Correlation, Spectrum, and Scales

Definition: Correlation Tensor (Two-Point)

u 2 ( x ) u 2 (x1 )
Consider a turbulent flow field as shown in
Figure 1. Here u 1 and u 1 are the components of
r
the velocity vectors and r = x 1 x 2 is the distance
between the two points. The two-point correlation
tensor is defines as u 1 ( x) u 1 ( x1 )
Qij ( x, x 1 ) = u i (x)u j (x 1 ) x x1

In a homogenous turbulent flow, the correlations Figure 1. Geometric features of two-point

(and all the statistics) are independent of the shift correlations in a turbulent flow field.
of space origin. That is,

Qij ( x, x 1 ) = Q ij (r )

Definition: Longitudinal Correlation Coefficient

The longitudinal correlation coefficient is defines as

Q11
f ( r) =
u 12 f (r)
where

Q11 = u 1 ( x)u 1 (x 1 ) , u12 = u 12 (x) = u 1 2 (x 1 )

Note that f(r) is an even function. That is

f ( r) = f ( r)
r
A typical longitudinal correlation coefficient is shown in Figure 2. Schematics of
Figure 2. longitudinal correlation
coefficient.

ME637 1 G. Ahmadi
Definition: Lateral Correlation Coefficient

The lateral correlation coefficient is defines as

Q 22 g (r)
g( r ) = , Q 22 = u 2 ( x)u 2 (x 1 )
u 22

The lateral correlation coefficient is also a symmetric

Function. That is

g(r ) = g( r)
r
8g
A typical lateral correlation coefficient is shown in
Figure 3. Figure 3. Schematics of lateral
correlation function and the
Definition: Taylor's Microscales corresponding Taylor Scale.

The Taylor microscales are defines as

2 2
2f = , 2g = ,
f (0) g (0)

where f and g are, respectively, the Taylor longitudinal and lateral microscales. The
microscales may be defined by fitting a parabola to the correlation coefficient curves at
r = 0 . That is,

1 2 r2


g( r ) = 1 + r g ( 0) + L 1 2
2! g

Definition: Integral Scales, Macroscales

The macroscales of turbulence are defined as

f = f (r)dr = Longitudinal Macroscale
0

g = g (r)dr = Lateral Macroscale
0

ME637 2 G. Ahmadi
Definition: Eulerian Time Correlation (stationary flows)

The Eulerian time correlation is defined as

u 1 (x, t) u 1 ( x, t + )
R E ( ) =
u12

The Eulerian time microscale E then is given by

2
2E =
R E (0)

The Eulerian time macroscale (integral scale) TE is defined as

TE = R E ()d
0

Using the uniform flow and frozen field approximations, the scales may be related. That
is

f UT E , f U E , f ( U) R E ( )

= U
t x

Definition: Lagrangian Time Correlation

The Lagrangian velocity correlation is defined as

v L ( t) vL (t + )
R L ( ) =
vL 2

where v L is the Lagrangian fluctuation velocity. The corresponding Lagrangian time

microscale L and the macroscale TL are given as

2
2L =
R L (0)
and

ME637 3 G. Ahmadi

TL = R L ()d
0

Definition: Energy Spectrum Tensor

Energy spectrum of is defined as the Fourier Transform of the correlation tensor.

That is

+ ++
1
E ij (k ) = 3
8 Q

ij (x)e ikx dx

+ ++
Qij ( x) = E

ij (k )e ikx dk

Definition: One Dimensional Energy Spectrum

The one dimensional energy spectrum is defined as

+
u12
f ( x )e
ik1x1
E l (k 1 ) = dx 1

1

1 +
u f ( x 1 ) = E 1 ( k 1 )e ik 1x1 dk 1
2
1
2

Symmetry of f ( x 1 ) implies that

2u12
0
E 1 (k 1 ) = f ( x 1 ) cos k 1 x 1dx 1
E(k1)

u 12 f ( x 1 ) = E1 ( k 1 ) cos k 1 x 1dk 1
0

A typical one dimensional energy spectrum is shown in

Figure 4.

Setting x 1 equal to zero, we find

k1
Figure 4. Schematics of one-
u12 = E 1 (k 1 )dk 1 . dimensional energy spectrum.
0

ME637 4 G. Ahmadi
Also

1 1 2f 1 2
= = 2 k 1 E (k 1 )dk 1
2f 2 x 12 x 1 =0
2u1 0

Example: The longitudinal correlation may be approximated as

r

f
f ( r) = e

The corresponding one dimensional spectrum is given as

u12 2 f
E 1 (k 1 ) =
1 + 2f k 12

ME637 5 G. Ahmadi
Estimates for Taylor Microscales

The energy dissipation is given as

u i u i
= (1)
x j x j

For isotropic turbulence, it can be shown that

u 1 2
= 15( ) = 15u 1 f (0)
2
(2)
x 1

It then follows that

u
2
u2 u2
= 30 12 = 30 2 = 15 2 (3)
f f g
Since

2f = 2g 2 (4)

Using the macroscopic estimate for the dissipation, we find

u3 u2
=A = 30 2 (5)
f

Therefore

f 30 1 / 2
= R (6)
A

Therefore,

f u
<< 1 since R = >> 1 (7)

Similarly

g 15 1/ 2
= R (8)
A

ME637 6 G. Ahmadi
 It may also be shown that

g 15 1 u
= R , R = (9)
A

and

g 225 1 / 4 1 / 4
=( ) R = 151 / 4 R 1 / 2 (10)
A

From Equation (3) it follows that

u 0.26
= 0.26 = (11)
g

where the Kolmogorov time scale is given by

= = (12)

This means that the Taylor microscale is not a characteristic length of the dissipation
eddies. It, however, provides a useful artificial length scale for estimating the velocity
gradients of the small eddies when macroscopic velocity scale is used for the velocity of
the eddies. That is, Equations (1) and (11) imply that

u i u i u
= ~ ( )2 (13)
x j x j g

Other useful estimates are

u i u i u u3 u2 u2
d ij d ij ~ ~ ( )2 , ~ 2 , ~ (14)
x j x j
Also

~ R 1 / 2 ~ R 1 , ~ R 3 / 4 ~ R 3 / 2 , ~ R 1 / 4 ~ R 1 / 2 (15)

and

2 = 3 (16)

ME637 7 G. Ahmadi