Você está na página 1de 29

Existence of vortex sheets with re ection

symmetry in two space dimensions

M C Lopes Filho

H J Nussenzveig Lopes

Zhouping Xin

The main purpose of this work is to establish existence of a weak

solution to the incompressible D Euler equations with initial vortic

ity consisting of a Radon measure with distinguished sign in

compactly supported in the closed right half plane superimposed to

its odd re ection in the left half plane We make use of a new a pri

ori estimate to control the interaction between positive and negative

vorticity at the symmetry axis We prove that a weak limit of a se

quence of approximations obtained by either regularizing the initial

data or by the vanishing viscosity method is a weak solution of the

incompressible D Euler equations We also establish the equivalence

at the level of weak solutions between mirror symmetric ows in the

full plane and ows in the half plane Finally we extend our exis

tence result to odd perturbations without distinguished sign of

our original initial vorticity

Con tents

Symmetric smooth solutions

A priori estimates in the inviscid case

Convergence theorem

The method of images

Viscous approximations

Extensions and conclusions

In troduction

In J M Delort proved existence of weak solutions to the incompressible

D Euler equations with vortex sheet initial data under the assumption that

the initial vorticity be the sum of a Radon measure with distinguished sign

and an arbitrary L p function p see This result was later proved in

di erent ways see extended to p see and to the conver

gence to a weak solution of approximations generated by vanishing viscosity

see Delort s result was also extended to convergence of approxima

tions obtained by vortex methods see The problem of existence

for vortex sheet initial data without the distinguished sign assumption has

remained open The distinguished sign hypothesis is needed because under

it conservation of energy implies the avoidance of concentrations in vorticity

see However the distinguished sign assumption is a signi cant

restriction on the scope of the available theory because important features

of irregular ow seem to be connected with intricate intertwining of regions

of positive and negative vorticity

In the present work we will prove an existence result for ows with initial

vorticity which is odd with respect to a straight line and which is of a

distinguished sign on each side of the line Setting the initial data in this

way creates a situation where vorticity of di erent signs is allowed to interact

but not to intertwine This is the rst instance where the global in time

existence of vortex sheet evolution is rigorously established with vorticity

density changing sign albeit under very special circumstances The di culty

is to show that vorticity concentration does not occur along the symmetry

axis The key new ingredient in our proof is an a priori estimate on the

velocity at the symmetry axis which allows us to control the interaction

between positive and negative vorticity Our result shows that the di culty

in the analytical treatment of vortex sheets without distinguished sign stems

from the intertwining of the positive and negative vorticity and not merely

from their interaction One concrete situation where a vortex sheet initial

data problem with re ection symmetry appears is the modelling of the wake

due to an elliptically loaded airplane wing in the Tre tz plane as done by

R Krasny in

The incompressible D Euler equations are covariant with respect to re

ection symmetry This means that a smooth solution of the equations on

a half plane may be extended by re ection to a smooth solution in the full

plane This observation is imbedded in the method of images see Thus

a natural approach to our existence problem would be to extend Delort s

existence theory to ows in the half plane and use the covariance of the D

Euler equations with respect to the re ection symmetry in order to obtain

a symmetric weak solution in the full plane This approach does not work

The extension of Delort s theorem for bounded domains to half plane ows

is rather routine without any new additional estimates compared with the

case without boundaries However the weak solution thus obtained assumes

the boundary condition velocity tangent to the boundary in a way that is

not strong enough to guarantee that the ow constructed by re ection is a

weak solution in the full plane

For ows in a bounded domain Delort s theorem guarantees the existence

of a weak solution in an interior sense i e the solution satis es the weak

formulation of the equations with respect to test functions that are com

pactly supported in the interior of the domain and it satis es the boundary

condition in a trace sense On one hand the boundary condition is linear

and hence it is well behaved with respect to weak convergence which is nec

essary in Delort s treatment of boundaries On the other hand this separate

treatment of the boundary condition leaves open the possibility of vortic

ity concentrating at the boundary We introduce another notion of weak

solution which we call boundary coupled weak solution where we use test

functions that vanish at the boundary but not their derivatives We prove

that the existence of a re ection symmetric weak solution to the full plane

problem is equivalent to the existence of a boundary coupled weak solution

to the half plane problem One corollary of this result together with the ex

istence of a weak solution to the full plane problem with re ection symmetry

is that there exists a boundary coupled weak solution to the incompressible

D Euler equations in the half plane with nonnegative measures as initial

vorticity

The problem of existence of vortex sheet evolution must be understood in

the context of the pioneering work of R DiPerna and A Majda in

The concern with concentrations in kinetic energy which lies at the core of

their analysis has played no role in our work in the same way that it was

not present in Delort s work Energy concentration is possible and in fact

it is an outstanding open problem whether it does actually occur dynami

cally Rather our analysis revolves around the possibility of concentrations

in vorticity This was already the case with Delort s work as explained by

S Schochet in one key fact in Delort s result is that concentrations in

vorticity are excluded by logarithmic decay of circulation in small circles

The main result in is a concentration cancellation theorem in which

approximate vorticities which concentrate at a single point x x t with

x aC function of time are shown to possess a weak limit which is a

weak solution In contrast our theorem is not a concentration cancellation

result Although vorticity concentrations are not ruled out the possibility

of their occurrence is bypassed We show that in a time averaged sense no

concentrations occur This is enough to pass to the limit in the nonlinear

term of the weak vorticity formulation

The remainder of this article is divided in six sections In the rst one we

construct symmetric smooth solutions In the second one we obtain the new

a priori estimate that makes our analysis possible In the third section we

construct inviscid approximate solution sequences and we apply the a priori

estimates obtained in the second section to the weak vorticity formulation

proving our main existence result In the fourth section we prove the validity

of the method of images obtaining existence of a boundary coupled weak

solution for the half plane problem In Section we prove the convergence

of a subsequence of the viscous approximations to a weak solution in the

full plane and in the last section we extend our main result to odd L

perturbations without sign restriction of the initial vorticities previously

considered and present our conclusions

Symmetric smooth solutions

In order to construct symmetric weak solutions to the D Euler system we

must rst be able to construct symmetric approximate solution sequences

In this article we consider two families of approximate solution sequences

obtained by mollifying the initial data and either exactly solving the Euler

equations or exactly solving the Navier Stokes equations It is easy to mollify

preserving the symmetry of the initial data so that we need to show that

the Navier Stokes and the Euler systems preserve the symmetry This can be

accomplished by a standard energy estimate argument which we will outline

below

The D incompressible Navier Stokes equations in the full plane and in

vorticity form are given by

t u r

in

div u

curl u in

x x in

juj x t

as jxj

The incompressible D Euler equations in vorticity form correspond to

If is smooth and compactly supported then for any there

exists a unique smooth solution to the problem above see It is easy

to see that the solution is also compactly supported in space for all time if

If x t is exponentially decaying as jxj This is a

consequence of the properties of the parametrix for the linearized parabolic

problem see

The elliptic system div u curl u in the full plane together

with the condition juj at can be inverted explicitly so as to express

velocity in terms of vorticity by

u x t K t x Z yj y t dy

jx

x y

with z z z z This identity is called the Biot Savart law Using

the Biot Savart law the system becomes a scalar nonlocal equation with

the vorticity as the single unknown

De nition A Radon measure BM is said to be NMS nonneg

ative mirror symmetric if restricted to the right half plane fx g is

nonnegative and is odd with respect to the fx g axis i e the duality

pairing h i vanishes for any C which is even with respect to the

rst variable

We will use the following notation for the mirror symmetry x x x

x x x We note that the restriction of a bounded Radon measure

in the plane to any open subset of the plane with smooth boundary gives

rise to a bounded Radon measure on This is an immediate consequence

of the characterization of BM as the dual of the space C the closure

of C with respect to the sup norm

c

Proposition L et and let C be NMS If x t is

c

the unique solution of with initial data then t is NMS for al l

t

Proof De ne x t x t x t so that x

e

e

Then satis es the following equation

e

t x t

e

u x t r x t

e

x t u x t r x t u x t r x t

e

We rewrite the r h s of this equation using the explicit form of the

Biot Savart law to obtain

t x t u x t r x t x t K x t r x t

e

e

e

e

Multiply this identity by integrate by parts over all to obtain

e

d

dt k

k

e

L Z x t r x t dx

K

e

e

k

k L kK

e

k L p p kr k L p

e

where p is arbitrary The exponent p p is precisely the critical

Sobolev exponent corresponding to p p p Therefore since p

one can use the Hardy Littlewood Sobolev inequality see Proposition

to get

kK

e k L p p C p k

e k L p

In the case t has compact support and hence the L p norm

e

above is dominated by the L norm so that by Gronwall s inequality

e

vanishes identically This means that t is odd with respect to the rst

variable which implies by the explicit form of the Biot Savart law that u is

tangent to the fx g axis so that each half plane is invariant under the

ow Hence t is nonnegative in the right half plane so that t is

NMS

If one can use the exponential decay of t at in nity and

e

Holder s inequality to conclude that

k

k L p C R k k L e KR

e

e

where C K are positive constants R R for R su ciently big and

p p so that One inputs this information into

the di erential inequality and uses Gronwall s inequality to conclude that

e

t k KR for arbitrary R R so that Finally we

k e

L Ce

e

observe that x t is now odd with respect to x so that x t

Hence satis es a linear Fokker Planck equation on the half plane

with a homogeneous Dirichlet boundary condition and nonnegative initial

condition By the maximum principle the nonnegativity is preserved for

future times so that is NMS in this case as well

for future times so that is NMS in this case as well A priori estimates in

A priori estimates in the inviscid case

We consider the problem of existence of a weak solution to the incompressible

D Euler equations in the full plane with initial velocity u L such

that the vorticity curl u is NMS and compactly supported We will

look for solutions satisfying the weak vorticity formulation of the equations

as was introduced by S Schochet in which we make explicit in the

de nition below

First we note that if BM then it is possible to make sense of K

as a distribution To see this observe that the map K is a continuous

linear operator from C into C where C denotes the continuous

c

functions vanishing at in nity Hence using that K z K z we may

de ne K as a distribution by the relation

hK i h K i

De nition We say that L BM is a weak solution of the

incompressible D Euler equations with initial data BM c H

if

a the velocity u K L L and

b

for any test function C we have

c

Z

Z t x t dxdt Z Z Z H x y t x t y t dydxdt

Z x x dx

wher e

H x y t r x t r y t x y

jx yj

jx yj

We wil l say that is a nite energy weak solution if it is a weak solution

with u L L

Remark The derivation of the relation above for smooth solutions is

contained in We abuse notation several times in the de nition since

is only a parametrized family of measures and not a function It is the

content of Lemma in that for a given weak solution in the sense above

one can construct a classical weak solution for the velocity formulation of the

Euler equations since the velocity being in L loc implies that the vorticity has

no discrete part

To prove our main existence result we will require a new a priori estimate

for the smooth symmetric solutions of the inviscid problem whose existence

we examined in Proposition

Lemma Let be a smooth compactly supported function which is NMS

Let u u u and be the smooth solution of the inviscid vorticity equation

with initial data If x x is a smooth function with

bounded derivatives up to second order in the closed right half plane then

d

dt Z x

Z

x x t dx u x t x x dx

Z x

u u x x u u x x x x dx

Proof We recall that since has compact support and total mass zero it

follows that u t L and that juj t O jxj as jxj With

this we have

d

dt Z

Z

x x t dx dx

Z Z

t dx dx Z div u dx dx

Z

Z Z

Z fjxj R x g u

u r dx dx lim

R

b ndS

Z

Z

u

r dx dx

since the boundary terms vanish due to t having compact support to

gether with the fact that u n u on the boundary x which also

b

vanishes We re write the term u r as u r div u and integrate by

parts to obtain

Z r u r u dx dx lim

Z

R

Z fjxj R x g u r u

ndS

b

Z Z

r u r u dx dx Z fx g u r u dS

Z Z

r u r u dx dx Z u x t x x dx

Z

r juj

Z

r u u

x x u u x x x x dx dx

Z u x t x x dx lim

R

Z fjxj R x g

juj

r

ndS

b

Z

Z

u u x x u u x x x x dx dx

Z

Z

Z u x t x x dx

u u x x u u x x x x dx dx

u x t x x dx

Z

as we wished

dx u x t x x dx Z as we wished R emark This lemma is

R emark This lemma is inspired on an a priori estimate derived by D

Chae and O Y Imanuvilov in for D axisymmetric inviscid ow

Lemma used with arctan x integrated in time yields for any

L the following a priori estimate on u x t

Z Z

L

T

L

ju x t j dx dt C

with C depending on k k L ku k L T and L We will use this a priori

estimate to show that the total mass of vorticity in a disk around a point on

the interface fx g decays as the disk shrinks to a point More precisely

we have

Lemma L et u be the smooth solution of in Lemma Set

x a If L and are such that a a L L

then

Z B x j y t jdy C p Z L ju x t j dx

L

where C is a universal constant

Proof Let us begin by noting that by the Biot Savart law the tangential

component of velocity on fx g under the symmetry considered is

u x t Z

Z

y

x y y t dy dy

y

and hence is nonpositive Therefore we have

Z

a

a

ju x t jdx Z

a Z

a

Z

y

x y y t dy dy dx

y

Z

Z

y t y

Z

a

a

Z

Z

 

y

x y dx dy dy

y t g y

y dy

 

where g y y arctan a y arctan a y

y

y

Z fjy x j y g y t g y y

dy

since in fy g

Next consider for any xed h the function f h z arctan z

h arctan z h with jz j h It is easy to check that in this range

f h z arctan h Now if y y y fjy x j y g then clearly

j a y y j and y If we set h y and z a y y then

y

we have g y y f h z arctan h arctan We have hence

Z

a

a ju x t jdx

C

Z B x j y t jdy

arctan

Z fjy x j y g y t dy

On the other hand using the Cauchy Schwartz inequality we obtain

Z

a

a

ju x t jdx C p Z ju x t j dx

a

a

which together with gives what we wish as long as a a L L

gives what we wish as long as a a L L Observe that the result in

Observe that the result in Lemma concerns only the Biot Savart law

Convergence theorem

The objective of this section is to prove the existence of an NMS weak solution

of the incompressible D Euler equations in the sense of De nition with

NMS initial vorticity We begin with the construction of an approximate

solution sequence by mollifying the initial vorticity and exactly solving the

Euler equations

Let r C be nonnegative monotonic decreasing inside

c

its support with total integral and x the Friedrichs molli er x

jxj

Let BM c H be NMS Consider the sequence of smooth

compactly supported functions f g obtained by convolving with n

n

n x n nx Let u n K and let u n n be the smooth solution

n

of with initial data

n

In Section C DiPerna and Majda proved that the sequences fu n g

and f n g are an approximate solution sequence in the sense of De nition

in Since is NMS the total mass of is automatically zero As

n

DiPerna and Majda observe is uniformly bounded in L and since

n

has total mass zero u n is uniformly bounded in L Due to our choice

n

of monotonic circularly symmetric molli ers the are NMS Indeed the

n

mirror symmetry is an obvious consequence of the circular symmetry of

whereas the sign condition follows from straightforward pointwise estimates

on n using both the symmetry and the monotonicity of Therefore

by Proposition the n t are NMS for all time

It was shown in that the following estimates hold for any T

E sup t T k n t k L C

E sup t T ku n t k L C

E There exists M such that fu n g is uniformly bounded in

M

Lip T H

loc

We x the approximate solution sequence fu n g f n g throughout this

section

The key issue in the proof of existence as formulated by Schochet in

is the possibility of concentrations in the sequence of vorticities In order to

control the occurrence of concentrations in the sequence f n g we will put

together the a priori estimate derived in the previous section with a version

of

the log decay of circulation in small circles rst observed by A Majda

in

The a priori logarithmic decay in circulation turns out to be a lo cal

feature of ows with distinguished sign vorticity which was pointed out by

Schochet in this locality is crucial to our analysis

Lemma For every T and K compact there exists a constant

C such that for every

Z T

sup Z B x j n y t jdy dt C j log j

x

K

Proof Let R be such that K B R We recall an estimate due to

S Schochet see Theorem estimate of

Z x y n y t dy Cku n t k L j log j

where was de ned as

z

if jz j

log p jzj

log p

if jz j p

if jz j p

It can be easily seen that if x b c with jbj p then B x p

fx g fx g and therefore since n t is of a distinguished sign in

this disk we get

Z B x j n y t jdy C j log j

b y estimate E

Let a and L be such that a a L L It

follows from Lemma and Lemma that

Z Z

L

T

L

ju n x t j dx dt C k k L T L C

n

k

L ku n

Z B a j n y t jdy C p Z x t j dx

L

L

ju n

for any T using E and E in

Now let x b c with jbj p Then

B x B c jbj and

jcj R Therefore using with a c and jbj p since

we have

Z B x j n y t jdy Z B c jbj j n y t jdy

C p Z x t j dx

R

ju

n

R

since c jbj c jbj R R

Therefore

sup Z B x j n y t jdy max n C j log j C p ku n

x

K

t k L R R o

C j log j ku n

t k L R R

We integrate in time over T and use to conclude the proof

in time over T and use to conclude the proof We recognize the result above as

We recognize the result above as describing the absence of concentra

tions in a time averaged sense If instead of the integral in time we had

the same estimate pointwise almost everywhere in time then our main ex

istence result would follow from Lemma in However the integral

estimate in Lemma does not imply the a e in time pointwise boundedness

of sup x K R B x j n y t jdy or in simpler terms one cannot from a sequence

bounded in L T extract a subsequence which is a e pointwise bounded

in T see