Unique continuation of microlocally analytic distributions and injectivity theorems for the ray transform (Q618993)

For a unit vector \(v\in \mathbb{R}^{n}\) and \(f\in L^{1}(\mathbb{R}^{n})\) the ray transform \(P_{v}\) is defined by \[ P_{v}f(y)=\int_{R}f\left( y+tv \right) dt,\quad y\in \mathbb{R}^{n}/\mathbb{R}v, \] where \(\mathbb{R}v\) is the subspace of \(\mathbb{R}^{n}\) generated by the vector \(v,\) and \(\mathbb{R}^{n}/\mathbb{R}v\) is the corresponding \(n-1\)-dimensional factor space. If \(\mu\in M(\mathbb{R}^{n})\), the space of (signed) measures on \(\mathbb{R}^{n}\) with finite total mass, and \(v\in \mathbb{R}^{n}\setminus 0,\) \(P_{v}\mu\) is the pushforward with respect to the canonical projection \ \(\pi_{v}:\mathbb{R}^{n}\rightarrow \mathbb{R}^{n} /\mathbb{R}v ,\) defined by \[ \left\langle P_{\upsilon}\mu,\varphi\right\rangle =\left\langle \pi_{\upsilon\ast}\mu,\varphi\right\rangle =\left\langle \mu,\pi_{\upsilon} ^{\ast}\varphi\right\rangle ,\;\;\;\varphi\in C_{b}\left( \mathbb{R}^{n} /\mathbb{R}v\right); \] here \(C_{b}\) denotes the space of bounded, continuous test functions and \(\pi_{v}^{\ast}\) is the pullback \(\varphi\mapsto\varphi\circ \pi_{v}\) from \(C_{b}(\mathbb{R}^{n}/\mathbb{R}v\)) to \(C_{b}(\mathbb{R}^{n})\). For \(\mu\in M(\mathbb{R}^{n})\), set \[ C_{\mu}=\left\{ \eta\in\widehat{\mathbb{R}^{n}};\int_{\mathbb{R}^{n}}e^{\eta\cdot x}\left| d\mu\right| <\infty\right\}, \] where \(\widehat{\mathbb{R}^{n}}\) is the dual space of \(\mathbb{R}^n.\) For \(v\in \mathbb{R}^{n}\backslash 0,\) the subspace of \(\widehat{\mathbb{R}^{n}}\) consisting of all linear forms that vanish on \(v\) will be denoted by \(v^{\perp}\). Using a vanishing theorem for microlocally real analytic distributions and a theorem on flatness of distributions vanishing on infinitely many hyperplanes, the author gives a new proof of the following theorem which was first proved by \textit{C. Bélisle, J.-C. Massé} and \textit{T. Ransford} [Ann. Probab. 25, No.~2, 767--786 (1997; Zbl 0878.60006)]: Theorem 1. Assume that \[ P_{v_k}\mu=0,\quad k=1,2,\dots, \tag{1} \] for an infinite sequence of (pairwise non-parallel) \(v_{k}\in \mathbb{R}^{n}\backslash 0\) with limit point \(v_{0}\in \mathbb{R}^{n}\;\backslash\;0\) such that \[ C_{\mu}\;\text{intersects both components of }\widehat{\mathbb{R}^{n}}\;\backslash \;v_{0}^{\perp}. \tag{2} \] Then \(\mu=0.\) The author then generalizes the above theorem by replacing the real analyticity by quasianalyticity as follows. Let \(M=\left\{ M_{k}\right\} \) be a logarithmically convex sequence of numbers satisfying \(M_{k}\geq k!.\) The Denjoy-Carleman class \(C^{M}(\mathbb{R}^{n})\) is defined as the set of \(C^{\infty}\) functions \(\varphi\) whose derivatives satisfy \[ \sup_{x\in K}\left| \partial^{\alpha}\varphi\left( x\right) \right| \leq C^{\left| \alpha\right| +1}M_{k},\quad \left| \alpha \right| =k, \] for all multi-index \(\alpha\) and all compact sets \(K.\) Assume that the sequence \(M_{k}\) satisfies \[ M_{k+1}\leq C^{k+1}M_{k}, \tag{3} \] which implies that the class \(C^{M}(\mathbb{R}^{n})\) is closed under differentiation and that it is invariant under real analytic coordinate transformations. Let \[ q_M(t)=\sum_{k=0}^\infty t^k/M_k,\quad t\geq 0, \text{ and } q_M(t)=1, \text{ for } t<0, \] and define \[ C_{M,\mu}=\left\{ \eta\in\widehat{\mathbb{R}^{n}};\int_{\mathbb{R}^{n}}q_M (\eta\cdot x)\left| d\mu (x) \right| <\infty\right\}. \] The following extension of Theorem 1 is proved. Assume that the sequence \((M_{k})\) satisfies ({3}) and that the class \(C^{M}(\mathbb{R}^n)\) is quasianalytic. \ Assume furthermore that ({1}) holds, \(v_{0}=\lim_{k\rightarrow\infty}v_{k}\in \mathbb{R}^{n}\;\backslash\;0,\) and that \[ C_{M,\mu}\;\text{intersects both components of }\widehat{\mathbb{R}^{n}}\;\backslash \;v_{0}^{\perp}. \] Then \(\mu=0.\) It is also shown that the assumption ({2}) in Theorem 1 cannot be omitted.