Effect of density gradients in confined supersonic shear layers. II.

Threedimensional
modes
Oshin Peroomian

Citation: Physics of Fluids (1994-present) 8, 241 (1996); doi: 10.1063/1.868831

View online: http://dx.doi.org/10.1063/1.868831
View Table of Contents: http://scitation.aip.org/content/aip/journal/pof2/8/1?ver=pdfcov
Published by the AIP Publishing

Articles you may be interested in

Poloidal shear flow effect on toroidal ion temperature gradient mode: A theory and simulation
Phys. Plasmas 3, 3689 (1996); 10.1063/1.871939

The threedimensional development of the shearing instability of convection

Phys. Fluids 8, 1350 (1996); 10.1063/1.868943

Effect of density gradients in confined supersonic shear layers. I. Twodimensional disturbances

Phys. Fluids 8, 225 (1996); 10.1063/1.868830

A numerical study of threedimensional Jeffery orbits in shear flow

J. Rheol. 38, 1829 (1994); 10.1122/1.550604

Integral eigenmode analysis of shear flow effects on the ion temperature gradient mode
Phys. Fluids B 5, 4072 (1993); 10.1063/1.860575

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
132.174.255.116 On: Sat, 20 Dec 2014 18:33:51
 Effect of density gradients in confined supersonic shear layers. II.
Three-dimensional modes
Oshin Peroomian
Metacomp Technologies Inc., Westlake Village, California 91362
~Received 11 April 1995; accepted 3 October 1995!
The effect of density gradients on three-dimensional ~3-D! supersonic acoustic modes and 3-D
subsonic modes in a compressible confined shear layer were investigated using linear analysis. The
compressible Rayleigh equation with the solution of the steady compressible boundary layer
equations as its basic profiles was solved for two different density ratios, 1.398 and 3.0, and several
spanwise wave numbers. For a density ratio of 1.398, the 2-D C 01 mode had the highest growth rate
for a fixed aspect ratio B/H52. At the higher-density ratio, the 3-D mode had a slightly higher
maximum growth rate than the two-dimensional ~2-D! mode, and the maximum growth rate was at
a location where the phase speed was supersonic with respect to the fast stream. © 1996 American
Institute of Physics. @S1070-6631~96!03201-X#

I. INTRODUCTION
The effect of basic flow density gradients on the two-
dimensional ~2-D! supersonic acoustic modes for this flow
were investigated in part I of this analysis. A detailed over-
view of previous research in this area can be found in part I
and will not be presented here. However, for the sake of
completeness, some results for three-dimensional ~3-D! dis-
turbances that were not covered in part I will be given here.
Tam and Hu1 investigated the 3-D modes in a confined vor-
tex sheet and reported that the first 2-D class A mode ~A 01!
has the highest growth rate compared to all other 2-D and
3-D modes present in the vortex sheet for their particular set
of flow parameters. They also showed that this result held
true for finite thickness confined shear layers with vorticity
thickness less than dv'0.125. For unbounded free shear
layers, Sandham and Reynolds2,3 showed that the 3-D
subsonic Kelvin–Helmholtz ~KH! mode becomes the domi-
nant mode for a convective Mach number M c 5(U 1 2U 2 )/
(a 1 1a 2 ).0.6. For a range of 0.6,M c ,1.6, they also gave
an approximate expression, M c cos u'0.6, for the wave
angle @u5tan21~k z /k x !# of the wave with the maximum
growth rate. Based on this expression and the figures pre-
sented in their analysis, it is clear that as the convective
Mach number increases the wave angle corresponding to the
maximum growth rate shifts from 0° toward 90°. This might
suggest that at very high convective Mach numbers, outside
the range considered by Sandham and Reynolds,2,3 2-D
modes might actually dominate, since the range of wave
angles for which the ‘‘apparent’’ flow is subsonic is only a
small region near 90°. Jackson and Grosch4 investigated the
effect of cross-flow and obliqueness on the slow and fast
modes present in a M c .1 environment, and showed that for
certain combinations of cross-flow and wave angles the
growth rates could be increased by up to a factor of 2 with
respect to the 2-D case. Two major differences between the
confined and unconfined cases are the type of modes present
in the two cases and the restriction on the spanwise wave
number in the confined case. The modes present in a con-
fined supersonic 2-D shear layer are often of the acoustic
type ~wall modes! as opposed to the vorticity modes present
in the free shear layer, and these acoustic modes can have
higher growth rates than the vorticity modes. Also, due to the
sidewalls, the confined shear layer only has discrete span-
wise wave-number solutions for moderate aspect ratios. As
the aspect ratio of the duct ~B/H! is increased, the spectrum
of the discrete spanwise wave numbers become closer to the
continuous case.
The case studied here is a confined shear layer shown in
Fig. 1 of part I. All solution procedures and basic flow pro-
files are the same as in part I. The effect of density gradients
on the 3-D modes present for the density ratios considered
primarily in part I are investigated.
II. FLOW MODEL AND GOVERNING EQUATIONS
As in part I, the model that is studied here is a confined
compressible shear layer formed by two gases with different
velocities, densities, and properties, but with the same con-
stant pressure. Figure 1 in part I shows the configuration
used in this analysis. The subscript 1 is used for the quanti-
ties related to the high-speed free stream and the subscript 2
for the quantities of the low-speed free stream. The stream-
wise coordinate is x, the spanwise coordinate is z, and the
cross-stream coordinate is y. The aspect ratio of the channel,
unless otherwise stated, is taken to be 2 ~B * 52H * ). For
simplicity, free slip wall boundary conditions are assumed at
the walls of the channel. It is also assumed that the flow is
inviscid, nonconducting, and nondiffusive. For this situation
the governing equations are the Euler equations for a two
species system.
The equations are nondimensionalized using the fast
~upper! free-stream quantities, r1* , U 1* , T 1* , and the height of
the channel, H * . Thermodynamic properties are also nondi-
mensionalized by the upper free-stream thermodynamic
properties, c * * . Based on this nondimensionalization,
p a and R a
the density ratio r * 2 r*
/ 1 and the velocity ratio U * 2 /U *
1 are
defined as r2 and U 2 , respectively. Once the equations are
nondimensionalized, they are linearized around a parallel ba-
sic flow @~r̄(y), P̄,Ū(y),C̄ a (y)#. These basic flow quantities
can be found by solving the compressible boundary layer
equations using a similarity variable for Pr5Sc5Le>1. This
procedure was discussed in part I.

Phys. Fluids 8 (1), January 1996 1070-6631/96/8(1)/241/7/$6.00 © 1996 American Institute of Physics 241
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
132.174.255.116 On: Sat, 20 Dec 2014 18:33:51
 Normal modes are assumed for these infinitesimal dis- Therefore Eq. ~2! can be written as

S D
turbances with the form
2k DŪ 1
q 8 5q̂ ~ y ! exp@ i ~ kx1 b z2 v t !# , ~1! D 2 p̂1 2 D r̄ Dp̂1k̃ 2 ~ M̄ 2c 21 ! p̂50. ~10!
~ v 2kŪ ! r
where q̂ is the eigenfunction, k and b are the streamwise and
spanwise wave numbers, and v is the frequency. In general, This form of the O.D.E. for the disturbance pressure
k, b, and v are complex. Once these normal modes are sub- eigenfunction clearly shows the role of the parameter M̄ c and
stituted into the equations of motion, a single O.D.E. can be its significance in the type of solutions possible. Now, out-
found for the disturbance pressure eigenfunction: side the shear layer for a neutral wave this convective Mach

S D
number has a constant real value, i.e., M̄ c1
2kDŪ 1
D 2 p̂1 2 D r̄ D p̂ 5M c1 , M̄ c2 5M c2 , where
~ v 2kŪ ! r̄

1 S
r̄ g 1 M 21
ḡ
~ v 2kŪ ! 2 2k 2 2 b 2 p̂50, D ~2!
M c1 5M 1 S kŪ2 v
k̃
D 5M 1 cosc ~ 12C phx ! ~11a!

and
where

D5
d
dy
, ḡ 5
c p a C̄ a 1c P b ~ 12C̄ a !
c v a C̄ a 1c v b ~ 12C̄ a !
,
M c2 5M 1 S D a 1*
a*
2
cos c ~ U 2 2C phx ! , ~11b!

and M 1 and g1 are, respectively, the Mach number and the where C phx 5 v /k r , and c5tan21 ( b /k) is the wave angle.
ratio of specific heats of the high-speed free stream. As shown for the 2-D case, when M c2 .1 noninflectional
The above equation is solved subject to the boundary neutral solutions of the type C phx 51 are possible, and when
conditions M c1 .1 noninflectional neutral solutions of the type
C phx 5U 2 are possible.
Dp̂50, at y56 21 , ~3a! Based on the above definitions and the results obtained
in part I, the following cases are investigated:
Dp̂50, at z561, ~3b!
which come from the y- and z-momentum equations by set- case I: r 2 51.398, m51;
ting the normal velocity at the walls to zero. It is the latter case II: r 2 51.398 m53;
boundary condition that leads to a discrete spectrum of span- case III: r 2 53.000, m51;
wise wave numbers. Therefore the spanwise wave number
becomes case IV: r 2 53.000, m53.

b 5m p ,
for an aspect ratio of 2, and the disturbance pressure takes
the form
p 8 ~ x,y,z,t ! 5p̂ ~ y ! cos~ m p z ! exp@ i ~ kx2 v t !# .
Outside the shear layer the vortex sheet solutions hold.
They can be written as
~4! In all the cases, the fast stream gas is taken as He and the
slow stream gas is N2 . Also, the convective Mach number is
1.836, and the velocity ratio is 0.276. These correspond to
the parameters used in the 2-D analysis and provide a basis
~5! for comparison. The mode labels used in the 3-D analysis are
the same as those used in the 2-D analysis and are described
in part I.

p̂ ~ y ! 5A cos@ l 1 ~ 21 2y !# , for d ,y< 21 , ~6a!

p̂ ~ y ! 5B cos@ l 2 ~ 21 1y !# , for 2 d .y>2 21 , ~6b!

where
l 1 5 @ M 21 ~ v 2kU 1 ! 2 2k 2 2 ~ m p ! 2 # 1/2, ~7a!

F S D
l 2 5 M 21
a 1*
a*
2
2
~ v 2kU 2 ! 2 2k 2 2 ~ m p ! 2 G 1/2
. ~7b!

Following Mack,5 we define a 3-D wave number as

k̃5 @ k 2 1 ~ m p ! 2 # 1/2. ~8!
One can also define a complex convective Mach number,
M̄ c (y), as

M̄ c ~ y ! 5 A r̄ g 1 M 21
ḡ k̃ 2
~ v 2kŪ ! 2 . ~9!
FIG. 1. Contours of uM c (y)umax in the frequency–real streamwise domain
for b5p and the test case parameters. Dashed lines represent contours be-
low unity.

242 Phys. Fluids, Vol. 8, No. 1, January 1996 Oshin Peroomian

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
132.174.255.116 On: Sat, 20 Dec 2014 18:33:51
 FIG. 3. Here C phx versus frequency of modes present for b5p and test case
FIG. 2. Real streamwise wave number versus frequency of modes present parameters.
for b5p and test case parameters. The triangular fine-dashed line is
uM c (y)umax51.

bounded region of supersonic flow with respect to their

III. RESULTS AND DISCUSSION
A. Case I
As shown in part I, the basic flow profiles have three
generalized inflection points giving rise to three separate
types of modes. These modes are all supersonic acoustic
modes and the necessary condition for their existence is a
phase speeds. In order to use this condition, the maximum
convective Mach number in the channel, u M̄ c (y) u max , is plot-
ted in the frequency and real wave-number plane. This is
done by first fixing the spanwise wave number at a specific
value, and then taking discrete points in the real v and real k
plane to calculate a phase speed and thus a convective Mach
number throughout the channel ~note that this phase speed

FIG. 4. ~a! Growth rate versus frequency of class A modes for b5p and test case parameters. ~b! Growth rate versus frequency of class B modes for b5p
and test case parameters. ~c! Growth rate versus frequency of class C modes for b5p and test case parameters. ~d! Growth rate versus frequency of class D
modes ~D 11! for b5p and test case parameters.

Phys. Fluids, Vol. 8, No. 1, January 1996 Oshin Peroomian 243

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
132.174.255.116 On: Sat, 20 Dec 2014 18:33:51
 FIG. 5. Contours of uM c (y)umax in the frequency–real streamwise domain FIG. 7. Here C phx versus frequency of modes present for b5p and test case
for b53p and the test case parameters. Dashed lines represent contours parameters.
below unity.

originally subsonic, and as it turns supersonic with respect to

does not have to correspond to a mode, since we are only the fast stream it is very much like the combination of the
interested in what the corresponding convective Mach num- first three 2-D class B modes ~the upper portion of the phase
ber would be if there is a mode with that phase speed!. Then speed curves in Fig. 6 of part I!. This mode will be called the
the maximum value of the convective Mach number D 11 mode. The first class C mode C 11 also starts from within
throughout the channel is taken as u M̄ c (y) u max . the wedge described above, but its phase speed exits the
Figure 1 shows the contours of this maximum convec- wedge very quickly and only slightly changes the character-
tive Mach number. The dashed lines correspond to istics of the mode. The rest of the class A, B, and C modes
u M̄ c (y) u max,1. Any modes that are present within the
‘‘wedge’’ ( u M̄ c (y) u max51! are subsonic with respect to both
streams and therefore are not acoustic in nature. The two
small imperfections seen in the contours are due to the coarse
grid used and disappear as the grid is refined. Figure 2 shows
the real streamwise wave number versus frequency of all the
modes present when the spanwise wave number is fixed to
b5p. The short dashed line in the form of a wedge is the
u M̄ c (y) u max51 line. Two modes originate in that wedge and
are subsonic with respect to both streams until they exit the
wedge. Figure 3 shows the phase speed of the modes present
for this case. Several differences arise when we compare Fig.
3 to Fig. 6 of part I. First, we now have a mode present in the
limit as the frequency goes to zero, which is characteristic of
the KH mode present in the fully subsonic case. This mode is

FIG. 6. Real streamwise wave number versus frequency of modes present FIG. 8. ~a! Growth rate versus frequency of class A and B modes for b5p
for b53p and test case parameters. The triangular fine-dashed line is and test case parameters. ~b! Growth rate versus frequency of class D modes
uM c (y)umax51. for b5p and test case parameters.

244 Phys. Fluids, Vol. 8, No. 1, January 1996 Oshin Peroomian

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
132.174.255.116 On: Sat, 20 Dec 2014 18:33:51
 FIG. 9. Real streamwise wave number versus frequency of modes present
for b5p and r253.0 parameters. The triangular fine-dashed line is
uM c (y)umax51.

are all entirely acoustic modes and their phase speeds are
similar to those in the 2-D case. One must note that the first
two class A and C modes are different in the sense that each
has ‘‘picked up’’ the other’s phase speed curve. Figures 4~a!–
4~d! show the growth rates of the class A, B, C, and D
modes, respectively. The growth rate curve for the A 12 mode
has only one peak and its maximum growth rate is a few
percent higher than that of the 2-D mode. The growth rate
curve for the A 13 mode is almost identical to its 2-D coun-
terpart, despite the fact that the wave angle at the maximum FIG. 11. ~a! Growth rate versus frequency of class A modes for b5p and
growth rate for this mode is about 30°. Mode A 11 has very r53.0 parameters. ~b! Growth rate versus frequency of class D modes for
small growth rates ~as do modes B 11, B 12, and B 13! and is not b5p and r53.0 parameters.
shown. Since the class A and C modes have switched por-
tions of their phase speed curves with one another, one must
also discuss the growth rates of the class C modes in con- ber has one growth rate peak whose corresponding phase
junction to the class A modes. Instead of identifying the 2-D velocity is subsonic with respect to both streams and three
and 3-D counterparts of each growth rate peak, one can make peaks with corresponding phase speeds that are supersonic
a general statement that, except for A 12, the growth rates are with respect to the fast stream. The supersonic portion of the
lower in the region where the waves are 3-D than those phase speed curve for this mode is very much like the curve
found for the 2-D case. In fact, mode C 11 has a growth rate for the B 01 mode for a density ratio of about 1.65, where the
that is about 25% lower than its 2-D counterpart. The growth first three class B modes have undergone a resonant interac-
rate curves of the class B modes are very much like the 2-D tion ~see part I!. In general, if one plots all the growth rates
modes. The D 11 mode present for this spanwise wave num-

FIG. 10. Here C phx versus frequency of modes present for b5p and r253.0 FIG. 12. Growth rate versus frequency of the B 01 mode from part I and the
parameters. D 11 mode for r53.0 parameters.

Phys. Fluids, Vol. 8, No. 1, January 1996 Oshin Peroomian 245

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
132.174.255.116 On: Sat, 20 Dec 2014 18:33:51
 FIG. 13. Real streamwise wave number versus frequency of modes present
for b53p and r253.0 parameters. The triangular fine-dashed line is
uM c (y)umax51.
of the 2-D and 3-D modes ~b5p!, one will find that the C 01
has the highest growth rate.
B. Case II
In this case, the density ratio is the same as above, but
the spanwise wave number is now equal to 3p. Figure 5
shows the maximum convective Mach number contours. The
wedge corresponding to the u M̄ c (y) u max51 contour is much
larger than that in case I. This implies that the frequency and
streamwise wave number region over which purely subsonic
modes can exist is much greater. Figure 6 shows the real
wave numbers on the modes present when the spanwise
wave number is set to 3p. Now both modes that start within
the subsonic wedge have zero-frequency limits and both will
be labeled as class D modes. Figure 7 shows the phase speed
of the modes for this case. Some of the class A, B, and C
modes within this frequency range are not shown because
they have very small growth rates. Figure 8~a! shows the
growth rates for some of the class A and B modes, and like-
wise in Fig. 8~b! for the class D modes. Very much like case
I, it can be concluded that the 3-D modes have lower growth
rates than the 2-D modes. Therefore, for the test case param-
eters the 2-D C 01 mode has the highest growth rate. Tam and
FIG. 14. Here C phx versus frequency of modes present for b53p and
r253.0 parameters.
246 Phys. Fluids, Vol. 8, No. 1, January 1996
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
132.174.255.116 On: Sat, 20 Dec 2014 18:33:51
FIG. 15. Growth rate versus frequency of class A modes for b53p and
r53.0 parameters.
Hu1 also found similar results for their A 01 mode, which had
the highest growth rate for their test case parameters.
C. Case III
This case corresponds to a density ratio of 3.0 for which
the basic flow profiles only have one generalized inflection
point. Thus, as shown in part I, only class A and B modes are
present in the 2-D analysis. Figure 9 shows the real wave
number for the modes present. Only one mode exists in the
subsonic wedge and is present in the zero-frequency limit.
The phase speed versus frequency is given in Fig. 10, and
Figs. 11~a! and 11~b! show the growth rates of the class A
modes and the D 11 mode, respectively. The 3-D class A
modes for this case have smaller growth rates than their 2-D
counterparts. Very much like case I, the D 11 mode’s super-
sonic continuation is very similar to the B 01 mode present for
these parameters. Figure 12 shows the growth rates of both
modes ~D 11, B 01! on the same graph. It can be seen from this
figure that the maximum growth rate peak for the 3-D mode
is slightly higher than the 2-D mode. More about the growth
rates of 3-D class D modes will be said in the next section.
The class B modes are not shown due to the fact that
they possess very small growth rates.
FIG. 16. Growth rate versus frequency for the D 31 mode for b53p and
r53.0 parameters.
Oshin Peroomian

FIG. 17. Growth rate versus frequency for class D modes at different span-
wise wave numbers for r253.0 parameters.
D. Case IV
The parameters for this case are the same as in case III,
however, the spanwise wave number is increased to 3p. Fig-
ures 13 and 14 show the wave number and phase speed of
the modes present. Again, like case II the sonic wedge is
much greater than for the smaller wave number ~case III!.
Figures 15 and 16 show the growth rates of the class A and
the D 11 modes. The class A modes have much lower growth
rates than their 2-D counterparts. However, the D 11 mode has
a higher growth rate ~by about 5%! than the B 01 mode. Also,
the frequency for the maximum growth rate has shifted to the
left, causing the most unstable wave to have a lower fre-
quency than the 2-D mode. In order to investigate this effect
further, the mode D 11 for b55p is calculated, and its growth
rate along with the D 11 modes for b5p and 3p, and the B 01
~2-D! modes are plotted in Fig. 17. It is clear from this figure
that the rise in the growth rate due to the larger growth rates
in the subsonic regions of these modes eventually stops, and
actually the growth rates start to decrease beyond a spanwise
wave number of 5p.
As the aspect ratio is increased, more and more discrete
3-D modes are possible in the wave number range of
0<b<5p. In order to check the above results, an aspect ratio
of 10 was considered, which corresponds to b5m p /5. The
maximum growth rate for the D m1 modes was obtained at
m516, which corresponds to b'3p, which agrees well with
the results given in the previous paragraph.
Phys. Fluids, Vol. 8, No. 1, January 1996
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
132.174.255.116 On: Sat, 20 Dec 2014 18:33:51
IV. CONCLUSIONS
Based on all the results shown for all the cases consid-
ered, one can conclude that 3-D effects are less important for
these acoustic modes and their subsonic continuations, i.e.,
the class D modes, at this convective Mach number than for
the KH modes at high subsonic convective Mach numbers or
the vorticity modes ~one of which is a continuation of the
KH mode into the supersonic region, as shown by Jackson
and Grosch6! present in the unbounded supersonic shear
layer, in the sense that they do not give rise to much larger
growth rates, as is the case for the latter two modes. How-
ever, 3-D modes do exist that have almost the same phase
speeds and growth rates as the 2-D modes and that therefore
can give rise to nonlinear resonant interaction during the
evolution of the full 3-D shear layer.
Of course, as stated in the Introduction, as the convec-
tive Mach number increases the wave angles at which the
rise in 3-D growth rates are seen shift. Therefore, it is con-
ceivable that the convective Mach number for the confined
case considered here is high enough that the increase in
growth rate is not seen for the lower-density ratio, and only a
slight increase is seen for the higher density ratio. One final
note is the fact that unlike the 2-D vorticity modes in the free
shear layer, the acoustic modes are not continuations of the
2-D KH modes, and their 3-D characteristics might be dif-
ferent than those found in the free shear layer.
ACKNOWLEDGMENTS
I would like to thank my thesis advisor Professor Robert
E. Kelly for his support and guidance. This work was per-
formed as part of requirements for the Ph.D. degree at UCLA
and was supported by NASA via Grant No. NCC2-374.
1
C. K. W. Tam and F. Q. Hu, ‘‘The instability and acoustic wave modes of
supersonic mixing layers inside a rectangular channel,’’ J. Fluid Mech.
203, 51 ~1989!.
2
N. D. Sandham and W. C. Reynolds, ‘‘A numerical investigation of the
compressible mixing layer,’’ Report No. TF-45, Thermosciences Division,
3
Mechanical Engineering Department, Stanford University, 1989.
N. D. Sandham and W. C. Reynolds, ‘‘Compressible mixing layer: Linear
theory and direct simulation,’’ AIAA J. 28, 618 ~1990!.
4
T. L. Jackson and C. E. Grosch, ‘‘Inviscid spatial stability of a compress-
ible three-dimensional mixing layer,’’ J. Fluid Mech. 231, 35 ~1991!.
5
L. M. Mack, ‘‘On the inviscid acoustic-mode instability of supersonic
shear flows—Part 1: Two-dimensional waves,’’ Theor. Comput. Fluid Dyn.
2, 97 ~1990!.
6
T. L. Jackson and C. E. Grosch, ‘‘Inviscid spatial stability of a compress-
ible mixing layer,’’ J. Fluid Mech. 208, 609 ~1989!.
Oshin Peroomian 247