A `large deviation' phenomenon in discrete geometry. I (Q1077727)

The author's abstract states: ``Let \(\mathcal D\) be a completely arbitrary family of circular discs of unit radius on the plane. We show that \(\mathcal D\) cannot be too evenly distributed relative to all directions. Actually we prove that there is a 'random size' oscillation''. More precisely, the following very interesting theorem is proved (quoted directly from the paper): Let \(\mathcal D=\{D_1,D_2,D_3,\ldots,\}\) be an arbitrary infinite family of circular discs on the plane having common radius one. We say that \(\mathcal D\) has density \(\alpha\) if \[ \lim_{y\to \infty}(y^2\pi)^{-1}\sum^\infty_{i=1} \mathrm{area}\left(D_i\cap \{\mathfrak x=(x_1, x_2)\in \mathbb R^2: x^2_1+x^2_2\leq y^2\}\right) \] exists and is equal to \(\alpha\). Let \(S\) be a straight line segment, and put \(\ell (\mathcal D,S)=\sum^\infty_{i=1}\mathrm{length}(D_i\cap S)\). Theorem. Let \(\mathcal D\) be a family of circular discs of unit radius on the plane. Suppose \(\mathcal D\) has positive density \(\alpha\). Given an arbitrarily large real number \(r>1\), one can find two parallel straight line segments \(S_1\) and \(S_2\) with \(\mathrm{length}(S_1)=\mathrm{length}(S_2)\leq r\) and with deviation \(| \ell (\mathcal D,S_1)-\ell (\mathcal D,S_2)| >c_0(\alpha)\cdot r^{1/2}\), where the positive constant \(c_0(\alpha)\) depends only on the density \(\alpha\). In part II of the paper (to appear) the author will discuss the sharpness of the theorem. For an interesting result in the opposite direction see \textit{P. Frankl}, \textit{J. Pach} and \textit{V. Rödl} [Monatsh. Math. 98, No. 2, 93--98 (1984; Zbl 0542.52010)].