Recovering Initial States in Certain Quasilinear Parabolic Problems from Time Averages
Bogdan–Vasile Matioc, Christoph Walker · 2025
Read Recovering Initial States in Certain Quasilinear Parabolic Problems from Time Averages by Bogdan–Vasile Matioc, Christoph Walker (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-2510-21687.txt.
Introduction
Bogdan-Vasile Matioc and Christoph Walker solve a detective problem for heat-like equations. The task is to recover the starting state of a system from averages of its behavior over time. The equations are quasilinear parabolic. That means their diffusion strength depends on the state itself. The class covers chemical and biological spreading processes.
Under mild conditions on the equation's structure, they prove the starting state is uniquely fixed by small time averages. The growth condition on the reaction part near zero is part of those conditions. The observation window can be any time period at all. That turns a weak measurement into a full fingerprint of the start. Two families of examples show the result at work. They are chemotaxis models, where organisms chase chemical signals, and reaction-diffusion systems.
Catalog record
| CREATOR | Bogdan–Vasile Matioc, Christoph Walker |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 84 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2510-21687
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Recovering Initial States in Certain Quasilinear Parabolic Problems from Time Averages Bogdan–Vasile Matioc, Christoph Walker Abstract . The inverse problem of reconstructing the initial state in quasilinear parabolic equations from time averages is investigated. Under suitable regularity assumptions on the quasilinear structure and a superlinear growth condition near zero for the semilinear part, it is shown that the initial state can be uniquely recovered from small time averages taken over an arbitrary time period. The applicability of the result is demonstrated for certain chemotaxis models and reaction–diffusion systems. 2020 Mathematics Subject Classification 35K59, 47D99 1. Introduction and Main Result Recovering the past state of a system from given observations is a central challenge in the study of evolution equations. In this work, we focus our attention on quasilinear parabolic problems of the form u^{\prime}=A(u)u+f(u)\,,\quad t\in(0,T]\,, (1.1a) subject to \int_{0}^{T}\mathrm{w}(t)\,u(t)\,\mathrm{d}t=M\,. (1.1b) We assume that E_{1} and E_{0} are Banach spaces such that E_{1} is continuously and densely embedded into E_{0} and that A(u)\in\mathcal{H}(E_{1},E_{0}) . That is, for fixed u , the operator A(u) with domain E_{1} generates a strongly continuous analytic semigroup on E_{0} . Hereby, A(u) may depend nonlinearly upon u . Moreover, f=f(u) is a nonlinear function, \mathrm{w} is a nontrivial weight function, and M\in E_{1} . In ( 1.1 ), the usual initial condition is replaced by a nonlocal condition ( 1.1b ). Roughly speaking, we shall show that if A=A(u) and f=f(u) depend Lipschitz continuously upon u and f has a superlinear growth near zero, then one can recover the (unique) initial state u(0) from small values of time averages over arbitrary time periods. Hence, problem ( 1.1 ) is well-posed for arbitrary T>0 and small values of M . The smallness condition is due to the fact that T>0 is a priori fixed and one thus seeks for global solutions to a nonlinear problem. We shall later also consider related problems, where the nonlocal condition ( 1.1b ) is replaced by conditions of the form u(0)+\int_{0}^{T}\mathrm{w}(t)\,u(t)\,\mathrm{d}t=M (1.2) or (with a constant \mathrm{w}_{0} ) u(0)-\mathrm{w}_{0}u(T)=M\,, (1.3) which, however, are significantly easier to handle. Before presenting our results in detail, let us mention that the semilinear counterpart of ( 1.1 ) (i.e., when A(u) is independent of u ) was recently investigated in [ 29 ] . As we shall see herein, however, the quasilinear setting turns out to be substantially more involved. The linear counterpart of problem ( 1.1 ) (also subject to the other nonlocal conditions ( 1.2 )-( 1.3 ) ) was considered in [ 16 ] in a Hilbert space setting. Further results on solvability, uniqueness, and regularity of solutions to linear parabolic problems with conditions of type ( 1.2 )–( 1.3 ) were obtained in [ 11 , 12 , 13 , 14 , 24 , 27 ] (see also the references therein). The connection to stochastic differential equations is explored in [ 12 , 13 , 15 ] . Moreover, both linear and nonlinear problems with alternative, and in some cases more general, nonlocal conditions have been studied in a number of works [ 7 , 6 , 18 , 19 , 20 , 21 , 32 , 31 ] , employing a wide range of analytical techniques in concrete functional settings [ 6 , 21 , 24 , 31 ] as well as in abstract frameworks [ 7 , 18 , 19 , 20 , 32 ] ; see also the survey [ 26 ] . These references are by no means exhaustive. It is worth noting, however, that only a few contributions explicitly deal with genuinely quasilinear structures (but see [ 18 , 31 ] ). In order to explain the difficulties associated with ( 1.1 ), we proceed formally and assume the existence of a (Hölder continuous in time) solution u such that the mapping t\mapsto A(u(t)) generates a parabolic evolution operator U_{A(u)}(t,s) , 0\leq s\leq t\leq T , on E_{0} with regularity subspace E_{1} in the sense of [ 5 ] . Then, u solving ( 1.1a ) can be written in form of the variation-of-constants formula u(t)=U_{A(u)}(t,0)u(0)+\int_{0}^{t}U_{A(u)}(t,s)f(u(s))\,\mathrm{d}s\,,\quad t\in[0,T]\,, and, when plugging this into ( 1.1b ), we obtain an equivalent reformulation of ( 1.1 ) as a fixed point problem u(t)=U_{A(u)}(t,0)\Xi(u)+\int_{0}^{t}U_{A(u)}(t,s)f(u(s))\,\mathrm{d}s\,,\quad t\in[0,T]\,, (1.4a) where u(0)=\Xi(u) is given by \Xi(u)=\Phi_{A(u)}^{-1}\big(M-\Psi(u)f(u)\big) (1.4b) with \Phi_{A(u)}:=\int_{0}^{T}\mathrm{w}(t)\,U_{A(u)}(t,0)\,\mathrm{d}t (1.5) and \Psi(u)g:=\int_{0}^{T}\mathrm{w}(t)\int_{0}^{t}U_{A(u)}(t,s)g(s)\,\mathrm{d}s\,\mathrm{d}t\,. (1.6) Clearly, solving the fixed point problem ( 1.4 ) requires on the one hand the existence of the evolution operator U_{A(u)} and, on the other hand, a careful analysis of the operator \Phi_{A(u)} that encodes the initial condition in ( 1.4 ), in particular concerning its invertibility and its dependence on u . It turns out that, under rather mild assumptions, \Phi_{A(u)} is an isomorphism between E_{0} and E_{1} , so that the inverse operator \Phi_{A(u)}^{-1}\in\mathcal{L}(E_{1},E_{0}) is of the same order as A(u) . Hence, one can only expect that the initial value u(0)=\Xi(u) in ( 1.4 ) lies in E_{0} , without any additional a priori regularity. It is well-known, however, that an initial value in the phase space E_{0} is usually insufficient to guarantee the well-posedness of quasilinear problems, e.g. see [ 4 , 2 , 28 ] . Furthermore, it is necessary that M\in E_{1} . In view of these restrictions, and in order to retain greater flexibility for later applications, it is natural to reformulate the problem in spaces of higher regularity within an interpolation–extrapolation scale (see below). Moreover, the use of time-weighted spaces enables us to handle the singular behavior of the evolution operator U_{A(u)}(t,s) in this scale at t=s and to accommodate a sufficiently broad class of semilinearities. A further crucial step in solving the fixed point problem ( 1.4 ) is then establishing the (Lipschitz) continuity of the mapping u\mapsto\Phi_{A(u)}^{-1} in the operator norm of \mathcal{L}(E_{1},E_{0}) , which will require some effort. Main Result To be more precise, certain technical assumptions are necessary. For this purpose, we assume for the (real or complex) Banach spaces E_{1} and E_{0} that \text{$E_{1}$ is compactly and densely embedded in $E_{0}$}\,. (1.7a) Further, let A(0)\in\mathcal{H}(E_{1},E_{0})\,, (1.7b) i.e. A(0)\in\mathcal{L}(E_{1},E_{0}) is the generator of a strongly continuous analytic semigroup (e^{tA(0)})_{t\geq 0} on E_{0} . Define then E_{2}:=\big(\mathrm{dom}(A(0)^{2}),\|(\omega-A(0))^{2}\cdot\|_{E_{0}}\big) for some fixed \omega\in\rho(A(0)) (i.e. in the resolvent set). We fix for each \theta\in(0,1) an exact admissible interpolation functor (\cdot,\cdot)_{\theta} of exponent \theta and introduce the Banach spaces 1 1 1 That is, [(E_{\theta},\tilde{A}_{\theta});0\leq\theta\leq 2] is (part of) the interpolation-extrapolation scale of order 0 generated by (E_{0},\tilde{A}) and (\cdot,\cdot)_{\theta} in the sense of [ 5 , V.Theorem 1.5.1] , where \tilde{A}:=\omega-A(0) . Then E_{\theta}\hookrightarrow E_{\vartheta} is a compact and dense embedding for 0\leq\vartheta\leq\theta\leq 2 . E_{\theta}:=(E_{0},E_{1})_{\theta}\,,\qquad E_{1+\theta}:=(E_{1},E_{2})_{\theta}\,. Fix 0\leq\beta<\alpha<\alpha_{0}\leq 1 (1.7c) and assume that there is r_{0}>0 such that \displaystyle A \displaystyle\in C^{1-}\big(\mathbb{B}_{E_{\beta}}(0,r_{0}),\mathcal{L}(E_{1},E_{0})\big)\cap C^{2-}\big(\mathbb{B}_{E_{\beta}}(0,r_{0}),\mathcal{L}(E_{1+\alpha},E_{\alpha})\big) (1.7d) \displaystyle\cap C^{1-}\big(\mathbb{B}_{E_{\beta}}(0,r_{0}),\mathcal{L}(E_{1+\alpha_{0}},E_{\alpha_{0}})\big)\,, where C^{1-} and C^{2-} refer to (local) Lipschitz continuity of a function and, respectively, its derivative. We also require for the scalar weight function that \mathrm{w}\in C^{1}([0,T])\,,\qquad\mathrm{w}(0)\not=0\,. (1.7e) As we shall later prove in Proposition 2.2 , assumptions ( 1.7b ) and ( 1.7e ) imply that \Phi_{A(0)}=\int_{0}^{T}\mathrm{w}(t)e^{tA(0)}\,\mathrm{d}t\in\mathcal{L}(E_{0},E_{1})\,. We then further suppose that \mathrm{ker}(\Phi_{A(0)})=\{0\} (1.7f) and thus obtain (again from Proposition 2.2 below) that \Phi_{A(0)}\in\mathcal{L}is(E_{0},E_{1}) . As for the semilinear part we fix \ell>0\,,\qquad\alpha_{0}<\gamma\leq 1\,,\qquad 0<\xi<\min\left\{\alpha+\frac{1}{\ell+1},1\right\}\,,