Mixed norm estimates for a restricted X-ray transform. (Q1809911)

The authors establish optimal (except for endpoints) mixed norm estimates for the restricted X-ray transform \[ Xf(\theta,y) := \int_{| s| \leq C} f(y + s \nu_\theta)\;ds \] in \(\mathbb R^n\), where \(\theta\) is in a fixed interval of \(\mathbb R\), \(C\) is a constant, and \(\nu_\theta\) is a generic polynomial curve such as \(\nu_\theta = (1,\theta,\ldots,\theta^{n-1})\). The authors study the question of when this transform maps \(L^p(\mathbb R^n)\) to \(L^q_\theta L^r_y\). Examples show that we have the three necessary conditions \[ n/p \leq (n-1)/r + 1, \] \[ n(n-1)/p \leq 2/q + n(n-1)/r, \] \[ (n-2)(n+1)/p \leq n(n-1)/r, \] determining a certain three-dimensional polytope. The authors establish strong type estimates on the interior of the polytope and restricted weak-type estimates on the boundary of the polytope. The most interesting vertex here is when \(p=q = (n+2)/n\) and \(r = (n^2 + n-2)/(n^2-n-2)\), which implies all the other results in this polytope. The method is a variant of that of \textit{M. Christ} [Int. Math. Res. Not. 1998, No. 19, 1033--1048 (1998; Zbl 0927.42008)], relying primarily on the combinatorial geometry of the complex of lines associated to the X-ray transform; the main point of the paper is that the presence of mixed Lebesgue norms causes no substantial additional difficulties.