On the singularity structure of Kahan discretizations of a class of quadratic vector fields (Q2049056)

Consider a system of differential equations defined by a quadratic vector field on \(\mathbb{R}^n\) \[\dot{x} = f(x) := Q(x) + Bx + c, \] where \(Q\) is an \(\mathbb{R}^n\)-valued quadratic form, \(B\in \mathbb{R}^{n\times n}\) and \(c\in \mathbb{R}^n\). The Kahan discretization method is defined by the numerical integration method \(x \mapsto \tilde{x}\) with step size \(\epsilon\) given by \[\frac{1}{2\epsilon} \left(\tilde{x} - x \right) = Q(x, \tilde{x}) + \frac{1}{2} B(x+\tilde{x}) + c. \] Here \(Q(x, \tilde{x})\) is the bilinear form associated to \(Q(x)\). It was shown in [\textit{E. Celledoni} et al., J. Phys. A, Math. Theor. 45, No. 45, Article ID 025201, 12 p. (2012; Zbl 1278.65192)] that this method is equivalent to a Runge-Kutta method. However, the focus of the present article is on the so-called Kahan map defined by \[ \tilde{x} = \Phi_\epsilon(x) = x + 2 \epsilon(\mathrm{Id} - \epsilon \nabla f)^{-1} f(x). \] Here \(\nabla f\) is the Jacobian of \(f\). The Kahan map is birational (when extended to \(\mathbb{C}^n\)) as can be seen from \(\Phi_\epsilon^{-1}(x) = \Phi_{-\epsilon}(x)\). The main result of this article is a statement about the integrability of the Kahan map for certain quadratic vector fields, i.e., for a special form of \(f\). These are defined by \[ \dot{x} = \ell_1^{1-\gamma_1}(x) \ell_2^{1-\gamma_2}(x) \ell_3^{1-\gamma_3}(x) J \nabla H(x), \quad x = (x_1, x_2) \in \mathbb{R}^2, \] where \(\ell_i(x_1, x_2) = a_i x_1 + b_i x_2\), \(J\) is the standard symplectic form on \(\mathbb{R}^2\) and \(\gamma_1, \gamma_2, \gamma_3 \in \mathbb{R} \setminus \{0\}\). These quadratic vector fields (i.e., differential equations) are generalizations of the two-dimensional reduced Nahm systems, see [\textit{N. J. Hitchin} et al., Nonlinearity 8, No. 5, 662--692 (1995; Zbl 0846.53016)]. The integrability of the Kahan maps \(\Phi\colon \mathbb{C}^2 \to \mathbb{C}^2\) has only been considered for a few cases of \((\gamma_1,\gamma_2,\gamma_3)\) before [\textit{E. Celledoni} et al., Proc. R. Soc. Lond., A, Math. Phys. Eng. Sci. 471, No. 2184, Article ID 20150390, 10 p. (2015; Zbl 1372.65201); \textit{M. Petrera} et al., Exp. Math. 18, No. 2, 223--247 (2009; Zbl 1178.14007); \textit{M. Petrera} and \textit{R. Zander}, J. Phys. A, Math. Theor. 50, No. 20, Article ID 205203, 13 p. (2017; Zbl 1377.37086)]. For example, for \((\gamma_1, \gamma_2, \gamma_3) = (1, 1, 1)\), one obtains the canonical Hamiltonian system on \(\mathbb{R}^2\) with homogeneous cubic Hamiltonian. In this article, the author proves a strong statement about the non-integrabilty of the Kahan maps depending on \((\gamma_1, \gamma_2, \gamma_3)\). To do so, the Kahan maps are considered as birational maps \(\Phi\colon \mathbb{CP}^2 \to \mathbb{CP}^2\) and their singularity structures are studied in a precise way (based, e.g., on [\textit{J. Diller} and \textit{C. Favre}, Am. J. Math. 123, No. 6, 1135--1169 (2001; Zbl 1112.37308)]). In this way, the author proves whether the sequence \(d(m)\) of the degrees of \(\Phi^m\) grows exponentially, quadratically, linearly or is bounded (these are all possible cases). In the first case, \(\Phi\) is nonintegrable. Some of the integrable cases are considered in detail in the later sections of the article, thus exemplifying the general methods.