You are on page 1of 23

This article was downloaded by: [WALDO - Full Library]

On: 26 July 2010


Access details: Access Details: [subscription number 918522989]
Publisher Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-
41 Mortimer Street, London W1T 3JH, UK

Journal of Turbulence
Publication details, including instructions for authors and subscription information:
http://www.informaworld.com/smpp/title~content=t713665472

On coherent-vortex identification in turbulence


Yves Dubiefa; Franck Delcayrea
a
LEGI-MOST, Grenoble-Cedex 9, France

First published on: 18 December 2000

To cite this Article Dubief, Yves and Delcayre, Franck(2000) 'On coherent-vortex identification in turbulence', Journal of
Turbulence, Volume 1, Art. No. N 11,, First published on: 18 December 2000 (iFirst)
To link to this Article: DOI: 10.1088/1468-5248/1/1/011
URL: http://dx.doi.org/10.1088/1468-5248/1/1/011

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.informaworld.com/terms-and-conditions-of-access.pdf

This article may be used for research, teaching and private study purposes. Any substantial or
systematic reproduction, re-distribution, re-selling, loan or sub-licensing, systematic supply or
distribution in any form to anyone is expressly forbidden.

The publisher does not give any warranty express or implied or make any representation that the contents
will be complete or accurate or up to date. The accuracy of any instructions, formulae and drug doses
should be independently verified with primary sources. The publisher shall not be liable for any loss,
actions, claims, proceedings, demand or costs or damages whatsoever or howsoever caused arising directly
or indirectly in connection with or arising out of the use of this material.
JOT J O U R N AL O F TUR BULEN C E
http://jot.iop.org/

On coherent-vortex identification in
turbulence
Yves Dubief† and Franck Delcayre‡

JoT 1 (2000) 011


LEGI-MOST, BP 53, 38041 Grenoble-Cedex 9, France
E-mail: meyd@engmail.newcastle.edu.au
Downloaded By: [WALDO - Full Library] At: 19:20 26 July 2010

Received 27 September 2000; online 18 December 2000

Abstract. The identification issue of coherent vortices is investigated on the


basis of direct numerical simulation (DNS) and large-eddy simulations (LES) of
turbulent flows. It is first shown that the pressure Laplacian is positive within
a low-pressure tube of small cross section enclosed by convex isobaric surfaces in
a uniform-density flow. Since this quantity is related to the second invariant Q
of ∇u, the Q criterion (region where Q is positive) is a necessary condition for
the existence of such tubes. This eduction scheme is compared to other classical
methods in incompressible simulations of isotropic turbulence: a mixing layer,
a channel flow and a backward-facing step. Q-isosurfaces turn out to display
very nice coherent vortices. This criterion is also used in combination with a
conditional sampling method to discuss the characteristics of quasi-longitudinal
vortices in a manipulated channel flow. The contribution of near-wall vortical
structures to velocity and vorticity fluctuations is clearly isolated. Finally, the Q
criterion is applied to the eduction of Λ vortices in separated flows. The dynamics
of these structures is discussed for an incompressible backward-facing step and a
transonic flow past a rectangular cavity.

† Author to whom correspondence should be addressed. Present address: Department of Mechanical Engineering,
University of Newcastle, NSW 2308, Australia.
‡ Present address: Von Karman Institute for Fluid Dynamics, Chaussee de Waterloo 72, 1640 Rhode, St Genese,
Belgium.

c
2000 IOP Publishing Ltd PII: S1468-5248(00)18318-3 1468-5248/00/000001+22$30.00
On coherent-vortex identification in turbulence

Contents

1 Introduction to vortex identification in turbulence 2


2 Computational details 5
2.1 DNS of isotropic turbulence . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.2 LES of a mixing layer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.3 LES of a backward-facing step flow . . . . . . . . . . . . . . . . . . . . . . . 6
2.4 DNS and LES of a channel flow . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.5 LES of the flow past a rectangular cavity . . . . . . . . . . . . . . . . . . . . . 7

JoT 1 (2000) 011


3 Vortical structures in various flows 7
Downloaded By: [WALDO - Full Library] At: 19:20 26 July 2010

3.1 Isotropic turbulence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7


3.2 Mixing layer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
3.3 Backward-facing step . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.4 Plane channel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
4 The Q criterion in LES and compressible flows 12
5 On the choice of the threshold 13
6 Λ vortices in separated flows 17
7 Conclusion 20

1. Introduction to vortex identification in turbulence

Turbulent flows involve such a variety of scales and structures that they seem impossible to study
without the help of statistics. Once a good comprehension of the mean structure of the flow is
achieved, the thrilling search for links between instantaneous events and particularities in the
statistics begins. This work depends strongly on the eduction scheme and the visualizations and
animations investigated. The identification method requires the definition of coherent structures.
Regarding coherent vortices, Lesieur [1] (pp 6–7) described them as regions of the flow satisfying
three conditions:
(i) the vorticity concentration ω should be high enough so that a local roll up of the surrounding
fluid is possible,
(ii) they should approximately keep their shape during a time Tc far enough in front of the local
turnover time ω −1 ,
(iii) they should be unpredictable.
In this context, a high vorticity modulus ω is a possible candidate for coherent-vortex
identification, especially in free shear flows. For instance, Comte et al [2] extensively discussed
the dynamics of streamwise vortices in a mixing layer on the basis of ω-isosurfaces. In
the presence of a wall, however, the mean shear created by the no-slip condition is usually
significantly higher than the typical vortical intensity of the near-wall vortices. A more
sophisticated criterion is therefore required to distinguish vortices from internal shear layers
in those types of flow.
Journal of Turbulence 1 (2000) 011 (http://jot.iop.org/) 2
On coherent-vortex identification in turbulence

Definitions (i)–(iii) can also be interpreted in terms of a local minimum of pressure. A fluid
parcel winding around a vortex needs to be (in a frame moving with the parcel) in approximate
balance between centrifugal and (static) pressure-gradient effects, according to the following
Euler equation:
1
∂t u + ω × u = − ∇P (1)
ρ
where ω is the vorticity vector, P = p + ρu2 /2 is the dynamic pressure and ρ is the density
(hereinafter, the density ρ is assumed to be uniform whenever an incompressible flow is
considered). In a frame moving with a coherent vortex and supposed locally to be Galilean,
the ratio of the second to the first terms on the left-hand side of (1) is of the order of Tc ω. Thus
the equation reduces to the cyclostrophic balance:

JoT 1 (2000) 011


1
ω × u = − ∇P (2)
Downloaded By: [WALDO - Full Library] At: 19:20 26 July 2010

ρ
if condition (ii) above is fulfilled. Under the assumptions implied by (2), the dynamic pressure
should decrease inside a vortex tube in order to counterbalance the centrifugal effects. Isosurfaces
of pressure have also been used by Comte et al [2], and Robinson’s [3] investigation of coherent
structures in a turbulent boundary layer suggests the superiority of pressure as a vortex eduction
criterion rather than the vorticity modulus. However, the threshold to be used for proper
isopressure surfaces strongly depends on the pressure surrounding the vortical structure. In
regions of high concentrations of vortices, this criterion may fail to capture the details of the
vortical organization.
The criterion which is preferred in the present study shares some properties with both the
vorticity and the pressure criterion. The Q-criterion was named after the second invariant of
velocity gradient tensor ∇u by Hunt et al [4]. The second invariant Q is:
Q = 12 (Ωij Ωij − Sij Sij ) (3)
where Ωij = (ui,j − uj,i )/2 and Sij = (ui,j + uj,i )/2 are respectively the antisymmetric and
the symmetric components of ∇u. In other words, Q is the balance between the rotation rate
Ω2 = Ωij Ωij and the strain rate S 2 = Sij Sij . The implication of the latter observation is fairly
straightforward: positive Q isosurfaces isolate areas where the strength of rotation overcomes
the strain, thus making those surfaces eligible as vortex envelopes. Further support can be found
by recasting (3) in a form which relates to the vorticity modulus (condition (i)):
Q = 14 (ω 2 − 2Sij Sij ). (4)
Since vorticity should increase as the centre of the vortex is approached, Q can be expected to
remain positive in the core of the vortex. This speculation, which arises from good sense, can be
proven providing a few approximations and subsequently linked to pressure lows. One should
be reminded that Q is equal to half the Laplacian of pressure:
11 2  1 1  1 
Q= ω − Sij Sij = − Sij + ijk ωk Sij + ijk ωk
2 2 2 2 2
1 1 1 1
= − ui,j uj,i = − ∂i [uj ui,j ] = − ∂i ∂j [ui uj ] = ∇2 p. (5)
2 2 2 2ρ
According to the maximum principle, the pressure maximum occurs only on the boundary if
Q is strictly positive and the pressure minimum occurs only on the boundary if Q ≤ 0. As
stated by Jeong and Hussain [5], there is no necessary implication for the pressure to reach a
Journal of Turbulence 1 (2000) 011 (http://jot.iop.org/) 3
On coherent-vortex identification in turbulence

Figure 1. Schematic diagram of a low-pressure tube.

JoT 1 (2000) 011


Downloaded By: [WALDO - Full Library] At: 19:20 26 July 2010

minimum within a region of positive Q. Although it has been suggested that a minimum of
pressure might not be appropriate within an agglomeration of vortices, it is important to check
the correspondence of the pressure criterion with the Q criterion for an isolated vortex tube
which contains a pressure low. Figure 1 represents a tube of small cross section around the axis
of the vortex. The lateral surface ∆Σ is assumed to be isobaric and convex. Let ∆V be the
volume of the tube portion bounded by the cross sections Σ1 and Σ2 which are supposed to be
locally normal to the axis of the tube. The pressure gradient on ∆Σ is normal to it and directed
outwards, while ∇p is a tangent to the cross sections. Thus, the pressure gradient flux through
the tube is equal to the flux through ∆Σ and positive. The integral of this flux can be recast into
the integral of ∇2 p on the volume ∆V in accordance with the divergence theorem. Therefore (5)
indicates that the volume integral of Q over ∆V is positive. If the variation of Q is assumed to
be small, it implies that Q is positive in ∆V . Since this procedure can be repeated all along the
tube’s length, the Q criterion (Q > 0) is a necessary condition for the existence of such a thin
low-pressure tube.
It is interesting to note that, for two-dimensional flows, the relation between coherent vortices
and convex regions of a positive Laplacian of pressure has been shown by Larchevêque [6]. This
author remarked that the stream function on a compact domain Ω of R2 is defined as a solution
of a Monge–Ampère equation. Larchevêque then proved that on a compact subdomain V whose
boundary is a closed streamline and inside which ∇2 p is positive, all streamlines inside V are
closed and V is therefore a vortical structure. The extension to the Q criterion is straightforward
using (5). Similar conclusions have been obtained by Weiss [7] to characterize ‘elliptic’ regions
in two-dimensional flows. If one looks at the inviscid vorticity-gradient equation in this case
D
∇ω = −∇u|t ⊗ ∇ω (6)
Dt
where ω is the vertical vorticity component, one can calculate the eigenvalues of ∇u|t (identical
to those of ∇u). Their square is −Q/2, and they are pure imaginaries if Q > 0. In this case the
vorticity gradient will rotate locally, a property which is expected from a vortex. It was checked
in Basdevant and Philopovitch [8] using DNS that the vortices in two-dimensional isotropic
turbulence were quite well represented by Weiss [7]’s elliptic regions (the quantity Weiss called
Q was in fact of the opposite sign with respect to the present notations). In fact, (6) is also valid
for a passive-scalar gradient ∇ρ, if one neglects molecular diffusion,
D
∇ρ = −∇u|t ⊗ ∇ρ (7)
Dt
Journal of Turbulence 1 (2000) 011 (http://jot.iop.org/) 4
On coherent-vortex identification in turbulence

and the scalar gradient will also rotate locally in elliptic regions. Note that (7) is still valid in
three dimensions, which unifies the two- and three-dimensional formalisms in terms of scalar
gradient. In three dimensions, (6) is no longer valid of course, but the inviscid vorticity equation
D
ω = ∇u ⊗ ω (8)
Dt
has been used in the literature (see Chong et al [9]) to characterize regions where eigenvalues
of ∇u may be complex, implying a local rotation of ω. This is obtained when the discriminant
 3
Q R
∆= + (9)
3 2
(where R is the third invariant of the velocity-gradient tensor) is positive. It can be shown to

JoT 1 (2000) 011


be equivalent to the antisymmetric part of ∇u prevailing over the symmetric one, as pointed
out by the authors, which obviously relates the ∆ and the Q criteria. Another criterion has
Downloaded By: [WALDO - Full Library] At: 19:20 26 July 2010

been developed by Jeong and Hussain [5] based on the following arrangement of the inviscid
Navier–Stokes equations:
D 1
Sij + Ωik Ωkj + Sik Skj = − p,ij . (10)
Dt ρ
The Hessian of pressure (p,ij = ∂i ∂j p) can provide information on the local extrema of pressure.
Assuming that unsteady straining (DSij /Dt) is negligible, the study of pressure minima can be
performed with S 2 +Ω2 . A local minimum of pressure exists if p,ij has two positive eigenvalues,
i.e. S 2 + Ω2 has two negative eigenvalues. The λ2 definition of [5] was named after the second
eigenvalues with λ1 ≥ λ2 ≥ λ3 . Cucitore et al [10] have shown that this definition is strongly
related to the Q criterion, since, in the reference frame of the vortex, λ2 can be written as a
balance between local straining and rotation.
In the present paper, an extensive database is used to investigate the behaviour of the
Q criterion in transitional and turbulent flows. After a brief presentation of the different
computational details involved in the study, section 3 presents a comparison of four vortex-
eduction methods (low pressure, high vorticity, negative λ2 , positive Q) for isotropic turbulence,
mixing layer, channel flow and backward-facing step flow. A more detailed investigation of
the threshold effects on the Q criterion is proposed in section 4. The last section is devoted to
flow animations of vortices in separated flows over expansions. In spite of its obvious link with
pressure (equation (5)), we often refer to the Q criterion as a ∇u-based vortex eduction method,
such as the ∆ and λ2 criteria and the vorticity (which relates to the antisymmetric part of ∇u).
The reason for such a classification comes from the use of the Q criterion in compressible flows,
where equation (5) no longer holds. Finally, we should point out that the Q criterion was found
to be inadequate for certain vortices, such as the Bödewadt vortex, by Jeong and Hussain [5] and
Cucitore et al [10]. However, this configuration, as well as others presented by these authors, are
very unlikely in common turbulent flows such as those presented here, as the similarity between
the λ2 definition and the Q criterion will show.

2. Computational details

Vortex eduction schemes have already been applied in a wide variety of flows in our group. Five
examples have been chosen and are detailed here. The validity of each simulation has been
assessed by comparing statistical results with either experimental or other numerical data.

Journal of Turbulence 1 (2000) 011 (http://jot.iop.org/) 5


On coherent-vortex identification in turbulence

2.1. DNS of isotropic turbulence

The direct simulation of isotropic turbulence was performed by Ossia [11] on a cubic box whose
dimensions and resolution are (2π)3 and 2563 , respectively. The flow is incompressible, and the
Navier–Stokes equations are solved in their rotational form. Pseudospectral methods are used to
perform spatial derivatives. Time advancement is carried out by a low-storage third-order Runge–
Kutta scheme. The initial velocity field is chosen so that its kinetic-energy spectrum peaks at a
given low wavenumber in order to study the infrared pressure spectrum. The turbulence decays
freely. This work is reported in Lesieur et al [12].

2.2. LES of a mixing layer

JoT 1 (2000) 011


Comte et al [2] used a parallel incompressible code to run LES of mixing layers at a very
high Reynolds number (ν = 0). The code is actually a modified version of the one presented
Downloaded By: [WALDO - Full Library] At: 19:20 26 July 2010

earlier to suit the requirements of massive-parallel computation. The spatial development of


the mixing layer is allowed by efficiently replacing the pseudospectral methods by sixth-order
compact schemes in the longitudinal directions. The inlet velocity profile consists of a hyperbolic
tangent profile perturbed by a small two-dimensional white noise. The size of the domain is
112δi ×28δi ×14δi , where δi is the vorticity thickness at the inlet. The resolution of the simulation
is 384 × 96 × 48. The subgrid-scale model is the filtered structure function model developed
by Ducros et al [13]. Hereafter, x, y and z refer to, respectively, the streamwise, vertical and
spanwise directions and the corresponding velocities are u, v and w.

2.3. LES of a backward-facing step flow

A LES of this separated flow was carried out by Delcayre [14, 15] in the same configuration as
Le et al [16]. The numerical method uses a finite-volume method on a staggered grid. The
velocities at cell interfaces which are required to evaluate convective fluxes are interpolated by
the QUICK-SHARP scheme (Leonard [17]). All the numerical details can be found in [15].
The step-height Reynolds number corresponding to this simulation is: ReH = U0 H/ν = 5100.
The total resolution of the computational domain is 97 × 34 × 46. The subgrid scale used is the
selective structure function model in its four-point formulation (see Lesieur and Métais [18]).

2.4. DNS and LES of a channel flow

Also, in section 3, is a DNS performed by Lamballais [19] with a code similar to that of the
mixing layer. The time step was modified into a semi-implicit fractional step scheme. In section 4,
channel flow data are extracted from a very well resolved LES aimed at the study of near-wall
manipulation [20]. One side of the channel is smooth whereas two spanwise cavities of small
dimensions are embedded in the opposite wall. The Reynolds number is low Reτ = 160 based on
the skin-friction velocity and the half-width of the channel. The simulation was performed using
a multi-domain compressible code whose accuracy is fourth order in space and second order in
time. The scheme is the predictor–corrector scheme of MacCormack [21] modified by Gottlieb
and Turkel [22]. The Mach number is sufficiently low (M a = 0.3) to neglect compressibility
effects. The dimensions of the channel in wall units (ν/uτ ) are 2000 × 320 × 500 and the width
of the grooves is equal to 40ν/uτ . The wavelength of the grooves is 25w, where w is the width
(and height) of a groove. The grid is stretched in both the streamwise and spanwise directions

Journal of Turbulence 1 (2000) 011 (http://jot.iop.org/) 6


On coherent-vortex identification in turbulence

to capture well the near-wall structures and events occurring in the vicinity of the cavities. The
minimum grid spacings are ∆+ +
x = 2 and ∆y = 0.4. The total resolution is 200 × 128 × 64. The
subgrid scale model is the filtered structure function model in its four-point formulation [13].
The model contribution was checked to be negligible in the near-wall region of the flow. Here,
we focus only on the flow developing on the smooth wall. For this half of the channel, the
perturbation induced by the grooves is found to have little effect on the structure of the near-wall
turbulence. Velocity statistics are in very good agreement with the DNS of [19].

2.5. LES of the flow past a rectangular cavity

The last database investigated here concerns the LES of a transonic flow past a rectangular
cavity [20]. The calculation reproduces an experiment of Tracy and Plentovich [23]. The cavity

JoT 1 (2000) 011


dimensions are 4H × H × H, where H is the depth of the cavity. The boundary layer thickness
Downloaded By: [WALDO - Full Library] At: 19:20 26 July 2010

upstream of the cavity is equal to 0.3H. The Reynolds number, Re∞ = 1.25 × 106 , is based
on the free-stream velocity and H. The corresponding Mach number is 0.95. The simulation is
carried out with the same code as for the channel flow. The resolution is fairly coarse with the
first node at y + = 70 in the vertical direction for the boundary layer domain. The inflow consists
of a mean profile perturbed by white noise. Further details can be found in [20].
Despite the lack of near-wall coherent structures, pressure spectra were found to predict
accurately the fundamental frequency of the vortex shedding. Two other low frequencies have
also been identified and are in good agreement with the experiment of Tracy and Plentovich [23],
within the convergence error of the spectra for the lowest frequency. The level of noise is
overestimated by 10–20 db, mainly because of the confinement of the domain and open boundary
conditions. This latter measurement is based on the level of the root-mean square (rms) of the
pressure fluctuations using the same relation as [23].

3. Vortical structures in various flows

For each flow presented here, the thresholds of the pressure, ω, λ2 and Q criteria are adjusted in
order to give the same representation of the flow, which relies on previous publications and the
authors’ experience. The picture obtained with Q isosurfaces serves as the reference.

3.1. Isotropic turbulence

Numerical simulations of isotropic homogeneous turbulence have shown the existence of vortical
structures called ‘worms’ [24, 25]. These structures possess a longitudinal scale of the order of
the integral velocity scale and a diameter of several Kolmogorov length scales it is possible that
a scaling based on the Taylor scale, which characterizes a kind of vorticity thickness on the local
mixing zones, would be more appropriate to define the diameter length scale of the worms. When
a numerical simulation of freely decaying isotropic homogeneous turbulence is undertaken, an
autosimilar regime is reached after 5Tret , where Tret is the turn over time of the large initial
eddies. At this time, when enstrophy reaches a maximum, turbulence becomes fully developed
and the energy
√ spectrum exhibits a broad range of scales. Here an√initial vorticity is be defined
as ωI = 2π/ 3 based on the initial energy spectrum where U0 = 3 is a characteristic velocity.
The four criteria are applied to a field at t = 5.5Tret extracted from the data of [11, 12].
The low-pressure field depicted in figure 2(a) effectively shows the existence of worms, yet

Journal of Turbulence 1 (2000) 011 (http://jot.iop.org/) 7


On coherent-vortex identification in turbulence

JoT 1 (2000) 011


(a) (b)
Downloaded By: [WALDO - Full Library] At: 19:20 26 July 2010

(c) (d)

Figure 2. Organized vortical activity in isotropic homogeneous turbulence.


(a) Pressure isosurfaces, P = −0.15ρ0 U02 . (b) Vorticity isosurfaces, ω = 0.36ωI .
(c) λ2 isosurfaces, λ2 = −0.25. (d) Q isosurfaces, Q = 0.5.

these are fewer and of larger dimensions than the structures isolated by ∇u-based criteria
(figures 2(b)–(d)). It is striking that the pressure fails to capture the fine detail of agglomerations
of vortices (for instance at the bottom left-hand side of the box in figure 2(a)). The topology
of the flow given by vorticity isosurfaces (figure 2(b)) is slightly different from that obtained
with Q. As observed by Kida and Miura [26], in addition to the expected tubular vortices, the
vorticity criterion isolates layered structures which are vorticity sheets, not vortices. Q and λ2
criteria give the same results, with both figures free of deformation regions. However it should
be noted that figure 2(c) is affected by a small noise in comparison to figure 2(d). This feature
is common to all flows treated in this study and cannot be explained.

3.2. Mixing layer

Large-scale vortices in a plane mixing layer are of two types: spanwise and streamwise
vortices. The first to appear are Kelvin–Helmholtz vortices which emerge from the growth
of small perturbations in a basic inflectional shear. Their wavelength λ is approximately equal
to 7δI where δI is the initial vorticity thickness. As the vortices are advected downstream,
some undergo a phenomenon called pairing. These events strongly depend on the initial

Journal of Turbulence 1 (2000) 011 (http://jot.iop.org/) 8


On coherent-vortex identification in turbulence

perturbations. When perturbations are two dimensional, so is the pairing process. In the case
of three-dimensional perturbations, Kelvin–Helmholtz billows undergo helical pairing which
is characterized by staggered stretchings and local reconnections. These events have been
highlighted by Comte et al [27] who investigated the evolution of a three-dimensional temporal
mixing layer with two- and three-dimensional white noise superimposed to the initial flow field.
The second type of vortices consists of counter-rotating streamwise vortices stretched between the
Kelvin–Helmholtz billows. The spanwise wavelength is found to be of the order of 2/3λ [28].
The mechanism which gives birth to such vortices can be understood by using the vorticity
equation (8) for an inviscid fluid, in a slightly different form:
 
Dω 1
= ∇u ⊗ ω = S ⊗ + ω× ω = S ⊗ ω. (11)
Dt 2

JoT 1 (2000) 011


Here S is the deformation tensor which is assumed not to vary too much between the spanwise
Downloaded By: [WALDO - Full Library] At: 19:20 26 July 2010

vortices. It implies that a vortex filament in such a straining field tends to be stretched in the
direction of the first principal axis of deformation of S, giving rise to alternate longitudinal
vortices.
The data of Comte et al [2] used here is a LES of a spatially developing mixing layer
perturbed by two-dimensional white noise at the inlet. The snapshot chosen depicts the early
stages of transition with the formation of Kelvin–Helmholtz vortices, a pairing and some
longitudinal vortices at the end of the computational domain. The major disadvantage of vorticity
isosurfaces (figure 3(b)) is shown at the entrance of the domain where the magnitude of shear is
close to the vorticity magnitude of vortices, hence the presence of a layered structure. Further
downstream, as the mean shear is smaller than the typical vorticity magnitude of the coherent
structures, ω isosurfaces are similar to λ2 (figure 3(c)) and Q (figure 3(d)) isosurfaces. All
these criteria isolate well the two Kelvin–Helmholtz vortices involved in a pairing process and
longitudinal vortices, the Q criterion giving the cleanest picture of the flow. Pressure isosurfaces
again educe structures of larger dimensions than those by the other criteria, especially in the
region of pairing. Figure 3(a) does not show any evidence of longitudinal vortices. This situation
underlines a major issue of pressure as a criterion used to follow vortices over time, for instance.
In mixing layers, the pairing is known to cause significant variations of pressure which are very
likely to affect the shedding region, as mentioned by Dimotakis and Brown [29] in a feedback
process. These large-scale variations may require a time, if not local, adjustment of the threshold
to make sure longitudinal vortices are identified. Therefore, the pressure criterion might not be
reliable for flow animations.

3.3. Backward-facing step

This flow presents several regions with very different characteristics. The detachment of the
boundary layer at the corner implies the generation of a strong shear region very similar to a plane
mixing layer. The problem of the separated–reattached flows is the reattachment region where
the shear layer reattaches and strongly interacts with the wall. The backward-facing step flow
gets together free-shear and wall flows, which is a severe test for the vortices eduction criteria. As
for the preceding cases, the same tendency is observed in this calculation. Using the Q criterion
(figure 4(d)) is much more appropriate for this flow and allows one to study the temporal evolution
of the flow. Large scales are approximately represented by pressure isosurfaces (figure 4(a)),
especially upstream of the reattachment and, more realistically, downstream of the reattachment.
The reattachment region is dominated by the impingement of the shear layer, hence the radiation
Journal of Turbulence 1 (2000) 011 (http://jot.iop.org/) 9
On coherent-vortex identification in turbulence

(a) (b)

JoT 1 (2000) 011


Downloaded By: [WALDO - Full Library] At: 19:20 26 July 2010

(c) (d)

Figure 3. Organized vortical activity in a mixing layer. Here ωI = 2U0 /δI .


(a) Pressure isosurfaces, P = −0.08ρ0 U02 . (b) Vorticity isosurfaces, ω = 0.75ωI .
(c) λ2 isosurfaces, λ2 = −0.06. (d) Q isosurfaces, Q = 0.5.

of pressure waves as shown in the middle of figure 4(a). The vorticity modulus (figure 4(b)) is
highly contaminated by the regions of strong shear. The criterion based on λ2 (figure 4(c)) gives
an aspect quite similar to the Q criterion, but contains more noise. It is then possible to carry out
a topological study from the development of the Kelvin–Helmholtz instabilities to their further
evolution with the aid of the Q criterion. An example is given in section 6.

3.4. Plane channel

The same procedure as before is applied to a channel flow field obtained from a direct numerical
simulation [19]. In this flow, turbulence is sustained by coherent structures confined in the inner
layer, i.e. the region close to the wall [30]. The dynamics of the flow seems to depend on near-wall
vortices and low- and high-speed streaks. The regeneration process of these structures is still far
from being understood. As near-wall turbulence produces high-magnitude skin friction, there is
an obvious interest in its dynamics with the perspective of attenuating its effects on drag. Since
the very first conceptual models of near-wall hairpin-like vortices appeared [31], it has been
suspected that vortices play a key role in the generation of skin friction. This has been confirmed
with in-depth investigations of organized vortical structures in boundary layers and channel flows
(see Robinson [3]). These vortices consist of a variety of hairpin-like vortices with symmetric
and non-symmetric cores and longitudinal vortices. It is still unclear whether all such structures
belong to the class of hairpin vortices, which would be degenerated by the ambient turbulence,
or are vortices emerging from different mechanisms. Nevertheless it seems obvious that the

Journal of Turbulence 1 (2000) 011 (http://jot.iop.org/) 10


On coherent-vortex identification in turbulence

JoT 1 (2000) 011


(a) (b)
Downloaded By: [WALDO - Full Library] At: 19:20 26 July 2010

(c) (d)

Figure 4. Vortical activity in a backward-facing step flow educed by various


criteria. (a) Pressure isosurfaces, P = −0.003ρ0 U02 . (b)Vorticity isosurfaces,
ω = 1.5U0 /h. (c) λ2 isosurfaces, λ2 = −0.075. (d) Q isosurfaces, Q = 0.25.

longitudinal portion of their core is responsible for most of the transverse turbulent activity in
the near-wall region (see Robinson [3], Dubief et al [32]). Thus we shall refer here to near-wall
vortical structures as quasi-streamwise vortices by taking into account only the dominant part of
their core.
The streamwise vorticity fluctuations induced by near-wall vortices are of the order of the
fluctuations in other directions due to streaks for instance, as shown in section 5. The eduction of
vortices by high-vorticity isosurfaces is therefore almost useless in discussing the topology of the
flow. Figure 5(b) does not allow any clear distinction between streaks and vortices, even though
some tubular structures are actually visible. As for the previous examples, the pressure criterion
(figure 5(a)) fails in regions of high concentrations of vortices, such as the bottom left-hand side
of the figure (to be compared with figure 5(d)). The picture obtained with pressure isosurfaces
is misleading regarding the density of vortices, but still manages to capture well some isolated
structures. The noise in the λ2 plot is critical in this configuration and may well depend on the
threshold. Jeong et al [33] suggested that the threshold should be adjusted to the maximum
of the rms of λ2 . Such a procedure has not been performed here; in the light of Jeong et al ’s

Journal of Turbulence 1 (2000) 011 (http://jot.iop.org/) 11


On coherent-vortex identification in turbulence

JoT 1 (2000) 011


(a) (b)
Downloaded By: [WALDO - Full Library] At: 19:20 26 July 2010

(c) (d)

Figure 5. Organized vortical activity in a plane channel flow. (a) Pressure


isosurfaces, P = −0.006ρ0 U02 . (b) Vorticity isosurfaces, ω = 3U0 /h. (c) λ2
isosurfaces, λ2 = −0.075. (d) Q isosurfaces, Q = 0.5.

results, it is believed that it could probably increase the quality of the picture. It should be
reminded here that we choose the threshold so that the visualization matches the one obtained
with Q isosurfaces. The Q criterion seems to capture well all the details of the flow, especially
the packets of vortices (see figure 5(d)).

4. The Q criterion in LES and compressible flows

As stated in the introduction, the choice of the Q criterion as a vortex-identification method is


strongly supported by: (i) its relation to pressure low, through equation (5) and (ii) the very
definition of Q (equation (3)) as the balance between the local rotation rate and the strain rate.
While the second definition can be applied to any velocity field, the theoretical basis (equation (5))
of definition (i) does not hold in incompressible LESs. The formalism of LES, which relies on a
turbulent viscosity νt , brings a slight modification to the incompressible Navier–Stokes equations:
∂i ui = 0

Journal of Turbulence 1 (2000) 011 (http://jot.iop.org/) 12


On coherent-vortex identification in turbulence

1
∂t ui + uj ∂j ui = − ∂i p + ν∂j ∂j ui + ∂j [νt (∂j ui + ∂i uj )] (12)
ρ
leading to the following relation between Q and ∇2 p:
 
1 1 1 2
Q = − ui,j uj,i = ∇ p − 2∂i νt ∂j ∂j ui − (∂i ∂j νt )(∂j ui + ∂i uj ) . (13)
2 2 ρ
Equation (13) introduces two terms which respectively involve the gradient and the Hessian of
νt . The importance of these terms obviously depends on the behaviour of subgrid-scale models
inside coherent vortices. In the variety of simulations presented here, it should be pointed out
that one model, at least, is designed to nullify νt in coherent vortices. The selective-structure
function model used in section 3.3 prescribes νt = 0 wherever the local vorticity vector is

JoT 1 (2000) 011


sufficiently aligned with the vorticity vector averaged on the neighbouring points. Thus, both
the gradient and the Hessian of νt are expected to vanish in the core of the vortices resolved by the
Downloaded By: [WALDO - Full Library] At: 19:20 26 July 2010

LES. As a matter of fact, the LES of a backward-facing step flow (section 3.3) shows very good
agreement between pressure lows (figure 4(a)) and Q isosurfaces (figure 4(d)) downstream of the
reattachment, for the largest scales. For other subgrid-scale models, the effect of the additional
terms in equation (13) are more difficult to anticipate. The filtered structure function model (used
for the other LESs) calculates the local turbulent viscosity on the basis of a filtered velocity field
which accentuates the weight on the scales in the vicinity of the cut-off length scale. In such a
case, it has been checked that the magnitude and the fluctuations of νt inside large-scale vortices
are fairly small. This issue has not received further attention in the present study, but we believe
that a statistical procedure, similar to that described in section 5, can be worthwhile in order to
assess the impact of any subgrid-scale modelling on coherent-vortex identification.
Compressible flows are also an example where equation (5) does not hold and, consequently,
the justification of the method must rely only on definition (ii). Simulations of compressible
mixing layers (not shown here) have shown that Q isosurfaces are well correlated with pressure
lows in the early stage of the transition, i.e. during the formation of Kelvin–Helmholtz billows.
The identification of longitudinal vortices with the pressure criterion has been found to be as
delicate as for the incompressible case discussed in section 3.2. The Q criterion and pressure
criterion have also been used in subsonic turbulent channel flows (see the lowest Mach number
investigated (M a = 0.3, in figure 10 below) with the same success as shown in section 3.4.
Thus, the most widely accepted property of the Q criterion seems to be definition (ii). Such a
definition was introduced by Truesdell in 1953 (as quoted by Jeong and Hussain [5], p 75) as
the kinematic vorticity number
Nk =

/
S
(14)
which gives a measure of ‘the quality of rotation’ [5]. However, it is interesting to point out that
pressure lows can still be used in LES and compressible flows to validate Q isosurfaces even
though the relation between Q and pressure could be partially established for uniform density
and viscosity, i.e. incompressible, flow (see the introductory section of the paper).

5. On the choice of the threshold

The main issue in vortex eduction is subjectivity, at least for the methods presented here. The
threshold is critical to the study of the details of a turbulent flow, as illustrated by figures 6(a)–(f).
The field is extracted from the database of a well resolved LES of a channel flow [20]. The density

Journal of Turbulence 1 (2000) 011 (http://jot.iop.org/) 13


On coherent-vortex identification in turbulence

(a) (b)

JoT 1 (2000) 011


Downloaded By: [WALDO - Full Library] At: 19:20 26 July 2010

(c) (d)

(e) (f)

Figure 6. Threshold effects on near-wall vortices educed by the Q criterion:


(a) Q = 0.2, (b) Q = 0.4, (c) Q = 0.6, (d) Q = 0.8, (e) Q = 1.0 and (f) Q = 1.2.

of structures isolated by the Q criterion decreases as the threshold increases. The theory behind
the eduction method offers no strong basis to support any one plot of figure 6 against the others
as a representative picture of coherent vortices. A selection can be made on vorticity since
the first impact of a vortex on turbulent quantities is to create vorticity fluctuations. Figure 7
shows the statistics of vorticity fluctuations in the present case. The comparison with the DNS
of Lamballais [19] indicates an underestimation of the resolved vorticity components which is a
mere consequence of the filtering process inherent to LES methods. As shown experimentally
(see Dubief et al [34]), the filtering velocity differences ∂j ui = ∆ui /∆xj by a too large, spatial

Journal of Turbulence 1 (2000) 011 (http://jot.iop.org/) 14


On coherent-vortex identification in turbulence

Figure 7. Rms vorticity fluctuations normalized by the wall scaling (ωI+ =


ωI ν/u2τ ): —— denotes a smooth wall, - - - - denotes a rough wall and
· · · · · · denotes DNS [19]. Extracted from [20, 32].

JoT 1 (2000) 011


Downloaded By: [WALDO - Full Library] At: 19:20 26 July 2010

step ∆xj produces a dramatic underestimation of (∂y ui ) , even for ∆xj ∼ 10η, η being the
local dissipative scale. Despite the discrepancy with DNS, the LES manages to reproduce the
qualitative trend and, more importantly, the ratio between components of vorticity. The shape
of the curve for streamwise vorticity fluctuations was related to quasi-longitudinal vortices by
Kim et al [35]. These authors used a two-dimensional streamwise vortex model above the
wall to find out the typical diameter and location of the near-wall vortices (y + 20). The rms
of the streamwise vorticity fluctuations being dominated by the other components for most of
the near-wall region (figure 7) highlights the difficulty of extracting such structures with the ω
criterion. It also raises the question: how significant are the fluctuations of vorticity induced by
quasi-longitudinal vortices?
A fairly efficient way to investigate both the latter issue and threshold effects is to use a
conditional sampling technique. A review of this concept can be found in [36]. In our case, the
intermittency function IQs is defined as a function of the second invariant:

1 if Q(x, t) ≥ Qs
IQs (x, t) = (15)
0 if Q(x, t) < Qs
where Qs is the threshold value. This function is subsequently applied to 700 instantaneous
fields of vorticity fluctuations. The statistical formulation of this procedure is
T
2 IQs (x, t)ωI2 (x, t) dt
ωI Q=Qs (x) = 0  T (16)
I
0 Qs
(x, t) dt
which, in physical terms, provides information on the magnitude of vorticity fluctuations in
regions of positive Q. The results for the three components of vorticity are displayed in figure 8.
For each plot the magnitude of ωI  ≡ ωI2  increases as the Qs goes from 0.2 to 0.8; it then
decreases as Qs further increases. Relating this phenomenon to instantaneous visualizations
(figures 6(a)–(f)) demonstrates that too large a threshold could hide structures which contribute
greatly to high vorticity fluctuations. Furthermore, the behaviour of ωi  shows that the large
vortices isolated in figures 6(e) and 6(f) do not provide the largest contribution to the vorticity
fluctuations. On the other hand, it has not been possible to verify the physical significance of
low-Q plots, such as figure 6(a). Vector plots in various cross sections have been used (not shown
here), and mostly confirm that the large-scale isosurfaces are vortices which create a spiral motion
around their axis. This method fails for the smallest scales of the plot. However, it does not

Journal of Turbulence 1 (2000) 011 (http://jot.iop.org/) 15


On coherent-vortex identification in turbulence

(a) (b)

JoT 1 (2000) 011


Downloaded By: [WALDO - Full Library] At: 19:20 26 July 2010

(c)

Figure 8. Conditional statistics of vorticity fluctuations for the six different


thresholds: , Q = 0.2; , Q = 0.4; , Q = 0.6; , Q = 0.8; ◦, Q = 1.0;
•, Q = 1.2; ——, ωI , extracted from figure 7. (a) Longitudinal vorticity.
(b) Vertical vorticity. (c) Spanwise vorticity.

prove that these structures are not vortices. It should be noted that this plot (figure 6(a)) would
be very unlikely to be proposed for a discussion on near-wall vortices by anyone who is aware
of the work of Robinson [3] since it contains too many structures. In addition, the conditional
statistics of the vorticity given by Qs = 0.2 has too small a level of Q, which suggests that the
impact of most isolated structures on the dynamic of flow is probably not significant. In order
to minimize the weight of subjectivity on the choice of the threshold, the conditional sampling
technique appears to provide a valuable hint through the threshold Qc which gives the highest
level of vorticity fluctuations. The threshold can eventually be adjusted to enlarge the core of
some vortices which could be quite small for Q = Qc . It should be noted that the fluctuations
of ωx  (quasi-streamwise vortices) have the the same order of magnitude as ωz (the average of
the contribution of streaks and vortices) in the near-wall region.
Another interesting aspect of figures 8 is the behaviour of ωI  in the range 5 ≤ y + ≤ 25. For
Qs = Qc = 0.8, the ratio ωx /ωx reaches a maximum of almost four while the corresponding
quantities for the vertical and spanwise directions are 2 and 1.5, respectively. Thus, around
y + 5–10, all conditional rmss have roughly the same intensity, hence the difficulty in
identifying the privileged orientation of the near-wall vortices with plots such as figures 8(a)–(c).
Visualizations of Q isosurfaces show no evidence for the existence of either vertical or spanwise
vortices in that region. A typical cross section of a vortex in the near-wall region is displayed
in figure 9. The two terms on the right-hand side of (3), namely Ω2 and S 2 , are very similar,
owing to high-deformation regions surrounding the vortex. The Q and ωx contours (figures 9(b)
and 9(e)) both exhibit a nice-looking circular pattern which denotes the existence of a vortex.
The intense red ωx contours at the wall show the counter-rotative vortex sheet generated by the

Journal of Turbulence 1 (2000) 011 (http://jot.iop.org/) 16


On coherent-vortex identification in turbulence

(a) (b) (c)

(d) (e) (f)

Figure 9. Cross section of a quasi-longitudinal vortex. The colour range extends

JoT 1 (2000) 011


from blue to red: 0 ≤ Ω2 ≤ 20, 0 ≤ S 2 ≤ 20, 0.2 ≤ Q ≤ 3, −2.5 ≤ ωy ≤ 2.5,
Downloaded By: [WALDO - Full Library] At: 19:20 26 July 2010

−4 ≤ ωx ≤ 4, −6 ≤ ωz ≤ 6. All quantities are normalized by Uc and h. (a) Ω2 ,


(b) Q ≥ 0.2, (c) S 2 , (d) ωy , (e) ωx and (f) ωz .

no-slip condition. The interpretation of the negative and positive regions of fluctuating ωy and
ωz is more delicate because of the combined effects of the mean shear along the y-direction, the
shear imposed by low- and high-speed streaks (mainly along z) and the presence of the wall.
The vorticity contours can easily be related to the conceptual model of Jeong and Hussain [5] of
an inviscid longitudinal vortex tube embedded in a uniform shear in the direction perpendicular
to the vortex axis. Figure 4(b) of [5] isolates two regions of high vorticity located on both sides
of the vortex, in a similar way as the two zones displayed on the lower left-hand side and upper
right-hand side of figure 9(a). For the conceptual model, this vorticity production arises from
the evolution of the longitudinal velocity field which is written as a simple advection equation,
Du/Dt = 0. Therefore, the regions of high vorticity outside the vortex core are shear layers of
fairly intense ωy and ωz , caused by the wrapping of the mean shear around the vortex.
An illustration of near-wall coherent vortices is provided by the animations in figures 10(a)
and (b). They represent the evolution of the flow on the upper and the lower parts of the channel,
respectively. The bottom wall contains two, transverse, square grooves to mimic a widely-spaced
d-type rough wall. The statistics displayed in this section are issued from the upper side of the
channel since this has been found to be similar to the unperturbed channel flow. The comparison
of the results from the conditional sampling based on Q indicates a reduction in the intensity of
the velocity and vorticity fluctuations of the order of 10% due to the groove. For more details, the
reader is referred to Dubief [20] and Dubief et al [32]. It should be noted here that the threshold
in figures 10(a) and 10(b) is half the critical threshold found earlier in figures 8(a)–(c). It was
found to give the most detailed picture of the flow without the addition of uncertain small-scale
vortices, as discussed earlier for figure 6(a).

6. Λ vortices in separated flows

The Q criterion is applied to two separated flows over moderately complex geometries where
the interaction of coherent vortices and walls is found to produce large-scale Λ vortices. The
following visualizations use a threshold which has been previously tested in the same way as
in section 3, i.e. by comparing pictures given by different criteria at various thresholds. No

Journal of Turbulence 1 (2000) 011 (http://jot.iop.org/) 17


On coherent-vortex identification in turbulence

(a) (b)

JoT 1 (2000) 011


Figure 10. Animations of coherent vortices educed by the Q criterion (Q = 0.4)
in a one-sided grooved channel flow. In the text, (a), a smooth wall, is also
Downloaded By: [WALDO - Full Library] At: 19:20 26 July 2010

referred to as the upper wall and (b), a grooved wall, as the lower wall.

conditional analysis, as in the previous section, has been performed. In the last animation, the
flow is compressible and therefore most of the theory presented in the introduction no longer
holds. In such a case the eduction of vortices relies on the balance between the rotational rate
and the strain rate (equation (3)). The physical significance of the resulting isosurfaces was
validated using vector plots in several cross sections of the large-scale vortices. It is interesting
to note that Q isosurfaces are also in good agreement with the pressure lows for instantaneous
visualizations. However, both local and time adjustments of the pressure threshold are required,
owing to pressure waves emanating from vortex impingements.
The animation in figure 11 shows the phenomenology of coherent vortices downstream of a
backward-facing step. These visualizations show the general tendency of the flow to reorientate a
spanwise vorticity field into streamwise vorticity. It clearly indicates how the Kelvin–Helmholtz
billows, shed downstream of the step, oscillate in the spanwise direction and are subject to helical
pairing. The oscillation of the spanwise vortices that are initiated by a staggered oscillation of
the Kelvin–Helmholtz vortices is amplified, foreshadowing the appearance of Λ vortices. Then,
after this pairing the three dimensionality of the flow increases. Thus, the topology of the flow is
completely changed 3H downstream of the step. Nevertheless, the large-scale coherence of the
flow is still observed with the nonlinear stretching of the Kelvin–Helmholtz billows into large Λ
vortices. These vortices impinge the wall and are stretched into large arch-like vortices as they
are convected downstream of the reattachment. The persistence of these Λ-shaped structures
downstream of the reattachment is very likely to be a key factor in the slow relaxation of the
mean streamwise profiles towards canonical turbulent boundary layer profiles. Some authors
have also experimentally highlighted the presence of large-scale structures far downstream of
the step and have shown their quasi-periodic occurrence.
The mechanism involved in the formation of the Λ vortices in the flow past a rectangular
cavity is far simpler than that described for the backward-facing step flow. The deformation of
the two-dimensional Kelvin–Helmholtz vortices is merely due to the shear of the mean velocity in
the spanwise direction due to the side-walls. The ends of a spanwise vortex shed at the upstream
edge of the cavity quickly become trailing legs (see figure 12(a)), and the centre, which turns into
the head of a new Λ vortex, is lifted upwards in a similar way as hairpin vortices in a turbulent
boundary layer [37], as shown by figure 12(b). Such a flow produces self-sustained oscillations

Journal of Turbulence 1 (2000) 011 (http://jot.iop.org/) 18


On coherent-vortex identification in turbulence

Figure 11. Animation of the evolution of coherent vortices downstream of a

JoT 1 (2000) 011


backward-facing step. Vortices are educed by Q isosurfaces coloured by ωx .
Downloaded By: [WALDO - Full Library] At: 19:20 26 July 2010

(a) (b)

(c)

Figure 12. Animations of the flow past a rectangular cavity of Q isosurfaces.


The time is given in milliseconds. (a) Top view, (b) side view and (c) rear view.

of the shear layer (see Rockwell [38]) which then induce low-frequency components. The main
frequency is the shedding frequency which is forced at a Strouhal number of one based on the
free-stream velocity and on the cavity length. This result is in very good agreement with the
experimental results of [23]. The forcing of the shedding is caused by the impingement of
vortices on the downstream edge of the cavity, which creates pressure waves. These pressure
waves travel upstream and perturb the shear layer near the shedding region in a feedback-loop
manner. The critical aspect of this flow is of aeroacoustical matter. The periodic variation of
the pressure at the downstream edge due to vortex impingement induces structural vibrations
which can result in a premature fatigue for the Reynolds number under consideration here. The
animations help in understanding the sequence of events borne by the structure of the cavity.

Journal of Turbulence 1 (2000) 011 (http://jot.iop.org/) 19


On coherent-vortex identification in turbulence

Primary vortices

Necklace vortices

Secondary vortices

JoT 1 (2000) 011


Figure 13. Top view of the main coherent vortices in a transonic flow past a
Downloaded By: [WALDO - Full Library] At: 19:20 26 July 2010

rectangular cavity. The vortical structures are identified using the Q criterion.

The side view (figure 12(b)) highlights the periodic generation of dense packets of vortices.
The vortices, shed at the upstream edge, appear to trigger this intense vortical activity, as
figure 12(a) shows. An interesting pattern of this flow is the creation of secondary Λ vortices at
a subharmonic frequency. Figure 13 is fairly representative of the occurrence of such structures.
From figure 12(a), it is possible to see that the secondary vortices are centred slightly on the inside
of the trailing legs of the primary vortices. Since the existence of secondary spanwise vortices in
the early stage of the shedding could not be detected, it has been speculated that their formation
is caused by the strong ejections induced by the trailing legs of primary vortices, not by a pairing
process. Thus, it could explain the similarity of figure 13 with the picture of a hairpin vortex
created by blowing from a slot. A description of such a flow can be found in Singer [39], which
has been used here for the nomenclature of figure 13. The necklace vortex is also associated with
localized upward fluid motion and can be found in junction flows. This structure is generated
upstream of the disturbance. In our case, only that half which is located outside the cavity is
depicted. The other half is probably hidden by the vortical activity around the centre of the
cavity. The organization and the periodicity of this flow is remarkable, and figure 12(c) gives an
idea of what the downstream wall experiences during a short period ( 20 ms). One should be
reminded that the upstream flow does not provide realistic perturbations of a turbulent boundary
layer, and it can be assumed that the streaky structure of wall turbulence might enhance the
three dimensionality of the flow. However, the good prediction of the fundamental frequencies
suggests that the large-scale structures, in a more realistic simulation, should retain most of their
coherence.

7. Conclusion

The Q criterion has been used to investigate a large panel of coherent vortices in DNS and
LES. Its superiority over ω and low-pressure criteria has been checked in representative free and
wall-bounded turbulent flows. The insensitiveness of Q to mean shear and large-scale variations
of pressure allows a precise eduction of agglomerations of quasi-longitudinal vortices in the
inner layer of a channel flow as well as streamwise vortices in the vicinity of Kelvin–Helmholtz
instabilities in a mixing layer, for instance. The good agreement with the λ2 criterion is expected

Journal of Turbulence 1 (2000) 011 (http://jot.iop.org/) 20


On coherent-vortex identification in turbulence

according to the mathematical relation between the two criteria. The main concern, when using
the Q criterion, is obviously the setting of the threshold. The easiest way, but also the most
subjective, consists of a systematic comparison of the pictures of the same field obtained by
different criteria, using a large range of thresholds for each. A second option has been proposed,
using the contribution of educed coherent vortices to vorticity fluctuations. The method requires
a simple conditional sampling based on Q, and has been applied to a turbulent channel flow. In
this flow, a critical threshold can be found which gives a maximum contribution to each vorticity
component. Choosing a threshold close to this critical value ensures the eduction of the most
energetic structures in terms of vorticity fluctuations.
The capabilities of flow visualizations to understand the dynamics of coherent structures have
been illustrated by animations of a grooved channel flow and separated flows. Regarding the latter
flows, the animation in figure 11 shows the generation of large-scale Λ vortices downstream of the

JoT 1 (2000) 011


reattachment region of backward-facing step flows. The existence of such structures explains
Downloaded By: [WALDO - Full Library] At: 19:20 26 July 2010

the slow recovery of the boundary layer in this region towards a canonical state. The other
animations (see figure 12) concern a transonic flow past a rectangular cavity. Kelvin–Helmholtz
vortices shed at the upstream edge of the cavity are rapidly distorted into Λ vortices by the
spanwise variation of the mean shear imposed by the side-walls. Different views have helped to
track the creation of secondary vortices in the wake of the trailing legs of the Λ vortices. Such
animations might be helpful in setting up a control strategy to reduce the vibrations induced by
the impingement of vortices on the cavity walls.

Acknowledgments

We are indebted to Patrick Bégou, Pierre Comte, Eric Lamballais, Jorge Silvestrini and Sepand
Ossia for their contribution (computational encyclopedia). Computational time has been freely
allocated by CNRS (Institut du Développement et des Ressources en Informatique Scientifique,
Paris).

References
[1] Lesieur M 1997 Turbulence in Fluids (Dordrecht: Kluwer)
[2] Comte P, Silvestrini J and Bégou P 1998 Streamwise vortices in large eddy simulation of mixing layers Eur. J. Mech. B
17 615–37
[3] Robinson K S 1991 The kinematics of turbulent boundary layer structure Technical Memorandum TM 103859 NASA
[4] Hunt J C R, Wray A A and Moin P 1988 Eddies, stream, and convergence zones in turbulent flows Report CTR-S88,
Center For Turbulence Research
[5] Jeong J and Hussain F 1995 On the identification of a vortex J. Fluid Mech. 285 69–94
[6] Larchevêque M 1990 Equation de monge–ampère et écoulements incompressibles bidimensionnels C. R. Acad. Sci.
Paris 311 33–6
[7] Weiss J 1991 The dynamics of enstrophy transfer in two-dimensional hydrodynamics Physica D 48 273–94
[8] Basdevant C and Philipovitch T 1994 On the validity of the ‘Weiss criterion’ in two-dimensional turbulence Physica D 73
17–30
[9] Chong M S, Perry A E and Cantwell B J 1990 A general classification of three-dimensional flow field Phys. Fluids A 2
765
[10] Cucitore R, Quadrio M and Baron A 1999 On the effectiveness and limitations of local criteria for the identification of a
vortex Eur. J. Mech. B 18 261–82
[11] Ossia S 2000 La dynamique des échelles infrarouges en turbulence isotrope incompressible PhD Thesis Institute National
Polytechnique de Grenoble
[12] Lesieur M, Ossia S and Métais O 1999 Infrared pressure spectra in two- and three-dimensional isotropic incompressible
turbulence Phys. Fluids 11 1535–43

Journal of Turbulence 1 (2000) 011 (http://jot.iop.org/) 21


On coherent-vortex identification in turbulence

[13] Ducros F, Comte P and Lesieur M 1996 Large eddy simulation of transition to turbulence in a boundary layer developing
over a flat plate J. Fluid Mech. 326 1–36
[14] Delcayre F 1997 Topology of coherent vortices in the reattachment region of a backward-facing step Proc. 11th Symp.
on Turbulent Shear Flows (Grenoble, France) vol 3, pp 24–6
[15] Delcayre F 1999 Etude par simulation des grandes échelles d’un écoulement décollé: la marche descendante PhD Thesis
Institute National Polytechnique de Grenoble
[16] Le H, Moin P and Kim J 1997 Direct numerical simulation of turbulent flow over a backward-facing step J. Fluid Mech.
330 349–74
[17] Leonard B P 1988 Simple high-accuracy resolution program for convective modelling of discontinuities Int. J. Numer.
Methods Fluids 8 1291–308
[18] Lesieur M and Métais O 1996 New trends in large eddy simulations of turbulence Ann. Rev. Fluid Mech. 28 45–82
[19] Lamballais E 1996 Simulations numériques de la turbulence dans un canal plan tournant PhD Thesis Institute National
Polytechnique de Grenoble
[20] Dubief Y 2000 Simulations des grandes echelles de couches limites et de sillages PhD Thesis Institute National

JoT 1 (2000) 011


Polytechnique de Grenoble
[21] MacCormack R W 1969 The effect of viscosity in hypervelocity impact cratering AIAA Paper 69 354
Downloaded By: [WALDO - Full Library] At: 19:20 26 July 2010

[22] Gottlieb D and Turkel E 1976 Dissipative two-four methods for time-dependent problems J. Comp. Phys. 30 703–23
[23] Tracy M B and Plentovich E B 1997 Cavity unsteady-pressure measurements at subsonic and transonic speed Technical
Paper 3669 NASA
[24] Siggia E D 1981 Numerical study of small-scale intermittency in three-dimensional turbulence J. Fluid Mech. 107
375–406
[25] Vincent A and Meneguzzi M 1994 The dynamics of vorticity tubes in homogeneous turbulence J. Fluid Mech. 258
245–54
[26] Kida S and Miura H 1998 Identification and analysis of vortical structures Eur. J. Mech. B 17 471–88
[27] Comte P, Lesieur M and Lamballais E 1992 Large- and small-scale stirring of vorticity and a passive scalar in a 3-d temporal
mixing layer Phys. Fluids A 4 2761–78
[28] Bernal L P and Roshko A 1986 Streamwise vortex structure in plane mixing layer J. Fluid Mech. 170 499–525
[29] Dimotakis P E and Brown G L 1976 The mixing layer at high Reynolds number: large-structure dynamics and entrainment
J. Fluid Mech. 78 535–60
[30] Jiménez J and Pinelli A 1999 The autonomous cycle of near-wall turbulence J. Fluid Mech. 389 335–59
[31] Theodorsen T 1952 Mechanism of turbulence Proc. 2nd Midwestern Conf. on Fluid Mechanics (Ohio State
University, OH) Bull. No. 149
[32] Dubief Y, Comte P and Lesieur M 2000 Response of near-wall vortical structures to a grooved surface, in preparation
[33] Jeong J, Hussain J, Schoppa W and Kim J 1997 Coherent structures near the wall in a turbulent channel flow J. Fluid
Mech. 332 185–214
[34] Dubief Y, Djenidi L and Antonia R A 1997 The measurement of ∂u/∂y in a turbulent boundary layer over a riblet surface
Int. J. Heat Fluid Flow 18 183–7
[35] Kim J, Moin P and Moser R 1987 Turbulent statistics in fully developed channel flow at low Reynolds number J. Fluid
Mech. 177 133–66
[36] Antonia R A 1981 Conditional sampling in turbulence measurement Ann. Rev. Fluid Mech. 13 131–56
[37] Smith C R and Walker J D A 1995 Turbulent Wall-Layer Vortices (Dordrecht: Kluwer) pp 235–90
[38] Rockwell D 1984 Oscillations of impinging shear layers AIAA J. 21 645–64
[39] Singer B A 1996 Characteristics of a young turbulent spot Phys. Fluids 8 509–21

Journal of Turbulence 1 (2000) 011 (http://jot.iop.org/) 22

You might also like