A comprehensive study of secondary and

tertiary vortex phenomena of flow past a

circular cylinder: A Cartesian grid approach
Cite as: Phys. Fluids 33, 053608 (2021); https://doi.org/10.1063/5.0042603
Submitted: 01 January 2021 • Accepted: 22 April 2021 • Published Online: 17 May 2021

Pankaj Kumar and Jiten C. Kalita

ARTICLES YOU MAY BE INTERESTED IN

Flow around a diamond-section cylinder at low Reynolds numbers

Physics of Fluids 33, 053611 (2021); https://doi.org/10.1063/5.0049811

Large-eddy simulation of flow past a circular cylinder for Reynolds numbers 400 to 3900
Physics of Fluids 33, 034119 (2021); https://doi.org/10.1063/5.0041168

Flow-induced vibrations of circular cylinder in tandem arrangement with D-section cylinder at

low Reynolds number
Physics of Fluids 33, 053606 (2021); https://doi.org/10.1063/5.0048580

Phys. Fluids 33, 053608 (2021); https://doi.org/10.1063/5.0042603 33, 053608

© 2021 Author(s).
 Physics of Fluids ARTICLE scitation.org/journal/phf

A comprehensive study of secondary and tertiary

vortex phenomena of flow past a circular cylinder:
A Cartesian grid approach
Cite as: Phys. Fluids 33, 053608 (2021); doi: 10.1063/5.0042603
Submitted: 1 January 2021 . Accepted: 22 April 2021 .
Published Online: 17 May 2021

Pankaj Kumar1,a) and Jiten C. Kalita2,b)

AFFILIATIONS
1
Department of Mechanical Engineering, Indian Institute of Technology Guwahati, Guwahati 781039, Assam, India
2
Department of Mathematics, Indian Institute of Technology Guwahati, Guwahati 781039, Assam, India

a)
Electronic-mail: pankaj.2013@iitg.ac.in
b)
Author to whom correspondence should be addressed: jiten@iitg.ac.in

ABSTRACT
This study envisages undertaking a comprehensive simulation of the flow past an impulsively started circular cylinder with special emphasis
on vortex dynamics in the secondary and tertiary level. A recently developed second order spatially and temporally accurate compact finite
difference scheme on a nonuniform Cartesian grid has been used to discretize the transient Navier–Stokes equations governing the flow. The
grid is generated in such a way that the cylinder boundary passes through the grid points, thus dispensing with the need to use the immersed
interface approach on a Cartesian grid. High quality simulations are accomplished for a wide range of Reynolds numbers (Re) from
5  Re  10 000 in the laminar regime, including the periodic flow characterized by von Karman vortex street. The a, b, sub-a, and sub-b
phenomena, which are the trademark of the secondary and tertiary vortex dynamics associated with such flows, are studied in detail. Our
results are compared with existing experimental and numerical results, and close comparison is obtained in all the cases exemplifying their
accuracy. In the process, for the first time, we also provide a tabular documentation of the early stages of the flow for Re  700.
Published under an exclusive license by AIP Publishing. https://doi.org/10.1063/5.0042603

I. INTRODUCTION impulsively started circular cylinder alone, literally hundreds of papers

Bodies kept in fluid flow are characterized as being streamlined
or blunt/bluff, depending on its overall shape and structure. A bluff
body can be defined as a body, which, owing to its shape, has separated
flow over a significant part of its surface; in other words, if the fluid
does not touch the whole boundary of an object1 while immersed in
fluid flow, it is a bluff body. Flow past bluff bodies is a very common
phenomenon, and it happens all around us. Some common examples
are the flows past an airplane, a submarine, an automobile, or wind
blowing past a high-rise building. Such flows are very complex and
involve the interactions of three shear layers in the same problem,
namely, a boundary layer, a separating free shear layer, and a wake.2
Moreover, there is a very strong interaction between the viscous and
inviscid regions. Because of the vortical instabilities in the wakes, they
are often difficult to predict, and their understanding poses a great
challenge to the scientific community.
Although bluff bodies can be of different shapes and sizes, the cir-
cular cylinder is considered to be the representative of all two dimen-
sional bluff bodies. On the study of the wake behind the flow past an
have been published, in part due to its engineering significance, and in
part due to the tempting simplicity in setting up such an arrangement
in an experimental or computational laboratory. In the simplest of
geometric settings, it exhibits almost all the characteristics of incom-
pressible viscous flow phenomena. As such it has attracted the atten-
tion of engineers, scientists, and mathematicians alike, who have been
engaged in carrying out theoretical, experimental, and numerical stud-
ies on this problem over the past few decades.2–5,9–15
A plethora of numerical studies on this problem attempted
through all three well known approaches of discretization:10–15 finite
element, finite volume, and finite difference, can be found in the exist-
ing literature. All these methods are based on the Eulerian description
of the flow. The mesh-free methods, which are based on the
Lagrangian description, have co-existed along with the finite difference
methods over the years. Although not used very frequently at the early
part of their evolution, of late these methods such as the vortex
method, smoothed-particle hydrodynamics method, or hybrid vortex
method16–18 are also getting recognized as a standard or mainstream

Phys. Fluids 33, 053608 (2021); doi: 10.1063/5.0042603 33, 053608-1

Published under an exclusive license by AIP Publishing
 Physics of Fluids ARTICLE scitation.org/journal/phf

tools in Computational Fluid Dynamics (CFD). Besides, the Lattice shedding. All the results are qualitatively and quantitatively validated
Boltzmann method has also become quite popular as a tool for the through comparison with experimental and well established numerical
simulation of such flows.19–22 results, and they are found to be extremely close. This is due to the fact
The finite difference approach, which is otherwise easy to imple- that the scheme utilizes the advantage of grid clustering in the regions of
ment, possesses the greatest challenge amongst all the approaches pri- small scales, which invariably requires more grid points to resolve the
marily because of the circular geometry, where a body fitted scale irrespective of the spatial accuracy of the scheme.
coordinate system is extremely difficult to construct. Consequently, The paper has been arranged in six sections. Section II deals with
majority of the simulations on this particular problem on a finite dif- the mathematical formulation and discretization procedures, Sec. III
ference set up were performed on polar grids.22–24 The very few simu- with the problem and grid generation, Sec. IV with the grid indepen-
lations performed on a finite difference set-up on Cartesian grid, dence and code validation, Sec. V discusses the results, and finally,
mainly utilized the immersed boundary or the immersed interface Sec. VI summarizes the whole work.
approach.25–33 However, all these approaches invariably result in addi-
tional computations compared to the methods employing body fitted II. GOVERNING EQUATIONS AND NUMERICAL
grids. For example, the discrete immersed boundary and other SCHEME
immersed interface methods33 use interpolation in the neighborhood The viscous fluid flow abides the Navier–Stokes (N–S) equations,
of the curved boundary where the grid lines intersect it, while the con- which are the governing equations for unsteady 2D incompressible flow.
tinuous immersed boundary method32 has to rope in an extra source In primitive variable (velocity–pressure) form, they can be written as
term in the form of a Dirac delta function. ux þ vy ¼ 0; (1)
In the current study, in order to discretize the equations governing 1
the flow, we employ a recently developed34 second order spatially and ut þ uux þ vuy ¼  px þ  ðuxx þ uyy Þ; (2)
q
temporally accurate compact finite difference scheme on a nonuniform 1
Cartesian grid without transformation for the biharmonic form of the vt þ uvx þ vvy ¼  py þ  ðvxx þ vyy Þ; (3)
q
Navier–Stokes equations35–39,41,49 for incompressible viscous flows. The
utmost practical utility of using a Cartesian grid is the implementation of where t is the time and u, v, p, q, and  are the x-, y-velocities, pres-
the boundary conditions without any complication.40 Also, being with- sure, density, and kinematic viscosity of the fluid.
out transformation, governing equations are discretized over the same The problem considered here is flow past an impulsively started
physical domain, yielding an easy passage to an efficient computation of cylinder of unit diameter D in a free stream. The non-
the flow field. Here, the grid is generated in such a way that the circular dimensionalization of fluid variables are done as follows:
geometry passes through the grid points, thus dispensing with the need x ¼ x=D; y ¼ y=D; u ¼ u=U1 ; v ¼ v=U1 ;
to use the immersed boundary or immersed interface approach. A quick
look at the numerical works on this problem reveals that very few studies t  ¼ tU1 =D; and p ¼ p=qU1
2
:
detail the flow description for the whole range of 100  Re  104 by Here, Re is defined as U1 D=. Where U1 is the free stream velocity.
employing a single numerical scheme. Particularly, the ones on Now the following Eqs. (2) and (1) reduce to (by removing  sign):
Cartesian grid through the immersed interface approach also dealt only
with a low Reynolds number regime. Keeping that in mind, rather than ux þ vy ¼ 0; (4)
describing the vortex dynamics in both unseparated (for example, 1
ut þ uux þ vuy ¼ px þ ðuxx þ uyy Þ; (5)
splitter-plate flows) and separated wakes for all the different types of 3D Re
and nominally 2D body shapes, the current attempt is an endeavor to 1
vt þ uvx þ vvy ¼ py þ ðvxx þ vyy Þ: (6)
present an overview of the vortex dynamics phenomena in the wake of a Re
circular cylinder over this entire range in the laminar regime. Equations (5) and (6) have three unknowns, which on using the conti-
Our study is mainly concerned with the vortex dynamics beyond nuity equation (4) and defining streamfunction w and vorticity x as
primary level that are typical of the laminar stage of flows at high uðx; yÞ ¼ wy ; vðx; yÞ ¼ wx ; x ¼ vx  uy ; (7)
Reynolds numbers and is characterized by the presence of a, b, sub-a,
and sub-b phenomena. The existence of a and b phenomena at the sec- reduce to the streamfunction–vorticity (w–x) formulation
ondary level was first observed in the path breaking experimental visual- 1
ization of Coutaceau and Bouard,5,6 which was later reported in several xt þ uxx þ vxy ¼ ðxxx þ xyy Þ; (8)
Re
numerical works.7,23,56 Recently, Kalita and Sen reported the presence of wxx þ wyy ¼ xðx; yÞ; (9)
the same phenomena at the tertiary level, which they termed as sub-a
and sub-b phenomena in.8,10 This study of the tertiary vortex dynamics thus eliminating the variable p from (5) and (6). Substitution of (9) in
is of particular interest to us, which is presented in detail for the early (8) yields the following equation:
stage of the flow for the moderately high Reynolds number 7500. The  
@4w @4w @4w @
subtle changes in flow evolution containing these phenomena in the ½ 2
þ 2 2 2 þ 4  Re vr u  ur v ¼ Re 2  2
ðr wÞ ; (10)
early stages for Re ¼ 5000 reported in Ref. 10 and the current computa- @x4 @x @y @y @t
tion for Re ¼ 7500 are also demonstrated in an accompanying movie. which Gupta and Kalita35 termed as the w–v formulation of the
For other Reynolds numbers, viz., Re  700 they are systematically Navier–Stokes equation. In the above, if u and v are also expressed in
organized in a tabular summary. In the process, we have also studied terms of w given in (7), it reduces to the pure streamfunction formula-
the periodic flow regime characterized by the von Karman vortex tion of the 2D steady Navier–Stokes equations,

Phys. Fluids 33, 053608 (2021); doi: 10.1063/5.0042603 33, 053608-2

Published under an exclusive license by AIP Publishing
 Physics of Fluids ARTICLE scitation.org/journal/phf

" !
@4w @4w @4w @w @ 3 w @3w Recently, we have proposed a second order time and space accu-
þ 2 2 2 þ 4  Re þ rate compact finite difference scheme for the biharmonic form of the
@x4 @x @y @y @y @x3 @x@y2
!#   transient-state Navier–Stokes (N–S) equations on nonuniform
@w @ 3 w @3w @ 2
Cartesian grids without transformation. The discretized form of the
 þ ¼ Re ðr wÞ : (11) Eq. (11) on a nine point stencil can be written in the form
@x @x2 @y @y3 @t

"   # "   #
2 wnþ1
iþ1;j wnþ1
i1;j 1 1 nþ1 2 wnþ1
i;jþ1 wnþ1
i;j1 1 1 nþ1
þ  þ w þ þ  þ w
xf þ xb xf xb xf xb i;j yf þ yb yf yb yf yb i;j
0:5Dt n n  o
¼ Awiþ1;jþ1 þ Bwni;jþ1 þ Cwni1;jþ1 þ Dwniþ1;j þ Ewni;j þ Fwni1;j þ Gwniþ1;j1 þ Hwni;j1 þ Iwni1;j1  /ni;j
Re
0:5Dt n nþ1  o
þ Awiþ1;jþ1 þ Bwnþ1 nþ1 nþ1 nþ1 nþ1 nþ1 nþ1 nþ1
i;jþ1 þ Cwi1;jþ1 þ Dwiþ1;j þ Ewi;j þ Fwi1;j þGwiþ1;j1 þ Hwi;j1 þ Iwi1;j1  /i;j
nþ1
Re "
wniþ1;j wni1;j  1  # " n
wi;jþ1 wni;j1  1  #
2 1 n 2 1
þ þ  þ w þ þ  þ wn
xf þ xb xf xb xf xb i;j yf þ yb yf yb yf yb i;j
"   # "   #
xf  xb vniþ1;j vni1;j 1 1 n yf  yb uni;jþ1 uni;j1 1 1 n
þ þ  þ vi;j  þ  þ u
xf þ xb xf xb xf xb yf þ yb yf yb yf yb i;j
"   # "   #
nþ1 nþ1 nþ1 nþ1
xf  xb viþ1;j vi1;j 1 1 nþ1 yf  yb ui;jþ1 ui;j1 1 1 nþ1
þ þ  þ v  þ  þ u ; (12)
xf þ xb xf xb xf xb i;j yf þ yb yf yb yf yb i;j

where xf, xb, yf, and yb are the nonuniform forward and backward condition @/ @/
@t þ U0 @x ¼ 0 is imposed. On the surface of the body
step-lengths in the x- and y-directions, and Dt is the uniform time u ¼ v ¼ w ¼ 0. At the other two boundaries, @u @y ¼ 0 ¼ v with
step. Making use of u ¼ @w @y , the horizontal velocity can be approxi- w ¼ yT and w ¼ yB at the top and bottom boundaries, respectively,
mated up to third order accuracy by the following implicit relation: where yT and yB are the y coordinates thereat.
yb yf For our simulations, we have used grids of size ranging from
ui;jþ1  2ui;j þ ui;j1 81  81 to 601  481, where the choice is made depending upon the
2k  2k
3 wi;jþ1  wi:j1 3ðyf  yb Þ Reynolds number and the flow situation, and the time step Dt is
¼  adjusted accordingly. As theory suggests the formulation of an impul-
2k 2k
" # sive start by the use of potential flow field as the initial condition,16,57
wi;jþ1 wi;j1  1 1   we have used the same as the initial data in our computation. The size
 þ  þ w þ O yf yb ðyf  yb Þ : (13)
yf yb yf yb i;j of the computational domain and the location of the circular cylinder
are chosen in such a way that ample space is available on either side of
The algebraic system resulting from (13) is solved using the famous
the cylinder in the stream wise and normal directions so that the com-
tridiagonal solver Thomas algorithm.42–44 The vertical velocity, v, can
puted flow is free from any entrance effect and a smooth shedding of
be approximated in a similar fashion. The details of the coefficients the vortices is guaranteed when the flow is unsteady. Note that when
A; B; C; D; E F; G; H; I and the function / along with the derivation the physical plane is considered in cylindrical polar coordinates, since
of Eqs. (12)–(13) can be found in Ref. 34. grid lines are aligned with the cylinder geometry, one is mainly con-
cerned with the grid clustering22–24 nearby the surface of the cylinder
III. SCHEMATIC OF THE PROBLEM AND THE GRID
or length of the outer radius of the annular region under consider-
USED
ation. However, if both the computational and physical planes are con-
The schematic of the problem including the boundary conditions sidered over a Cartesian grid along with the accommodation of the
and the computational domain is shown in Fig. 1. The computational circle geometry set in the grid passing exactly through grid points, grid
domain is considered as 8  x  25; 8  y  8, where l ¼ D ¼ 1 generation could be quite tricky. In the following, we provide a brief
so that H ¼ 16, xL ¼ 8, and xR ¼ 25. The cylinder was placed at description of how the grid is generated for computing the flow.
ðx; yÞ ¼ ð0; 0Þ as its center. On the surface of the cylinder / ¼ 0 (/ Since the current computation does not use any immersed inter-
representing either u, v, or w); the same conditions were imposed face approach, the grid essentially needs to be generated in such a way
inside the cylinder as well during computation. In the far-field region that the geometry of cylinder passes through the grid points (see the
in front of the cylinder, a uniform velocity u ¼ U0 ¼ 1 is prescribed close up of the grid around the cylinder in Fig. 3). Following is the pro-
while at the far-field behind the cylinder, a convective boundary cedure to accomplish this: first a square block is created (block 1 in

Phys. Fluids 33, 053608 (2021); doi: 10.1063/5.0042603 33, 053608-3

Published under an exclusive license by AIP Publishing
 Physics of Fluids
Fig. 2), where we generate the grid for the circular geometry by distrib-
uting the points along x-axis (0:5  x  0:5) through the function
cos h; 0  h  2p; likewise, along the y-axis (0:5  y  0:5), one
can distribute the grid points by using the sin h function. After that,
the neighboring blocks (2 to 9), created by using different stretching
function according to the computation requirement, are assembled.
Note that one can also generate the grid in block 1 by first using a
centro-symmetric distribution of points along x-axis and then distrib-ffi
pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ute the points along y-axis by using the formula y ¼ 6 a2  x2 ,
where a is the desired radius of the circle. However, the first approach
FIG. 2. Assembly of grids on the different parts of the computational domain for the flow past an impulsively started circular cylinder.
Phys. Fluids 33, 053608 (2021); doi: 10.1063/5.0042603
Published under an exclusive license by AIP Publishing
ARTICLE scitation.org/journal/phf
FIG. 1. Configuration of the flow past a
cylinder.
mentioned above generates the points along the surface of the cylinder
at a uniform interval.
During this task, it must be ensured that continuity of the grid
lines in each direction of the overall computational domain is main-
tained as the computation is extremely sensitive to the grid being used.
A demonstration of this fact would be shown in Sec. V C. The grid has
to be chosen in such a way that there is smooth transition of step
lengths as one moves from one point to the next one. This has been
illustrated in Fig. 3, where we show the close-up view of the grid
around the circular geometry. Here, one can see the circular geometry
33, 053608-4
 Physics of Fluids
passing through the grid points in block 1 along with the transition of
the grids around the junction of the blocks 1, 2, 7, and 8.
IV. GRID INDEPENDENCE AND CODE VALIDATION
In order to validate our code, we present the streakline simulation
from our computation depicting the well-known von Karman vortex
street for Re ¼ 140 with the experimental results of Taneda45,46 in
Fig. 4. It is heartening to see the striking similarity between our
Phys. Fluids 33, 053608 (2021); doi: 10.1063/5.0042603
Published under an exclusive license by AIP Publishing
ARTICLE scitation.org/journal/phf
FIG. 3. Representative grid and close-up
views of parts of the grid.
numerical results [Fig. 4(b)] and the experimental visualization for the
entire photograph available in the literature [Fig. 4(a)]. Note that while
we were able to accomplish such a comparison on a grid of size
361  241 over a domain of dimension 30:5l  13l, in Ref. 47, it could
be achieved on a 1400  700 grid over a domain of size 14l  13l, that
too, only for a small length behind the cylinder.
In Table I, we compare our computed results for Re ¼ 9500 on
three different grid sizes 451  361, 526  361 (with Dt ¼ 103 ), and
FIG. 4. Streaklines for the flow past an
impulsively started circular cylinder for
Re ¼ 140: (a) Taneda’s experiment45
[reprinted with permission from Kozlov
et al., “Application of RES methods for
computation of hydrodynamic flows by an
example of a 2D flow past a circular cylin-
der for Re ¼ 5  200,” Int. J. Heat Mass
Transfer 54(4), 887–893 (2011), Copyright
2011 Elsevier] and (b) current simulation
(at time t ¼ 300.0).
33, 053608-5
 Physics of Fluids ARTICLE scitation.org/journal/phf

TABLE I. Grid Independence of the computed results on three different grid of size both the grids resulted in simulations which are visually impossible to
451  361, 526  361, and 601  481 for Re ¼ 9500. differentiate. This can be seen from the streamlines for Re ¼ 5000 at
time t ¼ 2.5 in Fig. 6, where the polar grid computation is from
Mesh M1 M2 M3 Ref. 10.
Grid size 451  361 526  361 601  481 V. RESULTS AND DISCUSSION
t ¼ 0.5 hs 65 64.28 64.81 It is a well-known fact for an unsteady laminar flow around a cir-
L 0.031 0.028 0.030 cular cylinder that the flow is irrotational everywhere instantly after
(x, y) ð0:4307; 0:2831Þ ð0:4307; 0:2831Þ ð0:4307; 0:2831Þ the fluid is set into the motion. Momentarily after the fluid movement,
t ¼ 0.75 hs 74 72.85 73.695 the boundary layer starts to grow on the cylinder’s surface, leading to
L 0.0596 0.0554 0.0587 separation. Separation is first seen at the rear stagnation point. After a
(x, y) ð0:4145; 0:3345Þ ð0:4172; 0:3289Þ ð0:4176; 0:3301Þ short period, for flows with Re > 5, two symmetrical vortices appear
t¼1 hs 80 79.714 80.007 behind the cylinder and begin to grow. There is no flow distortion for
L 0.0889 0.0872 0.0865 Re  5, and our study confirms it. At Re  6, although boundary
layer separation takes place on the cylinder surface, the flow remains
(x, y) ð0:4373; 0:3473Þ ð0:4367; 0:3455Þ ð0:4380; 0:3468Þ
steady and laminar. The literature confirms that the flow becomes
t ¼ 1.25 hs 84 83.14 83.29
unsteady beyond some critical Reynolds number (Rec); numerical
L 0.1389 0.1317 0.1357 studies indicate this Rec to lie in the range 43  Rec  50, where Hopf
(x, y) ð0:5339; 0:3078Þ ð0:5197; 0:3175Þ ð0:5245; 0:3185Þ bifurcations occurs. With the increase in Re, the flow becomes more
t ¼ 1.5 hs 89 87.428 88.71 complicated with the appearance of secondary and tertiary vortices.
L 0.2333 0.222 0.228 For Re > Rec , the phenomenon of vortex shedding, characterized by
(x, y) ð0:5713; 0:2575Þ ð0:5545; 0:2754Þ ð0:5672; 0:2683Þ the von Karman vortex street, starts to appear. Keeping all the above
in mind and depending upon the characteristics observed within each
range, we have divided the flow regime into the following parts for the
601  481 (with Dt ¼ 104 ) at different time stations in order to study Reynolds numbers considered in this study, viz., 5  Re  9500:
the influence of grid size on the characteristics of the flow. Here, we • Flow structures for 5  Re  40, also known as a steady-state
present the wake length L, which is the distance between rear stagna- region.
tion point A of the cylinder and the end of the separation at the point • Flow structures for 100  Re  300. In this region, wake behind
B, the separation angle hs, which is the angle between the x-axis and the circular cylinder becomes unstable. Oscillations in the wake
the line joining the center of the cylinder and the separation point S grow in amplitude and finally leading to vortex shedding with
on the cylinder and the location of the primary vortex center ðx; yÞ in the formation of array of vortices, also known as the von Karman
the upper half of the cylinder (Fig. 5). The maximum differences in all vortex street.
the values of the parameters are found to below 5% signifying the grid
independence of the computed results. For Re > 300, we focus only on the early stage of the flow, i.e., in
We also endeavored to see how our computation on Cartesian the laminar regime.
grid fares with the one carried out in polar grid. Quite interestingly, • Flow structures for 300  Re  1000: In this regime, secondary
despite drastically different in nature and the polar grid having an vortices form during the initial stage but do not split up further.
obvious advantage of grid resolution on the surface of the cylinder, The secondary flow structure is characterized by the appearance
of: (i) bulge phenomenon and (ii) isolated secondary eddy.
• Flow structures for 3000  Re  9500: Here, the most compli-
cated flow properties are associated with the secondary phenom-
ena a, b and tertiary phenomena sub-a and sub-b.

FIG. 5. Flow parameters corresponding to Tables I and II for the flow past an impul-
sively started circular cylinder: A is the rear stagnation point, B, the wake stagnation
point, L, the wake length, S, the separation point, and hs, the angle of separation.
A. Flows for 5  Re  40
This is one of the most studied regions for the flow past a station-
ary circular cylinder. As mentioned above, two symmetrical vortices
appear behind the cylinder for 5  Re < Rec , and with time, they
grow in size until it reaches a steady-state.
In Fig. 7, we show the steady-state streamlines for Re ¼ 10, 20,
and 40, where one can see that as Re increases, the size of wake also
increases. We further present our computed L, hs (see Fig. 5), and CD
in Table II, which is calculated by utilizing the momentum balance
along the horizontal direction. We then compared our results with
established numerical as well as the experimental results of
Coutanceau and Bouard.48 We also compare the post-processed sur-
face vorticity for Re’s mentioned above in Fig. 8. In all the cases, our

Phys. Fluids 33, 053608 (2021); doi: 10.1063/5.0042603 33, 053608-6

Published under an exclusive license by AIP Publishing
 Physics of Fluids
results are found to be extremely close to the available numerical and
experimental results, both qualitative and quantitatively.
B. Flows for 100  Re  300
Beyond Re  Rec , the flow past an impulsively started circular
cylinder becomes periodic sooner or later in the flow evolution process
and is known to develop vortex shedding represented by the von
Karman vortex street. The range of Reynolds numbers chosen in this
section typically exhibits this phenomenon. The fundamental differ-
ence between the present Re range with the previous one is that the
FIG. 7. Steady-state streamlines behind the circular cylinder for (a) Re ¼ 10, (b) Re ¼ 20, and (c) Re ¼ 40.
Phys. Fluids 33, 053608 (2021); doi: 10.1063/5.0042603
Published under an exclusive license by AIP Publishing
ARTICLE scitation.org/journal/phf
FIG. 6. Comparison of Cartesian and
polar grid computations: (a) grid used in
Ref. 10, (b) grid used in the present com-
putation, (c) streamlines for Re ¼ 5000 at
time t ¼ 2.5 from Ref. 10, and (d) stream-
lines for Re ¼ 5000 at time t ¼ 2.5 from
the current computation.
wake behind the cylinder develops quickly because of the rapid devel-
opment of velocity with time, and there is a development of secondary
vortex in this region. As time progresses, the wake grows in size and
strength. The flow undergoes different phases throughout the evolu-
tion till the development of periodic vortex shedding. These phases are
portrayed in Figs. 9(a)–9(l), which show the evolution of streamlines
corresponding to different stages of flow for Re ¼ 200. This Re may be
considered as the representative of all the three Re’s presented in this
section, although the duration of the stages will vary for different Re.
The last three Figs. 9(j)–9(l) represent a typical vortex shedding cycle,
where one can see two eddies being shed behind the cylinder and
33, 053608-7
 Physics of Fluids ARTICLE scitation.org/journal/phf

TABLE II. Comparison of wake lengths, separation angles, and drag coefficients for different Reynolds numbers for the circular cylinder.

Re Ref. 48 (exp) Ref. 49 Ref. 25 Ref. 26 Ref. 27 Ref. 11 Ref. 50 Present

L 10 0.504 0.533 0.531

20 1.86 1.851 1.860 1.940 1.860 1.825 1.830 1.874
40 4.38 4.625 4.560 4.660 4.620 4.420 4.250 4.278
CD 10 2.699 2.629 2.690
20 1.949 2.16 2.07 2.100 2.052 2.172 2.160
40 1.439 1.61 1.55 1.580 1.534 1.590 1.576
hs 10 29.732 30 29.69
20 44.4 43.141 43.9 44.1 44.40 43.50 45 42.66
40 53.4 53.226 53.4 54.1 54.1 53.54 54 53.08

washed away into the wake region within each cycle. These three fig-
ures are half a vortex shedding cycle apart, and Fig. 9(k) is a mirror
image of Figs. 9(j) and 9(l).
We have tabulated the evolution of maximum width lmax and the
abscissa of this maximum xlmax of the wake for Re ¼ 200, 550, and
3000 in Table III and compared them with the experimental results of
Bouard and Coutanceau;5 one can see a very close match between all
of them. Figure 10 shows the evolution of drag and lift coefficients for
Re ¼ 100, 200, and 300 from the start of the flow to the establishment
of vortex shedding. For the range of Reynold number considered here,
a pair of symmetric vortices develop behind the cylinder shortly after
the flow has started [see Figs. 10(a) and 10(b)]. We can see from Fig.
10(a) that after the impulsive start of the flow, the drag coefficient rises
to a very high value. However as time progresses, the value of the drag
coefficient begins to fall, and the eddies start to grow in size [see Figs.
9(a)–9(f)]. As soon as the flow subsides, a quasi-steady state develops
with no lift force. The duration of this state decreases with the increase
FIG. 8. Surface vorticity distribution in the steady-state range.
in Re. From Fig. 9(f), we can see that waviness of the streamlines starts
to appear at the eddy trail and symmetry of the flow about y ¼ 0 line is
broken. With the onset of asymmetry, the lift force becomes nonzero
around t ¼ 60 for Re ¼ 200 as in Fig. 10(b) [also refer Fig. 9(d)]. A few
moments later, the vortices behind the cylinder start oscillating gradu-
ally [see Figs. 9(g)–9(j)], and vortex shedding starts leading to the
development of the so-called von Karman vortex street. Note that
throughout the computation of all the Reynolds numbers considered
in this study, no artificially induced perturbation was introduced to
break the symmetry; asymmetry sets in spontaneously for all
Re  Rec .
One of the objectives of the present work is to analyze the proper-
ties of the streamline and streakline patterns behind a circular cylinder
during vortex shedding. Keeping that in mind, we present the stream-
lines, post-processed vorticity contours and streaklines for Re ¼ 100;
200; and 300 in the left, middle, and right panels of Fig. 11, respec-
tively. These figures correspond to the instants when the lift coeffi-
cients reach their peak values during their own vortex shedding cycles.
In incompressible flow, vorticity is generated only at solid boundaries,
which, for this problem is the surface of the cylinder, and this vorticity
dwells within the fluid. Thus, the streaklines portrayed in Figs.
11(c1)–11(c3) provide a clear view of dominant places in the flow field
where the vorticity is inherent. Our computed streakline patterns for
the range of Re in these figures are in conjunction with those depicted
in Refs. 2 and 4. Note that the relationship between instantaneous
streamlines and streaklines is exceptionally complex, and visualization
of both is necessary for a proper understanding of flow field character.
In the initial stages, the wake behind the cylinder is in a closed-loop
for the streamline patterns [see Figs. 11(a1)–11(a3)]. However once
the vortex-shedding process begins, this closed wake becomes open.
Our computed streaklines, despite thinning down, remain continuous
without breakage. Streaklines now stand for a flexible barrier which a
fluid can never cross and is quite apparent from Figs. 11(c1)–11(c3).
The fluid entering the wake moves in and jumps toward the cylinder
surface sequentially from both sides and is eventually forced out of the
wake and roll-up. The two sets of vortex sheets braid with each other
in the far wake, and we are able to capture them entirely. Once shed-
ding starts, the vortices are shed systematically, alternatively from the
two sides of the cylinder. It is also apparent from the crests and
troughs of the sinuous waves in the streamlines shown in Figs.
11(a1)–11(a3), which reflect the alternatively positive and negative

Phys. Fluids 33, 053608 (2021); doi: 10.1063/5.0042603 33, 053608-8

Published under an exclusive license by AIP Publishing
 Physics of Fluids ARTICLE scitation.org/journal/phf

FIG. 9. Evolution of streamlines for Re ¼ 200 for flow past a circular cylinder at: (a) t ¼ 2, (b) t ¼ 4, (c) t ¼ 10, (d) t ¼ 55, (e) t ¼ 56, (f) t ¼ 58, (g) t ¼ 75, (h) t ¼ 80, (i)
t ¼ 110, (j) t ¼ 281, (k) t ¼ 283.5, and (l) t ¼ 286.

vorticities of the eddies presented in Figs. 11(b1)–11(b3). Shedding fre- maximum value for Re ¼ 100, 200, and 300 are shown in Fig. 12(a);
quency increases over time until a limiting condition is reached. With Fig. 12(b) displays the phase plane of the drag and lift coefficients for
the onset of vortex shedding, the drag coefficient starts decreasing the same time sample for Re ¼ 100, 200, and 300; it clearly establishes
again, and eventually, both the drag and lift coefficients become peri- that the frequency of drag coefficients is twice that of the lift coeffi-
odic as the flow becomes fully developed (see Fig. 10). We also calcu- cients, which is also exemplified by Fig. 10. In Table IV, we compare
late the Strouhal number from the formula St ¼ UfD0 , where f is the our computed Strouhal numbers and drag and lift coefficients for
dominant frequency of the lift variants, which characterizes the vortex these Re with established experimental and numerical results; for all
shedding process. the Re, we obtain very close comparisons. As can be seen from the
The dominant frequency of the drag and lift variations can be table, the frequency of vortex shedding increases with the increase in
computed by a spectral analysis of time samples of these coefficients. Re, which is also obvious from Fig. 11.
The power density spectra of this analysis normalized by the Note that for the results shown in this section, computations
were performed using a rectangular domain ½8D; 25D  ½8D; 8D
TABLE III. Comparison of the of maximum width lmax and the abscissa of this maxi- [where D ¼ 1 is the diameter of the cylinder with center (0, 0)] up to
mum xlmax of the wake with the experimental result of Ref. 5 for Re ¼ 200, 550, and non-dimensional time t ¼ 500. However, behavior of the flow at a
3000 at different time stations (the experimental data within parenthesis). much later time (say around t ¼ 5000) for Re ¼ 300 is known to be
fraught with several modes of vortex shedding far downstream the cyl-
t 1.0 1.5 2.0 2.5 3.0 inder, as suggested by some recent studies.51 However, the second
mode of vortex shedding appears much away from the downstream
Re ¼ 200 xlmax =D 0.39 0.42 0.62 0.73 0.85
length 25D considered in the current study.
(0.35) (0.51) (0.65) (0.76) (0.86)
lmax =D 0.86 0.92 0.98 1.02 1.06 C. Flows in the early stages for 300  Re  550
(0.94) (0.97) (1.02) (1.07) (1.10)
The significant feature of this range (300  Re  800) is the (i)
Re ¼ 550 xlmax =D 0.31 0.42 0.68 0.78 0.85
bulge and (ii) secondary eddy phenomenon, which is also corrobo-
(0.30) (0.50) (0.66) (0.76) (0.85)
rated by experimental studies. Initially, there is a deflection in stream-
lmax =D 0.93 0.98 1.02 1.08 1.14 lines between the rear stagnation and separation points. Some instants
(0.94) (0.98) (1.03) 91.10) (1.16) later, streamlines close to the cylinder’s wall deviate from the cylinder,
Re ¼ 3000 xlmax =D 0.19 0.28 0.41 0.69 0.83 causing a bulge pattern known as the bulging phenomenon. For
(0.19) (0.26) (0.40) (0.73) (0.84) Re ¼ 300, it appears at dimensionless time t ¼ 2.5 (see Fig. 13) and for
lmax =D 0.97 1.01 1.08 1.13 1.19 Re ¼ 550, at time t ¼ 1.5 approximately [see Fig. 15(b)]. For Re > 300,
(0.93) (0.98) (1.04) (1.11) (1.20) with time, this bulge triggers a second separation of flow leading to the
formation of a small secondary vortex known as the secondary eddy

Phys. Fluids 33, 053608 (2021); doi: 10.1063/5.0042603 33, 053608-9

Published under an exclusive license by AIP Publishing
 Physics of Fluids ARTICLE scitation.org/journal/phf

FIG. 10. Evolution of (a) drag coefficient

and (b) lift coefficient for Re ¼ 100, 200,
and 300.

FIG. 11. Vortex shedding behind the circular cylinder: streamlines [(a1)–(a3)], vorticity contours [(b1)–(b3)] and streaklines [(c1)–(c3)] for Re ¼ 100 (top), Re ¼ 200 (middle),
and Re ¼ 300 (bottom).

1 1

Drag, Re=100 0.8

Lift, Re=100
Drag, Re=200
Lift, Re=200
0.8 Drag, Re=300 0.6
Lift, Re=300

0.4
Power Spectrum

0.6 0.2
Lift

0 FIG. 12. (a) Power spectrum of drag and

lift coefficients and (b) phase diagram of
0.4 -0.2
drag and lift coefficients for Re ¼ 100,
-0.4 Re ¼ 200, and Re ¼ 300.

0.2 -0.6
Re=100
Re=200
-0.8 Re=300

0 -1
0 0.2 0.4 0.6 0.8 1 1.2 1.25 1.3 1.35 1.4 1.45 1.5 1.55
Frequency Drag

Phys. Fluids 33, 053608 (2021); doi: 10.1063/5.0042603 33, 053608-10

Published under an exclusive license by AIP Publishing
 Physics of Fluids ARTICLE scitation.org/journal/phf

TABLE IV. Comparison of Strouhal number and drag and lift coefficients of the periodic flow for Re ¼ 100, 200, and 300.

100 200 300

Re
Reference St CD CL St CD CL St CD CL
52
Frank et al. 0.194 1.31 60.65 0.205 1.32 60.84
Williamson2 0.163 0.185 0.203
Silva et al.29 0.160 1.39 0.180 0.200 1.27
Le et al.53 0.160 1.37 6 0.009 60.323 0.187 1.34 6 0.030 60.430 0.200
Berthelsen and Faltinsen54 0.169 1.38 6 0.010 60.340 0.200 1.37 6 0.046 60.700
Wang et al.55 0.170 1.379 60.357 0.195 1.262 60.708 0.206 1.174
Sen56 0.165 1.394 6 0.007 60.191 0.197 1.375 6 0.038 60.453 0.209 1.401 6 0.068 60.607
Present study 0.162 1.325 6 0.026 60.306 0.200 1.333 6 0.046 60.351 0.210 1.41 6 0.0645 60.62

phenomenon. This phenomenon can be seen at t ¼ 2.5 for Re ¼ 550

, which was clealy absent in Fig. 15(a) for Re ¼ 300.
FIG. 13. Streamlines for Re ¼ 300 at t ¼ 2.5, top, experimental5 [reprinted with per-
mission from Bouard and Coutanceau, “The early stage of development of the
wake behind an impulsively started cylinder for 40 < Re < 104,” J. Fluid Mech.
101(3), 583–607 (1980). Copyright 2006 Cambridge University Press], and bottom,
numerical.
[see Fig. 14(b)]. The strength of this secondary eddy increases [see Fig.
14(c)] as time progresses and has a rotation opposite to the primary
eddy. The comparison of our computed data of early flow evolution
for Re ¼ 550 with experimental results5 can also be found in Table III
presented in Sec. V B. One can also find the existence of these eddies
from the distribution of vorticities on the cylinder surface in Fig. 15,
where two successive change of signs indicates the presence of a vor-
tex. For example in Fig. 15(b), one can clearly see the proof of the exis-
tence of the secondary eddy phenomena at time t ¼ 2.5 for Re ¼ 550 at
around h ¼ 42:6
In Figs. 13 and 14(b), we also compare the computed streamlines with
the experimental visualization of Ref. 5 for Re ¼ 300 and 550, respec-
tively, at time t ¼ 2.5. These figures exemplify the extreme closeness of
our numerical results with the experimental ones, thus demonstrating
the efficiency of our scheme. We have further exhibited the sensitivity
of the grid distribution in the vicinity of the cylinder and its effect on
the computed solution for Re ¼ 550 in Fig. 16. First, in Figs. 16(a) and
16(b) we show close-up views the junctions of the blocks 1, 3, 7, and 9
of Fig. 2 (see the bottom right of Fig. 3 also) on two different grids,
viz., Grid1 and Grid2 of the same size 361  241. The comparison of
velocity distributions along the flow axis behind the cylinder resulting
from computations on these grids with the experimental results of
Ref. 5 is shown in Figs. 16(c) and 16(d), respectively. One can clearly

0.5
0.5

0 0
y

-0.5
-0.5

0 0.5 1 1.5 0 0.5 1 1.5 2

x x

FIG. 14. Streamlines for Re ¼ 550 at (a) t ¼ 1.5, (b) t ¼ 2.5, top, experimental5 [reprinted with permission from Bouard and Coutanceau, “The early stage of development of
the wake behind an impulsively started cylinder for 40 < Re < 104,” J. Fluid Mech. 101(3), 583–607 (1980). Copyright 2006 Cambridge University Press], and bottom, numer-
ical and (c) t ¼ 5.0.

Phys. Fluids 33, 053608 (2021); doi: 10.1063/5.0042603 33, 053608-11

Published under an exclusive license by AIP Publishing
 Physics of Fluids ARTICLE scitation.org/journal/phf

-30
-30

-20
-20

-10
Vorticity

Vorticity
-10
FIG. 15. Surface vorticity at different time
0 stations for (a) Re ¼ 300 and (b)
0 Re ¼ 550.
t=0.5 t=0.5
10 t=1.0
10 t=1.0
t=1.5 t=1.5
t=2.0 t=2.0
t=2.5 t=2.5
t=3.0 t=3.0

20 20
180 150 120 90 60 30 0 180 150 120 90 60 30 0
Angle in degree Angle in degree

FIG. 16. (a) Grid1 with severe clustering, (b) Grid2 with moderate clustering, and Comparison between experimental5 and numerical results for the velocity distribution on flow
axis for Re ¼ 550 on (c) Grid1 and (d) Grid2 (Dt ¼ 104 on both the grids).

Phys. Fluids 33, 053608 (2021); doi: 10.1063/5.0042603 33, 053608-12

Published under an exclusive license by AIP Publishing
 Physics of Fluids ARTICLE scitation.org/journal/phf

FIG. 17. Secondary vortex phenomena for Re ¼ 5000: (a) a-phenomenon at t ¼ 2.25 and (b) b-phenomenon at t ¼ 1.1.

FIG. 18. Tertiary vortex phenomena for Re ¼ 5000: (a) sub-a phenomenon at t ¼ 1.45 and (b) sub-b phenomenon at t ¼ 3.15.

see that the results produced by Grid2 are much closer to the experi- streamlines close to the cylinder initially deviate from the surface caus-
mental ones than Grid1. ing a bulge pattern eventually giving rise to a secondary eddy. This
eddy grows in size to such an extent that it touches the boundary of
the main eddy, thereby splitting the main one into two parts and iso-
D. Flows beyond Re ¼ 550
lating the region of the wake next to the separation point where
Flow around a cylinder eventually becomes three dimensional another secondary eddy becomes visible. When these two secondary
and turbulent at these Re’s. However, during the early stages, the flow eddies become equivalent in size and strength, the phenomenon is
is still laminar as suggested by laboratory experiments, and hence, we known as the a-phenomenon [see Fig. 17(a)]. On the other hand, for
limit our discussion only to the very early stages for this range of certain Re’s, just after the start of the flow, a very thin recirculating
Reynolds numbers. This range is characterized by the intriguing inter-
play between secondary and tertiary vortices leading to very interesting
flow phenomena such as a, b, sub-a, and sub-b. Before embarking
into flow description for this range of Re’s, we briefly explain these
phenomena; the readers may refer to Figs. 17 and 18. In these and sub-
sequent figures, only the upper part behind the cylinder is depicted as
the flow is still symmetric in the early stages of the flow.

E. a, b, sub-a, and sub-b phenomena

The terms a and b phenomena were coined by Bouard and
Coutanceau in their pathbreaking experimental study5 in 1980. For
some Reynolds numbers in the regime, when the primary vortex P is FIG. 19. Schematic diagram for a, b, sub-a, and sub-b phenomena. The letters P,
still growing behind the cylinder and stable, as time progresses, the S, and T stand for primary, secondary, and tertiary vortices, respectively.

Phys. Fluids 33, 053608 (2021); doi: 10.1063/5.0042603 33, 053608-13

Published under an exclusive license by AIP Publishing
 Physics of Fluids ARTICLE scitation.org/journal/phf

wake is formed; but soon afterwards, the core of this recirculating zone
rotates in one piece with a speed which is much faster than the other
part of the separated zone, forming a vortex which gains strength and
size with time. After a while, this vortex separates the initial wake into
two parts where the one situated near the point of separation is occu-
pied by a pair of secondary eddies, which is known as the b-phenome-
non [see Fig. 17(b)]. When the a and b phenomena are replicated at
the tertiary vortex level, it is termed as sub-a and sub-b phenomena,
respectively [see Figs. 18(a) and 18(b)], which was coined by Kalita
and Sen.10
We now provide a comprehensive description of the flow for
Re ¼ 7500 in the early laminar stage up to t ¼ 5.0. This description
is more or less representative of the flow structures during the early
stages for all the Reynolds numbers considered in this section as
would be seen in Sec. V F. For the benefit of the readers, we have
provided a schematic of the primary, secondary, and tertiary

FIG. 20. Evolution of streamlines for Re ¼ 7500 for flow past a circular cylinder at: (a) t ¼ 0.5, (b) t ¼ 0.85, (c) t ¼ 0.95, (d) t ¼ 1.05, (e) t ¼ 1.3, (f) t ¼ 1.75, (g) t ¼ 1.85, (h)
t ¼ 2.0, (i) t ¼ 2.05, (j) t ¼ 2.3, (k) t ¼ 2.4, (l) t ¼ 2.5, (m) t ¼ 2.6, (n) t ¼ 2.7, (o) t ¼ 3.10, (p) t ¼ 3.15, (q) t ¼ 3.75, (r) t ¼ 4.00, (s) t ¼ 4.25, and (t) t ¼ 5.00.

Phys. Fluids 33, 053608 (2021); doi: 10.1063/5.0042603 33, 053608-14

Published under an exclusive license by AIP Publishing
 Physics of Fluids ARTICLE scitation.org/journal/phf

vortices appearing on the surface of the cylinder in the upper half TABLE V. Streamfunction w values at the centers of the tertiary vortices constituting
in Fig. 19. the sub-a phenomenon for Re ¼ 7500.
In Figs. 20(a)–20(t), we portray the evolution of primary, second-
ary, and tertiary vortices for Re ¼ 7500, which in the process is charac- Vortex center t ¼ 2.4 t ¼ 3.15
terized by the occurrence of a-, b-, sub-a-, and sub-b phenomena. At T1 0.002 811 17 0.001 497 73
time approximately t ¼ 0.25, a pair of symmetric vortices (denoted by
T2 0.002 876 32 0.001 468 87
“P” in Fig. 19) is formed owing to typical boundary layer separation,
which can be seen more clearly behind the cylinder at t ¼ 0.50 [Fig.
20(a)]. As time progresses, they gain both size and strength maintain-
ing their symmetry alongside. Simultaneously, the core of the Primary one observed here, viz., t u 1:85 when the tertiary vortex T2 is formed
vortex travels upwards in the direction of the downstream. At around at the location h ¼ 42
[see Figs. 19 and 20(g)]. At this instant, the
t ¼ 0.85 [Fig. 20(b)], there is a secondary separation on the cylinder’s streamfunction value of the secondary vortex S1 shielding it reaches a
surface leading to the formation of a secondary vortex (S1 in Fig. 19). local maxima at its center, causing an adverse pressure gradient. A few
This vortex grows in size and strength with time and becomes promi- moments later, this tertiary vortex gains more strength and hence size,
nent at around t ¼ 0.95 [Fig. 20(c)] and breaks the primary vortex into breaking the secondary vortex S1 into two parts T1 and T2 such that
two parts, one of the parts being the core. We can also see this phe- these tertiary vortices T1 and T2 are almost equal in strength at time
nomenon from three alternating rotating zones of surface vorticity of t ¼ 2.4 [see Figs. 19, 20(j), and 20(k) and Table V]. This is the advent
the cylinder [see Figs. 20(b)–20(d)]. of the sub-a phenomenon.10 Soon afterwards, at time t ¼ 2.5, T2
However, the part with lesser strength (S2 in Fig. 19) still contin- merges with the parent secondary vortex S1, and its split part gets re-
ues its correspondence with its core. With time, both the core and the attached with S1 which can be seen in Fig. 20(l). This is followed by
secondary vortex gain size and strength, which trigger the narrowing of the merger of the parts of this secondary vortex to form one single vor-
the communication between the two parts of the core [see Fig. 20(e)]. tex. It resurfaces again at t ¼ 2.70 [Fig. 20(n)], leading to another sub-
However as time progresses, the primary vortex becomes more prom- a-phenomenon at t ¼ 3.15 [see Fig. 20(p) and Table V] persisting till
inent in size again. Both its parts (namely P and S2) start getting more t ¼ 3.45 after which S2 and its appendage coalesce back to form one
robust and start sandwiching the secondary vortex S1 from left and single vortex once again.
right, resulting in the reduction of its strength and size at time In conjunction with the observations made in Ref. 10, the tertiary
t u 1:05 [Fig. 20(e)]. This is the onset of the b-phenomenon. vortices T1, T2, T3, and T4 are much smaller in size and weaker in
Momentarily the size and strength of the appendage S2 of the pri- strength than their secondary counterparts S1 and S2. However, the
mary vortex gets reduced and reconnects with P. Following that, the instants and the frequency at which secondary and tertiary vortices
secondary vortex S1 regains strength, grows in size, and divides P again appear, their size, rate at which they grow and strengths are different
to recreate the appendage S2. It grows in size and intensity along with from the ones reported in Ref. 10 for Re ¼ 5000. Moreover, while for
the other secondary vortex S1 [Figs. 20(f)–20(h)], and at time t ¼ 2.05,
they become almost equal in size and strength [Fig. 20(i)], and as such,
constitute the a-phenomenon. Both the secondary vortices S1 and S2
maintain their growth, both in size and strength [Figs. 20(h) and
20(i)].
Next, we lay down the details of the tertiary vortex dynamics for
Re ¼ 7500; its first evidence on the cylinder’s surface is observed at
t u 1:05 as can be seen from Fig. 20(e), where the secondary vortex S2
acts like a shield providing a padding around it. An even closer look at
this figure divulges its location to be at h u 58
on the cylinder surface.
Meanwhile, the center of secondary vortex S1 travels down to the posi-
tion corresponding to h ¼ 45
. While slowly gaining size and inten-
sity, this new tertiary vortex breaks the primary vortex’s appendage S2,
into two parts viz. T3 and T4 [Figs. 19 and 20(e)]. This is nothing but
the sub-b phenomenon.10 This sub-b phenomenon lingers till
t ¼ 1.35, when the secondary vortex S2 and its appendage reunite
forming one single vortex. In the process, it pushes the tertiary vortex
T3 downwards reducing its size and strength. After some time, T3
regains size and strength [see Fig. 20(i) at t ¼ 2.05] leading to another
sub-b phenomenon at t ¼ 2.60 [see Fig. 20(m)] persisting till t ¼ 3.10
[see Fig. 20(o)]. The tertiary vortex T3 is visible till t ¼ 4.00 and dissi-
pates afterwards.
In the meantime, another interesting event homologous to the a-
phenomenon [Figs. 20(h)–20(i)] takes place at secondary and tertiary FIG. 21. The comparison of streamlines for Re ¼ 5000 and 7500 at time t ¼ 5.0.
levels around t u 2:00. This is similar to the observation made by Their evolution from t ¼ 0.0 to t ¼ 5.0 can be seen in the accompanying video.
Kalita and Sen10 for Re ¼ 5000, though at an instant much earlier than Multimedia View: https://doi.org/10.1063/5.0042603.1

Phys. Fluids 33, 053608 (2021); doi: 10.1063/5.0042603 33, 053608-15

Published under an exclusive license by AIP Publishing
Published under an exclusive license by AIP Publishing
Phys. Fluids 33, 053608 (2021); doi: 10.1063/5.0042603

Physics of Fluids
TABLE VI. Characteristics of the secondary and tertiary vortices for different Reynolds numbers for 0 < t  5:0.

Sub-a preceded
Re a b Sub-a Sub-b by sub-b Remarks

700 The secondary vortex S1 appears at around t ¼ 1.6 and grows in size
creating the appendage of the primary vortex P, but never splits into
two separate vortices.
750 S1 appears at around t ¼ 1.25 and eventually splits P into two parts;
however, the secondary vortices S1 and S2 are never equivalent in size
and strength. S1 does not reattach with P again.
800 S1 appears at around t ¼ 1.225; rest is similar to Re ¼ 750.
1000 S1 appears at around t ¼ 1.15, splits P into two parts; S1 and S2 are
equivalent in size from t ¼ 4.5 onwards; however, their strengths are
never close to each other. S1 does not reattach with P again.
2000 S1 appears at around t ¼ 1.05, splits P into two parts; S1 and S2 are
equivalent in size from t ¼ 2.25 onwards; like Re ¼ 1000 their
strengths are never close to each other. In fact, strength of S2 is almost
twice the strength of S1 . T1 and T2 make appearance between t ¼ 2.55
to 3.0; however, no sub-a phenomenon is observed. S1 does not reat-
tach with P again.
3000 t ¼ 3.05. S1 appears at around t ¼ 1.05, splits P into two parts; S1 and S2 are
equivalent in size from t ¼ 2.1 onwards; strengths are never close to
each other during this period. S1 does not reattach with P again.
4000 t ¼ 2.65. ð1:2; 2:2Þ t ¼ 2.75. ð1:5; 1:85Þ and Once the a phenomenon starts, it is persistent throughout the period.
ð3:25; 3:6Þ.
4500 t ¼ 2.55. ð1:1; 2:15Þ t ¼ 2.6. ð1:4; 1:5Þ and The a phenomenon is persistent from inception.
ð3:0; 3:2Þ.
5000 t ¼ 2.5. ð1:1; 2:1Þ t ¼ 1.45 and ð1:3; 1:45Þ and The a phenomenon is persistent from inception.
2.545. ð2:75; 3:15Þ.
6000 t ¼ 2.4. ð1:1; 2:05Þ t ¼ 1.35, 2.5 ð1:25; 1:4Þ and After inception, the a phenomenon breaks down at around t ¼ 3.55;
and 3.65. ð2:5; 2:7Þ. regroups again at 4.5. The sub-b phenomenon at t ¼ 2.7 almost leads

ARTICLE
to a sub-a phenomenon, but not yet.
7500 t ¼ 2.3, 3.9 ð1:1; 2:0Þ t ¼ 2.4 1.3 and 2.6. After inception, the a phenomenon breaks down at around t ¼ 2.6;
and 4.35. and 3.15. regroups again at 3.75. The sub-b phenomenon at t ¼ 1.3 almost leads
to a sub-a phenomenon, but not yet. At t ¼ 1.85, T1 splits S1 into two
parts creating T2 ; at t ¼ 2.4, T1 ; T2 , and S1 are almost of the same size

scitation.org/journal/phf
unlike what happens in sub-a phenomenon. After the occurrence of
sub-a phenomenon at t ¼ 3.15, T1 and T2 persists throughout.
8000 t ¼ 2.2. ð1:0; 1:5Þ 2.15, 2.85, ð1:15; 1:275Þ. ð2:0; 2:355Þ At t ¼ 1.2, T1 ; T2 , and S1 are almost of the same size like Re ¼ 7, 500
and 4.8. at t ¼ 1.3. The sub-a phenomenon at t ¼ 2.355 which is preceded by
the sub-b phenomenon with the formation of T3 at t ¼ 2.0, the tertia-
ries T3 ; T4 are persistent till t ¼ 4.65. In the meantime their strengths
keeps on reducing and increasing alternatively with T4 showing a ten-
dency of reattaching with S2 . However, T4 could ultimately reattach
33, 053608-16

with S2 only at t ¼ 4.9. After inception, the a phenomenon breaks

down at around t ¼ 2.8; regroups again at 4.35.
 Physics of Fluids ARTICLE scitation.org/journal/phf

Re ¼ 5000, the a-phenomenon persists till the symmetry of the flow

formation of T3 at t ¼ 2.0, the tertiaries T3 ; T4 are persistent through-

From t ¼ 1.9 onwards, S1 and S2 are equivalent in size; however, their

and S2 are never close to each other in strength. The sub-a phenome-
break down again at 3.1 after which no equivalence in size is seen. S1

Figs. 24(b)–24(d)]. The overall observation for Re ¼ 10 000 is similar

non at t ¼ 2.2 which is preceded by the sub-b phenomenon with the

out. As in Re ¼ 8000, after the ups and downs in their strengths, the

existent as can be seen from Table VII. Several tertiary recirculation

zones are observed on the surface of the cylinder intermittently [see
reattachment of T4 with S2 finally takes place at t ¼ 2.95. Although
equivalence breaks down at t ¼ 2.35 which regroups again at 2.4 to

the a-phenomenon is absent, the sub-a phenomenon is very much

about x-axis is broken (at some instant t > 5) from the time of its
inception, in the case of Re ¼ 7500, the a-phenomenon occurs repeat-
edly and gets diffused three times within the time interval 0  t  5.
The subtle differences between the flows for Re ¼ 5000 and Re ¼ 7500
during 0  t  5 can be seen in the accompanying video embedded
in Fig. 21 showing the streamline evolution. Note that the flow being
symmetric about the x-axis in the time range under consideration, we
have shown the streamlines behind the cylinder for Re ¼ 5000 and
Re ¼ 7500 in the lower and upper halves, respectively, in this video.

to Re ¼ 9500.
Remarks

F. Summary on the Reynolds numbers exhibiting

secondary and tertiary vortex phenomena
In this section, first we summarize the existence of secondary and
tertiary vortex phenomena for the Reynolds number in the range of
700  Re  9500 and provide a brief qualitative and quantitative
description of the early evolution of flow for Re ¼ 3000, 5000, and
9500.
In Table VI, we summarize the secondary and tertiary vortex
characteristics for the early stage of the flow till t ¼ 5.0 for all the
Reynolds numbers for which we have computed the flow. One should
refer to Fig. 19 for the description of the primary vortex P, secondary
vortices S1 and S2, and tertiary vortices T1, T2, T3, and T4. Here, the
first column represents the Reynolds number Re, the second, the
Sub-a preceded

instant of time at which the secondary vortices S1 and S2 become

ð2:0; 2:2Þ
by sub-b

equivalent in size and strength for the first time, thus auguring the
a phenomenon, the entries ðb1 ; b2 ; b3 Þ in the third column, respec-
tively, represent the instants at which S1 is clearly visible, when the
pair (S1 ; S2 ) is clearly visible subsequently and the instant at which
S2 gets reattached with P again, the fourth column represents the
times at which sub-a phenomenon occur, the entries ðb1 ; b2 Þ in the
fifth column, respectively, represent the instants when T3 is formed
ð1:1; 1:2Þ.
Sub-b

and the pair ðT3 ; T4 Þ is clearly visible, the entries ðb; aÞ in the sixth
column, respectively, represent the instant at which T3 emerges to
eventually form the sub-b phenomenon and the instant at which
the tertiary vortices T3 and T4 become equivalent in strength giving
rise to the sub-a phenomenon, and finally, we offer some remarks
1.85, 3.0, 3.45,

in the last column.

4.2 and 4.65.

In Figs. 22–24, we show the evolution of streamlines for

Sub-a

Re ¼ 3000, 5000, and 9500. For Re ¼ 5000 (Fig. 23) and 9500 (Fig. 24),
we compare our computed solutions with the experimental results of
Bouard and Coutanceau,5 where in the upper half contains the
experimental visualization, and the lower half, our simulated
results. Once again, our computed solutions are extremely close to
ð1:0; 1:45Þ

the experimental ones. The secondary and tertiary vortex phenom-

b

ena exhibited by these flows are described in Table VI. The exis-
tence of these phenomena is also obvious from the distribution of
vorticities on the cylinder surface in Fig. 25 for Re ¼ 3000, 5000,
7500, and 9500. The comparison between experimental and numer-
TABLE VI. (Continued.)

ical results for the velocity distribution on flow axis for these three
a

Reynolds number can be seen in Fig. 26, which is again evident

from the proximity of our numerical results with the experimental
ones. In Table VII, we present the strengths of the tertiary vortices
9500

T1 and T2, which validates the presence of sub-a phenomenon as

Re

tabulated in the summary Table VI.

Phys. Fluids 33, 053608 (2021); doi: 10.1063/5.0042603 33, 053608-17

Published under an exclusive license by AIP Publishing
 Physics of Fluids ARTICLE scitation.org/journal/phf

FIG. 22. Streamlines for Re ¼ 3000 at (a) t ¼ 0.5, (b) t ¼ 1.0, (c) t ¼ 1.5, (d) t ¼ 2.0, and (e) t ¼ 2.5.

VI. CONCLUSION
Finite difference simulation of the flow past an impulsively
started circular cylinder is well represented in the existing literature.
However, the majority of them are performed on polar grids. Most of
the simulations for this problem on Cartesian grid utilizes the
immersed interface approach, that too, up to Reynolds number regime
exhibiting the vortex shedding phenomena only. In the current study,
we carry out a comprehensive simulation of the flow past an impul-
sively started circular cylinder through a recently developed second
order spatially and temporally accurate compact scheme on

FIG. 23. Streamlines for Re ¼ 5000 at (a) t ¼ 1.0, (b) t ¼ 1.5, (c) t ¼ 2.0, and (d) t ¼ 2.5 {top half, experimental5 and bottom half, numerical [reprinted with permission from
Bouard and Coutanceau, “The early stage of development of the wake behind an impulsively started cylinder for 40 < Re < 104,” J. Fluid Mech. 101(3), 583–607 (1980).
Copyright 2006 Cambridge University Press]}.

Phys. Fluids 33, 053608 (2021); doi: 10.1063/5.0042603 33, 053608-18

Published under an exclusive license by AIP Publishing
 Physics of Fluids
nonuniform Cartesian grid. We provide a detailed account of the gen-
eration of the grid in such a way that the cylinder boundary passes
through the grid points, thus doing away with the need to use the
immersed interface approach. Along with a grid independence study,
we have also established the efficiency of the simulation on Cartesian
vis a vis polar grid. Results are presented for a wide range of Reynolds
numbers (Re) ranging from 5  Re  10 000 in the laminar regime.
This includes the periodic regime, which is characterized by the
Phys. Fluids 33, 053608 (2021); doi: 10.1063/5.0042603
Published under an exclusive license by AIP Publishing
ARTICLE scitation.org/journal/phf
FIG. 24. Streamlines for Re ¼ 9500 at (a)
t ¼ 0.75, (b) t ¼ 1.0, (c) t ¼ 1.25, and (d)
t ¼ 1.5 {top half, experimental5 and bottom
half, numerical [reprinted with permission
from Bouard and Coutanceau, “The early
stage of development of the wake behind
an impulsively started cylinder for
40 < Re < 104 ,” J. Fluid Mech. 101(3),
583–607 (1980). Copyright 2006
Cambridge University Press]}.
presence of von Karman vortex street highlighting the robustness of
the numerical scheme being used. They are validated through compar-
ison with experimental and well established numerical results and the
accuracy of the solutions is exemplified by their close proximity to
them. Not only the a, b, sub-a, and sub-b phenomena, which are the
trademark of the secondary and tertiary vortex dynamics associated
with such flows, are studied in detail, for the first time, they have also
been documented compactly in a tabular form for Re  700.
33, 053608-19
 Physics of Fluids ARTICLE scitation.org/journal/phf

FIG. 25. Time evolution of surface vorticity for (a) Re ¼ 3000, (b) Re ¼ 5000, (c) Re ¼ 7500, and (d) Re ¼ 9500.

FIG. 26. Comparison between experimen-

tal and numerical results for the velocity
distribution on flow axis for (a) Re ¼ 3000,
5000, and (b) 9500.

Phys. Fluids 33, 053608 (2021); doi: 10.1063/5.0042603 33, 053608-20

Published under an exclusive license by AIP Publishing
 Physics of Fluids ARTICLE scitation.org/journal/phf

17
TABLE VII. Streamfunction w values at the centers of the tertiary vortices constitut- Y. Wang and M. A. Maksoud, “Application of the remeshed vortex method to
ing the sub-a phenomenon for Re ¼ 9500. the simulation of 2D flow around circular cylinder and foil,” Ship Technol. Res.
65(2), 79–86 (2018).
18
Time T1 T2 G. H. Cottet and P. D. Koumoutsakos, Vortex Methods: Theory and Practice
(Cambridge University Press, 2000).
19
t ¼ 1.85 0.002 946 25 0.003 058 7 X. He and G. Doolen, “Lattice Boltzmann method on curvilinear coordinates
system: Flow around a circular cylinder,” J. Comput. Phys. 134(2), 306–315
t ¼ 3.0 0.002 577 13 0.002 739 3 (1997).
t ¼ 3.45 0.017 408 3 0.012 690 1 20
X. D. Niu, Y. T. Chew, and C. Shu, “Simulation of flows around an impulsively
t ¼ 4.2 0.004 767 4 0.004 372 7 started circular cylinder by Taylor series expansion and least squares-based lat-
t ¼ 4.65 0.002 869 96 0.003 074 5 tice Boltzmann method,” J. Comput. Phys. 188(1), 176–193 (2003).
21
Y. Li, R. Shock, R. Zhang, and H. Chen, “Numerical study of flow past an
impulsively started cylinder by the lattice-Boltzmann method,” J. Fluid Mech.
519, 273–300 (2004).
ACKNOWLEDGMENTS 22
X. He and G. Doolen, “Lattice Boltzmann method on a curvilinear coordinate
system: Vortex shedding behind a circular cylinder,” Phys. Rev. E 56(1),
The second author acknowledges Grant No. MTR/2017/ 434–440 (1997).
23
000482 from Science and Engineering Research Board of the J. C. Kalita and R. K. Ray, “A transformation-free HOC scheme for incom-
Department of Science and Technology, Government of India to pressible viscous flows past an impulsively started circular cylinder,”
carry out the current research. J. Comput. Phys. 228(17), 5207–5236 (2009).
24
B. N. Rajani, A. Kandasamy, and S. Majumdar, “Numerical simulation of laminar
DATA AVAILABILITY 25
flow past a circular cylinder,” Appl. Math. Model. 33(3), 1228–1247 (2009).
K. Taira and T. Colonius, “The immersed boundary method: A projection
The data that support the findings of this study are available approach,” J. Comput. Phys. 225(2), 2118–2137 (2007).
from the corresponding author upon reasonable request. 26
M. N. Linnick and H. F. Fasel, “A high-order immersed interface method for
simulating unsteady incompressible flows on irregular domains,” J. Comput.
REFERENCES Phys. 204(1), 157–192 (2005).
27
1 E. M. Kolahdouz, A. P. S. Bhalla, B. A. Craven, and B. E. Griffith, “An
P. Bearman, “Bluff body hydrodynamics,” Twenty-First Symposium on Naval immersed interface method for discrete surfaces,” J. Comput. Phys. 400(1),
Hydrodynamics (1997), pp. 561–579. 108854–108837 (2020).
2
C. H. K. Williamson, “Vortex dynamics in the cylinder wake,” Annu. Rev. 28
D. Russell and Z. J. Wang, “A Cartesian grid method for modeling multiple
Fluid Mech. 28(1), 477–539 (1996). moving objects in 2D incompressible viscous flow,” J. Comput. Phys. 191(1),
3
A. Roshko, “Perspectives on bluff body aerodynamic,” J. Wind Eng. Ind. 177–205 (2003).
Aerodyn. 49(1), 79–100 (1993). 29
A. L. F. L. E. Silva, A. Silveira-Neto, and J. J. R. Damasceno, “Numerical simu-
4
G. K. Batchelor, An Introduction to Fluid Dynamics (Cambridge University lation of two-dimensional flows over a circular cylinder using the immersed
Press, Cambridge, 1993). boundary method,” J. Comput. Phys. 189(2), 351–370 (2003).
5
R. Bouard and M. Coutanceau, “The early stage of development of the wake 30
D. Calhoun, “A Cartesian grid method for solving the two-dimensional
behind an impulsively started cylinder for 40 < Re < 104 ,” J. Fluid Mech. 101, streamfunction-vorticity equations in irregular regions,” J. Comput. Phys.
583–607 (1980). 176(2), 231–275 (2002).
6
T. P. Loc and R. Bouard, “Numerical solution of the early stage of the unsteady 31
R. Mittal and G. Iaccarino, “Immersed boundary methods,” Annu. Rev. Fluid
viscous flow around a circular cylinder: A comparison with experimental visu- Mech. 37, 239–261 (2005).
alization and measurements,” J. Fluid Mech. 160, 93–117 (1985). 32
M. Ma, W. X. Huang, and C. X. Xu, “A dynamic wall model for large eddy sim-
7
Y. V. S. S. Sanyasiraju and V. Manjula, “Flow past an impulsively started circu- ulation of turbulent flow over complex/moving boundaries based on the
lar cylinder using a higher-order semicompact scheme,” Phys. Rev. E 72(1), immersed boundary method,” Phys. Fluids 31(11), 115101 (2019).
016709 (2005). 33
S. Tao, Q. He, L. Wang, S. Huang, and B. Chen, “A non-iterative direct-forcing
8
J. C. Kalita and S. Sen, “a-, b-phenomena in the post-symmetry break for the immersed boundary method for thermal discrete unified gas kinetic scheme with
flow past a circular cylinder,” Phys. Fluids 29(3), 033603 (2017). Dirichlet boundary conditions,” Int. J. Heat Mass Transfer 137, 476–488 (2019).
9 34
D. J. Tritton, “Experiments on the flow past a circular cylinder at low Reynolds P. Kumar and J. C. Kalita, “An efficient w  v scheme for two-dimensional
number,” J. Fluid Mech. 6(4), 547–567 (1959). laminar flow past bluff bodies on compact nonuniform grids,” Int. J. Numer.
10
J. Kalita and S. Sen, “Unsteady separation leading to secondary and tertiary Methods Fluids 92(1), 1723–1752 (2020).
vortex dynamics: The sub-a and sub-b phenomena,” J. Fluid Mech. 730, 19–51 35
M. M. Gupta and J. C. Kalita, “A new paradigm for solving Navier-Stokes equa-
(2013). tions: Streamfunction-velocity formulation,” J. Comput. Phys. 207, 52–68 (2005).
11 36
J. C. Kalita, “Effect of boundary location on the steady flow past an impulsively M. M. Gupta and J. C. Kalita, “New paradigm continued: Further computation
started circular cylinder,” J. Comput. Sci. Math. 5(3), 252–279 (2014). with stream function-velocity formulation for solving Navier-Stokes equa-
12
C. Jackson, “A finite-element study of the onset of vortex shedding in flow past tions,” Commun. Appl. Anal. 10(4), 461–490 (2006).
variously shaped bodies,” J. Fluid Mech. 182, 23–45 (1987). 37
J. C. Kalita and M. M. Gupta, “A stream function-velocity approach for 2D
13
S. Sen, S. Mittal, and G. Biswas, “Steady separated flow past a circular cylinder transient incompressible viscous flows,” Int. J. Numer. Methods Fluids 62(3),
at low Reynolds numbers,” J. Fluid Mech. 620(2009), 89–119 (2009). 237–266 (2010).
14 38
D. Kim and H. Choi, “A second-order time-accurate finite volume method for M. Ben-Artzi, J.-P. Croisille, and D. Fishelov, “A high order compact scheme
unsteady incompressible flow on hybrid unstructured grids,” J. Comput. Phys. for the pure-streamfunction formulation of the Navier-Stokes equations,”
162(2), 411–428 (2000). J. Sci. Comput. 42, 216–250 (2010).
15 39
Y. Liu, W. Zhang, Y. Jiang, and Z. Ye, “A high-order finite volume method on S. Sen, J. C. Kalita, and M. M. Gupta, “A robust implicit compact scheme for
unstructured grids using RBF reconstruction,” Comput. Math. Appl. 72(4), two-dimensional unsteady flows with a biharmonic stream function for-
1096–1117 (2016). mulation,” Comput. Fluids 84, 141–163 (2013).
16 40
P. Koumoutsakos and A. Leonard, “High-resolution simulations of the flow S. Dutta, P. Kumar, and J. C. Kalita, “Streamfunction-velocity computation of
around an impulsively started cylinder using vortex methods,” J. Fluid Mech. natural convection around heated bodies place in a square enclosure,” Int. J.
296, 1–38 (1995). Heat Mass Transf. 152, 119550 (2020).

Phys. Fluids 33, 053608 (2021); doi: 10.1063/5.0042603 33, 053608-21

Published under an exclusive license by AIP Publishing
 Physics of Fluids ARTICLE scitation.org/journal/phf

41 50
J. C. Kalita and S. Sen, “The biharmonic approach for unsteady flow past an P. Kumar and J. C. Kalita, “A transformation-free w-v formulation of the
impulsively started circular cylinder,” Commun. Comput. Phys. 12(4), Navier-Stokes equations on compact nonuniform grids,” J. Comput. Appl.
1163–1182 (2012). Math. 353, 292–317 (2019).
42 51
J. D. Anderson, Computational Fluid Dynamics: The Basics with Applications H. Jiang and L. Cheng, “Transition to the secondary vortex street in the wake
(McGraw-Hill Education, 1995). of a circular cylinder,” J. Fluid Mech. 867, 691–722 (2019).
43 52
J. D. Hoffmann, Numerical Methods for Engineers and Scientists (Marcel R. Franke, W. Rodi, and B. Sch€ onung, “Numerical calculation of laminar
Dekker Inc., New York, 2001), pp. 297–298. vortex-shedding flow past cylinders,” J. Wind Eng. Ind. Aerodyn. 35, 237–257
44
D. Anderson, J. C. Tannehill, and R. H. Pletcher, Computational Fluid (1990).
53
Mechanics and Heat Transfer (CRC Press, 2012). D. V. Le, B. C. Khoo, and J. Peraire, “An immersed interface method for vis-
45
S. Taneda, “Flow visualization,” Proc. Jpn. Soc. Civil Eng. 387, 1–10 (1987). cous incompressible flows involving rigid and flexible boundaries,” J. Comput.
46
M. Van Dyke, An Album of Fluid Motion (The Parabolic Press, Stanford, 1982). Phys. 220, 109–138 (2006).
47 54
I. M. Kozlov, K. V. Dobrego, and N. N. Gnesdilov, “Application of RES meth- P. A. Berthelsen and O. M. Faltinsen, “A local directional ghost cell approach
ods for computation of hydrodynamic flows by an example of a 2D flow past for incompressible viscous flow problems with irregular boundaries,”
circular cylinder for Re ¼ 5–200,” Int. J. Heat Mass transfer 54, 887–893 J. Comput. Phys. 227, 4354–4397 (2008).
55
(2011). Z. Wang, J. Fan, and K. Cen, “Immersed boundary method for the simulation
48
M. Coutanceau and R. Bouard, “Experimental determination of the main fea- of 2D viscous flow based on vorticity-velocity formulations,” J. Comput. Phys.
tures of the viscous flow in the wake of a circular cylinder in uniform transla- 228, 1504–1520 (2009).
56
tion. Part 2. Unsteady flow,” J. Fluid Mech. 79(2), 257–272 (1977). S. Sen, “Compact biharmonic computation of the Navier-Stokes equations:
49
S. Sen and J. C. Kalita, “A 4OEC scheme for the biharmonic steady Navier- Extension to complex flows,” Ph.D. thesis, 2012.
57
Stokes equations in non-rectangular domains,” Comput. Phys. Commun. 196, S. Mittal and B. Kumar, “Flow past a rotating cylinder,” J. Fluid Mech. 476,
113–133 (2015). 303–334 (2003).

Phys. Fluids 33, 053608 (2021); doi: 10.1063/5.0042603 33, 053608-22

Published under an exclusive license by AIP Publishing