scientific article; zbMATH DE number 1738334

Publication date 7 May 2002

branching exist Markov systems diffusions elliptic equations and diffusions linear and semilinear differential equations Markov processes superdiffusions

Mathematics Subject Classification ID

Nonlinear boundary value problems for linear elliptic equations (35J65) Nonlinear parabolic equations (35K55) Research exposition (monographs, survey articles) pertaining to probability theory (60-02) Diffusion processes (60J60) Branching processes (Galton-Watson, birth-and-death, etc.) (60J80)

Review text

This monograph is devoted to connections between a broad class of linear and semilinear differential equations and corresponding Markov processes, which are diffusions and superdiffusions, respectively. Here, a diffusion \(\xi\) is related to a second-order uniformly elliptic differential operator \(L\) (like the Laplacian \(\Delta\)). It describes the random motion of a single particle (like Brownian motion if \(L=\Delta\)). A superdiffusion \(X\) is related to an operator \(u\mapsto Lu-\psi(u)\) with \(L\) as above and with \(\psi\) from a certain class of functions (like \(\psi(u)= u^\alpha\), \(1< \alpha\leq 2\)), and describes the random evolution of a ``cloud of mass. Heuristically, the spatial spread of mass is governed by the mass flow corresponding to the diffusion \(\xi\) (that is, the heat flow if \(L=\Delta\)). Additionally, each bit of mass branches according to a continuous-state branching mechanism described by \(\psi\) (Feller's branching diffusion if \(\psi(u)= u^\alpha\) and \(\alpha= 2\)). In particular, the operator \(u\mapsto\Delta u- u^2\) corresponds to the continuous super-Brownian motion \(X\). The content of the book is based to a large extent on the author's (and his collaborators') own research since about 1988, and detailed historical notes are given in this direction. The concluding chapter touches upon some related models and results. This systematic, rigorous presentation is well suited for researchers interested in any of the fields described above, we will give more details. The book consists of two parts: 1. Parabolic equations and branching exit Markov systems, 2. Elliptic equations and diffusions. It is necessary to begin with the parabolic setting, since it is used for the construction of the basic probabilistic model in the book, the so-called branching exit Markov system. The class of superprocesses and, in particular, superdiffusions under consideration are special cases of them. A branching exit Markov system is a certain family of laws \((P_\mu)_{\mu\in\mathbb{M}}\), each describing a family \(X= (X_Q)_{Q\in\mathbb{O}}\) of random measures on a measurable space \((S,{\mathcal B}_S)\). Here, \(S\) is a Borel subset of a compact metric space (for example, \(S= \mathbb{R}\times \mathbb{R}^d\), representing time \(r\in \mathbb{R}\) and location \(x\in\mathbb{R}^d\)). \(\mathbb{O}\) is a subset of \({\mathcal B}_S\) (e.g. a class of open subsets of \(\mathbb{R}\times \mathbb{R}^d\)), and \(\mathbb{M}\) is a class of measures on \(S\) (e.g. all finite point measures on \(\mathbb{R}\times \mathbb{R}^d\)) containing all \(\delta\)-measures. (For instance, \(\mu= \delta_{(r,x)}\) is interpreted as a unit mass (particle) at site \(x\in\mathbb{R}^d\) at starting time \(r\in\mathbb{R}\).) The system is required to have a Markov property: Roughly speaking, if \(Q'\subset Q\subset Q\), then, given \(X_Q\), the ``past \(X_{Q'}\) and the ``future \(X_{Q}\) are independent. Moreover, the system has the branching property, which basically means, that if \(\mu\) is decomposed in \(\mu= \mu^1+ \mu^2\) (with \(\mu^1, \mu^2\in\mathbb{M}\)), then \(X\) with law \(P_\mu\) can be decomposed into \(X^1+ X^2\), where \(X^1\) and \(X^2\) are independent and have distributions \(P_{\mu^1}\) and \(P_{\mu^2}\), respectively. Finally, \(X\) is assumed to be an exit system, that is, we have, for all \(Q\) and \(\mu\), that \(X_Q(Q)= 0\), \(P_\mu\)-a.s., as well as the fact that \(\mu(Q)= 0\) implies \(X_Q= \mu\), \(P_\mu\)-a.s. We give a particle system example. If \(\mu= \sum \delta_{(r_i, x_i)}\), then particles start at times \(r_i\) from \(x_i\), \((r_i, x_i)\in Q\). They move according to independent copies of a right continuous strong Markov process \(\xi\) in \(\mathbb{R}^d\). Additionally, after i.i.d. exponential times, they are independently replaced by \(n\) particles (again at the same site) with probabilities \(p_n\), where \((p_n)_{n\geq 0}\) is a nice offspring law. But if \(\tau\) denotes the first exit time from \(Q\) of the process \(\xi\), then the evolution of a (surviving) particle is stopped at time \(\tau\). Now, intuitively, the exit measure \(X_Q\) is the collection of all the surviving particles \(\delta_{(\tau,\xi_\tau)}\). Continuous state analogues lead to other branching exit Markov systems, as superprocesses and superdiffusions. In the superdiffusion case, for nice \(Q\) and \(f\), the log-Laplace transform \(-\log P_\mu\exp\langle X_Q,- f\rangle\) equals \(\langle\mu, u\rangle:= \int u d\mu\), where \(u\) is the unique nonnegative solution to the semilinear equation (i) \({\partial\over\partial r} u+ Lu= \psi(u)\) in \(Q\), satisfying \(u(z)\to f(\widetilde z)\) as \(z\to\widetilde z\in \partial Q\). The (longer) Part 2 is devoted to the characterization of nonnegative solutions to be related semilinear elliptic equation (ii) \(Lu= \psi(u)\), in fact, to the characterization by their traces on the boundary, and to the study of their boundary singularities. As for (i), a combination of probabilistic and analytic tools is used to deal with (ii). To get a rough impression, the chapter titles of Part 2 are listed here: 6. Linear elliptic equations and diffusions, 7. Positive harmonic functions, 8. Moderate solutions\dots, 9. Stochastic boundary values of solutions, 10. Rough trace, 11. Fine trace, 12. Martin capacity\dots, 13. Null sets and polar sets. In two appendices, surveys are given of some basic facts about Markov processes, martingales, and elliptic differential equations, used in the book. In an epilogue (3 pp.), a few challenging open problems and possible directions of research in line with the presented material are listed.

Recommendations

Cited in

(only showing first 100 items - show all)

This page was built for publication:

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4529787)