Rotational representations of stochastic matrices (Q5896465)

Let u denote Lebesgue measure on the interval \(X=[0,1)\). A stochastic matrix \(P=(p_{ij})\) is said to have a rotational representation if there exists a nonnegative number t and a partition \(S=\{S_ i\}^ n_{i=1}\) of X into nonempty sets \(S_ i\), which are each finite unions of intervals, such that \(p_{ij}=u(S_ i\cap f^{-1}(S_ j))/u(S_ i),\) 1\(\leq i,j\leq n\), where f is the u-preserving transformation of X into itself defined by \(f(x)=x+t(mod 1).\) It is shown that a stochastic matrix P has a rotational representation if and only if there exists a positive vector v such that \(vP=v\). This result implies the validity of the conjecture by \textit{J. E. Cohen} [ibid. 9, 899-901 (1981; Zbl 0469.15009)] that every irreducible stochastic matrix has a rotational representation.