A mathematical preprint presents a constructed smooth propagation speed c: (0, +∞) → [1/2, 3/2]. For every T > 0 and s > 1/2, that speed falls outside W^{s,1}((0,T)); W^{s,1} is a fractional Sobolev regularity class used in the analysis. Yet the associated abstract wave equation is well posed in D(A^{1/2}) × H for every Hilbert space and every nonnegative self-adjoint operator A.
The work is theoretical: it analyzes abstract wave equations and their scalar spectral-frequency reductions, and describes no empirical sample. The document is an arXiv version 1 preprint dated 25 Aug 2026.
The estimate has two moving parts
The equations involve a propagation speed that changes with time and are studied under strict hyperbolicity. That assumption means the speed is bounded above and below almost everywhere by positive constants. The equations are reduced to scalar frequency problems, and the high-frequency behavior of an energy-growth function, G, is used to classify outcomes as no derivative loss, arbitrarily small derivative loss, finite derivative loss or infinite derivative loss.
One organizing device is the approximate-energy method. It replaces 1/c with a positive smooth γ and treats two requirements as competing costs: γ must have enough derivative regularity to support the estimate, while remaining faithful to the reciprocal coefficient. The paper also constructs a hierarchy of frequency-dependent quadratic energies equivalent to the standard energy and increasingly close to conserved quantities.
Why second order matters
The variational results identify a dividing line at order two. All variational problems of order k ≥ 2 have the same admissible growth classes, while the first-order problem can behave differently.
The associated bounded-growth spaces show the same split: the first-order space is strictly contained in the second-order space, and the second-order space equals every higher-order space for k ≥ 2. The higher-order class is itself strictly contained in a fractional Sobolev class.
Membership of 1/c in S2 is sufficient for Sobolev well-posedness without derivative loss. The analysis presents this as a sufficiency statement, not as a general necessity or converse.
A separate oscillatory regime
In the oscillatory estimates, the proof separates a central near-resonant region from two non-resonant regions. It uses integration by parts to control boundary terms and differentiated amplitudes.
The framework is also applied to modulus-of-continuity, fractional Sobolev, derivative, three-region and Zygmund-type assumptions. In the bounded Zygmund scale, G is associated with bounded growth and no derivative loss. In the logarithmic Zygmund scale, G is associated with the bound C(1 + log λ) and finite derivative loss.
Power-like examples yield speeds that fail to belong to W^{s,1} above a threshold that can be made arbitrarily close to 1/2, but the threshold is never equal to 1/2.
The boundary is still open
The paper asks whether local W^{1/2,1} regularity is necessary for uniform energy estimates. The oscillatory construction does not settle that endpoint or characterize every coefficient outside the variational class.
These are deterministic mathematical statements about equations and function spaces, not empirical evidence. Their scope includes the strict-hyperbolicity assumption, so the results do not by themselves establish uniformly bounded energy for every rough propagation speed.
Paper data and sources
Original title: Fifty Years of Wave Equations with Time-Dependent Propagation Speeds: A Variational Perspective and New Frontiers
Authors: Marina Ghisi, Massimo Gobbino
Journal/Repository: arXiv
Status: Preprint, not yet peer-reviewed
First online: 2026-08-25
DOI: Not available
Original paper · Full text