When distribution measures are invariant for dynamical systems

From MaRDI portal





A topological dynamical system \(T\colon X\to X\) (where \(X\) is a compact metric space and \(T\) is a continuous map) always has a \(T\)-invariant Borel probability measure \(\mu\) by the \textit{N. Kryloff} and \textit{N. Bogoliouboff}'s theorem [Ann. Math. (2) 38, 65--113 (1937; Zbl 0016.08604)], because the space \(M(X)\) of Borel probability measures on \(X\) is weak* compact and any limit point of the sequence \[\left(\frac1n\sum_{j=0}^{n-1}\delta_{T^jx}\right)_{n\geqslant1}\] is automatically \(T\)-invariant. Here the exact orbit viewed as a sequence \((T^nx)_{n\geqslant0}\) is replaced with an arbitrary sequence \((x_n)_{n\geqslant0}\), and the question addressed is this: How close must this sequence be to a real orbit of \(T\) in order to guarantee that any associated `distribution' measure obtained as a limit point of \[\left(\frac1n\sum_{j=0}^{n-1}\delta_{x_j}\right)_{n\geqslant1}\] is automatically invariant? \N\N\textit{D. Kwietniak} et al. [Monatsh. Math. 183, No. 4, 625--648 (2017; Zbl 1377.37037)] showed that being an `asymptotic average pseudo-orbit' is sufficient, and the author [J. Differ. Equations 402, 443--462 (2024; Zbl 1557.37020)] showed that being an `asymptotic quasi-orbit' is sufficient (the exact definitions are interesting but not entertaining to explicate). Here some satisfyingly complete answers are found as follows: All distribution measures are invariant if and only if the sequence is an `asymptotic quasi-orbit'; at least one distribution measure is invariant if and only if the sequence is a `proximal quasi-orbit'; finally no distribution measure is invariant if and only if the sequence is a `distal quasi-orbit'.











This page was built for publication: When distribution measures are invariant for dynamical systems

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q7028325)