I'm looking to understand how to approximate certain countable alphabet subshifts by Markov shifts, and realised that I don't know how to do it even in the finite alphabet case. My guess is that the answer to the following question is well known, but I couldn't work it out this morning.

Let $X\subset\{0,1\}^{\mathbb Z}$ be closed and shift invariant. It is a classical result that there exist Markov shifts $X^n\supset X$ such that $$\lim_{n\to\infty} h_{top}(X^n)= h_{top}(X),$$ where

$$h_{top}(X):=\lim_{n\to\infty}\frac{1}{n}\log(\text{ no. of words of length $n$ in $X$}). $$

This is easy to see, one just lets $X^n$ be the space made by freely concatenating all words of length $n$ in $X$, which is $n$-step Markov.

Question: Can we do such an approximation from below? i.e. Can we find Markov shifts $X_n\subset X$ with $\lim_{n\to\infty} h_{top} X_n= h_{top}(X)?$


If $X$ is minimal and not a periodic orbit then it cannot contain a periodic orbit and hence in particular cannot contain a Markov shift. A classical construction by Grillenberger shows that one can construct uniquely ergodic (hence in particular minimal) subshifts with arbitrary entropy. (This result also follows from the symbolic version of the Jewett-Krieger Theorem.)

  • $\begingroup$ It should be Jewitt-Krieger. $\endgroup$ – MHS Nov 12 '14 at 1:36
  • $\begingroup$ My spelling of "Jewett" was correct but I had indeed mis-spelled "Krieger". Thanks. $\endgroup$ – Ian Morris Nov 12 '14 at 11:11

Nevertheless there are approximations from below for other classes of subshifts, like sofic or coded systems. Depends on what your subshift looks like whether there is a sequence of proper subshifts approximating it in entropy from below or not.

  • 1
    $\begingroup$ care to expand your answer? $\endgroup$ – Rodrigo A. Pérez Nov 12 '14 at 4:01
  • $\begingroup$ What do you want me to explain in more detail? $\endgroup$ – MHS Nov 12 '14 at 13:49
  • $\begingroup$ Thanks! In fact the article that Vaughn linked to is quite helpful for understanding when this can be done. $\endgroup$ – Tom Kempton Nov 14 '14 at 12:22

I guess that you can find such a statement in the first articles of Omri Sarig about thermodynamical formalism.

If your shift is mixing, then the pressure/entropy of your shift coincides with the supremum of pressure/entropy over all compact invariant subsets in restriction to which the shift is topologically mixing.

  • $\begingroup$ Thanks Barbara, unfortunately my shift needn't be Markov, or even recodable as a countable Markov shift, so applying the results of Omri seems difficult. I've been trying to apply his methods, sometimes you can but sometimes not... $\endgroup$ – Tom Kempton Nov 17 '14 at 10:13

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.