• Aucun résultat trouvé

2 Problem II

N/A
N/A
Protected

Academic year: 2022

Partager "2 Problem II"

Copied!
6
0
0

Texte intégral

(1)

Universit´e Paris-Dauphine 2016-2017

An introduction to evolution PDEs January 2017

1 Problem I

We consider the evolution PDE

tf = ∆f+ div(Ef), (1.1)

on the unknown f =f(t, x),t ≥0, x∈Rd, withE =E(x) a given smooth force field which satisfies for someγ ≥1

∀ |x| ≥1, |E(x)| ≤C|x|γ−1, divE(x)≤C|x|γ−2, x·E ≥ |x|γ. We complement the equation with an initial condition

f(0, x) = f0(x).

Question 1. Which strategy can be used in order to exhibit a semigroup S(t) in Lp(Rd), which provides solutions to (2.1) for initial data inLp(Rd)? Is the semigroup positive? mass conservative? Explain briefly why there exists a functionG=G(x) such that

0≤G∈L2(m), hGi:=

Z

G= 1, LG= 0.

We accept that G > 0. For any nice function f : Rd → R we denote h := f /G and, reciprocally, for any nice function h:Rd→R we denotef :=Gh.

In the sequel we will not try to justify rigorously the a priori estimates we will establish, but

(2)

Question 4. For some constant λ ≥ 0 to be specified later, we define W :=w+λ. Deduce from the previous question that

Z

h2w G≤ 1 α

Z

h2(b1BR

0 − LW)G, for any nice function h:Rd→R.

Question 5. Take a nice functionh:Rd→Rsuch thathhGi= 0 and denoteG(Ω) :=hG1i.

Prove that for any R≥R0 there exists κR∈(0,∞) such that Z

h21BRG≤κR

Z

BR

|∇h|2G+ 1 G(BR)

Z

BRc

h G 2

and deduce that Z

h21BRG≤ κR 1 +λ

Z

|∇h|2W G+ G(BRc) G(BR)

Z

h2w G.

Question 6. Establish finally that there exist some constants λ, K1 ∈(0,∞) such that 2

K1 Z

h2wG≤ Z

W|∇h|2− 1

2h2LW G=

Z

(−Lf)f G−1W, for all nice function h such that hhGi= 0.

Question 7. Consider a nice solution f to (2.1) associated to an initial datum f0 such that hf0i= 0. Establish thatf satisfies

1 2

d dt

Z

h2tW G=− Z

|∇h|2W G+1 2

Z

h2GLW.

Deduce that there exists K2 ∈(0,∞) such that f satisfies the decay estimate Z

ft2W G−1dx≤e−K2t Z

f02W G−1dx, ∀t≥0.

(3)

2 Problem II

We consider the Keller-Segel equation

tf = ∆f + div(Kff), (2.1)

on the unknown f =f(t, x),t≥0, x∈R2, with

Kf :=K ∗f, K:=∇κ = 1 2π

z

|z|2. We complement the equation with an initial condition

f(0, x) = f0(x).

We accept that for any initial datumf0 ≥0 with finite massM > 0, finite moment of order 2 and finite entropy, there exists at least one nonnegative solutionf ∈C([0, T);L1) for some T >0, and that any weak solution furthermore satisfies

f ∈C1((0, T);W1,p(R2)), ∀p∈[1,∞].

Question 1. Establish that any weak solution satisfies d

dtkfk2L2 + 2k∇xfk2L2 =kfk3L3 on (0, T).

Deduce that there exists some constantA ∈(0,∞) such that d

dtkfk2L2 +1

2k∇xfk2L2 ≤A2M on (0, T).

Deduce next that there exists a constant cM >0 such that d

dtkfk2L2 +cMkfk4L2 ≤A2M on (0, T).

and finally that exists a constant K (which only depends on c , A2M and T) such that

(4)

Question 3. Introducing the notation logf+g := 2 + (logg)+ and using an H¨older inequality, prove that

kgkL4/3 ≤C(H(g), M2(g)) Z

f2(logf+g)−2 1/4

.

Question 4. Establish that

t1/4kf(t, .)kL4/3 →0 as t→0.

Briefly explain how to deduce the uniqueness of weak solutions to the Keller-Segel equation.

(5)

3 Problem III -

We consider the relaxation equation

tf =Lf :=−v · ∇f +ρfM −f in (0,∞)×R2d, (3.1) on the unknown f =f(t, x, v),t≥0, x, v ∈Rd, with

ρf(t, x) = Z

Rd

f(t, x, v)dv, M(v) := 1

(2π)d/2 exp(−|v|2/2).

We complement the equation with an initial condition

f(0, x, v) = f0(x, v) in R2d. (3.2) Question 1. A priori estimates and associated semigroup. We denote by f a nice solution to the relaxation equation (3.1)–(3.2).

(a) Prove that f is mass conserving.

(b) Prove that

g| ≤ kgkL2

v(M−1/2), ∀g =g(v)∈L2v(M−1/2) and deduce that

kf(t,·)kL2

xv(M−1/2)≤ kf0kL2

xv(M−1/2).

(c) Consider m=hvik, k > d/2. Prove that there exists a constant C ∈(0,∞) such that

g| ≤CkgkLpv(m), ∀g =g(v)∈Lpv(m), p= 1,2, and deduce that

kf(t,·)kLpxv(m) ≤eλtkf0kLpxv(m), for a constant λ ∈[0,∞) that we will express in function of C.

(d) What strategy can be used in order to exhibit a semigroupS(t) inLpxv(m),p= 2,p= 1, which provides solutions to (3.1) for initial date inLpxv(m)? Is the semigroup positive? mass conservative? a contraction in some spaces?

(6)

Question 3. Establish thatA:L1xv(m)→L1xLv (m) where kgkL1

xLpv(m) :=

Z

Rd

kg(x,)kLp(m)dx.

Prove that d dt

Z Z

fpdx1/p

dv= Z Z

(∂tf)fp−1dxZ

fpdx1/p−1

dv.

Deduce thatSB satisfies a growth estimateO(e−t) in anyL1xLpv(m) space forp∈(1,∞), and then in L1xLv (m). Finally prove that SB(t)A is appropriately bounded in B(L1, L1xLv (m)) and that SL is bounded in L1xLv (m).

Question 4. We define u(t) :=ASB(t). Establish that (u(t)f0)(x, v) = M(v)e−t

Z

Rd

f0(x−vt, v)dv. Deduce that

ku(t)f0kLxv(m) ≤Ce−t

td kf0kL1xLv (m).

Question 5. Establish that there exists some constantsn ≥1 andC ∈[1,∞) such that ku(∗n)(t)kL1xv(m)→Lxv(m) ≤C e−t/2.

Deduce that SL is bounded in Lxv(m).

Question 6. How to prove that SL is bounded in L2xv(m) in a similar way? How to shorten the proof of that last result by using question (1b)? Same question for the spaceLxv(m).

Références

Documents relatifs

Abstract : We study families of complex Monge-Amp`ere equations, fo- cusing on the case where the cohomology classes degenerate to a non big class.. Kolodziej [K 1], [K 2] that

In turn, the decomposition of the solution into a regular and a singular part, together with interpolation estimates (Theorem 3.1), leads to establishing a priori error estimates

[4] Caffarelli L.A., Nirenberg L., Spruck J., The Dirichlet problem for nonlinear second order elliptic equations, III: Functions of the eigenvalues of the Hessian, Acta Math. J.,

In this paper we derive several a priori estimates for classical solutions of certain problems in two-dimensional compressible nonlinear elasticity.. We consider a

We show that if a positive Radon measure is suitably dominated by the Monge-Amp`ere capacity, then it belongs to the range of the Monge-Amp`ere op- erator on some class E χ ( X,

Under this assumption we prove interior a priori estimates of Schauder type for solutions of a class of second order partial differential equations with C α coefficients, which are

operators with constant coefficients and general boundary conditions. He does not, however, use a, priori estimates, but explicit formulas for the cor-.. responding

Average error between the finite element approximation u FEM discretized in time and space (this is the solution of Problem 2.3 in which we replace the POD space V p by the finite