Multifractal analysis of Birkhoff averages for typical infinitely generated self-affine sets (Q2444874): Difference between revisions

This long work studies the thermodynamic formalism for quasi-multiplicative potentials over countable alphabets. Several technical results are established, which lead to four main theorems: one variational principle and three multifractal dimension statements (which are novel even in the finitely generated setting). As an appetizer, we quote only Theorem B, a dimension result for an iterated function system constituted by a countable collection of contractive affine maps \(\{f_i\}_{i\in\mathbb{N}}\), \(f_i:\mathbb{R}^{d}\to\mathbb{R}^{d}\), acting on \(d\)-dimensional Euclidean space. Let \(f_i=T_i + a_i\), where the linear part \(T_i\) of \(f_i\) is uniformly contractive with \(\sup_{i\in\mathbb{N}} \|T_i\| < 1/2\), and \(a_i\in {[0,1]}^{d}\). Until the statement of Theorem B, \(a = (a_i)_{i\in\mathbb{N}}\) is fixed, though it appears as a subscript. Define the projection of the symbolic space \(\mathbb{N}^\mathbb{N}\) to be \(\pi_{a}:\mathbb{N}^\mathbb{N} \to \mathbb{R}^{d}\), \(\pi_{a}(\omega) = \lim_{n\to\infty} f_{\omega_{1}} \circ \dots\circ f_{\omega_{n}} (0)\) for \(\omega = (\omega_{1},\dots,\omega_{n},\dots)\in \mathbb{N}^\mathbb{N}\). Then the set \(F_{a} = \pi_{a}(\mathbb{N}^\mathbb{N})\) is self-affine, \(F_a=\bigcup_{i\in\mathbb{N}} f_{i}(F_a)\). The singular value function of \(T\in \roman{GL}_{d}(\mathbb{R})\) is set to be \[ \varphi^{s}(T) = \gamma_{1}(T)\dots \gamma_{m}(T)\gamma_{m+1}(T)^{\delta}, \] when \(0\leq s=m+\delta\leq d\), \(m\in\mathbb{Z}\), \(0< \delta \leq 1\), and \(\varphi^{s}(T) = |\det T|^{s/d}\), when \(s\geq d\). Here \(\gamma_{1}(T)\geq \dots\geq \gamma_{d}(T) >0\) are the singular values of \(T\), i.e., the square roots of the eigenvalues of \(T^{*}T\). Write \(\Sigma_{*} = \bigcup_{n\in\mathbb{N}} \mathbb{N}^{n}\) for the set of finite words. The map \(\varphi: \Sigma_{*} \to (0,\infty)\) is a sub-multiplicative potential, if \(\varphi({\tau}{\upsilon}) \leq \varphi(\tau)\cdot \varphi(\upsilon)\), where \(\tau, \upsilon\in \Sigma_{*}\) and \({\tau}{\upsilon}\) is a concatenation of \(\tau\) and \(\upsilon\). It is quasi-multiplicative, if additionally there exist a constant \(c\geq 1\) and a finite subset \(\Gamma\subset \Sigma_{*}\) such that for each \(\tau, \upsilon \in\Sigma_{*}\) one can find \(\varkappa\in \Gamma\) so that \(\varphi(\tau)\cdot \varphi(\upsilon) \leq c\cdot \varphi(\tau{\varkappa}\upsilon)\). Given a sub-multiplicative potential \(\varphi\) we define the pressure \[ P(\varphi) = \lim_{n\to\infty} \frac{1}{n} \log \sum_{\tau\in\mathbb{N}^{n}} \varphi(\tau). \] Theorem B: For almost all parameters \(a = (a_i)_{i\in\mathbb{N}} \in ([0,1]^{d})^{\mathbb{N}}\), the Hausdorff dimension of the self-affine set \(F_{a}\) of the IFS \(\{f_i = T_i + a_i\}_{i\in\mathbb{N}}\) on \(\mathbb{R}^{d}\), \(\sup_{i\in\mathbb{N}} \|T_i\| < 1/2\), can be expressed as \[ \text{(i)}\;\; \dim_{H}(F_a) = \min \{d, \inf \{s: P({\varphi}^s) \leq 0\} \}, \] \[ \text{(ii)}\;\; \dim_{H}(F_a) = \sup \{\dim_{H}(\,\pi_{a}(I^{\mathbb{N}})\,) : I\subset\mathbb{N} \;\text{is finite}\}, \] under the assumption that the associated potential \(\varphi^{s}(\tau_{1}\dots\tau_{n}) := \varphi^{s}(T_{\tau_{1}}\circ\dots T_{\tau_{n}})\) is quasi-multiplicative for all \(0\leq s\leq d\). Almost all in \(([0,1]^{d})^{\mathbb{N}}\) is meant in the sense of the countable product \({\bigotimes}_{\mathbb{N}}\; ([0,1]^{d}, \mathcal{L})\) of Lebesgue probability measures \(\mathcal{L}\) on \([0,1]^{d}\). Formula (ii) in Theorem B can be thought of as a variational principle for a countable system over its finite subsystems. According to Proposition 2.3.(3) in the article, \(\varphi^{s}\) is quasi-multiplicative provided \(T_{i} = \text{diag} (t_{1}^{i}, \dots, t_{d}^{i})\), \(1> |t_{1}^{i}| > \dots > t_{d}^{i} > 0\) for all \(i\in \mathbb{N}\). According to Remarks 2.4 in the article, quasi-multiplicativity may be a prevalent phenomenon in general, which is known to be true at least for \(d=2\).