Reconstruction of space-dependence and nonlinearity of a reaction term in a subdiffusion equation

Barbara Kaltenbacher, William Rundell · 2025

Read Reconstruction of space-dependence and nonlinearity of a reaction term in a subdiffusion equation by Barbara Kaltenbacher, William Rundell (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-2502-21243.txt.

Introduction

Barbara Kaltenbacher and William Rundell tackle, in a 2025 math paper, a hard reconstruction problem in slow-spreading systems. Their equation models reaction and subdiffusion, where spread crawls slower than usual. Two hidden traits must be found at once: how the reaction term bends, and how it varies across space. The two are multiplied together, which tangles any attempt to split them.

The pair test several data setups: readings along the edge over time, a snapshot of the inside at the final time, and a mix of both. They build fixed-point schemes and a frozen Newton method to do the rebuild. For final-time data, they prove the scheme converges and that the answer is unique. Computer tests show how the methods perform as the spread order changes.

Catalog record

CREATORBarbara Kaltenbacher, William Rundell
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE63 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2502-21243

Browse the full library · About the Tower of Babel

Full text

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

Reconstruction of space-dependence and nonlinearity of a reaction term in a subdiffusion equation Barbara Kaltenbacher, William Rundell Abstract In this paper we study the simultaneous reconstruction of two coefficients in a reaction-subdiffusion equation, namely a nonlinearity and a space dependent factor. The fact that these are coupled in a multiplicative matter makes the reconstruction particularly challenging. Several situations of overposed data are considered: boundary observations over a time interval, interior observations at final time, as well as a combination thereof. We devise fixed point schemes and also describe application of a frozen Newton method. In the final time data case we prove convergence of the fixed point scheme as well as uniqueness of both coefficients. Numerical experiments illustrate performance of the reconstruction methods, in particular dependence on the differentiation order in the subdiffusion equation. 1 Introduction In this paper, we study reconstruction of both q and f in a reaction-subdiffusion equation with the specific form \frac{\partial u}{\partial t}-\triangle u+q(x)f(u)=0 (1) defined in a cylindrical domain \Omega\times(0,T) where \Omega is a closed bounded domain in \mathbb{R}^{d} with smooth boundary \partial\Omega . Here q(x) is a spatially-dependent coefficient and the reaction term f(u) depends only on the solution u . Such equations are ubiquitious in applications throughout the biological and physical sciences. For example, in population dynamics the famous Verhulst ordinary differential equation model sets f(u)=au-bu^{2} to separate out the birth and death processes. This was extended to spatial regions and to include diffusion by Skellum in 1951, [ 28 ] and was one of the first models to incorporate random walk processes into the subject. The model therefore parallels the Fisher model for an advantageous gene in a population introduced slightly earlier [ 6 , 7 ] . Indeed, the general model ( 1 ) with q constant and f of the form f(u)=a(u-u^{2}) is now commonly referred to as the Fisher equation. There is a vast literature in this application context for the equation and allowing a spatial dependence on the coefficients is clearly of importance. The generalisation to include diffusion and hence spatial dependence, adds considerabe mathematical complexity including the possibility of travelling wave solutions of the above equation. See, for example, [ 20 , 1 ] . The form of the right hand side term here balances growth and decay with different power-law terms and clearly there are many possible modifications from a modelling standpoint. These might include different rates by using exponents other than linear and quadratic terms as well as higher order and fractional powers involving u , or indeed a functional form that goes beyond just power law behaviour. In perhaps the other most well-known application, that of chemical reactions, a similar form is used, [ 9 , 21 ] . A cubic polynomial for f in terms of u appears in the Rayleigh-Bérnard model of convection and in the FitzHugh–Nagumo model of 1961 modelling impulses along a nerve, [ 8 ] and also in the Zeldovich–Frank–Kamenetsky–Equation. More general models of the ZFK type involve exponential functions: for example, f(u)=\alpha\,u(1-u)e^{-\beta(1-u)} , [ 30 ] . Thus there is an application-driven reason to go beyond merely seeking f(u) as a combination of powers of u . Thus, although these models are based on specific polynomial assumptions there have been numerous modifications including assuming more general polynomials and with possibly non-integer powers. Our paradigm here allows us to recover the form of the reaction term from measured data; and, in fact, in addition to allow for a coupled inhomogeneous spatial environment that itself has to recovered. We note that this type of coupling is a major factor in the difficulty of determining both q and f . Undetermined coefficient problems for parabolic equations have a history going back more than 40 years. For the linear equation u_{t}-\triangle u+q(x)u=0 , Pierce showed a uniqueness result for recovering q(x) in one spatial variable from time trace data u(x_{0},t) with x_{0} on the boundary [ 22 ] . He also indicated a reconstruction method. For the case of final time data u(x,T) , Rundell proved unique identifiability for q and showed it could be determined by an iterative process from a fixed point scheme, [ 26 , 27 ] . For the case where q(x)=1 and the nonlinear term f was unknown Pilant and Rundell in a series of papers [ 25 , 23 , 24 ] , showed uniqueness and gave reconstruction methods for f(u) . See also the book by Isakov, [ 10 ] . More recently, due to its relevance in modeling, time-fractional versions of these problems have been considered. Also in this paper we will include the subdiffusion case where our model extends the differential equation for 0<\alpha\leq 1 by \partial_{t}^{\alpha}u-\Delta u+q(x)f(u)=r (2) The literature in this direction and from an inverse problems standpoint is growing rapidly and we reference the book [ 18 ] as a starting point. We here only give a small selection of related examples. The most closely related ones are identification of the spatially dependent potential q(x) in ( 2 ) for given f(u) in [ 11 , 13 ] or the determination of f(u) in ( 2 ) for given q(x) in [ 13 ] . Uniqueness and reconstructibility results for the pair \{a(x),f(u)\} in u_{t}^{\alpha}-\bigl(a(x)u_{x})_{x}-f(u) can be found in [ 17 ] . From the above it can be seen that our interest here lies in the inverse problem of recovering both the spatial term q(x) and the specific form of the nonlinearity f(u) . This is clearly an essential problem as while the general form of the model might be prescribed in an application the individual components are likely to be unknown and must be determined from additional (overposed) data measurements. The additional difficulty over previous combinations noted above is the fact that the two unknowns are directly coupled as a product as distinct from an unknown diffusion coefficient a(x) and reaction f(u) . Also the type of overposed data is highly relevant for reconstructibility of q(x) and f(u) . We here study on one hand measurements inside \Omega at a fixed time instance T , on the other hand, boundary observations over a time interval (0,T) , as well as a combination thereof. The plan of the paper is as follows. In section 1.1 we formulate the inverse problem in the three above mentioned observation settings. Section 2 is devoted to an analysis of the forward problem and a proof of well-definedness and Lipschitz continuous Fréchet differentiability of the operator mapping (q,f) to the solution of ( 2 ) with initial and boundary conditions. In section 3 we develop some iterative reconstruction methods for each of these scenarios. Some of them result from fixed point schemes that are based on a projection of the PDE on the observation manifold; for comparison, we also consider Newton type schemes. In the final time data case with \alpha=1 and \Omega\subseteq\mathbb{R}^{1} we provide a convergence result of the fixed point iteration and prove uniqeness in section 5 . 1.1 The inverse problem Our aim is to identify q(x) and f(u) in \displaystyle\partial_{t}^{\alpha}u-\triangle u+q(x)f(u) \displaystyle=r\quad \displaystyle\text{ in }\Omega\times(0,T) (3) \displaystyle Bu \displaystyle=0 \displaystyle\text{ on }\partial\Omega\times(0,T) \displaystyle u(t=0) \displaystyle=u_{0} \displaystyle\text{ in }\Omega from

Read the complete text