Normal sets related to non-commuting matrices (Q677638)

The \(r\times r\) real matrix \(A\) is ergodic if no eigenvalue of \(A\) is a root of unity, in which case for each \(v\in \mathbb{R}^r\), the sequence \((A^nv)_{n \geq 0}\) is uniformly distributed in \([0,1)^r\) and \(v\) is said to be normal with respect to the matrix \(A\). The \(r\times r\) rational matrix \(C\) is quasi-integral if its eigenvalues are algebraic integers. A monic integral polynomial \(P\) such that all its irreducible factors \(Q\) have one dominant root \(\lambda>1\) (for all other roots \(\mu\) of \(Q\), \(|\mu |< \lambda)\) is called a Perron polynomial. The collection of dominant roots is called the Perron spectrum of \(P\). The author proves the following theorem. Suppose that \(A\) is a quasi-integral ergodic matrix and \(C\) is a quasi-integral matrix with characteristic polynomial \(P\) a Perron polynomial. Suppose also that \(A\) has an eigenvalue \(\beta\) either in \(\mathbb{N}\) or a Pisot number such that \(\log\beta/ \log \theta\) is irrational, where \(\theta\) is in the Perron spectrum of \(P\). Then \(B(C) \subset B(A)\) where \(B(A)\) is the set of vectors in \(\mathbb{R}^n\) normal with respect to \(A\). When \(A\) and \(C\) commute, the relationship between \(B(C)\) and \(B(A)\) is better understood [\textit{G. Brown} and \textit{W. Moran}, Invent. Math. 111, 449-463 (1993; Zbl 0773.11049)].