A non-autonomous version of the Denjoy-Wolff theorem (Q371886)

The Denjoy-Wolff theorem asserts that if an analytic self-map \(\varphi\) of the unit disk \(\mathbb D\) is not an elliptic automorphism, then the sequence of iterates \((\varphi_n)\) converges to the Denjoy-Wolff point of \(\varphi\) in the closure \(\overline{\mathbb D}\) of \(\mathbb D\), locally uniformly in \(\mathbb D\). The authors of the paper deal with evolution families \((\varphi_{s,t})\) solving the differential equations \[ \frac{dz}{dt}=(z-\tau(t))(\overline{\tau(t)}z-1)p(z,t),\;\;t\geq s,\;\;z(s)=z, \] where \(s\geq0\), \(\tau:[0,\infty)\to\overline{\mathbb D}\) is measurable and \(\tau(t)\) is almost everywhere the common Denjoy-Wolff point of \((\varphi_{s,t})\), and \(p:\mathbb D\times[0,\infty)\to\mathbb C\) is measurable in \(t\), holomorphic in \(z\), and \(\mathrm{Re} p(\cdot,t)\geq0\) for all \(t\geq0\). The hyperbolic angular extent of a closed arc \(A\subset\mathbb D\) located on the boundary of a horocycle \(\text{Hor}(\xi,k)\) lied at \(\xi\in\partial\mathbb D\) and with factor \(k>0\) is defined as the hyperbolic length of \(A\) divided by \(k\). The main results consist of two parts: \(\tau\in\mathbb D\) and \(\tau\in\partial\mathbb D\). Theorem 1.1. Let \((\varphi_{s,t})\) be an evolution family in \(\mathbb D\) with a common Denjoy-Wolff point \(\tau\in\partial\mathbb D\). Then one and only one of the three mutually excluding situations can happen: (1) for every \(s\geq0\), \(\lim_{t\to\infty}\varphi_{s,t}=\tau\) uniformly on compact subsets of \(\mathbb D\), (2) for every \(s\geq0\), there exists a univalent self-map \(h_s\) of \(\mathbb D\) such that \(\lim_{t\to\infty}\varphi_{s,t}=h_s\) uniformly on compact subsets of \(\mathbb D\), (3) for every \(s\geq0\) and \(z\in\mathbb D\), the \(\omega\)-limit \(\omega(s,z)\) of the trajectory \(t\in[s,\infty)\to\varphi_{s,t}\in\mathbb D\) is a closed arc of the circumference defined by the boundary of a certain horocycle \(\text{Hor}(\tau,k(s,z))\) where \(0<k(s,z)\leq k_{\mathbb D}(\tau,z)=|\tau-z|^2/(1-|z|^2)\). Moreover, the case (3) holds if and only if one of the following three mutually excluding subcases holds: (a) for every \(s\geq0\) and \(z\in\mathbb D\), \(\omega(s,z)\) is exactly the whole circumference \(\partial\text{Hor}(\tau,k(s,z))\), (b) for every \(s\geq0\) and \(z\in\mathbb D\), \(\omega(s,z)\) is a proper closed arc of \(\partial\text{Hor}(\tau,k(s,z))\) and one of its extreme points is \(\tau\), (c) for every \(s\geq0\) and \(z\in\mathbb D\), \(\omega(s,z)\) is a proper closed arc of \(\partial\text{Hor}(\tau,k(s,z))\), it is contained in \(\mathbb D\), and all of those arcs have the same associated hyperbolic angular extent. Theorem 1.2 describes the case \(\tau\in\mathbb D\), and in comparison with Theorem 1.1, it requires that \(z\in\mathbb D\setminus\{\tau\}\), and the horocycle \(\text{Hor}(\tau,k(z,s))\) is substituted by a hyperbolic disk \(D_H(\tau,r(s,z))\) centered at \(\tau\) and of radius \(r(s,z)\).