On multidimensional record patterns

Multidimensional record patterns are random sets of lattice points defined by means of a recursive stochastic construction. The patterns thus generated owe their richness to the fact that the construction is not based on a total order, except in one dimension, where usual records in sequences of independent random variables are recovered. We derive many exact results on the statistics of multidimensional record patterns on finite samples drawn on hypercubic lattices in any dimension $D$. The most detailed analysis concerns the two-dimensional situation, where we also investigate the distribution of the landing position of the record point which is closest to the origin. Asymptotic expressions for the full distribution and the moments of the number of records on large hypercubic samples are also obtained. The latter distribution is related to that of the largest of $D$ standard Gaussian variables.