J . Fluid Mech. (1968), vo2. 32, purt 2, p p .

393-395 393
Printed in &eat Britain

The Rayleigh-Jeffreys problem with boundary slab

of finite conductivity
By D. A. N I E L D
Department of Mathematics, University of Auckland, New Zealand

(Received 11 September 1967)

Linear perturbation analysis is applied to the problem of the onset of convection

in a horizontal layer of fluid heated uniformly from below, when the fluid is
bounded below by a rigid plate of infinite conductivity and above by a solid
layer of finite conductivity and finite thickness. The critical Rayleigh number
and wave-number are found for various thickness ratios and thermal conductivity
ratios. Both numbers are reduced by the presence of a boundary of finite (rather
than infinite) conductivity in qualitative agreement with the observation of
Koschmieder (1966).

1. Introduction
The problem of determining the conditions for the onset of convection induced
by buoyancy effects resulting from the heating from below of a horizontal layer
of a viscous fluid, received its first theoretical treatment from Rayleigh (1916)
and Jeffreys (1926). (The problem has been referred to as the BBnard problem,
but this term is more appropriate for the related surface-tension problem.) Many
subsequent authors have extended the theory, a standard reference for which is
Chandrasekhar (1961).
Until recently attention has been concentrated on the case where both upper
and lower boundaries are of infinite thermal conductivity. Sparrow, Goldstein
& Jonsson (1964) extended the classical Rayleigh-Jefieys analysis to allow for
a ‘radiation’ type condition at a boundary, while Hurle, Jakeman & Pike (1967)
considered a layer of fluid bounded below and above by the same solid material,
of finite thermal conductivity, extending to infinity in each vertical direction.
I n the present paper we treat a configuration more practical than that of Hurle
et al. Our model is as follows. The upper surface of the fluid is adjacent to the lower
side of a horizontal slab of solid material whose upper side is in turn bounded by
a medium of infinite conductivity. The lower boundary of the fluid is assumed t o
be an infinitely conducting rigid plate. (The analysis below could, if desired,
easily be adapted for the case where this plate was replaced by a second slab of
finite conductivity.)
Most experiments on the Rayleigh-Jeffreys problem have been designed to
ensure good heat conduction on the boundaries. An exception is the experiment
described by Koschmieder (1966) who, in order to facilitate visual observation of
394 D. A . Nield
the convection patterns, used as a lid over the fluid a plate of glass whose thermal
conductivity was only an order of magnitude greater than that of the fluid.
Koschmieder found that when the Rayleigh number was continuously increased
beyond the critical value for the onset of convection, the wave-number of the
convection pattern decreased. This decrease, which is also evident from photo-
graphs published by Schmidt & Silveston (1959), is in disagreement with the
theory (for infinitely conducting boundaries) of Platzman (1965) or Schliiter,
Lortz & Busse (1965). As was pointed out by Dr Koschmieder in a private com-
munication to the author, the decrease in wave-number is apparently the result
of increasingly less effective transfer of heat perturbations in the media (glass
and cooling fluid) above the observed fluid layer in comparison with that in this
convecting fluid. The present work was undertaken in an attempt to throw some
light, as far as a linear theory can, on this phenomenon.

2. Analysis
We consider a fluid layer of depth d overlain by a solid layer of thickness d'.
Thus, with z indicating distances vertically upwards, the fluid will occupy the
region 0 < z < d and the solid the region d < z < d + d'. The planes z = 0 and
z = d + d' are held at uniform temperatures Toand Tl respectively, and each plane
is assumed to be a perfect thermal conductor. I n the steady state the fluid velocity
is then zero and the temperature distribution is given by

where

are the adverse temperature gradients in the fluid (of thermal conductivity K )
and the solid (of thermal conductivity K') respectively.
The analysis now follows that of Chandrasekhar (1961). The fluid is assumed to
be of the Boussinesq type (quasi-incompressible, and otherwise with constant
fluid properties). Perturbations from the steady-state solution are considered,and
the governing differential equations are linearized. An expansion in normal
modes, involving a separation of variables, is performed. The following equations
for a steady neutral disturbance are obtained.
In0 < z < d, v(d2/d$- k2)2W - gak20 = 0, (1)
and K(d2/dz2-k2)0+/3W = 0, (2)
whileind < z < d+d',
(d2/dz2- k2)0' = 0. (3)
Here a,v and K are the thermal expansion coefficient, kinematic viscosity and
thermal diffusivity of the fluid, g is the gravitational acceleration, and k is the
horizontal wave-number of the disturbance. W ( z )and 0(x)give the variation
 Rayleigh-Jeffreys problem with a boundary slab 395
with z of the vertical component of the fluid velocity, and the temperature per-
turbation in the fluid, respectively. I n the solid the corresponding quantities are
denoted by primes.
The thermal boundary conditions are
0 = 0 at z = O , (4)
o = O’ and K a q a z = K’aw/az at = a, (5)
0‘ = 0 at z = d+d’. (6)
(It seems that in the paper of Hurle et al. (1967), at this stage of the analysis
and in their table of results, it is the thermal conductivity K rather than the
thermal diffusivity K which should be used.)
The solution of (3) subject t o the condition (6) is
0’ = C sinh k(d+d’-z),
where C is a constant. Conditions ( 5 ) then imply that
1 a0 K
--=-kcothkd at z = d . (7)
0 ax K‘

CE, FE C( - 1)”E, F: Rb2C(- l),E, F,

C( - l)nE,F; CE,F; RbZCE, F, = 0, (15)
C( - l)nEmFn CEnFn Rb’CE, - $(L +nb coth n b )
396 D. A . Nield
The minimum value of R as b varies is then the critical value R,, which, for the
lowest mode, corresponds to the onset of convection. The numerical calculation
is straightforward.

3. Results and discussion

I n tables 1 ( a ) and 1 (b) are presented values of the critical Rayleigh number
R: = n4R, and the corresponding critical wave-number a, = n-be for the most
unstable mode. As expected, both R, and a, increase with K'/K and decrease
with increase of d'/d. The same trend is seen in the values given in tables 2 ( a )
and 2 ( 6 ) for the second mode of instability.
d'Jd
A
r 7
k"/K 0 0.01 0.03 0.1 0.3 1 2- 10
0 - 1295.8 1295.8 1295.8 1295.8 1295.8 1295.8
0.01 1707.8 1398.5 1337.9 1309.8 1301.4 1299.5 1299.4
0.03 1707-8 1497.6 1398.7 1334.9 1312.1 1306.6 1306.5
0.1 1707.8 1607.1 1508.3 1400.4 1345-1 1329-9 1329.6
0.3 1707.8 1667.1 1607.3 1500.3 1414.0 1384.1 1383.4
1 1707.8 1694.6 1670.9 1610.3 1529.7 1493.4 1492.7
3 1707.8 1703.3 1694.6 1668.2 1623.4 1599.3 1598.9
10 1707.8 1706.5 1704.0 1694.9 1678.1 1668.1 1668.0
100 1707.8 1707.6 1707.4 1706.4 17044 1703.4 1703.4
co 1707.8 1707.8 1707.8 1707.8 1707.8 1707.8 1707.8
TABLEl(a). Values of the critical Rayleigh number R,* = n4R for the lowest mode of
instability
d'ld
7
A _____ v
K'/K 0 0.01 0.03 0.1 0.3 1 2- 10
0 - 2.553 2,553 2.553 2-553 2.553 2.553
0.01 3.117 2.753 2.639 2.582 2.564 2.557 2.556
0.03 3.117 2.901 2.750 2.632 2.582 2-566 2.565
0.1 3.117 3-028 2.914 2.751 2.641 2.596 2.594
0.3 3.117 3.084 3.028 2.900 2.753 2.669 2.665
1 3.117 3.107 3-087 3.029 2.910 2.819 2.815
3 3.117 3.113 3.107 3.084 3-027 2.967 2.964
10 3.117 3.115 3.111 3.106 3.086 3.063 3.062
100 3.117 3.116 3.116 3.115 3.114 3.110 3-110
co 3.117 3.117 3.117 3.117 3.117 3.117 3-117
TABLEl(5). Values of the corresponding critical wave-number a, = nb, for the lowest
mode of instability
d'ld
7---
-h__- ~- 7
K'IK 0 0.1 1 2- 10 0 0.1 1 210
0 - 15,278 15,378 15,278 - 4.91 4.91 4.91
0.1 17,610 15,654 15,467 15,467 5.37 5.01 4.94 4.94
1 17,610 16,803 16,382 16,382 5.37 5.24 5.11 5.11
10 17,610 17,492 17,421 17,380 5.37 5.35 5.32 5.31
co 17,610 17,610 17,610 17,610 5.37 5.37 5.37 5-37
(a) (b)
2(ta) and 2(5). Values of the critical Rayleigh number R,* and the corresponding
TABLES
critical wave-number a,, for the second mode of instability
 Rayleigh-Jeffreys problem with ct boundary slab 397
The published experimental results which are most relevant to the present
theory are those of Koschmieder (1966), but his experiment was designed prim-
arily for the visual observation of the convection patterns and not for obtaining
precise quantitative data. By the time the motion of his fluid was sufficiently well
developed to be visible, it is probable that the Rayleigh number was substantially
above the critical value. Thus the agreement between the values R = 1730
and a = 3.3 from Koschmieder’s experiment for K’/K = 8 and d‘ld = 0.3, and
the value R, = 1650 and a, = 3.08 from the present theory, is as good as could be
expected.
When Koschmieder continued to increase the rate of heating of the fluid, he
observed that the wave-number a of the convective motion declined from the
value 3.3 to 2.3, but when the latter value of a had been reached the motion had
become markedly unsymmetrical. The reduction in wave-number cannot be
explained merely by noting that higher order modes of instability become in-
creasingly effective as the Rayleigh number is increased, since the wave-number
is larger for the higher order modes (for boundaries of given constant conduc-
tivity). It appears that the observed effect is a result of the effective decrease in
the appropriate value of our ratio K’/K (or effective increase in d’ld) as the water
circulating above Koschmieder’s glass plate became less and less effective in
transporting away the increased heat flux. I n other words, our assumption that
the surface z = d + d‘ is perfectly conducting becomes increasingly less appro-
priate. Presumably the same is true, but to a lesser extent, at the boundary
x = 0 ,which in Koschmieder’sexperiment was the upper surface of a copper plate.
A proper explanation of the reduction in wave-number with increase in heat
flux must, of course, invoke non-linear considerations, but our linear theory does
at least make plausible a reduction in wave-number from 3.1 to 2.5 (or to a
smaller value if the conductivity of the material below the fluid layer is given a
finite value rather than an infinite one). Further experimental and theoretical
work is highly desirable.

The author is grateful to Dr Koschmieder for drawing his attention to this

problem and the likely explanation.

REFERENCES
CHANDRASEKHAFC, S. 1961 Hydrodynamic and Hydromagnetic Stability. Oxford : Clarendon
Press.
HURLE, D. T. J.,JAKEMAN, E. & PIKE,E. R. 1967 On the solution of the BBnard problem
with boundaries of finite conductivity. Proc. Roy. SOC.A 296, 469-75.
JEFFREYS, H. 1926 The stability of a layer of fluid heated below. Phil. Mug. 2, 833-44.
KOSCHMIEDER,E. L. 1966 On convection on a uniformly heated plane. Beitruge zur
Physik der Atmosphure, 39, 1-11.
NIELD,D. A. 1967 The thennohaline Rayleigh-Jeffreys problem. J . FZuid Mech. 29,545-
58.
PLATZMAN, G. W. 1965 The spectral dynamics of laminar convection. J . FZuid Mech. 23,
481-5 10.
RAYLEIGH, LORD1916 On the convection currents in a horizontal layer of fluid when the
higher temperature is on the under side. Phil. Mag. 32, 529-46.
398 D.A . Nield
SCHLUTER, A., LORTZ,D. & BUSSE,F. 1965 On the stability of steady finite amplitude
convection. J . Fluid Mech. 23, 129-44.
SCHMIDT, E. & SILVESTON,P. L. 1959 Natural convection in horizontal liquid layers.
Chem. Engng Prog. Symposium Series, 55, 163-9.
SPARROW,E. M., GOLDSTEIN,R. J. & JONSSON, V. K. 1964 Thermal instability in a
horizontal fluid layer : effect of boundary conditions and non-linear temperature pro-
file. J . Fluid Mech. 18, 513-28.