12.07.2015 Views

Lecture Notes on Fluid Dynamics

Lecture Notes on Fluid Dynamics

Lecture Notes on Fluid Dynamics

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

<str<strong>on</strong>g>Lecture</str<strong>on</strong>g> <str<strong>on</strong>g>Notes</str<strong>on</strong>g> <strong>on</strong> <strong>Fluid</strong> <strong>Dynamics</strong>Alberto Bressan and Carlotta D<strong>on</strong>adello26th October 20051 Review of differential operatorsGiven a scalar functi<strong>on</strong> φ : R N ↦→ R, its gradient ∇φ is the vector of partial derivatives∇φ = . ( ∂ ∂∂)(φ x1 , φ x2 , . . . , φ xN ) = φ , φ , . . . , φ .∂x 1 ∂x 2 ∂x NFor a vector field u : R N ↦→ R N , its divergence isThe Laplace operator is defined asdiv u . =N∑u i x i.i=1∆φ . =N∑i=1∂ 2 φ∂x 2 i= div ∇φ .Ňow c<strong>on</strong>sider a velocity field u : R × R N ↦→ R N , so that u(t, x) = (u 1 , . . . , u N )(t, x) denotesthe velocity of the particle located at the point x = (x 1 , . . . , x N ) at time t. We then define theMaterial Derivative of a functi<strong>on</strong> φ = φ(t, x) asD t φ . = ∂ ∂t φ + N ∑i=1u i∂φ .∂x iThis provides the time-derivative of φ al<strong>on</strong>g particle trajectories. ˇWe recall that the wedgeproduct of two vectors v, w ∈ R 3 can be obtained as follows. Let e 1 , e 2 , e 3 be the vectors in thestandard orth<strong>on</strong>ormal basis of R 3 . Then⎛⎞e 1 e 2 e 3v ∧ w = det ⎝ v 1 v 2 v 3⎠ .w 1 w 2 w 3Given a velocity field u : R 3 ↦→ R 3 , its vorticity ω = curl u is defined asNote that, formally, <strong>on</strong>e can writeω = curl u . = ( ∂ 2 u 3 − ∂ 3 u 2 , ∂ 3 u 1 − ∂ 1 u 3 , ∂ 1 u 2 − ∂ 2 u 1) .⎛curl u = ∇ ∧ u = det ⎝⎞e 1 e 2 e 3∂ 1 ∂ 2 ∂ 3⎠ .u 1 u 2 u 31


2 Euler and Navier-Stokes equati<strong>on</strong>sfor a homogeneous incompressible fluidThe flow of a homogeneous incompressible, n<strong>on</strong>-viscous fluid in R N is modelled by the systemof Euler equati<strong>on</strong>s⎧⎨ D t u(t, x) = −∇p(t, x), balance of momentumdiv u(t, x) ≡ 0,incompressibility c<strong>on</strong>diti<strong>on</strong>(1)⎩u(0, x) = u 0 (x),initial dataWherelatex fluid.text is time,x ∈ R N is the Eulerian space variable,u = (u 1 , . . . , u N ) is the fluid velocity,p is the scalar pressure,D t = ∂ t + u · ∇ x is the material derivative.When viscosity is present, an additi<strong>on</strong>al diffusi<strong>on</strong> term is present in the momentum equati<strong>on</strong>.This leads to the Navier-Stokes equati<strong>on</strong>s:⎧⎨ D t u(t, x) = −∇p(t, x) + ν∆u, balance of momentumdiv u(t, x) ≡ 0,incompressibility c<strong>on</strong>diti<strong>on</strong>(2)⎩u(0, x) = u 0 (x),initial dataThe positive c<strong>on</strong>stant ν is the coefficient of kinematic viscosity. Its reciprocal is called theReynolds number: Re = 1/ν.Ǐf the fluid moti<strong>on</strong> takes place in a bounded set Ω ⊂ R N , <strong>on</strong>e must add suitable boundaryc<strong>on</strong>diti<strong>on</strong>s. In the case of the inviscid Euler equati<strong>on</strong>s, <strong>on</strong>e requires that the velocity u of thefluid is tangential to the boundary ∂Ω. Calling n the unit outer normal, we thus haven · u = 0 x ∈ ∂Ω . (3)In the presence of viscosity, instead of (3), the Navier-Stokes equati<strong>on</strong>s are supplemented bythe n<strong>on</strong>-slip boundary c<strong>on</strong>diti<strong>on</strong>su = 0 x ∈ ∂Ω . (4)2.1 Derivati<strong>on</strong> of Euler equati<strong>on</strong> from the balance law of momentumC<strong>on</strong>sider any regi<strong>on</strong> W 0 ⊂ R n . Denote by W t the regi<strong>on</strong> occupied at time t > 0 by those fluidparticles which are initially in W 0 . Assume that the <strong>on</strong>ly force acting <strong>on</strong> the fluid through theboundary ∂W t is the pressure. Then the balance law of momentum takes the formddt∫W tu dV = [total forces across the boundary]∫= − p n dA.∂W tIf e is any fixed vector in R N by the divergence theorem we get∫∫p e · n dA = div(p e) dV∂W t W∫ t= ∇p · e dV.W t2(5)(6)


Then:∫ ∫∫du dV = D t u dV = − ∇p dV, (7)dt W t W t W tfor any regi<strong>on</strong> W t in the fluid at time t. Hence the equality holds also in the differential form:D t u = −∇pEuler equati<strong>on</strong>.2.2 Symmetry groups for the Euler and Navier-Stokes equati<strong>on</strong>sAssume that u = u(t, x) and p = p(t, x) provide a soluti<strong>on</strong> to Euler or Navier-Stokes equati<strong>on</strong>.Further soluti<strong>on</strong>s can then be obtained by various variable transformati<strong>on</strong>s.• Translati<strong>on</strong> Invariance: For any c<strong>on</strong>stant vector c in R N{ u c (t, x) := u(t, x − ct) + c,p c (t, x) := p(t, x − ct),is another soluti<strong>on</strong>.• Rotati<strong>on</strong> Invariance: For any orthog<strong>on</strong>al matrix Q{ u Q (t, x) := Q T u(t, Qx),p Q (t, x) := p(t, Qx),is another soluti<strong>on</strong>.• Scale Invariance: If u, p provide a soluti<strong>on</strong> to the Euler equati<strong>on</strong>s, then for any λ, τ > 0the functi<strong>on</strong>s⎧⎨ u λ, τ (t, x) := λ τ u( t τ , x λ ),(A)⎩p λ, τ (t, x) := λ2τp( t 2 τ , x λ ),provide a 2-parameters family of soluti<strong>on</strong>s to the Euler equati<strong>on</strong>.If u, p are a soluti<strong>on</strong> to the Navier-Stokes equati<strong>on</strong>s, then for all τ > 0 the functi<strong>on</strong>s⎧⎨ u τ (t, x) := τ −1/2 u( t τ , x),τ 1/2(B)⎩p τ (t, x) := τ −1 p( t τ ,xτ 1/2 ).provide a 1-parameter family of soluti<strong>on</strong>s to the Navier-Stokes equati<strong>on</strong> for τ ∈ R. Observethat (A) coincides with (B) when λ = τ 1/23 Particle trajectoriesGiven a fluid flow with velocity field u (not necessarily incompressible), the particle trajectorymappingX : [0, ∞) × R N −→ R N(8)t , α ↦→ X(t, α)describes the trajectory of the particle which is initially located at point α = (α 1 , . . . , α N ) attime t = 0. The Lagrangian variable α, can be regarded as a particle marker. The functi<strong>on</strong>X(t, α) is determined by solving the Cauchy problem⎧⎨⎩∂∂tX(t, α) = u(t, X(t, α)),X(0, α) = α.3


An initial domain W 0 ⊂ R N in the fluid evolves in time toX(t, W 0 ) = W t = {X(t, α)| α ∈ W 0 }.We shall denote by ∇ α X the Jacobian matrix of first order partial derivatives of X w.r.t. α, sothat(∇ α X(t, α))ij = ∂Xi (t, α) .∂α jThe evoluti<strong>on</strong> of ∇ α X al<strong>on</strong>g a particle trajectory is described by the linear evoluti<strong>on</strong> equati<strong>on</strong>) ()D t(∇ α X(t, α) = ∇ x u(t, X(t, α) · ∇ α X(t, α) .The Jacobian determinant of X(t, ·) is denoted asJ(t, α) = det ( ∇ α X(t, α) ) .Clearly J(0, α) ≡ 1. If u is a smooth velocity field then the time evoluti<strong>on</strong> of J is given by∂[]∂t J(t, α) = div u(t, X(t, α)) J(t, α) .In particular, J(t, X(t, α)) ≡ 1 if the flow is incompressible.Theorem 1 The Trasport FormulaLet W be an open, bounded domain in R N with smooth boundary, and let X be the particletrajectory mapping of a given smooth velocity field u.For any smooth functi<strong>on</strong> f = f(t, x) we have∫ ∫df dx = [f t + div (fu)] dx.dt W t W tProof∫df dx = d ∫f(t, X(t, α))J(t, X(t, α)) dαdt W tdt W∫( )= Dt f J + f J t dα∫W(= ft + u · ∇f + f div u ) J dα∫W= [f t + div(fu)] dx.W tDefiniti<strong>on</strong> 1 A flow X is incompressible if for any subregi<strong>on</strong> W with smooth boundary, andfor any time t > 0, X is volume preserving:Vol ( X(t, W ) ) = Vol ( W ) .Applying the transport formula with f ≡ 1 <strong>on</strong>e obtainsPropositi<strong>on</strong> 1 The following are equivalent:• The fluid flow is incompressible,• div u = 0,• J(t, α) ≡ 1□4


4 VorticityIn the following, we c<strong>on</strong>sider a smooth velocity field u : R 3 ↦→ R 3 . We define the vorticity fieldω = as the curl of the velocity:ω = curl u . = ( ∂ 2 u 3 − ∂ 3 u 2 , ∂ 3 u 1 − ∂ 1 u 3 , ∂ 1 u 2 − ∂ 2 u 1) .4.1 Local behavior of an incompressible flowIn a neighborhood of any point x 0 , we now show that, up to higher order terms w.r.t. |x − x 0 |,a smooth incompressible velocity field u = u(x) can be written in a unique way as the sum ofinfinitesimal translati<strong>on</strong>, rotati<strong>on</strong> and deformati<strong>on</strong> field.Indeed, let x 0 be a given point in R 3 . A first order Taylor expansi<strong>on</strong> of u yieldsu(x 0 + h) = u(x 0 ) + ∇u(x 0 ) · h + O(h 2 ).The Jacobian matrix ∇u, can be written as the sum of its symmetric and antisymmetric parts:∇u =}{{} D + }{{} ΩSymm. Antisymm.WhereD = 1 2Ω = 1 2(∇u + (∇u) t ), hence tr D = 0(∇u − (∇u) t ), hence Ωh = 1 2 ω ∧ h.Observe that, since every symmetric matrix can be diag<strong>on</strong>alized and the trace is invariant underorthog<strong>on</strong>al trasformati<strong>on</strong>s, D can be written as⎛D = ⎝The Taylor expansi<strong>on</strong> thus takes the formγ 1 0 00 γ 2 00 0 −(γ 1 + γ 2 )⎞⎠u(x 0 + h) = u(x 0 ) + 1 } {{ } 2 ω ∧ h + Dh +O(h 2 ).} {{ }} {{ }translati<strong>on</strong>deformati<strong>on</strong>rotati<strong>on</strong> byvector ω4.2 Evoluti<strong>on</strong> equati<strong>on</strong> for the vorticityPropositi<strong>on</strong> 2 Let the velocity field u = u(t, x) provide a soluti<strong>on</strong> to the Navier-Stokes equati<strong>on</strong>s.Then the vorticity ω = curl u evolves according to the equati<strong>on</strong>Proof We use the vector identity:D t ω = (ω · ∇)u + ν∆ω. (9)12 ∇|u|2 = u ∧ curl u + (u · ∇)u5


in order to replace the term (u · ∇)u in the Navier-Stokes equati<strong>on</strong>:∂ t u + 1 2 ∇|u|2 − u ∧ ω = −∇p + ν∆u.We now recall that curl(∇ϕ) = 0 for every ϕ. Taking the curl of both sides of the above identity<strong>on</strong>e obtains∂ t ω − curl (u ∧ ω) = ν∆ω. (10)In order to write esplicitly the term curl (u ∧ ω) we recall the following general vector identity:Since div u ≡ 0, this reduces toTherefore, from (10) we obtaincurl (u ∧ ω) = (ω · ∇)u − ω div u − (u · ∇)ω + u div ω .curl (u ∧ ω) = (ω · ∇)u − (u · ∇)ω .∂ t ω + (u · ∇)ω = (ω · ∇)u + ν∆ω. (11)□5 Vorticity transport formula for the Euler equati<strong>on</strong>The map t ↦→ X(t, α) describes the trajectory of the particle that was in α at t = 0. C<strong>on</strong>siderany vector h 0 ∈ R 3 . At a given time t, the positi<strong>on</strong> of the particle which is initially at α + εhcan be described byX(t, α + εh) = X(t, α) + ε · ∇ α X(t, α) · h + O(ε 2 ) .In the present secti<strong>on</strong> we will show that the inviscid vorticity equati<strong>on</strong>, i.e. the vorticityevoluti<strong>on</strong> equati<strong>on</strong> associated to Euler equati<strong>on</strong>,can be integrated by means of the particle trajectory equati<strong>on</strong>D t ω = (ω · ∇)u, (12)dX(t, α) = u(t, X(t, α)).dtLet h(t, x) be a smooth vector field in R N , and c<strong>on</strong>sider the perturbed flowX(t, α + h) = X(t, α) + ∇ α X(t, α) h +O(h 2 ).} {{ }(⋆)(⋆) is the term which tell us how the perturbati<strong>on</strong> evolves in time. Then the evoluti<strong>on</strong> equati<strong>on</strong>of a first order perturbati<strong>on</strong> takes the formD t h = h · ∇u.The following lemma applies also to the more general case in which the velocity field u is notdivergence free.6


Lemma 1 Let u = u(t, x) be a smooth vector field with associed particle trajectory mappingX(t, α). The smooth vector field h = h(t, x) satisfiesD t h = h · ∇u (13)if and <strong>on</strong>ly ifh(t, X(t, α)) = ∇ α X(t, α)h(0, α). (14)Proof The particle trajectory mapping is the soluti<strong>on</strong> of⎧d⎨ dtX(t, α) = u(t, X(t, α)),⎩X(0, α) = α.By differentiati<strong>on</strong> with respect to α we obtainddt ∇ αX(t, α) = ∇ x u(t, X(t, α)) ∇ α X(t, α),henceddt ∇ αX(t, α) h(0, α) = ∇ x u(t, X(t, α)) ∇ α X(t, α) h(0, α) ,If (14) holds thenddt h(t, X(t, α)) = ∇ xu(t, X(t, α)) h(t, X(t, α))This implies (13). Viceversa, if (13) holds, then for a fixed α the vector h(t, X(t, α)) satisfiesddt h(t, X(t, α)) = ∇ xu(X(t, α)) h(t, X(t, α)).Since t ↦→ h(t, X(t, α)) and t ↦→ ∇ α X(t, α)h(0, α) satisfy the same linear ODE with initial datah(0, α), they coincide for all times t.□Propositi<strong>on</strong> 3 Vorticity Transport FormulaLet X(t, α) be the smooth particle trajectory mapping corresp<strong>on</strong>ding to the smooth divergencefreevector field u(t, x). Thensolves (12).ω ( t, X(t, α) ) = ∇ α X(t, α)ω 0 (α)Proof We observe that the vorticity equati<strong>on</strong> has the same form of the evoluti<strong>on</strong> equati<strong>on</strong>satisfied by a first order perturbati<strong>on</strong>.The proof of Propositi<strong>on</strong> 3 follows from the applicati<strong>on</strong> of the lemma in the case h = ω. □Definiti<strong>on</strong> 2 The smooth curveis a vortex line at time t ifγ = {γ(s) ∈ R N | 0 < s < 1}dγ(s) = λ(s)ω(t, γ(s)), for some λ(s) ≠ 0.ds7


Corollary 1 In a n<strong>on</strong>-viscous fluid flow, the vortex lines move with the fluid.Proof Let γ 0 be a vortex line at time t = 0. As a set of particles γ 0 evolves in time to γ tγ t := {X(t, γ 0 (s)) | 0 < s < 1},and there holdsdds γ t(s) = ∇ α X(t, γ 0 (s)) d ds γ 0(s)= ∇ α X(t, γ 0 (s))λ(s)ω(0, γ 0 (s))= λ(s)ω(t, X(t, γ 0 (s)))= λ(s)ω(t, γ t (s))(15)□Definiti<strong>on</strong> 3 Let u be a smooth divergence-free vector field and let γ be a smooth oriented closedcurve in the fluid. The circulati<strong>on</strong> of u around γ the line integral∮u · dl.γLemma 2 Transport theorem for curvesLet γ t := X(t, γ 0 ) be a closed curve transported by the flow. The following identity holds:∮ ∮du · dl = D t u · dl.dt γ t γ tProof Let s ↦→ γ 0 (s) be a parametrizati<strong>on</strong> of γ 0 , and let s ↦→ X(t, γ 0 (s)) = γ t (s) be theparametrizati<strong>on</strong> of γ t .∮du · dl = d dt γ tdt=∫ 10∫ 10u(t, X(t, γ 0 (s))) ∂ ∂s X(t, γ 0(s)) dsD t u(t, X(t, γ 0 (s))) ∂ ∂s X(t, γ 0(s)) ds∫ 1+ u(t, X(t, γ 0 (s))) ∂ ∂0∂t ∂s X(t, γ 0(s)) ds∮∫ 1∂ |u(t, X(t, γ 0 (s)))| 2= D t u · dl +ds.γ t∂s 2Since γ t is a loop the last term of the right hand side is null and we obtain the equality∮ ∮du · dl = D t u · dl.dt γ t γ t0(16)□Theorem 2 Kelvin’s circulati<strong>on</strong> theoremFor an inviscid, incompressible fluid the circulati<strong>on</strong> of the velocity u around any closed curveγ t moving with the fluid is c<strong>on</strong>stant in time.8


Proof∮du · dl =dt γ tsince the line integral of a gradient <strong>on</strong> a closed loop is zero.∮∮D t u · dl = −γ t∇p dl .γ t(17)□Corollary 2 The vorticity flux through a 2-dimensi<strong>on</strong>al surface moving with the fluid is c<strong>on</strong>stantin time.C<strong>on</strong>sider the 2-dimensi<strong>on</strong>al surface Σ t whose boundary is a closed curve γ t moving with thefluid. By Stokes formula we obtain∮ ∫u · dl = ω · dA,γ t Σ twhere n is a vector, ‖u‖ = 1, normal to Σ tAn example of this behavior is given by tornadoes: the surface of a horiz<strong>on</strong>tal secti<strong>on</strong> decreasesas the distance from the soil, then the angular velocity around the vertical axis has toincrease.6 C<strong>on</strong>served quantities for the Euler equati<strong>on</strong>A c<strong>on</strong>servati<strong>on</strong> law in N space dimensi<strong>on</strong>s takes the form:∂ t u(t, x) + div x F (u(t, x)) = 0 , (18)where t 0 is the time variable, x = (x 1 , . . . , x N ) ∈ R N is the space variable, u = (u 1 , . . . , u n )is a vector in R N and F is a smooth map defined <strong>on</strong> a c<strong>on</strong>vex neighborhood of the origin in R Nwith values in R N . The comp<strong>on</strong>ents of u are called the c<strong>on</strong>seved quantities while the comp<strong>on</strong>entsof F , are the fluxes.Let W ⊂ R N be a closed regi<strong>on</strong> with smooth boundary. Integrating (18) <strong>on</strong> W we obtain:ddt∫W∫u(t, x) dx = u t (t, x) dxW∫= − div F ( u(t, x) ) dxW∫= − F ( u(t, x) ) · n dA∂W= [Flow through the boudary].Observati<strong>on</strong> 1 Notice that∫du(t, x) dx = 0dt R N if |F | = ≀(|x| 1−N ).In fact∫∫ddu(t, x) dx = lim u(t, x) dx = limdt R N R→∞ dt− F (u(t, x)) · n dA,|x|RR→∞∫|x|=Ris zero if |F | = o(|x| 1−N ).9


The following are c<strong>on</strong>served quantities for the Euler equati<strong>on</strong>s.• Comp<strong>on</strong>ents of momentum The Euler equati<strong>on</strong> is equivalent to the momentum balancelaw. Then, if u is a soluti<strong>on</strong> to the Euler equati<strong>on</strong> its comp<strong>on</strong>ents satisfyNu i t + ∑u j u i x j= −p xi ,which can be written (div u=0) as a system of c<strong>on</strong>servati<strong>on</strong> law:j=1u i t + div (ui u + P i ) = 0,i = 1, . . . , Nwhere P i ≡ (0, . . . , 0, p i , 0, . . . , 0).• Vorticity The i-th comp<strong>on</strong>ent of the vorticity equati<strong>on</strong> isωt i + (u∇)ω i = (ω∇)u i .It can be written as a system of c<strong>on</strong>servati<strong>on</strong> laws::ωt i + div (ωi u + u i ω) = 0.• The energy By multipling Euler equati<strong>on</strong> by u we obtain the energy equati<strong>on</strong>( |u|22)t N + ∑u i u j u i x j+i,j=1N∑u i p xi = 0.i=1It can be written as a c<strong>on</strong>servati<strong>on</strong> law( |u|22)t + div ( |u|22 u + p v )= 0.• Helicity The helicity measures the comp<strong>on</strong>ent of velocity in the directi<strong>on</strong> of the vorticity:H . = u · ω.∂∂t H = u t · ω + u · ω t= [−(u · ∇)u − ∇p ] · ω + u · [−(u · ∇)ω + (ω · ∇)u ]N∑N∑N∑N∑= − u j u i x jω i − p xi ω i − u i u j ωx i j+ u i ω j u i x ji,j=1i=1i,j=1i,j=1This can be written asH t + div[(ω · u)u − |u|22 ω + p ω ]= 0.10


7 Leray formulati<strong>on</strong> of Navier-Stokesequati<strong>on</strong>Our aim in this secti<strong>on</strong> will be to eliminate the pressure from the explicit formulati<strong>on</strong> of Navier-Stokes equati<strong>on</strong> in order to obtain a system of closed evoluti<strong>on</strong> equati<strong>on</strong>s for u. Let’s assumethat u satisfies the equati<strong>on</strong>du= −(u · ∇)u + ν∆u − ∇p (19)dtThen u solves (2) if and <strong>on</strong>ly if the functi<strong>on</strong> p has been chosen in such a way that div u(t, x) = 0for all t 0.We define a subspace E ⊂ L 2 (R 3 ; R 3 ) c<strong>on</strong>taining all vector fields whose divergence (indistributi<strong>on</strong>al sense) vanishes identically.E . = { u ∈ L 2 ; div u = 0 } .Assume that t ↦→ u(t) ∈ E is a smooth soluti<strong>on</strong> of (19) Observe thatdiv u = 0 ⇒ div (∆u) = 0,div u = 0 div (u · ∇u) = 0.(20)More precisely, if the i-th comp<strong>on</strong>ent of (u · ∇)u is[(u · ∇)u] i = ∑ ju j u i x j,we can writediv (u · ∇)u = ∑ i,j(ujxiu i x j+ u j u i )x i x j .} {{ }A i,jSince div u = ∑ i ui x i= 0, we get ∑ i,j A i,j = 0. Therefore, in general,div (u · ∇)u = ∑ i,ju j x iu i x j= tr(∇u) 2 ≠ 0.We now need to choose p so that the sumlies in E.Sincedudt= −(u · ∇)u + ν∆u} {{ } }{{} −∇p/∈E∈Ediv [ (u · ∇)u + ∇p] = ∑ i,ju j x iu i x j+ ∆pthe pressure p satisfies the relati<strong>on</strong>= tr(∇u) 2 + ∆p,if and <strong>on</strong>ly if it solves the elliptic equati<strong>on</strong>:div [ (u · ∇)u + ∇p] = 0<strong>on</strong> the whole space R N .We now recall a soluti<strong>on</strong> formula for linear elliptic equati<strong>on</strong>s∆p = −tr(∇u) 2 (21)11


Lemma 3 Soluti<strong>on</strong> of the Poiss<strong>on</strong> equati<strong>on</strong>Let f be a smooth functi<strong>on</strong> in R N , vanishing sufficiently rapidly as|x| → ∞. Then the soluti<strong>on</strong> to the Poiss<strong>on</strong> equati<strong>on</strong>{ ∆z = f,∇z → 0, as |x| → 0,is given by∫z(x) = N N ∗ f(x) = N N (x − y)f(y) dy,R Nwhere the fundamental soluti<strong>on</strong> N is the Newt<strong>on</strong> potential{ 1N N (x) =2πln |x|, if N = 2,1(2−N) σ N|x| 2−N , if N 3(22)and σ N denotes the surface area of a unit sphere in R N .Observati<strong>on</strong> 2 N N is the soluti<strong>on</strong> in R N of∆N = δ 0where δ 0 is the Dirac mass c<strong>on</strong>centrated at the origin.By applying this Lemma we can solve the equati<strong>on</strong> (21)∫p(t, x) = N N (x − y) tr(∇u(t, y)) 2 dy.R NThe gradient of p takes the form:∇p(t, x) = ∇N N ∗ tr(∇u) 2 (x) = C N∫R Nx − y|x − y| N tr(∇u(t, y))2 dy,for a suitable c<strong>on</strong>stant C N . Using this result we can eliminate the pressure from the explicitformulati<strong>on</strong> of Navier-Stokes equati<strong>on</strong> and we obtain a system of closed evoluti<strong>on</strong> equati<strong>on</strong>s foru. This is the Leray’s formulati<strong>on</strong> of Navier-Stokes equati<strong>on</strong>∫x − yD t u(t, x) = ν∆u − C N|x − y| N tr(∇u(t, y))2 dy. (23)Observati<strong>on</strong> 3 The Navier-Stokes equati<strong>on</strong> can be written:R Nu t = Π E (u · ∇)u + ν∆u,where Π E is the perpendicular projecti<strong>on</strong> operator <strong>on</strong> the space E.To prove this claim, we have to show that, for any scalar functi<strong>on</strong> p vanishing sufficientlyfast as |x| → ∞, the gradient ∇p is orthog<strong>on</strong>al to all functi<strong>on</strong>s v ∈ E.Lemma 4 Let w be a smooth, divergence-free vector field in R N and let q be a smooth scalarfuncti<strong>on</strong> such that|w(x)||q(x)| = o(|x| 1−N ) as |x| → ∞.Then w and ∇q are orthog<strong>on</strong>al in L 2 :∫R N w · ∇q dx = 0.12


Proof A direct computati<strong>on</strong> yields∫div (w q) dx =|x|≤R∫lim (w q) · n dA = 0R→∞ |x|=R□8 Vorticity-stream formulati<strong>on</strong> of the Navier-Stokes equati<strong>on</strong>sTaking the curl of both sides of Navies-Stokes equati<strong>on</strong> we obtain the vorticity evoluti<strong>on</strong> equati<strong>on</strong>,in which the pressure does not appear. This provides an alternative way to close oursystem. In this case, the key step is to recover u from ω = . curl u).8.1 The 2-dimensi<strong>on</strong>al caseIf the flow lies <strong>on</strong> a plane, the velocity field has comp<strong>on</strong>ents u = (u 1 , u 2 , 0) and the vorticityfield is ω = (0, 0, u 2 x 1− u 1 x 2). We can denote them respectively asu =( u1u 2 )and ω = u 2 x 1− u 1 x 2.In this case the evoluti<strong>on</strong> of the vorticity is reduced to a scalar equati<strong>on</strong>∂ t ω + (u · ∇)ω = ν∆ω.To recover the velocity u from the scalar functi<strong>on</strong> ω, we first recall that the system of PDE’s{ψx1 = a(x 1 , x 2 ),ψ x2 = b(x 1 , x 2 ),has a soluti<strong>on</strong> if and <strong>on</strong>ly if a x2 = b x1 . Since div u = u 1 x 1+ u 2 x 2= 0, there exist a functi<strong>on</strong> ψ,called stream functi<strong>on</strong> , which solves{ψx1 = u 2 (x 1 , x 2 ),ψ x2 = −u 1 (x 1 , x 2 ).We then haveu =( −ψx2ψ x1).=∇ ⊥ ψ, (24)ω = curl u = curl (∇ ⊥ ψ) = ∆ψ. (25)Let ω : R 2 ↦→ R be a given vorticity functi<strong>on</strong>. We can determine ψ by solving the Poiss<strong>on</strong>equati<strong>on</strong> (25) as in lemma 3ψ(t, x) = N 2 ∗ ω(x) = 1 ∫ln(|x − y|)ω(t, y) dy .2π R 2We then obtain u from (24),u(t, x) = K 2 ∗ ω(x) =∫13R 2 K 2 (x − y)ω(y) dy. (26)


Here the kernel K 2 is the fundamental soluti<strong>on</strong> of the equati<strong>on</strong>its explicit form iscurl z = δ 0 ,K 2 (x) = 12π |x| 2 ( −x2x 1),Notice that the singularity at the origin has size|K 2 (x)| = 12π |x| .The formula (26) is called Biot-Savart equati<strong>on</strong> since it takes the same form of the Biot-Savartlaw for the magnetic field. Observe that also in this sec<strong>on</strong>d reformulati<strong>on</strong> of the Navier-StokesEquati<strong>on</strong> the pressure can be obtained from the the Poiss<strong>on</strong> equati<strong>on</strong>∆p = −tr(∇u) 2 .8.2 The 3-dimensi<strong>on</strong>al caseSince in the 3-dimensi<strong>on</strong>al case the term ω · ∇u in the vorticity equati<strong>on</strong> is n<strong>on</strong> zero, we haveto recover both u and ∇u from vorticity.Lemma 5 Let ω : R 3 → R 3 be a smooth vector field with a sufficiently fast decay as |x| → ∞.The overdetermined elliptic system { curl u = ω,(27)div u = 0,admits a soluti<strong>on</strong> if and <strong>on</strong>ly if div ω = 0.In the positive case, the soluti<strong>on</strong> is c<strong>on</strong>structed by means of the equati<strong>on</strong>where the stream functi<strong>on</strong> ψ solves the Poiss<strong>on</strong> equati<strong>on</strong>u = −curl ψ. (28)∆ψ = ω . (29)Observati<strong>on</strong> 4 We solve the equati<strong>on</strong> (29) using lemma 3ψ(x) = N 3 ∗ ω(x) = 1 ∫1ω(y) dy,2π R 3 |x − y|and then the explicit soluti<strong>on</strong> to (28) is provided by the Biot-Savart formula∫u(x) = K 3 ∗ ω = K 3 (x − y)ω(y) dy,R 3where the kernel K 3 is defined by the relati<strong>on</strong>K 3 (x) · h = 14πx ∧ h|x| 3 h ∈ R 3 .14


ProofNecessity: If ω = curl u then div ω = div (curl u) = 0.Sufficiency: Assume that div ω = 0. C<strong>on</strong>sider the vector identity∆ψ = ∇(div ψ) − curl (curl ψ) . (30)Take the inner product with ∇(div ψ) we obtain∫∫∇(div ψ) · ∆ψ dx = ∇(div ψ) · ∇(div ψ) dx −∫∇(div ψ) · curl (curl ψ) dx .The first and the third integral terms vanish because divergence-free fields are orthog<strong>on</strong>al togradient fields, as shown in lemma 4. Therefore∫R N |∇(div ψ)| 2 dx = 0 .This proves that ψ is divergence-free andω = ∆ψ = − curl (curl ψ).Settingu . = − curl ψwe obtain ω = curl u and of course div u = div (− curl ψ) = 0.□9 Particle trajectory formulati<strong>on</strong> of the Euler equati<strong>on</strong>In this secti<strong>on</strong> we show that the Euler equati<strong>on</strong>s can be reformulated as a family of integrodifferentialequati<strong>on</strong> for the particle trajectories. Given a smooth velocity field u, the particletrajectory mapping X satisfies the following Cauchy problem, for each value of the Lagrangianvariable α ∈ R N{ ddtX(t, α) = u(t, X(t, α)),X(0, α) = α.We would like to recover u from X and ∇X. We begin by recalling the vorticity transportformulaω(t, X(t, α)) = ∇ α X(t, α)ω 0 (α),and the Biot-Savart lawTogether, these equati<strong>on</strong>s yieldu(t, x) =∫R N K N (x − y)ω(t, y) dy.u(t, X(t, α)) = d dt X(t, α) = ∫R N K N (X(t, α) − y)ω(t, y) dy .C<strong>on</strong>sider the change of integrati<strong>on</strong> variable given byy = X(t, α ′ ), for a suitable Lagrangian variable α ′ ,15


Then∫u(t, X(t, α)) = K N (X(t, α) − X(t, α ′ ))ω(t, X(t, α ′ )) dα ′∫R N= K N (X(t, α) − X(t, α ′ ))∇ α X(t, α ′ )ω 0 (α ′ ) dα ′ .R NThe explicit form of this integral depends <strong>on</strong> N, we will c<strong>on</strong>sider the two cases separately.The 3-dimensi<strong>on</strong>al caseIn the 3-dimensi<strong>on</strong>al case the kernel K 3 is defined by the relati<strong>on</strong>K 3 (x) · h = 14πx ∧ h|x| 3 for x, h ∈ R 3 .Then the integral form of u(t, X(t, α))∫u(t, X(t, α)) = K 3 (X(t, α) − X(t, α ′ ))∇ α X(t, α ′ )ω 0 (α ′ ) dα ′R 3is a singular integral operator in a neighborhood of the origin.The 2-dimensi<strong>on</strong>al caseIn this case the situati<strong>on</strong> is simpler because vorticity is c<strong>on</strong>stant al<strong>on</strong>g particle trajectoriesω(t, X(t, α)) = ω 0 (α).In fact if the flow lies <strong>on</strong> the plane R 2 , then the velocity field is u = (u 1 , u 2 , 0) t , the vorticity isω = (0, 0, u 2 x 1− u 1 x 2) t , and we get:⎛⎞dX 1 dX 1dα 1 dα 20 ⎛ ⎞ ⎛ ⎞0 0∇ α X(t, α) ω =dX⎜ 2 dX 2⎝dα 1 dα 20 ⎟ ⎝ 0 ⎠ = ⎝ 0 ⎠ .⎠ω 3 ω 30 0 1Then the integral form of u(t, X(t, α)) becomes:∫u(t, X(t, α)) = K 2 (X(t, α) − X(t, α ′ ))ω 0 (α ′ ) dα ′ .R 210 Energy estimatesTaking the scalar product of the Navier-Stokes equati<strong>on</strong> with u and integrating by parts weobtain:d |u|dt∫R 2 ∫ ( |u|2 ) ∫∫N 2 dx + divR N 2 u dx = − div (p u) dx − ν |∇u| 2 dx.R N R NIf we assume, as in the previous secti<strong>on</strong>s, that u decays sufficiently fast as |x| → ∞, then thetwo integrals c<strong>on</strong>taining the divergence operator are zero because of Stokes theorem. We obtainan equality between the time derivative of the kinetic energy of the fluid and a negative termdepending <strong>on</strong> the viscosity coefficientd |u|dt∫R 2 ∫N 2 dx = − ν |∇u| 2 dx 0.R NThis equati<strong>on</strong> provides an a priori bound of the L 2 norm of the velocity.16


solves the ODE with initial data:{ ż(t) = 2‖∇v2 ‖ L ∞z(t),11 Uniqueness of soluti<strong>on</strong>sLet v 1 and v 2 be two smooth soluti<strong>on</strong>s of the same Navier-Stokes equati<strong>on</strong>, vanishing sufficientlyfast as |x| → ∞, and so, in particular, both in L 2 . If p 1 and p 2 are the associate pressures wedefineṽ . = v 1 − v 2 ,˜p . = p 1 − p 2 .These two functi<strong>on</strong>s ṽ, ˜p satisfy the evoluti<strong>on</strong> equati<strong>on</strong>ṽ t + (v 1 · ∇) ṽ + (ṽ · ∇) v 2 = −∇˜p = ν∆ṽ.Let’s note that the linear terms involve <strong>on</strong>ly ṽ and ˜p.Taking the inner product of this equati<strong>on</strong>s with ṽ we obtain:∫∫∫ṽ · ṽ t dx + ṽ · (v 1 · ∇)ṽ dx + ṽ · (ṽ · ∇)v 2 dx =R N R} N R{{ }N−∫(⋆)R N ṽ · ∇˜p dx} {{ }(⋆ ⋆)∫− ν ṽ · ∆ṽ dxR NWe observe that (⋆) and (⋆ ⋆) are equal to zero: (⋆ ⋆) is the scalar product in L 2 of a divergencefree vector field and the gradient of a scalar functi<strong>on</strong>, while integrating (⋆) by parts we get(⋆) =N∑∫ ( |ṽ i | 2 )v 1 · ∇ dx =R N 2i=1N∑∫div(v 1 · |ṽi | 2R N 2i=1)dx = 0because of the fast decay of v 1 as |x| → ∞.Integrating by parts the other terms the equati<strong>on</strong> (31) becomes:d |ṽ|dt∫R 2 ∫∫N 2 dx = − ṽ · (ṽ · ∇)v 2 dx − ν |∇ṽ| 2 dx .R N R} N {{ }0Then the following estimate holds:Comparis<strong>on</strong> with an ODEThe functi<strong>on</strong> defined asddt ‖ṽ‖2 L 2 ‖ṽ‖ 2 L 2 ‖∇v 2 ‖ L ∞.z(T ) = exp ( ∫ T2‖∇v 2 (t)‖ L 2 dt ) z(0)0(31)z(0) = ‖ṽ(0)‖ 2 L, 2and for any instant T 0 there holds:‖ṽ(T )‖ 2 L z(T ).2So if the soluti<strong>on</strong>s v 1 and v 2 coicide at t = 0, then they coincide for all t 0.17

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!