Finitary codings of finite expectation between shifts over free groups

James O'Quinn

2023-05-04

Abstract:We prove that finitary isomorphisms with finite expectation exist between Bernoulli shifts over the same free group only if the shifts have the same distribution. This generalizes the integer case result of Schmidt. We provide new proofs of some of the technical results of Parry and Schmidt which avoid the assumption of amenability.

Dynamical Systems,Probability

What problem does this paper attempt to address?

The core problem that this paper attempts to solve is how to generalize the concept of finitary codings with finite expected code length to Bernoulli shifts over free groups. Specifically, the author aims to prove that Bernoulli shifts that are finitary coding isomorphic with finite expected code length over free groups must have the same weight distribution. This can be regarded as a generalization of Schmidt's result on finitary codings with finite expected code length.

### Background and Motivation of the Main Problem

1. **Ornstein Isomorphism Theorem**:
- The Ornstein Isomorphism Theorem states that Bernoulli shifts on the integers can be classified by their entropy. Since Ornstein first proved this theorem in 1970, much research has been devoted to generalizing it to broader classes of groups.

2. **Finitary Equivalence**:
- A stronger form of equivalence is finitary equivalence, which requires a conjugacy mapping called finitary coding isomorphism between two Bernoulli shifts. The characteristic of this mapping is that the image of almost every point can be determined by an algorithm of finite - block codes.

3. **Finitary Codings with Finite Expected Code Length**:
- Schmidt et al. further studied finitary codings with finite expected code length and proved that Bernoulli shifts under such codings must have the same distribution. This result depends on the information cocycle and some measure - theoretic tools.

### Main Contributions of the Paper

- **Generalization to Free Groups**:
- The main contribution of this paper is to generalize the above results from the integer group to free groups. The author utilizes the special properties of free groups (such as being able to implement a Turing - machine - like scheme within the group) and, through the methods of information cocycle and asymptotic averaging, obtains a similar conclusion.

- **Technical Difficulties and Innovations**:
- The key technical difficulty lies in how to handle Krieger's observations in non - amenable groups. The author provides a new proof method that does not rely on the pointwise ergodic theorem, thus bypassing the tools related to amenability.

### Formula Summary

Some of the key formulas involved in the paper are as follows:

1. **Definition of Finite Expectation**:
\[
\int (a(x)+b(x))d\mu_p(x)<\infty
\]
where \(a(x)\) and \(b(x)\) respectively represent the smallest positive integers such that certain conditions given \(x\) are satisfied.

2. **Definition of Information Cocycle**:
\[
J(A, V)=I_{\mu_p}(A|V^{-1}A)-I_{\mu_p}(V^{-1}A|A)-\log E\left(\frac{d\mu_p V^{-1}}{d\mu_p}\bigg|A\right)
\]

3. **Definition of Beta Function**:
\[
\beta_{p,a}(t)=P_T(-tI_{\mu_p})
\]
where \(I_{\mu_p}\) is the conditional information function related to the past algebra \(A_{p,a}\).

4. **Specific Expression of Beta Function**:
\[
\beta_{p,a}(t)=\frac{P_1^t+\cdots+P_m^t}{1}
\]

These formulas show how to study the equivalence and invariants of Bernoulli shifts through information - theoretic and measure - theoretic tools.