REVIEW 3 major objections 4 minor 2 references
Extensions of homogeneous distributions on deformations to the normal cone
T0 review · 3 major / 4 minor · reviewed 2026-08-07 · deepseek-v4-flash
Pith's one-line read Homogeneous distributions extend across the deformation to the normal cone.
desk verdict The paper's central extension claim is false (e^{t^2} counterexample), but the polar decomposition lemma, the conditional extension theorems, and the classification of singular-supported homogeneous distributions are salvageable and worth a serious referee. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
The central object is the zoom action on the deformation to the normal cone, expressed in exponential charts as α_s(x,X,t)=(x,sX,$s^{{-1}}$t) on N^M_V×R. The key identity is the polar-coordinate representation of a-homogeneous distributions: u is a-homogeneous iff it is given by the weighted integral of a boundary value w on S^M_V×R against $s^{{a-1}}$, with the time dilation acting on w. This identity transfers the extension at the base of the normal bundle to the problem of integrating a weighted density across s=0, and the paper applies the classical extension theory for weakly homogeneous distributions to the time factor. The ambiguity of the extension is controlled by a direct-sum decomposition of homogeneous distributions supported on V×R, indexed by the order of normal derivatives.
What would settle it
Take the explicit homogeneous distribution from the paper's optimality remark, u(X,t)=|t|^{3/2}|tX|^{-n-1}dXdt, run it through the paper's own cutoff and polar-coordinate recipe, and compute the time support of the resulting boundary value w on S^M_V×R: the cutoff only bounds the product |X||t|, so w is not compactly supported in t. Then evaluate the defining Mellin integral for a test function f supported arbitrarily close to r=0 with f(0)≠0; if the limit (including any finite-part regularization) does not exist, the claimed universal extension fails for this example, and if it does exist, the proof needs a sharper temperedness argument than the one given.
Extended reading notes
Core claim
The central discovery is that the extension problem for the zoom action can be solved on the model space N^M_V×R by decomposing the action α_s: (x,X,t)↦(x,sX,$s^{{-1}}$t) into commuting dilations on the normal fiber and on time, then reading the distribution in polar coordinates (r,ω,t). An a-homogeneous distribution u is encoded by a distribution w on S^M_V×R through ⟨u,f⊗g⟩ = ∫_0^∞ $s^{{a-1}}$f(s)⟨w,γ^*_{$s^{{-1}}$}g⟩ds. Extending u across the exceptional locus amounts to making this integral well defined at r=0, which the paper achieves when w has suitable weak homogeneity under the time dilation by iterated integration by parts. The extension is non-canonical; two extensions differ by an a-homogeneous distribution supported on V×R, and the paper gives a complete formula for these discrepancy terms in terms of vertical homogeneous differential operators and powers of t.
Load-bearing premise
The final construction assumes that after cutting off the distribution and taking a partition of unity on the sphere bundle, the boundary values w_i on the sphere bundle times R are compactly supported and hence tempered; this is not automatic because the time factor R is not compactified and the paper gives no argument controlling the time support.
Editorial extensions
If this is right
- Every zoom-homogeneous distribution on the complement can be extended across the exceptional locus, so homogeneity alone does not block extension.
- The classification of supported homogeneous extensions gives a complete description of the ambiguity: any two extensions differ by an explicit sum of vertical homogeneous differential operators acting on distributions on V×R.
- For the tangent-groupoid case DNC(M×M,Δ_M)=TM, the π-transversal homogeneous distributions supported on the diagonal coincide with the differential-operator kernels used in the groupoid characterization of pseudodifferential operators.
- In the π-transversal case the extension is unique outside the critical degrees a+b∈-N*, which identifies exactly when a homogeneous extension is canonical.
- The same machinery describes homogeneous distributions supported on V×R, giving a model for the singular part of any homogeneous distribution on the whole space.
Reading between the lines
- The paper's final globalization step asserts that after multiplying by a cutoff and a partition of unity on the sphere bundle, the boundary values w_i are compactly supported; because the time factor R is not compactified, this claim is not justified by the displayed argument, and a temperedness or decay condition in t would be needed to complete the proof.
- The Mellin-type integral representation is reversible: sufficiently regular boundary data w on S^M_V×R produce homogeneous distributions u, so the paper's framework suggests a bijection between a class of boundary data and homogeneous distributions up to supported corrections.
- The same polar-coordinate reduction could be applied to homogeneous distributions for other one-parameter subgroups of the zoom action, or to quasi-homogeneous degrees, yielding a computational way to test extension statements on explicit models such as DNC(R,{0}).
- If the missing temperedness step can be patched, the dividing line between canonical and non-canonical extensions is the same as in the weak-homogeneity theory: canonical when c+Re(a) avoids -N, and governed by finite-part ambiguities otherwise.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper studies homogeneous distributions for the zoom action on the deformation to the normal cone DNC(M,V). The abstract claims that every a-homogeneous distribution on DNC(M,V)\V×R admits an a-homogeneous extension, and that all such extensions can be described. The proof strategy combines Meyer's weak-homogeneity extension theorems on vector bundles with a polar-coordinate decomposition on the complement of V, culminating in a partition-of-unity argument in the final section. The paper also contains conditional extension results (Theorem 3.3, Theorem 3.4) and a characterization of homogeneous distributions supported on V×R (Proposition 3.5, Corollary 3.5.1).
Significance. If the advertised unconditional extension theorem were correct, it would give a clean statement in the theory of tangent groupoids and could be relevant to the construction of homogeneous kernels in pseudodifferential calculus. The paper does contain a useful conditional theorem, Theorem 3.3, under explicit weak-homogeneity and exponent assumptions, as well as a detailed classification in Proposition 3.5. However, the central unconditional claim is false, and the proof's final step relies on a compact-support assertion that is incorrect. The paper as submitted does not deliver its main advertised result.
major comments (3)
- [§3, final construction (after Theorem 3.4)] The sentence 'We have that every wi is compactly supported, thus a tempered distribution' is false. A partition of unity (χ_i) on S_V only localizes the sphere directions; it does not make the support of w_i compact in the noncompact time factor R. Proper support of w for the projection S_V×R→S_V is not compact support in the total space. Concretely, take V={0}, N=R, S={±1}, and w(t)=e^{t^2} on {±1}×R. This is locally integrable and π(supp w)=S is compact, but w is neither compactly supported nor tempered. Substituting this w into Lemma 3.2 with a=-1 yields a -1-homogeneous distribution u∈D'((R\{0})×R) that admits no -1-homogeneous extension to R×R: for test functions g with nonzero integral, ⟨w, γ_{s^{-1}}g⟩ grows like s^{-1} as s→0, so the candidate extension integral diverges logarithmically. The same obstruction is acknowledged in Remark 3.3.1 for c+Re(a)∈-N, and the final argument does not address it.
- [Theorem 3.4 and its proof] The proof of Theorem 3.4 invokes Lemma 1.16 to decompose w into a finite sum of weakly-homogeneous distributions. However, Lemma 1.16 is stated and proved for a vector bundle W→V with a fiberwise dilation action. The space S_V×R is not a vector bundle over S_V (its fibers are S^n×R), and there is no global dilation action on the sphere factor. Thus the decomposition via Lemma 1.16 is not justified in this setting. Even if w were tempered, the application of Lemma 1.16 requires a vector-bundle structure that is absent.
- [Application of Theorem 3.3 in the final construction] The passage from the decomposition of w to the hypotheses of Theorem 3.3 is not established. Theorem 3.3 requires u∈E^c_γ, i.e. a weak-homogeneity bound with respect to the time-dilation γ, with c+Re(a)∉-N. The decomposition provided by Remark 1.16.1 yields summands in E^{n±ε}_α for the normal-bundle dilation α, not in E^c_γ. No argument shows that the w_i satisfy any γ-weak homogeneity estimate, so the applicability of Theorem 3.3 to the reconstructed summands v_i is unsupported.
minor comments (4)
- [Abstract] The sentence 'The technique used come from the results...' has a subject-verb agreement error; it should read 'The technique used comes from...'.
- [Remark 3.3.1] The notation F^{-n-1/2}_γ is used without definition; presumably it denotes the corresponding weak-homogeneity class E^c_γ or O^c_γ from Section 1.
- [Proof of Lemma 1.16] The symbol X is used both for the projection of a point in R^{m'+n} onto the vertical factor and later for the cutoff function χ pulled back to the bundle; this overloaded notation makes the argument harder to follow.
- [Throughout] There are numerous OCR-type artifacts in the LaTeX source, such as 'R∗+ ↷' and stray bracket fragments, which should be cleaned up before any resubmission.
Circularity Check
No significant circularity: the main extension theorem is derived from classical results and the paper's own lemmas, and the disputed compact-support assertion is a correctness gap rather than a circular reduction.
full rationale
The derivation chain is self-contained and non-circular. Lemma 3.2 characterizes a-homogeneous distributions on R^*_+ x S_V x R by a distribution w on S_V x R, and this is proved directly from the homogeneity definition. Theorem 3.3 constructs the extension from w under a weak-homogeneity estimate for the time action, using explicit integration-by-parts formulas. Theorem 3.4 then invokes the Schwartz-based decomposition of Lemma 1.16, which is proved from Schwartz's classical theorem, and applies Theorem 3.3 to each weakly homogeneous summand. No fitted parameter is renamed as a prediction, and no conclusion is obtained by assuming the target extension exists. The citations to the author's advisors ([LMV15], [Moh20], [Moh24]) concern auxiliary notions (pi-properly supported functions, iterated deformations, and the Wodzicki residue discussion) and are not load-bearing for the extension theorem; the core tools are Hormander's and Schwartz's classical results and Meyer's weak-homogeneity theorems. The one serious flaw is external to circularity: the final paragraph asserts 'We have that every wi is compactly supported, thus a tempered distribution' after taking a partition of unity on S_V, but the preceding argument establishes at most proper support of w for the projection S_V x R -> S_V, and a partition of unity on the compact sphere directions does not localize the noncompact R-factor. This is a substantive mathematical gap in the advertised unconditional theorem, but it is not a circular step: the extension is not shown to be equivalent to its inputs by definition, and no parameter is fitted. Therefore the circularity score is 0.
Assumptions & free parameters
assumptions (4)
- standard math Meyer's extension theorem for weakly homogeneous distributions (Theorem 1.9 and 1.10 from [MS97])
- standard math Hormander's classification of homogeneous distributions on the real line (Proposition 1.4 and Corollary 1.5.1)
- standard math Schwartz's characterization of tempered distributions (Lemma 1.15 from [Sch66])
- domain assumption Local triviality and the global exponential chart assumption (Section 2)
Cite this review
Pith. "Pith review of Extensions of homogeneous distributions on deformations to the normal cone." pith.science (2026). https://pith.science/paper/HZT4WFXH
@misc{pith2026250522885,
author = {Pith},
title = {Pith review of: Extensions of homogeneous distributions on deformations to the normal cone},
year = {2026},
howpublished = {\url{https://pith.science/paper/HZT4WFXH}},
note = {Machine review of arXiv:2505.22885}
}
abstract
On a deformation to the normal cone $\operatorname{DNC}(M,V)$ we show that given a distribution $u\in\mathcal{D}'(\operatorname{DNC}(M,V)\setminus V\times\mathbb{R})$ if $u$ is homogeneous of order $a$ for the zoom action, then it admits an $a$-homogeneous extension $\widetilde{u}\in\mathcal{D}'(\operatorname{DNC}(M,V))$. We describe all such extensions and discuss briefly about how it translates to the work of Van Erp and Yuncken in arXiv:2303.15787 . The technique used come from the results on the extension of weakly homogeneous distributions provided by Yves Meyer in the 90s.
Reference graph
Works this paper leans on
-
[1012]
Adiabatic groupoid, crossed product by $\R_+^*$ and Pseudodifferential calculus
doi: 10.1112/blms.12096. url: https://doi.org/10.1112%2Fblms. 12096. [DS13] C. Debord and G. Skandalis. Adiabatic groupoid, crossed product by R∗ + and Pseudodifferential calculus. 2013. arXiv: 1307.6320 [math.FA]. [Con94] A. Connes. Noncommutative Geometry. Elsevier Science, 1994.isbn: 9780121858605. url: https://books.google.de/books?id=SDot3ObNQwIC. [M...
work page Pith review arXiv 1977
-
[2017]
A groupoid approach to pseudodifferential operators
arXiv: 1511.01041 [math.DG]. [MS97] Y. Meyer and A. M. Society. Wavelets, Vibrations, and Scalings . CRM mono- graph series. American Mathematical Society, 1997.isbn: 9781470438555. url: https://books.google.fr/books?id=3-citAEACAAJ. [EY17b] E. van Erp and R. Yuncken. “On the tangent groupoid of a filtered manifold”. In: Bulletin of the London Mathematica...
work page Pith review arXiv 1997
Reviewed August 7, 2026 · model on record in the stance chip above.
Discussion (0). Sign in to comment.