Intersections of sets, diophantine equations and Fourier analysis

Abstract

A classical theorem due to Mattila (see \cite{Mat84}; see also \cite{M95}, Chapter 13) says that if $A,B \subset {\Bbb R}^d$ of Hausdorff dimension $s_A, s_B$, respectively, with $s_A+s_B \ge d$, $s_B>\frac{d+1}{2}$ and $dim_{{\mathcal H}}(A \times B)=s_A+s_B\ge d$, then $$ dim_{{\mathcal H}}(A \cap (z+B)) \leq s_A+s_B-d$$ for almost every $z \in {\Bbb R}^d$, in the sense of Lebesgue measure. In this paper, we replace the Hausdorff dimension on the left hand side of the first inequality above by the upper Minkowski dimension, replace the Lebesgue measure of the set of translates by a Hausdorff measure on a set of sufficiently large dimension and replace the translation and rotation group by a more general variable coefficient family of transformations. Interesting arithmetic issues arise in the consideration of sharpness examples. These results are partly motivated by those in \cite{EIT11} and \cite{IJL10} where in the former the classical regular value theorem from differential geometry was investigated in a fractal setting and in the latter discrete incidence theory is explored from an analytic standpoint. Fourier Integral Operator bounds and other techniques of harmonic analysis play a crucial role in our investigation. We also consider, in the spirit of the Furstenberg conjecture, inverse problems for intersections by asking how small a dimension of a set can be given that the dimension of its intersections with a suitably well-curved family of manifolds is bounded from below by a given threshold. Finally, we shall discuss applications of our estimates to the problem of estimating the number of solutions of systems of diophantine equations over integers.