Meromorphic Hamiltonian systems on complex surfaces (Q1317031)

A Poisson structure on a complex manifold \(M\) is given by a Lie bracket on the structure sheaf \({\mathcal O}\) of \(M\), \(\{\cdot,\cdot\}:{\mathcal O} \times {\mathcal O} \to {\mathcal O}\), satisfying the Leibniz identity. For any meromorphic function \(H\) on \(M\) there is associated the meromorphic vector field \(X_ H=\{H,\cdot\}\) on \(M\) which is called a meromorphic Hamiltonian system on \(M\). On a complex surface the notion of a Poisson structure reduces to an effective divisor \(S\) in the anti-canonical linear system, \(S \in |-K |\). Moreover, in blowing up the indeterminacies of the Hamiltonian \(H\), the system \((M,S,H)\) has a unique regularization \((\widehat M,\widehat S,\widehat H)\), where \(\widehat H\) is now a holomorphic function to the projective line, \(\widehat H:\widehat M \to \mathbb{P}^ 1\). The system \((M,S,H)\) is called algebraically completely integrable, if the general fiber of \(\widehat H\) is a union of elliptic curves and the vector field \(X_{\widehat H}\) is holomorphic on them (i.e. linear). Associated to a Hamiltonian system on \(M\) is a pencil \(\theta\) of divisors without fixed components on \(M\). More precisely, if \(D_ +\) is the zero divisor and \(D_ -\) is the pole divisor of \(H\), i.e. \((H)=D_ +-D_ -\), then \(\theta\) is given by the line through \(D_ +\) and \(D_ -\) in the linear system \(| D_ - |=| D_ +|\), \(\theta :=D_ + \vee D_ -\subseteq | D_ -|\). The notion of algebraic complete integrability depends only on \(\theta\). Associated to \(\theta\) is the base point graph \(B(\theta)\) of \(\theta\), i.e., the formal sum of the base points \(\{p_ 1,\dots,p_ r\}=B\) of \(\theta\) (including the infinitely near ones) counting with their multiplicities \(n(p) \in \mathbb{Z}_ +\), \(B(\theta)=\sum_{p \in B}n(p)p\). The degree of \(B(\theta)\) is given by \(\deg B (\theta)=\sum_{p \in B} n(p)\). The main theorem of the paper characterizes the algebraic complete integrability of a system \((M,S,\theta)\) in terms of the associated base point graph \(B(\theta)\). More precisely, \((M,S,\theta)\) is algebraically completely integrable if and only if (i) \(\deg B(\theta)=\theta\cdot S\) (intersection product) and (ii) \(B \subseteq S\) (in the appropriate sense).