Preprint

Preprint finds a constructed operator can pass a calculus test yet fail stochastic regularity

The theoretical analysis maps exact deterministic and stochastic conditions for diagonal operators and ties them to the geometry of Banach spaces.

An arXiv preprint reports that a constructed operator can pass a demanding calculus test and still fail a key form of stochastic regularity. In the example, the operator has a bounded H∞-calculus of angle zero but fails stochastic maximal L^p-regularity for every p≥2. The construction uses X=ℓ^2(ℓ^r) with r>2, a UMD Banach function space of type 2.

The finding is deliberately narrow, but important within the paper’s framework. It says that a bounded H∞-calculus alone is not enough for stochastic maximal regularity for this constructed operator; it does not say that every operator with such a calculus fails. The wider paper asks how Banach-space geometry constrains deterministic and stochastic maximal regularity—roughly, whether the mathematical system maintains the needed control when random input is present.

The diagonal test bed

The paper is a proof-driven analysis of a positive diagonal operator on mixed-norm sequence spaces X_{q,r}=ℓ^q(ℓ^r). A diagonal operator is specified coordinate by coordinate, allowing the authors to compare exact regularity conditions as the exponents p, q and r vary.

For the deterministic problem, the classification is exact for p,q,r≥1. Maximal L^p-regularity holds when p>1 and either q>1 or r=1; the all-one case p=q=r=1 also qualifies. Outside those cases, it fails for this diagonal operator.

A separate example shows the same gap between a favorable calculus property and maximal regularity in a deterministic setting. For r>1, the operator on X_{1,r} is R-sectorial, has a bounded H∞-calculus of angle zero, yet fails deterministic maximal L^p-regularity for every p≥1.

The stochastic version shifts the thresholds. For p,q,r≥2, stochastic maximal L^p-regularity holds exactly when p>2 and either q>2 or r=2, with the all-two case p=q=r=2 as an additional qualifying case.

A proof step called concavification links the two classifications: for p,q,r≥2, stochastic maximal L^p-regularity for the diagonal operator is equivalent to deterministic maximal L^{p/2}-regularity for a related construction on X_{q/2,r/2}.

A condition beyond the calculus

The authors then examine (S_p), a condition on stochastic-convolution kernels. For UMD Banach spaces of type 2 and p≥2, its interval-kernel and exponential-kernel formulations are equivalent.

At the endpoint p=2, (S_2) holds exactly when the space is isomorphic to a Hilbert space.

For p,q>2, the paper states an equivalence among (S_p), stochastic maximal L^p-regularity for the diagonal multiplier on Rad_p(X), and stochastic maximal L^p-regularity for the Laplacian on L^q(ℝ^d;X).

What the result does—and does not—say

A broader generalized diagonal construction separates two geometric requirements. On X(ℓ^2), R-sectoriality of angle zero is equivalent to finite cotype of the underlying space, while maximal L^p-regularity is equivalent to the UMD property.

The conclusions remain tied to specified constructions and parameter ranges. The full Laplacian equivalence is stated for p,q>2; when exactly one of p or q equals 2, the proof does not establish the converse implication. Likewise, the counterexample does not show that every operator with a bounded H∞-calculus fails stochastic maximal regularity.

The manuscript is identified as arXiv version 1, dated 20 August 2026. Its results are mathematical equivalences about the stated operators and Banach-space constructions, rather than findings from an empirical dataset.

Paper data and sources

Original title: Necessary conditions for deterministic and stochastic maximal regularity
Authors: Emiel Lorist, Jan van Neerven, Mark Veraar
Journal/Repository: arXiv
Status: Preprint, not yet peer-reviewed
First online: 2026-08-20
DOI: Not available
Original paper · Full text

Versions and corrections

  1. Published automatically after legal-source, freshness, evidence, and independent-verification gates passed.