We propose an Exponential DG approach for numerically solving partial differential equations (PDEs). The idea is to decompose the governing PDE operators into linear (fast dynamics extracted by linearization) and nonlinear (the remaining after removing the former) parts, on which we apply the discontinuous Galerkin (DG) spatial discretization. The resulting semi-discrete system is then integrated using exponential time-integrators: exact for the former and approximate for the latter. By construction, our approach i) is stable with a large Courant number (Cr > 1); ii) supports high-order solutions both in time and space; iii) is computationally favorable compared to IMEX DG methods with no preconditioner; iv) requires comparable computational time compared to explicit RKDG methods, while having time stepsizes orders magnitude larger than maximal stable time stepsizes for explicit RKDG methods; v) is scalable in a modern massively parallel computing architecture by exploiting Krylov-subspace matrix-free exponential time integrators and compact communication stencil of DG methods. Various numerical results for both Burgers and Euler equations are presented to showcase these expected properties. For Burgers equation, we present detailed stability and convergence analyses for the exponential Euler DG scheme.
翻译:我们提出一个指数化的 DG 方法,用于从数字上解决部分差异方程式(PDEs ) 。 想法是将执政的 PDE 操作员分解成线性( 通过线性化提取的快速动态)和非线性(在删除前一个部分后剩下的部分),我们在这些部分上应用不连续的 Galerkin (DG) 空间分解。 由此形成的半分解系统随后使用指数化时间分解器集成: 对前者的精确度和后者的近似值。 通过构建, 我们的方法i) 与一个大 Courant 数字( Cr > 1) 相稳定; ii) 在时间和空间上支持高分级解决方案; iii) 与IMEX DG 方法相比,在计算上有利,没有先决条件;iv) 需要可比的计算时间与明确的 RKGDG 方法相比,同时将定序的幅度大于最大稳定时间级数; v) 在现代的大规模平行计算结构中,通过利用Krylov- 子空基质矩阵分解指数化指数化指数化指数化指数化指数化指数化指数化和压缩通信分解器变等压的预期的公式。