rotating turbulence
Bernard J. Geurts, Darryl Holm, Arek Kuczaj
Multiscale Modeling and Simulation (Twente)
Anisotropic Turbulence (Eindhoven)
Mathematics Department, Imperial College (London)
IMS Turbulence Workshop
London, March 26, 2007
connect DNS to experiment - rotating decaying turbulence
show accuracy of regularization modeling - LES
apply regularization modeling to high Re flow
quantify cross-over from 3D to 2D dynamics as Ro
↓
connect DNS to experiment - rotating decaying turbulence
show accuracy of regularization modeling - LES
apply regularization modeling to high Re flow
quantify cross-over from 3D to 2D dynamics as Ro
↓
connect DNS to experiment - rotating decaying turbulence
show accuracy of regularization modeling - LES
apply regularization modeling to high Re flow
quantify cross-over from 3D to 2D dynamics as Ro
↓
connect DNS to experiment - rotating decaying turbulence
show accuracy of regularization modeling - LES
apply regularization modeling to high Re flow
quantify cross-over from 3D to 2D dynamics as Ro
↓
illustrate low Ro compressibility effects
connect DNS to experiment - rotating decaying turbulence
show accuracy of regularization modeling - LES
apply regularization modeling to high Re flow
quantify cross-over from 3D to 2D dynamics as Ro
↓
1
Rotation modulates turbulence
2
Rotation of incompressible fluid
3
Rotation at high Reynolds numbers
4
Rotation induced compressibility effects
Outline
1
Rotation modulates turbulence
2
Rotation of incompressible fluid
3
Rotation at high Reynolds numbers
4
Rotation induced compressibility effects
Taylor-Proudman: columnar structuring
Strong rotation/small friction:
“Motion of a homogeneous fluid will be the same in all
planes perpendicular to the axis of rotation”
Taylor’s experiment:
When an object moves in a rotating flow, it drags along
with it a column of fluid parallel to the rotation axis
Taylor-Proudman: columnar structuring
Strong rotation/small friction:
“Motion of a homogeneous fluid will be the same in all
planes perpendicular to the axis of rotation”
Taylor’s experiment:
When an object moves in a rotating flow, it drags along
with it a column of fluid parallel to the rotation axis
Taylor-Proudman: columnar structuring
Strong rotation/small friction:
“Motion of a homogeneous fluid will be the same in all
planes perpendicular to the axis of rotation”
Taylor’s experiment:
When an object moves in a rotating flow, it drags along
with it a column of fluid parallel to the rotation axis
Guadalupe Islands
Rotating turbulence in a container
Experiment:
Morize, Moisy, Rabaud, PoF, 2005
Ro
=
6.4
Ro
=
0.6
Ro
=
0.3
Vertical structuring
of flow as Ro
=
u
r
/
(
2
Ω
r
ℓ
r
)
decreases
Rotating turbulence in a container
Experiment:
Morize, Moisy, Rabaud, PoF, 2005
Ro
=
6.4
Ro
=
0.6
Ro
=
0.3
Vertical structuring
of flow as Ro
=
u
r
/
(
2
Ω
r
ℓ
r
)
decreases
Modulation:
Competition 3D turbulence – 2D structuring
Rotating turbulence in a container
×
Ro
=
5.5
+
Ro
=
1.2
*
Ro
=
0.35
Rotating Rayleigh-B ´enard problem
Rotating Rayleigh-B ´enard problem
Temperature:
buoyancy dominated (Ta
=
0 - left) and
rotation dominated (Ta
>
Ra)
Rotating Rayleigh-B ´enard problem
Temperature:
buoyancy dominated (Ta
=
0 - left) and
rotation dominated (Ta
>
Ra)
Outline
1
Rotation modulates turbulence
2
Rotation of incompressible fluid
3
Rotation at high Reynolds numbers
4
Rotation induced compressibility effects
Governing equations
Incompressible formulation:
rotating frame of reference
∇ ·
u
=
0
∂
t
u
+
∇ ·
(
u u
) +
∇
p
−
1
Re
∇
2
u
=
u
×
Ω
Ro
Characteristic numbers:
Re
=
u
r
ℓ
r
ν
r
;
Ro
=
u
r
2
Ω
r
ℓ
r
x z yNumerical method:
Pseudo-spectral discretization:
full de-aliasing, parallelized
Helical-wave decomposition:
Coriolis diagonalization
Governing equations
Incompressible formulation:
rotating frame of reference
∇ ·
u
=
0
∂
t
u
+
∇ ·
(
u u
) +
∇
p
−
1
Re
∇
2
u
=
u
×
Ω
Ro
Characteristic numbers:
Re
=
u
r
ℓ
r
ν
r
;
Ro
=
u
r
2
Ω
r
ℓ
r
x z yNumerical method:
Pseudo-spectral discretization:
full de-aliasing, parallelized
Helical-wave decomposition:
Coriolis diagonalization
Benefit of helical wave decomposition (HWD)
Decay of kinetic energy:
0
0.05
0.1
0.15
0.2
0.42
0.43
0.44
0.45
0.46
0.47
0.48
0.49
0.5
t
E
Rotation: Ro
=
10
−2
RK4 with HWD
(solid)
at
∆
t
=
3.10
−3
RK4 without HWD at
∆
t
=
7.10
−3
(+)
∆
t
=
5.10
−3
(
⊲
)
∆
t
=
5.10
−
4
(
◦
)
Vertical structuring - vorticity
(a)
(b)
(c)
(d)
Decaying turbulence at R
λ
=
92.
Snapshot t
=
2.5: Ro
=
∞
Geometrical statistics - alignment
Eigenvalues/vectors S
ij
:
(
λ
i
,
z
i
)
(
positive
)
λ
1
> λ
2
> λ
3
(
negative
)
−1 −0.8−0.6−0.4−0.2 0 0.2 0.4 0.6 0.8 1 1.8 2 2.2 2.4 2.6 2.8 3 3.2x 105 p df cos(ω, z1)(1)
−1−0.8−0.6 −0.4−0.2 0 0.2 0.4 0.6 0.8 1 0 2 4 6 8 10 12 14x 105 p df cos(ω, z2)(2)
−1 −0.8−0.6−0.4−0.2 0 0.2 0.4 0.6 0.8 1 0 1 2 3 4 5 6 7x 10 5 p df cos(ω, z3)(3)
cos
(
ω
,
z
i
)
i
=
1
Ro
→ ∞
– aligned or perpendicular, rotation reduces
parallel alignment
i
=
2
alignment
ω
and z
2
– reduced by rotation
i
=
3
Ro
→ ∞
– mainly perpendicular, reduced preference with
rotation
Geometrical statistics - vortex stretching
−1 −0.8 −0.6 −0.4 −0.2 0 0.2 0.4 0.6 0.8 1 0 2 4 6 8 10 12x 10 5p
df
cos(
ω, W
)
cos
(
ω
,
W
)
Ro
→ ∞
vortex stretching dominant over vortex compression
Depleted skewness at strong rotation
Experimental observation - reduced skewness at high 1
/
Ro:
Cambon et al.
S
=
A
(
1
+
2Ro
−2
)
1
/
2
;
S
∼
Ro
∼
Ω
Depleted skewness at strong rotation
Simulated skewness:
DNS at 512
3
0 1 2 3 4 5 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 t S(a)
10 −1 100 101 10−1 100 Roω S(b)
Velocity skewness versus t (a) and versus Ro (b)
Vorticity skewness
(a)
10 0 102 10−2 10−1 100 (Ωt)0.6 Ωt/(2π) Sω(b)
Experimental (a) and simulated (b) vorticity skewness at
various rotation rates
Ω
Algebraic decay of kinetic energy
DNS at 256
3
:
0 1 2 3 4 5 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 t E(a)
10 −1 100 101 10−2 10−1 t E(b)
Algebraic decay exponent
0 20 40 60 80 100 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 1/Ro n(a)
10 0 101 102 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1/Ro n(b)
Outline
1
Rotation modulates turbulence
2
Rotation of incompressible fluid
3
Rotation at high Reynolds numbers
4
Rotation induced compressibility effects
DNS and LES in a picture
Classical problem: wide dynamic range
capture both large and small scales: resolution problem
N
∼
Re
9
/
4
;
W
∼
Re
3
If Re
→
10
×
Re then N
→
175N and W
→
1000W
Coarsening/mathematical modeling instead: LES
Filtering Navier-Stokes equations
Convolution-Filtering:
filter-kernel G
u
i
=
L
(
u
i
) =
Z
G
(
x
−
ξ
)
u
(
ξ
)
d
ξ
Application of filter:
∂
t
u
i
+
∂
j
(
u
i
u
j
) +
∂
i
p
−
1
Re
∂
jj
u
i
−
1
Ro
(
u
×
Ω
)
i
=
−
∂
j
τ
ij
Turbulent stress tensor:
τ
ij
=
u
i
u
j
−
u
i
u
j
Shorthand notation:
Closure problem
NS
(
u
) =
0
⇒
NS
(
u
) =
−∇ ·
τ
(
u,
u
)
⇐ −∇ ·
M
(
u
)
Closed LES formulation
Find v
:
NS
(
v
) =
−∇ ·
M
(
v
)
Filtering Navier-Stokes equations
Convolution-Filtering:
filter-kernel G
u
i
=
L
(
u
i
) =
Z
G
(
x
−
ξ
)
u
(
ξ
)
d
ξ
Application of filter:
∂
t
u
i
+
∂
j
(
u
i
u
j
) +
∂
i
p
−
1
Re
∂
jj
u
i
−
1
Ro
(
u
×
Ω
)
i
=
−
∂
j
τ
ij
Turbulent stress tensor:
τ
ij
=
u
i
u
j
−
u
i
u
j
Shorthand notation:
Closure problem
NS
(
u
) =
0
⇒
NS
(
u
) =
−∇ ·
τ
(
u,
u
)
⇐ −∇ ·
M
(u)
Closed LES formulation
Find v
:
NS
(
v
) =
−∇ ·
M
(
v
)
Filtering Navier-Stokes equations
Convolution-Filtering:
filter-kernel G
u
i
=
L
(
u
i
) =
Z
G
(
x
−
ξ
)
u
(
ξ
)
d
ξ
Application of filter:
∂
t
u
i
+
∂
j
(
u
i
u
j
) +
∂
i
p
−
1
Re
∂
jj
u
i
−
1
Ro
(
u
×
Ω
)
i
=
−
∂
j
τ
ij
Turbulent stress tensor:
τ
ij
=
u
i
u
j
−
u
i
u
j
Shorthand notation:
Closure problem
NS
(
u
) =
0
⇒
NS
(
u
) =
−∇ ·
τ
(
u,
u
)
⇐ −∇ ·
M
(u)
Closed LES formulation
Find v
:
NS
(
v
) =
−∇ ·
M
(
v
)
Regularization models
Alternative modeling approach
‘first principles’ – derive implied subgrid model
achieve unique coupling to filter, e.g., Leray, LANS-α
Example:
Leray regularization - alteration convective fluxes
∂
t
u
i
+
u
j
∂
j
u
i
+
∂
i
p
−
1
Re
∆
u
i
=
0
LES template:
∂
t
u
i
+
∂
j
(
u
j
u
i
) +
∂
i
p
−
1
Re
∆
u
i
=
−∂
j
m
L
ij
Implied Leray model
m
L
ij
=
L
u
j
L
−1
(
u
i
)
−
u
j
u
i
=
u
j
u
i
−
u
j
u
i
Rigorous analysis available (
∼
1930s)
Provides accurate LES model (
∼
03)
Regularization models
Alternative modeling approach
‘first principles’ – derive implied subgrid model
achieve unique coupling to filter, e.g., Leray, LANS-α
Example:
Leray regularization - alteration convective fluxes
∂
t
u
i
+
u
j
∂
j
u
i
+
∂
i
p
−
1
Re
∆
u
i
=
0
LES template:
∂
t
u
i
+
∂
j
(
u
j
u
i
) +
∂
i
p
−
1
Re
∆
u
i
=
−∂
j
m
L
ij
Implied Leray model
m
L
ij
=
L
u
j
L
−1
(
u
i
)
−
u
j
u
i
=
u
j
u
i
−
u
j
u
i
Rigorous analysis available (
∼
1930s)
Provides accurate LES model (
∼
03)
Regularization models
Alternative modeling approach
‘first principles’ – derive implied subgrid model
achieve unique coupling to filter, e.g., Leray, LANS-α
Example:
Leray regularization - alteration convective fluxes
∂
t
u
i
+
u
j
∂
j
u
i
+
∂
i
p
−
1
Re
∆
u
i
=
0
LES template:
∂
t
u
i
+
∂
j
(
u
j
u
i
) +
∂
i
p
−
1
Re
∆
u
i
=
−∂
j
m
L
ij
Implied Leray model
m
L
ij
=
L
u
j
L
−1
(
u
i
)
−
u
j
u
i
=
u
j
u
i
−
u
j
u
i
Rigorous analysis available (
∼
1930s)
Provides accurate LES model (
∼
03)
Regularization models
Alternative modeling approach
‘first principles’ – derive implied subgrid model
achieve unique coupling to filter, e.g., Leray, LANS-α
Example:
Leray regularization - alteration convective fluxes
∂
t
u
i
+
u
j
∂
j
u
i
+
∂
i
p
−
1
Re
∆
u
i
=
0
LES template:
∂
t
u
i
+
∂
j
(
u
j
u
i
) +
∂
i
p
−
1
Re
∆
u
i
=
−∂
j
m
L
ij
Implied Leray model
m
L
ij
=
L
u
j
L
−1
(
u
i
)
−
u
j
u
i
=
u
j
u
i
−
u
j
u
i
Rigorous analysis available (∼
1930s)
Provides accurate LES model (∼
03)
NS-
α
regularization
Kelvin’s circulation theorem
d
dt
I
Γ(
u
)
u
j
dx
j
−
1
Re
I
Γ(
u
)
∂
kk
u
j
dx
j
=
0
⇒
NS
−
eqs
t
1
t
2
u
u
NS-
α
regularization
α
-principle: Fluid loop:
Γ(
u
)
→
Γ(
u
)
⇒
Euler-Poincar ´e
∂
t
u
j
+
u
k
∂
k
u
j
+
u
k
∂
j
u
k
+
∂
j
p
−
∂
j
(
1
2
u
k
u
k
)
−
1
Re
∂
kk
u
j
=
0
Rewrite into LES template: Implied subgrid model
∂
t
u
i
+
∂
j
(
u
j
u
i
) +
∂
i
p
−
1
Re
∂
jj
u
i
=
−∂
j
u
j
u
i
−
u
j
u
i
−
1
2
u
j
∂
i
u
j
−
u
j
∂
i
u
j
=
−∂
j
m
ij
α
Cascade-dynamics – computability
k
k
∼ 1/∆
−13/3E
k
−3M
k
k
k
k
L
DNS
−5/3NSa
NS-α,
Leray
are dispersive
Regularization alters spectrum –
controllable cross-over as
Confined and nonconfined decay
Decay exponent n and spectral exponent p
Nonconfined
Confined
Nonrotating
p
=
5/3
n
=
6/5
n
=
2
Rotating(I)
p
=
2
n
=
3/5
n
=
1
Rotating(II)
p
=
3
n
=
0
n
=
0
Rotating I – energy transfers time scale
Ω
−
1
LES: Algebraic decay at R
λ
=
92
DNS at 256
3
– LES at 64
3
:
0 20 40 60 80 100 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 1/Ro n(a)
10 0 101 102 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1/Ro n(b)
Decay exponent well captured:
DNS,
Leray
,
LANS-
α
LES: Algebraic decay at R
λ
=
200
LES at 128
3
:
0 20 40 60 80 100 0 0.5 1 1.5 2 2.5 1/Ro n(a)
10 0 101 102 0 0.5 1 1.5 2 1/Ro n(b)
Leray - dashed
,
LANS-
α
– solid
Leray closer to ‘nonconfined’, LANS-α
closer to ‘confined’
Evidence Rotating II - Kraichnan
LES: Leray spectra at R
λ
=
200
LES at 128
3
:
totally inhibited energy transfer regime
100 101 102 −1 0 1 2 3 4 5x 10 −3 k/(2π) Π