Method of analytic continuation for the inverse spherical mean transform in constant curvature spaces (Q355975)

Let \(f\) be a \(C^\infty\) function with compact support in the open ball \(B = \{ x \in \mathbb{R}^n: |x|< R\}\) and let \(\partial B\) denote the boundary of \(B\). The spherical mean Radon transform \(Mf\) integrates \(f\) over spheres with center lying on the boundary \(\partial B\); precisely, it is defined by \[ Mf(\xi,t) = \frac{1}{\sigma_{n-1}} \int_{S^{n-1}} f(\xi - t\theta) d\theta, \] where \(\xi \in \partial B\), \( t \in (0,\infty)\), and \(S^{n-1}\) is the unit sphere in \(\mathbb{R}^n\) with surface area \(\sigma_{n-1}\), and \(d \theta\) is the usual Lebesgue measure on \(S^{n-1}\). A major consideration regarding this transform is the following inverse problem: given the collection of data \(Mf(\xi,t)\) for all \(\xi \in \partial B\) and \(t \in (0,\infty)\), recover the original function \(f\). The ability to invert \(M\) and to reconstruct \(f\) has significant implications for certain imaging techniques (e.g. ultrasound). The paper under review considers this question in a more general setting of spaces of arbitrary constant curvature, and proves explicit inversion formulae in several cases. First, in the setting of \(\mathbb{R}^n\) and \(f\) supported in a ball \(B = \{ x \in \mathbb{R}^n: |x|< R\}\), the authors prove explicit inversion formulae for \(Mf(\xi,t)\); these break into three cases: dimension \(n \geq 3\) odd, \(n \geq 4\) even, and \(n=2\). For example, the case \(n \geq 3\) odd yields the inversion formula \[ f(x) = d_{n,1} \Delta \left. \int_{\partial B} D^{n-3} [t^{n-2}(Mf)(\xi,t)] \right|_{t=|x-\xi|} d\xi, \] for an explicit constant \(d_{n,1}\). Second, the paper considers the case of a spherical mean Radon transform \(Mf\) on an arbitrary space \(X\) of constant curvature. One then considers a smooth function \(f\) supported in \(B\), a geodesic ball centered at the origin, and the spherical means of \(f\) are evaluated over geodesic spheres, the centers of which are located on the boundary of \(B\). In the case where \(X=S^n\) is the \(n\)-dimensional unit sphere in \(\mathbb{R}^{n+1}\), this results in the following Radon transform: \[ Mf(x,t) = \frac{(1-t^2)^{(1-n)/2}}{\sigma_{n-1}} \int_{x \cdot y =t} f(y) d\sigma(y), \] where \(x\in S^n\), \(f \in C^\infty(S^n)\), \(t \in (-1,1)\), and the integral is therefore the mean value of \(f\) over the planar section \(\{y \in S^n : x \cdot y = t\}\). The authors prove explicit inversion formulae for such transforms, under two assumptions: (a) the support of \(f\) lies on a spherical cap (i.e., a geodesic ball lying on \(S^n\)), and (b) the mean value data \(Mf(x,t)\) is known for all \(x\) in the spherical cap and \(t \in (-1,1)\). In the case where \(X=\mathbb{H}^n\), a hyperbolic space with associated inner product \([x,y]\), the authors consider the spherical mean Radon transform defined for \(x \in \mathbb{H}^n\), \(t>1\), by \[ Mf(x,t) = \frac{(t^2-1)^{(1-n)/2}}{\sigma_{n-1}} \int_{[x,y]=t} f(y)d\sigma(y), \] which takes mean values of \(f\) over the planar sections \(\{ y \in \mathbb{H}^n: [x,y]=t\}\). Explicit inversion formulae are again proved in this case, under the assumptions that (a) \(f\) is supported in a certain type of a geodesic ball \(B\), and (b) the mean values \(Mf(x,t)\) are known for all \(x \in \partial B\) and all \(t>1\). The main method involved in proving these results is analytic continuation of corresponding operator families; a careful treatment is given to proving a number of integral formulae. Finally, certain of these results are applied to the Cauchy problem for the Euler-Poisson-Darboux equation in the settings considered above.