A note on the space-time variational formulation for the wave equation with source term in L 2 ​ ( Q ) L^{2}(Q)

Marco Zank · 2025

Read A note on the space-time variational formulation for the wave equation with source term in L 2 ​ ( Q ) L^{2}(Q) by Marco Zank (2025) 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-2512-23807.txt.

Introduction

Marco Zank writes a short mathematical note on the wave equation, setting up its variational formulation in space-time all at once rather than stepping through time. He treats the scalar wave equation in its second-order form on bounded Lipschitz domains, with zero initial conditions and a source term that is merely square-integrable over the space-time cylinder. Variational methods need the right pairing of solution and test spaces to work.

Zank investigates a framework pairing a new solution space with the plain square-integrable test space. Uses known existence and uniqueness results in a Sobolev space to prove the setting fits inf-sup theory, with the solution operator an isomorphism. He also shows the new solution space is not contained in the second-order Sobolev space. The result underpins space-time discretizations, least-squares methods, and space-time boundary element methods.

Catalog record

CREATORMarco Zank
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE19 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2512-23807

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

A note on the space-time variational formulation for the wave equation with source term in L 2 ​ ( Q ) L^{2}(Q) Marco Zank Abstract We derive a variational formulation for the scalar wave equation in the second-order formulation on bounded Lipschitz domains and homogeneous initial conditions. We investigate a variational framework in a bounded space-time cylinder Q Q with a new solution space and the test space L 2 ​ ( Q ) L^{2}(Q) for source terms in L 2 ​ ( Q ) L^{2}(Q) . Using existence and uniqueness results in H 1 ​ ( Q ) H^{1}(Q) , we prove that this variational setting fits the inf-sup theory, including an isomorphism as solution operator. Moreover, we show that the new solution space is not a subspace of H 2 ​ ( Q ) H^{2}(Q) . This new uniqueness and solvability result is not only crucial for discretizations using space-time methods, including least-squares approaches, but also important for regularity results and the analysis of related space-time boundary integral equations, which form the basis for space-time boundary element methods. Keywords: wave equation ⋅ \cdot space-time method ⋅ \cdot variational formulation ⋅ \cdot inf-sup conditions 1 Introduction The solutions to hyperbolic evolution problems are essential in many physical areas. As these solutions are often accessible only as approximations, different numerical schemes and their analysis are available. The most popular approaches are based on time-stepping schemes and spatial discretizations, e.g., finite element methods or finite difference methods. As time-stepping approaches are of pointwise-in-time-type, these methods are not based on a variational formulation in both, the spatial variable x and the temporal variable t . In recent years, approaches using also a variational setting with respect to t , have been developed, which we call space-time methods, see [ 16 ] for temporal first-order and second-order formulations. Not only for these methods but also in general, the underlying space-time functional framework used for hyperbolic evolution equations is crucial for, e.g., regularity results or the analysis of related boundary integral equations or any conforming discretization method. The model problem is the scalar wave equation in second-order formulation \left.\begin{array}[]{rcll}c^{-2}\partial_{tt}u-\Delta_{x}u&=&f&\mbox{in}\;Q=\Omega\times(0,T),\\[2.84526pt] u&=&0&\mbox{on}\;\Sigma=\partial\Omega\times[0,T],\\[2.84526pt] \partial_{t}u(\cdot,0)=u(\cdot,0)&=&0&\mbox{on}\;\Omega\end{array}\right\} (1) with a given source term f and the lateral boundary \Sigma=\partial\Omega\times[0,T] of the space-time cylinder Q=\Omega\times(0,T) , where \Omega\subset\mathbb{R}^{d} , d\in\{1,2,3\} , is a bounded Lipschitz domain, the constant wave speed c is set to 1 due to rescaling and T>0 . Many space-time approaches introduce additional unknowns, which represent first-order derivatives, see [ 16 , 3 , 5 , 6 , 8 ] . Other space-time variational settings in [ 1 ] require a star-shaped spatial domain \Omega with additional assumptions on the boundary, and besides this, the analysis of the ultraweak variational formulations of [ 7 ] are strongly based on isomorphisms between function spaces, which are pointwise in time. Alternatively, we recall the variational setting in H^{1}(Q) , analyzed in [ 9 , 12 , 16 ] , by introducing the space-time Sobolev spaces H^{1,1}_{0;0,}(Q) , satisfying the homogeneous initial condition for t=0 and the homogeneous Dirichlet boundary condition on \Sigma , and H^{1,1}_{0;,0}(Q) , fulfilling the homogeneous terminal condition for t=T and the homogeneous Dirichlet boundary condition on \Sigma , see Section 2 . With these spaces, we set the continuous bilinear form a_{H^{1}}(\cdot,\cdot)\colon\,H^{1,1}_{0;0,}(Q)\times H^{1,1}_{0;,0}(Q)\to\mathbb{R} by a_{H^{1}}(v,w)=\int_{0}^{T}\int_{\Omega}\left(-\partial_{t}v(x,t)\partial_{t}w(x,t)+\nabla_{x}v(x,t)\cdot\nabla_{x}w(x,t)\right)\mathrm{d}x\mathrm{d}t. (2) Then, the space-time variational formulation of finding u\in H^{1,1}_{0;0,}(Q) such that \forall w\in H^{1,1}_{0;,0}(Q):\quad a_{H^{1}}(u,w)=\int_{0}^{T}\int_{\Omega}f(x,t)w(x,t)\mathrm{d}x\mathrm{d}t (3) has a unique solution for each source term f\in L^{2}(Q) , fulfilling the stability estimate \sqrt{\|\partial_{t}u\|_{L^{2}(Q)}^{2}+\|\nabla_{x}u\|_{L^{2}(Q)}^{2}}\leq\frac{1}{\sqrt{2}}T\|f\|_{L^{2}(Q)} (4) and the regularity u\in C([0,T];H^{1}_{0}(\Omega)) and \partial_{t}u\in C([0,T];L^{2}(\Omega)). The variational formulation ( 3 ) is successfully used in the discretization schemes [ 4 , 10 , 12 , 14 , 15 , 16 ] . However, a natural question is whether the bilinear form a_{H^{1}}(\cdot,\cdot) in ( 2 ) fulfills the inf-sup theorem [ 2 , Thm. 25.9] . It turns out that the solution operator \mathcal{L}_{H^{1}}\colon\,L^{2}(Q)\to H^{1,1}_{0;0,}(Q) to the space-time variational formulation ( 3 ), defined by \mathcal{L}_{H^{1}}f=u , is not surjective. Therefore, \mathcal{L}_{H^{1}} is not an isomorphism and the bilinear form a_{H^{1}}(\cdot,\cdot) in ( 2 ) does not fit to the inf-sup theorem [ 2 , Thm. 25.9] . We summarize this result in the following theorem. Theorem 1 . [ 13 , Thm. 1.1] There does not exist a constant C>0 such that each source term f\in L^{2}(Q) and the corresponding solution u\in H^{1,1}_{0;0,}(Q) of ( 3 ) satisfy \|u\|_{H^{1,1}_{0;0,}(Q)}\,\leq\,C\,\|f\|_{[H^{1,1}_{0;\,,0}(Q)]^{\prime}}. In particular, the inf-sup condition \forall v\in H^{1,1}_{0;0,}(Q):\quad c_{S}\,\|v\|_{H^{1,1}_{0;0,}(Q)}\,\leq\,\sup_{0\neq w\in H^{1,1}_{0;\,,0}(Q)}\frac{|a_{H^{1}}(v,w)|}{\|w\|_{H^{1,1}_{0;,0}(Q)}} with a constant c_{S}>0 does not hold true. One possibility to derive a variational setting, which satisfies the inf-sup theorem [ 2 , Thm. 25.9] , is to enlarge the solution space H^{1,1}_{0;0,}(Q) of the variational formulation ( 3 ) to a Hilbert space \mathcal{H}_{0;0,}(Q)\supset H^{1,1}_{0;0,}(Q) and to enlarge also the space of the source term f to [H^{1,1}_{0;,0}(Q)]^{\prime} , leading to an isomorphism \mathcal{L}_{\mathcal{H}_{0;0,}}\colon\,[H^{1,1}_{0;,0}(Q)]^{\prime}\to\mathcal{H}_{0;0,}(Q) as solution operator, which is the main idea of [ 13 ] . Another possibility is to shrink the solution space H^{1,1}_{0;0,}(Q) of the variational formulation ( 3 ) to a Hilbert space H^{1}_{0;0,}(Q;\square)\subset H^{1,1}_{0;0,}(Q) , resulting in an isomorphism \mathcal{L}_{H^{1}_{0;0,}(Q;\square)}\colon\,L^{2}(Q)\to H^{1}_{0;0,}(Q;\square) as solution operator. The latter approach is the content of this work. We introduce the new solution space H^{1}_{0;0,}(Q;\square)\subset H^{1,1}_{0;0,}(Q) and a related bilinear form a(\cdot,\cdot)\colon\,H^{1}_{0;0,}(Q;\square)\times L^{2}(Q)\to\mathbb{R} . Using this, we prove that a(\cdot,\cdot) fits to the framework of the inf-sup theorem [ 2 , Thm. 25.9] , which includes the unique solvability of the related space-time variational setting to the wave equation ( 1 ) for each source term f\in L^{2}(Q) . Further, we prove that the new solution space H^{1}_{0;0,}(Q;\square) is not a subspace of H^{2}(0,T;L^{2}(\Omega)) . The rest of the paper is organized as follows: In Section 2 , we summarize notations of function spaces. Section 3 is the main part of this work, where we introduce the new space-time solution space and the new space-time variational formulation. We end this section by proving unique solvability with the help of the inf-sup theorem [ 2 , Thm. 25.9] . In Section 4 , we give some conclusions. 2 Functions spaces In this section, we introduce the used function spaces. For a bounded Lipschitz domain A\subset\mathbb{R}^{n} , n\in\mathbb{N} , we denote by C_{0}^{\infty}(A) the set of all real-valued functions with compact support in A , which are infinitely differentiable. Further, \mathcal{D}(A) is the space of test functions when the set C_{0}^{\infty}(A) is endowed with the standard locally convex topology. Further, we denote by L^{2}(A) the usual Lebesgue space of real-valued functions with inner product \langle\cdot,\cdot\rangle_{L^{2}(A)} and induced norm \|\cdot\|_{L^{2}(A)} . Similarly, we use the classical Sobolev spaces H^{k}(A) , k\in\mathbb{N} , and the subspace H^{1}_{0}(A) , satisfying homogeneous Dirichlet conditions, with dual space [H^{1}_{0}(A)]^{\prime} . We extend these spaces to vector-valued versions, e.g., L^{2}(A)^{m} for m\in\mathbb{N} . Further, for A=(0,T) , we consider the spaces H^{1}_{0,}(0,T) , satisfying the initial condition v(0)=0 , and H^{1}_{,0}(0,T) , satisfying the terminal condition v(T)=0 . Again, for A=(0,T) , we introduce the extensions to Bochner spaces L^{2}(0,T;X) and H^{k}(0,T;X) , k\in\mathbb{N} , with a real Hilbert space X . Analogously, we use the space C([0,T];X) of continuous functions. Moreover, for A=Q , we consider the Lebesgue space

Read the complete text