A criterion for the nonexistence of an additional analytic integral in Hamiltonian systems with \(n\) degrees of freedom (Q1594578): Difference between revisions

The author considers Hamiltonians of the form (1) \(H=\tfrac 12{\mathbf p}^2+V(\mathbf q)\), where \(\mathbf q,\mathbf p\in{\mathbf R}^n\), and \(V(\mathbf q)\) is homogeneous of degree \(k\) in \(\mathbf q\) and has the property that there is a solution \(\mathbf c\in {\mathbf R}^n\) of the equation \(\nabla V(\mathbf c)=\mathbf c\). Given the eigenvalues \(\lambda_i\) \((i=1,2,\cdots,n-1)\), \(\lambda_n:=k-1\) of the Hessian \(D^2H(\mathbf c)\) he constructs the \(n\) numbers \(\theta_i=\sqrt{(k-2)^2+ 8k\lambda_i}\), \(i=1,2,\cdots,n\), (in particular \(\theta _n:=3k-2)\) (which are differences of pairs of so-called Kovalevskaya exponents) and shows that if these numbers are independent over the rationals then the Hamiltonian (1) has no additional (meromorphic) integral besides the Hamiltonian itself. The proof is based on a famous theorem of Ziglin on the relationship between integrability and the structure of the monodromy group associated with the normal variational equation of some fixed solution \(\Gamma\) of (1). In the present case \(\Gamma\) is the straight line solution: \(\varphi(t)\mathbf c\), where \(\varphi(t)\) satisfies \(\dot\varphi^2=(2/k) (1-\varphi^2)\). Specifically, under the stated condition on the numbers \(\theta_i\) the group contains two nonresonant elements of type \(M_l=\roman{diag}(M_l(\lambda_1),M_l (\lambda_2),\cdots,M_l(\lambda_{n-1}))\) \((l=1,2)\), where \(M_1(\lambda_i),M_2(\lambda_i)\in \roman{SL}(2,\mathbf C)\) are certain noncommuting members of the monodromy group of the hypergeometric ODE with \(a+b=\tfrac 12-1/k\), \(ab=-\lambda_i/2k\), \(c=1-1/k\). According to the quoted theorem of Ziglin this suffices to rule out the existence of a meromorphic integral independent of \(H\). However, as stressed by the author, it is important that \(k\neq 2\) and that \(H\) is of the type (1) (i.e. kinetic energy \(+\) potential energy). In the case of more general homogeneous Hamiltonians counterexamples to the main statement of the present paper were first presented by the reviewer in joint work with \textit{R. C. Churchill} and {D. Rod} in their paper ``On Kovalevski exponents'' [in Saenz Festschrift, Gordon \& Breach, New York, (1990)]. The present author explains the existence of these counterexamples by showing that for them the noncommutativity property of the \(M\)'s is violated.