A fast spectral-multigrid Poisson solver in non-Cartesian geometries

Abstract

Accurate and efficient computation of self-gravity is essential in astrophysical fluid dynamics, particularly in spherical and cylindrical geometries where large radial dynamic ranges and non-axisymmetric structures arise. Poisson solvers in such settings must simultaneously achieve high accuracy, scalability, and flexibility across a wide range of grid configurations and physical regimes. We present a robust and scalable Poisson solver for three-dimensional non-Cartesian geometries, supporting both spherical and cylindrical coordinates with either uniform or logarithmic radial discretizations. The method employs azimuthal Fourier decomposition to transform the 3D Poisson equation into a set of independent 2D Helmholtz equations. These are solved using a geometrically consistent multigrid algorithm that preserves second-order accuracy on both uniform and non-uniform grids. Vacuum boundary conditions are implemented through a screening-mass approach, enabling accurate solutions in domains with open boundaries, inner cavities, and strongly non-axisymmetric mass distributions. Owing to the differing convergence rates of Fourier modes -- where higher-order modes converge more rapidly -- the solver allows efficient mode-by-mode treatment. The combination of spectral decomposition and multigrid acceleration provides an efficient and flexible computational framework. The solver is implemented in the PLUTO code and validated against both analytical solutions and dynamical test problems in spherical and cylindrical geometries. Results demonstrate second-order convergence and excellent agreement with reference solutions. Weak-scaling tests up to 4096 cores show strong parallel performance, with the Poisson solve remaining subdominant to magnetohydrodynamic update cost. This makes the method well suited for large-scale simulations of star formation, accretion disks, and gravitational instabilities.