Poisson systems as the natural framework for additional first integrals via Darboux invariant hypersurfaces (Q390825)

The theory of Darboux polynomials goes back to Darboux (1878) and applies to rather general evolutionary systems. However in the past 15 years there has been considerable interest in the special case of Hamiltonian mechanics. Here the Hamiltonian is usually chosen in the form \({\mathcal H}({\mathbf q},{\mathbf p})=\frac12 \sum_{i=1}^m\mu_i p_i^2+V({\mathbf q})\) where the \(\mu_i\) are complex constants and \(V\) is a polynomial. The time evolution of this system is governed by the Hamiltonian vector field \(\chi_{\mathcal H}=\sum_{i=1}^m \mu_ip_i\frac{\partial}{\partial q_i}-\sum_{i=1}^m\frac{\partial V}{\partial q_i}\frac{\partial}{\partial p_i}\). In this setting a ``proper Darboux polynomial'' \(F({\mathbf q},{\mathbf p}) \) is a nonzero polynomial in the variables \({\mathbf q},{\mathbf p}\) such that \(\chi_{\mathcal H}F=KF\), where \(K\) is a nonzero polynomial in \({\mathbf q},{\mathbf p}\). Thus \(F\) is not a constant of the motion but the hypersurface \(F({\mathbf q},{\mathbf p}) =0\) is invariant under the time evolution. This is an example of ``symmetry with a side condition''. A basic result proved by the authors in an earlier paper is that if \(F\) is a proper Darboux polynomial then \(F({\mathbf q},{\mathbf p})F({\mathbf q},-{\mathbf p})\) is a polynomial constant of the motion of the system. An important observation here is that the operator \(\chi_{\mathcal H}\) anticommutes with the operator involution \({\mathbf p}\rightarrow -{\mathbf p}\). Further, it can be shown that \(K\) must be a function of \({\mathbf q}\) alone. Thus \(\chi_{\mathcal H}\left(F({\mathbf q},{\mathbf p})F({\mathbf q},-{\mathbf p})\right)=0\). In the present paper this analysis is generalized to finite-dimensional Poisson systems where the time evolution is governed by the equation \({\dot {\mathbf x}}={\mathcal J}\cdot \nabla {\mathcal H}({\mathbf x})\), where \(\mathbf x\) is an \(n\)-tuple, \(\mathcal H\) is the generalized Hamiltonian, and \({\mathcal J}=(J_{ij}({\mathbf x}))\) is the \(n\times n\) skew-symmetric structure matrix which satisfies the \(n^3\) Jacobi conditions \(\sum_{\ell=1}^n\left( J_{\ell i}\partial_\ell J_{jk}+J_{\ell j}\partial_\ell J_{ki}+J_{\ell k}\partial_\ell J_{ij}\right)=0\). If the rank of \(\mathcal J\) is less than \(n\) there will be Casimir invariants \(D({\mathbf x})\) satisfying the equation \({\mathcal J}\cdot \nabla D=0\). The interest is in finding constants of the motion that are functionally independent of \({\mathcal H}\) and the Casimir invariants. The authors produce a number of examples where the Hamiltonian is greatly generalized, but still of the form kinetic energy + polynomial potential (in some sense). They use the property that if \(\mathcal J\) is of uniform rank \(n-s\) in some domain then there exists a change of coordinates such that the transformed structure matrix takes the canonical form \(\left(\begin{matrix} 0_m&I_m&\\ -I_m&0_m&\\ &&0_s\end{matrix}\right)\) where \(I_m,0_m\) are respectively the \(m\times m\) identity and null matrices, and there are \(s\) Casimir invariants \(D_1,\dots,D_s\). Once the problem is reduced to the Hamiltonian system \(\left(\begin{matrix} 0_m&I_m\\ -I_m&0_m\end{matrix}\right)\) the usual definitions and results for Darboux polynomials of Hamiltonian systems can be applied. The authors treat 3 successive generalizations of Hamiltonian systems: the first where a linear canonical transformation suffices to reduce the system to canonical form, the second where a diffeomorphic transformation is needed but \(s=0\), and the third where a diffeomorphic transformation is needed and \(s>0\). Several examples are given.