International Journal of Heat and Mass Transfer 88 (2015) 113

Contents lists available at ScienceDirect

International Journal of Heat and Mass Transfer

journal homepage: www.elsevier.com/locate/ijhmt

MHD ow and heat transfer behind a square cylinder in a duct under

strong axial magnetic eld
Dipankar Chatterjee a,, Satish Kumar Gupta b
a
b

Simulation & Modeling Laboratory, CSIR Central Mechanical Engineering Research Institute, Durgapur 713209, India
Department of Mechanical Engineering, National Institute of Technology, Durgapur 713209, India

a r t i c l e

i n f o

Article history:
Received 27 November 2014
Received in revised form 15 April 2015
Accepted 18 April 2015
Available online 29 April 2015
Keywords:
Square cylinder
MHD ow
Heat transfer
Axial magnetic eld
Quasi two-dimensional model

a b s t r a c t
We capture through numerical simulation the MHD (magnetohydrodynamic) vortex dynamics around a
square object in a square duct subjected to a strong externally imposed axial magnetic eld. A quasi twodimensional conditionality allows us to follow a two-dimensional modeling approach. The pertinent
MHD control parameters such as the Reynolds and Hartmann numbers are kept in the range
0 < Re 6 6000 and 0 < Ha 6 2160, respectively. The various regimes of the MHD wake are found in-line
with those obtained by Dousset and Pothrat under similar conditions using a circular object (Dousset
and Pothrat, 2008). Four different regimes are identied out of which the rst three regimes are the classical non-MHD 2D cylinder wakes. The transition from one to another regime is controlled by the friction
parameter Re/Ha. The fourth regime is characterized by the vortices evolved from the duct side walls due
to the boundary layer separation which strongly disturbs the Krmn vortex street. In order to explore
the thermal transport phenomena under the action of the axially imposed magnetic eld, the channel
bottom wall is considered heated while the top wall is maintained at the free stream temperature keeping the cylinder adiabatic. The heat transfer rate from the heated channel wall strongly depends on the
imposed magnetic eld strength as well as Reynolds number. Additionally, an enhancement in heat
transfer is experienced by placing the square cylinder in the channel over the bare channel, however,
simultaneously a degradation in heat transfer would occur if the square object is replaced by a same size
circular object.
2015 Elsevier Ltd. All rights reserved.

1. Introduction
Magnetohydrodynamics (MHD) has been a subject of intense
research for a long time due to its overwhelming importance in
numerous elds ranging from several natural phenomena like
geophysics, astrophysics to many engineering applications such
as plasma connement, liquidmetal cooling of nuclear reactors,
electromagnetic casting and so on. A strong external magnetic eld
is known to suppress the velocity uctuations because of its damping nature. This causes a reduction in the turbulent intensity of the
MHD ow, thereby affecting the thermouidic transport signicantly. In applications like nuclear fusion reactors, the liquid metal
blankets (required to extract heat to control the reactor temperature) are prone to such magnetic damping phenomena. Hence,
there is a chance for the degradation of the heat transfer in such
engineering devices. The heat transfer can be augmented by promoting turbulence by placing obstacles inside the blankets. The
Corresponding author. Tel.: +91 343 6510455; fax: +91 343 2548204.
E-mail address: d_chatterjee@cmeri.res.in (D. Chatterjee).
http://dx.doi.org/10.1016/j.ijheatmasstransfer.2015.04.053
0017-9310/ 2015 Elsevier Ltd. All rights reserved.

obstacles may have different geometric shapes such as circular or

square cross sections. The thermouidic transport around the
obstacles will strongly depend on the shape of the obstacles. It
needs to be emphasized that the ow patterns and the wake structures for the cases of ow over a square object are signicantly different from that over a circular one because of the fact that unlike
the circular object the square counterpart xes the separation
points, causing differences in the critical ow regimes.
Furthermore, the separation mechanisms depending on the
shedding frequencies and the aerodynamic forces also differ
considerably for the two geometries. These technological and fundamental issues set the motivation for carrying out the study of
MHD ow of liquid metal around a square obstacle in a square duct
subjected to strong external magnetic eld.
A plethora of research has been carried out on the non-MHD
ows past circular and square cylinders and excellent reviews
are available in [14]. In case of unconned ow past circular or
square cylinders at low to moderate Reynolds numbers (Re), three
two-dimensional regimes can be identied before the ow
becomes three-dimensional at Re = 150190. At very low Re6 1,

D. Chatterjee, S.K. Gupta / International Journal of Heat and Mass Transfer 88 (2015) 113

Nomenclature
a
B
CD
CL
cp
d
f
FD
FL
Ha
Ld
Lu
k
N
Nu
p
Pe
Pr
Re
St
T
Tf

height of the duct

magnetic eld strength
drag coefcient
lift coefcient
specic heat
size of the cylinder
vortex shedding frequency
drag force
lift force
Hartmann number
downstream length
upstream length
thermal conductivity of uid
Stuart number
Nusselt number
dimensionless pressure
Peclet number
Prandtl number
Reynolds number
Strouhal number
temperature
bulk uid temperature

the ow is attached on the bodies resulting in a creeping ow

regime. Subsequent increase in Re witnesses the formation of
steady twin symmetric vortices behind the bodies up to
Re 6 40  50 signifying the twin vortex regime. Beyond that range
of Re, the ow around the bodies exhibits an oscillatory behavior,
instantaneous ow becomes asymmetric with respect to the
incoming ow, the twin-vortex arrangement becomes unstable
and a time-periodic oscillation of the wake develops, which signies the onset of vortex shedding. This regime of Krmn vortex
street formation persists up to Re = 150190 beyond which the
ow becomes three-dimensional. When the ow is bounded, apart
from Re, another parameter, blockage ratio (b, dened as the ratio
of the cylinder size to the transverse length of the domain) strongly
inuences the uid dynamic transport [510]. The presence of the
connement stabilizes the ow and the regimes for unconned
ow shifts to higher Re values. A further stabilization in the ow
eld may be experienced when a strong external magnetic eld
is applied to the ow of an electrically conducting uid. The presence of the external magnetic eld in an electrically conducting
uid gives rise to the induced electric and magnetic elds. Under
low magnetic Reynolds number approximation, the induced magnetic eld can safely be neglected and the total magnetic eld
becomes the applied one. The external magnetic eld interacts
with the electrically conducting uid to produce electromagnetic
Lorentz force which is basically damping in nature. The Lorentz
force tends to damp out the velocity variations along the streamlines [11]. When the ratio of the Lorentz to inertia forces and the
ratio of the square of the Lorentz to viscous forces are large, the
hydromagnetic ow becomes invariant along the magnetic eld
lines, except in the proximity to the walls that intersect them.
This region is characterized by the development of the classical
Hartmann layer where the viscous forces produce strong velocity
gradients. The resulting ow involving the Hartmann layer and
the core ow where the velocity is invariant along the magnetic
eld lines is quasi-two-dimensional in nature. Hence, the orientation of the magnetic eld with respect to the cylinder axis plays a
crucial role. The streamwise [1214] and transverse [1520] orientations have received considerable attention whereas the axial orientation is infrequent. Few experimental investigations are

u, v
w
x, y

dimensionless velocity components

width of the duct
dimensionless rectangular coordinates

Greek symbols

a
b

m
dH ; ds

q
r
X; x

thermal diffusivity
blockage ratio
kinematic viscosity
boundary layer thickness
dimensionless temperature
density
electrical conductivity
vorticity

Subscripts
av
c
m
w
1

average
critical
maximum
cylinder surface
free-stream

available pertaining to the axially oriented magnetic eld [21

26]. Additionally, some studies are available on the MHD duct ow
for straight [2729] and toroidal ducts [30,31]. Mck et al. [32] carried out three-dimensional numerical simulation of the MHD liquid metal ow around a square cylinder with axially oriented
magnetic eld in a rectangular duct for two xed Reynolds numbers Re = 200 and 250. The hydrodynamic simulations showed a
characteristic intermittent pulsation of the drag and lift forces on
the cylinder. At Re = 200 a mix of secondary spanwise three-dimensional instabilities (A and B mode, rib vortices) was observed
and at Re = 250 the ow appeared more organized showing a regular B mode pattern. It was also observed that with very high magnetic elds the time-dependent vortex shedding could be almost
completely suppressed. For full 3D numerical investigation, the
challenge is that for a strong magnetic eld, a very thin
Hartmann layer develops which requires very ne resolution of
mesh making the computer storage and CPU performance prohibitive. This problem can be handled by exploiting the advantage
of the quasi-two-dimensional approximation [33] that assumes
that the ow outside the Hartmann layers is nearly two-dimensional. Using the model in [33], Dousset and Pothrat [34] carried
out extensive numerical simulations for a quasi-two-dimensional
ow of a liquid metal in a square duct past a circular cylinder in
a strong externally imposed magnetic eld. Apart from identifying
various ow regimes evolved around the circular object, they have
also reported the dynamics of the high Re KelvinHelmholtz vortices. The transient response of optimal linear perturbations of liquid metal ow past a circular object under a strong axial magnetic
eld in an electrically insulated rectangular duct was studied by
Hussam et al. [35]. The study has specic signicance in designing
of potential actuation mechanisms to promote vortex shedding
and thus enhancing heat transfer in the duct.
From the above discussion, it is clear that MHD ows in ducts
have been a subject of intense research in the past. We cannot deny
the tremendous importance it has because of its strong applications in various technological processes such as hydrodynamic
generators, pumps, blood ow meters, fusion reactors and metallurgical processing. The ow of electrically conducting uid under
a strong magnetic eld is known to be characterized by a laminar

D. Chatterjee, S.K. Gupta / International Journal of Heat and Mass Transfer 88 (2015) 113

ow structure as velocity uctuations in the direction of magnetic

eld are strongly damped. This phenomenon can particularly be
witnessed in case of nuclear fusion reactors as outlined earlier
where extremely hot plasma volumes are kept away from the walls
of the reactor by an intense magnetic eld generated by a set of
superconducting magnetic coils [34]. Due to the aforementioned
phenomenon of damping of the velocity uctuations, a dramatic
decrease in the heat transfer rate can be observed. Placing of obstacles in the ow eld is known to promote turbulence which in turn
leads to the augmentation in heat transfer. The behavior of a cylindrical obstacle under such situation is reported [34,35], however, a
square obstacle needs some more attention. In fact, to the best of
our knowledge, the only study on ow of liquid metals past a
square cylinder with magnetic eld aligned to the cylinder axis
has been carried out in [32]. However, the heat transfer aspects
were not taken into consideration. In a recent article, Hussam
et al. [36] proposed an approach for analyzing the side-wall heat
transfer enhancement in the magnetohydrodynamic ow of uid
in a rectangular duct that is damped by a strong transverse magnetic eld. The proposed mechanism employs the rotational oscillation of a circular cylinder placed inside the duct to encourage
vortex shedding, which promotes the mixing of uid near a hot
duct wall with cooler uid in the interior. A considerable increase
in the heat transfer from the heated channel wall due to rotational
oscillation of the cylinder was reported. In spite of that, the literature lacks a comprehensive parametric study that would reveal the
magnetohydrodynamic interaction for high values of Re and
Hartmann (Ha) numbers in liquid metal ows and heat transfer
around a square obstacle subjected to axial magnetic eld.
Furthermore, investigation of the various ow regimes and their
detailed description remains as an undisclosed proposition in the
present context. Accordingly, we study numerically the magnetohydrodynamic ow and heat transfer of liquid metal in a square
duct in presence of a square cylinder subjected to strong axial magnetic eld for a constant blockage ratio b 0:25. We report the
ow regime found with the help of extensive numerical simulations for Re up to 6000 and Ha in the range of 160 6 Ha 6 2160.
The forced convection heat transfer characteristics considering a
heated bottom channel wall are also analyzed under the prevailing
conditions.

2. Physical problem and mathematical formulation

The problem under consideration is shown schematically in
Fig. 1. We consider the ow of an electrically conducting, incompressible uid (an eutectic alloy GaInSn) in a duct of square cross
section past a square cylinder. The duct walls and the cylinder
are assumed to be electrically insulating. The cylinder is assumed
to be thermally insulated as well. The cylinder axis is at the centre
of the duct, and its axis is parallel to the side walls and orthogonal
to the streamwise direction. The bottom wall (in the equivalent 2D problem, refer to Fig. 1(b)) is considered heated (T w ) while the
top wall (Fig. 1(b)) is kept at the free stream temperature
T 1 < T w . The duct dimensions (a = w = 0.04 m) and the cylinder
size (d = 0.01 m) are so chosen that it produces a blockage ratio
b d=w 0:25. The upstream Lu and downstream Ld lengths
chosen for the present study are 12d and 42d respectively. These
values are in conformity with [34]. A steady homogeneous magnetic eld B with intensities between 0 and 1.35 T is imposed along
the cylinder axis.
In this study, the magnetic Reynolds number is considered
small which implies that the induced magnetic eld is negligible
and the resulting magnetic eld is the one that is imposed in the
z-direction (direction parallel to cylinder axis). Under such conditions, the ratios of the Lorentz force to inertia (Stuart number)

(a)

(b)
Fig. 1. (a) Actual conguration of the problem. (b) Equivalent 2D problem.

N raB2 =qum (um is the peak inlet velocity) and to viscous forces
Ha2 a2 B2 r=qm are large. Thus, the Lorentz force strongly damps
the velocity variations along the direction of B except in the vicinity of the walls normal to B. In these regions, viscous forces oppose
the Lorentz ones and maintain a strong velocity gradient. Thin
boundary layers called Hartmann layers thereby develop along
these walls and are known to have a characteristic thickness
dH a=Ha. Another boundary layers develop along the walls parallel to B known as Shercliff layers which are intrinsically 3D with a
p
characteristic thickness ds a= Ha. However, the three-dimensionality in the Shercliff layers is of little consequence on the
quasi-two-dimensional ow [34] since quasi-2D approximation
would involve a local error within 10% [37]. This error is signicant
only in the vicinity and inside the Shercliff layers.
With these justications, the quasi two-dimensional model proposed by Sommeria and Moreau [33] is used here which is derived
by averaging the ow equations along the direction of the magnetic eld. In this model, called thereafter SM82 model, the core
ow is perfectly two-dimensional, the Hartmann layers are modeled by the exponential prole and walls are electrically insulating.
SM82 model neglects the inertial effects occurring inside the
Hartmann layers so that this model is accurate down to the order
OHa1 ; N 1 . When applied to the Shercliff ow, the three-dimensional nature of the Shercliff layers is not specically addressed by
the SM82 model, as it assumes that diffusion along the eld lines is
an order of magnitude faster than the lateral diffusion of the angular momentum. Besides, the SM82 model requires electrically insulating perpendicular walls, as the assumption of a quasi-2D ow
fails if strong velocity jets at the lateral extremities of the
Hartmann layers are present. Finally, this model assumes that no
uid transfer exists at the interface between the Hartmann layers
and the core ow. As a result, Ekman-like recirculation ow, which
is the basic formation mechanism of cyclones for example, is forbidden in the SM82 model. In addition to this there are important
and more subtle constraints on the validity of the model. These
arise from inspection of the relevant timescales (which is concisely
described in Pothrat [38]). Essentially, quasi-two-dimensionality
is achieved when the timescale for the Lorenz force to diffuse
momentum of a structure of size l? along magnetic eld lines over
a length, lll ,

s2D rqB2 ll2ll , is shorter than the timescales for viscous

?

diffusion in the perpendicular and parallel planes (s?

m l? =m and

D. Chatterjee, S.K. Gupta / International Journal of Heat and Mass Transfer 88 (2015) 113

sllm l2ll =m, respectively), and the inertia timescale su l? =U (U is

3. Numerical method and verication

the base ow). The rst condition obtains a limiting length scale
on perpendicular structures dictated by l? =a > Ha1=2 (i.e. the
Shercliff layer scaling), and this constraint persists for all N. This
is important because in a conned ow, local accelerations around
the cylinder may create boundary and shear layers with scales of
this order or even smaller, so locally the SM82 model may not be
strictly valid. However, provided Ha  1 and N  1, the implementation of the SM82 model in a numerical code is an effective
and accurate method to investigate the ow dynamics [33]. The
involved CPU cost is much lower compared to 3D direct numerical
simulations.
Since the present study aims to investigate the effects of the
imposed magnetic eld on the forced convective transport around
a cylinder, the effect of buoyancy is ignored. Additionally, we
assume incompressible uid with constant thermophysical properties. The dimensionless governing equations in absence of any
phenomenological cross effects and neglecting viscous dissipation
and Joule heating can be prescribed as:

r:u 0

The conservation equations subjected to the aforementioned

boundary conditions are solved using a nite volume based CFD
solver Ansys Fluent (version 14.0) [39]. Ansys Fluent solves the governing system of partial differential equations in a collocated grid
system by constructing a set of discrete algebraic equations with
conservative properties. The numerical scheme solves the discretized governing equations sequentially. The sequence updates
the velocity eld through the solution of the momentum equations
using known values for pressure and velocity. Then, it solves a
Poisson-type pressure correction equation obtained by combining
the continuity and momentum equations. The QUICK (Quadratic
Upstream Interpolation Convective Kinetics) scheme is used for
spatial discretization of the convective terms and a central
difference scheme is used for the diffusive terms. PISO (PressureImplicit with Splitting of Operators) algorithm [40] is used as the
pressurevelocity coupling scheme which is known to perform
two additional corrections (neighbor and skewness corrections) in
order to improve the efciency of the calculation. A description in
regard to the implementation of the PISO algorithm in the present context is available in [41]. Another description of handling
2

@u
1
d Ha
u:ru rp r2 u  2 2
u
@t
Re
a Re

@H
1
u:rH r2 H
Pe
@t

where u, p and H are the dimensionless velocity, pressure and

temperature elds, respectively, projected onto the xy plane, t is
the dimensionless time, Re um d=m is the Reynolds number,
p
Ha aB r=qm is the Hartmann number and Pe Re Pr is the
Pclet number with Pr m=a being the Prandtl number
(a k=q cp is the thermal diffusivity). The non-dimensional form
of the model is obtained by scaling the lengths by the cylinder size
d, pressure by qu2m , time by d=um and temperature by DT T w  T 1 .
The uid properties are described by density q 6360 kg m3 ,
kinematic viscosity
1

k 16:5 Wm

m 3:4  107 m2 s1 , thermal conductivity

1

K , specic heat cp 152:6 J kg

6

1

1

K1 and electri-

cal conductivity r 3:46  10 X m . The Prandtl number is

obtained as Pr = 0.02. The ow is assumed to start impulsively from
rest with a velocity um everywhere in the domain.
At the inlet (x Lu ) the boundary condition is derived as the
exact parallel ow solution of Eqs. (1) and (2) for the duct ow
problem without the obstacle given by:
1

 p 
 p
3
cosh yd 2Ha=a  cosh h 2Ha=2a
5
 p

uLu ; y um 4
1  cosh h 2Ha=2a

the Hartmann damping term in Eq. (2) 2d=a Ha=Reu is

available in [34]. A suitable dimensionless time step
size dt 2  103 is used in the computation satisfying the
CourantFriedrichsLewy (CFL) and the grid Fourier criteria.
Finally, the algebraic equations are solved by using the Gauss
Seidel point-by-point iterative method in conjunction with the
Algebraic Multigrid (AMG) method solver. The convergence criteria
based on relative error for the inner (time step) iterations are set as
106 for the discretized equations.
A non-uniform structured grid distribution having a close clustering of grid points in the vicinity of the cylinder wall is used in
the present computation. Fig. 2 shows a representative grid distribution in the vicinity of the cylinder. A very high grid density is
adopted in the proximity of the object to capture the phenomena
of ow suppression at the rear stagnation zone. Also in order to
capture the Shercliff layers close clustering of grids is used near
the walls. 100 divisions on each side of the square object are used
with the rst cell height from the cylinder as 0.513469 and
0.00366 from the duct wall. Altogether, 154,125 cells (155,210
nodes) are used to discretize the entire computational domain.

The reference velocity um is the maximum of the velocity prole

imposed at the inlet boundary at y = 0. The exit boundary (x Ld ) is
located sufciently far downstream from the region of interest
hence an outow boundary condition is applied at the outlet. The
outow boundary condition is modeled by a homogeneous
Neumann condition for the velocity and temperature variables
as: @ n ujoutlet 0 and @ n Hjoutlet 0. No slip boundary conditions
are imposed on the side walls and on the cylinder surface
(ujsolid wall 0). Temperatures at the inlet and top wall are set at
H T  T 1 =T w  T 1 0 and on the bottom wall H 1. The
cylinder is thermally insulated (i.e. a zero normal temperature gradient is imposed on the cylinder surface, @ n Hjcylinder 0). Pressure
boundary conditions are not explicitly required since the solver
extrapolates the pressure from the interior.

Fig. 2. Closer view of the grid in the proximity of the cylinder.

D. Chatterjee, S.K. Gupta / International Journal of Heat and Mass Transfer 88 (2015) 113
Table 1
Grid sensitivity analysis showing characteristics of various meshes and errors in average drag coefcient and Strouhal number at Re = 100.
Meshes

M1

M2

M3

M4

M5

Total number of nodes on the surface of the cylinder

Number of nodes in the Shercliff layers at the top and bottom walls for Ha = 2160
Total nodes in the domain
eC D j1  C D Mi=C D M5j
eSt j1  StMi=StM5j
eNu j1  NuMi=NuM5j

220
12
42,400
0.058
0.006
0.062

280
16
71,600
0.046
0.009
0.052

340
20
112,000
0.0032
0.00024
0.044

400
24
155,210
0.0030
0.00022
0.042

460
28
196,200

In an effort to numerically verify the adopted computational

scheme, calculations are performed rst to obtain the critical
Reynolds numbers for the transition from steady to unsteady
regime for non-MHD conned ow over a square cylinder. Fig. 3
is plotted in this regard to show the critical Re as a function of
the blockage ratio. The results are compared with the available
results in the literature [42] and it is observed that the numerical
scheme adopted here correctly reproduces the stability properties
of the method. In another attempt, we compare the average drag
coefcient and RMS lift coefcient (the lift is given as
C L 2F L =qu2m d with F L being the lift force acting on the cylinder)
obtained from the numerical simulation in [32] for the MHD ow
of liquid metal around a square obstacle in a duct with the present
numerical technique. The comparison is shown in Fig. 4 with no
signicant discrepancies between the two sets of results.
However, at weak magnetic elds slight discrepancy can be witnessed which is attributable to the fact that the two-dimensional
Fig. 3. Critical Reynolds number for the onset of unsteady ow as a function of
blockage ratio b. The comparison is made with the numerical results of Patil and
Tiwari [42].

A comprehensive grid sensitivity analysis is also carried out to

understand the self consistency of the problem. Table 1 shows
the results of the grid sensitivity analysis. A long-time simulation
(t 400 t d , with td d=um ) is performed at Re = 100 for each mesh
size adopted. We compute the drag coefcient (CD) and the dimensionless vortex shedding frequency (Strouhal number, St) for each
case using the following denitions:

2F D

CD

qu2m d

St

fd
um

where F D is the drag force acting on the cylinder and f is the vortex
shedding frequency. The drag coefcient is nally time and surface
averaged. Next, we compute the local Nusselt number on the heated
bottom channel wall as

Nuw x; t


d
@ H
T f  T w @y wall

where T f is the bulk uid temperature calculated from

Tf

Z
u H dy

u dy

with u being the streamwise velocity component. Finally, the time

and surface average Nusselt number is obtained by rst taking
the time average of the local Nusselt number Nuw at each x location, and then integrating over the length of the heated bottom wall
L Lu Ld as

Nu

1
L

Nuw x dx

Next, we obtain the relative errors eC D , eSt and eNu in calculating

respectively CD, St and Nu with respect to the M5 mesh. The relative errors are observed to decrease with increasing nodes.
Finally, type (M4) mesh is preferred keeping in view the accuracy
of the results and computational convenience in the simulations.

Fig. 4. Comparative evolutions of CD (top) and CL RMS (bottom) vs. N. The

comparison is made with the numerical results of Muck et al. [32] (b = 0.1) for
Re = 200.

D. Chatterjee, S.K. Gupta / International Journal of Heat and Mass Transfer 88 (2015) 113

transition occurs at N  1 below which the ow eld is three-dimensional. Hence, the quasi-two-dimensional model overpredicts
at low magnetic eld strength. Fig. 4 conrms the validity of the
present numerical method for the MHD ow around obstacles.
The validations in regard to the heat transfer predictions for nonMHD ow around square object can be found in [43].
4. Results and discussion
4.1. Flow regimes
The unbounded non-MHD ow around a square cylinder may
exhibit transition to three-dimensionality from the two-dimensional Karman vortex street at Reynolds number in the range
150175 [44,45]. The blockage in case of bounded ow provides
additional stability which may delay that transition. However, no
comprehensive study to date exists for such data. When the ow
eld is severely perturbed by some strong external magnetic eld,
the velocity uctuations are damped resulting in more stabilized
ow in case both the conning walls and the magnetic eld act
in tandem. Under the action of the magnetic eld, the wake behind
the cylinder gets narrowed in the fashion as presented in Fig. 5.
Additionally, the distance between the vortices in the main ow
direction gets reduced and the vortices become organized in rows
travelling downstream.
A series of numerical computations are performed xing the
blockage ratio b 0:25 and taking B = 0.1, 0.2, 0.4, 0.7, 1.0, 1.35 T
corresponding to the Hartmann numbers Ha = 160, 320, 640,
1120, 1600, 2160 to compute the critical Reynolds numbers for
transitions from creeping ow to twin vortex (ReIc ) to primary vortex shedding regime (ReIIc ) and nally to the regime where secondary vortices are released (ReIII
c ). In order to ensure that the
quasi-two-dimensionality approximation remains valid, we intentionally choose N  1 (say, N = 10). Accordingly, we start from
Ha = 160 which is consistent with the choice of N.
The ow regime map is presented in Fig. 6 with clear identication of four different regimes evolved as an outcome of the
imposed magnetic eld. For a given magnetic eld strength (Ha)
as Re gradually increases, four different regimes evolve. The rst
three regimes are identical to the 2D non-MHD ow regimes, i.e.
the creeping ow regime I, the twin vortex regime II and the laminar periodic regime with regular Krmn vortex shedding III (refer
to Fig. 7). The fourth regime IV is different from the non-MHD ow
regime where the regular Krmn vortex street becomes irregular.
Special attention must be given to describe regime IV in an elaborated fashion.
As demonstrated in Fig. 8, the regime IV is characterized by the
formation of the regular Krmn vortices (Kc and Kac) from the
rolling-up of the free shear layers as in regime III along with the
formation of the secondary counter-rotating vortices (S1, S2 and
S3) from the separation of the Shercliff layer at the side walls.
These secondary vortices either cross the downstream wake obliquely and interact strongly with the adjacent Krmn vortices or
quickly dissipate as soon as they detach from the Shercliff layer
[34]. The phenomena can be better visualized from the movie clip
provided as a supplementary material [46]. The oblique trajectory
of the secondary vortices evolves from a combined action of the
free stream that drives them away downstream and of the regular
Krmn vortices at their origin of formation, those press them
towards the opposite side walls. As a consequence to this, rstly,
the Krmn vortices dissipate a large amount of energy during
the formation of the secondary vortices and the subsequent interaction with them. This lost energy reaches downstream to further
maintain the periodic vortex street. As a second consequence, the
formation process of the secondary vortices disturbs the Krmn

Fig. 5. Effect of magnetic eld on the wake of square cylinder, (a) no magnetic eld,
(b) with magnetic eld.

vortex street and this causes the Krmn vortices to become irregular rather than a regular procession of vortices. The resulting
chaotic vortex street is found to oscillate from one wall to the
other. Furthermore, since the regime IV results from the inuence
of the conning walls, the blockage ratio b is surely playing a crucial role in the transition to regime IV for a given Ha. With higher
degree of connement (larger b) is expected to yield a lower ReIII
c
[34].
Since regime IV is an outcome of the magnetic eld effect showing distinctly different behavior from the non-magnetic situation,
it is worthwhile to explore the ow dynamics in a greater detail
in absence of the magnetic eld in regime IV. However, the nonmagnetic ow characteristics strongly depend on the degree of
connement and the ow Reynolds number. In general, it is
observed that for b < 0:1, ow connement has little inuence
on the ow dynamics. For the range 0:1 < b < 0:6, the acceleration
due to ow connement causes noticeable changes in the ow.
Flow separation occurs at higher Re. At some specied Re, increasing b pushes downstream the point where the free shear layer separates from the cylinder surface. Furthermore, the connement
causes stabilization of the free shear layers. Accordingly, the oscillations along the transverse direction which initiates the onset of
vortex shedding is impeded by the presence of the conning walls.
Hence, as shown in Fig. 3, the transition to unsteadiness is shifted
to signicantly higher Re as b is increased. At somewhat larger
Reynolds number (e.g., Re P 200), the development of the
Krmn vortices initiates the separation of the side wall boundary
layers and the evolved secondary vortices then interact with the
Krmn vortices. This regime basically corresponds to the regime
IV of the MHD case. For a visual appreciation, we plot the nonMHD ow elds in Fig. 9 for representative Reynolds numbers,
Re = 500 and 1000 at the blockage b 0:25. Fig. 9 reveals that
the wake is asymmetric and a strong dominance of the vorticity
shed from the side farther from the wall is noticeable. Comparing
Fig. 7a (Re = 1000, Ha = 320) and Fig. 9b (Re = 1000, Ha = 0), it can
be noticed that a rather weak kind interaction between the secondary vortices emanating from Shercliff layer on the side walls
and the regular Krmn vortices can be observed for the MHD situation, whereas, a relatively strong interaction between the vortices originating from the conning walls and the Krmn vortex
is evident for the non-MHD situation.
In order to understand the differences in the MHD ow characteristics over a circular cylinder under similar conditions, the
square cylinder results are compared with those reported in Ref.
[34] for the circular cylinder. For that purpose Fig. 10 is reproduced
from Ref. [34]. Comparing Figs. 8 and 10, it can be understood that
there is no qualitative difference between the MHD ow characteristics around circular and square cylinders under the prevailing
conditions. However, the quantitative difference is observed to
be very much signicant. It is observed that the transition
Reynolds numbers for various regimes are much higher in case of

D. Chatterjee, S.K. Gupta / International Journal of Heat and Mass Transfer 88 (2015) 113

2000

Rec

N = 10
1500

IV

1000

III

500

II

500

1000

1500

2000

2500

Ha

Fig. 6. Flow regime map: sector I is the creeping ow regime, sector II represents the steady symmetric attached recirculation regions, sector III is the laminar periodic ow
regime with the regular Krmn vortex street formation, and sector IV is the ow regime where secondary vortices are emanated from the side walls. ReIc < ReIIc < ReIII
c are the
successive critical thresholds between the ow regimes.

(a)

the corresponding non-MHD cases. In order to understand the

damping nature of the imposed magnetic eld it is instructive to
look into Eq. (2) where the exponential growth rate of the ow
instabilities is shifted by Ha/Re through the linear damping action
of the Hartmann layers. Hence, a higher Re is required to effectively
reach the transition to another ow regime compared to the nonMHD cases.
4.2. Aerodynamic responses

(b)

(c)

The time history of the lift coefcient signal of the cylinder for
different Hartmann numbers and at a representative Re = 1000 is

(a)

Fig. 7. Dimensionless vorticity contours at an arbitrary time instant and Re = 1000

for different Hartmann numbers (a) Ha = 320, (b) Ha = 1120 and (c) Ha = 2160. In (a)
regular Krmn vortex street becomes irregular (regime IV), (b) regular Krmn
vortex shedding is observed (regime III) and (c) ow becomes steady with twin
vortex formation (regime II). The increasing magnetic eld strength stabilizes the
ow resulting in the suppression of vortex shedding. Vorticity elds: light and dark
gray contours show positive and negative vorticity, respectively.

(b)
circular cylinder in comparison to its square counterpart. This is
attributable to the existence of the sharp corners in case of the
square cylinder which promotes hydrodynamic instability
[45,47]. On the contrary, the smooth surface in case of the circular
object provides better stability to the ow. Hence, there is a further
requirement of larger inertial destabilizing forces to make the ow
unstable for the case of a circular cylinder. Consequently, the transition Reynolds number increases in the case of circular cylinder in
comparison to its square counterpart. Such inuence of the geometric shape on the ow transition can also be visible in case of
non-MHD and unbounded ow. The physical boundary (channel
wall) and the imposed magnetic eld provide additional stability
to the ow (when compared with unbounded non-MHD ow) for
which the transition further delays.
The regime diagrams presented above suggest that the imposed
magnetic eld due to its damping nature actually shifts the appearance of the ow instabilities to higher Re values and the resulting
ow regimes span over a wider range of Re than in comparison to

(c)

Fig. 8. Regime IV: dimensionless vorticity contours at different dimensionless time

instances (ac: incrementing the dimensionless time by 20) for Re = 2000, Ha = 640.
S1, S2 and S3 are secondary vortices, Kc and Kac are the clockwise and
anticlockwise Krmn vortices.

D. Chatterjee, S.K. Gupta / International Journal of Heat and Mass Transfer 88 (2015) 113
8

Re = 1000

(a)

Ha = 320
Ha = 640
Ha = 1120
Ha = 2160

(b)
Fig. 9. Dimensionless vorticity contours at Re = (a) 500 and (b) 1000 for Ha = 0 at an
arbitrary time instant. Vorticity elds: light and dark gray contours show positive
and negative vorticity, respectively.

CL

-2

shown in Fig. 11. The lift signals show time dependent oscillation
for Ha = 320, 640, 1120 and 1600 which is an outcome of the periodic ow feature. Additionally, we observe that the amplitude of
oscillation decreases gradually as we move towards higher magnetic eld strength. Finally, for Ha = 2160 no oscillation in the lift
signal is detected since the vortex shedding is completely suppressed at this magnetic eld strength.
The respective Fourier spectra of the lift coefcient signals presented in Fig. 11 are plotted in Fig. 12. For lower Ha values
(Ha = 320, 640), the ow is clearly in regime IV which is characterized by the formation of the regular Krmn vortices along with
the formation of the secondary counter-rotating vortices from
the separation of the Shercliff layer at the side walls. As a result
of the interaction of these vortices, secondary frequencies originate
simultaneously with the primary vortex shedding frequency.
Hence, multiple frequencies are observed in the spectra. The secondary frequencies gradually disappear when the magnetic eld
strength progressively increases. The resulting ow shifts towards
regime III and a single frequency persists at Ha = 1600 signifying a
highly periodic and uniform vortex shedding. It is further to be
noted that the spectrum energy gradually reduces as the magnetic
eld strength increases for the xed Reynolds number (Re = 1000).
Fig. 13 shows the variation of the Strouhal frequency with
Hartmann number for a representative Reynolds number
Re = 1000. The frequency is observed to drop to zero in regime II
where the ow becomes steady. Another drop in the frequency
can be observed in regime IV. This is attributable to the fact that

6000

IV
5000

-4

-6

40

50

60

70
Time

80

90

100

Fig. 11. Time history of the lift coefcient for different Hartmann numbers and at
Re = 1000.

in regime IV a secondary vortex develops in the Shercliff layer, it

sheds and begins to cross the wake during its downstream motion,
and creates an obstacle that impedes the ow of the incoming
Krmn vortices. As a consequence, the vortex shedding frequency
of the latter vortices decreases, leading to the observed drop in the
Strouhal number. In regime III, as the formation of the secondary
vortices occurs downstream of the cylinder, it does not affect that
of the Krmn vortices.
The effect of the magnetic eld on the time average drag coefcient is represented in Fig. 14. It is observed that CD rst decreases
then increases with an increase in the parameter Re=Ha4=5 . Dousset
and Pothrat [34] demonstrated for a circular cylinder that at low
values of Re=Ha4=5 , data collapses onto a universal curve. Similar
characteristic feature can also be observed for the present case
with the square cylinder. At higher Re=Ha4=5 the drag coefcient
increases, and displays a dependence on the Hartmann number.
However, when compared with the corresponding results in the
case of a circular cylinder under similar conditions as explored in
Ref. [34], the drag is found to be higher in the case of a square
cylinder compared to its circular counterpart under the prevailing
conditions.

N = 10

4.3. Heat transfer

Rec

4000

III

3000

2000

1000

II
I
0

500

1000

1500

2000

2500

Ha

Fig. 10. Flow regime map for a circular cylinder as reported in [34] (reproduced
with permission).

Fig. 15 presents the instantaneous dimensionless temperature

contours for different Hartmann numbers (Ha = 320, 1120 and
2160) at Re = 1000. At this Reynolds number for Ha = 320 and
1120 the ow dynamics respectively falls in regime IV and III
whereas for Ha = 2160 the ow dynamics corresponds to regime
II. Accordingly, for the lower magnetic eld strength Ha = 320
and 1120, the temperature elds are found time dependent since
the ow is unsteady periodic with vortex shedding behind the
cylinder. However, with further increase in the magnetic eld
strength (Ha = 2160), the damping action of the magnetic eld
reduces the ow velocity near the heated wall resulting the suppression of the unsteadiness in the ow. The ow becomes steady
at this Ha. This causes increase in the thickness of the thermal
boundary layer and hence the temperature gradient along the
heated wall decreases. Similar observations in regard to the MHD
thermo-convective transport in case of ow of liquid metal past

D. Chatterjee, S.K. Gupta / International Journal of Heat and Mass Transfer 88 (2015) 113
4000

3500

3500

3000

Ha = 320

3000

Ha = 640
2500

F{CL}

F{CL}

2500

2000

2000

1500

1500
1000

1000

500

500

10

-1

10

10

10-1

100

St

St

(a)

(b)

500

101

2.0

400

Ha = 1120

Ha = 1600

1.5

F{CL}

F{CL}

300
1.0

200

0.5
100

10

-1

10

10

0.0

10

-1

10

10

St

St

(c)

(d)

Fig. 12. Fourier spectra of lift coefcient at Re = 1000 for different Hartmann numbers, (a) Ha = 320, (b) Ha = 640, (c) Ha = 1120 and (d) Ha = 1600.

a circular cylinder placed centrally within a bottom heated channel

are made in Ref. [15]. Hence, there is qualitative similarity in thermal transport characteristics for MHD ow past circular and square
cylinders in heated channel. In Ref. [15], the inuence of blockage
ratio on the MHD thermo-uid-dynamics was also investigated
and it was observed that with increasing blockage the heat transfer
increases remarkably.
In order to delineate the special features of the heat transfer
characteristics in the regime IV, particular emphasis is given here
to analyze the inuence of secondary vortices on the heat transfer
phenomena. As discussed in Section 4.1, the secondary vortices
either interact strongly with the regular Krmn vortices or quickly
dissipate as soon as they detach from the Shercliff layer. Both the
cases result in the release of a large amount of energy which is
eventually transported downstream. This additional energy comes
specically from the secondary vortices originated from the
Shercliff layer on the side walls. As a consequence of this evolved
energy the thermal eld becomes irregular and chaotic in regime
IV. This would in turn cause more heat transfer to take place from
the heated wall in regime IV in comparison to other regimes. In an
effort to portray the heat transfer characteristics in regime IV when

there is no magnetic eld, Fig. 16 is plotted to show the instantaneous thermal elds at Re = 500 and 1000 for Ha = 0. Owing to
the relatively strong interaction between the secondary vortices
from the wall and the Krmn vortices in case of the non-magnetic
situation, the thermal elds become more irregular and chaotic
than the corresponding MHD situation. This eventually results a
relatively higher heat transfer rate from the heated wall in case
of the non-magnetic situation in comparison to the magnetic situation. The time and surface average Nusselt numbers from the
heated bottom wall at Ha = 0 for Re = 500 and 1000 are respectively
1.2478 and 1.8592, whereas the corresponding values for Ha = 320
are 1.194 and 1.7387. Hence, the non-magnetic situation produces
more heat transfer.
Fig. 17 shows the distribution of the local Nusselt number at an
arbitrary time instant along the heated wall as a function of the
streamwise coordinate for different Hartmann numbers at
Re = 1000. The effect of Hartmann number on the heat transfer rate
from the bottom heated wall can only be visible behind the cylinder. Upstream of the cylinder the ow is not changing with
Hartmann number (refer to Fig. 15) and hence the heat transfer
does not change. Downstream of the cylinder the ow shows

10

D. Chatterjee, S.K. Gupta / International Journal of Heat and Mass Transfer 88 (2015) 113

0.35

Re = 1000

(a)
0.30

(b)
0.25

III

St

(c)
II

IV

0.20

(d)
Fig. 15. Dimensionless temperature contours at an arbitrary time instant and
Re = 1000 for different Hartmann numbers Ha = (a) 320, (b) 640, (c) 1120 and (d)
2160. Temperature elds: dark and light contours show cold and hot uid,
respectively.

0.15

0.10

500

1000

1500

2000

Ha
Fig. 13. Strouhal frequency as a function of Hartmann number for Re = 1000.

substantial dependence on the Hartmann number. The shedding

phenomena causes the Nusselt number to uctuate along the
heated wall as can be observed for Ha = 320 and 1120. With
increasing magnetic eld strength the ow instability is suppressed and accordingly the amplitude of uctuation of the
Nusselt number decreases and at higher Hartmann number (e.g.,
Ha = 2160) when the ow becomes completely steady there is no
uctuation in the Nusselt number as can be observed from
Fig. 17. The inset of Fig. 17 shows the time average Nusselt number
values along the heated wall for the unsteady cases (Ha = 320 and
1120). The heat transfer progressively diminishes along the heated
wall.
The variation of the time and surface average heat transfer from
the heated bottom wall with Reynolds number for different
Hartmann numbers is presented in Fig. 18. The average Nusselt
number increases with Reynolds number and with increasing
Hartmann number it decreases. Similar increasing nature of the

20

Ha = 320
Ha = 640
Ha = 1120
Ha = 2160

CD

15

10

Nusselt number from the heated bottom wall with Reynolds number was observed by Hussam et al. [15] and Hussam and Sheard
[19] in their studies involving MHD ow past an adiabatic circular
cylinder within a bottom heated channel. The heat transfer is
found to be higher in the shedding ow regimes compared to the
steady state regime and this difference is better noticeable at larger
Reynolds numbers. With increasing Reynolds number, the ow
velocity near the heated wall increases. Accordingly, the cold uid
is transported toward the hot region of the channel and the hot
uid near the heated wall is convected away to mix with the cold
uid. Consequently, the heat transfer is enhanced at larger
Reynolds number. The Hartmann number dependence on the heat
transfer is noticeable at larger Reynolds number. At larger
Hartmann number, thicker thermal boundary layer develops which
causes the temperature gradient to decrease resulting in a corresponding decrease in the Nusselt number.
What happens when the square object is replaced by a circular
one in the channel under the prevailing conditions? We have
already demonstrated the qualitative similarities in the ow
dynamics for the two said geometries through numerical computation. Here we intend to nd out certain prospective differences
quantitatively. Table 2 is presented in this regard to show the time
and surface average Nusselt number for the two geometries at
some representative Reynolds and Hartmann numbers combination (Re = 2000, 3000 and Ha = 500, 1200). The circular cylinder
results are obtained from Ref. [15]. It is observed from Table 2 that
the circular object as a heat transfer promoter is preferable than a
square object as long as the imposed magnetic eld strength is
moderate. However, in presence of larger magnetic eld strength
both promoters act almost equivalently.
Finally, in order to understand the heat transfer enhancement
mechanism due to the addition of the square object in the channel,
Fig. 19 is plotted. Fig. 19 shows for a representative Hartmann
number (Ha = 320) and different Reynolds numbers the time and
surface average Nusselt number on the hot wall of the channel
for the cases when there is no object (Nu0 ) in the channel and a
square object as studied in the work. The overall increment in
the heat transfer is dened as:

DH
0

10

Re/Ha

4/5

10

Fig. 14. Time average drag coefcient as a function of Re=Ha4=5 .

Nu  Nu0
 100
Nu0

Fig. 20 shows the % heat transfer increment for different

Reynolds numbers for the xed Hartmann number. The presence
of the cylinder causes more heat transfer to take place and as the
Reynolds number increases this increment in heat transfer

11

D. Chatterjee, S.K. Gupta / International Journal of Heat and Mass Transfer 88 (2015) 113

Table 2
Quantitative comparison of the heat transfer prediction from the heated bottom wall
for the cases with circular and square objects in the channel (b 0:25).

(a)

Nuav
Ha = 500

(b)
Fig. 16. Dimensionless temperature contours at an arbitrary time instant and
Re = (a) 500 and (b) 1000 for Ha = 0. Temperature elds: dark and light contours
show cold and hot uid, respectively.

Re = 2000
Re = 3000

Ha = 1200

Square

Circular [15]

Square

Circular [15]

2.54
3.22

3.16
3.64

2.50
2.96

2.52
2.99

5.5
5.0

20

Ha = 320
Ha = 1120
Ha = 2160

18

Nu

Ha = 320

Nu0
Nu

4.5
4.0

14

3.5

Nu

16

12

3.0

10

2.5

2.0

1.5

1.0
1000

2000

3000

2
0

4000

5000

6000

Re
-10

10

20

30

40

Fig. 19. Variation of time and surface average Nusselt number on the heated
bottom channel wall with Reynolds number for Ha = 320, without cylinder (open
symbol, Nu0) and with cylinder (lled symbol, Nu).

Fig. 17. Distribution of the local Nusselt number at an arbitrary time instant on the
heated bottom wall of the channel as a function of streamwise coordinate for
different Hartmann numbers at Re = 1000 (inset: time average local Nusselt number
for unsteady cases).

Ha = 320
70

6.0

60

5.5

4.5

%H

Ha = 320
Ha = 1120
Ha = 2160

5.0

50

4.0

Nuav

3.5
3.0

40

2.5
2.0

1000

1.5

3000

4000

5000

6000

Re

1.0

Fig. 20. The percentage increase in heat transfer due to the cylinder in the channel
as a function of Reynolds number for Ha = 320.

0.5
0.0

2000

1000

2000

3000

4000

5000

6000

Re
Fig. 18. Variation of time and surface average Nusselt number on the heated
bottom channel wall with Reynolds number for different Hartman numbers.

becomes signicant. Hence, as mentioned in the introduction section, the heat transfer is proved to be augmented by inserting an
obstacle in the channel.

12

D. Chatterjee, S.K. Gupta / International Journal of Heat and Mass Transfer 88 (2015) 113

5. Conclusion
The thermouidic transport of liquid metal around a square
object placed in a square duct under a strong homogeneous magnetic eld aligned with the cylinder axis is numerically simulated
following a quasi two-dimensional model. A wide range of
Reynolds and Hartmann numbers are considered for characterizing
the various ow regimes originated as a consequence of the
imposed magnetic eld. In order to understand the thermal transport, the bottom channel wall is considered as heated. Four distinct
hydrodynamic regimes are observed to evolve under the action of
the magnetic eld. The rst three regimes are closely in resemblance to the typical non-MHD wake dynamics around a square
object. A creeping ow regime, followed by a steady twin vortex
regime and a vortex shedding regime subsequently evolve. The
fourth regime is signicantly different from that of the non-MHD
case. This regime is characterized by the formation of the secondary
counter-rotating vortices from the separation of the Shercliff layer
at the side walls along with the regular Krmn vortices from the
object. The transition Reynolds numbers for all these regimes are
obtained to develop the regime map. The most important observation is that the imposed magnetic eld due to its damping nature
actually shifts the appearance of the ow instabilities to higher Re
values and the resulting ow regimes span over a wider range of
Re than in comparison to the corresponding non-MHD cases.
The other observations from the present study are summarized as:
 The regime IV for both the cases of MHD and non-MHD situations is characterized by the formation of the irregular shedding
pattern and chaotic thermal transport. However, the non-MHD
cases show relatively stronger irregularity and more chaotic
nature compared to the corresponding MHD situation. Higher
heat transfer rate from the heated wall in case of the non-magnetic situation in comparison to the magnetic situation can also
be observed.
 Although qualitatively similar characteristic behavior can be
observed for the non-MHD ow past circular and square objects
in the channel under strong axial magnetic eld, the transition
Reynolds numbers for various regimes are much higher in case
of circular cylinder in comparison to its square counterpart.
 The uctuation in the lift coefcient is gradually suppressed by
the increasing magnetic eld strength and at larger eld strength
it is suppressed completely since the ow becomes steady.
 A drop in the Strouhal frequency is observed in the regime IV
due to the impeding action of the secondary vortex developed
in the Shercliff layer over the regular Krmn vortices. The frequency identically becomes zero at larger Hartmann number
also when the shedding is completely suppressed.
 The drag rst decreases then increases with an increase in the
friction parameter Re=Ha4=5 . Furthermore, at low values of
Re=Ha4=5 , the drag coefcients collapse onto a universal curve.
 Higher magnetic eld strength suppresses the convective transport resulting in a thicker thermal boundary layer which eventually reduces the heat transfer rate.
 The heat transfer rate from the heated channel wall increases
monotonically with Reynolds number and decreases with
increasing Hartmann number. This dependence of heat transfer
on Hartmann number is more profound at larger Reynolds
number. More heat transfer is observed to take place from the
heated wall in regime IV in comparison to other regimes.
 A circular object as a heat transfer promoter is preferable than a
square object as long as the imposed magnetic eld strength is
moderate. However, under stronger magnetic eld strength
both promoters act almost equivalently.

 The heat transfer rate is enhanced by inserting the square object

relative to an empty channel.

Conict of interest
None declared.

Appendix A. Supplementary data

Supplementary data associated with this article can be found, in
the online version, at http://dx.doi.org/10.1016/j.ijheatmasstransfer.2015.04.053.

References
[1] C.H.K. Williamson, Vortex dynamics in the cylinder wake, Annu. Rev. Fluid
Mech. 28 (1996) 477539.
[2] M.M. Zdravkovich, Flow Around Circular Cylinders, vol. 1, Oxford University
Press, Oxford, 1997.
[3] A. Sohankar, C. Norberg, Simulation of three-dimensional ow around a square
cylinder at moderate Reynolds numbers, Phys. Fluids 11 (1999) 288306.
[4] N. Kanaris, D. Grigoriadis, S. Kassinos, Three dimensional ow around a circular
cylinder conned in a plane channel, Phys. Fluids 23 (2011) 064106.
[5] M. Coutanceau, R. Bouard, Experimental determination of the main features of
the viscous ow of the wake of a circular cylinder in uniform translation. Part
1: steady ow, J. Fluid Mech. 79 (1977) 231256.
[6] J.H. Chen, W.G. Pritchard, S.J. Tavener, Bifurcation for ow past a cylinder
between parallel planes, J. Fluid Mech. 284 (1995) 2341.
[7] L. Zovatto, G. Pedrizetti, Flow about a circular cylinder between parallel walls,
J. Fluid Mech. 440 (2001) 125.
[8] W.A. Khan, J.R. Culham, M.M. Yovanovich, Fluid ow and heat transfer from a
cylinder between parallel planes, J. Thermophys. Heat Transfer 18 (2004) 395
403.
[9] M. Sahin, R.G. Owens, A numerical investigation of the wall effects up to high
blockage ratios on two-dimensional ow past a conned circular cylinder,
Phys. Fluids 16 (2004) 13051320.
[10] S. Mettu, N. Verma, R.P. Chhabra, Momentum and heat transfer from an
asymmetrically conned circular cylinder in a plane channel, Heat Mass
Transfer 42 (2006) 10371048.
[11] P.H. Roberts, Introduction to Magnetohydrodynamics, Longmans, New York,
1967.
[12] G. Mutschke, G. Gerbeth, V. Shatrov, A. Tomboulides, Two- and threedimensional instabilities of the cylinder wake in an aligned magnetic eld,
Phys. Fluids 9 (1997) 31143116.
[13] G. Mutschke, G. Gerbeth, V. Shatrov, A. Tomboulides, The scenario of threedimensional instabilities of the cylinder wake in an external magnetic eld,
Phys. Fluids 13 (2001) 723734.
[14] H.S. Yoon, H.H. Chun, M.Y. Ha, H.G. Lee, A numerical study on the uid ow
and heat transfer around a circular cylinder in an aligned magnetic eld, Int. J.
Heat Mass Transfer 47 (2004) 40754087.
[15] W.K. Hussam, M.C. Thompson, G.J. Sheard, Dynamics and heat transfer in a
quasi-two-dimensional MHD ow past a circular cylinder in a duct at high
Hartmann number, Int. J. Heat Mass Transfer 54 (2011) 10911100.
[16] D. Chatterjee, K. Chatterjee, B. Mondal, Control of ow separation around bluff
obstacles by transverse magnetic eld, J. Fluids Eng. Trans. ASME 134 (2012)
091102-1091102-11.
[17] D. Chatterjee, K. Chatterjee, Unconned ow and heat transfer around a square
cylinder at low Reynolds and Hartmann numbers, Int. J. Fluid Mech. Res. 40
(2013) 7190.
[18] D. Chatterjee, K. Chatterjee, Wall bounded ow and heat transfer around a
circular cylinder at low Reynolds and Hartmann numbers, Heat Transfer Asian
Res. 42 (2013) 133150.
[19] W.K. Hussam, G.J. Sheard, Heat transfer in a high Hartmann number MHD duct
ow with a circular cylinder placed near the heated side-wall, Int. J. Heat Mass
Transfer 67 (2013) 944954.
[20] D. Chatterjee, K. Chatterjee, B. Mondal, N.B. Hui, Wall conned ow and heat
transfer around a square cylinder at low Reynolds and Hartmann numbers,
Heat Transfer Asian Res. 43 (2014) 459475.
[21] M. Frank, L. Barleon, U. Mller, Visual analysis of two-dimensional
magnetohydrodynamics, Phys. Fluids 13 (2001) 22872295.
[22] G.G. Branover, Yu.M. Gelfgat, S.V. Turuntaev, A.B. Tsinober, Effect of a
transverse magnetic eld on velocity perturbations behind a circular
cylinder swept by an electrolyte, Magnetohydrodynamics (N.Y.) 5 (1969)
4146.
[23] O.V. Andreev, Yu.B. Kolesnikov, MHD instabilities at transverse ow around a
circular cylinder in an axial magnetic eld, in: Third International Conference
on Transfer Phenomena in Magnetohydrodynamics and Electroconducting
Flows, Aussois, France, 1997, pp. 205210.

D. Chatterjee, S.K. Gupta / International Journal of Heat and Mass Transfer 88 (2015) 113
[24] L.G. Kit, Yu.B. Kolesnikov, A.B. Tsinober, P.G. Shtern, Use of a conduction
anemometer in investigating the MHD wake behind a body,
Magnetohydrodynamics (N.Y.) 5 (1969) 4650.
[25] L.G. Kit, S.V. Turuntaev, A.B. Tsinober, Investigation with a conduction
anemometer of the effect of a magnetic eld in the wake of a cylinder,
Magnetohydrodynamics (N.Y.) 6 (1970) 331335.
[26] Y.B. Kolesnikov, A.B. Tsinober, Two-dimensional turbulent ow behind a
circular cylinder, Magnetohydrodynamics (N.Y.) 8 (1972) 2331.
[27] J. Hartmann, F. Lazarus, Hg-dynamics II: experimental investigations on the
ow of mercury in a homogeneous magnetic eld, K. Dan. Vidensk. Selsk. Mat.
Fys. Medd. 15 (1937) 145.
[28] E.C. Brouillette, P.S. Lykoudis, Magneto-uid-mechanic channel ow. Part 1.
Experiment, Phys. Fluids 10 (1937) 9951001.
[29] C.B. Reed, P.S. Lykoudis, The effect of a transverse magnetic eld on shear
turbulence, J. Fluid Mech. 89 (1978) 147171.
[30] J.A. Baylis, Experiments on laminar ow in curved channels of square section, J.
Fluid Mech. 48 (1971) 417422.
[31] P. Moresco, T. Alboussire, Experimental study of the instability of the
Hartmann layer, J. Fluid Mech. 504 (1971) 167181.
[32] B. Mck, C. Gnther, U. Mller, L. Bhler, Three-dimensional MHD ows in
rectangular ducts with internal obstacles, J. Fluid Mech. 418 (2000) 265295.
[33] J. Sommeria, R. Moreau, Why, how, and when, MHD turbulence becomes twodimensional, J. Fluid Mech. 118 (1982) 507518.
[34] V. Dousset, A. Pothrat, Numerical simulations of a cylinder wake under a
strong axial magnetic eld, Phys. Fluids 20 (2008) 017104.
[35] W.K. Hussam, M.C. Thompson, G.J. Sheard, Optimal transient disturbances
behind a circular cylinder in a quasi-two-dimensional magnetohydrodynamic
duct ow, Phys. Fluids 24 (2012) 024105.

13

[36] W.K. Hussam, M.C. Thompson, G.J. Sheard, Enhancing heat transfer in a high
Hartmann number magnetohydrodynamic channel ow via torsional
oscillation of a cylindrical obstacle, Phys. Fluids 24 (2012) 113601.
[37] A. Pothrat, J. Sommeria, R. Moreau, An effective two-dimensional model for
MHD ows with transverse magnetic eld, J. Fluid Mech. 424 (2000) 75100.
[38] A. Pothrat, Quasi-two-dimensional perturbations in duct ows under
transverse magnetic eld, Phys. Fluids 19 (2007) 074104.
[39] ANSYS Fluent 14.0 Theory Guide, ANSYS Inc., 2009.
[40] R. Issa, Solution of the implicitly discretized uid ow equations by operatorsplitting, J. Comput. Phys. 62 (1986) 4065.
[41] A. Pothrat, J. Sommeria, R. Moreau, Numerical simulations of an effective 2D
model for ows with a transverse magnetic eld, J. Fluid Mech. 534 (2005)
115143.
[42] P. Patil, S. Tiwari, Effect of blockage ratio on wake transition for ow past
square cylinder, Fluid Dyn. Res. 40 (2008) 753778.
[43] D. Chatterjee, B. Mondal, Effect of thermal buoyancy on vortex shedding
behind a square cylinder in cross ow at low Reynolds number, Int. J. Heat
Mass Transfer 54 (2011) 52625274.
[44] S.C. Luo, X.H. Tong, B.C. Khoo, Transition phenomena in the wake of a square
cylinder, J. Fluids Struct. 23 (2007) 227248.
[45] G.J. Sheard, M.J. Fitzgerald, K. Ryan, Cylinders with square cross-section:
wake instabilities with incidence angle variation, J. Fluid Mech. 630 (2009)
4369.
[46] Supplementary material, Movie clip representing the development of primary
Krmn vortices and secondary vortices from the Shercliff layer at the side
walls at Re = 2000 and Ha = 640.
[47] J.S. Leontini, M.C. Thompson, Vortex-induced vibrations of a diamond crosssection: sensitivity to corner sharpness, J. Fluids Struct. 39 (2013) 371390.