Information Limits for Stable Reduction of Coupled Diffusion Systems
Hiroki Ishizaka
Abstract
Eliminating an unresolved diffusion field from a coupled parabolic system produces an operator-valued memory term and a contribution from the hidden initial state. We study whether a reduced prediction of the retained field meets a prescribed tolerance. For known hidden initial data, the discarded response, damped by diffusion and transmitted through the coupling, gives an upper error bound; an auxiliary diffusion problem supplies a lower bound. Rigorous discretisation bounds then yield tests that accept, reject or remain undecided. If the discarded initial state is known only through a norm bound, the minimum worst-case error equals that bound multiplied by the time-uniform gain from the missing state to the retained trajectory. Our main result encloses this infinite-dimensional gain using a finite spectral Galerkin evolution, its residual in the full equations and an explicit diffusion tail. The enclosure converges without compactness or smallness of the coupling. Strict comparisons of the optimal radius with a tolerance have finite certificates if all endpoint quantities admit arbitrarily accurate rigorous enclosures. Multiplication-coupling examples with infinitely many dynamically coupled modes illustrate both settings, including explicit finite-element error budgets.
Create a lesson
Related papers
Convergence analysis of TriCG and TriMR for symmetric quasi-definite linear systems
Kui Du, Jia-Jun Fan
Adaptive data-driven reduced-order models for reciprocal MIMO port-Hamiltonian dynamical systems
Anarzhan Abilgazy, Mikhail Zaslavsky
Fixed-grid schemes for fractional moving-boundary problems
Conrard Giresse Tetsassi Feugmo
Initial-Trace Insufficiency and Entropy-Accounting Separation in Finite-Volume Coarse Graining
Antonis Polemitis, Nicholas Christakis, Dimitris Drikakis
A probabilistic score estimation using the Bismut--Elworthy--Li formula
Jiatong Sun, Long Nguyen, Ning Jiang et al.
Weighted Randomized Kaczmarz for Sparsely Corrupted Linear Systems
Jessie Chen, Ziyu Cheng, Yunhui He et al.