Do average Hamiltonians exist? (Q1976096)

The author shows that, even if generating functions are not used, given the Hamiltonian \[ H= H_0(J)+ \varepsilon R(\theta, J)\quad (\varepsilon\ll 1), \] it is not possible to transform it into a new Hamiltonian \(H^*(J^*)\) (depending only on the new actions \(J^*\)), through a canonical transformation given by zero-average trigonometrical series. Note that a general solution cannot be found even in a particular case, from celestial mechanics, in which the disturbing potential \(R(\theta, J)\) is a cosine series.