Inverse boundary problems for elliptic PDE and best approximation by - - PowerPoint PPT Presentation

inverse boundary problems for elliptic pde and best
SMART_READER_LITE
LIVE PREVIEW

Inverse boundary problems for elliptic PDE and best approximation by - - PowerPoint PPT Presentation

Inverse boundary problems for elliptic PDE and best approximation by analytic functions Juliette Leblond Sophia-Antipolis, France team APICS Joint work with L. Baratchart (team APICS), Y. Fischer (team Magique3D, INRIA Bordeaux) Overview


slide-1
SLIDE 1

Inverse boundary problems for elliptic PDE and best approximation by analytic functions

Juliette Leblond

Sophia-Antipolis, France team APICS Joint work with

  • L. Baratchart (team APICS), Y. Fischer (team Magique3D, INRIA Bordeaux)
slide-2
SLIDE 2

Overview

  • Boundary value problems

Dirichlet, Cauchy

  • Normed spaces of generalized analytic functions

Hardy

  • Application: a physical free boundary problem

plasma

  • Conclusion
slide-3
SLIDE 3

Conductivity equation

Let Ω ⊂ R2 with smooth boundary Γ = ∂Ω

(H¨

  • lder or Dini-smooth)

Ω simply connected: Ω ≃ disk D, Γ ≃ circle T

≃ conformally

  • r annular: Ω ≃ A, Γ ≃ T ∪ ̺T

(also in multiply connected domains) 0 < ̺ < 1

Conductivity coefficient σ Lipschitz smooth function in Ω

(known)

Consider solutions u to (u): div (σ grad u) = div (σ ∇ u) = 0 in Ω (u)

distributional sense 0 < c ≤ σ ≤ C second order elliptic equation ∆u + ∇(log σ).∇u = 0

slide-4
SLIDE 4

Boundary value problems

  • Cauchy (inverse) problem:

|I| , |J| > 0 partial overdetermined boundary data

Given measures u and σ ∂nu on I ⊂ Γ of a solution u to (u),

(u): div (σ ∇ u) = 0 in Ω

recover u, σ ∂nu on J = Γ \ I

(or ∂nu) n outer unit normal σ given pair of Dirichlet-Neumann data (φI , ψI ) on I, φI ∈ L2

R(I), ψI ∈ W −1,2 R

(I)... compatibility...

  • Dirichlet (direct) problem:

Given measures of u on Γ, recover u in Ω (and σ∂nu, on Γ)

well-posed for Dirichlet data φ ∈ L2

R(Γ)... (already for smooth data)

L2 boundary data smooth conductivity σ, tradeoff

practically: pointwise corrupted boundary measurements

slide-5
SLIDE 5

Ω = A I = T, J = ̺T

I J u, ∂nu u, ∂nu ? Ω = A ̺T T

u, ∂nu u, ∂nu ?

Ω = D, I ⊂ T, J = T \ I

slide-6
SLIDE 6

σ-harmonic conjugation

Generalized Cauchy-Riemann equations:

for Ω = D

u solution to (u): div (σ ∇ u) = 0 = ⇒ ∃v such that in Ω: ∂xv = −σ∂yu ∂yv = σ∂xu whence div 1 σ∇v

  • = 0

Function v: σ-conjugated to u

v unique up to additive constant

If u solution to (u) and its σ-conjugated v have L2(Γ) trace, then Cauchy-Riemann equations hold up to boundary Γ: ∂θv = σ∂nu

∂θ tangential derivative for Ω = A: ∃v if compatibility boundary condition

slide-7
SLIDE 7

Generalized analytic functions

In Ω ≃ D ⊂ R2 ≃ C complex plane

X = (x, y) ≃ z = x + iy , ∂ = ∂z = 1 2 (∂x − i ∂y ) , ¯ ∂ = ∂¯

z =

1 2 (∂x + i ∂y )

u solution to (u): div (σ ∇ u) = 0

∇ ≃ ¯ ∂, div ≃ Re ∂

⇔ f = u + i v satisfies conjugated Beltrami equation

pseudoanalytic

¯ ∂f = ν∂f (f) for ν = 1 − σ 1 + σ ∈ W 1,∞(Ω) , |ν| ≤ κ < 1 in Ω f solution to (f) ⇐ ⇒ u = Re f solution to (u)

(f) conformally invariant (R-linear, first order) = C-linear Beltrami equation: ¯ ∂g = ν∂g, quasi-conformal map. f (z, ¯ z), u(x, y), v(x, y) in Ω ≃ A, compatibility condition needed for ⇐

slide-8
SLIDE 8

Harmonic and analytic functions

Generalization of homogeneous situations σ = cst σ = 1, ν = 0 Holomorphic / complex analytic functions ¯ ∂F = 0 in D ⊂ C:

Ω = D unit disc

  • r Ω ≃ D conformally equivalent

X = (x, y) ≃ z = x + iy , ∂ = ∂z = 1 2 (∂x − i ∂y ) , ¯ ∂ = ∂¯

z =

1 2 (∂x + i ∂y ) Laplace operator ∆ = 4 ¯ ∂ ∂ = 4 ∂ ¯ ∂ = ∂2

x + ∂2 y

F(z) =

  • k≥0

ˆ Fk zk =

  • k≥0

ˆ Fk rk eikθ , z = r eiθ ∈ D , r < 1

(Fourier series, coefficients ˆ Fk )

¯ ∂ F = 0 (F holomorphic) ⇔ F = u + iv with ∆u = 0 and ∆v = 0: harmonic u and conjugate function v satisfying Cauchy-Riemann equations in D: ∂xv = −∂yu ∂yv = ∂xu

slide-9
SLIDE 9

Hardy spaces H2 of analytic functions in D

H2(D): solutions to ¯ ∂ F = 0 in D, F2 < ∞ F2

2 = ess sup 0<r<1

2π |F(reiθ)|2 dθ 2π =

  • k≥0

|ˆ Fk|2

Hilbert space ⊂ L2(D) Parseval p = 2, also Ω = A and Banach Hp

L2 boundary values on T: tr H2(D) ⊂ L2(T)

traces, non tg lim L2(T) = tr H2(D) ⊕ tr H2,0(C \ D) ⊥ decomposition, projection P+

equivalent boundary L2(T) norm: F2 = tr FL2(T) Cauchy-Riemann equation in D, up to boundary T: F = u + iv, ∂θv = ∂nu, ∂nv = −∂θu tr v = Htr u

also Cauchy integral formula, Poisson kernel, Hilbert-Riesz operator + further properties [Duren, Garnett] results for σ = 1, ν = 0, Laplace equations (dimension 2 or 3)

slide-10
SLIDE 10

Generalized Hardy space H2

ν

Hilbert space H2

ν = H2 ν(Ω):

Ω = D or A

also Ω ≃ D or A conformally, and Banach Hp

ν, 1 < p < ∞

  • solutions f to (f)

¯ ∂f = ν∂f in Ω

  • bounded in Hardy norm in Ω

f 2 < ∞

(sup of L2 norms on circles in Ω)

H2

ν shares many properties of H2 = H2

[Baratchart-L.-Rigat-Russ, 2010], [Fischer, 2011], [F.-L.-Partington-Sincich, 2011], [BFL, 2012]

slide-11
SLIDE 11

Properties of H2

ν

Generalize those of H2 Ω = D or A

also Hp

ν [BFL,F]

Theorem [BLRR] f ∈ H2

ν(Ω)

¯ ∂f = ν∂f , f 2 < ∞

  • f admits a non tangential limit tr f ∈ L2(Γ) on Γ
  • tr f = 0 a.e. on I ⊂ Γ, |I| > 0 implies that f ≡ 0

if f ≡ 0, then log |tr f | ∈ L1(Γ), and f admits isolated zeroes (+ Blaschke condition)

  • tr f L2(Γ) is equivalent to f 2 on H2

ν(Ω)

Hardy norm

  • Closedness of traces:

tr H2

ν(Ω) is closed in L2(Γ)

  • Re tr f = 0 a.e. on Γ implies that f ≡ 0 in Ω

(up to constant)

whenever

normalization on Γ

f ∈ H2,0

ν (Ω) = {f ∈ H2 ν(Ω) ,

  • T

Im tr f = 0}

Γ = T or T ∪ ̺T + maximum principle in modulus

slide-12
SLIDE 12

Properties of tr H2

ν(D)

Corollary [BLRR]

Dirichlet in H2

ν(D), density

  • ∀φ ∈ L2

R(T), ∃! f ∈ H2,0 ν (D) such that Re tr f = φ

moreover , tr f L2(T) ≤ cνφL2(T)

  • conjugation operator Hν bounded on L2

R(T)

Hilbert-Riesz transform, L2(T)

Re tr f = φ

− → Im tr f = Hνφ f ∈ H2,0

ν (D) ⇐

⇒ tr f = (I + iHν)φ, φ ∈ L2

R(T)

  • density:

let I ⊂ T, J = T \ I such that |J| > 0 then, restrictions to I of functions in tr H2

ν(D) dense in L2(I)

also in Hp

ν, 1 < p < ∞

slide-13
SLIDE 13

Other situations

  • Generalization to Ω = A annulus

A = D \ ̺D

  • r multiply connected smooth domains [BFL, F]

Dirichlet in H2

ν(A) for data in L2 R(A) ⊖ S

H2

ν(A) = solutions to (f): ¯

∂f = ν∂f in A with f 2 < ∞ S = {φ ∈ L2

R(∂A) s.t. φ|T = C, φ|̺T = −C, C ∈ R}

Density of restrictions on I ⊆ T of tr H2

ν(A) in L2(I)

(or I ⊆ ̺T)

  • Conformal invariance of (f): Ω ≃ D or Ω ≃ A
  • For H¨
  • lder smooth ν ∈ W 1,r(Ω), r > 2

(and σ)

in Hp

ν (Ω)

with ∞ > p > r/(r − 1)

slide-14
SLIDE 14

For related conductivity PDE

u solution to (u) in Ω:

Ω ≃ D or A

div (σ ∇ u) = 0 ⇐ u = Re f with f solution to (f) in Ω

if Ω ≃ D, ⇔

Dirichlet boundary value problems: from prescribed boundary data φ ∈ L2

R(Γ)

Ω ≃ A: φ ⊥ S

recover u in Ω solution to (u) such that tr u = φ on Γ From Dirichlet theorem in H2,0

ν (Ω):

∃! u in Lp

R(Ω) solution to (u) such that tr u = φ

tr f = φ + i

  • Γ

σ∂nu = φ + iHνφ , u2 = tr uL2(Γ) = φL2(Γ)

Also, unique continuation properties bounded conjugation operator stability properties for (u)... Dirichlet-Neumann map: Λφ = ∂θ Hνφ

slide-15
SLIDE 15

For related conductivity PDE

Cauchy inverse problems, I ⊂ T Ω = D or A Given φI and ψI in L2

R(I)

recover u solution to (u) in Ω such that tr u = φI, σ∂nu = ψI on I Let Φ = φI + i

  • I

ψI ∈ L2(I) Density results: [tr H2

ν]|I dense in L2(I)

Runge property

(compatible boundary data) ∃fk ∈ tr H2

ν, Φ − fkL2(I) → 0

(k → ∞)

either Φ ∈ tr H2

ν|I already

and Φ − fkL2(T) → 0 However

  • r Φ /

∈ tr H2

ν|I

and fkL2(J) → ∞

slide-16
SLIDE 16

For related conductivity PDE

Cauchy problem ill-posed for non compatible data φI, ψI on I Φ = φI + i

  • I ψI ∈ L2(I) \ (tr H2

ν)|I :

∃uk = Re fk solution to (u) in Ω tr uk − φIL2(I) − → 0

∂θHνuk − ψI L2(I) → 0

but tr ukL2(J) − → ∞ Look for tr u ≃ φI, σ∂nu ≃ ψI on I with tr u bounded on J... Bounded extremal problems (BEP) in tr H2

ν

best constrained approximation

slide-17
SLIDE 17

Best constrained approximation in H2

ν

Regularization: bounded extremal problems (BEP) Let I ⊂ Γ, |I|, |J| > 0, ε > 0

Ω = D, Γ = T, J = T \ I

B =

  • f ∈ tr H2

ν , Re f L2(J) ≤ ε

  • |I ⊂ L2(I) .

Theorem [BFL, FLPS] (BEP) well-posed

ν = 0: [BLP]

∀ function Φ ∈ L2(I), ∃ unique f∗ ∈ B such that Φ − f∗L2(I) = min

f ∈B Φ − f L2(I)

Moreover, if Φ / ∈ B, then Re f∗L2(J) = ε

Proof: bounded conjugation, density result also in Ω ≃ A, with I ⊂ T, J = (T \ I) ∪ ̺T also in Hp

ν, for Lp(I) data, or with other norm constraints

slide-18
SLIDE 18

Constructive issues in H2

ν

Computation algorithm, from Φ ∈ L2(I)

Ω = D, A (I ⊆ T) [AP,BFL,FLPS]

⊥ projection operator L2(Γ) → tr H2,0

ν

: Pνφ = 1 2(φ+iHνφ)

vanishing mean on T

Solution to (BEP): given Φ ∈ L2(I), M > 0

Toeplitz-Hankel operators on H2

ν

Pν(χIf∗) − γPν(χJf∗) = (I − (γ + 1)PνχJ) f∗ = Pν(Φ ∨ 0) for ! Lagrange parameter γ < 0 s.t. f∗L2(J) = M

minf ∈tr H2

ν Φ − f L2(I) + γ Re f L2(J)

γ % M smoothly decreasing

Complete families of solutions, for computations

in H2

ν(Ω) and L2(Γ)

Bessel/exponentials, toroidal harmonics (w.r.t. σ or ν, and Ω) polynomials? ν = 0: Fourier basis, polynomials [L.-P.-Pozzi]

slide-19
SLIDE 19

Plasma equilibrium model in a tokamak

In 2D poloidal sections, poloidal magnetic flux u: div 1 x ∇u

  • = div (σ∇u) = 0 in the vacuum Ω , conductivity σ = 1

x

Maxwell equations, cylindrical coordinates (x, y) = (R, Z), φ = cte

Ω ≃ A0 ⊂ R2 annular domain between plasma and chamber Γ = Γe ∪ Γp

limitor Γl ⊂ Ω inside plasma, Grad-Shafranov equation, control

1 1.5 2 2.5 3 3.5 4 −1 −0.8 −0.6 −0.4 −0.2 0.2 0.4 0.6 0.8 1 r(m) z(m)

θ ρT CT

Bt Bρ

From pointwise magnetic data

  • n outer boundary Γe

(tg poloidal mag. field)

u , Bρ = −1 x ∂tu , Bt = 1 x ∂nu recover plasma boundary Γp

free boundary problem

slide-20
SLIDE 20

Plasma in tokamak

σ(x, y) = 1 x = 2 z + ¯ z smooth in Ω ν(z, ¯ z) = z + ¯ z − 2 z + ¯ z + 2 Complete families of solutions to (u) and (f):

uk , Hνuk known

  • Bessel-exponentials Ω ≃ D0

[FLPS], also polynomials?

  • toroidal harmonics for Ω ≃ A0

[F], computations (BEP) associated Legendre functions In place of Fourier bases for σ = 1 or ν = 0 More about constructive issues, Toeplitz operators, algorithms GT-EP

slide-21
SLIDE 21

Toroidal harmonics

a −a

x y

M

η

A B

Toroidal coordinates τ = log MA MB

  • , η =

AMB

x y

τ = 0 τ = ∞

η = π 2 η = 3π 2

a

Annulus ≃ A0 between circles τ = cst

slide-22
SLIDE 22

Toroidal harmonics

Complete family T = (uj(τ, η))j≥0 in L2(∂A0)

(τ = cst) [F]

uj(τ, η) = a sinh τ √cosh τ − cos η    P1

j− 1 2

(cosh τ) Q1

j− 1 2

(cosh τ)    cos jη sin jη

  • x =

a sinh τ cosh τ − cos η , y = a sin η cosh τ − cos η

solutions to div ( 1

x ∇uj) = 0

P1

j− 1 2

, Q1

j− 1 2

associated Legendre functions

explicit σ-harmonic conjugate functions vj = Hν uj

  • n P0

j− 1 2

, Q0

j− 1 2

, div (x∇vj ) = 0

slide-23
SLIDE 23

Plasma in tokamak

From measurements of u, σ∂nu on outer boundary Γe, find level line Γp of associated solution u to (u), tangent to limitor Γl

Take a first such Γp,0 expand u on Γe, compute max u on limitor

Data transmission Γe Γp,0: u, σ∂nu on I = Γe u, σ∂nu on J0 = Γp,0, u in Ω0 Cauchy boundary inverse problem in Ω0 solve (BEP)

u , Bρ φI , Bt = ∂θv = σ∂nu ψI Cauchy data Φ on Γe constraint Re f∗ − cL2(J0) ≤ M small, c constant Free boundary problem Γp: u, ∂nu on Γl Γp,1 : {u = maxΓl u} iterate 1st step Γp, last closed level line tangent to Γl with shape optimization [Fischer-Privat]

slide-24
SLIDE 24

Approximation on Γe

Of given smooth data u, ∂nu by toroidal harmonics expansions

pi/2 pi 3*pi/2 2*pi −0.15 −0.1 −0.05 0.05 0.1 0.15 0.2 0.25 η H− ν (˜ v3

to,mo)

− uto,mo + L(uto,mo) pi/2 pi 3*pi/2 2*pi −0.95 −0.9 −0.85 −0.8 −0.75 −0.7 −0.65 −0.6 −0.55 η vto,mo ˜ v3

to,mo

slide-25
SLIDE 25

Plasma boundary recovery

pi/2 pi 3pi/2 2pi 0.65 0.7 0.75 0.8 0.85 0.9 Θ ρ(m) Γ(1),10

p

Γ(1),14

p

Γ(1),18

p

ΓEF IT

p

LIM APOLO

slide-26
SLIDE 26

Plasma boundary recovery

Poloidal section of tokamak Tore Supra

1.6 1.8 2 2.2 2.4 2.6 2.8 3 3.2 −0.8 −0.6 −0.4 −0.2 0.2 0.4 0.6 0.8 1.5 2 2.5 3 −1 −0.8 −0.6 −0.4 −0.2 0.2 0.4 0.6 0.8 1 R(m) Z(m) Γ(1),14

p

ΓEF IT

p

LIM APOLO

Reconstruction of plasma boundary Γp from measurements ◦ of poloidal flux u and × of magnetic field ∂nu on Γe with series of toroidal harmonics (18 terms)

slide-27
SLIDE 27

Application to plasma shaping in a tokamak

Tore Supra (CEA-IRFM Cadarache) magnetic field B, flux u

slide-28
SLIDE 28

Conclusion

Work in progress:

... or to be done?

  • More about generalized Hardy classes Hp

ν

factorization, operators; p = 1, ∞? density of traces for Ω = A; reproducing kernel in H2

ν?

extremal problems

  • min. w.r.t. c in constraint Re f − cL2(J)

solutions w = esF to related ¯ ∂w = α¯ w

α = ¯ ∂ log σ1/2

  • Other elliptic operators (and relat-ed/-ing PDEs)?

+ time t?

Schr¨

  • dinger ∆w ≃ |α|2w + (∂α)w, 3D Laplace + symmetry properties → 2D conductivity (u)
  • Stability estimates

unique continuation for (u) and Schr¨

  • dinger eq.?

Runge properties EIT issues

  • In higher dimensions?

H2

ν ↔ σ−1/2∇u?

  • Non smooth conductivity σ (or coefficients ν, α)?

anisotropic (matrix-valued)? (up to now, R-valued H¨

  • lder smooth σ, r > 2, in Hp

ν(Ω), p > r/(r − 1))

Also, geometrical issues: Bernoulli type (free boundary) problems

  • ther tokamaks, ITER: non smooth boundary (X point)
slide-29
SLIDE 29

Main references

[AP] Astala, P¨ aiv¨ arinta, Calder´

  • n inverse conductivity problem

(2006) [BLF] Baratchart, Fischer, Leblond, Dirichlet/Neumann problems and Hardy classes for the planar conductivity equation (subm.) [BLP] Baratchart, Leblond, Partington, Hardy approximation on subsets of the circle (1996-2000) [BLRR] Baratchart, Leblond, Rigat, Russ, Hardy spaces of the conjugate Beltrami equation (2010) [BN] Bers, Nirenberg, On a representation theorem for linear elliptic systems with discontinuous coeff. and appl. (1954) [FLPS] Fischer, Leblond, Partington, Sincich, BEP in Hardy spaces for the conjugate Beltrami equation in simply conn. dom. (2011) [SU] Sylvester, Uhlmann, A global uniqueness theorem for an inverse boundary value problem (1987)

and Astala, Iwaniec, Martin (2008), Kravchenko (2009), Vekua (1962), ...