Revisiting canonical integration of the classical solid near a uniform rotation, canonical action angle coordinates, hyperbolic and elliptic, are constructed in terms of various power series with coefficients which are polynomials in a variable r 2 depending on the inertia moments. Normal forms are derived via the analysis of a relative cohomology problem and shown to be obtainable without the use of ellitptic integrals (unlike the derivation of the action-angles). Results and conjectures also emerge about the properties of the above polynomials and the location of their roots. In particular a class of polynomials with all roots on the unit circle arises.
OverviewIntegration of a rigid body motions is well known. In this paper we recall in Sec.2 the kinematic description of the coordinates in which the system is obviously integrable by reduction to quadratures (Deprit's coordinates, described in Eq.(2.5)). We try to treat simultaneously the motions near all proper rotations, at the expense of having to append labels a = ±1 to most quantities thus making the notation a little heavier. This is done to keep always clear that the stable and unstable motions are in some sense described by the same analytic functions.Our first aim is to obtain canonical coordinates p, q, in the vicinity of the proper rotations, such that the Hamiltonian, at fixed total angular momentum A, becomes a function of the product pq (near unstable proper rotations) or of p 2 + q 2 (near the stable ones). Although existence of such coordinates is well known and their determination is in principle controlled by appropriate (elliptic) integrals, [1, Sec.50], it remains, to our knowledge, only implicit in the literature, [2, Sec.69].In Sec.3 explicit expressions for the elliptic integrals are derived, Eq.(3.7), which give the motions in terms of variables p ′ , q ′ evolving exponentially in time on the hyperbolae p ′ q ′ = const or rotating uniformly on the circles p ′2 + q ′2 = const and "reducing" the problem to determine the nontrivial scaling function C (depending on p ′ q ′ or p ′2 + q ′2 ) which leads to the above mentioned canonical coordinates via p = p ′ C, q = q ′ C. In Appendix A given details on the computation of appropriate elliptic integrals: this should also clarify the advantages of the cohomological analysis of the following Sec.4 and 6.In Sec.4 the scaling function C is related to a Jacobian determinant between two differential forms, and it is computed: this is done, avoiding evaluation of other elliptic integrals, by computing the cohomology of the forms dB ∧ dβ (where B, β are two Deprit's coordinates, see Sec.2) and dp ′ ∧ dq ′ relative to the functions p ′ q ′ or p ′2 + q ′2 . The canonical coordinates are therefore completely determined, Eq.(4.10),(4.11) (together with Eq.(3.7)).The latter expressions are still quite implicit and in Sec.5 we bring them to a form suitable for the evaluation of the integrating coordinates via computable power series; and as an example a few terms of the scaling functions C and of t...