Spatial evolution of supersonic streamwise vortices

Toshihiko Hiejima

Citation: Physics of Fluids (1994-present) 26, 074102 (2014); doi: 10.1063/1.4886097

View online: http://dx.doi.org/10.1063/1.4886097
View Table of Contents: http://scitation.aip.org/content/aip/journal/pof2/26/7?ver=pdfcov
Published by the AIP Publishing

Articles you may be interested in

Linear stability analysis on supersonic streamwise vortices
Phys. Fluids 25, 114103 (2013); 10.1063/1.4828055

On the feedback mechanism in supersonic cavity flows

Phys. Fluids 25, 056101 (2013); 10.1063/1.4804386

Wall pressure fluctuations beneath supersonic turbulent boundary layers

Phys. Fluids 23, 085102 (2011); 10.1063/1.3622773

Near-wall dynamics of compressible boundary layers

Phys. Fluids 23, 065109 (2011); 10.1063/1.3600659

Numerical study of screech generation in a planar supersonic jet

Phys. Fluids 19, 075105 (2007); 10.1063/1.2747225

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
130.191.17.38 On: Sat, 08 Nov 2014 19:06:30
 PHYSICS OF FLUIDS 26, 074102 (2014)

Spatial evolution of supersonic streamwise vortices

Toshihiko Hiejimaa)
Aerospace Engineering, Osaka Prefecture University, Sakai, Osaka 599-8531, Japan
(Received 8 January 2014; accepted 19 June 2014; published online 10 July 2014)

In this paper, the spatial developments of high Mach number streamwise vortices have
been investigated using direct numerical simulations for Mach numbers from 2.5 to
9.0. The particular streamwise vortices studied here are those with uniform stagnation
temperatures and those with constant entropy. For vortices with uniform stagnation
temperatures, the calculated spatial evolutions show that for small swirl parameters,
a bending wave mode is excited and spiral structures of low-wavenumber modes only
develop downstream. In contrast, for swirl parameters larger than a moderate amount
(approximately 0.5), relaminarization occurs downstream. This study also found that
the relation between the circulation and the freestream Mach number contributes to
a necessary condition as to whether perturbations can grow in supersonic flows. For
isentropic streamwise vortices, instability properties at high Mach numbers parallel
those in low-speed flows because the evolution can give rise to high-wavenumbers
within the core. The resulting entropy instability is analogous to the Rayleigh-Taylor
instability and has the advantage of being insensitive to compressibility effects in
supersonic flows.  C 2014 AIP Publishing LLC. [http://dx.doi.org/10.1063/1.4886097]

I. INTRODUCTION
Swirling flows occur in many situations. In terms of vorticity, it is meaningful to regard swirling
flows as streamwise vortices. An understanding of the dynamics of a vortex under axial flow is
useful to both fundamental physics and engineering; examples include tornados, trailing-line vortices
behind airplanes, and mixers in supersonic combustors. Herein, we focus on high-speed flows for
which the evolution of a vortex has never been examined. Such a study faces a difficult issue in that
fluctuations in supersonic flows are suppressed because of the compressibility effect. The decay of
fluctuating quantities and the possibility of generating new large-scale structures depend on stability
characteristics of the mean flow field. For turbulent transitions in low-speed flows, streamwise
vortices are essential for developing and maintaining the turbulent state. Moreover, in the processes
leading to turbulence, streamwise vortices play a crucial role at every spatial scale. Unlike in
subsonic flows, instability waves involved in the growth of perturbations cannot propagate upstream
in supersonic flows. Accordingly, understanding the instability features of streamwise vortices at
high speeds should help resolve critical issues. For practical applications, it is important to know
whether transitions in supersonic conditions can sustain turbulence or whether relaminarization
occurs. Elucidating relations between large and small scales in supersonic vortices would contribute
to research on high-speed problems.
For streamwise vortices, it is known that the Batchelor vortex,1 namely, the q-vortex, is a
convenient model for real columnar vortices with smooth velocity profiles. Lessen et al.2 defined
a swirl parameter as the ratio of the magnitude of the maximum azimuthal velocity to that of the
maximum axial velocity deficit or excess. The parameter Φ is introduced to clearly distinguish
between the traditional swirl number and the velocity deficit or excess. The purpose of the present
study is to investigate characteristics due to spatial development of Batchelor vortices as supersonic
streamwise vortices.

a) Electronic mail: hiejima@aero.osakafu-u.ac.jp

1070-6631/2014/26(7)/074102/19/$30.00 26, 074102-1 

C 2014 AIP Publishing LLC

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
130.191.17.38 On: Sat, 08 Nov 2014 19:06:30
 074102-2 Toshihiko Hiejima Phys. Fluids 26, 074102 (2014)

For incompressible flows, several researchers have used linear stability analysis to investi-
gate the instability of Batchelor-type vortices. At large Reynolds numbers, linear instabilities are
dominated by inviscid negative helical modes2–4 for swirl numbers 0 < Φ < 2.3; weak viscous
unstable modes exist at all swirl numbers.5–8 In this regard, the growth rates of viscous modes
are much smaller than those of strong inviscid modes. Moreover, the introduction of a moder-
ate amount of swirl strongly promotes absolute instabilities,9, 10 such as azimuthal wave modes
m = −2 and −3 near Φ = 0.4 and 0.6, respectively. On the basis of the perturbation energy dis-
tribution in the radial direction,11 the instability modes of swirling flows are classified into three
types: core modes, potential modes, and freestream modes. Core modes include inviscid helical
modes and viscous modes. Potential and freestream modes require three velocity components in
the basic flow; both modes are involved with transient growth12 contributed from an inviscid con-
tinuous spectrum. For compressible flows, inviscid linear stability analysis13, 14 shows that vortex
instabilities are suppressed as the Mach number increases. To develop fluctuations in high-speed
flows, only large-growth modes are interesting; therefore, the viscous mode is disregarded in this
study.
The nonlinear evolution of incompressible Batchelor vortices on perturbations has been investi-
gated in a periodic temporal development.15–18 Ragab and Sreedhar15 first showed that relaminariza-
tion of vortices with axial flows occurs at high swirl. Delbende and Rossi16 have studied the effects
of random disturbances on stability within a two-dimensional framework of Batchelor vortices at
Re = 1000. They found that the dynamics strongly depends on the initial swirl intensity Φ. For
low swirl Φ  0.4, it was found that the evolution gives birth to an array of dipoles and causes
a drastic increase in structure size. Abid17 considered the temporal nonlinear mode selection of
the Batchelor vortex for |Φ| = 0.4, 0.8, and 1.0, and showed that the long-wave mode is selected
nonlinearly, which differs from the short-wave mode of the maximum exponential growth rate. In
this way, transient growth modes seem to make an appearance when normal unstable modes are
small. Di Pierro and Abid19 showed that when the growth rates of a Batchelor vortex are small
(Φ = 0.1 and 1.2), linear instabilities are not sufficient for breakdown to occur. Duraisamy and
Lele18 studied the temporal evolution of a low-swirl Batchelor vortex using direct numerical simu-
lations (DNS). They demonstrated that the excitation of normal mode instabilities is the source of
coherent large-scale helical structures inside the vortex core and leads to the generation of a polar-
ized vortex layer. The largest growth rate was present in the vicinity of Φ ≈ 0.5. It was resolved
numerically that the vortex layer breaks down into multiple hairpin-shaped structures. From previous
studies, incompressible Batchelor vortices are known to possess large instabilities for Φ = 0.4–0.6.
However, it is necessary to know why moderate swirl (Φ ≈ 0.5) contributes to instabilities and
breakdown.
Little work has been done on the spatial development of Batchelor vortices even for incom-
pressible flows, although a few papers have been published on the vortex breakdown using DNS.
Broadhurst and Sherwin20 confirmed the connection between spiral-type instabilities and vortex
breakdown in that an axial deceleration develops into a stagnation point. Mao and Sherwin11 also
found that spiral-type breakdowns can be triggered by helical instabilities. For compressible flows,
there are a few published papers on the nonlinear evolution of the Batchelor vortex. Herrada et al.21
solved the axisymmetric Navier-Stokes equations to study the vortex breakdown in pipes for Mach
numbers from 0.3 to 0.6. It was shown that the structures of vortex breakdown in compressible
flows are not different from those in incompressible flows. Khorrami and Chang22 studied the linear
and nonlinear evolution of disturbances in supersonic streamwise vortices using both linear stabil-
ity analysis and parabolized stability equations (PSE) for Φ ≤ 0.4 at Mach numbers 1.5–3.5 and
Re = 30 000. Coherent large-scale structures, as well as mean flow distortions in the axial velocity
and temperature profiles, appeared in spatial developments. They stated that nonlinear effects tend
to stabilize the vortex because of the increase in the swirl intensity.
For supersonic streamwise vortices, whether or not evolutions lead to restabilization should
depend on the circulation, the velocity deficit, and the Mach number. What seems to be lacking is
a classification of developing vortical structures based on swirl parameters. Even in terms of the
inviscid helical mode, i.e., the strongest mode, the effect of Mach number on nonlinear evolution
remains unknown. At high Mach numbers, as normal mode instabilities become relatively weak,14

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
130.191.17.38 On: Sat, 08 Nov 2014 19:06:30
 074102-3 Toshihiko Hiejima Phys. Fluids 26, 074102 (2014)

transient growth of instabilities may appear in the nonlinear region. Therefore, under various condi-
tions, it is also necessary to examine nonlinear evolutions using the full equations as well as PSE. In
this paper, spatial direct numerical simulations of Batchelor vortices with random disturbances are
used to address some of the problems cited above. Even when Reynolds numbers are large enough,
the simulations have no small significance during the initial development of perturbations because
inviscid modes are dominant.
The paper is organized as follows. Section II defines the basic flow of supersonic streamwise
vortices. An instability condition for the basic flow is introduced in Sec. III. In Sec. IV, numerical
formulations for direct numerical simulations are introduced along with computational conditions.
Section V provides the results and discussions for nonlinear development of vortices at various Mach
numbers. Finally, conclusions are presented in Sec. VI.

II. SUPERSONIC STREAMWISE VORTICES

The supersonic streamwise vortices studied here are assumed to be steady and axisymmetric.
In cylindrical polar coordinates (r, θ , x), ur , uθ , ux , ρ, p, T, and S represent the three components of
velocity plus density, pressure, temperature, and entropy, respectively. Supersonic experiments of
streamwise vortices show that measured velocity profiles conform to that of a Batchelor vortex.23
Therefore, the axial vorticity ωx∗ and the axial velocity u ∗x can be represented by
 ∗ 2
− ar ∗
ωx∗ (r ∗ ) = ωmax
∗
e , (1)

 2
 ∗ − r∗
u ∗x ∗
r = U∞ −  Ux∗ e a∗
. (2)

Hereafter, dimensional quantities appear with the superscript ∗ . In (1) and (2), ωmax
∗
denotes the
∗ ∗
maximum axial vorticity, U∞ is the freestream velocity,  Ux is the axial velocity deficit, and
a∗ is the initial vortex core size. A freestream Mach number M∞ and Reynolds number Re are
given by
∗ ∗ ∗ ∗
U∞ ρ∞ U∞ a
M∞ =  , Re = ∗
, (3)
γ R ∗ T∞
∗ η∞
∗
where T∞ is the freestream temperature, R∗ is the gas constant, γ 
is the ratio of specific heats, and
∗ ∗ ∗ ) and the density ρ ∗ ,
η∞ is the viscosity. Using the freestream speed of sound c∞ (= γ R ∗ T∞ ∞
normalized physical variables can be defined as follows:
ρ∗ u r∗ u ∗θ u ∗x p∗
ρ= ∗
, u r = ∗
, u θ = ∗
, u x = ∗
, p = ∗ c∗2
,
ρ∞ c∞ c∞ c∞ ρ∞ ∞

T∗ S∗ r∗ x∗ ∗
c∞
T = ∗
, S = , r = , x = , t = t ∗, (4)
γ T∞ Cv∗ a∗ a∗ a∗
where Cv∗ represents the specific heat at constant volume.
The azimuthal velocity uθ calculated from the axial vorticity and the axial velocity ux are derived
from (1) and (2), respectively,
 
1 − e−r ,
2
u θ (r ) = = 2πq M∞ ,
2πr
(5)
u x (r ) = M∞ (1 − μ e−r ),
2

where
∗
 Ux∗
q= ∗ a∗
, μ= ∗
. (6)
2π U∞ U∞

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
130.191.17.38 On: Sat, 08 Nov 2014 19:06:30
 074102-4 Toshihiko Hiejima Phys. Fluids 26, 074102 (2014)

Here, q and μ represent the circulation and the velocity deficit, respectively, and Φ indicates their
ratio14
q ωx
Φ= =r . (7)
μ (−ωθ )
The parameter Φ also represents the ratio of axial and azimuthal vorticities, ωx and ωθ , respectively,
and is an important similarity parameter for streamwise vortices, as will be shown later.
For inviscid steady flows, suppose ur = 0, so that the density ρ (r) and pressure p (r) depend
on r and can be determined from the following equations:
 
dp uθ 2 dρ ρ 1 dp d S
=ρ , = − . (8)
dr r dr γ p dr dr
The normalized equation of state is p = ρ T. Over wide temperature ranges, a perfect gas with
simple molecules has constant specific heats. The entropy gradient can be described by the Crocco
equation,
1 dS γ dT0
T = − u θ ωx + u x ωθ . (9)
γ − 1 dr γ − 1 dr
Several experiments show that the stagnation temperature T0 is approximately uniform.14 Vortices
with uniform T0 can be estimated from (9) using d T0 /dr = 0. In addition, an isentropic vortex for
dS/dr = 0 is also studied in this paper.

III. STABILITY CRITERIA FOR INVISCID COMPRESSIBLE SWIRLING FLOWS

According to Eckhoff and Storesletten24 and Leblanc and Le Duc,25 the flow is unstable if the
basic flow obeys the following inequality (10) for some r:
 2  2    2  2
u2 d S du x du θ uθ 2u θ du θ uθ du x du θ u 2θ
− θ + − + − + −
γ r dr dr dr r r dr r dr dr r2
     
2u θ 2 2u θ 2u θ u θ 2  u θ 2
= N 2
ωθ 2
+ ωx − + ωx − ωθ + ωx −
2
− < 0, (10)
r r r r r
 
I II

where N is the Brunt-Väisälä frequency, which is peculiar to compressible flows

u 2θ d S du x du θ uθ
N2 = − , ωθ = − , ωx = + . (11)
γ r dr dr dr r
For term I in (10), as from (5), the following inequality holds for some r:
 
2u θ
ωx − ≤ 0. (12)
r
Substituting (−ωθ ) = (r/Φ) ωx obtained from (7) into term II of (10), we obtain
      
u θ 2  u θ 2 r 2 2u θ
ωθ 2 + ωx − − = ωx ωx − − ωx . (13)
r r Φ r

When N2 = 0, i.e., when dS/dr = 0 (isentropic vortices), if (13) is positive, then

r
Φ< . (14)
2u θ
−1
r ωx

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
130.191.17.38 On: Sat, 08 Nov 2014 19:06:30
 074102-5 Toshihiko Hiejima Phys. Fluids 26, 074102 (2014)

The inequality (10) follows from (13) on the condition that Φ < Φmax ≈ 1.404. Thus, the suffi-
cient condition for instability is satisfied. Although the Rayleigh-Taylor instability is effective with
increasing q or Φ, the above suggests there is an upper limit for the swirl parameter.
When N2 > 0, i.e., when dS/dr < 0 (total temperature is uniform), the inequality (10) can be
expressed as

 r 2      
2u θ 2 2u θ 2u θ  r 2 2 2u θ
N2 ωx2 + ωx − + ωx − ωx + ωx ωx −
Φ r r r Φ r

  2  r 2   
2u θ 2u θ 2u θ 2u θ
= N2 + ωx ωx − + ωx2 N2 + ωx − < 0. (15)
r r Φ r r

III

If term III in (15) is negative, the entropy gradient can be obtained from the following
inequality:

   
dS 1 1 du θ
− < 2γ − . (16)
dr r u θ dr

To satisfy the inequality (15), not only must the inequality (16) hold, but Φ must also be small with
increasing N2 . The entropy gradient is expressed as a function of M∞ , q, and μ from (9). The flow is
stable when M∞ and Φ are large, specifically, (−dS/dr)max > 2.749. This prediction also coincides
with features from linear stability analysis.14
However, if dS/dr > 0 (N2 < 0) or ωx < 0, it is easy to satisfy the inequality (15) for instability.
The former is a heavy-core vortex26–29 and the latter is a Taylor-type vortex30, 31 with a negative axial
vorticity for some r.
In terms of instability types, Di Pierro and Abid32 demonstrate that the swirl instability can
be classified based on the dominant term in the growth rate (17) derived from an asymptotical
analysis. Their result includes a sufficient condition for the destabilization in (10). For the function
χ corresponding to the temporal growth rate,

 u  du
θ x
m 2 G 2 + α 2r 2 H 2 + 2 m α r
χ= r dr , (17)
m 2 + α 2r 2

where m and α denote azimuthal and axial wavenumbers.

If the first term G2  −N2 is dominant for χ , the flow provides a Rayleigh-Taylor instability; if the
second term H2 = −N2 − 2(uθ /r) ωx is dominant, the flow develops a Taylor-Couette instability; if
the third term (uθ /r)dux /dr is dominant, the flow forms an instability due to the coupling between
the rotation and axial shear.27

IV. NUMERICAL FORMULATIONS

A. Governing equations
The governing equations are the three-dimensional, unsteady, compressible Navier-Stokes equa-
tions in general coordinates ξ i (i = 1–3)

 
∂ Q ∂ Fi ∂ Fv i
+ = , (18)
∂t J ∂ξi ∂ξi

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
130.191.17.38 On: Sat, 08 Nov 2014 19:06:30
 074102-6 Toshihiko Hiejima Phys. Fluids 26, 074102 (2014)

⎡ ⎤ ⎡ ⎤
ρ ρ Ui
⎢ ⎥ ⎢ ⎥
⎢ ρ u1 ⎥ ⎢ ρ u 1 Ui + p (J −1 ∂ξi /∂ x1 ) ⎥
⎢ ⎥ ⎢ ⎥  
⎢ ⎥ ⎢ ⎥ −1 ∂ξi
Q = ⎢ ρ u2 ⎥ , Fi = ⎢
⎢ ⎥
⎢ ρ u 2 Ui
−1 ⎥
+ p (J ∂ξi /∂ x2 ) ⎥ , Ui = J uk , (19)
⎢ ⎥ ⎢ ⎥ ∂ xk
⎢ ρ u3 ⎥ ⎢ ρ u 3 Ui + p (J ∂ξi /∂ x3 ) ⎥
−1
⎣ ⎦ ⎣ ⎦
e (e + p) Ui
⎡ ⎤
0
⎢ ⎥
⎢ τ11 (J −1 ∂ξi /∂ x1 ) + τ12 (J −1 ∂ξi /∂ x2 ) + τ13 (J −1 ∂ξi /∂ x3 ) ⎥
⎢ ⎥
⎢ ⎥
⎢ τ21 (J −1 ∂ξi /∂ x1 ) + τ22 (J −1 ∂ξi /∂ x2 ) + τ23 (J −1 ∂ξi /∂ x3 ) ⎥
Fv i =⎢ ⎥,
⎢ ⎥
⎢ τ31 (J −1 ∂ξi /∂ x1 ) + τ32 (J −1 ∂ξi /∂ x2 ) + τ33 (J −1 ∂ξi /∂ x3 ) ⎥
⎢ ⎥
⎣ −1 −1 −1
⎦
β1 (J ∂ξi /∂ x1 ) + β2 (J ∂ξi /∂ x2 ) + β3 (J ∂ξi /∂ x3 )
   
−1 ∂ x1 ∂ x2 ∂ x3 ∂ x2 ∂ x3 ∂ x1 ∂ x2 ∂ x3 ∂ x2 ∂ x3
J = − + −
∂ξ1 ∂ξ2 ∂ξ3 ∂ξ3 ∂ξ2 ∂ξ2 ∂ξ3 ∂ξ1 ∂ξ1 ∂ξ3
 
∂ x1 ∂ x2 ∂ x3 ∂ x2 ∂ x3
+ − ,
∂ξ3 ∂ξ1 ∂ξ2 ∂ξ2 ∂ξ1
where Q is the vector of conservative variables, F i are convective fluxes, Fv i are viscous fluxes, Ui
are the velocity components at the cell interface, J is the Jacobian that transforms the coordinates
from physical space to computational space, J−1 corresponds to the cell volume in physical space
provided that all cell volumes in computational space equal unity, and J−1 ∂ξ i /∂xk are the derivatives
for the coordinate conversion, namely, the metrics
p 1
p =ρT, e= + ρ uk uk ,
γ −1 2
 
η(T ) ∂u i ∂u j 2 ∂u k
τi j = + − δi j , (20)
Re M ∂ x j ∂ xi 3 ∂ xk
γ η(T ) ∂ T
qi = − , βi = u i τi j + qi .
(γ − 1) Re M Pr ∂ xi
Here, e is the total energy, τ ij is the viscous stress tensor, qi is the conductive heat flux, and ui are the
velocity components in Cartesian coordinates. The Reynolds number based on the sonic velocity is
∗ ∗
defined by ReM = (ρ∞ c∞ a ∗ )/η∞ ∗
(= Re /M∞ ), and the Prandtl number Pr = 0.72. The viscosity η
is calculated from Sutherland’s law,
3 1 + Tb
η(T ) = T 2 , (21)
T + Tb
where Tb depends on the freestream Mach number.

B. Numerical methods and computing conditions

The convective flux terms are discretized using the numerical flux F i, s+ 12 and the vector of
primitive variables q as follows:
⎡ ⎤
ρ
⎢ ⎥
⎢ u1 ⎥
∂ Fi F i, s+ 1 (q) − F i, s− 1 (q) ⎢ ⎥
⎢ ⎥
≡ 2 2
, q = ⎢ u2 ⎥ . (22)
∂ξi ξi ⎢ ⎥
⎢u ⎥
⎣ 3⎦
p

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
130.191.17.38 On: Sat, 08 Nov 2014 19:06:30
 074102-7 Toshihiko Hiejima Phys. Fluids 26, 074102 (2014)

A modified numerical flux F i, s+ 12 is obtained by multiplying F i, s+ 12 in (22) by a transformation

matrix T,
⎡ ⎤
ρ Ûi
⎢ ⎥
⎢ ρ Û Ûi + p δ1i ⎥
⎢ ⎥
1 ⎢ ⎥
F i, s+ 12 = √ ⎢
T F i, s+ 12 = ⎢ ρ V̂ Ûi + p δ2i ⎥
⎥ , (23)
(m i )k · (m i )k ⎢ ⎥
⎢ ρ Ŵ Ûi + p δ3i ⎥
⎣ ⎦
ρ H Ûi i, s+ 12

∂ξi (m i ) j 1
(m i ) j = J −1 , (m̂ i ) j = √ , Ûi = √ Ui ,
∂x j (m i )k · (m i )k (m i )k · (m i )k
where (mi )j is a vector of metrics and (m̂ i ) j is the unit vector of (mi )j . The unit vectors n̂ j , and lˆj
represent orthogonal vectors perpendicular to (m̂ i ) j . For each index i, a metric vector, which is a
normal vector on the cell interface, is defined as follows:
when i = 1 , m̂ j = (m̂ 1 ) j , m̂ j · n̂ j = n̂ j · lˆj = lˆj · m̂ j = 0,

when i = 2 , n̂ j = (m̂ 2 ) j , n̂ j · lˆj = lˆj · m̂ j = m̂ j · n̂ j = 0,

when i = 3 , lˆj = (m̂ 3 ) j , lˆj · m̂ j = m̂ j · n̂ j = n̂ j · lˆj = 0,

e+ p
Û = u k m̂ k , V̂ = u k n̂ k , Ŵ = u k lˆk , H = , (24)
ρ
⎡ ⎤ ⎡ ⎤
1 0 0 0 0 1 0 0 00
⎢ ⎥ ⎢ ⎥
⎢ 0 m̂ 1 m̂ 2 m̂ 3 0 ⎥ ⎢ 0 m̂ 1 n̂ 1 lˆ1 0 ⎥
⎢ ⎥ ⎢ ⎥
⎢ ⎥ ⎢ ⎥
⎢
T = ⎢ 0 n̂ 1 n̂ 2 n̂ 3 0 ⎥ , T = ⎢ 0 m̂ 2 n̂ 2 lˆ2 0 ⎥
⎥ −1 ⎢
⎥.
⎢ ⎥ ⎢ ⎥
⎢ 0 lˆ1 lˆ2 lˆ3 0 ⎥ ⎢ 0 m̂ 3 n̂ 3 lˆ3 0 ⎥
⎣ ⎦ ⎣ ⎦
0 0 0 0 1 0 0 0 01
The modified numerical flux in (23) is evaluated using the AUSMDV scheme.33 The scheme is
classified into the AUSM approach and is simple as well as highly accurate. Therefore, the numerical
flux is evaluated by substituting (25) into (22),

F i, s+ 12 = (m i )k · (m i )k T −1 F i, s+ 12 . (25)
Since this study intends to simulate, with satisfactory accuracy, the development of disturbances
for streamwise vortices, the primitive variables q maintain high-order spatial accuracy by using the
weighted interpolation at cell interfaces by Deng and Zhang;34 these are based on adaptive stencil
interpolations in combination with the AUSMDV scheme. The interpolated values are fifth-order
accurate in smooth regions and can capture large gradients, such as shock waves, without degrading
the precision of space differences. Unlike limiter functions, this non-oscillation scheme uses no
parameters for avoiding numerical oscillations. Although limiter functions are often used to capture
a discontinuous surface, such as a shock, their use degrades the accuracy on turning values for the
azimuthal velocities in a vortex, even in continuous regions. The viscous flux terms are calculated
with central differences to fourth-order accuracy. The temporal integration method is the four-step,
fourth-order accurate scheme.35 Since turbulence models have harmful effects on the static pressure
at the vortex center and on the axial velocity deficit, which are important in this study, direct
numerical simulations were conducted.

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
130.191.17.38 On: Sat, 08 Nov 2014 19:06:30
 074102-8 Toshihiko Hiejima Phys. Fluids 26, 074102 (2014)

FIG. 1. Computational domain in physical space.

LST(X = 0)
Numerical values
LST(X = 40)
LST: - αi x

Eigenfunction distributions
5
LST(X = 80)
15 10 X= 0
X = 40
)
x=0

3 X = 80
10
10
ln( |u~x| / |u~x|

1
10
5 -1
10

-3
0 10
0 100 200 0 1 2 3
x r

FIG. 2. Evolution of an unstable mode m = −2 for M∞ = 2.5, Φ = 0.32 (q = 0.16, μ = 0.5). (left) Growth rates and (right)
amplitude profiles of perturbations in axial velocity obtained from linear stability analysis14 (LST) and direct numerical
simulation (DNS).

The physical domain consisted of a rectangular box having sides Lx = 300, and Ly = Lz = 40,
with flow in the X-direction. The spatial resolution was Nx = 901, Ny = Nz = 171. (Note that
Lx = 500, and Nx = 1126 for M∞ = 7.5 and 9.0; Ly = Lz = 50, for the case q ≥ 0.32.) The grid
spacing was uniform within three times the diameter of the vortex core centered on the axis. (The
radius of the vortex core was close to the maximum azimuthal velocity.) On the outer side of the
region, the grid splayed out at irregular intervals. The computational space was also a rectangular
box that was composed of grids of uniform width ξ i = 1. As shown in Fig. 1, the inflow conditions
consisted of the supersonic streamwise vortices described in Sec. II with the addition of random
noise, given that the maximum amplitude of the noise was taken to be 5% of the freestream variables.
Since the flows were supersonic in any calculations at x = Lx , the outflow condition was extrapolated
at zeroth-order. Symmetry conditions were applied on the boundary surfaces in the Y-Z cross section.
The simulations were performed for M∞ = 2.5–9.0 and ReΓ = Re M · = 1.41× 105 q.
To validate the numerical procedure, we calculated the growth rate of flow perturbations obtained
from linear stability analysis. Figure 2 shows the evolution of an unstable mode m = −2 for
M∞ = 2.5, Φ = 0.32 (q = 0.16, μ = 0.5) in the axial velocity perturbation ux . The spatial growth
rate was evaluated by integrating the perturbation profile over a radius r. In the linear region,
the growth rate obtained from the numerical simulation was close to the spatial growth rate, −α i
= 0.1269, that was obtained from linear stability analysis.14

V. RESULTS AND DISCUSSION

This section describes the results for spatially developing streamwise vortices with small random
perturbations. To visualize vortex structures quantitatively, it is useful to have a formal definition of
such structures. The second invariant of the velocity gradient tensor Q is very sensitive to vortex
structures and can be used to identify vortical structures at early stages in their evolution,36
1  
Q= −Si j Si j + Ri j Ri j + C 2 , (26)
2

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
130.191.17.38 On: Sat, 08 Nov 2014 19:06:30
 074102-9 Toshihiko Hiejima Phys. Fluids 26, 074102 (2014)

   
1 ∂u j ∂u i 1 ∂u j ∂u i ∂u k
Si j = + , Ri j = − ,C = ,
2 ∂ xi ∂x j 2 ∂ xi ∂x j ∂ xk
where Si j and Ri j are the strain-rate and vorticity tensors, respectively; they are the symmetric and
asymmetric components of the velocity gradient tensor ∂ui /∂xj , and C is the divergence of velocity
vectors.
An unstable axial wavenumber α (axial wavelength λα ) is usually used as a guide for instabilities
of axial flows, such as jets and wakes, and an unstable azimuthal mode (azimuthal wavenumber) m
is an important factor that appears in the development of streamwise vortices. According to results
from linear stability analysis of vortices,14 the most unstable mode has a relation between the angular
frequency ω and the wavenumbers; ω ∝ − m and ω ∝ α. That is, α is proportional to −m. Therefore,
we can monitor the spatial properties of vortices by focusing on the integer mode of the azimuthal
wavenumber involved in the development of spiral structures. Let us represent any physical variable
by φ(x, y, z). To analyze azimuthal modes, the coordinates of such variables are transformed from
φ(x, y, z) to φ(x, r, θ ). Fluctuations in the flow are decomposed using a Fourier transform for m,

1 2π
m (x, r ) = Fm [φ] =
φ φ(x, r, θ )e imθ dθ, (27)
2π 0
where φm (x, r ) is the fluctuation of φ(x, r, θ ) on the azimuthal mode.
In terms of a property of streamwise vortex breakdown, the enstrophy provides useful
information.19 The enstrophy ε is given by

1
ε= ωi ωi d A, (28)
A 2

where ωi is a vector of vorticity, and A is the cross-sectional area in a plane perpendicular to the
axial flow.

A. Relation between swirl parameter and unstable modes

Figure 3 shows the isosurface of the second invariant of the velocity gradient tensor and the
contour lines of axial vorticity (Y-Z cross sections) for streamwise vortices with a uniform stagnation
temperature at M∞ = 2.5 and 5.0. In each case, unstable growth modes have developed from each
vortex at downstream locations. The vortical structures during development differ depending on the
parameter Φ. At M∞ = 2.5, the effect of the parameter Φ has been examined. When Φ = 0.16 (Fig.
3(a)), the vortex axis spirals on itself, causing growth of mode |m| = 1, which is the bending mode.
This structure is similar to the sketch described by Chanaud as a spiral-type vortex breakdown.37 It
is commonly observed in trailing-line vortices with weak swirling flows at subsonic speeds. When
Φ = 0.32 (Fig. 3(b)), although the growth mode has a long wavelength and includes the bending
mode, the vortex axis remains in the vicinity of the center. When Φ = 0.64 (Fig. 3(c)), the vortex
forms a weak spiral instability and becomes normal again. The axial vorticity near the axis strengthens
where the spiral mode appears outside the vortex core; this is due to circulation conservation. These
features strongly suggest relaminarization. (Even if a flow is initially unstable, there is no reason to
believe that it will remain unstable.30 ) In incompressible flows, Ragab and Sreedhar15 also reported
from time-evolution results that large circulation vortices with axial flow exhibit relaminarization;
in fact, they were repeatedly unstable and then stable. In spatially developing supersonic streamwise
vortices, this study first demonstrated relaminarization. Next, to investigate the effects of the Mach
number, we compare M∞ = 2.5 with M∞ = 5.0 at Φ = 0.32. From the visualization of Q, the
developing wavelength for M∞ = 5.0 differs from that for M∞ = 2.5. The Y-Z cross sections also
show that both are significantly different in the developing mode outside the vortex core. When M∞
= 5.0, mode |m| = 2 first develops in the core, after which the mode |m| = 3 appears. Additionally,
based on the visual appearance of large-scale structures, the spatial growth rate of the vortex decreases
with increasing Mach number. Therefore, the results show that the flow is stabilized by increasing
M∞ and Φ, which means that (−dS/dr|max ) becomes larger. This tendency is consistent with that
from the linear stability analysis.14

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
130.191.17.38 On: Sat, 08 Nov 2014 19:06:30
 074102-10 Toshihiko Hiejima Phys. Fluids 26, 074102 (2014)

(a) M∞ = 2.5, Φ = 0.16; q = 0.08, μ = 0.5 (c) M∞ = 2.5, Φ = 0.64; q = 0.32, μ = 0.5

(b) M∞ = 2.5, Φ = 0.32; q = 0.16, μ = 0.5 (d) M∞ = 5.0, Φ = 0.32; q = 0.16, μ = 0.5

X =0 X = 147 X = 159 X = 175 X = 200 X = 225

(a)

X =0 X = 80 X = 100 X = 140 X = 180 X = 220

(b)

X =0 X = 147 X = 159 X = 175 X = 200 X = 250

(d)

FIG. 3. Supersonic streamwise vortices with uniform stagnation temperature: (top) isosurface of the second invariant of
the velocity gradient tensor Q, (bottom) contour lines of axial vorticity ωx ; for M∞ = 2.5, (a) Φ = 0.16, (b) Φ = 0.32,
(c) Φ = 0.64; and for M∞ = 5.0, (d) Φ = 0.32.

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
130.191.17.38 On: Sat, 08 Nov 2014 19:06:30
 074102-11 Toshihiko Hiejima Phys. Fluids 26, 074102 (2014)

1.5 1.5 3

0
Mean Entropy ΔS = S - S
Mean Axial Velocity Ux
X= 0
X = 100

Mean Circulation Γ
2 X = 200
1.0 1.0 X = 300
X= 0
X= 0
X = 100 1
X = 100
0.5 X = 200 0.5
X = 200
X = 300
X = 300
0
0 0
0 1 2 3 4 0 1 2 3 4 5 6 0 1 2 3 4
r r r
(a)

1.5 1.5 3

0
Mean Entropy ΔS = S - S
Mean Axial Velocity Ux

X= 0
X = 100

Mean Circulation Γ
2 X = 200
1.0 1.0 X = 300
X= 0
X= 0
X = 100 1
X = 100
0.5 X = 200 0.5
X = 200
X = 300 X = 300
0
0 0
0 1 2 3 4 0 1 2 3 4 5 6 7 8 0 1 2 3 4
r r r
(b)

FIG. 4. (a) Φ = 0.64, M∞ = 2.5 and (b) Φ = 0.16, M∞ = 2.5: Mean profiles of (left) axial velocities, (middle) circulations,
and (right) entropies, at X = 0, 100, 200, and 300.

As described above, a transition occurs in vortices with uniform stagnation temperatures, which
have positive entropy profiles with convex distributions. To investigate causes for the transitions, it is
useful to monitor profiles of the flows because instabilities in supersonic streamwise vortices depend
mainly on three profiles: axial velocity, circulation, and entropy. Figure 4 shows mean profiles of
ux , /M∞ , and S for Φ = 0.16 and 0.64 at X = 0, 100, 200, and 300. For Φ = 0.64, the mean
profiles do not cause a major change in every streamwise position and seem to be stable on the
surface. However, Φ = 0.64 is characterized by relaminarization. This is partly because the initial
entropy on the axis is large; the positive entropy stabilizes the vortex. Delbende and Rossi16 showed
that the stabilizing effect leads incompressible vortices to relaminarize in the nonlinear regime to
the extent of 1 < Φ < 1.5. Hiejima14 demonstrated that supersonic vortices are stabilized by the
magnitude of a negative entropy gradient (−dS/dr) strengthened with increasing M∞ . Thus, it is
found that the vortex at M∞ = 2.5 has a low threshold for relaminarization. In contrast, the profiles
for Φ = 0.16 have some features of instability: (i) the axial velocity deficit weakens with distance,
(ii) while angular momentum is maintained, a circulation overshoot appears at X = 200 and the
overshoot is transferred outward at X = 300, and (iii) the entropy gradient approaches zero and
becomes unstable. The recovery of the axial velocity deficit arises from development of turbulence.
If turbulent diffusion is faster than laminar viscous diffusion, the vortex must develop an overshoot
in circulation.38 The circulation overshoot is associated with a centrifugal instability because the
circulation decays monotonically in the radial direction, i.e., d /dr < 0 for some r. Moreover,
outward propagation of the overshoot is a diffusion process by which angular momentum may
be transported outward.39 For vortices with uniform stagnation temperatures, flow instabilities are
influenced by centrifugal instabilities. This coincides with the asymptotical result32 in which the
growth rate is dominated by the third term (uθ /r)dux /dr in (17).
To elucidate spatial growth, dominant modes in fluctuation components were resolved by
Fourier transforms (27). Figure 5 shows streamwise variables of the perturbation amplitude of the
axial vorticity |x | normalized by the value at X = 0, for |m| = 1–8 and three values of Φ. The
vertical axes contain a measure of spatial growth rate, −α i = ln (|x |/|x |X=0 ). When Φ = 0.16,
only the mode |m| = 1 grows early, as shown in Fig. 5(a). When Φ = 0.32 (Fig. 5(b)), although |m|
= 1 is the most unstable, the other modes develop similarly. In Figs. 5(a) and 5(b), the |m| = 1 mode
is initially dominant. In both cases, the maximum growth rates are in agreement with results from

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
130.191.17.38 On: Sat, 08 Nov 2014 19:06:30
 074102-12 Toshihiko Hiejima Phys. Fluids 26, 074102 (2014)

|m| = 1 |m| = 5 |m| = 1 |m| = 5 |m| = 1 |m| = 5

q = 0.08 |m| = 2 |m| = 6 q = 0.16 |m| = 2 |m| = 6 q = 0.32
μ = 0.5 |m| = 3 |m| = 7 μ = 0.5 |m| = 3 |m| = 7 μ = 0.5 |m| = 2 |m| = 6
15 15 15 |m| = 3 |m| = 7
|m| = 4 |m| = 8 |m| = 4 |m| = 8
M = 2.5: M = 2.5: M = 2.5: |m| = 4 |m| = 8

8
8
)

)
T = const T = const T = const
xx=0

xx=0

xx=0
0 0 0

10 10 10
ln( |Ω | / |Ω |

ln( |Ω | / |Ω |

ln( |Ω | / |Ω |
x

x
5 5 5

0 0 0
0 100 200 300 0 100 200 300 0 100 200 300 400
X X X
(a) (b) (c)

FIG. 5. Streamwise variations in growth properties −α i x = ln (|x |/|x |x=0 ), for azimuthal wavenumbers |m| = −1 to −8
obtained from direct numerical simulations with (a) Φ = 0.16, (b) Φ = 0.32, and (c) Φ = 0.64. Straight lines in (a) and (b)
are from linear stability analysis.14

inviscid linear stability analysis,14 −α i = 0.0914 and 0.140. For an incompressible Batchelor vortex
at Φ = 0.5, Duraisamy and Lele18 also showed that the initial growth rate matched the results from
the normal mode stability analysis. For Φ = 0.64, there is no clear linear region and the plots in
Fig. 5(c) indicate a spike-like profile downstream where the sizes of fluctuations seem to increase
and decrease. This is the same behavior as relaminarization processes observed in incompressible
high swirls.
Circulation overshoot is related to the negative vorticity that is generated by the development
of streamwise vortices.16 Positive and negative circulations are defined by ( = + + − )
 
−
= ωx d A− , + = ωx d A+ . (29)
A− (ωx < 0) A+ (ωx > 0)

Figure 6(a) shows the ratio of positive to negative circulations − / + as a function of the streamwise
variation for M∞ = 2.5 and 3.3. The transition to small perturbations on streamwise vortices occurs
at shorter distances than wake flows without swirl components. (In Fig. 6(a), wake flows do not
appear because wake transitions start at approximately X > 450 for M∞ = 2.5.) For Φ = 0.16,
the generated negative circulation is larger than that for the others. For Φ = 0.48 and 0.64, both
production rates are small. Although the transition for Φ = 0.32 is faster than that for Φ = 0.16,
the production is lower. The transition position should depend on the linear growth rate. In short, it
follows that generation of negative circulation decreases with increasing Φ. Figure 6(b) shows the
enstrophy for various Φ. The presence or absence of a breakdown depends on Φ. For Φ = 0.16 and
0.32, a breakdown occurs in the vicinity of the location where the enstrophy is a local maximum, as
shown in Figs. 3(a) and 3(b). This is similar to the result addressed in the temporal evolution.19
Generated negative vorticity is also important for maintaining the turbulent state. Although
Figs. 3(b) and 3(d) have negative circulations after transitions, they show signs of relaminarization

M = 2.5, Φ =0.16: T = const. M = 2.5, Φ =0.16: T = const.

8

0 0
0.8 M = 2.5, Φ =0.32: T = const. M = 2.5, Φ =0.32: T = const.
8

0 0
M = 3.3, Φ =0.48: T = const. 2.0 M = 3.3, Φ =0.48: T = const.
8

0 0
M = 2.5, Φ =0.64: T = const. M = 2.5, Φ =0.64: T = const.
0.6
8

0 0
X=0
+
Γ /Γ

ε/ε

0.4
−

1.0

0.2

0 0
0 100 200 300 0 100 200 300
X X
(a) (b)

FIG. 6. Spatial development of (a) the circulation ratio −/ + and (b) the enstrophy ε/εX = 0 for various values of the swirl
parameter Φ.

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
130.191.17.38 On: Sat, 08 Nov 2014 19:06:30
 074102-13 Toshihiko Hiejima Phys. Fluids 26, 074102 (2014)

10
Disturbance
8 development

Mach number M
Laminar cases:
7.5 under X < 500.
M = 1 /q

8
5.0

2.5

0
0 0.1 0.2 0.3 0.4 0.5
q

FIG. 7. Relation between circulation and Mach number for development of disturbances in uniform stagnation temperature
vortices.

in the vicinity of X = 300. So, we considered the relation between relaminarization and the generated
negative circulation. We found that relaminarizations did not occur when − / + was greater than
approximately 0.2. Under these conditions, fluctuations might be sustained downstream.
As mentioned above, this study found that large values of circulation gradually dampen all
unstable modes. In general, the effects of high swirl vortices on stabilization have not been understood
in compressible flows. Accordingly, we first take into account the compressibility effect in the
azimuthal direction. To estimate azimuthal compressibilities, we define the azimuthal Mach number
Mθ derived from u ∗θ = ∗ /(2π a ∗ ) as
u ∗θ
Mθ = ∗
= q · M∞ < 1. (30)
c∞
Figure 7 shows the relation between M∞ and q obtained from the numerical results for μ = 0.5.
When Mθ > 1, the breakdown of streamwise vortices is unidentified at X < 500. This suggests that
amplification of unstable waves based on an azimuthal shear instability is suppressed because of
the compressibility in the azimuthal direction. (A shear-type instability is significantly affected by
the compressibility effect.) Therefore, the growth of streamwise vortices with uniform stagnation
temperature can be summarized in the simple formula (30). The present study agrees with the
literature14 in that growth rates of streamwise vortices with large circulations significantly decrease
at high Mach number.

B. Effects of isentropic vortices on developing vortical structures

We also consider isentropic vortices, which are often used as a theoretical model for compressible
vortices. Isentropic vortices satisfy the condition for instability in the range of Φ given by the
inequality (14). Results from linear stability analysis show that isentropic vortices have large growth
rates in supersonic flows.14
Figure 8 shows the isosurface of the second invariant of the velocity gradient tensor and contour
lines of axial vorticity in Y-Z cross sections for isentropic streamwise vortices with Φ = 0.32
(q = 0.16, μ = 0.5) at M∞ = 2.5 and 5.0. For M∞ = 2.5, the developing vortical structure clearly
differs from that of the uniform stagnation temperature in Fig. 3(b). The large helical vortex does not
appear, but a large-scale dipole grows within the vortex core. Amazingly, the downstream breakdown
may be due to a short-wavelength instability. For M∞ = 5.0, the instability resulting from the core
has a notable impact. From the contour lines at X = 145 and 155, high-wavenumber modes are
generated within the core, in spite of the high Mach number. The developing large-scale structures
are quite different from those in Fig. 3(d), and breakup of the streamwise vortex definitely occurs.
In addition, small-scale structures are generated as a result of secondary instabilities40 between the
large-scale structures. Note that it is essential for secondary instabilities to maintain fluctuations in
transient processes. As a characteristic of supersonic vortex breakdown, it is remarkable that both
breakdowns occurred without a stagnation point because subsonic regions were not present. Needless
to say, they are generated under the condition that perturbations cannot propagate upstream. So, we
ask why this type of vortex collapses and prevents the disturbance structure from relaminarizing.
One reason is that the dominant phenomenon is independent of the mainstream direction. In other

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
130.191.17.38 On: Sat, 08 Nov 2014 19:06:30
 074102-14 Toshihiko Hiejima Phys. Fluids 26, 074102 (2014)

(a) M∞ = 2.5, Φ = 0.32; q = 0.16, μ = 0.5, Δ S = 0 (b) M∞ = 5.0, Φ = 0.32; q = 0.16, μ = 0.5, Δ S = 0

X =0 X = 120 X = 130 X = 150 X = 170 X = 250

(a)

X =0 X = 145 X = 155 X = 175 X = 200 X = 250

(b)

FIG. 8. Supersonic streamwise vortices with constant entropy: (top) isosurface of the second invariant of the velocity gradient
tensor Q and (bottom) contour lines of axial vorticity ωx .

words, the behavior is based essentially on instabilities in a plane vertical to the axial flow28 and,
therefore, does not need an axial velocity deficit. The generating mechanism of the instability is
analogous to a Rayleigh-Taylor instability,29 which is caused by the density gradient and centrifugal
force, as shown in Fig. 9. Therefore, isentropic vortices seem to be insensitive to the dominant
compressibility effect in the streamwise direction, that is, to M∞ , because the instabilities, like the
ruffled shape in Fig. 9, arise in a plane perpendicular to the main stream.

FIG. 9. Schematic diagram of the Rayleigh-Taylor instability due to an entropy profile.

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
130.191.17.38 On: Sat, 08 Nov 2014 19:06:30
 074102-15 Toshihiko Hiejima Phys. Fluids 26, 074102 (2014)

M = 2.5, Φ =0.32: ΔS = 0 M = 2.5, Φ =0.32: ΔS = 0

8
0.8 M = 5.0, Φ =0.32: ΔS = 0 6 M = 5.0, Φ =0.32: ΔS = 0

8
M = 2.5, Φ =0.32: T = const. M = 2.5, Φ =0.32: T = const.
0

8
0
M = 5.0, Φ =0.32: T = const. 5 M = 5.0, Φ =0.32: T = const.
0

8
0.6 0

X=0
4

+
Γ /Γ

ε/ε
0.4 3

−
2
0.2
1
0 0
0 100 200 300 0 100 200 300
X X
(a) (b)

FIG. 10. Spatial development of (a) the circulation ratio −/ + and (b) the enstrophy ε/εX = 0 for various M∞ and Φ.

Figure 10 shows the negative circulation and the enstrophy produced in vortices with constant
S and T0 for Φ = 0.32. In Fig. 10(a), the negative circulations of isentropic vortices are larger than
those of the vortices with constant T0 . Since angular momentum for the vortices is conserved at high
Reynolds numbers, + must increase as − becomes large; this is overshoot production and is tied
to centrifugal instabilities. Thus, it follows that the former is highly unstable, compared to the latter.
Comparing the isentropic vortices at M∞ = 2.5 and 5.0 shows that the two have different transition
points, but they are very similar in the production of negative circulation. This also suggests that
instabilities of isentropic vortices are essentially independent of the freestream Mach number M∞ .
The cases in which the enstrophy profile changes from increasing to decreasing may be associated
with vortex breakdown in the nonlinear regime. The enstrophy of the isentropic vortices in Fig. 10(b)
is significantly large. This feature also indicates the strong instability effect of isentropic vortices.
The position of the breakdown that corresponds to the vortical structures in Figs. 8(a) and 8(b)
is approximately similar to that of the maximum enstrophy. Therefore, these results suggest that
breakdown occurs where the enstrophy is close to a maximum.
Now, we focus on differences between vortices of constant S and T0 for Φ = 0.32 at M∞ = 5.0.
As shown in Figs. 3(d) and 8(b), these differ in the degree of breakdown on streamwise vortices.
To view vortical structures within the core, contour lines of axial vorticity in X-Z cross sections
through the axis are shown in Fig. 11. When T0 = const (Fig. 11(a)), instability waves leisurely
develop within the core and form a hairpin-like structure with downstream regularity outside the
core. When S = const (Fig. 11(b)), instability waves grow drastically over short distances; they have
high spatial growth rates, and the developing random structures indicate breakup of the streamwise
vortex. Moreover, the vortical region in Fig. 11(b) spreads outward in the radial direction, compared
to that in Fig. 11(a). Mean profiles of ux , /M∞ , and S for Φ = 0.32 at X = 0, 100, 200, and 300

(a) M∞ = 5.0, Φ = 0.32; q = 0.16, μ = 0.5, T0 = const

(b) M∞ = 5.0, Φ = 0.32; q = 0.16, μ = 0.5, Δ S = 0

FIG. 11. Contour lines of axial vorticity ωx for supersonic streamwise vortices at M∞ = 5.0.

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
130.191.17.38 On: Sat, 08 Nov 2014 19:06:30
 074102-16 Toshihiko Hiejima Phys. Fluids 26, 074102 (2014)

1.5 1.5 3

0
Mean Entropy ΔS = S - S
Mean Axial Velocity Ux
X= 0
X = 100

Mean Circulation Γ
2 X = 200
1.0 1.0 X = 300
X= 0
X= 0
X = 100 1
X = 100
0.5 X = 200 0.5
X = 200
X = 300 X = 300
0
0 0
0 1 2 3 4 0 1 2 3 4 5 6 0 1 2 3 4
r r r
(a)

1.5 1.5 3

0
Mean Entropy ΔS = S - S
Mean Axial Velocity Ux

X= 0
X = 100

Mean Circulation Γ
2 X = 200
1.0 1.0 X = 300
X= 0
X= 0
X = 100 1
X = 100
0.5 X = 200 0.5
X = 200
X = 300 X = 300
0
0 0
0 1 2 3 4 0 1 2 3 4 5 6 0 1 2 3 4
r r r
(b)

FIG. 12. (a) T0 = const: uniform stagnation temperature and (b) S = 0: isentropic vortex: Mean profiles of (left) axial
velocities, (middle) circulations, and (right) entropies for M∞ = 5.0 and Φ = 0.32, at X = 0, 100, 200, and 300.

are shown in Fig. 12. For incompressible vortices having monotonically increasing circulation, the
axial velocity deficit is the only source of instabilities.15 The deficit is also an important indicator
of instability in spatially developing compressible flows. In the vortices of both constant T0 and
constant S, the recovery rates of the velocity deficits are in the same range, while the circulation
profiles are quite different. When T0 = const, the generated circulation overshoot is weak and the
negative entropy gradient in the core is still large at X = 200. In this case, the overshoot decreases at
X = 300. It is known that, although the presence of a circulation overshoot implies that the flow is
centrifugally unstable, instabilities are not always sustained in incompressible flows.18 Note that the
difference in incompressible and compressible flow is in the entropy profiles. In the compressible
vorticity equation, the baroclinic torque term, ∇S × ∇T, is famous as the cause of eddy-production.
However, the baroclinic torque term for T0 = const does not contribute to vorticity generation;
rather, it is involved in stabilization. In contrast, the factor for strong instabilities of an isentropic
vortex is clearly the entropy profile, which causes d /dr < 0. To investigate the dominant unstable
mode, azimuthal modes in the flow were resolved using Fourier transforms (27). Figure 13 shows
the spatial growth rate of the axial vorticity perturbation and the resolved perturbation profiles in
the radial direction at two spatial positions. Each straight line in Figs. 13(a) and 13(b) is the linear
growth rate obtained from linear stability analysis, −α i = 0.0900 (m = −2) and 0.1545 (m = −6).
When T0 = const, the multimode is initially competitive until at last the |m| = 2 and |m| = 3 modes
become dominant. Although it coincides with the contour lines in Fig. 3(d), the linear unstable mode
is not visible. In contrast, when S = const, the most dangerous mode is |m| = 6, which is consistent
with the growth rate obtained from linear stability analysis. A high-wavenumber mode, such as |m|
= 7, is also present in the fluctuations. Remarkably, the maximum perturbation amplitude leads to
the exponent 12 times the initial amplitude; that is, amplification becomes very large. Accordingly,
for instabilities, isentropic vortices are superior in that the high-wavenumber modes can develop,
compared with vortices for constant T0 where the growth of high-wavenumbers is suppressed at high
Mach numbers.
For an isentropic vortex, these results demonstrate that, at high Mach numbers, strong spiral
waves with high azimuthal wavenumbers support destabilization and breakup of the streamwise
vortex into small-scale eddies. We find that entropy instabilities can give supersonic streamwise
vortices the advantage of being independent of M∞ . (The instabilities develop on a plane vertical

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
130.191.17.38 On: Sat, 08 Nov 2014 19:06:30
 074102-17 Toshihiko Hiejima Phys. Fluids 26, 074102 (2014)

FIG. 13. Streamwise variations in growth properties, −α i x = ln (|x |/|x |x=0 ), Straight–lines obtained from linear stability
analysis.14 Curves are results from mode analysis for various azimuthal wave numbers obtained from direct numerical
simulations for amplitudes of axial vorticity disturbance |x | at M∞ = 5.0 and Φ = 0.32. (a) Uniform stagnation temperature
vortices and (b) isentropic vortices.

to the mainstream.) Accordingly, this type of vortex may be effective in promoting disruptions of
large-scale structures in supersonic flows.

C. Properties of instabilities due to entropy profiles

In swirling flows, Batchelor vortices with variable densities have been addressed by Coquart
et al.,26 Di Pierro and Abid,27 and Joly et al.28 The patterns of vortex evolution are interesting because
they involve multiple instabilities (Kelvin-Helmholtz, Rayleigh-Taylor, and centrifugal instabilities,
as pointed out by Sipp et al.29 ). Assuming that the density profile is correlated with the axial velocity,
they define a density profile by
2
ρ(r ) = 1 + (s − 1)e−( b ) ,
r
(31)
where s is the density ratio between the vortex center and the radial point at infinity and b is the
width of the internal core.27
In terms of instabilities, this study points out that the effects of density variations are more
generally summarized by the entropy profile described here. Figure 14 shows results with the
reference conditions26, 27 replaced by entropy profiles. Since the Reynolds numbers26, 27 are large
enough at Re = 10 000 and 1000, the pressure profiles can be calculated by
 ∞ 2
u
p(r ) = p∞ − ρ θ dξ. (32)
r ξ
Therefore, using (31) and (32), we can obtain the entropy profiles from the literature,26, 27
p
S ≡ S − S0 = ln γ , (33)
ρ
where S is the entropy variation and S0 is constant. When s < 1, the entropy profiles form a convex
distribution. This is stable for large Φ conditions, though it depends on the inequality (27). Moreover,
when s = 1, the entropy profile corresponds to normal incompressible flows (low-speed flows) and

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
130.191.17.38 On: Sat, 08 Nov 2014 19:06:30
 074102-18 Toshihiko Hiejima Phys. Fluids 26, 074102 (2014)

Entropy
s = 1.2, b profiles
= 0.3 (Coquart et al. 2005)
s = 2.0, b = 1.0 (Dipierro et al. 2012)

0
Mean Entropy ΔS = S - S
s = 0.5, b = 1.0 (Dipierro et al. 2012)
1 s = 1.0, b = 1.0
( ρ (r) =1; Incompressible case)

-1 ρ (r) = 1 + (s-1)exp(-(r/b)2)

0 1 2 3 4
r

FIG. 14. Entropy profiles for the incompressible Batchelor vortices with density profiles.26, 27

is nearly isentropic. When s > 1, the entropy gradient is positive and always satisfies the unstable
condition (27). In fact, the flow is unstable if and only if the entropy increases somewhere, dS/dr >
0. (This is the same condition on instability that arises if the density decreases somewhere, dρ/dr <
0.) In transition processes, this study found that vortices for dS/dr ≥ 0 have the same nature as high-
density vortices subjected to the Rayleigh-Taylor instability.28, 29 Thus, from an entropy perspective,
we can explain that the unstable mechanism is due to an entropy instability like the Rayleigh-Taylor
instability. It is significant that this concept can be universally applied to all compressible vortices
including supersonic flows.

VI. CONCLUSIONS
Spatially developing supersonic streamwise vortices have been numerically studied. Since insta-
bilities of streamwise vortices depend on profiles across the flow, it is important to clarify the response
to small random perturbations in linear and nonlinear states. Batchelor-type vortices with uniform
stagnation temperatures and constant entropies (homentropies) were assumed for basic upstream
flows. For uniform stagnation temperature vortices, this study found that occurrence of transitions
depends on the azimuthal Mach number, which involves both the circulation and freestream Mach
numbers. Moreover, when the swirl parameter Φ is small, the vortex axis spirals on itself with a
circulation overshoot, and when Φ is larger than approximately 0.5, the development leads to relami-
narization. For isentropic vortices, strong instabilities occur in which high wavenumber modes grow
inside the vortex core; this behavior differs from that in uniform stagnation temperature vortices.
Instabilities arise on a plane vertical to the streamwise direction and are similar to Rayleigh-Taylor
instabilities. Such instabilities are essentially ascribed to instabilities in the entropy. In addition,
this study demonstrated that instabilities due to entropy can also account for comparable studies on
variable-density incompressible Batchelor vortices, such as heavy-core vortices.26–29
Therefore, by controlling the entropy profiles, it is possible to promote supersonic growth of
fluctuations, which is known to be difficult at high Mach numbers. We find that vortices with
unstable entropy can generate breakdowns with only limited influence from the compressibility
because instabilities develop and are sustained independently of M∞ in surfaces perpendicular to
the streamwise direction.

ACKNOWLEDGMENTS
The author expresses his gratitude to the anonymous referees for helping to improve the paper
and thanks Emeritus Professor M. Nishioka for valuable discussions. This work was partly supported
by a Grant-in-Aid for Scientific Research (No. 20560739, 21226020) from the Ministry of Education,
Culture, Sports, Science and Technology, Japan.
1 G.
K. Batchelor, “Axial flow in trailing line vortices,” J. Fluid Mech. 20, 645–658 (1964).
2 M.Lessen, P. J. Singh, and F. Paillet, “The stability of a trailing line vortex. Part 1. Inviscid theory,” J. Fluid Mech. 63,
753–763 (1974).

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
130.191.17.38 On: Sat, 08 Nov 2014 19:06:30
 074102-19 Toshihiko Hiejima Phys. Fluids 26, 074102 (2014)

3 S. Leibovich and K. Stewartson, “A sufficient condition for the instability of columnar vortices,” J. Fluid Mech. 126,
335–356 (1983).
4 C. J. Heaton, “Centre modes in inviscid swirling flows and their application to the stability of the Batchelor vortex,” J.

Fluid Mech. 576, 325–348 (2007).

5 M. Lessen and F. Paillet, “The stability of a trailing line vortex. Part 2. Viscous theory,” J. Fluid Mech. 65, 769–779 (1974).
6 M. R. Khorrami, “On the viscous modes of instability of a trailing line vortex,” J. Fluid Mech. 225, 197–212 (1991).
7 E. W. Mayer and K. G. Powell, “Viscous and inviscid instabilities of a trailing vortex,” J. Fluid Mech. 245, 91–114 (1992).
8 L. Parras and R. Fernandez-Feria, “Spatial stability and the onset of absolute instability of Batchelor’s vortex for high swirl

numbers,” J. Fluid Mech. 583, 27–43 (2007).

9 I. Delbende, J.-M. Chomaz, and P. Huerre, “Absolute/convective instabilities in the Batchelor vortex: A numerical study

of the linear impulse response,” J. Fluid Mech. 355, 229–254 (1998).

10 C. Olendraru, A. Sellier, M. Rossi, and P. Huerre, “Inviscid instability of the Batchelor vortex: Absolute-convective

transition and spatial branches,” Phys. Fluids 11, 1805–1820 (1999).

11 X. Mao and S. J. Sherwin, “Spectra of swirling flow,” in Proceedings of the 7th IUTAM Symposium on Laminar-Turbulent

Transition (Springer, 2010), pp. 247–252.

12 C. J. Heaton and N. Peake, “Transient growth in vortices with axial flow,” J. Fluid Mech. 587, 271–301 (2007).
13 J. A. K. Stott and P. W. Duck, “The stability of a trailing-line vortex in compressible flow,” J. Fluid Mech. 269, 323–351

(1994).
14 T. Hiejima, “Linear stability analysis on supersonic streamwise vortices,” Phys. Fluids 25, 114103 (2013).
15 S. Ragab and M. Sreedhar, “Numerical simulation of vortices with axial velocity deficits,” Phys. Fluids 7, 549–558 (1995).
16 I. Delbende and M. Rossi, “Nonlinear evolution of a swirling jet instability,” Phys. Fluids 17, 044103 (2005).
17 M. Abid, “Nonlinear mode selection in a model of trailing line vortices,” J. Fluid Mech. 605, 19–45 (2008).
18 K. Duraisamy and S. K. Lele, “Evolution of isolated turbulent trailing vortices,” Phys. Fluids 20, 035102 (2008).
19 B. Di Pierro and M. Abid, “Energy spectra in a helical vortex breakdown,” Phys. Fluids 23, 025104 (2011).
20 M. S. Broadhurst and S. J. Sherwin, “Helical instability and breakdown of a Batchelor trailing vortex,” Progress in

Industrial Mathematics at ECMI 2006, Mathematics in Industry Vol. 12 (Springer, 2008), pp. 191–195.
21 M. A. Herrada, M. Perez-Saborid, and A. Barrero, “Vortex breakdown in compressible flows in pipes,” Phys. Fluids 15,

2208–2218 (2003).
22 M. Khorrami and C. Chang, “Linear and nonlinear evolution of disturbances in supersonic streamwise vortices,” Report

No. AFOSR-TR-97, 1997.

23 J. W. Naughton, L. N. Cattafesta, and G. S. Settles, “An experimental study of compressible turbulent mixing enhancement

in swirling jets,” J. Fluid Mech. 330, 271–305 (1997).

24 K. S. Eckhoff and L. Storesletten, “A note on the stability of steady inviscid helical gas flow,” J. Fluid Mech. 89, 401–411

(1978).
25 S. Leblanc and A. Le Duc, “The unstable spectrum of swirling gas flows,” J. Fluid Mech. 537, 433–442 (2005).
26 L. Coquart, D. Sipp, and L. Jacquin, “Mixing induced by Rayleigh-Taylor instability in a vortex,” Phys. Fluids 17, 021703

(2005).
27 B. Di Pierro and M. Abid, “Rayleigh-Taylor instability in variable density swirling flows,” Eur. Phys. J. B 85, 69 (2012).
28 L. Joly, J. Fontane, and P. Chassaing, “The Rayleigh-Taylor instability of two-dimensional high-density vortices,” J. Fluid

Mech. 537, 415–431 (2005).

29 D. Sipp, D. Fabre, S. Michelin, and L. Jacquin, “Stability of a vortex with a heavy core,” J. Fluid Mech. 526, 67–76 (2005).
30 M. S. Uberoi, “Mechanisms of decay of laminar and turbulent vortices,” J. Fluid Mech. 90, 241–255 (1979).
31 M. Sreedhar and S. Ragab, “Large eddy simulation of longitudinal stationary vortices,” Phys. Fluids 6, 2501–2514 (1994).
32 B. Di Pierro and M. Abid, “Instabilities of variable-density swirling flows,” Phys. Rev. E 82, 046312 (2010).
33 Y. Wada and M.-S. Liou, “A flux splitting scheme with high-resolution and robustness for discontinues,” AIAA Paper

94-0083, 1994.
34 X. Deng and H. Zhang, “Developing high-order weighted compact nonlinear schemes,” J. Comput. Phys. 165, 22–44

(2000).
35 A. Jameson, W. Schmidt, and E. Turkel, “Numerical simulation of the Euler equations by finite volume method using

Runge-Kutta time stepping schemes,” AIAA Paper 81-1259, 1981.

36 J. Jeong and F. Hussain, “On the identification of a vortex,” J. Fluid Mech. 285, 69–94 (1995).
37 O. Lucca-Negro and T. O’Doherty, “Vortex breakdown: A review,” Prog. Energy Combust. Sci. 27, 431–481 (2001).
38 S. P. Govindaraju and P. G. Saffman, “Flow in a turbulent trailing vortex,” Phys. Fluids 14, 2074–2080 (1971).
39 L. Jacquin and C. Pantano, “On the persistence of trailing vortices,” J. Fluid Mech. 471, 159–168 (2002).
40 M. Melander and F. Hussain, “Coupling between a coherent structure and fine-scale turbulence,” Phys. Rev. E 48, 2669–

2689 (1993).

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
130.191.17.38 On: Sat, 08 Nov 2014 19:06:30