Some issues on Electrical Impedance Tomography with complex coefficient

Similar documents
A CALDERÓN PROBLEM WITH FREQUENCY-DIFFERENTIAL DATA IN DISPERSIVE MEDIA. has a unique solution. The Dirichlet-to-Neumann map

Determination of thin elastic inclusions from boundary measurements.

is measured (n is the unit outward normal) at the boundary, to define the Dirichlet-

Complex geometrical optics solutions for Lipschitz conductivities

Stability and instability in inverse problems

On uniqueness in the inverse conductivity problem with local data

Put Paper Number Here

Laplace s Equation. Chapter Mean Value Formulas

Monotonicity arguments in electrical impedance tomography

Reconstructing conductivities with boundary corrected D-bar method

Inverse Electromagnetic Problems

Calderón s inverse problem in 2D Electrical Impedance Tomography

Inverse Transport Problems and Applications. II. Optical Tomography and Clear Layers. Guillaume Bal

Electric potentials with localized divergence properties

Approximation of inverse boundary value problems by phase-field methods

Non-uniqueness result for a hybrid inverse problem

Non-uniqueness result for a hybrid inverse problem

Stability Estimates in the Inverse Transmission Scattering Problem

Proceedings of the 5th International Conference on Inverse Problems in Engineering: Theory and Practice, Cambridge, UK, 11-15th July 2005

AN INVERSE PROBLEM FOR THE WAVE EQUATION WITH A TIME DEPENDENT COEFFICIENT

A uniqueness result and image reconstruction of the orthotropic conductivity in magnetic resonance electrical impedance tomography

A SHARP STABILITY ESTIMATE IN TENSOR TOMOGRAPHY

Nonlinear aspects of Calderón-Zygmund theory

AALBORG UNIVERSITY. A new direct method for reconstructing isotropic conductivities in the plane. Department of Mathematical Sciences.

Inverse parameter identification problems

Enhanced resolution in structured media

Homogenization and error estimates of free boundary velocities in periodic media

Inverse conductivity problem for inaccurate measurements

Fast shape-reconstruction in electrical impedance tomography

Reconstructing inclusions from Electrostatic Data

AALBORG UNIVERSITY. The Calderón problem with partial data for less smooth conductivities. Kim Knudsen. Department of Mathematical Sciences

Minimization problems on the Hardy-Sobolev inequality

A Direct Method for reconstructing inclusions from Electrostatic Data

Resistor Networks and Optimal Grids for Electrical Impedance Tomography with Partial Boundary Measurements

Electrical impedance spectroscopy-based nondestructive testing for imaging defects in concrete structures arxiv: v1 [math.

Unique Continuation Properties and Quantitative Estimates of Unique Continuation for Parabolic Equations

ANGKANA RÜLAND AND MIKKO SALO

On the definition and properties of p-harmonious functions

Lecture No 1 Introduction to Diffusion equations The heat equat

COMPLEX SPHERICAL WAVES AND INVERSE PROBLEMS IN UNBOUNDED DOMAINS

Piecewise Smooth Solutions to the Burgers-Hilbert Equation

Iterative regularization of nonlinear ill-posed problems in Banach space

Recent progress on the factorization method for electrical impedance tomography

A local estimate from Radon transform and stability of Inverse EIT with partial data

Novel tomography techniques and parameter identification problems

Potential Analysis meets Geometric Measure Theory

Inverse Gravimetry Problem

Reconstruction Scheme for Active Thermography

Green s Functions and Distributions

Detecting stochastic inclusions in electrical impedance tomography

Increasing stability in an inverse problem for the acoustic equation

R. M. Brown. 29 March 2008 / Regional AMS meeting in Baton Rouge. Department of Mathematics University of Kentucky. The mixed problem.

arxiv: v1 [math.na] 5 Jun 2018

LOCALIZED POTENTIALS IN ELECTRICAL IMPEDANCE TOMOGRAPHY. Bastian Gebauer. (Communicated by Otmar Scherzer)

Toward imaging modalities with high(er) resolution

Asymptotic Spectral Imaging

PROBING FOR INCLUSIONS IN HEAT CONDUCTIVE BODIES. Patricia Gaitan. Hiroshi Isozaki. Olivier Poisson. Samuli Siltanen.

Computational inversion based on complex geometrical optics solutions

SHARP BOUNDARY TRACE INEQUALITIES. 1. Introduction

Hyperbolic inverse problems and exact controllability

Regularization in Banach Space

DIRECT ELECTRICAL IMPEDANCE TOMOGRAPHY FOR NONSMOOTH CONDUCTIVITIES

TOPICS IN NONLINEAR ANALYSIS AND APPLICATIONS. Dipartimento di Matematica e Applicazioni Università di Milano Bicocca March 15-16, 2017

Survey of Inverse Problems For Hyperbolic PDEs

The Dirichlet problem for non-divergence parabolic equations with discontinuous in time coefficients in a wedge

ESTIMATES FOR ELLIPTIC HOMOGENIZATION PROBLEMS IN NONSMOOTH DOMAINS. Zhongwei Shen

The inverse conductivity problem with power densities in dimension n 2

MINIMAL SURFACES AND MINIMIZERS OF THE GINZBURG-LANDAU ENERGY

SYMMETRY RESULTS FOR PERTURBED PROBLEMS AND RELATED QUESTIONS. Massimo Grosi Filomena Pacella S. L. Yadava. 1. Introduction

The Factorization Method for the Reconstruction of Inclusions

MULTIPLE SOLUTIONS FOR CRITICAL ELLIPTIC PROBLEMS WITH FRACTIONAL LAPLACIAN

arxiv: v1 [math.ap] 21 Dec 2014

Simultaneous vs. non simultaneous blow-up

Besov regularity of solutions of the p-laplace equation

The Finite Difference Method for the Helmholtz Equation with Applications to Cloaking

REGULARITY RESULTS FOR THE EQUATION u 11 u 22 = Introduction

arxiv: v2 [math.ap] 9 Sep 2014

Mathematical analysis of the stationary Navier-Stokes equations

SEMILINEAR ELLIPTIC EQUATIONS WITH DEPENDENCE ON THE GRADIENT

Simultaneous reconstruction of outer boundary shape and conductivity distribution in electrical impedance tomography

arxiv: v1 [math.ap] 11 Jun 2007

Partial regularity for suitable weak solutions to Navier-Stokes equations

An inverse elliptic problem of medical optics with experimental data

STUDY OF THE BOUNDARY ELEMENT METHOD FOR THE DIRECT PROBLEM OF ELECTRICAL IMPEDANCE TOMOGRAPHY

On stable inversion of the attenuated Radon transform with half data Jan Boman. We shall consider weighted Radon transforms of the form

Estimation of transmission eigenvalues and the index of refraction from Cauchy data

Recovery-Based A Posteriori Error Estimation

S chauder Theory. x 2. = log( x 1 + x 2 ) + 1 ( x 1 + x 2 ) 2. ( 5) x 1 + x 2 x 1 + x 2. 2 = 2 x 1. x 1 x 2. 1 x 1.

The Method of Small Volume Expansions and Applications to Emerging Medical Imaging

Functions with orthogonal Hessian

EIT Reconstruction Notes

A new inversion method for dissipating electromagnetic problem

Heat equations with singular potentials: Hardy & Carleman inequalities, well-posedness & control

Liouville-type theorem for the Lamé system with singular coefficients

Partial regularity for fully nonlinear PDE

Symmetry of nonnegative solutions of elliptic equations via a result of Serrin

Detecting Interfaces in a Parabolic-Elliptic Problem

Frequency functions, monotonicity formulas, and the thin obstacle problem

ELECTROMAGNETIC SCATTERING FROM PERTURBED SURFACES. Katharine Ott Advisor: Irina Mitrea Department of Mathematics University of Virginia

arxiv: v1 [math.ap] 21 Dec 2018

Transcription:

Some issues on Electrical Impedance Tomography with complex coefficient Elisa Francini (Università di Firenze) in collaboration with Elena Beretta and Sergio Vessella E. Francini (Università di Firenze) EIT 1 / 42

Outline 1 Electrical Impedance Tomography with complex coefficient 2 Reconstruction from the DN map: a stability result. 3 Size estimate from a single boundary measurement. E. Francini (Università di Firenze) EIT 2 / 42

Electrical Impedance Tomography Electrical Impedance Tomography: Determine electrical conductivity and permittivity in the interior of a body by measurements of electric currents and voltages at the boundary of the body. Medical diagnostics: detection of pulmonary emboli, breast tumours, screening of organs in transplant surgery Nondestructive testing of materials: detection of corrosion or cracks Geophysical prospection: detection of underground mineral deposits E. Francini (Università di Firenze) EIT 3 / 42

Mathematical model γ : Ω R n C complex admittivity function γ = σ + iωɛ electrical conductivity frequency electrical permittivity σ λ 1 > 0 (dissipation of energy), γ λ The potential u H 1 (Ω, C) is a weak solution to equation div (γ u) = 0 in Ω. Boundary data: u Ω and γ u ν Ω. E. Francini (Università di Firenze) EIT 4 / 42

Mathematical model γ : Ω R n C complex admittivity function γ = σ + iωɛ electrical conductivity frequency electrical permittivity σ λ 1 > 0 (dissipation of energy), γ λ The potential u H 1 (Ω, C) is a weak solution to equation div (γ u) = 0 in Ω. Boundary data: u Ω and γ u ν Ω. E. Francini (Università di Firenze) EIT 4 / 42

Mathematical model γ : Ω R n C complex admittivity function γ = σ + iωɛ electrical conductivity frequency electrical permittivity σ λ 1 > 0 (dissipation of energy), γ λ The potential u H 1 (Ω, C) is a weak solution to equation div (γ u) = 0 in Ω. Boundary data: u Ω and γ u ν Ω. E. Francini (Università di Firenze) EIT 4 / 42

Part I Reconstruction from infinite boundary measurements E. Francini (Università di Firenze) EIT 5 / 42

DN map Given f H 1/2 ( Ω) there exists a unique weak solution u H 1 (Ω) of the problem { div (γ u) = 0 in Ω u = f on Ω. Consider the Dirichlet to Neumann map Λ γ : H 1/2 ( Ω) H 1/2 ( Ω) given by Λ γ f = γ u ν Ω, where ν is the exterior unit normal vector to Ω. EIT Inverse Problem: Determine γ from the knowledge of the Dirichlet-to Neumann map Λ γ E. Francini (Università di Firenze) EIT 6 / 42

DN map Given f H 1/2 ( Ω) there exists a unique weak solution u H 1 (Ω) of the problem { div (γ u) = 0 in Ω u = f on Ω. Consider the Dirichlet to Neumann map Λ γ : H 1/2 ( Ω) H 1/2 ( Ω) given by Λ γ f = γ u ν Ω, where ν is the exterior unit normal vector to Ω. EIT Inverse Problem: Determine γ from the knowledge of the Dirichlet-to Neumann map Λ γ E. Francini (Università di Firenze) EIT 6 / 42

DN map Given f H 1/2 ( Ω) there exists a unique weak solution u H 1 (Ω) of the problem { div (γ u) = 0 in Ω u = f on Ω. Consider the Dirichlet to Neumann map Λ γ : H 1/2 ( Ω) H 1/2 ( Ω) given by Λ γ f = γ u ν Ω, where ν is the exterior unit normal vector to Ω. EIT Inverse Problem: Determine γ from the knowledge of the Dirichlet-to Neumann map Λ γ E. Francini (Università di Firenze) EIT 6 / 42

Known results Uniqueness n 3 If γ is sufficiently smooth the results obtained by Sylvester and Uhlmann (1987, Ann. of Math.) for the conductivity case can be extended to the complex case. (Λ γ1 = Λ γ2 implies γ 1 = γ 2 ) n = 2 F.(2000, Inverse Problems) for small frequencies ω, Bukhgeim for arbitrary frequencies and smooth coefficients (2008, Inv. Ill Posed Problems). n 3, γ L (Ω) completely open n = 2 Astala and Päivärinta for real conductivities σ L (Ω) ( 2006, Ann. of Math.) E. Francini (Università di Firenze) EIT 7 / 42

Known results Stability n 3 Alessandrini (1988, Appl. Anal.) has proved for the conductivity case that if σ j W 2, (Ω) E for j = 1, 2, then σ 1 σ 2 L (Ω) ω( Λ σ1 Λ σ2 ) where ω(t) = C log t η. n = 2 Barcelo, Faraco and Ruiz (2007, Jour. Math. Pures Appl.) have proved for the conductivity case an analogous logarithmic estimate under the a priori assumption σ C α (Ω) E. Clop, Faraco, Ruiz (2009, IPI) for σ j W α,p, α > 0, j = 1, 2, and 1 < p < then where ω(t) = C log t η. σ 1 σ 2 L 2 (Ω) ω( Λ σ1 Λ σ2 ) E. Francini (Università di Firenze) EIT 8 / 42

A priori assumptions Mandache (2001, Inverse Problems) has proved that logarithmic stability is the best possible stability also using as a-priori assumption σ C k (Ω) E, k = 0, 1, 2... E. Francini (Università di Firenze) EIT 9 / 42

Strategy: Look for a-priori assumptions on γ that are physically relevant give rise to a better type of stability Assume: γ is piecewise constant on known domains. In the real case: Alessandrini and Vessella (2005, Adv. in Appl. Math.) E. Francini (Università di Firenze) EIT 10 / 42

Strategy: Look for a-priori assumptions on γ that are physically relevant give rise to a better type of stability Assume: γ is piecewise constant on known domains. In the real case: Alessandrini and Vessella (2005, Adv. in Appl. Math.) E. Francini (Università di Firenze) EIT 10 / 42

Strategy: Look for a-priori assumptions on γ that are physically relevant give rise to a better type of stability Assume: γ is piecewise constant on known domains. In the real case: Alessandrini and Vessella (2005, Adv. in Appl. Math.) E. Francini (Università di Firenze) EIT 10 / 42

Main assumptions (H1) Ω R n, bounded, Ω A, Lipschitz boundary (H2) γ(x) = N j=1 γ j1 Dj (x) unknown complex numbers known open sets (H3) D i D j =, i j, Lipschitz boundaries and N j=1 D j = Ω D 1 Ω contains an open flat portion Σ 1. For every j {2,..., N} there exists a chain of subdomains connecting D 1 to D j so that consecutive domains share a flat portion of boundary E. Francini (Università di Firenze) EIT 11 / 42

A possible configuration D j Σ 1 D 1 E. Francini (Università di Firenze) EIT 12 / 42

Chain D j Σ 1 D 1 E. Francini (Università di Firenze) EIT 13 / 42

[Beretta, F., Comm. PDE, 2011] E. Francini (Università di Firenze) EIT 14 / 42 Main result Theorem Let Ω satisfy assumption (H1). Let γ (k), k = 1, 2 be two complex piecewise constant functions of the form γ (k) (x) = N j=1 γ (k) j 1 Dj (x), where γ (k) satisfy for k = 1, 2 assumption (H2) and D j, j = 1,..., N satisfy assumption (H3). Then there exists a positive constant C such that γ (1) γ (2) L (Ω) C Λ γ (1) Λ γ (2) L(H 1/2 ( Ω),H 1/2 ( Ω)). The constant C depends on the Lipschitz constant of the boundaries, on the volume of Ω, on n, N and λ.

Main idea of the proof (uniqueness) Assume Λ γ (1) = Λ γ (2) ) ) div (γ (1) u 1 = 0 div (γ (2) u 2 = 0 in Ω 0 =< (Λ γ (1) Λ γ (2))u 1, u 2 >= We want to use this identity for singular solutions Ω ( γ (1) γ (2)) u 1 u 2 dx E. Francini (Università di Firenze) EIT 15 / 42

Main idea of the proof (uniqueness) Assume Λ γ (1) = Λ γ (2) ) ) div (γ (1) u 1 = 0 div (γ (2) u 2 = 0 in Ω 0 =< (Λ γ (1) Λ γ (2))u 1, u 2 >= We want to use this identity for singular solutions Ω ( γ (1) γ (2)) u 1 u 2 dx E. Francini (Università di Firenze) EIT 15 / 42

Main idea of the proof (uniqueness) Assume Λ γ (1) = Λ γ (2) ) ) div (γ (1) u 1 = 0 div (γ (2) u 2 = 0 in Ω 0 =< (Λ γ (1) Λ γ (2))u 1, u 2 >= We want to use this identity for singular solutions Ω ( γ (1) γ (2)) u 1 u 2 dx E. Francini (Università di Firenze) EIT 15 / 42

Main idea of the proof (uniqueness) Assume Λ γ (1) = Λ γ (2) ) ) div (γ (1) u 1 = 0 div (γ (2) u 2 = 0 in Ω 0 =< (Λ γ (1) Λ γ (2))u 1, u 2 >= We want to use this identity for singular solutions Ω ( γ (1) γ (2)) u 1 u 2 dx E. Francini (Università di Firenze) EIT 15 / 42

Main idea of the proof (uniqueness) D M γ (1) γ (2) L (D M ) = γ (1) γ (2) L (Ω) Extend γ (k) = 1 in D 0, an open cylinder connected to Σ 1. D 0 Set G y Σ k (x, y) the Green s function related to γ 1 z D and Ω D 0 1 For y, z D 0 ( ) D 1 γ (1) 1 γ (2) 1 G 1 (x, y) G 2 (x, z) dx ( ) + γ (1) (x) γ (2) (x) G 1 (x, y) G 2 (x, z) dx = 0. Ω\D 1 by letting y and z tend to Σ 1 γ (1) 1 = γ (2) 1. E. Francini (Università di Firenze) EIT 16 / 42

Main idea of the proof (uniqueness) D M γ (1) γ (2) L (D M ) = γ (1) γ (2) L (Ω) Extend γ (k) = 1 in D 0, an open cylinder connected to Σ 1. D 0 Set G y Σ k (x, y) the Green s function related to γ 1 z D and Ω D 0 1 For y, z D 0 ( ) D 1 γ (1) 1 γ (2) 1 G 1 (x, y) G 2 (x, z) dx ( ) + γ (1) (x) γ (2) (x) G 1 (x, y) G 2 (x, z) dx = 0. Ω\D 1 by letting y and z tend to Σ 1 γ (1) 1 = γ (2) 1. E. Francini (Università di Firenze) EIT 16 / 42

Main idea of the proof (uniqueness) D M γ (1) γ (2) L (D M ) = γ (1) γ (2) L (Ω) Extend γ (k) = 1 in D 0, an open cylinder connected to Σ 1. D 0 Set G y Σ k (x, y) the Green s function related to γ 1 z D and Ω D 0 1 For y, z D 0 ( ) D 1 γ (1) 1 γ (2) 1 G 1 (x, y) G 2 (x, z) dx ( ) + γ (1) (x) γ (2) (x) G 1 (x, y) G 2 (x, z) dx = 0. Ω\D 1 by letting y and z tend to Σ 1 γ (1) 1 = γ (2) 1. E. Francini (Università di Firenze) EIT 16 / 42

Main idea of the proof (uniqueness) Ω\D 1 ( ) γ (1) (x) γ (2) (x) G 1 (x, y) G 2 (x, z) dx = 0, y, z D 0. Can we let y, z into D 1? Yes because ( ) S 1 (y, z) := γ (1) (x) γ (2) (x) G 1 (x, y) G 2 (x, z) dx Ω\D 1 solves in D 0 D 1 D M and div y (γ (1) (y) y S 1 (y, z))) = 0 D 0 D 1 Σ 1 div z (γ (2) (z) z S 1 (y, z))) = 0 By letting y and z tend to D 2 γ (1) 2 = γ (2) 2. E. Francini (Università di Firenze) EIT 17 / 42

Main idea of the proof (uniqueness) Ω\D 1 ( ) γ (1) (x) γ (2) (x) G 1 (x, y) G 2 (x, z) dx = 0, y, z D 0. Can we let y, z into D 1? Yes because ( ) S 1 (y, z) := γ (1) (x) γ (2) (x) G 1 (x, y) G 2 (x, z) dx Ω\D 1 solves in D 0 D 1 D M and div y (γ (1) (y) y S 1 (y, z))) = 0 D 0 D 1 Σ 1 div z (γ (2) (z) z S 1 (y, z))) = 0 By letting y and z tend to D 2 γ (1) 2 = γ (2) 2. E. Francini (Università di Firenze) EIT 17 / 42

Main idea of the proof (uniqueness) Ω\D 1 ( ) γ (1) (x) γ (2) (x) G 1 (x, y) G 2 (x, z) dx = 0, y, z D 0. Can we let y, z into D 1? Yes because ( ) S 1 (y, z) := γ (1) (x) γ (2) (x) G 1 (x, y) G 2 (x, z) dx Ω\D 1 solves in D 0 D 1 D M and div y (γ (1) (y) y S 1 (y, z))) = 0 D 0 D 1 Σ 1 div z (γ (2) (z) z S 1 (y, z))) = 0 By letting y and z tend to D 2 γ (1) 2 = γ (2) 2. E. Francini (Università di Firenze) EIT 17 / 42

Main idea of the proof (uniqueness) Ω\D 1 ( ) γ (1) (x) γ (2) (x) G 1 (x, y) G 2 (x, z) dx = 0, y, z D 0. Can we let y, z into D 1? Yes because ( ) S 1 (y, z) := γ (1) (x) γ (2) (x) G 1 (x, y) G 2 (x, z) dx Ω\D 1 solves in D 0 D 1 D M and div y (γ (1) (y) y S 1 (y, z))) = 0 D 0 D 1 Σ 1 div z (γ (2) (z) z S 1 (y, z))) = 0 By letting y and z tend to D 2 γ (1) 2 = γ (2) 2. E. Francini (Università di Firenze) EIT 17 / 42

Main technical ingredients Construction of singular solutions and study of their asymptotic behavior near the discontinuity interfaces. Quantitative estimates of unique continuation. E. Francini (Università di Firenze) EIT 18 / 42

Main technical ingredients Construction of singular solutions and study of their asymptotic behavior near the discontinuity interfaces. Quantitative estimates of unique continuation. E. Francini (Università di Firenze) EIT 18 / 42

Singular solutions Difficulty: Lack of a result of existence of the Green s function in Ω for n 3 n 3 Hofmann and Kim (2007, Manuscripta Math. ) existence of Green s function for weak solutions satisfying interior Hölder continuity estimates. n = 2 Dong and Kim (2009, Trans. Amer. Math. Soc. ) existence of Green s function E. Francini (Università di Firenze) EIT 19 / 42

Singular solutions Difficulty: Lack of a result of existence of the Green s function in Ω for n 3 n 3 Hofmann and Kim (2007, Manuscripta Math. ) existence of Green s function for weak solutions satisfying interior Hölder continuity estimates. n = 2 Dong and Kim (2009, Trans. Amer. Math. Soc. ) existence of Green s function E. Francini (Università di Firenze) EIT 19 / 42

Walkway Define a walkway K M 1 j=0 D j that crosses only flat interfaces D M D 0 D 1 E. Francini (Università di Firenze) EIT 20 / 42

Tools Fundamental solutions at the interfaces Let γ, δ C. Then, if Γ(x, y) is the fondamental solution to the Laplace equation 1 γ Γ(x, y) + sγ(x, y ) if x n > 0, y n > 0, ( ) Γ γ,δ (x, y) = 1 γ + s Γ(x, y) if x n y n < 0, 1 δ Γ(x, y) + tγ(x, y ) if x n < 0, y n < 0, γ δ where y = (y 1,..., y n 1, y n ), s = γ(γ+δ) and t = δ γ γ(γ+δ), is a fundamental solution for the differential operator ) ) div ((δχ R + γχ n R in R n+ n. Li,Nirenberg, Comm. Pure Appl. Math., 2003 Hölder gradient estimates up to the interfaces for solutions to elliptic systems. E. Francini (Università di Firenze) EIT 21 / 42

Singular solutions Theorem Let γ satisfy assumptions (H1)-(H3) in Ω 0 = Ω D 0. For y K there exists a unique function G(, y), continuous in Ω 0 \ {y} Ω 0 γ G(, y) φ = φ(y), φ C 0 (Ω 0 ), G(, y) H 1 (Ω 0 \B r (y)) Cr 1 n/2, r < d(y, Ω)/2, G(x, y) = G(y, x) for every x, y K. Moreover for x D k, x 0 Σ k and y r = x 0 + rν, x G(x, y r ) = x Γ k (x, y r ) + O(1) E. Francini (Università di Firenze) EIT 22 / 42

Estimates of unique continuation Theorem Let v H 1 (K) be a solution to div (γ v) = 0 in K, such that v L (K 0 ) ɛ 0, v(x) (ɛ 0 + E 0 )dist(x, Σ M ) 1 n 2 for every x K. Then v( x r ) C ( ɛ0 E 0 + ɛ 0 ) τ (M)N 1 δ M 1 τ r (E 0 + ɛ 0 )r (1 n 2)(1 τ r ), where x r = P M 2rν (P M ), P M Σ M, r (0, r 0 ) and τ, τ r, δ 1 (0, 1), N 1 = A r0 n and C depends on r 0, L, A, n, λ, N. E. Francini (Università di Firenze) EIT 23 / 42

D M Σ M x r D 0 D 1 E. Francini (Università di Firenze) EIT 24 / 42

Some remarks We expect that Lipschitz continuous dependence of the admittivities on the data still holds replacing the flatness condition on the interfaces with C 1,α regularity. E. Francini (Università di Firenze) EIT 25 / 42

Some remarks Local Dirichlet to Neumann map γ 1 γ 2 L (Ω) C Λ Σ 1 1 Λ Σ 1 2 L(H 1/2 co,h 1/2 co ). { } Hco 1/2 = φ H 1/2 ( Ω) : suppφ Σ 1 Ω E. Francini (Università di Firenze) EIT 26 / 42

An important application V de Hoop, Qiu, Scherzer, Inverse problems, 2012 Hölder stability implies local convergence of Landweber iteration. The constant in Hölder stability determines the ray of convergence of iterations. Lipschitz continuity implies linear convergence. E. Francini (Università di Firenze) EIT 27 / 42

Part II Size estimates E. Francini (Università di Firenze) EIT 28 / 42

The problem D γ = γ 0 1 Ω\D + γ 1 1 D Ω Size estimate Determine D from FINITELY many boundary measurements. Motivation: Necrosis in liver tissue E. Francini (Università di Firenze) EIT 29 / 42

The solutions Let u 0 H 1 (Ω) be the background solution { div(γ0 u 0 ) = 0 in Ω, u γ 0 0 ν = h on Ω, γ 0 = σ 0 + iɛ 0, σ 0 c 0 > 0 Let u 1 H 1 (Ω) solution to { div(γ u1 ) = 0 in Ω, γ u 1 ν = h su Ω. γ = γ 0 1 Ω\D + γ 1 1 D γ 1 = σ 1 + iɛ 1, σ 1 c 0 > 0 Goal We want to bound D in terms of δw := Ω h(u 1 u 0 ) E. Francini (Università di Firenze) EIT 30 / 42

The solutions Let u 0 H 1 (Ω) be the background solution { div(γ0 u 0 ) = 0 in Ω, u γ 0 0 ν = h on Ω, γ 0 = σ 0 + iɛ 0, σ 0 c 0 > 0 Let u 1 H 1 (Ω) solution to { div(γ u1 ) = 0 in Ω, γ u 1 ν = h su Ω. γ = γ 0 1 Ω\D + γ 1 1 D γ 1 = σ 1 + iɛ 1, σ 1 c 0 > 0 Goal We want to bound D in terms of δw := Ω h(u 1 u 0 ) E. Francini (Università di Firenze) EIT 30 / 42

The solutions Let u 0 H 1 (Ω) be the background solution { div(γ0 u 0 ) = 0 in Ω, u γ 0 0 ν = h on Ω, γ 0 = σ 0 + iɛ 0, σ 0 c 0 > 0 Let u 1 H 1 (Ω) solution to { div(γ u1 ) = 0 in Ω, γ u 1 ν = h su Ω. γ = γ 0 1 Ω\D + γ 1 1 D γ 1 = σ 1 + iɛ 1, σ 1 c 0 > 0 Goal We want to bound D in terms of δw := Ω h(u 1 u 0 ) E. Francini (Università di Firenze) EIT 30 / 42

The conductivity case D γ 0 = σ 0 γ 1 = σ 1 Ω dist(d, Ω) d 0 > 0 Kang, Seo and Sheen, Alessandrini and Rosset, Capdeboscq and Vogelius (low volume fraction) Kang, Kim and Milton ( Different approach for the two dimensional case. Two boundary measurement. No need of background. Explicit constants) E. Francini (Università di Firenze) EIT 31 / 42

Strategy (Alessandrini, Seo, Rosset) Energy estimates δw := Ω h(u 1 u 0 ) C 1 δw u 0 2 C 2 δw D Lower and upper bounds for D in terms of D u 0 2 using respectively regularity estimates and quantitative estimates of unique continuation K 2 W 0 D p u 0 2 K 1 W 0 D D E. Francini (Università di Firenze) EIT 32 / 42

Strategy (Alessandrini, Seo, Rosset) Energy estimates δw := Ω h(u 1 u 0 ) C 1 δw u 0 2 C 2 δw D Lower and upper bounds for D in terms of D u 0 2 using respectively regularity estimates and quantitative estimates of unique continuation K 2 W 0 D p u 0 2 K 1 W 0 D D E. Francini (Università di Firenze) EIT 32 / 42

Strategy (Alessandrini, Seo, Rosset) Energy estimates δw := Ω h(u 1 u 0 ) C 1 δw u 0 2 C 2 δw D Lower and upper bounds for D in terms of D u 0 2 using respectively regularity estimates and quantitative estimates of unique continuation K 2 W 0 D p u 0 2 K 1 W 0 D D E. Francini (Università di Firenze) EIT 32 / 42

General assumptions (H1) Assumptions on Ω: Bounded with Lipschitz boundary. (H2) Assumptions on D: Lebesgue measurable subset of Ω; (H3) Assumptions on the coefficients: we assume the reference medium and the inclusion have admittivities γ 0 = σ 0 + iɛ 0 and γ 1 = σ 1 + iɛ 1 satisfying σ j c 0, γ j c0 1 in Ω, for j = 0, 1. (H4) Assumptions on the boundary data: h H 1/2 ( Ω), h 0 such that h = 0 Ω E. Francini (Università di Firenze) EIT 33 / 42

Assumptions for energy estimates (H3a) γ 0 and γ 1 are different constants with positive real part. (H3b) ɛ 0 (x) 0 in Ω, and there exists a positive constant µ 0 such that ɛ 1 (x) µ 0 or σ 1 (x) σ 0 (x) µ 0 in Ω. E. Francini (Università di Firenze) EIT 34 / 42

Energy estimates Proposition Assume (H3a) then 1 (1 + γ 1 γ 0 c 0 ) γ 1 γ 0 δw Proposition D u 0 2 Assume γ 0 and γ 1 satisfy assumption (H3b) then F 1 δw u 0 2 F 2 δw, where. F 1 = D c0 3 [ 1 2(2 + c0 2) and F 2 = 2 µ 0 c0 2 ( ) 1 2 + δw c 0 γ 1 γ 0 + 1 µ 0 + 1 c 0 ] E. Francini (Università di Firenze) EIT 35 / 42

The one dimensional case Let Ω = ( 1, 1) and let D = (a, b) Ω. Consider { (γ 0 u 0 ) = 0 in ( 1, 1) (γ 0 u 0 )( 1) = (γ 0u 0 )(1) = K C, u 0( 1) + u 0 (1) = 0 { (γu 1 ) = 0 in ( 1, 1) (γu 1 )( 1) = (γu 1 )(1) = K C, u 1( 1) + u 1 (1) = 0 E. Francini (Università di Firenze) EIT 36 / 42

The one dimensional case δw = K 2 2 b a ( 1 γ 1 1 γ 0 If ( 1 R 1 ) = σ 1 γ 1 γ 0 σ1 2 + σ 0 ɛ2 1 σ0 2 + ɛ2 0 or ( 1 I 1 ) = ɛ 1 γ 1 γ 0 σ1 2 + ɛ 0 ɛ2 1 σ0 2 + ɛ2 0 we can estimate b a from δw. ɛ 0 = 0 ) dx < (>)0 in ( 1, 1) < (>)0 in ( 1, 1) σ 1 (σ 1 σ 0 ) ɛ 2 1 σ 0 (σ 2 1 + ɛ2 1 ) < (>)0, ɛ 1 σ 2 1 + ɛ2 1 < (> 0) E. Francini (Università di Firenze) EIT 37 / 42

The one dimensional case δw = K 2 2 b a ( 1 γ 1 1 γ 0 If ( 1 R 1 ) = σ 1 γ 1 γ 0 σ1 2 + σ 0 ɛ2 1 σ0 2 + ɛ2 0 or ( 1 I 1 ) = ɛ 1 γ 1 γ 0 σ1 2 + ɛ 0 ɛ2 1 σ0 2 + ɛ2 0 we can estimate b a from δw. ɛ 0 = 0 ) dx < (>)0 in ( 1, 1) < (>)0 in ( 1, 1) σ 1 (σ 1 σ 0 ) ɛ 2 1 σ 0 (σ 2 1 + ɛ2 1 ) < (>)0, ɛ 1 σ 2 1 + ɛ2 1 < (> 0) E. Francini (Università di Firenze) EIT 37 / 42

The one dimensional case δw = K 2 2 b a ( 1 γ 1 1 γ 0 If ( 1 R 1 ) = σ 1 γ 1 γ 0 σ1 2 + σ 0 ɛ2 1 σ0 2 + ɛ2 0 or ( 1 I 1 ) = ɛ 1 γ 1 γ 0 σ1 2 + ɛ 0 ɛ2 1 σ0 2 + ɛ2 0 we can estimate b a from δw. ɛ 0 = 0 ) dx < (>)0 in ( 1, 1) < (>)0 in ( 1, 1) σ 1 (σ 1 σ 0 ) ɛ 2 1 σ 0 (σ 2 1 + ɛ2 1 ) < (>)0, ɛ 1 σ 2 1 + ɛ2 1 < (> 0) E. Francini (Università di Firenze) EIT 37 / 42

Estimate of D u 0 2 in terms of D (far from the boundary) Proposition Assume that σ 0 is real valued and Lipschitz continuous dist(d, Ω) d 0. There is a constant K 1 depending on the a priori constants and on d 0 and two constants K 2 and p > 1 depending on the a priori constants, on d 0 and also on F (h) = h H 1/2 ( Ω) h, such that H 1 ( Ω) K 2 W 0 D p u 0 2 K 1 W 0 D. D E. Francini (Università di Firenze) EIT 38 / 42

Ingredients Three Spheres Inequality Lipschitz Propagation of Smallness Doubling Inequality E. Francini (Università di Firenze) EIT 39 / 42

Size estimates Theorem Let Ω satisfy (H1) and let D be a measurable subset of Ω such that dist(d, Ω) d 0. Let γ 0 and γ 1 be coefficients satisfying assumptions (H3a) or (H3b) and let h with zero mean value. Then, δw C 1 W 0 D C δw 1/p 2 W 0 where C 1 depend on the a priori constants and on d 0 and the numbers p > 1 and C 2 depend on the a priori constants, on d 0 and, in addition, on F (h). [Beretta, Francini, Vessella, submitted] E. Francini (Università di Firenze) EIT 40 / 42

Inclusions touching the boundary D Ω P D E. Francini (Università di Firenze) EIT 41 / 42

Estimate of D u 0 2 in terms of D (touching the boundary) Assume D Ω, but there exist r 1 > 0 and P Ω such that D Ω\B r1 (P). h H 1/2 ( Ω) be a complex valued nontrivial current density at Ω such that Ω h = 0, and supp h Γ 0 := Ω B r1 /2(P). There is a constant K 1 depending on the a priori constants and r 1 and two constants K 2 and p > 1 depending on the same a priori constants, on r 1 and also on F (h) = h H 1/2 ( Ω) h, such that H 1 ( Ω) K 2 W 0 D p u 0 2 K 1 W 0 D. D E. Francini (Università di Firenze) EIT 42 / 42