International Journal of Heat and Fluid Flow 28 (2007) 638–652

www.elsevier.com/locate/ijhff

Vortex identification: New requirements and limitations

Václav Kolář *

Institute of Hydrodynamics, Academy of Sciences of the Czech Republic, 166 12 Prague 6, Czech Republic

Received 23 October 2006; received in revised form 3 March 2007; accepted 29 March 2007
Available online 24 May 2007

Abstract

Firstly, a brief survey dealing with popular vortex-identification methods is presented. The most widely used local criteria (applied
point by point) – sharing a basis in the velocity-gradient tensor $u – are treated more thoroughly to recall their underlying ideas and
physical aspects. A large number of recent papers have pointed out various applicability limitations of these popular schemes and for-
mulated (explicitly or implicitly) new general requirements, for example: validity for compressible flows and variable-density flows, deter-
mination of the local intensity of swirling motion, vortex-axis identification, non-local properties, ability to provide the same results in
different rotating frames, etc. Other quite natural requirements are pointed out and added to those already mentioned. Secondly, the
vortex-identification outcome of the proposed triple decomposition of the relative motion near a point is presented. The triple decom-
position of motion – based on the extraction of a so-called ‘‘effective’’ pure shearing motion – has been motivated by the fact that vor-
ticity cannot distinguish between pure shearing motions and the actual swirling motion of a vortex. This decomposition technique results
in two additive vorticity parts (and, analogously, in two additive strain-rate parts) of distinct nature, namely the shear component and the
residual one. The residual vorticity represents a direct measure of the actual swirling motion of a vortex. The new kinematic vortex-iden-
tification method is discussed on the background of previous methods and general vortex-identification requirements (illustrative exam-
ples are included).
 2007 Elsevier Inc. All rights reserved.

Keywords: Decomposition of motion; Flow kinematics; Vortex identification; Vortex-identification criteria; Vortical structures; Vorticity decomposition

1. Introduction cles around a common centre. These intuitive definitions

often depict a vortex in terms of closed or spiralling stream-
Flow modelling and numerical simulation of turbulent lines or pathlines, local pressure minima, and isovorticity
and/or complex vortical flows still lack a generally accept- contours and surfaces. However, spiralling streamlines or
able definition of a vortex though the understanding of pathlines are obtained just for an observer moving with
vortex dynamics (i.e. generation, evolution, interaction, the vortex to be identified, and the existence of a local pres-
and decay of vortical structures) should be based on objec- sure minimum does not guarantee the existence of a vortex
tive and unambiguous detection schemes. There is no (and vice versa).
doubt that the physical reasoning for these schemes plays Vorticity tensor, as a Galilean invariant quantity (i.e.
a crucial role. independent of the translational velocity of an observer)
What is a vortex? Though a vortex intuitively represents expressing an average angular velocity of fluid elements,
a distinct flow phenomenon, the answer to this question is appears as one of the most natural choices for a vortex-
neither simple nor unique, e.g. Lugt (1983). He presents identification criterial measure. However, it has been
one of the intuitive definitions of a vortex as follows: a vor- recently emphasized by many authors that vorticity is not
tex is the rotating motion of a multitude of material parti- suitable for the identification of a vortex as it cannot distin-
guish between pure shearing motions and the actual swirl-
*
Tel.: +420 233109095; fax: +420 233324361. ing motion of a vortex (Jeong and Hussain, 1995; Kida
E-mail address: [email protected] and Miura, 1998; Cucitore et al., 1999). This property of

0142-727X/$ - see front matter  2007 Elsevier Inc. All rights reserved.
doi:10.1016/j.ijheatfluidflow.2007.03.004
 V. Kolář / Int. J. Heat and Fluid Flow 28 (2007) 638–652 639

Nomenclature

A additive part of $u k1 ; k2 ; k3 ordered eigenvalues, k1 P k2 P k3

C vortex-strength parameter in (29) m kinematic viscosity
D jet-nozzle diameter q density
K shearing-strength parameter in (28) r shearing parameter in (28)
L, L1, L2 velocity magnitudes, see Fig. 6 x vorticity-tensor component in 2D, see (13) and
p pressure (15)
P arbitrary point of the flow field X, Xij vorticity tensor, antisymmetric part of $u
Q orthogonal linear transformation
Q second invariant of $u Subscripts and superscripts
r radius BRF basic reference frame
R third invariant of $u , subscript comma denoting differentiation (e.g.
s principal rate of strain in 2D, see (13) and (14) ui,j  oui/oxj)
S, Sij strain-rate tensor, symmetric part of $u EL elongation
t time SH shearing motion, shear component
u, ui velocity vector SHEARING shearing motion described by (28)
u, v, w components of the velocity vector RES residual component
UC crossflow velocity RR rigid-body rotation
Vtangential tangential velocity VORTEX vortical motion described by (29)
x; y; z coordinates x; y; z partial derivatives (e.g. ux  ou/ox)
T transpose
Greek symbols * new Cartesian coordinate (after rotation)
a; b; c rotation angles
a1, a2 characteristic angles in Fig. 2 associated with Other symbols and special functions
different vorticity components k  k absolute tensor value, the norm kGk of any ten-
b1, b2 characteristic angles in Fig. 2 associated with sor G is defined by kGk = [tr(GGT)]1/2
different strain-rate components $ nabla operator
C circulation MAX (. . .) returns the maximum from a given set
D vortex-identification criterial quantity defined MIN (. . .) returns the minimum from a given set
by (2) sgn . . . returns ±1 according to the sign of a given argu-
k eigenvalue ment

vorticity leads to the misrepresentation of vortex geometry. numerical simulation (DNS) and large-eddy simulation
Moreover, vortex geometry depends on the vorticity (LES) of turbulent flows has become particularly important
threshold applied. (e.g. Dubief and Delcayre, 2000; Lesieur et al., 2003; Garcı́a-
The most widely used local methods for vortex identifica- Villalba et al., 2006). In the following section the most pop-
tion are based on the analysis of the velocity-gradient tensor ular identification criteria are shortly described. Needless to
$u = S + X, its symmetric and antisymmetric parts, strain- say, these criteria are Galilean invariant.
rate tensor S and vorticity tensor X, respectively, and the In Section 3, some other – more recent – identification
three invariants of $u. Truesdell (1953) was the first to methods and discussions about vortex definition are sur-
describe a quantitative measure of rotation in terms of X veyed with a particular emphasis on the applicability limi-
and S normalizing the magnitude of the vorticity tensor by tations of various schemes. New general requirements for
the magnitude of the strain-rate tensor (the so-called kine- vortex identification are summarized and other natural
matic vorticity number). However, this measure does not requirements are pointed out.
discriminate between vortices with small vorticity in a flow In Sections 4 and 5, the vortex-identification outcome of
with small shear and vortices with large vorticity in a flow the proposed triple decomposition of the relative motion
with large shear (Jeong and Hussain, 1995; Geers et al., near a point is presented and discussed on the background
2005). The analysis of $u provides a rational basis for vortex of previous methods and general identification requirements.
identification and the general classification of 3D flow fields Illustrative examples of this novel approach are included.
(Chong et al., 1990). The application of complex measures
derived from $u has already revealed its importance for 2. The most widely used local vortex-identification criteria
the description of large-scale vortical structures in turbulent
free shear flows as well as turbulent boundary-layer flows. A vortex obviously represents a non-local flow phenom-
The identification of coherent vortices on the basis of direct enon in space and time. However, as noted by Chakraborty
640 V. Kolář / Int. J. Heat and Fluid Flow 28 (2007) 638–652

et al. (2005), the presence of viscosity in real fluids results in minimum in a plane across the vortex requires two positive
the continuity of the kinematic features of the flow field. eigenvalues of the tensor p,ij.
Consequently, a reasonable estimate of some non-local By removing the unsteady irrotational straining and vis-
vortical features can be inferred from the local (pointwise, cous effects from the strain-rate transport equation (3) one
applied point by point) methods and characteristics. yields the vortex-identification criterion for incompressible
fluids in terms of two negative eigenvalues of S2 + X2. The
2.1. Q-criterion existence of a local pressure minimum is neither a sufficient
nor a necessary condition for the presence of a vortex in
Hunt et al. (1988) identify vortices of an incompressible general, and the two removed terms from the Eq. (3) are
flow as connected fluid regions with a positive second found to be the main cause of this inaccuracy. Finally, a
invariant of $u (in tensor notation below the subscript vortex is defined as a connected fluid region with two neg-
comma denotes differentiation) ative eigenvalues of S2 + X2. Since the tensor S2 + X2 is
symmetric, it has real eigenvalues only. If these eigenvalues
1 1 1 are ordered as follows, k1 P k2 P k3, the vortex-identifica-
Q  ðu2i;i  ui;j uj;i Þ ¼  ui;j uj;i ¼ ðkXk2  kSk2 Þ > 0;
2 2 2 tion criterion is equivalent to the resulting condition k2 < 0.
ð1Þ
3. Other vortex-identification approaches,
that is, as the regions where the vorticity magnitude pre- new requirements and limitations
vails over the strain-rate magnitude. The norm (or absolute
tensor value) kGk of any tensor G is defined by The papers mentioned in the preceding section have
kGk = [tr(GGT)]1/2. In addition, the pressure in the vortex stimulated new and significant research activity during last
region is required to be lower than the ambient pressure. decade. Some other, more recent, vortex-identification
methods and discussions on vortex definition are briefly
2.2. D-criterion surveyed below. New requirements for vortex-identifica-
tion schemes are summarized at the end of this section.
Dallmann (1983), Vollmers et al. (1983), and Chong Kida and Miura (1998), similarly to Jeong and Hussain
et al. (1990) define vortices as the regions in which the (1995), point out that a swirling motion is not always asso-
eigenvalues of $u are complex (a pair of complex-conjugate ciated with a sectional pressure minimum. This aspect has
eigenvalues occurs) and the streamline pattern is spiralling motivated these authors to improve the pressure-minimum
or closed in a local reference frame moving with the point. scheme by imposing a kinematic swirl condition and they
Such points can be viewed within the critical-point theory – have constructed the central axes of vortices by the sec-
on a plane spanned by the complex eigenvectors – as ellip- tional-swirl-and-pressure-minimum scheme for the identifi-
tic ones (focus or centre). For incompressible flows, this cation of the low-pressure vortices in freely decaying
requirement reads homogeneous turbulence. As noted by Wu et al. (2005),
 3  2 an extremal condition adopted by Kida and Miura (1998)
Q R for low-pressure vortices is necessary for identifying a
D¼ þ > 0; ð2Þ
3 2 single line as the vortex axis.
In the eduction of longitudinal vortices in wall-turbu-
where Q and R are the invariants of $u, Q is given by (1),
lence, Jeong et al. (1997) employ a non-zero threshold for
R  Det(ui,j). Q and R play a key role in the reduced (due
k2 contrary to the k2-criterion. This practical aspect is
to incompressibility) characteristic equation for the eigen-
emphasized by Kida and Miura (1998), as well as by Lin
values k of $u: k3 + Qk  R = 0.
et al. (1996) in their study of a neutrally stratified planetary
boundary-layer flow.
2.3. k2-criterion For compressible flows, the Q-criterion suffers from
ambiguity as it offers two ways of extension which have dif-
The approach of Jeong and Hussain (1995) is formu- ferent physical meaning, the second invariant of $u and the
lated on dynamic considerations, namely on the search 2 2
quantity ðkXk  kSk Þ=2. Both choices cannot basically
for a pressure minimum across the vortex. By taking the avoid dependence on a non-zero divergence. The k2-crite-
gradient of the Navier–Stokes equations and by decompos- rion has been originally tailored for incompressible flows.
ing it into symmetric and antisymmetric parts they derive In the case of compressible fluids, additional terms occur
the well-known vorticity transport equation and the on the LHS of the Eq. (3) as shown by Cucitore et al.
strain-rate transport equation. The latter reads (1999) who examined the k2-criterion. The use of S2 + X2
DS ij 1 as an approximation of the pressure Hessian p,ij for com-
 mS ij;kk þ Xik Xkj þ S ik S kj ¼  p;ij ; ð3Þ pressible fluids requires discarding other terms besides the
Dt q
unsteady irrotational straining and viscous effects origi-
where the pressure Hessian p,ij contains information on nally removed from the strain-rate transport Eq. (3) valid
local pressure extrema. The occurrence of a local pressure for incompressible flows only. These additional terms are
 V. Kolář / Int. J. Heat and Fluid Flow 28 (2007) 638–652 641

related to a non-zero divergence and non-zero density Xiong et al. (2004) define a vortex as follows: if all the
gradients. fluid particles within the area in the plane normal to the
Another aspect worth mentioning is that all the methods vorticity direction have the rotational velocity components
based on the analysis of $u are pointwise providing local in the same direction around any point in the area, the area
vortex-identification criteria. This aspect has led Cucitore is identified as a part of a vortex. They clearly state that the
et al. (1999) to introduce the concept of non-locality for validation of the vortex definition using flow examples is
determining the vortices as structures. They suggest a not rigorous because of a prejudice of expecting what is a
non-local, Galilean-invariant identification technique. The priori a vortex. Examples only suffice to invalidate but
reason is that the use of local procedures often selects a not support an idea. Xiong et al. (2004) emphasize that a
particular privileged direction which is considered as the method of vortex identification should be evaluated by
vortex axis. For example, the k2-criterion captures the pres- whether there is a clear physical meaning in the vortex def-
sure minimum in a plane across the vortex but not along it. inition and that the vortex-identification method should be
The non-local criterion of Cucitore et al. (1999) is based on chosen by which characteristics the definition focuses on.
the intuitive notion that the particles inside a vortical struc- The fact that all of the most widely used vortex defini-
ture show small variations in their relative distance even tions are not objective relative to an arbitrarily rotating ref-
when following completely different trajectories. erence frame has motivated Haller (2005) to develop an
Zhou et al. (1999) use the imaginary part of the complex objective frame-independent definition of a vortex based
eigenvalue of $u to visualize vortices and to quantify the on Lagrangian stability considerations. An incompressible
strength of the local swirling motion inside the vortex vortex is defined as a set of fluid trajectories along which
(the so-called swirling-strength criterion). Their method is the strain acceleration tensor is indefinite over directions
based on the D-criterion, however, it identifies not only of zero strain. This definition should help in situations
the vortex region (equivalently as the D-criterion), but also where there is an unclear choice for a reference frame
the local strength and the local plane of swirling. For a sim- (for example, vortical flows in rotating tanks).
ilar approach based on complex eigenvalues, see Berdahl Zhang and Choudhury (2006) formulate a Galilean-
and Thompson (1993). The associated instantaneous- invariant scheme, eigen helicity density, based on the con-
streamline analysis is further developed in Chakraborty cept of Galilean-variant helicity density adopted earlier by
et al. (2005). Their criterion enhances the swirling-strength Levy et al. (1990) for the graphical representation of 3D
criterion by including a local approximation of the non- flow fields that contain concentrated vortices. Their scheme
local property proposed by Cucitore et al. (1999), requiring shows a promise to identify vortices in 3D compressible,
that the swirling material points inside a vortex have variable-density flows governed by the baroclinic term
bounded separation remaining small. Moreover, Chakra- (i.e. the normalized cross product of a density gradient
borty et al. (2005) study the relationship between local and pressure gradient) in the vorticity equation.
identification schemes. They show that all of the discussed The above mentioned new requirements for vortex-iden-
local criteria, given the proposed usage of threshold, result tification schemes and their underlying criterial quantities
in a remarkable vortex similarity. can be summarized as follows:
The evaluation of vortex-identification criteria by com-
paring the resulting vortex patterns in numerically simu- • validity for compressible flows,
lated complex vortical flows cannot lead to a final choice • validity for variable-density flows,
for the correct criterion as the judgment depends on the • avoidance of the subjective choice of threshold in the
adopted intuitive and subjective concept of the investigator vortex-boundary identification,
on what should be called a vortex, as pointed out by Wu • determination of the local intensity of the swirling
et al. (2005). Instead of numerical examples they make an motion (to describe the inner structure of a vortex),
analytical diagnosis of four local criteria, demonstrated by • vortex-axis identification,
the Burgers and Sullivan vortex, indicating that Q-criterion • allowance for an arbitrary axial strain,
and k2-criterion may cut a connected vortex into broken • non-local properties,
segments at locations with strong axial stretching. Conse- • ability to provide the same results in different rotating
quently, the following requirements are emphasized: a gen- frames (to fulfil material objectivity or frame indifference,
erally applicable vortex definition should be able to identify i.e. both translational and rotational independence, see
the vortex axis and allow for an arbitrary axial strain. Note e.g. Leigh, 1968).
that all of the local criteria described in Section 2 identify
just a vortex region through criterial inequalities without Though most of these requirements are intuitively clear,
specifying the vortex axis inside this region. Wu et al. some of them may need further justification. For example,
(2005) state that an equality is necessary for identifying a the allowance for an arbitrary axial strain has become a
single line as the vortex axis. Recall that the extremal con- subject of recent debate (Chakraborty et al., 2006; Wu
dition of Kida and Miura (1998) enables us to find the vor- et al., 2006).
tex skeleton by tracing the lines of the sectional pressure As to the local intensity of the swirling motion, engineer-
minimum, provided that a swirl condition is satisfied. ing practice may frequently need this local quantity to be
642 V. Kolář / Int. J. Heat and Fluid Flow 28 (2007) 638–652

easily integrable across the vortex region in order to obtain mechanics. The aim of the TDM is to decompose an arbi-
the integral strength of a vortex. However, the application trary instantaneous state of the relative motion near a
of conventional circulation C (calculated as a surface quad- point into three elementary motions, each described by
rature of vorticity) for this purpose is in fact misleading as an additive part of $u with a distinct tensor character,
the vorticity is misrepresenting the local intensity of the explicitly including a pure shearing motion. Therefore,
actual swirling motion of a vortex. For example, one the present decomposition method is – including its vor-
obtains a net circulation for the region of a simple linear tex-identification outcome – based on the extraction of a
shearing motion due to a net vorticity. so-called ‘‘effective’’ pure shearing motion. The TDM is
Furthermore, another requirement – much more trivi- expressed through the corresponding triple decomposition
ally looking than the integral strength of a vortex – is swirl of $u introduced in Kolář (2004).
orientation. Vorticity may answer the question regarding To discuss all of the main aspects of the TDM in 3D
the swirl orientation in simple problems, however, the local flows is far beyond the scope of this contribution and the
angle between the vortex-axis tangent and the vorticity vec- research in this regard is incomplete due to its complexity
tor may reach large values due to a strong shearing aligned (see also the final remark at the end of this subsection).
with the vortex axis (e.g. streamwise vortical structures in a However, its planar version – including the application to
turbulent boundary layer). This requirement becomes par- vortex identification – is very illustrative. Particularly, a
ticularly important in complex 3D vortical flows subjected straightforward comparison with the most widely used vor-
to high shear. tex-identification methods is provided for planar flows in
As noted by Kida and Miura (1998), it is impossible for Section 5.
the isosurface representation of a scalar field (applied in The qualitative model of three elementary motions of
vortex identification) to distinguish between individual vor- the TDM is depicted in Fig. 1. The deformable fluid ele-
tical structures. This is especially the problem of homo- ment in Fig. 1 consists of discrete undeformable material
geneous turbulence. To avoid ambiguity and vortex points in terms of which the local rate of deformation is
overlaps, the explicit vortex-axis requirements are pro- described through their relative motion. The material point
posed below. represents – in the present context – ‘‘much less than a fluid
These additional requirements are quite natural and, element’’ and generally allows translation and rotation
therefore, added to those already mentioned: only. A pure shearing motion near a point is interpreted
in terms of the parallel relative motion of non-rotating
• swirl orientation, undeformed shearing elements – planes, lines, or points
• determination of the (integral) vortex strength, (depending on flow complexity in 3D).
• vortex-axis requirements: existence and uniqueness for
each connected vortex region (to avoid ambiguity and
vortex overlaps).

4. Triple decomposition of the relative motion near a point

and vortex identification

4.1. Triple decomposition of the relative motion near a point

The conventional double decomposition of motion near deformable fluid

a point into a pure irrotational straining motion along the element as a set of pure shearing
principal axes of the rate of strain tensor (generally includ- discrete undeformable of planes (lines)
ing a uniform dilatation) and a rigid-body rotation material points
expressed by $u = S + X has a long history and stems (initial state)
from the Cauchy–Stokes decomposition theorem (first
explicitly stated by Stokes, 1845, according to Truesdell
and Toupin, 1960). Basic kinematics in this regard can be
found in many textbooks (e.g. Batchelor, 1967; Panton,
1984).
The triple decomposition of the relative motion near a
point (TDM) has been motivated by the fact that vorticity
cannot distinguish between pure shearing motions and the
actual swirling motion of a vortex. Analogously, strain rate
cannot distinguish between straining motions and shearing
rigid-body rotation irrotational straining
motions. These problems indicate that the double decom-
position may not satisfy all of today’s needs of fluid Fig. 1. Qualitative model of three elementary motions of the TDM.
 V. Kolář / Int. J. Heat and Fluid Flow 28 (2007) 638–652 643

A pure shearing motion in Fig. 1 is not a mere combina- ui;j ¼ uj;i ðfor all i; jÞ ð7Þ
tion of an irrotational straining motion with a rigid-body with the implication
rotation as in the case of the double decomposition. This
ui;j uj;i 6 0 ðno summationÞ
fact can be easily checked through the rotational change
of material points which remains zero for a pure shearing are fulfilled in an arbitrary reference frame rotated under an
motion within the proposed qualitative model of the orthogonal transformation, the condition (5) for the purely
TDM. The rotational change of material points is just asymmetric tensor is satisfied in a suitable reference frame
the quantity reflecting the actual swirling motion of a vor- only. In this reference frame, a straightforward consequence
tex: note that both an irrotational straining and a pure of (5) is that the strain-rate and vorticity magnitudes are in
shearing motion do not contribute to this rotational change strict equilibrium, component by component, as the rela-
(at least according to the present approach). tion jSijj = jXijj holds for all i, j. Further, the condition (5)
Although we focus below on planar flows, it is conve- implies the equality kSk = kXk valid in an arbitrary refer-
nient to introduce the quantitative TDM algorithm using ence frame (it does not hold vice versa). The condition (5)
3D formalism. The TDM distinguishes three different ele- defines – at least within the present paper – the purely asym-
mentary motions near a point, each defined in terms of a metric tensor with respect to second-order tensors and,
distinct tensor structure. The TDM reads moreover, it defines a general structure of a pure shearing
ru ¼ ðruÞEL þ ðruÞRR þ ðruÞSH ; ð4Þ motion with respect to flow kinematics near a point.
The above stated decomposition requirements are
where an irrotational straining motion is given by the sym- apparently insufficient. Considering ($u)SH, the condition
metric tensor ($u)EL (subscript ‘‘EL’’ denotes elongation), (5) is a necessary condition only as it characterizes just a
a rigid-body rotation (denoted by ‘‘RR’’) is given by the pure shearing motion without specifying the label ‘‘effec-
antisymmetric tensor ($u)RR, and an effective pure shearing tive’’. What we really need is a physically well-justified
motion (denoted by ‘‘SH’’) is described by the ‘‘purely algorithm leading to a unique decomposition. The ‘‘shear
asymmetric tensor form’’ ($u)SH its components ui,j fulfill- tensor’’ satisfying the condition (5) can be easily generated
ing in a suitable reference frame by the natural and straightforward decomposition scheme
which is applicable to an arbitrary reference frame
ui;j ¼ 0 OR uj;i ¼ 0 ðfor all i; jÞ ð5Þ 0 1
with the implication ux uy uz    
B C residual shear
ui;j uj;i ¼ 0 ðno summationÞ: ru  @ vx vy vz A ¼ þ ;
tensor tensor
wx wy wz
The condition (5) requires zeros on the leading diagonal
ð8aÞ
and, moreover, at least one off-diagonal element from each
pair must be zero as well. If the frame showing the tensor in where the residual tensor is given by
form (5) exists, the tensor is – by definition – purely asym- 0 1
  ux ðsgnuy ÞMINðjuy j; jvx jÞ 
metric. The tensor structures represented by (5) are, for residual B C
¼ @ ðsgnvx ÞMINðjuy j; jvx jÞ vy  A:
example, tensor
  wz
0 1 0 1 0 1
0  0 0 0 0 0 0 0 ð8bÞ
B C B C B C
@ 0 0 0 A ; @  0 0 A ; @  0  A; In (8a,b) the following simplified notation is employed:
0 0 0  0 0 0 0 0 u; v; w are velocity components, subscripts x; y; z stand
0 1 0 1 0 1
0 0  0  0 0 0  for partial derivatives. The remaining two non-specified
B C B C B C pairs of off-diagonal elements of the residual tensor in
@  0 0 A ; @ 0 0  A ; @ 0 0 0 A;
(8b) are constructed strictly analogously as the specified
0 0 0 0 0 0 0  0 one, each pair being either symmetric or antisymmetric.
0 1 0 1 0 1
0 0 0 0   0  0 Just a simple quantitative example:
B C B C B C 0 1 0 1 0 1
@ 0 0 A; @0 0 0 A; @0 0  A: 1 15 17 1 3 17 0 12 0
  0 0  0  0 0 B C B C B C
@ 3 8 14 A ¼ @ 3 8 14 A þ @ 0 0 0 A:
While the symmetric-tensor condition (specifying an 26 14 5 17 14 5 9 0 0
irrotational straining motion)
If considered separately, an arbitrary shearing motion
ui;j ¼ uj;i ðfor all i; jÞ ð6Þ should be, by (5), recognized in a suitable reference frame
with the implication as a third elementary part of the TDM. Finding a mini-
ui;j uj;i P 0 ðno summationÞ mum of the norm of the residual tensor within the scheme
(8a,b) by changing the reference frame under an orthogo-
and the antisymmetric-tensor condition (specifying a rigid- nal transformation guarantees to satisfy this necessary
body rotation) requirement and leads to the correct frame choice (to
644 V. Kolář / Int. J. Heat and Fluid Flow 28 (2007) 638–652
 
 shear 
perform the desired decomposition). This condition says 
zing  , leads to an ambiguous decomposition
tensor 
that the effect of the extraction of the shear tensor from
algorithm, even for 2D fluid motion. This criterion takes
$u is maximized.
into account only the magnitude of a pure shearing motion,
The TDM is closely associated with the so-called basic
not its effect (i.e. its impact on $u after the extraction) in
reference frame (BRF) where it is performed as the three
full which includes the shearing structure and orientation
separate parts of $u are generated just here through the
decomposition scheme (8a,b), where the residual tensor is of shearing elements (planes, lines, or points). Further, by
to be further decomposed into its symmetric and antisym- changing the present frame-choice criterion based on
jS 12 X12 j þ jS 23 X23 j þ jS 31 X31 j quantitatively to the other
metric parts representing ($u)EL and ($u)RR. However, the
extreme value, namely by choosing the shear-free frame
decomposition results generated in the BRF are valid for
by requiring jS 12 X12 j þ jS 23 X23 j þ jS 31 X31 j ¼ 0, the decom-
all other frames rotated (not rotating!) with respect to the
position scheme (8a,b) leads to the well-known double-
BRF under an orthogonal transformation. In the BRF,
decomposition results. In this case, the principal axes of
an effective pure shearing motion is shown in a clearly vis-
S represent the desired shear-free frame (though this frame
ible manner described by the form (5) on condition that the
norm of the residual tensor in (8a,b) is minimized. That is, is not always the only shear-free frame for a given tensor
consistently with the limiting case mentioned in the preced- data).
The TDM algorithm consists of the following three
ing paragraph, on condition that the effect of the extraction
steps (a uniform dilatation does not affect the interaction
of the shear tensor is maximized. This is the reason to label
scalar in the condition (10) and can be removed prior to
the obtained shearing motion with the term ‘‘effective’’.
a further analysis of $u without loss of generality and
Considering the following relation expressed in terms of
applicability to compressible flows):
the strain-rate and vorticity tensors, S and X, and valid
in an arbitrary reference frame,
  Step 1: Determination of the BRF satisfying the condi-
 residual 2 tion (10).
  2
 tensor  þ 4ðjS 12 X12 j þ jS 23 X23 j þ jS 31 X31 jÞ ¼ kruk ; Step 2: Decomposition of $u following the scheme
ð9Þ (8a,b); according to the initial scheme (4), the residual
tensor represents the sum ($u)EL + ($u)RR, and the
where k$uk remains unchanged under an orthogonal trans-
shear tensor represents ($u)SH.
formation and, consequently, the definition condition of
Step 3: Return to the original (e.g. laboratory) reference
the BRF takes the form
frame: any additive part Ai of $u is described in an arbi-
BRF
½jS 12 X12 j þ jS 23 X23 j þ jS 31 X31 j trary reference frame rotated (not rotating!) with respect
ALL FRAMES to the BRF under an orthogonal transformation Q by
¼ MAXð½jS 12 X12 j þ jS 23 X23 j þ jS 31 X31 j Þ:
QAiQT as
ð10Þ !
X X
The condition (10) says that – by changing the reference QðruÞQ ¼ QT
Ai QT ¼ QAi QT : ð11Þ
frame under an orthogonal transformation Q – we are i i
choosing from all frames mutually rotated (not rotating!)
the frame in which the quantity jS 12 X12 j þ jS 23 X23 j þ As mentioned earlier, the BRF is a local frame and its
jS 31 X31 j attains its maximum. In practice, though the num- orientation generally changes from point to point (at a
ber of frames is infinite, the determination of the BRF given instant in time). If we wish to see the whole field of
should be based on a finite set of discrete frame representa- relevant quantities in a common reference frame (e.g. labo-
tions. By considering a reasonable angle resolution (for ratory reference frame) we have to do Step 3.
example, one degree or less) we proceed to the BRF In the planar case treated below in detail, the uniqueness
approximation with reasonably high precision. The trans- of the TDM is obvious.
formation matrix Q for an arbitrarily rotated Cartesian From the viewpoint of the double decomposition, the
coordinate system can be obtained by a sequence of three TDM components of $u are certain products of interaction
rotational transformations, see Appendix A. between S and X. Unlike the two elementary parts of the
The orientation of the BRF (and, consequently, the ori- double decomposition, the three elementary parts of the
entation of an effective pure shearing motion) is a local TDM are mutually conditionally balanced. The term
aspect of the flow field, similarly as the orientation of prin- ($u)SH is responsible for a specific portion of vorticity
cipal axes. However, note that the BRF, unlike the system labelled ‘‘shear vorticity’’ and for a specific portion of
of principal axes of S, is determined simultaneously on the strain rate labelled ‘‘shear strain rate’’. The remaining por-
basis of S and X through the ‘‘interaction scalar’’ tions are called ‘‘residual vorticity’’ and ‘‘residual strain
jS 12 X12 j þ jS 23 X23 j þ jS 31 X31 j. rate’’. For the quantitative relation between the TDM
It is worth mentioning that the application of a qualita- and the double decomposition it holds (subscripts ‘‘SH’’
tively different frame-choice criterion maximizing directly and ‘‘RES’’ by S and X denote their shear and residual
the magnitude of a pure shearing motion, that is, maximi- components, respectively)
 V. Kolář / Int. J. Heat and Fluid Flow 28 (2007) 638–652 645

ru ¼ ðruÞEL þ ðruÞRR þ ðruÞSH dα 1

dα2
a
1 dα1 ~ ω RES
¼ ðruÞEL þ ðruÞRR þ ½ðruÞSH þ ððruÞSH ÞT  dα2 ~ ω SH
2
(dα1 + dα 2) ~ ω
1 T
þ ½ðruÞSH  ððruÞSH Þ 
2
¼ SRES þ XRES þ SSH þ XSH vxdx dx P P
dt
dy
¼ ½SRES þ SSH  þ ½XRES þ XSH  ¼ S þ X: ð12Þ
uydy
In the triple decomposition of $u, S and X are cut down shearing
shearing
in magnitudes to ‘‘share’’ their portions through the third plane
angle axis
term ($u)SH associated with a pure shearing motion as normal
$u = SRES + XRES + ($u)SH. The proposed tensor decom- dβ 1 dβ2
position is applicable to general second-order Cartesian
tensors and may prove its usefulness to problems outside b
fluid mechanics. d β 1 ~ sRES
dβ 2 ~ sSH
In planar incompressible flows, the velocity-gradient (dβ 1 + dβ 2 ) ~ s
tensor $u can be described in an arbitrary reference frame,
in the system of principal axes, and in the above mentioned
dx P P
BRF as follows: dt
vxdx dy 45˚
0 1 0 1PRINCIPAL 0 1BRF
ux uy 0 s x 0 AXES 0 sx 0
B C B C B C uydy
B vx ux 0 C ! B x s 0 C !B 0C
@ A @ A @s þ x 0 A ;

0 0 0 0 0 0 0 0 0
Fig. 2. Geometrical interpretation of 2D fluid motion (incompressible
ð13Þ flow) and the TDM outcome: (a) vorticity components, (b) strain-rate
components.
where s (i.e. the 2D principal rate of strain) and x (i.e. the
vorticity-tensor component in 2D) fulfil the relations
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi By combining the qualitative model of elementary
jsj ¼ 4u2x þ ðuy þ vx Þ2 2; ð14Þ motions of the TDM depicted in Fig. 1 with the situation
in 2D flows according to Fig. 2a, it follows that the resid-
x ¼ ðvx  uy Þ=2: ð15Þ ual vorticity can be interpreted in terms of (twice) the
angular velocity of material points due to the rigid-body
In the BRF, there are two different relative rotational rotation, the shear vorticity in terms of (twice) the average
orientations of uy dy and vx dx, the same and the opposite, angular velocity of the fluid element due to the shearing
see Fig. 2. In Fig. 2a vorticity dominates strain rate. The part of motion only while the conventional vorticity in
characteristic angles, a1 and a2, correspond to the residual terms of (twice) the average angular velocity of the fluid
vorticity xRES (associated with the rigid-body rotation) element.
and shear vorticity xSH (associated with the pure shearing Elementary parts of the TDM and flow patterns for
motion) while their sum is proportional to the total vortic- various flow situations in 2D fluid motion (incompressible
ity x. In Fig. 2b strain rate dominates vorticity. The char- flow) are shown in Fig. 3. All possible flow configurations
acteristic angles, b1 and b2, correspond to the residual near a point for fixed uy and variable vx are depicted in
strain rate sRES and shear strain rate sSH while their sum the corresponding BRFs showing an effective pure shear-
is proportional to the total strain rate s. Note that xRES ing motion in a clearly visible manner. The reference
and xSH have the same signs in the BRF due to the alge- points themselves can be described as critical points and
braic structure of (8a,b). The same holds for sRES and the local flow patterns correspond to the leading terms
sSH. The same signs indicate a non-destructive nature of of a Taylor series expansion for the velocity field in terms
the superimposing construction in Fig. 2. The virtual of space coordinates (Perry and Chong, 1987; Chong
superposition is applicable to infinitesimal motional et al., 1990).
changes only. For both rotational orientations, the magni- In 2D flows, with respect to (13) showing explicitly the
tude of the superimposed shearing motion is given by the desired tensor components in the BRF in terms of s and
difference of the absolute values of uy and vx. In planar x, with respect to the algebraic structure of the decompo-
flows, a non-zero xRES apparently existing only for the sition (8a,b), and assuming jsj P jxj or jsj 6 jxj the follow-
same rotational orientation of uy dy and vx dx, see Fig. 2, ing set of relations can be derived for s and x, and their
excludes the existence of a non-zero sRES existing only for residual and shear components (the case jsj = jxj represents
the opposite rotational orientation of uy dy and vx dx. a simple shear)
646 V. Kolář / Int. J. Heat and Fluid Flow 28 (2007) 638–652

The final remark in this subsection deals with general

3D data. Extensive numerical tests of 3D tensor data indi-
vxdx P rigid-body
dx
rotation
cate that the BRF is unique for local 3D flows near a point
dy without planar or rotational symmetries of $u at the point.
uydy However, even assuming tensor data leading to different
BRFs, the TDM may still be unique provided that different
BRFs lead to the same TDM results. For example, an arbi-
vorticity trary symmetric or antisymmetric tensor is associated with
rigid-body
dominates vxdx P an infinite number of BRFs, each showing the same decom-
dx rotation
strain rate &
dy pure shearing position results. The 3D aspects of the TDM need further
uydy investigation.

4.2. Vortex identification

dx P The vortex-identification outcome of the TDM reads: a

pure shearing
dy vortex is defined as a connected fluid region with a non-
zero magnitude of the residual vorticity. A vortex axis is
uydy
given in a plane across the vortex (but not along it) by
the point of a maximum magnitude of the residual vorticity
vector which is a tangent of the vortex axis at this point. In
P
irrotational the planar-flow context, using (26) and (27), the vortex
strain rate dx straining
dominates vxdx & regions are characterized by the non-zero residual vorticity
dy
vorticity pure shearing for jsj < jxj, see the application (Kolář, 2004) to the plane
uy dy turbulent near wake of two side-by-side square cylinders
(data from Kolář et al., 1997). It should be mentioned that
the residual vorticity in 2D problems does not change its
orientation perpendicular to the flow plane. Therefore
dx P irrotational
vxdx straining
Steps 1 and 3 of the TDM algorithm do not need to be per-
dy
formed. However, considering the strain-rate decomposi-
uy dy tion even for a planar case, if we are interested not only
in the magnitude of strain-rate components as given by
Fig. 3. Elementary parts of the TDM and flow patterns for various flow (18)–(21) but also in the orientation of shearing planes,
situations in 2D fluid motion (incompressible flow). we cannot avoid Steps 1 and 3 of the TDM algorithm.
Fig. 4 shows the quasi-2D secondary-flow vortex geom-
etry, namely the contrarotating vortex pair, in three down-
s ¼ sRES þ sSH ; ð16Þ stream cross-sections of a single jet in crossflow (JICF),
jsj ¼ jsRES j þ jsSH j; ð17Þ data from Kolář et al. (2000). A low threshold applied
sSH ¼ ðsgn sÞjxj for jsj P jxj; ð18Þ in Fig. 4 has no physical but numerical reasons only (to
avoid staircase outer contours of the residual vorticity).
sSH ¼ s for jsj 6 jxj; ð19Þ
Also note that for an arbitrarily chosen threshold level
sRES ¼ s  sSH ¼ ðsgn sÞ½jsj  jxj for jsj P jxj; ð20Þ the region of the residual vorticity forms a subdomain of
sRES ¼ s  sSH ¼ 0 for jsj 6 jxj; ð21Þ the vorticity region. The JICF vortex geometry is on aver-
x ¼ xRES þ xSH ; ð22Þ age nearly circular in terms of (streamwise) xRES some-
what resembling the geometry of an ideal isolated single-
jxj ¼ jxRES j þ jxSH j; ð23Þ
fluid viscous vortex. It can be inferred from Fig. 4 that
xSH ¼ x for jsj P jxj; ð24Þ a prolonged shape of (streamwise) vorticity contours is
xSH ¼ ðsgn xÞjsj for jsj 6 jxj; ð25Þ closely associated with the relatively strong shearing effect
xRES ¼ x  xSH ¼ 0 for jsj P jxj; ð26Þ of the flow through the ‘‘gap’’ between vortices along the
centreline of the contrarotating vortex pair. Unlike the
xRES ¼ x  xSH ¼ ðsgn xÞ½jxj  jsj for jsj 6 jxj: ð27Þ
residual vorticity, the total vorticity of the inner vortex
As can be inferred from (27), the magnitude of planar region (approximately between the residual-vorticity max-
residual vorticity represents nothing but the simplest mea- imum and the centreline of the contrarotating vortex pair)
sure of the dominance of planar vorticity magnitude over inevitably absorbs this shearing effect by increasing its
planar strain-rate magnitude. Analogously, from (20), the magnitude. Consequently, the transverse (i.e. y) vorticity
magnitude of planar residual strain rate is nothing but gradient of the inner vortex region is higher in comparison
the simplest measure of the dominance of planar strain-rate with the outer vortex region which is relatively less
magnitude over planar vorticity magnitude. affected by the ‘‘gap-like flow’’.
 V. Kolář / Int. J. Heat and Fluid Flow 28 (2007) 638–652 647

a 20 20 20

vorticity residual vorticity

18 18 18

16 16 16

0 .0
3
14 14 14

0 .1 2

0 .0 3
1
0 .2

z/D

z/D
z/D

12 12 0.3 0 12 0.03
0.12

0.
39
0.12

9
10 10 10

0.3
0.48

0 .3 0
0.57
0 .6 6 0 .4
8

2
30 9
0.2

0 .1
0 . 0 .3

12
1

0.
0.03
8 8 .2 1
8 3
0 0 .0
03
0.

6 6 6

UC
4 4 4
0 2 4 6 8 0 2 4 6 8 0 2 4 6 8
y/D y/D y/D

b 20 20 20

vorticity residual vorticity

18 18 18
0 .0
3

16 16 16
03
0.

14 14 0.21 14
0 .1 2

3
0 .0
0.3

0.1 2
0
z/D

z/D
z/D

12 12 12

0.2
0.12

1
0.
48

0 .3
2
0 .1
0 .3

0
10 10 30 10
9

0. 2
0 .1
03
0.

3
1

0 .0
0 .2

8 8 8
0 .0 3

6 6 6

UC
4 4 4
0 2 4 6 8 0 2 4 6 8 0 2 4 6 8
y/D y/D y/D

c 20 20 20

vorticity residual vorticity

18 18 18
03
0.

16 16 16
0. 03

0 .2

0 .1
1

2 0. 0 .0
14 14 14 12 3
0.30

0.21
z/D

0 .3 9

z/D
z/D

12 12 12
0.12

0 .0
0.0 3

3
0.12

0.03

0 .1 2
10 10 0 .2 1 10
0 .0

8 8 8
3

6 6 6

UC
4 4 4
0 2 4 6 8 0 2 4 6 8 0 2 4 6 8
y/D y/D y/D

Fig. 4. Secondary-flow vortex geometry of the JICF in terms of velocity, (negative) vorticity and residual vorticity in the cross-section at: (a) x/D = 10,
(b) x/D = 15, (c) x/D = 20. UC denotes crossflow velocity, x is crossflow direction (the origin of coordinates is located at the center of the jet exit), D is
jet-nozzle diameter.
648 V. Kolář / Int. J. Heat and Fluid Flow 28 (2007) 638–652

The employed additive vorticity decomposition implies tifier, positive x2  s2, and the proposed residual vorticity
a corresponding surface-quadrature decomposition. The are examined below as possible candidates for vortex-axis
circulation (strictly said, a portion of total circulation C) identification and for characterizing the local intensity of
based on residual vorticity can be obtained by integration. a vortex. Due to the two-dimensionality of the testcase
This quantity represents the (integral) vortex strength. examined below, the term ‘‘vortex centre’’ is employed
instead of ‘‘vortex axis’’.
5. Discussion The kinematically consistent model of a planar vortex-
shear interaction employed below provides a fair basis
The idea of vorticity decomposition has its own history, for obtaining correct qualitative results in vortex identifica-
though much shorter than the decomposition of motion. It tion. The examined planar $u-field is formed by super-
is almost 50 years old, see Astarita (1979), Wedgewood imposing two linear shearing effects of a Gaussian
(1999) and the references therein. The result of the Gies- distribution located symmetrically at y = ±0.5 given by
ekus–Harnoy–Drouot decomposition (this terminology is "  2 # "  2 #!
adopted following Wedgewood, 1999), namely the objective SHEARING y  0:5 y þ 0:5
uy ¼ K exp  þ exp  ;
portion of X obtained with respect to the principal axes of r r
S, is proposed by Astarita (1979) for a flow classification. ð28Þ
Wedgewood (1999) derives a new vorticity decomposition
into two parts, the so-called deformational vorticity and onto the $u-field of an ideal axisymmetric (counterclock-
the rigid vorticity. His analysis employs the cross product wise) Taylor vortex centred at (0, 0) describedpby the tan-
ffiffiffiffiffiffiffiffiffiffiffiffiffiffi
of a particle’s velocity and acceleration, u · Du/Dt, and gential velocity distribution of the form (r ¼ x2 þ y 2 )
leads to the evolution equation for the objective deforma- V VORTEX 2
tangential ¼ Cr expðr Þ: ð29Þ
tional vorticity. The solution of the ‘Wedgewood equation’
which depends on both space and time derivatives of $u is All parameters in (28) and (29) are positive. The vortex-
proposed to develop objective constitutive equations for strength parameter C was set for simplicity to be C = 1 and
complex rheological fluids. the shearing parameter r was set to be r = 1/4 (both
It should be emphasized that the present vorticity parameters are fixed within the present analysis). The plac-
decomposition into the shear vorticity and the residual ing of two shearing effects makes the examination proce-
one is based on the decomposition of motion through the dure much more illustrative than using only one.
decomposition of a $u-field ‘‘frozen’’ at a given instant in Three vortex characteristics – vorticity, positive x2  s2
time. The corresponding vortex-identification method rep- (i.e. the conventional planar-flow vortex-identification
resents a certain qualitative ‘‘comeback’’ of vorticity, measure) and the residual vorticity – are compared in
namely its specific portion, the residual vorticity. In view Fig. 5 for different values of the shearing-strength parame-
of the requirements summarized in Section 3, note that ter K, namely K = 0, 2, 10. The distribution of positive
the residual vorticity retains all of the very useful vorticity x2  s2 exhibits the formation of a double peak quite sim-
features – Galilean invariance, vectorial character (direc- ilar to that of vorticity distribution. For large shearing val-
tion and orientation), applicability to compressible flows ues, K  1, the peak magnitudes of both vorticity and
and variable-density flows, easy integrability across an positive x2  s2 are adequately large and the locations of
arbitrary surface area, etc. – and satisfies most of the gen- peak values ultimately attain the locations of shearing max-
eral requirements for vortex identification. Unlike vortic- ima at y = ±0.5. This causes an inevitable ambiguity in
ity, the residual vorticity distinguishes between pure defining the vortex centre in terms of vorticity as well as
shearing motions and the actual swirling motion of a vor- x2  s2 by using the natural extremal condition, as dis-
tex and, consequently, it correctly captures vortical motion cussed in Section 3. The identification of a vortex in terms
near a wall by diminishing to zero at the wall. On the other of the non-zero residual vorticity indicates that this criterial
hand, the residual vorticity is still a local quantity, more- quantity provides an identical vortex boundary as the posi-
over, not invariant with respect to rotating reference tive x2  s2. However, there is no ambiguity in defining the
frames. vortex-centre location as the characteristic peak value of
In planar incompressible flows, all of the local criteria the residual vorticity remains at the original centred posi-
described in Section 2 degenerate to the same one (Jeong tion (for K = 0) independently of the strength of a superim-
and Hussain, 1995; Wu et al., 2005) identifying the vortex posed shearing and its magnitude remains for K  1
region by the condition x2  s2 > 0 implying that vorticity unchanged as well.
dominates strain rate. This criterion corresponds to the It is shown in Fig. 5 that the conventional criterial quan-
Weiss criterion (Weiss, 1991; Basdevant and Philipovitch, tity, positive x2  s2, satisfactorily identifying the overall
1994) for elliptic flow regions and can be geometrically vortex region fails – unlike xRES – to describe the vortex
interpreted as the region of positive unnormalized Gauss- centre correctly (in terms of the location and magnitude
ian curvature of the stream function (Dresselhaus and of the peak value) provided that a pure shearing motion
Tabor, 1989). The swirling strength of Zhou et al. (1999)
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi superimposed onto an ideal axisymmetric vortex is not neg-
is equal in 2D motion to x2  s2 . The conventional iden- ligible. This quantity cannot represent the local intensity of
 V. Kolář / Int. J. Heat and Fluid Flow 28 (2007) 638–652 649

vorticity positive ω 2– s2 residual vorticity

1.0
K=0 1.0
K=0 K=0
1. 0
0.0

0.0
0.0 0.0
0.5 0.5 0. 5

0.0
0.5 0.5 0.5
y y 0.0 0.0
y
0.0 0. 0 0.0

0.5 0.5 0.5

-0.5 -0.5 0.0 -0.5 0.0

0.0 0.0

-1.0 -1.0 -1.0

-1.0 -0.5 0.0 0.5 1.0 -1.0 -0.5 0.0 0.5 1.0 -1.0 -0.5 0.0 0.5 1.0
x x x

K=2 K=2 K=2

1.0 1.0 1. 0
0.0 0.0
0.0 0.0
0.5
0.5
1.0 1.0
1.0 1.5
0.5 0.5 0. 5
1.5 1.5
0.0 0.0 0.0 0.0
0.5
0.5
1.0 0.5 0.5 0.5
1.0
y 0.0 0.5
y
0.0
1.0
0.0
y 0. 0 0.0

0.5 0.5
0.0 0.0
1.5
1.0
-0.5 1.5
-0.5 1.5 -0.5
1.0 1.0
0.0 0.5 0.0
0.5
0.0 0.0
-1.0 -1.0 -1.0
-1.0 -0.5 0.0 0.5 1.0 -1.0 -0.5 0.0 0.5 1.0 -1.0 -0.5 0.0 0.5 1.0
x x x

K=10 K=10 K=10

1.0 0.0 0.0 1.0 0.0 1. 0
0.0 0.0 0.0

0.5
2.5

5.0 0.0 0.0 0.0 0.0

0.5 0.5 0. 5
5.0 7.5 5.0 2.5

2.5 0.5

y y 0.0 0.0
y
0.0 0. 0 0.0

2.5
0.5
5.0
-0.5 -0.5 7.5 5.0 2.5 -0.5
5.0 0.0 0.0
0.0 0.0
2.5

0.0 0.0 0.0 0.0

-1.0 -1.0 -1.0
-1.0 -0.5 0.0 0.5 1.0 -1.0 -0.5 0.0 0.5 1.0 -1.0 -0.5 0.0 0.5 1.0
x x x

Fig. 5. Description of a planar vortex-shear interaction in terms of different vortex characteristics.

a vortex due to the strong inherent bias towards [jxj + jsj] of these measures requires specific conditions which guar-
expressed by antee the same local intensity of the swirling motion at dif-
ferent points of the flow field. Fig. 6 implies that a
x2  s2 ¼ ðsgn xÞ½jxj  jsj  ðsgn xÞ½jxj þ jsj superimposed linear shearing onto the rigid-body rotation
¼ xRES  ðx þ xSH Þ for jsj < jxj: ð30Þ near a point (i.e. let L1 > L2, L1 is otherwise arbitrary, L2
is fixed) clearly affects (makes greater) the intensity of the
The above mentioned bias can be inferred from the swirling motion measured by x2  s2 while xRES remains
geometry of the local velocity field near a point depicted unaffected.
in the BRF. The vortex intensity described in terms of The criteria given by positive x2  s2 and non-zero xRES
x2  s2 or, alternatively, xRES is examined in Fig. 6. Either are equivalent only at zero threshold as these criterial
650 V. Kolář / Int. J. Heat and Fluid Flow 28 (2007) 638–652

deformable fluid
element as a set of
discrete undeformable
L material points
P
reference intensity (initial state)

Conditions of vortex-intensity equivalence

circular
translation
for ω 2-s2:

L2 P
ω 2-s2 ∼ L1L2=L2

L1

for ω RES:
L2
P rigid-body
let L1>L2, ω RES ∼ L2=L rotation

L1
L1>L2
otherwise arbitrary

Fig. 6. Conditions of vortex-intensity equivalence: the comparison of

x2  s2 with xRES.

measures strongly differ in describing the inner structure of pure shearing

a vortex. The expression (30) explains the coincidence of
the zero contour for x2  s2 with that for xRES (Fig. 5).
In both cases, this contour represents the boundary of a Fig. 7. Virtual motional changes of the fluid element within an ideal
vortex. The boundary is, however, only roughly matching axisymmetric vortex (each at the final stage).
the original axisymmetric boundary of an isolated Taylor
vortex. The reason is that any local schemes based on
pointwise analysis can hardly reveal any virtual non-local perpendicular to a flow plane through a given point, is
flow components or features exactly. (twice) the mean angular velocity of any two instanta-
The TDM, as the local concept, cannot reveal the global neously mutually orthogonal line segments, within the flow
motion of an ideal axisymmetric vortex in concentric shear- plane, going through the given point and, consequently, it
ing layers. However, consistently with the TDM, the is (twice) the mean angular velocity of all line segments,
motion of an ideal vortex can be interpreted as the virtual within the flow plane, going through the given point (this
superposition shown in Fig. 7 (for clarity including a circu- result is attributed to Cauchy, 1841, according to Truesdell
lar translation) where the residual vorticity and shear vor- and Toupin, 1960; Truesdell, 1954; Green, 1995). Then the
ticity apparently have the same signs. Hence, equally as residual vorticity in 2D can be interpreted in terms of
in the case of Fig. 2, the superimposing construction in (twice) the least-absolute-value angular velocity of all line
Fig. 7 is of a non-destructive nature and is applicable to segments, within the flow plane, going through the given
infinitesimal motional changes only. point. Fig. 3 may help in this regard by considering dashed
Finally, let us have a look at a primary interpretation of lines as the line segments with the minimum and maximum
the residual vorticity. In the frame of the TDM, the residual angular velocities. The residual vorticity is clearly zero for
vorticity can be interpreted in terms of (twice) the angular all the planar cases where strain rate dominates over vortic-
velocity of material points (Figs. 1 and 7) due to the resid- ity as one portion of line segments rotates in one direction
ual rigid-body rotation. Now let us try to formulate a more while the other portion of line segments in the opposite
conventional interpretation, for simplicity in 2D. Vorticity, sense and two separating non-rotating line segments exist
 V. Kolář / Int. J. Heat and Fluid Flow 28 (2007) 638–652 651

(saddle separatrices in Fig. 3). Further, for a simple shear IAA2060302, and by the Acad. of Sci. of the Czech Rep.
one non-rotating line segment exists. The shear vorticity through Inst. Res. Plan AV0Z20600510.
represents just the difference between (twice) the mean
angular velocity and (twice) the least-absolute-value angu- Appendix A
lar velocity of all line segments. In 3D, the already stated
quasi-conventional interpretation of both the residual vor- The transformation matrix Q for an arbitrarily rotated
ticity and the shear one is for equal leading-diagonal ele- coordinate frame can be obtained by a sequence of three
ments applicable to the BRF coordinate directions and rotational (positive counterclockwise) transformations
planes, however, not to arbitrarily chosen directions and while the Cartesian coordinate system changes three times,
planes. Such universality is possessed by the conventional as follows:
interpretation of vorticity, strictly said, the vorticity projec- First rotation is around the z-axis by an angle
tion onto the normal of the examined flow plane. The a; 0 6 a 6 p; ðx; y; zÞ ! ðx ; y  ; z Þ; z  z,
above introduced interpretation of the residual vorticity is 0 1
cos a sin a 0
obviously a good argument for using this measure in vortex B C
identification. Q ¼ @  sin a cos a 0 A: ðA:1Þ
0 0 1
6. Conclusions Second rotation is around the y*-axis by an angle
b; 0 6 b 6 p; ðx ; y  ; z Þ ! ðx ; y  ; z Þ; y   y  ,
A brief survey of popular vortex-identification methods 0 10 1
is presented on the background of other vortex-identifica- cos b 0  sin b cos a sin a 0
tion schemes and relevant discussions on vortex definition. B CB C
Q¼@ 0 1 0 A@  sin a cos a 0 A
A particular emphasis is put on the summary of new gen- sin b 0 cos b 0 0 1
eral requirements and limitations for any vortex-identifica- 0 1
cos a cos b sin a cos b  sin b
tion schemes and their underlying criterial quantities. B C
The present paper further suggests that a specific por- ¼ @  sin a cos a 0 A: ðA:2Þ
tion of vorticity provides a proper physical quantity for cos a sin b sin a sin b cos b
the kinematic identification of a vortex. On the basis of
the triple decomposition of the relative motion near a Third rotation is around the z**-axis by an angle
point, the vorticity is decomposed into two parts, shear c; 0 6 c 6 p=2; ðx ; y  ; z Þ ! ðx ; y  ; z Þ; z  z ,

0 10 1
cos c sin c 0 cos a cos b sin a cos b  sin b
B CB C
Q ¼ @  sin c cos c 0 A@  sin a cos a 0 A
0 0 1 cos a sin b sin a sin b cos b
0 1
cos a cos b cos c  sin a sin c sin a cos b cos c þ cos a sin c  sin b cos c
B C
¼ @  cos a cos b sin c  sin a cos c  sin a cos b sin c þ cos a cos c sin b sin c A: ðA:3Þ
cos a sin b sin a sin b cos b

vorticity and residual vorticity. The latter is associated with An arbitrary vector v in old coordinates is given by Qv
the local residual rigid-body rotation near a point obtained in new coordinates, an arbitrary tensor G in old coordi-
after the extraction of an effective pure shearing motion nates is described by QGQT in new coordinates.
and represents a direct measure of the actual swirling
motion of a vortex. The residual vorticity retains all the
very useful vorticity features and satisfies most of the gen- References
eral requirements for vortex identification.
Astarita, G., 1979. Objective and generally applicable criteria for flow
classification. J. Non-Newtonian Fluid Mech. 6, 69–76.
Basdevant, C., Philipovitch, T., 1994. On the validity of the ‘‘Weiss
criterion’’ in two-dimensional turbulence. Physica D 73, 17–30.
Acknowledgements
Batchelor, G.K., 1967. An Introduction to Fluid Dynamics. Cambridge
University Press.
This work was supported by the Grant Agency of the Berdahl, C., Thompson, D., 1993. Eduction of swirling structure using the
Acad. of Sci. of the Czech Rep. through grant velocity gradient tensor. AIAA J. 31, 97–103.
652 V. Kolář / Int. J. Heat and Fluid Flow 28 (2007) 638–652

Cauchy, A.-L., 1841. Mémoire sur les dilatations, les condensations et les Kolář, V., Savory, E., Toy, N., 2000. Centerline vorticity transport within
rotations produits par un changement de forme dans un système de a jet in crossflow. AIAA J. 38, 1763–1765.
points matériels. Ex. d’An. Phys. Math. 2, 302–330 (Oeuvres (2) 12, Leigh, D.C., 1968. Nonlinear Continuum Mechanics. McGraw-Hill.
343–377). Lesieur, M., Begou, P., Briand, E., Danet, A., Delcayre, F., Aider, J.L.,
Chakraborty, P., Balachandar, S., Adrian, R.J., 2005. On the relationships 2003. Coherent-vortex dynamics in large-eddy simulations of turbu-
between local vortex identification schemes. J. Fluid Mech. 535, 189– lence. J. Turbulence 4, 1–24.
214. Levy, Y., Degani, D., Seginer, A., 1990. Graphical visualization of vortical
Chakraborty, P., Balachandar, S., Adrian, R.J., 2006. Comment on flows by means of helicity. AIAA J. 28, 1347–1352.
‘‘Axial stretching and vortex definition’’ [Phys. Fluids 17, 038108 Lin, C.-L., McWilliams, J.C., Moeng, C.-H., Sullivan, P.P., 1996.
(2005)]. Phys. Fluids 18, 029101-1–029101-2. Coherent structures and dynamics in a neutrally stratified planetary
Chong, M.S., Perry, A.E., Cantwell, B.J., 1990. A general classification of boundary layer flow. Phys. Fluids 8, 2626–2639.
three-dimensional flow fields. Phys. Fluids A 2, 765–777. Lugt, H.J., 1983. Vortex Flow in Nature and Technology. Wiley.
Cucitore, R., Quadrio, M., Baron, A., 1999. On the effectiveness and Panton, R.L., 1984. Incompressible Flow. Wiley.
limitations of local criteria for the identification of a vortex. Eur. J. Perry, A.E., Chong, M.S., 1987. A description of eddying motions and
Mech. B/Fluids 18, 261–282. flow patterns using critical-point concepts. Annu. Rev. Fluid Mech. 19,
Dallmann, U., 1983. Topological structures of three-dimensional flow 125–155.
separation. DFVLR-IB Report No. 221-82 A07. Stokes, G.G., 1845. On the theories of the internal friction of fluids in
Dresselhaus, E., Tabor, M., 1989. The persistence of strain in dynamical motion, and of the equilibrium and motion of elastic solids. Trans.
systems. J. Phys. A: Math. Gen. 22, 971–984. Cambridge Phil. Soc. 8, 287–319 (Math. Phys. Papers 1, 75–129).
Dubief, Y., Delcayre, F., 2000. On coherent-vortex identification in Truesdell, C., 1953. Two measures of vorticity. J. Rational Mech. Anal. 2,
turbulence. J. Turbulence 1, 1–22. 173–217.
Garcı́a-Villalba, M., Fröhlich, J., Rodi, W., 2006. Identification and Truesdell, C., 1954. The Kinematics of Vorticity. Indiana University Press.
analysis of coherent structures in the near field of a turbulent Truesdell, C., Toupin, R.A., 1960. The Classical Field Theories. In:
unconfined annular swirling jet using large eddy simulation. Phys. Flügge, S. (Ed.), Encyclopedia of Physics, Principles of Classical
Fluids 18, 055103-1–055103-17. Mechanics and Field Theory, vol. III/1. Springer-Verlag.
Geers, L.F.G., Tummers, M.J., Hanjalić, K., 2005. Particle imaging Vollmers, H., Kreplin, H.-P., Meier, H.U., 1983. Separation and vortical-
velocimetry-based identification of coherent structures in normally type flow around a prolate spheroid – Evaluation of relevant
impinging multiple jets. Phys. Fluids 17, 055105-1–055105-13. parameters. In: Proc. of the AGARD Symposium on Aerodynamics
Green, S.I., 1995. Introduction to vorticity. In: Green, S.I. (Ed.), Fluid of Vortical Type Flows in Three Dimensions, Rotterdam, AGARD-
Vortices. Kluwer. CP-342, pp. 14-1–14-14.
Haller, G., 2005. An objective definition of a vortex. J. Fluid Mech. 525, Wedgewood, L.E., 1999. An objective rotation tensor applied to non-
1–26. Newtonian fluid mechanics. Rheol. Acta 38, 91–99.
Hunt, J.C.R., Wray, A.A., Moin, P., 1988. Eddies, stream, and conver- Weiss, J., 1991. The dynamics of entrophy transfer in two-dimensional
gence zones in turbulent flows. Center for Turbulence Research Report hydrodynamics. Physica D 48, 273–294.
CTR-S88, pp. 193–208. Wu, J.-Z., Xiong, A.-K., Yang, Y.-T., 2005. Axial stretching and vortex
Jeong, J., Hussain, F., 1995. On the identification of a vortex. J. Fluid definition. Phys. Fluids 17, 038108-1–038108-4.
Mech. 285, 69–94. Wu, J.-Z., Xiong, A.-K., Yang, Y.-T., 2006. Response to ‘‘Comment on
Jeong, J., Hussain, F., Schoppa, W., Kim, J., 1997. Coherent structures ‘Axial stretching and vortex definition’’’ [Phys. Fluids 18, 029101
near the wall in a turbulent channel flow. J. Fluid Mech. 332, 185–214. (2006)]. Phys. Fluids 18, 029102-1–029102-2.
Kida, S., Miura, H., 1998. Identification and analysis of vortical Xiong, A.-K., Kobayashi, K., Izawa, S., Fukunishi, Y., 2004. Discussions
structures. Eur. J. Mech. B/Fluids 17, 471–488. on the methods for vortex identification. In: Abstracts of the IUTAM
Kolář, V., 2004. 2D velocity-field analysis using triple decomposition of Symposium on Elementary Vortices and Coherent Structures: Signi-
motion. In: Behnia, M., Lin, W., McBain, G.D. (Eds.), Proc. of the ficance in Turbulence Dynamics, Kyoto, pp. 90–91.
15th Australasian Fluid Mechanics Conference (CD-ROM), Univer- Zhang, S., Choudhury, D., 2006. Eigen helicity density: a new vortex
sity of Sydney, Paper AFMC00017. http://www.aeromech.usyd.edu. identification scheme and its application in accelerated inhomogeneous
au/15afmc. flows. Phys. Fluids 18, 058104-1–058104-4.
Kolář, V., Lyn, D.A., Rodi, W., 1997. Ensemble-averaged measurements Zhou, J., Adrian, R.J., Balachandar, S., Kendall, T.M., 1999. Mechanisms
in the turbulent near wake of two side-by-side square cylinders. J. for generating coherent packets of hairpin vortices in channel flow.
Fluid Mech. 346, 201–237. J. Fluid Mech. 387, 353–396.