Adenosine Triphosphate as a Function of pH, Mg2+ Ion phates with Monovalent Cations,” J. Phys. Chem.

, 60, 180
Concentration, and Ionic Strength,” J. Biol. Chem., 244, (1956).
3330 ( 1969). Stock, D. I., and C. W. Davies, “The Second Dissociation
Robinson, S. M., “Interpolative Solution of Systems of Non- Constant of Magnesium Hydroxide,” Trans. FUTU~CZY Soc., 44,
linear Equations,” SIAM J. NUVWT.Anal., 3, 650 (1966). 856 ( 1948).
Rose, I. A,, “The State of Magnesium in Cells as Estimated Taqui Khan, h4. M., and A. E. Martell, “Metal Chelates of
from the AdenyIate Kinase Equilibrium,” PTOC. Natl. Acad. Adenosine Triphosphate,” J . Phys. Chern., 66, 10 ( 1962).
Sci., 61, 1079 (1968). , “Thermodynamic Quantities Associated with
, M. Grunberg-Manago, S. R. Korey, and S. Ochoa, the Interaction of Adenosinediphosphoric and Adenosine-
“Enzymatic Phosphorylation of Acetate,” J. B i d . Chern., 211, monophosphoric Acids with Metal Ions,” ]. Am. Chem. SOC.,
734 ( 1954 ) . 89,5$85 (1967).
Satchell, D. P. N., and G. I;. White, “Kinetic Studies with Weitzel, G., and T. Spehr, “Zinc Complexes of Purines,
Acetate Kinase,” Uiochiin. Biophys. Acta, 212, 248 ( 1970). Nucleosides and Nucleotides,” Z. Physiol. Chem., 313, 212
Sigel, H., K. Becker, and D. B. McCorniick, “Ternary Com- 1958).
plexes in Solution. Influence of 2,2’ Bipyridyl on the Williamson, J. R., and E. B. Corkey, “Assays of Intermediates
Stability of 1 : l Complexes of Co2+, N?+, Lui+ and Zn2+ of the Citric Acid C cle and Related Compounds by Fluoro-
with Hydrogen Phosphate, Adenosine 5’-Monophosphate metric Enzyme Metiods,” in Methods in Enzymology, S. P.
and Adenosine 5’-Triphosphate,” ibitl., 148, 655 ( 1967). Colowick and N. 0. Kaplan, ed., Vol. 3, p. 434, Academic
Press, New York ( 1969).
Smith, R. M., and R. A. Alberty, “The Apparent Stability
Constants of Ionic Complexes of Various Adenosine Phos- Manuscript received July 28, and accepted October 8, 1976.

A Comprehensive Correlating Equation

for Laminar, Assisting, Forced
and Free Convection
A correlating equation for assisting convection was developed by com- STUART W. CHURCHILL
bining correlating equations for pure free and pure forced convection.
Department of Chemical and
These component equations are based on laminar boundary-layer theory
Biochemical Engineering
for an isothermal, vertical plate. Theoretical values for assisting convection
University of Pennsylvania
indicate that the third root of the sum of the third powers gives the best Philadelphia, Pennsylvonia I91 74
representation, as contrasted with the choice and rationalization of the
second or fourth power by prior investigators.
This expression was modified by the addition of a limiting value Nu, to
obtain a better representation below the domain of boundary-layer theory
and was generalized for uniform heating and for spheres and horizontal
cylinders by the appropriate choice of the characteristic length.

SCOPE

Heat transfer by forced convection implies a temper- general and accurate correlating equation for combined
ature difference and hence a density difference. The den- convection. Attention is confined to assisting convection,
sity difference gives rise to free convection. This coupled that is, to forced flow in the same direction as the buoyant
process has been the subject of both experimental and motion, and to the Iaminar regime but includes the ef-
theoretical investigations for a variety of geometries and fects of the Prandtl number, boundary conditions, and
conditions. A number of different correlating equations shape.
have been propoied for particular cases. However, the The correlating equation is based on theoretical results
accuracy, generality, and range of validity of these corre- for free convection, forced convection, and combined con-
lations have not been tested critically. vection for vertical plates but was tested with experi-
The objective of this work has been to develop a more mental data for plates, spheres, and horizontal cylinders.

CONCLUSIONS A N D SIGNIFICANCE
Equation (18), together with supplementary Equations uniform surface temperature and uniform heating. This
(11) and (12) and the coefficients in Table 1, Provide a expression appears to be valid for spheres and horizontal
good representatio11 for the data for both cylinders if the indicated characteristic lengths are used.
the local and overall Nusselt number for combined forced
and free convection from a vertical plate over the entire It appears to be applicable for other shapes, such as
potential flow and laminar boundary-layer regimes (Re wedges and vertical cylinders, based on the fragmentary
< 104, Ra < 109) for all Prandtl numbers and for both values which are available.

Page 10 January, 1977 AlChE Journal (Vol. 23, No. 1)

 Powers of 2 and 4 rather than the third power of Additional theoretical solutions for spheres and hori-
Equation (18) were chosen for correlation by earlier zontal cylinders are needed to provide firm expressions
workers, apparently because of physical rationalizations, for the dependence of free and forced convection on PT
imprecise data, uncritical tests, and insensitivity of this and to test Equations (10) and (18) for combined con-
type of expression to the value of the exponent. vection.
Equation (18) reduces to Equation (10) for the laminar Some experimental studies of forced, free, and com-
boundary-layer regime (15 < Re < lo4, 2 300 < Ra < bined convection from spheres have yielded values of N u
l o g ) . Equation (10) is based wholly on theoretical values. and Sh far greater than indicated by Equation (18). This
For very small wires, such that RePr and ( R u P T ) are~ discrepancy is worthy of resolution and explanation.
both less than ~ / 5 Equation
, (17), based on potential flow, Presentation of tabulated values of the individual di-
may be more reliable than Equation (18). Also, the ex- mensionless groups, N u , Re, PT and Gr or Ra, is desirable
perimental results of Gebhart and Pera (1970) for small in both theoretical and experimental studies. Graphical
wires reveal a significant dependence on L/d in this correlation of combined groups makes their accurate in-
regime, presumably due to three-dimensional effects. clusion in subsequent correlations difficult or impossible.
Equations (181, (lo), and (17) are presumed to be valid Attempts to generalize the correlating equation in some
for component transfer with Shy Sc, and Ra' substituted simple fashion to include forced convection oblique to or
for N u , PT, and Ra, although confirmatory data are not opposed to free convection were unsuccessful and, as sug-
available. gested by Weber (1976), may not be feasible.

PREVIOUS WORK with the plus sign for assisting and the minus sign for
opposing flow. Equation ( 3 ) corresponds to Equation ( 2 ) ,
McAdams (1942) simply recommended the use of the
again with n = 2, in terms of Re and Gr but yields the
higher of the heat transfer coefficients for free and forced
wrong dependence on Pr for large Re and Pr and for large
convection alone for the combined process. However, most
GTand small Pr.
of the correlating equations that have subsequently been
Other investigators have generally chosen n = 4. Borner
proposed have utilized the sum of some arbitrary power (1965) correlated data for flat plates, cylinders. and
of correlating equations for free and forced convection, spheres, with assisting and opposing convection, by adding
namely Re and ( G r / 2 )'h vectorially. For assisting convection this
NU" = NU$' +NUN" (1) yields n = 4, but again E / F = 2'14. Acrivos (1966) used
where the subscripts F and N indicate forced and free Equation ( 1 ) for correlation of his computed values for
(natural) convection, respectively. [Equation ( 1 ) has the assisting convection near the forward stagnation point of
same form as the general expression proposed by Churchill horizontal cylinders. He chose an exponent of 4 to yield
and Usagi (1972) for correlation, if NUF and NUN are in- a satisfactory numerical approximation of the more com-
terpreted as the asymptotic solutions for Ra/Re2 ap- plex theoretical dependence. Hatton et al. (1970) added
proaching zero and infinity.] vectorially Red and an effective Red for free convection
Many other correlating equations have been based on from electrically heated cylinders with the effective Red
the specification of an equivalent forced velocity for free obtained by equating the correlations for pure forced and
convection. Lemlich and Hoke ( 1956) demonstrated that pure free convection. The result for assisting flow is equiv-
this concept can be used to superimpose the correlations alent to Equation ( 2 ) with n = 4. Jackson and Yen
for pure free and pure forced convection to cylinders. (1971) recorrelated the data of Oosthuizen and Maden
However, the choice of an equivalent velocity is somewhat (1970) for assisting flow of air over heated horizontal
arbitrary. In addition, an arbitrary combining rule for the cylinders using the equivalent of Equation (2) with p =
imposed and equivalent velocities is necessary for mixed ?h, q = 1/4, n = 4, and E / F = 1. They rationalized the
convection. The usual combining rule for this case is again exponent of 4 on the basis of the additivity of the work
the nth root of the nth power of the components, namely of free and forced convection. Oosthuizen and Bassey
(1973) used the same exponent for their data for assisting
Nun = ( E Rep) + (FGT~) (2) flow of air over heated vertical plates but added a con-
stant term to both of the boundary-layer expressions to
where p and q for the laminar boundary-layer regime are account for the behavior in the conductive regime.
M and 1/4, respectively, and E and F are functions of Pr. A slight modification of Equations ( 1 ) and ( 2 ) leads to
Equation ( 2 ) can be recognized as a special case of ( 1 ) . the correlation of the effect of free convection superim-
The various prior correlating equations for assisting con-
posed on forced convection, or vice versa, in terms of the
vection can be reviewed concisely in terms of these two group GT/Re2, in an asymmetric form. Thus Sparrow and
models. Gregg (1959) proposed the expression
Krischer and Loos (1958) postulated that Re in graphi-
cal correlations for forced convection from air to various
bodies be replaced by Re +( G r / 2 ) ' h which is equivalent (4)
to using pn = 1, qn = 1/2 and E / F = 2"n in Equation
( 2 ) . If p = YZ and q = y4, n = 2 and E / F = 2'14. to represent the computed behavior for vertical plates at
Brdlik et al. (1974) postulated velocity and temperature GT,/ReX2 << 1, with the plus sign for assisting and the
distributions to derive an expression for N u , for vertical minus sign for opposing flow. Eshghy (1964) proposed the
plates. This expression was then utilized by an unexplained corresponding expression
procedure to obtain
(Nu/0.68) = (0.952 + Pr) %Gr" 2 PrRe (3)

AlChE Journal (Vol. 23, No. 1) January, 1977 Page 11

for Re/Gr'h << 1. Yuge (1960) correlated the deviations The initial development is carried out for isothermal
of his data for mixed convection with air and spheres from vertical plates because of the availability of analytical so-
the limiting correlations as exponential functions of lutions for pure convection for the limiting cases of large
0.493Red1/z - 0.392Crd". [However, Klyachko ( 1963) and smaIl Pr and of reliable theoretical values for inter-
contends that these data are in error on the high mediate Pr.
side owing to movement of air in the laboratory.
He then proposes a very complicated set of equa- LAMINAR BOUNDARY-LAYER REGIME ON
tions for the correlation of other data.] Oosthuizen VERTICAL PLATES
and Madan (1970) used a three-term power series in Uniform Wall Temperoture
Grd/Red2 to correlate their data for the assisting flow of
Churchill and Ozoe (1973b) and Churchill and Usagi
air over heated cylinders. Fand and Keswani (1973) de-
(1972) developed and confirmed the accuracy of the fol-
rived separate dependences on Grd/Red2 depending on the
lowing correlating equations for the local rate of heat
range of Grd/Redi3for their data for water and a heated
cylinder. For 0.5 < GTd/Red2 6 4, they added the term transfer by pure free and pure forced convection in the
laminar boundary-layer regime:
0.033 ( Grd/Red2)o.3Grd'/4 to the terms for forced convec-
tion. For 2 < Grd/Red2< 40, they used an expression NU,F = 0.339 R C , ~ / ~ P ~ ~ +2(0.0468/Pr)2/3]1/4
/~/[:I (6)
equivalent to Equation ( 2 ) with n = 2, p =.$, and q =
%. Nakai and Okazaki (1975) proposed correlations for
the deviations from the limiting correlations for very small These expressions are therefore postulated as the com-
cylinders in terms of the group PrRed3/NdGrd. Gebhart ponent parts of Equation ( 1 ) . In order to test Equation
and Pera (1970) noted for small wires that the effect of ( l ) ,it is convenient to rearrange it in the following two
superimposed free convection depended on B Gr/RedP/Q, forms.
with p / 9 and B varying with Pr and L/d. The ratio p / 9
varied from 2 to 4.
Morgan ( 1975) recently reviewed convection to cylin-
and
ders. However,his attention to assisting convection is es-
sentially limited to the conditions that produce a 5% in-
crease in Nu above the value for forced convection alone.
In summary, Equation ( 1 ) appears to be generally satis-
(Zy+ (Z)" =1 (9)

Figure 1 is a plot of the several sets of theoretical values

factory for correlation for combined convection. However, for combined convection in terms of the groupings in
some uncertainty exists concerning the component parts, Equations (8) and ( 9 ) and the details provided by Equa-
particularly with respect to Pr, the best value of the ex- tions ( 6 ) and ( 7 ) . The symmetry of Equations (8) and
ponent n, and the behavior for very small cylinders. ( 9 ) suggests the use of these split coordinates. The split
Solutions and correlating equations for pure forced and coordinates permit the use of arithmetic scales and hence
pure free convection, and experimental and theoretical the display of absolute deviations. Indeed, this plot pro-
values for assisting convection, will be discussed below in vides such a critical test that the computed, theoretical
connection with the development of the new correlation. values, as tabulated in the original references, demonstrate
some scatter. Nu,/Nu,F is plotted vs. N U , N / N U ~ Fon the
GENERAL DEVELOPMENT left side of Figure 1 and N u z / N u , ~vs. N u ~ F / N u ~on
N the
The development herein consists of selection of corre- right side. Curves corresponding to n = 2, 3, and 4 are
lating equations for pure free and pure forced convection included. The theoretical values follow these curves, con-
to isothermal, vertical plates; a test of the suitability of firming the applicability of Equation ( 1 ) for assisting con-
Equation ( 1 ) for combined convection; evaluation of the vection. The precise values of Merkin (1969) for Pr = 1
exponent n in Equation ( 1 ) using theoretical values for are represented by the curve for n = 3 within 1%except
assisting convection; generalization of the correlation for for two odd points. The values of Szewczyk (1964) for
other boundary conditions and shapes; and comparison of Pr = 0.72 agree equally well with the curve for n = 3
the resulting expression with experimental values. for small Ra, but are 5% high at higher Ra,. Merkin also
noted this discrepancy in Szewczyk's values. The values
of Lloyd and Sparrow (1970) for Pr A 0.72 generally
follow this same curve, but the values for Pr = 10 and
100 are more closely represented by the curve for n = 4.
On balance, a value of 3 appears to be the best choice for
n. The resulting correlation can be written as

(10)
with
fF{pr} = [1+ (cF/Pr)2/3]1'4 (11)
fN{Pr} = [1 + (CN/Pr)9/16]4/9 (12)
Equations (101, ( l l ) ,and (12) have a sufficiently gen-
eral form to apply to the mean as well as the local Nusselt
number of a vertical plate, to other boundary conditions,
and to other shapes. For example, an expression for Nzc
in the form of Equation (10) can be obtained simply by
integrating h in Equations ( 6 ) and (7) with respect to x

Page 12 January, 1977 AlChE Journal (Vol. 23, No. 1)

 TABLE
1. RECOMMENDED AND CHARACTERISTIC
CONSTANTS LENGTHS

Values from boundary-layer theory

1 AF AN CF CN
Vertical plate
local
uniform T X 0.339 0.503 0.0468 0.492 0
uniform i X 0.464 0.563 0.0205 0.437 0
mean
uniform T H 0.677 0.670 0.0468 0.492 0
uniform j H/2 0.656 0.669 0.0205 0.437 0
Horizontal cylinder
mean
uniform T ad 1.08") 0.690 0.412 0.559 1.0'2)
uniform i
Sphere
Ird - 0.694 0.442 - -
mean
uniform T
General approximate values
0.659
0.67(5)
-
0.45(6)
-
0.45

Notes
I-experimental
O f r o m the solution of Ring ( 1914) for potential flow
3-from the solution for pure conduction
4-except for Nu, and cylinders
5-except for Nu,
6-except use 0.035 for vertical plates
7-use 0.5 for vertical plates, 1.0 for cylinders and 7 for spheres

from 0 to H and repeating the rest of the derivation. The to use the higher of the limiting coefficients is equivalent
corresponding values of AF and AN are given in Table 1; n, to letting n + co and results in a maximum underestimate
CF and C N do not change. Derivations for another bound- of only 21%.
ary condition and for other shapes follow. Equations ( l o ) , ( l l ) , and (12) indicate that the di-
mensionless group that determines the relative importance
Uniform Heat Flux of free and forced convection in the laminar boundary-
Churchill and Ozoe (1973a, 1973c) developed the fol- layer regime is Ra/Re2Pr4/3 = Gr/Re2Pr1/3 in the limit
lowing correlating equations for uniform heat flux analo- of PT + co and Ra/Re2Pr = Gr/Re2 in the limit of Pr -+ 0.
gous to Equations ( 6 ) and ( 7 ) for uniform temperature: [These same groupings were noted by Acrivos (19SS)l.
Nu,F + (0.0205/Pr)2/3]1/4 (13)
= 0.464 Re2'/2Pr1/3/[1 Why have prior correlations generally been developed in
terms of Gr/Re2 instead of Gr/Ra2Pr1/3,which is a reason-
NuzN = 0.563 RaX1f4/[:l+ (0.437/Pr)9/1614/9(14) able approximation for most fluids? Apparently, this choice
was based on incomplete rationalization or the confinement
Figure 2 is a plot of the computed values for uniform heat of attention to one Auid.
flux in the same form as Figure 1. The precise values of
Wilks (1974) for Pr = 1.0 are generally represented
LAMINAR BOUNDARY-LAYER REGIME FOR OTHER
within 1% by the curve for n = 3. Wilks (1973) earlier SHAPES AND CONDITIONS
values for different Pr are also represented by this curve,
although a few values scatter, probably due to numerical Correlations for horizontal cylinders, equivalent to Equa-
error. Hence, Equation (10) is applicable for uniform heat- tions ( 7 ) , (8), and (14) for vertical plates, have been
ing as well as uniform temperature with the use of the ap- developed by Churchill and Bernstein ( 1976), Churchill
propriate values of A F , C F , A N , and CN.The choice of the and Chu ( 1975a), and Churchill ( 1974). Insufficient theo-
mean temperature difference at x = H/2 is recommended
since it yields coefficients AF and A N which d d e r negligibly
from those for uniform temperature. The values of CF and
CN do not differ greatly, and the correlation is quite in-
sensitive to these values for the practical range of P r . The
compromise values of AF, AN, CF, and CN suggested at the
bottom of Table 1 provide a good approximation for Nu
for both boundary conditions and probably for intermedi-
ate ones.
Interpretation
Why did most prior investigators choose n = 2 or 4
instead of 3? Apparently their choice was based on some
loose rationalization or cursory examination and was not
tested critically. Equation (10) is actually quite insensi-
tive to the choice of n in this range. A choice of n = 4
would result in a maximum decrease of 6% in the pre-
dicted value of Nu and a choice of n = 2 a maximum in- Fig. 2. Development of a correlating equation for ossisting forced
crease of 1270. The recommendation of McAdams (1942) and free convection from a uniformly heated plate.

AlChE Journal (Vol. 23, No. 1) January, 1977 Page 13

retical and experimental values are available to develop convection at very small Ra. The latest and most reliable
equivalent correlations for the effect of Pr on spheres. seem to be those of Nakai and Okazaki (1975), who ob-
However, Churchill and Churchill (1975) have shown that tained expressions which can be written as follows:
correlations for free convection from vertical plates can be -
adapted as reasonable approximations for cylinders and NuF = 2?r/ln [16.32/RePr], RcPr < n/5 (15)
spheres if 'is d and 'is d/S are used as the characteristic di-
mensions in N u and Ra. Krischer and Loos (1958) pro- = 6 ~ / l n[4 831( P r + 9.4)'h/NuRaPr],
ZN
posed a similar generalization for combined convection to ( R a P r ) s < ?r/5 (16)
many shapes but used n d/2 rather than T d for cylinders.
However, they apparently made a numerical error in com- Equations (15) and (16) yield much larger values of
puting values oi this characteristic dimension for cylinders. Nu than the components of Equation (10) as Re and Ra
In the absence of computed values, the same coeffi- approach zero. The experimental data of both Nakai and
cients, CF and CN, will be postulated for spheres as for Okazaki (1975) and Gebhart and Pera (1970) for assisting
cylinders. Experimental values for AF are listed in Table 1 convection indicate that a satisfactory correlating equation
for both cylinders and spheres. The use of the same for the creeping-flow regime can be constructed by add-
characteristic dimension as above for free convection yields ing GUF and RiuN to a high power such as 10. Such a large
a value for AF for spheres which is close to that of flat
plates, but the value for cylinders is much higher. All of power implies that the higher of the values of Nu given by
these values are summarized in Table 1. Together with Equations (15) and (16) may also be used as a very close
Equations (10) ( l l ) ,and (12) they constitute correla- approximation. An alternative correlation can be con-
tions for the laminar boundary-layer regime for all three structed by fmt inverting and then raising Equations (15)
shapes and for the two indicated boundary conditions. and (16) to the eth power before combining. This pro-
The computed values of Chen and Mucogh (1975) for cedure leads to a correlating power which can be rounded
combined convection to a vertical cylinder can be repre- off to -2, hence to
sented approximately by Equation (10) if N u , is replaced
by N u , - 0 . 8 8 ( x / d ) . The range of validity of this ex- 4n
--=In 1
pression is uncertain, since their results were obtained by -Nu RePr NuRaPr
the method of local nonsimilarity, are limited to x/d Rez?'z
< 0.5, Gr,/ReX2 < 2.0, and Pr = 0.7, and were not tested (=)'+ ( 4831(P~+9.4)'/~
with experimental data except for the limiting case of a (17)
flat plate. Equations (lo), ( l l ) ,and (12) are undoubt- The data of Gebhart and Pera indicate that L / d may be a
edly applicable for other geometrical configurations, such significant parameter in the creeping-flow regime, particu-
as inclined plates, wedges, and vertical cones, for other larly for large PT. This dependence apparently stems from
boundary conditions and for nonNewtonian fluids when the three-dimensional motion generated at the ends of the
limiting solutions for pure convection and computed values wire.
for combined convection become available. For example, Insufficient theoretical and experimental results have yet
Sparrow et al. (1959) give solutions for two particular been obtained for other shapes to support the development
wedge flows. Their values of Mu, can be represented very of a generalized correlation for the creeping-flow regime.
closely by Equation (13) simply by choosing an appropri- Unfortunately, a considerable gap exists between the
ate characteristic length. However, these two conditions indicated upper range of validity of Equations (15) and
are insufiicient to establish a general formulation. (16) and the indicated lower limit of applicability of the
The extension of the correlation to include forced con- components of Equation ( 10). One possibility would be to
vection oblique to or opposing free convection was at- develop an expression for interpolation between these
tempted, but a competely successful generalization was equations. However, the singularity in Equations (15) and
not discovered. Weber (1976) asserts that such a general- (16) is a serious impediment. A simpler alternative is to
ization would be inappropriate on theoretical grounds, approximate the limiting behavior in the creeping-flow
owing to the basic difference in the fluid motion. regime by a constant value N u , and then to interpolate
between N u , and the expressions for the laminar boundary-
CREEPINGFLOW REGIME (Re A N D Ra + 0 ) layer regime. The use of Nu, has a sound theoretical basis
and a value of v for pure conduction from spheres, with
Laminar boundary-layer theory eventually becomes in- ?i d/2 as a characteristic length. The theoretical solution
valid as Re and Ra approach zero. This limit occurs when of King (1914) for forced convection to cylinders yields
the thickness of the boundary layer approaches the charac- a value of 1.0 for Nu,, with ?i d as a characteristic length,
teristic dimension of the surface in magnitude. The regime but is based on a questionable boundary condition.
of lower Re and Ra is not of great practical importance for Churchill (1976) and Churchill and Chu (1975b) have
flat surfaces but may be for small horizontal cylinders shown that N u , = 0.5 provides a satisfactory empirical rep-
(wires) and for small particles. resentation for the experimental data for both forced and
The correlating equation developed by Churchill and free convection to flat plates. These authors and also
Chu (1975n) for free convection from horizontal cylinders Churchill and Chu ( 1 9 7 5 ~ )and Churchill and Bernstein
indicates that laminar boundary-layer theory may be used (1977) have demonstrated that a simple sum of Nu, and
with less than 10% error down to Ra = 72 000 for Pr + 00 the laminar boundary-layer equations provides a satisfac-
and to Ra Pr = 40 000 for Pr + 0. The correlating equa- tory interpolating equation for pure forced and free con-
tion of Churchill and Bernstein (1977) for forced convec- vection from both plates and cylinders. Rather than add
tion to cylinders indicates that the corresponding limits N u , to both components of Equation (lo), it can be
are Re Pr2/3 = 50 for Pr + 00 and Re Pr = 36 for Pr -+ 0. subtracted from N u , yielding
These limits are presumed to apply approximately to com-
bined convection and other geometries and hence to Equa-
tion (10).
A number of analytical solutions have been derived for
pure forced convection at very small Re and for pure free

Page 14 January, 1977 AlChE Journal (Vol. 23, No. 1)

 30
proper weighting factors, if any, to be applied to the sev-
eral sets of data and to the different ranges of Re, Ra, and
25 . A KLyschko - S p h e r e s Pr.
V Yuge - S p h e r e s
x Oorlhulzen & Madan -Cylindorr
In summary, Equation (18) is proposed as a working
0 K r i s i h c r 6 L o o s - F l a t Plates relationship for all shapes, boundary conditions, and Pr
4 K r i r c h e r h L O o r - Spheres
(or Sc) for the entire laminar regime. For the turbulent
~

+
f Krlrcher LLoor- Cylinders
. Fand A Kerxani - Cylinders
regime with R812 > lo6, it should serve as
a lower bound for Nu. For very small cylinders, such that
( R a P r ) and Re P r < w / 5 , Equation (17) or the tenth root
of the sum of the tenth powers of Equations (15) and
10
,- (16) may give a more accurate value for Nu. In this re-
+ ' ,,-Free Convection I Y : X I gime, Equation (18) appears to be an upper bound and
Equation (10) a lower bound.

ACKNOWLEDGMENT

Fig. 3. General correlation for assisting forced and free convection. The provision of tabulations of experimental vaIues for flat
plates by Professor P. H. Oosthuizen is greatly appreciated. The
criticisms and suggestions of the anonymous reviewers were
The recommended values of Nu, are included in Table 1. most helpful.
Equation (18) is proposed as a practical representation
for assisting convection for the entire laminar regime for NOTATION
all PT, all geometries, and all boundary conditions, with = coefficient in correlation for forced convection,
AF
the recognition that the values of N u , AF, AN, CF,and CN Equation (10)
and the expressions for I may be improved or others added . r

as additional theoretical or precise experimental work is AN = coefficient in correlation for free (natural) con-
vection, Equation (10)
carried out. B = function of Pr and L/d
For Re and Ra less than the values indicated as limits
for Equations (15) and (16), the tenth root of the sum of
CF = central value of Pr (or SC) for forced convection,
Equation ( 6 )
the tenth powers of Equations ( 15) and (16) or Equation CN = central value of Pr (or SC) for free (natural) con-
(17) may give more accurate values than Equation (18). vection, Equation ( 7 )
d = diameter, m
TURBULENT REGIME E = function of Pr, Equation (2)
F = function of Pr, Equation ( 2 )
The validity of the component parts of Equations (10) f ~ { P r }= function of Pr for forced convection, Equation
and (18) is presumed to be limited to approximately Re
< lo4 and Ra < lo9, respectively, owing to the onset of (11)
f ~ { P r }= function of Pr for free (natural) convection,
turbulent motion. Accordingly, these equations might be Equation (12)
presumed to have an approximate upper limit of validity = acceleration due to gravity, m/sz
corresponding to (Re"2 +
' ) (104)3/2= lo6.
R U ~ / ~ P T -= g
Gr = g@4T13//v2= Grashof number
This limit seems worthy of experimental investigation. H = height of plate, L
i = heat flux density, W/mZ s
COMPARISON WITH EXPERIMENTAL D A T A k = thermal conductivity, W/m2 s K -
Representative experimental data for flat plates, spheres,
'k = component transfer coefficient, m/s
I = characteristic length given by Table 1 or subscript
and cylinders are plotted in Figure 3. The theoretical m
values plotted in Figures 1 and 2 are not included, since L = length of cylinder, m
most of them would not show detectable scatter from n = exponent, Equation (1)
Equation (18) with these more condensed scales. The Nu = jl/khT = Nusselt number
experimental values for vertical plates scatter widely but -
Nu = mean Nusselt number
rather randomly about Equation ( 18). The data for spheres
Pr = v / a = Prandtlnumber
and horizontal cylinders are more limited and scatter more
widely, but again randomly. The results of Fand and p = exponept of Re, Equation (2)
Keswani (1973) for horizontal cylinders are plotted as a q = exponent of Gr, Equation ( 2 )
dotted line, since only smoothed, indirect values were Q = graphical function of Pr determined by Sparrow
and Gregg (1959)
given. The discrepancy may be due to this smoothing or Qr, Qz = graphical functions of Pr determined by Eshghy
to misinterpretation herein rather than to experimental
(1964)
error. The data of Garner and Hoffman (1960) for mass Rn = gp4T13/va = Rayleigh number
transfer from spheres were omitted since they converge to Ra' = g74013/vD = Rayleigh number for component
excessively high values of S h a s Re approaches zero. Like- transfer
wise, the data of Pei (1965) for heat transfer from spheres Re = Z U / V = Revnolds number
were omitted, since they converge to excessively high Sc = v / D = Schmidt number
values of N u i n both of the limits of pure convection. Re- Sh = k'l/D = Sherwood number
grettably, many of the published data, including the ex- u = velocity, m/s
tensive values of Borner ( 1965), Narasimhan et al. ( 1967), 1c = distance up plate, m
Wilhelm (1969), and Brdlik et al. (1974) are given only
in plots of combined variables which do not permit ac- Greek Letters
curate conversion to the coordinates of Figure 3. A sta- a = thermal diffusivity, m2/s
tistical analysis of the values plotted in Figure 3 does not p = volumetric coefficient of expansion with tempera-
appear to be justifiable in view of the uncertainty in the ture, K-l

AlChE Journal (Vol. 23, No. 1) January, 1977 Page 15

y = volumetric coefficient ofexpansion with composi- Forced and Natural Convection with Low-Speed Air Flow
tion over Horizontal Cylinders,” 1.Fluid Mech., 42, 17 ( 1970).
AT = temperature difference, K Jackson, T. W., and H. H. Yen, “Combining Forced and Free
AO = composition difference Convection Equations to Represent Combined Heat-Transfer
Coefficients for a Horizontal Cylinder,” 1. Heat Transfer,
v = kinematic viscosity, m2/s Tram. ASME, 93C, 247 ( 1971 ).
(p{Pr} =function of Pr, Equations (4)a n d (5) King, L. V., “On Convection of Heat from Small Cylinders in a
Subscripts Stream of Fluid,” PhiE. Trans. Roy. SOC. ( L o n d o n ) , 214A,
373 (1914).
d = based on diameter Kliegel, J. R., “Laminar Free and Forced Convection Heat
F = forced convection Transfer from a Vertical Flat Plate,” Ph.D. thesis, Univ.
N = free (natural) convection Calif., Berkeley (1959).
o = limiting value as Re a n d Ra + 0 Klyachko, L. S., “Heat Transfer between a Gas and A
x = local value based on x Spherical Surface with Coinbined Action of Free and Forced
Convection,” I . Heat Trunsfer, Trans. ASME, 85C, 355
LITERATURE CITED ( 1963).
Krischer, O., and G. Low, “WPrnie- und Stoffaustausch bei
Acrivos, A., “On the Combined Effect of Forced and Free erzwungener Stromung an Korpern verschiedener Form, Teil
Convection Heat Transfer in Laminar Boundary Layer 1,”Chem. Ing. Tech., 30, 31 ( 1958).
Flows,” Chem. Eng. Sci., 21, 343 (1966). Lemlich, R., and R. Hoke, “A Common Basis for the Correla-
Borner, H., “Uber den Warme- und Stoffubertragang urn tion of Forced and Natural Convection to Horizontal
umspiilten Einzelkorpern bei uberlaeerune von freier und Cylinders,” AIChE I., 2, 249 ( 1936).
erz4ungener Stromung,” Ver. Ueut. 1;g. b&chungsheft, No. Lloyd, J. R., and E. M. Sparrow, “Combined Forced and Free
512 (1965). Convection Flow on Vertical Surfaces,” Intern. J. Heat Mass
Brdlik, P. M., I. A. Kozhinov, and A. A. Petrova, “Approximate Transfer, 13, 434 ( 1970).
Solution to the Problem of Local Heat Transfer at a Vertical McAdams, W. H., ed., Heat Transmission, 2 ed., p. 217,
Plate under Conditions of Laminar Mixed Convection,” McGraw Hill, New York (1942).
Intern. Chem. Eng., 14, 65 (1974). Merkin, J . H., “The EEect of Buoyancy Forces on the
Chen, T. S., and A. Mucoglu, “Buoyancy Effects on Forced Boundary-Layer Flow over a Semi-Infinite Vertical Flat
Convection along a Vertical Cylinder,” 1. Heat Transfer, Plate in a Uniform Free Stream,” I. Fluid Mech., 35, 439
Trans. ASME, 97C, 198 ( 1975). (1969).
Churchill, S. W., “Laminar Free Convection from a Horizontal Morgan, V. T., “The Overall Convective Heat Transfer from
Cylinder with a Uniform Heat Flux Density,” Letters in Heat Smooth Circular Cylinders,” Aduances in Heat Transfer, 11,
Mass Transfer, 1, 109 ( 1974). 199 (1975).
-, “A Comprehensive Correlating Equation for Forced Nakai, S., and T. Okazaki, “Heat Transfer from a Horizontal
Convection from Flat Plates,” AIChE J., 72, 264 ( 1976). Circular Wire at Small Reynolds and Grashof Numbers-I1
-, and M. Bernstein, “A Correlation for Forced Con- Mixed Convection,” Intern. 1. Heat Mass Transfer, 18, 397
vection from Gases and Liquids to a Circular Cylinder in (1975).
Crossflow,” 1. Heat Trans., 99C, (1977), to be published. Narasimhan, C., and W. H. Gauvin, “Heat and Mass Transfer
Churchill, S. W., and H. H. -S. Chu, “Correlating Equations to Spheres in High Temperature Surroundings,” Can. J.
for Laminar and Turbulent Free Convection from a Hori- Chem. Eng., 45, 181 ( 1967).
zontal Cylinder,” Intern. 1. Heat Mass Transfer, 18, 1049- Oosthuizen, P. H., and M. Bassey, “An Experimental Study of
1053 ( 1 9 7 5 ~ ) . Combined Forced- and Free-Convection Heat Transfer from
, “Correlating Equations for Laminar and Flat Plates to Air at Low Reynolds Numbers,” I. Heat
Turbulent Free Convection from a Vertical Plate,” ibid., in Transfer, Trans. ASME, 95C, 120 (1973).
press ( 1975b 1. Oosthuizen, P. H., and S. Madan, “Combined Convective Heat
Churchill, S. W., and R. U. Churchill, “A Comprehensive Cor- Transfer from Horizontal Cylinders in Air,” ibid., 92C, 194
relating Equation for Heat and Component Transfer by (1970).
Free Convection,” AIChE I., 21, 604 ( 1975). Pei, D. C. T., “Heat Transfer from Spheres under Combined
Churchill, S. W., and H. Ozoe, “Correlations for Laminar Forced and Natural Convection,” Chem. Eng. Progr. Sym-
Forced Convection with Uniform Heating in Flow over a posium Ser. No. 59,61, 57 ( 1965).
Plate and in Developing and Fully Developed Flow in a Sparrow, E. M., and J . L. Gregg, “Buoyancy Effects in Forced-
Tube,” 1. Heat Transfer, Trans. ASME, 95C, 78 (1973a). Convective Flow and Heat Transfer,” I . Appl. Mech., 26,
, “Correlations f o r * Laminar Forced Convection Trans. ASME, 81E, 133 (1959).
in Flow over an Isothermal Flat Plate and in Developing and
Fully Developed Flow in an Isothermal Tube,” ibid., Sparrow, E. M., R. Eichhorn, and J. C. Gregg, “Combined
Forced and Free Convection in Boundary Layer Flow,”
416 (1973b). Phys. Fluids, 2, 319 ( 1959).
, “A Correlation for L,aminar Free Convection Szewczyk, A. A., “Combined Forced and Free-Convection
from a Vertical Plate,” ibid.,540 ( 1 9 7 3 ~ ) .
Laminar Flow,” I. Heat Transfer, Trans. ASME, 86C, 501
, and R. Usagi, “A General Expression for the Correla- (1964).
tion of Rates of lransfer and Other Phenomena,” AlChE J.,
Weber, M. R., “Letter to the Editor,” AlChE I., 22, 207
18, 1121 (1972).
(1976).
Eshghy, S., “Forced-Flow Effects on Free-Convection Flow
Wilhelm, R., “Der Stoffiibergang fest-flussig bei Einzelkorpern
and Heat Transfer,” J. Heat Transfer, Trans. ASME, 86C,
und Schiittungen in freier und erzwungener Stromung,”
290 ( 1964).
V D I Forschzmgshaft, No. 531 (1969).
Fand, R. W., and K. K. Keswani, “Combined Natural and
Forced Convection Heat Transfer from Horizontal Cylinders Wilks, G., “Combined Forced and Free Convection Flow on
Vertical Surfaces,” Intern. J. Heat Mass Transfer, 16, 1958
to Water,” Intern. 1. Heat Mass Transfer, 16, 1175 ( 1973).
( 1973 ).
Garner, F. H., and J. M. Hoffman, “The Transition from Free , “The Flow of a Uniform Stream over a Semi-Infinite
Convection to Forced Convection in Mass Transfer from Vertical Flat Plate with Uniform Surface Heat Flux,” ibid.,
Solid Spheres,” AlChE I., 6, 579 ( 1960). 17,743 (1974).
Gebhart, B., and L. Pera, “Mixed Convection from Long
Horizontal Cylinders,” 1. Fluid Mech., 45, 49 ( 1970). Yuge, T., “Experiments on Heat Transfer from Spheres In-
cluding Combined Natural and Forced Convection,” I . Heat
Gryzagoridis, J., “Combined Free and Forced Convection from Transfer, Trans. ASME, 82C, 214 (1960).
an Isothermal Vertical Plate,” Intern. 1. Heat Mass Transfer,
18,911 (1975). Manuscript received June 23, 1976; revision received September 28
Hatton, A. P., D. D. James, and H. W. Swire, “Combined and accepted October 7, 1976.

Page 16 January, 1977 AlChE Journal (Vol. 23, No. 1)