Mahfoud 2016
Mahfoud 2016
To cite this article: Brahim Mahfoud, Ali Bendjaghlouli & Rachid Bessaïh (2016):
Magnetohydrodynamic co-rotating flow in a vertical cylindrical container, Numerical Heat
Transfer, Part A: Applications, DOI: 10.1080/10407782.2015.1109383
Article views: 26
Download by: [Library Services City University London] Date: 27 March 2016, At: 21:23
NUMERICAL HEAT TRANSFER, PART A
http://dx.doi.org/10.1080/10407782.2015.1109383
1. Introduction
The influence of a magnetic field on the electrodepositing mass transfer process is well known from
the work of electrochemists [1] and crystal growth specialists [2]. The present study is focused on a
particular issue of the electromagnetic control of convective co-rotation flow in a cylindrical enclos-
ure. We studied the potential of stabilizing a mixed convective flow by an external axial magnetic
field. Several previous studies considered mainly electromagnetic damping of the flow and its effect
on heat/mass transfer and dopant segregation [3, 4]. In many practical applications, heat transfer is
associated with rotational fluid flows. Many previous investigations have focused on the convection
and instability in connection with crystal growth processing units where thermally unstable boundary
conditions are imposed (see Jaluria [5]). In the case of crystal growth, for example, magnetic fields are
used to suppress the convective motion induced by emergent strong fluxes in order to control the
flow in the melt and consequently crystal quality [6]. When the aspect ratio is greater than 1, as in
the present work (H/R ¼ 2), the influence of the lateral walls on flow can no longer be ignored.
The corresponding magnetohydrodynamic (MHD) problem has been studied in detail. Several pre-
vious studies can be found in the literature [7–11], describing asymptotic flow solutions with a strong
axial and horizontal magnetic field.
CONTACT Brahim Mahfoud [email protected] Department of Mechanical Engineering, Faculty of Science and
Applied Science, University of Bouira, Bouira 10000, Algeria.
Color versions of one or more of the figures in the article can be found online at www.tandfonline.com/unht.
© 2016 Taylor & Francis
2 B. MAHFOUD ET AL.
Nomenclature
B magnetic field, Tesla v dimensionless axial velocity
FLr dimensionless radial Lorentz force w dimensionless azimuthal velocity
FLz dimensionless axial Lorentz force α thermal diffusivity of the fluid, m2 s 1
FLθ dimensionless azimuthal Lorentz force β thermal expansion coefficient, K 1
g acceleration of gravity, m/s2 γ aspect ratio (¼ H/R)
H height of the cylinder, m pffiffiffiffiffiffiffiffiffiffiffi Θ dimensionless temperature, [¼ (T Tc)/(Th Tc)]
Ha Hartmann number ð¼ BR r=qnÞ ν kinematic viscosity of the fluid, m2/s
J dimensionless current density ρ density of the fluid, kg/m3
N interaction parameter, (¼Ha2/Re) σ electrical conductivity, Ω 1 · m 1
P dimensionless pressure τ dimensionless time
Pr Prandtl number (¼ν/α) Φ dimensionless electric potential
R radius of the cylinder, m Ω angular velocity, rad s 1
r dimensionless radial coordinate
Downloaded by [Library Services City University London] at 21:23 27 March 2016
In the present paper, we extend our previous study [12] and consider the effects of an axial
magnetic field on the stability of convective co-rotation flow in a vertical cylinder with a vertical
temperature gradient. A parametric study of stability of flow for increasing values of magnetic field
strength (Hartmann number, Ha) is performed for fixed values of both Prandtl number (Pr ¼ 0.015)
and a aspect ratio (γ ¼ H/R ¼ 2) of the cylinder, equal to 2. The corresponding stability diagram,
showing the dependence of the critical Reynolds number, Recr (transition from axisymmetric
to non-axisymmetric flow) on Hartmann number (Ha ¼ 0, 5, 10, and 20) and Richardson number
(Ri ¼ 0, 0.5, 1, and 2), is presented.
2. Problem formulation
We focus on mixed convection of an incompressible Boussinesq fluid in a vertical cylindrical
container of radius R and height H (H/R ¼ γ ¼ 2). The bottom disk rotates at constant angular
velocity Ω, and is maintained at a hot temperature Th, while the top disk is in co-rotating at the same
angular velocity and maintained at temperature Tc (Th > Tc); see Figure 1.
The sidewall is assumed to be adiabatic and the walls of the cylinder are electrically insulated,
subject to a constant external vertical and homogeneous magnetic field B(0, 0, Bez), with ez the unit
vector in the axial direction. The interaction of the magnetic field with the convective flow induces the
electric current with J ¼ σ( ∇ Φ þ Ux B), where U (u, v, w) is the velocity (u, v, and w are the radial,
axial, and azimuthal components of velocity, respectively), Φ is the electric potential, and σ is the
electrical conductivity of the fluid. Conservation of the induced electric current ∇J ¼ 0 yields the
equation for the electric potential, ΔΦ ¼ ∇ (U x B). The flow is described by the dimensionless
three-dimensional Navier–Stokes equation. By neglecting the induced magnetic field, the dissipation,
and Joule heating, and using R, 1/Ω, ΩR, ρ(ΩR)2, ΩR2 B, and (Th Tc) as typical scales, respectively,
for length, time, velocity, pressure, electric potential, and temperature, the continuity, momentum,
energy, and potential equations may be written in dimensionless form as follows.
Continuity equation:
1 qðruÞ 1 qw qv
þ þ ¼0 ð1Þ
r qr r qh qz
NUMERICAL HEAT TRANSFER, PART A 3
Downloaded by [Library Services City University London] at 21:23 27 March 2016
r-Momentum equation:
� � � �
qu qu w qu qu w2 qP 1 1q qu 1 q2 u q2 u u 2 qw
þu þ þv ¼ þ r þ 2 2þ 2 þ NFLr ð2Þ
qs qr r qh qz r qr Re r qr qr r qh qz r2 r2 qh
z-Momentum equation:
� � � �
qv qv w qv qv qP 1 1q qv 1 q2 v q2 v
þu þ þv ¼ þ r þ 2 2 þ 2 þ RiH þ NFLz ð3Þ
qs qr r qh qz qz Re r qr qr r qh qz
Swirl equation:
� � � �
qw qw w qw qw uw 1 qP 1 1q qw 1 q2 w q2 w w 2 qu
þu þ þv þ ¼ þ r þ 2 2þ 2 þ þ NFLh
qs qr r qh qz r r qh Re r qr qr r qh qz r2 r2 qh
ð4Þ
Energy equation:
� � � �
qH qH w qH qH 1 1q qH 1 q2 H q2 H
þu þ þv ¼ r þ 2 2 þ 2 ð5Þ
qs qr r qh qz RePr r qr qr r qh qz
Potential equation:
� �
1q qU 1 q2 U q2 U 1 qðrwÞ 1 qu
r þ 2 2þ 2 ¼ ð6Þ
r qr qr r qh qz r qr r qh
� �
1 qU
FLr ¼ u ð7Þ
r qh
4 B. MAHFOUD ET AL.
FLz ¼ 0 ð8Þ
� �
qU
FLh ¼ w ð9Þ
qr
FLr, FLz, and FLθ are the dimensionless Lorentz forces in the radial, axial, and azimuthal directions,
respectively. P is the dimensionless pressure, Ф is the dimensionless electric potential, and Θ is the
dimensionless temperature. In Eqs. (2)–(5) Re ¼ ΩR2/ν is the Reynolds number, Ri ¼ g · β(Th Tc)/
Ω2R is the Richardson number, and N ¼ Ha2/Re is the interaction
pffiffiffi parameter, which represents the
ratio of the electromagnetic to inertia forces. Ha ¼ BR r=qv is the Hartmann number, which
defines the ratio of the electromagnetic forces to viscous forces and Pr ¼ ν/α is the Prandtl number;
g, β, ρ, and σ are gravity acceleration, the thermal expansion coefficient, density, and the electrical
Downloaded by [Library Services City University London] at 21:23 27 March 2016
u ¼ 0; v ¼ 0; w ¼ 0; H ¼ 0; U ¼ 0
qH qU
u ¼ 0; v ¼ 0; w ¼ 0; ¼ 0; ¼0
qr qr
. At the rotating hot bottom disk: (z ¼ 1/2γ)
qU
u ¼ 0; v ¼ 0; w ¼ r; H ¼ 1; ¼0
qz
. At the co-rotating cold top disk: (z ¼ 1/2γ)
qU
u ¼ 0; v ¼ 0; w ¼ r; H ¼ 0; ¼0
qz
. Periodicity condition:
3. Numerical method
The governing Eqs. (1)–(6), with the associated boundary conditions, are solved using the finite-
volume method [10, 12]. The components of velocity (u, v, and w) are stored at the staggered loca-
tions and the scalar quantities (P, Θ, and Φ) are stored in the center of these volumes. A fully implicit
time-marching scheme is used. The numerical procedure, called SIMPLER [13], is used to handle the
pressure–velocity coupling. For treatment of the convection and diffusion terms in Eqs. (2)–(5), the
central difference scheme is adopted. The discretized algebraic equations are solved by the line-
by-line tridiagonal matrix algorithm (TDMA). Convergence at a given time step is declared when
the maximum relative change between two consecutive iteration levels fell below 10 5 for u, v, w,
and Θ. At this stage, a steady-state solution was obtained. A parallel test was made to guarantee
NUMERICAL HEAT TRANSFER, PART A 5
Table 1. Grid independence test for the hydrodynamic case (Re ¼ 1000) and values of Hartmann number Ha ¼ 0, 8, and 16.
Ha ¼ 0 Ha ¼ 8 Ha ¼ 16
Grid wmax wmax wmax
55 � 55 � 110 0.9863 0.9865 0.9865
60 � 60 � 120 0.9870 0.9870 0.9870
65 � 65 � 130 0.9870 0.9880 0.9893
that the energy balance between the hot and cold walls was less than a prescribed accuracy value,
i.e., 0.2%.
The solution was obtained as follows:
. Equations (1) and (2) are first solved to obtain P, u, v, and w.
. The temperature, Θ and potential, Φ are subsequently computed from Eqs. (5) and (6),
Downloaded by [Library Services City University London] at 21:23 27 March 2016
respectively.
The different quantities associated with the set of equations are discretized on non-uniform grids
and solved on (r; θ; z). For a properly resolved direct numerical simulation, especially in the case of
MHD flow, the characteristics of specific flow, heat transfer, and boundary layers must be considered.
In particular, increase in Ha results in thinner Hartmann layers at the walls normal to the magnetic
field of thickness ∼1/Ha. As a result, non-uniform grids were used for the simulation when Ha ¼ 0–
20. More specifically, a grid of 55 � 55 � 110 nodes was used when Ha ¼ 0–5, 60 � 60 � 120 nodes for
Ha > 5 to 10, and 65 � 65 � 130 nodes for Ha > 10. The above requirements for the construction of
the grid were tested by performing a grid independence test. For this purpose, the hydrodynamic
basic state of Re ¼ 1000 was chosen. The range of grids used in these tests and the individual results
are presented in Table 1. Comparison of the maximum values of the azimuthal velocity component
indicates that the changes are less than 0.5% and thus the grid selection is appropriate.
Figure 2. (a) Comparison between the present numerical results (lines) and experimental results (square) of [14]: dimensionless
azimuthal velocity for the case of a conducting top-rotating disk corresponding to conductance ratio k ¼ 0.1 for Re ¼ 6670 and Ha
¼ 260. (b) Comparison of the present model with the results of [11]: local Nusselt number distribution (Ra ¼ 104; Ha ¼ 25).
thickness O(Re 1/2) on both end disks (see Omi and Iwatsu [15]). An almost uniform axial flow is
observed in the vicinity of the rotating axis as a result of Ekman suction. These symmetries are
illustrated in Figure 3, which shows the plots of vertical velocity contours at Re ¼ 1000 (basic state
flow). Note that u and w are invariant under reflection in z, whereas v changes the signs.
Figure 3. The basic state at Re ¼ 1000, Ri ¼ 0: isocontours in the meridional planes, r ¼ 0 (top) and the views seen from above at
z ¼ 0.5 (middle) and z ¼ 0.5 (bottom) of radial (left), vertical (middle), and azimuthal (right) velocities: positive (negative) values
are indicated by solid (dashed) curves.
Figure 4. Circular patterns of axisymmetric vortices and spiral waves: vertical velocity contours at z ¼ 0.8 (left), z ¼ 0 (middle),
and z ¼ 0.8 (right).
8 B. MAHFOUD ET AL.
Figure 5. Vector field (v, w) showing a spiral wave at z ¼ 08 (left), z ¼ 0 (middle), and z ¼ 0.8 (right) when Re ¼ 2300.
Downloaded by [Library Services City University London] at 21:23 27 March 2016
Figure 6. Ri ¼ 1 and Re ¼ 200 (left), 225 (middle), and 250 (right): contour plots of vertical velocity at z ¼ 0.7 (top), z ¼ 0 (middle),
and z ¼ 0.7 (bottom).
Ri and that the basic axisymmetric state lost stability, to be replaced by circular patterns of axisym-
metric vortices and spiral waves when Ri ¼ 0. The numerical results obtained show the appearance
of mode m ¼ 1 for the values of critical Reynolds number, Recr ¼ 292, 221, and 162, corresponding
to Richardson number Ri ¼ 0.5, 1, and 2, respectively. We also show the cuts (r, θ) of the structure
of vertical velocity plotted in the z ¼ 0 plan (see Figure 8).
Figure 7. Temperature field at Ri ¼ 0.5: in a conductive state, Re ¼ 200 (left) and convective state, Re ¼ 250 (right).
NUMERICAL HEAT TRANSFER, PART A 9
Downloaded by [Library Services City University London] at 21:23 27 March 2016
Figure 8. Stability diagram in the (Re – Ri) plane for values of the Ri ¼ 0, 0.5, 1, and 2.
Figure 9. Effect of vertical magnetic field on (a) radial profiles of axial velocity at z ¼ 0.5 and θ ¼ 0 when Re ¼ 1000 and Ri ¼ 0;
(b) local Nusselt number at the bottom disk and θ ¼ 0 when Re ¼ 200 and Ri ¼ 0.5.
10 B. MAHFOUD ET AL.
Downloaded by [Library Services City University London] at 21:23 27 March 2016
Figure 10. Temperature field at Re ¼ 300, Ri ¼ 2: in a convective state Ha ¼ 10 (left) and conductive state Ha ¼ 30 (right).
values of Hartmann number. We can see a decrease in axial velocity when Hartmann number
increases. So for a moderate value of Hartmann number (Ha ¼ 30), the profile of axial velocity
becomes increasingly linear. The regions located just under the rotating disk and just above the
bottom co-rotating disk correspond to a concentration of the electric current lines perpendicular
to magnetic field B. As a result, at these locations the Lorentz force is large. As it concerns heat
transfer, the magnetic field suppresses fluid motion and as a result, Nusselt number decreases with
increasing Hartmann number. This is shown in Figure 9b, where the Nusselt number distribution
for Ha ¼ 5, 10, and 20 with Ri ¼ 0.5 at Re ¼ 200 is compared. Thus, an increase in magnetic field
favors conduction heat transfer (see Kakarantzas et al. [11]).
For a positive temperature difference (Ri > 0) with the magnetic field, we show that the branch of
steady state with m ¼ 1 is the first to bifurcate and it is the dominant structure (see Figure 13). We
observed that increase in Richardson number resulted in a decrease in critical Reynolds number. It
was found also that the critical Reynolds number corresponding to these bifurcations increased with
increase in magnetic field. Magnetic field also has a significant effect on the heat transfer mechanism,
so an increase in Hartmann number (Ha ¼ 30) resulted in a damping of the fluid motion and thus
heat conduction progressively dominated over convection heat transfer (Ha ¼ 5) (see temperature
distribution in Figure 10) at Re ¼ 300 and Ri ¼ 2. The iso-potential surfaces are shown in Figure 11,
their signs being opposite at the top bottom locations on the axis, and again between the inner and
outer radial rims. Under the effect of the rotation of the two disks with symmetry of revolution being
connected to the axial component azimuth velocity, the electrical potential gradient has no azimuth
component. The space structure obtained of the potential for induction stabilizes with increasing
Hartmann number.
Figure 11. Iso-potential surfaces when Re ¼ 450, Ri ¼ 0.5 for values of Hartmann number Ha ¼ 5 (left) and Ha ¼ 20 (right).
NUMERICAL HEAT TRANSFER, PART A 11
Downloaded by [Library Services City University London] at 21:23 27 March 2016
Figure 13. Stability diagram in the (Recr Ha) plane for values of Richardson number Ri ¼ 0.5, 1, and 2.
12 B. MAHFOUD ET AL.
Table 2. Critical Reynolds number Recr with increase in Hartmann number Ha ¼ 0, 5, 10, and 20 for values of Richardson number
Ri ¼ 0, 0.5, 1, and 2.
Ri 0 0.5 1 2
Ha ¼ 0 1,920 292 221 162
Ha ¼ 5 2,348 409 292 202
Ha ¼ 10 3,156 582 387 258
Ha ¼ 20 5,997 877 561 386
Under the curve, we present the structure of axisymmetric contours of vertical velocity at meridional
plane r ¼ 0, which correspond to an axisymmetric solution (m ¼ 0). The plots of iso-v above the curve
for z ¼ 0 are given with respect to the order of appearance of the corresponding modes.
The stability diagram presented in Figure 13 shows, for Ri ¼ 0.5, 1, and 2, the evolution of critical
Downloaded by [Library Services City University London] at 21:23 27 March 2016
Reynolds number Recr, corresponding to the primary bifurcations to the mode m ¼ 1, as a function of
the Ha, followed by cuts (r, θ) in the structure of vertical velocity at z ¼ 0, showing the appearance of
azimuth stable modes. The plots of iso-v in the horizontal plane are given with respect to the order of
appearance of the corresponding modes. Under the curve, we present the structure axisymmetric con-
tours of axial velocity at horizontal plane (z ¼ 0). With Ri ¼ 0.5, 1, and 2, the mode m ¼ 1 becomes
dominant even when the magnetic field exceeds Ha ¼ 20. All the primary thresholds increase with
Hartmann number, confirming the phenomenon of stabilization of convection by the application
of the magnetic field. Note that the results correspond to the critical value of Reynolds number
(axisymmetric to non-axisymmetric flow), to values of Richardson number Ri ¼ 0, 0.5, 1, and 2, and
to Hartmann number Ha ¼ 0, 5, 10, and 20 (see Table 2).
5. Conclusion
Co-rotation flow in a cylindrical enclosure filled with a liquid metal under vertical temperature
gradient and an axial magnetic field was numerically studied. The finite-volume method was used
to numerically solve the transport equations. The effects of both magnetic field and heat transfer were
studied. The key results obtained in this study are as follows.
. The computer code developed in this study was validated with results from the literature, and
good agreement was obtained.
. It was observed that when Reynolds number is increased, the axisymmetric basic state loses its
stability to circular patterns of axisymmetric vortices and spiral waves.
. In mixed convection case, the present results show that the mode m ¼ 1 becomes the dominant
structure and increase in Richardson number results in decrease in critical Reynolds. Therefore,
an increase in Ri destabilizes the flow at the same values of Recr.
. It was shown that when the magnitude of the magnetic field (Ha) exceeds a certain value,
the steady three-dimensional flow resulting from the co-rotation mechanism tends to be
axisymmetric and heat conduction progressively dominates over convection heat transfer.
References
[1] S. Mori, K. Satoh, and A. Tanimoto, Electrolytic Mass Transfer Around Inclined Cylinders in Static Magnetic
Field, Electrochim. Acta, vol. 18, pp. 2789–2794, 1994.
[2] R. W. Series and D. T. J. Hurle, The Use of Magnetic Fields in Semiconductor Crystal of Growth, J. Cryst. Growth,
vol. 113, pp. 305–310, 1991.
[3] M. Yao, A. Chait, A. L. Fripp, and W. J. Debnam, Magnetically Damped Convection and Segregation in Bridgman
Growth of PbSnTe, J. Cryst. Growth, vol. 173, pp. 467, 1997.
[4] N. Ma and J. S. Walker, A Parametric Study of Segregation Effects During Vertical Bridgman Crystal Growth with
an Axial Magnetic Field, J. Cryst. Growth, vol. 208, pp. 757, 2000.
[5] Y. Jaluria, Fluid Flow Phenomena in Materials Processing—The 2000 Freeman Scholar Lecture, Trans. ASME: J.
Fluids Eng, vol. 123, pp. 173–210, 2001.
NUMERICAL HEAT TRANSFER, PART A 13
[6] G. M. Oreper and J. Szekely, The Effect of a Magnetic Field on Transport Phenomena in a Bridgman–Stockbarger
Crystal Growth, J. Cryst. Growth, 67, vol. 3, pp. 405–419, 1984.
[7] R. Bessaïh, M. Kadja, and Ph. Marty, Effect of Wall Electrical Conductivity and Magnetic Field Orientation on
Liquid Metal Flow in Geometry Similar to the Horizontal Bridgman Configuration for Crystal Growth, Int. J. Heat
Mass Transfer, vol. 42, pp. 4345–4365, 1999.
[8] J. M. Jalil and K. A. Al-Taéy, Thermovibrational Convection in an Enclosure with Magnetic Field Damping,
Numer. Heat Transfer A, vol. 53, no. 7, pp. 766–786, 2008.
[9] S. Bouabdallah, R. Bessaïh, B. Ghernaout, and A. Benchatti, Effect of an External Magnetic Field on the 3-D
Oscillatory Natural Convection of Molten Gallium during Phase Change, Numer. Heat Transfer A, vol. 60,
no. 1, pp. 84–105, 2011.
[10] R. Bessaïh and S. Bouabdallah, Numerical Study of Oscillatory Natural Convection During Solidification of a
Liquid Metal in a Rectangular Enclosure with and without Magnetic Field, Numer. Heat Transfer A, vol. 54,
no. 3, pp. 331–348, 2008.
[11] S. C. Kakarantzas, I. E. Sarris, A. P. Grecos, and N. S. Vlachos, Magnetohydrodynamic Natural Convection in
a Vertical Cylindrical Cavity with Sinusoidal Upper Wall Temperature, Int. J. Heat Mass Transfer, vol. 52,
Downloaded by [Library Services City University London] at 21:23 27 March 2016