Rhodonea curves as sampling trajectories for spectral interpolation on the unit disk (Q2663136)

The aim of this article is spectral interpolation on the two-dimensional unit disk. For this, interpolation points have to be found that render the interpolation system uniquely solvable and, if possible, to provide stable and convergent approximations. It is shown in this paper that this is working well with so-called rose points created by rhodonea curves. These curves are defined by an angle \(\alpha\) and two natural numbers called a frequence vector by \[(\cos(m_2t)\cos(m_1t-\alpha\pi),\cos(m_2t)\sin(m_1t-\alpha\pi)).\] It is also shown that one can get continuous interpolants as well. The nodes are taken in an equally spaced way along the curves (equally spacing meaning in the same time steps along the rose curves), and the interpolation matrices are formed using these points and a Chebyshev-Fourier basis. Concretely, fast convergence is shown for sufficiently smooth approximants, and also the author finds that the Lebesgue numbers of the (interpolation) operators diverge logarithmically only, providing for a stable approximation. The interpolants are expressed using Lagrange functions, the most important theorem being Theorem 17 where this formula is provided. A method for computing these interpolants efficiently using FFT is developed (to be found in Section~7) too.