Algebraic non-integrability of the Cohen map
Rational and birational maps (14E05) Completely integrable finite-dimensional Hamiltonian systems, integration methods, integrability tests (37J35) Completely integrable infinite-dimensional Hamiltonian and Lagrangian systems, integration methods, integrability tests, integrable hierarchies (KdV, KP, Toda, etc.) (37K10)
The Cohen map \[ \phi(x,y)=(\sqrt{1+x^2}-y,x) \] is a curious example of an area-preserving plane diffeomorphism. Numerical studies show that the plane is foliated by smooth, possibly piecewise analytic invariant curves. However, no proof of this fact is known. The only information comes from KAM theory which asserts that invariant curves exist locally near the fixed point \((1/\sqrt{3}, 1/\sqrt{3})\). The current paper demonstrates that if an integral, that is a function \(G\) with \(G\circ \phi=G\), exists, it cannot be an algebraic function of \(x\) and \(y\). The proof of this fact uses the theory of iterations of algebraic relations developed by the first author. For example, the function \(\phi\) can be represented as an algebraic relation on \(\mathbb{C}\times \mathbb{C}\), namely \[ R= \{((x_1,y_1), (x_2,y_2));\;(x_2+y_1)^2= 1+x_1^2,\;y_2=x_1\}. \] Iterations of this relation can be considered in a natural way. From general considerations it is shown that if an algebraic integral existed, then the relation \(R\) restricted to any invariant variety would be periodic. But this contradicts KAM theory which asserts the existence of invariant curves with irrational rotation numbers.
This page was built for publication: Algebraic non-integrability of the Cohen map
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q1385438)
