Stability of the non-abelian \(X\)-ray transform in dimension \(\ge 3\) (Q2238550)

Non-abelian X-ray tomography seeks to recover a matrix potential \(\Phi\colon M \to \mathbb{C}^{m\times m}\), \(m\geq 1\), in a domain \(M\), from measurements of its so-called scattering data \(C_\Phi\) at \(\partial M\). It provides the mathematical basis for the imaging technology of ``polarimetric neutron tomography'' [\textit{N. M. Desai} et al., Inverse Probl. 36, No. 4, Article ID 045001, 17 p. (2020; Zbl 1476.44001)]. More precisely, \(M\) is a compact Riemannian manifold with strictly convex boundary, \(\Phi\) is a continuous matrix potential and \(C_\Phi(\gamma)=U(0)\in \mathrm{GL}(m,\mathbb{C})\) is the value at the boundary of the unique continuous solution \(U\) of the linear matrix differential equation \(\dot U(t)+\Phi(\gamma(t))U(t)=0\), \(U(\tau)=\mathrm{id}\), where \(\gamma\colon [0,\tau]\to M\) is a unit speed geodesic with endpoints on \(\partial M\). The inverse problem with access to partial data is to recover \(\Phi\) in an open set \(O\subset M\) from measuring \(C_\Phi(\gamma)\) for geodesics \(\gamma\) that do not leave \(O\). Even when \(M\) is a Euclidean ball, geodesics are straight lines and one has access to full data (that is, \(O=M\)), this inverse problem is non-linear, and for \(m\geq 2\) no explicit inversion formula is known. When \(\dim M \geq 3\) and \(O\subset M\) satisfies a ``foliation condition'', it is known that locally, smooth potentials are determined uniquely by their scattering data [\textit{G. P. Paternain} et al., Am. J. Math. 141, No. 6, 1707--1750 (2019; Zbl 1440.53083)]. The present article extends this result by proving a Hölder-type stability estimate. The main technical contribution of the paper is a more quantitative version of the microlocal technique (local inversion of scattering operators near elliptic points) introduced by \textit{G. Uhlmann} and \textit{A. Vasy} [Invent. Math. 205, No. 1, 83--120 (2016; Zbl 1350.53098)] As an application, a statistical consistency result for \(\dim M=2\) [\textit{F. Monard} et al., Commun. Pure Appl. Math. 74, No. 5, 1045--1099 (2021; Zbl 07363259)] is generalised to higher dimensions.