On Markov operators preserving polynomials (Q2338874)

Let \(K\) be a convex compact subset of \({\mathbb R}^{n} \) with nonempty interior. A positive linear operator \(T:C(K)\to C(K)\) is called a Markov operator if \(T(1)=1\). Let \(P_{m} (K)\) be the set of polynomials of degree \(\leq m\). The authors consider Markov operators which leave invariant \(P_{\, 1} (K)\) and prove various interesting results for them. In particular, it is shown that, if such an operator \(T\) exists with the additional property \(T(P_{\, 2} (K))\subseteq P_{\, 1} (K)\), then \(K\) is a simplex and \(T\) is the canonical projection associated with it. In this case, \(T(P_{\, m} (K))\subseteq P_{\, 1} (K)\) for every \(m\geq 2\) (Theorem 3.2). It is also shown that, if \(T\) is nontrivial and \(T(P_{\, 2} (K))\subseteq P_{\, 2} (K)\), then \(T(P_{\, m} (K))\subseteq P_{\, m} (K)\) for every \(m\geq 2\), and the boundary of \(K\) is an ellipsoid (Theorem 3.3).