Failure of the Gibbs inequality for continuous potentials

Arantha Ranu · 2026

Read Failure of the Gibbs inequality for continuous potentials by Arantha Ranu (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2601-19058.txt.

Introduction

Arantha Ranu shows a classic bound can break when the rules soften just a bit. The Gibbs inequality says a certain ratio stays trapped between two positive bounds. It holds for steady states of potentials that meet a smoothness test called Hölder continuity, on shift spaces.

For merely continuous potentials, Ranu shows the bound can fail. The gap between the two smoothness levels turns out to matter a great deal. Results of this kind map the true edge of theorems in dynamics, telling users exactly which assumptions carry the weight.

Catalog record

CREATORArantha Ranu
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE59 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2601-19058

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Failure of the Gibbs inequality for continuous potentials Arantha Ranu Abstract It is well known that the Gibbs inequality, which says that the Gibbs ratio is bounded above and below by positive constants, holds for the unique equilibrium states of Hölder continuous potentials on shift spaces, but it can fail for continuous potentials. In this article, we study the validity of a weaker form of the Gibbs inequality in this broader setting. 1 Introduction In [ 8 ] Walters proved the variational principle which states that if X is a compact metric space and T:X\to X is a continuous map on X and \psi is a continuous fucntion from X to \mathbb{R} , then the pressure P(\psi) of the the potential is given by, \displaystyle P(\psi)=\sup_{\eta}\left(h_{\eta}(\sigma)+\int\psi\,d\eta\right) where the supremum is taken over all T - invariant measures \eta and h_{\eta}(T) is the entropy of T with respect to \eta . In [ 2 ] Bowen showed that for continuous potentials \psi on shift spaces the supremum is a maximum. The measures at which the maximum is attained are called equilibrium states . In [ 2 ] it is further proved that for Holder continuous potential \psi , there exists a unique equilibrium state \mu_{\psi} . Also, there exists c>0 such that, \displaystyle\frac{1}{c}<\frac{\mu_{\psi}(W_{n}(z))}{e^{-nP(\psi)+S_{n}\psi(z)}}<c (1) for all n - length word W_{n} in the language of the finite shift and all z\in W_{n} . The unique equilibrium state is called a Gibbs state and the property ( 1 ) satisfied by it is called the Gibbs property This is a very important and useful property and has various uses. One of them is to find the Hausdorff dimensions of certain dynamical objects, an example of which can be found in the paper [ 3 ] . Now a natural question is to ask if such upper and lower bounds still exist if we weaken continuity condition on the potentials. In [ 6 ] Huyi Hu mentions the “weak Gibbs measures” for potentials \psi that are Holder at all but one point. The equilibrium states of these potentials satisfy, \displaystyle\frac{1}{p(z,n)}<\frac{\mu_{\psi}(W_{n}(z))}{e^{-nP(\psi)+S_{n}\psi(z)}}<p(z,n) (2) where \log_{n\to\infty}\frac{1}{n}\log p(z,n)=0 for all z except for the point at which the function is non Holder. In [ 5 ] Hofbauer gave an example of a continuous potential on the one-sided shift on two symbols that has two equilibrium states. One of the equilibrium states for this example is the Dirac measure at the point of all ones. This shows that there there is no suitable positive lower bound for every point in the shift space for the class of continuous potentials. Now the question arises, does there exist a suitable upper bound for the class of continuous potentials? For a shift space (Z,\sigma) and a Holder continuous potential \psi we have the distortion property i.e. for all n\in\mathbb{N} there exists C>0 such that \displaystyle\frac{1}{C}<\frac{e^{S_{n}\psi(z)}}{e^{S_{n}\psi(z)}}<C for all z,z^{\prime} in the same n- cylinder. A weak version of the distortion property holds for the class of continuous potentials as follows, for a continuous potential \psi there exists C>0 such that for all n\in\mathbb{N} . \displaystyle Ce^{-n\epsilon}<\frac{e^{S_{n}\psi(z)}}{e^{S_{n}\psi(z^{\prime})}}<Ce^{n\epsilon} for all z,z^{\prime} in the same n- cylinder. This weaker version of distortion motivates us to define the “very weak Gibbs inequality” and the above question about a suitable upper bound for the Gibbs ratio of continuous potentials will be answered by proving the following theorems. Definition 1.1 . Let \psi be a continuous potential on a shift space (Z,\sigma) that has an equilibrium state \mu . Let z\in Z and for n\in\mathbb{N} let [{W_{n}}] be the n -cylinder containing z . We say that \psi satisfies the “very weak Gibbs inequality” at z if for \epsilon>0 there exists C(z,\epsilon) such that the following is true, \displaystyle\frac{1}{C(z,\epsilon)e^{n\epsilon}}<\frac{\mu[W_{n}]}{e^{-nP(\psi)+S_{n}\psi(z)}}<C(z,\epsilon)e^{n\epsilon} (3) Theorem 1.1 . For a continuous potential \psi on a shift space (Z,\sigma) that has an ergodic equilibrium state \mu . The “very weak Gibbs inequality” is satisfied for \mu a.e. z . As we mentioned earlier the example in [ 5 ] has the Dirac measure as one of its equilibrium states which means the lower bound of the “very weak Gibbs inequality” is not true for every point in the shift space. The following Theorem states that \frac{\mu[W_{n}]}{e^{-NP(\psi)+S_{N}\psi(o)}} grows exponentially which shows that the upper bound is also not true for the point o of all \beta^{\prime} s. Theorem 1.2 . There exists a continuous potential \psi on a full shift that has a unique equilibrium state \mu such that for all n\geq 5 we have \displaystyle\frac{\mu[W_{n}]}{e^{-NP(\psi)+S_{N}\psi(o)}}>\left(\frac{e}{2}\right)^{n} where o is a fixed point of \sigma and W_{n} is the n - word containing o . Theorem 1.2 is proved in section 5 and Theorem 1.1 is proved in section 6 . 2 Preliminaries In this section, we will recall some properties of odometers and shift spaces. Properties of odometer maps:- Let X=\{0,1\}^{\mathbb{N}_{0}} where an element x\in X is of the form x=(x_{i})_{i\in\mathbb{N}_{0}} . Definition 2.1 . An n - cylinder in X denoted by [(a_{0}\ldots a_{n-1})]_{X} is defined as [a_{0}\ldots a_{n-1}]_{X}=\{x\in X:x_{0}=a_{0},\ldots x_{n-1}=a_{n-1}\} , for n\in\mathbb{N} and a_{0},\ldots,a_{n-1}\in\{0,1\} . Definition 2.2 . Let x\in[a_{0}\ldots a_{n-1}]_{X}\subset X . For k\in\mathbb{N} , we denote \mathcal{N}_{k}(x)=a_{k-1}a_{n-2}\ldots a_{0} which is the k - digit binary sequence that x ends with. Definition 2.3 . Define \mathcal{DN}_{k}(x)=2^{k-1}a_{k-1}+\ldots+2a_{1}+a_{0} which is the number corresponding to the binary sequence \mathcal{N}_{k}(x) . Definition 2.4 . The map T:X\to X called the odometer map is defined as, T(x)_{k}= \begin{cases}0&\text{if, }\mathcal{N}_{k}(x)=11\ldots 11\\ 1&\text{if, }\mathcal{N}_{k}(x)=01\ldots 11\\ x_{k}&\text{otherwise}\end{cases} Here are a few examples to show how the map works. \displaystyle T(\ldots 011110)=\ldots 011111 \displaystyle T(\ldots 011111)=\ldots 100000 The odometer map can be seen as “adding 1 with carry in binary” . We conclude the following from this:- • \mathcal{DN}_{k}(T(x))=\mathcal{DN}_{k}(x)+1\mod 2^{k} . • The set, \{\mathcal{DN}_{k}(x),\mathcal{DN}_{k}(T(x)),\ldots,\mathcal{DN}_{k}(T^{2^{k}-1}(x))\} is equal to \{0,..,2^{k}-1\} . • For the k - cylinder set [a_{0}a_{1}\ldots a_{k-1}]_{X} , for all x\in X , there exists i\in\{0,\ldots,2^{k}-1\} such that T^{i}(x)\in[a_{0}a_{1}\ldots a_{k-1}]_{X} . The inverse of the odometer map is defined as, T^{-1}(x)_{k}= \begin{cases}1&\text{if, }\mathcal{N}_{k}(x)=00\ldots 00\\ 0&\text{if, }\mathcal{N}_{k}(x)=10\ldots 00\\ x_{k}&\text{otherwise}\end{cases} Definition 2.5 . For x\in X , let \kappa(x) denote the number of zeros that x ends with. It is defined as follows: \displaystyle\kappa(x)=\begin{cases}n&\text{if }x\in[10^{n}]_{X}\\ \infty&\text{if }x=\overline{0}\text{ is the all zero element.}\end{cases} Definition 2.6 . Define the metric d_{X}(x,x^{\prime})=2^{-\min\{i:x_{i}\neq x_{i}^{\prime}\}} for x,x^{\prime}\in X . Note:- The Borel \sigma - algebra on X is generated by the cylinder sets and they are clopen sets. Definition 2.7 . We define a measure \nu on the Borel \sigma - algebra of X by specifying the measure of cylinder sets as, \displaystyle\nu([a_{0}\ldots a_{n-1}]_{X})=\frac{1}{2^{n}} for each n and each cylinder [a_{0},\ldots a_{n-1}]_{X} . It is a well known fact that the odometer map T is uniquely ergodic and \nu is the unique invariant ergodic measure [ 4 ] . Also from [ 9 ] (pg. 101 Theorem 4.25) the entropy of T w.r.t \nu is 0 . Properties of shift spaces:- Let Z=\{\alpha,\beta\}^{\mathbb{Z}} be the full shift on the alphabet \{\alpha,\beta\} . An element z\in Z is denoted as : \ldots z_{-2}z_{-1}.z_{0}z_{1}z_{2}\ldots where the “.” separates positive and negative indices. Definition 2.8 . The cylinder sets in Z are of the form [\omega_{0}\ldots\omega_{n}]_{Z}=\{z:z_{0}=\omega_{0},\ldots,z_{n}=\omega_{n}\} or [\omega_{-i}\ldots\omega_{-1}.\omega_{0}\ldots\omega_{j}]_{Z}=\{z:z_{-i}=\omega_{-i},\ldots,z_{-1}=\omega_{-1},z_{0}=\omega_{0},\ldots z_{j}=\omega_{j}\} where the “.” separates the positive and the negative

Read the complete text