Local uniqueness for the Dirichlet-to-Neumann map via the two-plane transform (Q1847850)

The authors are concerned with (partially) recovering the potential \(q\in H^s(\Omega)\), \(s > n/2\), in the Schrödinger equation \(\Delta u + q(x)u=0\) related to a bounded domain \(\Omega \subset {\mathbb R}^n\), \(n\geq 3\), with a Lipschitz boundary. They provide some additional information concerning the set of Cauchy data \[ {\mathcal C}(q) = \{ (v_{|\partial \Omega},D_nv_{|\partial \Omega})\in H^{1/2}(\partial \Omega) \times H^{-1/2}(\partial \Omega): v\in H^1(\Omega),\;\Delta u + q(x)u=0 \;\text{ in\;} \Omega\} \] on a ``small'' open subset of the boundary \(\partial \Omega\). \noindent From now on we shall assume that potentials \(q_1\) and \(q_2\) are in \(H^s(\Omega)\) for some \(s > n/2\). \noindent Denote by \(M_{2,n}\) the \((3n - 6)\)-dimensional Grassmannian of all affine two-planes \(\Pi \subset {\mathbb R}^n\) and by \(d\lambda^{\Omega}_{\Pi}\) the two-dimensional Lebesgue measure on \(\Pi\) relative to \(\Omega\) defined by \[ \langle f,d\lambda^{\Omega}_{\Pi}\rangle = \lim_{\varepsilon\to 0+}|B^{n-2}(0,\varepsilon)|^{-1} \int_{\Omega \cap \text{ dist}(x,\partial \Pi)\leq \varepsilon} f(x) dx. \] For each choice of an orthonormal basis for \(\Pi_0\), the translate of \(\Pi\) passing through the origin, as well as for other choices, the authors can construct a family \({\mathcal F}(q) = \{v_z(x) : z\in {\mathbb C},\;|z|\geq C\}\) of exponentially growing solutions of \((\Delta + q(x))v = 0\) concentrated near \(\Pi\). They introduce also the following definitions: (i) if \(U\subset \partial \Omega\) is open, \({\mathcal C}(q_1)\) and \({\mathcal C}(q_2)\) are equal on \(U\) at \(z\in {\mathcal F}(q_1)\cap {\mathcal F}(q_2)\) if the solutions \(v^{(1)}_z\) and \(v^{(1)}_{-z}\), corresponding to opposite exponential growths, have the same Cauchy data on \(U\): \[ ({v^{(1)}_{z}}_{|U},{D_nv^{(1)}_z}_{|U}) = ({v^{(2)}_{-z}}_{|U},{D_nv^{(1)}_{-z}}_{|U}); \] (ii) \({\mathcal C}(q_1)\) and \({\mathcal C}(q_2)\) are equal on \(U\) for a sequence of exponentially growing solutions if \({\mathcal C}(q_1)\) and \({\mathcal C}(q_2)\) are equal on \(U\) relative to some sequence \(\{z_j\}_1^\infty\) with \(|z_j|\to \infty\). The main results proved in the paper are the following: \noindent Theorem 1. Let \(\Pi\in M_{2,n}\), and let \({\mathcal F}_{q_1}\), and \({\mathcal F}_{q_1}\) be families of exponentially growing solutions associated to \(q_1\) and \(q_2\). If, for some fixed neighborhood \(U_\Pi\) of \(\Pi \cap \partial \Omega\) in \(\partial \Omega\) and \({\mathcal C}(q_1)\) and \({\mathcal C}(q_2)\) are equal on \(U_\Pi\) for a sequence of exponentially growing solutions, then \(\int q_1(y) d\lambda^{\Omega}_{\Pi}(y) = \int q_2(y) d\lambda^{\Omega}_{\Pi}(y)\). Theorem 2. Let \(C\subset \Omega\) be a closed, convex set. If, for all \(\Pi\in M_{2,n}\) such that \(\Pi \cap C = \emptyset\), there is some neighborhood \(U_\Pi\) of \(\Pi \cap \partial \Omega\) on which \({\mathcal C}(q_1)\) and \({\mathcal C}(q_2)\) are equal for some sequence of exponentially growing solutions, then \(q_1 = q_2\) on \(\Omega \backslash C\). Theorem 3. Suppose that \(\partial \Omega\) is \(C^2\) and strictly convex. If, for some \(r > 0\), \({\mathcal C}(q_1)\) and \({\mathcal C}(q_2)\) are equal for some sequence of exponentially growing solutions on all surface balls \(B = B^n(x_0;r)\cap \partial \Omega\), then \(q_1 = q_2\) on the tubular neighborhood \(\{ x\in {\overline \Omega}: \text{ dist}(x,\partial \Omega)\leq Cr^2\}\) of \(\partial \Omega\) in \({\overline \Omega}\).