A unified framework for the analysis, numerical approximation and model reduction of linear operator equations, Part I: Well-posedness in space and time
Moritz Feuerle, Richard Löscher, Olaf Steinbach, Karsten Urban · 2025
Read A unified framework for the analysis, numerical approximation and model reduction of linear operator equations, Part I: Well-posedness in space and time by Moritz Feuerle, Richard Löscher, Olaf Steinbach, Karsten Urban (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-2508-05407.txt.
Introduction
Moritz Feuerle and co-authors lay foundations for a unified equation framework in a 2025 paper. Large families of linear operator equations are covered at once. Elliptic, parabolic, and hyperbolic partial differential equations all fit inside.
Known weak formulations appear as special cases of the approach. New space-time forms for the wave equation are also derived. The core move completes and extends operators from the strong form. Well-posedness in space and time is the main result here. Later parts will handle numerical methods and model reduction. Readers get the theory base for one language across many equation types.
Catalog record
| CREATOR | Moritz Feuerle, Richard Löscher, Olaf Steinbach, Karsten Urban |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 115 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2508-05407
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
A unified framework for the analysis, numerical approximation and model reduction of linear operator equations, Part I: Well-posedness in space and time Moritz Feuerle, Richard Löscher, Olaf Steinbach, Karsten Urban Abstract We present a unified framework to construct well-posed formulations for large classes of linear operator equations including elliptic, parabolic and hyperbolic partial differential equations. This general approach incorporates known weak variational formulations as well as novel space-time variational forms of the hyperbolic wave equation. The main concept is completion and extension of operators starting from the strong form of the problem. This paper lays the theoretical foundation for a unified approach towards numerical approximation methods and also model reduction of parameterized linear operator equations which will be the subject of the following parts. 1 Introduction Linear operator equations can, e.g., be derived from partial differential and integral equations. Typical examples include elliptic, parabolic and hyperbolic second order partial differential equations (PDEs) as well as boundary integral equations. The analysis of well-posedness (existence, uniqueness and stability of solutions) as well as the construction and investigation of numerical approximation methods are usually done problem-specific and different for the above mentioned classes of problems. The aim of this paper is to provide a unified framework for the analysis, numerical approximation and model reduction of linear operator equations. In particular we aim at deriving a general approach which allows us to construct well-posed and optimally stable 1 1 1 In the sense that the norms of the operator and its inverse are equal to 1 . variational formulations for rather general linear operator equations. This is the scope of part I of this paper series. The forthcoming part will then concentrate on deriving corresponding discretizations and numerical approximations which will benefit from the general framework constructed in this first part. Finally, the close relation between the approximation error and the residual will then also allow us to construct a quite general approach towards model reduction of parameterized linear operator equations. In such a general framework we are able to cover stationary elliptic as well as time-dependent parabolic and hyperbolic problems. Elliptic problems are formulated in Sobolev and the time-dependent problems in Lebesgue-Bochner spaces using space-time variational formulations. The latter ones are well established for parabolic problems, but less is known for transport, wave and Schrödinger-type problems. The framework presented in this paper allows for well-posed variational formulations for all those equations. We shall detail the general approach to several examples, which will also include known weak and ultra-weak formulations. To the very best of our knowledge, this is a novel framework. It is, however, related to [ 5 ] , which focuses on ultra-weak formulations. We will comment on similarities and differences throughout the paper. Our approach also extends existing results in a unified manner, recovering also well-known formulations for elliptic and parabolic problems, while also allowing for well-posed formulations for hyperbolic problems. The remainder of this paper is organized as follows. After collecting some notation and guiding examples in the sequel of this section, Section 2 is devoted to the presentation of the theoretical foundation of our general framework. The main technical ingredient is completion and extension of operators to be presented in § 2.2 . We show applications to the Poisson problem, the heat equation and the wave equation in Section 3 . The paper ends with some conclusions and an outlook in Section 4 . 1.1 Notation and basic facts All spaces are assumed to be vector spaces over a common field \mathbb{F}\in\{\mathbb{R},\mathbb{C}\} . Thereby, for a normed vector space {X} , we denote by \lVert\cdot\rVert_{X} its norm and by {X}^{\prime} its dual space ( \mathbb{F}=\mathbb{R} ) or anti-dual space ( \mathbb{F}=\mathbb{C} ). If {X} is an inner product space, we denote its inner product by \lparen\cdot,\cdot\rparen_{X} . Hence, the inner product is a bilinear ( \mathbb{F}=\mathbb{R} ) or sesquilinear 2 2 2 linear in the first and anti-linear in the second argument form ( \mathbb{F}=\mathbb{C} ). Further, we denote by \langle\cdot,\cdot\rangle_{{X}^{\prime}\times{X}}:{X}^{\prime}\times{X}\rightarrow\mathbb{F} the duality pairing (or evaluation map), given by \langle f,x\rangle_{{X}^{\prime}\times{X}}\coloneqq f(x) for f\in{X}^{\prime} and x\in{X} , which is also a bilinear or sesquilinear form, respectively. For two Banach spaces {X},{Y} , we denote the space of all linear and bounded operators by \mathcal{L}({X},{Y}^{\prime})\coloneqq\{B:{X}\rightarrow{Y}^{\prime}\,:\,B\text{ is linear and }\lVert B\rVert_{\mathcal{L}({X},{Y}^{\prime})}<\infty\} , with \lVert B\rVert_{\mathcal{L}({X},{Y}^{\prime})}\coloneqq\sup\limits_{x\in{X}\setminus\{0\}}\frac{\lVert Bx\rVert_{{Y}^{\prime}}}{\lVert x\rVert_{X}}, and the space of all linear isomorphisms by \mathcal{L}_{\mathrm{is}}({X},{Y}^{\prime})\coloneqq\{B\in\mathcal{L}({X},{Y}^{\prime})\,:\,B\text{ bijective and }B^{-1}\in\mathcal{L}({Y}^{\prime},{X})\}. As usual, we call B\in\mathcal{L}({X},{Y}^{\prime}) isometric , if \lVert B\rVert_{\mathcal{L}({X},{Y}^{\prime})}=1 . Given B\in\mathcal{L}({X},{Y}^{\prime}) and f\in{Y}^{\prime} , seeking for an unknown x\in{X} , we call Bx=f in {Y}^{\prime} interpreted as \langle Bx,y\rangle_{{Y}^{\prime}\times{Y}}=\langle f,y\rangle_{{Y}^{\prime}\times{Y}} for all y\in{Y} , a linear operator equation . We call it well-posed (in the sense of Hadamard, [ 13 ] ) if B\in\mathcal{L}_{\mathrm{is}}({X},{Y}^{\prime}) and optimally stable if B is isometric. Further, for a Hilbert space {H} , we denote by R_{{H}}\in\mathcal{L}_{\mathrm{is}}({H},{H}^{\prime}) the isometric Riesz operator given by \langle R_{{H}}x,y\rangle_{{H}^{\prime}\times{H}}\coloneqq\lparen x,y\rparen_{H}\quad\forall x,y\in{H}. (1) In addition, it holds \lparen\cdot,\cdot\rparen_{{H}^{\prime}}=\langle\cdot,R_{{H}}^{-1}\cdot\rangle_{{H}^{\prime}\times{H}} , \lVert\cdot\rVert_{{H}}=\lVert R_{{H}}\cdot\rVert_{{H}^{\prime}} and \lVert\cdot\rVert_{{H}^{\prime}}=\lVert R_{{H}}^{-1}\cdot\rVert_{{H}} . For two normed vector spaces {X},{Y} with {X}\subseteq{Y} , we call {X} continuously embedded in {Y} (abbreviated “ {X}\hookrightarrow{Y} ” ) if there exists an embedding constant C<\infty , such that \lVert x\rVert_{Y}\leq C\lVert x\rVert_{X} for all x\in{X} . Further, we denote by {X}\subseteq_{d}{Y} , that {X} is dense in {Y} , and by {X}\hookrightarrow_{d}{Y} , that {X}\subseteq_{d}{Y} and {X}\hookrightarrow{Y} . We collect some well-known facts. Remark 1.1 . If {X}\subseteq_{d}{Y}\hookrightarrow_{d}Z , then {X}\subseteq_{d}Z and {X}\hookrightarrow Z if {X}\hookrightarrow{Y}\hookrightarrow Z . Remark 1.2 ( [ 3 , §5 Remark 3] ) . Let {X}\hookrightarrow_{d}{H} be Banach spaces with embedding constant C . (i) It holds {H}^{\prime}\hookrightarrow{X}^{\prime} with embedding constant C and {H}^{\prime}\hookrightarrow_{d}{X}^{\prime} , if {X} is reflexive. (ii) Let {H} be a Hilbert space. Identifying {H}\cong{H}^{\prime} using the Riesz operator, we get {X}\hookrightarrow_{d}{H}\cong{H}^{\prime}\hookrightarrow{X}^{\prime}, (2) as well as \lVert x\rVert_{H}\leq C\lVert x\rVert_{X} , x\in{X} , \lVert x\rVert_{{X}^{\prime}}\leq C\lVert x\rVert_{H} , x\in{H} and for the duality paring holds \displaystyle\langle f,x\rangle_{{X}^{\prime}\times{X}}=\lparen f,x\rparen_{H}\qquad\forall f\in{H},\ \forall x\in{X}. (3) Eq. ( 2 ) is usually called a Gelfand triple , denoted by ({X},{H},{X}^{\prime}) . Note: If {X} is a Hilbert space itself, there exists the natural Riesz isomorphism between {X} and {X}^{\prime} , but it is not viewed as the identity map, i.e., {X}\not\cong{X}^{\prime} . Instead, we use the Riesz isomorphism between {H} and {H}^{\prime} as the identity map to identify {H}\cong{H}^{\prime} . To make things clear, for x\in{X} , the dual norm is to be understood as \lVert x\rVert_{{X}^{\prime}}=\lVert R_{H}x\rVert_{{X}^{\prime}}=\sup\limits_{y\in X}\frac{\lparen x,y\rparen_{H}}{\lVert x\rVert_{X}} and not as \lVert R_{X}x\rVert_{{X}^{\prime}}=\sup\limits_{y\in X}\frac{\lparen x,y\rparen_{X}}{\lVert x\rVert_{X}} . Function spaces Let I\coloneqq(0,T) , 0<T<\infty be a time interval, \Omega\subset\mathbb{R}^{d} , d\in\mathbb{N} , be an open bounded spatial domain with Lipschitz boundary for d\geq 2 and let Q\coloneqq I\times\Omega denote the space-time domain. We denote by C^{k;\ell}(Q) , k,\ell\in\mathbb{N}_{0} , the set of all C(Q) functions where all partial derivatives w.r.t. t\in I up to order k and all partial derivatives w.r.t. x\in\Omega up to order \ell are in C(Q) , define C_{0}(\overline{\Omega})\coloneqq\{u\in C(\overline{\Omega}):u|_{\partial\Omega}=0\} and we denote the space of continuous functions with zero boundary in space and k^{\text{th}} -order zero initial conditions in time by C_{0,;0}^{(k)}(\mathmakebox[0pt]{\phantom{Q}\mkern 2.0mu\overline{\mkern-2.0mu\phantom{Q}}}Q)\coloneqq\{u\in C(I\times\overline{\Omega})\cap C^{k;0}([0,T)\times\Omega)\,:\,u|_{I\times\partial\Omega}=0,\ \partial^{\alpha}_{t}u(0)=0,\ \alpha=0,...,k\}. For function spaces in time on \bar{I} , we shall always denote homogeneous initial conditions