Realization of period maps of planar Hamiltonian systems (Q2465125): Difference between revisions

As described in a work of \textit{B. Fiedler, C. Rocha} and \textit{M. Wolfrum} [J. Differ. Equations 201, 99--138 (2004; Zbl 1064.35076)], the characterization of the global attractors of semilinear parabolic equations of the form \[ u_t = u_{xx} + f(u,u_x) \] where \(x \in \mathcal{S}^1\), leads to the consideration of the periodic solutions of planar Hamiltonian systems and the study of the corresponding period maps. In this paper the set of \(2\pi\)-periodic solutions of planar Hamiltonian systems of the form \[ u''+g(u) = 0,\tag{\(*\)} \] where \(g\in\mathcal{C}^1(\mathbb{R})\), is considered. Quoting the reviewed paper, the function \(g\) is assumed to satisfy: (i) a dissipative condition of the form \(g(u)/u<0\) for \(| u| >M\) and; (ii) the potential \(G(u) = \int_{0}^{u} g(s) \, ds\) is a Morse function. Thus, \(g\) has a finite odd number \(n\) of nondegenerated zeroes corresponding to an odd number of saddles and an even number of centers, which occur alternatively in the real line. The phase portrait of (\(*\)) contains a bounded open set of periodic orbits that decomposes into punctured disks (one disk for each center) and annular regions. For each periodic orbit the period map \(T\) is defined as the usual continuous positive function and if the point with \(u=e\) is one of the centers, its period is defined by continuity as \(2\pi/\sqrt{g'(e)}\). The domain of definition of the period map is denoted as \textit{cyclicity set} \(\mathcal{C}\) and we can restrict it to the open union of intervals \(D=\{r\in \mathbb{R} : (r,0) \in \mathcal{C}\}\). The \(2\pi\)-periodic orbits of (\(*\)) are related to the points \(r\in D\) for which \(T(r)=2\pi/k\) for some \(k\in \mathbb{N}\). Another generic assumption is made: (iii) all \(2\pi\)-periodic orbits are hyperbolic, that is, \(T'(r)\neq 0\) if \(r\in D\) corresponds to a \(2\pi\)-periodic orbit. Given a \(2\pi\)-periodic orbit, the integer \(k\) is called the \textit{periodic lap number} as it is the number of complete laps of the orbit in one period. As proved by \textit{M. Urabe} [Funkc. Ekvacioj, Ser. Int. 6, 63--88 (1964; Zbl 0127.04502)], the restriction of \(g\) to an interval of \(D\) uniquely determines the value of the period map \(T\) on this interval. And conversely, one can determine functions \(g\) realizing a specific period map \(T\). The realization problem solved in this paper is related to the periodic lap numbers. Given an interval of \(D\), the sequence \(S=\{k_1,k_2, \dots, k_N\}\) of lap numbers ordered by increasing values of \(r\), where \(r\) is situated at the left of the first center point of the corresponding nest, is called the \textit{signature} of the period map \(T\) on the corresponding interval. Which sequences \(S\) arise as signatures of the period map of an equation (\(*\))? Supported by several previous results cited in the paper, the following necessary conditions are shown. If the considered interval of \(D\) corresponds to a disk, then its signature \(S=\{k_1,k_2, \ldots, k_N\}\), in case of being nonempty, needs to satisfy \[ k_1=1, \qquad | k_{j+1}-k_j| \leq 1, \quad j=1,2,\dots,N_1,\tag{1} \] and, if \(k_{i-1} \neq k_i=k_{i+1}= \dots=k_j\neq k_{j+1}\) for \(1<i\leq j<N\), \[ (k_i-k_{i-1})(k_{j+1}-k_j)=(-1)^{i-j}.\tag{2} \] If the considered interval of \(D\) corresponds to an annular region, then its signature \(S=\{k_1,k_2, \dots, k_N\}\), in case of being nonempty, needs to satisfy (1), (2) and \[ k_N=1 \quad \text{with \(N\) even}.\tag{3} \] The main result of the paper, cf. Proposition 3 on page 587, ensures that these conditions are also sufficient.