arxiv: 2605.10479 · v1 · submitted 2026-05-11 · 🧮 math.PR · math.MG· math.NT

Recognition: no theorem link

Poisson approximation of random lattices

Boaz Klartag

Pith reviewed 2026-05-12 05:05 UTC · model grok-4.3

classification 🧮 math.PR math.MGmath.NT

keywords random latticesPoisson point processtotal variation distanceHaar measurehigh-dimensional geometrypoint processes

0 comments

The pith

A random covolume-one lattice restricted to a set S of volume at most linear in dimension n, with S disjoint from -S, approximates a unit-intensity Poisson point process in total variation distance at most C exp(-c' n).

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper establishes that in high dimensions a uniformly random lattice, when its points are collected inside a suitable bounded region S, produces a point pattern statistically close to that of a Poisson process. The closeness is measured by total variation distance and the error decays exponentially with dimension, provided S has volume at most linear in n and contains no pair of antipodal points. A sympathetic reader would care because the result supplies a concrete, quantitative bridge between deterministic lattice geometry and the simplest model of spatial randomness, showing that the two become practically interchangeable for many purposes once dimension is large.

Core claim

Fix a subset S of R^n of volume at most c n that satisfies S cap (-S) = empty. Let the random lattice be distributed according to the uniform (Haar) measure on the space of covolume-one lattices. Then the total variation distance between the Poisson point process of intensity one restricted to S and the point process given by the lattice points inside S is at most C exp(-c' n), where c, C, c' are positive universal constants.

What carries the argument

The uniform Haar probability measure on the space of covolume-one lattices in R^n, whose induced restriction to S is compared with the Poisson point process via the total variation metric.

If this is right

The two point processes become statistically indistinguishable with probability approaching one as dimension grows, inside any admissible S.
The exponential error bound holds uniformly over all admissible sets S of the given volume.
The approximation is insensitive to the detailed shape of S beyond the volume bound and the antipodal-disjointness condition.
Lattice-based constructions can replace Poisson processes in high-dimensional arguments without incurring more than exponentially small error.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

The result suggests that certain geometric or probabilistic statements proved for Poisson processes in high dimensions can be transferred to random lattices with negligible loss.
It would be natural to test whether the same exponential approximation persists for other metrics on point processes or for slightly larger volumes.
In applications where one needs both regularity from lattices and randomness from Poisson processes, the exponential closeness supplies a concrete justification for substituting one for the other.

Load-bearing premise

The set S must have volume at most linear in the dimension and must contain no vector together with its negative.

What would settle it

Construct a sequence of sets S_n in R^n, each of volume at most c n and disjoint from its negative, for which the total variation distance between the Poisson process and the random-lattice point process inside S_n stays bounded away from zero (or decays slower than any exponential) for infinitely many n.

read the original abstract

Fix a subset $S \subset \mathbb{R}^n$ of volume at most $c n$ that satisfies $S \cap (-S) = \emptyset$. We consider two point processes in $S$: the first is the Poisson point process of intensity one, and the second is the restriction of a random lattice to $S$, where the random lattice is distributed uniformly in the space of covolume-one lattices. We show that the total variation distance between these two point processes is at most $C e^{-c' n}$, where $c, C, c' > 0$ are universal constants.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

Klartag proves an exponential TV bound between random lattice points and Poisson on linear-volume antipodal-free sets in high dimensions.

read the letter

The central result here is that for a set S subset R^n with volume at most order n and S cap (-S) empty, the restriction of a Haar-random covolume-one lattice to S is within C exp(-c' n) total variation of a unit-intensity Poisson process on S, with universal constants. This quantitative rate stands out as the main new piece. Earlier work on random lattices and point processes tends to give qualitative limits or slower rates, so the exponential decay in n looks like a concrete advance under these tailored conditions. The setup stays grounded in the standard Haar measure on lattices and basic Poisson properties, with no visible circularity or parameter fitting. The conditions on S are natural: the volume bound keeps the expected count linear while the antipodal-free rule cuts out obvious lattice symmetries that would otherwise create dependence. The paper states the theorem cleanly with explicit constants, which is useful if the proof holds up. This could let people replace lattice points with Poisson points in some high-dimensional arguments without losing much error. The main soft spot is that we only have the abstract, so the actual derivation steps and the size of c' remain unchecked. High-dimensional concentration is probably doing the heavy lifting, and any gaps would likely show up in how the lattice measure is handled near the boundary of S. Nothing in the claim contradicts itself, but the constants need verification for applications. This is aimed at researchers in high-dimensional probability and geometry of numbers who already work with random lattices or point process approximations. A reader looking for explicit rates to simplify calculations would get direct value. It deserves a serious referee because the statement is sharp and the setting is standard, even if the write-up might need expansion on the proof.

Referee Report

0 major / 2 minor

Summary. Fix a subset S ⊂ R^n of volume at most c n that satisfies S ∩ (-S) = ∅. The paper considers two point processes in S: the Poisson point process of intensity one, and the restriction of a random lattice to S, where the random lattice is distributed uniformly in the space of covolume-one lattices. It shows that the total variation distance between these two point processes is at most C e^{-c' n}, where c, C, c' > 0 are universal constants.

Significance. If the result holds, it supplies a quantitatively sharp Poisson approximation for the local statistics of random covolume-one lattices in high dimensions, with an exponentially small total-variation error that depends only on dimension n. The derivation uses the Haar measure on the space of lattices together with high-dimensional concentration and avoids free parameters or fitted constants beyond the explicit volume and antisymmetry hypotheses on S; this parameter-free character and the explicit exponential rate are genuine strengths that could support applications in geometric probability, discrepancy theory, and statistical mechanics of point processes.

minor comments (2)

[Abstract] Abstract: the volume bound is stated as 'at most c n' but the admissible range of the constant c is not indicated; a short clarifying sentence would help readers assess the result's scope.
The manuscript would benefit from a brief discussion (perhaps in the introduction or a remarks section) of whether the antisymmetry condition S ∩ (-S) = ∅ is essential for the exponential rate or can be weakened while preserving the same order of approximation.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for their careful reading and for recommending minor revision. The referee's summary accurately captures the main theorem of the paper, including the hypotheses on the set S and the exponentially small total-variation bound. No specific major comments or requests for clarification were raised in the report.

Circularity Check

0 steps flagged

No significant circularity; derivation self-contained from definitions

full rationale

The paper establishes an exponentially small total variation distance bound between the restricted random lattice point process and a unit-intensity Poisson process on S. This follows directly from the Haar (uniform) measure on covolume-one lattices in R^n, combined with high-dimensional concentration and the explicit assumptions vol(S) ≤ c n and S ∩ (-S) = ∅ to control correlations and symmetry. No step reduces by construction to a fitted parameter, self-referential definition, or load-bearing self-citation; the argument uses standard properties of these objects without renaming known results or smuggling ansatzes. The central claim therefore has independent mathematical content.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The result rests on the existence and uniqueness of the Haar measure on the space of covolume-one lattices, standard properties of Poisson point processes, and basic facts from geometry of numbers; no free parameters or new entities are introduced.

axioms (2)

standard math Existence of a unique (up to scaling) Haar probability measure on the space of covolume-one lattices in R^n
Invoked to define the random lattice distribution
standard math The Poisson point process of intensity 1 is a well-defined simple point process on R^n
Used as the comparison object

pith-pipeline@v0.9.0 · 5386 in / 1258 out tokens · 47190 ms · 2026-05-12T05:05:56.468355+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

17 extracted references · 17 canonical work pages

[1]

M., Lekkerkerker, C

Gruber, P. M., Lekkerkerker, C. G.,Geometry of numbers.Second edition. North- Holland Publishing Co., 1987

work page 1987
[2]

Number Theory, V ol

Holm, K.,On the distribution of angles between increasingly many short lattice vec- tors.J. Number Theory, V ol. 240, (2022), 357–403

work page 2022
[3]

Z., V ol

Kim, S.,On the distribution of lengths of short vectors in a random lattice.Math. Z., V ol. 282, no. 3-4, (2016), 1117–1126

work page 2016
[4]

Kim, S.,Random lattice vectors in a set of sizeO(n). Int. Math. Res. Not. (IMRN), No. 5, (2020), 1385–1416

work page 2020
[5]

Y ., Margulis, G

Kleinbock, D. Y ., Margulis, G. A.,Logarithm laws for flows on homogeneous spaces. Invent. Math., V ol. 138, no. 3, (1999), 451–494

work page 1999
[6]

Last, G., Penrose, M.,Lectures on the Poisson process.Cambridge University Press, 2018

work page 2018
[7]

Marklof, J.,Random lattices in the wild: from P ´olya’s orchard to quantum oscillators. Lond. Math. Soc. Newsl., No. 493, (2021), 42–49

work page 2021
[8]

A.,Mean values over the space of lattices.Acta Math., V ol

Rogers, C. A.,Mean values over the space of lattices.Acta Math., V ol. 94, (1955), 249–287

work page 1955
[9]

A.,The number of lattice points in a set.Proc

Rogers, C. A.,The number of lattice points in a set.Proc. London Math. Soc. (3), V ol. 6, (1956), 305–320

work page 1956
[10]

A.,Lattice coverings of space

Rogers, C. A.,Lattice coverings of space. Mathematika, V ol. 6, (1959), 33–39

work page 1959
[11]

M.,Mittelwerte ¨uber Gitter.Monatsh

Schmidt, W. M.,Mittelwerte ¨uber Gitter.Monatsh. Math., V ol. 61, (1957), 269–276

work page 1957
[12]

M.,The measure of the set of admissible lattices.Proc

Schmidt, W. M.,The measure of the set of admissible lattices.Proc. Amer. Math. Soc., V ol. 9, No. 3, (1958), 390–403

work page 1958
[13]

M.,On the convergence of mean values over lattices.Canadian J

Schmidt, W. M.,On the convergence of mean values over lattices.Canadian J. Math., V ol. 10, (1958), 103–110

work page 1958
[14]

L.,A mean value theorem in geometry of numbers.Ann

Siegel, C. L.,A mean value theorem in geometry of numbers.Ann. of Math., V ol. 46, No. 2, (1945), 340–347

work page 1945
[15]

Z., V ol

S ¨odergren, A.,On the Poisson distribution of lengths of lattice vectors in a random lattice.Math. Z., V ol. 269, No. 3–4, (2011), 945–954. 16

work page 2011
[16]

S ¨odergren, A.,On the distribution of angles between theNshortest vectors in a ran- dom lattice. J. Lond. Math. Soc. (2), V ol. 84, no. 3, (2011), 749–764

work page 2011
[17]

Venkatesh, A.,A note on sphere packings in high dimension.Internat. Math. Res. No- tices (IMRN), no. 7, (2013), 1628–1642. School of Mathematical Sciences, Tel Aviv University, Tel Aviv 6997801, Israel; and De- partment of Mathematics, Weizmann Institute of Science, Rehovot 7610001, Israel. e-mail:klartagb@tau.ac.il 17

work page 2013