Integral invariants and limit sets of planar vector fields (Q2575069)

Two-dimensional autonomous systems of real differential equations of the form \[ \dot{x} = P(x,y), \quad \dot{y} = Q(x,y), \tag{1} \] where \(P\) and \(Q\) are \(\mathcal{C}^1\) functions in some nonempty open set \(\mathcal{U} \subseteq \mathbb{R}^2\), are considered. An inverse integrating factor for system (1) is a \(\mathcal{C}^1\) real function defined on \(\mathcal{U}\) which satisfies the following linear partial differential equation \[ P(x,y) \,\frac{\partial V}{\partial x} + Q(x,y)\,\frac{\partial V}{\partial y} = \left( \frac{\partial P}{\partial x} + \frac{\partial Q}{\partial y} \right) V(x,y). \tag{2} \] With equation (2) it is clear that the set of points \[ V^{-1}(0) = \left\{ (x,y)\in \mathcal{U} : V(x,y)=0 \right \} \] is formed by orbits of system (1). This work is an illustration of the narrow relationship between the properties of the set \(V^{-1}(0)\) and the qualitative structure of the orbits of system (1). This relationship is also exhibited in other previous papers, some of which are cited below, and forms an alternative and useful tool to tackle the qualitative study of such a system (1). It may also be a direction to approach substantial problems like Hilbert's 16th problem (part b). However, in order to follow this approach, an important difficulty must be overcome: the existence of a sufficiently smooth, inverse integrating factor, at least locally, in a neighborhood of a graphic or polycycle. This paper does not deal with existence problems but exhibits how an inverse integrating factor, if it exists, can be used to determine some qualitative properties of system (1). In several previous papers, see the works of \textit{L. R. Berrone} and \textit{H. J. Giacomini} [Qual. Theory Dyn. Syst. 1, 211--230 (2000; Zbl 1032.34027)], \textit{J. Chavarriga}, \textit{H. Giacomini}, \textit{J. Giné} and \textit{J. Llibre} [J. Differ. Equations 157, 163--182 (1999; Zbl 0940.37005)] and \textit{H. J. Giacomini, J. Llibre} and \textit{M. Viano} [Nonlinearity 9, 501--516 (1996; Zbl 0886.58087)], this relation between the orbits of system (1) belonging to \(V^{-1}(0)\) and their asymptotic behaviour is established. Two main examples of this fact are the following results. In the work of Giacomini, Llibre and Viano, it is shown that if there exists an inverse integrating factor defined in a neighborhood of a limit cycle, then the limit cycle must be contained in the set \(V^{-1}(0)\). In the work of Berrone and Giacomini, it is shown that if \(p_0\) is a hyperbolic saddle-point of system (1), then any inverse integrating factor \(V\) defined in a neighborhood of \(p_0\) vanishes on all four separatrices of \(p_0\), provided that \(V(p_0)=0\). The paper under review gives a generalization, with a more intrinsic point of view, of the previous results. The following statements are given and proved. The first result corresponds to a critical point of system (1) whose asymptotic behavior may be deduced from the vanishing or not of an inverse integrating factor. Theorem: Let \(p_0\) be a critical point of system (1) at which there is an elliptic or parabolic sector. If \(V\) is an inverse integrating factor defined on a neighborhood of \(p_0\), then \(V(p_0)=0\). A useful corollary of this theorem states that if system (1) is real and analytic and has an inverse integrating factor \(V\) defined in a neighborhood of a singular point \(p_0\) known to be an antisaddle and \(V(p_0) \neq 0\), then \(p_0\) must be a center. The second result given is related with a compact limit set of system (1). Theorem: Let \(K\) be any compact \(\alpha-\) or \(\omega-\)limit set of system (1) that contains a regular point, and let \(V\) be any inverse integrating factor of (1) defined in some neighborhood of \(K\). Then, there exists at least one point \(p\) in \(K\) such that \(V(p)=0\). Moreover, and depending on the smoothness of \(V\), the following statements hold: (i) If \(V\) is \(\mathcal{C}^2\), then either \(K\) contains a point which is an accumulation point of isolated critical points of \(V\), or \(K \subseteq V^{-1}(0)\). (ii) If \(V\) is real analytic or Morse, then \(K \subseteq V^{-1}(0)\). It is interesting to note that in the proof of the previous statements, some strong and nontrivial results are needed. The effect of the flow on area elements and the notion of integral invariant, introduced by Poincaré, as well as Sard's theorem related with regular points, are crucial for the proof. The third main result is related with a critical point of system (1) whose description of its separatrices may be given using the set \(V^{-1}(0)\). Theorem: Let \(p_0\) be an isolated singularity of system (1) whose associated linear part has exactly one zero eigenvalue and let \(V\) be an inverse integrating factor defined in a neighborhood of \(p_0\). Then \(V\) vanishes at any separatrix of system (1) at \(p_0\). This last theorem can be applied to several polycycles or graphics of a system (1). Finally, a set of examples to illustrate the theorems and a conjecture for a future work are given.