1. The case of uniform precession.

It was believed until very recently that a near-equatorial satellite would always keep up with the planet’s equator (with oscillations in inclination, but without a secular drift). As explained in Efroimsky and Goldreich (2004), this misconception originated from a wrong interpretation of a (mathematically correct) result obtained in terms of non-osculating orbital elements. A similar analysis carried out in the language of osculating elements will endow the planetary equations with some extra terms caused by the planet’s obliquity change. Some of these terms will be nontrivial, in that they will not be amendments to the disturbing function. Due to the extra terms, the variations of a planet’s obliquity may cause a secular drift of its satellite orbit inclination. In this article we set out the analytical formalism for our study of this drift. We demonstrate that, in the case of uniform precession, the drift will be extremely slow, because the first-order terms responsible for the drift will be short-period and, thus, will have vanishing orbital averages (as anticipated 40 years ago by Peter Goldreich), while the secular terms will be of the second order only. However, it turns out that variations of the planetary precession make the first-order terms secular. For example, the planetary nutations will resonate with the satellite’s orbital frequency and, thereby, may instigate a secular drift. A detailed study of this process will be offered in a subsequent publication, while here we work out the required mathematical formalism and point out the key aspects of the dynamics. ∗ In this article, we use the word “precession” in its most general sense which embraces the entire spectrum of changes of the spin-axis orientation – from the long-term variations down to the Chandler-type wobble down to nutations and to the polar wonder. 1 1 Physical motivation and the statement of purpose Ward (1973, 1974) noted that the obliquity of Mars may have suffered large-angle motions at long time scales. Later, Laskar and Robutel (1993) and Touma and Wisdom (1994) demonstrated that these motions may have been chaotic. This would cause severe climate variations and have major consequences for development of life. It is a customary assumption that a near-equatorial satellite of an oblate planet would always keep up with the planet’s equator (with only small oscillations of the orbit inclination) provided the obliquity changes are sufficiently slow (Goldreich 1965, Kinoshita 1993). As demonstrated in Efroimsky and Goldreich (2004), this belief stems from a calculation performed in the language of non-osculating orbital elements. A similar analysis carried out in terms of osculating elements will contain hitherto overlooked extra terms entailed by the planet’s obliquity variations. These terms (emerging already in the first order over the precession-caused perturbation) will cause a secular angular drift of the satellite orbit away from the planetary equator. The existence of Phobos and Deimos, and the ability of Mars to keep them close to its equatorial plane during obliquity variations sets constraints on the obliquity variation amplitude and rate. Our eventual goal is to establish such constraints. If the satellites’ secular inclination drifts are slow enough that the satellites stay close to Mars’ equator during its obliquity changes through billions of years, then the rigid-planet non-dissipative models used by Ward (1973, 1974), Laskar & Robutel (1993), and Touma & Wisdom (1994) will get a totally independent confirmation. If the obliquity-variation-caused inclination drifts are too fast (fast enough that within a billion or several billions of years the satellites get driven away from Mars’ equatorial plane), then the inelastic dissipation and planetary structure must play a larger role than previously assumed. Having this big motivation in mind, we restrict the current article to building the required mathematical background: we study the obliquity-variation-caused terms in the planetary equations, calculate their secular components and point out the resonant coupling emerging between a satellite’s orbiting frequency and certain frequencies in the planet axis’ precession. A more thorough investigation of this interaction will be left for our next paper. 2 Mathematical preliminaries: osculating elements vs orbital elements Whenever one embarks on integrating a satellite orbit and wants to take into account direction variations of the planet’s spin, it is most natural to carry out this work in a co-precessing coordinate system. This always yields orbital elements defined in the said frame, which are ready for immediate physical interpretation by a planet-based observer. A well camouflaged pitfall of this approach is that these orbital elements may come out non-osculating, i.e., that the instantaneous ellipses (or hyperbolae) parametrised by these elements will not be tangent to the physical trajectory as seen in the said frame of reference. 2 2.1 The osculation condition and alternatives to it An instantaneous orbit is parametrised with six independent Keplerian parameters. These include the three Eulerian angles i , ω, Ω which define the orientation of the instantaneous orbital plane, the eccentricity and semimajor axis of the instantaneous ellipse or hyperbola, and the mean anomaly at an epoch, Mo , which determines the initial position of the body. Often are employed sets of other variables. The variables are always six in number and are some functions of the Keplerian elements. These are, for example, the Delaunay set, the Jacobi set, and two sets offered by Poincare. More exotic is the Hill set which includes the true anomaly as one of the elements (Hill 1913). Systems of planetary equations in terms of the above sets of parameters may be derived not only through the variation-of-parameters (VOP) method but also by means of the Hamilton-Jacobi one. However, the latter approach, though fine and elegant, lacks the power instilled into the direct VOP technique and cannot account for the gauge invariance of the N-body problem (Efroimsky 2002a,b), important feature intimately connected with some general concepts in ODE (Newman & Efroimsky 2003). The Hamilton-Jacobi technique implicitly fixes the gauge, and thus leaves the internal symmetry heavily veiled (Efroimsky & Goldreich 2003). Below we shall present several relevant facts and formulae, while a more comprehensive treatment of the matter may be looked up in (Efroimsky & Goldreich 2003, 2004). As well known, a solution ~r = ~ f (C1, ..., C6, t) (1) to the reduced two-body problem ~̈r + Gm r2 ~r r = 0 (2) is a Keplerian ellipse or hyperbola parameterised with some set of six independent orbital elements Ci which are constants in the absence of disturbances. In the N-particle setting (or, more generally, in the presence of whatever disturbances) each body will be subject to some perturbing force ∆~ F(~r, ~̇r, t) acting at it from bodies other than the primary: ~̈r + Gm r2 ~r r = ∆~ F , (3) Solving the above equation of motion by the VOP method implies the use of (1) as an ansatz, ~r = ~ f (C1(t), ..., C6(t), t) , (4) the function ~ f being the same as in (1), and the “constants” Ci now being endowed with a time dependence of their own. Insertion of (4) into (3) is insufficient for determining the six functions Ci(t) . because the vector equation (3) comprises only three scalar equalities. To furnish a solution three more equations are needed. Since the age of Lagrange it has been advised in the literature to employ for this purpose the conditions of osculation,