ARTICLE IN PRESS

Journal of Fluids and Structures 26 (2010) 19–40

[Link]/locate/jfs

Vortex dynamics of a cylinder wake in proximity to a wall

S. Sarkar, Sudipto Sarkar
Department of Mechanical Engineering, Indian Institute of Technology Kanpur, Kanpur 208016, India
Received 1 November 2008; accepted 13 August 2009
Available online 6 November 2009

Abstract

Large-eddy simulations (LES) are used to investigate the modifications of wake dynamics and turbulence
characteristics behind a circular cylinder placed near a wall for varying gap-to-diameter (G/D) ratios (where G signifies
the gap between the wall and the cylinder, and D the cylinder diameter). The three-dimensional (3-D), time-dependent,
incompressible Navier–Stokes equations with a dynamic subgrid-scale model are solved using a symmetry-preserving
finite-difference scheme of second-order spatial and temporal accuracy. The immersed boundary (IB) method is
employed to impose the no-slip boundary condition on the cylinder surface. Flow visualizations along with turbulence
statistics are presented to gain insight into the flow structures that are due to interaction between the shear layers and
the approaching boundary layer. Apart from the vortex shedding mechanism, the paper illustrates the physics involving
the shear layer transition, stretching, breakdown and turbulence generation, either qualitatively or quantitatively, in the
presence of a wall for a Reynolds number of Re ¼ 1440 (based on D and the inlet free-stream velocity UN).
r 2009 Elsevier Ltd. All rights reserved.

Keywords: Wake–boundary layer interactions; Shear layer transition; LES; IB method

1. Introduction

Flow around a cylinder in unbounded condition has been studied extensively for over a century because of its
interesting flow features and practical applications as reflected in reviews by Roshko (1961), Berger and Willie (1972),
Norberg (1994) and Williamson (1996). In contrast, the flow around a cylinder in presence of a wall has received less
attention. However, this kind of study can directly be applied to many engineering problems, such as flow past a
suspension bridge, pipelines near the ground, flow past building construction, heat transfer enhancement in heat
exchangers, forced-air cooling of board-mounted electronic components, etc.
When the cylinder is placed close to a wall, the dynamics of vortex shedding changes appreciably. The cylinder is
under the influence of a non-uniform velocity profile owing to the flow over a rigid wall, which has a non-symmetric
influence on the body. This causes asymmetry in the strength of vortex shedding and generates a finite mean force
directed away from the wall. Further, the vorticity shed from the cylinder interacts with the vorticity of the boundary
layer resulting in an excited wake-induced boundary layer with complex flow structures. When the gap between the
cylinder and the wall is relatively low, this nonlinear interaction is strong, provoking the separation of the boundary
layer apart from a drastic change of the entire wake. Thus, the dynamics of vortex shedding in the presence of a wall is

Corresponding author. Tel.: +91 512 2597942; fax: +91 512 2590007.
E-mail address: subra@[Link] (S. Sarkar).

0889-9746/$ - see front matter r 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.jfluidstructs.2009.08.003
 ARTICLE IN PRESS
20 S. Sarkar, S. Sarkar / Journal of Fluids and Structures 26 (2010) 19–40

influenced by several factors, such as the approaching boundary layer thickness, free-stream disturbances, Reynolds
number and gap between the cylinder and the wall. These factors also control the evolution of the boundary layer and
its transition under excitation of the migrating wake. In brief, the major modifications of flow dynamics past a cylinder
in the vicinity of a wall can be identified as: (i) deflection of the approaching boundary layer away from the wall, (ii)
suppression of vortex shedding from the lower half of the cylinder, (iii) presence of separation bubble both upstream
and downstream of the cylinder and (iv) a significant change of flow parameters such as Strouhal number, lift and drag
coefficient.
The first significant experiment in this area was performed by Taneda (1965). A circular cylinder close to a wall was
towed through the stagnant water at a low Reynolds number Re ¼ 170, where there was no effect of boundary layer.
He observed a single row of vortices behind the cylinder for a small gap ratio of G/D ¼ 0.1 and the wavelength of the
vortices increased with downstream distance, which after a few wavelengths became unstable and broke down. As the
gap ratio was increased to 0.6, a regular double row of vortices was re-established. Several experiments for the flow
around a cylinder in the presence of a single plane boundary have been reported at moderately high Reynolds number,
Re ¼ 2  104 to 105 (Bearman and Zdravkovich, 1978; Grass et al., 1984; Taniguchi and Miyakoshi, 1990; Lei et al.,
1999). In most of these cases, the approaching boundary layer was turbulent. The results illustrated that the
aerodynamic forces on the cylinder were modified with a slight variation of the shedding frequency. The suppression of
vortex shedding was observed when the body was closer than a critical distance from the wall, denoted by (G/D)cr. This
critical gap height was found to be approximately 0.3–0.4 (Bearman and Zdravkovich, 1978; Grass et al., 1984;
Taniguchi and Miyakoshi, 1990; Buresti and Lanciotti, 1979). For smaller gaps, the wake was almost steady and the
periodic shedding was strongly inhibited with separation bubbles on the wall. Later, Lei et al. (1999) reported that the
suppression of vortex shedding occurred at a gap ratio of about 0.2–0.3 depending on the thickness of the approaching
boundary layer.
As stated earlier, the magnitude of aerodynamic forces on the cylinder and the shedding frequency are altered in
proximity to a wall. The frontal stagnation point moves towards the wall producing an upward lift (CL) on the cylinder
with the consequence of deviation of the wake away from the wall. This increases the base pressure, resulting in a low
value of coefficient of drag (CD), although these depend on the boundary layer, its character and thickness. The drag
coefficient remained largely unaffected by G/D when the cylinder was placed outside the boundary layer (Lei et al.,
1999; Zdravkovich, 1985). As the cylinder was moved into the boundary layer, CD decreased with G/d (where d is the
thickness of the boundary layer). However, Buresti and Lanciotti (1979) argued that there was hardly any correlation
between CD and G/d. Taniguchi and Miyakoshi (1990) and Lei et al. (1999) observed that when the lower surface of the
cylinder was placed within the critical gap ratio, the r.m.s. values of both lift ðC 0L Þ and drag ðC 0D Þ coefficients strongly
reduced. This might be attributed to the diffusion of velocity fluctuations because of the boundary layer. Regarding the
vortex shedding frequency (Strouhal number, St), Angrilli et al. (1982) found an influence of gap ratio on St if G/D was
less than 5.0, where Re was in the range of 2.8  103–7.6  103. However, others (Bearman and Zdravkovich, 1978;
Grass et al., 1984; Zdravkovich, 1985; Taniguchi and Miyakoshi, 1990) observed that the shedding frequency was
almost independent of gap ratio. A possible reason for this difference was the different range of Re used in the
experiment of Angrilli et al. (1982) as compared to others. Furthermore, the study of Lei et al. (1999) indicated that the
velocity gradient apart from the gap ratio also influenced the shedding frequency. The increase of St with reduction
of G/D could only be felt for a thinner boundary layer, whereas no significant change was observed for a thicker
boundary layer.
The flow visualization around a circular cylinder near a wall was performed by Price et al. (2002) for Re between 1200
and 4960 and G/D between 0 and 2. They concluded that the G/D ratio was a major factor for the change of flow
characteristics. The flow interactions between cylinder wake and flat plate boundary layer have been divided in four
zones based on the G/D ratio. For G/Do0.125, the flow beneath the cylinder was suppressed and separation of
boundary layer occurred both upstream and downstream of the cylinder. Although there was no regular vortex
shedding, a periodicity associated with the outer shear layer was found. For 0.25oG/Do0.375, the separation region
decreased, but the flow characteristics were the same as those of the previous case. A pronounced pairing between the
inner shear layer shed from the cylinder and the wall boundary layer was observed in this case. For 0.5oG/Do0.75, the
vortex shedding started and for G/D41.0 there was no separation of the wall boundary layer either upstream or
downstream of the cylinder. They observed that the variation of St with gap ratio was mainly dependent on Re. For low
Reynolds number flows (Reo2600), St was significantly greater than that of an isolated cylinder for G/Do2. But as Re
was increased (Re44000), St became insensitive to the G/D ratio. The paper presented the flow interactions with the
help of instantaneous vorticity and power spectra.
Numerical simulation of flow past a cylinder in the vicinity of flat plate is very limited. It was first performed by Vada
et al. (1989) using vortex in cell method for transcritical Reynolds number, Re ¼ 3.6  106 and G/D ¼ 0.4, 0.8 and 1.5.
They observed the appearance of mean lift force and increased drag due to the presence of a wall. The simulation
 ARTICLE IN PRESS
S. Sarkar, S. Sarkar / Journal of Fluids and Structures 26 (2010) 19–40 21

suffered from over-prediction of drag coefficient and under-prediction of lift coefficient as compared to the experiment
of Zdravkovich (1985) for a small gap ratio. Liou et al. (2002) performed a LES of the turbulent wake in proximity to a
wall for a Re ¼ 2.2  104, although a square cylinder was considered and the discussion was focused on the phase-
averaged vortex dynamics. The celerity of the positive vortex shed from the lower side of the cylinder was smaller than
that of the upper side shed vortex due to the interaction with the boundary layer. Recently, Dipankar and Sengupta
(2005) solved the cylinder boundary layer interaction considering two-dimensional co formulation. They have tried
to mimic the experiment of Price et al. (2002) considering Re ¼ 1200 and G/D ¼ 0.5 and 1.5. The computation
illustrated larger vortical structures as compared to the experiment with rollup of vortices closer to the cylinder
and failed to capture the breakdown. This was attributed to the fact that the absence of spanwise direction in 2-D
simulation failed to redistribute the energy in the third direction, developing unrealistic large structures (Mittal and
Balachandar, 1995).
The present paper will focus on the characteristics of the vortex shedding and its modifications behind a cylinder as
the wall is approached for a Reynolds number in the shear layer transition regime. Although there are experimental
investigations illustrating the vortex shedding process as well as the aerodynamic forces on the cylinder in proximity to
a wall, the physics of flow including the shear layer stretching, breakdown and turbulence generation are not very
complete. Most of the numerical simulations made so far on this flow configuration are in two-dimensional framework
and thus elucidate the flow characteristics with limited success, particularly for high Reynolds numbers. An appropriate
numerical tool in 3-D computational environment is very much needed for a complete study of the vortex dynamics and
turbulence characteristics. It should be noted that in complex turbulent flows, where an understanding of the eddy
motions and their interactions is the key to comprehension of the flow, a LES is often considered as an alternate root of
DNS because of its economy. Considering all these factors, a 3-D LES with a dynamic subgrid-scale model has been
used to investigate the flow around a cylinder near a wall at Re ¼ 1440 and for G/D of 0.25, 0.5 and 1.0. Here, the wake
dynamics and its nonlinear interactions with the approaching boundary layer involving the large- and small-scale
structures are illustrated. The present paper is organized as follows. We describe the governing equations, geometrical
configuration followed by flow validation and grid resolution study. Next we present the results and discuss the flow
physics with the help of instantaneous flow visualization along with the turbulence statistics. Finally, we end the paper
with a concluding note.

2. Numerical methods

2.1. Governing equations

In the present study, we perform LES of incompressible flow. The filtered mass and momentum equations can be
expressed as
@ūj
¼ 0, (1)
@xj

@ūi @ 1 @P̄ 1 2 @tij

þ ðūj ūi Þ ¼  þ r ūi  þ f̄ i , (2)
@t @xj r @xi Re @xj
where ui denotes the filtered velocity field and tij ¼ ui uj  ūi ūj is the residual stress tensor (also known as subgrid-scale
stress, SGS). The presence of the body forces f i is due to the IB method, which will be discussed later. The above
equations have been made dimensionless using the inlet free-stream velocity UN and the cylinder diameter D. The
resulting Reynolds number is defined as Re ¼ UND/n, n being the kinematic viscosity. The model proposed by
Germano et al. (1991) and modified by Lilly (1992) is used here to include the effect of subgrid motions in the resolved
LES, where the model coefficient C is dynamically calculated instead of input a priori. The least-squares estimate of the
optimal solution for C, following Lilly, is employed by averaging over the homogeneous z-direction. Thus, the resulting
C is the function of time and the inhomogeneous coordinates x and y. No damping of the turbulent viscosity, such as
the one proposed by van Driest, is applied as solid walls are approached. The momentum advancement is explicit using
the second-order Adams–Bashforth scheme, except for the pressure term, which is solved by a standard projection
method. The pressure equation is discrete Fourier transformed in one dimension (in which the flow can be considered
homogeneous and hence periodicity of flow can be imposed) and is solved by the BI-CG algorithm (Zhang, 1997) in the
other two dimensions. The spatial discretization is second-order accurate on a staggered mesh arrangement using a
symmetry-preserving central-difference scheme, which is widely used in LES owing to its non-dissipative and
 ARTICLE IN PRESS
22 S. Sarkar, S. Sarkar / Journal of Fluids and Structures 26 (2010) 19–40

conservative property (Mittal and Moin, 1997; Morinishi et al., 1998). However, as an alternative to the classical LES,
implicit LES methods have been proposed in recent years relating the dispersion and dissipation with turbulence
properties to encompass the constraints in near-wall flows (Drikakis, 2003).
Here, the simulations of flow past a cylinder in proximity of a wall have been performed in the Cartesian grid
considering the origin of axes lies on the flat plate and at a distance of 10D from the leading edge. The coordinates x, y,
z denote the streamwise, wall-normal and spanwise directions, respectively, and the corresponding velocities are
denoted by u, v and w. The computational domain and boundary conditions used here are illustrated in Fig. 1. The
domain extends from 5D at the inflow to 25D at the outflow with a spanwise length of 3D. It should be noted that a
spanwise length of pD was used by Beaudan and Moin (1994), Mittal and Moin (1997) and Kravchenko and Moin
(2000) while simulating the flow around a circular cylinder in unbounded condition for Re ¼ 3900 (the shear layer
transition region). It was also reported that doubling the domain in the spanwise direction did not affect the results
significantly (Kravchenko and Moin, 2000). The instant spanwise length of 3D appears reasonable in spite of an
influence of Reynolds number scaling.
A Blasius profile is imposed at the inlet considering the computational domain begins at a 5D distance downstream
from the leading edge of the flat plate, where the boundary layer thickness is considered as 0.295D. A non-reflecting
boundary condition (Orlanski 1976) is applied at the exit. The no-slip boundary condition (u ¼ v ¼ w ¼ 0) is applied on
the flat plate and on the cylinder surface. A free-slip velocity (v ¼ 0, qu/qy ¼ qw/qy ¼ 0) is imposed along the upper
surface of the domain, which does not allow outflow through the top surface. This causes the flow to accelerate above
the boundary layer in order to compensate for the growth of the boundary layer. In the spanwise direction, the periodic
boundary condition is used.
The IB method helps to impose boundary conditions on a given surface not coinciding with the computational grid as
in the present case. Thus, to satisfy the no-slip boundary conditions at the cylinder surface, the IB technique is
employed following Fadlun et al. (2000), which belongs to the class of ‘direct forcing method’. The velocity field near
the boundary of the body is modified at each step in such a way that the no-slip boundary condition is satisfied on the
surface. This is done using some interpolations, which is equivalent to including a body force f i in the momentum
equation. In this paper, a quadratic unidirectional interpolation by Muldoon and Acharya (2005) is used. The flow
solver has been extensively validated for a variety of transitional and turbulent flows (Sarkar and Voke, 2006; Sarkar,
2007, 2008) including simulation of vortex dynamics behind a circular cylinder for a wide range of Reynolds numbers
(Sarkar. and Sarkar, 2007a b).

2.2. Computational details

The LES of flow past a circular cylinder near a wall is performed at a Re ¼ 1440 and for three gap ratios,
G/D ¼ 0.25, 0.5 and 1.0 following the experiment of Price et al. (2002). In the streamwise direction, a uniform and
refined mesh is used near the cylinder to get adequate immersed-boundary points on the cylinder surface. Away from
the cylinder, the mesh is slowly stretched out. In the flow-normal direction, a refined and uniform mesh is employed
near the wall (up to y/D ¼ 2 for G/D ¼ 0.25, 0.5 and y/D ¼ 2.5 for G/D ¼ 1.0) to resolve the wake behind the cylinder
and its interaction with the boundary layer. A uniform mesh is used in the spanwise direction owing to the symmetry of
the body.
The evolution of boundary layer excited by the wake depends on a realistic representation of wake structures behind
a cylinder. To estimate the resolution in the vicinity of a cylinder, we first performed LES of flow past a cylinder in an

v = 0, ∂ u /∂ y = ∂ w / ∂ y = 0

Computational domain
Blasius
profile
v=w=0 u=v=w=0
D ∂ /∂ t + U∂ /∂ x = 0
Boundary Flat plate
layer
Y
δ G X

5D 5D 25 D

Fig. 1. Schematic diagram of the computational domain with the boundary conditions.
 ARTICLE IN PRESS
S. Sarkar, S. Sarkar / Journal of Fluids and Structures 26 (2010) 19–40 23

unbounded condition for two Reynolds numbers, Re ¼ 1000 and 3900. Here, the computational domain extends from
8.25D at the inflow to 25.75D at the outflow and from 8.25D to 8.25D in the cross-flow direction with a spanwise
length of 3D. A mesh of 360  192  32 points is used for Re ¼ 1000, while a mesh of 384  204  32 points is used for
Re ¼ 3900. For Re ¼ 1000, among 360 streamwise grid points, 64 points are distributed upstream of the cylinder
(8.25D to 0.75D), 64 points within 0.75D to +0.75D, 150 points in the near wake (+0.75D to +10.75D) and rest
are distributed in the far field downstream, whereas for Re ¼ 3900 the grid distributions are 64, 76, 150 and 82 in the
respective regions. The spacing between adjacent grid points near the cylinder becomes 0.0156D for Re ¼ 1000 and
0.013D for Re ¼ 3900. Ovchinnikov et al. (2006) illustrated that a grid spacing of 0.015D near the cylinder was
adequate to resolve flow in the wake for Re ¼ 3900, where a LES with the IB method was used. However, the near-wall
mesh distributions used here accommodate limited points within the laminar boundary layer on the windward side of
the cylinder (according to the boundary layer theory, Schlichting 1979). For Re ¼ 3900, a total of 146 IB-points for u-
velocity along the cylinder surface are obtained by the grid distribution as compared to 140 surface points used by
Mittal and Moin (1997) for a LES at the same Re. In the near wake region, non-dimensional mesh spacing varies
as Dx+ ¼ 0.13.5, Dy+ ¼ 0.051.8, Dz+ ¼ 0.12.9. The mean values of Dx+, Dy+ and Dz+ over a range of x/
D ¼ 1–25 along the cylinder axis of the cylinder are 1.4, 0.3 and 1.7, respectively.
Fig. 2 depicts the mean velocity profiles obtained at three different sections (x/D ¼ 3, 5, 7) for Re ¼ 1000, which is
the lower limit of the shear layer transition according to Williamson (1996). The velocities are normalized
by the maximum mean velocity deficit ūwake;max and the lengths are normalized by the wake half-width b, which is
defined as the distance between two points at which the mean velocity deficit becomes half of ūwake;max . The normal-
ized mean velocity profiles obtained from three different locations collapse almost to a single curve, which is
in good agreement with experimental measurements and data correlations of Schlichting (1979), i.e.,
ūwake =ūwake;max ¼ ½1  ðȳwake =1:1338bÞ1:5 2 . This demonstrates that the simulated wake has reached a self-similar state.
For Re ¼ 3900, the mean streamwise velocity profiles from the present simulation are compared with the
experimental data of Lourenco and Shih [data taken from Ma et al. (2000)] for the very near wake region (x/Dr3) and
the data of Ong and Wallace (1996) for the vortex shedding region (x/D ¼ 4 and 7) in Fig. 3. It is interesting
to note that velocity profiles agree well with the experimental data resolving a V-shaped profile in the recirculation
region immediately behind the cylinder at x/D ¼ 1.06 and 1.54. The streamwise velocity profile tends to attain
a U-shape downstream of the recirculation region, at x/D ¼ 2.02. The U-shaped profile in the vortex shedding
region (x/D ¼ 4, 7) may be attributed to the development of turbulence and enhanced mixing. The turbulence
statistics (not shown) obtained from the present LES are also in fair agreement with the experimental data of
Ong and Wallace and DNS of Ma et al. (2000). The study illustrates that the grid resolution chosen is good enough to
predict the physics of the flow past a cylinder, e.g. flow structure in the near wake and the shear layer transition
downstream.
In order to ensure the accuracy of calculations for the wake and boundary layer interactions, a grid resolution test
has been carried out for a gap ratio of G/D ¼ 0.25. The grid requirements to resolve the transitional flow or the
boundary layer excited by external disturbances, such as in the present case, are less well established. However, the grid
requirements for the simulation of fully turbulent boundary layer using a second-order accurate LES solver are well
known (Dx+C50, Dyþ min o1.0 and Dz C20). Jacobs and Durbin (2001) indicated that the prediction of boundary layer
+

0.1

-0.1

_ -0.3
u wake
___
_ -0.5 x/D = 3
uwake, x/D = 5
max -0.7 x/D = 7
Schlichting’s
-0.9 correlation

-1.1
-2 -1 0 1 2
y/b

Fig. 2. The wake deficit for flow over a circular cylinder at Re ¼ 1000.
 ARTICLE IN PRESS
24 S. Sarkar, S. Sarkar / Journal of Fluids and Structures 26 (2010) 19–40

x/D=1.06
1

x/D=1.54
0

8 x/D=2.02
-1

__
U
U
x/D=4.0
-2

x/D=7.0
-3

-4 -2 0 2 4
x/D

Fig. 3. Time averaged streamwise velocity at different x/D locations for flow over a circular cylinder at Re ¼ 3900: ––––, present
computation; &, experimental results.

transition is sensitive to streamwise resolution. Further, Ovchinnikov et al. (2006) illustrated that an under-resolved
calculation predicts a premature and abrupt transition. In the present computation, four levels of mesh, such as
288  192  32, 384  160  32, 384  192  32 and 384  192  64 (designated as grid 1, grid 2, grid 3 and grid 4) in the
streamwise, wall-normal and spanwise directions, respectively, are used. For grid 1, among 288 streamwise grid points,
32 points are distributed upstream of the cylinder (5D to D), 96 points surrounding the cylinder (D to +D), 108
points near wake region (+D to +10D) and the rest are distributed in the far field downstream (+10D to +25D),
whereas for other three mesh levels (384 points), the distribution is 40, 96, 160 and 88 in the respective regions. In the
wall-normal direction, a total of 128 grid points with equal spacing are used between 0 to 2D and remaining 64 points
are slowly stretched away from the cylinder (+2D to +8.5D) for grid levels 1, 3 and 4. For grid 2, the distribution is 96
and 64 in the respective regions. The grid resolution near the cylinder is almost the same as used for the LES in the
unbounded condition at Re ¼ 1000. It should be noted that the boundary layer thickness assuming a laminar flow at
the location of cylinder without its presence is 0.42D and approximately 27 grid points are employed within the
boundary layer for grid 3.
The profiles of mean (U=U 1 ) and r.m.s. (urms/UN) streamwise velocity at x/D ¼ 2, 3, 5, 7, 10 and 15 are
depicted in Fig. 4 for all four grid levels. The data are normalized by the free-stream velocity and the cylinder
diameter. No significant differences in the mean velocity profiles are observed between grid 1 and grid 4 except at x/
D ¼ 3 (Fig. 4(a)): the location of breaking down of shear layers to smaller eddies for G/D ¼ 0.25. However, urms profiles
exhibit appreciable differences between grid levels 1 and 4 (Fig. 4(b)). This indicates that there is a need of large
number of grid points in the near wake region and inside the boundary layer. For the subsequent analysis, grid 3 is
chosen considering the fact that there were no considerable changes of both the mean and r.m.s. velocity profiles by
further increasing the grid points. In the near wake (x/D ¼ 3–15) region, non-dimensional mesh spacing for
grid 3 varies as Dx+ ¼ 0.1515, Dy+ ¼ 0.071.5, Dz+ ¼ 0.17.5. For G/D ¼ 0.5, the same computational domain
and grid distribution are used as considered in G/D ¼ 0.25. The computational domain is extended by 0.5D in the
wall-normal direction for G/D ¼ 1 (to assure almost a similar free-stream condition) keeping the streamwise
domain unchanged. Here, a total of 160 grid points with equal spacing are used between 0 and 2.5D and 64 grid points
are distributed in the remaining (+2.5D to +9D) normal direction. Few parameters such as the domain, grid
distributions and the values of Dx+, Dy+ and Dz+ averaged between x/D ¼ 3 and 15 are summarized in Table 1 for
different cases.
The time step was about Dt ¼ 1.125  103(D/UN) in nondimensional units. This time kept the Courant number
below 0.2 for the entire simulation and the viscous stability number was much less. The flow field was allowed to evolve
for 15 vortex shedding cycles to get a dynamically steady state solution. Then the data were collected for statistics for a
period of 15 cycles that was found sufficient to ensure statistically converged results.
 ARTICLE IN PRESS
S. Sarkar, S. Sarkar / Journal of Fluids and Structures 26 (2010) 19–40 25

G/D = 0.25
4
x/D = 2 3 5 7 10 15
3

y/D
2

0
0 1 2 3 4 5 6
U /U

8
G/D = 0.25
4
x/D= 2 3 5 7 10 15
3
y/D

0
0 0.2 0.4 0.6 0.8 1 1.2
ur.m.s./U 8

Fig. 4. Grid-independent test for G/D ¼ 0.25 and Re ¼ 1440 considering (a) mean streamline velocity (U=U 1 ) and (b) r.m.s. of
streamline velocity (urms/UN) profiles at x/D ¼ 2, 3, 5, 7, 10 and 15: – . – . –, 288  192  32 (grid 1); – . – . –, 384  160  32 (grid 2);
——, 384  192  32 (grid 3);         , 384  192  64 (grid 4).

Table 1
Computational grid and box size.

G/D Dimension of box (Lx  Ly  Lz) Grid (nx  ny  nz) Dx+ Dy+ Dz+ Comments
0.25 30D  8.5D  3D 288  192  32(grid 1) 7.96 0.68 4.11 Test case
384  160  32(grid 2) 3.82 0.86 3.85 Test case
384  192  32(grid 3) 4.35 0.71 4.41 Used
384  192  64(grid 4) 4.23 0.74 2.25 Test case
0.5 30D  8.5D  3D 384  192  32 4.07 0.68 4.12 Used

1.0 30D  9D  3D 384  224  32 4.87 0.8 4.93 Used

3. Results and discussion

3.1. Time-averaged flow features

The modifications of wake dynamics behind a circular cylinder in the presence of a nearby boundary layer are
discussed in this section. Fig. 5 presents the iso-contours of time-averaged streamwise velocity along with a few selected
streamlines for the three gap ratios. When the cylinder is very close to the wall (G/D ¼ 0.25), the shear layers remain
frozen for a considerable time before they curl up to form the Kármán rolls, and these rolls lost their symmetry because
of suppression of vortex shedding. The front stagnation point moves towards the wall from the centreline of the
cylinder. Furthermore, the inner separation point shifts downstream along the shoulder of the lower half of the cylinder,
while the outer separation point moves upstream along the shoulder of the upper half that changes the wake size behind
the cylinder. The deflection of the wake and its departure from symmetry due to the wall proximity can be understood
by the iso-contours of streamwise velocity. Further, a strong coupling between the inner shear layer and the boundary
layer in this case accentuates the deflection of the boundary layer away from the wall forming a large downstream
separation (which is designated as the primary bubble). A secondary bubble rotating in the opposite direction is formed
beneath the primary bubble. The mean streamlines also depict the formation of an upstream separation for the low gap
 ARTICLE IN PRESS
26 S. Sarkar, S. Sarkar / Journal of Fluids and Structures 26 (2010) 19–40

G/D=0.25 Primary
2 2.64 D bubble bubble

y/D
Upstream
bubble
1
1.2 D

0
-3 -2 -1 0 1 2 3 4 5 6 7
1.87 D 0.75 D
8.37 D
G/D=0.5 1.8 D
2
y/D

0.71 D
0
-3 -2 -1 0 1 2 3 4 5 6 7
3.51 D
1.01 D
G/D=1
2
y/D

1
2.7 D 0.2 D
0
-3 -2 -1 0 1 2 3 4 5 6 7

-0.1 0.1 0.3 0.5 0.7 0.9 1.1

x/D

Fig. 5. Mean streamwise velocity contours along with few selected streamlines for different gap ratios.

ratio. As the gap ratio increases, the downstream separation bubble rapidly reduces and the upstream separation region
disappears. Moreover, the Kármán rolls from the curling of the inner and outer shear layer form closer to the cylinder
and appear symmetrical. The bubble length and its height along with the recirculation length of the shear layers are
marked in the figure. For an increased gap ratio (G/D ¼ 1), the recirculation length of the shear layer becomes 1.01D,
which approaches the value of recirculation length of flow past a circular cylinder in an unbounded condition (a
recirculation length of 0.93 is seen at Re ¼ 1000 in the present simulation). The deflection of the boundary layer is also
inhibited as compared to previous cases. The front stagnation points determined from the present LES are
approximately at yS ¼ 10.61, 9.01 and 4.91 for G/D ¼ 0.25, 0.5 and 1.0, respectively, where yS is measured from
the front point on the horizontal axis of the cylinder and in the clockwise direction. The variation of yS fairly agrees
with the experiment of Lei et al. (1999), where yS were 9.61, 6.81 and 3.31 for G/D ¼ 0.3, 0.5 and 1.0, respectively,
at Re ¼ 1.301.45  104 and boundary layer thickness d/D ¼ 0.48. It should be noted that in our case d/D ¼ 0.42, but
Re ¼ 1.44  103.
The lift coefficient (CL), drag coefficient (CD) and Strouhal frequency (St) are three non-dimensional parameters,
which are helpful to describe the effects of boundary layer on the wake dynamics. As the cylinder is moved towards the
wall, the mean lift coefficient (C L ) becomes positive with decrease of C D . The movement of the frontal stagnation point
towards the wall is associated with the generation of an upward lift, while the movements of separation points with an
increase of the base pressure result in a reduced drag coefficient. The lower shear layer of the cylinder is suppressed
because of the wall boundary layer, which also affects the base pressure. Thus, both the G/D ratio and the boundary
layer thickness influence the aerodynamic forces on the cylinder. The present LES predicts a value of C L as 0.124 and a
value of C D as 0.90 for a low gap ratio, G/D ¼ 0.25, which appears to be below the critical gap ratio (highly correlated
to the boundary layer thickness). As the gap ratio is increased to 1.0, C L approaches zero with a value of 0.0075, while
C D becomes 1.47. A similar trend in the variations of C L and C D was reported by Lei et al. (1999) and Zovatto and
Pedrizzetti (2001), although the magnitude was different because of differences in Reynolds number, the boundary
layer thickness and its character. Furthermore, the r.m.s. values of CL obtained from the LES are 0.008 and 0.27
for G/D ¼ 0.25 and 1.0, respectively. The reduced r.m.s. value also confirms that the vortex shedding is suppressed at
G/D ¼ 0.25. Above the critical gap ratio, the regular vortex shedding abruptly starts to develop that rapidly changes the
r.m.s. values (Taniguchi and Miyakoshi, 1990; Lei et al., 1999).
 ARTICLE IN PRESS
S. Sarkar, S. Sarkar / Journal of Fluids and Structures 26 (2010) 19–40 27

104 G/D = 0.25

G/D = 0.5
103 8
G/D = 1.0

102
(at x/D = 3.3, y/D = 1.5)
Present
6
1

PSD(arbitraryscale)
10 Experiment
Euu (Priceetal.2002)
____ (-5/3)
2 100
U D 4
8

10-1

10-2 2

10-3

10-4 0
10-1 100 101 0 0.2 0.4 0.6 0.8 1
St St

Fig. 6. Power spectra density of streamwise velocity at x/D ¼ 3.3, y/D ¼ 1.5 for Re ¼ 1440: (a) for all gap ratios; (b) a zoomed view
comparing with experiment for G/D ¼ 0.25.

0.5

0.45

0.4

0.35

0.3
St

0.25

0.2

0.15

0.1

0.05

0
0 0.5 1 1.5 2 2.5 3 3.5 4
G/D

Fig. 7. St plot for different G/D ratios: K, present LES (Re ¼ 1440); n, Price et al. (Re ¼ 1900); ,, Price et al. (Re ¼ 4900); &,
Angrilli et al. (Re ¼ 2860, 3820, 7640).

From the mean flow particularly at a low gap ratio, it is difficult to appreciate the vortex shedding. However, there
exists a strong periodicity component to the predicted velocity signal in the cylinder wake; almost certainly, this is
associated with the shear layer shed from the outer surface of the cylinder. A spectral analysis is performed after
collecting the velocity histories in the mid-span plane at a location of x/D ¼ 3.3 and y/D ¼ 1.5 from the cylinder center
over a time period of TUN/D ¼ 15 (T being the time period of vortex shedding). Fig. 6(a) depicts the power spectra of
the streamwise velocity fluctuations for all the gap ratios. This reveals that as the cylinder approaches the wall, the
Strouhal frequency increases due to the suppression of vortex shedding. For a low gap ratio (G/D ¼ 0.25), a dominant
wake frequency occurs at St ¼ 0.4; however, there is also a secondary peak at St ¼ 0.23. These two peaks are attributed
to the difference in celerity of the positive vortex shed from the lower side of the cylinder than that of the negative
vortex shed from the upper side: result of the vortex-boundary layer interactions. The prominent shedding frequency
shifts to 0.29 and 0.25 for G/D ¼ 0.5 and 1.0, respectively. Furthermore, the power spectra illustrate that the effect of
the wall boundary layer reduces as the gap ratio increases and the two peaks in the shedding frequency appear to merge
into one peak, the value of which approaches 0.2. A zoomed view of power spectra for G/D ¼ 0.25 is shown in Fig. 6(b)
 ARTICLE IN PRESS
28 S. Sarkar, S. Sarkar / Journal of Fluids and Structures 26 (2010) 19–40

and compared with the experimental data (Price et al., 2002), which demonstrate a good agreement. A wide range of
shedding frequency with more energetic high-frequency harmonic is also observed from Fig. 6(a), which is attributed to
the breaking down of primary vortices to small-scale energetic eddies downstream. The slope of the resolved inertial
range appears to follow 5/3 power law correlation between the energy and the frequency. A summary of St data
(considering prominent wake frequency) as a function of G/D obtained from the present simulations along with the
experiments (Angrilli et al., 1982; Price et al., 2002) is presented, in Fig. 7. The LES at Re ¼ 1440 agrees reasonably well
with the available experimental data, where the Re was 1900. The validation illustrates that the present LES can be used
to describe the evolution of shear layers shed from the cylinder under the influence of wall boundary layer.
The changes of flow features due to the presence of a wall are also reflected on the surface characteristics. The
evolution of the mean skin friction (C f ¼ tw =0:5rU 21 , tw being the wall shear stress) along the wall for the three gap
ratios is shown in Fig. 8. The separation and reattachment points along with the length of bubbles as marked in Fig. 5
are obtained from the distributions of C f . For G/D ¼ 0.25, C f becomes negative at x/D ¼ 2.58 to 0.71, indicating
an upstream separation, which is absent in the other two cases. The drop followed by a sharp rise of C f upstream
of x/D ¼ 0 is due to the combination of wall proximity and flow acceleration beneath the cylinder. The peak value of
C f becomes maximum for this gap as the flow below the cylinder is under the maximum negative pressure. Downstream
of the cylinder, C f indicates an onset of separation at x/D ¼ 1.69 with a large bubble length of 8.37D, whereas the
corresponding values are 1.06, 4.57D and 1.34, 3.04D for the gap ratio of 0.5 and 1.0, respectively. The appearance of a
secondary bubble downstream for G/D ¼ 0.25 and 0.5 is also reflected in the streamwise evolution of C f . When
compared with the skin friction values evaluated from the correlation of turbulent flow, Fig. 8 further illustrates the
relaxation and transition of flow near the wall. For G/D ¼ 1, an overshoot of C f is observed after reattachment of flow

0.04

G/D = 0.25
0.03
G/D = 0.5
G/D = 1.0
0.02 turbulent Cf
Cf

0.01

-0.01
-5 0 5 10 15 20 25
x/D

2.88 X (104) 5.04

Rex

Fig. 8. Profiles of mean coefficient of friction (Cf) on the wall for different G/D ratios. Turbulent Cf profile obtained from the
correlation Cf ¼ 0.058/Rex1/5 is also superimposed.

G/D = 0.25
0.5
G/D = 0.5
G/D = 1.0
0.25

0
Cpw

-0.25

-0.5

-0.75
-5 0 5 10 15 20 25
x/D

Fig. 9. Profiles of mean pressure coefficient on the wall (Cpw) for different G/D ratios. Minimum and maximum Cpw points are joined
by dashed lines.
 ARTICLE IN PRESS
S. Sarkar, S. Sarkar / Journal of Fluids and Structures 26 (2010) 19–40 29

and it approaches the level of turbulent Cf downstream. For the other two cases, large regions of negative C f are seen
that slowly attain the level of turbulent Cf.
The distributions of time-averaged dimensionless pressure (Cpw ¼ (PwPN)/0.5rU2N, Pw being the wall pressure) on
the wall are shown in Fig. 9 as a function of the gap ratio. There is a drop in pressure beneath the cylinder due to the
flow acceleration with a plateau in Cpw that tends to recover downstream. With a decrease of G/D ratio, the magnitude
of negative Cpw value increases with a large plateau illustrating a large separation region. Further to note that the
location of minimum Cpw moves a little downstream, whereas the maximum Cpw moves upstream with the increase in
gap ratio. The movements of these peaks (indicated by joining the locations of minimum and maximum Cpw) are
attributed to the shift of front stagnation point towards the wall and the rear separation point along the shoulder of the
cylinder as stated before. A similar trend was also found in the experiment of Choi and Lee (2000).

3.2. Instantaneous vortex dynamics

To visualize the vortex dynamics in proximity to a wall during the shedding cycle, the snapshots of iso-contours of
spanwise vorticity (oz) and streamlines are presented in Figs. 10–15 for different gap ratios. These figures are drawn
between a time period (T) of vortex shedding that has been calculated assuming the Strouhal frequency as 0.2. However,
it should not be interpreted as St ¼ 0.2 for all cases; indeed there is no such periodicity of the wake at all for a low gap
ratio, G/D ¼ 0.25 at Re ¼ 1440. The Strouhal number as 0.2 represents an approximate average value over the range of
G/D considered in the present study. A brief description of the vortex formation, its downstream convection, interaction
with the boundary layer and breaking down to smaller eddies are described in this section.
Fig. 10 depicts the iso-contours of instantaneous oz (t/T ¼ 0.0–0.8) along the mid-span section, indicating the flow
behaviour for a small gap ratio, G/D ¼ 0.25, which corresponds to G/d ¼ 0.6 (d being the boundary layer thickness at
the location of cylinder without its presence). For t/T ¼ 0.0, the outer shear layer shed from the cylinder is seen to curl
up forming a vortex (denoted by A) over the inner shear layer, which is deflected away from the wall. The inner shear
layer from the cylinder remains frozen for a long period of time and there is no apparent rollup. A strong pairing
between the inner shear layer and the boundary layer is observed being of opposite sign that suppresses the shedding of
the inner shear layer. This coupling also contributes to the lift-off of the boundary layer forming a separation bubble
downstream of the cylinder. The boundary layer appears to roll up (denoted by B) beneath the inner shear layer. For
the next few steps (t/T ¼ 0.2–0.6), the inner shear layer is stretched further developing a Kármán-vortex sheet-like
structure (denoted by C): one end of it remains attached to the cylinder to form a quasi-steady attached wake, while the
other end engulfs the outer roll A. This roll of negative vorticity with a patch of positive vorticity convects downstream
and breaks down to smaller eddies. The mutual interaction of the wall vortices with the shed vortices creates a
remarkable difference in the wake dynamics: the trajectories of shed vortices cross each other and the final vertical
position is opposite with respect to the unbounded case. Thus, the clockwise vortices shed from the upper side of the
cylinder occupy the lower position, while the counterclockwise vortices shed from the inner side of the cylinder occupy
the upper position in the sheet; a similar trend is reported by Zovatto and Pedrizzetti (2001). At t/T ¼ 0.8, neither
vortex A nor vortex B is identified downstream, but a negative vortex sheet appears that convects with a speed slower
than the free-stream. These migrating vortices impart momentum to the wall boundary layer and the small-scale
perturbations are observed near the wall and downstream of x/D ¼ 5.
Instantaneous span-averaged vorticity at two time instants (t/T ¼ 0.4 and 0.8) and the corresponding streamlines
downstream of cylinder are shown in Fig. 11 for G/D ¼ 0.25. The span-averaging removes the small-scale fluctuations
and indicates the coherent structures. The span-averaged results also illustrate the coupling between the inner-side shed
vortices and the wall vortices, the suppression of the shedding of inner shear layer and deflection of wake trajectory. A
prominent feature of flow visualization with the span-averaging is the appearance of a wake sheet of negative vorticity
downstream that has lost its discrete vortical structure. The roll of negative vortices from the upper shear layer merges
with the boundary layer vortices and convects downstream before breakdown to form this wake sheet of the same-sign
vorticity. The span-averaged streamlines indicate the appearance of a large separation bubble, unsteady in nature,
downstream of the cylinder.
For G/D ¼ 0.5 that corresponds to G/d ¼ 1.2, both the inner and outer shear layers shed from the cylinder curl up in
an almost alternating fashion producing two rolls (denoted by A and C), although lack of symmetry is evident at t/
T ¼ 0 as shown in Fig. 12. The lift-off of wall boundary layer occurs with a tongue-like vortex sheet that forms a roll
denoted by B. However, this deflection of the boundary layer decreases, which in turn reduces the size of the
downstream separation bubble as compared to the earlier. The curling of the shear layers also appears to form closer to
the cylinder. As time progresses (t/T ¼ 0.2–0.6), a rapid stretching of the inner shear layer is observed, forming an
elongated vortex sheet, and its coupling with the wall boundary layer leads to the formation of a large negative roll as
 ARTICLE IN PRESS
30 S. Sarkar, S. Sarkar / Journal of Fluids and Structures 26 (2010) 19–40

3 t / T = 0.0 A
Inner shear
2

y/D
1

0 B

0 2 4 6 8 10

3
t / T = 0.2 B
2

y/D
1

0
0 2 4 6 8 10
Stretching of
3 inner vortex sheet (C)
t / T = 0.4
2
y/D

A
1

0
0 2 4 6 8 10

3
Separation C
t / T = 0.6
2
y/D

A
1

0
0 2 4 6 8 10

3 A+B
C
t / T = 0.8
2
y/D

0
0 2 4 6 8 10
x/D

Fig. 10. Instantaneous oz contours for G/D ¼ 0.25. Total 40 non-dimensional contours are considered in between 10 and +10. The
negative vorticity is represented by the dotted line.

3 t/ T = 0.4 Coherent negative vortex 3 t/ T = 0.8

A
Pairing
2 2
y/D

y/D

1 1

0 0
0 2 4 6 8 10 0 2 4 6 8 10
3 3

2 2
y/D

y/D

1 1

0 0
0 2 4 6 8 10 0 2 4 6 8 10
x/D x/D

Fig. 11. Iso-contours of span-averaged oz and corresponding streamlines for G/D ¼ 0.25. Total of 40 nondimensional contours are
considered in between 5 and +5 for oz.
 ARTICLE IN PRESS
S. Sarkar, S. Sarkar / Journal of Fluids and Structures 26 (2010) 19–40 31

3 t/T = 0.0 A
B
2

y/D
C
1

0
0 2 4 6 8 10

3 t/T = 0.2 Separation

bubble
2

y/D 1

0
0 2 4 6 8 10

3 Stretched inner shear layer

t/T = 0.4

2
y/D

0
0 2 4 6 8 10

3 Negative roll
t/T = 0.6

2
y/D

0
0 2 4 6 8 10

3 t/T = 0.8 A+B

C
2
y/D

0
0 2 4 6 8 10
x/D

Fig. 12. Instantaneous oz contours for G/D ¼ 0.5. For details refer to Fig. 10.

t/ T = 0.4 3 t / T = 0.8
3
y/D

2 2
y/D

1 1
Negative roll Negative roll
0 0
0 2 4 6 8 0 2 4 6 8
10 10
3 3

2 2
y/D
y/D

1 1

0 0
0 2 4 6 8 0 2 4 6 8 10
10
x/D x/D

Fig. 13. Iso-contours of span-averaged oz and corresponding streamlines for G/D ¼ 0.5. For details refer to Fig. 11.
 ARTICLE IN PRESS
32 S. Sarkar, S. Sarkar / Journal of Fluids and Structures 26 (2010) 19–40

4 t/T = 0.0
B
3 A

y/D
2

1 A
C
0
0 2 4 6 8 10

4 t/T = 0.2

3
y/D
2

1 A

0
0 2 4 6 8 10

4 t/T = 0.4 Vortex C engulfs

vortex B
3
y/D

1 A

0
0 2 4 6 8 10

4 t/T = 0.6

3
y/D

1 A

0
0 2 4 6 8 10
x /D

Fig. 14. Instantaneous oz contours for G/D ¼ 1.0. For details refer to Fig. 10.

Interaction of vortex C
and boundary layer
3 t/T=0.0
C
3 t/T = 0.4
A

2 2
y/D

y/D

B
1 1

0 0
0 2 4 6 8 10 0 2 4 6 8 10
3 3

2 2
y/D

y/D

1 1

0 0
0 2 4 6 8 10 0 2 4 6 8 10
x /D x/D

Fig. 15. Iso-contours of span-averaged oz and corresponding streamlines for G/D ¼ 1.0. For details refer to Fig. 11.

before. The celerity of negative roll and positive vortex sheet appears marginally different: the positive vortex sheet
stays above the negative roll. The stretching process is expected to decay the circulation of the wake vortex because of
dissipation. The positive vortex sheet appears to interact with the boundary layer periodically and the perturbations of
 ARTICLE IN PRESS
S. Sarkar, S. Sarkar / Journal of Fluids and Structures 26 (2010) 19–40 33

the wall boundary layer downstream of x/D ¼ 5 are higher in this case. The span-averaged results seen in Fig. 13
for t/T ¼ 0.4 and 0.8 illustrate the similar flow visualization with alternating vortex shedding, stretching of the inner
shear layer and the wall boundary layer, resulting in convection of a dominant negative vortex. Streamlines illustrate
the appearance of downstream separation bubble, which is unsteady and coupled with vortex shedding from the
cylinder in a periodic manner.
As the gap ratio increases, the snapshots of spanwise vorticity (Fig. 14 for G/D ¼ 1.0 that corresponds to G/d ¼ 2.4)
over a time period of vortex shedding illustrates that the cylinder behaves closer to an isolated cylinder with an alternate
vortex shedding. However, this shedding of Kármán vortex, particularly the inner shear layer and the associated
positive vortex rolls (denoted by C), influences the wall boundary layer in a different fashion. The excitation of the
boundary layer downstream of x/D ¼ 5 appears to be substantial by this positive migrating rolls. In detail, downstream
of the cylinder a negative vortex from the wall boundary layer eventually elongates (denoted by B), forming a vortex
sheet that is seen to be engulfed by the positive convective vortex roll (Fig. 14, t/T ¼ 0.6). This creates an unsteady
bubble beneath the negative vortex sheet, the periodicity of which appears to be the same as that of the vortex shedding.
The span-averaged vorticity (Fig. 15) illustrates the coherent vortex rolls depicting the dynamics as explained before. It
explains a three-layer structure, which is slightly deflected away from the wall, where the inner shear layer periodically
interacts with the wall boundary layer. The rollup of boundary layer forming small-scale eddies because of excitation of
the convective coherent vortices is apparent downstream of x/D ¼ 5. The corresponding streamlines also suggest the
formation of a small unsteady bubble downstream of the cylinder.
To validate the wake dynamics evaluated in the presented LES, the iso-contours of spanwise vorticity (oz) are
compared in Fig. 16 to the corresponding experiment (Price et al., 2002) for G/D ¼ 0.25 and 0.5. The wake–boundary
layer interactions, formation of separation bubbles both upstream and downstream of the cylinder, stretching of shear
layers shed from the cylinder, rollup and then breaking down into smaller eddies are well represented. The numerical
visualization of instantaneous flow field illustrates a close resemblance with the experimental observations (Price et al.,
2002) in terms of vortex dynamics and flow structures.
Figs. 17 and 18 illustrate the overall trajectory of peak vorticity and the variation of vortex peak values along the
streamwise direction for G/D ¼ 0.5 and 1.0. Three vortices emanating from the outer shear layer, the inner shear layer
and the wall boundary layer are traced to reveal the vortex motion. For G/D ¼ 0.5, vortices from the inner shear layer
and the wall boundary layer travel in an almost parallel trajectory owing to the coupling between them (Fig. 17(a)) and
move in the wall-normal direction behind the cylinder (x/Do3.0). The outer shear layer vortices move almost parallel
to the wall. Near x/D ¼ 4.5, a strong mutual interaction between vortices is reflected by the cross-over of trajectories of
the shed vortices emerging a prominent negative vortex that travels downstream and remains almost parallel to the wall.
Fig. 17(b) indicates that the peak values of the shed vortices decay at a higher rate than that of the wall vortices near the
cylinder (x/Do4). Furthermore, the wall vortices decay rapidly in the vicinity of the cylinder and at a slower rate
downstream of x/D ¼ 4. This is attributed to the breakdown of the larger-scale vortex to the random small-scale eddies
resulting in a decrease of the magnitudes of lower frequency and the increase of higher frequency energy-content eddies.
It should be further noted that the inner shear layer and the wall boundary layer do not cancel each other behind the
cylinder, although of different signs of vorticity. This supports the experimental observation of Price et al. (2002), while
it contradicts the suggestion of Grass et al. (1984) and Taniguchi and Miyakoshi (1990).

Fig. 16. Instantaneous spanwise vorticity (oZ): (a) and (b) experiment at Re ¼ 1900 (Price et al., 2002); (c) and (d) present LES at
Re ¼ 1440. (a) and (c) G/D ¼ 0.25; (b) and (d) G/D ¼ 0.5.
 ARTICLE IN PRESS
34 S. Sarkar, S. Sarkar / Journal of Fluids and Structures 26 (2010) 19–40

5
G/D = 0.5
4

y/D
2

0
0 2 4 6 8 10
x/D

101 G/D = 0.5 Wall vortex

Inner vortex
Outer vortex
100
|ωz|max

10-1

10-2 10
0 1 2 3 4 5 6 7 8 9
x/D

Fig. 17. (a) Trajectory of vortex peak and (b) variation of vortex peak values for G/D ¼ 0.5 and Re ¼ 1440.

5
Wall vortex
Inner vortex
4 G/D=1.0
Outer vortex

3
y/D

0
0 2 4 6 8 10
x/D

101 G/D=1.0
Wall vortex
Inner vortex
Outer vortex
100
|ωz |max

10-1

10-2
0 2 4 6 8 10
x/D

Fig. 18. (a) Trajectory of vortex peak and (b) variation of vortex peak values for G/D ¼ 1.0 and Re ¼ 1440.
 ARTICLE IN PRESS
S. Sarkar, S. Sarkar / Journal of Fluids and Structures 26 (2010) 19–40 35

10 10

G / D=0.5 G / D=1.0
8 8

6 6
x/D

Wall vortex

x /D
Inner vortex
4 Outer vortex 4

2 2

0 0
0 0.4 0.8 1.2 1.6 2 0 0.4 0.8 1.2 1.6 2
Δt/ T Δt/ T

Fig. 19. Time history of spanwise vorticity peak along the streamwise direction for Re ¼ 1440; (a) G/D ¼ 0.5 and (b) G/D ¼ 1.0.

As the cylinder is moved away from the wall (G/D ¼ 1), the outer- and inner-side-shed vortices along with the wall
vortex can be separately identified far downstream of the cylinder (Fig. 18). The trajectories of vortex peak illustrate
that the vortices migrate parallel to each other. In this case, the effect of boundary layer is minimum and so both the
outer- and inner-shear layers roll up to produce almost identical vortices. Further, the outer and inner vortices have the
same decay profile (rapidly in the vicinity of the cylinder and at a slower rate downstream of x/D ¼ 4), which mean that
the interaction between them is a mutual process. The wall vortex appears to decay continuously, but at a slower rate
behind the cylinder.
Fig. 19 depicts the streamwise locations of the vortex peak with time for G/D ¼ 0.5 and 1.0. The slope of
the curve gives the vortex celerity. For G/D ¼ 0.5, the inner vortex travels with a velocity of 0.38UN up to t/T ¼ 0.2
and then suddenly accelerates due to attraction of outer vortices and travels with a speed of UN. This slower
celerity of the inner vortex at the beginning is attributed to the wall effect imposing on alternating vortex formation
near the base region. The outer vortex travels with a constant speed of 0.9UN. The wall vortex travels almost
with the same celerity of the lower vortex illustrating a pairing between them as explained earlier. Time history
of vortex celerity changes for G/D ¼ 1.0, which depicts that the wall vortex travels at a speed of 0.52UN, while the
outer- and inner-vortex travel at speeds of 0.65UN and 0.7UN, respectively. The speeds of the outer- and inner-vortex
come close to the speed of the Kármán vortex from an isolated cylinder [0.78UN, Lin et al. (1995)] as the gap ratio
increases.
The iso-surface of the spanwise component of instantaneous vorticity for different gap ratios is depicted in Fig. 20 to
visualize the 3-D flow structures and internal growth mechanism of shear layers after instability in proximity to a wall.
For a low gap ratio (G/D ¼ 0.25), the outer shear layer shed from the cylinder appears to remain laminar up to x/
D ¼ 2.5 and then the shear layer instability sets in with the appearance of longitudinal streaks that quickly break down
to turbulent flow downstream. To distinguish between the inner shear layer, which is of positive vorticity, and the
boundary layer of negative vorticity, different colours are used. As stated earlier, the boundary layer in this case is
highly deflected away from the wall, producing a large separation region due to its strong coupling with the inner shear
layer. It is commonly accepted that a separated laminar layer is inherently unstable and promotes the growth of
disturbances. A rapid growth of disturbances appears after the point of separation that breaks down to random
structures near x/D ¼ 3. The flow is turbulent downstream with the appearance of small-scale, higher frequency energy-
content eddies and loss of coherent structures. Thus, the generation of turbulence is prominent away from the wall,
leaving a relatively calm near-wall because of the detached layer.
With increase of the G/D ratio the effect of wall proximity is decreased and the breakdown of both the outer and
inner shear layers occurs relatively closer to the cylinder. The flow tends to retain its coherent structures with the
appearance of spanwise small-scale eddies. Further, to note that the defection of the boundary layer is inhibited as the
gap ratio is increased. For G/D ¼ 1, a thin separation bubble is observed as seen in Fig. 5 and the boundary layer
behaves almost like an attached layer. In this case, the receptivity to external disturbance is very prominent near the
wall with the appearance of longitudinal streaky structures. The perturbations created by the turbulent shear layers are
 ARTICLE IN PRESS
36 S. Sarkar, S. Sarkar / Journal of Fluids and Structures 26 (2010) 19–40

Fig. 20. Iso-surfaces of spanwise vorticity (oz ¼ 72.0) for (a) G/D ¼ 0.25, (b) G/D ¼ 0.5 and (c) G/D ¼ 1.0.

Fig. 21. Iso-surfaces of streamwise vorticity (ox ¼ 74.0) for G/D ¼ 0.5 and G/D ¼ 1.0.
 ARTICLE IN PRESS
S. Sarkar, S. Sarkar / Journal of Fluids and Structures 26 (2010) 19–40 37

amplified near the wall due to the inflexional layer, and then the nonlinear interactions of vortex stretching process
create these streaks, which are characteristics of turbulent flow.
Fig. 21 illustrates the iso-surface of ox for G/D ¼ 0.5 and 1.0 at the same time instant. The figure illustrates the
instability of the shear layers with the formation of streaky structures having a spanwise wavelength of 0.5D and this
instability occurs earlier for a larger gap ratio. The streaks become three-dimensional and break down to small-scale
eddies downstream. These small-scale vortices appear to overshadow the main vortices and the flow turns out to be
turbulent. It should be noted that Mansy et al. (1994) have given an empirical formula for the wavelength of the
streamwise structures in the near wake as 20Re0.5D using the data in the range of 300rRer1200. This empirical
relation estimates a wavelength of 0.52D at Re ¼ 1440, which is very close to the value of wavelength obtained from the
present simulation. Hence, the mode of instability of the shear layer in proximity to a wall would be close to that of the
unbounded condition, although the breakdown and generation of the small-scale eddies would differ as these are
sensitive to external perturbations.

3.3. Turbulence statistics

To understand the relative effects of G/D ratio on the wake and boundary layer interactions, the iso-contours of
Reynolds normal (u0 u0 =U 21 ) and shear (u0 v0 =U 21 ) stress behind the circular cylinder are presented in Fig. 22(a) and (b)
for the three gap ratios, G/D ¼ 0.25, 0.5 and 1.0. A few selected streamlines representative of trajectories of the shear
layers and the approaching boundary layer are superimposed to illustrate the correspondence between the mean flow
and the stresses.
In unbounded condition, the contours of time-averaged turbulent stresses of flow past a cylinder are found to be
symmetric about the centreline with a double-peak feature. The symmetry in normal and shear stress distributions is
lost completely as the cylinder is placed close to a wall, G/D ¼ 0.25 corresponding to G/d ¼ 0.6. In this case, the
stresses appear to develop towards the end of the recirculation region, indicating a delay in transition because of frozen
shear layers. The breakdown of the shear layers is reflected by the enhanced stress levels near x/D ¼ 3. The deflection of
the boundary layer due to the pairing with the inner shear layer, its stretching and convection over the downstream

0.002 0.004 0.006 0.008 0.010 0.012 0.014 0.016 0.018 0.020 -0.025 -0.019 -0.014 -0.008 -0.003 0.003 0.008 0.014 0.019 0.025
3 __ 3 __
/ / G/D=0.25 / / G/D=0.25
uu uv
2 2
y/D

y/D

1 1

0 0
0 3 6 9 12 0 3 6 9 12

3 G/D=0.5 3 G/D=0.5

2 2
y/D

y/D

1 1

0 0
0 3 6 9 12 0 3 6 9 12

G/D=1.0 G/D=1.0
3 3

2 2
y/D

y/D

1 1

0 0
0 3 6 9 12 0 3 6 9 12
x/D x/D

Fig. 22. Iso-contours of (a) Reynolds normal stress and (b) Reynolds shear stress for G/D ¼ 0.25, 0.5 and 1.0.
 ARTICLE IN PRESS
38 S. Sarkar, S. Sarkar / Journal of Fluids and Structures 26 (2010) 19–40

0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1 0.005 0.016 0.026 0.037 0.047 0.058 0.068 0.079 0.089 0.100

3 3
TKE Production
G/D=0.25 G/D=0.25
2 2

y/D
y/D

1 1

0 0
0 3 6 9 12 0 3 6 9 12

G/D=0.5 3 G/D=0.5
3

2 2

y/D
y/D

1 1

0 0
0 3 6 9 12 0 3 6 9 12

G/D=1.0 G/D=1.0
3 3

2 y/D 2
y/D

1 1

0 0
0 3 6 9 12 0 3 6 9 12
x/D x/D

Fig. 23. Iso-contours of (a) turbulent kinetic energy and (b) production for G/D ¼ 0.25, 0.5 and 1.0.

bubble are resolved in the evolution of enhanced stresses. In downstream, high values of the normal stresses occur away
from the wall along the trajectories of the shear layer. However, it should be noted that shear stress is not maximum on
the separating streamlines: the maximum value occurs a bit away where the velocity gradient is maximum. For
G/D ¼ 0.5 or G/d ¼ 1.2, the enhancement of turbulent stresses occurs along the trajectories of the shear layer, leaving a
relatively low value of stresses near the wall. It is interesting to note that the turbulent stresses appear to re-establish the
symmetry with the double-peak features as the cylinder is placed relatively away from the wall at G/D ¼ 1, which
corresponds to G/d ¼ 2.4. Here, the transition of the shear layers occurs much earlier with the formation of rolls closer
to the cylinder at x/D ¼ 1.5. In this case, the enhancement of both the normal and shear stresses is found to occur near
the wall, which initiates at the reattachment point of the bubble and becomes maximum at x/D ¼ 5.25. This illustrates
the excitation of boundary layer by the convective wake.
In addition to Reynolds stresses, the evolution of TKE along with its production ½PTKE ¼ u0i u0j ð@ūi =@xj Þ is shown in
Fig. 23 to highlight the influence of G/D. As expected, the enhancement of turbulence occurs along the trajectories of
the shear layers, leaving a relatively calm near-wall region for low gap ratios (G/D up to 0.5), which is very consistent
with the distribution of the normal stresses. This is attributed to the strong coupling between the approaching boundary
layer and the inner shear layer, leading to a rapid growth of disturbances with the appearance of small-scale eddies in
regions away from the wall: an example of mutual interactions between the boundary layer and the shear layer. In these
cases, the production occurs in the same locations of high TKE, where the turbulence extracts work from the mean
flow. This happens in regions of high turbulent stress and high spatial velocity gradient, aligned in the same direction.
Thus, the high production along the path of the wake is due to the coexistence of turbulent stress and shear in the wake
fluid. The delay in transition of shear layers following breakdown is also reflected in iso-contours of TKE and
production. For a large gap ratio (G/D ¼ 1.0), the cylinder remains well outside the boundary layer and the wake
dynamics emerges to be different from that of the other two cases, which has been reflected in the TKE and production
contours. In this case, the high TKE and production are not confined along the path of the shear layers; rather, the
augmentation is migrated towards the wall. The inner shear layer shed from the cylinder perturbs the inflexional
boundary layer generating higher frequency energy-content eddies and the turbulence: an example of receptivity of
external disturbances. The high production near the surface can be considered as the combined effect of a local
 ARTICLE IN PRESS
S. Sarkar, S. Sarkar / Journal of Fluids and Structures 26 (2010) 19–40 39

concentration of vorticity and a growing level of turbulence in the boundary layer. The almost symmetrical distribution
of TKE and production behind the cylinder further confirm that the effect of boundary layer on the shear layers is
minimal in this case, and the shear layers roll close to the cylinder.

4. Conclusions

Large-eddy simulations of flow past a circular cylinder in the vicinity of a wall have been carried out for three
different gap ratios at a Reynolds number in the shear layer transition region, following the experiment of Price et al.
(2002). An attempt is made to describe the physics of flow involving the nonlinear interactions of shear layers shed from
the cylinder and the approaching boundary layer. The present LES reveals that the G/D ratio has a strong influence on
shear layer instability, modification of wake dynamics apart from its coherent structure and small-scale motion.
When the inner shear layer lies within the boundary layer (Grd), a strong coupling between the inner shear layer and
the approaching boundary layer is observed, with suppression of vortex shedding and stretching of the shear layers.
Furthermore, the front stagnation point shifts towards the wall, generating an upward lift that deflects the boundary
layer, resulting in a relatively large separation behind the cylinder. The mutual interactions of the wall vortices with the
shed vortices create a remarkable difference in the wake dynamics: trajectories of shed vortices cross each other and an
inversion on the position of vortices occurs with respect to the unbounded case. The coupling between the inner shear
layer and the boundary layer also delays the shear layer instability and the appearance of a spanwise wavy structure,
which becomes three-dimensional and breaks down at the end of the recirculation region. In this case, the generation of
turbulence is mainly due to shear in the wake fluid, illustrating a highly active outer region along the wake trajectory
and a relatively calm near-wall.
The wake dynamics appears to be different as the cylinder is kept outside the boundary layer (G ¼ 2.4d). In this case,
the shear layers and the boundary layer can be separately identified. The breakdown of the shear layers occurs relatively
closer to the cylinder and the wake tends to retain its coherent structures with the appearance of small-scale eddies. The
Strouhal frequency becomes close to 0.2, although asymmetry in the wake shape is observed. What is interesting is that
the vortices shed from the cylinder inject momentum to the boundary layer during their convection, triggering its
transition. Thus, the generation of turbulence does not occur along the wake trajectories; rather, the augmentation of
turbulent stresses is found near the wall, initiating at the reattachment point of the downstream bubble: an example of
receptivity of external disturbances. The appearance of streaky structures near the wall is also evidence of transitional
boundary layer. Thus, it is worthwhile to note the differences in response of the boundary layer close to a migrating
wake as the gap ratio changes.

References

Angrilli, F., Bergamaschi, S., Cossalter, V., 1982. Investigation of wall induced modifications to vortex shedding from a circular
cylinder. ASME Journal of Fluids Engineering 104, 518–522.
Bearman, P.W., Zdravkovich, M.M., 1978. Flow around a circular cylinder near a plane boundary. Journal of Fluid Mechanics 89,
33–47.
Beaudan, P., Moin, P., 1994. Numerical experiments on the flow past a circular cylinder at sub-critical Reynolds number. Report No.
TF-62, Department of Mechanical Engineering, Stanford University.
Berger, E., Willie, R., 1972. Periodic flow phenomena. Annual Review of Fluid Mechanics 4, 313–340.
Buresti, G., Lanciotti, A., 1979. Vortex shedding from smooth and roughened cylinders in cross-flow near a plane surface. The
Aeronautical Quarterly 30, 305–321.
Choi, J.H., Lee, S.J., 2000. Ground effect of flow around an elliptic cylinder in a turbulent boundary layer. Journal of Fluids and
Structures 14, 697–709.
Dipankar, A., Sengupta, T.K., 2005. Flow past a circular cylinder in the vicinity of a plane wall. Journal of Fluids and Structures 20,
403–423.
Drikakis, D., 2003. Advances in turbulent flow computations using high-resolution methods. Progress in Aerospace Sciences 39,
405–424.
Fadlun, E.A., Verzicco, R., Orlandi, P., Mohd.-Yusof, J., 2000. Combined immersed boundary finite difference methods for three
dimensional complex flow simulations. Journal of Computational Physics 161, 35–60.
Germano, M., Piomelli, U., Moin, P., Cabot, W.H., 1991. A dynamic subgrid-scale eddy viscosity model. Physics of Fluids A 3,
1760–1765.
Grass, A.J., Raven, P.W.J., Stuart, R.J., Bray, J.A., 1984. The influence of boundary layer velocity gradients and bed proximity on
vortex shedding from free spanning pipelines. ASME Journal of Energy Resources Technology 106, 70–78.
 ARTICLE IN PRESS
40 S. Sarkar, S. Sarkar / Journal of Fluids and Structures 26 (2010) 19–40

Jacobs, R.G., Durbin, P.A., 2001. Simulations of bypass transition. Journal of Fluid Mechanics 428, 185–212.
Kravchenko, A.G., Moin, P., 2000. Numerical studies of flow over a circular cylinder at ReD ¼ 3900. Physics of Fluids 12, 403–417.
Lei, C., Cheng, L., Kavanagh, K., 1999. Re-examination of the effect of a plane boundary on force and vortex shedding of a circular
cylinder. Journal of Wind Engineering and Industrial Aerodynamics 80, 263–286.
Lilly, D.K., 1992. A proposed modification of the Germano subgrid-scale closure method. Physics of Fluids A 4, 633–635.
Lin, J.C., Towfighi, J., Rockwell, D., 1995. Instantaneous structure of the near-wake of a circular cylinder: on the effect of Reynolds
number. Journal of Fluids and Structures 9, 409–418.
Liou, T.M., Chen, S.H., Hwang, P.W., 2002. Large eddy simulation of turbulent wake behind a square cylinder with a nearby wall.
ASME Journal of Fluids Engineering 124, 81–90.
Lourenco, L.M., Shih, C., Characteristics of the plane turbulent near wake of a circular cylinder, a particle image velocimetry study,
data taken from Ma et al. (2000).
Ma, X., Karamanos, G.-S., Karniadakis, G.E., 2000. Dynamics and low-dimensionality of a turbulent near wake. Journal of Fluid
Mechanics 410, 29–65.
Mansy, H., Yang, P.-M., Williams, D., 1994. Quantitative measurements of three-dimensional structures in the wake of a circular
cylinder. Journal of Fluid Mechanics 270, 227–296.
Mittal, R., Balachandar, S., 1995. Effect of three-dimensionality on the lift and drag of no two-dimensional cylinders. Physics of Fluids
7, 1841–1865.
Mittal, R., Moin, P., 1997. Suitability of upwind-biased finite-difference schemes for large-eddy simulation of turbulent flows. AIAA
Journal 35, 1415–1417.
Morinishi, Y., Lund, T.S., Vasilyev, O.V., Moin, P., 1998. Fully conservative higher order finite difference schemes for incompressible
flow. Journal of Computational Physics 143, 90–124.
Muldoon, F., Acharya, S., 2005. Mass conservation in immersed boundary method. In: Proceedings of FEDSM 2005, FEDSM, 77301,
pp. 1–9.
Norberg, C., 1994. An experimental investigation of the flow around a circular cylinder: influence of aspect ratio. Journal of Fluid
Mechanics 258, 287–316.
Ong, L., Wallace, J., 1996. The velocity field of the turbulent very near wake of a circular cylinder. Experiments in Fluids 20, 441–453.
Orlanski, I., 1976. Simple boundary condition for unbounded hyperbolic flows. Journal of Computational Physics 21, 251–269.
Ovchinnikov, V., Piomelli, U., Choudhari, M.M., 2006. Numerical simulations of boundary-layer transition induced by a cylinder
wake. Journal of Fluid Mechanics 547, 413–441.
Price, S.J., Sumner, D., Smith, J.G., Leong, K., Paidoussis, M.P., 2002. Flow visualization around a circular cylinder near to a plane
wall. Journal of Fluids and Structures 16, 175–191.
Roshko, A., 1961. Experiments on the flow past a circular cylinder at very high Reynolds number. Journal of Fluid Mechanics 10,
345–356.
Sarkar, Sudipto, Sarkar, S., 2007. Immersed boundary method for simulating complex flows. In: Proceedings of ICFD’07, University
of Reading, UK.
Sarkar, S., Sarkar, Sudipto, 2007. Large-eddy simulations of cylinder boundary layer interactions. In: Proceedings of ACFD 7,
pp. 1349–1360.
Sarkar, S., Voke, P.R., 2006. Large-eddy simulation of unsteady surface pressure on a LP turbine blade due to interactions of passing
wakes and inflexional boundary layer. ASME Journal of Turbomachinery 128, 221–231.
Sarkar, S., 2007. The effects of passing wakes on a separating boundary layer along a low-pressure turbine blade through large-eddy
simulation. IMechE Journal of Power Energy 221, 551–564.
Sarkar, S., 2008. Identification of flow structures on a LP turbine blade due to periodic passing wakes. ASME Journal of Fluids
Engineering 130, 1–10.
Schlichting, H., 1979g. Boundary-Layer Theory, 7th ed. McGraw-Hill.
Taneda, S., 1965. Experimental investigation of vortex streets. Journal of the Physics Society of Japan 20, 1714–1721.
Taniguchi, S., Miyakoshi, K., 1990i. Fluctuating fluid forces acting on a circular cylinder and interference with a plane wall.
Experiments in Fluids 9, 197–204.
Vada, T., Nestegard, A., Skomedal, N., 1989. Simulation of viscous flow around a circular cylinder in the boundary layer near a wall.
Journal of Fluids and Structures 3, 579–594.
Williamson, C.H.K., 1996. Vortex dynamics in the cylinder wake. Annual Review of Fluid Mechanics 28, 477–539.
Zdravkovich, M.M., 1985. Forces on a circular cylinder near a plane wall. Applied Ocean Research 7, 197–201.
Zhang, S.L., 1997. GPBi-CG: generalized product-type methods based On Bi-CG for solving nonsymmetric linear systems. SIAM
Journal of Scientific Computing 18, 537–551.
Zovatto, L., Pedrizzetti, G., 2001. Flow about a circular cylinder between parallel walls. Journal of Fluid Mechanics 440, 1–25.