Temporal distributional limit theorems for dynamical systems (Q526586): Difference between revisions
From MaRDI portal
Latest revision as of 19:07, 13 July 2024
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | Temporal distributional limit theorems for dynamical systems |
scientific article |
Statements
Temporal distributional limit theorems for dynamical systems (English)
0 references
15 May 2017
0 references
The authors consider a Borel flow \((T^s)\) and a Borel function \(f\) on a compact manifold \(X\), the ergodic sums \(S_n(x):=\sum\limits^{n-1}_{k=0}f(T^kx)\) or integrals \(J_t(x):=\int^t_0f(T^sx)ds\), and the following five different kinds of distributional limit theorems (DLT): {\parindent = 0.6 cm \begin{itemize}\item[(1)] The ergodic sums \(S_n(x)\) satisfy a \textit{spatial} DLT on a probability space \((X,\mu)\) when there are real constants \(B_n\to\infty\) and \(A_n\) such that the law of \(x\mapsto\frac{S_n(x)-A_n}{B_n}\) under \(\mu\) converges as \(n\to\infty\) (towards some non-Dirac law). \item[(2)] The ergodic sums \(S_n(x)\) satisfy a \textit{temporal} DLT on the orbit of \(x\) when there are real sequences \(B_N(x)\to\infty\) and \(A_N(x)\) such that the law of \(n\mapsto\frac{S_n(x)-A_N(x)}{B_N(x)}\) under the uniform law \(\mathcal U\{0,\dots,N\}\) converges as \(N\to\infty\) (towards some non-Dirac law). \item[(3)] The ergodic sums \(S_n(x)\) satisfy a \textit{strong temporal} DLT when they satisfy a temporal DLT on the orbit of \(\mu\)-almost every \(x\) and the normalizing sequence \(B_N\) can be chosen to be independent of \(x\). \item[(4)] The ergodic sums \(S_n(x)\) satisfy an \textit{almost sure} DLT on a probability space \((X,\mu)\) when there are real constants \(B_N\to\infty\) and \(A_N\) such that for \(\mu\)-almost every \(x\) the law of \(n\mapsto\frac{S_n(x)-A_N}{B_N}\), under the law on \(\{1,\dots,N\}\) prescribing a mass \(\frac{c}{n}\) at \(n\), converges as \(N\to\infty\) (towards some non-Dirac law). \item[(5)] The ergodic sums \(S_n(x)\) satisfy a \textit{spatio-temporal} DLT on a probability space \((X,\mu)\) when there are real constants \(B_n\to\infty\) and \(A_ n\) such that the law of \((n,x)\mapsto\frac{S_n(x)-A_n}{B_n}\) under \(\mathcal U\{0,\dots,N\}\otimes\mu\) converges as \(n\to\infty\) (towards some non-Dirac law). \end{itemize}} The analogous notions (and results as well) regarding the ergodic integrals \(J_t(x)\) are very similar. The main aim of the authors is to specify, in three classical cases, which among the spatial, strong temporal, almost sure and spatio-temporal DLT hold and which do not. A consequence will be that a large diversity of possible scenarios emerges, so that generally none of theses notions implies another one. The three classical cases which the authors analyze are Anosov flows, irrational rotations on the circle and horocycle flows of a Riemannian surface. Some results presented in this interesting article are adapted from preceding ones, by a series of different authors, some are new extensions of previously known theorems, and some are new. It is proved that for Anosov flows, the spatial, almost sure temporal and spatio-temporal DLT hold while the strong temporal DLT does not; for irrational rotations on the circle, the strong temporal and spatio-temporal DLT hold while the spatial and almost sure temporal DLT do not; for horocycle flows, depending on the class of Sobolev regularity of \(f\), a series of various answers emerge, from which apparently no generality derives.
0 references
ergodic sum
0 references
ergodic integral
0 references
distributional limit theorem
0 references
renormalization
0 references
Anosov flow
0 references
irrational rotation
0 references
horocycle flow
0 references
0 references
0 references
0 references
0 references
0 references
0 references
0 references
0 references