We prove the following conjecture of Furstenberg (1969): if A, B ⊂ [0, 1] are closed and invariant under ×p mod 1 and ×q mod 1, respectively, and if log p/ log q / ∈ Q, then for all real numbers u and v,We obtain this result as a consequence of our study on the intersections of incommensurable self-similar sets on R. Our methods also allow us to give upper bounds for dimensions of arbitrary slices of planar self-similar sets satisfying SSC and certain natural irreducible conditions. 1 or combinatorial reasons. For A p , B q as in Conjecture 1.1, the set A p × B q is such an example, for which it is clear that certain lines parallel to the axes are exceptional for the slice result, and Conjecture 1.1 predicts that these lines are the only exceptions.There is a rich literature about generic slices of various fractal sets, see e.g. [2,7,8,20,26,28,29,37]. However, very little is known about specific slices, and there were few partial results concerning Conjecture 1.1 before the present paper. The first and perhaps also the best one is due to Furstenberg [16, Theorem 4]. His result states that under the assumption of the conjecture, if dimFrom the last assertion, it is not hard to deduce that in this case, we must have dim H A p + dim H B q > 1/2 (see [22, Theorem 7.9] for the deduction). Thus, under the assumption dim H A p + dim H B q ≤ 1/2, Furstenberg's result confirms Conjecture 1.1. We will return back to [16, Theorem 4] in Subsection 4.2. We would like to mention that the technique (namely, CP-process) Furstenberg introduced and used in [16] is also important for the present work, it will be one of the main ingredients for our proof of Conjecture 1.1.Recently, Feng, Huang and Rao [13] studied affine embeddings between incommensurable self-similar sets and, as a consequence, they showed that if p ≁ q, then for T p -invariant self-similar set E and T q -invariant self-similar set F , there exists a (non-effective) positive constant δ depending on E and F such that the Hausdorff dimension of the intersection Later, Feng [12] obtained some effective versions of the results of [13], but these effective versions are still far from sufficient for proving Conjecture 1.1. Feng [12] also constructed, for any s, t ∈ (0, 1) and ǫ > 0, a T p -invariant set A of dimension s and a T q -invariant set B of dimension t which verify Conjecture 1.1 with a loss of ǫ.Finally, we note that the slice problem may be considered as "dual" to the projection problem for fractal sets. In that direction, there is a dual version of Conjecture 1.1, also due to Furstenberg and recently settled by Hochman and Shmerkin [23] (some special cases by Peres and Shmerkin [31]), which asserts that under the assumptions of Conjecture 1.1, for each orthogonal projection P θ from R 2 to R with direction θ not parallel to the axes, we haveRecently, there has been considerable interest in the study of projections of dynamically defined Cantor sets, see for instance the survey paper of Shmerkin [35] and the references therein for more details.1.2. Statem...