Under consideration for publication in J. Fluid Mech.

Effects of Prandtl number and a new instability

mode in a plane thermal plume
arXiv:0802.0059v1 [physics.flu-dyn] 1 Feb 2008

By R. LAKKARAJU AND MEHEBOOB ALAM†

Engineering Mechanics Unit, Jawaharlal Nehru Center for Advanced Scientific Research
Jakkur P.O., Bangalore 560064, India

(Received 1 July 2007 and in revised form 10 September 2007)

The effect of Prandtl number on the linear stability of a plane thermal plume is analyzed under
quasi-parallel approximation. At large Prandtl numbers (P r > 100), we found that there is an ad-
ditional unstable loop whose size increases with increasing P r. The origin of this new instability
mode is shown to be tied to the coupling of the momentum and thermal perturbation equations.
Analyses of the perturbation kinetic energy and thermal energy suggest that the buoyancy force
is the main source of perturbation energy at high Prandtl numbers that drives this instability.

1. Introduction
The classic problem of natural-convection flow above a horizontal line heat source has re-
ceived considerable attention during the last few decades (Batchelor 1954; Fujii 1963; Gebhart
et al.1988). The temperature of the heat source is larger than that of the ambient fluid, and the
resulting density difference creates a plume that rises up against the gravity. For steady laminar
plumes, the similarity solutions of the pertinent boundary layer equations have been published by
many researchers (Fujii 1963; Gebhart, Pera & Schorr 1970; Riley 1974); experimental studies
on laminar plane plumes are in good agreement with similarity solutions (Riley 1974). Exper-
iments (Pera & Gebhart 1971) have confirmed that the laminar plumes are unstable, and they
sway in a plane perpendicular to the axis of the source. Pera & Gebhart (1971) have shown that
the initial instability of plane plumes to two-dimensional disturbances can be analyzed by the
linear stability theory and developed the coupled Orr-Sommerfeld type equations using a quasi-
parallel flow approximation. Since Squire’s theorem holds for natural convection flows (Gebhart
et al.1988), it is sufficient to consider two-dimensional disturbances for the stability analysis of
a thermal plume.
Strictly speaking, the thermal plume is a non-parallel flow field, and the streamwise variations
of both the laminar and disturbed flows should be incorporated in the stability analysis (Hieber
& Nash 1975; Wakitani 1985). From a weakly non-parallel spatial stability analysis (Wakitani
1985), it has been shown that the critical Grashof number of a plane thermal plume is slightly
larger than that predicted from the quasi-parallel theory, even though its precise value depends
on the flow quantity (fluctuating kinetic energy or thermal energy, etc.) that is being monitored to
calculate non-parallel corrections. It was shown that a lower branch of the neutral stability curve
in the (frequency, Grashof number)-plane exists when the non-parallel corrections are taken into
account. The upper branch of the neutral curve at moderate-to-large values of Grashof number
remains relatively unaffected, however, with non-parallel corrections.
Two non-dimensional numbers involved in natural convection phenomena are the Grashof
number (Gr), the ratio of the buoyancy force and the viscous force, and the Prandtl number
† Author to whom correspondence should be addressed: Email: [email protected]
Published in J. Fluid Mech., vol. 592, p. 221-231 (2007)
2 R. Lakkaraju and M. Alam
(P r), the ratio of the kinematic viscosity and the thermal diffusivity. In terms of P r, there are two
limiting cases: zero Prandtl number limit (e.g. molten metals P r ∼ 10−2 ) and infinite Prandtl
number limit (e.g. P r ∼ 103 for magmas, and P r ∼ 1021 for Earth’s mantle plume). Geological
flows involve fluids with large Prandtl numbers (Worster 1986; Lister 1987; Grossman & Lohse
2000; Kaminski & Jaupert 2003; Majumder, Yuen & Vincent 2004) and are studied in the limit
of infinite Prandtl number (Wang 2004) for which the inertial terms in the momentum equations
are neglected.
The goal of the present work is to understand the stability characteristics of high Prandtl num-
ber plane thermal plumes. To the best of our knowledge, all stability analyses of plane thermal
plumes (Pera & Gebhart 1971; Hieber & Nash 1975; Wakitani 1985) are confined to that of air
(P r = 0.7) and water (P r = 6.7). We use the quasi-parallel approximation to analyse the linear
stability of a thermal plume which is found to be unstable for very small Grashof numbers at any
Prandtl number. At high Prandtl numbers, we find a new instability loop which is shown to be
tied to the coupling of the hydrodynamic and thermal perturbation equations. An analysis of the
perturbation energy unveils the driving mechanism of this instability.

2. Governing equations and base flow

We consider the convective flow generated above a line heat source in an otherwise stagnant
fluid which is maintained at a constant temperature T∞ . Let the Cartesian coordinate system is
(x, y), with x being directed along the flow direction (i.e. against gravity) and y is the transverse
direction, and u and v are the corresponding velocity components along x and y directions,
respectively, and t is the time. With Boussinesq approximation, the governing equations for the
velocity and the temperature fields are given by
∂u ∂u ∂u 1 ∂p
+u +v = ν∇2 u − + gβ(T − T∞ ) (2.1)
∂t ∂x ∂y ρ ∂x
∂v ∂v ∂v 1 ∂p
+u +v = ν∇2 v − (2.2)
∂t ∂x ∂y ρ ∂y
∂T ∂T ∂T ∂u ∂v
+u +v = κ∇2 T , + = 0 (2.3)
∂t ∂x ∂y ∂x ∂y
Here ρ is the mean density of the fluid, p is the pressure, g is the acceleration due to gravity;
the thermo-physical properties of the fluid are the thermal expansion coefficient β, the kinematic
viscosity ν, the thermal conductivity k, the specific heat at constant pressure cp and the thermal
diffusivity κ = k/ρcp . The boundary conditions on velocity and temperature are: u = v =
0, T = Ts at x = y = 0 and u = v = T = 0 at x2 + y 2 → ∞.

2.1. Base flow: similarity solution

The steady laminar base flow is given by the leading-order boundary-layer equations (Fujii 1963;
Gebhart et al.1970; Pera & Gebhart 1971; Riley 1974) that can be expressed in terms of a stream
function: u = ∂ψ/∂y, v = −∂ψ/∂x. The resulting partial differential equations (not shown)
can be transformed into a set of ODEs in terms of a similarity variable η = y/δ, with δ = 4x/G,
where
1/4
gβ(T0 (x) − T∞ )x3 Gr
Gr = and G = 4 (2.4)
ν2 4
are the local Grashof number and the ‘modified’ Grashof number, respectively, and T0 (x) ≡
T (x, y = 0) is the local centerline temperature. The non-dimensional stream-function and tem-
New instability mode in a plane thermal plume 3
1 1
(a) (b)
0.8 0.8
Pr=0.7
0.6 0.6
Pr=0.7
f’ 6.7 h
0.4 0.4
100
0.2 1000 0.2 1000

0 −3 −2 −1 0 1 2
0 −3 −2 −1 0 1 2
10 10 10 10 10 10 10 10 10 10 10 10
η η

F IGURE 1. Variations of the base-state (a) velocity (f ′ ) and (b) temperature (h) with Prandtl number.

perature are defined via

ψ T − T∞ T − T∞
f (η) = , h(η) = = , (2.5)
Uc δ T0 (x) − T∞ (νUc /gβδ 2 )
where Uc = νG2 /4x is the local convective velocity and Tc = (T0 (x) − T∞ ) = νUc /gβδ 2 is
the local excess temperature at the centerline of the plume. With the assumption of power-law
variation of Tc (∼ x−3/5 ), the similarity equations can be obtained as
12 ′′ 4 ′ 2
′′′ ′′ 12 ′
f +
f f − f + h = 0, h + P r (f h) = 0, (2.6)
5 5 5
where the prime denotes differentiation with respect to η, and the related boundary conditions
′′ ′ ′
are (Gebhart et al.1970): f (0) = f (0) = h (0) = 0, h(0) = 1, f (∞) → 0, h(∞) → 0.
These equations have been solved by using the fourth-order Runge-Kutta method with Newton-
Raphson correction. The far-field boundary condition was implemented at η = 12, and the results
were checked by using different values of η = 8, 16, 20, 50. Figure 1 shows the velocity and the
temperature profiles for a range of Prandtl numbers. With increasing P r, the velocity profile
flattens across the plume width, and the temperature boundary layer becomes narrower.

3. Linear stability analysis: quasi-parallel approximation

To analyse the stability of a thermal plume, we decompose each dynamical variable into a
mean part (base flow) and a small-amplitude perturbation:
u(x, y, t) = u(x, y) + ũ(x, y, t), v(x, y, t) = v(x, y) + ṽ(x, y, t) (3.1)
T (x, y, t) = T (x, y) + T̃ (x, y, t), p(x, y, t) = p(x, y) + p̃(x, y, t) (3.2)
with the base flow being taken as that given by the similarity solution (2.6) over which the per-
turbation equations are linearized. The base flow quantities v and the x-derivatives of u and T
are taken as zero in the linearized perturbation equations– this is called the quasi-parallel ap-
proximation. The perturbations are assumed to be of the form such that their amplitudes depend
on the similarity variable η, as does the base flow.
As in the case of the base flow, it is straightforward to show that the perturbation equations
can be expressed in terms of a stream function: ũ = ∂ ψ̃/∂y, ṽ = −∂ ψ̃/∂x. The resulting
perturbation equations (not shown) are amenable to normal-mode analysis:
(ψ̃, T̃ )(x, η, t) = (φ̃, s̃)(η)ei(αx−ωt) , (3.3)
4 R. Lakkaraju and M. Alam
where α and ω are the non-dimensional wavenumber and frequency, respectively, with δ and
τ = δ/Uc being the reference length and time scales. The amplitudes for the perturbation stream
function and temperature are made dimensionless via φ = φ̃/Uc δ and s = s̃/Tc . Substituting
the normal-mode decomposition (3.3) into the linearized perturbation equations, the coupled
Orr-Sommerfeld stability equations are obtained:
′′′′ ′′ ′
h ′ ω ′′ ′′′
i
φ − 2α2 φ + α4 φ + s = iαG f − (φ − α2 φ) − f φ (3.4)
′′
h ′ α ω ′
i
s − α2 s = iαP rG f − s−h φ (3.5)
α
The boundary conditions on φ(η) and s(η) are:
′
φ(±∞) = φ (±∞) = s(±∞) = 0. (3.6)
3.1. Varicose and sinuous modes
For varicose modes, both the velocity and temperatures are symmetric about the mid-plane (η =
0) which can be translated into following conditions on φ and s:
′′ ′
φ(0) = φ (0) = s = 0, (3.7)
whereas for sinuous modes both the velocity and temperatures are asymmetric about the mid-
plane (η = 0) for which the conditions on φ and s are:
′ ′′′
φ (0) = φ (0) = s(0) = 0. (3.8)
It is known that the thermal plumes are more unstable to sinuous modes (Pera & Gebhart 1971)
which has been confirmed in our study too. Hence, all results are presented only for sinuous
perturbations (3.8).
3.2. Generalized eigenvalue problem and numerical method
The linear stability equations (3.4-3.5), along with boundary conditions (3.6, 3.8), constitute a
generalized eigenvalue problem:
AΦ = λBΦ. (3.9)
T
For the temporal stability analysis, λ = ω is the eigenvalue, Φ = (φ, s) is the eigenfunc-
tion and A and B are 2 × 2 matrix differential operators whose elements can be easily ob-
tained from (3.4-3.5). For the spatial stability analysis, the spatial eigenvalue λ = α appears
nonlinearly in (3.4-3.5) which are subsequently transformed into a linear problem in wavenum-
ber (α) by using the ‘companion-matrix’ method (Bridges & Morris 1984). For this case, Φ =
(α3 φ, α2 φ, αφ, φ, αs, s)T is the eigenfunction and A and B are 6 × 6 matrix differential oper-
ators; the non-zero elements of A are (with D = d/dη): A11 = −iGf ′ , A12 = iωG + 2D2 ,
′′′ ′ ′
A13 = iGf ′ D2 − iGf , A14 = −D4 − iωGD2 , A16 = −D, A53 = iP rGh , A55 = −iP rGf ,
A56 = D2 + iωP rG, A21 = 1 = A32 = A43 = A65 , and B is an unit diagonal operator.
For the temporal stability, the wavenumber, α, is real and the frequency, ω = ωr + iωi , is
complex, with ωi being the ‘temporal’ growth/decay rate of the perturbation. For the spatial
stability, the frequency, ω, is real and the wavenumber, α = αr + iαi , is complex, with αi
being the ‘spatial’ growth/decay rate. In either case, the flow is said to be stable/unstable if ωi or
−αi >, < 0, respectively, and neutrally stable if ωi or αi = 0.
For both temporal and spatial analyses, the differential eigenvalue problem (3.9) is transformed
into an matrix eigenvalue problem by discretizating the related differential operators along the
non-periodic η-direction. We have used two numerical methods for discretization (Malik 1990):
(1) the finite difference method with second-order accuracy; (2) the Chebyshev spectral collo-
cation method. The resulting matrix-eigenvalue problem has been solved by the QZ-algorithm
New instability mode in a plane thermal plume 5
3 2
(a) (b)
0 0
2.5
0.005 1.5
2 0.005
α 0.01 α 0.01
1.5 0 1
0.015
1
0.5 cr>0
0.5 c >0 0.015
r cr<0
0 0
0 50 100 150 200 250 0 5 10
G G
F IGURE 2. For the temporal analysis, the stability diagrams in the (α, G)-plane at P r = 200; panel b is
the zoomed part of low-G region.

0.6 0.15
(a) (b)
0
0.5
0.1
0.4 0.02
ω −α
i
0.3 0.05 0.05

0.2
0
0.1 G ~ 0.185
0.1 αr ~ 0.47
0 −0.05
0 50 100 150 200 250 0 5 10
G G
F IGURE 3. For the spatial analysis, (a) the stability diagram in the (ω, G)-plane at P r = 200; (b) the
variation of spatial growth-rate (−αi ) with G for ω = 0.

of Matlab. The growth-rate and the phase speed of the least-stable mode, obtained from finite
difference and spectral methods, were compared for a few test cases with different number of
grid/collocation points (N = 101 and 151). We found that both the growth rate and the phase
speed agreed upto the third decimal place for two methods. Moreover, from a comparison with
published literature, we found that our neutral stability curve for air (P r = 0.7) agrees well with
that of Pera & Gebhart (1971).

4. Results and discussion

We have carried out both temporal and spatial stability analyses of a thermal plume, and most
of the results are presented for the temporal case (except in figure 3).

4.1. Results for various Prandtl numbers

Figure 2(a) displays a typical stability diagram in the (α, G)-plane at high Prandtl numbers
(P r = 200) for the temporal stability analysis; figure 2(b) is the zoomed part of the low-G
6 R. Lakkaraju and M. Alam
0.2 0.2
(a) (b)
0.15 0.15

0.1 0.1
r

r
New Mode
ω ,c

ω ,c
0.05 0.05
New Mode
i

i
0 10ω 0 10ω
i i

c c
−0.05 r −0.05 r

−0.1 −0.1
0 1 2 3 0 1 2 3
α α

F IGURE 4. Variations of the ‘temporal’ growth rate (ωi ) and the phase speed (cr = ωr /α) of the least
stable mode with wavenumber α for (a) G = 50 and (b) G = 100, with P r = 200.

region of figure 2(a). In each panel, the neutral contour (ωi = 0) is marked by ‘0’, and the flow is
unstable inside it (see the positive growth rate contours) and stable outside. There are two distinct
unstable zones: (a) one at low wavenumbers (α) that spans the whole range of Grashof number
(G > Gcr ), and (b) the other at relatively higher wavenumbers that spans a limited range of G.
Figure 2(b) suggests that there is a minimum value of G below which the plume is stable.
The thick line in figure 2(a-b) demarcates the regions of downstream-propagating (phase
speed, cr = ωr /α > 0) and upstream-propagating (cr < 0) modes in the (α, G)-plane. The
origin of such upstream-propagating modes (at low α) remains unclear to us at present. We have
checked that the locus of cr = 0 line in the (α, G)-plane does not change by increasing the size
of the computational domain from η = 12 to η = 100 or by increasing the number of collocation
points. We should point out that the possibility of having upstream-propagating modes in a plane
thermal plume (which exist for any P r at very small values of α) has not been mentioned in
previous works (Pera & Gebhart 1971; Hieber & Nash 1975; Wakitani 1985). Such modes might
be analogous to certain backward-propagating modes in Ekman boundary layer (Lilly 1966) –
this issue is relegated to a future study.
Figure 3(a) displays the analogue of figure 2(a) in (ω, G)-plane for the spatial stability analy-
sis. As expected, the stability diagram in the (ω, G)-plane also contains two unstable loops which
are analogues of the two-loops of the temporal case, figure 2(a). Focussing on the zero-frequency
modes (ω = 0) in figure 3(a), we plot the variation of the spatial growth rate of the least-stable
mode (−αi ) with G in figure 3(b). It is seen that the flow is unstable to ω = 0 modes beyond
a minimum Grashof number, G ∼ 0.185, for P r = 200, and the corresponding real wavenum-
ber is αr ∼ 0.47. This critical point (G, αr ) = (0.185, 0.47) from the spatial analysis exactly
matches with the intersection point between the neutral curve and the locus of cr = 0 modes in
figure 2(b) for the temporal analysis. This result establishes that the upstream propagating modes
(cr < 0) for the temporal case are not an artifact of the numerical method.
Here onwards, we present results only for temporal stability. Focussing on figure 2(a), we show
the variations of the growth-rate and phase speed of the least-stable mode with wavenumber in
figures 4(a) and 4(b) for G = 50 and G = 100, respectively. Two humps in each growth-rate
curve correspond to two unstable loops in figure 2(a), and the second hump is referred to as new
mode since it does not have an analogue in low-P r fluids. The discontinuities in each phase-
speed curve correspond to crossing of different modes.
For a range of Prandtl numbers (P r = 0.7, 100, 200 and 500), the stability diagrams, contain-
ing the neutral contour (ωi = 0) along with a few positive growth-rate contours (ωi > 0), are
New instability mode in a plane thermal plume 7
1.5 3
(a) 0 (b)
2.5 0
0.02
1 0.04 2
α α 0.01
0.06 1.5

0.5 1 0.02

0.5 0.02

0 0
0 50 100 150 200 250 0 50 100 150 200 250
G G
3 4
(d) 0
(c) 0
2.5
New Mode 3 New Mode
2 0.005
α 0.01 α
1.5 2

1
1
0.5 0.015
0.01
0 0
0 50 100 150 200 250 0 100 200 300
G G
F IGURE 5. Stability diagrams at various Prandtl numbers: (a) P r = 0.7; (b) P r = 100; (c) P r = 200; (d)
P r = 500. Thick line in each panel corresponds to the neutral contour for ‘uncoupled’ stability equations
(see §4.2 for related explanation).

compared in the (G, α)-plane in figure 5(a-d). For P r = 0.7 (air), the stability diagram has one
loop, and the upper branch of the neutral curve is well defined and has an inviscid asymptotic
limit: α = 1.3847 (Pera & Gebhart 1971). In the limit G → ∞, there exists a range of wave-
numbers over which the flow is unstable. At high Prandtl numbers (P r > 100), as in figure 5(c),
the neutral curve contains a kink, and there is an additional unstable loop at large α and low G.
The size of this new unstable loop increases with increasing Prandtl number, see figure 5(d). As
mentioned before, this new unstable loop is referred to as a new mode since it does not appear in
low-P r fluids. Comparing the growth rate contours for different P r in figure 5, we find that the
growth rate of the least-stable mode decreases with increasing Prandtl number, even though the
size of the unstable zone in the (α, G)-plane increases in the same limit. The thick solid contour
in each panel of figure 5 is explained in the next section.
Now we suggest one possible experiment to realize this new instability mode. Suppose a lam-
inar plume is disturbed by a small-amplitude sinusoidal excitation of the source with a specified
frequency, f = ω/2π (the temperature at the source is constant such that G = 100, say, in figure
3a). For small enough f the plume will show a wavy instability according to the lower instability
loop in figure 3(a), however, for relatively larger f the plume is unstable to our new instability
loop and there is a window of frequencies between two instability modes over which the plume
8 R. Lakkaraju and M. Alam
remains stable. It would be interesting to verify this transition scenario at a given Grashof num-
ber, ‘unstable→stable→unstable’ with increasing frequency, in experiments of high P r-fluids.

4.2. Origin of new instability loop: coupling of hydrodynamic and thermal fluctuations
To shed light on the origin of the new instability loop at high P r, here we assume that the
velocity and the temperature perturbations are decoupled from each other. The full set of stability
′ ′
equations, (3.4-3.5), can be made independent from each other by dropping s and h φ from
equations (3.4) and (3.5), respectively. These two sets of equations can now be solved separately
to determine the least stable eigen-value – we have verified that the least stable mode belongs
to the Orr-Sommerfeld equation (i.e. a purely hydrodynamic mode, eqn. 3.4), and the energy
equation (3.5) always yields a stable mode.
For the uncoupled perturbations, the neutral stability curve for each P r is superimposed as a
thick solid contour in figure 5. (The flow is unstable inside the thick contour and stable outside.)
A comparison of the thick line in each panel with the corresponding neutral contour of coupled
stability equations (denoted by the thin curve 0) clearly reveals that the coupling between the
hydrodynamic and the thermal disturbance equations is responsible for the origin of our new
instability mode at high P r. It is observed that the lower parts of the instability loops in fig-
ures 5(c-d) closely follow the instability loop of the ‘uncoupled’ Orr-Sommerfeld equation, and
are, therefore, purely hydrodynamic in origin.
It is clear that the coupling terms in the stability equations (3.4-3.5) are responsible for ap-
pearance of the additional instability loop at high Prandtl numbers, and solving the uncoupled
perturbation equations would lead to incorrect results. The importance of this coupling between
hydrodynamic and thermal perturbations at high P r can be understood from the fact that the
′
gradient of the base-flow temperature, h , (which appears in the energy perturbation equation)
increases with Prandtl number: h′ (η) ∼ P rf (η)h(η) ∼ P rǫ , with 0 < ǫ < 1, and hence cannot
be neglected at large P r. (From an order-of-magnitude analysis of the pertinent boundary-layer
equations, we find ǫ = 1/2.)

4.3. Analysis of perturbation energy: instability mechanism

Lastly, to understand the underlying instability mechanism, we analyse different components of
perturbation energy. The time-evolution equations of perturbation kinetic energy and thermal en-
ergy are obtained from (3.4-3.5) by multiplying them with the corresponding complex conjugate
quantity φ† and s† , respectively, and integrating them from η = 0 to η = ∞. Considering the
real parts, the resulting evolution equations boil down to (Nachtsheim 1963; Gill & Davey 1969)
Z ∞ Z ∞ Z ∞ Z ∞
dEK
≡ ωi eK dη = etrK dη + eV D dη + eB dη, (4.1)
dt
Z0 ∞ Z 0∞ Z 0∞ 0
dET
≡ ωi eT dη = etrT dη + eT D dη, (4.2)
dt 0 0 0

where
′′
′ ′
′′
eK = | φ |2 + α2 | φ |2 , etrK = αf φr φi − φr φi , eV D = G−1 | φ − α2 φ |2 ,

′ ′

eB = G−1 sr φr + si φi ,
′
′
eT = | s |2 , etrT = αh (φr si − φi sr ) , eT D = −P r−1 G−1 | s |2 + α2 | s |2 ,

and the suffixes r and i denote the real and imaginary parts, respectively. For hydrodynamic
R fluc-
tuations, dEK /dt represents the rate of change of perturbation kinetic energy, EtrK = etrK
the rate of transfer of kinetic energy from the mean flow to perturbations via the Reynolds stress,
New instability mode in a plane thermal plume 9
−3
x 10
5 0.2
(a) (b) E
E T
K 0.15
EtrT
E
trK 0.1
E
Ei EVD E TD
i
0 0.05
E
B
0

−0.05

−5 −0.1
0 1 2 3 4 0 1 2 3 4
η η
F IGURE 6. Distributions of (a) kinetic and (b) thermal energies across the plume width for P r = 0.7 (air),
G = 100, α = 1.2923. Vertical line indicates the location of the critical layer.

−4
x 10
2 0.02
(a) (b) E
E T
K
1 E
E 0.01 trT
trK
E
Ei 0 EVD E TD
i
0
−1 EB

−0.01
−2

−0.02
0 0.5 1 0 0.5 1
η η
F IGURE 7. Same as figure 6, but for P r = 200, G = 100, α = 2.74645.

EV D the rate of viscous dissipation, and EB the rate of gain of kinetic energy through the buoy-
ancy force. For thermal fluctuations, dET /dt is the rate of change of perturbation thermal energy,
EtrT is the rate of gain of thermal energy from the mean temperature field and ET D is the rate
of dissipation of thermal energy.
The variations of different kinetic and thermal energy components across the plume width η
are displayed in figures 6 and 7 for P r = 0.7 (α = 1.2923) and 200 (α = 2.74645), respectively,
at G = 100. (These two cases correspond to the neutral modes on the upper branch of the neutral
contour in figures 5a and 5c.) In each figure, the location of the corresponding critical layer
is indicated by the vertical line. Since the instability
R mode is neutral (ωi = 0), the net rate of
gain of kinetic/thermal energy is EK = ωi eK dη = 0 = ET , denoted by the dotted zero-
line in each panel. For P r = 0.7, the kinetic energy gained by the perturbation mainly comes
from the Reynold’s stress term (EtrK ) and a small amount is contributed from the perturbation
buoyancy force (EB ); the maximum amount of energy is dissipated at the center line (η = 0) by
viscous forces (EV D ). With increasing P r, the rate of gain of kinetic energy by Reynolds stress
becomes progressively smaller, and the buoyancy force (EB ) takes over as the main source of
10 R. Lakkaraju and M. Alam
perturbation kinetic energy (see figure 7a at P r = 200) which balances the energy lost due to
viscous dissipation. We conclude that at high P r the contribution from the Reynold’s stress term
is negligible compared to the gain in kinetic energy by buoyancy force which drives instability.

5. Conclusions
Based on a quasi-parallel stability analysis of a plane thermal plume, we have uncovered a
new instability loop at high Prandtl numbers. The origin of this new mode is shown to be tied
to the coupling between the hydrodynamic and thermal fluctuations. The importance of this cou-
pling is tied to the increasing magnitude of the base-state temperature gradient with increasing
Prandtl number. It is shown that the perturbation kinetic energy gained from the buoyancy force
drives this instability at high P r. The underlying instability mechanism differs from the well-
known hydrodynamic instability mechanism for which the perturbation energy is gained from
the mean flow via the Reynolds stress. In future, it would be interesting to analyse the effects of
non-parallel corrections as well as the temperature-dependent transport coefficients on our new
instability loop.

We acknowledge financial support from two grants (DRDO/RN/4124 and PC/EMU/MA/35).

M.A. acknowledges Prof. Vijay Arakeri for motivating this research.

REFERENCES
• Batchelor, G. K. (1954) Heat convection and buoyancy effects in fluids. Q. J. R. Met. Soc.
80, p. 339-359.
• Bridges, T. J. & Morris, P. J. (1984) Differential eigenvalue problems in which the parameter
appears nonlinearly. J. Comp. Phys. 55, p. 437-460.
• Fujii, T. (1963) Theory of the steady laminar natural convection above a horizontal line heat
source and a point heat source. Int. J. Heat Mass Transfer 6, p. 597-606.
• Gebhart, B., Pera, L. & Schorr, A. (1970) Steady laminar natural convection plumes above
a horizontal line heat source. Int. J. Heat Mass Transfer 13, p. 161-171.
• Gebhart, B., Jaluria, J., Mahajan, R. L. & Sammakia, B. (1988) Buoyancy Induced Flows
and Transport. Hemisphere Publishing, Washington.
• Gill, A. E. & Davey, A. (1969) Instabilities of a buoyancy-driven system. J. Fluid Mech. 35,
p. 775-798.
• Grossmann, S. & Lohse, D. (2000) Scaling in thermal convection: a unifying theory. J. Fluid
Mech. 407, p. 27-56.
• Hieber, C. A. & Nash, E. J. (1975) Natural convection above a line heat source: Higher
order effects and stability. Int. J. Heat Mass Transfer 18, p. 1473-1479.
• Kaminski, E. & Jaupart, C. (2003) Laminar starting plumes in high Prandtl number fluids.
J. Fluid Mech. 478, p. 287-298.
• Lilly, K. L. (1966), On the instability of Ekman boundary flow. J. Atmos. Sci. 21, p. 481-494.
• Lister, J. R. (1987), Long-wavelingth instability of a line plume. J. Fluid Mech. 175, p. 571-
591.
• Majumder, C. A. H., Yuen, D. A. & Vincent, A. P. (2004) Four dynamical regimes for a
starting plume model. Phys. Fluids 16, p. 1516-1531.
• Malik, M. R. (1990) Numerical methods for boundary layer stability. J. Comp. Phys. 86,
p. 376-413.
• Nachtsheim, P. R. (1963) Stability of free convection boundary layers. NASA Tech. Note,
TN D-2089.
New instability mode in a plane thermal plume 11
• Pera, L. & Gebhart, B. (1971) On the stability of laminar plumes: Some numerical solutions
and experiments. Int. J. Heat Mass Transfer, 14, p. 975-984.
• Riley, N. (1974) Free convection from a horizontal line source of heat. ZAMP 25, p. 817-
828.
• Wakitani, S. (1985) Non-parallel flow instability of a two dimensional buoyant plume. J.
Fluid Mech. 159, p. 241-258.
• Wang, X. (2004) Infinite Prandtl number limit of Rayleigh-Benard convection. Comm. Pure
Appl. Math. 57, p. 1265-1282.
• Worster, M. G. (1986) The axisymmetric laminar plume: Asymptotic solution for large
Prandtl number. Stud. Appl. Math. 75, p. 139-152.