Numerical study of vortical structures around a wall-mounted cubic

obstacle in channel flow

Jong-Yeon Hwang and Kyung-Soo Yang

Citation: Phys. Fluids 16, 2382 (2004); doi: 10.1063/1.1736675

View online: http://dx.doi.org/10.1063/1.1736675
View Table of Contents: http://pof.aip.org/resource/1/PHFLE6/v16/i7
Published by the American Institute of Physics.

Related Articles
Multi-time delay, multi-point linear stochastic estimation of a cavity shear layer velocity from wall-pressure
measurements
Phys. Fluids 25, 017101 (2013)
Effect of acoustic coupling on power-law flame acceleration in spherical confinement
Phys. Fluids 25, 013602 (2013)
Rebound and jet formation of a fluid-filled sphere
Phys. Fluids 24, 122106 (2012)
Effect of orifice inner lip radius on synthetic jet efficiency
Phys. Fluids 24, 115110 (2012)
Effect of fluctuations on the onset of density-driven convection in porous media
Phys. Fluids 24, 114102 (2012)

Additional information on Phys. Fluids

Journal Homepage: http://pof.aip.org/
Journal Information: http://pof.aip.org/about/about_the_journal
Top downloads: http://pof.aip.org/features/most_downloaded
Information for Authors: http://pof.aip.org/authors

Downloaded 29 Jan 2013 to 138.250.72.107. Redistribution subject to AIP license or copyright; see http://pof.aip.org/about/rights_and_permissions
PHYSICS OF FLUIDS VOLUME 16, NUMBER 7 JULY 2004

Numerical study of vortical structures around a wall-mounted cubic

obstacle in channel flow
Jong-Yeon Hwang and Kyung-Soo Yanga)
Department of Mechanical Engineering, Inha University, Incheon, 402-020, Korea
共Received 30 January 2003; accepted 10 March 2004; published online 27 May 2004兲
Vortical structures around a wall-mounted cubic obstacle in channel flow are studied using
numerical simulation. Flows of low-to-moderate Reynolds numbers up to Re⫽3500 are considered.
The objective of this work is to elucidate characteristics of coherent vortical structures produced by
the presence of the wall-mounted cubic obstacle, including horseshoe vortex systems upstream of
the obstacle, lateral vortices in the vicinity of the two lateral faces of the cube, and hairpin vortices
in the near-wake region. As the flow approaches the cube, the adverse pressure gradient produces
three-dimensional boundary-layer separation, resulting in the formation of laminar horseshoe
vortices. As the Reynolds number increases, the structure of the horseshoe vortex system becomes
complex and the number of vortices increases in pairs. The distribution of skin friction on the
cube-mounted wall reflects the effect of the horseshoe vortices. Unsteady horseshoe vortex systems
are hardly found as long as the upstream flow is fully viscous; they are obtained when the cube is
placed in the entrance region of a developing channel flow. The unsteady horseshoe vortex systems
are characterized by a repeated process of generation, translation, and mutual merging of the
vortices. The laminar wake is characterized by a pair of spiral vortices behind the obstacle; distinct
singular points are identified leading to consistent flow topology. In the case of a turbulent wake,
however, it is observed that the flow becomes less coherent in the near-wall region downstream of
the obstacle. Instead, coherent structures such as lateral vortices and hairpin vortices are found in the
vicinity of the two lateral faces of the cube and in the turbulent near-wake region, respectively.
Quasiperiodic behaviors of those vortices are noticed and their frequencies are computed. The
translating speed of the head portion of a hairpin vortex is lower than the streamwise mean velocity
at that location. In the vicinity of the lower wall downstream of the cube, vortical structures of
various length scales are identified; they become gradually elongated downstream of the flow
reattachment. © 2004 American Institute of Physics. 关DOI: 10.1063/1.1736675兴

I. INTRODUCTION with rotor and blades, to name a few. Flows of low-to-

Vortical structure around a wall-mounted obstacle has
been a long-lasting research topic not only for academic but
also practical reasons. Since various types of vortices pro-
duced by the protruding surface of an obstacle significantly
enhance momentum and heat transfer downstream of the ob-
stacle, understanding their characteristics is essential for
many engineering applications where the vortices play an
important role. Most of the research on this topic, however,
has been carried out in an open flow domain, even though
there exist a number of practical applications where a wall-
mounted obstacle is present in internal flow.
In this study, physical aspects of the vortical structures
generated by a wall-mounted obstacle in internal flow are
investigated using numerical simulation. Especially, the
horseshoe vortices and the hairpin vortices associated with a
cubic obstacle mounted on one wall of a channel 共Fig. 1兲 are
studied. This specific internal flow has particular importance
in association with printed circuit boards with chips
mounted, heat exchangers with cooling fins, and turbines
a兲
Author to whom all correspondence should be addressed. Electronic mail:
ksyang@inha.ac.kr
moderate Reynolds numbers 共Re兲 up to Re⫽3500 are con-
sidered; here Re is based on the inlet streamwise mean ve-
locity and the spanwise width of the cubic obstacle 共b兲 which
is the same as one half of the channel height.
Horseshoe vortices are produced by three-dimensional
boundary-layer separation near the junction of the lower
channel wall and the obstacle. The boundary layer experi-
ences an adverse pressure gradient in the main flow direction
due to the blockage effect of the obstacle. When the adverse
pressure gradient is significant, the boundary layer separates
and a vortex which wraps around the obstacle and trails off is
formed. Its overall shape resembles a horseshoe from which
the name of the vortex originates. Horseshoe vortices have
been attracting many researchers’ attention for decades as
they are often observed in junction flows of various types of
engineering application; for example, junction of fuselage
and wing of an airplane, junction of bridge pier and river
bed, and junction of turbine blade and rotor, among others.
Horseshoe vortices yield large shear stresses near the junc-
tion, and unsteady horseshoe vortices may cause serious
damages on the structure due to fatigue and erosion.
It has been revealed by many researchers that horseshoe
vortices exhibit diverse characteristics depending on flow pa-

1070-6631/2004/16(7)/2382/13/$22.00 2382 © 2004 American Institute of Physics

Downloaded 29 Jan 2013 to 138.250.72.107. Redistribution subject to AIP license or copyright; see http://pof.aip.org/about/rights_and_permissions
Phys. Fluids, Vol. 16, No. 7, July 2004
FIG. 1. Physical configuration.
rameters such as geometry, Re, boundary-layer thickness,
and so forth. Baker1 experimentally showed that as the Rey-
nolds number increases, the number of horseshoe vortices
also grows and the vortex system becomes more complex
and unsteady. Kubendran and Harvey2 made an attempt to
control the horseshoe vortex systems. Kelso and Smits3 stud-
ied the unsteady characteristics of the horseshoe vortices pro-
duced by a jet instead of an obstacle. Besides these, other
experimental investigations were also performed on the
horseshoe vortices.4 – 6
Technical limitations of experimental methods, however,
keep researchers from obtaining a clearer understanding of
the flow physics. With the recent advent of powerful com-
puters and efficient numerical algorithms, researchers have
been carrying out numerical simulations to supplement the
drawbacks of experimental techniques.7–9 Kaul and Kwak7
studied horseshoe vortices around a circular cylinder
mounted on a flat plate, and classified their mechanisms.
Visbal8 considered the same geometry, and claimed that the
far upstream saddle point could be an attachment point.
Coon and Tobak9 investigated Re dependency of laminar
horseshoe vortices.
Hunt et al.10 carried out kinematical studies of the flows
around free or surface-mounted obstacles by applying topol-
ogy to experimental flow visualizations. They suggested a
simple mathematical relation to be satisfied by singular
points in a section of a flow field, and applied it to the flow
around a cubic obstacle mounted on a flat plate. For a de-
tailed survey of horseshoe vortices, Ballio, Bettoni, and
Franzetti11 systematically analyzed the time-averaged char-
acteristics of the horseshoe vortex system associated with a
symmetric obstacle for both laminar and turbulent flows.
They attempted to correlate the maximum value of the shear
stress at the intersection of the symmetry plane with the bot-
tom surface in front of the body, the position of the primary
separation, and the longitudinal and vertical coordinates of
the main vortex center, with the flow parameters such as
h/b, ␦ * /b, and Re␦ , where h, ␦*, b, and Re␦ represent
* *
obstacle height, displacement thickness of undisturbed
boundary layer at the position of the obstacle axis, spanwise
width of the obstacle, and the Reynolds number based on the
freestream velocity and ␦*, respectively. They concluded that
laminar junction vortices are more evidently influenced by
the characteristics of the flow than turbulent ones.
Even though the horseshoe vortex systems remain lami-
nar for Re currently considered, the flow region downstream
of the obstacle is quite complex and can become even turbu-
lent for moderate Re. The characteristics of coherent struc-
tures 共CS兲 associated with the free-shear layer downstream
of the obstacle, especially those of hairpin vortices, are also
studied in detail.
Downloaded 29 Jan 2013 to 138.250.72.107. Redistribution subject to AIP license or copyright; see http://pof.aip.org/about/rights_and_permissions
Numerical study of vortical structures 2383
Hairpin vortices were studied by many researchers in
order to reveal the flow mechanisms related to generating
turbulence. Haidari and Smith12 performed experiments to
study generation and development of hairpin vortices. They
reported that flow injection can induce disappearance and
regeneration of hairpin vortices. Elavarasan and Meng13 car-
ried out experimental investigation on CS in the turbulent
wake downstream of a tab. They observed counter-rotating
vortex pairs 共CVP兲 past the tab and hairpin vortices far
downstream, and reported that the frequency of the hairpin
vortices is higher than that of CVP. Perry and Steiner14 car-
ried out a topological study on the large-scale CS in turbulent
wakes behind bluff bodies using phase-averaged vector
fields. Numerical studies on CS have been performed also.
Yang, Spalart, and Ferziger15 described the generation of ⌳
vortices and their subsequent development into hairpin vor-
tices in a decelerating boundary layer using direct numerical
simulation. Lesieur16 studied mutual interactions of coherent
structures in shear layers using large eddy simulation 共LES兲.
Suksangpanomrung, Djilali, and Moinat17 performed LES
and reported that hairpin vortices are generated near the re-
attachment point in backward-facing step flows.
Even though flows with turbulent inlet-velocity fields for
the same flow geometry as the current one were studied by
some authors,18 –23 a comprehensive study on flows with
laminar inlet-velocity fields and subsequent development of
laminar horseshoe vortices and unsteady wakes is very rare.
Unlike previous numerical simulations in an open domain,7,8
which imposed a symmetric condition on the center plane,
present simulation does not have such restriction so that an-
tisymmetric modes, if they exist, can be captured. The objec-
tive of this investigation is to accurately describe the vortical
structures around a wall-mounted cubic obstacle in channel
flow, including the steady and unsteady behaviors of the
horseshoe vortex systems upstream of the obstacle and gen-
eration and development of hairpin vortices in the near-wake
region downstream of the obstacle. Particular attention is
paid to the unsteadiness of the vortices because presence of
the upper wall can have more influence on the unsteady be-
havior of the vortices than on the steady one. A parametric
study shows a growing kinematical complexity of the flow
field as the Reynolds number increases.
II. FORMULATION AND NUMERICAL METHOD
All the variables except pressure and wall-shear stress
are nondimensionalized by the inlet bulk mean velocity (U B )
between the two channel plates and b. Pressure and wall-
shear stress are nondimensionalized by 1/2␳ U B2 where ␳ de-
notes fluid density. It is implied that the variables indicated
in all the following figures are nondimensionalized accord-
ingly, unless specified otherwise. The code uses a nonuni-
form Cartesian staggered grid in a finite-volume approach.
The governing equations are the continuity and the incom-
pressible Navier–Stokes equations
⳵u j
⫽0, 共1兲
⳵x j
2384 Phys. Fluids, Vol. 16, No. 7, July 2004 J.-Y. Hwang and K.-S. Yang

TABLE I. Summary of the computational cases. Here, ⌬y min denotes the minimum cell size near the solid
walls.

Inlet conditions Re Total resolution L/h Resolution around the cube ⌬y min /h W/h

5 128⫻64⫻96 10 32⫻32⫻32 0.006 7

15 128⫻64⫻96 10 32⫻32⫻32 0.006 7
25 128⫻64⫻96 10 32⫻32⫻32 0.006 7
50 128⫻64⫻96 10 32⫻32⫻32 0.006 7
Parabolic 100 128⫻64⫻96 10 32⫻32⫻32 0.006 7
150 128⫻64⫻96 10 32⫻32⫻32 0.006 7
250 128⫻64⫻96 10 32⫻32⫻32 0.006 7
350 224⫻96⫻96 15 32⫻48⫻32 0.005 7
900 224⫻96⫻96 15 32⫻48⫻32 0.005 7
1500 224⫻96⫻96 15 32⫻48⫻32 0.005 7

Uniform 1000 224⫻96⫻144 15 48⫻48⫻48 0.004 7

3500 256⫻96⫻144 18 48⫻48⫻48 0.004 7

⳵ui ⳵ui ⳵p 1 ⳵ 2u i ⬎4, which is normally considered sufficient to exclude major

⫹u j ⫽⫺ ⫹ ,
⳵t ⳵x j ⳵ x i Re ⳵ x j ⳵ x j
where u 1 , u 2 , and u 3 共or u, v , w兲 are velocities in x 1
共streamwise兲, x 2 共normal兲, and x 3 共spanwise兲 directions 共or x,
y, z兲, respectively, and p is pressure.
All the spatial derivatives including convective terms are
discretized by the second-order central differencing. To ad-
vance the solution in time, a fractional step method24 is em-
ployed; the resulting Poisson equation is solved by a multi-
grid method. The time-advancement of the momentum
equations is hybrid; the convective terms are explicitly ad-
vanced by a third-order Runge–Kutta scheme and the vis-
cous terms implicitly by the Crank–Nicolson method. The
Courant number is fixed 关Courant–Friedricks–Lewey 共CFL兲
⫽1.5兴, and time-step size is dynamically determined accord-
ingly. Choi and Moin25 reported that the time step corre-
sponding to CFL⬇3 was fine enough to resolve the
Kolmogorov time scale and to sustain turbulent fluctuations
in their riblet computations. For a detailed description of the
numerical method used in the current study, see Refs. 20
and 26.
III. CHOICE OF PARAMETERS AND BOUNDARY
CONDITIONS
In all the cases considered in this work, the length of
each side of the cubic obstacle is fixed as one half of the
channel height (h/H⫽0.5). For low Re equal to or below
250, the streamwise 共L兲, normal 共H兲, and spanwise 共W兲
lengths of the computational domain are 10h, 2h, and 7h,
respectively. For high Re above 250, the streamwise length
of the domain is increased by 5h or 8h, while the other
lengths remain unchanged.
Periodic boundary condition is employed in the span-
wise direction. This, in fact, implies that obstacles are peri-
odically mounted in z. However, the spanwise size of the
computational domain is selected such that the influence of
the neighboring obstacles is insignificant. A parametric study
revealed that in the current flow configuration the vortical
structures around the cubic obstacle are not affected by W,
provided W/h⬎5. Ballio, Bettoni, and Franzetti11 also re-
ported that for most of the data they analyzed, it is W/b
共2兲 effects of W. At all the solid walls, the no-slip boundary
condition is imposed; at the outlet of the domain, a convec-
tive boundary condition27 is used.
A special care is necessary at the inlet of the domain.
When a fully developed laminar parabolic profile is imposed,
a life-cycle of horseshoe vortices characterized by formation
and mutual merging in a time-periodic fashion is not ob-
served. To obtain such a vortex system, however, it turns out
that it is necessary to use a developing channel flow which
contains an inviscid core. It will be discussed in more detail
in the next section. In that case, a simple uniform flow is
imposed at the inlet.
The front face of the cube is located at x⫽3 for all the
cases except the case with the uniform inlet profile at Re
⫽3500 (x⫽5 in this particular case as explained later兲. The
number of control volumes was determined by grid-
refinement study. At Re equal to or below 250, the finest
numerical resolution employed was 128⫻64⫻96 in x, y, and
z directions, respectively. For high Re above 250, the grid
was refined up to 256⫻96⫻144 until the change in the skin-
friction coefficient which is very sensitive to the numerical
resolution was unnoticeable. The current numerical resolu-
tion is approximately 6.75 times finer than that of Shah20 at
Re⫽3000. For the case with the finest resolution, about
15 000 time steps are typically required for one flow-through.
Key numerical parameters for all the cases reported in this
article are summarized in Table I.
IV. RESULTS AND DISCUSSION
A. Steady horseshoe vortices
With laminar parabolic velocity profiles as inlet condi-
tion, computations were carried out until the steady state was
reached. Figures 2共a兲 and 2共b兲 present streamlines at Re
⫽350 on the center x-y plane (z⫽3.5) and on the x-z plane
corresponding to the first grid point away from the wall,
respectively. In Fig. 2共a兲, clearly seen is the flow separation
far upstream of the obstacle; the corresponding point in Fig.
2共b兲 is often called a saddle point of attachment (S a ). One
can also notice an attachment point just in front of the ob-
stacle; the corresponding point in Fig. 2共b兲 is often called a

Downloaded 29 Jan 2013 to 138.250.72.107. Redistribution subject to AIP license or copyright; see http://pof.aip.org/about/rights_and_permissions
Phys. Fluids, Vol. 16, No. 7, July 2004 Numerical study of vortical structures 2385

FIG. 3. Streamlines at ⌬y min away from the lower channel wall 共see Table
I兲: 共a兲 Re⫽5, 共b兲 Re⫽25, 共c兲 Re⫽50, 共d兲 Re⫽100, 共e兲 Re⫽150, and 共f兲
Re⫽250.

Figure 3 shows streamlines near the lower wall for vari-

ous cases of Re. In the case of very low Re 关Fig. 3共a兲兴, the
FIG. 2. Streamlines at Re⫽350; 共a兲 on the center plane (z⫽3.5) and 共b兲 on flow does not separate; consequently no horseshoe vortex is
the x-z plane at y⫽0.0053. formed. For larger Re 关Figs. 3共b兲 through 3共f兲兴, horseshoe
vortices do appear. The LS which segregates the near-
obstacle region from the incoming flow is clearly seen in
nodal point of attachment (N a ). Separation and the subse- Figs. 3共b兲 through 3共f兲, and the flow pattern inside the LS
quent vortex roll-up are the essential procedures for the for- becomes more complex with increasing Re. The distance be-
mation of a horseshoe vortex. At Re⭐20, the flow separation tween the far-upstream saddle point and the upstream face of
does not occur; no horseshoe vortex is observed. the cube grows as Re increases. So does the distance be-
Downstream of the obstacle, a three-dimensional wake is tween the far-downstream nodal point and the rear face of
formed 关Figs. 2共a兲 and 2共b兲兴. The streamline which passes the cube.
through the saddle point far upstream of the obstacle and The blockage effect of the obstacle creates an adverse
trails around the obstacle forms a ‘‘line of separation’’ 共LS兲 pressure gradient which makes the flow separate and trail off
on the lower wall. This means that in the three-dimensional along the obstacle, forming a horseshoe vortex. In Fig. 4,
flow region upstream of the obstacle, streamlines approach- contours of magnitude of the shear stress tangential to the
ing LS turn their directions upward away from the wall. Op- lower wall at Re⫽900 are shown. The values were obtained
posite is the case near the ‘‘line of attachment’’ 共LA兲 formed by taking the magnitude of the vectorial sum of the shear
by the streamline passing through the nodal point located stresses in x and z. One can identify a horseshoe-shaped re-
immediately upstream of the obstacle. Consequently, on the
x-z plane very near the lower wall, streamlines converge into
LS, and diverge from LA. Since formation of a horseshoe
vortex is preceded by flow separation, it must be confined
inside the corresponding LS. It should be noted, however,
that downstream of the obstacle, LS and LA are the results of
a complex pattern of streamwise vortical structures. In Fig.
2共b兲, one can identify another LS passing through the saddle
point (S s ) closer to the cube. In fact, there are two very
closely spaced nodal (N a ) and saddle (S s ) points near x
⫽2.15 and z⫽3.5 in Fig. 2共b兲. They are caused by the pres-
ence of the secondary vortex there which is too weak to be
visible in Fig. 2共a兲. The two nodal points symmetrically lo-
cated downstream of the obstacle are foci; the flow ap-
proaches these points in a spiral fashion and simultaneously FIG. 4. Contours of magnitude of tangential shear stress on the lower wall
moves upwards to form spiral vortices. Re⫽900.

Downloaded 29 Jan 2013 to 138.250.72.107. Redistribution subject to AIP license or copyright; see http://pof.aip.org/about/rights_and_permissions
2386 Phys. Fluids, Vol. 16, No. 7, July 2004
FIG. 5. Normalized skin-friction coefficient on the lower wall along the
centerline z⫽3.5; 共a兲 upstream of the obstacle and 共b兲 a close-up view near
the obstacle.
gion of strong shear stress around the obstacle, of which
magnitude is approximately seven to ten times larger than
that of the surrounding. The intense shear stress may cause
erosion of the lower wall, and a safety issue must be seri-
ously considered for any construction on the floor of a river
or of the ocean.
Figure 5 exhibits the skin-friction coefficient (C f
⫽␶w/1/2 ␳ U B2 ) normalized with that of the plane channel flow
corresponding to the same Re (C f 0 ) along the center line
(z⫽3.5) at some Reynolds numbers considered in this study.
Here, ␶ w is the shear stress on the lower wall. Figure 5共b兲 is
a close-up view of C f /C f 0 near the obstacle. The skin-
friction coefficient vanishes at the singular points. The zero-
friction point near the obstacle 关Fig. 5共b兲兴 corresponds to
N a , and the far-upstream one corresponds to S a in Figs. 2共b兲
and 3. There exists no zero-friction point at Re⫽5. There
thus appears no singular point nor horseshoe vortex. The
number of singular points are doubled in the case of Re
⫽350, which implies a qualitative change in the horseshoe
vortex system. It will be addressed in more detail later. As
noticed in Fig. 4, the magnitude of wall-shear stress is large
beneath the horseshoe vortex. This is consistent with Baker’s
experimental results.1 Compare Fig. 5共a兲 with Fig. 2共a兲 for
Re⫽350. It is also interesting to note that the location of
maximum skin-friction does not shift much with increasing
Re.
The distance between the far-upstream saddle point and
the upstream face of the obstacle (l sp) is in general a func-
tion of several factors as follows:1
l sp /h⫽ f 共 Re,y/h,H/h, ␦ * /h 兲 .
Since the obstacle here has a fixed geometry and is contained
in a fully viscous channel flow, Eq. 共3兲 can be reduced to
Downloaded 29 Jan 2013 to 138.250.72.107. Redistribution subject to AIP license or copyright; see http://pof.aip.org/about/rights_and_permissions
J.-Y. Hwang and K.-S. Yang
FIG. 6. Correlation of the distance between the far-upstream saddle point
and the upstream face of the obstacle (l sp) with Re on the lower wall. Here,
l sp was computed at ⌬y min away from the lower wall 共see Table I兲.
l sp /h⫽F 共 Re,y/h 兲 . 共4兲
A least-squares fitting can be employed to obtain a relation
between l sp and Re on the lower wall. That is
l sp /h⫽ 再 0.595 log Re⫺0.339,
0.770 log Re⫺0.564,
Re⬍300;
300⬍Re⬍1500.
共5兲
Figure 6 exhibits these lines along with the numerical values;
a qualitative change can be identified near Re⫽300. The vor-
tex system develops into a more complex one near this value
of Re. Compare Figs. 2共b兲 with 3共f兲. One can conjecture that
this development may be related to the appearance of the
new vortices near the lower channel wall. Ballio, Bettoni,
and Franzetti11 reported similar correlations 共see p. 236 of
Ref. 11兲, but for the sake of comparability of the different
obstacle shapes they measured the ‘‘saddle-point distance’’
from an ideal axis placed inside the obstacle at a distance of
one half of the spanwise width from the leading edge.
As Re increases, the number of the horseshoe vortices
also increases in pairs. Figure 7 shows streamlines on the
center plane for four different values of Re. In the case of
Re⫽50 关Fig. 7共a兲兴, the streamline pattern does not show a
definite vortex system, although separation takes place at x
⫽2.32 and a far-upstream saddle point can be identified in
Fig. 3共c兲. At Re⫽250 关Fig. 7共b兲兴, a complete horseshoe vor-
tex system can be identified. At higher Re 关Figs. 7共c兲 and
7共d兲兴, new primary vortices 共Vortex 2 and Vortex 3兲 are cre-
ated along with the corresponding secondary vortices be-
tween the primary ones. The direction of rotation of the sec-
ondary vortices is opposite to that of the primary ones.
At Re⫽1500, a six-vortex system was observed 关Fig.
7共d兲兴. It should be noted that in all the cases reported so far
in this article the horseshoe vortices were found to be steady.
An unsteady horseshoe vortex system for which we specifi-
cally mean a horseshoe vortex system undergoing cycles of
formation and merging1 was not detected in the range of Re
we investigated. We further increased Re up to 3500 which is
close to the theoretical value of the critical Re for the inlet
parabolic profile. However, the ‘‘unsteady’’ behavior of the
horseshoe vortices was not observed either, and the number
of vortices remained as six. This obstructed channel flow is
共3兲
expected to be less stable than the inlet channel flow. Thus
our simulation leads one to conjecture that unsteady laminar
horseshoe vortices can hardly appear in the fully developed
Phys. Fluids, Vol. 16, No. 7, July 2004 Numerical study of vortical structures 2387

FIG. 8. Streamlines on each face of the obstacle, Re⫽250; 共a兲 front, 共b兲
lateral, 共c兲 top, and 共d兲 rear, taken at 0.01h away from each face of the
obstacle.

the front face 关Fig. 8共a兲兴 and wraps around the obstacle 关Figs.
8共b兲, 8共c兲, and 8共d兲兴. The foci on the lateral 关Fig. 8共b兲兴 and
the top faces 关Fig. 8共c兲兴 represent spiral flow patterns. At a
lower Reynolds number 共Re⫽150, Fig. 9兲, however, the foci
disappear and the flow becomes much simpler.

B. Unsteady horseshoe vortices

According to Baker’s experiment,1 there exists a certain
FIG. 7. Streamlines on the center plane (z⫽3.5) upstream of the obstacle; range of ␦* for a given Re, in which the horseshoe vortex
共a兲 Re⫽50, 共b兲 Re⫽250, 共c兲 Re⫽900, and 共d兲 Re⫽1500. system is unsteady. Although his result was obtained for a
circular cylinder mounted on a flat plate, one can get an
insight from it, which can be applied to the internal flow
laminar internal flows, although the wake region downstream under consideration. In Fig. 10, the dashed line, taken from
of the obstacle tends to become turbulent even at moderate Baker’s experiment, indicates the boundary between the
Re between Re⫽350 and Re⫽900. This conjecture is sup- steady and unsteady regions. From this figure, one can notice
ported by the fact that most of the studies on unsteady lami- that unsteady horseshoe vortex systems can be obtained only
nar horseshoe vortices1,5,7,8 were conducted for the junction when Re⬎2800 and the displacement thickness of the
flows where the boundary layer thickness is thin compared
with the height of an obstacle, and also by Baker’s
observation1 that the thicker the displacement thickness of
the boundary layer, the higher the threshold value of Re for
the appearance of unsteady horseshoe vortices. This point
will be addressed in more detail in the next section. Shah20
presented his LES results on a fully turbulent flow of Re
⫽3000 in the same geometry as ours. He separately com-
puted the fully developed turbulent channel flow at the same
subcritical Re, and used the instantaneous velocity fields at a
given streamwise location as the inlet conditions of the LES
at the corresponding times. Thus his turbulent horseshoe vor-
tices at Re⫽3000 共which is lower than the maximum Re we
tried with the parabolic laminar inlet velocity profile兲 cannot
be regarded as ‘‘natural’’ because turbulence was ‘‘supplied’’
from the inlet.
Compared with the flow around a wall-mounted circular
cylinder, the flow structure around the cube is quite compli-
cated. In Fig. 8, the streamline pattern on each face of the FIG. 9. Streamlines on each face of the obstacle, Re⫽150; 共a兲 front, 共b兲
obstacle at Re⫽250 is shown. From this figure, one can de- lateral, 共c兲 top, and 共d兲 rear, taken at 0.01h away from each face of the
duce the separation structure around the cube. The flow hits obstacle.

Downloaded 29 Jan 2013 to 138.250.72.107. Redistribution subject to AIP license or copyright; see http://pof.aip.org/about/rights_and_permissions
2388 Phys. Fluids, Vol. 16, No. 7, July 2004 J.-Y. Hwang and K.-S. Yang

FIG. 10. Dependency of the unsteady region on Re and displacement thick-

ness 共␦*兲 for Baker’s experiment 共Ref. 1兲. Here, h denotes the diameter of
the circular cylinder in his experiment. Symbols indicate h/ ␦ * -Re relations
at various distances from the leading edge in Blasius flow.

boundary layer is much smaller than the diameter of the ob-

stacle. In the range of Re⬍2800, it is not possible to obtain
an unsteady system regardless of ␦*.
In the case of fully developed parabolic inlet profiles, the
magnitude of the ‘‘effective’’ ␦* is comparable to that of the
spanwise width of the cubic obstacle. Therefore, according to
Fig. 10, the vortex system is always steady in the laminar
range of Re. Figure 10, in turn, suggests that an unsteady
horseshoe vortex system should be obtained if the obstacle is
placed in the entrance region of the channel flow where an
inviscid core exists and ␦* on each wall is very thin. To FIG. 11. Repeated quasiperiodic behavior of an unsteady horseshoe vortex
determine a proper location of the obstacle from the inlet system on the center plane (z⫽3.5); Re⫽3500.
where a uniform flow is imposed, the relations between h/ ␦ *
and Re at various distances from the leading edge in Blasius
flow were computed and indicated by the symbols in Fig. 10.
horseshoe vortices around a wing-body junction with St
Based on this calculation, was selected one target case of
⫽0.05 based on the undisturbed boundary-layer thickness
Re⫽3500 and an obstacle location of x⫽5, which belongs to
and the approaching main-stream velocity. They reported
the unsteady region. It should be recalled that this value of
that this unsteadiness seems to be produced by fluctuations in
Re was the maximum Re we tried with the parabolic inlet
the momentum and vorticity of fluid from the outer part of
velocity profile with the intention of obtaining an unsteady
the boundary layer which is recirculated as it impinges on
horseshoe vortex system. Obviously, the flow under current
the leading edge of the wing. Larousse, Martinuzzi, and
investigation is quite different from Blasius flow. However,
Tropea28 and Martinuzzi and Tropea29 also presented experi-
in the entrance region where the boundary layers are devel-
mental results on the unsteadiness of turbulent horseshoe
oping at both walls, one could make a reasonable guess in
vortices for the same flow configuration as ours at Re
selecting the proper location of the obstacle using Baker’s
⫽50 000 and Re⫽40 000, respectively. How the unsteadiness
result 共Fig. 10兲. It turned out that was the case.
of the turbulent horseshoe vortices mentioned above is re-
Figure 11 shows a quasiperiodic behavior of the horse-
lated to ours is by no means clear, but their findings may
shoe vortex system obtained at this particular condition. Vor-
shed light on explaining why a high-inertia core 共the inviscid
tex 2 approaches Vortex 1 with approximately the same
core兲 is needed to obtain unsteady horseshoe vortices in the
speed as the mean velocity, and then finally merges with
current flow configuration.
Vortex 1. Vortex 1⬘ is a secondary vortex and must be present
for the no-slip condition to be satisfied on the wall. Concur-
C. Horseshoe vortices in a developing channel flow
rently, a new vortex, ‘‘Vortex 3’’ is born, and repeats the
processes of translation and merging with Vortex 1 approxi- As indicated in the previous section, the horseshoe vor-
mately at St⫽0.25 based on U B and h. This repeated behav- tices behave differently depending upon the type of the
ior is different from other researchers’ previous observations; boundary layer where they are in, even at the same value of
Baker1 reported that Vortex 1 and Vortex 2 periodically move Re. This fact motivates us to further investigate the charac-
back and forth, and Visbal’s numerical result8 revealed a pe- teristic differences between the two cases; one is ‘‘Type 1’’
riodic production and disappearance of Vortex 1. The differ- where the obstacle is located in a fully developed channel
ence may be due to the facts that their geometries are differ- flow, and the other is ‘‘Type 2’’ where the obstacle is located
ent from ours and that Visbal used a symmetric assumption in a developing channel flow where the boundary layer thick-
which is rather restrictive. It should be noted that Devenport ness is small compared with h. All of the steady results
and Simpson6 found low-frequency oscillations of turbulent shown so far belong to Type 1. For direct comparison with

Downloaded 29 Jan 2013 to 138.250.72.107. Redistribution subject to AIP license or copyright; see http://pof.aip.org/about/rights_and_permissions
Phys. Fluids, Vol. 16, No. 7, July 2004
FIG. 12. Streamlines in a developing channel flow, Re⫽250; 共a兲 on the
center x-y plane (z⫽3.5) and 共b兲 at y⫽0.006.
those results, some additional simulations were carried out
using the same geometry as Type 1 and an uniform inlet
velocity profile.
Figures 12共a兲 and 12共b兲 show streamlines at Re⫽250 on
the center x-y plane (z⫽3.5) and at y⫽0.006, respectively.
Figure 12共a兲 reveals l sp⫽0.79 which corresponds to Re⫽80
of Type 1 共Fig. 6兲. Furthermore, one can notice a significant
qualitative difference between Figs. 12共b兲 and 3共f兲, while the
streamline pattern in Fig. 12共b兲 is quite similar to a low-Re
case of Type 1 关Fig. 3共d兲兴. In Fig. 13, the distribution of
skin-friction coefficient corresponding to each of the three
cases mentioned above is shown along the centerline of the
lower wall. Here, one can also identify a similarity between
Re⫽250 of Type 2 and its lower-Re counterpart of Type 1.
All these observations suggest that the current definition
of Re be inappropriate for the horseshoe vortices of Type 2;
a ‘‘local’’ length scale such as boundary layer thickness
would be more suitable for the definition of Reynolds num-
ber. Furthermore, as mentioned in the previous section, a
horseshoe vortex system of Type 2 at high Reynolds numbers
exhibits quasiperiodic characteristics which cannot be found
for that of Type 1. The dependency of the quasiperiodic be-
FIG. 13. Comparison of normalized skin-friction coefficient (C f
⫽ ␶ w /1/2 ␳ U 2B ) along the center line (z⫽3.5) for fully developed flow 共Type
1兲 and developing flow 共Type 2兲.
Downloaded 29 Jan 2013 to 138.250.72.107. Redistribution subject to AIP license or copyright; see http://pof.aip.org/about/rights_and_permissions
Numerical study of vortical structures 2389
havior on the type of flow is unique to horseshoe vortices in
internal flows, and remains to be further explored for the
purposes of not only practical application but also basic un-
derstanding of horseshoe vortices.
D. Hairpin and lateral vortices
The flow region upstream of the obstacle remains lami-
nar for all Re considered in this work. In the near-wake
region, however, the flow becomes turbulent at lower Re.
This downstream flow region contains various flow struc-
tures such as free-shear layer, recirculating flow, redevelop-
ing turbulent boundary layer, and so forth; little research
work has been done on their mutual interactions and subse-
quent developments because of the complexity involved. In
this section, generation and subsequent development of the
hairpin vortices associated with the free-shear layer are dis-
cussed in detail. The free-shear layer, produced by the flow
separation at the front edge of the obstacle, becomes unstable
and yields vortex roll-ups leading to hairpin-like vortices.
Two cases of Re⫽1000 共Case 1兲 and Re⫽3500 共Case 2兲
are considered in this section; the corresponding uniform
laminar flow is imposed at the inlet for each case. The flow
in the near-wake region is turbulent in both cases, and qua-
siperiodic vortex roll-ups are noticed. In Case 1, a steady
four-horseshoe vortex system is observed upstream of the
obstacle, while a quasiperiodic six-horseshoe vortex system
is obtained in Case 2. Numerical parameters for each case
are listed in Table I.
In Fig. 14, instantaneous contours of positive values of
J⫺0.25⌬ 2 on the center plane are presented for Case 1 dur-
ing one quasiperiod of generating a hairpin vortex at Re
⫽1000; frame 1 is at the same phase as frame 8. Here, J
represents Jacobian of shear-stress vectors and ⌬ denotes di-
vergence of shear stresses. This method is based on the to-
pological consideration that a vortex core constitutes a focus
which satisfies J⬎0.25⌬ 2 共see Ref. 10兲. In Fig. 14, the as-
terisk 共*兲 represents the center of the head portion of the
hairpin vortex. The flow field near the front edge of the ob-
stacle is nearly steady; in fact, that is the case in the region
up to x⫽3.8. Past the rear edge of the obstacle, the shear
layer becomes unstable and a new vortex head 共Vortex 2兲 is
born 共Frames 7 and 8兲 while the existing vortex head 共Vortex
1兲 is convected downstream 共Frames 4, 5, and 6兲 and dissi-
pated 共Frames 7 and 8兲. It turned out that this process is
repeated quasi-periodically. In the region where Vortex 1 dis-
appears (5.0⭐x⭐6.0), one can notice eddies of a wide range
of length scale, which is an indication of turbulent flow. In
order to quantify the quasiperiodicity of the hairpin vortices,
temporal variation of the streamwise velocity component
(u c ) at the center of a hairpin vortex head 共*兲 is presented in
Fig. 15; approximately five periods are identified. Here, the
instants of minimal u c correspond to those at which the hair-
pin vortex 共Vortex 1兲 is about to disappear 共Frame 6兲, while
the instants of maximum u c correspond to those at which a
new vortex 共Vortex 2兲 is just formed 共Frame 7兲. Figure 15
reveals that u c decreases as the hairpin vortex is convected
downstream. It should be noted that the Strouhal number of
the current hairpin vortex shedding 共St⫽0.8兲 is consistent
2390 Phys. Fluids, Vol. 16, No. 7, July 2004
with the experimental value 共St⫽0.84兲 of Elavarasan and
Meng13 measured in a tab wake. In fact, the head portion of
a hairpin vortex moves more slowly than u c . The averaged
value of u c nondimensionalized by U B is 0.6 while the av-
eraged speed of a typical hairpin vortex head nondimension-
alized by U B is 0.5.
Figure 16 presents instantaneous contours of positive J
⫺0.25⌬ 2 on the x-z plane at y⫽0.5h during one quasip-
eriod, identifying vortex shedding from the leading lateral
edges of the obstacle. Frame 1 is at the same phase as Frame
8. A discrete vortex, hereafter called ‘‘lateral vortex’’ is qua-
siperiodically rolled up from each leading lateral edge of the
cubic obstacle. One can notice that the vortex shedding in
each side is locked in phase. Furthermore, its Strouhal num-
ber 共St⫽0.8兲 turns out to be the same as that of the hairpin
vortex shedding. In order to elucidate the dynamical charac-
teristics of the flow structures around the obstacle, one can
define a coefficient for spanwise force loading as follows:
Fz
C z⫽ , 共6兲
1 tionship to the observed frequencies of oscillation of the
␳ U B2 h 2
2 horseshoe vortex systems. Since the ratio of cylinder height
FIG. 15. Temporal variation of the streamwise velocity component (u c ) at
the center of a hairpin vortex head 共*兲, Re⫽1000.
Downloaded 29 Jan 2013 to 138.250.72.107. Redistribution subject to AIP license or copyright; see http://pof.aip.org/about/rights_and_permissions
J.-Y. Hwang and K.-S. Yang
FIG. 14. Instantaneous contours of
positive values of J⫺0.25⌬ 2 on the
center plane during one quasiperiod of
generating a hairpin vortex, Re⫽1000;
frame 1 is at the same phase as frame
8. Here, J represents Jacobian of
shear-stress vectors and ⌬ denotes di-
vergence of shear stresses. The aster-
isk 共*兲 represents the center of the
head portion of the hairpin vortex. The
increment nondimensionalized by U B
and h is 0.83.
where F z denotes the instantaneous spanwise net force com-
puted on the obstacle faces excluding the top one. This co-
efficient is quite analogous to the lift coefficient in the case
of a two-dimensional cylinder immersed in a freestream. A
typical time history of C z is presented in Fig. 17; here t is
synchronized with the time variable in Fig. 15. Figure 17
reveals that high-frequency modes are superimposed on a
dominant low-frequency mode. The Strouhal number corre-
sponding to the dominant low-frequency mode is 0.13 which
is surprisingly similar to the typical Strouhal numbers
共0.124 –0.13兲 observed in the flows around a square cylinder
for a Re range compatible with the current Re.30 It should be
noted that a similar analogy also holds between the junction
flow of a finite circular cylinder 共St⫽0.16 –0.17兲1 and the
flow past a two-dimensional circular cylinder in a freestream
共St⫽0.16 –0.21兲.31 However, the flow mechanism causing
the spanwise low-frequency force loading on the cubic ob-
stacle is not clearly understood yet. Baker1 reported that the
range of St for the wake vortex shedding of the circular
cylinder is approximately 0.16 –0.17, which bears no rela-
to cylinder diameter was large 共4.76兲 in Baker’s setup, one
may conjecture that the low-frequency wake vortex shedding
would be visible in the cases of flows past a ‘‘tall’’ wall-
mounted rectangular cylinder situated in a ‘‘thin’’ boundary
layer. It should be also noted that in the current flow there are
three distinct frequencies present 共St⫽0.13, 0, 0.8兲 corre-
sponding to C z , the horseshoe vortex system, and hairpin/
lateral vortices, respectively, implying that the origin of each
oscillation is quite different.
Vortical structure in the near-wake region 共behind the
obstacle and below the free-shear layer兲 is very complex;
there is an overall flow recirculating in the clockwise direc-
tion along with spiral vortices near the rear face of the cubic
obstacle and the legs of horseshoe vortices near the lower
Phys. Fluids, Vol. 16, No. 7, July 2004
channel wall. It is very difficult to quantify their characteris-
tics. Figure 18 reveals the regions of negative streamwise
velocity very near the lower wall (y⫽0.005) at arbitrary
three instants during one period of hairpin vortex shedding
for Case 1. One can identify a steady four-horseshoe-vortex
system upstream of the obstacle while the downstream flow
is quite unsteady and non-periodic. The unsteadiness in the
FIG. 17. Time history of C z . Here, t is synchronized with the time variable
in Fig. 15.
Downloaded 29 Jan 2013 to 138.250.72.107. Redistribution subject to AIP license or copyright; see http://pof.aip.org/about/rights_and_permissions
Numerical study of vortical structures 2391
FIG. 16. Instantaneous contours of positive J
⫺0.25⌬ 2 on the x-z plane at y⫽0.5h during one qua-
siperiod of generating a lateral vortex, Re⫽1000;
Frame 1 is at the same phase as Frame 8. Here, J rep-
resents Jacobian of shear-stress vectors and ⌬ denotes
divergence of shear stresses. The increment nondimen-
sionalized by U B and h is 0.85.
near-wake region is not correlated with the hairpin vortices
at all. The mean reattachment length on the centerline (z
⫽3.5) is 2.08h downstream of the rear face of the obstacle.
In Fig. 19, instantaneous contours of the streamwise compo-
nent of vorticity ( ␻ x ) are presented at the same y and at the
same instants as in Fig. 18. The steady horseshoe vortex
system is clearly seen upstream of and wrapping around the
obstacle. Near the mean reattachment points (x⫽6.08 on the
centerline兲, one can identify relatively small eddies of 0.2h
⫺0.8h, while somewhat slant and longer eddies 共about h兲
are noticed further downstream. Figure 20 presents instanta-
neous contours of the second invariant of the velocity gradi-
ent tensor32 共Q兲 for Case 1 in a perspective view. This
method is useful in identifying three-dimensional vortical
structures. All the vortices discussed in this study 共horseshoe,
hairpin, and lateral vortices兲 are clearly identified. In particu-
lar, hairpin and lateral vortices evolve far downstream of the
cubic obstacle in contrast to those in a high-Re flow. See Fig.
9 of Krajnović and Davidson23 共Re⫽40 000兲.
In Fig. 21, instantaneous contours of positive values of
J⫺0.25⌬ 2 for Case 2 are presented on the center plane dur-
ing one quasiperiod of hairpin vortex shedding; Frame 1 is at
the same phase as Frame 6. Unlike Case 1, the flow becomes
2392 Phys. Fluids, Vol. 16, No. 7, July 2004 J.-Y. Hwang and K.-S. Yang

FIG. 18. Regions of negative streamwise velocity very near the lower wall
(y⫽0.005) during one period of hairpin vortex shedding 共T兲 for Re⫽1000; FIG. 19. Instantaneous contours of the streamwise component of vorticity
共a兲 t⫽t 1 , 共b兲 t⫽t 1 ⫹0.64T, and 共c兲 t⫽t 1 ⫹T. ( ␻ x ) at y/h⫽0.005, Re⫽1000; 共a兲 t⫽t 1 , 共b兲 t⫽t 1 ⫹0.64T, and 共c兲 t⫽t 1
⫹T. The increment nondimensionalized by U B and h is 2.77.

turbulent immediately after separation at the front edge of

the obstacle, and the vortices are generated in the free-shear
layer near the front edge. The developed vortex head is
clearly noticed in the region 0.4h⫺0.6h downstream of the
front edge; in Case 1 it was 1.9h – 2.1h downstream of the
front edge. Vortex 1 which corresponds to the head portion of
a hairpin vortex is convected downstream 共Frames 1 and 2兲
and dissipated 共Frames 3, 4, and 5兲. Concurrently, a new
vortex 共Vortex 2兲 is generated 共Frame 2兲 and convected
downstream 共Frames 3– 6兲. This process is repeated quasip-
eriodically. The life-cycle of hairpin vortex described above
is quite similar to that of Case 1, but St is almost doubled
共1.6兲 in Case 2, indicating that not only length scale but also
time scale of hairpin vortex are much finer in high-Re flows
compared with those in low-Re flows.
Stokes and continuity equations. A parametric study reveals
both qualitative and quantitative characteristics of the vorti-
cal structures associated with this particular geometry.
Flow separation occurs and a subsequent horseshoe vor-
tex system is developed due to the adverse pressure gradient
incurred by the obstacle. The number of vortices increases up
to six in pairs as the Reynolds number increases. In the case
of fully developed channel flow, the unsteady horseshoe vor-
tex system is not found in the range of Reynolds number up
to the critical value for the inlet parabolic velocity profile.

V. CONCLUSION
In this investigation, numerical simulations were carried
out to elucidate the characteristics of the vortical structures
associated with the internal junction flow with a cubic ob-
stacle mounted on one wall of a channel. Especially, the
steady and unsteady laminar horseshoe vortex systems up-
stream of the obstacle and the hairpin vortices in the free-
shear layer were studied in detail. Flows with low-to-
moderate Reynolds numbers up to Re⫽3500 were
considered. The horseshoe vortex systems are laminar in all
the cases considered here, but flow becomes turbulent in the
near-wake region including the free-shear layer at relatively
low Reynolds numbers. Unphysical assumptions were ruled FIG. 20. Instantaneous contours of the second invariant of the velocity
out in directly solving the governing incompressible Navier– gradient tensor 共Q兲, Re⫽1000. Q⫽7.5, nondimensionalized by U B and h.

Downloaded 29 Jan 2013 to 138.250.72.107. Redistribution subject to AIP license or copyright; see http://pof.aip.org/about/rights_and_permissions
Phys. Fluids, Vol. 16, No. 7, July 2004
FIG. 21. Instantaneous contours of positive values of J⫺0.25⌬ 2 on the
center plane during one quasiperiod of generating a hairpin vortex, Re
⫽3500; Frame 1 is at the same phase as Frame 6. Here, J represents Jaco-
bian of shear-stress vectors and ⌬ denotes divergence of shear stresses. The
increment nondimensionalized by U B and h is 14.37.
The unsteady horseshoe vortex systems, characterized by a
repeated pattern of generation, translation, and merging of
the vortices, can be obtained only when the obstacle is
placed in the entrance region of a developing channel flow
where the boundary layers on the channel walls are thin.
Life-cycles of hairpin vortex generated in the free-shear
layer downstream of and above the obstacle were also re-
ported at Re⫽1000 and Re⫽3500, respectively; those pro-
cesses turned out to be quasiperiodic. In the former case,
hairpin vortices are generated in the near-wake region with
relatively lower frequency 共St⫽0.8兲, while in the latter case
they appear in the free-shear layer close to the front edge of
the obstacle with almost doubled frequency 共St⫽1.60兲. This
fact indicates that not only length scale but also time scale of
hairpin vortex become finer as the Reynolds number in-
creases. Lateral vortices shed from the leading lateral edges
of the cubic obstacle were also identified. In the vicinity of
the channel walls, small-scale eddies are observed especially
near the reattachment points about two obstacle widths
downstream of the rear face of the obstacle; the eddies are
slant and elongated further downstream.
ACKNOWLEDGMENTS
This work was supported by Inha University Research
Grant 共INHA-30217兲. The authors are grateful to Dr. Kishan
Shah for useful discussions.
Downloaded 29 Jan 2013 to 138.250.72.107. Redistribution subject to AIP license or copyright; see http://pof.aip.org/about/rights_and_permissions
Numerical study of vortical structures 2393
1
C. J. Baker, ‘‘The laminar horseshoe vortex,’’ J. Fluid Mech. 95, 347
共1979兲.
2
L. R. Kubendran and W. D. Harvey, ‘‘Flow control in a wing/fuselage-type
juncture,’’ AIAA Paper 88-0614 共1988兲.
3
R. M. Kelso and A. J. Smits, ‘‘Horseshoe vortex systems resulting from
the interaction between a laminar boundary layer and a transverse jet,’’
Phys. Fluids 7, 153 共1995兲.
4
A. S. W. Thomas, ‘‘The unsteady characteristics of laminar juncture flow,’’
Phys. Fluids 30, 283 共1987兲.
5
C. J. Baker, ‘‘The oscillation of horseshoe vortex systems,’’ J. Fluids Eng.
113, 489 共1991兲.
6
W. J. Devenport and R. L. Simpson, ‘‘Time-dependent and time-averaged
turbulence structure near the nose of a wing-body junction,’’ J. Fluid
Mech. 210, 23 共1990兲.
7
U. K. Kaul and D. C. Kwak, ‘‘A computational study of saddle point
separation and horseshoe vortex system,’’ AIAA Paper 85-0182 共1985兲.
8
M. R. Visbal, ‘‘Structure of laminar juncture flows,’’ AIAA J. 29, 1273
共1991兲.
9
M. D. Coon and M. Tobak, ‘‘Experimental study of saddle point of attach-
ment in laminar juncture flow,’’ AIAA J. 33, 2288 共1995兲.
10
J. C. R. Hunt, C. J. Abell, J. A. Peterka, and H. Woo, ‘‘Kinematical studies
of the flows around free or surface-mounted obstacles; applying topology
to flow visualization,’’ J. Fluid Mech. 86, 179 共1978兲.
11
F. Ballio, C. Bettoni, and S. Franzetti, ‘‘A survey of time-averaged char-
acteristics of laminar and turbulent horseshoe vortices,’’ J. Fluids Eng.
120, 233 共1998兲.
12
A. Haidari and C. Smith, ‘‘The generation and regeneration of single hair-
pin votices,’’ J. Fluid Mech. 277, 135 共1994兲.
13
R. Elavarasan and H. Meng, ‘‘Flow visualization study of role of coherent
structures in a tab wake,’’ Fluid Dyn. Res. 27, 183 共2000兲.
14
A. E. Perry and T. R. Steiner, ‘‘Large-scale vortex structures in turbulent
wakes behind bluff bodies. Part 1. Vortex formation,’’ J. Fluid Mech. 174,
233 共1987兲.
15
K.-S. Yang, P. R. Spalart, and J. H. Ferziger, ‘‘Numerical studies of natural
transition in a decelerating boundary layer,’’ J. Fluid Mech. 240, 433
共1992兲.
16
M. Lesieur, ‘‘Large-eddy simulations of shear layers,’’ Exp. Therm. Fluid
Sci. 12, 197 共1996兲.
17
A. Suksangpanomrung, N. Djilali, and P. Moinat, ‘‘Large-eddy simulation
of separated flow over a bluff rectangular plate,’’ Int. J. Heat Fluid Flow
21, 655 共2000兲.
18
H. J. Hussein and R. J. Martinuzzi, ‘‘Energy balance for turbulent flow
around a surface mounted cube placed in a channel,’’ Phys. Fluids 8, 764
共1995兲.
19
R. Martinuzzi and C. Tropea, ‘‘The flow around surface-mounted, pris-
matic obstacles placed in a fully developed channel flow,’’ J. Fluids Eng.
115, 85 共1993兲.
20
K. Shah, ‘‘Large eddy simulation of flow past a cubic obstacle,’’ Ph.D.
thesis, Department of Mechanical Engineering, Stanford University, Cali-
fornia, 1998.
21
ERCOFTAC/IAHR/COST Workshop on Refined Flow Modeling, edited by
K. Hanjalić and S. Obi 共Delft University of Technology, Delft, The Neth-
erlands, 1997兲, p. 131.
22
W. Rodi, ‘‘Comparison of LES and RANS calculations of the flow around
bluff bodies,’’ J. Wind. Eng. Ind. Aerodyn. 69–71, 55 共1997兲.
23
S. Krajnović and L. Davidson, ‘‘Large-eddy simulation of the flow around
a bluff body,’’ AIAA J. 40, 927 共2002兲.
24
J. Kim and P. Moin, ‘‘Application of a fractional-step method to incom-
pressible Navier–Stokes equations,’’ J. Comput. Phys. 59, 308 共1985兲.
25
H. Choi and P. Moin, ‘‘Effects of the computational time step on numeri-
cal solutions of turbulent flow,’’ J. Comput. Phys. 113, 1 共1994兲.
26
K.-S. Yang and J. H. Ferziger, ‘‘Large-eddy simulation of turbulent ob-
stacle flow using a dynamic subgrid-scale model,’’ AIAA J. 31, 1406
共1993兲.
27
L. L. Pauley, P. Moin, and W. C. Reynolds, ‘‘A numerical study of un-
steady laminar boundary layer separation,’’ Report No. TF-34, Thermo-
sciences Division, Department of Mechanical Engineering, Stanford Uni-
versity, 1988.
28
A. Larousse, R. Martinuzzi, and C. Tropea, ‘‘Flow around surface-
mounted, three-dimensional obstacles,’’ Selected Papers of the 8th Turbu-
lent Shear Flow Symposium 共Springer-Verlag, Berlin, 1992兲, p. 127.
2394 Phys. Fluids, Vol. 16, No. 7, July 2004 J.-Y. Hwang and K.-S. Yang

29
R. Martinuzzi and C. Tropea, ‘‘Flow around a surface-mounted cube,’’
Selected Papers of the 6th Symposium on Applied Laser Technology
共Springer-Verlag, Berlin, 1993兲.
30
D.-H. Kim, K.-S. Yang, and M. Senda, ‘‘Large eddy simulation of turbu-
lent flow past a square cylinder confined in a channel,’’ Comput. Fluids 33,
81 共2004兲.
31
P. Beaudan and P. Moin, ‘‘Numerical experiments on the flow past a cir-
cular cylinder at sub-critical Reynolds number,’’ Report No. TF-62, Ther-
mosciences Division, Department of Mechanical Engineering, Stanford
University, California 1994.
32
J. Jeong and F. Hussain, ‘‘On the identification of a vortex,’’ J. Fluid
Mech. 285, 69 共1995兲.

Downloaded 29 Jan 2013 to 138.250.72.107. Redistribution subject to AIP license or copyright; see http://pof.aip.org/about/rights_and_permissions