Random parking, sequential adsorption, and the jamming limit (Q5940429)

Consider the following multidimensional generalisation of the random car-parking model of Rényi: Take a fixed bounded open subset \(D\) of \(\mathbb{R}^d\) containing 0, a large bounded Borel subset \(A\) of \(\mathbb{R}^d\), and an i.i.d. random sequence of uniformly distributed points \(\{x_j\mid j\in \mathbb{N}\} \subset A\). The point \(x_j\) is accepted iff the ``car'' \(x_j+D\) is included in \(A\setminus \bigcup_{i<j} (x_i+D)\); and the problem is to get knowledge on the number \(N(A):= \max\{j\in \mathbb{N}\mid x_j\) is accepted\}. The author establishes the following spatial law of large numbers: there exists a constant \(\lambda= \lambda (d,D)\) such that, for any sequence \(\{A_n\}\) of bounded Borel subsets of \(\mathbb{R}^d\) with \(\varepsilon\)-neighbourhoods \(\partial_\varepsilon A_n\) satisfying: \(|\partial_\varepsilon A_n|/ |A_n|\) goes to 0 for any \(\varepsilon>0\), \(N(A_n)/ |A_n|\) converges towards \(\lambda\) in any \(L^p\). Moreover this convergence is almost pure at least if the \(A_n\)'s are boxes increasing to \(\mathbb{R}^d\). The proof is valid also for several non-uniform variants of the above model.