On refinable sets (Q939683): Difference between revisions

This paper is devoted to the study of refinable sets which are generalizations of self-affine tiles \(T\), i.e., compact sets with positive Lebesgue measure such that \(A(T)= \bigcup_{j=1}^q (T+d_j)\), where \(d_j\in\mathbb{R}^d\) and \(A\in M_d(\mathbb{R})\) is an expanding matrix. This is closely connected to the refinement equation for \(f=\chi_T\). In general, this is the functional equation \[ f(x)=\sum_{j=1}^n c_j f(Ax-d_j)\quad\text{a.e.} \] with \(c_j\in\mathbb{R}\). The function \(f\) is called \(A\)-refinable accordingly. In particular, if \(A\in M_d(\mathbb{Z})\) and \(d_j\in\mathbb{Z}\), then it is called integral refinement equation and \(f\) integrally refined. The starting point for this paper is the common misconception, as pointed out in the introduction, that if \(\chi_T\) is refinable, then \(T\) is a self-affine tile. The purpose now is to study integrally refinable sets in the above sense in further detail. The main results are listed in Section~1, first collecting properties of refinable sets \(T\) such that \(\chi_T\) satisfies the above refinement equation (Theorem~1.1), then providing equivalent characterizations for integral refinable \(T\) (Theorem~1.2) and similarly for \(2\)-refinable sets (Theorem~1.3). Finally, there are given necessary and sufficient conditions for \(T=T_0+\mathcal{A}\) to be \(\mathcal{A}\)-refinable (Theorem~1.5) where \(T_0\) is an integral self-affine tile and \(\mathcal{A}\subset \mathbb{Z}^d\). The arguments to prove these outcomes include some further interesting properties of refinable sets. The paper ends with some examples and a list of open questions.