Vortex Control of Bifurcating Jets: A Numerical Study: Carlos B. Da Silva and Olivier Me Tais

PHYSICS OF FLUIDS VOLUME 14, NUMBER 11 NOVEMBER 2002
Vortex control of bifurcating jets: A numerical study

Carlos B. da Silva and Olivier Métaisa)
L.E.G.I./Institut de Mécanique de Grenoble, B.P. 53, 38041 Grenoble Cedex 09, France
共Received 9 January 2002; accepted 24 July 2002; published 12 September 2002兲
Direct and large-eddy simulations 共DNS/LES兲 are performed to analyze the vortex dynamics and the
statistics of bifurcating jets. The Reynolds number ranges from ReD⫽1.5⫻103 to ReD⫽5.0⫻104 .
An active control of the inlet conditions of a spatially evolving round jet is performed with the aim
of favoring the jet spreading in one particular spatial direction, thus creating a bifurcating jet. Three
different types of forcing, based on the information provided by a LES of a natural 共unforced兲 jet,
are superimposed to the jet inlet in order to cause its bifurcation. The different forcing types mimic
the forcing methods used in experimental bifurcating jets 共Lee and Reynolds, Parekh et al., Suzuki
et al.兲, but using excitations with relatively low amplitudes, which could be used in real industrial
applications. The three-dimensional coherent structures resulting from each specific forcing are
analyzed in detail and their impact on the statistical behavior of bifurcating jets is explained. In
particular we focus on the influence of the forcing frequency and of the Reynolds number on the jet
control efficiency. Analysis of the coherent vortex dynamics shows that an inlet excitation which
combines an axisymmetric excitation at the preferred frequency and a so-called flapping excitation
at the subharmonic frequency is the most efficient strategy for jet control, even at high Reynolds
numbers. © 2002 American Institute of Physics. 关DOI: 10.1063/1.1506922兴
I. INTRODUCTION diameter and U 1 is the maximum axial velocity at the inlet兲.

Within the range of Reynolds numbers considered, this ob-
Jet control has attracted a great deal of attention over the servation was found to be independent of the Reynolds num-
past decades due to the wide range of its potential industrial ber. Consequently, this particular frequency was denomi-
applications: it can be used to improve industrial processes nated the jet preferred mode. It was further noticed that the
like mixing, heat transfer, and combustion. Since the jet dy- natural jet 共unforced兲 has a peak frequency right at this pre-
namics is strongly influenced by the presence of coherent ferred mode which corresponds to the periodic passage of
vortices, particularly in the transition region near the inlet vortical structures at the end of the potential core. Although
nozzle, jet control can be achieved by manipulating these these structures could no longer be discerned for Reynolds
coherent structures. It is therefore of capital importance to numbers greater than ReD⬎7⫻104 , the spectral peak located
understand the dynamics and the topology of these structures at the preferred mode could still be noticed at high Reynolds
and the way they affect the flow behavior. In particular, an numbers. However, through the analysis of the jet stream-
important application of jet control can be found in noise wise velocity fluctuation at the centerline, they noticed that
reduction since it is recognized that the vortex pairings con- the jet could no longer be controlled for longitudinal dis-
stitute an important source of noise generation. Both passive tances such that x/D⬎8. Finally, they noticed that another
and active control can be used to modify the jet dynamics.
effect of this specific jet forcing was a big increase of the jet
Passive jet control refers to the control of the jet spatial evo-
spreading rate between 0⬍x/D⬍8.
lution through the use of particular 共in general noncircular兲
Many experimental works on jet control have focused on
shapes of the inlet nozzle 共Gutmark and Grinstein4兲. Active
the enhancement of the jet spreading rate 共Lee and
jet control is obtained through energy consuming devices
Reynolds,1 Longmire and Duong,8 Corke and Kusek,9 Su-
which create a deterministic perturbation at the jet inlet. In
zuki et al.3兲. Of particular interest for potential industrial ap-
practice, this can be obtained through the use of loud-
plications are the so-called bifurcating jets. Bifurcating jets
speakers 共Crow and Champagne,5 Zaman and Hussain6兲 or
display a spectacular increase of their spreading rate along
flap actuators 共Zaman et al.,7 Suzuki et al.3兲.
one particular plane 共the bifurcating plane兲 while maintain-
Among the first important results related to jet control
ing a standard or even reduced spreading rate along a plane
are those obtained by Crow and Champagne.5 They used a
loudspeaker to superimpose several forcing frequencies f, to perpendicular to the latter 共the bisecting plane兲. Lee and
the jet inlet and observed that for a forcing frequency corre- Reynolds1 showed how simultaneous axial and orbital exci-
sponding to a Strouhal number of StrD ⫽ f D/U 1 ⫽0.3 the jet tations can be used to dramatically increase the growth of the
attained the maximum amplification of the initial perturba- jet shear layer in the bifurcating plane. The orbital forcing
tion, at the end of the potential core x/D⫽4 共D is the jet was produced by the orbital motion of the tip of the nozzle.
For the range of Reynolds numbers considered (2800⬍ReD
⬍10 000) the spreading of the jet was found to depend solely
a兲
Electronic mail: olivier.metais@hmg.inpg.fr on the ratio between the axial frequency and the orbital fre-
1070-6631/2002/14(11)/3798/22/$19.00 3798 © 2002 American Institute of Physics
Downloaded 12 Sep 2002 to 194.254.66.60. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/phf/phfcr.jsp
Phys. Fluids, Vol. 14, No. 11, November 2002 Vortex control of bifurcating jets: A numerical study 3799
quency R⫽ f a / f o . The range of Strouhal numbers used was at low Reynolds numbers (ReD⫽1500). They recovered the
0.35⬍ f a D/U 1 ⬍0.75 and 0.15⬍ f o D/U 1 ⬍0.35 for the axial results from Urbin and Métais10 with a similar forcing
and orbital excitations, respectively. For R⫽2 a bifurcating method called the ‘‘flapping excitation.’’ Furthermore, Dan-
jet is obtained in which the primary vortex rings are expelled aila and Boersma12 introduced the so-called ‘‘bifurcating ex-
towards two opposite directions in the bifurcating plane. The citation’’ resulting from the combination of both an axisym-
resulting shear layer displays a Y-shaped form. For R⫽3 the metric excitation and a flapping excitation. Although both
jet spreads into three branches 共trifurcating jet兲. For noninte- flapping and bifurcating excitations cause a drastic increase
ger values of R in the interval 1.6⬍R⬍3.2 the jet ‘‘blooms’’ of the spreading rate in the bifurcating plane, the statistical
into all directions. Notice that these results were obtained and topological features of both jets were found to be quite
using a quite large forcing amplitude 共about 17% of the ini- distinct. Furthermore, the bifurcating excitation was found to
tial mean velocity value兲. induce the largest spreading rate in the bifurcating plane. Let
Suzuki et al.3 experimentally studied a bifurcating jet re- us emphasize that Urbin and Métais10 used a much lower
sulting from the presence of small flap actuators placed on excitation amplitude than Danaila and Boersma.12 5% of the
the side walls of the inlet nozzle. They found that maximum mean inlet velocity against 15% for Danaila and Boersma.12
spreading in the bifurcating plane was obtained for a Strou- Other numerical works dealt with the application of
hal number based on the flapping frequency equal to Sta ‘‘stochastic optimization procedures’’ to the determination of
⫽0.25. This value is about one-half the preferred Strouhal the most efficient forcing frequencies for jet mixing
measured in the natural 共nonexcited兲 jet (StrD ⫽0.52). The 共Hilgers,13,14 Hilgers and Boersma,15 Koumoutsakos et al.16兲
range of Reynolds numbers investigated went up to ReD in both low 共DNS兲 and high 共LES兲 Reynolds numbers flows,
⫽13 000 only. Once more, the flapping forcing amplitude but they do not analyze in detail the vortex dynamics result-
was quite high: it was associated with a streamwise velocity ing from their excitation procedures, and use also very high
fluctuation about 10% larger than in the natural jet case. forcing amplitudes.
Another experimental study of bifurcating jets at high Rey-
The present work was motivated by the possibility to
nolds numbers was performed by Parekh et al.2 using four
dramatically enhance the jet spreading rate by the sole im-
loudspeakers to force the jet at the inlet nozzle. Using a
position of well chosen perturbations at the inlet nozzle as
combination of axial and flapping forcings through acoustic
demonstrated in the experimental works of Lee and
excitation, they showed that bifurcating jets can still be gen-
Reynolds,1 Parekh et al.,2 and numerical simulations of
erated at very high Reynolds numbers (1.0⫻104 ⬍ReD⬍1.0
Urbin and Métais10 and Danaila and Boersma.12 From the
⫻105 ). Once again, these experiments were performed with
discussion above on bifurcating jets, several questions arise.
quite a large forcing amplitude 共15% of the mean inlet ve-
One may in particular wonder what is the role played by the
locity profile兲.
coherent vortices to make certain jet excitations more suc-
There are comparatively less numerical than experimen-
cessful than others in controlling the jet dynamics. Most of
tal works related to jet control. Urbin and Métais10 were the
first to apply specific perturbations at the inlet of a spatially the experimental works on bifurcating jets using active con-
growing round jet to control its dynamics. They performed trol require the application of very high levels of forcing
large-eddy simulations with various inlet perturbations and amplitude 共about 15 to 20% of the inlet velocity兲 to achieve
managed to trigger axisymmetric, helicoidal, and what they jet bifurcation: this may not be a realistic value which could
called the ‘‘alternated pairing’’ mode. In this latter case, the be applied in practical engineering applications. Therefore,
forcing consisted in a deterministic perturbation at the jet in the present study we investigate the possibility of control-
inlet consisting in the imposition of a speed excess for one ling the jet using much lower amplitudes. Furthermore, as
half of the jet nozzle, while a speed defect is imposed on the previously recalled, the forcing frequency is a determining
other half, and this alternatively. An impressive spreading of parameter to efficiently achieve a specific type of jet control.
the jet along a preferential plane was observed in strong Therefore, several forcing frequencies are considered in this
similarity with experimental bifurcating jets. In particular, work. Finally, the use of LES allows the study of the control
the coherent structures numerically obtained were very simi- efficiency as a function of increasing Reynolds number. To
lar to the ones observed by Lee and Reynolds1 and Parekh summarize, the present numerical simulations of spatially
et al.2 With this excitation, the primary rings of the jet were growing round jets are aimed at 共i兲 studying the vortex dy-
seen to be alternatively inclined with respect to their axis, namics of the bifurcating jets, 共ii兲 investigating the possibil-
creating a zigzag arrangement and localized pairings of con- ity of controlling the jet with relatively small inlet perturba-
secutive rings. The strong jet spreading along the bifurcating tions, and 共iii兲 analyzing the influence of the forcing
plane was shown to be caused by the self-induced radial frequency and of the Reynolds number on the control effi-
共outwards兲 motion of the inclined rings and the resulting jet ciency.
was shown to exhibit the characteristic Y shape found by Lee This paper is organized as follows. The next section 共II兲
and Reynolds.1 However, the high dissipative nature of the reviews the governing equations and the subgrid-scale
numerical scheme used in their work casts some doubt upon model. Section III details the numerical methods used in the
some of their conclusions, namely in the possibility of con- various calculations. Section IV presents the different types
trolling the jet with this forcing scheme at high Reynolds of imposed excitations and lists the numerical and physical
numbers. The work of Urbin and Métais10 was revisited by parameters of the various simulations. The results obtained
Danaila and Boersma11,12 using direct numerical simulations for natural jets without deterministic inlet forcing at low and
3800 Phys. Fluids, Vol. 14, No. 11, November 2002 C. B. da Silva and O. Métais
high Reynolds numbers are described in Sec. V. Section VI model 共see Ducros et al.18 and Lesieur and Métais17 for de-
focuses on the forced jets. tails兲, which constitutes an improved version of the structure
function model originally proposed by Métais and Lesieur.19
Indeed, the filtered structure-function model removes the
II. GOVERNING EQUATIONS
large-scale inhomogeneities before the computation of the
This section deals with the numerical discretization of structure function. This gives very good results and makes
the three-dimensional Navier–Stokes equations with con- the model particularly suited for flows with a transition re-
stant density ␳ 0 and the continuity equation to study the tran- gion 共see Ducros et al.18兲. For the FSF model, the filtered
sition in natural and forced round jets at low Reynolds num- field uជ̄ is submitted to a high-pass filter L consisting in a
bers: Laplacian operator discretized by second-order centered fi-
Du i ⳵ u i ⳵ 共 u i u j 兲
Dt
⫽
⳵t
⫹
⳵x j
⫽⫺
1 ⳵p
␳0 ⳵xi
⫹␯
⳵ ⳵ui ⳵u j
⫹
⳵x j ⳵x j ⳵xi
, 冉冊 nite differences and iterated three times 共see Ducros et al.18
for the formulation兲. The eddy-viscosity given by the FSF
model is then given by
共1兲
⳵ui ␯ FSF ជ ⫺3/2 ˜ ជ
⫽0. 共2兲 t 共 x ,t 兲 ⫽0.0014C K ⌬ 关 F̄ 2 共 x ,⌬,t 兲兴 1/2, 共8兲
⳵xi
where
x i is the ith component of the position vector xជ and u i is the
ith component of the velocity vector u, ␯ ⫽ ␮ / ␳ 0 is the kine- ˜
F̄ 2 共 xជ ,⌬x,t 兲 ⫽ 具储 L3 共 uជ̄ 共 xជ ⫹rជ ,t 兲兲 ⫺L3 共 uជ̄ 共 xជ ,t 兲兲储典储 r 储 ⫽⌬ 共9兲
matic viscosity and p is the pressure.
For the high-Reynolds number cases, the DNS approach is the second-order velocity structure function of the high-
is no longer feasible and LES were performed 共Lesieur and pass filtered velocity field L3 (uជ̄ ). C K is the Kolmogorov
Métais17兲. The LES equations are found by applying a low- constant, here taken to be 1.4. ⌬ is the mesh size, which is
pass filter G ⌬ of width ⌬ to the Navier–Stokes equations. uniform in all our computations, ⌬⫽⌬ x ⫽⌬ y ⫽⌬ z .
The scales of motion are then decomposed into a grid-scale
part ( f̄ ) corresponding to scales larger than ⌬ and a subgrid-
scale part ( f ⬙ ) corresponding to scales smaller than ⌬. The
III. NUMERICAL METHOD
filtered field is defined for any quantity f 共scalar or vectorial兲,
as All the simulations presented here are performed with
f̄ 共 xជ ,t 兲 ⫽ 冕
⍀
f 共 yជ ,t 兲 G ⌬ 共 xជ ⫺yជ 兲 dyជ , 共3兲
the same numerical code. It is an highly accurate Navier–
Stokes solver in which spatial derivatives are discretized us-
ing a sixth-order-compact scheme 共Lele20兲 in the streamwise
where the integration is carried out in the whole computa- 共x兲 direction and pseudospectral methods 共Canuto et al.21兲
tional domain ⍀. After applying the filter to the Navier– are used for the normal 共y兲 and spanwise 共z兲 directions. Time
Stokes equations, one gets advancement is made with an explicit, three step, third order,
low storage Runge–Kutta time stepping scheme
Dū i ⳵ ū i ⳵ 共 ū i ū j 兲
⫽ ⫹ 共Williamson22兲. For pressure-velocity coupling a fractional
Dt ⳵t ⳵x j step method is used, in which a Poisson equation is solved to
⫽⫺
1 ⳵ p̄
␳0 ⳵xi
⫹␯ ⫹
⳵x j ⳵x j ⳵xi冉
⳵ ⳵ ū i ⳵ ū j
⫹
⳵Tij
⳵x j
, 冊共4兲
insure incompressibility at each substep of the Runge–Kutta
time advancing scheme 共for details see da Silva23兲.
In a spatial simulation code, special care has to be taken
⳵ ū i at the outflow boundary condition. In particular one must be
⫽0. 共5兲 sure that the coherent structures leave the computational do-
⳵xi
main without being affected by the outflow boundary condi-
The influence of the subgrid scales on the grid scale vari- tion. For this purpose the code uses a nonreflective
ables, appears through the subgrid-scale tensor 共Orlansky24兲 outflow condition in which not only the convec-
T i j ⫽ū i ū j ⫺u i u j 共6兲 tive 共as it is done for instance in Akselvoll and Moin25兲 but
also the viscous terms of the Navier–Stokes equations are
which has to be modeled. explicitly advanced. It was observed that the coherent struc-
Here we use the classical eddy-viscosity assumption to tures leave the computational domain without being distorted
model the subgrid-scale tensor: by the presence of the outflow boundary condition
T i j ⫺ 13 ␦ i j T kk ⫽2 ␯ t S̄ i j , 共7兲共Gonze26兲. The boundaries in the two transverse directions y
and z are taken as periodic. It is clear that such boundary
where ␯ t is the turbulent viscosity and conditions, although permitting the use of very accurate
S̄ i j ⫽ 冉
1 ⳵ ū i ⳵ ū j
⫹
2 ⳵x j ⳵xi 冊 pseudospectral schemes, do not allow the entrainment natu-
rally expected in a jet. However, if the lateral boundaries are
placed sufficiently far away, it was checked 共see below兲 that
is the large scale deformation tensor. The subgrid-scale those do not perturb the flow with spurious reflections. Fur-
model used here is the filtered structure function 共FSF兲 thermore, as was previously shown 共Stanley et al.,27 da
Silva23兲 a very small co-flow 共less than 10%兲 superimposed tional deterministic perturbation Uជ (xជ ,t) is imposed at
forc 0
to the inlet does not perturb the flow dynamics and allows the inlet nozzle. The exact formulation of these various de-
the natural growth of the jet by entrainment. For inflow terministic perturbations will be given in Sec. VI. The so-
boundary condition, the instantaneous velocity profile is de- called natural jet case corresponds to an unforced jet, in
scribed at the inlet. This will be detailed in the next section. which U ជ (xជ ,t)⫽0ជ .
forc 0
Several DNS and LES of both natural 共unforced兲 and
IV. PHYSICAL AND COMPUTATIONAL PARAMETERS forced jets were performed at various Reynolds numbers in
Apart from the Reynolds number, the major difference the range 1.5⫻103 ⬍ReD⫽U1D/␯⬍5⫻104 . All the simula-
between the various simulations carried out in this study tions reported here were carried out on the same mesh. The
comes from their different inlet velocity conditions. For each number of points is 201⫻128⫻128 and the mesh size is
time step, a given velocity profile is prescribed as an inlet taken uniform in all three directions. The computational do-
condition. Its general shape is main extends to 12.25D⫻7D⫻7D along the streamwise
and the two transverse directions. The ratio of the jet radius
ជ 共 xជ ,t 兲 ⫽U
ជ 共 xជ 兲 ⫹U
ជ ជ 共 xជ ,t 兲 ,
ជ 0 ,t 兲 ⫹U 共10兲
U 0 med 0 noise共 x forc 0 to the initial shear layer momentum thickness is R/ ␪ 0 ⫽20
ជ
where U (xជ 0 ,t) is the instantaneous inlet velocity vector. We and the initial maximum and minimum 共co-flow兲 jet veloci-
will use both Cartesian and cylindrical coordinates to repre- ties are U 1 ⫽1.075 and U 2 ⫽0.075. The co-flow is then very
ជ ⫽(U,V,W), small (U 2 /U 1 ⫽0.069⬍10%) and has only a minor influ-
sent the velocity vector. In the former case U
ence on the jet dynamics as will be demonstrated below. All
where U stands for the streamwise velocity component and V
the computations described in this paper were de-aliased us-
and W the two transverse components. We will also use the
ជ ⫽(u ,u ,u ), where u , u , ing the 2/3 law 共see Canuto et al.21兲. Finally, the maximum
cylindrical decomposition U ␪
x r x r amplitude of the random noise is set to A n ⫽1.0%.
and u ␪ represent the axial, radial, and tangential velocity
components, respectively. V. NATURAL JET
ជ (xជ )⫽(U ,0,0) is the mean streamwise velocity
U We first focus on the statistics and on the various coher-
med 0 med
which is given by a hyperbolic-tangent profile 共Michalke and ent structures present in a jet with a white noise inlet pertur-
Herman28兲. It is known to be a good approximation of the bation only. We will refer to this case as the ‘‘natural jet.’’
inlet velocity profiles found in measured experimental round The goal is twofold: first, validate our numerical procedure
jets 共Freymuth29兲, by comparing our results with experimental data and provide
冋冉冊册
reference cases with which to compare the forced jet cases.
U 1 ⫹U 2 U 1 ⫺U 2 1 R r R
U med共 xជ 0 兲 ⫽ ⫺ tanh ⫺ . 共11兲 Second, investigate the natural jet coherent structures to de-
2 2 4 ␪0 R r sign efficient control strategies applicable for high Reynolds
Here U 1 is the jet centerline velocity, U 2 is a small co-flow numbers flows.
and ␪ 0 is the momentum thickness of the initial shear layer. We first consider a DNS, called NLR, consisting in a
Notice that the mean transverse velocity components were natural jet at low Reynolds number (ReD⫽1500). In Fig.
set to zero at the inlet, 1共a兲, the flow structures are identified using the so-called Q
criterion 共Hunt et al.30兲 which is based on the second invari-
V med共 xជ 0 兲 ⫽W med共 xជ 0 兲 ⫽0. 共12兲 ant of the velocity gradient tensor, Q⫽1/2(⍀ i j ⍀ i j ⫺S i j S i j ),
ជ
U noise(xជ 0 ,t) is the inlet noise profile which is given by where ⍀ i j is the antisymmetrical part and S i j the symmetri-
cal part of the velocity gradient tensor. The positive Q re-
ជ ជ 0 ,t 兲 ⫽A n U base共 xជ 0 兲 ជf ⬘ . 共13兲
U noise共 x gions have proven to be good indicators of the coherent vor-
A n is the maximum amplitude of the incoming noise and tices in various wall-bounded or free-shear flows 共see, e.g.,
U base(xជ 0 ) is a function that sets the noise location mainly in Dubief and Delcayre31兲. Well-defined vortex rings become
the shear layer gradients: visible around x/D⫽3. These are shed with a characteristic
再
frequency corresponding to the preferred mode of the jet.
1 if 0.8⬍r/D⬍1.2, These axisymmetric vortex rings result from the develop-
U base共 xជ 0 ,r 兲 ⫽ 0.2 if r/D⬍0.8, ment of the axisymmetric mode called varicose mode pre-
0 otherwise. dicted by linear stability studies. Indeed, works by Michalke
and Herman28 and Cohen and Wygnanski32 pointed out
ជf ⬘ is a three component random noise designed to satisfy a
clearly the capital importance of the inflow momentum
given energy spectrum, thickness ␪ and of the ratio R/ ␪ : for R/ ␪ ⬎6.5, the axisym-
s
冋
E 共 k 兲 ⬀k s exp ⫺ 共 k/k 0 兲 2 .
2 册共14兲
metric 共varicose兲 mode was shown to be the most unstable
mode. Our results 共for a jet with initial R/ ␪ ⫽20) are there-
fore consistent with the linear stability predictions. Around
k⫽(k 2y ⫹k z2 ) 1/2 is the wave number norm in the (y,z) plane. x/D⬇9, the flow seems to exhibit helical-like structures,
The exponent s, and peak wave number k 0 , were chosen to suggesting the emergence of an helical mode: this is also
have an energy input at small scales 共high k 0 ) and a large- consistent with the stability analysis which predicts the heli-
scale spectral behavior typical of decaying isotropic turbu- cal mode of azimuthal wave number m⫽1 to be the most
lence (s⭐4). Note that the random noise is imposed on the unstable mode for R/ ␪ ⬍6.5. This value is reached here for
three velocity components. In the excited jet cases, an addi- x/D⬎10 共see da Silva23兲 and may explain why helical struc-
FIG. 1. 共a兲 Isosurface of positive Q criterion for the natural, low Reynolds number jet 共NLR兲. The chosen threshold is 0.10(U 1 /D) 2 with Q varying between
Q Min⫽⫺8.5(U 1 /D) 2 and Q Max⫽⫹16.8(U 1 /D) 2 ; 共b兲 isosurface of positive Q criterion for the natural high Reynolds number jet 共NHR兲. The chosen
threshold is 5.54(U 1 /D) 2 with Q varying between Q Min⫽⫺48.7(U 1 /D) 2 and Q Max⫽⫹118.4(U 1 /D) 2 ; 共c兲 isosurface of low pressure p for the natural high
Reynolds number jet, NHR. The chosen threshold is ⫺0.07␳ 0 U 21 with p varying between p Min⫽⫺0.3␳ 0 U 21 and p Max⫽⫹0.1␳ 0 U 21 . 共b兲 and 共c兲 correspond to
the same instant. Note that the figures do not show the total extent of the lateral computational domain.
tures are encountered in the far field of our low Reynolds nitude reaches its higher values 共1.5 to 2 times the vorticity
number jet. of the primary rings兲 dominate the jet dynamics for x/D
Visualizations of a LES of the natural jet at ReD ⲏ5. From here onwards, the coherent structures show a high
⫽25 000 are shown at Figs. 1共b兲 and 1共c兲. This simulation is level of small scale turbulence and the flow seems to be
referred to as NHR in the course of the text. In Fig. 1共b兲, the evolving into a statistical isotropic state, as no preferential
flow structures are identified with the Q criterion, while Fig. direction can be distinguished in the coherent vortices from
1共c兲 shows the corresponding low pressure isosurfaces at the x/D⬎8.
same instant. Note that Q and the pressure p are related The present simulation is now compared with the mea-
through Q⫽“ 2 p/2␳ 0 : this implies that the vortex- surements of Hussein et al.33 in the self-similarly evolving
identification criterion based upon Q involves much more region of the round jet. These measurements are today
small scale activity than the pressure. Around x/D⫽2 the among the best data banks for the round jet self-similar re-
growth of the inlet perturbation begins to induce a distortion gion. Within this region, round jets obey the following linear
of the initial shear layer. Vortex rings 共see pressure isosur- relationships:33
冋册
faces兲 are periodically shed near the inlet nozzle (x/D⬇2)
and keep their shape until the end of the potential core lo- ␦ 0.5共 x 兲 x x0
⫽C d ⫺ , 共15兲
cated around x/D⫽4: their shedding frequency corresponds D D D
冋册
to the jet-preferred mode. Further downstream, the transition
towards a fully developed turbulent state takes place very U1 1 x x0
⫽ ⫺ . 共16兲
rapidly. Strong streamwise vortices, where the vorticity mag- 具 u x 共 x,r⫽0 兲典 B u D D
FIG. 2. 共a兲 Streamwise velocity pro-

files in the far field of the high Rey-
nolds number, natural round jet
共NHR兲. U 2L is the local co-flow veloc-
ity and x 0 is the virtual jet origin. 共b兲
Axial normal stresses; 共c兲 streamwise
and radial cross stresses; 共d兲 down-
stream evolution of the axial normal
stresses.
␦ 0.5(x) is the jet half-width, defined as the distance from The streamwise mean velocity profiles at several down-
the jet centerline at which the mean velocity excess equals stream locations near the end of the computational domain
the mean centerline velocity excess, are compared with the experimental self-similar data from
Hussein et al.33 in Fig. 2共a兲. The superposition of the various
具 u x 共 x,r⫽ ␦ 0.5共 x 兲兲典 ⫺U 2L ⫽0.5共具 u x 共 x,r⫽0 兲典 ⫺U 2 兲 . 共17兲
profiles clearly shows that a self-similar state is reached in
U 2L ⫽ 具 u x (x,r⫽⬁) 典 is the local co-flow, while U 2 is the the LES in good correspondence with the experimental data.
inlet co-flow from Eq. 共11兲. The ratio U 1 / 具 u x (x,r⫽0) 典 is Profiles from two of the components of the Reynolds
known as the centerline velocity decay. Here the brackets 具典 stress tensor, at several downstream locations, are shown in
represent a time average such that the instantaneous velocity Figs. 2共b兲 and 2共c兲. Overall, the agreement with the experi-
field may be decomposed as u i ⫽ 具 u i 典 ⫹u ⬘i where u ⬘i stands mental results is quite good. The normal streamwise Rey-
for the fluctuating part. nolds stresses 关see Fig. 2共b兲兴 are very slightly under pre-
The downstream evolution of these two quantities does dicted for a radial distance smaller than ␩ ⫽r/(x⫺x 0 )
exhibit a linear evolution typical of the fully developed tur- ⬍0.08. However, the peak typically present in this compo-
bulent regime 关see Fig. 8共a兲兴. The decay of the centerline nent is found at the right radial location, that is r/(x⫺x 0 )
mean velocity occurs at x/D⬇4, which corresponds to the ⬇0.06. The radial normal stresses and tangential normal
end of the potential core. This value falls well within the stresses 共not shown兲 show even a better agreement with the
classical experimental data range for the potential core experiments. The same is true for the cross stresses as indi-
length, 4⬍x/D⬍5.5 共Crow and Champagne5兲. Further cated in Fig. 2共c兲 which shows the streamwise-radial cross
downstream, the slope of the centerline velocity decay is stresses profile at several downstream locations.
found to agree very well with the measured values for the Figure 2共d兲 displays the downstream evolution of the
round jet self-similar region. To confirm this point, we used root mean square 共rms兲 of the axial velocity at the centerline,
data between x/D⫽6.5 and x/D⫽11.5 from the simulation which is compared with the experiments by Zaman and
NHR to compute the centerline velocity decay rate 1/B u , in Hussain6 and Crow and Champagne.5 Very close to the
the far field region. We obtain B u ⫽5.7, which is very close nozzle, the flow becomes unstable within the jet shear layer
to the value of B u ⫽5.8, obtained by Hussein et al.33 This leading to a strong increase of the velocity fluctuations.
shows that despite the limited lateral size of the computa- These reach a maximum intensity around x/D⫽7. Further
tional domain (L y ⫽7D), the jet does not exhibit any con- downstream, their amplitude decays at a constant rate, sign
finement effect, even at the very end of the domain, L x of a fully developed turbulent flow. The compilation of vari-
⫽12.25D. ous laboratory experiments reveals an important scatter in
FIG. 3. 共a兲 Anisotropy invariant map

关Lumley and Newman 共Ref. 34兲兴 for
the Reynolds stresses. Each symbol
corresponds to a given flow location.
These were taken for several 共5兲 radial
coordinates (r/D⫽0.0 to r/D⫽1.0)
along an array of 14, equally spaced,
downstream locations from X MIN /D
⫽0 until X MAX /D⫽11.7; frequency
spectra of the axial velocity signal at
several locations for the natural, high
Reynolds number jet 共NHR兲. 共b兲 In the
far field (x/D⫽10), at r⫽D/2 共log–
log plot兲; 共c兲 after the end of the po-
tential core (x/D⫽4.4), at r⫽D/2
共lin–log plot兲.
the experimental data, which is illustrated by the two particu- I⫽b ii ,

lar experiments considered here, and indicates a very strong
sensitivity of the jet development to the inlet fluctuations. II⫽⫺b i j b ji /2, 共19兲
Considering the difficulty to represent numerically the exact III⫽b i j b jk b ki /3.
experimental inflow conditions in which the mean inlet ve-
locity profile and the inflow perturbations are issued from a Figure 3共a兲 displays the curve of II as a function of III for the
boundary-layer development, we consider that the evolution NHR simulation. The figure shows that a statistically isotro-
of the rms of axial velocity away from the nozzle is well pic state 共as far as second order moments are concerned兲 is
reproduced by our numerics using ‘‘artificial’’ inflow condi- reached very early downstream of the nozzle.
To provide a detailed analysis of the unstationary flow
tions 共white noise兲. Indeed, the computed prediction exhibits
characteristics we considered also frequency spectra at dif-
a satisfactory agreement with the experimental measure-
ferent flow locations. Figure 3共b兲 shows a log–log plot of the
ments of Zaman and Hussain6 both in terms of the amplitude
frequency spectrum computed from the axial velocity signal
of the rms maximum and its axial position.
taken at one point in the far field of the turbulent round jet.
A detailed analysis of the flow anisotropy may be per-
The spectrum displays a wide range of scales with a spectral
formed through the anisotropy invariant map proposed by behavior very close to an inertial range behavior with a ⫺5/3
Lumley and Newman.34 This map characterizes the various law extending over more than one decade. This further sup-
possible states of the turbulence from the anisotropy tensor ports the previous findings showing that the flow has indeed
b i j deduced from the Reynolds stress tensor: reached a fully developed turbulent state before the end of
the computational domain. Next we concentrate on the
downstream evolution of the various characteristic frequen-
具 u i⬘ u ⬘j 典 ⫺2/3K ␦ i j cies. Streamwise velocity spectra taken from the shear layer
bi j⫽ , 共18兲 region (r/D⫽0.5) and near the inlet (x/D⫽1) exhibit a
2K
well-defined peak frequency at about f ␪ 0 /U 1 ⫽0.015. This
peak corresponds to the instability of the initial jet shear
layer and is often referred to as the shear-layer mode. The
where K⫽1/2具 u i⬘ u i⬘ 典 . The three invariants 共I, II, and III兲 are present value is in good agreement with the typical values
given by found experimentally for this instability (0.01⬍ f ␪ 0 /U 1
⬍0.023, Gutmark and Ho35兲. Further downstream, at the jet

centerline (r/D⫽0) and near the end of the potential core
(x/D⫽4), the peak frequency f p associated with the so-
called preferred mode could also be clearly seen. This peak
is located around f p D/U 1 ⫽0.38. This is also well within
experimental values which predict 0.24⬍ f D/U 1 ⬍0.5, Gut-
mark and Ho.35 Note that shear-layer mode corresponds to
f D/U 1 ⫽0.6 that is to say to a frequency which is close to
the double 共⬇1.6兲 of the frequency of the preferred mode:
this may indicate that the axisymmetric vortex rings result
from pairing processes of smaller structures issued from the
shear-layer region close to the nozzle. Figure 3共c兲 represents
a frequency spectrum taken from a velocity signal just after
the end of the potential core region. This spectrum shows
that, superimposed to the preferred mode, several low fre-
quency modes are also excited including the subharmonic
frequency f s ⫽ f p /2 associated with a Strouhal number StrD
⫽ f s D/U 1 ⬇0.19. Further downstream, although the flow be- FIG. 4. Isosurfaces of low pressure showing two ‘‘alternated pairing’’
comes more and more complex due to the emergence of an events. Note that the figures show only a zoom of the flow around the end of
important level of background turbulence, the growth of this the potential core region 共between x/D⬇3.5 and x/D⬇8). 共a兲 Pairing be-
tween a recently formed 共smaller兲 ring and a strong ring from the end of the
subharmonic frequency could still be observed. This is ac- potential core region; 共b兲 pairing between two vortex rings after the potential
companied by a decrease of the relative importance of the core region. The chosen threshold is ⫺0.05␳ 0 U 21 with p varying between
preferred frequency. p Min⫽⫺0.3␳ 0 U 21 and p Max⫽⫹0.1␳ 0 U 21 .
We have seen in Fig. 1共c兲 that the Kelvin–Helmholtz
instability along the jet border yields vortex structures
mainly consisting in axisymmetric rings as a result of the the jet mean velocity profile, the outer part of the ring is then
varicose mode growth. We now follow these structures fur- advected downstream at a speed which is lower than U c
ther downstream. After the end of the potential core, subse- while the inner part is advected at a speed larger than U c 关see
quently to the varicose mode growth, the 3D visualizations Fig. 5共c兲兴. The latter then tends to catch up the ring which
reveal the presence of the ‘‘alternated pairing’’ vortex ar- was previously shed and this ultimately yields a localized
rangement already observed by Urbin and Métais.10 This pairing of the two consecutive rings 关see Fig. 5共d兲兴.
vortex topology appears intermittently and is not always dis-
tinguishable. Note that it is the first time that it is numeri- VI. EXCITED JETS
cally observed in a high Reynolds number jet since the ‘‘ef-
A. Forcing types
fective’’ Reynolds number used by Urbin and Métais10 was
much lower. Figures 4共a兲 and 4共b兲 are zooms of the jet re- We have seen in the preceding section that the preferred
gion situated between x/D⬇3.5 and x/D⬇8, at two different mode appears before the end of the potential core and that
instants. Two consecutive rings are shown to locally connect subsequently the subharmonic mode dominates before a fully
forming a zigzag vortex structure. Since the jet is statistically nonlinear multimode regime. The growth of the subharmonic
axisymmetric, this zigzag structure may appear in any spatial mode is intermittently associated with well recognizable
direction. Such vortex topology was also observed in the alternated-pairing arrangement of the vortex rings. We then
direct numerical simulation of a temporally evolving round use these two modes 共preferred and subharmonic兲 to design
jet at low Reynolds number (ReD⫽2000) by Comte et al.36 specific inlet perturbations to achieve jet control and in par-
Note that vortex loops inclination at the end of the potential ticular to make the jet bifurcate along a chosen direction. As
core was also experimentally observed by Petersen,37 and shown by Urbin and Métais,10 the jet bifurcation is strongly
that an experimental evidence of ‘‘alternated pairings’’ was linked with the efficient triggering from the jet nozzle of the
also shown by Broze and Hussain.38 alternated pairing arrangement. Two types of excitation are
As suggested by Urbin and Métais10 these alternated considered here. The first excitation 共Flap兲 is similar to the
pairing events are strongly linked with the growth of the excitation designed by Urbin and Métais10 and subsequently
subharmonic mode. This is schematically explained on the used by Danaila and Boersma.12 We use the terminology
sketches of Figs. 5共a兲–5共d兲. Let us call L the distance of the introduced by Danaila and Boersma12 and designate by flap-
primary vortex rings which are periodically shed at some ping excitation this particular forcing. Unlike these previous
distance from the jet nozzle 关see Fig. 5共a兲兴. Their convection authors, in the present work the perturbations are applied
speed U c and their preferred frequency f p are then related mainly in the region of highest shear. Indeed, when small
through f p ⫽U c /L. Let us assume that a subharmonic mode flap actuators are used 共Suzuki et al.3兲, these are placed on
grows in a given plane containing the jet axis. This longitu- the wall of the inlet nozzle and they only affect the boundary
dinal subharmonic mode is associated with a streamwise layers developing inside the nozzle. Therefore, the applica-
wavelength 2L. It then induces an alternated radial shift of tion of this excitation into localized gradient regions rather
the vortex rings away from the jet axis 关see Fig. 5共b兲兴. Due to than into the whole jet is more representative of the physical
FIG. 5. Sketch of the vortex rings arrangement leading

to the occurrence of alternated pairing events. The se-
quence is from 共a兲 to 共d兲. See text for details.
reality than previous numerical simulations 共e.g., Danaila f pD

and Boersma12兲. The form of the second excitation 共Variflap兲 S trp D ⫽ . 共21兲
U1
is similar to Danaila and Boersma’s12 so-called bifurcating
excitation and is constructed using a combination of a vari- The inlet excitation is therefore a combination of a time
cose 共axisymmetric兲 excitation and of a flapping excitation. periodic perturbation at the frequency f p / ␣ and a sinusoidal
The axisymmetric forcing is applied to the whole inlet perturbation along the y direction. Due to the sinusoidal per-
nozzle, in order to mimic the acoustic excitation made with a turbation, each half of the jet, y⬎0 or y⬍0, will either
loudspeaker. As opposed to the previous numerical works, present an axial velocity excess or a deficit, when compared
for the two types of perturbation flapping and varicose flap- to the other part. The plane containing the jet axis and the y
ping, we investigate both the influence of the frequency of direction will be called the bifurcating plane. The plane per-
the flapping excitation and of the Reynolds number. Note pendicular to the bifurcating plane and containing the jet axis
also that in the present work the maximum amplitude of the is often referred to as the bisecting plane. Figure 6共a兲 sche-
imposed excitation is only 5% of the inlet velocity, unlike
most experimental and numerical works which use between
15–20% 共Lee and Reynolds,1 Parekh et al.,2 Danaila and
Boersma12兲. As stressed above, we have decided to use a
relatively low level of amplitude forcing, since it is more
realistic in terms of potential industrial applications.
1. Flapping excitation
The flapping excitation is given by
冉
U forc共 xជ 0 ,t 兲 ⫽ ⑀ U base共 xជ 0 兲 sin 2 ␲
␣ D
冊冉冊
S trp D U 1
t sin
2␲y
D
, 共20兲
FIG. 6. 共a兲 Sketch of the flapping excitation; 共b兲 sketch of the varicose
where, S trp D designates the preferred mode frequency, excitation.
TABLE I. List of the excitation types.
Forcing Description
冉
U forc共 xជ 0 ,t 兲 ⫽ ⑀ U med共 xជ 0 兲 sin 2 ␲ S trp D
U1
D
t 冊
Flap1
Flap2
Variflap1
Flapping at the preferred mode frequency
Flapping at the subharmonic frequency
Varicose at the preferred mode frequency⫹

冉
⫹ ⑀ U base共 xជ 0 兲 sin 2 ␲
S trp D U 1
␣ D
t⫹
␲
4
冊冉冊
sin
2␲y
D
.
Flapping at the preferred mode frequency 共22兲

Variflap2 Varicose at the preferred mode frequency⫹
Flapping at the subharmonic frequency As stressed above, the varicose excitation is applied to
Variflap4 Varicose at the preferred mode frequency⫹ almost all the inlet nozzle through the U med(xជ 0 ) profile
Flapping at half the subharmonic frequency
whereas the flapping excitation exists only within the shear
layer gradients. Similarly to the flapping excitations, we
study the influence of the forcing frequency by varying ␣.
matically describes this flapping excitation. The value of ⑀ Since the alternated-pairing arrangement was seen to be as-
controls the maximum forcing amplitude and is taken equal sociated with the growth of a subharmonic frequency, the
to 0.05 for all the simulations. As stressed above, this exci- most natural choice for ␣ is ␣⫽2. This particular excitation
tation is mainly applied to the region of strong mean velocity will be designated as Variflap2. We also considered the two
gradients, since it is multiplied by the U base(xជ 0 ) profile. The extra values ␣⫽1 with both the varicose and the flapping
influence of the forcing frequency is investigated by varying excitations excited at the preferred frequency 共Variflap1兲 and
the parameter ␣. S trp D corresponds to the frequency of the ␣⫽4 corresponding to an excitation of the flapping mode at
preferred mode: S trp D ⫽0.38. Motivated by the results of the half the subharmonic mode 共Variflap4兲.
natural jet, two values of ␣ are considered here: ␣⫽1 corre- Table I summarizes all the different excitations used in
sponds to forcing at the preferred frequency and ␣⫽2 corre- this study. Note that the various excitations concern only the
sponds to a subharmonic forcing. These two excitation types axial velocity component:
are called Flap1 and Flap2, respectively.
V forc共 xជ 0 ,t 兲 ⫽W forc共 xជ 0 ,t 兲 ⫽0. 共23兲
2. Varicose-flapping excitation It is important to realize that the inlet energy input is
The varicose-flapping forcing consists in a combination very close for the various excitations. Figures 7共a兲 and 7共b兲
of varicose 共axisymmetric兲 and flapping excitations. The show time spectra of the axial velocity signal taken at the
varicose excitation is obtained by applying a time periodic inlet for the excitations Flap2 and Variflap2. For the Flap2
excitation to the whole inlet nozzle with a forcing frequency excitation one can notice the existence of a well-defined peak
corresponding to the preferred mode. Figure 6共b兲 illustrates (S trs D ⫽0.19) associated with the flapping forcing. In the
the varicose excitation. This is equivalent to the excitation Variflap2 excitation one can see the presence of two peaks
produced by loudspeakers in experimental studies of forced (S trp D ⫽0.38 and S trs D ⫽0.19), respectively, associated with
jets 共Crow and Champagne,5 Zaman and Hussain6兲 and the varicose and the flapping excitations. On both figures one
greatly enhances the formation of axisymmetric vortex rings. observes also the contribution of the white noise, spanning
With this sole varicose forcing formula, Urbin and Métais10 over a large range of frequencies. This figure shows that the
and Danaila and Boersma12 obtained vortex structures re- amount of total energy injected at the inlet is very close for
markably similar to the ones found in ‘‘varicose’’ experimen- both excitations. This issue will be analyzed in greater detail
tal works 共Crow and Champagne,5 Zaman and Hussain6兲. below when studying the near field evolution of the axial
The final forcing formula is then the following: Reynolds stresses for all the simulations.
FIG. 7. 共a兲 Time spectrum of the axial

velocity signal taken at the inlet for
excitation Flap2; 共b兲 idem for Vari-
flap2.
TABLE II. Summary of all round jet simulations. the potential core are similar to the natural jet case at ReD
Simulation Reynolds Forcing DNS/LES ⫽25 000 共NHR兲. As pointed out in the preceding section, the
initial intensity of the axial Reynolds stresses is similar for
NLR 1500 共Natur.兲 DNS
all the simulations near the inlet 关see Fig. 8共b兲 for x/D⬍1].
FLA1-A 1500 Flap1 DNS
In particular, the difference in the overall level of energy fed
FLA1-B 5000 Flap1 LES at the inlet in simulations NLR and FLA1-A is slim, and
FLA2-B 5000 Flap2 LES does not explain the much faster transition to turbulence of
VFLA1-B 5000 Variflap1 LES
VFLA2-B 5000 Variflap2 LES
the excited jet. We will show below that the reason for the
VFLA4-B 5000 Variflap4 LES much faster transition comes from the nature of the coherent
structures which are triggered in the forced case.
NHR 25 000 共Natur.兲 LES
We now analyze the control efficiency of the Flap1 forc-
FLA2-C 25 000 Flap2 LES
VFLA2-C 25 000 Variflap2 LES ing to make the jet bifurcate along the bifurcating plane. We
quantify the spreading difference between the bifurcating
VFLA2-D 50 000 Variflap2 LES and the bisecting planes by considering both shear layer
thicknesses ␦ 00(x) in the bifurcating plane and ␦ 90(x) in the
bisecting plane:
B. Flapping excitation at the preferred mode
For this particular excitation, two numerical simulations 具 u x 共 x,y⫽ ␦ 00共 x 兲 ,z⫽0 兲典 ⫺U 2LY ⫽0.5共具 u x 共 x,r⫽0 兲典 ⫺U 2 兲 ,
共24兲
are performed: a DNS at low Reynolds number (ReD
⫽1500) and a LES at higher Reynolds number (ReD
⫽5000). These two runs are referred to as FLA1-A and 具 u x 共 x,y⫽0,z⫽ ␦ 90共 x 兲兲典 ⫺U 2LZ ⫽0.5共具 u x 共 x,r⫽0 兲典 ⫺U 2 兲 ,
FLA1-B, respectively 共see Table II for the description of the 共25兲
various runs兲.
where U 2LY ⫽ 具 u x (x,y⫽⬁,z⫽0) 典 and U 2LZ ⫽ 具 u x (x,y⫽0,z
1. Statistics ⫽⬁) 典 are the local co-flow velocities in the bifurcating and
Let us first consider the influence of this forcing on the bisecting planes, respectively.
jet statistics. The excited jets are compared with the natural Figure 9共a兲 shows the downstream evolution of ␦ 00(x)
jets at ReD⫽1500 共DNS, run NLR兲 and ReD⫽25 000 共LES, and ␦ 90(x) for simulations FLA1-A, FLA1-B, and also for
run NHR兲 for which the inlet perturbation consists in a white the natural jet at high Reynolds number 共NHR兲. Note that,
noise of weak amplitude 共1% of the mean axial velocity兲. for the natural unforced jet, the growth of ␦ 00(x) and of
Compared to the natural jet simulation at the same Reynolds ␦ 90(x) are absolutely identical. One of the striking features
number 共NLR兲, the excited jet undergoes a much faster tran- of the Flap1 excitation is its immediate influence on the spa-
sition to turbulence. This can be seen in Figs. 8共a兲 and 8共b兲 tial development of the jet. The bifurcating shear layer thick-
displaying, respectively, the downstream evolution of the ness ␦ 00(x) begins to grow as soon as x/D⬇2 reaching a
axial mean velocity decay and the rms of the axial velocity at first peak around x/D⬇5. During the same stage, ␦ 90(x) is
the centerline for all the natural and excited jet simulations to significantly reduced showing an effective bifurcating effect
be reported here. For the natural jet at low Reynolds number of the Flap1 excitation. Further downstream, a sharp transi-
共run NLR兲, the potential core extends to around x⫽10D and tion takes place between x/D⫽5 and x/D⫽7 with a sudden
no increase of the turbulent kinetic energy can be observed decrease of ␦ 00(x) accompanied with a strong increase of
until x/D⬇4. A drastic change is observed due to the jet ␦ 90(x). In the later stage of the jet development, contrarily to
excitation. Indeed, with the Flap1 excitation at ReD⫽1500 what was expected, ␦ 90(x) dominates ␦ 00(x) except near the
共FLA1-A兲, the growth of the perturbation and the length of very end of the computational domain.
FIG. 8. 共a兲 Axial velocity decay in the

centerline for the Flap1, Flap2, Vari-
flap2 and natural simulations, at sev-
eral Reynolds numbers; 共b兲 down-
stream evolution of the axial Reynolds
stresses in the centerline for the Flap1,
Flap2, Variflap2 and natural simula-
tions, at several Reynolds numbers.
bifurcating plane the jet spreads much more rapidly at least

in the early phase of its development (x/Dⱗ7). Second, the
jet suffers a much faster transition to turbulence, if compared
to the natural jet at the same Reynolds number. This is
clearly confirmed by Fig. 1共a兲 which shows isosurfaces of
positive Q for the natural jet at identical Reynolds number
ReD⫽1500 共white noise of 1% amplitude兲. The faster transi-
tion towards turbulence in the excited case is caused by the
particular shape of the coherent structures present in the ini-
tial transition stage 共up to x/D⬇5). Contrary to what was
expected, the imposed excitation does not trigger the
alternated-pairing arrangement observed in the natural jet
case. Indeed, a zoom on the inlet region shown in Fig. 11
shows that the primary vortex rings are folded in their
middle along an axis perpendicular to the bifurcating plane.
FIG. 9. Bifurcating 关 ␦ 00(x) 兴 and bisecting 关 ␦ 90(x) 兴 shear layer thicknesses Moreover, strong streamwise vortices are seen to emerge
for the FLA1-A and FLA1-B simulations. The evolution of the shear layer
from the crests of this azimuthal deformation 共both upstream
thickness for the natural jet 共NHR兲 is also given for comparison.
and downstream兲. It seems that the deformation imposed to
the rings and the associated longitudinal vortices lead to a
2. Coherent structures fast three dimensionalization of the primary structures and
To understand the previous observations, we investigate therefore to this early transition to turbulence. Indeed, once
the flow coherent structures. Figures 10共a兲 and 10共b兲 show the primary vortex rings are deformed by the imposed exci-
isosurfaces of positive Q for the simulation FLA1-A. The tation, the two parts of the folded ring tend to travel in op-
first important observation clearly visible in the figures is the posite directions due to their self-induced velocity and they
marked difference between the spreading rate undergone by eventually separate: this mechanism explains the significant
the jet in the bifurcating and bisecting planes. Almost no increase of the spreading rate in the bifurcating plane. At this
spreading is observed in the bisecting plane whereas in the stage it is interesting to note that the primary structures ob-
FIG. 10. Isosurfaces of Q criteria for the Flap1 jet at ReD⫽1500 共FLA1-A兲. The chosen threshold is 0.125(U 1 /D) 2 and Q varies between Q Min
⫽⫺37(U 1 /D) 2 and Q Max⫽130(U 1 /D) 2 . 共a兲 View of the bifurcating plane; 共b兲 view of the bisecting plane.
FIG. 11. Isosurfaces of low pressure

showing the primary structures for the
Flap1 excitation at ReD⫽1500 共FLA1-
A兲. The chosen threshold is
⫺0.04␳ 0 U 21 and p varies between
p Min⫽⫺0.39␳ 0 U 21 and p Max
⫽0.099␳ 0 U 21 .
tained in the present case are slightly different from the ones lence is now much higher. It speeds up the fragmentation of
found for instance by Danaila and Boersma12 in their flap- the flow structures and quickly suppresses the control effi-
ping excitation. The difference can be attributed to the fact ciency of the flapping excitation.
that in the present simulation the flapping excitation is ap-
plied mainly in the shear layer zone and also because a much C. Flapping excitation at the subharmonic mode
lower forcing amplitude is used 共5% against 15%兲.
As demonstrated above, the efficiency of a given excita-
tion is highly sensitive to the jet Reynolds number. We next
vary the frequency of this inflow excitation to investigate if a
3. Influence of the Reynolds number more effective forcing can be designed by simply changing
the forcing frequency. Since the subharmonic frequency
Control of low-Reynolds number jets can be easily ob- naturally emerges in the natural jet, it is physically founded
tained because these flows are characterized by well-marked to use this particular frequency to control the jet. Indeed, if a
primary vortices, which are less quickly fragmented by the first vortex ring appears at a given time t 0 , the next ring will
growth of small-scale three-dimensional perturbations than form at time t 0 ⫹1/f p , due to the vortex shedding at the
in high-Reynolds number flows. A forcing procedure can preferred jet frequency, f p . One may then easily show from
therefore be efficient at low-Reynolds number and be inap- the expression 共20兲 for the subharmonic excitation, that
plicable when the Reynolds number is increased. To analyze
this issue, the Flap1 excitation is now applied to a jet at U forc共 xជ 0 ,t 0 ⫹1/f p 兲 ⫽⫺U forc共 xជ 0 ,t 0 兲共26兲
ReD⫽5000 共run FLAP1-B兲. The earlier transition to turbu-
lence caused by the increase in the Reynolds number can be so that two consecutive rings are expected to be forced in an
seen first through the downstream decay of the centerline alternated fashion.
axial mean velocity 关see Fig. 8共a兲兴 and through the down- Two LES are performed with the subharmonic excita-
stream evolution of the rms of the axial velocity 关see Fig. tion, one at ReD⫽5000 共FLA2-B兲 and the other at ReD
8共b兲兴. However, the increase of the Reynolds number signifi- ⫽25 000 共FLA2-C兲. For both values of the Reynolds num-
cantly diminishes the efficiency of the flapping excitation to ber, the classical jet statistics such as the downstream evolu-
control the jet dynamics. This clearly appears in Fig. 9 which tion of the centerline axial velocity decay and rms are close
shows that the difference between the bifurcating and bisect- to the one found for the jet at ReD⫽25 000 excited with the
ing shear layer thicknesses is limited only to the very early flapping excitation at the preferred mode 关see Figs. 8共a兲 and
transition stages (x/D⬍6), as in the lower Reynolds number 8共b兲兴. However, as shown below, a very distinct statistical
case. We have checked through flow visualizations 共not behavior is observed in the bisecting and the bifurcating
shown兲 that the coherent vortices in the neighborhood of the planes in the low-Reynolds number case 共FLA2-B兲. Indeed,
inlet (x/D⬍4) are not very different from the same case at at ReD⫽5000 共FLA1-B and FLA2-B兲, the shear layer thick-
lower Reynolds number but the level of small scales turbu- nesses for the preferred-mode and subharmonic-mode exci-
FIG. 12. Bifurcating 关 ␦ 00(x) 兴 and bisecting 关 ␦ 90(x) 兴 shear layer thicknesses
for the FLA2-B, FLA2-C, and natural 共NHR兲 jet simulations. D 00 and D 90
represent the bifurcating and bisecting planes, respectively.
tations can be compared by looking at Figs. 9 and 12. The

figures show that the division by two of the forcing fre-
quency makes the jet control possible and significantly im-
proves the efficiency of the flapping excitation, at least at this
Reynolds number. With the subharmonic excitation 关see Fig.
FIG. 13. Isosurfaces of low pressure showing the primary structures for the
12兴, the bifurcating and bisecting shear layer thicknesses Flap2 excitation at ReD⫽5000 共FLA2-B兲. The chosen threshold is
evolve similarly near the inlet until x/D⬇8, but, further ⫺0.04␳ 0 U 21 and p varies between p Min⫽⫺0.47␳ 0 U 21 and p Max
downstream, ␦ 00(x) starts dominating ␦ 90(x) showing that ⫽0.104␳ 0 U 21 .
the action of the inlet perturbation is felt near the end of the
computational domain. This is the opposite of what happens
with the preferred-mode excitation which yields a significant
modification of the jet dynamics very close to the inlet in Fig. 12 which shows that for ReD⫽25 000 this excitation
nozzle and becomes inefficient further downstream. has lost most of its effect on the spatial jet development.
The investigation of the coherent vortices obtained with Pressure isosurfaces for this simulation 共not shown兲 suggest
the subharmonic excitation confirms 共not shown兲 the identi- that the early primary vortex rings are ‘‘eaten’’ by the high
cal spreadings of the jet in both planes at the beginning of level of small scale turbulence present at this relatively high
the jet development and the enhanced jet broadening in the Reynolds number. Without strong initial coherent structures,
bifurcating plane for x/Dⲏ8. The reason for the increased it is not surprising that the flow cannot be controlled. Note
efficiency of the subharmonic excitation is that the primary that the experimental results of flapping jets from Suzuki
rings are less distorted and less three-dimensionalized as et al.3 were also made at low and moderate Reynolds num-
compared with the case of the excitation at the preferred bers (1800⬍ReD⬍13 000).
mode: this allows the jet control to be felt further down-
D. Varicose-flapping excitation
stream. This is confirmed by low-pressure isosurfaces: Fig-
ure 13 shows, for the subharmonic excitation, a zoom of the From the results of the preceding sections, it seems that
near-inlet region, which has to be compared with the similar any process able to increase the ‘‘strength’’ of the primary
zoom for preferred-mode excitation in Fig. 11. Note that it is vortex rings before throwing them away into the two bifur-
still difficult to distinguish the expected alternated-pairing cating directions, will increase the forcing performance. This
arrangement since the rings seem to be three- idea is explored with the Variflap2 forcing, which forces both
dimensionalized before such a structure may form. The en- the preferred and its sub-harmonic mode. As pointed out ear-
hanced control achieved by the subharmonic excitation is lier, the preferred frequency applies to the varicose mode
consistent with the experimental results of Suzuki et al.3 In- while the subharmonic frequency concerns the flapping ex-
deed, these authors found that the flapping excitation creates citation. Note that this perturbation is similar to the pertur-
the maximum spreading rate of the jet when applied at a bation used in the low-Reynolds number DNS performed by
frequency corresponding to the Strouhal number StrD Danaila and Boersma12 but our forcing amplitude is much
⫽ f a D/U 0 ⫽0.25 where f a is the flapping frequency, while lower and our flapping forcing concentrated within the high
the preferred mode of their jet in natural 共nonforced兲 con- shear region of the jet border. Furthermore, we vary the Rey-
figuration was StrD ⫽0.53. nolds number with three different values: ReD⫽5000
Unfortunately, the subharmonic flapping excitation re- 共VFLA2-B兲, ReD⫽25 000 共VFLA2-C兲, and ReD⫽50 000
mains highly Reynolds number dependent. This can be seen 共VFLA2-D兲.
FIG. 14. 共a兲 Bifurcating 关 ␦ 00(x) 兴 and

bisecting 关 ␦ 90(x) 兴 shear layer thick-
nesses for the VFLA2-B and
VFLA2-C simulations. The evolution
of the shear layer thickness for the
natural jet 共NHR兲 is also given for
comparison; 共b兲 bifurcating
( 具 u x00(x) 典 ) and bisecting ( 具 u x90(x) 典 )
profiles of the mean streamwise veloc-
ity for the VFLA2-B simulation; 共c兲
bifurcating ( 具 u r00(x) 典 ) and bisecting
( 具 u r00(x) 典 ) profiles of the mean radial
velocity for the VFLA2-B simulation.
D 00 and D 90 represent the bifurcating
and bisecting planes, respectively.
1. LES at ReDÄ5000 a downstream evolution which is closer to the one found in

We first concentrate on the ReD⫽5000 case in order to the natural jet 共NHR兲. It is interesting to note that, despite the
make direct comparisons with the two cases 共Flap1 and early divergence, all three excitations eventually lead to
Flap2兲 with pure flapping excitation at similar Reynolds more or less the same maximum ( 具 u x⬘ 2 典 1/2/U 1 ⬇0.17 at x/D
number. Figures 8共a兲 and 8共b兲 show the axial velocity decay ⬇7).
and rms at the centerline for the three excitation types The growth of the bifurcating and bisecting shear layers,
共Flap1, Flap2, and Variflap2兲 at several Reynolds numbers. in the Variflap2 case, can be seen in Fig. 14共a兲. Unlike the
At low Reynolds number (ReD⫽5000), we see that the axial previous excitation methods, the Variflap2 excitation yields
velocity decay is similar for the three excitations until x/D similar downstream evolutions of the bifurcating and bisect-
⬇7. It is only after that location that one sees that the Flap2 ing shear layer thicknesses until the end of the potential core
excitation leads to a faster decay than Flap1 and that the (x/D⬇4). However, for x/D⬎5 onwards, the shear layer
Variflap2 forcing exhibits a still greater decay than both flap- thickness exhibits a spectacular growth in the bifurcating
ping excitations. Since the axial velocity decay is directly plane. Note that the bisecting shear layer grows with almost
associated with the ‘‘speed’’ of the jet spreading, it indicates the same speed as the natural shear layer at the same Rey-
a greater efficiency of the Variflap2 excitation as compared nolds number 共run NHR兲. To confirm these observations,
to both flapping excitations. It is important to point out that Fig. 14共b兲 shows that the difference between the bifurcating
this enhanced spreading does not result from a greater inlet and bisecting streamwise mean velocity profiles is small be-
energy input associated with the Variflap2 excitation. This fore the end of the potential core (x/D⭐4) but becomes
can be demonstrated by looking at the rms value of the axial increasingly important afterwards. Figure 14共c兲 shows the
velocity in the near field (x/D⬍1). Indeed, Fig. 8共b兲 shows radial velocity profiles at several downstream locations for
that, for x/D⬍1, the energy level is similar for the three VFLA2-B simulation (ReD⫽5000). At x/D⫽8 and x/D
types of excitations. Conversely, for x/Dⲏ1, the down- ⫽11, high positive radial velocities are present in the bifur-
stream evolution of the rms varies with the nature of the inlet cating plane creating strong outward fluid motion. At the
perturbation. In the Variflap2 excitation case, the rms dis- same locations, the radial velocities in the bisecting plane are
plays a double peak shape with a first local maximum around negative indicating inward fluid motion and they tend to zero
x/D⬇3. This strong and fast amplification of the initial per- as one moves downstream. This suggests that for x/Dⲏ4,
turbation is characteristic of jets with a varicose inlet excita- the natural 共although enhanced兲 mechanism of entrainment
tion 共see, e.g., Crow and Champagne5 and Zaman and of fluid into the shear layer takes place in the bisecting plane
Hussain6兲. The rms with the Flap1 and Flap2 excitations has whereas, in the bifurcating plane, the shear layer grows pri-
FIG. 15. Isosurfaces of Q criteria for the Variflap2 jet at ReD⫽5000 共VFLA2-B兲. The chosen threshold is 0.25(U 1 /D) 2 and Q varies between Q Min
⫽⫺36(U 1 /D) 2 and Q Max⫽124(U 1 /D) 2 . 共a兲 View of the bifurcating plane. 共b兲 View of the bisecting plane.
marily due to the radial spreading of fluid from the rotational the bifurcating and bisecting plane is very pronounced 关see
flow region of the jet core. Figs. 15共a兲 and 15共b兲兴. Second, the coherent structures of the
Visualizations of VFLA2-B simulation at ReD⫽5000 can jet keep their cohesion until well after the potential core re-
be seen in Figs. 15 and 16. The former shows isosurfaces of gion 共see Fig. 16兲. This is caused by the primary vortex
positive Q criteria in the whole jet, while the latter shows rings, which are forced by the varicose excitation before the
isosurfaces of pressure in a zoom of the flow in the middle of flapping forcing turned them alternatively into two opposite
the computational domain situated between x/D⫽2 and directions: the alternated inclination of the rings observed for
x/D⫽8. The jet development exhibits two striking features: x/Dⲏ6 is the consequence of the flapping excitation and
first, the difference between the shear layers development in yields the formation of alternated pairings. These events are
FIG. 16. Isosurfaces of low pressure showing vortices from the primary structures until the middle of the computational domain. For the Variflap2 excitation
for ReD⫽5000 共VFLA2-B兲. The chosen threshold is ⫺0.05␳ 0 U 21 and p varies between p Min⫽⫺0.64␳ 0 U 21 and p Max⫽0.109␳ 0 U 21 .
FIG. 17. 共a兲 Downstream evolution of

E r and E ␪ 关defined in Eqs. 共27兲 and
共28兲兴 for the simulations FLA1-B,
FLA2-B, and VFLA2-B; 共b兲 anisot-
ropy invariant map for the simulations
FLA1-B, FLA2-B, and VFLA2-B with
the invariants II and III computed
from the Reynolds stresses tensor at
the centerline r/D⫽0, and in stations:
x/D⫽2.0; 2.5; 3.0; 3.5; 4.0; 4.5; 5.0.
forced to take place always in the bifurcating plane and are E r 共 x 兲 ⫽ 共 E r00共 x 兲 ⫹E r90共 x 兲兲 /2, 共27兲
directly responsible for the high spreading rate of the jet.
It is quite striking to see that a relatively low level of E ␪ 共 x 兲 ⫽ 共 E ␪00共 x 兲 ⫹E ␪90共 x 兲兲 /2, 共28兲
forcing yields, at the beginning of the jet evolution, so well
where
formed vortex rings which are not distorted by the small
scale turbulence structures which are quite energetic at high
Reynolds numbers. One may also observe that, unlike with
the pure flapping 共Flap1 and Flap2兲 excitations, the primary
E r00共 x 兲 ⫽ 冑冉冕1
L yLz
2␲
D 00
冊
具 u r⬘ 2 共 x,r 兲典 r dr , 共29兲
rings show no sign of any azimuthal perturbation. To analyze

this point further we computed the quantities, E r90共 x 兲 ⫽ 冑冉冕1
L yLz
2␲
D 90
冊
具 u r⬘ 2 共 x,r 兲典 r dr , 共30兲
FIG. 18. Isosurfaces of low pressure for the Variflap2 jet at ReD⫽25 000 共VFLA2-C兲. The chosen threshold is ⫺0.05␳ U 21 and p varies between p Min
⫽⫺0.44␳ U 21 and p Max⫽⫹0.11␳ U 21 . 共a兲 View of the bifurcating plane. 共b兲 View of the bisecting plane.
FIG. 19. Isosurfaces of Q criteria for the Variflap2 jet at ReD⫽50 000 共VFLA2-D兲. The chosen threshold is 0.42(U 1 /D) 2 and Q varies between Q Min
⫽⫺66(U 1 /D) 2 and Q Max⫽120(U 1 /D) 2 . 共a兲 View of the bifurcating plane; 共b兲 view of the bisecting plane.
E ␪00共 x 兲 ⫽ 冑冉冕
1
L yLz
2␲
D 00
冊
具 u ␪⬘ 2 共 x,r 兲典 r dr , 共31兲
firms the visual observations discussed above for Fig. 16.
Figure 17共a兲 furthermore shows that E r , which is associated
with the primary instabilities and the formation of vortex
E ␪90共 x 兲 ⫽ 冑冉冕
1
L yLz
2␲
D 90
冊
具 u ␪⬘ 2 共 x,r 兲典 r dr , 共32兲
rings,39,40 largely dominates the flow in the case of the
varicose-flapping excitation, conversely to what is observed
with the other 共flapping兲 excitations. A direct consequence of
D 00 and D 90 mean that the integration is carried out in the this observation is the stronger anisotropy of the flow with
bifurcating and bisecting plane, respectively. The quantities the varicose-flapping excitation. This can be seen in Fig.
E r (x) and E ␪ (x) represent the respective contribution of the 17共b兲 which displays the invariant anisotropy map for the
radial and azimuthal Reynolds stresses to the turbulent ki- tree excitation types, along the centerline. The figure shows
netic energy at a given x location. Figure 17共a兲 shows the
that with the Flap1 and Flap2 excitations the flow attains
downstream evolution of E r and E ␪ for the Flap1, Flap2, and
isotropy very fast compared to the Variflap2 excitation in
Variflap2 excitations, at the same Reynolds number, ReD
which the flow remains axisymmetric on much larger dis-
⫽5000. The graph shows that until x/D⬇5, E ␪ is smaller in
tances from the nozzle. The weak ‘‘three dimensionality’’ of
the Variflap2 than for the Flap1 and Flap2 simulations. More-
over, in the initial stages of transition (x/D⬍3), the growth the primary rings in the varicose-flapping case implies that
rate ⳵ E ␪ / ⳵ x is much smaller in the case of the varicose- these structures survive longer than with the other 共flapping兲
flapping excitation than with a pure flapping excitation. As excitations. Due to their long lifetime, the primary vortices
shown in previous works39,40 E ␪ is a way of representing the then undergo partial vortex merging events 共alternated pair-
departure from axisymmetry of the flow. Figure 17共a兲 there- ings兲 well after the potential core region: this yields a stron-
fore demonstrates that the vortex rings from Variflap2 are ger forcing efficiency.
less strongly three dimensionalized than the ones found with Note that, although the excitation used by Danaila and
the other 共flapping兲 excitations. Since the degree of three Boersma12 was similar to our Variflap2 excitation, it was not
dimensionality of an isolated vortex ring is related to its identical since a much larger amplitude was used and it was
azimuthal instability 共whatever its mode is兲, this graph con- applied over the whole inlet diameter. Consequently, no al-
TABLE III. Decay rates 关 B u from Eq. 共16兲兴; spreading angles 关 ␤ 00 and ␤ 90 to simulation VFLA2-D. Only visualizations of this simula-
from Eqs. 共33兲 and 共34兲兴; forcing areas A J ; and forcing thicknesses D J for tion are shown here 关see Figs. 19共a兲 and 19共b兲兴. The pictures
the three excitation types, at several Reynolds numbers. The decay rates and
spreading angles were computed from data in the region 6.5⬍x/D⬍11.5.
of positive Q criterion show that Variflap2 is still able to
control the jet spatial development, since the enhanced
Simulation Reynolds Forcing Bu ␤ 00 ␤ 90 AJ DJ spreading in the bifurcating plane can still be observed. The
NHR 2500 共—兲 5.7 0.104 0.104 共—兲共—兲 isosurfaces of low pressure 共not shown兲 showed that the pri-
mary vortex rings maintain their cohesion until around x/D
FLA1-A 1500 Flap1 5.4 0.206 0.122 7.0 196 ⬇4 and these again give rise to large scale structures distin-
FLA1-B 5000 Flap1 6.3 0.098 0.126 2.6 72 guishable until the end of the computational box.
FLA2-B 5000 Flap2 5.8 0.125 0.111 2.7 88 Table III summarizes the results concerning the influence
VFLA2-B 5000 Variflap2 5.9 0.184 0.109 11.7 620 of the Reynolds number on the control efficiency by listing
FLA2-C 25 000 Flap2 4.1 0.110 0.116 0.8 22 the decay rate, B u 关see Eq. 共16兲兴 and the spreading angles
VFLA2-C 25 000 Variflap2 4.2 0.167 0.115 11.2 546 along the bifurcating ␤ 00 , and bisecting ␤ 90 planes,
VFLA2-D 50 000 Variflap2 共—兲共—兲共—兲 11.1 537 ⳵ ␦ 00共 x 兲
␤ 00⫽ , 共33兲
⳵x
ternated pairings were observed in the low-Reynolds DNS ⳵ ␦ 90共 x 兲

␤ 90⫽ . 共34兲
performed by these authors. ⳵x
The difference of the jet spreading in the bifurcating and
2. Influence of the Reynolds number bisecting planes has been characterized both by the shear-
The same excitation method was used at a higher Rey- layer thicknesses ␦ 00(x) and ␦ 90(x) and by the longitudinal
nolds number, ReD⫽25 000, which corresponds to the simu- mean velocity profiles 具 u x00(x,r) 典 and 具 u x90(x,r) 典 . Based on
lation VFLA2-C. We have checked that the various statistical these quantities, we introduce what we call the forcing thick-
quantities both in the bifurcating and bisecting planes show a ness,
very weak dependence on the Reynolds number when com-
pared with those of Figs. 14共b兲 and 14共c兲. In particular, the
shear layers growth in the bifurcating and bisecting planes
D J⫽
1
D2
冕 0
12D
兩 ␦ 90共 x 兲 ⫺ ␦ 00共 x 兲兩 dx, 共35兲
关see Fig. 14共a兲兴 is as impressive as in the lower Reynolds
number case. and the forcing area,
The flow visualizations by Q criterion 共not shown兲 con-
firmed the efficiency of Variflap2 excitation by showing that,
although the level of small scale turbulence is higher and
A J⫽ 兺冕兩具ux 90共 x,r 兲典⫺ 具ux 00共 x,r 兲典兩dr,
x/D⫽4,8,11 U 1 D 0
1 ⬁
共36兲
affects more than previously the primary vortex rings, the
latter still exhibit shapes similar to the ones found at lower to assess the performance of the forcing method. The values
Reynolds numbers in simulation VFLA2-B. In particular the of both quantities are also shown in Table III and in Figs.
formation of alternated pairings around the middle of the 20共a兲 and 20共b兲 for the simulations Flap1, Flap2, and Vari-
computational domain with their characteristic large-scale flap2 at several Reynolds numbers. The numerical results
zigzag arrangement in the bifurcating plane, can still be ob- show that below a forcing thickness of about D J ⬇40 the
served 共see Fig. 18兲. possibility of controlling jet can no longer be felt neither for
To further investigate the Reynolds number dependency, the shear layers growth nor for the mean axial and radial
another LES was carried out at ReD⫽5⫻104 corresponding velocity profiles. The strong dependence of the Flap1 forcing
FIG. 20. Forcing area A J , and forcing

thickness D J as a function of the Rey-
nolds number, for the three excitation
methods. 共a兲 Forcing area A J ; 共b兲
forcing thickness D J .
3. Influence of the forcing frequency
To demonstrate that the varicose-flapping excitation at-

tains its maximum efficiency if the flapping excitation is ap-
plied at the subharmonic mode, two extra LES are carried
out at Reynolds number ReD⫽5000. In both cases, the exci-
tation is composed of a combination of a varicose and of a
flapping excitation. The varicose mode is always excited at
the preferred frequency, while the excitation frequency of the
flapping mode is varied: the simulation VFLA1-B corre-
sponds to a flapping excitation at the preferred mode and
VFLA4-B to a flapping excitation at half the subharmonic
mode. The growth of the bifurcating 关 ␦ 00(x) 兴 and bisecting
关 ␦ 90(x) 兴 shear layer thicknesses are shown in Fig. 21. The
analogous curves for the Variflap2 excitation at the same
Reynolds number 共simulation VFLA2-B兲 are also given for
FIG. 21. Bifurcating 关 ␦ 00(x) 兴 and bisecting 关 ␦ 90(x) 兴 shear layer thicknesses comparison. One can see that, for the VFLA1-B simulation,
for the Variflap1, Variflap2, and Variflap4 simulations at the same Reynolds
number (ReD⫽5000).
a high spreading rate is achieved along the bifurcating plane
关see ␦ 00(x) for VFLA1-B兴, but both the bifurcating and bi-
secting spreading rates are significantly smaller than the ones
efficiency on the Reynolds number is particularly clear in obtained with the Variflap2 excitation. The value of the forc-
Figs. 20共a兲 and 20共b兲. As previously shown, even for mod- ing thickness computed for VFLA1-B is D J ⫽552, whereas
erate Reynolds numbers (ReD⫽5000) the Flap1 excitation VFLA2-B has D J ⫽620.
looses its efficiency. Conversely, the effect of the Flap2 ex- Figure 22共a兲 displays, through low-pressure isosurfaces,
citation is felt at moderate Reynolds number with quite a the coherent structures present at the beginning of the com-
large value of D j . At ReD⫽25 000, the low values of A j and putational domain for the simulation VFLA1-B. Due to the
D j show that the Flap2 excitation has lost most of its effect flapping excitation at the preferred mode, the consecutive
on the jet spatial development. For the Variflap2 excitation vortex rings are now inclined in the same direction and do
cases, both the values of A J and D J are quite high compared not exhibit the alternated arrangement achieved with the
with the other forcing methods at same Reynolds number. Variflap2 forcing 共see Fig. 16兲: the alternated pairings are no
Furthermore, one can notice that this excitation method is longer efficiently triggered and the jet spreading in the bifur-
very weakly sensitive to the value of the Reynolds number. cating plane is less pronounced than with the subharmonic
Indeed, A J is almost identical in the two cases ReD⫽5000 excitation of the flapping mode.
and ReD⫽25 000 while D J decreases slightly with increasing The excitation of the flapping mode at half the subhar-
Reynolds number. Note that Crow and Champagne5 found a monic frequency 共simulation VFLA4-B兲 has even a smaller
critical Reynolds number for which the varicose excitation effect on the jet spatial development. This is apparent from
could not affect the jet any more. In their case, this limit was the downstream evolution of its shear layer bifurcating and
around ReD⫽7⫻104 . Further investigations at higher Rey- bisecting thicknesses 共also shown in Fig. 21兲. There is almost
nolds numbers should therefore be performed to determine if no difference between the two shear layer thicknesses which
such a limiting Reynolds number exists in the present case. shows that this forcing type is not capable of creating a bi-
FIG. 22. 共a兲 Isosurfaces of low pressure for the Variflap1 jet at ReD⫽5000 共VFLA1-B兲. The chosen threshold is ⫺0.07␳ 0 U 21 and p varies between p Min
⫽⫺0.36␳ 0 U 21 and p Max⫽0.10␳ 0 U 21 . The figure shows only part of the computational domain from x/D⫽0 to x/D⫽6 seen from the bifurcating plane; 共b兲
isosurfaces of positive Q criteria for the Variflap4 jet at ReD⫽5000 共VFLA4-B兲. The chosen threshold is ⫹3.67(U 1 /D) 2 and Q varies between Q Min
⫽⫺46(U 1 /D) 2 and Q Max⫽⫹74(U 1 /D) 2 . The figure shows only part of the computational domain from x/D⫽2 to x/D⫽5 seen from the bifurcating plane.
furcating jet. Visualizations of the near field coherent struc- Flap2 has lost all its bifurcating effect. The reason for this
tures for Variflap4 关see Fig. 22共b兲兴 reveal that the primary comes from the intensification of small scale turbulence as
vortices, at x/D⬇2.5, are very close to the ones which would the Reynolds number increases. This small-scale turbulent
have been found if a purely varicose excitation was applied activity causes the destruction of the larger flow vortices
共see, e.g., Urbin and Métais10兲. However, the flapping exci- which leads quickly to the loss of flow control.
tation seems to yield an early breakdown of the rings, which The second forcing method 共Variflap兲 consists in the
makes the jet control inefficient. combination of a varicose mode excitation at the preferred
jet frequency and a flapping mode excitation. The frequency
of the flapping excitation was first chosen equal to the sub-
VII. CONCLUSIONS
harmonic frequency 共Variflap2 simulation兲: this constitutes
The goal of the present study was to design efficient the most natural choice to efficiently trigger the alternated
control strategies for spatially evolving round jets. Our meth- pairing structure. It reproduces the type of forcing used in the
odology was first to reach a good understanding of the co- experimental works of Lee and Reynolds1 and Parekh et al.2
herent structures dynamics in the natural, unforced jet and This forcing method was also tested at low Reynolds number
second to use this information to design deterministic exci- in the DNS by Danaila and Boersma.12 From all the forcing
tations of the jet inlet to efficiently control the flow even at analyzed, the Variflap2 causes by far the highest bifurcating
high Reynolds number. Our LES of the natural jet at ReD spreading rates while being also insensitive to the Reynolds
⫽25 000 have confirmed the results obtained by previous number increase, at least for the range of Reynolds number
authors 共Comte et al.36 and Urbin and Métais10兲: the growth investigated here 共up to ReD⫽50 000). The analysis of the
of the varicose mode at a preferred frequency StrD ⫽0.38 coherent structures reveals that the Variflap2 excitation man-
gives rise to axisymmetric vortex rings near the jet nozzle. ages to create vigorous primary vortex rings which can travel
Further downstream, the growth of a sub-harmonic fre- far away downstream while preserving their coherence, since
quency yields vortex structures consisting in alternated local- these are weakly affected by the growth of small scale tur-
ized pairings between consecutive rings: the alternated- bulence. Their long lifetime makes the action of the flapping
pairing structure. Urbin and Métais10 have designed a excitation more effective by triggering, near the end of the
specific excitation aimed at triggering this alternated-pairing potential core, a zigzag arrangement of the rings characteris-
structures and shown that, which such an excitation, the jet tics of the alternated-pairing structure. The influence of the
behavior is greatly modified. Indeed, the jet expansion was flapping frequency excitation has been studied: the greater
found to be greatly enhanced along one given radial plane in efficiency of the subharmonic forcing clearly shows that the
similarities with the so-called bifurcating jets. This specific jet is far more responsive to inlet excitations in good corre-
excitation was later called the flapping excitation by Danaila spondence with its natural vortical structures.
and Boersma.12 The dissipative nature of the numerical Finally, we want to stress that the present analysis is
scheme used by Urbin and Métais10 casts some doubt on the restricted to incompressible flows. At high Mach numbers
effective Reynolds number of their simulations. Therefore, 共such as used in Zaman et al.7兲 the compressible Navier–
we have here revisited their work with a very precise numeri- Stokes equations have to be used instead. As pointed out by
cal code. Moreover, we focused on the influence of the Rey- a reviewer it would be interesting to analyze also the acoustic
nolds number on the excitations through the use of LES with field resulting from each excitation type.
ReD ranging from 1.5⫻103 to 5.0⫻104 . The influence of the
forcing frequency was also carefully examined. Furthermore, ACKNOWLEDGMENTS
we have used forcing methods in closer correspondence with
experimental bifurcating jets 共Lee and Reynolds,1 Parekh Carlos B. da Silva is supported by the Portuguese gov-
et al.,2 Suzuki et al.3兲, although using a lower amplitude ernment under the scholarship PRAXIS XXI 共5726/95兲. The
forcing, more realistic for practical industrial applications. simulations reported in the present paper were carried out in
Two different types of forcing were investigated. The a machine NEC SX5 from the Institut du Developpement et
first forcing 共Flap兲 consists in an excitation of the flapping des Ressources en Informatique Scientifique 共IDRIS兲.
mode only. Similarly to Urbin and Métais,10 the forcing fre-
quency for this mode was first taken equal to the preferred 1
M. Lee and W. C. Reynolds, ‘‘Bifurcating and blooming jets at high Rey-
mode. This forcing type succeeded in causing the jet bifur- nolds number,’’ 5th Symposium on Turbulent Shear Flows, New York,
cation, but only at low Reynolds numbers. The reason comes 1992.
2
D. Parekh, A. Leonard, and W. C. Reynolds, ‘‘Bifurcating jets at high
from the nature of the primary vortex rings which, being Reynolds numbers,’’ Report No. TF-35, Department of Mechanical Engi-
highly distorted by the forcing method, cause a quick transi- neering, Stanford University, 1988.
3
tion to turbulence with the consequent fragmentation of the H. Suzuki, N. Kasagi, and Y. Suzuki, ‘‘Active control of an axisymmetric
jet with an intelligent nozzle,’’ 1st Symposium on Turbulent Shear Flow
bigger vortices and loss of flow control. We next applied the
Phenomena, 2000.
flapping forcing at a the subharmonic frequency 共Flap2 4
E. J. Gutmark and F. F. Grinstein, ‘‘Flow control with noncircular jets,’’
simulation兲. This corresponds to the forcing type used in the Annu. Rev. Fluid Mech. 31, 239 共1999兲.
experiments of Suzuki et al.3 and in the numerical simula-
5
S. C. Crow and F. H. Champagne, ‘‘Orderly structure in jet turbulence,’’ J.
Fluid Mech. 48, 547 共1971兲.
tions of Danaila and Boersma.12 This second method 6
K. B. M. Q. Zaman and A. K. M. F. Hussain, ‘‘Vortex pairing in a circular
achieves greater jet bifurcation and works at higher Reynolds jet under controlled excitation. Part 1. General jet response,’’ J. Fluid
numbers. However, for a Reynolds number of ReD⫽25 000, Mech. 101, 449 共1980兲.
7 24
K. B. M. Q. Zaman, M. F. Reeder, and M. Samimy, ‘‘Control of an axi- I. Orlansky, ‘‘A simple boundary condition for unbounded hyperbolic
symmetric jet using vortex actuators,’’ Phys. Fluids 6, 778 共1994兲. flows,’’ J. Comput. Phys. 21, 251 共1976兲.
8 25
E. K. Longmire and L. H. Duong, ‘‘Bifurcating jets generated with stepped K. Akselvoll and P. Moin, ‘‘Large-eddy simulation of turbulent confined
and sawtooth nozzles,’’ Phys. Fluids 8, 978 共1996兲. coannular jets,’’ J. Fluid Mech. 315, 387 共1996兲.
9 26
T. C. Corke and S. M. Kusec, ‘‘Resonance in axisymmetric jets with M. A. Gonze, ‘‘Simulation Numérique des Sillages en Transition à la
controlled helical-mode input,’’ J. Fluid Mech. 249, 307 共1993兲. Turbulence,’’ Ph.D. thesis, Institut National Politechnique de Grenoble,
10
G. Urbin and O. Métais, ‘‘Large-eddy simulations of three-dimensional 1993.
spatially-developing round jets,’’ in Direct and Large-Eddy Simulations II, 27
S. Stanley, S. Sarkar, and J. P. Mellado, ‘‘A study of the flowfield evolu-
edited by J. P. Chollet, P. R. Voke, and L. Kleiser 共Kluwer Academic, New tion and mixing in a planar turbulent jet using direct numerical simula-
York, 1997兲. tion,’’ J. Fluid Mech. 450, 377 共2002兲.
11
I. Danaila and J. Boersma, ‘‘Mode interaction in a forced homogeneous jet 28
A. Michalke and G. Hermann, ‘‘On the inviscid instability of a circular jet
at low Reynolds numbers,’’ Center for Turbulent Research, Proceedings of with external flow,’’ J. Fluid Mech. 114, 343 共1982兲.
the Summer Program 1998 共Stanford University, Stanford, 1998兲. 29
P. Freymuth, ‘‘On transition in a separated laminar boundary layer,’’ J.
12
I. Danaila and J. Boersma, ‘‘Direct numerical simulation of bifurcating Fluid Mech. 25, 683 共1966兲.
jets,’’ Phys. Fluids 12, 1255 共2000兲. 30
13 J. C. R. Hunt, A. A. Wray, and P. Moin, ‘‘Eddies, stream, and convergence
A. Hilgers, ‘‘Parameter optimization in jet flow control,’’ Annual Research
zones in turbulent flows,’’ Annual Research Briefs 共Center for Turbulence
Briefs 共Center for Turbulence Research, Stanford University, Stanford,
Research, Stanford University, Stanford, 1988兲.
1999兲. 31
14 I. Dubief and F. Delcayre, ‘‘On coherent-vortex identification in turbu-
A. Hilgers, ‘‘Control and optimization of turbulent jet mixing,’’ Annual
lence,’’ J. Turbulence 011 共2000兲.
Research Briefs 共Center for Turbulence Research, Stanford University, 32
J. Cohen and I. Wygnaski, ‘‘The evolution of instabilities in the axisym-
Stanford, 2000兲.
15 metric jet. Part 1. The linear growth of the disturbances near the nozzle,’’
A. Hilgers and B. Boersma, ‘‘Optimization of turbulent jet mixing,’’ Fluid
J. Fluid Mech. 176, 191 共1987兲.
Dyn. Res. 29, 345 共2001兲. 33
16
P. Koumoutsakos, J. Freund, and D. Parekh, ‘‘Evolution strategies for H. J. Hussein, S. P. Capp, and W. K. George, ‘‘Velocity measurements in
parameter optimization in jet flow control,’’ Center for Turbulent Re- a high-Reynolds-number, momentum-conserving, axisymmetric, turbulent
search, Proceedings of the Summer Program 1998 共Stanford University, jet,’’ J. Fluid Mech. 258, 31 共1994兲.
34
Stanford, 1998兲. J. L. Lumley and G. W. Newman, ‘‘The return to isotropy of homogeneous
17
M. Lesieur and O. Métais, ‘‘New trends in large-eddy simulations of tur- turbulence,’’ J. Fluid Mech. 82, 161 共1977兲.
35
bulence,’’ Annu. Rev. Fluid Mech. 28, 45 共1999兲. E. Gutmark and C.-M. Ho, ‘‘Preferred modes and the spreading rates of
18
F. Ducros, P. Comte, and M. Lesieur, ‘‘Large-eddy simulation of transition jets,’’ Phys. Fluids 26, 2932 共1983兲.
36
to turbulence in a boundary layer developing spatially over a flat plate,’’ J. P. Comte, Y. Fouillet, and M. Lesieur, ‘‘Simulation numerique des zones
Fluid Mech. 326, 1 共1996兲. de mèlange compressibles,’’ Revue Scientifique et Technique de la De-
19
O. Métais and M. Lesieur, ‘‘Spectral large-eddy simulation of isotropic fense, 3ème Trimestre, 1992.
37
and stably stratified turbulence,’’ J. Fluid Mech. 239, 157 共1992兲. R. A. Petersen, ‘‘Influence of wave dispersion on the vortex pairing in a
20
S. K. Lele, ‘‘Compact finite difference schemes with spectral-like resolu- jet,’’ J. Fluid Mech. 89, 469 共1978兲.
38
tion,’’ J. Comput. Phys. 103, 15 共1992兲. G. Broze and F. Hussain, ‘‘Transition to chaos in a forced jet: inter-
21 mitency, tangent bifurcations and hysteresis,’’ J. Fluid Mech. 311, 37
C. Canuto, M. Y. Hussaini, A. Quarteroni, and T. A. Zang, Spectral Meth-
ods in Fluid Dynamics 共Springer, New York, 1987兲. 共1996兲.
22 39
J. H. Williamson, ‘‘Low-storage Runge–Kutta schemes,’’ J. Comput. P. Brancher, J. M. Chomaz, and P. Huerre, ‘‘Direct numerical simulations
Phys. 35, 48 共1980兲. of round jets: Vortex induction and side jets,’’ Phys. Fluids 6, 1768 共1994兲.
23 40
C. B. da Silva, ‘‘The role of coherent structures in the control and inter- I. Danaila, J. Dusek, and F. Anselmet, ‘‘Coherent structures in a round,
scale interactions of round, plane and coaxial jets,’’ Ph.D. thesis, Institut spatially evolving, unforced, homogeneous jet at low Reynolds numbers,’’
National Politechnique de Grenoble, 2001. Phys. Fluids 9, 3323 共1997兲.

Vortex Control of Bifurcating Jets: A Numerical Study: Carlos B. Da Silva and Olivier Me Tais

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Vortex Control of Bifurcating Jets: A Numerical Study: Carlos B. Da Silva and Olivier Me Tais

Uploaded by

Copyright:

Available Formats

PHYSICS OF FLUIDS VOLUME 14, NUMBER 11 NOVEMBER 2002

Vortex control of bifurcating jets: A numerical study

I. INTRODUCTION diameter and U 1 is the maximum axial velocity at the inlet兲.

1070-6631/2002/14(11)/3798/22/$19.00 3798 © 2002 American Institute of Physics

FIG. 2. 共a兲 Streamwise velocity pro-

FIG. 3. 共a兲 Anisotropy invariant map

the experimental data, which is illustrated by the two particu- I⫽b ii ,

⬍0.023, Gutmark and Ho35兲. Further downstream, at the jet

FIG. 5. Sketch of the vortex rings arrangement leading

reality than previous numerical simulations 共e.g., Danaila f pD

TABLE I. List of the excitation types.

Varicose at the preferred mode frequency⫹

Flapping at the preferred mode frequency 共22兲

FIG. 7. 共a兲 Time spectrum of the axial

FIG. 8. 共a兲 Axial velocity decay in the

bifurcating plane the jet spreads much more rapidly at least

FIG. 11. Isosurfaces of low pressure

tations can be compared by looking at Figs. 9 and 12. The

FIG. 14. 共a兲 Bifurcating 关 ␦ 00(x) 兴 and

1. LES at ReDÄ5000 a downstream evolution which is closer to the one found in

FIG. 17. 共a兲 Downstream evolution of

rings show no sign of any azimuthal perturbation. To analyze

ternated pairings were observed in the low-Reynolds DNS ⳵ ␦ 90共 x 兲

FIG. 20. Forcing area A J , and forcing

3. Influence of the forcing frequency

To demonstrate that the varicose-flapping excitation at-

You might also like