• Aucun résultat trouvé

boundary for compressible and incompressible fluids On the local existence for the Euler equations with free C.R. Acad. Sci. Paris, Ser.I

N/A
N/A
Protected

Academic year: 2023

Partager "boundary for compressible and incompressible fluids On the local existence for the Euler equations with free C.R. Acad. Sci. Paris, Ser.I"

Copied!
6
0
0

Texte intégral

(1)

Contents lists available atScienceDirect

C. R. Acad. Sci. Paris, Ser. I

www.sciencedirect.com

Partial differential equations

On the local existence for the Euler equations with free boundary for compressible and incompressible fluids

Sur l’existence locale de solutions des équations d’Euler pour les fluides compressibles et incompressibles, avec frontière libre

Marcelo M. Disconzi

a

, Igor Kukavica

b

aDepartmentofMathematics,VanderbiltUniversity,Nashville,TN37240,USA

bDept.ofMathematics,UniversityofSouthernCalifornia,LosAngeles,CA90089-2532,USA

a rt i c l e i n f o a b s t ra c t

Articlehistory:

Received31July2017 Accepted5February2018 Availableonline21February2018 PresentedbytheEditorialBoard

We consider the free boundary compressible and incompressible Euler equations with surfacetension.Inbothcases,weprovidea prioriestimatesforthelocalexistencewiththe initialvelocityinH3,withthe H3conditiononthedensityinthecompressiblecase.An additionalconditionisrequiredonthefreeboundary.Comparedtotheexistingliterature, bothresults lower the regularityof initial data for the Lagrangian Euler equationwith surfacetension.

©2018Académiedessciences.PublishedbyElsevierMassonSAS.Allrightsreserved.

ré s u m é

Nous considéronsles équations d’Euler compressibles et incompressibles avec frontière libreettension desurface.Danslesdeux cas, nousfournissonsdes estimationsapriori pourl’existencedesolutionslocalesavecvitesseinitialedans H3 etlacondition H3sur ladensité dansle cascompressible.Unecondition supplémentaireest nécessaire surla frontièrelibre.Parcomparaisonaveclalittérature,lesdeuxrésultatsabaissentlarégularité desdonnéesinitialespourleséquationsd’Eulerencoordonnéeslagrangiennes,avectension desurface.

©2018Académiedessciences.PublishedbyElsevierMassonSAS.Allrightsreserved.

1. Introduction

In this note, we address the waterwave problem, which hasbeen studied extensively. Our settingis rather general:

we consider the Euler equations witha free surface and allow initial data to have nonzero curl, i.e. the initial data is rotational.Thedomainisassumedtobeoffinitedepth,buttheresultscanbeeasilyadaptedtotheinfinitedepthaswell.

E-mailaddresses:marcelo.disconzi@vanderbilt.edu(M.M. Disconzi),kukavica@usc.edu(I. Kukavica).

https://doi.org/10.1016/j.crma.2018.02.002

1631-073X/©2018Académiedessciences.PublishedbyElsevierMassonSAS.Allrightsreserved.

(2)

We consider both thecompressible andincompressiblecases; forthe compressibleEuler equations,we assume that the densityisboundedfrombelow,i.e.weconsiderthecaseofaliquid.

Ouraim inthis note isto announce tworecent resultson the localexistence withnon-zero surface tension.In both maintheorems(thecompressibleandincompressiblecases,respectively),weassumethattheinitialdataisofH3regularity intheinterior.Thislowerstheregularity ofexistingresultsforthecompressibleequations.Intheincompressiblesetting, ourresultlowerstheknownregularityinLagrangian coordinates,albeititdoesnotimproveoverwhathasbeenobtained inEuleriancoordinates.

The historyof the Eulerequations withfree interface is rich—we refer the readerto [2,4,5] fora more complete ac- count. We only mentiona few important worksdealing withnon-zero surface tension.The first general well-posedness resultfortheincompressiblefree-boundaryEulerequationswithsurface tensionis[2],followedby[9,10] and[3].Earlier, well-posednesshadbeenestablishedundertheassumptionthattheinitial vorticityvanishesontheboundary[8].Forthe compressibleequationswithsurfacetension,thefirstresultwas[1].

Inthefirstpartofthenote,weaddressthecompressibleEulerequations,whileinthesecondpartwetreattheincom- pressibleversion.

2. Compressiblecase

IntheLagrangiansetting,thefree-surfacecompressibleEulerequationsread

R

tvα

+

aμα

μq

=

0 in

[

0

,

T

) × ,

(1a)

tR

+

Raμα

μvα

=

0 in

[

0

,

T

) × ,

(1b)

taαβ

+

aαγ

μvγaμβ

=

0 in

[

0

,

T

) × ,

(1c)

q

=

q

(

R

)

in

[

0

,

T

) × ,

(1d)

aμαNμq

+ σ |

aTN

|

g

η

α

=

0 on

[

0

,

T

) ×

1

,

(1e)

vμNμ

=

0 on

[

0

,

T

) ×

0

,

(1f)

η (

0

,·) =

id

,

R

(

0

,·) =

0

,

v

(

0

, ·) =

v0

.

(1g) Above, v, R,andq denote theLagrangian velocity, density,andthe pressure,respectively; N is theunit outer normalto

,aistheinverseof∇

η

,

σ

=constant>0 isthecoefficientofsurfacetension,andg istheLaplacianofthemetricgi j inducedon∂(t)bytheembedding

η

,i.e.gi j=i

η

·j

η

=i

η

μj

η

μ.

Weconsiderthedomain0=T2×(0,1).Wenotethatusingthestraighteningmapin[7,Remark 4.2],itiseasyto modifytheapproachto considerageneralcurveddomain=R2×(0,h(x1,x2)).Applyingthechangeofvariablein[7], weget

R

tvα

+

aμαbβμ

βq

=

0

insteadof(1a);herebisthecofactorinversematrixofthestraighteningmap.Theother equationsinthesystem(1a)–(1g) aremodifiedsimilarly.Themethodsoutlinedheretheneasilycarryoverforthenewsystemaswell,providedisatleast H5regular.

Denotingthecoordinateson by(x1,x2,x3),wehave1=T2× {x3=1}asthefreeboundaryand0=T2× {x3=0} asthestationaryone.Onthepressurefunction,weassume

q

(

R

)

R

Aq

=

constant

>

0

,

(2)

whichissatisfiedbyalargeclassofequationsofstate.

We denoteby thecanonical projection,on

η

(1),fromthetangent bundle of

η

() toits normalbundle, whichis givenby αβ=δαβgklk

η

αl

η

β.Werecall thatinitial datafor(2) isrequiredto satisfycompatibilityconditions(cf. [5]).

Thefollowingisthemainresultinthecompressiblecase.

Theorem2.1.Letv0beasmoothvectorfieldonand 0asmoothpositivefunctiononboundedawayfromzerofrombelow.

Assumethatv0and 0satisfythecompatibilityconditions.Letq:(0,)(0,)beasmoothfunctionsatisfying(2)inaneigh- borhoodof 0.Then,thereexista T>0andaconstant C,dependingonlyon

σ

>0, v0 3, v0 3,1,

0 3,

0 3,1,and (divv0)11,1 suchthatanysmoothsolution(v,R)to(1),withinitialcondition(v0,

0)anddefinedonthetimeinterval [0,T),satisfies

N

(

t

) =

v 23

+ ∂

tv 22

+ ∂

t2v 21

+ ∂

t3v 20

+

R 23

+ ∂

tR 22

+ ∂

t2R 21

+ ∂

t3R 20

+ ∂∂

t2v 20,1

+ ∂

2

tv 20,

1

C

,

(3)

where∂standsforthetangentialderivative.

(3)

Next,wedescribe thestrategy oftheproof anddiscussthetreatmentofdifficultterms.Themainpartisbasedonthe energyestimateforthreetimederivatives,

1 2

d dt

R

(

0

)∂

t3vβ

t3

+

1 2

d dt

R

(

0

)

R q

¯

(

R

)(∂

t3R

)

2

+

1

t3

(

Jaαβq

)∂

t3vβNα

= −

R

(

0

)

R

t3

(

Raαβ

αvβ

)

Raαβ

t3

αvβ

t3

q

R

+

R

(

0

)

t3

aαβq R

aαβ

t3

q

R

t3

αvβ

3

R

(

0

)

q

¯

(

R

)

R

t4R

t2R

tR

R

(

0

)

q

¯

(

R

)

R

t4R

(∂

tR

)

3

+

1 2

R

(

0

)∂

t

q

¯

(

R

)

R

(∂

t3R

)

2

(4)

where

¯

q

(

R

) =

q

(

R

)

R

and J=det(∇

η

). The first two terms on the left provide a coercive term ∂t3v 20+ ∂t3R 20. In the third term I1 =

1t3(J aαβq)∂t3vβ ontheleftside of(4),weuse(1e) forthepressure,leadingto I1=t

0

1t3(

gg

η

α)∂t3,where weset

σ

=1 forsimplicity.Thismayberewrittenas

I1

= −

t

0

1

g gi j

αμ

j

t2vμ

i

t3vα

t

0

1

i

(

g gi j

αμ

)∂

j

t2vμ

t3vα

= −

1 2

1

g gi j

αμ

j

t2vμ

i

t2vα

+

1 2

t

0

1

t

g gi j

αμ

j

t2vμ

i

t2vα

t

0

1

i

(

g gi j

αμ

)∂

j

t2vμ

t3vα

+

1 2

1

g gi j

αμ

j

t2vμ

i

t2vα

|

0

=

I11

+

I12

+

I13

+

I14

.

(5)

Theintegral I11leadstoacoerciveterm 14 ∂∂t2v 20,

1.Thehighest-ordertermresultswhenifallson√

g gi j,whichgives theintegral

I131

= −

t

0

1

i

(

g gi j

nαn

ˆ

μ

j

t2vμ

t3vα

,

forwhichweneedtoexploreitsspecialstructure.Notethatweusedtheidentityαμ= ˆnαnˆμ,nˆ beingtheunitouternormal tothemovingboundary.ItturnsoutthatthisintegralcancelsapartoftheintegralresultingfromI14(cf. [5] forcomplete details).

Itremains todiscussthetreatmentoftherightsideof(4).Denotethetermsontherightsideof(4) byJ1J5.Allthe integralsresultingfromJ1,J3,J4,andJ5 canbeestimatedusingintegrationby partsintime andspaceandemploying theHölderandSobolevinequalities.Ontheotherhand,whenexpandingJ2,thereisatrickytermT=t

0

t3Aμα∂t3μvαq where A= J a,whichcannotbetreatedinthisway,andspecialcancellationsareneededtocontrolit.Wenamelyclaim

|

T

|

N

+

P0

+

P

t

0

P

,

(6)

wherePdenotesagenericpolynomialinN,P0agenericpolynomialofthenormsoftheinitialdata.

(4)

Herewesketchthemainideasusedinproving(6).UsingtheformulaforthecofactormatrixA= J a,werewrite

T

=

t

0

q

αλτ

2

t2vλ

3

η

τ

1

t3vα

+

t

0

q

αλτ

2

η

λ

3

t2vτ

1

t3vα

t

0

q

αλτ

1

t2vλ

3

η

τ

2

t3vα

t

0

q

αλτ

1

η

λ

3

t2vτ

2

t3vα

+

t

0

q

αλτ

1

t2vλ

2

η

τ

3

t3vα

+

t

0

q

αλτ

1

η

λ

2

t2vτ

3

t3vα

+

L

=

T1

+ · · · +

T6

+

L

(7)

where A1α =

αλτ2

η

λ∂3

η

τ, A2α= −

αλτ1

η

λ∂3

η

τ, A3α=

αλτ1

η

λ∂2

η

τ, and L stands for the lower-order terms. Also,

ε

αβγ isthetotallyanti-symmetricsymbolwith

ε

123=1.Toexplorethecancellation,wegroupthetermsin(7) asT1+T3, T2+T5,andT4+T6.Allthreepairsaretreatedbyintegratingbypartsintime.WhenwritingoutT1+T3,thehighest-order termsare

t

0

q

αλτ

2

t2vλ

3

η

τ

1

t3vα

+

t

0

q

αλτ

1

t3vλ

3

η

τ

2

t2vα

andtheleadingtermscancelbyrelabelingtheindices

α

λ inthesecond integral.Thesamecancellationoccursalsoin theothergroupingsT2+T5andT4+T6.However,thereweintegratethederivatives2and3;thelastsumproducesthe boundaryterm

1q∂t2v22t2v3.Toboundthisterm,weusetheprojection identityαβ=δαβgklk

η

αl

η

β,whichallows ustorelate∂∂t2vand∂∂t2v3andwrite

1

q

t2v2

2

t2v3

=

1

q

t2v2

(

3λ

2

t2vλ

+

gkl

k

η

3

l

η

λ

2

t2vλ

).

Thelastexpressioncanthenbe controlledusingthecoercivetermsandthetrace inequality.Collectingalltheinequalities leadsto

t3v 20

+

t3R 20

+ ∂∂

t2v 20,1

N

+

P0

+

P

t

0

P

.

Theenergyestimatesfortwotimederivativesandonetimederivativearesimilar.Forinstance,usinganalogousmethods, weobtain

∂∂

t2v 20

+ ∂∂

t2R 20

+ ∂

2

tv 20,

1

N

+

P0

+

P

t

0

P (8)

and

2

tv 20

+

2

tR 20

+

3v 20,

1

N

+

P0

+

P

t

0

P

.

(9)

We emphasize however, that unlike [1], we cannot continue and perform the same estimate for 3v 20+ 3R 20. The reasonisthattheanalogoftheterm(7) cannotbetreatedthesamewayduetolackoftimederivatives.

Instead,weobtainthefullcontrolofN byusinganewCauchyinvariancepropertyforthecompressibleEulerequations, statednext.NotethattheCauchyinvarianceprovidesa3Danalogofthe2Dvorticitypreservation.

Theorem2.2.(CompressibleCauchyInvarianceFormula)Let(v,R)beasmoothsolutionto(1)definedon[0,T).Then

ε

αβγ

βvμ

γ

η

μ

= ω

α0

+

t

0

ε

αβγaλμ

λq

γ

η

μ

βR

R2

,

(10)

for0≤t<T ,where

ω

0isthevorticityattimezero.

(5)

Here,weprovideasketchoftheproof(cf. [5] forcompletedetails).First,wehave

t

( ε

αβγ

βvμ

γ

η

μ

) = ε

αβγ

βvμ

γvμ

+ ε

αβγ

β

tvμ

γ

η

μ

= −

1

R

ε

αβγ

β

(

aλμ

λq

)∂

γ

η

μ

+

1

R2

ε

αβγaλμ

λq

βR

γ

η

μ

,

by theanti-symmetryof

ε

αβγ and(1a).Sincea=(∇

η

)1,weget∂β(aλμ∂γ

η

μ)=∂βaλμ∂γ

η

μ+aλμ∂γ∂β

η

μ=0,andafter somesimplealgebra,t(

ε

αβγ∂βvμ∂γ

η

μ)= R12

ε

αβγaλμ∂λq∂βR∂γ

η

μ.Theformula(10) thenfollowsbyintegratingintime.

TherestoftheproofofTheorem2.1maynowbeperformedsimilarlyto [6].Namely,controlofthecurlisprovidedby theCauchyinvarianceformula(10).Thenthedivergenceofvanditstimederivativesareobtainedfrom(1b) inaninductive fashion, usingthe factthatthethird timederivativesareindependentlyestimatedfrom(4).Finally,theboundaryintegral isestimatedusing(5),(8),and(9).Theexception isthemissingestimate forv3,whichisobtainedfromtheequation(1e) (cf. [5] fordetails).

Collectingalltheestimatesleadsto

N

(

t

)

C0P

(

N

(

0

)) +

P

(

N

(

t

))

t

0

P

(

N

(

s

))

ds

,

where P isapolynomial,andtherestfollowsbyastandardGronwallargument.

3. Incompressiblecase

In this section, we describe the a priori estimates forthe local existence forthe incompressible Eulerequations. We addresstheequationsintheLagrangiancoordinates,wheretheytaketheform

tvα

+

aμα

μq

=

0 in

[

0

,

T

) × ,

(11a)

aαβ

αvβ

=

0 in

[

0

,

T

) × ,

(11b)

taαβ

+

aαγ

μvγaμβ

=

0 in

[

0

,

T

) × ,

(11c)

aμαNμq

+ σ |

aTN

|

g

η

α

=

0 on

[

0

,

T

) ×

1

,

(11d)

vμNμ

=

0 on

[

0

,

T

) ×

0

,

(11e)

η (

0

, · ) =

id

,

v

(

0

, · ) =

v0

.

(11f)

All thequantitiesare asinthecompressiblecase, exceptforq,whichisnow aLagrange multiplier enforcingtheincom- pressibilityconstraint.Thenwehavethefollowinga prioriestimatessupportingthelocalexistencewiththevelocity H3 in theinterior,withaconditiononthetraceatthefreeboundary.

Theorem3.1.Assumethat

σ

>0in(11).Letv0beasmoothdivergence-freevectorfieldon.ThenthereexistT>0andaconstant C0,dependingonlyon v0 3, v0 4,1,and

σ

>0,suchthatanysmoothsolution(v,q)to(11)withinitialconditionv0anddefined onthetimeinterval[0,T),satisfies

v 3

+

tv 2.5

+

t2v 1.5

+

t3v 0

+

q 3

+

tq 2

+

t2q 1

C0

.

(12)

Observethat theSobolevexponentsin(3) and(12) differ.Theproof ofTheorem3.1isdifferentthan theoneofTheo- rem2.1;however,themainstepsalsorelyontheenergyinequalityforthethirdtimederivativeofthevelocity

1

2

t3v 20

=

1

2

t3v

(

0

)

20

t

0

t3

μ

(

aμαq

)∂

t3vα

.

Themaindifferencewiththecompressiblecaseareinthetreatmentofthepressureandinboundingtheboundaryintegral

t

0

1

g

(

gi jgkl

gljgik

)∂

j

η

α

k

η

λ

t2

lvλ

t3

ivα

.

(13)

Forthepressure,themaininequalitiesaresummarizedinthefollowingstatement.

(6)

Lemma3.2.Wehavetheestimates

q 3

C

v 2 v 2

+

C

tv 1.5,1

+

C

,

tq 2

C

v 1.5

(

q 2.5

+ ∂

tv 1.5

) +

C

( ∇

v 1.5

v L

+ ∇∂

tv 1.5

)

v 1.5

+

C

t2v 1,1

+

C v 2.5

+

C

tv 2.5

,

t2q 1

C

(

v 1.5

v L

+ ∂

tv 1.5

)(

q 2

+ ∂

tv 1

) +

C

v L

( ∂

tq 1

+ ∂

t2v 0

) +

C

(

v 1.5

v 2L

+ ∂

tv 1.5

v L

+ ∂

t2v 1.5

)

v 1

+

C

t3v 0

+

C

( ∂

tv 2.5

+

v 3

tq 1

) +

C

(

1

+

v 23

+ ∂

tv 2

)( ∂

tv 2

+

v 23

),

fort(0,T)andδ >0asmallnumber.

Thelemmaisobtainedbysolvingtheellipticboundaryvalueproblemsforq,∂tq,andt2q.Forq andqt,weconsiderthe Neumannproblem,whileforthesecondderivative,weusetheDirichletproblem,togetherwithanestimate

t2q 0,1

t2q 1

+

C

( ∂

tv 2.5

+

v 3

tq 1

) +

C

(

1

+

v 23

+ ∂

tv 2

)( ∂

tv 2

+

v 23

)

(cf. [4] for details). The treatment for the boundary integral (13) is involved and it relies on its determinant structure (cf. [2]). The restof the proof is obtained by the H1 estimate for t2v. However, in order to control v 3, we use the additionalboundaryregularityprovidedbythesurfacetension,togetherwiththediv-curlestimatesprovidedbytheCauchy invariance[7].

Acknowledgements

We thank the referee for corrections and useful suggestions. MMD was partially supported by the NSF grant DMS-1305705,whileIKwaspartiallysupportedbytheNSFgrantDMS-1615239.

References

[1]D.Coutand,J.Hole,S.Shkoller,Well-posednessofthefree-boundarycompressible3-DEulerequationswithsurfacetensionandthe zerosurface tensionlimit,SIAMJ.Math.Anal.45 (6)(2013)3690–3767.

[2]D.Coutand,S.Shkoller,Well-posednessofthefree-surfaceincompressibleEulerequationswithorwithoutsurfacetension,J.Amer.Math.Soc.20 (3) (2007)829–930.

[3]M.M.Disconzi,D.G.Ebin,ThefreeboundaryEulerequationswithlargesurfacetension,J.Differ.Equ.261 (2)(2016)821–889.

[4] M.Disconzi,I.Kukavica,Aprioriestimatesforthefree-boundaryEulerequationswithsurfacetensioninthreedimensions,preprint,2017.

[5] M.Disconzi,I.Kukavica,Aprioriestimatesforthe3dcompressiblefree-boundaryEulerequationswithsurfacetensioninthecaseofaliquid,preprint, 2017.

[6]M.Ignatova,I.Kukavica,Onthelocalexistenceofthefree-surfaceEulerequationwithsurfacetension,Asymptot.Anal.100(2016)63–86.

[7] I.Kukavica,A.Tuffaha,V.Vicol,Onthelocalexistenceforthe3dEulerequationwithafreeinterface,Appl.Math.Optim.76 (3)(2017)535–563, https://doi.org/10.1007/s00245-016-9360-6.

[8]B.Schweizer,Onthethree-dimensionalEulerequationswithafreeboundarysubjecttosurfacetension,Ann.Inst.HenriPoincaré,Anal.NonLinéaire 22 (6)(2005)753–781.

[9]J. Shatah,C.Zeng,GeometryandaprioriestimatesforfreeboundaryproblemsoftheEulerequation,Commun.PureAppl.Math.61 (5)(2008) 698–744.

[10]J.Shatah,C.Zeng,Localwell-posednessforfluidinterfaceproblems,Arch.Ration.Mech.Anal.199 (2)(2011)653–705.

Références

Documents relatifs

1. Paley-Zygmund theorems on Riemannian compact manifolds. An application to the cubic wave equation with random initial data. Paley-Zygmund theorems for the superquadratic

The internal energy balance features a corrective source term which is needed for the scheme to compute the correct shock solutions: we are indeed able to prove a Lax consistency

We finally obtain a class of schemes which offer many interesting properties: both the density and internal energy positivity are preserved, unconditionnally for the pressure

(2010) [2] we show that, in the case of periodic boundary conditions and for arbitrary space dimension d 2, there exist infinitely many global weak solutions to the

In order to find a priori estimates in spaces of higher regularity (as they are needed for the treatment of the free boundary), it is standard to try the following: One

After having found an optimal algorithm which converges in two steps for the third order model problem, we focus on the Euler system by translating this algorithm into an algorithm

us naturally to the notion of boundary-coupled weak solution. We have defined boundary-coupled weak solution using the vorticity formulation of the Euler equations. However, in order

In this article, we show a local exact boundary controllability result for the 1d isentropic compressible Navier Stokes equations around a smooth target trajectory.. Our