Source-linked AI summary
Computing Nearly Singular Solutions Using Pseudo-Spectral Methods
Thomas Y. Hou, Ruo Li
TL;DR
The paper asks how pseudo-spectral methods can accurately compute nearly singular fluid-flow solutions despite aliasing errors. It compares 2/3 dealiasing with 36th-order Fourier smoothing in 1D Burgers and 3D Euler computations, finding that smoothing retains more effective modes and delivers more accurate, localized-error solutions without observed high-frequency instability.
Problem
Aliasing errors pollute high-frequency modes in pseudo-spectral computations, motivating better control than the traditional 2/3 dealiasing rule for nearly singular solutions.
Method
The paper systematically compares 2/3 dealiasing with 36th-order Fourier smoothing for nearly singular 1D Burgers and 3D incompressible Euler solutions.
Results
The Fourier smoothing method gives more accurate approximations than 2/3 dealiasing, capturing 12–15% more effective modes per dimension in Burgers and about 20% more for 3D Euler.
Takeaways & Limitations
Fourier smoothing localizes error near the most singular region, reduces error in smooth regions by several orders of magnitude, and shows no observed high-frequency instability.
Abstract
from arXiv · showhide
In this paper, we investigate the performance of pseudo-spectral methods in computing nearly singular solutions of fluid dynamics equations. We consider two different ways of removing the aliasing errors in a pseudo-spectral method. The first one is the traditional 2/3 dealiasing rule. The second one is a high (36th) order Fourier smoothing which keeps a significant portion of the Fourier modes beyond the 2/3 cut-off point in the Fourier spectrum for the 2/3 dealiasing method. Both the 1D Burgers equation and the 3D incompressible Euler equations are considered. We demonstrate that the pseudo-spectral method with the high order Fourier smoothing gives a much better performance than the pseudo-spectral method with the 2/3 dealiasing rule. Moreover, we show that the high order Fourier smoothing method captures about $12 \sim 15%$ more effective Fourier modes in each dimension than the 2/3 dealiasing method. For the 3D Euler equations, the gain in the effective Fourier codes for the high order Fourier smoothing method can be as large as 20% over the 2/3 dealiasing method. Another interesting observation is that the error produced by the high order Fourier smoothing method is highly localized near the region where the solution is most singular, while the 2/3 dealiasing method tends to produce oscillations in the entire domain. The high order Fourier smoothing method is also found be very stable dynamically. No high frequency instability has been observed.
1 Introduction
The paper compares traditional 2/3 dealiasing with 36th-order Fourier smoothing for nearly singular Burgers and 3D Euler solutions. Fourier smoothing retains more effective modes, improves accuracy, localizes error, and remains stable in the reported computations.
- Problem and methods: Aliasing errors contaminate high-frequency Fourier modes, while the 2/3 rule removes the final one-third of modes unchanged elsewhere.Fourier smoothing instead applies a smooth cutoff, although many existing approaches also damp the final one-third.
- Problem and methods: The study applies both methods to nearly singular solutions of the 1D Burgers and 3D incompressible Euler equations.The Burgers equation permits a semi-analytic solution for convergence studies near singularity time.
- Results: 12–15% more effective Fourier modes are captured by Fourier smoothing in each dimension than by 2/3 dealiasing for the 1D Burgers study.The method also approximates unfiltered high-frequency coefficients accurately.
- 3D Euler results: 20% more effective Fourier modes are captured by Fourier smoothing than by 2/3 dealiasing for the 3D Euler computations.Both methods converge under mesh refinement, but abrupt spectral cutoff causes oscillations in vorticity contours for 2/3 dealiasing.
- Stability: Fourier smoothing is reported as stable and robust, with no observed high-frequency instability in either equation.The resolution study is based on accuracy rather than stability.
2 Convergence study of the two pseudo-spectral methods for the 1D Burgers equation
The study compares 2/3 dealiasing with high-order Fourier smoothing for nearly singular 1D Burgers solutions. Fourier smoothing retains more high-frequency information and achieves lower, more localized errors with stable computations.
- Burgers equation and setup: The 1D Burgers equation provides a prototype with a quadratic nonlinear convection term and finite-time shock formation.
- Filtering methods: 12 ∼15% more modes are retained by Fourier smoothing than by the 2/3 dealiasing rule.The smoothing remains close to 1 below approximately 4N/5 and decays rapidly toward machine precision near N.
- Convergence results: Fourier smoothing produces smaller L∞ errors and several-orders-of-magnitude smaller errors in smooth regions than 2/3 dealiasing.The comparison uses three resolutions and examines both physical and spectral convergence.
- Error structure and stability: The 2/3 method generates oscillatory errors across the domain, while Fourier smoothing localizes errors near the shock singularity and shows no observed high-frequency instability.The widespread oscillations are attributed to Gibbs phenomenon and loss of L2 energy from the abrupt spectral cut-off.
- Convergence results: Fourier smoothing retains about 20% more Fourier modes for a given resolution, and its additional modes accurately approximate the exact spectrum as singularity develops.The two spectra are nearly indistinguishable when the solution is sufficiently smooth and resolved.
3 Computing nearly singular solutions of the 3D Euler equations using pseudo-spectral methods
The study compares 2/3 dealiasing and Fourier smoothing for nearly singular 3D Euler computations, finding that Fourier smoothing retains more effective modes and delivers greater accuracy and stability. The computations also find no finite-time blowup through t = 19 for the tested Kerr initial condition.
- Method: The 3D Euler study uses Kerr’s initial condition and compares the two pseudo-spectral methods through careful convergence tests in physical and spectral space.The computations use periodicity and exploit the stated vorticity symmetries to compute one quarter of the periodic cell.
- Spectral comparison: 12 ∼15% of modes in each dimension remain accurate beyond the 2/3 cut-off, yielding about 20% more effective modes in 3D with Fourier smoothing.At the largest resolution, Fourier smoothing retains more than 320 million effective modes, 140 million more than 2/3 dealiasing.
- Spectral comparison: Fourier smoothing produces smoother spectra beyond the 2/3 cut-off, whereas 2/3 dealiasing introduces noticeable oscillations near its cut-off as the solution becomes less resolved.The two spectra are nearly identical at early times when the solution is smooth, but diverge near the cut-off from t = 12 onward.
- Physical-space comparison: For a given resolution, Fourier smoothing is more accurate, while the 2/3 method becomes significantly under-resolved after t = 18 at resolution 768 × 512 × 1536.The two largest resolutions give relatively small differences for vorticity and nearly indistinguishable maximum velocity results.
- Stability: Fourier smoothing remains stable and robust throughout the 3D computations, with no observed high-frequency instability.The resolution study was conducted for accuracy rather than stability, while the method nevertheless showed stable dynamics.
- Finite-time singularity: Maximum vorticity rises from 0.669 to 23.46 by t = 19, but the computations show no finite-time blowup up to that time.The vortex stretching term exhibits cancellation and grows more slowly than C∥ω∥∞log(∥ω∥∞); maximum velocity also remains bounded.
4 Conclusion Remarks
The paper’s systematic studies find that high-order Fourier smoothing more accurately and robustly computes nearly singular Burgers and 3D Euler solutions than traditional 2/3 dealiasing.
- High-order Fourier smoothing retains significant Fourier modes beyond the 2/3 cutoff while smoothly attenuating high frequencies.The method uses a 36th-order Fourier smoothing function.
- 1D Burgers equation: 13 digits of accuracy are obtained for the 1D Burgers solution sufficiently close to the singularity time.This enables accurate numerical-error estimation and supports the convergence study.
- Fourier smoothing gives more accurate approximations than 2/3 dealiasing for both 1D Burgers and 3D incompressible Euler equations.The comparison uses convergence studies in physical and spectral spaces, with the highest affordable resolution for 3D Euler.
- Error distribution: Its error is concentrated near the most singular region, whereas 2/3 dealiasing produces broader pointwise errors and oscillations in smooth regions.In smooth regions, Fourier smoothing errors are several orders of magnitude smaller than those from 2/3 dealiasing.
- Effective Fourier modes: Fourier smoothing retains about 12 ∼15% more effective Fourier modes per dimension, with gains as large as 20% for 3D problems.The paper identifies this gain as significant for large-scale computation.
- 3D incompressible Euler equations: Both methods produce qualitatively similar 3D Euler results, but 2/3 dealiasing develops relatively large late-time oscillations from the Gibbs phenomenon.