Stability of oscillating canonical systems in nondegenerate case with application to celestial mechanics (Q5933576)

Kolmogorov asserted that most tori are not destroyed by small perturbations of the Hamiltonian. Subject to a very particular ``non-resonance'' restriction, Arnol'd and Moser proved this assertion. From several concrete examples it is known that this restriction is sufficient but not necessary. The particular example of \textit{V. I. Arnol'd} [Sov. Math., Dokl. 5, 581--585 (1964); translation from Dokl. Akad. Nauk SSSR 156, 9--12 (1964; Zbl 0135.42602)] allows distinct interpretation in a mathematical content and in a physical one, where certain state variables are less ``significant'' than others, for example ``amplitudes'' and ``angles''. To be more explicit, the Arnol'd example admits for \(\mu=0\) (unperturbed Hamiltonian) the integrable solution (A1) \(x_2(t)= C_2\), \(y_2(t)= - C_2t+ C_3\), \(y_1(t)= G(t,\varepsilon, C_2, C_4)\), \(x_1(t)= {dG\over dt}\), where \(G\) is the general solution of the ``ideal pendulum'' equation (A2) \({d^2G\over dt^2}+ \varepsilon\sin G= 0\), \(\varepsilon> 0\), \(C_1, C_2, C_3, C_4=\) integration constants. It is well known that \(G\) can be expressed by means of elliptic functions. The solution (A1), (A2) is unstable because not all state variables are bounded. Since the full Hamiltonian depends smoothly on the parameter \(\mu\) (at least near \(\mu=0\)), the ``perturbed'' solution is also unstable (convergence of classical ``Abschätzungssätze'' on solutions of close ODE-s). In a polynomial interpretation unbounded ``angles'' dot not matter, and (A2) admits in the \(x_1\), \(y_1\) phase place a centre-saddle structure, repeated \(\text{mod}(2\pi)\). The oscillatory cell is a (degenerate) torus. If \(\mu\neq 0\), \(x_1(t)\) and \(y_1(t)\) become coupled to other state variables and the forms (isolated for \(\mu= 0\)) is qualitatively changed''. This change has become known as ``Arnol'd diffusion''. The author, and many physicists, consider such an occurrence to be a ``key'' counterexample to Kolmogorov's assertion. Semantically this may be true, but not factually. It is by now common knowledge that Hamiltonian dynamics can be ``orderly'', ``chaotic'', or partly both. The history of ``chaos theory'' is summarized in the booklet of \textit{R. Abraham} and \textit{Y. Ueda} [World Scientific Series on Nonlinear Science. Series A 39, Singapore: World Scientific (2000; Zbl 0958.00013)]. Arnol'd diffusion is not related to ergodic theory. Ref. 8 of the paper is therefore irrelevant. It is also unrelated to any physically observable diffusion process. The statement on p. 369 that Hamiltonian dynamics can also be studied in the Hamilton-Jacobi PDE is uncountable. The fact that extremals (i.e. trajectories of Hamiltonian ODEs) are transversals to the 5-surfaces forms the foundation of \textit{R. Bellman's} ``Dynamic Programming'' [Princeton, New Jersey: Princeton University Press (1947; Zbl 0077.13605)], although the notion of chaos is not mentioned in this book. It is also not even vaguely alluded to in the author's paper. A comprehensive answer to the question: ``Do chaotic dynamics constitute a major or a minor texture of the \(n\)-body problem?'' would be very useful in guiding future research. No light is shed on this problematic in the paper. It must be decided by a professional astronomer whether the estimates given in the paper do increase the understanding of at least ``our'' planetary movements.