Numerical Heat Transfer

ISSN: 0149-5720 (Print) (Online) Journal homepage: http://www.tandfonline.com/loi/unht19

ANALYSIS OF NATURAL CONVECTION IN THE SPACE

BETWEEN CONCENTRIC VERTICAL CYLINDERS OF
DIFFERENT HEIGHT AND DIAMETER

M. Charmchi & E. M. Sparrow

To cite this article: M. Charmchi & E. M. Sparrow (1982) ANALYSIS OF NATURAL CONVECTION
IN THE SPACE BETWEEN CONCENTRIC VERTICAL CYLINDERS OF DIFFERENT HEIGHT AND
DIAMETER, Numerical Heat Transfer, 5:2, 119-144, DOI: 10.1080/10407788208913439

To link to this article: http://dx.doi.org/10.1080/10407788208913439

Published online: 26 Jan 2017.

Submit your article to this journal

View related articles

Citing articles: 8 View citing articles

Full Terms & Conditions of access and use can be found at

http://www.tandfonline.com/action/journalInformation?journalCode=unht19

Download by: [New York University] Date: 05 February 2017, At: 12:35
 Numerical Heat Transfer, vol. 5 , pp. 119-144,1982

ANALYSIS OF NATURAL CONVECTION I N THE

SPACE BETWEEN CONCENTRIC VERTICAL
CYLINDERS OF DIFFERENT HEIGHT
AND DIAMETER

M. Charmchi and E. M. Sparrow

Department o f Mechanical Engineering, University o f Minnesota.
Minneapolis, Minnesota 55455

Numeriml finite-difference solutions were carried out for natural convection in the
enclosed space between two concentric vertical cylinders, with eoch cylinder being
maintained at a different unifom temperature. In the investigated configurations.
the inner cylinder was of lesser height than the outer cylinder. For the computations,
the height of each cylinder was set equal to its diameter. The investigated parameters
included the ratio of the inner-cylinderdiameter to the outer-cylinderdiameter, the
vertical position of the inner cylinder, and the Rayleigh number of the inner cylin-
der: the Prandtl number was fixed at 0.7 for all cases. The presentation of results
encompassed streamline and isothem maps, velocity profiles, per-sutface and overall
heat tronrfer rates and Nusselt numbers, and local Nusselt numbers. It was found
that as the natukl convection grows stronger with increasing Rayleigh number, the
most vigorous motions tend to be concentrated in the upper portion of the en-
closure, with an accompanying bottom-to-top stratifimrion At a given Rayleigh
number, the natural convection is stronger for smaller diameter ratios In the pre-
sence of vigorous natural convection, the ovemN inner-cylinder Nusselt number
corresponding to a given Raylelgh number is nearly independent of the geometric
parameters (diameter ratio, inner-cylinder elevation) in the investigated range. On
the other hand, rhe local Nusselt number distributions are more responsive to the
geomelry of the configuration

INTRODUCTION
During the past two decades, there has been a marked upsurge of interest in natural
convection in enclosed spaces. This interest and the research that it engendered have been
sparked by applications such as compact packaging o f electronic equipment, nuclear
reactor safety, shipping containers for spent nuclear fuel, cryogenic fuel tanks, energy
conservation (multipane windows), and alternative energy sources (solar collectors).
Globally speaking, the existent literature may be classified in two broad categories.
In one, the participating fluid is contained within an enveloping bounding surface, as
occurs in a hollow rectangular enclosure o r in a hollow horizontal cylinder. These con-
figurations are, in fact, the most commonly studied cases among externally bounded
enclosures. The state of knowledge with respect to such enclosures was reviewed by
Ostrach in 1972 [ I ] , and the substantial literature published since that time is identified
in the annual reviews of heat transfer literature published in the InternatiomlJournalof
Heat and Mass Transfer (e.g., [ 2 ]).
In the other category-the one of particular interest here-the participating fluid
is contained between an inner bounding surface and an outer bounding surface. In this
Copyright O 1982 by'tiemisphere Publishing Corporation
0149-5720/82/020119-26$2.75
I20 M. CHARMCHI AND E. M. SPARROW

NOMENCLATURE
heat transfer surface area temperature of outer cylinder
diameter of inner cylinder dimensionless velocities [Eq. (811
diameter of outer cylinder velocity components
acceleration due to gravity dimensionless axial coordinate
height of cylinder (= z/D,)
local heat transfer coefficient axial coordinate
average heat transfer coefficient axial coordinate of center of inner
t h e n a l conductivity cylinder
local inner-cylinder Nusselt number axial coordinate measured from bot-
local outer-cylinder Nusselt number tom of inner cylinder
average Nusselt number for cylinder in coefficient of thermal expansion
i n f ~ t domain
e dimensionless temperature
overall inner-cylinder Nusselt number I= ( T - To)I(Ti- T o ) ]
average Nusselt number for side surface viscosity
of inner cylinder kinematic viscosity
overall outer-cylinder Nusselt number density
dimensionless pressure [Eq. ( 9 ) ] density corresponding to To
Prandtl number dimensionless stream function
pressure (= J,lvDo)
reduced pressure (= p + P&) stream function
surface-integrated heat transfer rate
overall heat transfer rate
4 local heat flux Subscripts
R dimensionless radial coordinate
(= d o o ) b bottom surface
Rat inner-cylinder Rayleigh number i inner cylinder
[=gP(Ti - T O P ! Pdv' 1 o outer cylinder
r radial coordinate s side surface
Ti temperature of inner cylinder t top surface

category, the cases most often studied are t h e annulus between t w o horizontal cylinders
and the annulus between two spheres. Representative recent work for the intercylinder
annulus is described in [3] and [ 4 ] , while [5] and [6] convey information and addi-
tional references for the intersphere annulus.
With this general review as background, attention will now b e turned t o the physi-
cal situation t o be studied here.
The problem t o be analyzed is the natural convection flow and heat transfer in the
enclosed space between two vertical cylinders, with t h e special feature that the inner
cylinder is not only smaller in diameter b u t also o f lesser height than t h e outer cylinder.
A schematic diagram o f t h e problem is presented in the left-hand portion o f Fig. 1. T h e
figure shows a side view o f t h e two-cylinder assembly, with t h e plane of the figure coin-
ciding with a vertical plane that encompasses t h e diameters of t h e cylinders. As seen in
the figure, the two cylinders are concentric, with respective heights and diameters Hi, Di
and H o , D , (i = inner, o = outer). T h e elevation of the center o f t h e inner cylinder is
defined b y the coordinate z,, s o that t h e value of z, controls the relative sizes o f the
clearance spaces above and below the inner cylinder. T h e surface o f t h e inner cylinder
is maintained isothermal at T i , while t h e temperature o f the outer cylinder is uniform and
equal t o T o . For concreteness, Ti has been taken t o be larger than T o , but the results
also apply when To > T i . Owing t o the temperature difference, a recirculating flow is set
 CONVECHON IN THE SPACE BETWEEN CYLINDERS 121

up in the space between the cylinders. The general direction of the recirculation is up-
ward adjacent to the axis of the enclosure and downward along the outer wall of the
enclosure.
The work reported here constitutes the analytical portion of a two-part research
effort [7] that included both analysis and experiment. Ideally, it would have been pre-
ferable t o have presented the entire work in one paper. However, the analytical portion
alone, even with certain results omitted and referenced to [7], gives rise to a longer-than-
average paper, necessitating that the experimental work be presented elsewnere.
Examination of Fig. 1 indicates that the system is defined by five dimensions: Hi,
Di, Ho, D o , and z,. On a nondimensional basis, there are four geometric parameters:
Hi/Di,Ho/Do, Di/Do, and z c / D o . To these are to be added the Rayleigh number and the
Prandtl number, yielding a grand total of six parameters that must be prescribed for
each solution. To reduce this number to a level that is realistic for numerical computa-
tions, it was decided to focus attention on the case of unit aspect ratio for both the inner
and outer cylinders, that is,

Furthermore, the Prandtl number was fixed at 0.7 for all the solutions (air and other
gases). The three remaining parameters, along with their range of variation, are as follows:

where

The solutions were obtained numerically by means of an elliptic finite-difference

method. The results to be presented include local, per-surface, and overall heat transfer
rates and coefficients, streamline and isotherm maps, and velocity profiles. As noted
earlier, additional results that supplement those presented here are available in [ 7 ] .

i i
Fig. 1 Schematic diagram of investigated enclosure
122 M. CHARMCHI AND E. M. SPARROW

A survey of the literature failed t o disclose any prior analytical o r experimental

work for enclosures of the type considered here. Furthermore, there has been relatively
little study of the somewhat related problem of the vertical annulus (both inner and
outer cylinders of the same height). For that problem, numerical solutions corresponding
t o different uniform temperatures on the inner and outer cylinders and t o adiabatic end
faces are reported in [8-101. In [ I I ] , solutions were obtained for the case in which the
end faces are thermally active, while the cylindrical walls are adiabatic (the vertical
annulus was a special case of the inclined annulus treated in [I I]).

ANALYSIS AND SOLUTION

Governing Equations and Boundary Conditions
The first step in the analysis is to deduce the form of the buoyancy term that is
appropriate t o the problem. To begin, consideration may be given t o the terms
(-ap/az -pg) that appear in the z-momentum equation, t o which the terms p o g and
-pog may be appended. With these, and after regrouping, there is obtained

The quantity (p + p o p ) may be regarded as a reduced pressure and denoted by the sym-
bol p', while the density difference @, -p) is expressible via the Boussinesq equation
of state as

so that Eq. (5) becomes

Equation (7) is then substituted into the z-momentum equation.

The dimensionless form of the conservation equations will now be written. For
these equations, it will be assumed that axisymmetry prevails and that the thermophysical
properties are constant. The dimensionless variables and coordinates t o be used in these
equations are as follows:

so that
Mass conservation:
 CONVECTION IN THE SPACE BETWEEN CYLINDERS

Z momentum:

v -aa~v,
+v -av,
= - - + vap
az 2vz+-
az
R momentum:

Energy:

In these equations, V 2 denotes the Laplace operator in R , Z cylindrical coordinates.

From an examination of Eqs. (1 1)-(14), it is seen that Rai, Pr, and Do/Di appear
as parameters. Since R ~ , ( D , / D ~ ) is
~ equal t o the Rayleigh number Ra, based o n the
diameter of the outer cylinder, it would appear expeditious t o use it as a parameter
instead of separately specifying Rai and Do/Di. In actuality, however, Do/Di must already
be separately specified t o define the geometry of the problem. Furthermore, somewhat
greater emphasis will be placed o n the results for the inner cylinder than on those for the
outer cylinder, so that the use of Rai as a parameter appears more appropriate than the
use of Ra,.
For the statement of the boundary conditions, it is convenient to refer t o the right-
hand diagram of Fig. 1, where the solution domain is illustrated. With regard to the
velocity, both velocity components Vrand V, are zero along the solid surfaces ABCD and
A'B'C'D' that bound the intercylinder space, while Vr and aVz/aR are zero along the
symmetry line AD. For the temperature, 0 = 1 on the surface A'B'C'D' of the inner
cylinder, and 0 = 0 o n the surface ABCD of the outer cylinder. On the symmetry line,
a e l a ~= o.
Computational Aspects
Strictly speaking, the solution domain does not contain the region A'B'C'D'A',
which lies within the inner cylinder. This region is, in effect, an indentation into the regu-
lar cylindrical domain ABCDD'A'A. Many general-purpose computer programs, including
that employed here, are not equipped t o routinely handle such indentations. To avoid the
difficulty caused by the indentation, A'B'C'D'A' will be included in the solution domain,
so that the solution domain is the entire cylindrical region ABCDD'A'A.
T o suppress motion within A'B'C'D'A' and t o attain 0 = 1 therein, it is assumed
that A'B'C'D'A' is filled with a substance whose viscosity p and thermal conductivity k
are both infinite. To demonstrate the implementation of this approach, consider the dif-
ference equation for Vrwritten for a nodal point P with neighbors N, S, E, and W

The specification of the coefficient 4 as a very large number and the source nras zero
causes Vr t o be zero everywhere within A'B'C'D'A'. A similar approach is employed to
obtain V, = 0 in the region.
124 M. CHARMCHI AND E. M. SPARROW

For the energy equation, the difference equation may be written as

(aTO)P = + . . .+ + aT
Here, 4 and O T are set equal to the same large number at each point in A'B'C'D'A', so

-
that Eq. (16) reduces to Op = 1.
The use of k = p ~ 1 - in A'B1C'D2', while the k and p values are finite in the re-
mainder of the solution domain, leads to discontinuities in these quantities at surfaces
A'B', B'C', and C'D'. In [12] it was shown that in the case of nonuniform viscosity (or
conductivity), the momentum (or heat) flux across the face between two control volumes
can be accurately represented when the harmonic mean viscosity (or conductivity) is
used. If the respective viscosities of the fluid in the neighboring control volumes are p'
and p H ,the harmonic mean is given by

The harmonic mean viscosity and conductivity were included in the adopted solution
method [note that Eq. (17) reduces t o p = p' = pN when p' = p"] .
The numerical scheme used to obtain solutions is that of Patankar. Since the under-
lying basis of the method is described in a book-length exposition [13], little elaboration
is needed here. The method is based o n a control volume formulation. In order to set up
the difference equations, the governing differential equations are integrated over a typical
control volume, and simple algebraic representations are used to express the variations of
the functions between neighboring nodal points. The method works with the so-called
primitive variables (velocities, pressure, and temperature). A special feature is that the
grid points at which the velocity components are computed are displaced (i.e., staggered)
with respect to the grid points at which the pressure and temperature are computed.
A grid consisting of 3 0 X 3 2 points, respectively, for the R and Z directions was
employed for all the final computer runs. The grid-point layout was carefully tailored to
accommodate the relatively steep gradients adjacent t o the inner cylinder. Owing to the
changing nature of the flow field with increasing Rayleigh number, the tailoring opera-
tion, which involved a manual selection of the grid-point locations based on inspection
of velocity and temperature profile graphs, was performed at each Rayleigh number. At
higher Rayleigh numbers, the flow took on a distinct boundary-layer character adjacent
to the walls, especially at the side wall of the inner cylinder. This necessitated a closer
packing of the grid points in these regions.
For each configuration defined by the geometric parameters Di/Do and z,/Do, the
calculations were started at Rai = 0 , where conduction is the only mode of heat transfer.
Once the Rat = 0 solution had been obtained, the Rai = 10 case was undertaken, and so
on through Rai = l o a , lo3, l o 4 , and l o 5 . The solution for each case was iterative,
starting with an initial guess for the distributions of the dependent variables. Convergence
at the higher Rayleigh numbers was found to be slow. In order to speed up the high Rai
solutions, the converged solutions corresponding to the two preceding lower Rayleigh
numbers were extrapolated linearly to the higher Rayleigh number and were used for the
initial guess.
It was found that without under-relaxation of the momentum and energy equa-
 CONVECTION IN THE SPACE BETWEEN CYLINDERS 125

tions, the computational procedure would not converge. The employed relaxation factors
depended on the Rayleigh number. They were generally in the range 0.2-0.4 at the initia-
tion of the conlputations for any given case and were increased, whenever possible, as
the solution proceeded.

Heat Transfer and Other Presentation Parameters

The distributions of the local heat flux at the surfaces o f the inner and outer cylin-
ders were evaluated from Fourier's law q = -k(aT/an). The inner-cylinder heat flux will
be taken as positive in the direction from the surface into t h e fluid, while at the outer
cylinder positive heat flux will be defined as passing from the fluid t o the surface. Corre-
spondingly, for both cylinders, aT/an = aT/ar, aT/az, and -aT/az, respectively, at the
side, top, and bottom surfaces. In dimensionless form, Fourier's law becomes

The local Nusselt numbers for the inner and outer cylinders can then be expressed as

where it is understood that M/aN is a function of position on the respective surfaces of

the cylinders.
The surface-integrated heat transfer rate for each of the three surfaces of the respec-
tive cylinders will also be presented. For instance, for the side surface (subscript s) of the
inner cylinder

where z' is defined in Fig. 1. Since Hi = Difor the present solution and q i l Sis expressed
by Eq. (18),there follows

Qi. s

Similar expressions can be derived for the top and bottom surfaces of the inner cylinder,
Qi,, and Q i t b ,respectively, and for the three surfaces of the outer cylinder.
The overall heat transfer rate, denoted by Q x , is common t o both cylinders, so that
126 M. CHARMCHI AND E. M . SPARROW

With these, average heat transfer coefficients and average Nusselt numbers for the respec-
tive cylinders are evaluated from their definitions

- i i ~-
i h0 Do
Nui = - Nu, =-
k k

Patterns of fluid flow within the enclosure will be presented via streamline contours
and the corresponding temperature fields will be portrayed in isotherm maps. For these,
a dimensionless stream function and dimensionless temperature will be employed

Dimensionless velocity profiles will be plotted in the form V , versus R .

RESULTS AND DISCUSSION

To provide insights into the physical procesies that take place in the fluid situated
in the intercylinder space, the presentation of results will begin with the streamline and
isotherm maps. Supplementary flow field information will then be conveyed by velocity
profile plots. Attention is next turned to the heat transfer results, which will include the
surface-integrated heat transfer rates at the individual surfaces of the cylinders, the
average Nusselt numbers for each cylinder, and the local Nusselt numbers at the surfaces
of the inner cylinder.

Streamline Maps

Representative streamline maps, taken from a more extensive collection available in

[ 7 ] , are presented in Figs. 2-4. Each figure consists of four diagrams, with each diagram
depicting a side view of the two-cylinder assembly. Owing t o symmetry about the vertical
axis, only the region between the axis and the outer cylinder need be considered. The
right-hand boundary of each diagram is the symmetry axis, while the left-hand boundary
is the side wall of the outer cylinder. In each diagram, the inner cylinder is seen as a rec-
tangular region adjacent t o the symmetry line.
Figure 2 shows results for the inner cylinder positioned at midheight, i.e., z,/D, =
0.5. The two leftmost diagrams are for the smallest of the investigated diameter ratios
DJD, = 0.1, while the two rightmost diagrams are for the largest ratio DJD, = 0.3.
Within each pair for a f x e d diameter ratio, the left-hand and right-hand diagrams give
results for Rai = 10 and l o 5 , respectively. Results similar t o those of Fig. 2 but for
DJD, = 0.2 and for Rat = 103 may be found in [ 7 ] .
From an overview of Fig. 2, it is seen that the main feature of the flow field is a
general recirculation that is downward-directed adjacent t o the outer cylinder and upward-
directed adjacent t o the symmetry axis of the enclosure. The numerical values of the
dimensionless stream function q provide a direct index of the velocities in the recircu-
 CONVECTION IN THE SPACE BETWEEN CYLINDERS 127

Fig. 2 Streamline maps for the inner cylinder positioned at midheight. The leftmost pair of
diagrams is for Di/Do = 0.1, and the righ!most pair of diagrams is for DJD0 = 0.3. Within each
pair, the left and right diagrams correspond to Rai = 10 and l o s , respectively.

lating flow. The q values for Rai = 10 are substantially smaller than those for Rai = l o 5 ,
indicating an increasingly stronger recirculation as the Rayleigh number increases.
For very low Rayleigh numbers, where conduction is virtually the sole mechanism
of heat transfer, top-to-bottom symmetry about the horizontal line z/Do = 0.5 is to be
expected when the inner cylinder is at midheight. The Rai = 10 flow field for the larger
inner cylinder (Di/Do = 0.3) is, indeed, virtually symmetric. However, in the case of the
smaller inner cylinder (Di/Do = 0.1), the streamline pattern for the Rai= 10 case is
slightly unsymmetric, with the eye of the recirculating flow displaced upward. This indi-
cates that natural convection is already playing a role in the heat transfer process. In
general, at a given Rai the strength of the natural convection flow is greater for small
Di/Do than for large Di/Do, as witnessed both by the shape of the streamline pattern and
by the decreasing magnitude of the stream function q at larger Di/Do.
As Ra, increases, the eye of the recirculating flow is displaced upward. Along with
this, a zone of vigorous fluid motion tends to develop in the upper portion of the en-
closure, while in the lower portion the fluid motions are generally sluggish. These dif-
ferences in the vigor of the flow are evident from a comparison of the values in the
upper and lower parts of the enclosure. The kidney-shaped eye and the nearly horizontal
streamline lay above and below the eye, in evidence for Rai = l o 5 and Di/Do = 0.1, are
characteristic of fully developed natural convection in the intercylinder space.
The horizontal flow below the eye is induced by the demands of the boundary
layer at the side of the inner cylinder and the plume that billows up above the cylinder.
Both the boundary layer and the plume call for continuous infusions of fluid, and to ful-
fill this need they draw fluid inward from the enclosure proper. This process intercepts
fluid that might otherwise find its way t o the bottom portion of the enclosure. In the
bottom region, fluid is drawn toward the lower surface of the inner cylinder and then
flows radially outward along that surface.
As a final note with regard to Fig. 2, attention may be called to the small recircula-
tion zone situated atop the inner cylinder in the rightmost diagram. This is a low-velocity
region, as indicated by the small value of q associated with the recirculation eddy. A
128 M. CHARMCHI AND E. M. SPARROW

similar eddy may have existed for the case DJD, = 0.1, Rai = l o 5 , but it could not be
detected with the available number of grid points.
Attention wiU now be turned t o the effect of inner-cylinder elevation on the flow
pattern in the enclosure, and Figs. 3 and 4 have been prepared in this regard. The respec-
tive figures pertain t o Di/Do = 0.1 and 0.3. In each figure there are four diagrams. The
two leftmost diagrams are for Rat = l o 3 and show results for inner-cylinder elevations
z,/Do = 0.25 and 0.75, while the two rightmost diagrams show results for the same
elevations but for Rai = 10'.
A superficial examination of Fig. 3 suggests that for a given Rai the overall shape
of the flow field is not greatly affected by the upward shift of the inner cylinder from
zJD, = 0.25 t o 0.75. There are, however, various aspects of the flow pattern that
differ for the two elevations. First, when the inner cylinder is positioned low in the
enclosure, it draws fluid downward to it, so that there is a vigorous flow in the lower
reaches of the enclosure. Consequently, the vigor of the natural convection flow is more
uniform throughout the enclosure at low inner-cylinder elevations than at high inner-
cylinder elevations. This characteristic can be verified by examining the spatial distribu-
tions of the *values for the two elevations (at the same Rai).
Such an inspection also shows that higher mass flows (i.e., higher 9 vvalues) are
attained for lower inner-cylinder elevations. This behavior occurs because for low eleva-
tions the plume that rises from the inner cylinder is very long, and consequently the fluid
in the plume has a greater length of run in which to accelerate. The demands of the long,
accelerating plume for infusions of fluid are clearly evidenced for Rai = l o s by the hori-
zontal streamlines that lead into the plume.
The effects of inner-cylinder elevation that were just described for Di/Do = 0.1 are
also applicable t o the DJD, = 0.3 case, as can be verified by examination of Fig. 4. As
before, more uniform velocities prevail for low inner-cylinder elevations, and the mass
flow rate is higher. Furthermore, since the inner cylinder intrudes more deeply into the
enclosure at DJD, = 0.3 than at Di/Do = 0.1, the effect of elevation changes on the
overall streamline pattern is more in evidence in Fig. 4 than in Fig. 3.

Fig. 3 Effect of inner-cylinder elevation on the streamline pattern for Di/Do = 0.1. The leftmost
pair of diagrams is for Rai = lo3, and the rightmost pair of diagrams is for Rai = 10'. Within
each pair, the left and right diagrams correspond to zdD, = 0.25 and 0.75, respectively.
 CONVECTION IN THE SPACE BETWEEN CYLINDERS 129

F
ig.4 Effect of inner-cylinder elevation on the streamline pattern for Di/Do = 0.3. The left-
most pair of diagrams is for Rai= lo3,and the rightmost pair of diagrams is for Rai= 10'.
Within each pair, the left and right diagrams correspond t o z d D o = 0.25 and 0.75,respectively.

Especially noteworthy in Fig. 4 are marked elevation-related differences in the

approach flow beneath the inner cylinder and in the flow above the cylinder. At low
elevations, the main recirculating flow experiences difficulty in penetrating the narrow
gap beneath the cylinder. Similarly, at high elevations, the main flow is unable to pene-
trate the gap above the cylinder. Rather, the space above the cylinder is filled with a
local recirculation zone, the outer leg of which is an upflow and the inner leg of which
is a downflow.
On the basis of the foregoing observations, it might be expected that the heat
transfer results for Di/Do = 0.3 would be more sensitive to inner-cylinder elevation
than would be the results for Di/Do = 0.1. This issue will be revisited when the heat
transfer results are presented.

Isotherm Maps
Temperature field information, conveyed by isotherm maps, will now be presented
for representative cases ([7] contains a more complete set of isotherm maps). This infor-
mation is displayed in Figs. 5-7. The format of these figures is identical to that of Figs.
2-4, the difference between the two sets of figures being that 2-4 are for the dimension-
*
less stream function while 5-7 are for the dimensionless temperature 0.
In Fig. 5, attention is focused on the situation where the inner cylinder is at mid-
height. The leftmost pair of diagrams is for Di/Do = 0.1 with Rai = 0 (pure conduction)
and Rat = l o 5 , respectively. The rightmost pair is for Di/Do = 0.3, where the first dia-
gram is for both Rai = 0 and 10 and the second diagram is for Rai = l o 5 . In interpreting
these diagrams, it is helpful to recall that 0 = 1 on the surface of the inner cylinder and
0 = 0 on the surface of the outer cylinder.
By examining Fig. 5, it is seen that the conduction cases are characterized by con-
tour lines that are symmetric from top to bottom and are, in general form, flattened
circles. On the other hand, at high Rayleigh numbers the temperature field is charac-
terized by stratification, with the entire lower part of the enclosure filled with fluid
130 M. CHARMCHI AND E. M. SPARROW

Fig. 5 Isotherm maps for the inner cylinder positioned at midheight The leftmost pair of dia-
grams is for Di/Do = 0.1, and the rightmost pair of diagrams is for DtID, = 0.3. W~thineach
pair, the left and right diagrams correspond to Rat = 0 (pure conduction) and lo4,respectively.

whose temperature is very near near that of the outer cylinder. Above the nearly iso-
thermal low-temperature zone, the fluid temperature increases with height.
In addition to the stratification, another noteworthy feature of the high-Rayleigh-
number results is the presence of thermal boundary layers whose characteristic signature
is a crowding together of the isotherms adjacent t o a surface. The greater the crowding
of the isotherms, the higher the heat transfer rate. Boundary layers are seen to be present
adjacent to the side and lower walls of the inner cylinder and adjacent to the top wall
and the top-adjacent side wall of the outer cylinder. Yet another feature of the high-
Rayleigh-number results is the concentration of relatively hot fluid in the space above
the inner cylinder.
The degree of stratification is greater for the DdD, = 0.1 case than for the Di/Do =
0.3 case, and thinner boundary layers are also in evidence for the former. These compari-
sons affirm the presence of more vigorous natural convection at a futed Rat when the
diameter of the inner cylinder is small.
Figures 6 and 7 deal with the effect of inner-cylinder elevation on the temperature
field in the fluid. The isotherm maps presented in Fig. 6 are for the DJD, = 0.1 case,
where the two leftmost diagrams are for Rat = lo3 and the two rightmost diagrams for
Rat = 10" Each pair of diagrams shows results for zc/Do = 0.25 and 0.75. A similar
presentation is made in Fig. 7 for DJD, = 0.3.
From Fig. 6, it is seen that when the inner cylinder is positioned at a low elevation,
the vertical extent of the low-temperature zone adjacent to the bottom of the enclosure
is diminished. On the other hand, with the cylinder positioned in the upper portion of the
enclosure, the low-temperature zone occupies more and more of the total height of the
enclosure. Above the low-temperature zone, the fluid temperature increases, signaling the
presence of stratification. The horizontal lay of the isotherms is accentuated at the higher
cylinder elevation (and, as noted earlier, at higher Rayleigh numbers).
In Fig. 6, careful inspection of the isotherm patterns adjacent to the surface of the
inner cylinder indicates only minor changes with zc/Do at a given Ra,. Thus, at least for
DJD, = 0.1, the heat transfer rates at the individual surfaces of the inner cylinder should
be relatively insensitive to zc/Do.
 CONVECXION IN THE SPACE BETIYEEN CYLINDERS 131

Fig. 6 Effect of inner-cylinder elevation on the isotherm pattern for Di/Do = 0.1. The leftmost
pair of diagrams is for Raj= lo', and the rightmost pair of diagrams is for Raj= 10'. Within
each pair, the left and right diagrams correspond t o rdDo = 0.25 and 0.75,respectively.

The trends in evidence in Fig. 7 will now be discussed. For inner-cylinder diameters
such as Di/Do = 0.3, conduction continues to play an important role at Rai = 10" so
that stratification is not as strongly in evidence as at smaller Di/Do or at larger Rai. The
upward displacement, of the inner cylinder from zc/Do = 0.25 to 0.75 shifts the
conduction-affected isotherm lay from the gap beneath the inner cylinder to the gap
above the inner cylinder.
The Rat = lo5 results in Fig. 7 bear the stamp of natural convection, with stratifi-
cation clearly in evidence and trends with zc/Do similar to those already discussed in
connection with Fig. 6. Of particular note is the isotherm pattern in the gap above the
inner cylinder for the high-elevation case, which is unique among those displayed in
Figs. 5-7. This unique isotherm lay is the result of the strong local recirculation zone
in evidence in the right-hand diagram of Fig. 4.

Fig. 7 Effect of inner-cylinder elevation on the isotherm pattern for Dj/Do = 0.3. The leftmost
pair of diagrams is for Rat= lo3,and the rightmost pair of diagrams is for Rai= l o 5 . Within
each pair, the left and right diagrams correspond to zc/Do = 0.25 and 0.75,respectively.
132 M. CHARMCHI AND E. M. SPARROW

Velocity Profiles
To provide additional insights about the patterns of fluid flow, representative pro-
files of the axial velocity will now be presented. This information is displayed in Figs. 8
and 9, respectively, for Di/Do = 0.1 and 0.3. Each figure contains two graphs, one for
Rat = 10 and the other for Rai = 10'. For all cases shown in these figures, the cylinder
is at the midheight position z,/Do = 0.5.
Each graph displays velocity profiles at five different axial stations z/Do. These
correspond, respectively, to a station in the lower region of the enclosure, one just below
the bottom of the inner cylinder, one at the midheight of the imer cylinder, one just
above the top of the inner cylinder, and one in the upper region of the enclosure.
Attention will first be directed to Fig. 8. At Rai= 10, the velocity profiles are
rounded curves that consist of an upflow (i.e., positive) lobe that extends out t o r/Do E
0.27 and a downflow (i.e., negative) lobe that spans the remainder of the cross section.
In sharp contrast to these, the profiles for Rai = l o 5 are of the boundary-layer type,
with all of the action confined to narrow zones adjacent to the bounding surfaces and
the symmetry line. Over most of the cross section, the axial velocity is negligibly small,
that is, the velocity outside the boundary layer is zero.
By comparing the ordinates of the left- and right-hand graphs of Fig. 8, it is seen
that the increase in Rayleigh number is accompanied by an enormous increase in the
magnitude of the velocities in the boundary-adjacent regions. Other changes that ac-
company the Rayleigh number increase are the tendencies toward relatively low velocities
in the lower reaches of the enclosure (undetectably small at Rai = l o s ) and toward rela-
tively high velocities in the region above the inner cylinder.
The velocity profiles for Di/Do = 0.3 presented in Fig. 9 confirm the trend toward

Fig. 8 Profiles of the axial velocity for Di/Do = 0.1 and the inner cylinder positioned at
midheight.
 CONVECTION IN THE SPACE BETWEEN CYLINDERS

-- 1m. -- mI S

-- ZlD.
- L 1-

--
zm. L sm
zm. Pam
I n . m a

Fig. 9 Profiles of the axial velocity for DJDo = 0.3 and the inner cylinder positioned at mid-
height.

boundary-layer type flow with increasing Rayleigh number. However, fully established
boundary-layer flow is not yet attained at Rai= 10' when Di/Do = 0.3. This finding,
taken together with the smaller V, magnitudes in Fig. 9 (relative to those of Fig. 8),
offers further testimony to the lesser vigor of the natural convection flow as Di/Do in-
creases at a futed Rai.
Another manifestation of this characteristic can be seen in the left-hand diagram
of Fig. 9, which pertains to Rai = 10. There, the profiles at the five axial stations blend
together into three curves. Whereas the profile for z/Do = 0.5 maintains a separate iden-
tity, the curves for z/Do = 0.150 and 0.850 are coincident, as are those for z/Do = 0.333
and 0.667. This top-to-bottom symmetry indicates that conduction holds full sway. In
contrast, the velocity profiles for Rai = 10 in Fig. 8 (Di/Do = 0.1) do not display top-to-
bottom symmetry, demonstrating that natural convection is already significant.

Per-Surface and Overall

Heat Transfer
The surface-integrated heat transfer rates for each of the surfaces of the respective
cylinders are listed in Table 1 along with the overall rate of heat transfer between the
cylinders. The table is subdivided into parts (a), (b), and (c), respectively, for Di/DO =
0.1,0.2, and 0.3. In turn, each part is further subdivided into three sectionsparameterized
by zJD0 = 0.25, 0.5, and 0.75. In each section, the respective columns give the Rayleigh
number Rai, the overall rate of heat transfer (common t o both cylinders), the per-surface
heat transfer rates for the bottom, side, and top (b, s, and t) of the inner cylinder, and the
corresponding per-surface rates for the outer cylinder.
 Table 1 Per-Surface and Overall Heat Transfer Rates, Q/21rkDo(TI - T o )

Inner cylinder Outer cylinder

z,./Do Rai Overall b s t b s t

(a) Di/Do = 0.1

 CONVECTION IN THE SPACE BETWEEN CYLINDERS 135

The heat transfer rates listed in Table 1 are dimensionless and are given by

It should be noted that the factor 2nkDo(Ti - To) is common to all the entries in the
table so that a comparison of the dimensionless values is equivalent to a comparison of
the actual heat transfer rates.
The information provided by Table 1 can be employed, if desired, to compute Nus-
selt numbers. For example, for the side of the inner cylinder

Since Ai,, = nDIHi and Hi = Di for the present solutions, the quantity in the final paren-
theses of Eq. (28) equals 2D0/Di. The overall Nusselt numbers for the respective cylin-
ders, Nui and G o , can be obtained by using Table 1 in conjunction with Eqs. (24) and
(25). Shortly, Kiwill be tabulated and its main characteristics identified.
The main trends exhibited by the results of Table 1 will now be identified. In
interpreting the results, it is helpful to recall that the surface area of the side of either
cylinder is four times the surface area of the top or the bottom of that cylinder. Atten-
tion is first focused on the inner cylinder. It is seen that both the side- and bottom-
surface heat transfer rates respond in a regular manner to the Rayleigh number, increasing
as the Rayleigh number increases. The top-surface heat transfer, however, shows a mixed
trend. It tends to decrease moderately with increasing Rayleigh number at the smaller
Rayleigh numbers but increases sharply at the larger Rayleigh numbers. With a few
exceptions (at low Rai for the highest elevation and at Rai = 0 for the middle elevation),
the bottom-surface heat transfer exceeds the top-surface heat transfer. Furthermore, in
general terms, the side-surface heat transfer is in the range of four times the bottom-
surface heat transfer (the area ratio is 4).
For the smallest of the inner cylinders (Di/Do = 0.1), the heat transfer results
are quite insensitive to cylinder elevation, and the results for Di/Do = 0.2 display a
similar characteristic. However, greater sensitivity to elevation is exhibited by the results
for the largest of the inner cylinders (Di/Do = 0.3). For this case, the bottom-surface heat
transfer decreases markedly with elevation at the smaller Rayleigh numbers, while the
upper-surface heat transfer is sensitive to elevation at all Rayleigh numbers-but particu-
larly at the smaller Rar values. The side-surface heat transfer for Di/Do = 0.3 is only
moderately affected by cylinder elevation.
Attention will now be turned to the results given in Table I for the outer cylinder.
Here, the top-surface heat transfer increases monotonically with the Rayleigh number
while the bottom-surface heat transfer decreases. The side-surface results show nonmono-
tonic trends, but overall, from the lowest t o the highest Rayleigh numbers, there is a
marked increase.
Another noteworthy feature is the high rate of heat transfer at the top surface of
the outer cylinder at intermediate and high Rayleigh numbers, for all elevations of the
inner cylinder. This high rate results from the impingement of the upwelling stream of
heated fluid on that surface. There is also appreciable heat transfer at the top surface at
136 M . CHARMCHI AND E. M. SPARROW

low Rayleigh numbers when the inner cylinder is at its highest elevation. In contrast to
this, the bottom surface experiences relatively low rates of heat transfer except when the
inner cylinder is at its lowest elevation and when the Rayleigh number is low.
In general, the per-surface heat transfer results for the outer cylinder are more
sensitive to the elevation of the inner cylinder than are the inner-cylinder heat transfer
results. Furthermore, there is no simple numerical relationship, such as the area ratio,
between the per-surface heat transfer rates.at the respective surfaces of the outer cylinder.
It is relevant to consider possible comparisons between the results of Table 1 and
the literature. As noted in the Introduction, a literature search did not reveal any experi-
mental or analytical work for the configuration studied here. There is, however, a sub-
stantial literature on boundary-layer type natural convection at the outer surface of a
vertical cylinder situated in an infinite domain. On the basis of physical reasoning, it
may be expected that the vertical cylinder heat transfer characteristics will be approached
by the present results for the side surface of the inner cylinder as Di/Do + 0.
To examine the validity of this expectation, use will be made of a correlation [14]
of the available experimental data for the average Nusselt number Kcyl for a vertical
cylinder. The correlation of [14] was evaluated for the Rayleigh number range of interest
here, and the numerical values are listed in the second column of Table 2. The table also
lists the present average Nusselt numbers GLSfor the side surface of the inner cylinder.
From the table, it is seen that the Nui,, results for Di/Do = 0.1 are very close
(within 24%) to Kcyl when Rai> lo3. At lower Rayleigh numbers, heat conduction
effects that occur in the present enclosure configuration cause deviations between Kivs
and Kc,l. It can also be seen from the table that the general level of agreement between
Kit,and Nucyl diminishes as Di/Do increases.

Table 2 Comparison of & s with GcYl

from Vertical Cylinder Experiments
-
Nui,, for zJD0 =
 CONVECTION IN THE SPACE BETWEEN CYLINDERS

Table 3 Overall InnerCylinder Nusselt Numbers Ki

Overall Nusselt Numbers

It has already been pointed out that overall Nusselt numbers for each cylinder can
be evaluated by using the information listed in the third column of Table 1. Since

- Di -
Nu, = - Nui
Do

from Eqs. (24) and (25), it is sufficient to focus attention on the inner cylinder Nusselt
number Ki.
The numerical values of Giare listed in Table 3. As seen there, at any fued Ray-
leigh number and for DJD, = 0.1, Nul is remarkably insensitive to the elevation of the
inner cylinder over the range investigated. For DJD, = 0.2, there is a modest effect of
zc/Do at low Rayleigh numbers, but at larger Rayleigh numbers there is only a very weak
dependence of on zc/Do. This same behavior is also in evidence for Di/D, = 0.3, but
the insensitivity sets in at a higher Rayleigh number.
It can also be seen that the variation of mi with DJD, at fued values of Rar and
z,/D, generally diminishes with increasing Rayleigh number. Thus, in the presence of
vigorous natural convection (i.e., high Rayleigh number), Ki is very weakly dependent
on both of the geometric parameters Di/Do and zc/Do.
To provide a further perspective on these results, Siis plotted in Fig. 10 as a
function of Rai for each of the three investigated diameter ratios. The figure pertains to
the inner cylinder situated at the midheight position z,/D, = 0.5.
Figure 10 shows that for sufficiently small values of the Rayleigh number, the
Nusselt number for each case approaches a constant value-that for pure conduction. In
138 M. CHARMCHI AND E. M. SPARROW

Fig. 10 Overall innercylinder Nusselt number results corresponding to the inner cylin-
der at the midheight position.

the conduction-dominated regime, %i increases with Di/Do. The onset of significant

convective effects is seen to occur at larger Rayleigh numbers as Di/Do increases, which
confirms the conclusions drawn from the streamline and isotherm maps. In the range of
Rayleigh numbers where convection dominates, the Nusselt number decreases with
increasing diameter ratio.
A concerted effort to obtain algebraic correlations for Ei was made in [7]. The
objective of the work was to correlate Ei with Rai, with account being taken of the
effects of DJD, and of the inner-cylinder elevation zc/Do. For Rai > l o 3 , a power law
of the form & = CRal was employed. For Di/Do = 0.1, the values C = 0.850 and
n = 0.204 yielded an algebraic representation valid to within -1.0 and +0.8% for all
zc/Do between 0.25 and 0.75. This same n and C = 0.820 yielded a correlation for
DJD, = 0.2 valid over the same range of zc/Do and accurate to within - 1.5 and + 1.8%.
For D,/D, = 0.3, a single power-law representation is not able to correlate the results
for all zc/Do between 0.25 and 0.75. Specific power-law equations for the various in-
vestigated zc/Do are available in [7, p. 881 .
To span the entire range of Rayleigh numbers between 0 and l o s , the correlation
procedure of Churchill and Usagi [15] was employed. Excellent success was achieved
for Di/Do = 0.1, where the equation

represented the computed results to within k 1.6% for 0.25 < zc/Do < 0.75. For Di/Do =
0.2, two equations, each having a form similar to Eq. (30), had to be used, one for
z,/Do = 0.5 and another for zc/Do = 0.25 and 0.75, and a similar state of affairs holds
for DJD, = 0.3. These correlating equations are given in [7, pp. 86-90] .

Local Nusselt Number Distributions

A representative sampling of the available distributions of the local Nusselt number

will now be presented in Figs. 11-13. These figures pertain to the inner cylinder. Addi-
 CONVECHON IN THE SPACE BETWEEN CYLINDERS 139

tional results for the inner cylinder may be found in [7], as are a set of distribution
curves for the outer cylinder.
Figure 11 corresponds to the smallest diameter ratio Di/Do = 0.1 and to the inner
cylinder at midheight, zc/Do = 0.5. The figure contains two columns of graphs, with
three graphs in each column. The graphs of the left-hand column convey the distribution
curves for the top and bottom surfaces ( t and b, respectively), while the graphs of the
right-hand column are for the side surface (s). Note that the z' coordinate measures
distances upward from the lower edge of the inner cylinder. In addition to the local distri-
butions, there is a dashed line in each graph corresponding to G i , the overall Nusselt
number for the inner cylinder as a whole. The uppermost graph in each column is for
Rai = 10, the middle pair of graphs is for Rai = lo3, and the lower pair is for Rai = lo5.
In the uppermost pair of graphs, it is seen that despite the low value of Rai the
bottom- and top-surface distribution curves are not coincident and that the side-surface
distribution curve is not symmetric about the midheight position z1/D0 = 0.05. These
characteristics indicate that natural convection is already operative.
At any Rayleigh number, the bottom-surface distribution curve falls above that for
the top surface, consistent with the impingement of relatively cool fluid on the former
while the latter is washed by fluid that has already been heated by contact with the side
surface of the cylinder. The shapes of the distribution curves for the top and bottom
surfaces are generally the same for all the cases illustrated, except that for the top at
Rai = l o 5 . In that instance, the strong recirculation zone in the upper part of the en-
closure alters the flow pattern above the inner cylinder. With increasing Rayleigh number,
the local Nusselt numbers for the top and bottom surfaces lie progressively lower with

Fig. 11 Distributions of the local innercylinder

Nusselt number for Di/Do = 0.1 and the inner cylin-
der positioned at midheight.
140 M. CHARMCHI AND E. M. SPARROW '

respect to the overall Nusselt number &, with the opposite behavior for the local Nus-
selt numbers on the side surface.
With regard to the side-surface Nusselt number distribution, it develops a definite
boundary-layer character as the Rayleigh number increases. Very high Nusselt numbers
occur near the lower edge of the surface and decrease along the flow direction (i.e., in-
creasing 2 ' ) . This decrease is interrupted and reversed as the boundary layer is pushed
toward the surface by the recirculation eddy that holds sway in the upper portion of
the enclosure.
Figure 12 has a structure similar to that of Fig. 11 but corresponds to the largest
of the investigated diameter ratios Di/Do = 0.3, rather than to Di/Do = 0.1. The inner-
cylinder elevation remains at zJD0 = 0.5, as in Fig. 11. For Rai = 10 (uppermost pair
of graphs), the distribution curves for the bottom and top surfaces are coincident, and
the side-surface distribution curve is symmetric about the midheight point z'/Do = 0.15.
These characteristics signal the total dominance of heat conduction. As the Rayleigh
number increases, the distribution curves evolve in a manner similar to that of Fig. 11,
but there are some differences in detail. In particular, for all Rayleigh numbers the
bottom-surface Nusselt numbers are generally of the same magnitude as the overall
Nusselt number Gi.In addition, the Nusselt numbers in the central region of the top
surface remain relatively low, indicating a minimal contribution of natural convection.
The last of the figures, Fig. 13, provides a glimpse into the effect of inner-cylinder

Fig. 12 Distributions of the local inner-cylinder

Nusselt number for Di/Do = 0.3 and the inner cylin-
der positioned at midheight.
 CONVECTION IN THE SPACE BETWEEN CYLINDERS

Fig. 13 Effect of innercylinder elevation on the

distribution of the local inner-cylinder Nusselt num-
ber; D$D0 = 0.3 and Rai = l o 5 .

elevation on the local Nusselt number distributions. The figure corresponds to the diam-
eter ratio Di/Do = 0.3 and to a Rayleigh number Rai= l o 5 . Results for elevations
zc/D, = 0.25 and 0.75 are presented in the lower and upper graphs, respectively.
The distribution curves for zc/Do = 0.25 are quite similar to those already pre-
sented for zc/Do = 0.5 in the lowermost graphs of Fig. 12, and no further elaboration
is needed. On the other hand, the results for zc/Do = 0.75 merit elucidation. At the top
surface, the distribution curve shows that the Nusselt number decreases from the center
to the edge, which is just opposite to the trend displayed in all the other graphs thus far
presented. This novel result is caused by the local recirculation zone situated above the
top surface of the cylinder. As illustrated earlier, the recirculation carries fluid radially
outward along the top of the cylinder, giving rise to a thickening boundary layer and to
decreasing heat transfer coefficients.
Furthermore, the radial outflow prevents the main recirculation zone, which fils
the upper part of the enclosure, from driving the side-surface-adjacent streamlines into
the space above the cylinder. As a consequence, the acceleration of the flow adjacent to
the upper portion of the side surface, which occurred in the other cases dominated by
natural convection, does not occur here. As a result of the absence of this acceleration,
the side-surface heat transfer coefficients do not increase near the upper end of the
surface, as occurred in the other cases.

CONCLUDING REMARKS
Numerical solutions were obtained for natural convection fluid flow and heat
transfer in the space between two vertical, concentric cylinders of fmite length. The
142 M. CHARMCHI AND E. M. SPARROW

special feature of the investigated configuration is that the inner cylinder is not only
smaller in diameter than the outer cylinder but also of lesser height. The numerical work
was focused on the case in which both cylinders are of unit aspect ratio (height = diam-
eter). Three parameters were varied during the course of the computations. These include
the diameter ratio Di/Do, the inner-cylinder elevation z,/Do, and the inner-cylinder
Rayleigh number Rai. The Prandtl number was futed at 0.7 for all cases.
Streamline maps were presented to display the nature of the natural convection
flow field. At low Rayleigh numbers, the fluid motions are relatively feeble. As the Ray-
leigh number increases, a zone of vigorous fluid motion develops in the upper portion of
the enclosure, while in the lower portion the fluid motions are generally sluggish, espe-
cially when the inner cylinder is positioned at intermediate and high elevations. Also, at
high Rayleigh numbers, a local recirculation zone tends to develop above the upper sur-
face of the inner cylinder.
At a given Rayleigh number, the vigor of the natural convection is greater for small
Di/Do than for large Di/Do. On the other hand, the fluid flow is more affected by changes
of inner-cylinder elevation when Di/Do is large.
When natural convection is significant, the temperature field is characterized by a
bottom-to-top stratification, with the lower part of the enclosure filled with fluid whose
temperature is very near that of the outer cylinder. With the inner cylinder positioned at
a low elevation, the vertical extent of the low-temperature zone is diminished. At any
ftved Rayleigh number, the degree of stratification is greater for smaller Di/Do than for
larger Di/Do.
Representative cross-sectional distributions of the axial velocity were presented to
highlight the major changes in the flow field that occur with increasing Rayleigh number.
At low Rayleigh numbers, the velocity profiles are rounded curves consisting of upflow
and downflow lobes that, taken together, span the entire cross section of the enclosure.
On the other hand, at high Rayleigh numbers, the profdes are of the boundary-layer type,
with all of the action confined to narrow zones adjacent to the bounding surfaces and the
symmetry line.
An extensive tabulation was presented for the per-surface heat transfer rates for
both the inner and outer cylinders and for the overall rate of heat transfer. For the inner
cylinder, both the side- and bottom-surface heat transfer rates increase monotonically
with increasing Rayleigh number. The top-surface heat transfer decreases at the smaller
Rayleigh numbers and then increases sharply at the larger Rayleigh numbers. With a few
exceptions, the bottom-surface heat transfer rates exceed those for the top surface. For
the smallest of the inner cylinders (Di/Do = 0.1), the heat transfer results are insensitive
to the cylinder elevation, but the sensitivity becomes more marked for the largest cylin-
der.
With regard to the outer cylinder, the top-surface heat transfer increases mono-
tonically with the Rayleigh number, while the bottom-surface heat transfer decreases.
The side-surface results are not monotonic but, overall, the heat transfer increases from
the lowest to the highest Rayleigh number. In general, the per-surface heat transfer results
for the outer cylinder are more sensitive to the elevation of the inner cylinder than are
the inner-cylinder heat transfer results.
The side-surface Nusselt numbers for the inner cylinder were compared with avail-
able experimental results for boundary-layer type natural convection at the outer surface
of a vertical cylinder situated in an i n f ~ t domain.
e For Rai > lo3 and Di/Do = 0.1, the
side-surface results were within 24 %of those for the vertical cylinder.
 CONVECTION IN THE SPACE BETWEEN CYLINDERS 143

Overall Nusselt numbers K i and Gofor the inner and outer cylinders were evalu-
ated. Since Go= (Di/Do)Ki,it was necessary only t o deal with K i .For sufficiently
small values of the Rayleigh number, approaches the pure conduction value that
corresponds t o the given values of DJD, and zc/Do. The onset of significant convective
effects occurs at larger Rayleigh numbers as Di/Do increases. Algebraic correlation equa-
tions were developed for characterizing Si.
For the smallest diameter ratio, Ki was found t o be virtually independent o f
cylinder elevation over the range investigated. Similar insensitivity is in evidence at the
other diameter ratios when the Rayleigh number is sufficiently large. Indeed, at high
Rayleigh numbers (i.e., vigorous natural convection), Ki is very weakly dependent on
both of the geometric parameters DJD, and zc/Do.
A presentation of local Nusselt numbers for the inner cylinder showed that for
most cases there is a radially outward increase along the top and bottom surfaces. An
interesting exception occurs when a strong local recirculation zone is situated above the
t o p of the inner cylinder; in this instance, the top-surface Nusselt numbers decrease in
the radial outward direction. The side-surface Nusselt numbers decrease in the upward
direction, but in most cases the decrease is interrupted and reversed near the upper end
o f the surface.

REFERENCES
1. S. Ostrach, Natural Convection in Enclosures, Adv. Heat Transfer, vol. 8, pp. 161-
227, 1972.
2. E. R. G. Eckert, Heat Transfer-A Review of 1979 Literature, Int. J. Heat Mass
Transfer, vol. 24, pp. 1-34, 1981.
3 . T. H. Kuehn and R. J . Goldstein, A Parametric Study of Prandtl Number and Diam-
eter Ratio Effects on Natural Convection Heat Transfer in Horizontal Cylindrical
Annuli, J. Heat Transfer, vol. 102, pp. 768-770, 1980.
4. M. C. Charrier-Mojtabi, A. Mojtabi, and J. P. Caltagirone, Numerical Solution of a
Flow due t o Natural Convection in Horizontal Cylindrical Annulus, J. Heat Transfer,
vol. l O l , p p . 171-173, 1979.
5. R. E. Powe, R. C. Baughman, J. A. Scanlan, and J. T. Teng, Free Convection Flow
Patterns between a Body and Its Spherical Enclosure, J. Heat Transfer, vol. 97, pp.
296-298, 1975.
6. S. H. Yin, R. E. Powe, J. A. Scanlan, and E. H. Bishop, Natural Convection Flow
Patterns in Spherical Annuli, Int. J. Heat Mass Transfer, vol. 16, pp. 1785-1795,
1973.
7. M. Charmchi, Experimental and Analytical Study of Natural Convection in the En-
closed Space between Finite Length Cylinders, Ph.D. thesis, Dept. of Mechanical
Engineering, Univ. of Minnesota, Minneapolis, Minn., 198 1.
8. G. de Vahl Davis and R. W. Thomas, Natural Convection between Concentric Verti-
cal Cylinders, Phys. Fluids, vol. 12, suppl. 11, pp. 11-198-11-207, 1969.
9. R. W. Thomas and G. de Vahl Davis, Natural Convection in Annular and Rectangular
Cavities-A Numerical Study, Proc. 4th Int. Heat Transfer Conf., vol. 4 , paper NC2.4,
1970.
10. T. H. Schwab and K. J. DeWitt, Numerical Investigation of Free Convection between
Two Vertical Coaxial Cylinders, AIChE J.,vol. 16, pp. 1005-101 0, 1970.
11. H. Ozoe, T. Shibata, and S. W. Churchill, Natural Convection in an Inclined Circular
Cylindrical Annulus Heated and Cooled on Its End Plates, Int. J. Heat Mass Transfer,
vol. 24, pp. 727-737, 1981.
144 M. CHARMCHI AND E. M. SPARROW

12. S. V. Patankar, A Numerical Method for Conduction in Composite Materials, Flow in

Irregular Geometries, and Conjugate Heat Transfer, PToc. 6th Int: Heat Transfer
C o n f , Toronto, vol. 3 , pp. 297-302, 1978.
13. S. V. Patankar, Numerical Heat Transfer and Fluid Flow, Hemisphere, Washington,
D.C., 1980.
14. T. H. Kuehn, Natural Convection Heat Transfer from a Horizontal Cylinder t o a
Surrounding Cylindrical Enclosure, Ph.D. thesis, Dept. of Mechanical Engineering,
Univ. of Minnesota, Minneapolis, Minn., 1976.
15. S. W. Churchill and R. Usagi, A General Expression for the Correlation of Rates of
Transfer and Other Phenomena, AIChE J., vol. 18, pp. 1121-1 128, 1972.

Received December 11, 1981

Accepted January 26,1982

Requests for reprints should be sent t o E. M. Sparrow.