An introduction to inverse elliptic and parabolic problems (Q731485)

The book is devoted to the determination of unknown coefficients or unknown -- i.e., inaccessible -- subdomains or parts of the boundary related to both elliptic and parabolic equations. The basic questions dealt with are uniqueness and stability, i.e., continuous dependence on the data. Generally speaking, such problems are nonlinear, with the exception of those involved with the determination of sources. In the nonlinear case the metrics needed to solve the problem and suitable to measure the data are very weak, vanishing usually as a negative power of a logarithm, sometimes as a Hölder positive power and very seldom linearly. The text is divided into three parts, the first being a survey of results needed for the following two -- very rich in applications -- chapters related to elliptic and parabolic inverse problems. In chapter 2 two methods are highlighted to determine -- in the sense explained above -- the potential \(q\) in a Schrödinger equation (SE). The first is concerned with the so-called optimal geometric technique, consisting in constructing suitable solutions of (SE), the product of which is dense in the \(L^1\)-space so that to ensure uniqueness. The latter method consists in obtaining such a result under the knowledge of the so-called Dirichlet-Neumann operator (DN), i.e., the operator associating with any function \(f\in H^{3/2}(\partial \Omega)\) the normal derivative \(D_nu\in H^{1/2}(\partial \Omega)\) of the solution \(u\) to (SE) in \(\Omega\) such that \(u=f\) on \(\partial \Omega\). A different problem is related to the determination of \(q\) in (SE) by means of the eigenvalues of (SE), endowed with homogeneous Dirichlet conditions. While the uniqueness is implied by the knowledge of the (DN)-operator, the continuous dependence is deduced by passing to the wave equation with potential \(q\) and using a suitable density result for such an equation. In this case the reconstruction of \(q\) makes use of all eigenvalues as well as of all traces on \(\partial \Omega\) of the normal derivatives of the eigenfunctions. The determination of the inverse conductivity problem for the equation \(\text{div}(a(x)\nabla u)=0\) is faced with, assuming that the map \(u_{|\partial \Omega}\to D_nu_{|\partial \Omega}\) is known, constructing suitable singular solutions for such an equation. In this case a stability of Hölder type can be established. In case of corrosion problems, consisting in determining an unknown boundary coefficient, some stability results are proved taking advantage of only one boundary measurement by suitable Carleman inequalities. Finally, the determination of a subdomain \(D\subset \subset \Omega\) is proved by taking into account the (DN)-map and using a technique of differentiation with respect to the domain, while the determination of cracks is carried out using quasi-conformal mappings and the previous technique. The third chapter is devoted to uniqueness and stability for \(a\) or \(f\) in the parabolic problem \(D_tu-\text{div}(a(x)\nabla u)+f(x,u)=0\) in \((0,t)\times \Omega\), \(u(0,\cdot)=u_0\) in \(\Omega\), \(u=g\) on \(\partial \Omega\), \(a(x)D_nu=g\) on \((0,T)\times \partial \Omega\), when suitable pieces of additional information are provided. The first problem deals with the determination of function \(q\) when \(a\equiv 1\) and \(f(x,u)=q(x)u\) and \(u(T,\cdot)\) is prescribed. In this case an existence result is obtained by a recursive procedure. Relying on this result and the observability of \(u(T,\cdot)\) via \(D_nu\) on a open subset \((0,T)\times \gamma\) of the lateral surface \((0,T)\times \partial \Omega\), \(q\) can be determined. When \(f(x,u)\equiv f(u)\) and \(f(0)=0\), \(f\) can be uniquely determined by the means of \(D_nu\) on \((0,T)\times \gamma\) provided it is a priori known that two possible solutions \(f_1\) and \(f_2\) satisfy the sign condition: \(f_2-f_1\) changes sign no more than \(N\) times. In this case a stability of Hölder type can be proved. When \(f(x,u)\equiv f(x)\), i.e., when the source is unknown, a Fourier series techniques allows to determine \(f\), the stability being of logarithmic type. In the case when \(a\equiv 1\), \(f(x,u)\equiv 0\), but the initial value \(u_0\) is unknown, \(u_0\) can be determined in \(L^2(\Omega)\), but only in terms of a Bergman-Selbarg norm of \(D_nu\) on \((0,T)\times \gamma\). Finally, in the case of the heat equation, a source of the form \(\sum_{i=1}^m\,\lambda_i(t)\delta_{a_i}\), \(\delta_{a_i}\) denoting the Dirac measure concentrated at \(t=a_i\), or of the form \(\sum_{i=1}^m\,\lambda_i(t)\chi_{B_i}\),\(B_i\) being balls contained in \(\Omega\), can be determined in terms of \(D_nu\) on \((0,T)\times \gamma\). The last section of the book lists seven problems that the reader is invited to solve step by step. In conclusion, the reviewer not only recommends this well-written book as a good trainer for young mathematicians wishing to be acquainted with inverse problems for PDEs, but also recommends it to researchers, since they can find in it a lot of stimulating ideas and indispensable techniques nowadays needed to solve inverse problems.