In an astrophysical context (cf., §2.1 in Binney & Tremaine 1987),
the Newtonian potential is determined
by solving the appropriate boundary-value problem that is given by
the three-variable Poisson equation,
(1) |
where is the Laplacian, and is the gravitational constant. Exterior to an isolated mass distribution, physically correct boundary conditions may be imposed on Poisson's equation through the following integral formulation of the potential problem,
(2) |
Due to the long-range nature of the Newtonian/Coulomb interaction, a physically correct solution of Poisson's equation and, hence, an accurate determination of the potential ``interior'' to the boundary can be highly boundary value dependent. If inaccurate potential values are given on the boundary, the interior solution will reflect these defects.
In numerical simulations of astrophysical systems, a standard procedure for determining the value of the gravitational potential along the boundary of a computational mesh has been the multipole method. In this method, an approximate solution to equation (2) is obtained by summing up higher and higher order spherical multipole moments (monopole, dipole, quadrupole, etc.) of the mass distribution. The multipole moments themselves are normally cast in terms of two quantum numbers -- one meridional and one azimuthal m -- and for practical reasons the series summation is truncated at a finite value of and m. An alternative to the standard multipole expansion has been given by Cohl & Tohline (1999) in terms of a single sum over the azimuthal quantum number m. This new expression yields the pure azimuthal mode contribution to the Newtonian potential. In terms of the standard multipole description, this new expansion is equivalent to completing the infinite sum over the quantum number for each given value of the quantum number m (Cohl et al. 2000). It is now clear to us that this new expansion can be extended to every rotationally invariant coordinate system which separates the three-variable Laplace equation.
In §2 we describe the rotationally invariant coordinate systems which separate Laplace's equation and briefly describe some of the properties of these systems related to the double summation/integration expressions for the reciprocal distance between two points (hereafter, Green's function for Laplace's equation, or just Green's function) in these coordinate systems. In §3 we describe the key expressions for the Green's function in circular cylindrical coordinates, and show how these expressions are consistent with an azimuthal Fourier series expansion for the Green's function whose coefficients are given in terms of toroidal functions. In §4 we describe some of the highly symmetric properties of toroidal functions, such as their behavior for negative degree and order. We also present toroidal function implications of the Whipple formulae for associated Legendre functions. Using these Whipple formulae for toroidal functions, we then obtain a new expansion which is shown to be equivalent to the expansion given in Cohl & Tohline (1999). In §5 we briefly summarize several key mathematical implications of these expansions as they apply to prolate spheroidal, oblate spheroidal, and, parabolic coordinates. In §6 we present some key mathematical implications of these expansions as they apply to bispherical, and toroidal coordinates. All of the coordinate systems presented in §§5 and 6 have known double integration/summation expansions for the Green's function and therefore can be easily related to these new alternative expansions. A detailed treatment in spherical coordinates is being presented elsewhere (Cohl et al. 2000). The mathematical relations we derive below are either in the form of an infinite integral or an infinite series expansion over the set of basis functions which separate Laplace's equation.
Since toroidal functions can be used to unify all of the rotationally invariant coordinate systems which separate Laplace's equation, we suspect that they represent a basis set that is better suited for general studies of nonaxisymmetric mass/charge distributions than, for example, spherical harmonics. It is with this in mind, that the relations in this paper are presented for use in numerical and analytical solutions to any theoretical physics problem which requires accurate evaluation of the gravitational and/or Coulomb potential.