Some Old and Some New Results in Inverse Obstacle Scattering Rainer - - PowerPoint PPT Presentation

some old and some new results in inverse obstacle
SMART_READER_LITE
LIVE PREVIEW

Some Old and Some New Results in Inverse Obstacle Scattering Rainer - - PowerPoint PPT Presentation

Some Old and Some New Results in Inverse Obstacle Scattering Rainer Kress, Gttingen Analysis and Numerics of Acoustic and Electromagnetic Problems Linz, October 2016 Rainer Kress Some Old and Some New Results in Inverse Obstacle Scattering


slide-1
SLIDE 1

Some Old and Some New Results in Inverse Obstacle Scattering

Rainer Kress, Göttingen Analysis and Numerics

  • f Acoustic and Electromagnetic Problems

Linz, October 2016

Rainer Kress Some Old and Some New Results in Inverse Obstacle Scattering

slide-2
SLIDE 2

Inverse obstacle scattering

ν

❳❳ ❳ ③ ✲ ✄ ✄✄ ✄

ui = eik x·d us, u = ui + us D ∆u + k2u = 0 in R3 \ ¯ D u = 0

  • n ∂D

∂us ∂r − ikus = o 1 r

  • ,

r = |x| → ∞ us(x) = eik|x| |x|

  • u∞

x |x|

  • + O

1 |x|

  • ,

|x| → ∞ Given: Far field u∞ for one (or several) incident plane wave Find: Shape and location of scatterer D

  • 1. Uniqueness, i.e., identifiability
  • 2. Reconstruction algorithms

Rainer Kress Some Old and Some New Results in Inverse Obstacle Scattering

slide-3
SLIDE 3

Outline

1

Two old uniqueness proofs

2

Two new uniqueness results for the generalized impedance boundary condition

3

A recent iterative reconstruction algorithm for the generalized impedance boundary condition

4

Factorization method and transmission eigenvalues

Rainer Kress Some Old and Some New Results in Inverse Obstacle Scattering

slide-4
SLIDE 4

Uniqueness, i.e., identifiability

Rellich’s Lemma: u∞ = 0

  • n S2

⇔ us = 0 in R3 \ D

Rainer Kress Some Old and Some New Results in Inverse Obstacle Scattering

slide-5
SLIDE 5

Uniqueness, i.e., identifiability

Rellich’s Lemma: u∞ = 0

  • n S2

⇔ us = 0 in R3 \ D Far field u∞ uniquely determines total field u = ui + us D u = 0 ui us

❳❳ ❳ ③ ✲ ✄ ✄✄ ✄

Question of uniqueness: ⇔ Existence of additional closed surfaces with u = 0

Rainer Kress Some Old and Some New Results in Inverse Obstacle Scattering

slide-6
SLIDE 6

Uniqueness, i.e., identifiability

Question of uniqueness: ⇔ Existence of additional closed surfaces on which u = 0 D u = 0 ui us

❳❳ ❳ ③ ✲ ✄ ✄✄ ✄

u = 0 No!!

Rainer Kress Some Old and Some New Results in Inverse Obstacle Scattering

slide-7
SLIDE 7

Uniqueness, i.e., identifiability

Question of uniqueness: ⇔ Existence of additional closed surfaces on which u = 0 D u = 0 ui us

❳❳ ❳ ③ ✲ ✄ ✄✄ ✄

u = 0 Do not know???

Rainer Kress Some Old and Some New Results in Inverse Obstacle Scattering

slide-8
SLIDE 8

Schiffer’s theorem

Theorem (Schiffer ≈ 1960) For a sound-soft scatterer, assume that u∞,1(ˆ x, d) = u∞,2(ˆ x, d) for all observation directions ˆ x and all incident directions d. Then D1 = D2.

D1 D2

Rainer Kress Some Old and Some New Results in Inverse Obstacle Scattering

slide-9
SLIDE 9

Schiffer’s proof

D1 D2 D∗ = unbounded component

  • f R3 \ (D1 ∪ D2)

us

1(· , d) = us 2(· , d)

in D∗

Rainer Kress Some Old and Some New Results in Inverse Obstacle Scattering

slide-10
SLIDE 10

Schiffer’s proof

D1 D2 D∗ = unbounded component

  • f R3 \ (D1 ∪ D2)

us

1(· , d) = us 2(· , d)

in D∗ u1(x, d) = eik x·d + us

1(x, d)

In shaded domain: △u1 + k2u1 = 0 On boundary: u1 = 0 {u1(· , d) : d ∈ S2} linearly independent

Rainer Kress Some Old and Some New Results in Inverse Obstacle Scattering

slide-11
SLIDE 11

Schiffer’s proof

D1 D2 D∗ = unbounded component

  • f R3 \ (D1 ∪ D2)

us

1(· , d) = us 2(· , d)

in D∗ u1(x, d) = eik x·d + us

1(x, d)

In shaded domain: △u1 + k2u1 = 0 On boundary: u1 = 0 {u1(· , d) : d ∈ S2} linearly independent Lax and Philipps 1967

Rainer Kress Some Old and Some New Results in Inverse Obstacle Scattering

slide-12
SLIDE 12

Schiffer’s proof

Correct shaded domain: (R3 \ D∗) \ D1 D1 D2 D∗ Incorrect shaded domain: D2 \ (D1 ∩ D2) D1 D2

Rainer Kress Some Old and Some New Results in Inverse Obstacle Scattering

slide-13
SLIDE 13

Schiffer’s proof

Schiffer’s proof does not work for other boundary conditions! D1 D2 Finite multiplicity of eigenvalues is based on the compact embedding of H1

0(Ω) into L2(Ω) for the shaded domain Ω.

However, in general, the embedding H1(Ω) into L2(Ω) is not compact.

Rainer Kress Some Old and Some New Results in Inverse Obstacle Scattering

slide-14
SLIDE 14

Other boundary conditions

u =

  • n ∂D,

sound-soft ∂u ∂ν =

  • n ∂D,

sound-hard ∂u ∂ν + ikλ =

  • n ∂D,

impedance, ℜλ ≥ 0

Rainer Kress Some Old and Some New Results in Inverse Obstacle Scattering

slide-15
SLIDE 15

Uniqueness of obstacle plus boundary condition

Theorem (Kirsch, K. 1992) Assume that u∞,1(ˆ x, d) = u∞,2(ˆ x, d) for all observation directions ˆ x and all incident directions d. Then D1 = D2 and B1 = B2.

D1 B1 D2 B2

Rainer Kress Some Old and Some New Results in Inverse Obstacle Scattering

slide-16
SLIDE 16

Idea of proof

z∗ x∗ D wi(x, z) = eik|x−z| |x − z| = incident field, point source ws(x, z) = scattered field, w∞(ˆ x, z) = far field

Rainer Kress Some Old and Some New Results in Inverse Obstacle Scattering

slide-17
SLIDE 17

Idea of proof

z∗ x∗ D wi(x, z) = eik|x−z| |x − z| = incident field, point source ws(x, z) = scattered field, w∞(ˆ x, z) = far field reciprocity: u∞(ˆ x, d) = u∞(−d, −ˆ x), ws(x, z) = ws(z, x) mixed reciprocity: us(z, d) = w∞(−d, z)

Rainer Kress Some Old and Some New Results in Inverse Obstacle Scattering

slide-18
SLIDE 18

Idea of proof

mixed reciprocity: us(z, d) = w∞(−d, z) D1 D2 D∗ x⋆ ⋆z B1 B2

Rainer Kress Some Old and Some New Results in Inverse Obstacle Scattering

slide-19
SLIDE 19

Idea of proof

mixed reciprocity: us(z, d) = w∞(−d, z) D1 D2 D∗ x⋆ ⋆z B1 B2 u∞,1(ˆ x, d) = u∞,2(ˆ x, d) for |ˆ x| = |d| = 1 us

1(z, d)

= us

2(z, d)

for z ∈ D∗, |d| = 1 w∞,1(d, z) = w∞,2(d, z) for z ∈ D∗, |d| = 1 ws

1(x, z)

= ws

2(x, z)

for x, z ∈ D∗

Rainer Kress Some Old and Some New Results in Inverse Obstacle Scattering

slide-20
SLIDE 20

Idea of proof

D1 D2 D∗ x∗ ⋆ ⋆z B1 B2 ws

1(x, z) = ws 2(x, z)

for x, z ∈ D∗

Rainer Kress Some Old and Some New Results in Inverse Obstacle Scattering

slide-21
SLIDE 21

Idea of proof

D1 D2 D∗ x∗ ⋆ ⋆z B1 B2 ws

1(x, z) = ws 2(x, z)

for x, z ∈ D∗ lim

z→x∗ B1ws 1(x∗, z) = ∞,

lim

z→x∗ B1ws 2(x∗, z) = finite

⇒ D1 = D2 Holmgren’s theorem yields B1 = B2

Rainer Kress Some Old and Some New Results in Inverse Obstacle Scattering

slide-22
SLIDE 22

Idea of proof

D1 D2 D∗ x∗ ⋆ ⋆z B1 B2 ws

1(x, z) = ws 2(x, z)

for x, z ∈ D∗ lim

z→x∗ B1ws 1(x∗, z) = ∞,

lim

z→x∗ B1ws 2(x∗, z) = finite

⇒ D1 = D2 Holmgren’s theorem yields B1 = B2 Use of mixed reciprocity: Potthast 1999

Rainer Kress Some Old and Some New Results in Inverse Obstacle Scattering

slide-23
SLIDE 23

Generalized impedance condition

∂u ∂ν + ik (λu − div∂D µ grad∂D u) = 0

  • n ∂D,

ℜλ, ℜµ ≥ 0 Theorem (Bourgeois, Chaulet, Haddar 2012) Both the shape and the impedance functions of a scattering

  • bstacle with generalized impedance condition are uniquely

determined by the far field patterns for an infinite number of incident waves with distinct incident directions and one fixed wave number.

Rainer Kress Some Old and Some New Results in Inverse Obstacle Scattering

slide-24
SLIDE 24

Uniqueness for impedance functions in R2

∂u ∂ν + ik

  • λu − d

ds µ du ds

  • = 0
  • n ∂D

Theorem (Cakoni, K. 2013) In two dimensions, for a given shape ∂D, three far field patterns corresponding to the scattering of three plane waves with different incident directions uniquely determine the impedance functions µ and λ.

Rainer Kress Some Old and Some New Results in Inverse Obstacle Scattering

slide-25
SLIDE 25

Uniqueness for impedance functions in R2

∂u ∂ν + ik

  • λu − d

ds µ du ds

  • = 0
  • n ∂D

Theorem (Cakoni, K. 2013) In two dimensions, for a given shape ∂D, three far field patterns corresponding to the scattering of three plane waves with different incident directions uniquely determine the impedance functions µ and λ. ik d ds µ

  • u1

du2 ds − u2 du1 ds

  • W(u1, u2)

= u1 ∂u2 ∂ν − u2 ∂u1 ∂ν

  • n ∂D

⇒ α3 W(u1, u2) = α1 W(u2, u3) = α2 W(u3, u1)

Rainer Kress Some Old and Some New Results in Inverse Obstacle Scattering

slide-26
SLIDE 26

Iterative methods versus qualitative methods

Iterative methods: Reformulate inverse problem as nonlinear ill-posed operator equation. Solve by iteration methods such as regularized Newton methods, Landweber iterations or conjugate gradient methods

Rainer Kress Some Old and Some New Results in Inverse Obstacle Scattering

slide-27
SLIDE 27

Iterative methods versus qualitative methods

Iterative methods: Reformulate inverse problem as nonlinear ill-posed operator equation. Solve by iteration methods such as regularized Newton methods, Landweber iterations or conjugate gradient methods Qualitative methods: Develop criterium in terms of behaviour

  • f certain ill-posed linear integral equations that decide on

whether a point lies inside or outside the scatterer. Linear sampling, factorization, probe methods, etc

Rainer Kress Some Old and Some New Results in Inverse Obstacle Scattering

slide-28
SLIDE 28

Single-layer approach

Assume that k2 is not a Dirichlet eigenvalue for −∆ in D us(x) = 1 4π

  • ∂D

eik|x−y| |x − y| ϕ(y) ds(y), x ∈ R3 \ D

Rainer Kress Some Old and Some New Results in Inverse Obstacle Scattering

slide-29
SLIDE 29

Single-layer approach

Assume that k2 is not a Dirichlet eigenvalue for −∆ in D us(x) = 1 4π

  • ∂D

eik|x−y| |x − y| ϕ(y) ds(y), x ∈ R3 \ D Data equation u∞(ˆ x) = 1 4π

  • ∂D

e−ik ˆ

x·y ϕ(y) ds(y),

ˆ x ∈ S2 Field equation ui(x) = − 1 4π

  • ∂D

eik|x−y| |x − y| ϕ(y) ds(y), x ∈ ∂D Inverse problem is equivalent to solving the two equations for the two unknowns ∂D and ϕ

Rainer Kress Some Old and Some New Results in Inverse Obstacle Scattering

slide-30
SLIDE 30

Parameterized equations

∂D = {p(ˆ x) : ˆ x ∈ S2}, set ψ = ϕ ◦ p Define S, S∞ : C2(S2, R3) × L2(S2, C) → L2(S2, C) by S(p, ψ)(ˆ x) := 1 4π

  • S2

eik|p(ˆ

x)−p(ˆ y)|

|p(ˆ x) − p(ˆ y)| J(ˆ y) ψ(ˆ y) ds(ˆ y), ˆ x ∈ S2 S∞(p, ψ)(ˆ x) := 1 4π

  • S2 e−ik ˆ

x·p(ˆ y)ψ(ˆ

y) J(ˆ y) ds(ˆ y), ˆ x ∈ S2

Rainer Kress Some Old and Some New Results in Inverse Obstacle Scattering

slide-31
SLIDE 31

Parameterized equations

∂D = {p(ˆ x) : ˆ x ∈ S2}, set ψ = ϕ ◦ p Define S, S∞ : C2(S2, R3) × L2(S2, C) → L2(S2, C) by S(p, ψ)(ˆ x) := 1 4π

  • S2

eik|p(ˆ

x)−p(ˆ y)|

|p(ˆ x) − p(ˆ y)| J(ˆ y) ψ(ˆ y) ds(ˆ y), ˆ x ∈ S2 S∞(p, ψ)(ˆ x) := 1 4π

  • S2 e−ik ˆ

x·p(ˆ y)ψ(ˆ

y) J(ˆ y) ds(ˆ y), ˆ x ∈ S2 Data equation S∞(p, ψ) = u∞ Field equation S(p, ψ) = −ui ◦ p Linearize both equations with respect to p and ψ (and iterate). Ivanyshyn, K., Rundell, Serranho 2005 . . .

Rainer Kress Some Old and Some New Results in Inverse Obstacle Scattering

slide-32
SLIDE 32

Example

One incident wave with k = 1 Reconstruction of a cushion for 5% noise

Rainer Kress Some Old and Some New Results in Inverse Obstacle Scattering

slide-33
SLIDE 33

Linearization of the data equation

Step 1. Given an approximation for p solve the field equation S(p, ψ) = −ui ◦ p for the density ψ. Step 2. Keeping ψ fixed, linearize the data equation S∞(p, ψ) = u∞ to obtain S′

∞(p, ψ; q) = u∞ − S∞(p, ψ)

and solve for q to update p into p + q. Johansson, Sleeman 2007

Rainer Kress Some Old and Some New Results in Inverse Obstacle Scattering

slide-34
SLIDE 34

The two-dimensional case

∂D = {z(t) : 0 ≤ t ≤ 2π} (Sψ)(t) := i 4 2π H(1)

0 (k|z(t)−z(τ)|) |z′(τ)| ψ(τ) dτ,

t ∈ [0, 2π] (S∞ψ)(ˆ x) := ei π

4

√ 8πk 2π e−ik ˆ

x·yψ(τ) dτ,

ˆ x ∈ S1

Rainer Kress Some Old and Some New Results in Inverse Obstacle Scattering

slide-35
SLIDE 35

The two-dimensional case

∂D = {z(t) : 0 ≤ t ≤ 2π} (Sψ)(t) := i 4 2π H(1)

0 (k|z(t)−z(τ)|) |z′(τ)| ψ(τ) dτ,

t ∈ [0, 2π] (S∞ψ)(ˆ x) := ei π

4

√ 8πk 2π e−ik ˆ

x·yψ(τ) dτ,

ˆ x ∈ S1 Data equation S∞ψ = u∞ Field equation Sψ = −ui ◦ z

Rainer Kress Some Old and Some New Results in Inverse Obstacle Scattering

slide-36
SLIDE 36

The two-dimensional case, impedance condition

(Nψ)(t) := ik 4 2π ν(z(t)) · [z(τ) − z(t)] |z(t) − z(τ)| H(1)

1 (k|z(t)−z(τ)|) |z′(τ)| ψ(τ) dτ

Rainer Kress Some Old and Some New Results in Inverse Obstacle Scattering

slide-37
SLIDE 37

The two-dimensional case, impedance condition

(Nψ)(t) := ik 4 2π ν(z(t)) · [z(τ) − z(t)] |z(t) − z(τ)| H(1)

1 (k|z(t)−z(τ)|) |z′(τ)| ψ(τ) dτ

Data equation S∞ψ = u∞ Field equation 1 2 ψ − Nψ − ik

  • λ Sψ + 1

|z′| d dt µ |z′| d dt Sψ

  • = g ◦ z

where g := ∂ui ∂ν

  • ∂D

+ ik

  • λ − d

ds µ d ds

  • ui|∂D

Inverse problem is equivalent to solving the two equations for the two unknowns z and ϕ

Rainer Kress Some Old and Some New Results in Inverse Obstacle Scattering

slide-38
SLIDE 38

The two-dimensional case, impedance condition

(Sψ)(t) := i 4 2π H(1)

0 (k|z(t) − z(τ)|) |z′(τ)| ψ(τ) dτ

(Nψ)(t) := ik 4 2π ν(z(t)) · [z(τ) − z(t)] |z(t) − z(τ)| H(1)

1 (k|z(t)−z(τ)|) |z′(τ)| ψ(τ) dτ

Field equation 1 2 ψ − Nψ − ik

  • λ Sψ + 1

|z′| d dt µ |z′| d dt Sψ

= g ◦ z

Rainer Kress Some Old and Some New Results in Inverse Obstacle Scattering

slide-39
SLIDE 39

The two-dimensional case, impedance condition

(Sψ)(t) := i 4 2π H(1)

0 (k|z(t) − z(τ)|) |z′(τ)| ψ(τ) dτ

(Nψ)(t) := ik 4 2π ν(z(t)) · [z(τ) − z(t)] |z(t) − z(τ)| H(1)

1 (k|z(t)−z(τ)|) |z′(τ)| ψ(τ) dτ

Field equation 1 2 ψ − Nψ − ik

  • λ Sψ + 1

|z′| d dt µ |z′| d dt Sψ

= g ◦ z A : H1/2[0, 2π] → H−1/2[0, 2π] is Fredholm with index zero (provided |µ| > 0). Solve via trigonometric quadrature and differentiation with spectral convergence Cakoni, K. 2013 for Laplace, K. 2016 for Helmholtz

Rainer Kress Some Old and Some New Results in Inverse Obstacle Scattering

slide-40
SLIDE 40

Example

The impedance functions

1 2 3 4 5 6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5 7 6 Rainer Kress Some Old and Some New Results in Inverse Obstacle Scattering

slide-41
SLIDE 41

Example

Two incident waves with k = 2

  • 1
  • 0.5

0.5 1

  • 1
  • 0.8
  • 0.6
  • 0.4
  • 0.2

0.2 0.4 0.6 0.8 1 reconstructed exact initial guess

  • 1
  • 0.5

0.5 1

  • 1
  • 0.8
  • 0.6
  • 0.4
  • 0.2

0.2 0.4 0.6 0.8 1 reconstructed exact initial guess

Reconstruction of an apple for exact data and for 5% noise

Rainer Kress Some Old and Some New Results in Inverse Obstacle Scattering

slide-42
SLIDE 42

Example

Two incident waves with k = 2

  • 1
  • 0.5

0.5 1

  • 1
  • 0.8
  • 0.6
  • 0.4
  • 0.2

0.2 0.4 0.6 0.8 1 reconstructed exact initial guess

  • 1
  • 0.5

0.5 1

  • 1
  • 0.8
  • 0.6
  • 0.4
  • 0.2

0.2 0.4 0.6 0.8 1 reconstructed exact initial guess

Reconstruction of a peanut for exact data and for 5% noise

Rainer Kress Some Old and Some New Results in Inverse Obstacle Scattering

slide-43
SLIDE 43

Factorization method for obstacle scattering

Assume that k2 is not a Dirichlet eigenvalue of −∆ for D. Define far field operator F : L2(S2) → L2(S2) by Fg(ˆ x) :=

  • S2 u∞(ˆ

x, d)g(d) ds(d), ˆ x ∈ S2 and recall the far field of a point source w∞(ˆ x, z) = e−ik ˆ

x·z,

ˆ x ∈ S2, z ∈ R3

Rainer Kress Some Old and Some New Results in Inverse Obstacle Scattering

slide-44
SLIDE 44

Factorization method for obstacle scattering

Assume that k2 is not a Dirichlet eigenvalue of −∆ for D. Define far field operator F : L2(S2) → L2(S2) by Fg(ˆ x) :=

  • S2 u∞(ˆ

x, d)g(d) ds(d), ˆ x ∈ S2 and recall the far field of a point source w∞(ˆ x, z) = e−ik ˆ

x·z,

ˆ x ∈ S2, z ∈ R3 Theorem (Kirsch 1998) The equation (F ∗F)1/4g(· , z) = w∞(· , z) is solvable in L2(S2) if and only if z ∈ D. Use Picard criterium for numerical implementation

Rainer Kress Some Old and Some New Results in Inverse Obstacle Scattering

slide-45
SLIDE 45

Factorization method for inhomogeneous medium

∆u + k2nu = 0 in R3 u = ui + us ∂us ∂r − ikus = o 1 r

  • ,

r = |x| → ∞ where n > 0 is the refractive index with supp(n − 1) = D Theorem (Kirsch 1999) The equation (F ∗F)1/4g(· , z) = w∞(· , z) is solvable in L2(S2) if and only if z ∈ D. Here k is required not to be an interior transmission eigenvalue

Rainer Kress Some Old and Some New Results in Inverse Obstacle Scattering

slide-46
SLIDE 46

Transmission eigenvalues

k is called a transmission eigenvalue it there exist nontrivial functions v, w ∈ H2(D) such that ∆v + k2v = 0, ∆w + k2nv = 0 in D and v = w, ∂v ∂ν = ∂w ∂ν

  • n ∂D

Rainer Kress Some Old and Some New Results in Inverse Obstacle Scattering

slide-47
SLIDE 47

Transmission eigenvalues

k is called a transmission eigenvalue it there exist nontrivial functions v, w ∈ H2(D) such that ∆v + k2v = 0, ∆w + k2nv = 0 in D and v = w, ∂v ∂ν = ∂w ∂ν

  • n ∂D

Can boundary integral equations be used to analyze and compute transmission eigenvalues in the case n = const in D?

Rainer Kress Some Old and Some New Results in Inverse Obstacle Scattering

slide-48
SLIDE 48

Transmission eigenvalues

k is called a transmission eigenvalue it there exist nontrivial functions v, w ∈ H2(D) such that ∆v + k2v = 0, ∆w + k2nv = 0 in D and v = w, ∂v ∂ν = ∂w ∂ν

  • n ∂D

Can boundary integral equations be used to analyze and compute transmission eigenvalues in the case n = const in D? Cossonnière, Haddar 2013 System of two integral equations for v and ∂νv from Green’s representation formula

Rainer Kress Some Old and Some New Results in Inverse Obstacle Scattering

slide-49
SLIDE 49

Transmission eigenvalues

Alternative approach using the Dirichlet-to-Neumann operator DtNk,n : ϕ → ∂u ∂ν , where u is the unique solution to ∆u + k2nu = in D u = ϕ

  • n ∂D

assuming that k2 is not an eigenvalue for this problem. If we further assume that k2 is not an eigenvalue for the case when n = 1, then k is a transmission eigenvalue if and only if the kernel of the operator M(k) := DtNk,n − DtNk,1 is nontrivial.

Rainer Kress Some Old and Some New Results in Inverse Obstacle Scattering

slide-50
SLIDE 50

Transmission eigenvalues

Theorem (Cakoni, K. 2016) Let κ > 0 and κn = κ√n. Then (κ2 − κ2

n)M(iκ) : H− 1

2 (∂D) → H 1 2 (∂D)

is coercive. Further let kn = k√n. Then the operator M(k) + k2 − k2

n

|k|2 − |kn|2 M(i|k|) : H− 1

2 (∂D) → H 1 2 (∂D)

is compact. Idea of proof: Write Mk = (I + Nk)S−1

k

− (I + Nkn)S−1

kn and use

Green’s integral theorem and mapping properties for difference

  • f the two boundary integral operators.

Rainer Kress Some Old and Some New Results in Inverse Obstacle Scattering

slide-51
SLIDE 51

Transmission eigenvalues

Corollary M(k) : H− 1

2 (∂D) → H 1 2 (∂D) is a Fredholm operator with index

zero and analytic in C \ E E is the set of all positive k such that k2 or k2

n is a Dirichlet

eigenvalue for −∆ in D. Can use the effective numerical method for non-linear eigenvalue problems proposed by Beyn 2012 to compute transmission eigenvalues. The drawback that Dirichlet eigenvalues have to be excluded can be remedied by an analogous (but more involved) analysis using a Robin-to-Dirichlet map.

Rainer Kress Some Old and Some New Results in Inverse Obstacle Scattering

slide-52
SLIDE 52

Numerical example

Transmission eigenvalues for ellipse with major axis a = 1

n = 2 b = 1 b = 0.8 b = 0.5 b = 0.3 7.37512 7.63521 8.99951 12.32237 7.39666 8.13114 9.08377 12.34655 7.39666 8.43084 10.97143 16.48683 7.98435 8.51387 11.03990 16.49174 7.98435 8.95319 12.30361 17.01368 8.02926 9.00645 12.31572 17.14467 8.02926 9.24582 13.40175 19.44708 8.21647 9.34204 13.41828 19.44871 8.21647 9.47369 14.21896 19.92760 8.67540 9.76915 14.47108 19.94432

Rainer Kress Some Old and Some New Results in Inverse Obstacle Scattering

slide-53
SLIDE 53

References

Cakoni, F . and Kress, R.: Integral equation methods for the inverse obstacle problem with generalized impedance boundary condition. Inverse Problems 29, 015005 (2013). Cakoni, F ., Hu, Y. and Kress, R.: Simultaneous reconstruction of shape and generalized impedance functions in electrostatic imaging. Inverse Problems 30, 105009 (2014). Cakoni, F . and Kress, R.: A boundary integral equation method for the transmission eigenvalue problem. Applicable Analysis (2016). Kress, R.: Integral equation methods in inverse obstacle scattering with a generalized impedance boundary condition. Submitted

Rainer Kress Some Old and Some New Results in Inverse Obstacle Scattering

slide-54
SLIDE 54

Thanks

THANKS

Rainer Kress Some Old and Some New Results in Inverse Obstacle Scattering