Global continua of rapidly oscillating periodic solutions of state-dependent delay differential equations (Q5961951)

Periodic solutions of systems with state-dependent delay of the form \[ \begin{aligned} \dot x(t) & = f(x(t), x(t-\tau(t), \sigma),\\ \dot \tau(t) & = g(x(t), \tau(t), \sigma) \end{aligned} \] are considered, and described in the form \((x,\tau,\sigma,p)\), where \(p\) is the minimal period and \(\sigma\) the parameter. A previous theorem of the authors (Theorem 2) says that a continuum \(C\) of such solutions is unbounded, or meets the set of equilibria in finitely many Hopf bifurcation points, in which case a sum \( \sum_i \varepsilon_i\gamma_i\) of signed local degrees at these points is zero. If, for an unbounded \(C\), a-priori-bounds on the \(x\)- and \( \tau\)-components and also on the periods \(p\) along \(C\) are known, then \(C\) must be unbounded in the \( \sigma\)-direction, and in this sense global in parameter space. Finding such continua is the main interest of the paper. The approach is to provide `delay-period-disparity' (dpd), i.e., the situation \(\tau(t) \neq mp\) for some \(t \in \mathbb{R}\) and all \(m \in \mathbb{N}\). Assumptions on \(g\), in particular, \(\frac{\partial g} {\partial \tau}(x_{\sigma}, \tau, \sigma) \neq 0\) for equilibrium values \(x_{\sigma}\), together with a definite feedback condition on \(f\) (namely \(xf(x,x,\sigma) >0\) or \(xf(x,x,\sigma) < 0\) if \(f(x,x,\sigma) \neq 0\)) imply that (dpd) is in some sense an open-dense property along continua of periodic solutions. The proof of this fact uses, for one subcase, the iteration behavior of the map \(l(t) = t - \tau(t) + \tau(t_0)\) \((t_0 \) in \(\mathbb{R}\) fixed), which maps \([t_0, t_0 + \tau(t_0)]\) to itself and has only finitely many fixed points \(t_1,\dots,t_k\) to which the iterates \(l^j(t)\) converge. If none of the values \( x(t_i)\) is an equilibrium value, the feedback condition on \(f\) results in monotonicity of \(t \mapsto \|x(t)\|^2\), contradicting periodicity. In Theorem 6, the local (dpd) property in the form \(p< \tau(t)\) is extended from a Hopf bifurcation point to the whole connected component \(C\) of the solution set containing this point. The argument uses Zorn's lemma and reminds of the construction of maximal existence intervals for solutions of ODE. Thus, a bound for \(\tau\) along \(C\) implies the same bound for \(p\). Section 3 derives a-priori bounds for \( \tau\) and \(\|x\|\) from assumptions of the type `vector field pointing inward/outward'. The Minkowski functional is used here. In Section 4, a class of examples is discussed in detail, to which the a-priori bound method of Section 3 applies. Hopf bifurcation is obtained from analysis of the linearization at equilibria, which also yields the property \(p< \tau(t)\) locally, so that Theorem 6 applies. It is shown that, in the `finitely many Hopf points' case of Theorem 2, the sum \( \sum_i \varepsilon_i\gamma_i\) would be nonzero (a contradiction), so the global continuations of the local Hopf branches must be unbounded and, in view of the bounds on \( \tau, p\), and \(x\), unbounded in the parameter direction.