Sie sind auf Seite 1von 14

The oscillatory instability of plane variable-density jets

L. Raynal, J-L. Harion, M. Favre-Marinet,a) and G. Binder

Laboratoire des Ecoulements Geophysiques et Industriels-IMG, UJF-CNRS-INPG,b) B.P. 53 X-38041


Grenoble Cedex, France

~Received 4 May 1994; accepted 8 November 1995!


The results of an experimental investigation of the instability of variable-density plane jets issuing
into ambient air are reported. When the jet to ambient fluid density ratio S(S5 r j / r ` ) is less than
a critical value, an intense oscillating instability is observed. This instability is characterized by
sharp peaks in the power spectral density measured in the near field of the jet. The effects of the
control parameters S, Re, and H/u ~jet width to exit momentum thickness! on the instability regime
are determined. It is shown that Re is a better scaling parameter than H/u. The Strouhal number of
the dominant mode StH increases with S and Re up to a constant value of 0.25, which is in rather
good agreement with the theory and the experiments of Yu and Monkewitz @Phys. Fluids A 2, 1175
~1990!; J. Fluid Mech. 255, 323 ~1993!#. In the present experiments the critical value S c above
which the oscillating regime disappears is an increasing function of Re and S c seems to reach a
limiting value in the neighborhood of 0.7, which does not agree well either with the theory or with
the experiments of Yu and Monkewitz @Phys. Fluids A 2, 1175 ~1990!; J. Fluid Mech. 255, 323
~1993!#. This difference is in qualitative agreement with the results of linear stability computations,
also reported in the paper, which take into account differences in shape and relative positions of the
density and velocity profiles. 1996 American Institute of Physics. @S1070-6631~96!00703-2#

I. INTRODUCTION

The stability of variable-density jets has attracted much


attention in recent years. Two striking phenomena are observed in plane and round nonhomogeneous jets. Power
spectra present sharp peaks when the density ratio is small
enough and flow visualizations reveal regular symmetric
structures and simultaneously violent lateral ejections of
fluid in the form of side jets in the initial region of the jets.
These phenomena have first been observed in variable density round jets.1 4 They may, to a certain extent, be interpreted as a consequence of absolute instability of the flow.
The general concept of absolute or convective instability is
not new ~see, for instance, Drazin and Reid5!, but the existence of such an instability regime in mixing layers and
variable-density jets has been shown possible only in 1985
by Huerre and Monkewitz.6 The problem of the impulse response of a round heated jet has been addressed by Monkewitz and Sohn.7 Their frequency predictions from the linear
theory of nonviscous parallel flow compare rather well with
the experimental data obtained in variable density round jets
by Kyle and Sreenivasan8 ~referenced later as KS8!.
The present investigation originated from a research
project on the effects of large-density differences on turbulent transport properties within the defunct European space
shuttle program HERMES. Our attention to the strongly oscillating instability mode in variable-density jets was drawn
by the clearly audible pure tone sound emitted under certain
conditions by the jet facility built for the calibration of hot
wires versus density and velocity.
The aim of the present work is to determine the influence
of various parameters-S, H/u, Reon the nature of the inTel: ~33! 76.82.50.49; fax: ~33! 76.82.52.71; electronic mail: favre@img.fr
Universite Joseph Fourier, Centre National de la Recherche Scientifique,
Institut National Polytechnique de Grenoble.

a!

b!

Phys. Fluids 8 (4), April 1996

stability in plane jets ~S5 r j /r` , ratio of jet to ambient fluid


density; H, jet width; u momentum thickness in the exit section; Re5U j H/ n j , jet Reynolds number based on the velocity, U j , and the viscosity, n j , of the jet fluid in the exit
section!. When the investigation was started, the only experimental published results were on nonhomogeneous round
jets. Meanwhile, this paper was preempted by the paper of
Yu and Monkewitz9 ~referenced YM9 hereafter!, dealing with
the oscillations of a two-dimensional heated jet. Fortunately
the two investigations only partially overlap and... partly disagree! In YM9 the main emphasis is on an extensive spatial
exploration of the instability characteristics, a detailed analysis of the critical value of S, nonlinear interactions, and results on the entrainment, but only within a limited range for
S ~0.731! and at nearly fixed Reynolds number. As will be
discussed later, the main disagreement concerns the critical
value of S above which the oscillating mode disappears. The
observations of YM9 yield 0.9, which is close to the predictions of YM10 and Monkewitz et al.,11 while in the present
work no oscillations were observed when S.0.7. As shown
by linear stability calculation results, this difference seems to
be mainly due to the effect of the relative positions of the
density and velocity profiles, and also to the difference in the
shape of these profiles.
II. EXPERIMENTAL SETUP

Density differences were obtained by mixing air and helium in varying proportions. The mixing device is made up
of 20 calibrated sonic throats of different diameters ~0.153
mm for air and 0.12 mm for helium! supplied by a tank of
compressed air and by a compressed-helium storage bottle
via regulating valves. The choice of the opened or closed
throats and of the generating pressures upstream of the sonic
throats determines the mass flow rate and the composition of
the flow. This system has the advantage of producing flows

1070-6631/96/8(4)/993/14/$10.00

1996 American Institute of Physics

993

Downloaded30Sep2010to202.3.77.11.RedistributionsubjecttoAIPlicenseorcopyright;seehttp://pof.aip.org/about/rights_and_permissions

FIG. 2. Definition of quality factor.

FIG. 1. Experimental facility.

with rigorously constant velocities and concentrations and to


block perturbations from the regulating valves or from the air
compressor. The jet to ambient air density ratio S could thus
be varied continuously from 0.14 ~pure helium! to 1 ~pure
air! and the velocity of the narrowest jet from 0.2 to 120 m/s.
The mixed gas enters the settling chamber ~12 cm324
cm35 cm! through a conic porous diffuser, then flows
through three grids, a honeycomb, a porous plate, and issues
into the ambient air after a converging nozzle ending in a slot
of width H ~Fig. 1!. The contraction is designed by using two
matched cubic equations ~Morel12!. The jet facility permits
the change of the jet width from 1 to 11 mm by adding
inserts in the settling chamber. Three different values of H
were used in this study: 2.07, 3.04, and 4.28 mm, which
correspond to contraction ratios of 49, 34, and 24, respectively. The length of the slot ~in the z direction; x,y,z represent the streamwise, normal, and spanwise coordinates! is 50
mm. A porous plate is placed just upstream the contraction to
prevent excitation of the jet by resonant oscillations of the
settling chamber. Two glass plates parallel to the x-y plane at
the ends of the nozzle guide the flow in order to improve its
two-dimensionality. The solid surface of the exit plane ~x50!
extends to 20 cm from the jet axis in order to provide welldefined boundary conditions. Note that the Reynolds number
Re is computed with the viscosity of the jet fluid at the
nozzle exit, contrary to YM,9 who use the viscosity of the
ambient fluid. The values of the Richardson number being
very low @R i 5( r ` 2 r j )Hg/ r j U 2j , R imax 5 4.9 3 1023 !, buoyancy effects are negligible.
Measurements of the longitudinal velocity U were performed with a hot wire ~Wollaston wire 6 mm diam and 1
mm active length! and a DISA 55MO1 constant temperature
anemometer. Data processing was done with an analog
digital converter ~IOtech 488/16, 16 bit conversion, maximum sampling frequency 100 kHz! connected to a Macintosh IIci computer. The bucking voltage fed to the A/D
converter in order to improve the resolution was provided by
digitalanalog converter ~IOtech 488/4! controlled by the
994

Phys. Fluids, Vol. 8, No. 4, April 1996

computer. The probe was mounted on a sting whose position


was also controlled by the computer. For measurements of
velocity profiles through the jet, the hot wire was located at
about 0.6 mm downstream of the nozzle exit plane.
The spectra shown are obtained directly from the anemometer output signal. Since the amplitudes of the velocity
fluctuations are small in the potential core where the spectra
were measured, they are a good approximation of true velocity spectra. They were computed with a 1024 fast Fourier
transform ~FFT! algorithm using a flat top window. The
spectra shown are an average of eight realizations. This number is large enough to obtain a stable average. Indeed, when
the number of realizations was increased from 5 to 100, there
were no changes in the frequency value of the peak and in
the peak-to-base ratio, but only a smoothing of the averaged spectra. This has been checked for various values of S.
Some spectra were obtained by using a 10 mm diam microphone located outside the flow ~x/H50, y/H'25!. The microphone signal was amplified ~3100! and filtered between
50 and 5000 Hz.
A spectral peak may be characterized by a quality factor
Q defined as Q5 f 0 /D f , where f 0 is the peak frequency and
D f is the width where the spectral level E( f ) has dropped
by a factor 10 ~see Fig. 2!. Because of the finite record length
T ~here T51024/f s , where f s is the sampling frequency! the
measured spectrum is the result of the convolution of the true
spectrum by the function @sin( p f T)/ p f t#.2 Hence, for a
purely sinusodal signal, D f is not zero but corresponds to
a5p(D f /2)T, such that ~sin a!/a51/~10!1/2. This yields
a50.738p, i.e D f /250.738/T. The measured quality factor
for
a
purely
sinusoidal
signal
would
be
1024/~2*0.738!f 0 / f s 5694 f 0 / f s . Furthermore, because of
the discrete description of the spectrum, the true value of the
finite record peak is a function of its location between two
successive multiples of the sampling step ~D f s 5 f s /1024!.
The peak value is further reduced when f 0 falls between the
values n D f s and (n11)D f s , the reduction being maximum
when f 0 5(n11/2)D f s . This reduction was determined experimentally with a wave generator. It was found that
Q discrete/Q continuous'0.5 when f 0 5(n1 21 )D f s , and 0.8 when
f 0 5n D f s . Since the exact value of f 0 is not known, all that
can be said is that the reduction is between 0.5 and 0.8. For
want of better, the average value 0.65 was selected. The
quality factor is, therefore, affected by an uncertainty of
625%.
For a purely sinusoidal signal the measured quality factor is consequently 0.65*694f 0 / f s 5450 f 0 / f s . In order to
take into account the finite record length and discrete deRaynal et al.

Downloaded30Sep2010to202.3.77.11.RedistributionsubjecttoAIPlicenseorcopyright;seehttp://pof.aip.org/about/rights_and_permissions

FIG. 3. Frequency of oscillations versus jet velocity, S50.14. Apparatus


without porous plate. j, peaks of maximum intensity; 1, peaks of medium
intensity; m, peaks of minimum intensity.

scription of the spectrum, we consider the normalized quality


factor Q N :
Q N5

Q
,
450f 0 / f s

where Q is the measured quality factor. Q N varies then from


0 to 1.3. Thus, if in a given instance, the measurements yield
Q N '1, it means that the corresponding oscillation is purely
sinusoidal. This is precisely the case for the main peak of the
spectrum of Fig. 6 measured in the center of the mixing layer
~Q5101, Q N 51.26!.
III. PRELIMINARY MEASUREMENTS
A. Two-dimensionality checks and cavity resonance
effects

The preliminary measurements had a two-fold purpose:


to perform experimental checks of the facility and to gather a
database for the determination of flow parameters such as u
and dv ~the vorticity thickness! versus Re.
The two-dimensionality checks showed that z variations
of the mean velocity were less than 0.4% over 80% of the
total central span. The two-dimensionality of the jet is also
quite satisfactory.
The porous plate upstream of the contraction ~Fig. 1!
was added later after having observed an abnormal behavior
of the jet. Indeed, as illustrated by Fig. 3, without the porous
plate the frequencies of the instability were not proportional
to the jet velocity, as they should be, but varied stepwise. The
instability actually locked onto the cavity resonance frequency f 1 , determined independently by exciting the cavity
with a loudspeaker, or its higher harmonics. This locking was
facilitated by the fact that the cavity resonance frequency fell
into the range of the natural frequencies of the jets produced
by the facility. The purpose of the porous plate inserted just
upstream of the contraction was to reduce the volume of the
cavity and, thus, to push the resonance frequency to suffiPhys. Fluids, Vol. 8, No. 4, April 1996

FIG. 4. Frequency of oscillations versus jet velocity, S50.14. Lines are


determined by the least square method. ), H52.07 mm; 1, H53.04 mm;
s, H54.28 mm.

ciently high values out of this range. This goal was achieved
since with the porous plate the frequency varies linearly with
the jet velocity ~Fig. 4!.
B. Probe position

The convective heat transfer from the hot wire is sensitive to both fluid velocity and density. In order to avoid the
complication of separating velocity and density variations in
the anemometer output, which would have required information from two sensors, it was desirable to place the hot wire
in the potential core of the jet where the density is constant,
so as to have a purely velocity-dependent anemometer output
signals. It was, therefore, necessary to check how well the
instabilities that occur in the shear layers may be captured by
a hot wire located in the potential core, where only the potential fluctuations of the velocity induced by the vorticity
fluctuations in the shear layers are present.
Measurements with the hot wire located in the potential
core ~x/H51, y/H50! and with a microphone located in the
farfield yield spectra ~Fig. 5! with peaks at the same frequencies and with amplitudes with the same relative order. The
small difference between the frequencies of the dominant
peaks, 923 and 918 Hz, respectively, may be accounted for
by small changes in the flow rate of the jet. Although the
flow rate through the sonic throats of the mixing apparatus is
blocked, it remains proportional to the upstream generating
pressure. The level of this pressure ~read on a pressure transducer! is controlled by the setting of the regulating valve
whose reproducibility is, of course, not perfect. However,
while the ratio of the f 0 to f 0/2 peak ~f 0 being the frequency
of the dominant mode; here f 05920 Hz! is nearly 100 in the
local hot-wire spectrum, it is only about 3 in the microphone
spectrum, which corresponds to an integration over the
whole flow field. This is coherent with the variations of Fig.
7, which shows that the f 0/2 mode takes over with increasing
downstream distance. Since the f 0 peak in the microphone
Raynal et al.

995

Downloaded30Sep2010to202.3.77.11.RedistributionsubjecttoAIPlicenseorcopyright;seehttp://pof.aip.org/about/rights_and_permissions

FIG. 5. Spectra obtained ~a! with a microphone, ~b! with the hot wire at
x/H51, y/H50 in the same conditions Re5500, S50.14, and H53.04
mm.

spectrum is nevertheless the highest, we consider it as the


dominant mode. The frequency f 0 of this mode is designated
2 f 0 by YM.9 When comparing the dominant Strouhal numbers from the two investigations, the values of YM9 ~Fig. 11,
p. 333 of their paper! have consequently been multiplied by
2.
Since the microphone captures the dominant global jet
instability mode, so does, therefore, also the hot wire in the
potential core. The difference in level between the two spectra is due to the amplification of the microphone signal
~3100!. It may be seen that the dominant peak obtained with
the microphone is sharper than the one from the hot wire.
Both quality factors are high, respectively, 72 and 54, the
normalized corresponding values are 0.87 and 0.65. As discussed above, this difference is not really significant. The
broadband noise in the microphone spectrum is, however,
also much higher than in the hot-wire spectrum, especially in
the low-frequency range from 200 to 600 Hz. In this range,
the difference nearly reaches two orders of magnitude. It is
due to the acoustical background noise of the room. Since
this factor is difficult to control in the present experimental
conditions, the hot wire was preferred to the microphone for
the subsequent measurements.
Further checks showed that the microphone spectra are
insensitive to the hot-wire probe position. This latter conclusion was confirmed by lateral and longitudinal explorations.
Indeed, Figs. 6 and 7 show that the frequency of the dominant peak is not affected by the probe position. It is the same
within 60.5% uncertainty, which may be accounted for by
the small variations in the generating pressure of the mixing
apparatus, as discussed above, and the discrete description of
5000
'5.0 Hz!. These
the spectra ~interval between points5 1024
figures also illustrate the changes in the peak value with
probe position. Thus when the probe is in the center of the
mixing layer ~y/H5 21, Fig. 6!, where the perturbations occur,
the peak is about two orders of magnitude higher than when
the probe is outside the layer ~either at y/H50 or 1!, where
it only sees the induced potential fluctuations. This was also
observed by YM,9 whose results indicate that the maximum
temperature amplitude is located at y/H5 21 for 0<x/H<2.
996

Phys. Fluids, Vol. 8, No. 4, April 1996

FIG. 6. Velocity spectra for various positions of the hot wire across the jet
at x/H51. Here Re5500, S50.14, and H53.04 mm. Here Q is the quality
factor defined in Sec. II by Fig. 2.

But, since the broadband turbulence increases in nearly the


same proportion, the quality factors are not dramatically different. It may be concluded that the presence of the hot wire
placed at x/H51, y/H50 has little effect on the flow stability when S50.14, as well as in the case of higher-density
ratios, as discussed in Sec. IV B. When the probe is moved
downstream, the evolution of spectra, as shown by Fig. 7, is
mainly characterized by the rise of the subharmonic and by
the increase of the spectral level, which reflects the spatial
amplification of the perturbations. The subharmonic growth
is probably associated to the pairing process.
C. Exit flow profiles

Since H/u is a major parameter of the stability analysis,


velocity profiles in the exit section were measured for various Reynolds numbers. Simultaneously, the rms-fluctuation
intensities were also determined. Figure 8 shows the normalized mean velocity profiles measured in air obtained at x/H
'0.3 with the H52.04 mm slot width for five values of the
Reynolds number. The corresponding rms-intensity profiles
are shown in Fig. 9. The fluctuation intensity in the core of
Raynal et al.

Downloaded30Sep2010to202.3.77.11.RedistributionsubjecttoAIPlicenseorcopyright;seehttp://pof.aip.org/about/rights_and_permissions

FIG. 9. The rms intensity profiles for selected values of Reynolds number.
Here S51, H52.07 mm; the same symbols as in Fig. 8.

boundary layer of an axisymmetric jet. These fluctuations


may be due to time variations of the boundary layer thickness, as suggested by these authors.
The momentum thickness u and the vorticity thickness
dv were determined from mean velocity profiles and calculated by the following relations:

u5
FIG. 7. Velocity spectra for various x positions of the hot wire, Re5500,
S50.14, and H53.04 mm.

the jet is 0.3%, which is satisfactory for such a small-scale


facility. It must be noted that the peaks of u 8 /U max increase
with Re and reach a maximum value of 4%, comparable to
the results of Hussain and Zedan13 in the initial laminar

d v5

y1

y 0.1

U
U
1
dy,
U max
U max

~1!

U max
,
~ ] U/ ] y ! max

~2!

where U max is the average of five velocity values around the


midplane, y 0.1 and y 1 are the y locations, where U/U max is
0.1 and 1, respectively. The momentum thickness is inversely proportional to the square root of the Reynolds number, as expected for laminar flow:
A~ H !
u
5
.
H/2 ARe

~3!

The momentum and vorticity thickness data give


A ~ H52.07 mm! 52.1,
A ~ H53.04 mm! 51.8,

~4!

A ~ H54.28 mm! 51.4,

dv
~ H52.07 mm! 54.0,
u
dv
~ H53.04 mm! 54.6,
u
dv
~ H54.28 mm! 55.0.
u

FIG. 8. Mean velocity profiles for selected values of Reynolds number. Here
S51; H52.07 mm; ), Re5500; 1, Re5750; s, Re51000; L, Re52000,
n, Re53000.
Phys. Fluids, Vol. 8, No. 4, April 1996

~5!

The values of A have been obtained by forcing the least


square straight line to pass through the origin. Since the exit
velocity profiles depend only on the Reynolds number for a
given slot width, it is assumed that these relations also apply
to the jets of helium/air mixture emerging into the ambient
air. A few checks proved this to be true. Indeed, the values of
u measured in helium jets were close to those predicted with
the above values of A(H). We note that the product H * A(H)
Raynal et al.

997

Downloaded30Sep2010to202.3.77.11.RedistributionsubjecttoAIPlicenseorcopyright;seehttp://pof.aip.org/about/rights_and_permissions

FIG. 10. Strouhal number StH vs Re. Here S50.14. ), H52.07 mm; 1,
H53.04 mm; s, H54.28 mm.

equals 4.35, 5.47, and 5.99 for H52.07, 3.04, and 4.28 mm,
respectively. u is, thus, an increasing function of H for a
given Re. This is in keeping with the decrease of the nozzle
contraction ratio with H increasing and consequently with
the decrease of the favorable pressure gradient in the contraction. In the various plots involving u, the measured values of u are used and not those computed from the above
formula.
IV. EXPERIMENTAL RESULTS
A. Frequency of the dominant mode

1. The pure helium jet, S 50.14

As the flow is characterized by two length scales, the


slot width H and the initial momentum thickness u, it is
possible to define two Strouhal numbers StH ~5f 0 H/U j ! and
Stu ~5f 0 u /U j !. Moreover, effects of viscosity may be taken
into account by the global Reynolds number, Re, or by the
more local parameter H/u. The oscillating mode with a sharp
peak in the spectrum only exists when Re>300, 250, and
250 for H52.07, 3.04, and 4.28 mm, respectively. The minimum corresponding values of H/u are, respectively, 15.4,
17.6, and 22.6. The fairly small variations in the transitional
Re is a first indication that Re may be a better scaling parameter than H/u.
Is is observed ~Fig. 10! that the StH values match well
for the different slot widths when plotted vs Re. StH tends to
0.2 for the highest Re values reached in the present study.
The near constancy of StH at large values of Re is a result of
the linear variations of f 0 with U j : f 0 5aU j 1b ~Fig. 4!. For
large values of U j , f 0 'aU j , and consequently StH 5cte.
We remark in passing that there is no simple explanation to
the facts that b0 and to the variations of b with the slot
width. When StH is plotted versus H/u ~not shown here! the
collapse is much less satisfactory, which is contrary to the
observations of KS8 in the round jet who use the same definition for Re.
998

Phys. Fluids, Vol. 8, No. 4, April 1996

FIG. 11. Strouhal number Stu vs ~a! Reu , ~b! H/u. Here S50.14. We see ),
H52.07 mm; 1, H53.04 mm; s, H54.28 mm.

Figure 11~a! clearly shows that the Strouhal number Stu


does not scale with Reu . The collapse of Stu on a single
curve for the three nozzles is improved when plotted against
H/u @Fig. 11~b!#. Since H/ u 5(2/A) 2 Reu , the plots of Figs.
11~a! and 11~b! display the same Re dependence of Stu but
not the same nozzle dependence reflected in the coefficient
A(H).
Figures 10 and 11 clearly show that best collapses are
obtained when plotting a nondimensional frequency against a
flow control parameter based on the same length scale ~StH
vs Re or Stu vs H/u!. But considering the Re dependence of
the onset of the oscillating mode, the quality of the collapse
on a single curve of StH vs Re and, above all, the trend
toward a constant value of StH , it is shown that the best
scaling for the dominant frequency of the oscillating instability is provided by the jet width H and not by the initial
shear layer thickness. It may, thus, be concluded as YM9 that
this instability clearly corresponds to an oscillatory behavior
of the whole jet, and not to a shear layer instability. These
authors also point out that the most amplified shear layer
mode has a frequency f 0 d v /U j 50.07, which leads to
Raynal et al.

Downloaded30Sep2010to202.3.77.11.RedistributionsubjecttoAIPlicenseorcopyright;seehttp://pof.aip.org/about/rights_and_permissions

FIG. 12. Influence of density ratio S on the Strouhal number, StH . Here
H53.04 mm; the same symbols as in Fig. 8; m, result of the computation of
YM.10

21/2
f 0 }U 3/2
. As shown by Fig. 4, it is found
j , since dv} U j
here that f 0 }U j and not f 0 a U 3/2
j . Furthermore, since
dv /u'4.5 ~average value for the three nozzles!,
f 0 d v /U j 50.07 yields Stu5( f 0 d v /U j ) ~u/dv!50.016, which
is a factor 2 larger than the maximum value of the measured
Stu . These facts reinforce the conclusion that the oscillating
mode of the inhomogeneous plane jet is not a shear layer
mode.

2. Influence of S on the frequency of the dominant


mode

Having established that, for a constant value of S, StH is


better scaled by Re than by H/u, its simultaneous dependence on Re and S was investigated. The results ~StH vs S
for various values of Re, H53.04 mm, Fig. 12! show that
StH first increases slightly with S before reaching a plateau,
which is itself an increasing function of Re from 0.14 to
0.24. One notes that the curves obtained for the two highest
values of Re are close, suggesting that StH reaches a limiting
value of about 0.25. This value is comparable to, but clearly
distinct from, the 0.32 value measured by YM9 for S50.76
and for a Reynolds number, based on the ambient viscosity
of about 4000. A tentative explanation for this differences
and others between the present results with those of YM9 is
proposed below.
The oscillating mode disappears for a critical value S c of
the density ratio that increases from 0.25 to 0.7 when Re
varies from 500 to 3000. This transition is the topic of the
next section.
B. The critical density ratio S c and the domain of
instability

A typical evolution of the spectra with increasing density


ratio at a fixed value of the Reynolds number, Re51000, is
shown in Fig. 13. When S increases, the amplitude of the f 0
peak decreases by several orders of magnitude and the peak
broadens. The resulting decline of the quality factor by two
Phys. Fluids, Vol. 8, No. 4, April 1996

FIG. 13. Evolution of velocity spectra with density ratio, Re51000,


H52.07 mm.

orders of magnitude is much larger than the 625% experimental uncertainty and is, therefore, truly significant. Once
the normalized quality factor Q N ,0.2, the peak becomes indistinguishable from the random fluctuations. The value of S
for which Q N '0.2 was, therefore, considered as the critical
value corresponding to the limit of existence of the oscillating mode. A similar method was used by Sreenivasan et al.1
We note in Fig. 13 that f 0 decreases from 2400 to 1600
Hz when S increases from 0.31 to 0.51. This is contrary to
the observations of YM,9 who found that, in their experimental conditions, f 0 was nearly independent of S. This constancy can be explained by the narrow range of S used in
their experiments ~S varies from 0.73 to 0.92!. Monkewitz
et al.2 observed in a heated round jet an increase of the dominant frequency when S decreases from 0.78 to 0.47 for a
constant Reynolds number. This is consistent with the fact
that, for high enough Re values, St is constant when S increases. One can actually check that St5const leads to
S * f 05const when Re is kept constant. The most important
observation is, however, the critical value of S, S c 50.51 for
Re51000. This is much lower than the value S c 50.92 obRaynal et al.

999

Downloaded30Sep2010to202.3.77.11.RedistributionsubjecttoAIPlicenseorcopyright;seehttp://pof.aip.org/about/rights_and_permissions

FIG. 14. Hopf bifurcation diagram for peak amplitude versus density ratio,
for Re51000. Here ), H52.07 mm; 1, H53.04 mm; s, H54.28 mm.

served by YM,9 which is in agreement with the stability calculation of YM.10


Since the preceding method for determining S c depends
to a certain degree upon some subjective appreciation, the
2
A u 8 ( f 0 ) vs S plot was also applied, A u 8 being the area under
the peak of the power spectral density curve. Here A u 8 is
thus a true velocity amplitude. As pointed out by YM,9 this is
more satisfactory than using the value of the spectral peak
~A u 8 corresponds to the square root of what these authors
designate by D P!. YM9 integrated the spectra over a bandwidth 61.94 Hz centered on the frequency f 0 for f 0'50 Hz.
In order to facilitate comparisons, we chose a bandwidth
proportional to f 0 : 1.943f 0/50. Figure 14 shows the variations of the f 0 peak amplitude versus the density ratio S for
Re51000, and confirms the previous value S c 50.51 with an
experimental uncertainty about 60.02. Let us note at once in
Fig. 14 that A u 8 u S2S c u n with n.1 near the critical density
ratio, in contrast with variations as u S2S c u 1/2, required if A u 8
were the saturation amplitude of an instability modeled by a
Landau amplitude equation with zero noise ~a f 50!:
dA
5 s u A u 2C n u A u 3 1 a f .
dt

~6!

The small amount of data in Fig. 14 in the range ~0.4 0.5!


does not allow us to conclude on the agreement of our experimental data with this prediction. YM9 interpret their oscillations as being described by a supercritical Hopf bifurcation. Considering the experimental scatter of their data in the
neighborhood of S c ~see Figs. 12 and 13 of their paper! and
the departures of the data points from the fitted u S2S c u 1/2
curve near S c , this interpretation is open to question. Nevertheless, their experimental data clearly indicate a critical
value between 0.9 and 0.95.
Two further comments on the present results are of interest. First, it is observed that the measured variations of
A u 8 ( f 0 )/U j stretch over a range of nearly two orders of magnitude. These variations are, at best, of one order of magnitude in the heated jet experiment of YM.9 Second, the points
for the three nozzles collapse remarkablyand
1000

Phys. Fluids, Vol. 8, No. 4, April 1996

FIG. 15. Domain of the oscillating mode in the S-Re plane. Here ),
H52.07 mm; 1, H53.04 mm; s, H54.28 mm; 3, Yu and Monkewitz.9

surprisinglywell on a single curve. This is a further and


rather strong proof that the oscillating mode does scale with
the jet Reynolds number and that it corresponds to a very
stable limit cycle.
The question that arises then is whether the critical density ratio depends upon the Reynolds number. The preceding
discussion having shown that the determination of S c based
on the quality factor yields the same result as the peak amplitude decrease, the first method that is somewhat faster was
used to investigate the Re dependence of S c . The results are
summarized in Fig. 15 for the full experimental range, which
could be covered in the present experiments, i.e., S in the
range 0.14 1 and Re in the range 2503000. Larger values
of Re could not be reached, because, with the flow rates
required, the generating pressures could not have been maintained constant for sufficiently long times to make the measurements. It is nevertheless noteworthy to remark that, for
H53.04 mm and S50.14, Re53000 requires already a velocity U j 5120 m/s, so that appreciably larger Reynolds
numbers would have introduced complications due to compressibility effects. Furthermore, Re53000 based upon the
viscosity n j of the jet fluid corresponds to a Reynolds number based on the viscosity n` of the ambient fluid of more
than 20 000 for S50.14 with the pure helium jet.
As reported in Fig. 15, the highest quality factors are
obtained for small values of S. For S'0.14 there are two
small areas in the S-Re plane, where the oscillations display
normalized quality factors larger than 0.9 and correspond to
nearly pure tones. The spectral peaks broaden with increasing Re, a trend that was also observed by KS8 in the round
helium jet and by Monkewitz et al.2 in the heated round jet.
Russ and Strykowski14 showed that peaks in the power spectra were suppressed in a heated round jet with turbulent exit
conditions. The disappearance of the peaks with increasing
Re could thus be explained by the rise of the intensity of the
small-scale turbulence.
The most important results of Fig. 15 are the monotonous increase of S c from 0.14 to 0.7 with increasing Re and
the continuous decrease of the slope of the S c vs Re curve,
which seems to indicate an upper limit of S c only slightly
Raynal et al.

Downloaded30Sep2010to202.3.77.11.RedistributionsubjecttoAIPlicenseorcopyright;seehttp://pof.aip.org/about/rights_and_permissions

FIG. 16. Influence of the probe presence. Spectra obtained with a microphone without the presence of the hot-wire probe ~above! and with it
~moved below!. Here Re5750, H53.04 mm.

higher than 0.7. This is a major difference with the observations of YM,9 whose measurements unquestionably lead to
S c 50.92 for a Reynolds number based on the ambient viscosity, Re` , of about 4000, a value, moreover, in close agreement with their previous stability calculations ~YM10!, which
show that S c 50.95 is the upper limit for the existence of
absolute instability. Let us note that this discrepancy in S c
cannot be accounted for by the differences in Re. Indeed, the
dynamic viscosity is nearly independent of temperature, so
that Re`5U j H/ n ` 5 n j / n ` U j H/ n j ' r ` / r j Re5Re/S. The
critical condition Re530002S50.7 corresponds to a Reynolds number based on n` of about 4300, very close to the
value of 4400 used in the YM9 experiments ~total head of jet
of 1.26 mm H2O!. Figure 15 clearly shows that the point
~Re544002S50.92! is largely above the present S c curve.
This difference cannot be accounted for by the influence of
the probe position on the jet stability. Indeed, for conditions
near the critical value of S, there is no observable change in
the microphone measurements when the hot wire is present
in the jet at x/H51, y/H50 or absent ~see Fig. 16!. Moreover, peak-amplitude diagrams, such as Fig. 14, show quite a
negligible influence of the hot-wire probe position ~on the jet
axis, inside the mixing layer, and outside of the jet in the
entrainment flow!, and give the same critical value of S in
the three cases. No damping effect, such as that observed by
Sreenivasan et al.1 in a round air/helium jet, could be detected. This could be explained by the fact that spatial coherence is stronger in a plane than in a round jet.
Phys. Fluids, Vol. 8, No. 4, April 1996

There are, then, several differences between the present


experimental results and those of YM,9 namely in the Strouhal number, the variation of the Strouhal number with the
density ratio, and, most importantly, in the upper limit of the
critical density ratio. The following reasons may tentatively
be put forward to explain these differences. Let us first stress
the differences in the domain in the Re-S plane covered by
the two investigations, bounded by ~3000, 5000! and ~0.73,
1! in the YM9 investigation, by ~250, 3000! and ~0.14, 1! in
the present one. The limited range of the Re and S variations
in the YM9 experiment could explain why these authors did
not observe a noticeable change in StH with S. But this cannot account for the discrepancies in StH and especially in S c .
There are two differences in the two experimental realizations that may be of physical significance. They concern the
exit condition in the x50 plane and the generation of the
density inhomogeneity. Figure 1 of the YM9 paper seems to
indicate that there is no baffle in their setup, contrary to the
present one. It is well known ~Kotsovinos15! that the entrainment is not the same with or without a plate in the x50
plane, especially in the neighborhood of the nozzle, where
the flow conditions are critical for the stability. On the other
hand the way in which the density difference is generated
affects the r profile with respect to the U profile in the exit
section. Indeed, in the heated jet there is a thermal boundary
layer of finite thickness at x50, whatever the efficiency of
the thermal insulation of the settling chamber and of the
nozzle may be. In the jet of a helium mixture, on the contrary, the density profile is truly in the form of an inverted
top hat. Since the stability calculations of YM10 were performed with identical profiles for U and r, we were interested in exploring the effects of differences in shape and
position ~e.g., the position of the inflection point! of the velocity and density profiles on the stability.

V. RESULTS FROM STABILITY COMPUTATIONS


A. Comparison of experimental results with
the stability computations of Yu and Monkewitz10

Yu and Monkewitz10 determined the domain of absolute


instability of the plane inhomogeneous jet in the parameter
space (S,N), S being, as above, the density ratio, and N 21
being a measure of the maximum velocity and density gradients. They solved the eigenvalue problem by assuming parallel inviscid flow without buoyancy and considering small
perturbations of the usual normal mode form
q 8 5q exp@i(kx2 v t)#. Under these restrictions, the equations of motion reduce to the simple second-order differential
equation for the disturbance pressure amplitude p :

2 du d p
d 2 p
1 d r
1
2k 2 p 50,
22
dy
r dy u 2c dy dy

~7!

with the following boundary conditions:


Raynal et al.

1001

Downloaded30Sep2010to202.3.77.11.RedistributionsubjecttoAIPlicenseorcopyright;seehttp://pof.aip.org/about/rights_and_permissions

FG
FG
p
dp
dy
p
dp
dy

F G

FG

1 2ky
e ,
2k

y`

~8!

1
0

~ varicose mode! .

y50

The complex frequency v0~5v 0r 1i v 0i ! and wave number k 0


subjected to the condition of zero group velocity d v /dk50
are computed. The flow is absolutely unstable if the temporal
growth rate v0i is positive ~v0r is the frequency v0r 52pStH !.
The authors further assume that the nondimensional velocity
and density profiles u(y) and r(y) are described by the following functions:
u ~ y ! 52F ~ y ! ,

r ~ y ! 5 @ 11 ~ S 21 21 ! F ~ y !# 21 ,

~9!

where F(y)511sinh2N @y arcsinh~1!#21; u(y) and r(y) are


related to the dimensional quantities ~marked by *! by
y5

y*
y*
1/2

u~ y !5

u *~ y *
1/2 ! 5

2u * ~ y !
,
u *~ 0 !

u *~ 0 !
,
2

r~ y !5

r *~ y !
.
r *~ ` !

~10!

The boundary conditions,


u ~ 0 ! 52,

r ~ 0 ! 5S

u ~ ` !@ 0 # ,

r ~ ` ! 51,

~ jet centerline! ,

~11!

are thus satisfied by the above functional forms. The stability


diagrams computed by Yu and Monkewitz consist of v0r and
v0i curves versus N 21 for various values of S in the range
0.71. Direct comparison of these theoretical predictions
with present experimental observation are not always possible because the parameter spaces only overlap slightly.
Before discussing these comparisons, it seems appropriate to point out the main differences between the real and
computed flows. First, in the oscillatory regime the perturbations are not really small. The velocity oscillations normalized by the local mean values reach intensities of about 8%
at the nozzle exit and are amplified farther downstream. Second, the flow is not parallel. The jet spread is actually even
faster than in the homogeneous case ~see, e.g., Refs. 1, 2, 8,
and 16!. Moreover, we have only determined the Reynolds
dependence of u/H in the exit section since measurement at
different x-stations would have required the separation of the
velocity and density effects on the hot wire. Third, the shape
and relative positions of the velocity and density profiles
could not be independently controlled in the experiment.
These profiles can clearly not be described by a single exponent N and by the functional forms assumed in the theory.
This point will be discussed in detail below. Finally, as
pointed out by YM,9 a regime with large-amplitude oscillations is merely compatible with the existence of absolute
instability, but does not really prove it, since this would,
strictly speaking, require the experimental realization of an
ideal jet devoid of the slightest perturbation, and to which a
1002

Phys. Fluids, Vol. 8, No. 4, April 1996

FIG. 17. Comparison between experimental and velocity profiles as modeled by YM,10 ~a! ), U j53.7 m/s, Re5500;, profile calculated with Eq.
~9!, N54; ~b! ), U j522.2 m/s, Re53000;, profile calculated with Eq.
~9!, N510.

local transient perturbation could then be applied. It is, of


course, not possible to perform such an experiment, except
perhaps by numerical simulation.
In order to obtain the frequencies predicted by the theory
from the v0r vs N 21 diagram, it is necessary to determine the
values of the shape parameter N from the experimental velocity profiles. These may easily be computed by combining
Eqs. ~3!, ~4!, and ~5! with the formula

dv
H/2

'

&
,
N arcsinh~ 1 !

~12!

for N>2,

obtained from the function u(y) above. This yields


N5B ARe,

~13!

with the numerical values of the constant B50.19, 0.2, and


0.23, respectively, for H52.07, 3.04, and 4.28 mm. Figure
17 illustrates the fit of the modeled and measured velocity
profiles for two cases.
There is a good agreement between the predicted and the
measured values of the frequency. For instance, for S50.7
and N510 ~2500<Re<3000!, the predicted Strouhal number is 0.23 while our experimental value is 0.24 ~Fig. 12!.
Raynal et al.

Downloaded30Sep2010to202.3.77.11.RedistributionsubjecttoAIPlicenseorcopyright;seehttp://pof.aip.org/about/rights_and_permissions

The theory also predicts a decrease of StH with decreasing S,


in qualitative agreement with the present experiment.
Since the temporal growth rate v0i cannot be measured,
only the limits of the domain of absolute instability may be
compared. The trend of v0i vs N 21 curves for increasing
values of S ~Fig. 3 of YM10! indicates that for S50.6, according to the theory, the jet should be absolutely unstable
when N 21,0.4, i.e. Re.150 from Eq. ~13! for the present
experiment ~for H53.04 mm!. This theoretical critical value
of Re is far from the observed value of about 2000 ~Fig. 15!.
Another difference concerns the upper limit of Re values for
which instabilities seem to disappear. Although the assumption of infinite Reynolds numbers is approached, we observe
that the oscillating mode disappears for large values of Re
~'3000!.
The major disagreement between our measurements and
the theory is the critical value S c , respectively, 0.7 and 0.94,
as underscored earlier. It is appropriate to mention here that
in the round jet of the heliumair mixture of KS8 there is
also an upper limit of Re and a critical density ratio S c that is
lower than that predicted by the theory for a round jet,7 the
values being, respectively, 0.61 and 0.72. But as with the
plane jet, the critical value of the density ratio observed in
the heated round jet ~S c 50.737! is also in good agreement
with the stability theory.
B. Influence of the differences in shape and position
of the velocity and density profiles on the
stability

FIG. 18. Influence of N r on the density profile. Here S50.3, N53, N r53,
10, 30, and 150.

amplification rate v0i , but has no significant effect on the


frequency v0r . In order to determine the influence of the density ratio and the shape parameter of the velocity profile on
the increase of v0i , computations were also performed for
N55 and 10, S50.3. They show that the relative increase of
v0i with N r is sensitive to N, being, respectively, 50%, 30%,
and 15% for N53, 5, and 10 while changing S from 0.3 to
0.7 has no perceptible effect on this increase. In physical
terms, this means that the steepening of the density profile,
as would result from the change of a heated to an air/helium

Stability computations were undertaken in order to try to


account for the disparity in the behaviors of the heated jets
and the airhelium mixture jets. This disparity could, a priori, only come from the differences in the density profile of
the initial mixing layer, which has a finite thickness in the
first instance but is infinitely thin in the second one. Therefore, the same problem was solved as by YM,10 except for
changes in the mean density profile by introducing two additional control parameters, a shape parameter N r independent from that of the velocity profile N and a shift in position
Dy around y51, so that instead of Eq. ~9!, we have

r ~ y2Dy,N r ,S ! 5 @ 11 ~ S 21 21 ! F r ~ y2Dy !# 21 ,
F r ~ y ! 5 $ 11sinh2N r @ y arcsinh~ 1 !# % 21 .

~14!

The branch point v0 is determined by the method described


6
by Monkewitz and Sohn7 from four eigenvalues k 6
1 ,k 2 corresponding to two frequencies as close as possible to v0,
where 6 refers to the two branches of the solution in the
neighborhood of the saddle point k 0~v0!. The discretized
equation and the associated boundary conditions were solved
by an implicit scheme and a shooting technique. The code
was tested by comparing the results with those of YM10 for
identical conditions.
1. Influence of the shape of the density profile

We first performed the computation by varying the density profile shape parameter, N r , as illustrated by Fig. 18 and
by keeping the other parameters constant: N53, S50.3, and
Dy50. Figures 19~a! and 19~b! show that the steepening of
the shape of the density profile produces an increase in the
Phys. Fluids, Vol. 8, No. 4, April 1996

FIG. 19. Influence of the density profile shape parameter ~a! on the frequency, ~b! on the amplification rate of the oscillations. Here N53, S50.3.
Raynal et al.

1003

Downloaded30Sep2010to202.3.77.11.RedistributionsubjecttoAIPlicenseorcopyright;seehttp://pof.aip.org/about/rights_and_permissions

FIG. 20. Sketch of the shift between the velocity and the density profiles.
Here N510 and N r5150.

jet, should increase v0i , i.e. make the jet more unstable. This
is clearly contrary to the experimental observations.
Figure 18 shows that, with the functional representation
of r, Eq. ~14!, there is a shift Dy i between the velocity and
the density profiles inflection points, which varies with N r
and S. In the variations of v0i with N r of Fig. 19~b!, as well
in those of v0i with N of YM,10 there are therefore combined
effects of changes in shape and in the position of the inflection point of the density profile. We may, however, note that,
for a given value of N r , a significant shift Dy i is only possible for low values of S. For instance, for N r as low as 3,
Dy i .0.1 requires S,0.48. In order to determine to which
extent the shift of the density profile affects the results of
Figs. 19~a! and 19~b!, some of the preceding computations
were repeated by forcing Dy i 50 and using the following
expression of r:

r ~ y ! 5 ~ 12S !@ 12F r ~ y !# 1S,

~15!

so that the position of the density profiles inflection point


becomes independent of N r and S. The new results are not
significantly different from the earlier ones, the increase of
v0i with N r being only somewhat smaller.
It may thus be concluded that the steeper slope of the
density profile in the air/helium jet as compared to that of the
heated jet cannot account for the more stable behavior of the
former with respect to the latter, nor can it account for the
discrepancies of the airhelium jet with the theory.
2. Influence of a lateral shift of the density profile with
respect to the velocity profile

The second step of the computation consisted in shifting


the location of the density profile discontinuity by using Eq.
~15! for y2Dy i and N r5150 while keeping the same velocity profile. This is illustrated by Fig. 20 for N510. Figures
21~a! and 21~b! show the frequency and the amplification
rate against Dy i for different values of S. Similar evolutions
were obtained for N55 and N53. Once again, no major
influence is observed on the frequencies values @Fig. 21~a!#.
A comparison with the present experimental value of v0r
seems possible for S50.7 that is near the critical value of S
for N510 ~Re52500!, where the fluctuations are supposed
1004

Phys. Fluids, Vol. 8, No. 4, April 1996

FIG. 21. Influence of the shift between the velocity and the density profiles
~a! on the frequency. Symbols: present calculations, straight line: present
experimental result; ~b! on the amplification rate of the oscillations. Here
N510 and N r5150.

to be low enough for the linear theory to apply. The value of


v0r , deduced from Fig. 12 ~v0r 52p StH !, and shown by a
straight line in Fig. 21~a!, is found to be in close agreement
with the predicted value.
Figure 21~b! shows the significant role of the distance
Dy i between the density and the velocity profiles inflection
points. One can note that v0i reaches its maximum value
when the inflection points of the two profiles coincide. It
decreases almost linearly when shifting the density profile.
The main point of this figure is that for sufficiently large
interval Dy i the amplification rate becomes negative. It is
worthwhile emphasizing that the influence of Dy i on v0i is
much more pronounced than that of the density profile shape.
For the four density ratios studied here ~S50.14, 0.3,
0.5, and 0.7!, the change of sign occurs when the density
profile inflection point is located in the range @1.2, 1.3#
~Dy ic '0.25!. The corresponding local velocities are between
0.9% and 0.1% of the maximum velocity for N510. They
are much higher for N53 and N55 ~20% for N53!, whereas
Dy ic is still about 0.25 in these cases. These computations
indicate that the change of sign of the amplification rate is
not related to the precise shape of the velocity and of the
density profiles separately, but to the shift between their inflection points with a critical value Dy ic of 0.25.
These results can explain the differences between the
Raynal et al.

Downloaded30Sep2010to202.3.77.11.RedistributionsubjecttoAIPlicenseorcopyright;seehttp://pof.aip.org/about/rights_and_permissions

FIG. 22. Evolution of the amplification rate along the axis for S50.7. Solid
line: the Yu and Monkewitz10 calculations, dotted line: supposed evolution
deduced from present results.

heated air jet and the air/helium jet. Figure 22 shows the
evolution of the amplification rate along the jet axis, corresponding to the YM10 calculation for a density ratio of 0.7
~solid line!. As N 21 is proportional to the vorticity thickness
@Eq. ~12!# and the latter to the streamwise distance ~see, e.g.,
Brown and Roshko17!, one can consider that v0i evolution
with N 21 is representative of v0i evolution along the axis. A
curve corresponding to an air/helium jet at a Reynolds number of about 3000 is tentatively drawn on the same figure by
using the preceding results in a qualitative way. In the exit
section ~present experimental conditions correspond to
N 2150.1, Dy i 50.2!, the preceding calculation @Fig. 21~b!#
gives a very low value of v0i ~51.331023!. Along the axis,
due to the mixing process, the density profile becomes less
and less steep and its inflection point gets closer and closer to
that of the velocity profile. There are therefore also less and
less differences between the heated jet density profile and the
air/helium jet density profile evolutions and, consequently,
for sufficiently high N 21 values, there should be no more
significant differences between the corresponding v0i evolutions. In the development of the flow, the initial Dy i between
the inflection points decreases downstream, and according to
Fig. 21~b!, v0i tends to increase. This effect is balanced by
the decrease of the profile velocity shape factor N, which
induces a decrease of v0i . It is likely that the first effect
should be dominant near the nozzle, giving an increase in the
amplification rate in the initial region of the jet. The curve
~dotted line in Fig. 22! has been drawn with this hypothesis.
The main point is, of course, the important reduction of the
area for which v0i is positive. Thus, it is suspected that such
a pocket of local absolute instability would not be of sufficient size to influence the whole jet, which is consequently
globally stable ~see Huerre and Monkewitz18!. In order to
compute the exact v 0i (x/H) curve, one should make velocity
and density profiles measurements through the flow for different axial locations to obtain the curves N 21 (x/H) and
N r 21 (x/H); this has not been made, due to the experimental
Phys. Fluids, Vol. 8, No. 4, April 1996

difficulty of measuring separately the velocity and the density in the mixture.
The influence of the interval Dy i may explain why, for
low values of the Reynolds number, instabilities occur in a
small range of the density ratio. Actually Dy i increases with
decreasing exit velocity and then the flow becomes much
more stable. In this light, one can understand why
Sreenivasan et al.1 observed the most important differences
between their experimental round air/helium jet results and
the theoretical curve of Monkewitz and Sohn7 in the lower
part of their M-S diagram ~Fig. 4 of their paper!.
Raghu and Monkewitz19 observed that an annular cold
air flow around a round hot jet could suppress the oscillations of the hot jet. The presence of this coflow shifts the
inflection points of the velocity profile outward, and the suppression of the oscillations could be explained by the above
argument.
Figure 22 could also explain why oscillation frequency
values are lower in the case of the air/helium jet than that of
the heated air jet ~St50.25 and 0.32, respectively; see Sec.
IV A!. It is likely that the maximum amplification rate along
the curve v0i vs N 21 is actually moved to greater N 21 values,
so that the corresponding frequency value, v0r , is lowered. In
the heated jet case, the most amplified mode is constant if the
Reynolds number is sufficiently large, so that the N 21 value
in the exit section is lower than approximately 0.15, a value
that corresponds to the maximum of the curve in the solid
line of Fig. 22. In the case of the air/helium jet, the initial
conditions, in particular Dy i , depend on the Reynolds number. Consequently, the v0i ~N 21! curve depends on Re, even
for high values of this parameter. It is likely that the location
of the most amplified mode moves slightly upstream with
increasing Re. The corresponding frequency, v0r , should increase accordingly, and it explains why the observed Strouhal number increases with Re ~Fig. 10!.
Another point of interest can be deduced from the
present computations. The evolution of the maximum amplification rate in the conditions of Fig. 21~b! ~at Dy i 50! with
the density ratio is illustrated by Fig. 23. It clearly shows the
existence of a maximum obtained for S50.2. The amplification rate decrease for density ratio values smaller than 0.2
can be explained by the fact that density differences between
the jet and the ambient fluid become so important that perturbations cannot propagate into the too heavy ambient fluid.
The limit S0 actually corresponds to a flow between two
plates. Russ et al.20 made measurements for very low density
ratios down to S50.07 by heating a helium jet. It is interesting to note that the peak to base ratio of the main peak in
their power spectra is more than six orders of magnitude in
the case of the helium jet ~S50.14! and is only about 3 for
S50.07. Although those measurements are made in a round
jet, the decrease of the oscillations intensity is consistent
with the decrease of v0i for too small density ratio values, as
shown by the present computations.
In summary, linear stability calculations point out the
importance of the relative position of the velocity and the
density profiles. The main parameter is actually given by the
interval Dy i between their respective inflection points. The
maximum of the amplification rate is obtained when the two
Raynal et al.

1005

Downloaded30Sep2010to202.3.77.11.RedistributionsubjecttoAIPlicenseorcopyright;seehttp://pof.aip.org/about/rights_and_permissions

constitutes a major difference with the previous theoretical


result. Yet, this difference can be explained by linear stability
calculations, taking into account the density profile differences existing between the heated air jet and the air/helium
jet. The relative positions of the velocity and the density
profiles inflection points are actually of main importance.
The stability of the flow is increased by increasing the interval between the two inflection points. This corresponds to the
physical situation of the air/helium jet, which is consequently
less unstable than the heated air jet.
ACKNOWLEDGMENTS

FIG. 23. Evolution of the absolute amplification rate with the density for the
same position of the velocity and the density profiles inflexion points. Here
N510 and N r5150.

inflection points coincide. When increasing the parameter


Dy i , one observes an important decrease of v0i , even down
to negative values for Dy i .0.25. This result can therefore
explain why the air/helium jets are less unstable than heated
air jets, and why differences between air/helium experiments
and YM10 calculations differ largely for low Reynolds number values. Besides, as the parameter Dy i does not really
influence the frequency of the oscillations, the rather good
agreement between YM10 calculations and the present experimental results is confirmed by the present calculations.
VI. CONCLUSION

The experiments on inhomogeneous plane jets generated


with airhelium mixtures described in the paper cover a
range of density ratios from 0.14 to 1 and a range of Reynolds numbers based on the slot width H and the jet fluid
viscosity n j from 250 to 3000. Measurements were carried
out for three different values of H in order to investigate the
influence of this parameter on the flow.
The existence of self-excited strong sinusoidal oscillations, first reported by Yu and Monkewitz9 in 1993, which
have features of absolute instability modes, has been confirmed. The frequency and the amplitude of these oscillations
scale with the nozzle width and not with the thickness of the
shear layer, which proves that they are the result of an oscillatory behavior of the whole jet. The frequencies of these
oscillations are only in fair agreement with those observed
by Yu and Monkewitz9 and in a better one with their
computations.10
The new features revealed by the present experiments
concern the variations of the oscillation frequency with the
Reynolds number and the density ratio, and especially the
dependence of the critical density ratio S c with the Reynolds
number. The fact that in the present experiments S c '0.7

1006

Phys. Fluids, Vol. 8, No. 4, April 1996

The authors would like to thank Professor P. A. Monkewitz for very helpful discussions. This work was supported
by the HERMES R&D Program through the Centre National
dEtudes Spatiales and Dassault Aviation.
1

K. R. Sreenivasan, S. Raghu, and D. M. Kyle, Absolute instability in


variable density round jets, Exp. Fluids. 7, 309 ~1989!.
2
P. A. Monkewitz, W. B. Bechert, B. Barsikow, and B. Lehmann, Selfexcited oscillations and mixing in a heated round jet, J. Fluid Mech. 213,
611 ~1990!.
3
P. A. Monkewitz and E. Pfizenmaier, Mixing by side-jets in strongly
forced and self-excited round jets, Phys. Fluids A 3, 1356 ~1991!.
4
R. Riva, Ecoulements de fluides inhomoge`nes: Stabilite des jets, transferts turbulents dans les couches limites, Ph.D. thesis, INPG, Grenoble,
1991.
5
P. G. Drazin and W. H. Reid, Hydrodynamic Stability ~Cambridge University Press, Cambridge, England, 1981!.
6
P. Huerre and P. A. Monkewitz, Absolute and convective instabilities in
free shear layers, J. Fluid Mech. 159, 151 ~1985!.
7
P. A. Monkewitz and K. D. Sohn, Absolute instability in hot jets, AIAA
J. 26, 911 ~1988!.
8
D. M. Kyle and K. R. Sreenivasan, Instability and breakdown of a round
variable-density jet, J. Fluid Mech. 249, 619 ~1993!.
9
M-H. Yu and P. A. Monkewitz, Oscillations in the near field of a heated
two-dimensional jet, J. Fluid Mech. 255, 323 ~1993!.
10
M-H. Yu and P. A. Monkewitz, The effect of nonuniform density on the
absolute instability of two-dimensional inertial jets and wakes, Phys.
Fluids A 2, 1175 ~1990!.
11
P. A. Monkewitz, P. Huerre, and J. M. Chomaz, Global linear stability
analysis of weakly non-parallel shear flows, J. Fluid Mech. 213, 611
~1993!.
12
T. Morel, Design of two-dimensional wind tunnel contractions, J. Fluid
Eng. 371 ~1977!.
13
A. K. M. F. Hussain and M. F. Zedan, Effects of the initial condition on
the axisymmetric free shear layer: Effects of the initial momentum thickness, Phys. Fluids 21, 1100 ~1978!.
14
S. Russ and P. J. Strykowski, Turbulent structure and entrainment in
heated jets: The effect of initial conditions, Phys. Fluids A 5, 3216
~1993!.
15
N. E. Kotsovinos, A note on the conservation of the volume flux in free
turbulence, J. Fluid Mech. 86, 201 ~1978!.
16
P. A. Monkewitz, B. Lehmann, B. Barsikow, and D. W. Bechert, The
spreading of self-excited hot jets by side-jets, Phys. Fluids A 1, 446
~1989!.
17
G. L. Brown and A. Roshko, On density effects and large structure in
turbulent mixing layers, J. Fluid Mech. 64, 775 ~1974!.
18
P. Huerre and P. A. Monkewitz, Local and global instabilities in spatially
developing flows, Annu. Rev. Fluid Mech. 22, 473 ~1990!.
19
S. Raghu and P. A. Monkewitz, The bifurcation of a round jet to limitcycle oscillations, Phys. Fluids A 3, 501 ~1991!.
20
S. Russ, P. J. Strykowski, and E. Pfender, Mixing in plasma and low
density jet, Exp. Fluids 16, 297 ~1994!.

Raynal et al.

Downloaded30Sep2010to202.3.77.11.RedistributionsubjecttoAIPlicenseorcopyright;seehttp://pof.aip.org/about/rights_and_permissions

Das könnte Ihnen auch gefallen