Unconditionally stable space–time isogeometric discretization for the wave equation in Hamiltonian formulation
Matteo Ferrari, Sara Fraschini, Gabriele Loli, Ilaria Perugia · 2024
Read Unconditionally stable space–time isogeometric discretization for the wave equation in Hamiltonian formulation by Matteo Ferrari, Sara Fraschini, Gabriele Loli, Ilaria Perugia (2024) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2411-00650.txt.
Description
Abstract We consider a family of conforming space–time discretizations for the wave equation based on a first-order-in-time formulation employing maximal regularity splines. In contrast with second-order-in-time formulations, which require a CFL condition to guarantee stability, the methods we consider here are unconditionally stable without the need for stabilization terms. Along the lines of the work by M. Ferrari and S.
Catalog record
| CREATOR | Matteo Ferrari, Sara Fraschini, Gabriele Loli, Ilaria Perugia |
|---|---|
| YEAR | 2024 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 151 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2411-00650
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Unconditionally stable space–time isogeometric discretization for the wave equation in Hamiltonian formulation Matteo Ferrari, Sara Fraschini, Gabriele Loli, Ilaria Perugia Abstract We consider a family of conforming space–time discretizations for the wave equation based on a first-order-in-time formulation employing maximal regularity splines. In contrast with second-order-in-time formulations, which require a CFL condition to guarantee stability, the methods we consider here are unconditionally stable without the need for stabilization terms. Along the lines of the work by M. Ferrari and S. Fraschini (2024), we address the stability analysis by studying the properties of the condition number of a family of matrices associated with the time discretization. Numerical tests validate the performance of the method. 1 Introduction We consider the following acoustic wave problem with Dirichlet boundary conditions: \begin{cases}\partial_{t}^{2}U(\boldsymbol{x},t)-\text{div}_{\boldsymbol{x}}(c^{2}(\boldsymbol{x})\nabla_{\boldsymbol{x}}U(\boldsymbol{x},t))=F(\boldsymbol{x},t)&(\boldsymbol{x},t)\in Q_{T}:=\Omega\times(0,T),\\ U(\boldsymbol{x},t)=g(\boldsymbol{x},t)&(\boldsymbol{x},t)\in\Sigma_{T}:=\Gamma\times[0,T],\\ U(\boldsymbol{x},0)=U_{0}(\boldsymbol{x}),\quad\partial_{t}U(\boldsymbol{x},t)_{|_{t=0}}=V_{0}(\boldsymbol{x})&\boldsymbol{x}\in\Omega,\end{cases} (1) where \Omega\subset\mathbb{R}^{d} ( d=1,2,3 ) is a bounded Lipschitz domain with boundary \Gamma:=\partial\Omega , T>0 is a finite time, F\in L^{2}(Q_{T}) is a given source term, c\in L^{\infty}(\Omega) is a positive wave velocity independent of t , and the given boundary and initial data satisfy g\in H^{\nicefrac{{1}}{{2}}}(\Sigma_{T}),\quad U_{0}\in H^{1}(\Omega),\quad V_{0}\in L^{2}(\Omega),\quad\text{~and~}\quad{U_{0}}=g{}_{|_{t=0}}\quad\text{on~$\Gamma$}. Here and throughout the paper we follow standard notation for differential operators, function spaces and norms that can be found, for example, in [ 8 ] . Various discretization techniques are available to compute an approximate numerical solution of problem ( 1 ). In contrast to the more standard approaches based on separate discretizations of space and time, space–time methods, introduced in the seminal papers [ 20 , 27 , 28 ] , provide a simultaneous discretization of space and time variables. Thanks to their features (e.g. high-order approximation in space and time, space–time unstructured meshes, and parallelization) and to advances in computer technology, the investigation of space–time methods has increased recently, leading to the development of several space–time discretizations for wave propagation problems. Among them, we mention space–time discontinuous Galerkin methods (see e.g. [ 3 , 4 , 5 , 14 , 32 , 33 , 34 ] ), and conforming space–time discretizations (see e.g. [ 2 , 26 , 6 , 23 , 36 , 30 ] ). Here, we focus on the latter class of methods. A second-order-in-time space–time variational formulation is considered in [ 36 , 37 ] . It is obtained by integrating by parts ( 1 ) both in time and space. Unique solvability of that formulation is proven (see e.g. [ 37 , Theorem 5.1] for c\equiv 1 and V_{0}\equiv 0 ), but an inf-sup stability in standard Sobolev norms in not satisfied (see [ 38 , Theorem 4.2.24] ). Therefore, stability of conforming space–time finite element discretizations may be achieved under suitable CFL conditions; see [ 17 ] for explicit and sharp results on this. In order to recover second-order-in-time unconditionally stable methods, various possibilities have been explored. One consists of testing with optimal test functions written in terms of suitable operators (e.g. the modified Hilbert transform [ 30 ] ). Another possibility is to stabilize the corresponding bilinear form by adding appropriate (non-consistent) penalty terms. The different choices of this stabilization depend only on the discretization of the temporal part. In [ 36 ] , a stabilization for continuous piecewise linear functions in time has been proposed and analyzed. That idea has been then generalized to higher order continuous piecewise polynomials in [ 39 ] , and to higher order maximal regularity splines in [ 19 , 17 ] . In this paper, we study a numerical method based on a first-order-in-time formulation of ( 1 ) obtained by introducing the auxiliary unknown V:=\partial_{t}U . Both U and V are discretized in the same discrete space, while test functions are taken in a space with lower polynomial degree and regularity. In [ 2 , 21 ] , such a space–time method was proposed, with discretization in time performed with continuous piecewise polynomials as trial functions, and discontinuous piecewise polynomials of one degree less as test functions. In [ 2 ] and [ 21 ] , via matricial and variational arguments, respectively, unconditional stability, as well as error estimates, have been proved and numerically verified for c\equiv 1 . Note that testing with discontinuous piecewise polynomial functions in time guarantees a time-stepping procedure. The temporal part of these schemes is actually equivalent to a Runge-Kutta Gauss-Legendre method, see [ 22 , Section 2] . In this work, we analyze a discretization of the first-order-in-time formulation with high order maximal regularity splines in time. We discretize both U and V in the same spline spaces. Test functions are taken in the spline space of one less polynomial degree and regularity. Although a complete well-posedness and error analysis is still out of reach at the moment, via a matricial argument along the lines of [ 17 ] , the resulting method is shown to be unconditionally stable without the need of additional stabilization terms. The analysis is focused on a system of ordinary differential equations, which is strongly related to the time part of the wave equation, and stability is derived by exploiting the algebraic structure of the matrices involved. This analysis combines two main tools: properties of the symbols [ 25 ] of the matrices associated with spline discretizations, and the behaviour of the condition number of general Toeplitz band matrices characterized in [ 1 ] . With the same techniques, we also show that a CFL condition is required when both test and trial spaces in time are splines of the same degree and maximal regularity. This CFL condition turns out to be sharp. The paper is structured as follows. In Section 2 , we introduce the discrete first-order-in-time variational formulation of ( 1 ), its discretization with maximal regularity splines in time, and we discuss the properties of the associated Galerkin matrices for the temporal part. Furthermore, we recall results on the conditioning of families of Toeplitz band matrices. In Section 3 , we present our main results on the stability of proposed method. In Section 4 , we consider the numerical scheme with equal trial and test spaces in time, and we explain mathematically why it is only conditionally stable. Finally, in Section 5 , we present an efficient algorithm for the solving the linear system involved, and various numerical tests on the full space–time problem with isogeometric discretization also in the space variables, which demonstrate the performance of the method and its unconditional stability. 2 First-order-in-time variational formulation and temporal discretization with maximal regularity splines With the auxiliary unknown V:=\partial_{t}U , problem ( 1 ) is reformulated as follows: \begin{cases}\partial_{t}U(\boldsymbol{x},t)-V(\boldsymbol{x},t)=0&(\boldsymbol{x},t)\in Q_{T},\\ \partial_{t}V(\boldsymbol{x},t)-\text{div}_{\boldsymbol{x}}(c^{2}(\boldsymbol{x})\nabla_{\boldsymbol{x}}U(\boldsymbol{x},t))=F(\boldsymbol{x},t)&(\boldsymbol{x},t)\in Q_{T},\\ U(\boldsymbol{x},t)=g(\boldsymbol{x},t)&(\boldsymbol{x},t)\in\Sigma_{T},\\ U(\boldsymbol{x},0)=U_{0}(\boldsymbol{x}),\quad V(\boldsymbol{x},0)=V_{0}(\boldsymbol{x})&\boldsymbol{x}\in\Omega.\end{cases} (2) From now, we restrict, for simplicity, to the case of U_{0}\equiv 0 , V_{0}\equiv 0 , and g\equiv 0 . The general case can be readily considered with a suitable lifting. For the spatial discretization, let us consider a discrete space V_{h_{\boldsymbol{x}}}(\Omega)\subset H_{0}^{1}(\Omega) depending on a spatial parameter h_{\boldsymbol{x}} (e.g. piecewise linear, continuous functions over a triangulation of \Omega of mesh size h_{\boldsymbol{x}} ). For the temporal discretization, we introduce the space S_{h_{t}}^{(p,k)}(0,T) of splines of polynomial degree p\geq 0 and regularity C^{k} , with k\geq-1 ( C^{-1} allowing for discontinuous functions) over a uniform mesh of [0,T] made of