Reconstruction of a potential parameter in subdiffusion via a Kohn–Vogelius type functional: Theory and computation
Hamza Kahlaoui, Mourad Hrizi, Abdessamad Oulmelk, Xiangcheng Zheng, Mahmoud A. Zaky, Ahmed Hendy · 2025
Read Reconstruction of a potential parameter in subdiffusion via a Kohn–Vogelius type functional: Theory and computation by Hamza Kahlaoui, Mourad Hrizi, Abdessamad Oulmelk, Xiangcheng Zheng, Mahmoud A. Zaky, Ahmed Hendy (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-2509-19260.txt.
Introduction
Hamza Kahlaoui and five co-authors recover a hidden property of a material from measurements at its boundary. The setting is subdiffusion, where particles spread slower than normal diffusion predicts, and the unknown is a potential that varies through space. Their method minimizes a Kohn-Vogelius cost function, which compares the responses to two different excitations. Where the two responses disagree, the guessed potential must be wrong.
They prove the minimization problem is well-posed, with a unique best potential that depends stably on noisy data. They also show the cost function is smooth enough for a conjugate gradient algorithm, and prove that the algorithm converges. Numerical tests in one and two dimensions, including noisy cases, confirm the recovery works. Inverse problems of this kind matter wherever interiors cannot be probed directly, from imaging to materials testing.
Catalog record
| CREATOR | Hamza Kahlaoui, Mourad Hrizi, Abdessamad Oulmelk, Xiangcheng Zheng, Mahmoud A. Zaky, Ahmed Hendy |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 150 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2509-19260
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Reconstruction of a potential parameter in subdiffusion via a Kohn–Vogelius type functional: Theory and computation Hamza Kahlaoui, Mourad Hrizi, Abdessamad Oulmelk, Xiangcheng Zheng, Mahmoud A. Zaky, Ahmed Hendy Abstract . This work considers the identification of spatial-dependent potential from boundary observations in subdiffusion using a stable and robust recovery method. Specifically, we develop an algorithm to minimize the Kohn-Vogelius cost function, which measures the difference between the solutions of two excitations. The inverse potential problem is recast into an optimization problem, where the objective is to minimize a Kohn-Vogelius-type functional within a set of admissible potentials. We establish the well-posedness of this optimization problem by proving the existence and uniqueness of a minimizer and demonstrating its stability under perturbations in the boundary data. Furthermore, we analyze the Fréchet differentiability of the Kohn-Vogelius functional and prove the Lipschitz continuity of its gradient. These theoretical results enable the development of a convergent conjugate gradient algorithm for numerical reconstruction. Numerical examples in one and two dimensions, including scenarios with noisy data, validate the effectiveness and robustness of the proposed method. 1. Introduction The reconstruction of coefficients in normal or anomalous diffusion processes has a wide history in the literature of inverse problems from theoretical and numerical perspectives. The reconstruction of time dependent parameters in subdiffusion problems has been extensively investigated in literature [ 26 , 27 , 36 ] . In particular, some studies have investigated the reconstruction of space-dependent potentials in subdiffusion inferred from lateral Cauchy data [ 34 , 37 , 45 , 69 ] . In this paper, we aim to address the inverse problem of reconstructing a space-dependent potential q^{*} that arises in subdiffusion, utilizing lateral Cauchy data. Our primary objective is to successfully reconstruct q^{*} using a stable numerical method. In most of recent studies, the numerical algorithms developed for identifying the potential q^{*} rely on a least-squares approach. However, this paper focuses on the mathematical analysis of an algorithm aimed at minimizing the so-called Kohn-Vogelius cost function with a Tikhonov regularization term that penalizes the potential to be reconstructed. The Kohn–Vogelius cost functional quantifies the discrepancy between two auxiliary boundary value problems: one driven by boundary measurements and the other by boundary excitations. In the noise-free setting, this functional attains its minimum value of zero, and the corresponding minimizer coincides with the exact solution of the inverse problem. When the data are noisy or incompatible, the inverse problem itself may not admit a solution, yet the minimization problem remains well-posed and solvable. A key advantage of Kohn–Vogelius functionals lies in their robustness: they yield stable reconstructions even for ill-posed problems, owing to their definition over the entire domain, in contrast to classical least-squares functionals that are restricted to the boundary. This robustness has been confirmed in several studies, including those by Afraites et al. [ 4 , 5 ] and Chaabane et al. [ 17 ] . For completeness, we briefly recall the historical development of Kohn–Vogelius type functionals. The foundational idea dates back to Wexler, Fry, and Neuman [ 72 ] , who in 1985 proposed a method for detecting unknown impedance from boundary measurements. In 1987, Kohn and Vogelius [ 50 ] refined this approach by introducing a novel misfit gap-cost functional in an alternating-choice framework. Later, in 1990, Kohn and McKenney [ 49 ] implemented this formulation numerically for inverse conductivity problems. Since then, the method has been extended to a wide range of inverse problems [ 1 , 2 , 3 , 8 , 12 , 15 , 23 , 28 , 38 , 62 , 65 , 67 ] , further establishing its versatility and effectiveness. Recovering a potential in subdiffusion through the inverse problem has been extensively studied. In the one-dimensional setting, Rundell and Yamamoto [ 68 ] established the uniqueness of a non-negative potential q^{*}\in C[0,1] using Gelfand-Levitan theory. They further proposed a Newton-based iterative reconstruction procedure and provided an empirical analysis of the singular value spectrum of the linearized forward map, underscoring the problem’s severely ill-posed nature. This analysis was later extended to a broader class of solutions within an appropriate Sobolev space in time [ 69 ] . Beyond recovering a single parameter, Jing and Yamamoto [ 37 ] demonstrated the simultaneous identifiability of multiple parameters—including the fractional order, initial value, spatial potential, and Robin coefficients—from lateral Cauchy data at both endpoints in a one-dimensional subdiffusion model with homogeneous boundary conditions and no source term. A significant advancement was made by Jin and Zhou [ 34 ] , who achieved the simultaneous reconstruction of both the fractional order and the space-dependent potential q^{*}\in\mathcal{Q} using only endpoint data. Remarkably, their results hold even with only partial knowledge of the initial data or source, requiring only a mild constraint on the Neumann boundary condition for uniqueness. More recently, the focus has shifted to geometric inverse problems. In [ 40 ] , the authors developed a one-shot algorithm based on topological derivatives to identify the location and shape of a potential q^{*}=\chi_{\omega^{*}} supported on an unknown domain \omega^{*}\subset\Omega from partial boundary measurements. For a comprehensive overview of related work, including the recovery of time- or space-dependent potentials from other data types such as terminal observations, over-posed final time data, or average flux measurements, we refer the reader to [ 33 , 41 , 46 , 58 , 62 , 75 , 76 ] . On the numerical front, various approaches have been explored. For instance, Oulmelk et al. [ 64 ] investigated the use of Artificial Neural Networks to solve the inverse potential problem for the standard diffusion equation, offering insights applicable to the fractional case. Finally, it is worth noting that for the classical parabolic case ( \alpha=1 ), this inverse problem has been thoroughly studied by numerous authors employing a wide array of methods [ 6 , 9 , 10 , 13 , 14 , 29 , 48 , 60 , 71 ] . In the present work, building on the Kohn–Vogelius formulation, we recast the inverse reconstruction problem as a regularized optimization problem. Specifically, the unknown potential q^{*} is identified as the minimizer of a Kohn–Vogelius type cost function augmented with a Tikhonov regularization term. Our main contribution is to establish existence and uniqueness results for the Kohn–Vogelius formulation with Tikhonov regularization in the subdiffusion setting, and to propose a convergent conjugate gradient method (CGM) specifically adapted to this structure. We establish theoretical results showing that this regularized minimization problem admits a unique solution. Moreover, we prove that if a sequence of minimizers corresponds to perturbed observation data converging to the exact (noise-free) data, then the associated sequence of potentials converges to the solution of the considered inverse problem. To compute the minimizer, we derive the Fréchet gradient of the Kohn–Vogelius functional explicitly and employ an iterative CGM. We rigorously prove the Lipschitz continuity of the Fréchet gradient, which underpins the convergence analysis of the CGM. A distinctive and challenging numerical aspect of our method is the selection of the optimal step size within the CG iteration. Unlike standard implementations that solve a linear equation, our formulation necessitates solving a non-trivial quadratic equation, a complication we address with a carefully designed and efficient numerical procedure. For completeness, it