REVIEW 2 major objections 6 minor 1 cited by
The Nikolskii Constant in Arbitrary Dimension
T0 review · 2 major / 6 minor · reviewed 2026-08-27 · deepseek-v4-flash
Pith's one-line read The paper proves that the sharp Nikolskii constant in every dimension reduces to a one-dimensional spectral problem, via a factorization of the extremal function into two entire factors of half the exponential type.
desk verdict New even-dimensional machinery and the first d=2 value, but the proof leans on imported facts about the extremal function that still need checking. 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 load-bearing mechanism is the Cauchy transform of the derivative of a modified sign function. One sets $\tilde\sigma(x)=\mathrm{sign}(x)^{\epsilon}\sigma(x)$ (with $\epsilon=0$ for odd $d$ and $1$ for even $d$) so that $|x|^{d-1}\sigma=x^{d-1}\tilde\sigma$ is a monomial; then $\nu=\tilde\sigma'$ is a discrete measure supported at the zeros $\tau_k$. Its Cauchy transform $M(z)=\mathrm{p.v.}\int d\nu(t)/(z-t)$ has boundary values whose Fourier transforms vanish on $(-1,1)$ after a carefully chosen polynomial correction $R_d$. The function $\Theta_d(z)=z^dM(z)-2\epsilon z^{d-1}+4R_d(z)/z$ therefore has a spectral gap, and the defect $H_d=\varphi\Theta_d-4a_d/z$, being entire with spectrum concentrated at the origin, is a polynomial; a classical sampling inequality for bandlimited functions forces it to vanish identically. That vanishing is exactly the functional identity (16), and differentiating it pins down the differential system.
What would settle it
Compute the zeros $\tau_k$ for $d=2$ by an independent numerical method, form $\Phi_1,\Phi_2$ from (14), and evaluate identity (16) at a regular point such as $z=1.7$. Any difference between the two sides beyond round-off error would falsify the central claim.
Extended reading notes
Core claim
The central discovery is Theorem 1: for every $d\ge1$, the radial extremal function $\varphi$ of the extremal problem splits as $\varphi=\Phi_1\Phi_2$, where the factors are entire of exponential type $1/2$ and, for all $z\in\mathbb C$, satisfy $z^{d+1}(\Phi_1'(z)\Phi_2(z)-\Phi_1(z)\Phi_2'(z))-2R_d(z)\Phi_1(z)\Phi_2(z)=-2a_d$, with $a_d=1/(2v_d C_d)$ and $R_d$ an even polynomial built from the distributional data of the sign of $\varphi$. The same identity, differentiated, yields the second-order system (18) for $\Phi_1,\Phi_2$; after an exponential gauge change the system becomes a single one-dimensional equation $u''+V_d(x)u=0$ on the positive half-line, with a rational potential $V_d$ fixed by at most $d+1$ parameters. Thus the sharp constant is encoded in the spectral data of one ordinary differential equation, and the paper works out the new case $d=2$ numerically.
Load-bearing premise
Everything rests on the previously established profile of the extremal function: it is the unique radial minimizer, entire of exact exponential type one, bounded by one on the real line, with simple positive zeros spaced roughly by $\pi$, and it satisfies a prescribed balance condition after Fourier transformation; the present paper uses this input but does not prove it.
Editorial extensions
If this is right
- For every fixed $d$, the sharp constant $C_d$ is determined by the one-dimensional spectral problem (19); no search over the whole space $\mathrm{PW}_1^1(\mathbb R^d)$ remains.
- The factors $\Phi_1,\Phi_2$ have exact exponential type $1/2$ and decay as $|x|^{-(d+1-\epsilon)/2}$ and $|x|^{-(d+1+\epsilon)/2}$, so the extremal function itself decays like $|x|^{-(d+1)}$.
- The coefficients of $R_d$ are regularized alternating power sums over the zeros: $\eta_{\tau,d}(-d+2j)=r_j$, with $\eta_{\tau,d}(0)=1/2$.
- The zero equilibrium condition $a_d=\tau_k^d\prod_{j\ne k}|1-\tau_k^2/\tau_j^2|$ gives an explicit finite-dimensional algorithm for the zeros and hence for $C_d$, valid uniformly in $d$.
- For $d=2$ the method yields the new numerical value $L^*(2)=2/a_2=0.28214506418970912412\ldots$.
Reading between the lines
- The author leaves implicit that the spectral reduction can be used to test the conjectured large-dimension asymptotics $L^*(d)=2^{-d}(1+o(1))$ and the empirical refinement $2^dL^*(d)=\pi/2-cd^{-1/3}$ with $c\approx0.777$ by solving equation (19) numerically across a range of $d$.
- A natural testable extension is whether the same factorization and spectral gap appear when the weight $|x|^{d-1}$ is replaced by other monomial-compatible weights; the construction in Section 3 uses only the polynomial character of the weight after the signature trick.
- The zeta formula $\eta_{\tau,d}(-d+2j)=r_j$ suggests that the dependence of the constant on $d$ might be studied by analytic continuation in the dimension, a route the paper does not itself pursue.
Signed reviews
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper studies the radial extremal function φ for the sharp Nikolskii constant C_d in PW_1^1(R^d). It claims Theorem 1: φ factors as Φ1Φ2, with Φ1 and Φ2 satisfying the functional identity (16) and the second-order differential equations (17)-(18); this reduces the original extremal problem to a one-dimensional spectral problem depending on at most d+1 parameters. The paper also derives a zeta interpretation of the coefficients of R_d, sign properties and a zero equilibrium condition, and it computes the case d=2 numerically. The main proof proceeds by replacing the weight |x|^{d-1} with x^{d-1}σ̃(x), forming the Cauchy transform M of ν=σ̃′, constructing Θ_d with one-sided Fourier support, proving the defect H_d vanishes by a Plancherel-Pólya argument, and then extracting the differential equations by zero-location and coefficient comparison.
Significance. If Theorem 1 is established, it is a substantial advance: it unifies the odd- and even-dimensional treatments, gives a concrete spectral route to the sharp constants, and provides new structural information about the extremal function (factorization, a third-order equation, and a zero equilibrium condition). The internal algebra is detailed and largely checkable: the cancellation of the defect in §3.5-3.7, the Plancherel-Pólya step, and the degree/leading-coefficient argument in §4.5-4.6 are coherent. The explicit d=2 computation is a useful concrete demonstration. The central weakness is not an internal contradiction but the paper's reliance on imported, unproved profile and extremality facts for φ, so the true scope of the theorem is conditional on the status of those inputs.
major comments (2)
- [§1.1-1.2, Eq. (3)] The proof of Theorem 1 is conditional on the variational extremality condition (3), the canonical product (2), and the zero asymptotics τ_k∼πk, none of which are proved in this manuscript. These facts are taken from [2] and [6], and (3) is the only step connecting the minimization problem (1) to the distribution identity (4) that generates P_d and hence the entire spectral construction. Since the manuscript does not state these inputs as explicit hypotheses of Theorem 1, the theorem as stated is not self-contained. The paper must either prove these facts, or state them as assumptions, or give precise statements of the theorems in [6] that establish them. In particular, because [6] is titled for odd dimensions, the manuscript must specify whether (3) and the zero profile are proved in [6] for all d; if they are only proved for odd d, then the even-dimensional cases, including the d=2 computation in Section 6, are not covered by Theorem 1 as stated.
- [§4.5-4.6, Eq. (56) and Lemma 9] The proof that deg K_d=d+1 and that its leading coefficient is 1/4 depends on the zero-counting relation n_{u1}(X)=X/(2π)+o(X), which is imported from τ_k∼πk, and on Lemma 9, which is quoted from [6]. These are parts of the same external dependency identified above. If the zero asymptotics or Lemma 9 are not available for every d, the degree and leading coefficient of K_d are not established, and part (b) of Theorem 1 does not follow. The manuscript should state these imports precisely or move them into an appendix with proofs.
minor comments (6)
- [§6, Table and numerical computation] The numerical values for a2, β, C2, and L*(2) are quoted at 120-digit precision, but no code, stopping criterion, or rigorous error bound is supplied; the use of Newton's method and Miller's algorithm is described qualitatively. A reproducibility appendix or a statement of the convergence criterion would strengthen the computational claim.
- [§7, Eq. (70)-(72)] The zeta interpretation is presented as a consequence, but the key asymptotic expansion (70) is asserted with 'some integer L≥0' and the analytic continuation step is sketched rather than proved in detail. Since Section 7 is advertised as a main result in the introduction, the proof of (70) and of the continuation should be written out or the section should be labeled as a sketch.
- [§8.3, Eq. (78)] The asymptotic statements in Section 8.3 are not theorems of this paper: L*(d)=2^{-d}(1+o(1)) is announced for a separate paper, and (78) is supported only by numerical experiments. Please label these as conjectures or future work in the text.
- [§8.3, figure] The graph in Section 8.3 has no axis labels or caption; as printed it is not self-contained. Please add a proper caption and axis labels.
- [Abstract and §1.6] The abstract says the spectral problem depends on 'at most d+1 parameters', while Section 1.6 gives d+1-ϵ parameters (d parameters when d is even). Please reconcile this discrepancy.
- [Lemma 1 proof] The sentence 'Thustheidentityisprovedforeventestfunctions' contains typographical spacing errors and should read 'Thus the identity is proved for even test functions.'
Circularity Check
The derivation chain is internally coherent and does not assume Theorem 1; the only caveat is that the arbitrary-dimension proof is conditional on self-cited prior facts (the extremality condition (3) and the zero profile of φ) imported from [6]/[2].
-
self citation load bearing
[Sections 1.1–1.2, Eq. (3); used in §3.2 and §4.5–4.6]
"In [6], we extended the results of [1] to odd dimensions. We will make frequent use of the notation and results of [6]. In particular, the problem (1) has a unique radial extremal function φd(|·|); see also [2]."
This passage is the only stated source for the variational extremality condition (3) and for the entire-type-1 profile with τk∼πk. Equation (3) is then used to derive (4) via Lemma 1, and (4) together with (7) and (27) produces P_d, R_d, Θ_d, and finally the functional identity (16). The zero asymptotics τk∼πk is used essentially for the convergence of M in §3.3, for the separated subsequence in §3.6, and to fix deg K_d=d+1 and c=1/4 in §4.5–4.6. The present paper does not re-derive these premises, so Theorem 1 is conditional on the same-author preprint [6]. This is a load-bearing dependency rather than an internal circular use of the target identity; the subsequent algebra is not circular given the imported premises.
full rationale
The paper's main contribution is a reduction, not a prediction of a new constant from independent data. The factorization φ=Φ1Φ2 is a construction from the zeros of φ, and identity (16) is derived through the Cauchy transform and spectral-gap argument, not assumed. The differential system (18) follows from (16) and the zero relations, and the polynomial K_d is identified with the explicitly defined K_d in (11) after determining its degree and leading coefficient. The one-dimensional spectral problem is an equivalent reformulation whose parameters p_n and λ_q are defined in terms of φ, so it is self-consistent rather than a parameter-free prediction. The only circularity-adjacent issue is the load-bearing import of the extremality condition (3) and the zero asymptotics from [6]/[2]; those are prior results cited rather than conclusions of this paper, so the internal derivation does not reduce to its own target. Score 2 reflects this dependency without treating it as a logical loop.
Assumptions & free parameters
free parameters (3)
- a_d (target constant; C_d = 1/(2 v_d a_d)) =
d = 2: 7.08855214513069412139...
- lambda_q, q = 0..N (alternating power sums over zeros) =
d = 2: lambda_0 = beta/(2 a_2) = -0.02149...
- p_n, n = 0..N (coefficients of P_d) =
d = 2: p_0 = 2 a_2; p_N from (29) for general d
assumptions (4)
- domain assumption Existence, uniqueness, and analytic profile of the radial extremal function phi: exact exponential type 1, |phi| <= 1, canonical product (2), zeros tau_k ~ pi k.
- domain assumption Variational extremality condition (3) and distribution identity (4).
- standard math Lemmas 4-7 from [6]: derivatives of truncated powers, support of a product, Plancherel-Polya inequality, and the logarithmic-derivative polynomial lemma.
- standard math Paley-Wiener-Schwartz theorem and standard distribution Fourier formulas.
Cite this review
Pith. "Pith review of The Nikolskii Constant in Arbitrary Dimension." pith.science (2026). https://pith.science/paper/Q7DRCP4E
@misc{pith2026260822578,
author = {Pith},
title = {Pith review of: The Nikolskii Constant in Arbitrary Dimension},
year = {2026},
howpublished = {\url{https://pith.science/paper/Q7DRCP4E}},
note = {Machine review of arXiv:2608.22578}
}
abstract
We study the radial extremal function $\varphi(|{\,\cdot\,}|)$ arising in the problem of finding the sharp Nikolskii constant $\mathcal C_d$ in $\mathit{PW}_1^1(\mathbb R^d)$ for arbitrary dimension $d\ge1$. We prove a factorization $\varphi=\Phi_1\Phi_2$, where $\Phi_1$ and $\Phi_2$ are entire functions of exponential type $1/2$ satisfying a functional equation and second-order differential equations with polynomial coefficients. As a result, the original extremal problem is reduced to a one-dimensional spectral problem depending on at most $d+1$ parameters. We also obtain a zeta interpretation of the coefficients of the polynomial appearing in the functional equation and a multiplicative equilibrium condition for the zeros of the extremal function. These results can be used to construct several algorithms for computing $\mathcal C_d$.
Forward citations
Cited by 1 Pith paper
-
Exact Asymptotics of the Multidimensional Nikolskii Constant
As d tends to infinity, 2^d times the normalized L1 Nikolskii constant converges to π/2, and the rescaled zeros of the extremal function converge to an explicit Bessel-type measure.
Reference graph
Works this paper leans on
-
[2]
F. Dai, D. Gorbachev, S. Tikhonov,Estimates of the asymptotic Nikolskii constants for spherical polynomials, J. Complexity65(2021), 101553
work page 2021
-
[6]
D. V. Gorbachev,The Nikolskii constant in odd dimensions, arXiv:2608.15674, 2026. Saint Petersburg State University Email address:dvgmail@mail.ru
work page Pith review arXiv 2026
-
[1]
A. Bondarenko, J. Ortega-Cerdà, D. Radchenko, K. Seip,The Hörmander–Bernhardsson extremal function, arXiv:2504.05205v2
-
[3]
Gautschi,Computational aspects of three-term recurrence relations, SIAM Rev.9(1967), 24–82
W. Gautschi,Computational aspects of three-term recurrence relations, SIAM Rev.9(1967), 24–82
work page 1967
-
[4]
I. M. Gel’fand and G. E. Shilov,Generalized functions. Vol. 1: Properties and operations, Academic Press, New York–London, 1964. 22 D. V. GORBACHEV
work page 1964
-
[5]
D. V. Gorbachev,Sharp Bernstein–Nikolskii inequalities for polynomials and entire functions of exponential type, Chebyshevskii Sb.22(2021), no. 5, 58–110. (In Russian)
work page 2021
Reviewed August 27, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.