Poisson boundary

Poisson boundary
Name	Poisson boundary
Field	Probability theory; Ergodic theory; Geometric group theory
Introduced	1960s
Notable contributors	Furstenberg; Kaimanovich; Vershik; Dynkin; Doob; Grigorchuk; Kesten; Varopoulos

Poisson boundary

The Poisson boundary is a measure-theoretic construction that captures the asymptotic behavior of random walks on groups, graphs, and Markov chains, and identifies bounded harmonic functions via a boundary representation. It plays a central role in ergodic-theoretic descriptions of entropy, harmonic analysis on groups, and rigidity phenomena in geometric group theory and probability on networks.

Introduction

The Poisson boundary originated in work by Leonid Khinchin-era probability theorists and was systematized by Harry Furstenberg in studies of random walks on Lie groups and homogeneous spaces, with later contributions from Eugene Dynkin, Joseph Doob, and Andrei Kolmogorov. It provides a link between asymptotic sigma-algebras of Markov processes, the space of bounded harmonic functions examined by David Blackwell, and entropy notions developed by Yakov Sinai and Anatole Katok. Applications span from rigidity results for lattices in Semisimple Lie groups investigated by Grigory Margulis to boundary theory for random walks on groups studied by Mikhael Gromov and Rostislav Grigorchuk.

Definitions and constructions

One standard construction starts with a probability measure on a countable group such as a finitely generated Free group or a lattice in a Lie group, defines a right random walk, then considers the tail sigma-algebra of path space as in the work of Furstenberg and represents bounded harmonic functions by integrals over the tail space. Alternative approaches use compactifications: the Martin compactification, introduced by Robert Martin, and the minimal Martin boundary relate to potential-theoretic kernels first studied by Andrey Kolmogorov and Eugene Dynkin. Measurable models include the space of sample paths with shift-invariant measures studied by Donald Ornstein and Benjamin Weiss, and the identification of the Poisson boundary with a quotient of the path space under the shift.

Examples and computations

For simple random walk on the integer lattice studied by George Pólya, the Poisson boundary is trivial for recurrent cases like the one- and two-dimensional Euclidean plane lattice, while for nonamenable graphs such as Cayley graphs of nonamenable Free groups it is nontrivial. For random walks on Hyperbolic groups in the sense of Mikhael Gromov, the Poisson boundary often coincides with the Gromov boundary as shown in work by Vladimir Kaimanovich and Wolfgang Woess. Specific computations include the identification of boundaries for random walks on SL(2,R), lattices in SO(n,1), Lamplighter groups studied by Yuval Peres, and branching random walks linked to results by Russell Lyons and Luczak Martin.

Properties and invariants

The Poisson boundary is invariant under factors of the underlying Markov chain and under measure-preserving isomorphisms of the path-space flow, akin to rigidity phenomena proved by Margulis for higher-rank lattice actions. Entropy and the asymptotic entropy of random walks, concepts developed by Avez and Furstenberg-entropy frameworks, give numerical invariants that determine boundary triviality in many settings; Kesten's spectral radius theorem relates spectral radius invariants to boundary behavior in works by Harry Kesten. Amenability for groups introduced by John von Neumann yields trivial Poisson boundary in many symmetric measures, while nonamenability frequently produces rich boundary structures as in Kesten's criteria and Kaimanovich-Vershik entropy criteria.

Connections to harmonic functions and potential theory

The fundamental representation theorem asserts that bounded harmonic functions correspond bijectively to L∞ functions on the Poisson boundary, generalizing the classical Poisson integral representation on the unit disc studied by Gaston Julia and Pierre Fatou. The Martin boundary refines this connection by representing positive harmonic functions via minimal harmonic kernels, following classical potential theory developed by Riesz and Choquet. Doob's h-transform and conditioned processes provide probabilistic constructions of harmonic functions used in works by Eugene Dynkin and James L. Doob to interpret boundary limits and relative entropy along conditioned paths.

Applications in probability and group theory

Poisson boundaries appear in classification problems for group actions, rigidity theorems for lattices in Semisimple Lie groups, and characterization of random walk behavior on groups such as Free groups, Hyperbolic groups, and solvable groups like the Lamplighter group. They underpin entropy-based proofs of nonelementary action properties used by Kaimanovich and Vershik and inform spectral gap results related to expanders constructed via work of Margulis and Lubotzky. In ergodic theory, Poisson boundary techniques intersect with orbit equivalence studied by Miller and Hjorth and with stationary measure classification problems addressed by Benoist and Quint.

Techniques and key theorems

Key tools include entropy theory for random walks developed by Kaimanovich and Vershik, the strip criterion and ray criterion for identification of boundaries, and the use of coupling and martingale convergence theorems traced to Doob and Kolmogorov. The identification of Poisson and Martin boundaries uses Harnack inequalities and Green kernel estimates from potential theory, often relying on geometric estimates from Gromov hyperbolicity or isoperimetric inequalities from Varopoulos and Coulhon. Prominent theorems include Furstenberg’s boundary theory for random walks on Lie groups, Kaimanovich-Vershik entropy criteria for triviality, and the Kesten amenability criteria linking spectral radius to return probabilities.

Category:Probability theory Category:Ergodic theory Category:Geometric group theory