Chapter 1. Background and results Chapter 2. Kepler maps and the Perihelia reduction 2.1. The P-map vs rotations and reflections Chapter 3. The P-map and the planetary problem 3.1. A general property of Kepler maps 3.2. The case of the P-map Chapter 4. Global Kolmogorov tori in the planetary problem 4.1. A domain of holomorphy 4.2. A normal form for the planetary problem 4.3. A "multi-scale" KAM Theorem and proof of Theorem A Chapter 5. Proofs 5.1. Normalization of fast angles 5.2. Secular normalizations Appendix A. Computing the domain of holomorphy A.1. On the analyticity of the solution of Kepler equation A.2. Proof of Proposition 4.2 Appendix B. Proof of Lemma 3.2 Appendix C. Checking the non-degeneracy condition Appendix D. Some results from perturbation theory D.1. A multi-scale normal form theorem D.2. A slightly-perturbed integrable system Appendix E. More on the geometrical structure of the P-coordinates, compared to Deprit's coordinates Bibliography iii CHAPTER 1 1. BACKGROUND AND RESULTSFirstly, the P-map is well defined in the case of the planar problem, while JRBD coordinates are not. Everyone knows, in fact, that the starting point for the Radau-Jacobi reduction is the so-called "line of the nodes", the straight line determined by the intersection between the planes of the two orbits. When the orbits of the two planets belong to the same plane, this is not defined. A similar circumstance arises for Boigey-Deprit's coordinates, since their construction relies on certain straight lines in the space, which again lose their meaning in case of co-planarity.The proof of Arnold's theorem given in [27,9] is not affected by such singularity, since, as said, it relies on RPS coordinates, which, at the expense of one more degree of freedom, are well defined for co-planar motions -in that case they reduce to the classical Poincaré coordinates. It has its consequences when one wants to compare results for the fully reduced systems, in space or in the plane. The singularity of the chart does not allow one to state that motions in the spatial problem with minimum number of independent frequencies starting with very small inclinations stay close to the corresponding planar motions. Notwithstanding further studies appearing in [28], where this problem is partially overcome (via the construction of regular coordinates for coplanar motions defined locally), it would be nice, in principle, to handle a global system of action-angle coordinates which completely reduces rotations and is shared simultaneously by the planar and the spatial problem.Secondly, the P-map is well adapted to reflection symmetries of the problem, while JRBD coordinates are not, as discussed in [25,29].1 In the framework of the study of canonical coordinates for the planetary system, by "partial reduction", we mean a system of canonical coordinates where a couple of conjugated coordinates consists of integrals (e.g., functions of the three components of the total angular momentum). By "full reduction", we mean a partial reduction where also another...