Zeta measures and thermodynamic formalism for temperature zero (Q711434): Difference between revisions
From MaRDI portal
Added link to MaRDI item. |
Removed claim: author (P16): Item:Q1118228 |
||
Property / author | |||
Property / author: Artur Oscar Lopes / rank | |||
Revision as of 10:00, 22 February 2024
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | Zeta measures and thermodynamic formalism for temperature zero |
scientific article |
Statements
Zeta measures and thermodynamic formalism for temperature zero (English)
0 references
26 October 2010
0 references
Let \(X= \{1,\dots,d\}^\mathbb{N}\) denote the Bernoulli space with \(d\) symbols. This is a compact metric space with metric \( d(x,y) = d_\theta(x,y)=\theta^N\) where \(N\) is given by \(x_i=y_i\) (\(i < N\)) and \(x_N \neq y_N\) for the sequences \(x=(x_i)_{i \in \mathbb{N}}\) and \(y=(y_i)_{i \in \mathbb{N}}\). Consider the Borel \(\sigma\)-algebra over \(X\) and \(\mathcal{M}\) the set of probability measures on it which are invariant under the shift map \(\sigma\). \(F_\theta\) denotes the set of real Lipschitz functions over \(X\). For strictly positive \(f \in F_\theta\) denote furthermore \[ \beta(f) = \sup_{\nu \in \mathcal{M}} \int f \, d\nu \quad \text{and} \quad M_{\max} (f) = \bigg\{ \nu \in \mathcal{M};\;\int f \, d\nu = \beta(f)\bigg\}. \] For real continuous functions \(k:X \to \mathbb{R}\) define the notation \[ k^n(x) := k(x) + k\big( \sigma(x)\big) + k\big( \sigma^2(x)\big) + \dots + k\big( \sigma^{n-1}(x)\big). \] \(\text{Fix}_n= \{x\); \(\sigma^n(x)=x\}\) denotes the set of \(n\)-periodic points and \(P(f)\) denotes the pressure for \(f\). The main object of the paper are probability measures \(\mu_{c,s} \in \mathcal{M}\) which satisfy for any continuous function \(k : x \to \mathbb{R}\) \[ \int k \, d \mu_{c,s} = \frac{\sum_{n\in \mathbb{N}} \sum_{x \in \text{Fix}_n} e^{cs f^n(x) - n P(cf) \frac{k^n(x)}{n}} }{\sum_{n\in \mathbb{N}} \sum_{x \in \text{Fix}_n} e^{cs f^n(x) - n P(cf)} }, \] where \(c >0\) and \(0 < s <1\). Define the function \(I(x)\) on periodic \(x\) for a strictly positive Lip\-schitz function \(f\) by \[ I(x) = n_x \left( \beta(f) - \frac{f^{n_x}(x)}{n_x} \right), \] where \(n_x\) denotes the minimum period of \(x\). The authors show \(\inf_{x \text{ periodic}} I(x) = 0\) and the following Theorem. Suppose \(f > 0\) is Lipschitz. When \(c \to \infty\) and \(s \to 1\), any accumulation point of \(\mu_{c,s}\) is in \(M_{\max}(f)\). Moreover, if \(g : X \to \mathbb{R}\) is a continuous function, \(c_j \to \infty\) and \(s_j \to 1\) are such that there exists \(\lim_{j \to \infty} \mu_{c_j,s_j}(g)\), then this limit is \(\int g \, d\mu\) for some accumulation point \(\mu\) of \(\mu_{c,s}\) in the weak star topology. The main result of the paper under review is the following large deviation principle for \(\mu_{c,s}\). Theorem. Suppose \(f >0\) is Lipschitz. Then, for any fixed \(0< L \leq \infty\), and for all cylinders \(k \subset X\) \[ \lim_{c(1,s) \to L} \frac{1}{c} \log \big( \mu_{c,s}(k) \big) = - \inf_{x \in k, x \text{ periodic}} I(x). \] The same is true if we have \(\liminf_{c \to \infty, s \to 1} c(1-s) = L\). It is also shown that the requirement of periodic \(x\) can be dropped in the previous large deviation principle.
0 references
zeta measure
0 references
maximizing probability
0 references
large deviation principle
0 references
Gibbs measure
0 references