Fast high order ADER schemes for linear hyperbolic equations
Introduction
Finite volume schemes have become very popular in computational fluid dynamics due to their robustness and flexibility. They consist of two steps: the reconstruction and the flux calculation. The discrete values are approximations of cell averages. By the reconstruction step, local values are interpolated from the average values to calculate the numerical flux between the grid cells. The construction of high order schemes concerning spatial discretization has been introduced with the idea of ENO and WENO interpolation [4], [5]. But the time integration usually becomes a limiting factor for accuracy. All these schemes are generally discretized in time with Runge–Kutta (RK) schemes. These time integration schemes become inefficient for orders of accuracy higher than four. It can be proven, that all explicit RK time integration schemes of order higher than four need more integration stages than their order. This is the so-called Butcher barrier [2].
The idea of the ADER approach of Toro (see, e.g., [7], [9], [11]) is to circumvent this efficiency barrier for the time discretization by considering finite space–time volumes, where the temporal evolution of the fluxes over the borders of the finite volumes is estimated by a Taylor series in time. The time derivatives are then replaced by space derivatives using the Lax–Wendroff procedure. With this approach it is (at least theoretically) possible to construct schemes of arbitrary high order in space and time with Δt∼Δx close to the stability limit. However, the Lax–Wendroff procedure becomes cumbersome already in the case of linear systems of PDE's for very high order ADER schemes. If the ADER approach is applied without any special treatment, it is not competitive with e. g. finite difference schemes for the solution of linear PDE's with respect to the computational effort. In this paper we develop a particular formulation of the ADER approach, which makes use of simplifications that can be taken into account for linear PDE's on structured meshes. A typical field of application would be computational aeroacoustics to simulate noise propagation in the time domain or electromagnetic wave propagation.
The scope of the paper is as follows. In Section 2 we first give the formulas which are essential for the reformulation. We derive then a finite difference-like formulation of the ADER approach which significantly enhances efficiency with respect to computational effort. In Section 3 we show the convergence rates obtained with the reformulated ADER schemes in numerical experiments for the two-dimensional linearized Euler equations up to 16th order of accuracy in space and time.
Section snippets
The fast-ADER approach
The ADER approach is a scheme developed to calculate numerical solutions of systems of hyperbolic PDE's up to an, at least theoretically, arbitrary order. Toro and Millington [11], [13] first developed the idea in one space dimension. Here, the 1st order scheme reduces to the Godunov scheme and the 2nd order scheme is equivalent to the MGRP approach by Toro [12], [13], which is a simplified GRP scheme of Ben-Artzi and Falcovitz [1]. The extension to multi-dimensions is straightforward and can
Numerical results
In Computational AeroAcoustics (CAA) good wave propagation properties are a crucial point for numerical schemes. The dispersion and dissipation errors must be as low as possible in order to provide accurate wave propagation over long distances on reasonably coarse grids. In [7], [9] we plotted the amplitude- and phase errors for the ADER approach. As nothing is neglected, they apply without any change to the fast-ADER approach. An ADER scheme has approximately the same phase error as a sixth
Conclusions
If the following assumptions hold: The mesh sizes Δx and Δy is constant and the PDE is linear with constant coefficients, then the fast formulation of the ADER scheme leads to a single-step scheme in time of arbitrary order of accuracy which is dramatically faster than the original formulation. For the 20th order scheme the speed-up is about 2000. We implemented the scheme in such a way, that the order of accuracy in space and time becomes only a parameter to be specified, so a really arbitrary
References (13)
- et al.
A second-order godunov-type scheme for compressible fluid dynamics
J. Comput. Phys.
(1984) - et al.
A finite-volume high-order ENO scheme for two-dimensional hyperbolic systems
J. Comput. Phys.
(1993) - et al.
Uniformly high order accurate essentially non-oscillatory schemes III
J. Comput. Phys.
(1987) The Numerical Analysis of Ordinary Differential Equations: Runge–Kutta and General Linear Methods
(1987)- et al.
Essentially non-oscillatory and weighted essentially non-oscillatory schemes for hyperbolic conservation laws
Numerical Computation of Internal and External Flows Vol II: Computational Methods for Inviscid and Viscous Flow
(1988)
Cited by (95)
On improving the efficiency of ADER methods
2024, Applied Mathematics and ComputationTent-pitcher spacetime discontinuous Galerkin method for one-dimensional linear hyperbolic and parabolic PDEs
2023, Computers and Mathematics with ApplicationsHigh resolution numerical algorithm of order (2, 4, 4) for 2D quasi-linear hyperbolic equations based on half-step discretization on an unequal grid
2023, Partial Differential Equations in Applied MathematicsA new Mean Preserving Moving Least Squares method for Arbitrary Order Finite Volume schemes
2023, Applied Mathematics and ComputationExtended stress gradient elastodynamics: Wave dispersion and micro-macro identification of parameters
2021, International Journal of Solids and Structures