Fast and backward stable transforms between spherical harmonic expansions and bivariate Fourier series

Richard Mikael Slevinsky
DOI: https://doi.org/10.48550/arXiv.1705.05448
2017-11-05
Abstract:A rapid transformation is derived between spherical harmonic expansions and their analogues in a bivariate Fourier series. The change of basis is described in two steps: firstly, expansions in normalized associated Legendre functions of all orders are converted to those of order zero and one; then, these intermediate expressions are re-expanded in trigonometric form. The first step proceeds with a butterfly factorization of the well-conditioned matrices of connection coefficients. The second step proceeds with fast orthogonal polynomial transforms via hierarchically off-diagonal low-rank matrix decompositions. Total pre-computation requires at best $\mathcal{O}(n^3\log n)$ flops; and, asymptotically optimal execution time of $\mathcal{O}(n^2\log^2 n)$ is rigorously proved via connection to Fourier integral operators.
Numerical Analysis
What problem does this paper attempt to address?
This paper attempts to solve the problem of fast and backward - stable conversion between spherical harmonic expansion and bivariate Fourier series. Specifically, the author proposes a two - step method to achieve this conversion: 1. **Step 1**: Convert the expansion of the normalized associated Legendre functions of all orders into the expansion of order zero and order one. This step processes the well - conditioned connection coefficient matrix through butterfly factorization. 2. **Step 2**: Re - expand these intermediate expressions into trigonometric forms. This step is achieved through fast orthogonal polynomial transformation, making use of hierarchical low - rank matrix decomposition. ### Specific Problem Description Spherical harmonic functions are the natural basis for square - integrable functions on the sphere. They are very elegant mathematically and stable in numerical calculations. However, in some application scenarios, such as the fast Fourier transform (FFT), bivariate Fourier series can provide faster calculation speed and better flexibility. Therefore, how to perform efficient and stable conversion between spherical harmonic expansion and bivariate Fourier series has become an important research topic. ### Main Contributions of the Paper - **Fast Conversion Algorithm**: The paper proposes a new algorithm that can complete the conversion from spherical harmonic expansion to bivariate Fourier series within the time complexity of O(n^2 log^2 n), which is asymptotically optimal. - **Backward Stability**: By using butterfly factorization and fast orthogonal polynomial transformation, the backward stability of the algorithm is ensured, thus avoiding the need for high - precision calculations. - **Theoretical Proof**: Rigorously prove the rank properties of the connection coefficient matrix, which is applicable not only to the algorithm proposed in this paper but also to the acceleration of other related methods. ### Application Background This conversion has important applications in solving partial differential equations, time - evolution problems, and multi - sphere scattering problems. For example, in time - evolution problems, the spherical harmonic basis can diagonalize certain linear differential operators (such as the Laplace - Beltrami operator), while the bivariate Fourier series can be quickly evaluated at grid points. Therefore, a method that can stably convert between these two representations is very valuable. ### Summary This paper solves the problem of fast and backward - stable conversion between spherical harmonic expansion and bivariate Fourier series, proposes a new algorithm based on butterfly factorization and fast orthogonal polynomial transformation, and strictly proves its complexity and stability. This result is of great significance in multiple scientific computing fields.