Corresponding author.
0169-5983/98/$19.00 c 1998 The Japan Society of Fluid Mechanics and
Elsevier Science B.V. All rights reserved.
PII: S0 1 6 9 - 5 9 8 3 ( 9 8 ) 0 0 0 0 4 - 5
344 H.P. Mazumdar, H.I. Andersson / Fluid Dynamics Research 23 (1998) 343348
turbulence in a uniform shear ow. He proposed that pressure uctuations remain highly corre-
lated right across a free shear ow. George et al. (1984) presented spectral models for turbulent
pressure uctuations in a homogeneous constant mean-shear ow. They showed that the pressure
spectrum exhibits k
7}3
and k
11}2
forms in the inertial subrange, respectively, for the cases when
turbulenceturbulence interaction and the turbulence-mean shear contribution dominate. Following a
review of the more important studies in this eld, Larchev eque (1990) applied eddy damped quasi-
normal Markovian closure approximations to obtain an equation for the pressure spectrum, whereas
Pullin (1995) calculated pressure spectra at large wave numbers for LundgrenTownsend vortex
models of the ne scales of homogeneous turbulence. Hill and Wilczak (1995) made assumptions
of local homogeneity, local isotropy, and compressibility and used the NavierStokes equation to
determine relationships between several fourth-order velocity statistics and the pressure structure
function.
In this paper, we calculate the pressure spectrum following RDT in a homogeneous low Reynolds
number turbulent ow subjected to uniform shear and weak rotation. This is a sequel to our earlier
work (Andersson and Mazumdar, 1993) for such a turbulent ow eld wherein we followed a 2D
rapid distortion and obtained turbulence energy spectra.
2. Mathematical formulation: pressure spectrum
Let us consider the uctuating pressure eld of an incompressible, homogeneous turbulent ow
which is subjected to a uniform shear and spanwise rotation. The equation governing the covariance
of pressure uctuations pertaining to the points P(x, t) and P
(x
cr
i
cr
i
=2[
c
cr
1
u
2
c
2
cr
i
cr
k
u
i
u
k
+ 2O
3
c
cr
1
u
2
2O
3
c
cr
2
u
1
, (1)
where is the pressure uctuation, j the density, x
i
a position coordinate, u
i
a uctuating component
of velocity, [ =dU
1
}dx
2
is the constant mean shear and r
i
is a component of the vector r extending
from P to P
1
as
1
j
c
2
u
1
cr
i
cr
i
=2[
c
cr
1
u
2
u
1
2O
3
c
cr
1
u
2
u
1
+ 2O
3
c
cr
2
u
1
u
1
2
c
2
u
i
u
k
u
1
cr
i
cr
k
, (2)
and an analogous equation for u
2
is obtained by replacing u
1
with u
2
.
The Fourier transforms of the quantities
, u
1
, u
2
, u
1
u
1
, u
1
u
2
, u
2
u
1
and u
2
u
2
are denoted by
, z
0,1
, z
0,2
, [
11
, [
12
, [
21
, and [
22
, where for instance, [
22
is dened according to
u
2
u
2
=
_
[
22
e
ikr
dk. (3)
Now, neglecting triple correlation terms in the equations for
, u
1
and u
2
and then taking their
Fourier transforms, we obtain
1
j
k
2
=2[ik
1
z
0,2
2ik
1
O
3
z
0,2
+ 2O
3
ik
2
z
0,1
, (4)
H.P. Mazumdar, H.I. Andersson / Fluid Dynamics Research 23 (1998) 343348 345
k
2
j
z
0,1
=2[ik
1
[
21
2O
3
ik
1
[
21
+ 2O
3
ik
2
[
11
, (5)
k
2
j
z
0,2
=2[ik
1
[
22
2O
3
ik
1
[
22
+ 2O
3
ik
2
[
12
. (6)
The continuity equation for an incompressible ow, k
i
[
i)
=0, is to be satised for all time t0.
While Deissler (1962) explored the evolution of initially three-dimensional turbulence, we consider
the evolution with time of a plane-wave disturbance (Hasen, 1967) in the x
1
x
2
-plane. That is, we
require that
[
13
=[
23
=[
33
=0. (7)
It can be easily shown that
[
12
=[
21
=
k
2
k
1
[
22
and [
11
=
k
2
k
1
[
21
=
_
k
2
k
1
_
2
[
22
. (8)
Eliminating z
0,1
and z
0,2
from Eq. (4) with the help of (5) and (6), and taking into account the
relation [
12
=[
21
, we obtain
=[
22
_
4[
2
k
2
1
j
2
k
4
8[O
3
k
2
1
j
2
k
4
+
4O
2
3
k
2
1
j
2
k
4
_
+[
12
_
8[O
3
k
1
k
2
j
2
k
4
8O
2
3
k
1
k
2
j
2
k
4
_
+ [
11
4O
2
3
k
2
2
j
2
k
4
. (9)
In view of the relations (8), Eq. (9) reduces nally to
=4[
2
j
2
[
22
_
1 +
R
2
4
R
_
k
2
1
k
4
+ 4[
2
j
2
[
22
_
R
_
R
2
1
__
k
2
2
k
4
+ [
2
j
2
[
22
R
2
k
4
2
k
2
1
k
4
, (10)
where R is a rotation number, dened as R=2O
3
}(dU
1
}dx
2
), or a reciprocal Rossby number. The
particular spectrum component [
22
was given explicitly by Andersson and Mazumdar (1993) as
[
22
(k, t) =[(k
1
, k
2
+ k
1
[t, k
3
) exp
_
2v
_
k
2
t + k
1
k
2
[t
2
+
1
3
k
2
1
[
2
t
3
_
_
k
2
1
+ (k
2
+ k
1
[t)
2
+ k
2
3
k
2
1
+ k
2
2
+ k
2
3
_
2{1+(O
3
}[)(k
2
3
}k
2
1
)}
, (11)
where [ is a function that depends on the initial conditions at t =0, i.e. [
22
(k, 0) =[(k
1
, k
2
, k
3
). In
their analysis, Andersson and Mazumdar (1993) considered an initial spectrum tensor satisfying the
relations (8) as
[
i)
(k, 0) =A[o(k + k
0
) + o(k k
0
)]{(o
i)
k
2
k
i
k
)
) o
i3
o
)3
(k
2
1
+ k
2
2
)}, (12)
where k
0
=(k
1
, k
2
, 0) and A is a constant. It follows from Eq. (12) that [
22
(k, 0) =2Ak
2
1
. Let us
now express the wave number vector in spherical coordinates, i.e. k =(k cos [ sin 0, k sin [ sin 0,
k cos 0). Eq. (10) can thus be written as
(k, [, 0) =(4[
2
j
2
}k
2
) sin
2
0
_
cos
2
[ +
1
4
R
2
sec
2
[ R
_
[
22
, (13)
346 H.P. Mazumdar, H.I. Andersson / Fluid Dynamics Research 23 (1998) 343348
which, in the absence of rotation, reduces to Deisslers (1962) expression for (k, [, 0), e.g.,
(k, [, 0) =(4[
2
j
2
}k
2
) cos
2
[sin
2
0 [
22
(k, [, 0).
Integrating (k, [, 0) over all directions in wave number space, we obtain the spectrum of
2
as
(
2
)
k
=
_
0
_
2
0
(k, [, 0)k
2
sin 0 d[d0, (14)
where
2
=
_
0
(
2
)
k
dk. (15)
Finally, introducing the dimensionless variables
[
22
=[
22
(vt)
2
}A,
k =(vt)
1}2
k,
t =[t, (16)
we calculate the nondimensionalized quantities.
3. Results and discussion
Initially the spherical averaging in Eq. (14) can be accomplished analytically, leading to
(
2
)
k
vt
4A[
2
j
2
=
4
5
_
1 +
2
3
R(R 2)
_
2
k
2
exp(2
k
2
), (17)
and the one-point correlation (15) at
t =0 becomes
2
(vt)
3}2
4A[
2
j
2
=
3
50
_
1 +
2
3
R(R 2)
_
. (18)
Here, it is noteworthy that the pressure spectrum (17) and the pressure moment (18) depend explicitly
on the rotation number R. This contrasts with the initial velocity spectra derived by Andersson and
Mazumdar (1993), which turned out to be independent of R.
Spectra of
2
were computed for
t0 and the results for R=0 are presented in Fig. 1. The
numerical integration was accomplished by repeated one-dimensional integrations using an adapted
version of the QUAD3D routine of Press et al. (1986). The QTRAP-routine based on the extended
trapezoidal rule was adopted as the one-dimensional integrator. The accuracy of the integration was
veried by comparison with the analytical solutions (17) and (18), and both were reproduced to
within 10
4
. The spherically averaged spectra in Fig. 1 resemble those for the spherical averages
of [
22
reported earlier by Andersson and Mazumdar (1993), but the maximum values are shifted
to the left, i.e. towards lower wave numbers. Information about the most inuential scales of the
uctuating pressure eld has obvious relevance for the understanding of turbulence-generated noise.
Spectra of
2
were computed also for R=0.01 i.e. the same rotation numbers as those considered
by Andersson and Mazumdar (1993). The eect of weak rotation would scarcely be discernible if
these spectra had been included in Fig. 1. This is because the imposed background vorticity 2O
3
corresponds only to 1% of the mean shear vorticity dU
1
}dx
2
when the magnitude of R is 0.01.
H.P. Mazumdar, H.I. Andersson / Fluid Dynamics Research 23 (1998) 343348 347
Fig. 1. Dimensionless spectra of
2
at various dimensionless times
t without rotation.
Table 1
Variation of
2
(vt)
3}2
}4A[
2
j
2
with time
t for dierent rotation numbers