Multifractal formalism for self-affine measures with overlaps (Q2390920): Difference between revisions

The authors of this interesting paper essentially contribute to the study of different spectra of self-affine measures as well as to the so-called multifractal formalism. The main results are the following. For a \(d\times d\) integral matrix \(A\) all of whose eigenvalues have the same modulus \(\lambda>1\), and the unique self-affine measure \(\mu\) defined via \[ \mu(\cdot)= \sum_{i=1}^m p_i \mu(S_i^{-1}(\cdot)), \] where \(S_i(x)=A^{-1}(x+d_i)\), \(d_i\in\mathbb{Z}^d\), and \(p_i\) are probability weights, the corresponding \(L^q\)-spectrum \(\tau(q)\) exists for all \(q\geq 0\) and is differentiable in \(q>0\). Moreover, for the dimension spectrum \(f_\mu(\alpha)\) of \(\mu\), it can be proved that \(f_\mu(\alpha)=\tau^\ast(\alpha) = \inf\{\alpha q-\tau(q), q\in\mathbb{R}\}\) for all \(\alpha=\tau'(q)\). This is the well-known multifractal formalism, i.e., the celebrated heuristic principle \(f_\mu(\alpha)=\tau^\ast(\alpha)\) known to be false in general, but verified in many cases. However, for self-affine measures as studied here little was known before.