Pith. sign in

REVIEW 3 major objections 3 minor 10 references

Subtlety of oscillation indices of oscillatory integrals of real analytic functions

T0 review · 3 major / 3 minor · reviewed 2026-08-03 · deepseek-v4-flash

Pith's one-line read For a class of even-dimensional real analytic phases, the oscillation index is strictly below the negative of the real log canonical threshold, contradicting a standard formula in the literature.

desk verdict Theorem 3 is false — f(x,y)=x^3+y^3 gives β=-2/3, so the claimed strict inequality β<-γ fails; the proof drops the absolute value in the Jacobian. read the letter →

arxiv 2511.16257 v3 pith:3GFIHVZA submitted 2025-11-20 math.CV math.AG

classification math.CVmath.AG MSC 14B0532S2542B20
keywords oscillationindexoscillatoryintegralreallogcanonicalthresholdNewtonnondegenerateanalyticfunctionasymptoticexpansionhomogeneouspolynomialblowup
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

This paper studies when the oscillation index β(f) of an oscillatory integral equals the negative of the real log canonical threshold, −γ(f). For Newton R-nondegenerate convenient homogeneous functions in an even number of variables, the authors prove that equality fails — β(f) is strictly smaller than −n/d — when the degree exceeds the dimension, or when the zero set is just the origin with even degree. This conflicts with a widely used formula from the literature, and the authors explicitly acknowledge that the discrepancy means some error exists somewhere, though they do not identify whether it lies in their proof or in the earlier formula. If the paper is right, the oscillation index carries finer information than the real log canonical threshold and depends on parity in a way previously missed.

What carries the argument

The key machinery is a blowup of the origin combined with the Newton polytope of f. The oscillation index β(f) is the leading exponent in the asymptotic expansion of ∫ e^{iτf}φ dx, while the real log canonical threshold γ(f) is the minimal real jumping number, computable from a resolution of singularities; for homogeneous degree d it equals n/d. Newton R-nondegenerate means the principal part on each face of the Newton polytope has no nonzero real critical points away from coordinate hyperplanes, and 'convenient' means the polytope meets every coordinate axis. The decisive calculation is the inner integral over the exceptional coordinate y: for even n and d, the integrand y^{n-1} e^{iτy^d h(

What would settle it

Compute the asymptotic expansion of ∫ e^{iτ(x^2+y^2)} φ dx dy for a bump φ supported near 0 with φ(0) ≠ 0. Stationary phase gives a nonzero coefficient of τ^{-1}, so β = −1 = −γ, directly violating the strict inequality β < −1 that Theorem 3 predicts for this phase.

Watch

Extended reading notes

Core claim

The central claim is Theorem 3: assume f is Newton R-nondegenerate, convenient, and homogeneous of degree d, with n even and either d > n or f^{-1}(0) = {0} (and d even). Then the real log canonical threshold is γ(f) = n/d, but the oscillation index satisfies the strict inequality β(f) < −n/d. The proof uses the blowup of the origin, where the integral decomposes over charts; when d and n are even, the integral along the exceptional coordinate is an integral of an odd function and vanishes, and a symmetry argument with x ↦ −x extends the conclusion to odd d in the zero-set-origin case. The paper also notes that in the odd-dimensional case with f^{-1}(0) = {0}, equality holds, so the even/odd

Load-bearing premise

The authors' proof of Theorem 3 is sound, and the contradiction with the earlier formula is due to an error in that formula rather than in this paper; if the standard formula is correct, Theorem 3 is false.

Editorial extensions

If this is right

  • If Theorem 3 holds, the decay of oscillatory integrals in these homogeneous cases is faster than the exponent suggested by the real log canonical threshold alone.
  • The strict inequality means any formula asserting β(f) = −γ(f) for all Newton R-nondegenerate convenient functions must be restricted by dimension parity and by the structure of the zero set.
  • The paper's Remark 3.1 indicates that in the covered cases one can in fact derive β(f) = −(n+1)/d from the same Newton-polyhedron techniques, providing a concrete corrected value.
  • The odd-dimensional singleton-zero-set case preserves equality, so the phenomenon is a parity effect, not a general failure of the bound.
  • The sketched extension to nonnegative non-homogeneous functions suggests that parity of certain exponents on Newton faces controls whether equality holds there as well.

Reading between the lines

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

  • If the strict inequality is genuine, the asymptotic expansion in these cases has no term at the threshold exponent, a property that could be checked numerically for low-degree phases and would immediately distinguish the two competing formulas.
  • The quadratic phase x^2 + y^2 on R^2 (n=2, d=2, zero set the origin) is the sharpest test case: standard stationary phase gives β = −1 = −γ, which appears to contradict Theorem 3; re-evaluating this one example with the paper's blowup method could reveal whether the error lies in the older formula or in the paper's own argument.
  • The parity phenomenon connects to a general principle: odd/even symmetry of the phase under negation can annihilate leading-order contributions to oscillatory integrals, a mechanism that may apply to other symmetric phase families.
  • The authors' inability to locate the error suggests the contradiction is subtle; a systematic comparison between the blowup computation and Newton-polytope estimates on a single explicit example would settle which side is mistaken.
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.

Referee Report

3 major / 3 minor

Summary. The paper studies the relation between the oscillation index β(f) of an oscillatory integral and the real log canonical threshold γ(f)=rlct(f) for a real analytic phase f. It recalls Theorem 1 (γ(f)=rlct(f)) and Theorem 2 (β(f)≤−γ(f)), then states Theorem 3: if f is Newton R-nondegenerate, convenient, homogeneous, n is even, and either deg f>n or f^{-1}(0)={0} with d even, then γ(f)=n/d and β(f)<−n/d. The proof in Section 3 uses blowups, a partition of unity, and a parity/cancellation argument in the coordinates of (3.1). The abstract and introduction explicitly acknowledge that the strict inequality seems incompatible with standard formulas and that 'there must be some error somewhere', without locating it.

Significance. If Theorem 3 were correct, it would identify a new class of phases where Varchenko's bound is strict, contradicting standard Newton-polyhedron formulas; that would be a substantial result. The paper also contains useful review material, and Corollaries 2.1–2.2 on independence of cutoff and of the Taylor expansion of the amplitude are interesting. However, the central theorem is false: f=x^3+y^3 is a direct counterexample satisfying all hypotheses of Theorem 3. The parity proof rests on a signed-form/density error, and the paper's own admission of an unlocated contradiction, together with the conceded non-cancellation in Remark 3.2, means the main claim is not established and cannot be repaired by a small local correction.

major comments (3)
  1. [Theorem 3 / §3] Theorem 3 is false as stated. Let n=2 and f(x,y)=x^3+y^3. This is homogeneous of degree d=3>2, convenient, and Newton R-nondegenerate: for the face x^3+y^3, ∂f=0 gives (0,0)∈{xy=0}; for the vertex faces the critical sets are {x=0} or {y=0}, also contained in {xy=0}. For any φ∈C_c^∞ with φ(0)≠0, the scaling u=τ^{1/3}x, v=τ^{1/3}y gives I(τ,φ)=τ^{-2/3}∫ e^{i(u^3+v^3)} φ(τ^{-1/3}u,τ^{-1/3}v) dudv → τ^{-2/3} φ(0) C, where C=(∫_{-∞}^{∞} e^{it^3}dt)^2=Γ(1/3)^2/3≠0. Hence β(f)≥−2/3. Since (3.2) gives γ(f)=n/d=2/3 and Theorem 2 gives β(f)≤−2/3, we obtain β(f)=−2/3, contradicting the strict inequality in (6).
  2. [§3, Eq. (3.5)] The parity/cancellation step is invalid because π^*dx is a signed n-form, not a density. The oscillatory integral (1) is an integral of the density |π^*dx|. In the chart (3.1), for n even the factor y_i^{n-1} changes sign under y_i↦−y_i, and π is orientation-reversing on {y_i<0}. Equation (3.5) integrates the signed form y_i^{n-1}dy_i∧..., so oddness in y_i can make the integral zero, but the original integral uses |y_i|^{n-1}. The correct integrand is even in y_i, and no cancellation follows. This is precisely the mechanism in the counterexample f=x^3+y^3.
  3. [§3, d odd case] For odd d the argument is also not valid. The text says to apply the previous argument to the real part and then to show vanishing of the leading imaginary terms using the involution ι with ι^*x_i=−x_i. For n even, ι^*dx=dx; after symmetrizing χ′, I(τ,χ′)=∫ cos(τf)χ′ dx. This real part is not zero. In the counterexample f=x^3+y^3, the real part of the leading term is φ(0)C with C>0. The sentence in the same paragraph that a symmetry of f does not imply vanishing does not repair the gap, because the surviving term is a cosine integral over the density, not an odd integral.
minor comments (3)
  1. [Abstract/Introduction] The text states that 'there must be some error somewhere, although it does not seem easy to find it inside this paper.' This is an explicit acknowledgment that the contradiction with [Va76]/[CKN13] is not resolved. A refereed theorem should either identify the error in the cited literature or be presented conditionally.
  2. [Remark 3.1] Remark 3.1 suggests that under the hypotheses of Theorem 3 one could get β(f)=−(n+1)/d from [CKN13, Section 7.1]. This is inconsistent with Theorem 3 and false for f=x^3+y^3, where β(f)=−2/3. The paper should reconcile these assertions.
  3. [Remark 3.2] Remark 3.2 concedes that without a nonnegativity assumption 'it does not seem easy to show the non-cancellation among the integrals associated with different faces'. Since Theorem 3 does not assume nonnegativity, this is a gap in the odd-d part of the proof that is acknowledged in the text.

Circularity Check

0 steps flagged · score 0.0 of 10

No load-bearing circularity: the paper's derivation imports prior theorems and computes the claimed oscillation-index bound, rather than assuming or fitting it.

full rationale

The derivation chain in Theorem 3 is not circular. It computes gamma(f) = n/d from the definition of the real log canonical threshold via a resolution of singularities, imported as Theorem 1 from Saito's earlier arXiv paper [Sa07]; it then applies the external Newton-polyhedron bound from [CKN13, Theorem 2.1] to reduce the desired strict inequality beta(f) < -n/d to a vanishing/parity calculation in Section 3. No parameter is fitted to the target value, no quantity is defined in terms of the claimed conclusion, and the conclusion beta(f) < -n/d is not an input to any cited theorem. The only author-overlapping citation, [Sa07], is a prior theorem with independent content (rlct(f) = gamma(f)) and is not invoked to forbid alternatives or to smuggle in the strict inequality. The paper's own admission that its result 'does not seem compatible' with standard formulas in [Va76] and [CKN13] is a correctness concern, not circularity: it indicates possible mathematical error, but not that the reasoning reduces to its inputs by construction. Therefore the circularity score is 0.

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

No free parameters are fitted and no new entities are introduced. The central derivation depends on four imported theorems and one parity/cancellation assertion the authors do not fully justify. The unresolved contradiction with the standard formula is the main burden.

assumptions (5)
  • standard math Hironaka resolution of singularities exists and gives monomial π*f and Jacobian (used for asymptotic expansion (2) and γ(f)).
    Invoked around eqs. (2)-(3); standard theorem assumed without proof.
  • standard math rlct(f) = γ(f) (Theorem 1 of [Sa07]).
    Used at (3.2) to identify γ(f) = n/d; prior theorem by one of the authors, not the target result.
  • standard math β(f) ≤ −γ(f) (Theorem 2 of [Va76]).
    Baseline inequality; assumed from the literature.
  • standard math β(f, φ) ≤ −1/d(f, φ) (Theorem 2.1 of [CKN13]).
    Used for the reduction to β(f, χ') < −n/d.
  • ad hoc to paper In the d odd case, the leading imaginary-part terms cancel under x → −x and no relevant non-cancellation occurs.
    The proof of Theorem 3 relies on this; Remark 3.2 admits non-cancellation is not easy to show in related settings.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Subtlety of oscillation indices of oscillatory integrals of real analytic functions." pith.science (2026). https://pith.science/paper/3GFIHVZA

@misc{pith2026251116257,
  author       = {Pith},
  title        = {Pith review of: Subtlety of oscillation indices of oscillatory integrals of real analytic functions},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/3GFIHVZA}},
  note         = {Machine review of arXiv:2511.16257}
}
abstract

For a locally defined real analytic function $f$, we study the relation between the oscillation index of oscillatory integrals and the real log canonical threshold. The former is always negative, and its absolute value is greater than or equal to the latter. They coincide very often, but there are certain exceptional cases, and it is not very clear when the equality holds. In this note we give some sufficient conditions for the coincidence to hold or to fail. In the Newton-nondegenerate convenient homogeneous case, we show that the strict inequality holds if the number of variables $n$ is even and smaller than the degree $d$ of $f$ (or $f^{-1}(0)=\{0\}$), and the equality holds if $n$ is odd and $f^{-1}(0)=\{0\}$ (in particular, $d$ is even). The first assertion does not seem to be compatible with some standard formula in the literature, and there must be some error somewhere, although it does not seem easy to detect it inside this paper.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

10 extracted references · 2 linked inside Pith

  1. [1]

    Arnold, V.I., Gusein-Zade, S.M., Varchenko, A.N., Singularities of Differentiable Maps II, Birk\-h\"auser, Boston, 1988

  2. [2]

    Budur, N., Saito, M., Multiplier ideals, V -filtration, and spectrum, J.\ Alg.\ Geom.\ 14 (2005), 269--282

  3. [3]

    Cho, K., Kamimoto, J., Nose, T., Asymptotic analysis of oscillatory integrals via the Newton polyhedra of the phase and the amplitude, J. Math. Soc. Japan 65 (2013), 521--562 (or arxiv:1203.0808)

  4. [4]

    Decker, W., Greuel, G.-M., Pfister, G., Sch\"onemann, H., Singular 4.3.2 --- A computer algebra system for polynomial computations, available at http://www.singular.uni-kl.de (2023)

  5. [5]

    Hironaka, H., Resolution of singularities of an algebraic variety over a field of characteristic zero, Ann.\ Math.\ 79 (1964), 109--326

  6. [6]

    Jung, S.-J., Kim, I.-K., Saito, M., Yoon, Y., Hodge ideals and spectrum of isolated hypersurface singularities, Ann.\ Inst.\ Fourier 72 (2022), 465--510

  7. [7]

    62 Part 1, (1997), 221--287

    Koll\'ar, J., Singularities of pairs, Proc.\ Symp.\ Pure Math., A.M.S. 62 Part 1, (1997), 221--287

  8. [8]

    (4), 7 (1974), 405--430

    Malgrange, B., Int\'egrales asymptotiques et monodromie, Ann.\ Sci.\ Ecole Norm.\ Sup. (4), 7 (1974), 405--430

Show all 10 references
  1. [9]

    Saito, M., On real log canonical thresholds, arxiv:0707.2308

  2. [10]

    10 (1976), 175--196

    Varchenko, A.N., Newton polyhedra and estimation of oscillating integrals, Funct.\ Anal.\ Appl. 10 (1976), 175--196

Pith tools

Reviewed August 3, 2026 · model on record in the stance chip above.