Limit cycle bifurcations by perturbing a quadratic integrable system with a triangle (Q2634274)

The authors consider a planar quadratic system of the form \[ \dot{x}=x[\beta-\beta x-(\beta+1)y]+\varepsilon p(x,y,\lambda),\quad \dot{y}=y[-\alpha +(\alpha +1)x+\alpha y]+\varepsilon q(x,y,\lambda) \] with positive parameters \(\alpha\), \(\beta\) not equal to one, and small parameters \(0<\varepsilon\ll\lambda\ll1\). As known, the unperturbed system has a first integral \(H=x^\alpha y^\beta(1-x-y)\) and an integrating factor \(x^{\alpha-1}y^{\beta-1}\); there are three hyperbolic saddles at \((0,0), (1,0), (0,1)\) and a center inside the triangle contour connecting the saddles. The main result of the paper states that the cyclicity of the triangle contour is at least three within the class of quadratic systems. The proof is based on asymptotical expansions and follows the methods developed by \textit{Y. Wang} and \textit{M. Han} [J. Math. Anal. Appl. 418, No. 1, 357--386 (2014; Zbl 1318.34050)]; \textit{M. Han} and \textit{Y. Xiong} [``Limit cycle bifurcations in a class of near-Hamiltonian systems with multiple parameters'', Chaos Solitons Fractals 68, 20--29 (2014; \url{doi:10.1016/j.chaos.2014.07.005})]; \textit{M. Han} et al. [Int. J. Bifurcation Chaos Appl. Sci. Eng. 19, No. 12, 4117--4130 (2009; Zbl 1183.34054)]. Remark. The case \(\alpha=\beta\) had to be removed from consideration too. The reason is that the cases \(\alpha=1\) or \(\beta=1\) or \(\alpha=\beta\) are degenerate (have symmetry), they are equivalent to each other by suitable affine change in the \((x,y)\)-plane and moreover they cannot be studied by using first-order Melnikov theory only.