Parseval frames from compressions of Cuntz algebras (Q6038747)

Bounded operators \(T_1, \dots , T_n \in B(H)\) acting on a Hilbert space \(H\) are called a row co-isometry if \(\sum_{i=1}^n T_i T_i^* =1\). It is known that then there exists a so-called Cuntz dilation, that is, a larger Hilbert space \(H_2 \supseteq H\) and standard generators \(S_1,\dots , S_n\) of the Cuntz algebra \({\mathcal O}_n\) represented on \(H_2\) such that each \(T_i\) is the restriction of \(S_i\) onto the subspace \(H\). It may happen that the operators \(T_i\) act as a random walk on some vectors of \(H\). The authors assume such an assumption by considering given a map \(e: X \rightarrow H\) from a complete metric space \(X\) to a set of unit Hilbert space vectors in \(H\) which generate \(H\), and contractive transition maps \(g_1, \dots,g_n :X \rightarrow X\) and transition amplitudes \(\lambda_1, \dots,\lambda_n : X \rightarrow \mathbb{C}\) satisfying \(T_i^* e_x = \lambda_i(x) e_{g_i(x)}\) for all \(1 \le i \le n\) and \(x \in X\), and thus forming a random walk on \(X\) with transition probabilities \(|\lambda_i(x)|^2\) for the transitions \(x \rightarrow g_i(x)\), and then keeping track how these vectors evolve under the \(T_i^*\)s. At the end, they show for example that \(H_2\) can be decomposed into a finite direct sum of Hilbert subspaces generated by \(S_w(e(M))\) (\(w\) words) associated to a finite collection of minimal compact invariant subsets \(M\) of \(X\) with respect to the random walk. In a finite section, their results will be applied in a broad discussion to affine iterated function systems. A Parseval frame in a Hilbert space \(H\) is a collection of vectors \((e_i)_{i \in I}\) in \(H\) such that \(\|x \|^2 = \sum_{i \in I} |\langle x,e_i \rangle|^2\) for all \(x \in H\). They also show that the image under a certain element \(x_0 \in H\) of certain products \(T_w\) (\(w\) words) forms a Parseval frame in \(H\) if and only if \(S_w x_0\) forms an ONB in \(H_2\) if and only if the commutant of the Cuntz dilation is trivial. The paper is quite technical and the above represents only a glimpse about its content.