Inverse problem to determine simultaneously several scalar parameters and a time-dependent source term in a superdiffusion equation involving a multiterm fractional Laplacian
Hany Gerges, Jaan Janno
TL;DR
We study the inverse problem of simultaneously identifying a scalar coefficient $a$, the time-fractional order $\alpha$, the multiterm Laplacian parameters $(b_j,\beta_j)$, and a time-dependent source term from time-trace data of solutions to a superdiffusion equation. The forward model employs a multiterm spectral Laplacian $L=\sum_{j=1}^m b_j(-\Delta)^{\beta_j}$ with $\alpha\in(1,2)$ and a source $\chi=g(t)f(x)+z(t,x)$; observations are given by $\Phi u(t,\cdot)=h(t)$ on $(0,T+\delta)$. A novel pole-asymptotics approach analyzes the Laplace transform $H(s)$ of the observed functional to establish uniqueness, by relating poles to the spectral data via $\mu_k=\tfrac{1}{a}\sum_{j=1}^m b_j\lambda_k^{\beta_j}$ and the Mittag-Leffler structure of the solution. The main contributions include a rigorous uniqueness theorem for the inverse problem under precise regularity and spectral assumptions, and a detailed pole-based framework that accommodates nonlocal spatial operators and a time-dependent source, supported by eigenvalue asymptotics and an established direct-problem theory. The results enhance identifiability in fractional-elliptic-diffusion models and provide a principled route to reconstruct nonlocal operators from time-trace measurements.
Abstract
Inverse problem to recover simultaneously a scalar coefficient, order of a time-fractional derivative, parameters of multiterm fractional Laplacian and a time-dependent source term occurring in a superdiffusion equation from measurements over the time is considered. Uniqueness of a solution is proved. The proof uses asymptotics of poles of the Laplace transform of a measured function.
