Remarks on the life and research of Roland L. Dobrushin (Q675246)

This obituary notice is one in the series dedicated to the memory of Professor Roland L. Dobrushin (1929-1995), whose life and research work had a profound influence on various areas of probability theory, information theory, and mathematical physics. The first part of the paper contains a brief biography of Dobrushin. Paying special attention to the origins of his main ideas and to retrospective analysis of his methods, the authors present also a short story of his confrontation with the Soviet officialdom and its supporters. That were his very strong and independent personality and belief in democratic principles that put him on a collision course with the repressive machine. As a result, the authorities classified him as a ``non-voyager'' and did not allow him to travel beyond the Iron Curtain until 1988. Despite his scientific fame (Dobrushin was elected a Honorary Member of the American Academy of Fine Arts and Sciences in Boston (1982), an Associated Foreign Member of the USA National Academy (1993), a Member of the European Academy (1995)), he was treated as an outsider by the Soviet academic elite. The second part of the paper presents a detailed survey of main results obtained by Dobrushin, which is splitted into five sections: Markov processes, information theory, equilibrium statistical mechanics and the theory of random fields, non-equilibrium statistical mechanics and processes with local interaction, and queueing network theory. The authors present the material in a way accessible to a large probabilistic audience, maintaining at the same time the necessary level of mathematical rigor. In his first series of papers, Dobrushin studied the central limit theorem for a class of non-homogeneous Markov chains. He invented a natural specific parameter known as an ergodicity coefficient and used it to obtain the results, that are to a certain degree complete. Listing his main contributions to information theory, the authors discuss extensively his work related to the Shannon's theorem about the error probability in decoding a long message at the output port of a ``noisy'' channel. Dobrushin found the condition of information stability and showed that it is the main condition for the validity of Shannon's theorem in the case of arbitrary finite alphabets. The main problem he focused on at the beginning of sixties was the construction of probabilistic models of matter that exhibit the phenomenon of phase transition. According to him, the specification (the system of conditional distributions of a random field), not a field itself, is the primary object. From this point of view, the Gibbs' random field can be determined as a solution of the so-called Dobrushin-Lanford-Ruelle equation. He suggested a general approach to the concept of phase transition as the non-uniqueness of a random field with a given system of conditional distributions. In 1968, Dobrushin gave a mathematically rigorous proof of the existence of phase transition in the Ising model with zero magnetic field. He formulated a general sufficient condition of existence and uniqueness of a Gibbs' random field for a general state space and applied it to a large variety of problems. In a subsequent series of papers he gave a constructive criterion and showed that the corresponding Gibbs' random field has many nice properties known as ``complete analyticity''. In the non-uniqueness direction, Dobrushin generalized the concept of a contour and stated in explicit form the so-called Peierls condition, that is one of the cornerstones of the Pirogov-Sinai theory. One of his most impressive results is the theorem about the existence of non-translation-invariant Gibbs' random fields for the Ising model with a zero magnetic field in dimensions three and above. The authors list also his achievements in other directions of equilibrium statistical mechanics and related areas of probability theory. Dobrushin was the first who gave a formal construction of a process with local interaction and established a sufficient condition for its convergence to an invariant distribution; he showed that the reversible invariant distributions are precisely the Gibbs' random fields on \(\mathbb{Z}^d\) with a potential, which is naturally calculated in terms of the corresponding conditional probabilities. He proposed a new construction of the solution of the non-equilibrium dynamical systems of statistical mechanics, which allowed him to include (in dimensions \(d=1\) and \(2\)) some ``realistic'' potentials and to establish the existence and uniqueness of solution for a set of initial conditions of probability one. Based on his mathematically correct definition of the hydrodynamical limit, he managed to get the rigorous proof of the existence of such a limit in some relatively simple situations (called by him ``caricatures of hydrodynamics''). He proposed the so-called mean-field approach to the queueing network theory, developed the concept of an infinite network and treated the instability of the stationary regime of the latter as the non-uniqueness of an invariant distribution in a network. In the late eighties, Dobrushin started an active research of a large deviation approach to various problems, in particular, in queueing theory. He discovered the so-called bottleneck phenomenon for the waiting time in a tandem single-server network (that shows that the asymptotics of large deviation probabilities of this time is determined by the ``slowest'' server). The paper is completed by a list of Dobrushin's publications of 151 items [that, however, does not contain the recent lecture notes of his course given at the Saint-Flour Summer School in Probability Theory, in: Lectures on probability theory and statistics. Lect. Notes Math. 1648, 1-66 (1996; Zbl 0871.60086)] and a list of relevant references. Despite a small number of typographical errors (like the absence of any notations in Figure 4), the paper presents an informative and engaging account on Dobrushin's life and research and is of considerable value to anyone who is interested in contemporary mathematics.