A natural order in dynamical systems based on Conley-Markov matrices (Q665978): Difference between revisions

Consider the dynamical system on \(\mathbb{R}^n\) generated by \[ dx=F(x)dt \tag{1} \] and suppose it has a compact global attractor \(\mathcal{A}\) with a Morse decomposition \(\mathcal{M}=\{M_1,\ldots, M_N\}\). The main goal of the paper is to introduce a natural order on the set \(\mathcal{M}\) from the least to the most stable invariant set in a probabilistic sense. The authors perturb the system (1) by white noise and consider the stochastic differential equation \[ dx=F(x) dt+\sqrt{2\epsilon}\;dW(t) \tag{2} \] where \(W(t)\) is an \(n\)-dimensional Brownian motion and \(\epsilon>0\) is a constant. They extend the concept of connection matrix from Conley index theory to the random dynamical system (2) and introduce Conley-Markov matrices which combine topological information and probabilistic information. The Conley-Markov matrix depends on \(\epsilon\), on a chosen period [0,T], and on isolating neighbourhoods \(\mathcal{N}_i\) of \(M_i\). The entry \(m_{ij}^\epsilon\) in the Conley-Markov matrix \(M^\epsilon(T)\in\mathbb{R}^{N\times N}\) of (2) is the average probability that the trajectories starting from \(\mathcal{N}_i\) enter \(\mathcal{N}_j\) in the period \([0,T]\). These probabilities are now used to order the set \(\mathcal{M}\). The paper also contains some numerical examples to illustrate the Conley-Markov matrix and its properties.