Beruflich Dokumente
Kultur Dokumente
D. R. BARTZ
PAUL M. CHUNG
F. M. DEVIENNE
A. B. METZNER
E. M. SPARROW
Advances in
HEAT
TRANSFER
Edited by
Volume 2
OF CONGRESS
LIBRARY CATALOG
CARDNUMBER:63-22329
[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
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
C. BACKGROUND
OF ANALYSES
OF THE PROBLEM
11. Analyses
A. INTEGRAL
MOMENTUM SOLUTION
AND ENERGYEQUATION
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
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
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
FROM RAPIDLYACCELERATING
jlux of the realflow for the same massflux. Hence, the physical definition
of the momentum thickness is
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:
0
pa
-(1 -
PU
tb - Tw
To - Tw
) dY 113)
(16)
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
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
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:
U dx M ( l ++MY) dx
pU dx M ( l + q M 2 ) dx
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
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
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.
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
-e -
[ M (1
1
+
+ @*P)
M2) "1
D. R. BARTZ
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
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.
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
FROM RAPIDLYACCELERATING
FLOWS
where
(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
FROM RAPIDLYACCELERATING
Pr2/9(pU ~ / p ) f ( + ~ / r ~ ) 'Pr2/3RDt(q5,/ro)+
(45b)
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
.-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
.-c
I-
oo
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
FROM RAPIDLYACCELERATING
HEATTRANSFER FLOWS
Nu = C(RD)o.ePro.a (47)
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
FIG. 10. Heat-transfer coefficient for nozzle and conditions of Fig. 5 ; closed-form
approximation comparison.
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
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
FROM RAPIDLYACCELERATING
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
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
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)
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
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
FROM RAPIDLYACCELERATING
and where 1: is the heat of formation of the ith species. In the special
case of a binary mixture, Ficks law states that
FROM RAPIDLYACCELERATING
HEATTRANSFER FLOWS
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
1 + (Pr, - 1 ) a i k 1net I c / ay
az,/ay
FROM RAPIDLYACCELERATING FLOWS
HEATTRANSFER
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,
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)
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
FROM RAPIDLYACCELERATING
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
FIG. 1 1 . Flow and instrumentation diagram; heated-air investigation of Back et al. (59).
All dimensions in inches.
0042 D
J M E L THERMOCOUPLE
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
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).
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
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
FIG. 16. Ratio of local to one-dimensional mass flux along the nozzle (of Back et al., 59)
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
FROM RAPIDLYACCELERATING
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.
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).
0.043
0.046
0.031
FIG. 18. Heat-transfer coefficient versus axial distance ratio with 18-in. cooled
approach length (59).
D. R. BARTZ
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
FROM RAPIDLYACCELERATING
-
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
AXIAL DISTANCE RATIO z/L
FIG. 21. Comparison of measured and predicted heat-transfer coefficients for various
boundary-layer thicknesses at nozzle inlet for heated air (59).
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).
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
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
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
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.
22v dUldx
x = --- TudP
(87)
' 6 I ' O
N
N
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
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
THERMOCOUPLE PLUG
/
PLATING 0005 THICK (MAXI
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.
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
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
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
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;\
~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
-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
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
FROM RAPIDLYACCELERATING
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
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
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
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
FROM RAPIDLYACCELERATING
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
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
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
and
,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
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
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
U = y < 6,
and
i = =P= = a
P
where
where
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
(C.2)
O P
and similarly with the enthalpy flux of the real flow near the wall,
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
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
FROM RAPIDLYACCELERATING
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
PO81
Chemically Reacting
Nonequilibrium Boundary Layers
.
PAUL M CHUNG
.
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
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.
Energy
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.
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):
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
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
NONEQUILIBRIUM LAYERS
BOUNDARY
and
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
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
REACTINGNONEQUILIBRIUM
CHEMICALLY BOUNDARYLAYERS
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
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
-(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.
A. HOMOGENEOUS
CHEMICAL
REACTION
Chemical reactions are of two types: homogeneous reactions and
heterogeneous reactions. T h e former occurs in a homogeneous phase
<<
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,
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
i54)
or
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
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
C airj"
n n
A E O = bjrjo - (60)s
3=1 3=1
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
NONEQUILIBRIUM LAYERS
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
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
REACTION COORDINATE -
- ----
-
Y
REACTION COORDINATE
2ajMj
j=l
X+
+ j=i
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)
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
= Z E ( T )exp[--dE"/(RT)] (74)
A, + A , + X*A, + x
kf
kb (77)
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
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
=-
1 exp (16.2) exp (- --)ARTE o =
4p
16.2 - --)ARTE o (88)
TABLE I1
MAJORNEUTRAL
G A S REACTION
RATESFOR AIR BOUNDARY ( I 7)
LAYERS"
A" X
Reaction (eV) Catalyst
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.
REACTINGNONEQUILIBRIUM
CHEMICALLY BOUNDARYLAYERS
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
k, = 10'
4F
Licr
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:
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
and
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
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
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
( 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)
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)
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
REACTINGNONEQUILIBRIUM
CHEMICALLY BOUNDARYLAYERS
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
For constant 1, we redefine the similarity variable and the stream function
as
and
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
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
and for n = a
. r. sc f,,, 1
0.8 - -
0.4 -
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
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
NONEQUILIBRIUM LAYERS
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,
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
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
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
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)
-
- 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
REACTING BOUNDARYLAYERS
NONEQUILIBRIUM
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,
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
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
NONEQUILIBRIUM LAYERS
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.
(179)O
(2%)
qwo frozen or equ11.
-F(X)
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
-------2---
b5
0.4 -
O ' I
0.2
U
1 2 3 4
X
100. 1
10.
L'u
I1
' 0
v)
hi
0.
0.0
15 00 200, OOO 250, OOO
ALTITUDE, f t
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
CHEMICALLY BOUNDARYLAYERS
REACTINGNONEQUILIBRIUM
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.
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.
a solution of
with Yn(7)
1
- Y; +fYn - 2ynJYn= 0 (193)
sc
a1 = b*/Yl'(O) (198)
...
an = b,"/II% Yi'(0)
a = 6
(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
1 I 1 I
=
0 1 2 3
z
1
(0.339) \rz- SC
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
and
Conservation of reactant
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).
(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__
~
((5) = z b n P / a (220)13
n=l
+Z a n P a
m
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
fi4 Sd 4n 2
ID
= - (0.339) a, ( E ) B (jr;,3) $ l a
n=l
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
NONEQUILIBRIUM LAYERS
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
+ 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
1
Cd(4 =
(0.339) .\/2 Sch
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)
0.1 - -
1 m
I I
- -
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.
where
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
-
9=
C-Cw
ce - c
w
, F, =I
m
0
U(l - 8 ) d t
See Section E.
~831
PAULM. CHUNG
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:
when n = 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
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
e
' Y,"
////
ARB1TRARY ORlG IN
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
2 [at,yn(.;i)+
oc
c(.i,5) =
n=O
bnwn(.i)~~n(O (266)
1
- W,"
sc
+ (7 - f)W,' - n( 1 - f ' ) Y n = 0 (269)
CHEMICALLY
REACTING BOUNDARYLAYERS
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
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
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
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.
For uniform Ci,, the species conservation equation (16) and the energy
equation (27) based on temperature are transformed respectively, as
+ 2 (-e)f(j;O,
C
Cve
-
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 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
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
D. SELF-SIMILAR
SOLUTIONS REGION
AT STAGNATION
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
Energy
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
f' =1
m=l
6=1
Le = 1.4
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".
REACTINGNONEQUILIBRIUM
CHEMICALLY BOUNDARYLAYERS
I I I I I
x t
0 2 2
-
0
2 NOIl2W?1JS S V W W O l V
PAULM. CHUNG
mzo' in the above equation is zero for noncatalytic surfaces. Defining the
Nusselt and Reynolds numbers as
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
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
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.
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' = 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'
0.16 I I I
0.12
c 0.08
0.04
0 I I -
tl
FIG. 19. Mass fraction of premixed fuel (54).
E0 /RTe= 1
Ah"/cpTe=1
C
e, = 0.25
I n the limit, the quantity within the braces becomes one showing that
all the combustion energy returns to the surface.
where the subscripts 1 and 2 denote oxygen and fuel respectively. Again
employing the Arrhenius form we let
*3 Very recently, analogous nonequilibrium problems have been analyzed for certain
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)
and at 77 = 03
(337b)
I
1
_---
1
FIG. 21. Development of diffusion flame.
REACTINGNONEQUILIBRIUM
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
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
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
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)
T h e boundary conditions are as follows:
at q = O
f =f' = 0
8 = 0," (354)
m, = 0 (no surface reaction)
REACTINGNONEQUILIBRIUM
CHEMICALLY BOUNDARYLAYERS
(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:
Im;
0; + Pr fe,'
+ Sc fm,l
(359)
PAULM. CHUNG
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
REACTINGNONEQUILIBRIUM
CHEMICALLY BOUNDARYLAYERS
O L
" Pr = Sc = 0.65
8, = 3.9
I
.I
1 I I 1
1 2
rl
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
=
1
-2 (kRoF ) (RT)1
P C2
[i-i.l; - -
1
4P
exp (- RT ) (1
AE"
-- - C)] (363)
where
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).
T h e boundary conditions are as follows:
e(o) = e w , e(4 = 1
m,(O) =0 (fork, = 0)
(369)25
m(0) = 0 (fork, 3 00)
m(a)= I
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
REACTINGNONEQUILIBRIUM
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
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
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
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)
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
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
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
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
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
100 I I I I I 1 I
iB M,= 15
C, (AT EQUILIMIW)=0.527
-
0 2 4 6 8 10 12 14 16
x (feet)
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.
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
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.
I I I 1 I
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).
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:
(
KD3= exp 16.2 -
T
~ 3 x l0ls
2 k R , = 2kK3 = 2 k = (-)300
T
1.5 cm6
mole2 sec (403)
Total energy
REACTINGNONEQUILIBRIUM
CHEMICALLY BOUNDARYLAYERS
(%)= 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
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
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
atY=l
REACTINGNONEQUILIBRIUM
CHEMICALLY LAYERS
BOUNDARY
4w = ($)ah W
1 .o
- I - -
0.8 -
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).
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
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
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
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
REACTINGNONEQUILIBRIUM LAYERS
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
It is seen from Eq. (409) that the gas phase Damkohler number along
the surface varies as
(437)
With the aid of Eqs. (438) and (439), Eq. (436) is modified for all X as
CHEMICALLY
REACTINGNONEQUILIBRIUM LAYERS
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
6, = l/L(X) (44 1 )
Eq. (366) is then rewritten here as
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
?@ w is assumed to be 2.
(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
REACTINGNONEQUILIBRIUM LAYERS
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
"i".)]
fi3
(448)
where
+
d2(F p) + + =o
d(F If')
(452)
dz2 dz
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
REACTINGNONEQUILIBRIUM
CHEMICALLY BOUNDARYLAYERS
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
outer regions as shown in Fig. 33. We then let the dependent variables
in the outer regions be expanded into the following series forms.
.........
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 )
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
REACTINGNONEQUILIBRIUM
CHEMICALLY BOUNDARY
LAYERS
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)
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)
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
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.
REACTINGNONEQUILIBRIUM
CHEMICALLY BOUNDARYLAYERS
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
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
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
I. Introduction
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
'6 p2
FIG. 1
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
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
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
Low DENSITYHEATTRANSFER
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
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)
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)
AE = El -
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
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:
c,=++-dUi
dT
F. M. DEVIENNE
since
c, + 31 2(y
=1 + l)C, with y = C,/C,,
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:
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:
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:
Finally, from these equations we can obtain the heat conducted across
unit area per second from the inner cylinder:
with
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,
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:
V dS = n dS lirn
0
V,f( V )dVz
-m
dVv
-m
dV, (33)
with
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:
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
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
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:
Low DENSITYHEATTRANSFER
whence
""q= Y" [m -
u2
2
+ +"AT1 (45)
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
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.
A. TEMPERATURE
JUMP REGIME
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
Finally, the heat transfered per unit area from the inner cylinder is given
by the formula:
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
I
g/Lvalues obtained using Smoluchowski Gherke Lazareff
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
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
~
Low DENSITYHEATTRANSFER
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
TABLE I1
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:
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.
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
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
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)
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)
) T, - (1 - u)m[i
T ; ~= + (1 - a) + (1 - ... + (1 - ay-11
A. DIFFERENT
METHODS
OF MEASUREMENT
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
Low DENSITYHEATTRANSFER
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
.
Q~ = 1.5 10-4sl
1
( T , - T,) cal/cm2//sec
-
1
1 + (z-
1 81
1) 3;,
d j r = 4.71 * - T,)
10-4~lZd(T1
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
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 :
Helium MacLeod
reservoir Pump gage
Molybdenum J(
spring
I
FIG. 7.
FIG. 8.
2 I 3
FIG. 9.
Sealed electrode!S
wire wound
Tungsten (12110 m m 41
W = glass welding
FIG. 10.
conductivity cell
FIG. 11.
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
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
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.
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
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.
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 .
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.
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
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
Low DENSITYHEATTRANSFER
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
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.
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
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.
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
THEORY
B. CLASSICAL REPORTEDBY BAULE
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
or
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
Low DENSITYHEATTRANSFER
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:
with
if 0<3T
zo = T
zo = 3 if 0>3T
and
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:
E. THEORY
OF LANDAU
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)
a =-
7~
3pmt
(--)2d2c2
~kT
(133)
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
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
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)
whence finally
a=
E,'- E,
E, - Ez
=
h
k3T
sp aJ
( h ~dv) ~ G(E,E
0
+ hv)e-E/kr2dE (145)
TABLE I11
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
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
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
W41
Low DENSITYHEATTRANSFER
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
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
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
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
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
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.
T~~ = +-
1
B
sinh-l(I,,/AEp) (4)
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:
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
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
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
and
which, when combined with Eq. ( 9 ) and put into dimensionless form
yields
""
U"
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
r w -
3n'
--.-.--
+ 1 8V
4n' D
Combination of Eqs. (14) and (15) gives
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
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(*)
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
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.
and
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.
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
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:
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
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
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
.-.
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.
NOTATION
3n + I t
( 7 (applies to)
flow through round tubes).
2n 1 4
( 7
-C
(applies to)
flow through slits formed by parallel plates).
REFERENCES
W I
A. B. METZNER
C3961
HEATTRANSFER FLUIDS
IN NON-NEWTONIAN
.
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
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
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
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
,.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
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.
d A,
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
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)
(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'.
RADIATIONHEATTRANSFER
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
lpy)/l = -Fl-j + zN
i-1
( 1 - i)(v;J)/a)<-i (144
...
...
or, in general,
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
RADIATION BETWEEN SURFACES
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
RADIATIONHEATTRANSFER
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-
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
-
x2 + 3
____
FdA1- A 2-
(XA + 1)h -Xx, X = x/2R
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
RADIATION BETWEEN SURFACES
(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
I h
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
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
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:
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
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)
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
RADIATION BETWEEN SURFACES
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.
e (DEGREES)
FIG. 12. Emission and absorption characteristics of conical and vee-groove cavities.
C$T DEGREES
5 10 15 20 25 30 35 40
-
L
r,
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
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).
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.
2kt(dZT/dx2)
dx (53)
wherein
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)
1.0
.a
.6
.4
*
00 I 2 3 4
Nc
FIG. 18. Effectiveness of a plate-type radiating fin.
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
RADIATION BETWEEN SURFACES
-4r +-----2~-4
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
r/ L
FIG.20. Effect of tube surface on fin heat loss.
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
(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
C . RADIANTINTERACTION
BETWEEN FIN 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
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
RADIATIONHEATTRANSFER
BETWEEN SURFACES
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.
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
RADIATION BETWEEN SURFACES
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
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
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
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
RADIATION BETWEEN SURFACES
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
REFERENCES
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
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
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