• Aucun résultat trouvé

One- Dimensional Convex Integration

N/A
N/A
Protected

Academic year: 2022

Partager "One- Dimensional Convex Integration"

Copied!
33
0
0

Texte intégral

(1)

Dimensional Convex Integration

V.Borrelli

Bibliography Introductory Examples Fundamental Lemma C0-Density

One-Dimensional Convex Integration

Vincent Borrelli

Université Lyon 1 - France

(2)

Dimensional Convex Integration

V.Borrelli

Bibliography Introductory Examples Fundamental Lemma C0-Density

Bibliography

Mikhail Gromov

(3)

Dimensional Convex Integration

V.Borrelli

Bibliography Introductory Examples Fundamental Lemma C0-Density

Bibliography

Yakov Eliashberg

(4)

Dimensional Convex Integration

V.Borrelli

Bibliography Introductory Examples Fundamental Lemma C0-Density

Bibliography

David Spring

(5)

Dimensional Convex Integration

V.Borrelli

Bibliography Introductory Examples Fundamental Lemma C0-Density

An Elementary Example

Exercice.– Let f0: [0,1] −→ R3 t 7−→ (0,0,t)

Findf : [0,1]−→C1 R3such that :

i) ∀t ∈[0,1], |cos(f0(t),e3)|<

ii) kf −f0kC0 < δ

where >0 andδ >0 are given.

e3

e1

e2 f 0

?

(6)

Dimensional Convex Integration

V.Borrelli

Bibliography Introductory Examples Fundamental Lemma C0-Density

An Elementary Example

Exercice.– Let f0: [0,1] −→ R3 t 7−→ (0,0,t)

Findf : [0,1]−→C1 R3such that :

i) ∀t ∈[0,1], |cos(f0(t),e3)|<

ii) kf −f0kC0 < δ

where >0 andδ >0 are given.

e3

e1

e2

f 0 f

(7)

Dimensional Convex Integration

V.Borrelli

Bibliography Introductory Examples Fundamental Lemma C0-Density

Rephrasing...

Condition(i)means that the image off0 lies inside the blue cone :

e

3

R

Ris thedifferential relationof our problem.

(8)

Dimensional Convex Integration

V.Borrelli

Bibliography Introductory Examples Fundamental Lemma C0-Density

... and new understanding

The mapf must beC0near off0:

f ’0 f ’

The average off0 for eachloopoff isf00(t): 1

Long(I) Z

I

f0(u)du =f00(t)

(I=the preimage of one loop byf).

(9)

Dimensional Convex Integration

V.Borrelli

Bibliography Introductory Examples Fundamental Lemma C0-Density

A more general problem

Problem.– LetR ⊂R3be a path-connected subset (=our differential relation) andf0: [0,1]−→C1 R3be a map such

∀t ∈[0,1], f00(t)∈ Conv(R).

Findf : [0,1]−→C1 R3such that : i) ∀t ∈[0,1], f0(t)∈ R ii) kf −f0kC0 < δ

withδ >0 given.

(10)

Dimensional Convex Integration

V.Borrelli

Bibliography Introductory Examples Fundamental Lemma C0-Density

How to build a solution ?

f ’ 0

R

(11)

Dimensional Convex Integration

V.Borrelli

Bibliography Introductory Examples Fundamental Lemma C0-Density

How to build a solution ?

(12)

Dimensional Convex Integration

V.Borrelli

Bibliography Introductory Examples Fundamental Lemma C0-Density

How to build a solution ?

f ’ 0

R

f ’

(13)

Dimensional Convex Integration

V.Borrelli

Bibliography Introductory Examples Fundamental Lemma C0-Density

Construction of the solution : step 1

Choose a continuous family of loops

h: [0,1] −→ C0(R/Z,R)

u 7−→ hu

such that

∀u∈ [0,1], Z

[0,1]

hu(s)ds=f00(u).

h u

f ’ (u)0

s

u

(14)

Dimensional Convex Integration

V.Borrelli

Bibliography Introductory Examples Fundamental Lemma C0-Density

Construction of the solution : step 2

Then set

f(t) :=f0(0) + Z t

0

hu({Nu})du

whereN∈Nand{Nu}is the fractional part ofNu.

({Nu}) f’(u):=h u

s

u

(15)

Dimensional Convex Integration

V.Borrelli

Bibliography Introductory Examples Fundamental Lemma C0-Density

Construction of the solution : step 2

Then set

f(t) :=f0(0) + Z t

0

hu({Nu})du

whereN∈Nand{Nu}is the fractional part ofNu.

We say thatf is obtained fromf0by aconvex integration.

({Nu}) f’(u):=h u

s

u

(16)

Dimensional Convex Integration

V.Borrelli

Bibliography Introductory Examples Fundamental Lemma C0-Density

Fundamental Lemma

Notation.– LetR ⊂Rn andσ ∈ R.We denote by IntConv(R, σ)the interior of the convex hull of the connected component ofRto whichσ belongs.

Definition.– A (continuous) loopg : [0,1]→Rn, g(0) =g(1),strictly surrounds z ∈Rnif

IntConv(g([0,1]))⊃ {z}.

(17)

Dimensional Convex Integration

V.Borrelli

Bibliography Introductory Examples Fundamental Lemma C0-Density

Fundamental Lemma

Fundamental Lemma.–LetR ⊂Rnbe an open set,σ∈ R and z∈IntConv(R, σ)There exists a loop h: [0,1]−→ RC0 with base pointσ that strictly surrounds z and such that :

z = Z 1

0

h(s)ds.

(18)

Dimensional Convex Integration

V.Borrelli

Bibliography Introductory Examples Fundamental Lemma C0-Density

Fundamental Lemma

Remark.– A priori h∈Ωσ(R),but it is obvious that we can choosehamong "round-trips"i.ethe space :

ARσ (R) ={h∈Ωσ(R)| ∀s∈[0,1] h(s) =h(1−s)}.

The point is that the above space is contractible. For every u∈[0,1]we then denote byhu: [0,1]−→ Rthe map defined by

hu(s) =

h(s) if s∈[0,u2]∪[1−u2] h(u) if s∈[u2,1− u2].

This homotopy induces a deformation retract ofΩARσ (R)to the constant map

σe: [0,1] −→ R

s 7−→ σ.

(19)

Dimensional Convex Integration

V.Borrelli

Bibliography Introductory Examples Fundamental Lemma C0-Density

Parametric Fundamental Lemma

Parametric version of the Fundamental Lemma. –Let E = [a,b]×Rn−→π [a,b]be a trivial bundle andR ⊂E be an open set. Letσ ∈Γ(R)and z ∈Γ(E)such that :

∀p∈[a,b], z(p)∈IntConv(Rp, σ(p)) whereRp :=π−1(p)∩ R.Then, there exists h: [a,b]×[0,1] C

−→ Rsuch that :

h(.,0) =h(.,1) =σ∈Γ(R),

∀p ∈[a,b], h(p, .)∈ΩARσ(p)(Rp) and

∀p∈[a,b], z(p) = Z 1

0

h(p,s)ds.

(20)

One- Dimensional

Convex Integration

V.Borrelli

Bibliography Introductory Examples Fundamental Lemma C0-Density

Proof

Observation.– The parametric lemma still holds if the parameter space[a,b]is replaced by a compact manifoldP.

(21)

Dimensional Convex Integration

V.Borrelli

Bibliography Introductory Examples Fundamental Lemma C0-Density

Proof

Observation.– The parametric lemma still holds if the parameter space[a,b]is replaced by a compact manifoldP.

(22)

Dimensional Convex Integration

V.Borrelli

Bibliography Introductory Examples Fundamental Lemma C0-Density

The Convex Integration Process

LetR ⊂Rnbe a arc-connected subset,f0∈C(I,Rn)be a map such thatf00(I)⊂ IntConv(R).We set

∀t∈I, F(t) :=f0(0) + Z t

0

h(s,Ns)ds

withN ∈N.

Definition.– We say thatF ∈C(I,Rn)is a solution ofR obtained fromf0by aconvex integration process.

ObviouslyF0(t) =h(t,Nt)∈ Rand thusF is a solution of the differential relationR.

(23)

Dimensional Convex Integration

V.Borrelli

Bibliography Introductory Examples Fundamental Lemma C0-Density

C

0

-Density

One crucial property of the convex integration process is that the solutionF can be made arbitrarily close to the initial mapf0.

Proposition (C0-density).–We have

kF −f0kC0 ≤ 1 N

2khkC0+k∂h

∂tkC0

wherekgkC0 =supp∈Dkg(p)k

E3 denotes the C0norm of a function g:D→E3.

(24)

One- Dimensional

Convex Integration

V.Borrelli

Bibliography Introductory Examples Fundamental Lemma C0-Density

C

0

-Density

Remark.– Even iff0(0) =f0(1), the mapF obtained by a convex integration fromf0does not satisfyF(0) =F(1)in general. This can be easily corrected by defining a new map f with the formula

∀t∈[0,1] , f(t) =F(t)−t(F(1)−F(0)).

Proposition (C0-density).–We have

kf −f0kC0 ≤ 2 N

2khkC0+k∂h

∂tkC0

(25)

Dimensional Convex Integration

V.Borrelli

Bibliography Introductory Examples Fundamental Lemma C0-Density

C

0

-Density

Remark.– Even iff0(0) =f0(1), the mapF obtained by a convex integration fromf0does not satisfyF(0) =F(1)in general. This can be easily corrected by defining a new map f with the formula

∀t∈[0,1] , f(t) =F(t)−t(F(1)−F(0)). Proposition (C0-density).–We have

kf −f0kC0 ≤ 2 N

2khkC0+k∂h

∂tkC0

(26)

Dimensional Convex Integration

V.Borrelli

Bibliography Introductory Examples Fundamental Lemma C0-Density

C

0

Density, N = 3

(27)

Dimensional Convex Integration

V.Borrelli

Bibliography Introductory Examples Fundamental Lemma C0-Density

C

0

Density, N = 5

(28)

Dimensional Convex Integration

V.Borrelli

Bibliography Introductory Examples Fundamental Lemma C0-Density

C

0

Density, N = 10

(29)

Dimensional Convex Integration

V.Borrelli

Bibliography Introductory Examples Fundamental Lemma C0-Density

C

0

Density, N = 20

(30)

Dimensional Convex Integration

V.Borrelli

Bibliography Introductory Examples Fundamental Lemma C0-Density

Multi Variables Setting

•In a multi-variable setting, the convex integration formula takes the following natural form :

f(c1, ...,cm) :=f0(c1, ...,cm−1,0)+

Z cm

0

h(c1, ...,cm−1,s,Ns)ds

where(c1, ...,cm)∈[0,1]m.

A corrugated plane

(31)

One- Dimensional

Convex Integration

V.Borrelli

Bibliography Introductory Examples Fundamental Lemma C0-Density

Multi Variables Setting

•TheC0-density property can be enhanced to a

C1,bm-density property where the notationC1,bmmeans that the closeness is measured with the following norm

kfkC1,bm =max(kfkC0,k ∂f

∂c1kC0, ...,k ∂f

∂cm−1

kC0).

Proposition (C1,bm-density).–We have

kf−f0kC1,mb =O 1

N

.

(32)

Dimensional Convex Integration

V.Borrelli

Bibliography Introductory Examples Fundamental Lemma C0-Density

Multi Variables Setting

•TheC0-density property can be enhanced to a

C1,bm-density property where the notationC1,bmmeans that the closeness is measured with the following norm

kfkC1,bm =max(kfkC0,k ∂f

∂c1kC0, ...,k ∂f

∂cm−1

kC0).

Proposition (C1,bm-density).–We have

kf−f0kC1,bm =O 1

N

.

(33)

Dimensional Convex Integration

V.Borrelli

Bibliography Introductory Examples Fundamental Lemma C0-Density

Mikhaïl Gromov

Références

Documents relatifs

Practical approaches consist either of solving the above kind of problem with a finite number of possible initial data, or of recasting the optimal sensor location problem for

We consider the Finite Volume method for conservation laws with high order polynomial reconstruction in the context of one dimensional problems.. A particular reconstruction based on

It is well known that closed subgroups of minimal Abelian groups are mini- mal ([P2], [P5]) and all compact Hausdorff groups are minimal1. Stojanov and the first

In particular, we define and completely solve a convexified version of this problem, underline a generic non existence result with respect to the values of the measure constraint L,

We first prove existence and uniqueness of optimal transportation maps for the Monge’s problem associated to a cost function with a strictly convex constraint in the Euclidean plane

We show that the solution of the two-dimensional Dirichlet problem set in a plane do- main is the limit of the solutions of similar problems set on a sequence of

Keywords: Wasserstein space; Geodesic and Convex Principal Component Analysis; Fréchet mean; Functional data analysis; Geodesic space;.. Inference for family

Keywords: Riemannian manifold, gradient flow, steepest descent, convex- ity, minimization.. AMS Subject Classification: