twist & shout: maximal enstrophy generation in 3-d...

Twist & Shout: Maximal EnstrophyGeneration in 3-D Navier-Stokes

LU LU, Ph.D. (2006)Wachovia Investments

CHARLES R. DOERINGUniversity of Michigan

Outline

I. Enstrophy & regularity and uniqueness of solutionsII. Analytic estimate on rate of enstrophy generationIII. Variational formulation of maximal productionIV. Computational results & a reality checkV. Conclusions, remarks & laments

Navier-Stokes equations:

• Periodic box:

• Initial condition:

Open ($1M) question:

Does a smooth solution exist for all t > 0?

!

r ˙ u +

r u "

r #

r u +

r # p = $%

r u

0 =r # "

r u

!

r u (

r x ,0) =

r u

0(r x )

WLOG r u

0(r x )d

3x = 0"( )

!

r x " [0,L]

3

Some definitions & some things we know:

• Kinetic energy:

• Vorticity:

• Enstrophy:

• If solution is smooth enough, dK/dt = –νE.

!

K(t) " 1

2

r u (

r x ,t)

2d3x# = 1

2

r u (

r x ,t)

2

2

!

r " =# $

r u %

r ˙ " +

r u &

r #

r " = '(

r " +

r " &

r #

r u

!

E(t) "r # (

r x ,t)

2

2=

r $

r u (

r x ,t)

2

2

% 8& 2

L2 K(t)

Global (in time) weak solutions exist:

• If K0<∞, there are weak solutions with finite energy,

• … and with finite integrated enstrophy,

• … but only known to satisfy an energy inequality,

• … and there is no assurance that they are unique.

!

K(t) " K0 #t $ 0

!

E(t)dt < "

ta

tb

# $0 % ta% t

b

!

K(tb) " K(t

a) #$ E(t)dt for a.e. t

a> 0

ta

tb

%

Local (in time) strong solutions exist:

• For (8π2/L2 )K0 ≤ E0 < ∞,

• Fact:

• And strong solutions are unique.

!

" T K0,E0,#( ) > 0 $ E(t) <% for 0 & t < T.

!

E(t) <" for ta# t # t

b

cr u ($,t)% C

"([0,L]

3) for t

a< t # t

b.

As long as the enstrophy is finite …

⇑Enstrophy generation rate G{u} = production – dissipation

!

dK

dt= "#E

dE

dt= "2#

r $

r %

2

2

+ 2r % &

r $ '

r u &

r % d3

x

!

= "2# $r u

2

2

+ 2r u %

r & '

r u % $

r u d

3x

1 2 4 4 4 4 4 3 4 4 4 4 4

Vortex stretching & enstrophy production:

Vorticity can be amplified; enstrophy can be produced.

Does this nonlinear process get out of control?

Can G{u} be estimated in terms of K and E?

Hölder & Cauchy-Schwarz⇓

Agmon-Sobolev-Gagliardo-Nirenberg (in 3D)

⇓

!

G{r u } = "2# $

r u

2

2+ 2

r u %

r & '

r u % $

r u d

3x

!

r u

"# ˜ c

r $

r u

2

1/ 2

%r u

2

1/ 2

= ˜ c E1

4 %

r u

2

1/ 2

!

" #2$ %r u

2

2

+ 2r u

&

r '

r u

2

%r u

2

= #2$ %r u

2

2

+ 2r u

& E

12

%r u

2

Hölder-Young (ab ≤ ap/p + bq/q when 1/p + 1/q=1)⇓

Integration by parts & Cauchy-Schwarz⇓

!

E =

r "

r u

2

2

= #r u $% &

r u d

3x '

r u

2 &

r u

2 = 2K &

r u

2

!

G{r u } " #2$ %

r u

2

2+ 2 ˜ c E

34

%r u

2

3 / 2

!

2˜ c E3

4 "

r u

2

3 / 2

#

c

$ 3E

3+ $ "

r u

2

2

System of differential inequations:

!

dK

dt= "#E

dE

dt$ "

#

2

E2

K+c

# 3E3

!

" E # E0

K

K0

$

% &

'

( )

1

2

1+2cK0E0

*4

K

K0

$

% &

'

( )

3

2

+1

,

-

.

.

.

/

0

1 1 1

$

%

& & &

'

(

) ) )

+1

... as long as RHS 2 0.

!

dK

dt= "#E

dE

dt$ "

#

2

E2

K+c

# 3E

3

!

dK

dt= "#E

dE

dt$ "

#

2

E2

K+c

# 3E

3

!

dK

dt= "#E

dE

dt$ "

#

2

E2

K+c

# 3E

3

Enstrophy decreases if E0K0 ≤ ν4/2c.

!

dK

dt= "#E

dE

dt$ "

#

2

E2

K+c

# 3E

3

Global existence and uniqueness if E0K0 ≤ 3ν4/2c.

!

dK

dt= "#E

dE

dt$ "

#

2

E2

K+c

# 3E

3

But does not prevent finite-time singularity if E0K0 > 3ν4/2c.

!

dK

dt= "#E

dE

dt$ "

#

2

E2

K+c

# 3E

3

!

dK

dt= "#E

dE

dt$ "

#

2

E2

K+c

# 3E

3

But does not prevent finite-time singularity if E0K0 ≥ 3ν4/2c.

Question:

How big can G{u} really get in terms of K and E?

• Analytic estimates don’t account for div u = 0 …

• or total competition between production & dissipation.

• Would like to solve the variational problem for max rate:

!

M(K,E;",L) = supr # $

r u = 0

G{r u } |

1

2

r u

2

2= K and

r #

r u

2

2

= E{ }

Settle for slightly less:

• We know that ℜ ≤ cE3/ν3 … but that ≠ $1M.

• “Critical” behavior is ℜ ~ E2 as E → ∞.

• Solve the Euler-Lagrange equations:

Computationally … via gradient ascent method.

!

"(E;#,L) = supr $ %

r u = 0

G{r u } |

r $

r u

2

2

= E{ }

so dE

dt&"(E)

!

0 ="

"r u

G{r u } + p

r # $

r u d

3x% + &

r #

r u

2

d3x%{ }

Starting from exact solution as E → 0 …

• Large E behavior is ℜ ~ E 1.78 (= 7/4?) … subcritial!

What do the maximizers look like?

periodic array of “vortex stretchers”

But this is a non-convex variational problem

• Euler-Lagrange solutions are local extrema …• So must see if there are other, global, maxima.

… another branch emerges at high E:

• Large E behavior is ℜ ~ E 2.997 (= 3?) … as estimated.

What do these maximizers look like?

colliding vortex rings

Another view …

Vorticity in a plane slice

Reality check:

!

Gr u { } = "2#

r $

r %

2

2

+ 2 %&2'

ur

r d

3x

z

r

For velocity fields with cylindrical symmetry …

!

" Gr u { } ~

E3

# 3

Then maximize over a …… max occurs at a ~ ν2/E

!

Circulation "

#

r $ ~

"

a2

%

& '

(

) *

E ~"

a2

%

& '

(

) *

2

a3

="

a

2

Reality check (continued):

!

Gr u { } = "#

r $

r %

2

2

+ 2 %&2'

ur

r d

3x

~ "#1

a

(

a2

)

* +

,

- .

2

a3 +

(

a2

)

* +

,

- .

2

1

a

(

a

)

* +

,

- . a

3

!

~ "#E

a2

+E3 / 2

a3 / 2

Conclusions, remarks & laments:

• Conclusion #1: The analytic asymptotic high-E estimateℜ(E;ν,L) ≤ cE3/ν3 can be saturated by divergence-free fields.

• Lament #1: no $1M to be found down this road!

• Conclusion #2: This “most dangerous” velocity field will

not produce a singularity in N-S.

• Lament #2: no $1M to be found down that road!

• Conclusion #3: K ~ 1/E for the optimizer, so we’re not sureif knowing the full upper limit M(K,E;ν,L) will help …

• Lament #3: so $1M not clearly down that road, either!

Maybe Lu will find $1M in Manhattan …

Just for fun … what do colliding vortex rings do?

(from website of Dr. T.B. Nickels <http://www2.eng.cam.ac.uk/~tbn22/Mov.html>)

The End

Twist & shout:Well, shake it up, baby, nowTwist and shout.C'mon c'mon, c'mon, c'mon, baby, now,Come on and work it on out.

Well, work it on out, honey.You know you look so good.You know you got me goin', now,Just like I knew you would.

Well, shake it up, baby, now,Twist and shout.C'mon, c'mon, c'mon, c'mon, baby, now,Come on and work it on out.

You know you twist your little girl,You know you twist so fine.Come on and twist a little closer, now,And let me know that you're mine.

Well, shake it, shake it, shake it, baby, now.Well, shake it, shake it, shake it, baby, now.Well, shake it, shake it, shake it, baby, now.Ahhhhhhhhhh(low) Ahhhhhhhhhh(higher)Ahhhhhhhhhh(higher) Ahhhhhhhhhhh(high)