Status quo:
We consider a irreducible, aperiodic Markov chain $(X_n)_{n\in\mathbb{N}}$ on a countable set $S$ with tranistion function $p(\cdot,\cdot)$.

Now we want to examine $\underset{n\rightarrow\infty}{\text{lim}}~p^n(i,j)$ for arbitrary $i,j\in S$.

If $j$ is transient we can show that $\sum_{n=0}^\infty p^n(i,j)<\infty$ and therefore we know that $\underset{n\rightarrow\infty}{\text{lim}}~p^n(i,j)=0$. If $j$ is positive recurrent we can show by nice coupling argument (see for example Durrett) that $\underset{n\rightarrow\infty}{\text{lim}}~p^n(i,j)=\pi(j)$, where $\pi$ is the unique stationary distribution of the chain.

The question:
My concern is the null recurrent case. I looked through a bunch of textbooks and I could not find a proof for the fact that $\underset{n\rightarrow\infty}{\text{lim}}~p^n(i,j)=0$ in this case. Most textbooks mention it but none of them gives a proof and I can't proof it either.
Any ideas, thoughts and especially references are highly appreciated. Thanks!

  • 1
    $\begingroup$ ?? Huh? Which textbooks did you check? $\endgroup$ – Did Mar 12 '15 at 17:46
  • $\begingroup$ @Did The usual suspects (Durrett, Klenke, Levin, lecturenotes) - maybe I was blind. But now I found it (see my answer). $\endgroup$ – MathProb Mar 12 '15 at 20:48
  • $\begingroup$ @Did Another thing since you have always been a help with my probability questions. Any thoughts on this: math.stackexchange.com/questions/1184926/… ? I would be very grateful. $\endgroup$ – MathProb Mar 12 '15 at 21:43

After some more research I found a nice answer in the book by Norris Theorem 1.8.5. It also makes use of a Coupling argument with the choice of a shifted measure.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.