Two remarks on the interpolation space

Mohammad Daher

REVIEW 3 major objections 3 minor 4 references

Two remarks on the interpolation space

T0 review · 3 major / 3 minor · reviewed 2026-08-14 · deepseek-v4-flash

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

Pith's one-line read For 0<θ<1, interpolating measures on the circle with null sequences gives exactly the L1 interpolation space; every element is integrable.

desk verdict Theorem 0.3 is a solid little observation; the paper's headline equality is not actually proved—Proposition 0.1 only shows E⊂Xθ. read the letter →

arxiv 1908.02977 v2 pith:Z2FRAPQD submitted 2019-08-08 math.FA

Mohammad Daher This is my paper · ORCID

classification math.FA MSC 46B7046E2746B45

keywords interpolationspacesmeasuresonthetorusFouriercoefficientsc0sequencesL1functionsGarling-Smithcouplecomplexapproximateidentity

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 establishes that the interpolation space between M(𝕋), the complex measures on the circle, and c0(ℤ), the null sequences, coincides with the interpolation space between L¹(𝕋) and c0(ℤ) for every 0<θ<1. The interest is that interpolation normally enlarges a space, yet here the measure space collapses to its integrable-function subspace. The proof works by showing that every measure whose Fourier coefficients vanish at infinity can be approximated by L¹ convolutions inside the interpolation space. A second remark constructs an isomorphism for the Garling–Smith interpolation couple that sends the θ-index subspace to the 1−θ subspace.

What carries the argument

The load-bearing object is the space $E$ of measures on $\mathbb{T}$ whose Fourier coefficients vanish at infinity, together with the interpolation spaces $X_\theta=(L^1(\mathbb{T}),c_0(\mathbb{Z}))_\theta$ and $Y_\theta=(M(\mathbb{T}),\ell^\infty(\mathbb{Z}))_\theta$. The approximation mechanism is convolution with a bounded approximate identity $(K_n)$: translating $\mu$ by $t$ is continuous in $Y_\theta$ because its Fourier coefficients decay uniformly at infinity, so $K_n\ast \mu\in L^1$ converges to $\mu$ in $Y_\theta$, and the isometric inclusion $X_\theta\subset Y_\theta$ transfers the limit into $X_\theta$. For the second result, the machinery is the explicit isomorphism $U_0:C_0\to C_1$ defined on the Garling–Smith blocks and its extension $U_1$ to $C_0+C_1$, which interpolation theory converts into $U_\theta$ and gives the parameter-reversing restriction.

What would settle it

Find a singular measure on $\mathbb{T}$ whose Fourier coefficients tend to $0$ and show it belongs to $(M(\mathbb{T}),c_0(\mathbb{Z}))_\theta$; such a measure cannot be an $L^1$ function, contradicting the asserted equality. More directly, test whether $X_\theta=(L^1,c_0)_\theta$ is closed in $Y_\theta=(M,\ell^\infty)_\theta$ by computing the interpolation norm of an $L^1$ function against a sequence of approximations; failure of closure would break Proposition 0.1.

Watch

Extended reading notes

Core claim

The central theorem is $(M(\mathbb{T}), c_0(\mathbb{Z}))_\theta = (L^1(\mathbb{T}), c_0(\mathbb{Z}))_\theta$ with equality of norms, for $0<\theta<1$. Writing $X_\theta=(L^1,c_0)_\theta$, $Y_\theta=(M,\ell^\infty)_\theta$, and $Z_\theta=(E,c_0)_\theta$ with $E$ the measures whose Fourier coefficients tend to $0$, the author proves $X_\theta=Z_\theta$ isometrically. The inclusion $X_\theta\subset Z_\theta$ is immediate; for the reverse inclusion, a measure $\mu\in E$ is convolved with an approximate identity $K_m$ after showing the translation map $t\mapsto \mu_t$ is continuous into $Y_\theta$, and the approximating functions $K_m\ast \mu$ lie in $L^1\subset X_\theta$. Because $X_\theta$ is an isometric subspace of $Y_\theta$ by the author's earlier lemma, the limit $\mu$ belongs to $X_\theta$. The second remark shows that for the Garling–Smith couple $(C_0,C_1)$ there is an isomorphism $U_\theta:(C_0,C_0+C_1)_{\theta,p}\to(C_1,C_0+C_1)_{\theta,p}$ whose restriction maps $C_{\theta,p}$ isomorphically onto $C_{1-\theta,p}$, and the complex-method analogue $C_\theta\to C_{1-\theta}$.

Load-bearing premise

The proof relies on the author's earlier lemma [Da1, Lemma 3.8] that $X_\theta=(L^1,c_0)_\theta$ is an isometric, hence closed, subspace of $Y_\theta=(M,\ell^\infty)_\theta$; if this embedding is false or not closed, the approximation argument that places $E$ inside $X_\theta$ collapses.

Editorial extensions

If this is right

For each $0<\theta<1$, the space obtained by interpolating $M(\mathbb{T})$ with $c_0(\mathbb{Z})$ contains no singular measures; every element is an integrable function.
The equality $X_\theta=Z_\theta$ gives a canonical identification of the interpolation norm on the Fourier-vanishing measures, so $E$ becomes a dense subspace of $(L^1,c_0)_\theta$.
The isomorphism $U_\theta$ makes $(C_0,C_0+C_1)_{\theta,p}$ and $(C_1,C_0+C_1)_{\theta,p}$ isomorphic and identifies the subspaces $C_{\theta,p}$ and $C_{1-\theta,p}$, so the Garling–Smith couple is symmetric under $\theta\mapsto 1-\theta$.
For the complex interpolation method, the same construction gives an isomorphism $C_\theta\to C_{1-\theta}$.

Reading between the lines

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

A testable next step would be to compute the interpolation norm of a concrete Rajchman measure in $(M(\mathbb{T}),c_0(\mathbb{Z}))_\theta$; the theorem predicts it equals the $L^1$ norm of a function, giving a numerical check of Proposition 0.1.
If the main equality extends to other compact abelian groups, interpolation between $M(G)$ and $c_0(\widehat G)$ would collapse to $L^1(G)$; the proof's translation-continuity step suggests the group structure, not the dimension of the torus, is what matters.
The parameter reversal in the second theorem hints that any invariant attached to $C_\theta$ must be symmetric in $\theta$, a constraint the paper does not explore.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

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

Two remarks, two very different fates. The Garling-Smith remark (Theorem 0.3) is a clean piece of interpolation theory: U0 is an explicit isomorphism, U1 is well-defined and injective, and the Bergh–Löfström interpolation theorem does the rest. The symmetry Cθ,p ≅ C1−θ,p comes out directly. I'd be happy to see this part as a short note. The first remark, however, is not proved. Proposition 0.1 claims Xθ = Zθ isometrically, and the abstract claims (M,c0)_θ = (L1,c0)_θ. The written proof of Proposition 0.1 starts with μ ∈ E and shows, after a continuity argument and convolution, that μ ∈ Xθ, using [Da1, Lemma 3.8] to make Xθ a closed isometric subspace of Yθ. That is an argument for E ⊂ Xθ. It is not an argument for Zθ ⊂ Xθ, nor for (M,c0)_θ ⊂ Xθ. Interpolation spaces are sensitive to the endpoint: E ⊂ Xθ and c0 ⊂ Xθ do not imply (E,c0)_θ ⊂ Xθ unless the inclusions are bounded and the norms are handled, and replacing E by the larger M raises the same issue again. The reverse inclusion—the entire content of the abstract equality—is never derived. The dependence on [Da1, Lemma 3.8] is load-bearing and unaudited: it's self-cited, not restated, and it's exactly the step that turns approximation in Yθ into membership in Xθ. So the state of play: second remark solid, first remark a plausible conjecture with an incomplete proof. The paper will be read by interpolation theorists; the second part is a useful lemma, and the first is a nice open problem if it isn't true already. I don't think this is a desk-reject situation—the second part is refereeable, and the first part might be repairable—but the referee should demand either a correct proof of Proposition 0.1 or a substantial rewrite that downgrades the abstract to a question. As it stands, the central theorem should not be accepted.

Referee Report

3 major / 3 minor

Summary. The manuscript, written in French, contains two remarks. The first, Proposition 0.1, claims that for 0<θ<1 the complex interpolation space Xθ=(L1(T),c0(Z))θ coincides isometrically with Zθ=(E,c0(Z))θ, where E is the space of complex measures on the torus whose Fourier coefficients tend to zero. The abstract further asserts the stronger equality (M(T),c0(Z))θ=(L1(T),c0(Z))θ, i.e., interpolation of the measure space with c0 produces only integrable functions. The proof of Proposition 0.1 shows that each μ∈E can be approximated in Yθ=(M(T),ℓ∞(Z))θ by L1 functions, whence μ∈Xθ, using a self-cited lemma [Da1, Lemme 3.8] that Xθ is an isometric subspace of Yθ. The second remark, Theorem 0.3, constructs, for a Garling–Smith interpolation couple (C0,C1), an isomorphism Uθ between (C0,C0+C1)θ,p and (C1,C0+C1)θ,p whose restriction to Cθ,p is an isomorphism onto C1−θ,p, with a corresponding statement for the complex method.

Significance. If the first claim were established, it would be a striking rigidity result: interpolation against c0 would force elements of the measure-based interpolation space to be actual integrable functions, a strong analogue of the Riemann–Lebesgue lemma at the level of interpolation spaces. The second remark, Theorem 0.3, is an elegant and seemingly correct construction for the Garling–Smith couple; it is largely self-contained and provides a concrete isomorphism between certain interpolation spaces with a symmetry property. The paper is concise and transparent about its reliance on the author's earlier work, but the central equality for measures is not proven in the present text: the written argument establishes only a one-sided inclusion, and the jump from the auxiliary space E to the full measure space M is missing. Because the abstract's main theorem is unsupported, the paper requires a substantial revision rather than minor polishing.

major comments (3)

[Proposition 0.1] The proof establishes at most the inclusion E⊂Xθ. It shows that every μ∈E can be approximated in Yθ by L1 functions, and since Xθ is claimed to be an isometric (hence closed) subspace of Yθ, concludes μ∈Xθ. This is a valid argument for that inclusion. However, the proposition states the isometric equality Xθ=Zθ. To obtain Zθ⊂Xθ, one must control the interpolation of the pair (E,c0) into Xθ; the argument only handles the first endpoint E and never addresses the second endpoint c0 or the sum of the two components. The manuscript does not show that the canonical copy of c0 in the couple (E,c0) is contained in Xθ, nor that an interpolation inequality for the couple follows. Thus the equality Xθ=Zθ is not derived.
[Abstract / Introduction] The theorem announced in the abstract, (M(T),c0(Z))θ=(L1(T),c0(Z))θ, is never stated as a theorem in the body and is not derived from Proposition 0.1. Even if Proposition 0.1 were fully proved, it would give Xθ=Zθ, which by monotonicity yields Xθ⊂Zθ⊂(M(T),c0)θ. The reverse inclusion (M(T),c0)θ⊂Xθ is the essential content of the abstract claim, and no argument for it appears. The space E is a proper subspace of M (it requires Fourier coefficients to tend to zero), and the manuscript gives no density or approximation argument for arbitrary measures in the interpolation norm. Consequently, the central claim of the paper is unsupported.
[Proposition 0.1, after (0.1)] The proof depends in an essential way on the self-cited result [Da1, Lemme 3.8], which asserts that Xθ is an isometric subspace of Yθ. This lemma is not reproduced or proved in the manuscript, and it is load-bearing: it supplies both the inclusion and the closedness that convert the L1 approximation of μ into membership in Xθ. Because this result is taken from the author's earlier work and is not independently verified here, the proof is not fully auditable. The author should either include a proof of the lemma or state it in sufficient detail to allow verification.

minor comments (3)

[Introduction, notation] The manuscript silently identifies measures with their Fourier transforms when discussing the couples (E,c0), (M,ℓ∞) and the spaces Xθ, Yθ, Zθ. This identification should be stated explicitly, along with the norm carried by E; otherwise the interpolation inequalities, such as the estimate for ||μt−μt′||Zθ after (0.1), are ambiguous.
[Proof of Theorem 0.3] In the demonstration of Theorem 0.3, the text says 'U1 : C0+C1 → C0+C1 sont isomorphismes' before proving that U1 is surjective and bounded. The surjectivity is true (for w=c0+c1 take x=U0−1(c0) and y=U0(c1)), and boundedness follows from the quotient norm, but these steps should be written out. As it stands, the assertion outruns the displayed computation of injectivity.
[Throughout] There are several typographical errors: 'spus-espace' should be 'sous-espace', 'applicationet' should be 'application', 'isomrphisme' should be 'isomorphisme', and 'pour pour tout' has a duplication. In addition, the final phrase of the injectivity argument, 'c'est-à-dire que U0 est une application et injective', is garbled and should read 'que U1 est bien définie et injective'.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity; the self-cited embedding lemma is standard and not equivalent to the target, though Proposition 0.1 has a separate proof gap.

full rationale

Walking the derivation: the abstract claim is (M(𝕋), c0(ℤ))_θ = (L^1(𝕋), c0(ℤ))_θ. The introduction reduces this to Proposition 0.1, Xθ = Zθ, where Xθ=(L^1,c0)_θ, Zθ=(E,c0)_θ, and E is the space of measures with Fourier coefficients tending to zero. The proof of Proposition 0.1 takes μ∈E, shows the translation map t↦μ_t is continuous in Yθ=(M,ℓ∞)_θ, and approximates μ by the L^1 convolutions K_m∗U(0). It then invokes [Da1, Lemme 3.8] to assert that Xθ is an isometric (hence closed) subspace of Yθ, so the Yθ-limit μ lies in Xθ. This is the only step that relies on the author's own previous work. The lemma is a standard interpolation embedding (the inclusions L^1⊂M and c0⊂ℓ∞ give an isometric inclusion of the interpolation spaces) and it does not assert the target equality; it is therefore independent evidence rather than a definitional equivalence. There are no fitted parameters, no prediction renamed as a theorem, and no uniqueness theorem imported to force a choice. The second theorem is an explicit construction from Garling–Smith's couple and the standard interpolation theorem [Ber-Lof, th.3.1.2]. A separate mathematical concern is that the proof of Proposition 0.1 as printed appears to establish only E⊂Xθ; one still needs an argument that Zθ=(E,c0)_θ is contained in Xθ (or that (M,c0)_θ⊂Xθ) to complete the stated equality. That is a potential gap in the proof, but it is not circularity: the conclusion is not assumed among the inputs, and the missing step is an interpolation argument, not an identity-by-definition. Accordingly, no specific circular step can be exhibited, and the circularity score is 0.

Assumptions & free parameters 0 free parameters · 5 assumptions · 0 invented entities

The paper introduces no free parameters and no invented entities. It depends on five external inputs, one of which ([Da1, Lemma 3.8]) is self-cited and load-bearing for the first theorem. The rest are standard theorems in interpolation theory.

assumptions (5)

standard math There exists a sequence (K_n) bounded in L^1(𝕋) that acts as an approximate identity for continuous functions with values in any Banach space.
Used at the start of the proof of Proposition 0.1 to approximate a measure by L^1 convolutions.
domain assumption By [Da1, Lemma 3.8], X_θ=(L^1,c0)_θ is an isometric subspace of Y_θ=(M,ℓ∞)_θ.
Load-bearing, self-cited result; the proof does not reproduce it.
domain assumption By [Blas-Xu], X_θ contains c0 isomorphically.
Cited to explain the structure of X_θ; not used directly in the proof except as background.
domain assumption By [Gar-Smi, Theorem 2], there is a couple (C0,C1), each isomorphic to ℓ1, with the coordinate form in (0.2)-(0.3).
The explicit formulas for C0 and C1 are taken from the cited theorem.
standard math By [Ber-Lof, Theorem 3.1.2], interpolation of the isomorphism U1 yields U_θ with the stated restriction property.
Standard interpolation theorem for operators; not proved in the text.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Two remarks on the interpolation space." pith.science (2026). https://pith.science/paper/Z2FRAPQD

@misc{pith2026190802977,
  author       = {Pith},
  title        = {Pith review of: Two remarks on the interpolation space},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/Z2FRAPQD}},
  note         = {Machine review of arXiv:1908.02977}
}

abstract

Dans ce travail, on montre que $(M(\mathbb{T}),c_0(\mathbb{Z}))_\theta = (L^1,c_0(\mathbb{Z}))_\theta$, $0<\theta <1$. Dans la suite on montre pour le couple d'interpolation $(C_0,C_1)$ trouv\'e par Garling-Smith qu'il existe un isomorphisme $U_\theta: (C_0,C_0+C_1)_{\theta ,p}\rightarrow (C_1,C_0+C_1)_{\theta, p}$ (resp. $U_\theta : (C_0,C_0+C_1)_\theta \rightarrow (C_1,C_0+C_1)_\theta)$ tel que sa restriction \`a $C_{\theta, p}$ (resp. \`a $C_\theta)$ est un isomorphisme : $C_{\theta, p} \rightarrow C_{1-\theta, p}$ (resp. $C_\theta \rightarrow C_{1-\theta })$. -- In this work we show that $(M(\mathbb{T}),c_0(\mathbb{Z}))_\theta = (L^1,c_0(\mathbb{Z}))_\theta$, $0<\theta <1.$ In the following we show for the interpolation couple found by Garling-Smith that there exists an isomorphism $U_\theta: (C_0,C_0+C_1)_{\theta ,p}\rightarrow (C_1,C_0+C_1)_{\theta, p}$ (resp. $U_\theta : (C_0,C_0+C_1)_\theta \rightarrow (C_1,C_0+C_1)_\theta)$ such that its restriction to $C_{\theta ,p}$ (resp. to $C_\theta)$ is an isomorphism : $C_{\theta, p} \rightarrow C_{1-\theta, p}$ (resp. $C_\theta \rightarrow C_{1-\theta })$.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

4 extracted references · 4 canonical work pages

[1]

o fstr\

J. Bergh, J. L\" o fstr\" o m, \ Interpolation spaces an introduction, Springer-Verlag-Berlin Heidelberg New York, (1976)

work page 1976
[2]

Blasco, Q

O. Blasco, Q. Xu, Interpolation between vector valued Hardy spaces, J. Func. Anal. 102, 331-359, (1991 )

work page 1991
[3]

Daher, Interpolation des espaces de Hardy vectoriels, Annales de Toul

M. Daher, Interpolation des espaces de Hardy vectoriels, Annales de Toul. Vol. XXIV, n ^ 2, 389-425, (2015)

work page 2015
[4]

Garling-Montgomery-Smith, Complemented subspaces of spaces obtained by interpolation, J. Lond. Math. Soc. II, Ser. 44, No. 3, 503-513, (1991)

work page 1991

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.

[1] [1]

o fstr\

J. Bergh, J. L\" o fstr\" o m, \ Interpolation spaces an introduction, Springer-Verlag-Berlin Heidelberg New York, (1976)

work page 1976

[2] [2]

Blasco, Q

O. Blasco, Q. Xu, Interpolation between vector valued Hardy spaces, J. Func. Anal. 102, 331-359, (1991 )

work page 1991

[3] [3]

Daher, Interpolation des espaces de Hardy vectoriels, Annales de Toul

M. Daher, Interpolation des espaces de Hardy vectoriels, Annales de Toul. Vol. XXIV, n ^ 2, 389-425, (2015)

work page 2015

[4] [4]

Garling-Montgomery-Smith, Complemented subspaces of spaces obtained by interpolation, J. Lond. Math. Soc. II, Ser. 44, No. 3, 503-513, (1991)

work page 1991