Random r-flats meeting a ball (Q1110179)

The paper deals with hitting probabilities in stochastic geometry. Let us consider a random r-flat meeting a fixed ball B in the d-dimensional Euclidean space (d\(\geq 2\), \(1\leq r\leq d-1)\). The distribution of the random r-flat is derived from a translation-invariant measure and is completely described by the so-called direction distribution \(\theta\). The probability that two independent, identically distributed random r- flats have a common point in B depends on the direction distribution \(\theta\) and is denoted by p(\(\theta)\) (2r\(\geq d).\) In the case \(r=d-1\) (hyperplanes), \textit{R. Schneider} [Z. Wahrscheinlichkeitstheor. Verw. Geb. 61, 379-387 (1982; Zbl 0492.60011)] proved the conjecture of \textit{R. E. Miles} [Adv. Appl. Probab. 1, 211-237 (1969; Zbl 0184.215)] that p(\(\theta)\) becomes maximal if and only if \(\theta\) is isotropic. \textit{C. Thomas} and the author [Commun. Stat., Stochastic Models 2, 273-280 (1986; Zbl 0618.60017)] pointed out the fact that for \(r<d-1\) the isotropic distribution is no longer maximal, but they did not determine the extremal distributions. In the present paper only the case \(d=2r\) is considered (r\(\geq 2)\). The probability p(\(\theta)\) attains its maximal value if there is an r- dimensional subspace S such that the random r-flats are parallel to S with probability 1/2 and perpendicular to S with probability 1/2. If \(r\geq 3\), there are no other extremal distributions \(\theta\). The remaining exceptional case \(r=2\), \(d=4\) is more exhaustively treated in the paper ``An extremal property of random flats'', which will appear in Vol. 150 of the Journal of Microscopy containing the proceedings of 4th Int. Conf. Stereol. Stoch. Geom., Berne 1987. Related results for intersection intensities of stationary r-flat processes in the 2r- dimensional Euclidean space are explained.