The Lyapunov dimension of a nowhere differentiable attracting torus

From MaRDI portal
Publication:3223586





The authors study a very particular discrete dynamic system of order \(k+1:\) (1) \(x_{n+1}=Ax_ n\), \(A=(a_{ij})\), \(a_{ij}= integers\), det A\(=1\), x: mod 1, (2) \(y_{n+1}=\lambda y_ n+p(x_ n)\), \(0<\lambda <1\), \(p= smooth\), periodic of period 1. No distinction is made between recurrences and difference equations, R and D for short. The system (1) being autonomous, locally linear and decoupled from (2), its general solution \(x_ n=g(n,C_{n0})\) can be expressed explicitly in terms of generalized exponentials. \(C_{n0}\) is a constant if \((1)=R\), and an arbitrary periodic function of period 1 if \((1)=D\). x: mod 1 implies that g is not smooth as a rule. Inserting g into (2) yields a linear nonhomogeneous R or D, solvable in an elementary way. All properties of (1),(2) are more easily deducible from those explicit solutions than from the indirect analysis of the paper. In an applied context the attractors and graphs are unrealistic, and the claimed results trivial or misleading. Example: \(k=1\), \(a_ n=1\), \(x_ n=g=C_ 0\), \(p(C_ 0)= known\). If \((1),(2)=R\), then \(p=C=const.\), g is smooth (exceptionally), and \(y_ n=\bar C\lambda^ n+C/(1- \lambda)\), \(\bar C=const\). For \(n\to \infty\) the asymptotic state \((x_ n,y_ n)\) is attractive in y, but not in x! If \((1),(2)=D\), \((x_ n,y_ n)\) contains arbitrary functions and is undefined for \(n\to \infty\) without additional conditions.




Cited in
(69)








This page was built for publication: The Lyapunov dimension of a nowhere differentiable attracting torus

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3223586)