An asymptotic analysis in thermodynamic formalism (Q652544)

From MaRDI portal
scientific article
Language Label Description Also known as
English
An asymptotic analysis in thermodynamic formalism
scientific article

    Statements

    An asymptotic analysis in thermodynamic formalism (English)
    0 references
    0 references
    14 December 2011
    0 references
    Let \(S=\{1,2,\ldots,d\}\) and let \(A\) be an irreducilbe \(d\times d\) matrix with entries in \(\{0,1\}\). Consider the subshift of finite type \((\Sigma_A^+,\sigma)\) with transition matrix \(A\). On \(\Sigma_A^+\), consider the distance function defined by \(d_\theta(\omega,\omega')=\theta^{\min\{n\geq0\colon \omega_n\neq\omega'_n\}}\), where \(0<\theta<1\). Let \(C(\Sigma_A^+)\) be the totality of continuous functions with supremum norm and \(F_\theta(\Sigma_A^+)\) be the totality of \(d_\theta\)-Lipshitz continuous functions with \(||f|_\theta=||f||_\infty+[f]_\theta\), where \([f]_\theta=\max_{i\in S}\sup\{|f(\omega)-f(\omega')|/d_\theta(\omega,\omega')\colon w_0=\omega_0'=i\}\). Let \(C(\Sigma_A^+\to\mathbb R)\) and \(F_\theta(\Sigma_A^+\to\mathbb R)\) be the real-valued functions belonging to \(C(\Sigma_A^+)\) and \(F_\theta(\Sigma_A^+)\), respectively. Denote by \(M(\Sigma_A^+)\) the totality of the complex Borel measures on \(\Sigma_A^+\). Now suppose that the following two assumptions hold: \begin{itemize}\item[1)] {\((\Phi)_n\colon\)} There exists functions \(\phi=\phi_0,\phi_1,\ldots,\phi_n,\Gamma_n(\epsilon,\cdot)\in F_\theta(\Sigma_A^+,\to\mathbb R)\) such that \(\phi(\epsilon,\cdot)\) has the form \(\phi(\epsilon,\cdot)=\phi+\phi_1\epsilon+\cdots+\phi_n\epsilon^n+\Gamma_n(\epsilon,\cdot)\epsilon^n\) and \(||\Gamma_n(\epsilon,\cdot)||_\infty\to0\) as \(\epsilon\to0\). \item[2)] {\((\Phi)_n'\colon\)} The condition\((\Phi)_n\) is satisfied and \(\sup_{\epsilon>0}|\Gamma_n(\epsilon,\cdot)|_\theta<\infty\). The main tool of this article is the Ruelle operator with potential \(\phi\) \[ \mathcal Lf(\omega)=\sum_{i\in S,A(i,\omega_0)=1}e^{\phi(i\cdot\omega)}f(i\cdot \omega), \] where \(i\cdot\omega\) is the concatenation of \(i\) and \(\omega\). We have the decomposition \[ \mathcal L=\lambda\mathcal P+\mathcal E, \] where \(\mathcal P\) is the projection onto the maximal eigenvalue \(\lambda\), and it has the form \(\mathcal P f=\nu(f)h\), \(h\in F_\theta(\Sigma_A^+)\), and \(\nu\in M(\Sigma_A^+)\) with \(\nu(h)=\nu(\Sigma_A^+)=1\). \item[4)] \(\mathcal E\in\mathcal L(C(\Sigma_A^+))\) satisfies \(\mathcal E\mathcal P=\mathcal P\mathcal E=O\), and the eigenvalues of \(\mathcal E|_{F_\theta(\Sigma_A^+)}\) is contained in \(|z-\lambda|>\eta\) for some \(\eta>0\). The topological pressure is defined by \[ P(\sigma,f)=\lim_{n\to\infty}\log\sum_{w\in S^n} \exp\left(\sup_{\omega\colon \omega_0\cdots\omega_n=w}\sum_{k=0}^{n-1}f(\sigma^k\omega)\right). \] Denote also the Gibbs measure with potential \(\phi\) by \(\mu=\mu_\phi\), and the measure-theoretical entropy by \(h(\sigma,\mu)\). The main theorems of this article are the following. \textbf{Theorem 2.1.} Assume that \(A\) is irreducible and the condition \((\Phi)_n\) is satisfied. Then there exist \(p_1,\ldots,p_n\in\mathbb R\) and linear functionals \(\nu_1,\ldots,\nu_n\) on \(F_\theta(\Sigma_A^+)\) such that \[ \begin{aligned} & P(\sigma,\phi(\epsilon,\cdot))=P(\sigma,\phi)+p_1\epsilon+\cdots+p_n\epsilon^n +o(\epsilon^n),\\ & \nu(\epsilon, f)=\nu(f)+\nu_1(f)\epsilon+\cdots+\nu_n(f)\epsilon^n+o(\epsilon^n), \quad(f\in F_\theta(\Sigma_A^+)).\end{aligned} \] \textbf{Theorem 2.4.} Assume that \(A\) is irreducible and the condition \((\Phi)_n'\) is satisfied. Then there exist functions \(h_1,\ldots,h_n\in F_\theta(\Sigma_A^+)\) such that \[ \begin{aligned} & h(\epsilon,\cdot)=h+h_1\epsilon+\cdots+h_n\epsilon^n)+o(\epsilon^n) \hbox{ in } C(\Sigma_A^+);\\ & \mu(\epsilon,\cdot)=\mu(f)+\mu_1(f)\epsilon+\cdots+\mu_n(f)\epsilon^n)+o(\epsilon^n) \quad(f\in F_\theta(\Sigma_A^+));\\ & h(\epsilon,\mu(\epsilon,\cdot))= h(\sigma,\mu)+h_1(\sigma)\epsilon+\cdots+h_n(\sigma)\epsilon^n)+o(\epsilon^n).\end{aligned} \] \textbf{Theorem 2.6.} Assume that \(A\) is irreducible and the condition \((\Phi)_n\) is satisfied. Assume moreover that \(\phi<0\). Then there exist \(s_1,\ldots,s_n\in\mathbb R\) such that \[ s(\epsilon)=s+s_1\epsilon+\cdots+s_n\epsilon^n)+o(\epsilon^n) , \] where \(s\) is an unique solution of the equation \(P(\sigma,s\phi+\xi)=0\).\end{itemize}
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    Ruelle operator
    0 references
    symbolic dynamics
    0 references
    thermodynamic formalism
    0 references
    asymptotic perturbation
    0 references
    0 references