Borel–Cantelli lemma

Borel–Cantelli lemma

The Borel–Cantelli lemma is a pair of results in probability theory that relate the sum of probabilities of events to the probability that infinitely many of those events occur. It connects measures of sequences of measurable sets and almost sure convergence, and is fundamental in the study of limit laws, ergodic theory, and number theory. The lemma informs results in the work of Paul Erdős, Andrey Kolmogorov, Émile Borel, and Francesco Cantelli, and underpins techniques used by researchers affiliated with institutions such as Princeton University, École Normale Supérieure, and University of Cambridge.

Statement

The classical pair of statements known collectively as the lemma are usually given for a probability space (Ω, F, P) and a sequence of events A1, A2, A3, ... . The first (easy) direction asserts that if the series Σ P(An) converges, then the probability that infinitely many of the An occur is zero. The second (stronger) direction provides a partial converse: if the events are independent and Σ P(An) diverges, then the probability that infinitely many An occur equals one. These assertions appear in expositions by Andrey Kolmogorov in his work on limit theorems and in treatments by William Feller in his textbooks, and they are applied in texts influenced by Kolmogorov's zero–one law and techniques from Srinivasa Ramanujan-related problems.

Proofs

The proof of the convergent-series direction uses the union bound and monotone convergence arguments; such proofs appear in lecture notes from Harvard University, Massachusetts Institute of Technology, and Stanford University. The divergent independent-events direction is traditionally proved via the Borel-Cantelli second lemma using independence to estimate complements and apply the inequality (1 − x) ≤ e−x, an approach visible in presentations by Norbert Wiener and Andrey Kolmogorov. Alternative proofs exploit martingale convergence theorems credited to Jean Ville and connections to the strong law of large numbers developed by Émile Borel and S. R. Srinivasa Varadhan; ergodic-theoretic proofs reference results by George David Birkhoff and John von Neumann.

Extensions and Generalizations

Several extensions relax the independence assumption using notions introduced by Paul Lévy, Egon Pearson, and Kolmogorov: pairwise independence, various mixing conditions, and the concept of quasi-independence. The Kochen–Stone theorem, named for Joseph Leo Doob's collaborators and related to results by Paul Kochen and Charles J. Stone, gives quantitative replacements for independence in the divergent case. Generalizations connect to the ergodic theorems of George David Birkhoff and the metric theory of Diophantine approximation developed by Kurt Mahler and Vojtěch Jarník; these include the Gallagher and Duffin–Schaeffer results in metric number theory. Another direction uses maximal inequalities and martingale methods from the work of Joseph L. Doob and Donald Burkholder to obtain Borel–Cantelli-type conclusions under dependence structures studied by Olivier Garet and researchers at Institut des Hautes Études Scientifiques.

Applications

The lemma is widely used across probabilistic and analytic areas: in proofs of almost sure convergence in the strong law of large numbers as treated by Andrey Kolmogorov and Alexander Yakovlevich Khinchin; in ergodic theory arguments inspired by George David Birkhoff and John von Neumann; in metric number theory influenced by Émile Borel and Srinivasa Ramanujan; and in random graph threshold phenomena investigated by Erdős–Rényi and researchers at Bell Labs. It appears in studies of percolation by researchers at University of Cambridge and Perimeter Institute, in limit behaviors of stochastic processes developed by Karatzas and Shreve, and in statistical mechanics contexts linked to work at Los Alamos National Laboratory and Princeton Plasma Physics Laboratory.

Examples and Counterexamples

Standard examples illustrating the two parts include coin-tossing sequences related to results by Émile Borel and sequences of rare events constructed by Paul Erdős to show sharpness. A classical counterexample to a naive converse uses dependent events such as those arising from overlapping intervals on the unit circle studied by Kurt Mahler and Vojtěch Jarník, which demonstrate that divergence of Σ P(An) need not imply almost sure occurrence without independence or extra hypotheses; this theme appears in work by A. Ya. Khintchine and Alexander Ostrowski. More refined counterexamples underpin the necessity of conditions in the Kochen–Stone theorem and related results by Paul Erdős and Alfréd Rényi.

Category:Probability theorems