Abstract
Several numerical problems, such as radiative transfer, require the interpolation of discrete data containing various types of discontinuities. This study proposes and tests a novel fourth-order Weighted Essentially Non-Oscillatory (WENO) interpolation technique for both uniform and nonuniform grids. Numerical tests demonstrate that this method guarantees fourth-order accuracy in smooth regions and maintains a non-oscillatory property near discontinuities, avoiding the "clipping" of smooth extrema common in other high-order monotonic interpolations.
Introduction
In computational astrophysics, specifically in the study of radiative transfer within stellar atmospheres, researchers frequently encounter data with both smooth gradients and sharp discontinuities. Traditional high-order interpolation methods, such as Cubic Splines, often suffer from the Gibbs phenomenon, producing spurious oscillations (ringing) near jumps. Conversely, "monotonicity-preserving" methods like Bézier or Hermite interpolations often sacrifice accuracy at smooth peaks, degenerating toward lower-order accuracy. This paper introduces a fourth-order WENO scheme designed to provide high accuracy in smooth regions while remaining stable near shocks.
Methodology
The proposed technique utilizes a fourth-order WENO interpolation framework. The core of the method involves:
- Adaptive Stencils: The algorithm evaluates multiple candidate substencils to determine local data behavior.
- Smoothness Indicators: New smoothness indicators are derived for both uniform and nonuniform grids to measure local volatility.
- Weighted Averaging: A convex combination of candidate substencils is used. In smooth regions, weights are balanced to achieve fourth-order accuracy. Near discontinuities, the weight of the "jumpy" stencil is minimized to prevent ringing.
- Symmetry: A symmetric four-point stencil is employed around the target interval, ensuring computational balance and ease of integration into existing simulation codes.
Numerical Results
The algorithm was evaluated using synthetic benchmark functions, including Exponential, Heaviside, Discontinuous Sine, and Gaussian functions. These tests were conducted on both uniform and nonuniform grids to assess the Experimental Order of Convergence (EOC).
As shown in Table 1, the WENO 4 method consistently achieves an EOC of approximately 4.0 in smooth regions (Exponential and Gaussian). Notably, in the Gaussian test, WENO 4 maintains high accuracy (EOC 4.371) where Monotonic Hermite methods drop to an EOC of 2.922, confirming that the new method does not "clip" smooth extrema. In the presence of discontinuities (Heaviside), the EOC naturally drops to approximately 1.0, but the method successfully avoids the oscillations characteristic of standard cubic methods.
Discussion and Limitations
A key advantage of the WENO 4 method is its local nature, making it highly efficient for parallel computing environments. Furthermore, its ability to handle nonuniform grids makes it a versatile tool for real-world physics simulations where grid spacing is often irregular.
However, some limitations persist:
- Differentiability: The interpolation is not $C^1$ differentiable at grid points due to shifting weights.
- Positivity: Near sharp drop-offs, the method may occasionally produce slightly negative values for strictly positive functions.
- Computational Overhead: The calculation of smoothness indicators and weights makes WENO more computationally intensive than simpler linear or cubic schemes.
Conclusion
The novel fourth-order WENO interpolation guarantees high accuracy in smooth regions while effectively handling discontinuities without the drawbacks of traditional monotonic or high-order methods. This technique is particularly suitable for multidimensional radiative transfer problems, multigrid methods, and formal solutions in dynamic astrophysical environments.
Summary updated