In this paper, we use Fourier analysis to study the superconvergence of the semi-discrete discontinuous Galerkin method for scalar linear advection equations in one spatial dimension. The error bounds and asymptotic errors are derived for initial discretization by $L_2$ projection, Gauss-Radau projection, and other projections proposed by Cao et. al. For pedagogical purpose, the errors are computed in two different ways. In the first approach, we compute the difference between the numerical solution and a special interpolation of the exact solution, and show that it consists of an asymptotic error of order $2k+1$ and a transient error of lower order. In the second approach, as by Chalmers and Krivodonova, we compute the error directly by decomposition into physical and nonphysical modes, and obtain agreement with the first approach. We then extend the analysis to vector conservation laws, solved using the Lax-Friedrichs flux. We prove that the superconvergence holds with the same order. The error bounds and asymptotic errors are demonstrated by various numerical experiments for scalar and vector advection equations.
翻译:在本文中,我们使用 Fourier 分析法来研究半分解的Galerkin 方法在一个空间维度中用于 scalar 线性对映方方程式的超共性。 错误界限和无症状错误是用美元=2美元的投影、 Gauss- Radau 投影和Cao 等人提议的其他预测为初始分解而得出的。 为了教学目的,错误是用两种不同的方法计算的。 在第一个方法中, 我们计算数字解决方案和精确解决方案的特殊内插之间的差别, 并显示它是由2k+1美元的顺序和较低顺序的瞬态错误组成的。 在第二个方法中, 我们用物理和非物理模式解析直接计算出错误, 并取得与第一种方法的一致。 我们然后将分析扩大到矢量保护法, 使用 Lax- Friedrichs 通量来解析。 我们证明, 超相趋同的顺序是2k+1美元, 和低顺序的中位错误。 在第二个方法中, 正如Chalmermal 和 Kental 实验所显示的数值和矢量性误。