), In Rarefied Gas Dynamics, Vol. I. Academic Press,
New York, 1963.
88. P. Lazareff, Ann. Physik [4] 37, 233 (1912).
89. J. E. Lennard-Jones, Nature 137, 969 (1936).
90. J. E. Lennard-Jones and A. F. Devonshire, Proc. Roy. SOC.A156, 6 (1936).
91. J. E. Lennard-Jones and A. F. Devonshire, Proc. Roy. SOC.158, 242 (1937).
92. J. E. Lennard-Jones and A. F. Devonshire, Proc. Roy. SOC.158, 253 (1937).
93. J. E. Lennard-Jones and C . Stracham, Proc. Roy. SOC.A150, 442 (1935).
94. L. B. Loeb, Kinetic Theory of Gases, 2nd ed. McGraw-Hill, New York, 1934.
95. W. Mandell and J. West, Proc. Roy. SOC.37, 20 (1925).
W41
96. W. B. Mann, Proc. Roy. SOC.A146, 776 (1934).

97. W. B. Mann and B. G. Dickins, Proc. Roy. SOC.A134, 77 (1931).
98. W. B. Mann and W. C . Newell, Proc. Roy. SOC.A158, 397 (1937).
99. W. C. Michels, Phys. Rev. 40, 472 (1932).
100. W. C. Michels, Phys. Rev. 52, 1067 (1937).
101. R. A. Millikan, Phys. Rev. 21, 217 (1923).
102. J. L. Morrisson and J. K. Roberts, Proc. Roy. SOC.A173, 13 (1939).
103. J. L. Morrisson, J . Chem. Phys. 14, 466 (1946).
104. J. L. Morrisson and W. E. Grummitt, J . Chem. Phys. 21, 654 (1953).
105. W. Nothdurft, Ann. Physik [5] 28, 2-137 (1937).
106. R. N. Oliver and M. Farber, Jet Propulsion Lab., California Institute of Technology,
Memorandum No. 9-19, 1950.
107. L. S. Ornstein and W. R. Van Wyck, Z . Physik [5] 7 8 , 734 (1932).
108. R. E. Peck, W. S. Fagan, and P. P. Werlein, Trans. ASME pp. 281-287 (1951).
109. M. PolPnyi and E. P. Wigner, Z. Physik. Chem. 139, 439 (1928).
110. B. Raines, Phys. Rev. 56, 691 (1939).
1 1 I . J. K. Roberts, Proc. Roy. SOC.A129, 146 (1930).
112. J. K. Roberts, Proc. Roy. SOC.A135, 192 (1932).
114. J. K. Roberts, Proc. Cambridge Phil. SOC.30, 74 (1933).
116. J. K. Roberts, Some Problems in Adsorption. Cambridge Univ. Press, London
and New York, 1939.
117. J. K. Roberts, Heat and Thermodynamics. 3rd ed. Blackie, London, 1940.
118. H. H. Rowley and K. F. Bonhoeffer, Z . Physik. Chem. 21, 84 (1933).
119. K. Schafer, W. Rating, and A. Eucken, Ann. Physik [5] 42, 176 (1942).
120. K. Schafer, Fortschr. Chem. Forsch. 1, 61 (1949).
121. K. Schafer, Anales SOC.ESP.Fis.Quim. 48, 361 (1952).
122. M. von Smoluchowski, Sitzber. Akad. Wiss. Wien 107, 304 (1898).
123. M. von Smoluchowski, Wiedermanns Ann. 64, 101 (1898).
124. M. von Smoluchowski, Sitzber. Akad. Wiss. Wien 108, 5 (1899).
125. M. von Smoluchowski, Ann. Physik [4] 35, 983 (1911).
126. M. von Smoluchowski, Phil. Mag. [6] 21, 11 (191 1).
127. F. Soddy and A. J. Berry, Proc. Roy. SOC.83, 254 (1910).
128. F. Soddy and A. J. Berry, Proc. Roy. SOC.84, 576 (191 1).
129. G. V. Spivak, Uch. Zap. Leningr. Gos. Univ. Ser. Fiz. Nauk 5 (38), 7-11 (1939).
130. G. V. Spivak, Khim. Ref. Zh. No. 3, 15-16 (1940).
131. G. V. Spivak, Chem. Abstr. 36, 3995 (1942).
132. J. R. Stalder and D. Jukoff, J. Aeron. Sci. 15, 388 (1948).
133. J. R. Stalder and D. Jukoff, Natl. Advisory Comm. Aeron. Rept. 944 (1948).
134. M.W. States, Phys. Rev. 21, 662 (1923).
135. C. Stracham, Proc. Roy. SOC.A150, 456 (1935).
136. L. Talbot (eds), Rarefied Gas Dynamics. Academic Press, New York, 1961.
137. H. S. Taylor and S. Glasstone (eds.), Treatise on Physical Chemistry, 3rd ed.,
Vol. 2. Van Nostrand, New York, 1951.
138. W. J. Taylor, Phys. Rev. 35, 375 (1930).
139. W. J. Taylor and H. L. Johnston, J . Chem. Phys. 14, 219 (1946).
140. L. B. Thomas and F. Olmer, J . Am. Chem. SOC.65, 1036 (1943).
141. L. B. Thomas and R. E. Brown, J . Chem. Phys. 18, 1367 (1950).
142. L. B. Thomas and R. C . Golike, J. Chem. Ph. 22, 300 (1954).
13551
F. M. DEVIENNE
143. A. B. Van Cleave and 0. Maass, Can. J. Res. 12, 372 (1935).
144. A. B. Van Cleave, Trans. Faraday SOC.34, 1174 (1938).
145. K. S. Van Dyke, Phys. Rev. 21, 250 (1923).
146. K. S. Van Dyke, Phys. Rev. 21, 239 (1923).
147. L. von Ubisch, Appl. Sci. Res. B2, 364 (1951).
148. Wartenstein, J . Phys. Radium 4, 281 (1923).
149. S. Weber, Ann. Physik [4] 54, 325 (1917).
150. S. Weber, Ann. Physik [4] 54, 437 (1917).
151. S. Weber, Commun. Kamerlingh Onnes Lab. Univ. Leiden 246b (1937).
152. S. Weber, Kgl. Danske Videnskab. Selskab, Mat.-Fys. Medd. 16 (9) (1939).
153. S. Weber, Kgl. Danske Videnskab. Selskab, Mat.-Fys. Medd. 19 (11) (1942).
154. S. Weber, Kgl. Danske Videmkab. SeZskob, Mat.-Fys. Medd. 24 (4) (1947).
155. F. R. Whaley, J . Chem. Phys. 1, 186 (1933).
156. M. L. Weidmann, Trans. A S M E 68 57 (1946).
157. R. W. Wood, Phil. Mag. [6] 30, 300 (1915).
158. R. W. Wood, Phil. Mag. [6] 32, 364 (1916).
159. C. Zener, Phys. Rev. 40, 178 (1932).
160. C. Zener, Phys. Rev. 40, 375 (1932).
161. R. W. Zwanzig, J. Chem. Phys. 32, 1173 (1960).
Heat Transfer in Non-Newtonian Fluids
A. B. METZNER
University of Delaware, Newark, Delaware
I. Scope . . . . . . . . . . . . . . . . . . . . . . .. . 357
11. Fluid Property Considerations . . . . . . . . . . . . . . 358
A. Classification of Fluids: Rheological Properties . . . . . . 358
B. Thermal Properties of Non-Newtonian Fluids . . . . . . 364
C. Summary . . . . . . . . . . . . . . . . . . . . . . 365
111. Heat Transfer in Steady Ducted Flow Fields (The Internal
Problem) . . . . . . . . . . . . . . . . . . . . . . .. 366
A. Laminar Flow Conditions . . . . . . . . . . . . . . . 366
B. Laminar Flow with Interrial Heat Generation, Round Tubes 376
C. Laminar Flow with Internal Heat Generation, Other Geom-
etries .. . .
. . . . . . .. . . . . . . . . . . . . 379
D. Transition and Turbulent Flow Conditions. . . . . .. 38 1
E. S u m m a r y . . . . . . , . . . . . ..
. . . . . . . . 388
IV. Boundary Layer Problems , . . . . . . . . . . . . . . . 389
A. Laminar. . . . . . . . . . . . . . . . . . . . . . . 389
B. Turbulent. . . . . . . . . . . . . . . . . . . . . . 392
V. Miscellaneous Heat Transfer Problems . . . . . . . . . . . 392
Notation . . .
. . . . .. . . . . . . . . . . . . . . . 392
References . , . . . . . . .... .. . . ..., .. . 394
I. Scope
T h e productive period of research into nonlinear rheological phenom-

ena (as distinguished from studies of linear viscoelasticity and related
subjects which have led, to date, at best to only minor engineering
applications) did not begin until the late 1940s save for a very small
number of earlier pioneering contributions, largely during World
War 11. It is not surprising, then, that useful engineering studies which
deal with transport processes in non-Newtonian fluids have been under-
way for little more than a decade. Earlier heat transfer reviews (65, 127)
therefore served primarily to separate a very small number of useful
13571
A. B. METZNER
contributions from the substantial mass of trivial, misleading or even

incorrect concepts and interpretations available at that time. As a result,
no further mention of studies prior to about 1955 will be made unless
the contribution is judged to be of permanent value. Conversely, an
attempt is made to discuss, at least briefly, all of the more recent publi-
cations so that the present bibliography will constitute an encyclopedic
compilation of work on heat transfer to non-Newtonian fluids since 1954,
as well as serving the primarily purpose of describing the current state
of the art.
T h e reader who may be interested in corresponding summaries or des-
criptions of fluid mechanical studies dealing with non-Newtonian
materials may wish to refer to the reviews or books by Bernhardt (4),
Wilkinson (127), Fredrickson (30),and Metzner (65, 68).
11. Fluid Property Considerations
OF FLUIDS:
A. CLASSIFICATION PROPERTIES
RHEOLOGICAL
1. General Considerations
An engineering approach suggested several years ago (25, 67) now
appears to have been generally adopted. T h e following paragraphs are
not intended to present this in detail but only to provide the general
survey of rheological behavior necessary for an understanding of trans-
port processes in non-Newtonian fluids; the reader may wish to refer
to discussions in the areas of fluid mechanics and rheology in order to
fill in necessary details1
The fluid classification system of interest is based upon the form of the
constitutive equations used to describe the properties of real fluids
quantitatively. These equations relate the components of either the
total stress tensor T or of the deviatoric stress tensor P to the corre-
sponding components of the deformation rate tensor d and any other
relevant kinematic variables such as the time or history of the deforma-
tion rate process. Substitution of an appropriate constitutive equation
into the stress equations of motion gives a set of equations analogous
The classification system is discussed in detail by Metzner (68) and the reader is
referred to that work for a more complete description. Other specific discussions of
fluid behavior which may be especially helpful in the present context may be found in
the papers of Markovitz (61-63) and White and Metzner (122, 123) and in the books
by Reiner (88) and Fredrickson (30). Bird et al. (7) present solutions of many engineering
problems in which simple non-Newtonian constitutive equations are employed.
W I
HEATTRANSFER FLUIDS
IN NON-NEWTONIAN
to the Navier-Stokes equations for Newtonian fluids. Solution of such

equations for chosen initial and boundary conditions, as in the analogous
Newtonian case, leads to a solution of the flow problem at hand.
Table I summarizes the several fluid classifications. T h e simplest pos-
sible category, that of purely viscous materials, is seen to include
TABLE I
TYPES
OF FLUIDBEHAVIOR
Category I:
Purely Viscous Fluids
P,j = pdti
p = p (I, 11, 111)
Special cases
(a) p constant (Newtonian)
=
(b) p decreasing function of the invariants of the strain rate tensor I, 11, 111;
=
includes materials commonly termed pseudoplastic, Bingham plastic,
or shear-thinning.
(c) p = increasing function of I, 11, 111; commonly termed dilatant or shear
thickening.
Category 2:
Time-dependent Fluids
Psj = P du
= (I, 11, 111, v * )
Category 3:
Viscoelastic Fluids
k 0
P<i ~AYA~)I
i=-w
~
Category 4
More complex materials
Newtonian fluids as its simplest special case. T h e properties of both

Category 2 and Category 3 fluids are dependent upon the strain rate
history, as well as upon its instantaneous value. Fundamentally, there is
perhaps no major reason to subdivide such materials into two separate
groups. However, the engineering advantages of such a subdivision are
compelling ones: the time-dependent systems of Category 2, frequently
described as either thixotropic or rheopectic, normally appear to be sub-
W I
A. B. METZNER
stantially free of elastic effects. For example, most such fluids exhibit
little or no tendency to return to their original configuration when an
applied stress is released. By contrast, viscoelastic systems develop
pronounced normal or elastic stresses even under conditions of steady
laminar shearing flow, and exhibit the phenomena of strain recovery and
finite rates of stress relaxation.2
Solid and molten polymers and their solutions are usually strongly
viscoelastic. Category 2 (thixotropic and rheopectic) behavior is common
to slurries and suspensions of solids or colloidal aggregates in liquids. In
this case it is frequently found that the time-dependent function is only
of significance for relatively short times of deformation. As a result,
time dependent systems may frequently be treated as purely-viscous
materials as a good approximation; this is especially likely to be the case
if the fluid passes through a pump or through fittings in which it is
vigorously sheared before entering the particular piece of equipment
being designed.
Corresponding to this simplification in the case of time-dependent
systems, the presence of viscoelastic effects is also frequently irrelevant
in solution of engineering problems. For example, in the case of steady
isothermal flow through round tubes any elastic stresses arising will not
change in magnitude with axial position along the tube, beyond the inlet
region. Thus they do not influence either the velocity profile or the
pressure drop in this portion of the duct. Therefore from the viewpoint
of these problems it is irrelevant whether one is designing for such
fluids or simply for purely viscous non-Newtonians. Under flow condi-
tions which are nonsteady, from a Lagrangian viewpoint, the same is
not true of course: the behavior of viscoelastic materials in the entrance
region of a duct or under turbulent flow conditions, to take two specific
examples, differs grossly from that of purely viscous materials.
No problems involving heat transfer to either time-dependent or
viscoelastic materials, under conditions in which these complications
may be of significance, appear to have been studied as yet. I n fact, any
such studies would have been premature in the sense that experimental
methods for measuring the properties of these materials under engi-
neering conditions have only been perfected recently in the case of time-
dependent systems and are still under evaluation for viscoelastic fluids.
As a result constitutive equations which are likely to portray the fluid
properties adequately, yet with sufficient simplicity to enable their use
The components of the stress tensor are functionals of the deformation gradients
yDr in agenera1 sense, for viscoelastic fluids, as discussed by No11 (81) and reviewed by
Rivlin (92) and Fredrickson (30).
HEATTRANSFER FLUIDS
IN NON-NEWTONIAN
in reasonably sophisticated engineering problems, are also only in their

early stages of development for these system^.^
It is thus seen that few materials other than gases or low molecular
weight liquids (which are Newtonian) actually exhibit purely viscous
properties under all conditions. However, many time-dependent systems
approximate such behavior closely under processing conditions and,
in other cases, the design in question must be carried out so as to
accomodate the limiting conditions of purely viscous behavior. Thus,
the approximation of purely viscous non-Newtonian behavior is of
rather broad interest in applications in which slurries are processed.
Similarly, under steady, laminar shearing flow conditions the proper-
ties of viscoelastic fluids which are important are frequently just those
of purely viscous non-Newtonians. I n this latter case it should be noted,
however, that flow conditions which are actually perfectly steady, from
a Lagrangian viewpoint, are seldom encountered under conditions of
nonisothermal flow as when heat is being transferred. Therefore, the
approximation to purely viscous behavior is not as good for visco-
elastic fluids as it is for time-dependent systems under laminar flow
conditions; it cannot, of course, be made at all under conditions of
turbulent flow for viscoelastic fluids.
All available literature on heat transfer in non-Newtonian fluids
considers only purely viscous fluids, although this restriction is seldom
stated. It should be clear from the preceding discussion what the
limitations of this assumption are.
2. Constitutive Equutions f o r Description of Viscous Fluids

As discussed elsewhere (I, 68) the following four equations, all
empirical or at best only semitheoretical, have enjoyed considerable use.
T o facilitate their application to the usual two-dimensional flow
problems having only one nonzero velocity component, they are written
in this form.*
Constitutive equations and experimental results related to them are discussed by

Hahn et al. (40), Denney and Brodkey (23) and Ritter (91) for time-dependent systems
and by Markovitz and Brown (63), Tanner (104), Shertzer and Metzner (102), Ginn
and Metzner ( 3 6 ) and White and Metzner (122, 123) for viscoelastic fluids.
Properly invariant forms of several of these equations are available [see, for example,
Fredrickson (30)].However, as the equations have not been applied to solution of fluid
mechanics problems having more than one nonzero velocity component and for which
experimental results free from miscellaneous aberrations, such as viscoelasticity, were
available, there is as yet little reason to accept the validity of such generalizations of
these equations.
A. B. METZNER
a . The Power Law.

712 = Kr:z
provides a simple (two-constant) empirical equation; at moderately

high shear rates it is frequently valid over several decades of shear rate
and may serve as an excellent approximation formula. It always fails
at the extremes of both very low and very high shear rates in that it
does not predict the behavior (usually Newtonian) observed at these
extremes.
b. The Ellis Model. T h e Ellis model attempts to correct for the
deficiency of the power law at low shear rates by addition of a Newtonian
term:
and, as a result, is more useful when low to moderate shear rates are of
interest, as in the case of highly viscous systems.
c. The Bingham Plastic Equation.
is frequently employed with slurries which exhibit either a true or

apparent yield stress T ~ .
d . The Three-Parameter Eyring-Powell Equation.
T~~ = +-
1
B
sinh-l(I,,/AEp) (4)
is inconvenient to use analytically because it is not explicit in shear

rate but is the only equation of this group which may be used to predict
correctly the fluid properties at both extremes of high and low shear
rates. T h e intermediate non-Newtonian region also appears to be
approximated more closely by this equation than by most others; as a
result its use may be strongly recommended when one requires an
equation valid over wide ranges. T h e increased complexity of the equa-
tion may, of course, not be significant in those problems which require
numerical solutions in any event.
Some brief discussion of the implications of the fluid properties under
consideration may be helpful in developing a better understanding of
heat transfer phenomena. Using the power law, Eq. ( I ) , as a frame of
reference, most non-Newtonian slurries, as well as ordinary polymer
~3621
HEATTRANSFER
IN NON-NEWTONIAN
FLUIDS
solutions and molten polymers, exhibit power law exponents (also
termed flow behavior indices) of less than unity, the numerical value
decreasing from unity toward zero as the concentration of the material
added to a solvent or other continuum increases. Thus, values of the
flow behavior index of the order of 0.5 are common and values as low
as 0.02 have been reported (see, for example, Metzner and Reed (73)
for a fairly extensive compilation). Such fluids are termed pseudo-
plastic or shear thinning, as the resistance to deformation, if measured
by an apparent viscosity ( p a = T ~ ~ / . T ~decreases
~), with increasing
deformation rate.
On the other hand, highly concentrated slurries and polymeric solu-
tions which are highly structured or made up using a very viscous
solvent (13, 69) may exhibit, over limited ranges of deformation
rate, highly dilatant behavior, i.e., flow behavior indices well in excess
of unity. I n this latter case the resistance to deformation as measured
by an apparent viscosity increases rapidly with increasing deformation
rate, leading, in extreme cases, to a fracture of the continuum into
discrete particles if sufficiently high stresses are applied.
The consequences of flow indices approaching zero are flatter velocity
profiles (as compared to the Newtonian parabola) under laminar flow
conditions in well-developed ducted flow fields, and a comparative
insensitivity of pressure drop to flowrate. Correspondingly, as the flow
index approaches infinity the predicted velocity profile becomes peaked
(triangular) and small increases in flowrate are accompanied by very
large increases in pressure drop.
An apparent omission in the above listing of equations must be justi-
fied. I n recent years several applied mathematics journals have published
large numbers of papers dealing with heat transfer and boundary layer
analyses using the constitutive equation of a Reiner-Rivlin fluid.6
The development of this equation clearly represents a historically-
significant event in nonlinear rheology but it is quite evident that no
real materials have been found to date which behave in the manner
predicted. As a result, it seems reasonable to treat this development as one
of historical interest only (116) and to disregard the solutions of applied
problems which have been based on this constitutive equation.6
For discussions of this equation, see Rivlin (92), Markovitz (61), and Fredrickson (30).
The author appreciates the strong indictment offered by these comments and would
be pleased to reconsider them in the light of any evidence which would indicate them
to be incorrect. However, as even the mathematical problems encountered when handling
more sophisticated constitutive equations are distinct from those involved with Reiner-
Rivlin fluids it is simply not clear at present that studies of engineering problems such
as boundary layer development, for such fluids, are of any value.
fW
A. B. METZNER
B. THERMAL
PROPERTIES
OF NON-NEWTONIAN
FLUIDS
I , Thermal Conductivity
T h e conductivity of suspensions of inert solids in liquids may be
estimated by means of the Maxwell (64) equation under the assumed
conditions of no relative motion of solid and fluid, symmetrical particles
and concentrations sufficiently low to preclude significant interparticle
contacts:
which reduces (65) to
when kp > k, .
As discussed by Orr and DallaValle (83), Jefferson et al. (51), and
Thomas (109), a considerable body of experimental evidence is available
in support of these predictions. However, above volume fractions x,,
of the suspended particles of about 0.10 modifications dependent on
particle shape are necessary. Such modifications as are available appear
to represent only marginal improvements in that significant errors may
still be incurred at high concentration levels; accordingly direct experi-
mental measurements of the thermal conductivities of slurries appear
to be required except in the case of dilute suspensions.
Obviously the assumed model of a continuum conduction process
would not apply under turbulent conditions if the particles were large
or if the particle and fluid densities were not identical; in this event the
concept of a homogeneous non-Newtonian continuum would also be
inapplicable for analysis of the fluid mechanical considerations, however.
While discrete suspensions of inert solids may occasionally exhibit
highly non-Newtonian behavior, more usually the non-Newtonian fluid
consists of a continuous solvent phase in which the added polymeric
molecules or solid particles have become at least partially dissolved or
solvated. No methods for predicting the conductivities of these systems,
in terms of the conductivities of the individual components, are available.
While the numerous experimental studies of heat transfer rates to be
discussed later usually included direct measurements of the thermal
conductivity no general conclusions may be drawn from this work in
view of the fact that the usual experimental conditions involved dilute
suspensions or solutions made up of components having nearly identical
conductivities.
W I
HEATTRANSFER FLUIDS
I N NON-NEWTONIAN
Since the momentum transport coefficient of purely viscous non-

Newtonians is a function of the fluid deformation rate (through the
invariants of the strain rate tensor, Table I), a phenomenon which
presumably is at least in part due to the alignment and/or disentangle-
ment of the particles or macromolecules involved, it has been suggested
that thermal conductivity might similarly be a function of fluid deforma-
tion rate, and, in the case or oriented solid polymers such an effect has
been found (84, although its magnitude (1-4y0) was quite small. In the
case of fluids, however, the agreement between the rates of heat transfer
predicted on the basis of conductivities which were measured in a cell
in which the fluid was stagnant and the rates observed under laminar
flow conditions, as will be discussed later, constitutes strong circum-
stantial evidence to the contrary. I n addition the one direct study (27)
of the effect of shear rate on thermal conductivity led to a similarly
negative conclusion. Finally, mass diffusivities in non-Newtonian fluids
have been studied by a number of independent investigators and also
found to be independent of deformation rate (3, 18, 48). In all cases it
must be admitted that effects of the order of 1-4% would not have been
measurable in these experiments, hence they only show conclusively the
absence of major effects.
It is concluded that thermal conductivities of non-Newtonian materials
will have to be evaluated experimentally unless all components separately
exhibit substantially identical conductivities or, as in the case of suspen-
sions of discrete particles, Eqs. ( 5 ) or (6) or one of their modifications
may be used. Presently available data suggest the absence of measurable
effects of fluid deformation rate upon thermal conductivity although,
following the fluid structural considerations of Burrow et a/. (13) and
the effects noted in solids, one may expect to find small anisotropies in
highly viscous systems.
2. Heat Capacity
I n the case of inert suspended solids a weighted mean may obviously
be used; in all other cases experimental determinations are required
unless the heat capacities of the separate components are virtually identi-
cal or the differing component is present in only negligibly small quanti-
ties.
C. SUMMARY
Most non-Newtonian fluids may be expected to exhibit either time-
dependent or viscoelastic rheological properties, as defined in Table I.
W I
A. B. METZNER
Under fully developed, steady laminar flow conditions, such complica-

tions are fortunately of little or no import, in many instances, and these
materials may be assumed to behave as purely viscous fluids from the
viewpoint of heat transfer rates. T h e same appears to be true of time-
dependent systems under conditions of turbulent flow. These justifiable
simplifications are important because no heat transfer literature exists
for non-Newtonian materials having properties more complex than those
of purely viscous fluids.
Equations ( 1)-(4) represent relatively simple empirical approxima-
tions to the behavior of purely viscous fluids and have been used (with
the stress equations of motion and appropriate energy balance equations)
to obtain solutions to heat transfer problems.
Equations (5) and ( 6 ) , or one of several available modifications of
these, may be used to estimate the thermal conductivities of dilute
suspensions. T h e conductivities of other non-Newtonian fluids must be
determined experimentally, as must the specific heats of these systems,
unless arbitrary estimates are likely to be sufficiently precise.
Thermal and mass diffusivities in non-Newtonian fluids, unlike the
momentum diffusivity, do not appear to be appreciably dependent upon
the fluid deformation rate.
111. Heat Transfer in Steady Ducted Flow Fields (The Internal Problem)
FLOWCONDITIONS
A. LAMINAR
1. Heat Transfer Inside Round Tubes
a. Constant Wall Temperature Conditions. Two fairly general and
complementary approaches of practical value will be discussed in detail
in this section. I n the first, solutions are obtained to the energy balance
equations coupled with the equations of motion to allow for temperature-
dependent material properties; specific choices of the constitutive equa-
tions and of the temperature-dependency of the physical properties
must obviously be made. Depending on these choices the results may be
of fairly general utility or quite restricted. In the second case, the
asymptotic Leveque (56) approach is extended to non-Newtonian
systems. This approach is more general than the first in that no specific
rheological models need be chosen but it is limited by the fact that
corrections for temperature-dependent properties are only made empiri-
cally, hence more approximately. I n addition to these two rather general
attacks, a large number of papers have appeared in which neither the
temperature-dependency of the fluid properties nor general rheological
~3661
HEATTRANSFER FLUIDS
IN NON-NEWTONIAN
relationships have been considered. These occupy the bulk of the

literature in this area; they are tabulated herein to provide a measure
of completeness to the review but will not be discussed in detail.
(i) Solutions of the coupied energy and momentum equations. Earlier
work in this area has been largely eclipsed by the recent contributions
of Christiansen and Craig (15) and Christiansen and Jensen (16).
T h e assumptions employed by Christiansen and Craig and which serve
to define the problem solved are as follows:
(i) T h e velocity profile of the fluid entering the heat transfer section
is that of isothermal, well-developed flow.
(ii) T h e fluid density, thermal conductivity and heat capacity are
assumed to be independent of temperature.
(iii) Radial motion of the fluid and axial heat conduction are neglected.
(iv) There is negligible energy generation within the fluid.
(v) T h e flow properties of the fluid correspond to that of a tempera-
ture dependent purely viscous power law fluid obeying the equation
T~~ = K[I',,exp(AH/HT)ln (7)
in which AHIR, K , and n are assumed to be temperature-independent
constants. This equation is shown to represent experimental data for
a number of fluids very well over the temperature ranges normally
encountered in heat exchangers; obviously for fluids for which this
which this is not the case an alternate analysis will be required. T h e
limitations of Eq. (7) with respect to shear rate are those of the usual
power law formulation, Eq. (l), hence it will depict fluid properties most
accurately at moderate to high deformation rates (flow rates).
Natural convection effects are ruled out by the second and third
of the above assumptions; under these conditions the equations of
motion reduce to
Neglecting the dependence of aPlaz and of 7 1 2 upon z as an approxi-

mation Eq. (8) may be integrated successively to give
u =0.5D(~,/K)~/~l, (9)
- 2w I ,
rrD2p I 2
in which
A. B. METZNEH
and
are dimensionless integrals.

Under the conditions postulated by the previous assumptions the
energy equations [see, for example, Bird et al. (7)] reduce to
which, when combined with Eq. ( 9 ) and put into dimensionless form
yields
in which T odenotes the dimensionless temperature ( T - Ti)/( Tu, - Ti).

Numerical integration of Eq. (11) was carried out for boundary
conditions of constant wall temperature to give a result of the form
""
FIG. I . Nusselt number-Graetz number relationships for fluids having temperature

dependent properties. Taken from Christiansen and Craig ( 15). (a) Newtonian Fluids.
[3681
HEATTRANSFER FLUIDS
IN NON-NEWTONIAN
U"
FIG. 1 . (b) and ( c ) non-Newtonian Fluids.

A. B. METZNER
but the results also indicated the influence of the term ( T , - T,)!T,
to be negligibly small under the usual conditions. Therefore, deleting
this term one obtains
u 4GZ * n, +(W1 (13)
in which
Equation (13) is depicted graphically in Fig. 1 a-c. As would be expected,

the flatter velocity profiles of non-Newtonians having flow behavior
indices below unity lead to increased heat transfer rates because of the
steeper velocity profiles near the wall, the usual controlling region.
T h e experimental data obtained to verify Eq. (13) and Fig. 1 (1.5)
were quite limited as to fluid type, hence it is premature to judge the
utility of Eq. (13) except to note that it appears to represent a distinct
improvement over the prior art, particularly at moderate to high
Graetz numbers and moderately low temperature differences. Three
limitations remain
(i) No curves are given for negative values of the modulus
dH/R((l/Ti) - ( I / T , ) ) ; as a result predictions are possible only under
conditions of heating, for the usual case in which the fluid consistency
decreases with increasing temperature.
(ii) Significant errors may be incurred, especially at high heat fluxes,
because of natural convection effects.
(iii) Some fluids, especially those exhibiting either true or apparent
yield values (the term T~ in Eq. (3)) may deviate considerably from
the assumed power law behavior (Eq. (7)).
T h e third of these problems is usually not nearly as serious as might
be expected at first glance since the principal region of importance in
defining both the velocity and temperature fields is in the immediate
vicinity of the tube wall. Providing Eq. (7) defines the fluid properties
well in this region, its invalidity at lower shearing stresses is of lesser
consequence. Nevertheless, some problems remain; to this end an
analysis similar to the above but in which a temperature-dependent
generalization of the Eyring-Powell constitutive equation (Eq. (4)) was
employed instead of Eq. (7) has been published by Christiansen and
Jensen (26). Only data obtained using dilute polymer solutions were
employed to check this latter analysis; since Eq. (7) is a good approxima-
tion for these materials no significant improvement over Eq. (13) was
observed as expected. I n other cases, however, this work should prove
to be of significant value.
WOI
HEAT TRANSFER FLUIDS
IN NON-NEWTONIAN
Attention will now be turned to an alterate approach which has led

to some consideration of the effects of natural convection and of heat
transfer under conditions of cooling as well as heating.
(ii) Extension of the Leveque Approach to Non-Newtonian Systems. At
flow rates high enough to confine the thermal boundary layer to a region
sufficiently close to the tube wall to enable the use of a linear velocity
profile in this region, one obtains [see, for example, Pigford (86)]
For the well-developed isothermal flow of any non-Newtonian fluid

inside round tubes the wall shear rate is given by7
r w -
3n'
--.-.--
+ 1 8V
4n' D
Combination of Eqs. (14) and (15) gives
NNu = 1.756t(NG,)t (16)

which is valid for NGz > 20 and n' > 0.10. For NGz < 20 and/or
n' < 0.1
NNU = d'(NNu)Newt (164
wherein (N,JNe,, denotes the value of the Nusselt number for New-
tonian fluids at the same Graetz number; A* is given as a function of
n' and Graetz number (74).
I n the event that fluid properties do not change significantly over the
temperature ranges involved, Eqs. (16) and (16a) are of more general
applicability than is Eq. (13) since no specific choice of any constitutive
equation was made. While non-Newtonian systems quite frequently are
'Standard references in fluid mechanics may be referred to for a development of

this relation; see, for example, Bird et al. (7).
As shown in the Notation, the definition of the flow behavior index n', while analogous
to that of n, involves only the average (integrated) flowrate or mean velocity. This is a
real advantage when calculations involving only flow through round tubes are being
considered, as it is not necessary that a constitutive equation valid over the entire tube
cross section be used for entirely rigorous results. For the special case in which the
power law is valid over the entire cross section, n and n' become identical. In matter
of fact the numerical difference between the two in other cases is usually too small to
evaluate, so the practical difference between the two terms is not great. Conceptually
the difference is finite, however, and since equations based on n' are the more rigorous
they should be used whenever a choice is possible.
Similar comments apply to n", introduced in Eq. (19).
P711
A. B. METZNER
less temperature sensitive than are Newtonian fluids of comparable

viscosity levels, these effects are still significant and, in fact, at high
temperature differences these effects overshadow the importance of
non-Newtonian behavior (71). Thus, to be applicable to real systems
showing temperature-dependent properties Eqs. (1 6) and (1 6a) must be
corrected to account for radial variations in fluid consistency and for
natural convection effects introduced by fluid density variations. Using
the framework of previous work in Newtonian systems, a consistency
ratio correction ( y w / y B ) 0 . 1has
4 been introduced to account for the former
effect (71, 74) and a term containing the Grashof number for the latter
(71, 82). Oliver and Jenson suggest use of the equation
which fits the critical data obtained recently by these workers more
precisely than an equation proposed earlier by Metzner and Gluck.
In Eq. (17) the physical properties appearing in the Grashof and
Prandtl number terms are evaluated at the wall conditions of both
temperature and shear rate: since the apparent viscosity of pseudo-
plastic or shear thinning non-Newtonians decreases with increasing
stress the fluid would normally be least viscous in this region, hence
any natural convection effects would tend to be maximized close to the
wall in such fluids unless the temperature gradients are of such a
magnitude and direction as to completely offset the normally strong
effects of shear stress.
T h e experimental data available in support of Eq. (17) are of high
precision but quite limited in both quantity and range of variables
covered. Additionally, the omission of the 61 or Ab correction term from
Eq. (17) means that it does not reduce properly to Eq. (16) or (16a)
under conditions of negligible natural convection effects. Finally,
Christiansen and Craig have shown the use of a Sieder-Tate type of
correction factor ( y w / y B ) 0 . 1to
4 be incorrect in principle: instead, a
function of this form must incorporate some dependence on Graetz
number. In its present form, they show that for one particular set of
conditions it is approximately correct at Graetz numbers of 500, 12%
too high at NGz = 50 and 11 yo too low at N,, = 5000. Thus it is clear
that Eq. (17) can only be recommended under conditions when the more
specific approaches of Christiansen and Craig, and Christiansen and
Jensen are inapplicable.s In view of the importance of natural convection
T h e use of the 0.10 power in place of the Sieder-Tate 0.14 exponent has been suggested
by I<reith and Summerfield (55). T h e conclusions drawn from the Christiansen-Craig
analysis suggest that no careful consideration need be given to the question of which
is more nearly correct, as neither can be of general value under all conditions.
13721
HEATTRANSFER FLUIDS
IN NON-NEWTONIAN
effects and radial consistency variations this is likely to be the case

frequently however.
I t is relevant to pause and to assess the relative importance of
the three corrections to the equations for heat transfer rates to
isothermal Newtonian fluids: in the data available the maximum
effects due to natural convection, temperature dependence of rheological
properties and non-Newtonian properties are 300 yo, 43 yo and 30 yo,
respectively. I t is clear that the non-Newtonian effects are of least
significance and that further studies must necessarily come to grips with
the difficult question of rigorous consideration of effects simultaneously
due to natural convection and to temperature dependent rheological
properties if they are to be of any value to engineers.
In summary of these papers, the results of Christiansen and Craig (25)
and Christiansen and Jensen (26)provide accurate predictions for heating
at high Graetz numbers or modest temperature differences (so that
natural convection effects will be unimportant) and are recommended
for use under these conditions. Under the more usual conditions when
natural convection effects are important or when cooling rates are to be
predicted, used of the modified Leveque approach is recommended,
e.g., Eq. (16) or (16a) (if natural convection is unimportant) and Eq.
(17) or analogous equations given by Metzner and Gluck when natural
convection effects are to be considered. Unfortunately, sufficient results
are not available to make a clear choice possible in this latter case, nor
to assess the probable accuracy of the prediction unless the conditions
used fall within the experimental ranges studied by either Metzner and
Gluck or Oliver and Jensen, in which event predictions good to 10-1576
are to be expected.
(iii) Other studies. Using the above analyses as points of departure, the
additional contributions available may be reviewed briefly as follows.
Thomas (106) has employed equations of the form of Eq. (16) and
( 1 6a) to interpret heat transfer measurements on slurries which approxi-
mate Bingham plastic properties. T h e term @(=[(3n + l)/4nlA)in
terms of the Bingham plastic parameters of Eq. (3) may be shown
(47, 66) to be
Thomas applied a Sieder-Tate correction similar, but not identical to,

that of Eq. (17); deviations of the data from the predicted relationships
were greatest for the least viscous fluid, indicating that some correction
for natural convection effects similar to that of Eq. (17) would have been
P731
A. B. METZNER
helpful. T h u s these results generally support equations of the form of

Eq. (17) and further suggest the need for an accurate analysis of the
effects of natural convection.
Hanks and Christiansen (43) have considered pressure drop relation-
ships for non-isothermal fluids flowing through round tubes; radial
variations of fluid consistency are considered rigorously. As natural
convection effects are neglected the analysis is likely to be of greatest
value at relatively high flowrates. Gill (35) published an interesting
method of analysis which avoids the excessively laborious computations
required by eigenvalue problems but which apparently is limited to
constant (i,e., temperature independent) properties. Hirai (46, 47)
repeated earlier approximate analyses in his first two papers and
presents extensive data on heat transfer rates in slurries in the two later
ones. Perhaps because the slurries used were difficult to maintain in a
homogeneous state the data scatter excessively.
Murakami (79) and Charm (14) developed approximate procedures
which enable simple and fairly reliable estimates of the temperature
profile^.^ Lyche and Bird (60) and Whiteman and Drake (125) present
extensions of the constant-properties Graetz solution to power law
fluids; this is a much more restrictive approach than that used to develop
Eqs. (16) and (16a). Furthermore the analysis of Lyche and Bird is
restricted to low Graetz numbers and that of Whiteman and Drake is
inaccurate except when the non-Newtonian effects are small.
b. Constant Heat Flux Conditions. T h e papers by Grigull(38), Schenk

and Van Laar (98),and Bird (6) all consider power law fluids; Schenk
and Van Laar also considered a special case of the Eyring-Powell
equation, Eq. (4), and included internal heat generation due to viscous
dissipation. All assume that effects due to temperature dependence of
the physical properties are negligible; the applicability of the results
to real problems is therefore only very limited.
No experimental data appear to have been taken under conditions of
constant heat flux.
2. Heat Transfer in Other Geometries.

Flat plate heat exchangers, frequently employed in the plastics
industry, have been considered by Crozier et al. (22). Using the Leveque
More rigorous estimates are possible, of course, by referring to the original theses
upon which the Christiansen and Craig, and Christiansen and Jensen papers are based.
W41
HEATTRANSFER FLUIDS
I N NON-NEWTONIAN
equation, Eq. (14), as a point of departure for NGz > 20, noting that
the shear rates at the surfaces of these ducts is given by (78,96, Z24)lo
r, = ( 2n;nf )
and inserting the hydraulic or equivalent diameter
D, =4 (area)/(perimeter) = 2B,,
one obtains
he.mDe -
--
k
(
1.86 De2GC,
kL )*(a)*
as compared to
kmD D2GC tar
-k
= 1.62(*)
for round tubes, wherein 6 = (372 +

1)/4n (as before) and 6 =
(2n + 1)/3n for the case of parallel plates.
Similarly, at very low flowrates (NGz< 3) the limiting equation may
be shown to be, for the case of constant wall temperature under consider-
ation,
Crozier et al. obtained experimental data using extremely viscous

polymeric solutions. Under these conditions natural convection effects
(Eq. 17) are negligible but the radial variations of the physical properties
are not. Therefore, the (yw/ye)0.14Sieder-Tate correction was applied.
Conditions of both heating and cooling were studied. T h e resulting
experimental check obtained was excellent, the principal limitation
being that most of the data were taken under conditions in which the
heat transfer rates are defined either Eq. (16), together with the Seider-
Tate consistency ratio, or by Eq. (20a) so that the actual check of Eq.
(20) was limited.
No analysis in which a more rigorous consideration of the radial
variation of physical properties was considered appears to have been
undertaken for this geometry.
lo The term n is defined in the Notation section; its significance and relationship
to the power law exponent n are comparable to that of the term n, which is discussed
in footnote 7.
W I
A. B. METZNER
For the special case of isothermal power law fluids an approximate

eigenvalue solution has been carried out by Tien (110). T h e range of
Graetz numbers covered (3-25) is precisely that required to aid in the
interpolation between Eqs. (20) and (20a).
No other geometries appear to have been considered.
R. LAMINAR HEATGENERATION,
FLOWWITH INTERNAL
ROUNDTUBES
Consideration of internal heat generation rates during transfer of
energy to or from the fluid may be conveniently divided into either of
two categories depending upon the intended application of the analysis.
I n the first instance one may assume the internal heat generation to be
solely due to viscous dissipation of energy. This corresponds realistically
to the behavior of viscous oils, some polymeric solutions and molten
polymers under actual processing conditions. Secondly, one may neglect
viscous dissipation processes and consider internal heat generation due
to molecular, atomic or biological processes, as in the design of reactors.
1. Heat Generation due to Viscous Dissipation.

Several experimental observations are of assistance in defining the
problem:
(1) A simple macroscopic energy balance reveals that bulk tempera-
ture increases of about 5F are generated during adiabatic flow through
round tubes for each 1000 psi of pressure drop, employing the physical
properties of typical organic fluids. Since pressure changes of this
magnitude are commonly encountered with viscous oils, and may reach
levels of at least 20,000 psi during extrusion of molten polymers, it is
seen that the rates of heat generation by viscous dissipation may affect
fluid properties significantly, hence also the velocity profile of the
fluid and the rates of heat transfer to the tube wall.
(2) Even more importantly, Gerrard and Philippoff (34) have shown
experimentally that the local temperature rise in the fluid may exceed
the above bulk temperature changes by as much as an order of magni-
tude, hence the distortion of the velocity profiles may be severe even when
the bulk temperature is changed only moderately or even slightly.
(3) Toor (ZZ4)has shown that when frictional heat generation rates
are appreciable real liquids cannot be considered to be incompressible-
since appreciable thermal effects accompany small volumetric changes.
P761
HEATTRANSFER FLUIDS
I N NON-NEWTONIAN
For purely viscous fluids as defined in Table I the'viscous dissipation

term reduces to p ( a ~ / a yand
) ~ the energy equation under steady-state
conditions may be written (33)
T h e momentum equations become
and
if radial flow is neglected and the fluid is assumed to be incompressible

insofar as momentum and continuity considerations are concerned.
Assuming the temperature dependence of the specific heat and thermal
diffusivity terms to be linear, the viscosity to vary exponentially with
temperature and, with shear rate, by an equation of the form of Eq. (2),
and further assuming the volumetric expansion term TP to be independ-
ent of temperature and position, Gee and Lyon solved the above equa-
tions numerically for several specific flow conditions. An isothermal
wall was assumed in all cases, and both the temperature and velocity
profiles were assumed flat as the fluid entered the heat exchanger tubes.
Most unfortunately, the results are not presented in detail and no attempt
was made to present the calculations in dimensionless form. Consequently
the results are of little design value and of no generality, nevertheless
several points may be noted.
(a) When the tube wall and inlet fluid temperatures are held at
identical values the velocity profile is not grossly affected by heat gener-
ation (at least under the conditions studied) since the effects of decreased
viscosity due to heat generation and consequently increased viscosity
levels because of lower shear stress (lower pressure gradients) are
counterbalancing.
(b) Very appreciable temperature differences between the inlet fluid
and the wall must be maintained to develop appreciable heat transfer
rates because the large effect of heat generation upon the temperature
profile overshadows low heat conduction rates.
While the solution of Eqs. (21) and (22) over some range of the dimen-
sionless variables involved may represent a large expenditure of com-
W'I
A. B. METZNER
puter time it would appear to be warranted in view of the importance

of the problem and the excellent agreement between theory and experi-
ment shown by Gee and Lyon for their data. For greater generality the
dissipative term in Eq. (21), [~./(pC,R)][(au/ay)~], should be replaced
by terms which include the higher order contributions to the dissipation
function in complex flow fields (121).
A number of more specific solutions have also been reported. While
usually involving more stringent assumptions, several of these may be
of interest under specific conditions. Brinkman ( l o ) , Kearsley (54),
and Gill (35) assume temperature independent viscosities and neglect
fluid compressibility; the analyses of Bird (5) and Gill are for power
law fluids while those of Brinkman and Kearsley are for Newtonian
systems.l Schenk and Van Laar (98) considered one special case of a
temperature-independent Eyring-Powell constitutive equation. Toor
(115) considered power law fluids and included effects of fluid compress-
ibility. While not capable of predicting real fluid behavior widely these
papers represent worthwhile analyses in that they lead to an under-
standing of the relative importance of the separate effects in a very
complex area. T h e discrepatlcy between theory and experiment which
Toor discusses in relation to Birds study-in polyethylene, a highly
viscoelastic fluid-suggests the importance of consideration of all
energetic processes through use of more erudite constitutive equations.
Gerrard and Philippoff (34) have developed what might perhaps be
described as a sophisticated intuitive approach to the heat transfer
problem of special interest in viscometry: that of correcting pressure
drop-flowrate measurements for the effects of viscous heat generation.
I t is interesting to note that they were unable to achieve the isothermal
wall conditions assumed in most analyses in their experimental work,
as contrasted to Gee and Lyon, who apparently had little difficulty.
T h e answer may lie in the comparative rates of heat generation: Gee and
Lyon, working with viscous molten polymers, were restricted to modest
shear rates. Gerrard and Philippoff, studying oils having viscosities of
only 1-10 poise employed shear rate levels of 104-105 sec-, thereby
developing viscous dissipation rates which were as much as several
orders of magnitude higher than those involved with the more viscous
molten polymers. Simple heat transfer rate calculations show that in the
case of molten polymers the isothermal wall assumption may frequently
be a realistic one; the reverse is true of polymeric solutions and other
l1 A more extensive analysis for Newtonian fluids in which an exponential viscosity-

temperature relationship was used but compressibility effects were neglected has recently
been published by Gruntfest et al. (39).
C3781
HEATTRANSFER
IN NON-NEWTONIAN
FLUIDS
materials which may be processed at very high shear rate levels. In this
case the experimental data of Gerrard and Philippoff show the adiabatic
wall case to be approached reasonably closely even under conditions of
intentional cooling.
In summary, it is seen that the available analyses of heat transfer under
conditions of internal generation due to dissipative processes are of
interest principally from a research viewpoint: they provide a worthwhile
framework for further analyses but the results available are generally
too restrictive to apply directly to the solution of real problems. Further,
the recent study of Gerrard and Philippofi shows that the problem of
choice of the proper boundary conditions requires a more careful
consideration than it appears to have been given in the past.
2. Heat Generation Due to Internal Sources

Dilute slurries as might be used in nuclear and biological reactors are
occasionally appreciably less temperature sensitive than other non-
Newtonian fluids [see, for example, Vaughn ( I l l ? ) ] . Under these condi-
tions the velocity profile corresponds approximately to that of fully
developed isothermal flow provided aberrations due to natural convec-
tion are absent. As discussed in Section 111, A, frequently this latter
assumption will not be a realistic assumption, yet solutions based on this
premise represent a useful point of departure for further consideration
(perhaps empirical) of this complication.
Table I1 reviews the several analytic and numerical studies available
for purely viscous fluids. No experimental results of any kind appear to
be available in this area; further analysis should probably be delayed
until the indicated importance of natural convection effects has been
defined experimentally. T h e readers attention is drawn to the availability
of a well-developed and interesting experimental technique for internal
generation of heat in the fluid stream: the passage of electric current
through an electrolytic fluid, as described by Inman (49) and Sparrow
et al. (103).
C. LAMINAR
FLOWWITH INTERNAL
HEATGENERATION,
OTHERGEOMETRIES
No studies of non-Newtonian fluids in which well-developed consti-
tutive equations are employed appear to be available but two excellent
analyses which anticipate and solve a number of the problems likely to be
encountered are available. In the first, Turian and Bird (117) studied the
effects of viscous dissipation in temperature-sensitive Newtonian fluids
P791
A. B. METZNER
TABLE I1
HEATTRANSFER
UNDER CONDITIONS
OF INTERNALHEATGENERATION,
PURELYVISCOUSFLUIDSINSIDE ROUNDTUBES
Assumed Assumed Assumed distribution

Authors boundary rheological of internal heat
conditions properties" generation rates
-
(1) Wissler and T, = const. Bingham (a) Uniform: considered with some
Schechter( 129) plastic completeness
(b) Variable with either radial or
axial position, or both: equa-
tions formulated and methods
of solution indicated
(2) Schechter and (q/A)waii= 0 Bingham Uniform
Wissler (97) plastic
(3) Gill (35) Both T, = const. Power law Any function of radial position
and (q/A)wnll= const. may be accomodated. Obtains
asymptotic solutions for very long
tubes only
(4) Michiyoshi (a) (q/A)wall= 0 Bingham Uniform in all cases
et al. (77) (b) (q/A)wail= const. plastic
= const.
(c) T,,,
( 5 ) Foraboschi T, = const. Power law Linearly dependent on local tem-
and perature
di Federico (29)
- ~~
a Fluid properties are assumed constant (independent of temperature) in a11 analyses

listed.
between one stationary and one moving flat plate; the resulting analysis
was applied to the flow field in a cone-and-plate rotational viscometer.
Clearly extensions to the analogous non-Newtonian problems are of
interest and the expository nature of their paper provides an excellent
framework for such extensions.
Jain (50) considers dissipation of energy in the same (parallel plate)
geometry for viscoelastic fluids having a constitutive equation of the
form
67
rii +h 6t
= 2pdj' + 4p,d,idja (23)
in which djidenotes the rate of deformation tensor, T~~ the stress tensor
and &;/st a convected derivative of the stress tensor. T h e material
parameters A, p, and pc are assumed to be independent of temperature
in the subsequent analysis. In this case the dissipation term in the energy
equation assumes the form
=p + ~ ~ ~ 7 4 (24)
~3801
HEATTRANSFER FLUIDS
IN NON-NEWTONIAN
instead of just the first term as in Eq. (21). T h e temperature profile now
depends on an additional dimensionless group
---
AtLc ugz
P L2
which the author has (incorrectly) termed a viscoelasticity number.
As a matter of fact this dimensionless group does not represent a ratio
of elastic to viscous forces but merely the separate contributions to the
viscosity function. Thus, as might perhaps be expected intuitively,
the dissipation function depends only upon the effective fluid viscosity,
just as in the case of Newtonian fluids, and not upon fluid elasticity
at all.
AND TURBULENT
D. TRANSITION FLOWCONDITIONS
1. Flow through Round Tubes
a . Prediction of the Onset of Turbulence. In the case of Bingham
plastic behavior the Reynolds number is usually based upon the coeffi-
cient of rigidity; as this, in itself, does not completely define the magni-
tudes of the viscous stresses the laminar-turbulent transition depends
upon both the Reynolds and Hedstrom numbers. As shown in Fig. 2
-g 0.100
b
L
P
.-
E
=
e 0.010
0.005
10 10 lo4
Reynolds number, -
DVP
7
FIG. 2. Drag coefficient-Reynolds number-Hedstrom number relationships for
Bingham plastic fluids. Taken from Hedstrorn (45).
A. B. METZNER
the end of the stable laminar region occurs at (or shortly after) the inter-
section of the laminar friction factor-Reynolds number curves with the
usual turbulent curve for Newtonian fluids (106, 128).12 T h e friction
factors in the region of fully developed turbulent flow fall slightly below
the turbulent Newtonian line for highly non-Newtonian Bingham
plastics (107, log), the difference decreasing with increasing Reynolds
numbers. Equations of the Blasius form have been given for this region
by Thomas; they are not reproduced here as an alternate analysis (73)
may be of somewhat greater generality.
I n view of the fact that Eq. (1 5 ) defines the fluid shear rate at the wall
of the tube for all fluids under well-developed laminar flow conditions,
and since the shear stress is a function of only the shear rate for purely
viscous fluids (see Table I), one may define a generalized Reynolds
number applicable to all purely viscous fluids on the basis of these con-
. ~h~e definition used requires a single line cf = 16/Nk,)
s i d e r a t i o n ~T
under laminar flow conditions, regardless of the fluid properties. Figure 3
depicts the friction factor-Reynolds number relationships using this
approach; the curves shown are based upon the above considerations
under conditions of laminar flow and upon an experimentally supported,
semitheoretical analysis for the turbulent region (26).T h e analysis of the
turbulent region was based upon the power law constitutive equation, but
it was shown analytically that deviations of real purely viscous fluids
from this model are too small to be significant under turbulent flow
conditions. Thus, the models used to develop Fig. 3 are such that all fluids
may be expected to conform to the relationships shown: under laminar
conditions the requirement is rigorous and, under turbulent conditions,
while not rigorous, comes as an excellent approximation. T h e transitional
generalized Reynolds number is seen to vary between about 2100 and
4000 for fluids studied, depending upon the value of the flow behavior
index n.
While the exact form of the constitutive equation of the fluid may be
either totally or substantially irrelevant under fully laminar or fully
turbulent conditions, respectively, the same may not be expected to be
true under conditions of incipient instability as the transition region is
approached. T h e development of instabilities may depend critically upon
local conditions-such as the exact shape of the velocity profile-hence
upon the detailed rheological propertiesof the fluids used. This suggestion
T h e work of Winning has been discussed in detail by Metzner (65).
lS Aberrations such as separation of the fluid into two distinct phases are obviously
not included, i.e. a single constitutive equation must be capable of predicting the shear
stress-shear rate behavior of the fluid. However, while it must be assumed that such a
unique relationship exists its form or complexity are irrelevant.
r3821
HEATTRANSFER FLUIDS
IN NON-NEWTONIAN
DO
Dn' 2- n'
Reynolds number, Nk, =
Y P
FIG. 3. Drag coefficient-Reynolds number-flow behavior index relationships for
purely viscous fluids. Taken from Dodge and Metzner (26).
was first made by Ryan and Johnson (93),and has since been extended
by Hanks and Christiansen (44), and by Hanks ( 4 / , 42). The relation
proposed by these investigators reduces to the following for the special
case of power law fluids:
This equation predicts critical Reynolds numbers increasing from 2100

to about 2400 as the flow behavior index decreases from unity to about
0.5. For still lower flow behavior indexes the critical Reynolds number is
predicted to decrease, reaching 1600 at n = n' = 0.10. It is not known
whether such peculiar trends are correct or not; however, it may be noted
that the relation analogous to Eq. (25) for Bingham plastic fluids also
predicts critical Reynolds numbers which appear to be somewhat too
low. Obviously the precise behavior in the transition region is not yet
well-defined. T h e analyses of Hanks and Christiansen have been shown
to provide an opportunity for making predictions for other geometries
and under highly nonisothermal conditions with what may be expected
to be superior precision and are recommended in those cases. For tube
flow, however, there seems to be little advantage at present to use of
P831
A. B. METZNER
relationships more complex than locating the intersection of the laminar

and nonlaminar curves on Figs. 2 or 3 and, at least in the case of Bingham
plastics having high yield values, the differences between this procedure
and that of Hanks and Christiansen may be significant. Thus, use of
Figs. 2 and 3 for smooth round tubes is recommended at the present.
b. Turbulent Flow Characteristics. T h e turbulent-region drag coefl-
cients (which will subsequently be required to enable heat transfer rate
predictions) may be taken from Fig. 3 for all purely viscous fluids,
according to Dodge and Metzner (26). T h e standard deviation of their
fairly extensive data from the curve was only 2.4%. Subsequent studies
lead to the following comments on this correlation.
(i) T h e pressure loss data of Bogue and Metzner ( 9 ) taken on a
different apparatus but for similar suspensions and polymeric solutions
support Fig. 3 fully over the ranges indicated.
(ii) An extensive analysis of the applicability of these relationships
to Bingham plastic slurries, such as used in developing Fig. 2, revealed
a substantial deviation from the indicated relationships (107, 109).
However, the rheological data used to arrive at this conclusion appear
to be subject to more than a single interpretation (9) hence it is not clear
whether the differences observed are real or not.
(iii) Clapp (17) reports higher friction factors than those predicted
by Fig. 3, using similar solutions. The published discussion of his paper
includes a suggestion that his fluids may have been more nearly New-
tonian than he assumed, however, since the rheological data apparently
had to be extrapolated to reach the shear stresses encountered under
turbulent conditions. While the authors reply to this suggestion is in
the negative, the data he cites reveal that this was indeed the case.14
Since the fluids used approach Newtonian behavior more closely at
higher shear rates, the deviations reported are qualitatively those
expected due to inadequate rheological data; in any event the con-
clusions reached by Clapp must be disregarded because of his inadequate
rheological measurements.
(iv) Recent data, as yet unpublished, suggest the correctness of the
curves of Fig. 2 over the ranges studied but indicate tentatively that the
l4 Shear stresses as high as 3.4 psf are cited by the author under turbulent conditions;
the measurements extend only to about 1.5 psf. While the author also states that shear
rates extended only to 2 x los sec- this is incompatible with the shear stresses cited
unless one assumes he is confusing the term 8V/D with shear rate. Under laminar condi-
tions the two are, of course, related by Eq. (15) but elementary considerations reveal
that no such relation exists under turbulent conditions.
P I
HEATTRANSFER FLUIDS
IN NON-NEWTONIAN
extrapolated theoretical predictions (shown as dashed lines) may be

too low.
(v) Recent measurements [some of which are reported by Metzner
et al. (76)] reveal the existence of finite normal stresses (viscoelastic
effects) in some of the fluids used to develop the curves of Fig. 3. How-
ever, slurries, in which long-range viscoelastic effects should be absent,
were also used to assure the applicability of the results to purely viscous
fluids. Thus this complication should not be significant but admittedly
more experimental evidence would be desirable.
In summary, each of the suggestions that Fig. 3 may not represent
accurately the turbulent behavior of all purely viscous fluids is incom-
plete or controversial. Taken together, however, these questions cannot
be ignored. On the other hand, some decision is required and since
a very substantial body of experimental evidence of high precision is
available in support of the curves shown they will to be correct until
a further clarification is obtained.
Turbulent velocity projiles have been predicted from the analysis of
turbulent momentum transport rates which led to Fig. 3. However,
one of the simplifying assumptions which is not critical in the corre-
sponding siniilarity analysis for Newtonian fluids has been shown to
become of mcjor importance in the non-Newtonian case ( 9 ) . Thus, the
Dodge-Metzner velocity profile predictions must be considered to be
invalid. Experimental measurements of the turbulent profiles, fargely
in the turbulent core region of the flow field, by Bogue and Metzner
( 9 ) and Eissenberg and Bogue (28), show that turbulent non-Newtonian
profiles are substantially identical to those of Newtonian fluids, although
perhaps slightly flatter. Brodkey (12) has correlated many of these data
using the Pai power series analysis.
T h e turbulent behavior of viscoelastic non-Newtonian fluids deviates
markedly from that of purely viscous materials. However, no definitive
analyses of the momentum transport characteristics are as yet available
and no analysis or study of turbulent heat transfer processes in these
systems has as yet been reported. T h e reader is referred to the literature
(26, 59, 72, 87, 95, 101, 120) for further details concerning turbulence
in viscoelastic fluids.
c. Turbulent Heat Transfer Rates. T h e analogy between turbulent
heat, mass and momentum transport processes in non-Newtonian systems
may be developed by extension of the corresponding analysis for New-
tonian fluids. Noting that the region of high Prandtl numbers is of
primary interest because of the viscous nature of most slurries and solu-
tions which are concentrated enough to show significant departures from
[3851
A. B. METZNER
Newtonian behavior, the sublayer region immediately adjacent to the

tube wall, and the small turbulent fluctuations in it, are of primary
interest. Under these conditions Friend and Metzner (32) suggest,
for Newtonian fluids, use of the equation
with the limitation that
to justify assumptions concerning the relative importance of turbulent

and molecular transport processes in the wall region.
T h e choice of eq. (26) as a point of departure for subsequent considera-
tion of non-Newtonian effects, in preference to the widely quoted
analogies of Deissler (22), and Lin et al. (58), requires justification.
However, both of these latter analogies were developed with emphasis
upon the experimental data of Lin et al. (57) which are believed to be
both inapplicable and incorrect: inapplicable because they were taken
under conditions of annular flow, not flow through round tubes, and
no well-developed means is available for transposing these to the
desired tube flow relationships under turbulent conditions. They appear
to be incorrect since, under laminar flow conditions, the data obtained
by Lin et al. do not conform to the theoretical equations which may be
derived simply and rigorously in this case. I n view of these rather obvious
shortcomings it is somewhat surprising that these results, and analogies
based upon them, continued to be cited.
Turning now to an extension of Eqs. (26) and (27) to non-Newtonian
systems, and again noting that the thermal boundary layer for these
materials will be largely confined to the wall region, one may note that
the shearing stress levels will be approximately constant within the
thermal boundary layer. Under these conditions the viscosity of the non-
Newtonian fluid must also be approximately constant (since it is depend-
ent upon shear stress) and the distinction between Newtonian and purely
viscous non-Newtonian fluids has been lost. As a result, Eq. (26) would
be expected to be applicable to these systems also provided the Prandtl
number (or Schmidt number, in the case of mass transfer) is evaluated
using the non-Newtonian viscosity as determined at the shear stress
levels existing at the tube wall. Differences in momentum transport rate
(velocity profile at the wall) between Newtonian and non-Newtonian
systems must, of course, be considered. Metzner and Friend, (70) using
~3861
HEATTRANSFER FLUIDS
IN NON-NEWTONIAN
Fig. 3 to obtain the necessary drag coefficients, showed these considera-

tions to be valid for a variety of non-Newtonian systems. Generalized
Reynolds numbers were varied from transitional levels to beyond
70,000; the wall Prandtl numbers varied from 6.3 to 186 and the flow
behavior indices from 0.39 to 0.92. Studies by Thomas (106) on slurries
quite different from those used by Friend and Metzner generally
support Eq. (26) also, although at the lower Prandtl numbers used by
Thomas this equation becomes indistinguishable from most other
analogies. Eoth authors recommend the use of Sieder-Tate or Kreith-
Summerfield correction factors (see Eq. (17)) to account for highly non
isothermal conditions as a first approximation. T h e limits of applicability
of this analysis should be more greatly dependent on the turbulent velo-
city profiles than are the heat transfer rates, but, as discussed earlier,
these profiles differ but little from the Newtonian ones. Hence Eq. (27)
may also represent a fair approximation for non-Newtonian systems; in
this case the generalized Reynolds number and the wall Prandtl number
should be used. This is at variance with the conclusions reached by
Metzner and Friend (70), who, however, in the absence of velocity
profile measurements, used the incorrect predictions of Dodge and
Metzner (26).
Limited heat transfer data reported by Clapp (17) were correlated
using an empirical equation of highly unusual form which cannot be
recommended. T h e difficulties encountered may stem from inadequate
rheological data, as discussed earlier in Section C, 1, b. This is most
unfortunate as it makes his excellent and interesting temperature profile
measurements, the only ones reported to date for non-Newtonian
systems, quantitatively uninterpretable. However, qualitatively they do
support the earlier suggestion that the major portion of the temperature
change occurs in a region very close to the tube wall.
An empirical modification of Eq. (26) has been proposed by Petersen
(85);the available data are insufficient to assess its necessity and success
critically. This author also discusses the difficult problem of heat transfer
under laminar-turbulent transitional conditions.
I n summary, Eq. (26) is recommended for design purposes; while
it has been checked over fairly wide ranges of conditions the standard
deviation of all the available data-20-25 %-is considerably more than
desirable. As a result some further study of this problem is necessary.
Since in recent years significant strides have been made in under-
standing the unsteady turbulent processes in the sublayer region
immediately adjacent to the tube wall, such studies might constitute
a reasonable point of departure for further analyses of the heat transfer
problem. T h e pertinant recent developments in this area are sum-
W I
A. B. METZNER
marized concisely by Wasan et al. (119) and by Reiss and Hanratty

(89)-
2. Other Geometries
No heat transfer rate measurements appear to be available for non-
Newtonian fluids flowing through ducts of other geometries. One of the
first significant problems, under turbulent conditions, is that the proper
choice of the terms in a Reynolds number is not at all clear. While a
number of fluid mechanical analyses based on simple constitutive equa-
tions (power law, Bingham plastic) have been carried out for laminar
flow in an annulus, it is not expected that either of these would lead to
precise predictions (this is especially true of the power law; as the
region in which this equation breaks down, at low shear rates is not, as
fortuitously occurs in flow through round tubes, of negligible impor-
tance). More importantly, an analog of the Weissenberg-Mooney
relation (Eq. (15)) which led to the generalized Reynolds number for
tube flow of all purely viscous fluids does not appear to be available.
Thus intelligent studies of turbulent heat transfer in annuli must be
preceeded by more thorough fluid mechanical analyses than have been
available. T h e current state of the art, from a fluid mechanics viewpoint,
is summarized in the publications of Brodkey ( I ] ) , Savins ( 9 4 , Wilcox
(Z26), and Fredrickson and Bird (31).
E. SUMMARY
With some qualifications as noted in the text heat transfer rates to
purely viscous fluids in steady ducted flow fields may be predicted with
nearly the same degree of assurance as for Newtonian materials. In the
case of laminar $ow, Eq. (13) and Fig. 1 may be expected to enable
excellent estimates for heating of power law fluids whenever effects due
to natural convection may be ignored. When these limitations are un-
acceptable, Eqs. (17) may be used to obtain estimates nearly as reliable
as those for Newtonian fluids under the same conditions. These equa-
tions all apply to conditions of constant wall temperature; constant
heat flux conditions have not been studied experimentally although
several approximate analytic solutions are available.
Corresponding to these analyses for flow through round tubes, Eqs.
(20) and (20a) enable approximate predictions for flow between parallel
plates; these predictions would be expected to be good to l0-15%
provided complications due to natural convection are not present.
Under turbulent $ow conditions, Eqs. (26) and (27), to be used with
~3881
HEATTRANSFER FLUIDS
IN NON-NEWTONIAN
Fig. 3, may evidently be used to obtain design estimates good to within

about 20% for purely viscous fluids. In this case the assurance of ade-
quate predictive abilities is not as high as under laminar conditions
because of the more limited nature and greater scatter of available data.
As compared to Newtonian fluids the problem of internal heat
generation in non-Newtonian systems (due either to internal sources
of energy or to viscous dissipation effects) occupies a position of much
greater importance. Solutions for a large number of special cases are
already available but all are predicated upon simplifying assumptions
which may frequently be invalid. These solutions are likely to be ap-
proached closely under conditions of moderate temperature differences,
when the neglected effects of temperature-dependent properties and of
natural convection are likely to be unimportant. T h e equations which
represent more rigorous attacks have been formulated and solved for
several special cases but not in general. Thus considerable further ana-
lysis and experimental confirmation are required in this area.
This summary would be incomplete if the relative importance of
non-Newtonian effects, natural convection and variation of viscous
properties with temperature, under laminar flow conditions, were not
reiterated. T h e first effect is frequently the least important of the three;
the law of diminishing returns is clearly applicable to the many analytic
and numerical analyses of heat transfer rates which continue to empha-
size non-Newtonian behavior and ignore these other more important
effects. I t is to be hoped that applied mathematicians of the future may
seek more renumerative problems than many which have been chosen
for study in this area in the past.
IV. Boundary Layer Problems
A. LAMINAR
If a sufficiently simple constitutive equation is employed, the order
of magnitude analysis of the stress equations of motion which may be
used to obtain the usual boundary layer equations is identical to that for
Newtonian fluids. Then, following substitution of the chosen constitutive
relationships for the remaining viscous stress terms solutions of these
boundary layer equations may be obtained as usual. Depending on the
geometry chosen the actual solution may be more difficult to carry out
than in the linear Newtonian case but the basic approach is the same.
A veritable burst of activity has occurred in this area in recent years;
the problems considered are summarized in Table 111.
W I
A. B. METZNER
TABLE I11
BOUNDARY FOR POWERLAWFLUIDS
LAYERSOLUTIONS
Problem Investigators
.-.
Fluid mechanics problems

(boundary layer thickness, drag coefficients, velocity profile relationships)
Boundary layer development in flow over Tomita (113)
flat plates
Boundary layer development on flow into (1) Bogue (8)
a round tube (2) Tomita (112, 113)
(3) Collins and Schowalter (20)
(4) Kapur and Gupta (52)
(3) Boundary layer development on flow into a Collins and Schowalter (19)
channel consisting of parallel +t plates
(4) Flow into a convergent channel Kapur and Srivastava (53)
Consideration of both j u i d mechanics and boundary layer heat transfer rates

Natural convection, two dimensional surfaces Acrivos ( I )
Forced convection, flow over flat plates Acrivos et al. (2)
Forced convection from cylinders Shah et al. (100)
Forced convection during flow into a channel Yau and Tien (130)
consisting of parallel flat plates
Bogue, Tomita, and Yau and Tien employed approximations based

either upon variational methods or upon the von Karman-Pohlhausen
integral equations. These appear to be'satisfactory in the corresponding
analyses for Newtonian systems; the other solutions listed are rigorous
within the framework of the original boundary layer approximations.
However, it is not clear whether the discrepancies between the analyses
of Bogue, Tomita, and Collins and Schowalter, for the round tube entry
region are due to the use of such approximations in the earlier work
or to computational errors in one analysis or the other. In other words,
some further examination of the calculations available may be in order.
A much more important general question arises and does not appear
to have been considered: viz., are solutions to boundary layer problems
based on the power law constitutive equation likely to be of permanent
value, on the one hand, or, at the other extreme, relatively useless?
Napolitano (80) has drawn attention to the fact that since real fluids
never follow power law behavior over wide ranges of shear stress (in
particular, the fluid behavior must always deviate from this approxima-
tion at low stresses-see Section 11, A), it is clear that the above power
POI
HEATTRANSFER FLUIDS
IN NON-NEWTONIAN
law solutions must be in error over some fraction of the boundary layer
region. If the power law parameters are fitted to depict the fluid proper-
ties correctly adjacent to the surface and near its leading edge, then the
outer region of the boundary layer will be in error, as will the entire
boundary layer further downstream, and vice versa. Obviously both
numerical solutions based upon more realistic approximations and experi-
mental studies are in order, to determine whether these shortcomings are
of great significance or not. No such calculations appear to have been
made and only two sets of experimental results are available, neither
being conclusive. Measurements reported by Dodge (24) of the pressure
profiles in the inlet regions of round tubes have been used both by
Bogue (8), and Collins and Schowalter (2O), to verify their analyses.
Only weak conclusions may be drawn: in the first place, such pressure
drop measurements do not reflect sensitively the behavior of the entire
boundary layer but only of the wall region. Secondly, Dodges measure-
ments were incidental to other studies and are incomplete. T h e only
experimental program undertaken to study boundary layer phenomena
directly is the study of Shah et al. (ZOO) of heat transfer rates from
round cylinders. While the results generally appear to check the pre-
dictions the experimental data are reported to be accurate to only about
10yo as compared to calculated maximum deviations from Newtonian
boundary layer behavior of 25-50 yo for the fluids used. Additionally, the
C M C solutions employed in the experimental studies almost surely
exhibited some viscoelastic effects (75, 76, 90) and, if present, these
would in principal void the entire analysis as additional acceleration
tensors would be introduced into the equations of motion by the con-
stitutive relation. I n practice the influence of these additional terms
may not be large (121). However, their precise influence is unknown
at present so that the data may not be cited in support of the power
law analysis in any event.
I n addition to the above solutions Tomita (111)has carried out dimen-
sional analysis approaches to several boundary layer problems. More
importantly, Schowalter (99) has inquired into the types of two- and
three-dimensional flow fields for which similarity transformations
are possible when the power law is used, and White (121) has extended
this to other constitutive equations. T h e general conclusions are that
with more complex fluid property relationships or with more complex
geometries similarity transformations and therefore general solutions
appear to be possible only for flow fields which are so specialized as to
be degenerate.
I n summary, some further analysis and a considerable amount of
experimental work is seen to be necessary to determine the ranges of
E3911
A. B. METZNER
validity of the various available power law solutions. It is not at all clear
at the present that an area of analysis comparable to that of Newtonian
boundary layer theory exists: when extremely simple constitutive equations
are employed there is some question as to whether or not the results
will depict the behavior of real fluids correctly. On the other hand, when
more complex constitutive equations are employed, it is not apparent
that general solutions of the resulting equations of motion are obtainable.
A productive intermediate area of analysis remains to be defined and
developed.
B. TURBULENT
A useful beginning on application of the similarity laws to a study of
momentum transfer in turbulent non-Newtonian fluids has been
reported by Granville (37).No experimental results appear to be available
to enable any check of the predictions and no heat transfer studies of
any kind, analytic or experimental, have been published.
V. Miscellaneous Heat Transfer Problems
Studies of nucleate boiling rates in non-Newtonian suspensions of

thoria in water showed that both the dependence of heat flux on tem-
perature difference and the peak heat flux were appreciably reduced
from the levels observed in water alone. These observations were attri-
buted to the formation of a loosely adhering sludge phase of the slurry
particles on the heat transfer surfaces and no separate effects due to the
non-Newtonian properties of the slurry could be identified (105).
NOTATION
Surface area for heat transfer

Physical property parameter in Eyring-Powell constitutive equation
Physical property parameter in Eyring-Powell constitutive equation
Separation of parallel plates
Specific heat
Deformation rate tensors
Tube diameter
Drag coefficient (Fanning friction factor)
Symbol denoting a functional of the quantity within the square brackets
Acceleration of gravity and dimensional conversion factor, respectively
Mass velocity
Heat transfer coefficient ; ham denotes the coefficient based on an arithmetic
mean temperature difference
HEATTRANSFER
IN NON-NEWTONIAN
FLUIDS
AH Enthalpy of activation (Eq. (7))

k Thermal conductivity of fluid; k , refers to the conductivity of the continuous
fluid, k , to that of suspended particles
K Index of fluid consistency in power law model (Eq.(1))
K Index of fluid consistency defined, at a particular stress, by the relation:
7w = K(8V/D)"'
L Length of heated surface
n Flow behavior index or power law exponent (Eq. (1)):
d(ln 7 1 2 ) . analogously, n' = -

n=------- -d(ln 7 , ) and n" = ___-d(ln 7,)
d(ln ' d(ln SV/D) d(ln V / B d
Grashof number with fluid properties
. - evaluated at wall conditions of tempera-
ture and fluid deformation rate: N G = ~ /3 dTD3pzg/p2
Graetz number: NGz = wC,/kL
Hedstrom number: N H = ~ 7vD2pg,/pip
Nusselt number, hDjk or hamDjk
Prandtl number, C,p/k. For non-Newtonian fluids Npru,denotes this dimension-
less group with the viscosity term evaluated at the wall conditions of shear
rate and temperature.
NRe Reynolds number. For Newtonian fluids, N R = ~ DVp/p. For non-Newtonians
the generalized Reynolds number NK, denotes D"'V*-"'p/y and NKee
denotes this same group evaluated at the critical velocity of the laminar-
turbulent transition. For Bingham plastic fluids, ??&, = DVp/p~p
Stanton number, hjC,G
Pressure
+
Components of the deviatoric stress tensor P defined by Pij = 7 i 1 p6,,
Heat transfer rate. (q/A), = 0 denotes the case of an adiabatic wall.
Radial distance
Gas law constant (appears only in Eqs. (7) and (9b) and Fig. 1)
Radius of tube or duct
Time
Temperature. T i , T, , and T denote the initial fluid temperature, wall tempera-
ture and average fluid temperatures, respectively
Local velocity; u, denotes the velocity at the centerline of the duct
Characteristic velocity
Average or bulk fluid velocity
Total mass rate of flow
Volume fraction of suspended particles
Reduced radial coordinate, r / R
Axial distance
Reduced axial distance, z/L
Thermal diffusivity
Coefficient of thermal expansion; fl denotes an average value over the tempera-
ture range of interest
YDtr Deformation rate tensor in convected coordinates (Table I only).
Y Denominator of generalized Reynolds number. y = g,K'8"'-' ; yw denote
this term evaluated at the wall temperature, JB at the bulk remperature
Component of the deformation rate tensor r. In simple shearing flows simply
denotes the fluid shear rate. T, denotes this term evaluated at the wall
conditions of the shear rate or deformation rate.
Kronecker delta
P931
A. B. METZNER
3n + I t
( 7 (applies to)
flow through round tubes).
2n 1 4
( 7
-C
(applies to)
flow through slits formed by parallel plates).
Ratio of non-Newtonian to Newtonian heat transfer rates at low Graetz numbers.

Tabulated by Metzner et al. (74).
Function of time of shear
Material (property) parameter in Eq. (23)
Viscosity; pw denotes viscosity evaluated at wall conditions of shear rate and
temperature
Coefficient of rigidity of Bingham plastic fluids (Eq. (31))
Coefficient of cross-viscosity (Eq. (23))
Viscosity of fluid at low shear rates (Eq. (2))
3.14...
Fluid density
Components of total stress tensor T. rL1denotes the conventional shear com-
ponent; when evaluated at wall conditions this is abbreviated as T~
Yield value of Bingham plastic fluids (Eq. (3))
Symbol denoting function of the following groups
Energy dissipation rate per unit volume of fluid
REFERENCES
1. A. Acrivos, A.Z.Ch.E. J . 6 , 584 (1960).

2. A. Acrivos, M. J. Shah, and E. E. Petersen, A.Z.Ch.E. J . 6 , 312 (1960).
3. G. Astarita, Znd. Eng. Chem. Fundamentals 4, (1965).
4. E. C. Bernhardt, Processing of Thermoplastic Materials. Reinhold, New York,
1959.
5. R. B. Bird, SPE (SOC.Plastics Engrs.) J 1 1 ( I ) 35 (1955).
6. R. B. Bird, Chem. Zngr.-Tech. 31, 569 (1959).
7. R. B. Bird, W. E. Stewart, and E. N. Lightfoot, Transport Phenomena. Wiley,
New York, 1960.
8. D. C. Bogue, Znd. Eng. Chem. 51, 874 (1959).
9. D. C. Bogue and A. B. Metzner, Znd. Eng. Chem. Fundamentals 2, 143 (1963).
10. H. C. Brinkman, Appl. Sci. Res. A2, 120 (1951).
11. R. S. Brodkey, Ind. Eng. Chem. 54 (9), 44 (1962).
12. R. S. Brodkey, A.Z.Ch.E. J. 9, 448 (1963).
13. S. Burow, A. Peterlin, and D. T. Turner, Polymer Letters 2 , 67 (1964).
14. S. E. Charm, Znd. Eng. Chem. Fundamentals 1, 79 (1962).
15. E. B. Christiansen and S. E. Craig, Jr., A.Z.Ch.E. J. 8, 154 (1962).
16. E. B. Christiansen and G. E. Jensen, in Progress in International Research on
Thermodynamic and Transport Properties (J. F. Masi, ed.), p. 738. Academic
Press, New York and Am. SOC.Mech. Engrs., New York, 1962.
17. R. M. Clapp, in International Developments in Heat Transfer pp. 652, D21 I ,
Am. SOC.Mech. Engrs. New York, 1963.
18. S. B. Clough, H. E. Read, A. B. Metzner, and V. C. Behn, A.Z.Ch.E. J. 8, 346 (1 962).
19. M. Collins and W. R. Schowalter, A.1.Ch.E. J . 9, 98 (1963).
20. M. Collins and W. R. Schowalter, A.Z.Ch.E. J . 9, 804 (1963).
P941
HEATTRANSFER
IN NON-NEWTONIAN
FLUIDS
21. R. D. Crozier, J. R. Booth, and J. E. Stewart, Chem. Eng. Progr. 60 (8), 43 (1964).
22. R. G. Deissler, Natl. Adwisory Comm. Aeron. Rept. 1210 (1955).
23. D. A. Denney and R. S. Brodkey, J. Appl. Phys. 33, 2269 (1962).
24. D. W. Dodge, Ph.D. thesis, University of Delaware, Newark, Delaware, 1957.
25. D. W. Dodge, Ind. Eng. Chem. 51, 839 (1959).
26. D. W. Dodge and A. B. Metzner, A.I.Ch.E. J. 5, 189 (1959); for corrections see
8, 143 (1962).
27. R. M. Drake, Jr., Ph.D. thesis, University of Tennessee, Knoxville, Tennessee, 1961.
28. D. M. Eissenberg and D. C. Bogue, A.I.Ch.E. J. 10, 723 (1964).
29. F. P. Foraboschi and I. di Federico, Intern. J. Heat Mass Transfer 7, 315 (1964).
30. A. G. Fredrickson, Principles and Applications of Rheology. Prentice-Hall,
Englewood Cliffs, New Jersey, 1964.
31. A. G. Fredrickson and R. B. Bird, Ind. Eng. Chem. 50, 1599 (1958).
32. W. L. Friend and A. B. Metzner, A.I.Ch.E. J. 4, 393 (1958).
33. R. E. Gee and J. B. Lyon, Ind. Eng. Chem. 49, 956 (1957).
34. J. E. Gerrard and W. Philippoff, Trans. SOC.Rheol. (in press).
35. W. N. Gill, Appl. Sci. Res. A l l , 10 (1962); A.I.Ch.E. J. 8, 137 (1962).
36. R. E. Ginn and A. B. Metzner, Trans. SOC.Rheology (in press)
37. P. S. Granville,Report 1579, David Taylor Model Basin, Washington, D. C. (1962);
see also J. Ship Res. 6, October (1962).
38. U. Grigull, Chem. Ingr.-Tech. 28, 553, 655 (1956).
39. I. J. Gruntfest, J. P. Young, and N. L. Johnson, J. Appl. Phys. 35, 18 (1964).
40. S. J. Hahn, T. Ree, and H. Eyring, Ind. Eng. Chem. 51, 856 (1959).
41. R. W. Hanks, A.1.Ch.E. J. 9, 45 (1963).
42. R. W. Hanks, A.1.Ch.E. J. 9, 306 (1963).
43. R. W. Hanks and E. B. Christiansen, A.1.Ch.E. J. 7, 519 (1961).
44. R. W. Hanks and E. B. Christiansen, A.I.Ch.E. J. 8, 467 (1962).
45. B. 0. A. Hedstrorn, Ind. Eng. Chem. 44, 651 (1952).
46. E. Hirai, Chem. Eng. (Tokyo) 20, 440 (1956).
47. E. Hirai, Chem. Eng. (Tokyo) 21, 17, 26, 792 (1957).
48. J. R. Hopper, M.Ch.E. thesis, University of Delaware, Newark, Delaware (1964).
49. R. M. Inman, Intern. J. Heat Mass Transfer 5, 1053 (1962).
50. M. K. Jain, A w l . Sci. Res. A l l , 295 (1963).
51. T. B. Jefferson, 0. W. Witzell, and W. L. Sibbitt, Ind. Eng. C h . 50, 1589 (1958).
52. J. N. Kapur and R. C. Gupta, Z. Angew. Math. Mech. 43, 135 (1963).
53. J. N. Kapur and R. C. Srivastava, Z. Angew. Math. Phys. 14, 383 (1963).
54. E. A. Kearsley, Trans. SOC.Rheol. 6, 253 (1962).
55. F. Kreith and M. Summerfield, Trans. A S M E 72, 869 (1950).
56. M. A. Leveque, Ann. Mines 13, 201 (1928).
57. C. S. Lin, E. B. Denton, H. S. Gaskill, and G. L. Putnam, Ind. Eng. Chem. 43, 2136
(1951).
58. C. S. Lin, R. W. Moulton, and G. L. Putnam, Ind. Eng. Chem. 45, 636 (1953).
59. J. L. Lumrnus, J. E. Fox, Jr., and D. B. Anderson, Oil GasJ. Dec. 11, p. 87 (1961)
60. B. C. Lyche and R. B. Bird, Chem. Eng. Sci. 6, 35 (1957).
61. H. Markovitz, Trans. SOC.Rheol. 1, 37 (1957).
62. H. Markovitz, in Viscoelasticity (J. T. Bergen, ed.), p. 133. Academic Press,
New York, 1960.
63. H. Markovitz and D. R. Brown, Trans. SOC.Rheol. 7, 137 (1963).
64. J. C. Maxwell, Electricity and Magnetism, Vol. I. Oxford Univ. Press (Clarendon),
London and New York, 1873; see also Dover Books, New York, 1954.
W I
A. B. METZNER
65. A. B. Metzner, Advan. Chem. Eng. 1, 79 (1956).

66. A. B. Metzner, Znd. Eng. Chem. 49, 1429 (1957).
67. A. B. Metzner, Rheol. Acta 1, 205 (1958).
68. A. B. Metzner, in Handbook of Fluid Dynamics (V. L. Streeter, ed.),
McGraw-Hill, New York (1961).
69. A. B. Metzner and M. Whitlock, Trans. Sor. Rheol. 2, 239 (1958).
70. A. B. Metzner and P. S. Friend, Ind. Eng. Chem. 51, 879 (1959).
71. A. B. Metzner and D. F. Gluck, Chem. Eng. Sci. 12, 185 (1960).
72. A. B. Metzner and M. G. Park, J . Fluid Mech. 20, 291 (1964).
73. A. B. Metzner and J. C. Reed, A.I.Ch.E. J . 1, 434 (1955).
74. A. B. Metzner, R. D. Vaughn, and G. L. Houghton, A.Z.Ch.E. J. 3, 92 (1957).
75. A. B. Metzner, W. T. Houghton, R. A. Sailor, and J. L. White, Trans. SOC.Rheol.
5 , 133 (1961).
76. A. B. Metzner, W. T. Houghton, R. E. Hurd, and C. C. Wolfe, in Proceedings
of International Symposium on Second Order Effects, (M. Reiner and D. Abir,
eds.), p. 650. Macmillan (Pergamon), New York, 1964.
77. I. Michiyoshi, R. Matsumoto, and M. Hozumi, Bull. J S M E (Japan Soc. Mech.
Engrs) 6, 496 ( I 963).
78. M. Mooney, and S. A. Black, J . Colloid Sci. 7, 204 (1952).
79. Y. Murakami, Chem. Eng. (Tokyo) 26, 873 (1962).
80. L. G. Napolitano, Private communication, 1964.
81. W. Noll, Arch. Ratl. Mech. Anal. 2, 197 (1958).
82. D. R. Oliver and V. G. Jenson, Chem. Eng. Sci. 19, 115 (1964).
83. C. Orr, Jr. and J. M. DallaValle, Chem. Eng. Progr. Symp. Ser. 50 (9), 29 (1954).
84. A. D. Pasquino and M. N. Pilsworth, Polymer Letters 2, 253 (1964).
85. A. W. Petersen, Ph.D. thesis, University of Utah, Salt Lake City, Utah, 1960.
86. R. L. Pigford, Chem. Eng. Progr. Symp. Ser. 51 (17), 79 (1955).
87. G. T. Pruitt and H. R. Crawford, Paper presented at December meeting, (Houston,
Texas), A.I.Ch.E., 1963.
88. M. Reiner, Lectures on Theoretical Rheology. North-Holland Publ., Amsterdam,
1960.
89. L. P. Reiss and T. J. Hanratty, A.I.Ch.E. J . 9, 154 (1963).
90. J. F. Ripkin and M. Pilch, Technical paper 42B, St. Anthony Falls Hydraulic
Laboratory, University of Minnesota, Minneapolis, Minnesota, 1963.
91. R. A. Ritter, Ph.D. thesis, University of Alberta, Edmonton, Canada (1962).
92. R. S . Rivlin, in Phknomknes de Relaxation et de Fluage en Rhkologie Non-LinCaire,
p. 83. C.N.R.S., Paris, 1961.
93. N. W. Ryan and M. M. Johnson, A.I.Ch.E. J . 5, 433 (1959).
94. J. G. Savins, A.Z.Ch.E. J . 8, 272 (1962).
95. J. G. Savins, SOC.Petrol. Engrs J . 4, 203 (1964).
96. J. G. Savins, G. C. Wallick, and W. R. Foster, SOC.Petrol. Engrs. J . 2, 211 (1962).
97. R. S. Schechter and E. H. Wissler, Nucl. Sci. Eng. 6, 371 (1959).
98. J. Schenk and J. Van Laar, Appl. Sci. Res. A7, 449 (1958).
99. W. R. Schowalter, A.1.Ch.E. J . 6, 24 (1960).
100. M. J. Shah, E. E. Petersen, and A. Acrivos, A.Z.Ch.E. J. 8, 542 (1962).
101. R. G. Shaver and E. W. Merrill, A.I.Ch.E. J . 5, 181 (1959).
102. C. R. Shertzer and A. B. Metzner, Trans. SOC.Rheology (in press).
103. E. M. Sparrow, J. L. Novotny, and S. H. Lin, A.1.Ch.E. J. 9, 797 (1963).
104. R. I. Tanner, Rheol. Acta 3, 21, 26 (1963).
105. D. G. Thomas, Chem. Eng. Progr. Symp. Ser. 57 (32), 182 (1960).
C3961
HEATTRANSFER FLUIDS
IN NON-NEWTONIAN
106. D. G. Thomas, A.1.Ch.E. J. 6, 631 (1960).

107. D. G. Thomas, in Progress in International Research on Thermodynamic and
Transport Properties (J. F. Masi, ed.), p. 669. Academic Press, New York and
Am. SOC.Mech. Engrs., New York, 1962.
108. D. G. Thomas, A.l.Ch.E. J. 8, 266 (1962).
109. D. G. Thomas, Ind. Eng. Chem. 55 ( 1 I), 18; (12), 27 (1963).
110. C. Tien, Can. r. Chem. Eng. 40, 130 (1962).
I I 1. Y. Tomita, Chem. Eng. (Tokyo) 23, 66 ( I 959).
112. Y. Tomita, Chem. Eng. (Tokyo) 23, 525 (1959).
113. Y. Tomita, Bull. J S M E (Japan Soc. Mesh. Engrs.) 4, 77 (1961).
114. H. L. Toor, Ind. Eng. Chem. 48, 922 (1956).
115. H. L. Toor, Trans. SOC.Rheot. 1, 177 (1957).
I 16. C. Truesdell, Remarks offered during the Symposium on Second-Order Effects
in Elasticity, Plasticity and Fluid Dynamics, Haifa, 1962.
117. R. M. Turian and R. B. Bird, Chem. Eng. Sci. 18, 689 (1963).
118. R. D. Vaughn, Ph.D. thesis, University of Delaware, Newark, Delaware, 1956.
119. D. T. Wasan, C. L. Tien, and C. R. Wilke, A.Z.Ch.E. J, 9, 567 (1963).
120. C. S. Wells, Jr. Preprint 64-36; presented at January Meeting A.I.A.A., 1964.
121. J. L. White,Ph.D. thesis, Universityof Delaware, Newark,Delaware (in preparation).
122. J. L. White and A. B. Metzner, in Progress in International Research on Thermo-
dynamic and Transport Properties (J. F. Masi, ed.), p. 748. Academic Press,
New York and Am. SOC.Mech. Engrs., New York, 1962.
123. J. L. White and A. B. Metzner, J. Appl. Polymer Sci. 7, 1867 (1963).
124. J. L. White and A. B. Metzner, Trans. SOC.Rheol. 7, 295 (1963).
125. I. R. Whiteman and D. B. Drake, Trans. A S M E 80, 728 (1958).
126. W. R. Wilcox, Ind. Eng. Chem 50, 1600 (1958).
127. W. Wilkinson, Non-Newtonian Fluids. Macmillan (Pergamon), New York,
1960.
128. M. D. Winning, M.Ch.E. thesis, University of Alberta, Edmonton, Canada, 1948.
129. E. H. Wissler and R. S. Schechter, Chem. Eng. Progr. Symp. Ser. 55 (29), 203 (1958).
130. J. Yau and C. Tien, Can. J . Chem. Eng. 41, 139 (1963).
Radiation Heat Transfer
between Surfaces
.
E. M SPARROW
Heat Transfer Laboratory. Department of Mechanical Engineering
University of Minnesota. Minneapolis. Minnesota
I . Introduction . . . . . . . . . . . . . . . . . . . . . 400
.
I1 The Gray. Diffuse Enclosure . . . . . . . . . . . . . . 401
A. The Basic Postulates of the Computation Method . . . . 401
I11. Interchange Computation for Gray. Diffuse Enclosures . . . 407
.
IV Diffuse Angle Factors . . . . . . . . . . . . . . . . . . 411
A. Shorthand Methods . . . . . . . . . . . . . . . . . 412
B. Contour Integral Method . . . . . . . . . . . . . . . 414
V. Generalization of Enclosure Computations: Integral Equations 41 8
VI . Solution of the Integral Equations . . . . . . . . . . . . 420
A . Numerical Methods . . . . . . . . . . . . . . . . . 421
B . Approximate Analytical Methods . . . . . . . . . . . 423
VII . Radiant Emission and Absorption Characteristics of Cavities . 425
.
.
A . Reciprocity Theorem for Cavities . . . . . . . . . . .
B . Results for LY and . . . . . . . . . . . . . . . .
VIII Radiant Transmission Characteristics of Passages . . . . . .
426
421
430
IX . Radiating-Conducting Fins . . . . . . . . . . . . . . . 432
A. The Single Plate-Type Fin . . . . . . . . . . . . . . 433
B. Radiant Interaction between Fin and Tube Surfaces . . . 436
C . Radiant Interaction between Fin Surfaces . . . . . . . 440
D . Radiant Interaction between Fin and Tube. and between
Neighboring Fins . . . . . . . . . . . . . . . . . . 443
E . Related Problems . . . . . . . . . . . . . . . . . . 444
.
X Specularly Reflecting Surfaces . . . . . . . . . . . . . . 445
A . Enclosures Containing Plane Specular-Diffuse Surfaces . . 445
.
B Nonplanar Specular-Diffuse Surfaces . . . . . . . . . 449
References ..................... 450
E. M. SPARROW
I. Introduction
Renewed interest in radiation heat transfer processes has been sti-

mulated by modern technological developments, especially in connection
with space flight and with high-temperature energy sources. For instance,
in atmosphere-free space, thermal radiation would be the basic transfer
mechanism by which waste heat could be dissipated from a space
vehicle. As other examples, one may cite the thermal radiation from hot
gas streams such as are generated in arc-jets and at a blunt nose cone
during atmospheric re-entry.
T h e aforementioned illustrations include two basic types of radiant
transfer processes. T h e first type relates to radiant interchange between
surfaces that are separated by a medium which does not participate
in the radiation process (e.g., a transparent gas or a vacuum). T h e second
type relates to radiant transport in participating media such as an
emitting-absorbing gas or a semitransparent solid. T h e analysis and
solution of radiation problems involving participating media has been
discussed by Cess ( 1 ) in an article that appeared in the preceding volume
of this series. T h e present paper will be concerned with radiant inter-
change between surfaces which are separated by nonparticipating media.
T h e aim of this article is to present a coherent picture of calculation
methods and results for radiant interchange between surfaces. We begin
with a discussion of the analysis for gray, diffuse enclosures which
includes a generalization of presently available methods. Angle factors,
which are necessary for such calculations, are considered in some detail.
T h e gray, diffuse model is extended to systems in which there is a con-
tinuous variation of the radiant flux along the surfaces. Such a general-
ization requires that the radiant interchange be governed by integral
equations rather than algebraic equations as is the case when the radiant
flux is uniform over each surface in the enclosure. Various solution
methods for such integral equation systems are discussed. Results are
presented for the emission and absorption characteristics of cavities
and also for the energy transport through various passages.
T h e use of extended surfaces (i.e., fins) has been proposed for increasing
the radiating area of space vehicles. The heat transport in such situa-
tions is controlled by the interaction between the heat conduction in the
fin and the radiation from the fin surface. The theory of radiating-
conducting fins will be described in detail and typical results presented.
As will be demonstrated later, the reflection of radiant energy need
not necessarily be diffuse. Indeed, there may be a significant specular
component. Methods of analysis for specularly-reflecting surfaces are
being devised and these will be discussed here. T h e forthcoming
presentation represents the author's view of his subject as of the time
of writing, July, 1963.
[4001
RADIATIONHEATTRANSFER
BETWEEN SURFACES
11. The Gray, Diffuse Enclosure
I n calculating the radiant interchange at any surface, it is necessary

to take account of radiation arriving at that surface from all directions
in space. T o make certain that all radiation, whether outgoing or in-
coming, is fully accounted for, one figuratively constructs an enclosure
and specifies the thermal state and radiation properties on each surface
of the enclosure. All of the surfaces need not be actual material surfaces.
For instance, an open window could be a surface of an enclosure; it
would have the property that all of the radiant energy incident upon it
is absorbed, while its effective emissive power would be equal to the
radiant energy streaming into the enclosure through the open window.
Radiant interchange within an enclosure is typically computed by
applying a model based on five postulates. Before going on to a descrip-
tion and generalization of the actual computation procedure, these
postulates will be discussed in detail.
A. THEBASICPOSTULATES METHOD
OF THE COMPUTATION
1. Isothermal Surfaces
T h e first postulate is that each one of the surfaces in the enclosure is
isothermal. In practice, physical surfaces which are nonisothermal can
be subdivided into smaller sections, each of which is approximately
isothermal. Thus, there is no difficulty in satisfying the first postulate.
2. Gray Surfaces
It is postulated that the absorptance, 3, of a given surface is equal to
its emittance, z. T o illuminate the significance of this assumption, it is
helpful to consider the definitions of the emittance and absorptance.
It is known that a black body at surface temperature T, emits an energy
quantity ebA(Ts)dh per unit time and area in a wavelength band dh.
Similarly, a nonblack body emits el( T,)e,,( Ts)dh, wherein cA(Ts) is
the monochromatic emittance at temperature T, . Over the entire wave-
Jr
length range, the black and nonblack bodies, respectively, emit
ebA(T,J dh and J," el( T,)e,,( Ts)dh. T h e emittance, z, of a surface is
defined as the ratio of the energy emitted by the surface to that emitted
by a black body at the same temperature
E. M. SPARROW
This is called the hemispherical emittance inasmuch as it relates to

energy emitted into the entire hemispherical space above the surface.
T o derive an expression for the absorptance a, one begins with an
incoming stream of radiation which brings to the surface an energy
quantity ei, dh per unit time and unit area in a wavelength band dh. Of
this, al(T,)eil dh is absorbed by the surface, wherein ad(T,) is the mono-
energy absorbed over all wavelengths is Jr

chromatic absorptance of a surface at temperature T, . T h e amount of
ad( T,)eia dh. Then, forming
the ratio of the absorbed to the incoming radiant flux, one has
Inspection of Eqs. (1) and (2) reveals a basic difference between E

and 3 namely, that E is a property of the surface while 01 depends both
on the characteristics of the surface and on the nature of the incident
radiation ei . T h e conditions under which 01 = E may now be investigated.
According to Kirchoffs law, the monochromatic emittance and absorp-
tance are equal, al(T,) = E~(Ts).With this, it is clear that LY = E under
two conditions. T h e first corresponds to the situation in which
eia = eba(T,s);i.e., the incoming radiant flux is black body radiation at
the surface temperature T, . This situation is of limited technical im-
portance. The second is the gray body condition, where in the mono-
chromatic values are independent of wavelength, i.e., aJ.(T,) = Ed( Ts) =
const.
Materials characterized by values of Ed or CYI which are independent
of wavelength over the entire range 0 < < h 00 are typically not found
in nature. Strictly speaking therefore, there are no gray bodies. However,
many engineering materials are partially gray; i.e., they are gray for
some range of wavelengths. In particular, many materials are nearly gray
in the infrared range. If such materials were operated at moderate tem-
peratures, then the ebl which appears in the integrals of Eq. ( 1 ) is nearly
zero for A values which lie below the infrared range. Therefore, under
these conditions E q (T , ). Furthermore, if such a surface were to be
irradiated from a source of moderate temperature, it follows from Eq. (2)
by the same reasoning that 01 C I ~ ( T,). Therefore, the condition LY = E
can also apply to partially gray surfaces.
3. DifJuse Rejection
The directional distribution of the radiant energy reflected from each
surface in the enclosure is assumed to follow Lamberts cosine law.
Such surfaces are referred to as diffusely reflecting inasmuch as the
~4021
HEATTRANSFER
RADIATION BETWEEN SURFACES
intensity of the reflected radiation is independent of the angle of reflec-

tion. Since all radiation is reflected uniformly regardless of the direction
from which it came, it is clear that contact with a diffusely reflecting
surface completely obliterates the past history of the radiation. There-
fore, from the computational point of view, there is no need to keep
account of specific rays as they inter-reflect between the surfaces of the
enclosure.
Experiments by Eckert (2), carried out with a beam of black-body
radiation impinging normal to a variety of surfaces, indicate a preferential
reflection at the specular angle1 as well as a general background of dif-
fusely reflected radiation. Very recently, detailed experiments by
Birkebak (3) have shown that a given surface may be either a diffuse
reflector or a specular reflector depending upon the wavelength of the
incident radiation. His investigation was carried out with ground nickel
surfaces of controlled roughness, the incident beam arriving at the
surface under an angle of 10 degrees to the normal. T h e reflected
radiation was measured at a large number of angular positions through-
out the hemisphere above the surface and analyzed according to wave-
length by a monochrometer. A sampling of Birkebak's results2 is shown
in Fig. 1. T h e figure has five parts, each corresponding to a given
+
wavelength. T h e angles and 8 are explained in sketches at the lower
right of the figure. T h e parameter R is the ratio of the reflected radiant
energy at angle +, 8 to that reflected at the specular angle = 0", +
8 = 10". Inspection of the figure reveals that for the shortest wavelength,
the angular distribution is essentially independent of 4, although it
falls somewhat below the Lambert cosine curve. As the wavelength
grows longer, a definite specular peak emerges. At wavelengths of IOp and
higher, one would have to regard the surface as specular.
T h e foregoing discussion suggests that the diffuse surface may be
more of an abstraction than a reality. T h e realization that this is indeed
so, has prompted the formulation of analytical methods for computing
radiant interchange between specularly reflecting surfaces. These cal-
culation methods will be described later.
4. Dzffuse Emission
If it is assumed that the emitted radiation is diffusely distributed as
is the reflected radiation, then there is no need to make a distinction
between these two energy fluxes as they leave a surface. It is therefore
Angle of reflection equals angle of incidence.

a The measured surface roughness was 0.86 p.
E. M. SPARROW
,.o
8
6
R
4
SURFACE
10
6
R
4
10
6
R
4
R':L
2
GI O"1
x=loP
4 -
2 70- 60 50 40 30
e DEGREES
FIG. I . Directional reflectance of a nickel surface.

HEATTRANSFER
convenient to introduce a quantity which is the sum of the emitted and

reflected radiation. This sum is sometimes called the radiosity and is
denoted by the symbol B. If H represents the incident radiant flux,
then
B = coT4 p H + (3)
wherein the units of B and H are energy per unit time and area. T h e
reflectance is denoted by p. I n general, p = 1 - a , and for a gray surface
p=l--E (4)
Experiments relating to the directional distribution of emitted radia-
tion for several metallic and nonmetallic materials are reported by
Eckert (2). T h e total radiation (in contrast to monochromatic) was
measured at various angles and compared with that of a black body. T h e
results indicate that metallic surfaces follow the cosine law fairly well for
angles up to about 50" (measured with respect to the surface normal);
for larger angles, the emission is larger than that predicted by the cosine
law. For nonmetallic surfaces, the cosine law is closely followed up to
angles of 60 to 70", whereupon the emission drops below that of the
cosine law. These experiments indicate that the diffuse surface is also an
abstraction as far as emission is concerned, but it appears to be a more
realistic abstraction than is the cosine law description of the reflection.
5 . Surface Distribiction of Leaving Radiant Flux.

Perhaps least known of the five basic postulates is the requirement
that the radiant flux leaving any surface of the enclosure is uniformly
distributed over that surface. This condition arises through the use of
diffuse angle factors for interchange between finite surfaces. T o illumi-
nate this, it is useful to return to the definition of the angle factor. This
is a quantity which represents the fraction of the radiant energy leaving
one surface which arrives at a second surface. In particular, for diffusely
distributed radiant energy leaving a surface 1 and incident upon a
surface 2, the angle factor is given by
wherein B, is the radiosity of surface 1 which, in general, can be a

function of position across that surface. As indicated on Fig. 2, r is
the length of a connecting line between elements dA, and d A z , and
W I
E. M. SPARROW
d A,
FIG.2. Nomenclature and coordinates for computation of diffuse angle factors.
and & are angles formed by the respective surface normals and the
connecting line.
If the radiosity were to vary across surface A , , then it can be seen
from Eq. ( 5 ) that the angle factor would depend upon the thermal and
radiation parameters of the system as well as on the geometrical con-
figuration. On the other hand, if B, were uniform over A , , the angle
factor expression reduces to
In this form, the angle factor is a purely geometrical quantity. Equation

(6) is the standard representation for the angle factor for diffuse inter-
change between two finite surfaces.
T h e use of angle factors based on Eq. (6) implies the assumption that
the radiosity is uniformly distributed over the surface, Additionally,
it may be noted from Eq. (3) that an isothermal surface is a surface of
uniform radiosity only if the incident radiant energy is uniformly
distributed.
T h e postulate of uniform surface distribution of radiosity is tacitly
imposed whenever radiant interchange between finite surfaces is com-
puted by standard methods. However, this restriction may be removed
if the computation is made for interchange between infinitesimal surface
elements. Computations of this type have been recently carried out and
will be dzscribed later.
~4061
HEATTRANSFER
111. Interchange Computation for Gray, Diffuse Enclosures
Several computational procedures are available for determining the

radiant interchange in gray, diffuse enclosures: the radiosity method (2),
Hottels method (4), and Gebharts method (5). Additionally, Oppen-
heim (6) has devised an electric circuit analogy based on the radiosity
method. All of the aforementioned computational procedures are based
on the five postulates which have been discussed in the preceding
section of this article.
A thoroughgoing evaluation of the various computational methods
has recently been made (7). I t was demonstrated that when each of the
methods was extended to its most generalized form, there were no
essential differences between them. Indeed, the final set of equations
from which numerical results are extracted were shown to be identical.
I t is the present intent to develop an alternate computational procedure
for the gray, diffuse enclosure which has the fetaure that the heat fluxes
(or temperatures) are computed directly. This is in contrast to the prior
methods wherein intermediate quantities such as the radiosity, reflected
radiant flux, and absorption factors are computed.
T h e derivation begins with a radiant flux balance at a typical surface
K of an enclosure, for example, Fig. 3. T h e radiosity of such a surface
FIG. 3. Enclosure with N finite surfaces.
follows, from Eq. (3) as

B, = kUTk4 f (1 - k ) H k (7)
wherein the gray body condition (4) has already been introduced. T h e
incident radiant flux Hkis quite evidently related to the radiation which
leaves the other surfaces of the enclosure. For instance, the rate at which
radiant energy leaves surface (1) in all directions is B,A, . Of this, an
14071
E. M. SPARROW
amount B,A,F,-,/A, arrives per unit time and area at surface k . However,
according to the reciprocity relation for diffuse angle factors
With this, the radiant energy leaving surface (1) and arriving at k per
unit time and area becomes B,F,-, . But, radiant er.ergy may arrive
at surface k from all of the surfaces of the enclosure, and the total is
simply a summation of terms such as the foregoing contribution from
surface (1). From this, it follows that
+
BP = v T k 4 ( 1 - E , ) 2 B,F,-,
i=l
, k = 1,2,3, ..., N (9)
A relationship such as Eq. (9) may be applied at every one of the N

surfaces in the enclosure.
T h e local rate of heat transfer q per unit area at a typical surface in
the enclosure is the difference between the emitted radiant flux and the
absorbed portion of the incident radiation.
qk I D
= E L U T ~- L~H~ (10)
Upon eliminating the incident energy H , between Eqs. (7) and (10) and
introducing the gray body condition (Y = E , there follows
B k = uTk4 - [(I - Ek)/Ek]q1; (11)

I n turn, the radiosity may be eliminated from (9) and this gives, after
some rearrangement,
Y N
-q k= ; . C ( T k 4 - ~ ~ ) F k . . i f C (- lE , 9) . ~ F ~ - ~k ,= 1 , 2 , 3,..., N
Ek i=l i=l 1
(12)
wherein the identity ZE, FkPi = 1 has been used.
T h e foregoing constitute N linear3 algebraic equations, which inter-
relate the 2N quantities of interest: N surface temperatures and N heat
fluxes. In general, one would prescribe temperatures or heat fluxes to
the extent that N quantities would be known. Then, the other N quan-
tities could be readily solved. In particular, most electronic computers
have standard programs for solving linear algebraic equations.
* Since the temperatures enter only to the fourth power, the equations can be regarded
as being linear in T'.
BETWEEN SURFACES
Equation (12) separates the heat transfer process into two basic com-
ponents. T h e first summation is the heat flux due to black-body inter-
change, while the second summation represents the increment in the
heat flux due to interreflections.
According to the foregoing formulation [i.e., Eq. (12)], it would
be necessary to re-solve the set of linear algebraic equations for every
group of thermal boundary conditions which might be of interest. It
will now be demonstrated that the thermal boundary conditions can be
separated from the radiative and geometrical properties of the system.
As a consequence, the solution of the linear equations can be carried
out independently of the particular thermal boundary conditions.
To begin the generalization, consider an enclosure in which all sur-
faces are at zero temperature except surface j which is at temperature T j .
Let ~ ( jrepresent
) a dimensionless heat flux defined as
wherein the superscript ( j )is meant to delineate the case: Ti # 0, Ti= 0

for all i # j. For these conditions, Eqs. (12) reduce to
lpy)/l = -Fl-j + zN
i-1
( 1 - i)(v;J)/a)<-i (144
...
...
or, in general,
wherein the aj, = 0 when j # k, and aj, = 1 when j = k. This set

of N linear algebraic equations can be solved for the N values of q k j )
without any recourse to the surface temperatures.
A set of N equations as represented by Eqs. (15) can be written for
each of the surfaces j in the enclosure, i.e., j = 1, 2, 3, ..., N . Therefore,
there are such N sets in all, each of which can be solved independently
of the surface temperatures.
~4091
E. M. SPARROW
Next, consider a general enclosure in which the surface temperatures

T , , T , , ..., TN may take on arbitrary values. This general problem may
be subdivided into N subproblems, each characterized by the condition
that one surface is at its temperature T while the other surfaces are at
zero temperature. Because of the linearity of the situation, the solution
for the general problem may be found by superposing the solutions of
the subproblems,
qI = v1 ) u T1 4 + vi2)uT24 + ..* = z
N
j=1
v:j)uT; (164
...
wherein qdl) refers to the case in which T , is the only nonzero tempera-
ture, @) refers to the case in which T2 is the only nonzero temperature,
and so forth.
T h e computation is further simplified by making use of a reciprocity
relationship between the q~ functions. To derive this relationship,
suppose again that all surfaces are at zero temperature except surface K,
which has temperature Tk . For this condition, the net heat transfer at
surface j is Aj#)aTk4. Next, consider the case in which all surfaces are
at zero temperature except surface j, which has temperature T i . Cor-
respondingly, the net heat transfer at surface k is Akv$)aTj4.But, when
Tj = T k , there can be no net interchange between surfaces j and k,
thus
With this result, one can return to Eqs. (16) and write
...
HEATTRANSFER
T h e essential point to be noted is that the q1 computation requires only

the rp(l) functions, the q2 computation requires only the qP)functions,
and so forth. Consequently, if one is interested in the heat flux at a
particular surface in the enclosure, say surface k, then the linear equa-
tions (1 5 ) need be solved only once for the corresponding
To achieve an alternate and perhaps more satisfactory form for the
heat transfer results (18), one may note that energy conservation requires
Upon introducing this into (18c), we get
Tk4), k = 1,2, 3, ..., N (20)
T h e net heat flux at any surface is thus seen to be a weighted sum of

fourth power temperature differences. T h e weighting factors are the
dimensionless heat fluxes r#) which are obtained by solving the linear
equations (1 5). Readers who are familiar with Hottel's method may note
a similarity in the role played by the qjk)and that played by Hottel's 3.
T o summarize the present formulation, it may be noted that the essen-
tial computational equations (12), (15), and (20) involve only heat
fluxes and temperatures. I n prior methods, the computational equations
give either radiosities, reflected radiant fluxes or absorption factors
which are subsequently utilized to determine the heat fluxes.
IV. Diffuse Angle Factors
A knowledge of diffuse angle factors is a requisite for the execution

of radiant interchange calculations. Such information is available from a
variety of sources. Among these, perhaps the most extensive compila-
tions are to be found in the texts of Eckert and Drake (2), Hottel (4,
Jakob (8),and Kreith (9); and in the reports of Hamilton and Morgan (10)
and of Leuenberger and Pearson (ZI).
T h e standard definitions of diffuse angle factors are restated here for
completeness.
(a) Between a pair of infinitesimal surfaces
E. M. SPARROW
(b) Between an infinitesimal and a finite surface
(c) Between a pair of finite surfaces

cos p, cos pz
dA, dA,
T h e angles 8, and /3, and the distance Y are illustrated in Fig. 2.

To obtain numerical results from any of the foregoing equations, the
first step is to evaluate cos PI, cos P 2 , and r in terms of a convenient
coordinate system. Thus, considering rectangular coordinates, the nor-
mal to dA, may be characterized by the direction cosines I, , m, , and n , ,
where I , is the cosine of the angle between the normal and the x-axis,
and so forth. Similarly, the normal to dA2 is characterized by I,, m 2 ,
and n 2 . T h e connecting line Y may be thought of as a vector whose
components are (xz - x,), (y, - y,), and (zZ- x,). With these, one can
derive
cos B1 = [4(%- x1) + +
%(Y2 - Y l ) -4 l / Y (244
cos P 2 = [4hl - x2)+ m2(y1 - Y2) + nz(z1 - z2W (24b)
+
= (Jcz - X J 2 +
(y2 - y d 2 ( z 2 - zJ2 (24c)
These equations are in themselves sufficient for evaluating Eq. (21a).
Additionally, if the direction cosines are specified functions of position
across the finite surfaces A , and A , , the foregoing equations also serve
as a sufficient description of the integrands in Eqs. (22a) and (23a).
I n practice, the integrations involved in the latter equations are fre-
quently quite difficult.
It is the present purpose to summarize some shorthand analytical
methods for determining angle factors and also to describe a recently
formulated approach which appears to simplify the angle-factor cal-
culation. Graphical and analogy methods are available (2, 8, 10) but
will not be treated here.
A. SHORTHAND
METHODS
A well-known and useful technique for extending the utility of
available information is angle-factor algebra (e.g., 8, ZO). By this techni-
[4121
BETWEEN SURFACES
que, the unknown angle factor for some new configuration is determined
by adding together (or subtracting away) the angle factors for known
configurations. Recently, angle-factor algebra has been used to consider-
able advantage for differential as well as finite surfaces. An illustration
of differential angle-factor algebra will now be given. Suppose that it
is desired to compute the angle factor FdAlPA, from a differential ring
element dA, to a disk A , as pictured in Fig. 4. T o facilitate the deriva-
FIG.4. Angle factor derivation for ring-disk system.
tion, it will be convenient first to find the reciprocal angle factor

dFA,-dA,. It is intuitively clear that the radiant energy leaving the disk
and striking the ring element is simply the difference between the energy
passing through a disk at x and a second disk at (x dx), thus +
dFA2-dAl = Fd-d(x) - Fd-d(x + dx) (25)
wherein FdPnis the angle factor between two coaxial disks. From the
definition of the derivative, it follows that F d - d ( X ) - F d - , ( X +
dx) =
-(dFtlPd/dx) dx, and correspondingly
dFAZ-dA, - -(dFd-d/dX) dx
(26)
Then, from the reciprocity relation for diffuse angle factors, there follows
Upon introducing the disk-to-disk angle factor from ( 8 ) , Eq. (31-45),

and differentiating, one obtains
-
x2 + 3
____
FdA1- A 2-
(XA + 1)h -Xx, X = x/2R
A remarkably simple method applies when the surfaces involved are

greatly elongated in one of their dimensions. For example, the surfaces
pictured in Fig. 5 are very long in the direction into and out of the plane
[4131
E. M. SPARROW
FIG. 5. Angle factor derivation for surfaces which are long in the direction normal
to the page.
of the page. For this condition, the angle factor for interchange between
two infinitesimal surfaces (left-hand sketch of Fig. 5 ) can be computed
from
dFdAl-dAz= @(sin 0) (294
wherein 0 is formed by the normal at dA, and the connecting line between
dA, and dA, . T h e angle factor for interchange between an infinitesimal
surface and a finite surface (center sketch) follows directly by integration
of Eq. (29a) as
FdAl-A, = ;(sin e, + sin e,) (29b)
T h e angles 6, and 6, are formed by the normal at dA, and the connecting
lines drawn to the extremities of A, . T h e angle factor for interchange
between finite surfaces (right-hand sketch) is most easily computed by
Hottels string method (4). I n this method, strings are imagined to be
tightly stretched between the end points of A, and A , . T h e angle factor
FA,-A,is then determined by adding the lengths of the crossed strings
and subtracting away the lengths of the uncrossed strings, and then
dividing by twice the length of A , . I n terms of the nomenclature of the
figure, this becomes
FAl-A2
1
= - [(UC
2L
-
+ a) (a+ G)]
- (294
B. CONTOUR
INTEGRAL
METHOD
T h e presently standard representations of the angle factors for finite
surfaces require that integrations be carried out across the participating
areas. Considering Eq. (22a), it is seen that the angle factor between an
infinitesimal and a finite surface involves a double integral, while the
angle factor for a pair of finite surfaces involves a quadruple integral
[Eq. (23a)l. By reformulating the angle factors in terms of contour
PI41
HEATTRANSFER
(i.e., line) integrals, the order of the integration is reduced, the double
integral being reduced to a single and the quadruple to a double. This
reduction of order is generally advantageous in the analytical evaluation
of angle factors and has even greater utility in the numerical evaluation
of angle factors.
Attention may first be directed to the angle factor between an in-
finitesimal area dA, and a finite area A,. With the aid of Stokes' theorem,
the standard representation, Eq. (22a), can be transformed (12) to a
contour-integral representation
T h e integrals appearing in Eq. (30) are to be taken around the contour C

which bounds the area A , .
I t is desirable to explain in detail the meaning of the quantities
appearing in the contour integral expression. T h e Z, , m, , and n, are the
direction cosines of the infinitesimal area dA,; these are constant during
the integration. T h e coordinates x1 , y1 , and x1 represent the location
of d A , , and these are also fixed during the integration. Additionally,
x 2 , y 2 , and x2 are coordinates representing any point on the contour C,
while Y is the distance between dA, and points on the contour C. T h e
quantities x 2 , y 2 , z 2 , and Y may vary during the contour integration.
I n practice, it may be convenient to orient the coordinate axes so that
the normal to dA, lies precisely along a coordinate line. Then, two of the
three direction cosines are zero and correspondingly, two of the three
integrals appearing in Eq. (30) may be deleted. Furthermore, additional
simplifications occur in situations in which one or more of the coordi-
nates are constant along any part of the contour C ;then, the corre-
sponding differentials are zero and part of the integrand is eliminated.
T h e direction of travel around C is :he same as that of an observer who
walks along the contour with his body aligned with the normal and
keeps the interior of A , at his left.
T o fix these ideas, consideration may be given to deriving the angle
factor between an infinitesimal area element and a circular disk as
pictured in Fig. 6. T h e element is located off-axis with respect to the
disk. Inasmuch as the normal to dA, points in the positive z-direction,
it follows that Z, = m, = 0, n, = + 1. Additionally, x 1 = R , ,
PI51
E. M. SPARROW
I h
FIG.6. Element and circular disk for illustrating contour-integral me.--od.
y, = z1 == 0. If polar coordinates are introduced, then x, = R, cos 8,

dx, = -R, sin 8 do, y , = R, sin 8, dy, = R, cos 8 d8; also x, = h.
With these, the contour integral (30) becomes
T h e integral is an elementary type which appears in standard tables.

After integration and substitution of limits, there follows
T h e foregoing example serves to illustrate the ease with which the

contour integral method may be applied. It has been the experience of
the present author that integrals of simpler form are usually encountered
in evaluating the contour-integral representation than would be encount-
ered in evaluating the hitherto standard area-integral representation.
T h e contour-integral representation (30) naturally separates itself into
three parts. Each one of the contour integrals can be evaluated indepen-
dently of the other two and of the values of l, , m, , and nl . Indeed,
except for sign, each one of the integrals appearing in Eq. (30) is itself
an angle factor. T h e integral multiplying lI is, when taken with the proper
sign, the angle factor between an element dA, having its normal along
the x-axis (m,= n, = 0) and A,. Thus, one may write
HEATTRANSFER
Returning to Eq. (30) with these definitions, there is obtained
FdA,-A, = * IlF(1,0,O)* m,F(O, * n,F(O, 0,

150) 1) (33)
T h e form of Eq. (33) immediately suggests a superposition principle

which may be interpreted as follows. Given an element dA, located
at x1 , y1 , z1whose orientation is defined by direction cosines 1, , m, , n,
and a finite area A , . T h e angle factor FdA,-A, is computed by linearly
superposing the angle factors associated with three basic elements
located at xI , y l , zl.T h e basic elements are oriented so that each one
has its normal lying along one of the coordinate axes. T h e weighting
factors in the superposition are the direction cosines.
T h e choice of the plus or minus signs in Eq. (33) is made according
to the following rules: the plus sign is affixed to the first term if dA, can see
+
A, when its direction cosines are l1= 1 , m, = 0, n1 = 0. Contrariwise,
the minus sign is affixed if dA, has these same direction cosines and
cannot see A,. Similar rules are formulated for the second and third
terms by rotating 1, m, and n in the foregoing statement.
T o illustrate the superposition principle, consider the computation of
the angle factor between an element with direction cosines I, = 0, m, ,
n, and a circular disk as pictured in Fig. 7. T h e angle factors for the
basic elements oriented as shown in the lower portion of the figure
are available in the literature (2, 10). Then, utilizing Eq. (33), there
follows
F d A l - A 2 = -m,F(O, 130)+n,F(O,O, 1) (34)
Since m, is negative for the inclined element shown in the figure, the
two contributions will add.
T h e contour-integral representation of the angle factor for interchange
between a pair of finite surfaces has also been formulated (12). T h e
FIG. 7. Illustration of the superposition principle.

PI71
E. M. SPARROW
derivation is carried out by applying Stokes' theorem to the double

area integral which appears in Eq. (23a). The end result of the trans-
formation is
In Y dx, dx, + In Y dy, dy, + In Y dz, dz, i

\ (35)
wherein C, and C, are, respectively, the bounding curves of the areas

A, and A , , and r is the distance between points on the contour C, and
points on the contour C, . It is seen from Eq. (35) that the calculation
of the angle factor requires successive integrations around the contours
C , and C, . If any of the coordinates are constant over some portion of
either C, or C, , then the corresponding differentials are zero and the
integration is thus simplified.
Application of Eq. (35)has been made in (12) with a view to illustrating
the technique of evaluating the successive contour integrals. Interested
readers are referred to that paper for additional details.
V. Generalization of Enclosure Computations: Integral Equations

It has already been noted that standard methods for computing
radiant interchange include the tacit assumption that the radiant flux
leaving each surface of the enclosure is uniformly distributed over that
surface. This assumption is, however, no longer necessary if the radiant
interchange calculation is carried out for infinitesimal surfaces rather
than for finite surfaces. Such an analysis permits not only a continuous
variation of the leaving radiant flux, but also permits a continuous
variation of the surface temperature and the surface heat flux.
T h e relationship between surface temperature and heat flux that cor-
responds to interchange between infinitesimal surface elements is most
easily found by generalizing Eq. (12), the latter being based on inter-
change between finite surface elements. T h e generalization can be
carried out by replacing summation with integration and by replacing
the finite angle factors F with infinitesimal angle factors dF. I n addition,
a running variable f is introduced (Fig. 8) to identify positions on the
surface of the enclosure [corresponding to the summation index i of
Eq. (12)]; also, f = x represents the particular position at which the
heat flux is being calculated [corresponding to the subscript k in Eq. (12)].
With these modifications, the generalization of Eq. (12) for continuously
varying radiosity, temperature, and heat flux is
HEATTRANSFER
FIG. 8. Enclosure with continuous variations of radiosity, surface temperature, and

surface heat flux.
wherein the identity
has been used to extract T4(x)from under the integral. T h e integration

in 5 is extended over all parts of the enclosure. T o achieve a solution,
Eq. (36) is to be satisfied at every surface location x in the enclosure
(corresponding to the requirement that Eq. (12) is to be applied for all
k = 1 , 2 , 3,..., N . )
This equation provides a means of determining the surface distri-
bution of the heat flux if the temperature is prescribed. Alternatively,
if the heat flux is prescribed, then the surface distribution of the tempera-
ture can be determined. In either case, the unknown appears under the
integral sign. Such equations are called integral equations. Therefore,
the generalization which permits continuous variations of the radiosity
transforms the mathematical structure of the problem from algebraic
equations to integral equations. A variety of methods have been utilized
in solving the integral equations of the radiant interchange problem.
Before going on to a discussion of these methods, it may be worthwhile,
for purposes of illustration, to write down the integral equations which
arise in a specific physical problem.
Consideration may be given to a circular cylindrical cavity of length L
and diameter d as pictured in Fig. 9. Suppose that the surfaces of the
cavity are isothermal (a condition which one tries to achieve in construct-
ing sources approximating black-body radiation) and that the emittance
E is uniform over all the surfaces. It will also be assumed that no radia-
tion enters the cavity through its opening.
Let 5 be the running coordinate along the lateral surface and let x
denote a particular position on this surface; correspondingly, 71 is the
running coordinate on the base surface and y denotes a particular
PI91
E. M. SPARROW
FIG. 9. Circular-cylindrical cavity.
position on this surface. It is convenient to write two integral equations

for this problem, one for the lateral surface and one for the base surface.
Upon specializing Eq. (36) to the cylindrical cavity, one finds
wherein = q/uT4and A, is the area of the cavity opening. In deriving

Eq. (37), the following identities have been used:
corresponds to interchange between the element

T h e angle factor FdA,-Ao
dA, and the cavity opening, and similarly for FdA,-A,.T h e element
dA, exchanges heat with both the lateral and the base surfaces, and this
accounts for the presence of two integral terms on the right side of
Eq. (37a). O n the other hand, the element dA, exchanges heat only with
the lateral surface and correspondingly, only one integral term appears
on the right side of Eq. (37b). Further details relating to this problem
may be found in (13).
Attention may now be given to solution methods which have been
utilized in solving the integral equations of radiant interchange.
VI. Solution of the Integral Equations
As a prelude to a general discussion of solution methods, mention

may be made of two geometrical configurations for which solutions can
~4201
HEATTRANSFER
be obtained with relative ease. T h e first of these is the spherical cavity.

For this case, the angle factor dFaAZpdA is a constant (14) and can thus
be extracted from under the integral &gns of Eq. (36). What remains
is an equation which is readily solvable in closed form, with perhaps
a numerical quadrature needed at the end if the prescribed q or T
distributions cannot be integrated analytically.
T h e second relatively simple configuration is the cylindrical-arc
cavity pictured in Fig. 10. T h e extension of the surface normal to the
FIG. 10. Circular-arc cavity.
plane of the figure is very long. For this case, the angle factor appearing
in Eq. ( 3 6 ) is (15)
dFdA,-d+ = a sin 3 I fi -f 1 d[ (394
and the integrals which must be executed are of the type
If this integral were twice differentiated with respect to x (with proper

cognizance of the absolute magnitude signs), it would be found that
i.e., the integral is exactly restored after the differentiations. T o proceed,

Eq. ( 3 6 ) may be twice differentiated with respect to x. Then, the integraIs
can be eliminated between the undifferentiated and twice differentiated
forms of Eq. (36). This leaves a second-order, ordinary differential
equation which is readily solvable (15).
A. NUMERICAL
METHODS
For more general configurations, there are a variety of solution
methods available. T h e most obvious procedure is to retrace the steps
14211
E. M. SPARROW
previously followed in deriving Eq. (36) from Eq. (12); in other words,
employing Eq. (12) as the finite difference form of Eq. (36). Perhaps the
only distinction between such a finite-difference representation and the
standard computation procedure is that now special effort would be
made to use a sufficient number of mesh points to insure close approxi-
mation of the integrals. Many present-day electronic computers have
standard programs which are capable of solving on the order of one
hundred linear algebraic equations.
Equation (12) is a trapezoidal-rule type of approximation of Eq. (36)
and is in this sense the simplest finite-difference representation. T h e
next order of approximation is provided by Simpsons rule:
f f ( x ) dx = dx [f(4+ 4 f ( 4 + 2f(2 A x ) + *.-- t . f ( b ) ] (40)

a 3
T h e Simpsons rule representation of Eq. (36) leads to a set of linear
equations as does the trapezoidal-rule representation, the essential
distinction being that the unknowns are multiplied by different numer-
ical coefficients. T h e Simpson rule representation has been employed
successfully by Perlmutter and Siege1 (16, 17, 18) for solving radiant
interchange problems in cylindrical tubes. Higher-order approximations
(four, five, etc., point formulas) may also be employed; these are available
in standard texts on numerical analysis.
Another numerical method is that of successive iterations. This is, in
a sense, the numerical analog of the analytical technique of successive
substitutions (19). T h e iterative technique for solving Eq. (36) will
now be described. Suppose, for concreteness, that the temperature distri-
bution is prescribed* and it is desired to determine the heat flux. One
begins by guessing a distribution of the heat flux q(@. For a given x,
the angle factor is a function of f alone. Thus, with the guessed distri-
bution of q ( [ ) and a prescribed x, the integration in the last term of
Eq. (36) can be carried out numerically. This gives a new value of q
at x. Next, another location x is selected and the integration repeated,
and a new value of q is thus obtained at this second x. Proceeding in
this way, a new q distribution is generated along the entire surface. T h e
new distribution of q is then compared with the previously guessed
distribution. If the agreement is not satisfactory, then the procedure
is repeated again and again until the q distributions from successive
outputs are essentially the same.
Experience with radiation problems has indicated that the direct
iteration procedure will converge without difficulty provided the equa-
All integrals in Eq. (36) involving T are therefore known.

r4221
BETWEEN SURFACES
HEATTRANSFER
RADIATION
tions involved are linear, either in q or T4.This condition is fulfilled by

Eq. (36). However, when both T and T4appear in the same equation,
as in problems of combined radiation and conduction, then instability
may occur. This difficulty can frequently be circumvented by employing
a weighted average of the former input and present output as input to
the next cycle of iteration.
B. APPROXIMATE
ANALYTICAL
METHODS
Aside from the purely numerical methods of solution, there are
available another group of methods which employ approximate repre-
sentations for either the unknown function or the angle factor. One of
the more promising among these is the Taylor's series method (18).
T o facilitate the discussion, it may be assumed again, for concreteness,
that the temperature distribution is prescribed. Therefore, all terms in
Eq. (36) involving T are known, and it remains to solve for q. T h e heat
flux q( t)is expanded in Taylor's series
This is then introduced into the last term of Eq. (36), and upon noting
that q ( x ) , (dqldx), , (d2q/dx2), depend on x and not 5, there follows
T h e braced quantities in the foregoing are functions of x alone

(after the integration in .$ is performed). Upon substituting Eq. (42)
into (36), it may be noted that the series expansion has transformed the
integral equation into a differential equation, the order of which depends
upon the number of terms which were retained in the series. The
differential equation is then solved for q, either analytically or numer-
ically. Boundary conditions are obtained by imposing a conservation con-
dition on the overall heat flux, i.e., [J, q dx = 01, by using whatever
symmetries are present, and by requiring that the original integral
equation be satisfied at selected points. Further details may be found in
(18).
W I
E. M. SPARROW
A second analytical procedure is based on the approximation of the

actual angle factor by a more manageable function. However, this
method appears to show promise only for angle factors of the form
(dFdA,-d+)/d[ = function Of I f -X I (43)

Such angle factors have been found to occur in cylindrical cavities and
tubes and in parallel-walled cavities and passages. T h e actual angle
factors are approximated by exponentials
(dFdA,-d..,t)/d[ = Ae-alE-xl + Be-bIt-SI + (44)

Consider first the case where one exponential is employed. With this,
a typical integral term of Eq. (36) takes the form
If I , is twice differentiated with respect to x, with proper cognizance

taken of the absolute magnitude signs, one obtains
Thus, the integral is restored by the double differentiation. I t follows

that the integral terms can be eliminated between the undifferentiated
and twice-differentiated forms of Eq. (36). By this, Eq. (36) is trans-
formed into a second order, ordinary differential equation. When two
exponential terms are employed as in Eq. (44),it can be shown that
integrals such as
f([)[Ae-al 5-rl + ge-blE-zl 1 df (47)
are restored by a combination of quadruple and double differentiation.

Correspondingly, Eq. (36) would be transformed into a fourth-order,
ordinary differential equation.
Boundary conditions are derived by the same means as has already been
discussed for the Taylor series method. T h e foregoing exponential
method appears to have first been employed in 1927 by Buckley (22)
for interchange in a cylindrical cavity. More recently, it has been applied
to a series of radiant interchange problems involving the cylindrical tube
(26, 27, 28, 20) and the rectangular groove cavity (22). As a recent
innovation, Perlmutter and Siege1 (28) have devised a correction proce-
F241
HEATTRANSFER
dure to improve the accuracy of the solutions, the accuracy of the un-
corrected solutions having been found inadequate for long tubes.
T h e last of the analytical methods to be discussed here is based on
the calculus of variations. T h e class of radiation problems to which
variational methods have thus far been applied (23, 20, 18) are character-
ized by the following form of integral equation:
= Q(4+
4%) I 4 5 ) K ( x ,5 ) d5 (48)
where Q(x) is a prescribed function of x and h is a known constant. T h e

function K(x, 6) is restricted to be symmetric in x and f , i.e., the
function is no way altered if x and ( are interchanged. T h e angle factors
for diffuse interchange in a circular cylindrical tube (20) and a parallel-
walled passage (23) display this symmetry property.
Corresponding to the integral Eq. (48) there is a variational expression
I (Eq. 13 of 23), which has a very special property. Namely, that when
a function A is found which makes I an extremum, this same function
is also a solution of the integral equation. T o find the function which
provides the extremum is very difficult. Instead, one attempts to come
as close as possible to the extremum, and in this way to get a close
approximation of the solution for A .
There are systematic procedures for seeking the extremum condition.
One such procedure is the Rayleigh-Ritz method. T h e detailed steps for
applying the variational method to problems of radiant interchange
are given in (23).
VII. Radiant Emission and Absorption Characteristics of Cavities
When radiant energy from an external source enters a cavity, it may

undergo a sequence of interreflections which provide additional oppor-
tunities for absorption at the cavity walls. Consequently, the radiant
energy absorbed within a cavity will exceed that of a plane area of identical
absorptance stretched tightly over the cavity opening. This is called
the cavity effect. Correspondingly, for similar reasons, the radiant
energy streaming out of a heated cavity will exceed that of a plane area
of identical emittance and temperature stretched across the opening
of the cavity.
T h e magnitude of the cavity effect is generally characterized by an
apparent emittance E, and an apparent absorptance a, . These quantities
are defined and discussed in the following paragraphs. Numerical
results will be presented in a later section.
P I
E. M. SPARROW
T h e apparent emittance is defined as:

rate of radiant efflux from a cavity
c, = (49)
rate of efflux from a black-walled cavity
wherein the nonblack and black cavities have the same shape and tem-
perature. Generally, the apparent emittance is computed for cavities
whose surfaces are isothermal. I n addition, radiant energy entering the
cavity opening from an external source is not considered in the computa-
tion of the apparent emittance. I n almost all instances, the cavity surfaces
have been regarded as gray, diffuse emitters and reflectors. Considera-
tion has been extended to nondiffuse conditions by DeVos (24) and
by Krishnan (25, 26,27). T h e first of these investigators presents numer-
ical results, but they are of uncertain validity; the second does not give
numerical results.
T h e definition of the apparent absorptance is
rate of absorption of radiant energy in cavity
a, =
rate of incoming radiant energy
I n computing a,, it is customary to consider the radiant interchange
process only for energy which enters the cavity opening from an external
source; the energy emitted by the cavity walls is not included, i.e., the
walls may be regarded as being at absolute zero. There are two limiting
distributions of the incoming radiant stream that have been considered:
either diffusely distributed across the cavity opening or as a bundle of
parallel rays. For each of these, consideration has been given to cavities
whose walls are either diffusely or specularly reflecting.
There is an interesting reciprocity relationship between E, and a,
which will now be discussed.
THEOREM
A. RECIPROCITY FOR CAVITIES
Consider a cavity whose walls are gray and into which enters a radiant
stream which is diffusely distributed across the opening. T h e a, for
such a cavity is equal to E , under any of the following conditions:
(1) T h e cavity walls are diffuse emitters and diffuse reflectors.
(2) T h e cavity walls are diffuse emitters and specular reflectors.
(3) T h e emittance and reflectance can have any directional distribu-
tion consistent with the requirement that the radiant flux leaving
a surface element be black and diffusely distributed when the
radiation incident upon the element is black and diffusely distri-
buted.
14261
HEAT TRANSFER
Cases (1) and (2) are especially important since they permit generaliza-
tion of available information.
T h e proof of the theorem is quite simple. Imagine a heated cavity
having a temperature T ; radiant energy streams out of the cavity at the
rate Qo . Next, suppose the cavity opening is closed by a black surface
at the same temperature T which radiates into the cavity. T h e presence
of the black surface creates a complete isothermal enclosure. I t is well
known that there is no net heat transfer at any surface of an isothermal
enclosure and also that the radiation within the enclosure is black and
diffusely distributed at all points.
Now, the isothermal cavity just described can be thought of as the
superposition of two physical situations: (1) a cavity with unblocked
opening having walls at temperature T ; and (2) a cavity having walls
at absolute zero with opening blocked by a black surface at temperature
T . As already noted, the net heat transfer is zero when situations
( 1 ) and (2) are superposed; also, the net heat transfer in situation ( I )
is Qo . Therefore, a cavity having walls at absolute zero will absorb Q,,
when irradiated by diffusely distributed radiation of temperature T ,
such as that provided by a black surface. From this, it follows that
a, = E,. T h e three conditions stated in the theorem follow directly
from the requirement that the radiation in an isothermal enclosure
must be everywhere black and diffusely distributed.
B. RESULTSFOR E, AND LY,
T h e apparent emittance E, has been determined by solving the

appropriate integral equations for the circular-cylindrical cavity (13),
the rectangular-groove cavity (28),the conical cavity (29),the vee-groove
cavity (30),and the spherical cavity (14). These results apply for cavities
with gray, isothermal walls which are diffuse emitters and reflectors.
According to the aforementioned reciprocity theorem for cavities, these
E , are equal to a, for diffusely distributed radiation entering a cavity
with diffusely reflecting walls.
T h e E , , LY, results for the circular-cylindrical cavity and the rectan-
gular-groove cavity have been brought together in Fig. 11. T h e former
is pictured in Fig. 9 along with dimensional nomenclature. T h e latter
is a groove of rectangular profile, depth L and height h, whose extension
in the direction normal to the plane of the paper is very great. I n Fig. 1 I,
E , is plotted against the ratio of L/r (cylindrical cavity) or L/h (rectangular
groove). Inspection of the figure reveals that when plotted in this
way, the E , is essentially the same for the cylindrical cavity as for
the rectangular groove. Thus, h appears to be an appropriate
W I
E. M. SPARROW
FIG. 1 1 . Emission and absorption characteristics of circular-cylindrical and rectan-

gular-groove cavities.
"hydraulic radius," just as it is in fluid mechanics for a parallel-plate

channel.
T h e departure of E , from E provides a measure of the cavity effect.
It is seen from the figure that the cavity effect is most pronounced for
surfaces of relatively low emittance. Additionally, the cavity effect is
accentuated in configurations of greater depth; however, for sufficiently
deep cavities, E, becomes independent of further increases in depth.
It is interesting to note how quickly E , approaches its limiting value.
For instance, for E = 0.9, there is little change beyond Llh or L p = 2;
a corresponding statement applies for the E = 0.5 results when L/ h or
L / r exceeds 5.
T h e cLI, CY,
results for the conical cavity and the vee-groove cavity are
plotted together in Fig. 12 as a function of the opening angle 6. T h e
results for the two cavities are seen to be essentially the same. T h e
cavity effect is most evident for surfaces of low emittance, and this is
accentuated at small opening angles, that is, closed-in cavities, T h e
limiting values at 0 = 0" for cone and vee-groove correspond respectively
to those of the cylindrical and the rectangular groove cavities of infinite
depth.
Results for the spherical cavity are presented in Fig. 13. T h e figure
also includes a small sketch which defines the opening angle $*. T h e
trends evidenced by the figure are essentially the same as those seen
in the prior figures. However, the a, for the sphere has a broader scope
than did the au for the previous cases. I n particular, the LY, applies
for any arbitrary spatial distribution of the radiation entering the spheri-
[4281
BETWEEN SURFACES
e (DEGREES)
FIG. 12. Emission and absorption characteristics of conical and vee-groove cavities.
C$T DEGREES
FIG. 13. Emission and absorption characteristics of a spherical cavity.

[4291
E. M. SPARROW
cal cavity. Additionally, the following closed-form expressions are
available
E, 1 0.5(1 - a)(l
=~/[- + COS+*)] (514
a,' = ./[I - 0.5(1 - a)(l + COS~*)] (51b)
I t is expected that the foregoing figures would be useful in connection

with the design of black-body cavities.
I n addition to the foregoing, one may find further information on the
absorption characteristics of cavities in the literature. T h e absorption of
a parallel-ray bundle in both diffusely and specularly reflecting cavities
has been determined for the vee groove (30) and rectangular groove (22).
Diffusely distributed radiation incident upon specularly reflecting sur-
faces has also been considered for these configurations. For the circular-
arc cavity, Fig. 10, computations of aa have been made for a ray bundle
incident upon diffusely or specularly reflecting surfaces (25); analysis
was also performed for diffusely distributed incoming radiation and
diffuse surfaces, but or, values are not reported (31).
VIII. Radiant Transmission Characteristics of Passages
T h e problem of radiant transport through conduits, open at both

ends, is closely related to the radiant interchange within cavities, and
the analysis and computations are carried out in a similar manner. Solu-
tions of the appropriate integral equations have provided results for the
transport of radiant energy between isothermal environments connected
by tapered tubes or gaps. T h e physical situation is shown schematically
in sketches appearing at the top of Figs. 14 and 15 which also include
dimensions and nomenclature. T h e first of these corresponds to the
tube, while the second is for the gap. T h e results apply to the case
wherein the walls of the connecting passages are adiabatic and are diffuse
emitters and reflectors of radiant energy.
T h e net rate of heat transfer Q between the isothermal environments
is plotted on the ordinate as a function of the aspect ratio of the passage,
with the half angle of taper appearing as a parameter on the curves.
T h e results are independent of the emittance and absorptance of the
passage. Inspection of the figures reveals that the energy throughflow
decreases with increasing passage length for a given angle of taper.
However, the rate of decrease depends quite strongly on the angle of
taper. I n particular, for passages of moderate and large taper angle,
W I
BETWEEN SURFACES
5 10 15 20 25 30 35 40
-
L
r,
FIG. 14. Radiant energy transport through a tapered tube.
15"
-
2
.-
-
*I -
I l l 1 1 I j I I
5 10 15 20 25 30 35 L
L
-
h
FIG. 15. Radiant energy transport through a tapered gap.
E. M. SPARROW
the energy throughflow decreases with increasing length only when the
passage is short; for longer passages, further increases in length have
very little effect on the throughflow. On the other hand, when the taper
angle is small, the energy throughflow is more sensitive to the length
of the passage. I t is also seen from the figures that for a given opening
dimension (rl or h ) and given length, the energy throughflow is larger
when the taper angle is large.
T h e dashed lines on the figures give asymptotic results for long
circular tubes and parallel-walled gaps. For the tube, the asymptote is
expressed by
and correspondingly for the parallel-walled gap
QPh 2 ln(L/h)- 1
(52b)
a(T14 -T,4) - L/h
In a recent paper, Perlmutter and Siege1 (18) have solved for the
radiant transport through a circular tube whose walls are specular
reflectors and diffuse emitters. It was found that the energy transport
through a specularly reflecting tube always exceeded that through a
diffusely reflecting tube. T h e same authors have also investigated the
combined transport of energy through a tube by convection and radia-
tion (16, 17).
IX. Radiating-Conducting Fins
I t is widely contemplated to employ extended surfaces (i.e., fins)

to augment the area available for radiating heat away from space vehicles.
T h e heat transfer in such situations is controlled by the interaction
between heat conduction within the fin material and radiation at the
fin surface. T h e conductive transport depends linearly on the tempera-
ture, while the radiative transport varies with the fourth power of the
temperature. T h e result of coupling these processes is a nonlinear
dependence of heat transfer upon the temperatures of the problem.
I n response to the need for heat transfer information, analytical stu-
dies have been carried out for a range of radiator configurations.
Typical among these is the fin-tube radiator pictured in Fig. 16. T h e
tubes contain a flowing fluid from which heat is to be extracted. The
present purpose is to -bring together and systematically describe the
L-4321
HEATTRANSFER
FIG. 16. Fin-tube radiator.
various analytical treatments and the results which have been obtained
from these. T h e presentation will proceed from relatively simple fin
configurations to more complex systems in which there is mutual
irradiation between radiator elements.
FIN
A. THESINGLEPLATE-TYPE
In the initial analyses of radiating fins, it was postulated that the radiant
interchange between the tube and fin could be neglected. Such a con-
dition is evidently achieved when the tube radius is much smaller than
the spacing between fins. As a result of this assumption, the problem
is reduced to the consideration of a single plate-type fin which radiates
freely to space and which in turn may be irradiated from space. Such
a fin is shown schematically in Fig. 17 along with dimensional nomen-
clature. I t is typically assumed that identical thermal conditions exist
along x = 0 and x = 2L; consequently, the line x = L is a symmetry line.
FIG. 17. Plate-type radiating fin.

E. M. SPARROW
I n the analytical formulation of the heat transfer problem, it has been

standard to assume that the heat conducted in the x-direction is much
smaller than that conducted in the x-direction (Fig. 17). This supposition
has recently been shown to be valid for aspect ratios L*/L which are in
the practical range for fin-tube radiators (32). In addition, because the
fin thickness, 2t, is generally very much less than the height 2L, the
dependence of the temperature on the transverse coordinate y is relatively
small and is neglected (thin-fin assumption). T h e physical system just
described has been analyzed in a number of investigations of which
(33)-(38)are representative.
Using the foregoing simplifying assumptions as a basis, one may for-
mulate the energy conservation principle for the typical element of fin
which is shown shaded in Fig. 17. T h e rate at which heat is conducted
into the element is given by Fouriers law as Q , = --k(2t)dT/dx per
unit length in the z-direction; further, the heat conducted out of the
element at x + dx is QZz+dz = Q, + (dQ,/dx) dx. It follows from these
that the net inflow by conduction is
2kt(dZT/dx2)
dx (53)
Under steady-state conditions, this must be balanced by the net outflow

of radiant energy q dA. If both the upper and lower surfaces of the fin
are participating in the radiant exchange, then dA = 2dx per unit
length in the z-direction. Therefore, upon equating the net conduction
to the net radiation and introducing dimensionless variables, one has
wherein
9 = T/Tb, x = XIL, (55)
T h e temperature Tb is the temperature at the fin base x = 0. T h e

parameter N , is frequently designated as the conductance parameter.
It is seen to represent a ratio of two conductances: that due to radiation
divided by that due to conduction. It is quite unlikely that N , values in
excess of five would be encountered in practice, with values of one or
two being realistic.
T h e net radiant outflow q per unit area is the difference between the
emitted radiant flux and the absorbed portion of the incident radiant
flux. T h e former is simply eaT4. T h e incident radiation is that coming
from space. I n the usual formulation, it is assumed that this radiation is
W41
HEATTRANSFER
uniformly distributed across the fin, (i.e., independent of x). If the in-
coming radiation is from a high-temperature source such as the sun,
then there would be a significant amount of energy contained in the near
infrared and visible wavelength ranges. T h e absorption properties of the
surface relative to such radiation can be different from the infrared
emittance, E , of the surface. T o account for such differences, it is common
to define an equivalent temperature of space T* such that E U T *repre- ~
sents the energy absorbed by the element per unit area. I n addition,
if radiation is incident on only one side of the fin, this is accounted for
by an adjustment in T*. With this, the net radiant outflow is
q = e 4 T 4 - T**], = a4 - (56)
wherein 9.* = T * / T b . Correspondingly, the energy balance equation
(54) becomes
dz'
___ - $*4
- NC[9.4(X)
1 (57)
in which the functional dependence upon X is explicitly displayed. T h e

appropriate boundary conditions are
19 = 1 at X = 0, dS./dX = 0 at X = 1 (58)
the latter being a statement of symmetry.

T h e governing differential equation for 6 reflects the nonlinearity of
the simultaneous process of internal heat conduction and surface
radiation. Solutions of Eq. (57) subject to the boundary conditions of
Eqs. (58) depend upon two parameters, N, and 6*. For the special case of
9." = 0, closed-form solutions in terms of complete and incomplete beta
functions can be derived. However, most investigators have chosen to use
numerical means which may be applied for any value of I?*.For example,
by employing a forward-integration scheme such as the Runge-Kutta
method, one may determine which of a succession of guessed values of
d9./dX at X = 0 leads to a solution satisfying the boundary condition
d6/dX = 0 at X = 1. When programmed for an electronic computer,
such a calculation proceeds quite rapidly.
T h e result of greatest practical interest is the overall rate of heat
transfer Qt from the fin. This information is usually presented in terms
of a fin effectiveness 7 defined as the ratio of the actual fin heat loss to
that from an ideal fin having infinite thermal conductivity. The tempera-
ture of such an ideal fin would be equal to Tb throughout. Thus,
E. M. SPARROW
in which both Q, and Qltlealrefer to a unit length of fin extending from

x = o to x = L .
A presentation of effectiveness results is made in Fig. 18. Inspection
of the figure reveals that the fin effectiveness decreases quite rapidly
with increasing N,. when N,. is small. For N, = 1, the effectiveness has
1.0
.a
.6
.4
*
00 I 2 3 4
Nc
FIG. 18. Effectiveness of a plate-type radiating fin.
already dropped to about 50 percent. This suggests that there is little

motivation for using fins having N,. values in excess of one. Indeed, the
optimum fin which provides the maximum heat transfer for a given
fin weight is characterized by N, values ranging from 0.87 to 0.6 as
19"ranges from 0 to 0.9.
B. RADIANTINTERACTION
BETWEEN FIN AND TUBE
SURFACES
I n investigating the effects of radiant interaction between fin and tube,
it is convenient to consider separately the cases wherein the participating
surfaces are black or nonblack. T h e black case will be treated first.
An end view of a fin-tube radiator is pictured in Fig. 19. T h e exten-
sion of the surfaces in the direction normal to the plane of the figure is
taken sufficiently large so that end effects are negligible; temperature
variations in this direction are also neglected. T h e effect of these simpli-
fying assumptions on the heat transfer characteristics of the fin has
recently been demonstrated to be of second order (32).I n addition, the
tube surfaces are assumed to be circumferentially isothermal at tempera-
ture T b .
WI
HEATTRANSFER
-4r +-----2~-4
FIG. 19. Fin-tube radiator (end view).
T h e energy conservation Eq. (54) is in no way affected by the radiant

interaction and continues to apply as before. However, the net radiant
heat flux, v, must be modified to include the radiant energy arriving
at an element of fin from the tube surfaces. T h e expression for y is
most easily derived by specializing Eq. (36) for E = 1. One begins by
associating dA, with an element of fin dA, . T h e integration over 5 is
extended over both tube surfaces (1) and (2) and over space. Inasmuch
as the tubes are isothermal, it follows that
for the tubes, while
for space. I n the foregoing, the equivalent space temperature T* is

taken to be a function of X since the tube may well shield certain portions
of the fin from radiant energy arriving from space.
With these, the expression for y~ follows directly by introducing
dimensionless variables into Eq. (36). Then, upon returning to the
energy conservation Eq. (54), one finds
T h e angle factors FdA,--A,and FdA,-A,!have been derived in (39), where

they are shown to be functions of X with parametric dependence on r/L.
T h e energy equation (61) for the black fin-tube radiator may now be
compared with Eq. (57) for the single plate-type fin. From such a com-
parison, it is seen that the only difference lies in the nature of the in-
homogeneous term which appears in the respective equations; that of
Eq. (61) being a function of X while that of Eq. (57) is a constant.
Further, a new parameter r/L has been added. Mathematically speaking,
there are no major structural differences between Eq. (57) and (61),
W I
E. M. SPARROW
and the numerical solution method outlined for the former continues
to apply to the latter.
Numerical solutions of Eq. (61) have been carried out in (39) for the
limiting case in which 9* is zero. T h e fin heat transfer results are reported
as a ratio
QtlQro
in which Qfo is the fin heat transfer rate when r/L = 0. This information
is reproduced in Fig. 20, wherein the foregoing heat transfer ratio is
10
.
Q9
0.8
-
Q'
Q'o
0.7
0.6
' ' 1 ' 1 ' ' 1 1 ) I ' ) I '

0 0.2 0.4 0.6 0.8 1.0
r/ L
FIG.20. Effect of tube surface on fin heat loss.
plotted as a function of r/L for parametric values of N , . T h e departure

of the curves from unity immediately indicates the reduction in fin
heat transfer due to the presence of the black tube surfaces. For realistic
values of N , (-1 or 2), it is seen that relatively large values of r/L are
required to cause significant reductions in fin heat transfer. Practical
design considerations suggest that r/L values in excess of 0.4are quite
unlikely. Therefore, for practical values of N , and r/L, the reductions
in fin heat loss due to radiant interaction with black tube surfaces are
limited to 10-15 percent.
T h e net heat transfer from each one of the tubes is affected by incident
radiant energy which comes from the fin and from the opposite tube.
When the system is black and the tube surface temperature is uniform,
this radiant interaction is readily formulated in terms of an algebraic
equation which involves a quadrature of the 9 solution of Eq. (61).
14381
BETWEEN SURFACES
Numerical results for the tube heat loss are presented in Fig. 4 of ref.
(39). For typical values of the parameters, the tube heat loss is reduced
by 10 to 20 percent due to the presence of the fin.
Consideration may now be given to the situation wherein the surfaces
of the fin and the tubes are nonblack. T o introduce a degree of generality,
it will be assumed that the radiation arriving from space may contain
substantial amounts of energy in the range of short wavelengths (for
example, solar energy) as well as in the infrared range. Correspondingly,
the surfaces may have different radiation properties in these wavelength
ranges. I n the infrared, the surfaces will be assumed to have gray-body
properties a = (E = 1 - p ; while for the short-wavelength range, the
properties will be described by & =.1 - p. I n general, 01 # Z. T h e
radiation emitted by the surfaces is assumed to be confined essentially
to the infrared range. As in the preceding, the tube surfaces are regarded
as isothermal at temperature T b . All quantities which relate to the
short-wavelength range will be designated by a tilde, N. Additionally,
the subscripts f and t will be used to respectively denote quantities
relating to the fin and the tubes.
I n carrying out the analysis, it is convenient to separate the net radiant
heat flux into two components q and 9, respectively corresponding to
the infrared and short-wavelength ranges. With these, the energy
conservation equation (54) becomes
T h e boundary conditions remain as before.

T h e dimensionless heat flux quantities vt and +, can be derived without
difficulty by specializing Eq. (36):
(63b)
T h e gray body assumption need not be made in deriving Eq. (63b),
and it is for this reason that 01 appears rather than E . T h e Sl and 6, are
angular coordinates on the tube surfaces as shown in Fig. 19.
It may be noted that Eq. (63a) and (63b) contain terms involving
cpf and+ f ,these being the heat fluxes at the tube surfaces. Consequently
W I
E. M. SPARROW
radiation balances for these quantities must also be derived. By applying

Eq. (36), one finds
Additionally, pi(&,) = p,(S,) when 6 - 6,.

.-
Inspection of the governing equations reveals that the unknowns
appear under the integral signs as well as outside of the integrals. Also,
the temperature 9 appears under a derivative operator. Therefore, the
governing equations are integro-differential. They are also highly non-
linear inasmuch as a4 appears both under and outside of the integrals,
while 9. appears under the derivative operator.
While Eqs. (62), (63), and (64) represent a highly complex mathe-
matical system, there is at least one element of simplification; namely,
that Eqs. (63b) and (64b) can be solved fully independently of the other
equations of the system. Then, the q t and solutions are used as input to
Eqs. (62), (63a), and (64a), which must be solved simultaneously.
T o the knowledge of the author, numerical solutions for this problem
have not yet appeared in the literature.
As a final note, it is interesting to contrast the mathematical problem
for the radiatively interacting nonblack system with that for the non-
interacting plane fin and for the radiatively interacting system with
black surfaces. All of these problems are nonlinear in the fin temperature
distribution. However, the essential difference is that the former is
governed by integro-differential equations, while in the latter problems
only differential equations appear.
C . RADIANTINTERACTION
BETWEEN FIN SURFACES
There are technically attractive radiator configurations in which there

may be radiant interchange between neighboring fins of an ensemble.
Such a situation is illustrated in the left-hand sketch of Fig. 21, where
there is pictured a tube to which are attached longitudinal fins. As a
W I
HEATTRANSFER
FIG. 21. Radiatively interacting longitudinal fins.
limiting case of such a configuration, consideration has been given to

the condition in which the radius of the tube is much smaller than the
height of the fins (40).Provided that the thermal conditions are the same
along each fin, one may single out for analytical treatment a pair of fins
as pictured in the right-hand sketch of Fig. 21. Clearly, results for such
a physical model should also apply to corrugated fins attached to a plane
wall.
T h e investigation reported in (40)was carried out for the limiting case
in which the incoming radiation from space is negligible. T h e structure
of the analysis closely parallels that which has just been described for
the radiant interchange between nonblack fins and tubes. T h e energy
conservation principle as stated in Eq. (54)continues to apply
and the y function follows from Eq. (36) as
Owing to symmetry, y ( X J = p)(X2)and #(X,) = #(X,) when X , = X , .

T h e required angle factor is given in (40) and is a function of X , and
X , with parametric dependence on the opening angle y.
T o complete the statement of the problem, it remains to specify the
boundary conditions. One may continue to state that 9 = 1 at X = 0
as before. However, the position X = I is not a symmetry line as in
the previous problems. Consequently, the condition d$/dX = 0 at
X = 1 must now be justified on the basis of a negligible heat loss from
W11
E. M. SPARROW
the fin tip (this condition is commonly used for conducting-convecting

fins).
T h e numerical solutions of (40) provide a wide range of results,
including overall fin heat transfer, local fin heat transfer, optimum fin
configuration, and fin temperature profiles. Of these, the overall heat
transfer is perhaps the quantity of greatest interest and a sampling of this
information is presented in Fig. 22. T h e figure shows the heat loss Q,
from the two surfaces which form the vee-cavity as pictured in the right-
hand sketch of Fig. 21. As plotted on the ordinate of Fig. 22, Q, is
normalized by the quantity 2aTb4Lsin(y/2), which represents the heat
FIG.22. Radiating effectiveness of radiatively interacting fins.

P421
BETWEEN SURFACES
loss from the same surface configuration if the fins were black and had
infinite thermal conductivity. T h e abscissa is the N , parameter without E .
T h e figure holds for c = 0.75, and curves are given for opening angles y
ranging from 45" to 120". Inspection of the figure reveals that the fin
heat loss decreases quite rapidly with increasing N,; from a considera-
tion of the numerical values of the heat loss parameter on the ordinate,
it would appear that fins characterized by N, values in excess of 1 cv 1.5
would not be very advantageous. Comparable information is also pre-
sented in (40) for fins having emittances of 1.0 and 0.5. Additionally,
solutions for interacting fins of trapezoidal profile have been worked
out and results are reported in (41).
D. RADIANTINTERACTION
BETWEEN FIN AND TUBE,
FINS
AND BETWEEN NEIGHBORING
A somewhat more complex radiant interaction situation occurs when

there is radiant interchange between fin and tube as well as between
neighboring fins of an ensemble. Such a situation occurs in the longitu-
dinal fin-tube configuration (pictured in the left-hand sketch of Fig. 21)
when the tube radius is not small relative to the fin height. Another
technically interesting configuration in which such interactions occur is
the annular-finned space radiator shown in schematic view in Fig. 23.
This arrangement has beer, analyzed in some detail in (42).
FIG. 23. Annular fin-tube radiator.
T h e most interesting and perhaps the most challenging aspect of the

analysis of the annular fin-tube arrangement is the shielding effect of the
tube surface. Because of this shading, radiant energy leaving an element
on one of the fins is unable to fall upon some finite area on the surface
of the opposite fin. T h e size and shape of the zone thus excluded,
depends upon the position of the radiating element as well as on the
separation distance between fins and on the tube radius. From the point
of view of the analysis, the shielding action just described introduces
significant difficulties in the derivation of the angle factors. These
W I
E. M. SPARROW
difficulties have been overcome by employing the contour-integration

method as previously discussed.
Aside from the angle-factor derivation, the analytical formulation
follows along lines which are similar to those described in preceding
sections and will not be repeated here. The specific problem treated in
(42) involved black surfaces and neglected radiation arriving from space.
A sampling of the results is presented i n Fig. 24, which shows the heat
FIG.24. Ratio of heat loss for fin-tube system to that for unfinned tube, Ri/R, = 1.
loss Q from the fin-tube system as a function of an appropriately defined
N,. parameter for various values of spacing to radius ratio L / R , . T h e
figure applies for R,R, = i. On the ordinate, the heat loss, Q, is
normalized by the quantity 2rrR,LaTb4 which represents the heat loss
from the tube surface in the absence of the fin. Consequently, the
departure of the curves from unity indicates the increase in system heat
loss due to the presence of the fins. I t is seen from the figure that the
fins are highly effective at small values of N , .
E. RELATEDPROBLEMS
Mention may be made of a class of problems which is closely related
to the radiating-conducting fin. These have to do with thin-walled
bodies subjected to incident thermal radiation over a portion of their
P441
BETWEEN SURFACES
external surfaces, for instance, the solar heating of a space vehicle.

T h e distribution of temperature along the surface of the body is controll-
ed by the interaction of thermal radiation with t h e heat conduction within
the wall. References (43) and (44) are representative of the studies
which have been carried out in this connection.
X. Specularly Reflecting Surfaces
Although radiant interchange calculations are generally carried out

under the assumption of diffusely-reflecting surfaces, evidence has
indicated (for example, Fig. 1) that there may be a significant specular
contribution. T h e most straight-forward procedure for analyzing radiant
interchange with specular reflection would be to keep account of the
individual rays as they reflect and re-reflect from surface to surface.
Such an approach would be quite tedious. However, it has recently been
shown (45, 46, 18) that under certain conditions, the interchange com-
putation involving specular reflections is no more difficult than that
involving diffuse reflections.
I n particular, a systematic and convenient procedure has been devised
(45, 46) for calculating interchange in an enclosure containing diffusely
emitting, specularly reflecting plane surfaces in addition to other sur-
faces (plane or nonplane) which are diffuse emitters and reflectors.
An analysis has also been made (18) for the interchange in a circular
tube having diffusely emitting, specularly reflecting walls.
I n the foregoing, consideration has been given to surfaces which are
either pure specular reflectors or pure diffuse reflectors. T h e behavior of
actual surfaces falls somewhere between these limits. It was suggested
by Seban, in a formal discussion of (46),that a first approximation to the
reflection characteristics of real surfaces could be achieved by sub-
dividing the hemispherical reflectance, p, into specular and diffuse com-
ponents,
+
,O = P s Pd (66)
By employing this concept, the analytical procedures previously devel-
oped for purely specular reflection can be generalized to apply to sur-
faces having both specular and diffuse components of reflection. Such
a generalization will be employed in the forthcoming presentation.
A. ENCLOSURES
CONTAINING
PLANESPECULAR-DIFFUSE
SURFACES
For the case of plane surfaces, the computation of the specular com-
ponent of the reflected radiation makes use of the basic property of plane
~ 5 1
E. M. SPARROW
mirrors; namely, that the radiant energy (or light) reflected from a plane
mirror appears to come from an image located behind the mirror. T h e
distance between the image and the mirror is identical to the distance at
which the object is placed in front of the mirror. This principle is well-
known in optics and may also be observed in everyday experience.
T o introduce the computation method for specular-diffuse enclosures,
it is convenient first to deal with a specific situation. Later, a completely
general formulation will be described.
For concreteness, consideration will be extended to the rectangular
enclosure pictured in Fig. 25. T o avoid difficulties in visualizing three-
3: S P E C U L A R - D I F F U S E
1,2,4: DIFFUSE
FIG.25. Rectangular enclosure containing a specular-diffuse surface.
dimensional geometries, the surfaces of the enclosure are taken to be of

infinite extent in the direction normal to the plane of the figure. T h e
usual postulates of the gray body enclosure theory will be invoked,
except that the surfaces may now have specular components of reflection.
I n particular, surfaces (l), (2), and (4) are diffuse reflectors, while sur-
face (3) is a specular-diffuse reflector with reflectance pa = pad f pas=
1 - e 3 . All surfaces are gray diffuse emitters.
As a first step, a radiant flux balance at one of the diffuse surfaces
may be carried out by applying Eq. (3); for instance, for surface (4),
B4 = 64uTp4 i-
p4H4 (67)
T h e incident energy H4 comes from the other surfaces (I), (2), and (3)
of the enclosure.,For example, from a fully-diffuse surface such as (I),
there arrives directly at surface (4) a quantity
per unit time and area. I n addition to this, energy leaving surface (1)
may reach surface (4) indirectly due to specular reflection on the partial
14461
HEATTRANSFER
mirror 3.5 T h e image of surface (1) as reflected in mirror (3) is shown

as l(3) in the figure. Thus, as far as surface (4)is concerned, diffuse
energy leaving surface (1) and reflected in the mirror appears to originate
as diffuse energy at the image surface l(3). This is illustrated by the
ray leaving point a, reflecting at point b, and finally impinging at point c.
An observer at c sees the ray as if it comes from the image point a on
the surface l(3). Thus, in analogy with expression (68a), the energy
passing indirectly from surface (1) to surface (4)via the mirror (3) is
P38[BlF4-1(3)1 (68b)
T h e multiplying factor p38 indicates the fraction of the radiant energy
which is reflected specularly, while F 4 - 1 ( 3 ) is the angle factor between
surface (4)and the image surface l(3).
Expressions (68a) and (68b) represent the radiation passing between
surfaces (1) and (4)either directly or by specular reflection. T h e sum
of these can be written as
V 4 - 1 (694
wherein E4-1 is termed an exchange factor and is given by
E4--1 = F4--1 f P3SF4-l(3) (69b)
Utilizing this concept, the radiant energy passing from surface (2) to
surface (4)can be represented as
Diffusely-distributed radiation in the amount B,A3 leaves surface (3)

in all directions, and of this there arrives per unit time and area at
surface (4)
R3E,-, , E4-3 = F4-3 (71)
T h e incident flux H4 is found by summing expressions (69a), (70), and
(71). Upon substituting these into the radiant flux balance (67), one
gets
B, = e40T44 4- p4 5 BiE4_,
i=l
Identical expressions apply for the other diffuse surfaces (1) and (2)
provided that the subscript 4 is respectively changed to 1 and 2 thus
B, = ek0Tk4f P k 2 B,Ek-i , k = 1, 2,4 (72)
Energy which is diffusely reflected at surface (3) will be included in the radiosity B,.
~ 7 1
E. M. SPARROW
O n the other hand, for the specular-diffuse surface (3), the radiosity
equation is
wherein the diffuse component p3d of the reflectance multiplies the

incident radiant flux.
Upon considering Eq. (72) and (73), it is seen that the use of the
exchange factor in lieu of the angle factor reduces the system of radiosity
equations to the same form as that for a purely diffuse enclosure. In
addition, it may be reiterated that the diffuse component p d must be
used in the radiosity equation for the specular-diffuse surface. I n the
event that p3d = 0, then B, = c , u T ~ and ~ , the foregoing formulation
reduces identically to that of (46) wherein consideration was confined
to purely specularly or purely diffusely reflecting surfaces.
For the fully diffuse surfaces, the relationship between the radiosity,
temperature, and heat flux applies as stated in Eq. (1 1)
B k = UTk4 - [(I - E k ) / k ] q k , h = 1, 2, 4 (744

whereas for the specular-diffuse surface (3)
B 3 = (1 - P 3 s ) u T 3 4 - [(I - 3 - P3.9)/31!?3 (74b)
With these, the radiosity may be eliminated from Eqs. (72) and (73),
leaving four algebraic equations for the four unknown heat fluxes.
- [(I - P 3 s ) a T 3 4 - (1 - 3 - P a s ) 1"
3
4 - 3 (753)
and similarly for the other diffuse surfaces (2) and (4). For the specular-
diffusive surface (3), one finds
T h e foregoing can be solved for the heat fluxes of the problem with no
greater difficulty than would be encountered for the case of purely
diffuse reflections.
With this specific problem as background, one can now state the
general computational equations for an enclosure in which there are Nl
W81
HEATTRANSFER
specular-diffuse surfaces and N , purely diffuse surfaces. At any one of

the surfaces of the enclosure, the heat flux is given by
k = 1, 2, ..., N , , ...,N , (76)

T h e first bracket on the right-hand side represents the contribution of
the purely diffuse surfaces and the second bracket is the contribution of
the specular-diffuse surfaces. The EkPi and Ek-j are exchange factors
which include the direct diffuse interchange between the surfaces and
the indirect interchange due to specular reflections. I t can be readily
verified that when pis = 0, Eq. (76) reduces identically to the computa-
tion Eq. (12) for the case of purely diffuse reflection.
T h e key to the application of Eq. (76) is the determination of the
exchange factors E. When the specular-diffuse surfaces are plane, the
exchange factors are readily derived by the image method illustrated
in the foregoing example. Additional illustrations of the application of
the image method may be found in (46),wherein exchange factors have
been worked out for several plane-surfaced enclosures that are more
complex than that of Fig. 25.
T h e generalization of Eq. (76) to include continuous surface variations
of radiosity, temperature, and heat flux can be carried out without
difficulty
T h e first integral extends over that part of the enclosure which reflects
diffusely, while the second integral extends over that part which is
specular-diffuse. T h e exchange factors needed in evaluating the integrals
can be readily derived by the image method provided that the specular
surfaces are plane.
B. NONPLANAR SURFACES
SPECULAR-DIFFUSE
T h e foregoing computational Eqs. (76) and (77) apply, at least in prin-
ciple, to radiant interchange in enclosures wherein the specular-diffuse
[4491
E. M. SPARROW
surfaces are nonplanar. However, the task of determining the exchange

factors is very formidable indeed. For the case of the circular tube, Perl-
mutter and Siegel (14, following the work of Krishnan (25, 26, 27),
have shown how to derive the exchange factors for purely specular
reflection. However, these also apply to the case of specular-diffuse
reflections. Interested readers are referred to these papers for additional
details.
REFERENCES
I . R. D. Cess, The interaction of thermal radiation with conduction and convection

heat transfer. Adwan. Heat Transfer I, 1 (1963).
2. E. R. G. Eckert and R. M. Drake, Heat and Mass Transfer. McGraw-Hill,
New York, 1959.
3. R. C. Birkebak, Monochromatic directional distribution of reflected thermal radiation
from roughened surfaces. Ph.D. thesis, Univ. Minnesota, Minneapolis, Minnesota,
1962.
4. H. C. Hottel, Radiant heat transmission. I n Heat Transmission (W. H. McAdams,
ed.), 3rd ed., Chapt. 3. McGraw-Hill, New York, 1954.
5. B. Gebhart, Heat Transfer. McGraw-Hill, New York, 1961.
6. A. K. Oppenheim, Radiation analysis by the network method. Trans. Am. SOC.
Mech. Engrs. 18, 725-735 (1956).
7. E. M. Sparrow, On the calculation of radiant interchange between surfaces. In
Modern Development in Heat Transfer (W. E. Ibele, ed.), p. 181. Academic
Press, New York, 1963.
8. M. Jakob, Heat Transfer, Vol. 2. Wiley, New York, 1957.
9. F. Kreith, Radiation Heat Transfer. International Textbook, Scranton, Pennsylvania,
1962.
10. D. C. Hamilton and W. R. Morgan, Radiant interchange configuration factors.
NACA T N 2836 (1952).
1 1 . H. Leuenberger and R. A. Pearson, Compilation of radiation shape factors for
cylindrical assemblies. A S M E Ann. Meeting, New York, 1956. A S M E Paper no.
56-A-144 (1956).
12. E. M. Sparrow, A new and simpler formulation for radiative angle factors. J. Heat
Transfer C85, 81 -88 ( I 963).
13. E. M. Sparrow, L. U. Albers, and E. R. G. Eckert, Thermal radiation characteristics
of cylindrical enclosures. J . Heat Transfer C84, 73-81 (1962).
14. E. M. Sparrow and V. K. Jonsson, Absorption and emission characteristics of diffuse
spherical enclosures. J. Heat Transfer C84, 188-189 (1962).
15. E. M. Sparrow, Radiant absorption characteristics of concave cylindrical surfaces.
J . Heot Transfer C84, 283-293 ( I 962).
16. M. Perlmutter and H. Siegel, Heat transfer by combined forced convection and
thermal radiation in a heated tube. J . Heat Transfer C84, 301 -3 1 I ( I 962).
17. R. Siegel and M. Perlmutter, Convective and radiant heat transfer for flow of a
transparent gas in a tube with a gray wall. Intern. J . Heat Mass Transfer 5, 639-660
(1 962).
18. M. Perlmutter and R. Siegel, Effect of specularly-reflecting gray surface on thermal
WOI
BETWEEN SURFACES
radiation through a tube and from its heated wall. J. Heat Tranffer C85, 55-62
(1963).
19. F. B. Hildebrand, Methods of Applied Mathematics, pp. 421-427, 501-502.
Prentice-Hall, Englewood Cliffs, New Jersey, 1952.
20. C. M. Usiskin and R. Siegel, Thermal radiation from a cylindrical enclosure with
specified wall heat flux. J. Heat Transfer C82, 369-374 (1960).
21. E. M. Sparrow and V. K. Jonsson, Thermal radiation absorption in rectangular-
groove cavities. J. Appf. Mech. E30, 237-244 (1963).
22. H. Buckley, On the radiation from the inside of a circular cylinder. Phil. Mag.
[7] 4 (23), 753-762 (1927).
23. E. M. Sparrow, Application of variational methods to radiation heat transfer calcula-
tions. J. Heat Transfer C82, 375-380 (1960).
24. J. C. De Vos, Evaluation of the quality of a black body. Physica 20, 669-689 (1954).
25. K. S. Krishnan and R. Sundaram, The distribution of temperature along electrically
heated tubes and coils, I. Theoretical. Proc. Roy. SOC.A257, 302-315 (1960).
26. K. S. Krishnan, Effect of specular reflexions on the radiation flux from a heated
tube. Nature 187, 135 (1960).
27. K. S. Krishnan, Nature 188, 652 (1960).
28. E. M. Sparrow and J. L. Gregg, Radiant emission from a parallel-walled groove.
J. Heat Transfer C84, 270-271 (1962).
29. E. M. Sparrow and V. K. Jonsson, Radiant emission characteristics of diffuse conical
cavities. J . Opt. SOC.Am. 53, 816-821 (1963).
30. E. M. Sparrow and S. H. Lin, Absorption of thermal radiation in V-groove cavities.
Intern. J . Heat Mass Transfer 5, 11 11- 111 5 ( I 962).
3 I . M. A. Heaslet and H. Lomax, Radiative heat-transfer calculations for infinite shells
with circular arc sections, including effects of an external source field. Intern. J.
Heat Mass Transfer 5, 445-457 (1962).
32. E. M. Sparrow, V. K. Jonsson, and W. J. Minkowycz, Heat transfer from fin-tube
radiators including longitudinal heat conduction and radiant interchange between
longitudinally-nonisothermal finite surfaces. NASA T N D-2077, December, 1963.
33. S. Lieblein, Analysis of temperature distribution and radiant heat transfer along a
rectangular fin. NASA T N D-196 (1959).
34. D. B. Mackay and E. L. Leventhal, Radiant heat transfer from a flat plate uniformly
heated on one edge. North American Aviation, Inc., Rept. MD 58-187, August, 1958.
35. J. G. Bartas and W. H. Sellers, Radiation fin effectiveness. J . Heat Transfer C82,
73-75 (1960).
36. C. Y. Liu, On minimum-weight rectangular radiating fins. J. Aerospace Sci. 27,
871-872 (1960).
37. J. E. Wilkins, Jr., Minimizing the mass of thin radiating fins. J . Aerospace Sci.
27, 145-146 (1960).
38. J. W. Tatom, Steady state behavior of extended surfaces in space. A R S (Am. Rocket
Soc.) J. 30, 118-121 (1960).
39. E. M. Sparrow and E. R. G. Eckert, Radiant interaction between fins and base
surfaces. J. Heat Transfer C84, 12-18 (1962).
40. E. M. Sparrow, E. R. G. Eckert, and T. F. Irvine, Jr., The effectiveness of radiating
fins with mutual irradiation. J . Aerospace Sci. 28, 763-772 (1961).
41. B. V. Karlekar and B. T. Chao, Mass minimization of radiating trapezoidal fins
with negligible base cylinder interaction. Intern. 3. Heat Mass Transfw 6, 33-48
( 1963).
42. E. M. Sparrow, G. B. Miller, and V. K. Jonsson, Radiating effectiveness of annular-
W11
E. M. SPARROW
finned space radiators, including mutual irradiation between radiator elements.
J. Aerospace Sci. 29, 1291-1299 (1962).
43. L. D. Nichols, Surface-temperature distribution on thin-walled bodies subjected
to solar radiation in interplanetary space. NASA T N D-584 (1961).
44. W. E. Olmstead and S. Raynor, Solar heating of a rotating spherical space vehicle.
Intern. J. Heat Mass Transfer 5, 1 1 65-1 177 ( I 962).
45. E. R. G. Eckert and E. M. Sparrow, Radiative heat exchange between surfaces with
specular reflection. Intern. J. Heat Mass Transfer 3, 42-54 (1961).
46. E. M. Sparrow, E. R. G. Eckert, and V. K. Jonsson, An enclosure theory for radiative
exchange between specularly and diffusely reflecting surfaces. J . Heat Transfer
C84, 294-300 (1962).
Author Index
Numbers in parentheses are reference numbers and indicate that an authors

work is referred to although his name is not cited in the text. Numbers in italic
show the page on which the complete reference is listed.
A Birkebak, R. C., 403,450

Black, S. A., 375(78), 396
Acrivos, A., 166, 170, 172,269, 361(1), Blankenstein, E.,353
390,391 (loo),394,396 Blodgett, K.B., 353
Adamson, T. C., Jr., 213,269 Bloom, M. H., 266,270
Albers, L. U.,420(13), 427(13),450 Blottner, F. G.,194(53), 230,231,
Amdur, I., 326,352 233 (53),234 (53),235,269,270
Anderson, A. D.,134(19),157, Blottner, G. F., 266,270
162(31), 163(31), 170(31),179, Bogue, D.C., 384,385,390,391,
182(31),227,230,234,235(19), 394, 395
239 (19),247( 19),268,269 Bonhoeffer, K. F., 330,355
Anderson, D. B., 385(59), 395 Bonilla, C. F.,38(36),106
Andrussow, L.,37,106 Booth, J. R.,374(21),395
Archer, C. T.,352,354 Bremner, J. G. M., 353
Astarita, G.,365(3), 394 Brinkman, H.C., 378,394
Austin, L.G.,138(22),268 Brodkey, R. S.,361,385,388,394,
395
B Brokaw, R. S., 38,106
Back L. H., 53, 54, 55,56,57, 60(59), Brown, D.R.,358(63),361,395
61(59), 62(59). 64(59),65(59), Brown, R.E.,326,355
66(59),68, 69(59), 70, 72, 73, 92; Brun, E.A.,322(24),334(24),
94,108 337(24),353
Baron, J. R.,53,107 Brush, C.F.,293,353
Bartas, J. G.,434(35),451 Buckley, H., 424,451
Bartlett, E.P.,2(2), 105 Budenberg, J. W., 38,106
Bartz, D.R.,6(9), 7(9, 13, 14), 8(9), Burow, S.,363 (13),365,394
9(9), 14, 17(9), 22(9), 23, zQ(9), C
27(9, 14),29(9),30(9),32(9),
33(13),34(9,13),35(13), Calder, P. H., 52,107
37(26,27),38(13, 26,27), Camm, J. C., 127(16), 131(16),
40(9,13),62(14), 63(9), 95(79), 134(16),268
105,106 Carman, E.H., 297(65),354
Baule, B., 339,352 Cess, R.D.,400,450
Beckett, C. W., 37(28), 38(28),106 Chambre, P.L.,166,170,172,269
Beek, O.,353 Chan, H.W., 25,106
Behn, V. C., 365(la),394 Chao, B. T.,443(41),451
Benedict, W.S.,37(28), 38(28),106 Chapman, D.,269
Benser, W.A., 95,108 Chariton, J., 353
Bernhardt, E. C., 358,394 Charm, S. E.,374,394
Berry, A. J., 325,355 Cheng, H.K.,266,270
Bertram, A., 353 Cheng, S. I., 213,269
Bird, R. B., 38(35),47(35),106, Chiu, H.H., 213,269
114(8),268,358,368,371,374, Christiansen, E. B., 367, 368, 370(15,
378,379,388,394,395, 397 16),373,374,383,394,395
C4531
AUTHORINDEX
Chu, S. T., 168,269 Doughty, D. L., 45 (48), 107
Chung, P. M., 134(19),146,152,153, Drake, D. B., 374,397
157,162(31), 163(31), 170(31), Drake, R. M., Jr., 23(23), 106,365,
172,175,176,179,182 (31), 187, 395, 403(2), 405(2), 407(2), 411,
189,190(51), 192,227,230,234, 412(2), 417(2), 450
235 (19),239 (19), 247,249 (75), Durgin, F. H., 53,107
255,263,267,268,269,270 Dushman, S., 353
Churchill, S. W., 16(21), 43,106 Duval, X., 353
Clapier, R., 324(28), 353
Clapp, R. M., 384,387,394 E
Clarke, J. F., 266,270
Clawing, P., 353 Eckert, E. R. G., 17(22), 23 (23), 39,40,
Clough, S. B., 365 (18). 394 41,45,106,107,207,211,269,403,
Cockroft, J. O., 353 405, 407(2), 411, 412(2), 417(2),
Colburn, A. P., 42,107 420 (13), 427 (13), 437 (39),
Coles, D. E., 7,44,97,98,105 439 (39), 441(40), 443 (40),
Collins, M., 390,391,394 445(45, 46), 448(46), 449(46),
Craig, S. E., Jr., 367, 368, 370 (15), 373, 450, 451, 452
394 Economos, C., 265 (94),270
Crave, B., 324(28), 353 Edmond, A., 322(24), 334(24),
Crawford, H. R., 385 (87), 396 337 (24), 353
Crocco, L., 41,106 Eggleton, A. E. J., 353
Crozier, R. D., 374,395 Ehlers, R. C., 53,56,69,71,94,108
Curtiss, C. F., 38(35),47(35), 106, Eissenberg, D. M., 385,395
114(8), 115(10), 268 Elliott, D. G., 7 (14), 9,27,62 (14), 105
Epstein, P., 353
Erdos, J., 231(65), 233(65), 269
D Eucken, A., 298(119), 338(119), 353,
355
Dalla Valle, J. M., 364,396 Eyring, H., 361(40), 395
Da Riva, I., 247,270
de Boer, J. H., 353 F
Deissler, R. G., 42,107, 386,395
Denison, M. R., 46,107 Fagan, W. S., 316(108), 355
Denney, D. A., 361,395 Fage, A., 172,269
Denton, E. B., 386(57), 395 Falkner, V. M., 172,269
Desmon, L. G., 42(42), 107 Fano, L., 37(28), 38(28), 106
Devienne, F. M., 284(26), 317(20), Fappo, M., 353
319 (22), 322,324(28). 334,337, Farber, M., 326,327,332,355
353 Fay, J. A., 49,107,139,196,198,
Devonshire, A. F., 346,353,354 199(24), 200(24), 201(24), 268
DeVos, J. C., 426,451 Feiler, C. E., 4(5), 86(5), 105
Diaconis, N. S., 156,269 Fendell, F. E., 255,263(89), 263,270
Dickins, B. G., 325,327(97), 353, 355 Foraboschi, F. P., 380,395
di Federico, I., 380,395 Fortini, A., 53,56,69,71,94,108
Dock, E. H., 354 Foster, W. R., 375 (96),396
Dodge, D. W., 358 (25), 382 (26), Fox, J. E., Jr., 385(59), 395
383(26), 384,385(26), 387,391, Fraser, R. G. J., 353
395 Fredrickson, A. G., 358,360, 361, 363,
Doetsch, G., 244 (68), 270 388, 395
Dooley, D. A., 46,107,213,269 Freeman, N. C., 270
Dorrance, W. H., 111(3), 268 Friend, P. S., 386,387,396
C4541
AUTHORINDEX
Friend, W. L., 386,395 Hansen, C. F., 123(14), 133(141,268

Frish, S., 353 Harper, E. Y ., 75,76 (68),88(68).
Frost, A. A., 123(11), 268 89(68),90(68), 91(68), 92,93(68),
108
G Harrje, D. T.,4(3,4),86(3), 105
Hartnett, J. P., 207,211,269
Gaede, W., 353 Hartunian, R. A., 164,190,269
Garsuch, P. D., 156 (29), 269 Heaslet, M. A. 430(31), 451
Gaskill, H. S., 386 (57), 395 Hedstrom, B. 0. A., 381,395
Gebhart, B., 407,450 Hildebrand, F. B., 422 (19), 451
Gee, R. E., 377(33),395 Hilsenrath, J., 37(28), 38(28), 106
Gehrke, E., 293,353 Hinze, J. O., 72,108
Gerrard, J. E., 376,378,395 Hirai, E., 373(47), 374,395
Gibson, W., 266,270 Hirschfelder, J. O., 38,47,106, 114(8),
Gier, H. L., 53 (59), 54 (59), 55 (59), 115(lo), 253,268,270
56(59), 57(59), 60(59), 61(59), Hoge, H. J., 37(28), 38(28), 106
62(59), 64(59), 65(59), 66(59), Hopper, J. R., 365(48), 395
68(59), 69(59), 70(59), 72(59), Hottel, H. C., 407,411,414,450
73(59), 94(59), 108 Houghton, G. L., 371 (74), 372(74), 396
Gill, W. N., 374,378,380,395 Houghton, W. T., 385!76), 391(75,76),
Gilli, M., 322, 327, 353 396
Ginn, R. E., 361,395 Howarth, A., 343,354
Glasstone, S., 355 Hozumi, M., 380(77), 396
Gluck, D. F., 372(71), 396 Humble, L. V., 42,107
Golike, R. C., 295, 327, 337, 355 Hurd, R. E., 385(76), 391(76),396
Gomf, G. E., 6(11), 105
Goodwin, G., 247,249(75), 270 I
Goulard, R. J., 136(20), 146,268
Inger, G. R., 166, 169,170(36),175,
Graham, R. W., 95,108 186,219,224,225 (62), 226 (62),
Granville, P. S., 392,395 246,247,251 (70,74), 269,270
Greer, S. E., 38(36), 106 Inman, R. M., 379,395
Gregary, H. S., 354 Inouye, M., 172,173,269
Gregg, J. L., 427 (28), 451
Irvine, T. F., Jr., 441(40), 443(40), 451
Gregory, H. S., 353, 354 Irving, J. P., 75(69), 95(69), 108
Gretsov, V. K., 68,108
Griffin, R. N., 164,269 J
Grigull, U., 374, 395
Grilly, E. R., 327,330,354 Jackson, J. M., 343,354
Grummitt, W. E., 355 Jaeckel, R., 354
Grunfest, I. J., 378,395 Jain, M. K., 380,395
Guenther, M.E., 75(69),95(69), 108 Jaivin, G. I., 75,79,86,108
Gupta, R. C., 390,395 Jakob, M., 411,412(8), 413 ( 8 ) ,450
Janssen, E., 41, 106
H Jarre, G., 243,269
Jefferson, T. B., 364,395
Hahn, S. J., 361,395 Jensen, G. E., 367,370,373,394
Hamilton, D. C., 411,412(10), 417(10), Jenson, V. G., 372,396
450 Johnson, H. A., 41,106
Hammerling, P., 127(15), 131(15), Johnson, M. M., 383,396
134 (15), 268 Johnson, N. L., 378(39), 395
Hanks, R. W., 374,383,395 Johnston, H. L., 327(57), 330(57),
Hanratty, T. J., 388,396 354,355
[4551
AUTHORINDEX
Jones, M. C., 352 Libby, P. A., 242(66), 265,269,270

Jonsson, V. K., 421 (14), 424 (21), Lieblein, S., 434 (33), 451
427(14,29),430(21), 434(32), Lightfoot, E. N., 358(7), 368(7), 394
436(32),443(42),444(42), Lighthill, M. J., 133,172,268,269
445 (46), 448 (46), 449 (46), 450 Lin, C. S., 386,395
451,452 Lin, S. C., 131(17),134(17),268
Jukoff, D., 287,355 Lin, S. H., 379 (103), 396,427 (30),
430 (30), 451
K Linan, A., 247,263,270
Liu, C. Y., 434(36), 451
Kannuluik, W. G.. 297 (65), 354 Liu,S. W., 172(45,46),175,176(45),
Kaplun, S., 257(91), 270 179(45). 247,269,270
Kapur, J. N., 390,395 Lobb, R. K., 16(18), 43,105
Karlekar, B. V., 443(41), 451 Loeb, L. B., 354
Kays, W. M., 16(19), 25 (19), 43,105, Lomax, H., 430(31), 451
107 Lowdermilk, W. H., 42(42), 107
Kearsley, E. A., 378,395 Ludwieg, H., 16,105
Keesom, W. H., 297,354 Lummus, J. L., 385 (59), 395
Kennard, E. H., 296,325,354 Lyche, B. C., 374,395
Kenty, C., 354 Lyon, J. B., 377(33), 395
Kestin, J., 94(76,77), 108
Kistiakowski, G. B., 354 M
Kivel, B., 127(15,16), 131(15,16),
134(15, 16), 268 Maass, O., 356
Klebanoff, P. S., 73,108 McAdams, W. H., 5,105
Kline, S. J., 16(19), 25 (19), 105 Mackay, D. B., 434 (34), 451
Knudsen, M., 297,354 Mandell, W., 294,354
Kolozsi, J. J., 53,68,69(57), 70,107 Mann, W. B., 327,355
Kovitz, A. A., 213,269 Marble, F. E., 213,269
Kreith, F., 372,395,411,450 Markovitz, H., 358, 361,363,395
Krishnan, K. S., 426,450,451 Marrone, P. V., 190,269
Krome, H., 353 Marshall, J. H., 354
Kundt, A., 354 Masi, J. F., 37 (28,30), 38 (28), 106
Kutateladze, S. S., 25,106 Massier, P. F., 53 (59), 54 (59), 55 (59),
56 (59), 57 (59), 60 (59), 61 (59),
L 62(59), 64(59), 65(59), 66(59),
68 (59), 69 (59), 70 (59), 72 (59).
Lacher, J. R., 354 73(59), 94(59), 107,108
Landau, H. G., 344,354 Matsumoto, R., 380(77), 396
Langmuir, I., 353, 354 Maxwell, J. C., 364,395
Laurmann, J. A., 354 Meader, P. F., 94(76,77), 108
Lazareff, P., 293,354 Merrill, E. W., 385(101), 396
Lees, L., 46,47,48,49,107,111(1), 139, Metzner, A. B., 357(65), 358,361,363,
148(1,23), 149(1), 150(1,23), 364(65), 365(18), 371(74), 372(71,
154(1), 157(1), 158, 159(23), 74), 373(66), 375(124), 382(26,65,
161(23),201,202(1,23),268 73). 383 (26), 384,385,386,387,
Lennard- Jones, J. E., 354 391 (75,76), 394,395, 396,397
Leontev, A. I., 25,106 Michels, W. C., 355
Leuenberger, H.. 411,450 Michiyoshi, I., 380,396
Leventhal, E. L., 434(34), 451 Miller, G. B., 443(42), 444(42), 451
Leveque, M. A., 366,395 Millikan, R. A., 355
[4W
AUTHORINDEX
Minkowycz, W. J.,434(32), 436(32), Peterlin, A., 363 (13), 365 (13), 394
451 Petersen, A. W., 387,396
Mirels, H., 172(45),176(45),179(45), Petersen, E. E., 390(2, loo), 391(100),
269 394,396
Mooney, M., 375(78), 396 Philippoff, W., 376, 378,395
Moore, F. K., 111(6), 266,268 Pigford, R. L., 371,396
Moore, J., 231 (65), 233 (65), 269 Pilch, M., 391 (go), 396
Morgan, W. R., 411,412(10),417(10), Pilsworth, M. N., 365,396
450 Polanyi, M., 355
Morrisson, J. L., 355 Powell, W. B., 75, 76,77,95,108
Mott, N. F., 343,354 Price, T. T., 76,77,108
Moulton, R. W., 386(58), 395 Pruitt, G. T., 385 (87), 396
Murakami, Y., 374,396 Putnam, G. L., 386 (57,58), 395
N R
Napolitano, L. G., 390,396 Rae, W. J., 219, 231,243,269
Nazmi, F., 354 Ragsdale, W. C., 52,107
Neu, R. F., 74,108 Raines, B., 312,327,355
Newell, W. C., 327(98), 355 Rating, W., 298(119),338(119),355
Nichols, L. D., 445(43), 452 Raynor, S., 445 (44),452
Niclause, M., 353 Read, H. E., 365 (18), 394
Nicoll, W. B., 43,107 Ree, T., 361 (40), 395
Noll, W., 360,396 Reed, J. C., 363,382(73), 396
Northdurft, W., 355 Reiner, M., 358,396
Novotny, J. L., 379 (103), 396 Reiss, L. P., 388,396
Nutall, R. L., 37 (28), 38 (28), 106 Reynolds, W. C., 16(19),25(19), 105
Riddell, F. R., 49,107,139, 196, 198,
0 199(24), 200 (24), 201 (24), 268
Oliver, D. R., 372,396 Ripkin, J. F., 391(90), 396
Oliver, R. N., 326,327,332,355 Ritter, R. A., 361,396
Olmer, F., 355 Rivlin, R. S., 360,363,396
Olmstead, W. E., 445(44), 452 Roberts, J. K., 310,355
Oppenheim, A. K., 407,450 Rose, R. K., 74, 108
Ornstein, L. S., 317,355 Rosner, D. E., 4,46,47,48,49,50,51,
Orr, C., Jr., 364,396 86,105, 111(4, 7), 164, 172, 183,
Oswatitsch, K., 58,108 185(49), 247,268,269,270
Rothstein, W., 58,108
P Rowley, H. H., 330,355
Rowley, R. W., 2(1), 105
Pallone, A., 231,233(65), 266,269,270 Rubesin, M. W., 41,106,172,173,269
Park, M. G., 385(72), 396 Rupe, J. H., 75, 79,86,108
Pasquino, A. D., 365,396 Rusinko, F., 138(22),268
Pearlman, H., 352 Ryan, N. W., 383,396
Pearson, R. A., 411,450
Pearson, R. G., 1231l l ) , 268 S
Peck, R. E., 316,355
Penner, S. S., 114(9), 123(9),268 Sailor, R. A., 391 (75), 396
Perlmutter, M., 422,423 (18).424 (16, Saunders, 0. A., 52,107
17,18), 425 (18), 432,445 (18). Saunders, R. C., 4(3), 86(3), 105
450 (18),450 Savins, J. G.,375(96), 385(95), 388,
Persh, J., 16(18),43( 18),105 396
[4571
AUTHORINDEX
Scala, S. M., 137,146,152,196,247, Stitt, F., 354
251(25), 266,268,270 Stoddard, F. J., 266,270
Schafer, K., 298,338,355 Stracham, C., 354,355
Schechter, R. S., 380,396,397 Summerfield, M., 372,395
Schenk, J., 374,378,396 Sundaram, R., 426(25), 450(25), 451
Schlichting, H., 255(90), 270 Sutera, S. P., 94(77),108
Schmidt, G., 297,354 Sutton. G. W., 202,205(54),206(54),
Schowalter, W. R., 390,391,394,396 269
Seban, R. A., 25,45(48), 94(75), 106, Svehla, R. A., 37, 39(31), 106
107,108
Sellers, J. P., Jr., 75,108 T
Sellers, W. H., 434 (35), 451
Semenov, N. I.,353 Taikeff, E. A., 38(36), 106
Sergienko, A. A,, 68,108 Talbot, L., 355
Shah, M. J., 390,391,394,396 Tanner, R. I., 361,396
Shapiro, A. H., 13(16), 17(16), 18(16), Tatom, J. W., 434(38), 451
105 Taylor, H. S., 355
Shaver, C. R., 385(l O l ) , 396 Taylor, R. L., 127(16), 131(16),
Sheridan, R. A., 156(29), 269 134(16), 268
Shertzer, C. R., 361,396 Taylor, W. J., 327(57), 330(57), 354
Sibbitt, W. L., 364(51), 395 Teare, J. D., 127(15,16),131(15,16,
Sibulkin, M., 6(8),9,23,24(8),105 17), 134(15,16,17),268
Siegel, R., 422,423(18),424 (16,17,18, Tetervin, N., 6( 12), 105
20), 425 (18,20), 432,445 (18), Thomas, D. G., 364,373,382(106), 384,
450, 450, 451 387,392(105), 396,397
Silver, S., 7 (14), 9 (14), 27 (14), 62 (14), Thomas, L. B., 295,326,327,337,355
105 Thompson, W. P., 164(33),269
Simpkins, P. G., 270 Tien, C., 376,390,397
Smith, J. M., 52,107 Tien, C. L., 388(119), 397
Soddy, F., 325,355 Tifford, A. N., 168,269
Souquet, J., 324 (28),353 Tillman, W., 16, 105
Spalding,D. B., 111(2), 123(12), 268 Tomita, Y., 390,391,397
Sparrow, E. M., 379,396,407(7), Tompkins, F. C., 353
415 (12), 417 (12),418 (12), Toong, T., 213,217(61), 218(61),269
420( 13),421 (14,15),424(21), Toor, H. L., 376, 378,397
425 (23),427 (13,14,28,29,30), Touloukian, Y. S.,37(28),38(28),106
430(15, 30, 21), 434(32), 436(32), Truesdell, C., 363 (116), 397
437(39), 439(39), 441(40), 443(40, Tsai, D. H., 37(30),106
42), 444(42), 445 (45,46),448 (46), Tucker, M., 6(10), 105
449 (46). 450,451,452 Turian, R. M., 379,397
Spivak, G. V., 355 Turner, D. T., 363 (13), 365 (13),394
Squire, H. B., 45(47), 107 Turner, L. A., 354
Srivastava, R. C., 390,395 Tyson, H., 354
Stalder, J. R., 287,355
States, M. W., 355 U
Steiger, M. H., 266,270
Usiskin, C. M., 424(20), 425(20), 451
Stephens, R. W. B., 354
Stern, O., 353
V
Stewart, J. E., 374(21), 395
Stewart, W. E., 358(7), 368(7), 371(7), Van Cleave, A. B., 356
394 Van Driest, E. R., 41,42,45,107
C4581
AUTHORINDEX
Van Dyke, K. S., 356 Wilke,C. R.,38,106,388(119),397

Van Dyke, M.,257(92),260(92),270 Wilkins, J. E.,Jr., 434(37),451
Van Laar, J., 374,378,396 Wilkinson, W., 357(127),358,397
Van Wyck, W. R., 317,355 Wilson, A. H., 123(13),268
Vaughn, R. D., 371(74), 372(74), 379, Winkler, E. M.,16(18), 43(18), 105,
396,397 164,269
von Smoluchowski, M., 292,293,355 Winning, M. D., 382(128), 397
von Ubisch, L., 356 Wissler, E.H., 380,396,397
Witte, A.B.,51,52,68,75,76(68),
W 78(52),80(52),81(52),83 (52),
84(52,73),87,88(68),89(68),
Walker, P. L., 138(22),268 90(68),91(68),92,93(68),107,
Wallick, G.C., 375(96),396 108
Wanford, D.W. B., 353 Witzell, 0.W., 364(51),395
Wang, H.E.,94(76),108 Wolf, H.,16(20), 25,43,106
Warburg, E.,354 Wolf,C. C.,385(76),391(76),396
Wartenstein, 356 Wood, R. W., 356
Wasan, D. T.,388,397 Woolley, H. W., 37(28),38(28), 106
Weber, S.,294,356 Wray, K.,127(15), 131(15). 134(15),
Weidmann, M.L.,328,356 268
Wells, C. S.,Jr., 385 (120),397
Welsh, W. E.,Jr., 51,52,68,75, Y
78(52), 80(52),83(52), 84(52),
87,107,108,153,156,269 Yau, J., 390,397
Werlein, P.P.,316(108),355 Yeager, E. B.,4(5), 86(5), 105
West, J., 294,354 Young, G.B.W., 41,106
Whdey, F. R.,356 Young, J. P.,378(39), 395
White J. L.,358,361,375(la),
378(121), 391(75,121),396,397 Z
Whiteman, I. R., 374,397
Whitlock, M.,363(69), 396 Zellnik, H. E., 16(21), 43,106
Wigner, E.P.,355 Zener, C., 341,343,356
Wilcox, W. R.,388,397 Zwanzig, R. W., 349,366
Subject Index
A Bingham plastic fluid, 359,373, 381f,

384
Absorptance, 401f equation, 362, 383
apparent, 425ff parameters, 373
Absorption characteristics of cavities, Blasius equation 21,142,203f
see Cavities Blunt body (hypersonic), see Locally
Acceleration, effects of on turbulence, similar solution for . . .
67f, 72ff Blunt body flows, 234ff, 246ff
Accommodation coe5cient (s),273, Boltzman constant, 279,288,345
277ff, 340,342,344,345f, 348 Boundary layer (s)
devices employed in measuring, 311ff at nozzle inlet, 59,63f
effects on, chemically reacting, 11Off
internal energy, 337ff shape parameters, 99ff
rotational energy, 338 turbulent in rocket nozzle, 2ff
vibrational energy, 338 with gas phase reactions, 1926
experimental determination of, 304ff Boundary layer equations, transforma-
factors modifying, 328f tion into incompressible form, 170ff
influence of pressure on, 3352 Boundary layer flows, non-Newtonian
measurements, 304ff, 307ff, 311ff, fluid mechanics in, 389f
317ff, 319ff, 322ff, 324ff, 326ff non-Newtonian heat transfer in, 389ff
of Boundary layer thickness, 101f, 195
air, 327 energy, 14, 15
argon, 327 momentum, 13f, 15
carbon monoxide, 327 temperature, 11
helium, 326f, 330 velocity, 11
hydrogen, 327
nitrogen, 327
oxygen, 327 C
triatomic gases, 328
surface temperature effects, 329ff Catalytic reaction, see Locally similar
theoretical calculations of, 339ff solution for . . .
Activated complex, 127ff Catalytic surface reactions, 135ff, 146ff
Air experiments, 52ff Catalytic wall, 46
nozzle configurations, 52f Cavities
Angle factors, 400,405f, 420f absorption characteristics of, 425ff
diffuse, 411ff circular arc, 421
Apparent absorptance, 425ff circular cylinder, 419f, 424, 425, 427f
Apparent emittance, 425ff conical, 427ff
Arrhenius type reaction, 214 emission characteristics of, 425ff
Atoms conservation equation, see parallel-walled, 424, 425, 428
Conservation of. . . reciprocity theorem for, 426f
rectangular-groove, 427f
spherical, 421,427ff
B vee-groove, 427ff
Chapmans formula, 297
Bades theory, 339ff Characteristic dissociation tempera-
Binary scaling law, 246 ture, 134
[4601
INDEX
SUBJECT
Chemical diagnostics, 110,164 Deviennes method, 304,319ff,324
Chemical kinetics, 122ff Devonshires theory, 346ff
Chemical reaction Diatomic gases, dissociation-recombi-
heterogeneous, 135ff nation of, see Dissociation-recom-
homogeneous, 123ff bination of. . .
Chemically frozen, 46 Diffuse emission, 403,405,4262
Chemically reacting boundary layers, Diffuse reflection, 402f, 426f
see Boundary layers . . . Diffusion coefficient, 48,114
Circular-arc cavities, see Cavities Diffusion flames, development of, 210f
Circular cylindrical cavities, see Diffusion velocity, 113ff
Cavities Dilatant fluid, 359,363
Circular tube passages, see Passages Displacement thickness, 13
Colburn form of Reynolds Analogy, see Dissociation-recombination,223ff
Reynolds Analogy. . . kinetics, 195f
Coles skin-friction coefficient, see Skm- of diatomic gases, 132f
friction coefficient Dissociation temperature, see Charac-
Combustible gases, reaction of injected, teristic dissociation temperature
see Reaction of . . . Dissociative relaxation, 227ff, 231ff,
Combustion of injected premixed fuel, 243ff
202ff Dorodnibyn integral method, 231f
Combustion chamber heat flux, 79fT Drag coefficient, 381ff, 384,387
Combustion efficiency, 51 Driving potential, 44ff
Combustion sheet without chemical reaction, 44f
off the wall, 211f with chemical reaction, 45ff
on the wall, 212f
Concentration-integral method, 138ff E
Conductance parameter, 434
Conduction, free-molecule, 273,275f Ellis model, 362
Cone, supersonic flow about, 167f, 171, Emission characteristics of cavities, see
l73,174ff, 231ff Cavities
Cone (supersonic), initial relaxation Emittance, 401
near leading edge of, 219ff apparent, 425ff
Conical cavities, see Cavities Emitting-absorbing gas, 400
Conservation of atoms equation, 197, Energy equation, 8f, 19f,28f, 112,
227,237 116, 140, 147, 171, 193f, 197, 214,
Conservation of reactants equation, 214 236, 377
Conservation of species equation, 112, Energy thickness, see Boundary
116,140,147,171,193 layer. . .
nonsimilar solutions of, 164ff Enthalpy, 112
self-similar oht ti on^ of, 144f Enthalpy flux,19,102
Continuity equation, 112,171 Enthalpy state, 46f
Contour integral method, 414ff, 444 Enzian injector, 77ff
Equation of state, 112
Equilibrium flow singularities, see
D Singularities . . .
Equilibrium flows, 252ff
Damkohler number, 143f, 147,154,156, Extended surfaces, see Fins
158, 185, 194f, 197, 208, 219, 220, Eyring-Powell equation, 362, 370, 378
228, 237, 247, 249f
Deficiency thicknew, llff F
Deformation rate tensor, 358
Deviatoric stress tensor, 368 Falkner-Skan flows, 142,167f
C4611
INDEX
SUBJECT
Ficks law, 46 I
Ficks relation, 114f
Fin effectiveness, 435f Ignition, see Thermal ignition
Fin (s) Incident radiant flux,405
radiating-conducting,400,432ff Incipient ignition, see Thermal
radiant interaction between, 44Off ignition
single plate-type, 433ff. 437 Initial relaxation near the leading edge
Fin-tube radiator, 433f, 436ff of flat plates, and supersonic
interactions, 443f wedges and cones, 219ff
annular, 443 Injectors, 77f
Flame sheet, 256ff, 262f Internal energy, 281
Flat plate flows Internal energy and the accommoda-
chemically reacting boundary layers, tions coefficients, 337ff
167f, 171, 173, 174ff, 183, 184f, Integral equations, 418ff
192, 213ff, 227ff, 231ff, 243ff solutions of, 420ff
initial relaxation near leading edge, Integral method, see Dorodnitsyn . . .
219ff Pohlhausen . . .
non-Newtonianheat transfer, 390 Integro-differentialequations, 440
Flow behavior indices, 363 Interaction methods, 422f
Fouriers law, 434 Intermediate regime
Fracture of a fluid, 363 between parallel plates, 300ff
Free-module conduction, see Conduc- between unspecified surfaces, 303f
tion Internal heat generation and non-
Friction factor, 382 Newtonian heat transfer, 376ff.
Frozen flow, 111 379ff
Isothermal surfaces, 401
G
J
Graphite, see Surface combustion of
Grahof number, 372 Jacksons theory, 343ff
Gray diffuse enclosure, 401,407ff
Gray surfaces, 401 K
Graetz number. 368ff
Kinetic energy of molecular
rotation, 277
H translation, 276f
vibration, 277
Heat conduction of a highly rarefied Kinetic energy, turbulent, 72ff
gas, 276ff Kirchoffs law, 402
Heat flux in chemical mixtures, 115f Knudsen number, 273,275
Heat flux measurement techniques, 55f Knudsens method, 304,317ff
Heat transfer coefficient in turbulent Kreith-Summerfieldcorrection, 387
boundary layers, 33ff, 45,59ff
Heat transfer measurements in
L
turbulent boundary Iayers, 59ff
Heterogeneous chemical reaction, see
Chemical reactions Lamberts cosine law, 402f, 405
Homogeneous chemical reaction, see Laminar flow
Chemical reactions external, 389ff
Hydraulic radius, 428 internal, 388
[4623
INDEX
SUBJECT
Landaus theory, 344 transition to turbulence, 381ff
Laplace method, 244f heat transfer, 358ff, 389ff
Latent heat of vaporization, 121 between parallel flat plates,
Law of mass action, 123f 374f, 380
Leaving radiant flux, 405 from cylinders, 390f
Lees approximations, 201f, 237 in ducts, 366ff
Leveque approach, 366,371ff from flat plate, 390
Lewis number, 47f, 117 forced convection, 390
Lighthill gas, 133 in round tubes, 3666,380
Leveque equation, 371,375 transition to turbulence, 381ff
Line integral method, see Contour natural convection, 390
integral method nucleate boiling rates, 392
Local chemical equilibrium, 46 with internal heat generation,
Locally similar solution, 145ff 376ff, 379ff
for catalytic reaction around a with viscous dissipation, 376ff
hypersonic blunt body, 157ff thermal properties, 364ff
heat capacity, 365
M thermal conductivity, 354f
turbulent flow
Mass action law, see Law of. .. external, 392
Mass diffusion coefficient, 46ff internal, 388f
Mass flux in nozzle, 58,102 Non-similar solutions of conservation
Mass fraction of species, 114,133 of species equation, see
Mass production rate, 153 Conservation of species. . .
Mass rate of formation, 113f Nozzle configurations, see Air
Maxwellian distribution, 278,285, experiments
317,323 Nozzle heat flux, 87ff
Maxwells law. 278 Nozzle inlet, see Boundary layers at. . .
Mean free path, 273 Nusselt number, 29,368ff
Molecular heat capacity, 298 Nusselt-Reynoldstype equation, 32ff
Molecular heat conduction coefficient,
297 0
Momentum equation, 8f, 18f, 112,140,
147,171,197,214,227,236,377 Ornstein and Van Wycks method, 304
Momentum flux, 18
Momentum thickness, see Boundary P
layer. . . Parallel-walled cavities, see Cavities
Morse potential, 346 Parallel-walled passages, see Passages
Partition function, 125f, 133
N Passages
circular tube, 432
Near-equilibrium flows, 251ff parallel-walled, 432
Newtonian fluid, 359f radiant transmission characteristics,
Non-equilibrium flows of high energy 430ff
gases with dissociation and recom- tapered gap, 430f
bination, 218ff tapered tube, 430f
Non-Newtonian fluids Perturbation solution methods, 220ff
fluid mechanics, 389f Perturbation techniques, 253ff
in entrance region, 390 Planck constant, 345
between parallel flat plates, 390 Pohlhausen integral method, 179ff,
in round tubes, 390f 2272 234ff
[4631
SUBJECT
INDEX
Potential-flownozzle, 11 Recombination, see also Dissociation-
Power law, 362 recombination
Prandtl number 15, 23, 117, 372, 385, arbitrary, 46
3861 Recombination rate coefficient, 197
frozen, 48 Recovery factor, adiabatic, 15
Premixed fuel, combustion of injected, Rectangular-groovecavities,see
see Combustion of.. . Cavities
Pressure, influence of on accommo- Reflectance, 405
dation coefficient, see Accommo- Reiner-Rivlinfluid, 363
dation coefficient Relaxation, see also Initial relaxation
Pressure oscillations, dissociated, see Dissociated
effects of, 82ff, 94 relaxation
Probability integral, 286 Relaxation time, 298
Probe, boundary layer, 54f Reynolds analogy, 23f, 47,96,98f
Property variations,effects of, 39ff, 96 Colburn form, 23
Pseudoplastic fluid, 359,363,372 von K h n h form, 16,23,98f
Reynolds number, 4,381f
Rheological properties of fluids, 358ff
R Rheopectic fluid, 359f
Robertsexperiments, 312f. 325f, 328,
Radiating-conductingfins, see Fins 343,348f
Radiation heat transfer between Rocket thrust-chamber measuremente,
surfaces, 4OOff 74ff
Radiation heat transfer cavities, Rotational energy and the accommo-
see Cavities dation co&cient, 338
Radiation heat transfer p m g e s , Runge-Kutta method, 435
see Parceages
Radiosity, 405 S
Radiometricapparatus, 317f
Rarefied gases, heat transfer Schleiermachers method, 304ff
between coaxial cylinders,282ff, 292f Schmidt number, 117
between parallel plates, 280ff, 292 Self-similarflows, 142
between unspecified surfaces, 284 Self-similar solutions of species conser-
from a flat plate, 284 vation equation,see Conservation
by heat conduction, 291ff of species equation
from a moving plate, 279 Semitransparent solid,400
Rate of production of species, 120,135f Shear-thickeningfluid, 359
Rayleigh-Ritz method, 425 Shear-thinning fluid, 359,363,372
Reactanta conservationequation,see Shock tube, 187f
Conservation of. . . Shock wave, see Surface reaction
Reaction, see Catalytic . . . behind.. .
at surface,see Surface reaction Showerhead injection, 78f
of injected combmtiblegases, 207ff SiedemTate correction, 3721 375,387
Reaction coordinate, 127f Similarity variable, 146
Reaction rate, 126f, 134 Simpsons rule, 422
Reactions,see also Chemical reactions Single plate-type fin,see Fin
in boundary layer, see Boundary Singular perturbation method, 255ff
layer Singularitiesassociated with equilib-
Reciprocity rium flows, 252ff
relation, 408 Skin-friction coefficient, 14,15f,20ff,
theorem for cavities,426f 39,98f
E4641
INDEX
SUBJECT
adiabatic, 21 Thermal conduction, mechanism of,
Coles, 21,98f 274
diabatic, 20ff Thermal conductivity, 274
low speed, 21 Thermal ignition
Slender body flows, 246 incipient, 216ff
Species conservation equation, see problems, 213ff
Conservation of. . . Thermocouple plug fabrication, 55,77
Species production rate, see Rate of. . Thin-fin assumption, 434
Specific heat, 277 Thixotropic fluid, 3591
Spectral emission measurements, 317 Total enthalpy, 116,207
Specular-diffuse surfaces, 445ff Transformation of boundary layer
nonplanar, 449f equations into incompressible
Specularly-reflecting surfaces, 400, form, see Boundary layer. . .
426f, 445ff Transport properties, 36ff
Spherical cavities, see Cavities Trapezoidal rule, 422
Stagnation pressure, 53 Turbulence, effects of acceleration on,
Stagnation region flows, 195ff 67f, 72ff
Stagnation region heat flux, 200f Turbulent decay, 9Off
Stagnation temperature, 53 Turbulent flow characteristics, 384f
Stanton number, 14,15f, 20,22ff, 27, Turbulent heat transfer rates, 385f
29,39,96,99,386
Static pressure, 15
dbtributions, 57ff V
Stefans constant, 321
Stoichiometric equation, 124 Vee-groovecavities, see Cavities
Stokes theorem, 415,418 Velocity, characteristic, 35
Stream function, 139,146,170,187f Vibrational energy and the accommo-
Stress tensor, 358 dation coefficient, 338
Sublayer, 386,387 Viscoelastic fluid, 359ff
Superposition principle, 417 Viscoelasticitynumber, 381
Surface combustion of graphite, 152ff Viscosity, 15,38f
Surface heat flux, 120f, 160,205,239 Volterra integral equation representa-
Surface reaction, 135f, 138ff tion of surface reaction, 172ff
and heat transfer from high temper- von Kbrmbn form of the Reynolds
ature dissociated air, 161ff analogy, see Reynolds analogy . . .
behind strong moving shock, 185ff von KbrmAn-Polhausenequations, 390
Surface reaction solutions, 172ff
Surface reactions, catalytic, see Cata-
lytic surface reactions W
Surface recombination, 246ff
Wall, catalytic, see Catalytic wall
T Wedge, supersonic flow about, 167f,
l71,173,174ff, 183,184f
Tapered gap passages, see Passages Wedge (supersonic), initial relation
Tapered tube passages, see Passages near leading edge of, 219ff
Taylors series method, 423,424 Weissenberg-Modneyrelation, 388
Temperature, 277
Temperature jump
distance, 275,292,296,299 Z
measurements, 293f
regime, 291f Zeners theory, 341ff
theory, 296ff Zwanzigs theory, 349f
[4651

Advances in Heat Transfer 2 (1965)

Hochgeladen von

Dokumentinformationen

Originaltitel

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

Advances in Heat Transfer 2 (1965)

Hochgeladen von

Copyright:

Verfügbare Formate

Contributors to Volume 2

James P. Hartnett Thomas F. Irvine, Jr.

ACADEMICPRESS New York London

ACADEMIC PRESS INC.

United Kingdom Edition published by

PRINTED I N THE UNITED STATES OF AMERICA

D. R. BARTZ, Jet Propulsion Laboratory, California Institute of Tech-

PAUL M. CHUNG, Aerospace Corporation, San Bernardino, California

F. M. DEVIENNE, Laboratoire Miditerranken de Recherches Thermo-

A. B. METZNER, University of Delaware, Newark, Delaware

E. M. SPARROW, Heat Transfer Laboratory, Department of Mechanical

Since the appearance last year of Volume 1 of Advances in Heat

April, 1965 JAMES P. HARTNETT

added weight of the wall. Hence, it becomes important to ascertain the

transport except in an heuristic fashion. Another significant complexity

are made. However, critical experiments, such as reaction rate measure-

Before proceeding with a current version of the analysis of the problem

which served as the basis for the analytical prediction, several

development in a nozzle (9, Fig. 3) showed that the local boundary-layer

dimension in computing the skin-friction coefficient since this thickness

T h e only method of approach amenable to analysis known to this

energy equations. As mentioned in the Introduction, Sibulkin (8) and

Nomenclature for real flow.

FIG. 2. Nomenclature for adiabatic potential flow.

Equations (4)-(6) are approximate in that a product of mean values

6,' - 6', = 1;' (1 -

T h e integral above is customarily defined as the displacement thick-

while the integral definition is

Because of the approximate nature of Eq. (7) (resulting from the

$l, - $lr = 27rrpuz J

T h e integral above is customarily defined as the momentum thickness 8.

and the integral definition is

&- H , = 2i7rpUc,(T0 - T W ) 1 6,'

T h e integral above is customarily defined as the energy thickness 4.

and the integral definition is

b. Coejicients. T h e skin-friction coefficient C, is defined as the ratio

T h e stanton number C, is defined as the ratio of the wall heat flux

T h e adiabatic wall temperature Taw is the wall temperature for zero

where R is the adiabatic recovery factor.

(12) Heat transfer affects the skin-friction coefficient in either one of

T h e usual approach to the derivation of the integral momentum and

layer (26, p. 1090). Through this process, one arrives at integrals of

dxd (rpU28)= 17, - rpU6*

Under Assumptions 5 and 7, the expressions involving p and U can

Substituting these expressions into Eq. (22) and transforming the

Subtracting Eq. (27) from Eq. (26) yields

Equation (28) is the integral energy equation for thin axisymmetric

Substituting Eqs. (23) and (24) and transforming the independent

Note that the effect of variable surface temperature is accounted for in

T h e correlation of adiabatic turbulent boundary-layer skin-friction

and m is the exponent of the viscosity relationship adopted, p

T h e second relationship, Assumption 12b, is that of accounting for

Or, again making use of Eq. (32) and assuming m = 0.6,

and C,/2 in a straightforward manner because the similarity between

A Prandtl-number correction of this form has been utilized in all of the

It is important to note at this point that relationships such as Eqs. (38)

in the momentum equation and the term

in the energy equation. Thus, it was apparent in the analyses (8, 9)

Noting that C, increases sharply as the downstream distance approaches

This form suggests that a Reynolds-number-like parameter R, = p U $ / p

tion of C, . Such a suggestion has been made previously by Seban and

7 . Solution of Integral Equations

I n fact, the shape parameters are quite insensitive to the arbitrary