Counting and Cutting Rich Lenses in Arrangements of Circles

Abstract

We show that the maximum number of pairwise nonoverlapping $k$-rich lenses (lenses formed by at least $k$ circles) in an arrangement of $n$ circles in the plane is $O(n^{3/2}\log(n / k^3)/k^{5/2} + n/k)$, and the sum of the degrees of the lenses of such a family (where the degree of a lens is the number of circles that form it) is $O(n^{3/2}\log(n/k^3)/k^{3/2} + n)$. Two independent proofs of these bounds are given, each interesting in its own right (so we believe). The second proof gives a bound that is weaker by a polylogarithmic factor. We then show that these bounds lead to the known bound of Agarwal et al. [J. ACM, 51 (2004), pp. 139--186] and Marcus and Tardos [J. Combin. Theory Ser. A, 113 (2006), pp. 675--691] on the number of point-circle incidences in the plane. Extensions to families of more general algebraic curves and some other related problems are also considered.