Experimental results for the Poincaré center problem (Q2474195)

The Poincaré center problem (1885) is a question when the autonomous system \[ \dot x = y + q(x,y), \quad \dot y = -x - p(x,y), \quad (x,y) \in \mathbb{R}^2, \tag{1} \] where \(p(x,y)\) and \(q(x,y)\) are polynomials or convergent power series starting with quadratic terms, has stable solutions in a neighborhood of the origin on the \((x,y)\)-plane. Poincaré showed that one can iteratively find a formal power series \(F(x,y) = x^2+y^2 + \cdots\), where dots are terms of degrees greater than \(2\), such that \[ (y + q) F'_x - (x + p) F'_y = \sum_{i=1}^{\infty} s_i (x^{2(i+1)} + y^{2(i+1)}). \tag{2} \] The coefficients \(s_i\) in (2) are polynomials of the coefficients of \(p(x,y)\) and \(q(x,y)\), they are called the focal values of (1) at origin. If all focal values \(s_i=0\), \(i \in \mathbb{N}\), and the series \(F\) is convergent, then \(F\) is a first integral of (1). Poincaré proved that if the system (1) has an analytic first integral, than all \(s_i=0\), and later Frommer proved that the system (1) is stable if and only if all \(s_i=0\) even without the assumption of convergence of \(F\). Unfortunately, it is in general impossible to check the condition \(s_i=0\) for all \(i \in \mathbb{N}\). Let \(p(x,y)\) and \(q(x,y)\) be polynomials of degree \(d \geq 2\), then the focal values \(s_i\) are polynomials in finitely many unknowns. Let \(I_{\infty} = (s_1, \dots)\) be the ideal generated by \(s_i\), \(i \in \mathbb{N}\), in the ring of polynomials of relevant number of variables. Hilbert's Basis Theorem implies that the ideal \(I_{\infty}\) is finitely generated, i.e., there exists an integer number \(m=m(d)\) such that the condition \(s_1 = \cdots = s_m = 0\) implies \(s_i=0\) for all \(i \in \mathbb{N}\). This shows that a finite criterion for stability exists, nevertheless no value for \(m(d)\) was obtained. In fact, today only the case \(d=2\) is studied: \(m(2) = 3\). Even for \(d=3\) only partial results are known, because the polynomials \(s_i\) are very large (for example, \(s_5\) has already \(5348\) terms, and \(s_i\) with \(i \geq 6\) can not be determined by contemporary computer systems). In particular, Żołądek and Christopher showed that \(m(3) \geq 11\). One new feature of the Poincaré center problem with \(d=3\) is that a new type of systems (1) appeared, namely the rationally reversible centers. Żołądek has classified all rationally reversible centers which are not of Darboux type. He also conjectured that all systems (1) with \(d=3\) with stable solutions near the origin are either of Darboux type or rationally reversible. In this paper, the author applies a heuristic method based on counting points over finite fields to the Poincaré center problem. He shows that this method gives the correct results for the systems (1) with homogeneous polynomials \(p, q\) of degree \(d=2,3\). He provides additional statistical evidence for Żołądek's conjecture in the whole space of the systems (1) with \(d=3\) up to codimension \(7\). For the case of general polynomials \(p, q\) of degree \(3\), the author gives a new conjecture about the stratification of the subset \(X_{\infty}\) of the systems (1) with zero focal values, he checks it at different examples. Finally, the author applies his methods for a heuristic estimate of the number \(m(d)\). This leads to the correct estimates for polynomials \(p, q\) of degree \(d=2,3\). For general polynomials \(p, q\) of degree \(3\) he does not manage to give a reliable estimate, the computer experiments give a lower bound \(m(3) \geq 10\), which does not quite reach the known bound \(m(3) \geq 11\).