Self-affine sets with positive Lebesgue measure (Q2252942)

Let \(\beta_2>\beta_1>1\) be two real numbers. For \(i\in\{-1,1\}\), define contractions \(T_i: \mathbb{R}^2\rightarrow \mathbb{R}^2\) by \(T_i(x,y)=({x+i \over \beta_1}, {y+i \over \beta_2})\). The present paper studies the self-affine attractors \(A_{\beta_1, \beta_2}\) of \(T_i\), i.e., the unique non-empty compact set satisfying \(A_{\beta_1,\beta_2}=T_{-1}(A_{\beta_1,\beta_2})\cup T_1(A_{\beta_1,\beta_2})\). Notice that when \(\beta_2<2\), the two contracted copies \(T_{-1}(A_{\beta_1,\beta_2})\) and \(T_1(A_{\beta_1,\beta_2})\) overlap, which makes the study of \(A_{\beta_1,\beta_2}\) complicated. By studying the absolute continuity of a certain measure on \(A_{\beta_1,\beta_2}\), \textit{P. Shmerkin} [Indiana Univ. Math. J. 55, No. 4, 1291--1331 (2006; Zbl 1125.28013)] proved that there exists an open set \(K\subset (1,2)^2\) such that for almost \((\beta_1, \beta_2)\in K\), the set \(A_{\beta_1,\beta_2}\) has positive Lebesgue measure. The present paper shows that \(A_{\beta_1,\beta_2}\) contains a neighborhood of \((0,0)\) for all \((\beta_1, \beta_2)\in (1,1+C)^2\), where \(C=\root{3}\of{\sqrt{10}-2} \approx 0.05\). To prove their main result, the authors follow the techniques of \textit{C. S. Güntürk} [``Simultaneous and hybrid beta-encodings'', in: Proceedings of the 42nd annual IEEE conference on information science and systems, CISS'08. Los Alamitos, CA: IEEE Computer Society. 743--748 (2008; \url{doi:10.1109/CISS.2008.4558620})] which are different from the usual fractal geometry techniques for studying self-affine sets. In fact, the authors prove that with the above mentioned constant \(C\), for any \(1<\beta_1<\beta_2<1+C\), there exists \(\delta=\delta(\beta_1, \beta_2)\) such that for any \((x_1, x_2)\in [-\delta, \delta]^2\), one can find a sequence \((a_n)\in\{-1, 1\}^{\mathbb{N}}\) such that \(\sum_{n=1}^\infty a_n \beta_k^{-n}=x_k\) (\(k=1,2\)). Using the same techniques, the authors also show that for all \(\beta\in (1, 1+C_1)\), where \(C_1>0\) satisfies \((1+C_1)+2(1+C_1)^3=6\), there exists \(\delta=\delta(\beta)\) such that for any \(x\in [-\delta, \delta]\) there exists a sequence \((a_n)\in\{-1, 1\}^{\mathbb{N}}\) satisfying \(\sum_{n=1}^\infty a_n \beta^{-n}=x=\lim_{n\to\infty} (a_1+\cdots + a_n)/n\). As pointed out by the authors, the constants \(C, \delta(\beta_1, \beta_2), C_1, \delta(\beta)\) are unlikely optimal. They strongly depend on some well-chosen polynomials in the proofs. Some better polynomials might lead to better bounds of the constants.