On the shape of the typical Poisson-Voronoi cell in high dimensions

Matthias Irlbeck; Tobias M\"uller; Zakhar Kabluchko

REVIEW 2 major objections 4 minor 23 references

On the shape of the typical Poisson-Voronoi cell in high dimensions

T0 review · 2 major / 4 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read In high dimensions, the typical Poisson-Voronoi cell has inradius half its outradius, diameter twice its outradius, and all four basic size measures converge to explicit constants under a natural scaling.

desk verdict Strong, likely-correct main results, but two supporting lemmas—one false as stated—need fixing before the face-count theorems are reliable. read the letter →

arxiv 2506.02607 v1 pith:IMJWKIMI submitted 2025-06-03 math.PR

Matthias Irlbeck , Zakhar Kabluchko , Tobias Müller This is my paper

classification math.PR MSC 60D0560G55

keywords Poisson-Voronoitessellationtypicalcellhighdimensionsinradiusoutradiusdiametermeanwidthfacecounts

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 asks what the typical cell of a Poisson-Voronoi tessellation looks like when the dimension is large. It proves that after dividing by $\left(\lambda\,\mathrm{vol}(B)\right)^{1/d}$, the cell's inradius, outradius, diameter, and mean width converge in probability to $1/2$, $1$, $2$, and $2$ as the dimension grows. Because the inradius is half the outradius and the diameter is twice the outradius, the limiting shape is far from a ball: the Hausdorff distance from the cell to any ball is at least a constant fraction of the cell's diameter. The paper also shows that faces of non-constant codimension have vanishingly small diameter, while the number of faces of fixed codimension $k$ grows like a constant to the $d$-th power, and almost all such faces are defined by near-regular simplices.

What carries the argument

The central mechanism is the representation of the typical cell as an intersection of half-spaces, $V_{\mathrm{typ}}=\bigcap_{z\in Z}\{x: x\cdot z \le \|z\|^2/2\}$, together with the scaling choice $\lambda=1/\kappa_d$ that makes the expected number of Poisson points in a radius-$s$ ball equal to $s^d$. Two identities carry the argument: the polar duality $V_{\mathrm{typ}}^{\circ}=\mathrm{conv}(f[Z])$ with $f(z)=2z/\|z\|^2$, which converts width into radial distances of the polar hull, and the Blaschke-Petkantschin formula, which reduces face counts to an integral over $k$-tuples of vectors weighted by a determinant power. The determinant's maximum is governed by the extremal fact that regular simplices maximize volume among simplices inscribed in the unit sphere, which yields the explicit constant in the face-count theorem.

What would settle it

Run a high-dimensional simulation of the Poisson-Voronoi tessellation at $d\ge 500$ and examine the normalized vertex norms: if a vertex with norm outside $1\pm\varepsilon$ appears with non-negligible frequency, or if a ball of radius $(1/2+\varepsilon)(\lambda\,\mathrm{vol}(B))^{1/d}$ fits inside the typical cell, Theorem 1 fails. Alternatively, an explicit counterexample to the inscribed-simplex volume extremum would change the face-count constant in Theorem 6.

Watch

Extended reading notes

Core claim

Under the rescaling by $\left(\lambda\,\mathrm{vol}(B)\right)^{1/d}$, the typical Poisson-Voronoi cell $V_{\mathrm{typ}}$ satisfies $\mathrm{outr}(V_{\mathrm{typ}})\to 1$, $\mathrm{inr}(V_{\mathrm{typ}})\to 1/2$, $\mathrm{diam}(V_{\mathrm{typ}})\to 2$, and $\mathrm{meanw}(V_{\mathrm{typ}})\to 2$ in probability as $d\to\infty$, for any intensity $\lambda=\lambda(d)>0$. The width is shown to lie between $2\sqrt{5}/(2+\sqrt{5})-o_d(1)$ and $3/2+o_d(1)$ with probability $1-o_d(1)$. Consequently the cell is asymptotically not close to any ball in Hausdorff distance, even though its volume concentrates at the value that a ball of the same scale would have. All vertices concentrate on a sphere of radius $1\pm o_d(1)$, all faces of super-constant codimension have diameter $o_d(1)$ on the chosen scale, and the number of faces of fixed codimension $k$ satisfies $\sqrt[d]{f_{d-k}(V_{\mathrm{typ}})}\to \left((k+1)^{(k+1)/2}/k^{k/2}\right)$ in probability.

Load-bearing premise

The argument leans on two cited, unproved facts: that regular simplices maximize volume among all simplices inscribed in a unit sphere, and that the polar of the typical cell equals the convex hull of the inverted Poisson process.

Editorial extensions

If this is right

With probability $1-o_d(1)$, the typical cell's inradius is $1/2\pm o_d(1)$ times its outradius, so the cell is asymptotically far from any ball in Hausdorff distance.
Every face whose codimension grows with $d$ lies in the annulus between radii $1-\varepsilon$ and $1+\varepsilon$ (on the scale $\left(\lambda\,\mathrm{vol}(B)\right)^{1/d}$) and has diameter $o_d(1)$ of that scale.
The $d$-th root of the number of $(d-k)$-faces converges in probability to $((k+1)^{(k+1)/2}/k^{k/2})$; for facets ($k=1$) this means about $2^d$ facets.
Almost all faces of fixed codimension $k$ are defined by $k$-tuples of Poisson points that form an $\varepsilon$-near regular simplex, for any fixed $\varepsilon>0$.

Reading between the lines

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

The four convergences together suggest that all intrinsic linear size measures of the typical cell share one scale in high dimension; the unresolved width constant, if it exists, would be pinned down by the interval in Proposition 2.
The same rescaling and polar-duality scheme could be applied to section volumes, projection volumes, or other translation-invariant functionals of the typical cell to produce additional explicit high-dimensional limit constants.
A direct simulation at moderate dimension (say $d\ge 100$) could test both the conjectured width convergence and how quickly the width bounds narrow as the dimension grows.
The near-regular simplex structure of face-defining points implies that the high-dimensional Delaunay graph of a homogeneous Poisson process is locally built from nearly regular simplices, which may be useful for geometric inference algorithms.

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

The paper settles the first-order shape of the typical high-dimensional Poisson-Voronoi cell: inradius, outradius, diameter, and mean width converge to 1/2, 1, 2, 2 after normalization. It also gives the first exponential-order face-count asymptotics for fixed codimension. These are real advances, and the proofs for Theorem 1 are mostly clean—empty-ball events, Chernoff bounds, and a nice polarity argument for the width. The authors are also honest about the Persian-thesis overlap for the inradius/outradius ratio.

The soft spots are localized but real. Lemma 29's proof uses a false angle relation: for a chord's midpoint, the angles to the origin are not supplementary. The statement itself is true and can be saved by a two-line algebra argument, so this is a patchable flaw. More serious is Lemma 31(ii). As written, it claims second-moment concentration for every fixed r > 1/√2. That is false: for k = 2 and r = 0.75, the overlap term EX_1 has leading base r^3 v_2 v_1 ≈ 2.19, bigger than (r^2 v_2)^2 ≈ 2.13, so the ratio grows exponentially. The proof's inequality (v_ℓ/v_{ℓ−1})^2 ≥ 2 is simply wrong (for ℓ = 2 it is 27/16). The good news is that Theorem 6 only uses r = 1 ± δ with δ small, and the needed concentration does hold when r is close enough to 1 depending on k. So the main face-count theorem is plausibly correct, but the lemma as stated and its proof must be corrected before the claim is accepted. The reliance on cited extremal and duality results (Propositions 33 and 20) is standard and acceptable.

This paper deserves a serious referee: the main results are significant and the defects are fixable, not structural. I would send it to review with the clear instruction that Lemma 31(ii) and Lemma 29's proof be repaired. Once those are fixed, I would cite it in my own work.

Referee Report

2 major / 4 minor

Summary. This paper studies the typical cell V_typ of a Poisson-Voronoi tessellation in R^d with intensity λ. The main theorem (Theorem 1) states that after normalization by (λ κ_d)^{1/d}, the outradius, inradius, diameter, and mean width converge in probability to 1, 1/2, 2, and 2 respectively. Proposition 2 gives non-matching lower and upper bounds for the width, and this implies that V_typ is far in Hausdorff distance from any ball. The paper also proves that all faces of dimension k with d−k→∞ lie in a thin annulus and have diameter o(1) after normalization (Proposition 3, Corollary 4), that almost all faces of constant codimension k are microscopic (Theorem 5), that the d-th root of the number of such faces converges to v_k = ((k+1)^{k+1}/k^k)^{1/2} (Theorem 6), and that almost all such faces are defined by near-regular simplices (Proposition 7). The proofs use Poisson empty-ball probabilities, Chernoff bounds, the Blaschke–Petkantschin formula, and polarity.

Significance. If the face-count theorem is repaired, this is a valuable and significant contribution. The sharp shape constants and the contrast between volume concentration and non-ball-like geometry are clean and falsifiable, and the proof strategy is mostly elementary. The paper is honest about cited external facts (Proposition 20 and Proposition 33) and does not fit free parameters. However, Theorem 6 currently rests on Lemma 31(ii), which is false as stated, so the significance is conditional on a correction.

major comments (2)

[3.7, Lemma 31(ii) and proof of Theorem 6] Lemma 31(ii) is false as stated, and the error is load-bearing for Theorem 6. The inequality chain 'if r>1/√2 then 1<rv_1<r^2v_2<...' in the proof of Lemma 31 is not correct: for ℓ=1 the needed inequality is r>v_1/v_2=4/(3√3)≈0.7698, not merely r>0.707. More importantly, the bound in (22) is too crude to decide the issue. A genuine counterexample to the statement is given by k=6 and r=3/4: a regular simplex with k+2 vertices and circumradius r produces configurations counted by X_1 in (20), giving EX_1 ≥ c (r^{k+1}v_{k+1})^{d−O(1)} with r^7v_7 ≈0.603 > (r^6v_6)^2 ≈0.559, the latter being the base of (EX(r))^2. Thus EX(r)^2/(EX(r))^2 does not tend to 1 for this parameter range. The application in Theorem 6 uses r=1−δ with δ chosen small, so the theorem may survive, but Lemma 31(ii) must be replaced by a correct statement, for example with r sufficiently close to 1 depending on k, and the proof must be repaired.
[3.6, Lemma 29] The proof of Lemma 29 contains a false geometric assertion: for p=(1,0), q=(0,1), and m=(p+q)/2, the angles ∠p0m and ∠q0m are both π/4, so α+β≠π. This invalidates the displayed proof of the diameter bound. The statement itself is true and can be proved directly from ∥p−q∥² = 2∥p∥²+2∥q∥²−∥p+q∥² ≤ 4(1+ε)²−4(1−ε)²=16ε, using that the midpoint m lies in F and hence in the annulus. Since Lemma 29 is used in Corollary 4 and Theorem 5, the proof should be corrected.

minor comments (4)

[3.4, proof of Lemma 22] In the computation of EX, the displayed exponent should be sin^{d−1} α rather than sin^d α; with the corrected exponent, Lemma 8 indeed gives the claimed o((4/r²)^d sin^{d−1} α) = o_d(1).
[Equation (3)] The definition of H_x contains an extra comma after the set-braces; it should read H_x := {y ∈ R^d : x^t y ≤ x^t x/2}.
[Section 4, paragraph on facet distances] The text refers to 'the proof of Lemma 14', but no Lemma 14 exists; the intended reference is presumably Lemma 13 or Corollary 14.
[3.7, Proposition 33] Since the constant in Theorem 6 is inherited from Proposition 33, it would be helpful to state the exact version of the extremal simplex result being cited and to verify that references [20] and [22] cover the equality case used in Lemma 32.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: the target limits are derived from Poisson-process concentration and standard convex-geometric facts, not from fitted inputs or self-referential uniqueness claims.

full rationale

The derivation chain is self-contained in the relevant sense. The normalization λ=1/κ_d is an explicit scaling choice, not a fitted parameter; every theorem is then proved by direct estimates such as Poisson void probabilities, Chernoff bounds, the Blaschke–Petkantschin formula, and Chebyshev's inequality. The constants in Theorem 1 arise from the annulus [1−ε, 1+ε] for all vertices, and the constants in Theorem 6 arise as the d-th root of the Blaschke–Petkantschin integral EX(r), whose growth r^k v_k is computed rather than imposed. The only load-bearing self-citation by the present authors is Proposition 20, V_typ^o = conv(Y), cited from [10]; this is a polar-duality identity whose statement does not contain the target limits, so it is independent support rather than a circular premise. Proposition 33 is cited from [20,22], which are not authored by the present authors. The paper also explicitly flags Conjecture 35 as unproved, but no theorem depends on it. The skeptic's concern about Lemma 31(ii) is a correctness issue concerning the claimed parameter range r>1/√2; even if that lemma needs repair, the instance used in Theorem 6 has r=1−δ with δ small, and a faulty lemma does not by itself make the derivation circular. No equation is defined in terms of the quantity it predicts, and no fitted parameter is renamed as a prediction.

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

No free parameters: all constants (1/2, 1, 2, 2, v_k) emerge from the proofs and are not fitted to data. No new physical or geometrical entities are introduced. The assumptions are standard background facts in stochastic geometry, apart from the two proof defects noted in the red flags.

assumptions (5)

standard math Proposition 33: among all simplices inscribed in the unit sphere, only regular simplices maximize volume.
Used in Lemma 32 to bound the determinant D and to identify the constant v_k in Theorems 5 and 6; cited from [20,22].
domain assumption Proposition 20: almost surely V_typ^o = conv(Y), where Y = f[Z] and f(x) = (2/||x||^2)x.
Used for the width lower bound via Lemma 21; cited from [10] and standard polar duality for Poisson-Voronoi cells.
standard math Lemma 8 angle concentration bound for a uniform random point on the unit sphere.
Cited from [4] and used in Lemma 22 to control angular separation of Poisson points near the sphere.
standard math Blaschke-Petkantschin formula and Mecke formula.
Used to compute moments EX(r) and EY in Section 3.7; standard integral geometry tools.
domain assumption Under the scaling λ = 1/κ_d, the expected number of Poisson points in a ball of radius s equals s^d.
The proof uses this scaling for convenience; the results for arbitrary λ follow from scale invariance of the Voronoi typical cell.

how reviews work

0 comments

Cite this review

Pith. "Pith review of On the shape of the typical Poisson-Voronoi cell in high dimensions." pith.science (2026). https://pith.science/paper/IMJWKIMI

@misc{pith2026250602607,
  author       = {Pith},
  title        = {Pith review of: On the shape of the typical Poisson-Voronoi cell in high dimensions},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/IMJWKIMI}},
  note         = {Machine review of arXiv:2506.02607}
}

abstract

We study the typical cell of the Poisson-Voronoi tessellation. We show that when divided by the $d$-th root of the intensity parameter $\lambda$ of the Poisson process times the volume of the unit ball, the inradius, outradius, diameter and mean width of the typical cell converge in probability to the constants $1/2, 1, 2, 2$ respectively, as the dimension $d\to\infty$. We also show that the width of the typical cell, when rescaled in the same way, is bounded between $2\sqrt{5}/(2+\sqrt{5})-o_d(1)$ and $3/2+o_d(1)$, with probability $1-o_d(1)$. These results in particular imply that, with probability $1-o_d(1)$, the Hausdorff distance between the typical cell and any ball is at least of the order of the diameter of the typical cell. In addition, we show that for all $k$ with $d-k\to\infty$, with probability $1-o_d(1)$, all faces of dimension $k$ have a diameter that is of a much smaller order than the diameter, inradius, etc., of the full typical cell. The same is true for ''almost all'' faces of dimension $d-k$ with $k$ fixed. And, we show that the number of such faces is $\left( (k+1)^{(k+1)/2} / k^{k/2} \pm o_d(1) \right)^d$ with probability $1-o_d(1)$.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

23 extracted references · 23 canonical work pages

[1]

Alishahi

K. Alishahi. Poisson-Voronoi tessellation in high dimensions. PhD thesis, Sharif Univer- sity of Technology, 2008. (In Persian)

work page 2008
[2]

Alishahi and M

K. Alishahi and M. Sharifitabar. Volume degeneracy of the typical cell and the chord length distribution for Poisson-Voronoi tessellations in high dimensions. Adv. in Appl. Probab., 40(4):919–938, 2008

work page 2008
[3]

Bonnet and E

G. Bonnet and E. O’Reilly. Facets of spherical random polytopes. Math. Nachr. , 295(10):1901–1933, 2022

work page 1901
[4]

Brieden, P

A. Brieden, P. Gritzmann, R. Kannan, V. Klee, L. Lov´ asz, and M. Simonovits. Deter- ministic and randomized polynomial-time approximation of radii. Mathematika, 48(1- 2):63–105 (2003), 2001

work page 2003
[5]

Brøndsted

A. Brøndsted. An introduction to convex polytopes , volume 90 of Graduate Texts in Mathematics. Springer-Verlag, New York-Berlin, 1983

work page 1983
[6]

P. Calka. Tessellations. In New perspectives in stochastic geometry , pages 145–169. Oxford Univ. Press, Oxford, 2010

work page 2010
[7]

Calka and N

P. Calka and N. Chenavier. Extreme values for characteristic radii of a Poisson-Voronoi tessellation. Extremes, 17(3):359–385, 2014

work page 2014
[8]

Fejes T´ oth.Regular figures

L. Fejes T´ oth.Regular figures. The Macmillan Company, New York, 1964

work page 1964

Show all 23 references

[9]

H¨ orrmann, D

J. H¨ orrmann, D. Hug, M. Reitzner, and C. Th¨ ale. Poisson polyhedra in high dimensions. Adv. Math., 281:1–39, 2015

2015
[10]

Kabluchko

Z. Kabluchko. Angles of random simplices and face numbers of random polytopes. Adv. Math., 380:Paper No. 107612, 68, 2021

2021
[11]

Kabluchko

Z. Kabluchko. Face numbers of high-dimensional Poisson zero cells. Proc. Amer. Math. Soc., 151(1):401–415, 2023

2023
[12]

Kabluchko, A

Z. Kabluchko, A. Marynych, D. Temesvari, and C. Th¨ ale. Cones generated by random points on half-spheres and convex hulls of Poisson point processes.Probab. Theory Related Fields, 175(3-4):1021–1061, 2019

2019
[13]

Kabluchko, C

Z. Kabluchko, C. Th¨ ale, and D. Zaporozhets. Beta polytopes and Poisson polyhedra: f -vectors and angles. Adv. Math., 374:107333, 63, 2020

2020
[14]

J. F. C. Kingman. Poisson processes, volume 3 of Oxford Studies in Probability . The Clarendon Press, Oxford University Press, New York, 1993. Oxford Science Publications. 26

1993
[15]

Matouˇ sek

J. Matouˇ sek. Lectures on discrete geometry , volume 212 of Graduate Texts in Mathe- matics. Springer-Verlag, New York, 2002

2002
[16]

Meijering

J.L. Meijering. Interface area, edge length, and number of vertices in crystal aggregates with random nucleation. Philips Res. Rep. 8, 1953

1953
[17]

M. D. Penrose. Random geometric graphs , volume 5 of Oxford Studies in Probability . Oxford University Press, Oxford, 2003

2003
[18]

Schilling

R.L. Schilling. Measures, integrals and martingales . Cambridge University Press, New York, 2005

2005
[19]

Schneider and W

R. Schneider and W. Weil. Stochastic and integral geometry . Probability and its Appli- cations (New York). Springer-Verlag, Berlin, 2008

2008
[20]

D. Slepian. The content of some extreme simplexes. Pacific J. Math. , 31:795–808, 1969

1969
[21]

Stoyan, W.S

D. Stoyan, W.S. Kendall, and J. Mecke. Stochastic geometry and its applications . Wiley Series in Probability and Mathematical Statistics: Applied Probability and Statistics. John Wiley & Sons, Ltd., Chichester, 1987. With a foreword by D. G. Kendall

1987
[22]

R.M. Tanner. Some content maximizing properties of the regular simplex. Pacific J. Math., 52:611–616, 1974

1974
[23]

Y.-C. Yao. On variances of partial volumes of the typical cell of a Poisson-Voronoi tessellation and large-dimensional volume degeneracy. Adv. in Appl. Probab., 42(2):359– 370, 2010. A The Hausdorff distance between Vtyp and any ball is large Here we substantiate the claim fro...

2010

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.

[1] [1]

Alishahi

K. Alishahi. Poisson-Voronoi tessellation in high dimensions. PhD thesis, Sharif Univer- sity of Technology, 2008. (In Persian)

work page 2008

[2] [2]

Alishahi and M

K. Alishahi and M. Sharifitabar. Volume degeneracy of the typical cell and the chord length distribution for Poisson-Voronoi tessellations in high dimensions. Adv. in Appl. Probab., 40(4):919–938, 2008

work page 2008

[3] [3]

Bonnet and E

G. Bonnet and E. O’Reilly. Facets of spherical random polytopes. Math. Nachr. , 295(10):1901–1933, 2022

work page 1901

[4] [4]

Brieden, P

A. Brieden, P. Gritzmann, R. Kannan, V. Klee, L. Lov´ asz, and M. Simonovits. Deter- ministic and randomized polynomial-time approximation of radii. Mathematika, 48(1- 2):63–105 (2003), 2001

work page 2003

[5] [5]

Brøndsted

A. Brøndsted. An introduction to convex polytopes , volume 90 of Graduate Texts in Mathematics. Springer-Verlag, New York-Berlin, 1983

work page 1983

[6] [6]

P. Calka. Tessellations. In New perspectives in stochastic geometry , pages 145–169. Oxford Univ. Press, Oxford, 2010

work page 2010

[7] [7]

Calka and N

P. Calka and N. Chenavier. Extreme values for characteristic radii of a Poisson-Voronoi tessellation. Extremes, 17(3):359–385, 2014

work page 2014

[8] [8]

Fejes T´ oth.Regular figures

L. Fejes T´ oth.Regular figures. The Macmillan Company, New York, 1964

work page 1964