Testing compactness of linear operators

Timo S. H\"anninen; Tuomas V. Oikari

REVIEW 4 major objections 5 minor 4 references

Testing compactness of linear operators

T0 review · 4 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read The universal test for compactness is weak nullity of every subsequence of test vectors.

desk verdict A clean abstract framework for testing compactness with a genuinely new paraproduct/VMO characterization, but the sufficiency proof has a load-bearing gap that likely needs one real fix. read the letter →

arxiv 2411.17654 v1 pith:C5YKR6NM submitted 2024-11-26 math.FA math.CA

Timo S. Hänninen , Tuomas V. Oikari This is my paper · ORCID

classification math.FAmath.CA MSC 46B2046E3047B0742B20

keywords compactlinearoperatorsadmissiblesequencesweaklynulldyadicparaproductsvanishingmeanoscillationBanach-SakspropertyT(1)theoremlocallyfiniteBorelmeasures

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper asks which sequences of test vectors are powerful enough to detect compactness universally: given sets $(F_i)$ in a Banach space $X$, when does $\limsup_i \sup_{f_i \in F_i} \|T f_i\|_Y = 0$ hold for every Banach space $Y$ and every compact operator $T\colon X\to Y$? It proves that this happens exactly when every subsequence chosen from the sets is weakly null, a purely Banach-space condition that makes no reference to the operator. This gives a common template for the testing conditions that appear in $T(1)$-type compactness theorems. As the main application, the paper characterizes the $L^p\to L^p$ compactness of dyadic paraproducts on general locally finite Borel measures by the vanishing of mean oscillations along heavy, light, and distant cubes.

What carries the argument

The load-bearing object is the notion of an admissible sequence: $(F_i)$ is admissible if the displayed limit vanishes for every compact $T$ and every $Y$. Theorem 2.1 identifies admissibility with weak nullity of every subsequence, which is the engine of the paper. In the paraproduct application, the machinery is the dyadic paraproduct itself, $P_{b,D} f = \sum_{Q\in D} D_Q b \,\langle f\rangle_Q 1_Q$, together with the mean-oscillation testing bound $\|b-\langle b\rangle_Q\|_{L^p(Q,\mu)} \le 2\|P_b(1_Q/\mu(Q)^{1/p})\|_{L^p}$, and a telescoping identity (Lemma 6.11) that reduces sums of martingale differences over a connected chain of cubes to differences of averages $\langle b\rangle_R-\langle b\rangle_S$. The sufficiency proof repeatedly splits a paraproduct into a finite-rank part and a piece whose norm is controlled by vanishing mean oscillations; since finite-rank operators are compact and the compact operators form a closed subspace, each reduction step preserves compactness.

What would settle it

Take a purely atomic locally finite Borel measure on $\mathbb{R}$ whose atoms accumulate at the origin, and define $b$ so that its mean oscillations over heavy, light, and distant dyadic cubes vanish while $\int_{\bigcup Q_k}|b|\,d\mu$ grows without bound along a nested chain $Q_1\subset Q_2\subset\cdots$ of cubes of comparable measure. Computing $\|P_{b,D}(1_{Q_k})\|_{L^p(\mu)}$ along this chain would then test the unproved Lemma 6.14 estimate: a nonzero limit would give a symbol in VMO whose paraproduct is not compact, contradicting Theorem 6.7.

Watch

Extended reading notes

Core claim

The central discovery is a dichotomy: a sequence $(F_i)$ of sets in a Banach space $X$ is admissible for testing compactness if and only if every subsequence $f_{i(k)} \in F_{i(k)}$ converges weakly to zero. The forward direction uses compactness of the operator; the reverse direction is a uniform-boundedness and subsequence argument, and it is fully general. With this criterion in hand, the paper proves that a dyadic paraproduct $P_{b,D}$ on $L^p(\mathbb{R}^d,\mu)$ is compact exactly when the symbol $b$ belongs to the generalized VMO space defined by vanishing mean oscillations over cubes that are heavy ($\mu(Q)\ge M$), light ($\mu(Q)\le 1/M$), or distant from the origin. Necessity follows from the universal test; sufficiency is obtained by reducing any collection of dyadic cubes to finite-rank pieces plus an arbitrarily small error, using a telescoping identity for martingale differences over connected cube chains.

Load-bearing premise

The sufficiency half of the paraproduct characterization rests on two technical estimates that the paper invokes without proof: an integral bound for $|b|$ over a nested chain of cubes in Lemma 6.14, and the interchange of a limit with an $L^p$-norm of a supremum in Lemma 6.15; if either fails for some locally finite Borel measure, the claimed equivalence between compactness and vanishing mean oscillations would collapse.

Editorial extensions

If this is right

In any $T(1)$-type compactness theorem, the testing condition $\lim_i \sup_{Q\in\mathcal{Q}_i}\|T f_{Q,i}\|=0$ can be verified by showing the test functions $f_{Q,i}$ are weakly null along every subsequence, with no reference to the structure of $T$.
For dyadic paraproducts on $L^p(\mathbb{R}^d,\mu)$ with $\mu$ locally finite Borel, compactness is equivalent to $b\in \mathrm{VMO}^p(\mathbb{R}^d,D,\mu)$, the space of bounded mean oscillation whose oscillations vanish over heavy, light, and distant cubes.
On $\mathbb{R}^d$ with Lebesgue measure, this recovers the classical result that $P_{b,D}$ is compact if and only if $b$ has vanishing mean oscillations over large, small, and distant cubes.
Under a dyadic doubling condition on $\mu$, the VMO class is independent of $p$: $\mathrm{VMO}^p = \mathrm{VMO}^1$ for all $p\in(1,\infty)$.
The admissibility criterion also gives necessary conditions for compactness in other settings, such as commutators and Calder\'on\textendash Zygmund operators, where test sequences are often built from normalized cube indicators.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

Editorial inference: the abstract criterion suggests a general recipe\textemdash whenever a compactness theorem is phrased as a vanishing condition over a family of test functions, the real content is that those functions form a weakly null family; operator-specific work can be shifted to verifying weak nullity.
Editorial inference: the two unproved analytic steps in the sufficiency proof (the integral bound over a nested chain of cubes and the dominated-convergence interchange) are where a counterexample would likely appear for non-doubling measures; testing them on atomic or exponentially decaying measures could decide whether the general-measure characterization survives.
Editorial inference: the measure-theoretic replacement of \textquotedblleft distant cubes\textquotedblright by an exhausting sequence of sets suggests the same compactness characterization should transfer to purely measure-theoretic martingale filtrations, which would make the result independent of the Euclidean structure.
Editorial inference: the Ces\`aro-cancellativity viewpoint implies that in Banach spaces where weak convergence does not force norm-Ces\`aro summability, admissible sequences need not be cancellative, so compactness-testing sets can exist even when no asymptotic cancellation in norm is available.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Tim, here's my read. The abstract machine is simple and correct: Theorem 2.1 just repackages the standard weak-null characterization of compact operators, which is fine, and Section 4 gives genuinely useful geometric criteria, especially the Banach function space version. The real news is the VMO characterization for paraproducts on general locally finite Borel measures, replacing side length with heavy/light/distant cubes. That is a natural and worthwhile extension, and the necessity direction is clean. The proof strategy by reduction steps is also nice. The soft spot is exactly where the stress-test note lands: Lemma 6.15. The line where the limit is moved inside the L^p norm via dominated convergence is not justified. No dominating function is exhibited, and the pointwise limit is attributed to the Lebesgue differentiation theorem, which does not apply pointwise for arbitrary Borel measures. The natural fix—use dyadic martingale convergence on the finite-measure union and dominate by something like the BMO maximal function—probably works, but it is not in the paper. This is not a cosmetic gap; the sufficiency direction of Theorem 6.7 rests on it. Lemma 6.14 looks fixable too, but the stress-test note is right that it is easier. The self-citation of Lemma 6.9 is fine; it is published and not circular, though it does reduce self-containedness. I agree with the reader that the paper deserves conditional acceptance rather than rejection: the framework is solid, the new characterization is likely true, and the missing estimates are probably fillable. But a referee should insist on seeing them. Recommend: send to peer review, and make the main referee report focus on Lemma 6.15 and the DCT interchange. If the authors fill that gap, this is a nice paper for harmonic analysis and operator theory audiences.

Referee Report

4 major / 5 minor

Summary. The paper studies sequences of sets (F_i) in a Banach space X for which the condition limsup_i sup_{f_i in F_i} ||T f_i||_Y = 0 holds for every compact linear operator T : X to Y (admissible sequences). Theorem 2.1 characterizes admissible sequences as exactly those for which every subsequence is weakly null; Lemma 2.2 adds that admissible sequences are uniformly bounded. Sections 3 and 4 connect this notion to cancellativity and the (weak) Banach–Saks property, and give verifiable criteria in Hilbert spaces (Theorem 4.1) and in Banach function spaces with an improved triangle inequality (Theorem 4.2). Section 5 explains the relevance to T(1)-type compactness criteria. Section 6 applies the framework to dyadic paraproducts P_{b,D} on L^p(R^d, mu) for arbitrary locally finite Borel measures mu: Theorem 6.6 shows that compactness forces b to have vanishing mean oscillations along heavy, light and distant cubes, and Theorem 6.7 claims the converse. The sufficiency proof proceeds by three reduction steps (Lemmas 6.12 and 6.13) that isolate outer chain-like collections and inner fine-scale collections, whose compactness is delegated to Lemmas 6.14 and 6.15.

Significance. The abstract part of the paper (Sections 2–4) is clean, correct, and genuinely useful: Theorem 2.1 identifies the universal testing condition for compactness with a simple weak-nullness condition, and the geometric criteria (Theorems 4.1 and 4.2) are self-contained and likely to be citable. The necessity direction of the paraproduct application (Theorem 6.6) is an elegant demonstration of the machinery. However, the sufficiency direction (Theorem 6.7) is not established as written: the key auxiliary Lemma 6.11 is false in its stated generality, and two analytic estimates in Lemmas 6.14 and 6.15 are asserted without proof. A repaired proof would yield a substantial extension of the classical Chao–Peng characterization of compact dyadic paraproducts to general locally finite Borel measures, but that claim is currently not proved.

major comments (4)

[Lemma 6.11 (§6.4)] The telescoping identity used to prove this lemma is not valid for the notion of connectedness in Definition 6.10. Take d = 1, the standard dyadic grid, Q = {[0,1), [0,1/2), [1/2,1)} (connected in the sense of Definition 6.10), and b = 10(1_{[0,1/4)} - 1_{[1/4,1/2)}). For x in [0,1/4), the sum over R in Q with R subset [0,1) of D_R b(x) equals _{[0,1/4)} - _{[0,1)} = 10, whereas sup_{R,S in Q, R subset S subset [0,1)} |_R - _S| 1_S(x) = 0, because all Q-averages of cubes containing x coincide (they are all 0). Hence the stated L^p inequality fails even for bounded locally integrable b. The reason is that the dyadic children of cubes in Q need not belong to Q, so the martingale differences do not telescope to differences of Q-averages. Since Lemmas 6.14 and 6.15 both rely on this bound to control the discarded operator norms, the sufficiency direction of Theorem 6.7 is not proved as it stands.
[Lemma 6.14 (proof)] The estimate integral over the union of the chain of cubes of |b| dmu, bounded by the integral over Q_1 of |b| plus the BMO_p(R^d,Q,mu) norm, is asserted without proof. With only mu(Q_k) ~ 1, the averages _{Q_k} can drift by a bounded amount at each step of the chain, and it is not immediate, for arbitrary locally finite Borel mu, that the resulting linear drift is excluded, nor that b is integrable over the (possibly unbounded) union of the chain. This estimate is load-bearing: it is what forces the limit lim_{L -> -infinity} sup_{m,n <= L} |_{Q_m} - _{Q_n}| = 0 that makes the outer-collection error term vanish. A complete proof, or an additional hypothesis on mu (for example dyadic doubling), is needed.
[Lemma 6.15 (proof)] The displayed interchange lim_{L -> infinity} || sup_{R,S in Q cap D_{>=L}, R subset S} |_R - _S| 1_S ||_{L^p} = || lim_{L -> infinity} sup_{R,S in ...} ... ||_{L^p} = 0 is justified only by 'the dominated convergence theorem' and 'the Lebesgue differentiation theorem'. Two required ingredients are missing. First, no dominating function in L^p is identified; the natural candidate C ||b||_{BMO_p(Q)} 1_{union Q} follows from |_R - _S| <~ ||b||_{BMO_p(Q)} when mu(R) ~ mu(S) ~ 1, but this estimate is not shown. Second, pointwise convergence of dyadic averages along nested cubes is a martingale convergence statement on the finite-measure space union Q, which requires b in L^1(union Q, mu); the Lebesgue differentiation theorem is not valid pointwise for arbitrary Borel measures, and b in L^1(union Q, mu) is not proved (the same integrability gap as in Lemma 6.14).
[Theorem 6.7, Step 3] The collections Q_out^{Q_*}(M) produced by Lemma 6.13 are connected and exhaustible by a chain, but they are not closed under dyadic children; Lemma 6.11 therefore cannot be applied to them as stated, and the error bounds in Lemmas 6.14 and 6.15 do not follow. Because of this, the reduction chain used to conclude compactness in Theorem 6.7 breaks at the last step. The authors should either prove a corrected version of Lemma 6.11 (for example, by adding a BMO-norm term or by assuming closure under children) and rework Lemmas 6.14 and 6.15 accordingly, or restrict Theorem 6.7 to a class of measures for which the needed estimates hold.

minor comments (5)

[Sections 6.1 and 6.4] Lemmas 6.14 and 6.15 use the notation BMO_p(R^d, Q, mu) for the dyadic BMO norm relative to a collection Q, but only BMO_p(R^d, D, mu) is defined in Section 6.1; please add the definition.
[Remark 6.4] The claim that b in L^1_loc together with the three vanishing conditions implies b in VMO_p is deferred to the reader; since this implication would need the same estimates that are currently missing from the proof of Theorem 6.7, either prove Remark 6.4 or state explicitly that VMO_p is a subspace of BMO_p and keep that assumption in Theorems 6.6 and 6.7.
[Section 3] The statement that ell^infty lacks the weak Banach–Saks property is unlikely to be documented in [RN12]; please provide a precise reference for this standard fact.
[Throughout] The source file contains many encoding artifacts (for example '/parallel.alt1', '/summation.disp', and 'W e' in the abstract), which make parts of the text hard to read; please ensure a clean compilation.
[Theorem 2.1 (proof)] In the admissibility-implies-weak-convergence direction, it would be cleaner to state explicitly that one tests admissibility against the rank-one compact operator x -> <x, f*> y_0 for a fixed unit vector y_0; as written, the step is implicit.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the admissibility characterization is proved directly from the definitions, and the one cited prior-work lemma in the paraproduct application is independent support rather than a reduction to the target conclusion.

full rationale

Theorem 2.1 is derived from the definition of admissibility and standard Banach-space facts: compact operators map weakly null bounded sequences to norm-null sequences, and the uniform-boundedness principle supplies boundedness. No equation in the proof assumes the conclusion. The paraproduct application (Theorems 6.6 and 6.7) proves necessity by testing the compact operator on normalized cube indicators and verifying their admissibility via Theorem 4.2, and proves sufficiency by a sequence of reduction lemmas. Lemma 6.9 is quoted from the first author's earlier work [HH16], but it is a published, parameter-free norm comparison for paraproducts over arbitrary cube collections whose assumptions do not include compactness; it is therefore independent support and not a circular self-citation. The unproved domination and dominated-convergence interchanges in Lemmas 6.14 and 6.15 are potential analytic gaps in the sufficiency proof, not circular reductions: no fitted quantity is renamed a prediction, no uniqueness theorem is imported from the authors, and no definition is fixed in terms of the result being proved. Remark 6.4 also defers details to the reader, but that omission is a completeness matter, not a circular step. Consequently the derivation chain is not circular.

Assumptions & free parameters 0 free parameters · 6 assumptions · 3 invented entities

The paper introduces no fitted parameters. It relies on standard theorems in functional analysis and on prior results in harmonic analysis, one of which (Lemma 6.9) is from the first author's own earlier work. The main analytic estimates used in the sufficiency proof are supported only by assertion.

assumptions (6)

standard math Uniform boundedness principle
Used in Theorem 2.1 and Lemma 2.2 to conclude that weakly null subsequences are norm bounded.
standard math Riesz representation theorem
Used in Theorem 4.1 to represent functionals by inner products in Hilbert space.
standard math Lebesgue differentiation theorem
Used in Lemma 6.15 to compute the pointwise limit of mean oscillations over shrinking cubes.
domain assumption Boundedness of dyadic paraproducts (Theorem 6.1)
Quoted from [Mey90] and [CL92]; it is an external benchmark, not derived in this paper.
domain assumption Improved triangle inequality condition on Banach function spaces (existence of phi)
Assumed in Theorem 4.2; it holds for L^p, p>1, but is not derived from first principles.
domain assumption Lemma 6.9 (operator norm over any collection of cubes)
Quoted from the first author's prior work [HH16]; the proof is not reproduced and is load-bearing for the reduction arguments.

invented entities (3)

Admissible sequence
purpose: Central object of study: sequences of sets that vanish under every compact operator.
New definition in Section 1; a mathematical construction, not an empirical entity.
Cancellative sequence
purpose: Links admissibility to the Banach-Saks property.
New definition in Section 3.
VMO_p with heavy, light and distant cubes
purpose: Generalized vanishing mean oscillation for locally finite Borel measures.
New definition in Section 6.2; its falsifiable handle is the compactness characterization, but that is the theorem under test.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Testing compactness of linear operators." pith.science (2026). https://pith.science/paper/C5YKR6NM

@misc{pith2026241117654,
  author       = {Pith},
  title        = {Pith review of: Testing compactness of linear operators},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/C5YKR6NM}},
  note         = {Machine review of arXiv:2411.17654}
}

abstract

Let $(F_i)$ be a sequence of sets in a Banach space $X$. For what sequences does the condition $$ \limsup_{i\to \infty} \sup_{f_i\in F_i} \|Tf_i\|_Y=0 $$ hold for every Banach space $Y$ and every compact operator $T:X\to Y$? We answer this question by giving sufficient (and necessary) criteria for such sequences. We illustrate the applicability of the criteria by examples from literature and by characterizing the $L^p\to L^p$ compactness of dyadic paraproducts on general measure spaces.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

4 extracted references · 3 canonical work pages

[1]

Sur la convergence forte dans les c hamps Lp

[BS30] S. Banach and S. Saks. “Sur la convergence forte dans les c hamps Lp” . fre. In: Stud. Math. 2.1 (1930), pp. 51–57. [JN61] F. John and L. Nirenberg. “On functions of bounded mean os cilla- tion” . In: Comm. Pure Appl. Math. 14 (1961), pp. 415–426. issn: 0010- 3640,1097-0312. doi: 10.1002/cpa.3160140317. [Joh65] F. John. “Quasi-isometric Mappings, S...

work page arXiv 1930
[197]

On the $T1$ theorem for compactness of Calder\'on-Zygmund operators

[Naz+17] F. Nazarov et al. “Convex body domination and weighted es timates with matrix weights” . In:Advances in Mathematics 318 (2017), pp. 279–306. issn: 0001-8708. doi: 10.1016/j.aim.2017.08.001. [Guo+21] W. Guo et al. “Boundedness and compactness of commuta tors associ- ated with Lipshitz functions” . In: Anal. Appl. 20.1 (2021), pp. 1–37. doi: 10.114...

work page Pith review arXiv 2017
[544]

Ondelettes et op´ erateurs

issn: 0022-2518,1943-5258. doi: 10.1512/iumj.1979.28.28037. url: https://doi.org/10.1512/iumj.1979.28.28037. [Mey90] Y. Meyer. “Ondelettes et op´ erateurs” . In: I: Ondelettes (1990). [CL92] J.-A. Chao and R.-L. Long. “Martingale Transforms with Unbo unded Multipliers” . In:Proceedings of the American Mathematical Society 114.3 (1992), pp. 831–838. [CP96]...

doi:10.1512/iumj.1979.28.28037 1990
[2012]

A characterization of compactness for singula r integrals

isbn: 9780486162140. [Vil15] P. Villarroya. “A characterization of compactness for singula r integrals” . In: J. Math. Pures Appl. 104.3 (2015), pp. 485–532. issn: 0021-7824. doi: https://doi.org/10.1016/j.matpur.2015.03.006. [HH16] T. S. H ¨anninen and T. P. Hyt ¨onen. “Operator-valued dyadic shifts and the T (1) theorem” . In:Monatsh. Math. 180.2 (2016)...

work page doi:10.1016/j.matpur.2015.03.006 2015

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.

[1] [1]

Sur la convergence forte dans les c hamps Lp

[BS30] S. Banach and S. Saks. “Sur la convergence forte dans les c hamps Lp” . fre. In: Stud. Math. 2.1 (1930), pp. 51–57. [JN61] F. John and L. Nirenberg. “On functions of bounded mean os cilla- tion” . In: Comm. Pure Appl. Math. 14 (1961), pp. 415–426. issn: 0010- 3640,1097-0312. doi: 10.1002/cpa.3160140317. [Joh65] F. John. “Quasi-isometric Mappings, S...

work page arXiv 1930

[2] [197]

On the $T1$ theorem for compactness of Calder\'on-Zygmund operators

[Naz+17] F. Nazarov et al. “Convex body domination and weighted es timates with matrix weights” . In:Advances in Mathematics 318 (2017), pp. 279–306. issn: 0001-8708. doi: 10.1016/j.aim.2017.08.001. [Guo+21] W. Guo et al. “Boundedness and compactness of commuta tors associ- ated with Lipshitz functions” . In: Anal. Appl. 20.1 (2021), pp. 1–37. doi: 10.114...

work page Pith review arXiv 2017

[3] [544]

Ondelettes et op´ erateurs

issn: 0022-2518,1943-5258. doi: 10.1512/iumj.1979.28.28037. url: https://doi.org/10.1512/iumj.1979.28.28037. [Mey90] Y. Meyer. “Ondelettes et op´ erateurs” . In: I: Ondelettes (1990). [CL92] J.-A. Chao and R.-L. Long. “Martingale Transforms with Unbo unded Multipliers” . In:Proceedings of the American Mathematical Society 114.3 (1992), pp. 831–838. [CP96]...

doi:10.1512/iumj.1979.28.28037 1990

[4] [2012]

A characterization of compactness for singula r integrals

isbn: 9780486162140. [Vil15] P. Villarroya. “A characterization of compactness for singula r integrals” . In: J. Math. Pures Appl. 104.3 (2015), pp. 485–532. issn: 0021-7824. doi: https://doi.org/10.1016/j.matpur.2015.03.006. [HH16] T. S. H ¨anninen and T. P. Hyt ¨onen. “Operator-valued dyadic shifts and the T (1) theorem” . In:Monatsh. Math. 180.2 (2016)...

work page doi:10.1016/j.matpur.2015.03.006 2015