A method of reduction for invariant curves of quasiperiodically forced maps

Amadeu Delshams, Rafael Ortega · 2026

Read A method of reduction for invariant curves of quasiperiodically forced maps by Amadeu Delshams, Rafael Ortega (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2603-26912.txt.

Introduction

Delshams and Ortega study translated curves for maps forced in a quasi-periodic way. They prove such curves exist under very mild smoothness demands. The result covers rotation numbers of constant type. Among the translated curves, the truly invariant ones solve one scalar equation. That bifurcation equation makes existence easy to read off. It also shows stability and how the curves bifurcate. The approach cuts a two-dimensional curve problem down to a single scalar condition.

Catalog record

CREATORAmadeu Delshams, Rafael Ortega
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE42632
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2603.26912

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

A method of reduction for invariant curves of quasiperiodically forced maps Amadeu Delshams, Rafael Ortega Abstract The existence of translated curves for quasiperiodically forced maps is established, under very mild regularity hypotheses, for rotation numbers of constant type. Among the translated curves, the invariant curves are characterized as the solutions of a scalar bifurcation equation, from which their existence, stability as well as bifurcation can be easily described. 1 Introduction This paper deals with the existence of invariant curves of skew-product dynamical maps (r_{1},\theta_{1})=f_{\varepsilon}(r,\theta) of the form \begin{array}[]{rcl@{\qquad}l}r_{1}&=&r+\varepsilon F(r,\theta;\varepsilon),&r\in\mathbb{R}\\ \theta_{1}&=&\theta+2\pi\alpha,&\theta\in\mathbb{T}=\mathbb{R}/2\pi\mathbb{Z},\end{array} (1) defined on the cylinder \mathcal{C}:=\mathbb{R}\times\mathbb{T} , where \varepsilon is a small parameter. For an irrational frequency \alpha , such kind of maps are called quasiperiodically forced maps and arise in a number of well-known examples. As a motivation, let us mention the so-called quasiperiodically forced Arnold circle map \begin{array}[]{rcl@{\qquad}l}x_{1}&=&x+\omega+k\sin x+b\sin\theta,&x\in\mathbb{R}\,,\\ \theta_{1}&=&\theta+2\pi\alpha,&\theta\in\mathbb{T}=\mathbb{R}/2\pi\mathbb{Z}\,,\end{array} (2) which has been studied by many authors as a model of quasiperiodically forced oscillators and as a paradigm for the investigation of strange non-chaotic attractors (SNA, see, for instance, [ 13 , 5 , 21 ] and references therein). This paper focuses on finding smooth and invariant invariant curves for map ( 1 ) in a simple way, which will also allow us to address the problem of their existence and stability. In this direction, let us begin by noting that to find these invariant curves in the system above ( 2 ), in [ 7 , 6 ] the authors dealt with small \omega and k , and thus introduced expansions in a small parameter \varepsilon k=\varepsilon,\quad\omega=\omega_{1}\varepsilon+\omega_{2}\varepsilon^{2}+\cdots,\quad b=b_{0}+b_{1}\varepsilon+b_{2}\varepsilon^{2}+\cdots, and searched for invariant curves of the map above ( 2 ) of the type x=G(\theta)=G_{0}(\theta)+G_{1}(\theta)\varepsilon+G_{2}(\theta)\varepsilon^{2}+\cdots, (3) where all functions G_{i} are 2\pi -periodic. After substituting this expansion into the mapping ( 2 ), the function G_{0} can be explicitly computed for \alpha\not\in\mathbb{Z} . Namely, G_{0}(\theta)=g_{0}-\frac{b_{0}}{2\sin(\pi\alpha)}\cos(\theta-\pi\alpha), (4) where g_{0} is a constant to be adjusted later (in the computation of G_{1} ). The functions G_{1} , G_{2},\dots, can be determined in a recursive way. They are solutions of linear functional equations of the type G_{n}(\theta+2\pi\alpha)=G_{n}(\theta)+p_{n}(\theta), (5) where p_{n} depends upon the parameters \omega_{1},\dots,\omega_{n} , b_{1},\dots,b_{n} , and the previous functions G_{0},G_{1},\dots,G_{n-1} (see [ 7 ] for more details of these formal computations). The solutions to equation ( 5 ) can be obtained using Fourier analysis and, for \alpha irrational, the condition \int_{0}^{2\pi}p_{n}(\theta)\,\mathrm{d}\theta=0 becomes necessary and sufficient for their solvability. Sometimes, these solutions are only formal because small divisors appear. To overcome this objection one has to assume that \alpha has good arithmetic properties. However, even when \alpha is the golden mean, it is not obvious how to prove the convergence of expansion ( 3 ). Before coming back to the general case ( 1 ), we notice that map ( 2 ) can also be written in this form, simply by performing the change of variables x=r+G_{0}(\theta) where G_{0} is given by ( 4 ). Indeed, under this change, map ( 2 ) becomes \begin{array}[]{rcl}r_{1}&=&r+\varepsilon\left(\omega_{1}+\widetilde{F}(r,\theta;\varepsilon)\right),\\ \theta_{1}&=&\theta+2\pi\alpha,\end{array} (6) where \widetilde{F}(r,\theta;\varepsilon)=\sin(r+G_{0}(\theta))+b_{1}\sin\theta+\mathop{\mathrm{}}{O}\nolimits\left(\varepsilon\right) . In this paper we propose a different approach for studying the existence of invariant curves. First, we shall show that the cylinder \mathcal{C} can be foliated by translated curves of the map ( 1 ). By a translated curve, we understand a curve of the type r=G(\theta) which is mapped under map ( 1 ) onto r_{1}=G(\theta_{1})-\lambda , for some \lambda\in\mathbb{R} , that is, G(\theta) and \lambda are solutions of the equation G(\theta+2\pi\alpha)=G(\theta)+\varepsilon F\left(G(\theta),\theta;\varepsilon\right)+\lambda\,. (7) Once we know that such a foliation exists, it is enough to discuss the equation \lambda=0 among all translated curves to get invariant curves. Let us be more concrete. We will assume that F\in C^{m}(\mathcal{C}\times[0,1]) in equation ( 1 ), for m\geq 3 . More precisely, we assume that \left|\partial^{\ell_{1}}_{\theta}\partial^{\ell_{2}}_{r}\partial^{\ell_{3}}_{\varepsilon}F(r,\theta;\varepsilon)\right|\leq k_{F}^{(m)} (8) for all (r,\theta;\varepsilon)\in\mathcal{C}\times[0,1] , \ell=(\ell_{1},\ell_{2},\ell_{3})\in\mathbb{N}^{3} , \left|\ell\right|:=\ell_{1}+\ell_{2}+\ell_{3}\leq m . We also assume that the frequency \alpha has good arithmetic properties, to be precise, \alpha is assumed to be of constant type ; see ( 11 ). Under these two hypotheses, our main result (Theorem 1 ) is that for \varepsilon small enough ( \varepsilon\in[0,\varepsilon^{*}) , with \varepsilon^{*} depending only on k_{F}^{(3)} and \alpha )), there exists a foliation of \mathcal{C} by translated curves of the map ( 1 ). More precisely, there exists a C^{m-2} map \psi:\mathcal{C}\times[0,\varepsilon^{*})\to\mathbb{R} of the form \psi(c,\varepsilon,\theta)=c+\mathop{\mathrm{}}{O}\nolimits\left(\varepsilon\right) , such that, for every (c,\varepsilon)\in\mathbb{R}\times[0,\varepsilon^{*}) , r=\psi(c,\theta,\varepsilon) is a translated curve of ( 1 ), satisfying c=\frac{1}{2\pi}\int^{2\pi}_{0}\psi(c,\theta,\varepsilon)\,\mathrm{d}\theta\,. Moreover, for every fixed \varepsilon\in[0,\varepsilon^{*}) , (c,\theta)\mapsto(\psi(c,\theta),\theta) is a diffeomorfism in the cylinder \mathcal{C} and therefore the family of translated curves r=\psi(c,\theta,\varepsilon) defines a foliation of the cylinder. The most delicate aspect of the method presented here is the construction of translated curves. To this end, instead of dealing with standard KAM rapidly convergent iterations or invoking “hard” implicit function theorems in scaled spaces, we propose a simple iteration scheme where a linear functional-differential equation is solved at each step. This procedure is certainly more artificial than the method proposed in [ 7 ] , but the iteration scheme is very easy to state and we can justify its convergence in a rigorous way. Our approach is based on Herman’s ideas in [ 10 , 12 ] and is also related to the papers of de la Llave [ 15 , 16 ] . Once a lemma for the solution of a linear difference equation in a suitable Sobolev space and a suitable frequency \alpha (Lemma 6 , due to Herman) is taken for granted, the proof of Theorem 1 is completely self-contained. We note again that one could also employ an approach based on Newton’s Method as in KAM theory, and even consider a multi-dimensional \theta\in\mathbb{T}^{d} . This would relax the conditions on \alpha but it probably would require more regularity on F and, of course, this would complicate the proof a lot. Once the existence of a foliation of translated curves r=\psi(c,\theta,\varepsilon) is proved, the existence of invariant curves among them is reduced to searching zeros of the “translation” function c\mapsto\lambda=\lambda(c,\varepsilon) , which has the form \lambda(c,\varepsilon)=-\varepsilon\Phi(c,\varepsilon) , with \Phi(c,\varepsilon)=\frac{1}{2\pi}\int^{2\pi}_{0}F(\psi(c,\theta,\varepsilon),\theta;\varepsilon)\,\mathrm{d}\theta\,. (9) Equivalently, for \varepsilon\in[0,\varepsilon^{*}) , the invariant curves of map ( 1 ) are precisely given by r=\psi(c^{*},\theta,\varepsilon) for every zero c^{*} of the real function c\mapsto\Phi(c,\varepsilon) . Therefore, the search for invariant curves is reduced to finding the zeros of \Phi . The reader who is familiar with the methods of bifurcation theory will notice that this strategy is reminiscent of Lyapunov-Schmidt and alternative methods (see [ 9 ] ). Actually, equation ( 7 ) for the translated curves can be seen as the auxiliary equation , while \Phi(c,\varepsilon)=0 becomes the bifurcation equation . From the study of the scalar equation \Phi(c,\varepsilon)=0 , global and local results about the existence of invariant curves can be easily deduced. For example, each simple zero c^{*} of \Phi will give rise to a hyperbolic invariant curve, attractor for \partial_{c}\Phi(c^{*},\varepsilon)<0 , or repeller otherwise. More interestingly, multiple zeros of \Phi give rise to bifurcating invariant curves. On the other hand, for a circle map ( 1 ), that is, for a function F periodic in the variable r , it turns out that \Phi is also a periodic function of c , and

Read the complete text