Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Recently I've found lecture notes on "Analysis on Graphs" where the potential theory methods were used to study discrete-time, time-reversible Markov chains (i.e. the state space is countable).

Among such methods there was the Dirichlet problem for Laplace equations; Cheeger and isoperimetric inequalities were used to lower-bound second eigenvalue of the Laplace operator. On of the main tool used was discrete Green's formula: $$ \int\limits_A\Delta f(x)g(x)\pi(dx) = -\frac12\int\limits_{A}\nabla_{xy}f\cdot\nabla_{xy}g\;\pi(dx)+\int\limits_{A}\int\limits_{A^c}(\nabla_{xy}f)g(x)P(x,dy)\pi(dx). $$ where the set $A$ is measurable, $\Delta = I-P$ is a Laplace operator ($P$ is a Markov operator, $I = P^0$ is the identity one) and $\nabla_{xy}f = f(x) - f(y)$ is the discrete gradient.

Clearly, for the uncountable state space the same formula holds (for any measurable $A$) if the Markov process is time-reversible w.r.t. measure $\pi$. I am interested if in the general case (not time-reversible) Green's formula is still so powerful tool, and which measure $\pi$ should be taken.

With this thoughts I started looking for the book on potential theory for discrete-time, general-space Markov processes. Currently I am reading 'Markov Chains' by D. Revuz, there is a chapter on potential theory - but methods used there are different and based mostly on potential operator $$ G = I + P + P^2+\dots $$

My questions are the following:

  1. I am looking for the reference which treats potential theory for Markov processes in a discrete time with the use of Dirichlet forms (even if it is fruitful only for time-reversible processes). Especially I am interested in the functions which are minimums for these forms. Again, in the general case, which measure $\pi$ should be used?

  2. Maybe, there is book devoted to potential theory of discrete-time Markov processes in general, because usually I saw only chapters in other books.

  3. In his book, Revuz considers only non-negative superharmonic functions. This does not seem to be a restriction (w.r.t. considering bounded from below superharmonic functions) since e.g. Riesz decomposition holds also in the latter case - it just brings a bit of an inconvenience, since for the supmarkovian kernels constants are not always harmonic, so simple shift does not help. Is there any reason to consider non-negative superharmonic functions rather than bounded from below?

The part of this question I asked on MO but didn't get an answer.

share|cite|improve this question
    
+1 Hope you can get an answer soon. – Tim Jan 21 '12 at 17:16

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Browse other questions tagged or ask your own question.