Spline approximations in the problem of estimating the number of limit cycles of autonomous systems on the plane (Q2461863)

The paper deals with the problem of estimating the number of limit cycles of planar systems \[ (S):\quad \frac{dx}{dt}=P(x,y),\quad \frac{dy}{dt}=Q(x,y), \] defined in a domain \(\Omega \subset \mathbb{R}^2,\) with \(P,Q\in C^{1}(\Omega)\) (for polynomials \(P,Q,\) this is the well-known Hilbert's 16th problem). The authors developed in a series of previous papers a numerical approach we pass to describe for the sake of clarity. They try to find a function \(\Psi\in C^{1}(\Omega)\) such that \[ (D):\quad \Phi\equiv D(\Psi)\equiv k\Phi\text{div}f +P\frac{\partial \Psi}{\partial x}+ Q\frac{\partial \Psi}{\partial y} >0 (<0)\quad \forall(x,y)\in\Omega, \quad f=(P,Q), \] where \(k<0\) is a given number, and \(D\) is the Dulac operator. If this is achieved, a upper bound of the number of limit cycles can be given by analyzing the curve \(\Psi=0\) in the boundary \(\partial \Omega.\) In some cases, the authors have proved in previous papers that \(\Psi\) can be chosen as a linear combination of basic polynomials. Now, they consider the possibility of using a function \(\Psi\) obtained as a linear combination of basic spline functions \(\Psi_j\) such that \(\Psi_j\) is equal to \(1\) in the \(i\)-th grid point of a rectangular domain \(x_0\geq x\geq x_k, \) \(y_0\geq y\geq y_k, \) and vanishes at all other nodes. Then \(\Phi=\Phi(x,y,C)=\sum_{j=1}^{n}C_j\Phi_j(x,y)\) (with \(\Phi_j=D(\Psi_j)\)) will satisfy \((D)\) for a given \(k<0\) if and only if the inequality \[ (C):\quad \max_{\| C\| \leq 1} \min_{(x,y)\in\Omega}\Phi(x,y)>0 \] holds, where \(\| C\| \) means the norm of \(C=(C_1,\dots,C_n)\). To attain the maximin \((C)\) the authors obtain the values of suitable constants \(C_j\) by solving a linear programming problem. This strategy is numerically applied, among others, to several examples of Liénard systems, a van der Pol system, and a predator-prey system (\(P,Q\) are not necessarily polynomials). The computation was done by employing \textit{Mathematica}. As a conclusion, let me paraphrase the authors: ``Nowadays, the use of spline approximations for constructing the function \(\Psi\) is most preferable for systems \((S)\) with nonpolynomial right-hand sides. For polynomial systems \((S)\) the choice of the construction of the function \(\Psi\) depends on the form of the system itself.''