Local limits of determinantal processes

Asaf Nachmias; Yuval Peled

REVIEW 2 major objections 5 minor 26 references

Local limits of determinantal processes

T0 review · 2 major / 5 minor · reviewed 2026-08-04 · deepseek-v4-flash

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

Pith's one-line read For a wide class of determinantal processes, local limits are a single Poisson branching tree, determined by one integer k.

desk verdict Genuinely new unification, probably correct, but Corollary 2.14 is false as stated and needs a gamma->0 fix before the exclusion bound is rigorous. read the letter →

arxiv 2510.19563 v2 pith:7RJDAHMB submitted 2025-10-22 math.PR math.CO

Asaf Nachmias , Yuval Peled This is my paper · ORCID

classification math.PRmath.CO MSC 60C0505C80

keywords determinantalprocesseslocalweakconvergenceC4-freebipartitegraphsPoissonbranchingprocessuniformspanningtreeshypertreesspectralconcentrationmatchings

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 establishes a universality theorem for the local geometry of determinantal processes. Whenever the process is attached to the row space of a signed adjacency matrix of a C4-free bipartite graph that is bi-regular with right degree k+1 and left degree d going to infinity, the neighborhood of a uniformly random left vertex converges in the local topology to a fixed infinite tree T_k: a variant of a Poisson(k) branching process conditioned to survive. The limit depends only on k, not on the host graph or the signs. This one result recovers the known local limit of uniform spanning trees of high-degree regular graphs, extends the known local limit of determinantal hypertrees from the complete complex to arbitrary high-degree cell complexes, and gives the first local limit statements for processes such as discrete Grassmannians and incidence matroids. The limiting probability of seeing any finite rooted tree has an explicit formula in terms of matchings and automorphisms.

What carries the argument

The argument revolves around the family of (epsilon,delta)-structured vertices: right-side vertices u at which the eigenvectors of the operator L_- = B^T B / d with eigenvalues far from 1 carry more than delta of their squared mass. The identity Tr((L_+ - I)^2) = k n / d, valid exactly when the graph has no 4-cycles, forces all but o(n) of the eigenvalues of L_- to lie within sqrt(epsilon) of 1, so the structured set has o(m) vertices. The proof then couples T_n with the determinantal process F generated by the spectral subspace with eigenvalues near 1, which is stochastically dominated by T_n, and shows the two agree locally with high probability. The remaining inclusion-exclusion estimates

What would settle it

Take G_n to be the disjoint union of n/3 copies of K_{3,d(n)}, the complete bipartite graph with parts of sizes 3 and d(n), with d(n) to infinity. This is simple, bipartite, (d,3)-bi-regular and full of 4-cycles. The determinantal process draws one right vertex uniformly, so a uniform left root has a singleton radius-1 ball with probability 1-o(1), while the root of T_2 has at least one neighbor with probability 1; moreover Tr((L_+ - I)^2) equals 2n here, whereas the C4-free identity would give k n / d = 2n/d tending to 0. This discrepancy isolates why C4-freeness is needed.

Watch

Extended reading notes

Core claim

Let G_n be a sequence of simple C4-free bipartite (d(n), k+1)-bi-regular graphs with signed adjacency matrix B_n, and let T_n be sampled from the determinantal measure assigning weight det(P_H restricted to T) to each subset T of the right-side vertices of size rank(B_n). The paper proves that, with a uniformly random left vertex o_n, the rooted induced graph (G[T_n], o_n) converges in the local topology to T_k, the multi-type branching process with four particle types: a-even, b-even, a-odd, b-odd. The limit is universal: it is independent of the host graph and of the signs, and it is a conditioned variant of a Poisson(k) branching process. For a fixed rooted tree (T,o), the limiting probab

Load-bearing premise

The load-bearing premise is that the host bipartite graph has no pair of vertices with more than one common neighbor (no 4-cycles), which is exactly what forces the relevant Laplacian eigenvalues to sit near 1 and the structured vertices to be rare; without it, the theorem is not claimed and the proof's key estimate fails.

Editorial extensions

If this is right

The uniform spanning tree of any high-degree regular graph has local limit equal to the Poisson(1) branching process conditioned to survive; this is the k=1 special case.
Determinantal hypertrees in any regular high-degree cell complex, including colorful simplicial complexes and hypercube skeletons, converge to the same T_k that was previously known only for the complete complex.
Processes with no explicit description of their bases or determinants, such as discrete Grassmannians over finite fields and incidence matroids, still get a fully explicit local limit.
Convergence is quenched: the fraction of left vertices whose r-ball is isomorphic to a fixed tree converges in probability to the explicit formula, not just in expectation.
The explicit formula for each finite rooted tree gives asymptotic probabilities for every finite local pattern at once.

Reading between the lines

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

If C4-freeness is dropped, the spectral concentration step breaks: for a disjoint union of copies of K_{3,d}, Tr((L_+ - I)^2) is linear in the number of copies rather than vanishing, so the local limit, when it exists, would have to depend on the host graph's cycle structure. This is our reading, not a claim of the paper.
The matching-count form of the limit is a natural template for other tree-indexed determinantal processes: any sequence of kernels whose squared deviation from the identity has vanishing normalized trace should have a limit of the same multiplicative shape.
The quenched result indicates local statistics are self-averaging; a natural next step, not taken in the paper, is to quantify the rate at which the error terms shrink as a function of d and epsilon.
One might expect bi-regularity itself could be relaxed to average-degree conditions, since the proof uses only the trace concentration and local regularity of the right side; this is speculation beyond the paper's assumptions.

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 short version: this paper deserves a serious referee, but not in its current form. Theorem 1.1 is a real advance — a single convergence theorem for determinantal processes on C4-free (d,k+1)-bi-regular signed incidence graphs, recovering Nachmias–Peres for UST and Mészáros for hypertrees, and covering Grassmannians, incidence matroids, and regular cell complexes. The proof is structured and mostly sound: the spectral concentration from C4-freeness, the tightness/tree-ness argument, the inclusion bound via Cauchy-Binet, and the exclusion bound via the Nash-Williams-type Lemma 2.13. The alternate derivation of the limit distribution in §1.3 is a nice extra, and the target law is independently established, so there is no circularity.

The soft spots are concentrated in §2.3.2–2.3.3. Corollary 2.14 as printed is false: the prefactor (1−γk)/(1−γk) is identically 1, and the proof only supports (1−2kγ)/(1−kγ). For k=2, ξ1=(1,√3,0,0), ξ2=(1,0,√3,0), γ=1/4, the claimed lower bound is 1/2 but 1ᵀM⁻¹1 = 0.4. That is not just a typo: in Lemma 2.16 the vectors have norms Ω(d) and inner products O(1), so the right way to apply the corollary is with γ=γ_n→0, which the statement as written does not permit. Without a limit version, the product in Lemma 2.8 picks up a fixed loss e^{-c k |I(T)|} and the upper bound does not match the lower bound. This is repairable — the proof of the corollary works for any γ<1/k and gives the correct factor with γ→0 — but it is a real gap and should be fixed before publication. There are also minor slips in the same section: the garbled phrase about u∈U0 in Lemma 2.16's norm estimate and a few other typos.

On the substance, I don't see a load-bearing flaw beyond the corollary. C4-freeness is genuinely needed for the spectral trace and path-counting; the authors are upfront about that. The counting lemma 2.18 is standard. The quenched theorem 1.3 follows the announced pattern. Formal verification is absent, and given the length, an independent check of Claims 2.10–2.11 and Lemma 2.16 would be worthwhile.

Who this is for: anyone working on local limits, determinantal processes, or random geometric complexes. I would cite it once the corollary is fixed. My recommendation: send it to a good journal, ask the authors to correct Corollary 2.14 (or replace it with a γ_n→0 statement) and to clean up §2.3.3. The main theorem is likely true and important.

Referee Report

2 major / 5 minor

Summary. The paper studies local limits of determinantal processes on the right side U of a C4-free bipartite graph G=(V,U,E) that is (d,k+1)-biregular with d→∞. The process is the row-space determinantal process of a signed adjacency matrix. The main theorem (Thm 1.1/1.2) states that the rooted random graph (G[T],o) with o uniform in V converges to a Poisson(k) branching process T_k conditioned to survive, with explicit probabilities (1.2). A quenched version (Thm 1.3) is also proved. The proof uses spectral concentration of the Laplacians L± to compare T with a truncated process F, inclusion bounds via Cauchy-Binet, exclusion bounds via a new determinantal Nash-Williams inequality (Lemma 2.13/Cor. 2.14), and a counting argument. Examples include uniform spanning trees, Kalai hypertrees, hyperforests in cell complexes, discrete Grassmannians, and incidence matroids.

Significance. If the result stands, this is a significant unifying advance: it recovers the UST local limit of Nachmias–Peres and Mészáros's hypertree limit, and extends them to a broad class of determinantal processes. The explicit limiting distribution, the independent derivation of (1.2), and the quenched theorem are assets. The spectral approach and the determinantal Nash-Williams lemma are likely to be useful. However, the written proof contains a false statement in Corollary 2.14 that is load-bearing for the exclusion argument; because the issue is local and repairable, the appropriate disposition is major revision.

major comments (2)

[§2.3.2, Corollary 2.14] Corollary 2.14 is false as stated. The displayed factor (1−γk)/(1−γk) equals 1, but the proof's final estimate gives only (1−2kγ)/(1−kγ), because \tilde R_{i,j} ≥ −γ/(1−γk) yields 1 + Σ_j \tilde R_{j,i} ≥ (1−2kγ)/(1−kγ). Concretely, for k=2, ξ1=(1,√3,0,0), ξ2=(1,0,√3,0), γ=1/4, the hypotheses hold, yet 1^T M^{-1}1 = 0.4 < 1/4+1/4 = 0.5. Since Lemma 2.16 and hence Lemma 2.8 rely on this inequality, the statement must be corrected (likely to (1−2kγ)/(1−kγ)).
[§2.3.3, Lemma 2.16] In Lemma 2.16 the vectors ξ0,…,ξk have norms (1+O(d^{-1}))d and pairwise inner products O(1); the only admissible parameter in Corollary 2.14 is therefore γ = γ_n → 0. The corollary is stated only for fixed γ. Its proof is uniform in γ, so an extension to γ_n → 0 is immediate, but it must be stated explicitly; without it, the application to Lemma 2.8 gives the wrong exponential constant (for fixed γ>0 the factor (1−2kγ)/(1−kγ) is bounded below 1). This is a repairable but load-bearing gap.

minor comments (5)

[§2.1, Lemma 2.5] Typo: 'there there must exists'. Also the notation lim_{t→∞} lim_{n→∞} should be a limsup or the existence of the inner limit should be justified.
[§2.3.2, Corollary 2.14] The letter W is used both for the subspace and for a sample; this is confusing in the proof. Suggest using S for the sample.
[Claim 2.3, Eq. (2.3)] The step from Tr ≥ (r−|I|)ε + n−r to ≥ (n−|I|)ε is not immediate; expanding n−r = n−|I| − (r−|I|) makes it clear.
[§1.2.2] 'Grimmet' should be 'Grimmett' (also in the reference list entry [6]).
[§2.4, Proof of Theorem 1.2] The contradiction argument after (2.18) is terse; a sentence explaining why the assumed strict inequality contradicts (2.17) would improve readability.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: the target distribution is independently re-derived and the main proof compares against that independent expression.

full rationale

The central theorem is not obtained by assuming its own conclusion. The target formula (1.2) is quoted from Mészáros [15] and then independently re-derived in Section 1.3 from the 1-out model (Lemma 1.5), with the paper explicitly noting that this alternate proof “is not used in the main results of this paper.” The main proof then bounds the probability of a fixed rooted ball from above by the same expression (Corollary 2.17, via Lemmas 2.7 and 2.8), counts embeddings (Lemma 2.18), and uses the fact that the T_k probabilities sum to 1 to force the matching lower bound. Nothing is fitted to the claimed limit: the auxiliary parameters ε, δ are free and only required to satisfy the stated diverging/converging rates. The self-citations ([11] for the 1-out model and the definition of T_k, and [22] for the already-known UST special case) are not load-bearing inputs to the new theorem; [22] is presented as a recovered example. The possible typo/correctness issue in Corollary 2.14’s displayed factor is a proof defect, not a reduction of the conclusion to its assumptions, and therefore does not constitute circularity. Overall the derivation chain is self-contained apart from standard known facts and independently reproved ingredients.

Assumptions & free parameters 1 free parameters · 4 assumptions · 0 invented entities

The theorem rests on standard determinantal-measure facts, the paper's structural C4-free bi-regular assumptions, and the known/independently re-proved distribution of T_k. The only hand-chosen quantities are the proof parameters ε and δ. No new particles, forces, dimensions, or other postulated entities are introduced.

free parameters (1)

ε, δ decay parameters = e.g. ε = d^{-1/2}, δ = d^{-5/4}
Chosen by hand in (2.1) to satisfy ε→0, εd→∞, δd→0, εδd²→∞. They appear only in the proof, vanish asymptotically, and do not enter the limiting object.

assumptions (4)

domain assumption G is C4-free, (d,k+1)-bi-regular with d→∞ and k fixed.
Core hypothesis of Theorem 1.1; enables the trace bound Tr((L_+−I)²)=kn/d in Claim 2.3 and the path/cycle counting and norm estimates throughout.
standard math Determinantal measure facts from Lyons [12]: finite marginals (1.5), negative correlation (1.6), stochastic domination (1.7), and the conditional subspace formula (1.8).
Imported without proof and used throughout, e.g., in Lemma 2.5, Lemma 2.7, and Lemma 2.8.
standard math The limiting tree T_k and its explicit ball distribution (1.2) are valid; the authors give an alternate proof in §1.3.
The distribution formula is originally due to Mészáros [15, Lemma 1.4]; the paper re-proves it independently, so it is not assumed in a circular way.
standard math Cauchy–Binet, the generalized matrix determinant lemma, Minkowski's determinant inequality, and standard spectral facts for PSD matrices.
Used in the bounds of Lemma 2.7 and throughout the spectral arguments; standard background.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Local limits of determinantal processes." pith.science (2026). https://pith.science/paper/7RJDAHMB

@misc{pith2026251019563,
  author       = {Pith},
  title        = {Pith review of: Local limits of determinantal processes},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/7RJDAHMB}},
  note         = {Machine review of arXiv:2510.19563}
}

abstract

Let $H_n$ be the row space of a signed adjacency matrix of a $C_4$-free bipartite bi-regular graph in which one part has degree $d(n)\to\infty$ and the other part has degree $k+1$ where $k\geq 1$ is a fixed integer. We show that the local limit as $n\to \infty$ of the determinantal process corresponding to the orthogonal projection on $H_n$ is a variant of a Poisson$(k)$ branching process conditioned to survive. This setup covers a wide class of determinantal processes such as uniform spanning trees, Kalai's determinantal hypertrees, hyperforests in regular polytopal complexes, discrete Grassmanians and incidence matroids, as long as their degree tends to $\infty$.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

26 extracted references · 4 linked inside Pith

[1]

R. M. Adin. Counting colorful multi-dimensional trees. Combinatorica, 12(3):247– 260, 1992

1992
[2]

Benjamini and O

I. Benjamini and O. Schramm. Recurrence of distributional limits of finite planar graphs. Electron. J. Probab., 6:no. 23, 13, 2001

2001
[3]

N. Curien. Random graphs, the local convergence point of view. Lec- ture notes, https: // www. dropbox. com/ s/ llq1oreh7m25dad/ cours-RG-online. pdf? dl= 0, 2024

2024
[4]

A. M. Duval, C. J. Klivans, and J. L. Martin. Cellular spanning trees and laplacians of cubical complexes. Advances in Applied Mathematics , 46(1-4):247–274, 2011

2011
[5]

A. M. Duval, C. J. Klivans, and J. L. Martin. Simplicial and cellular trees , pages 713–752. Springer International Publishing, Cham, 2016. 22

2016
[6]

G. R. Grimmett. Random labelled trees and their branching networks. Journal of the Australian Mathematical Society , 30(2):229–237, 1980

1980
[7]

D. A. Harville. Matrix algebra from a statistician ’s perspective. Springer-Verlag, New York, 1997

1997
[8]

R. A. Horn and C. R. Johnson. Matrix analysis . Cambridge University Press, Cam- bridge, second edition, 2013

2013

Show all 26 references

[9]

Kahle and A

M. Kahle and A. Newman. Topology and geometry of random 2-dimensional hyper- trees. Discrete & Computational Geometry , 67(4):1229–1244, 2022

2022
[10]

G. Kalai. Enumeration of Q-acyclic simplicial complexes. Israel J. Math., 45(4):337– 351, 1983

1983
[11]

Linial and Y

N. Linial and Y. Peled. Enumeration and randomized constructions of hypertrees. Random Structures & Algorithms , 55(3):677–695, 2019

2019
[12]

R. Lyons. Determinantal probability measures. Publ. Math. Inst. Hautes ´Etudes Sci., (98):167–212, 2003

2003
[13]

R. Lyons. Random complexes and ℓ2-betti numbers. Journal of Topology and Anal- ysis, 1(02):153–175, 2009

2009
[14]

Lyons and Y

R. Lyons and Y. Peres. Probability on trees and networks , volume 42 of Cambridge Series in Statistical and Probabilistic Mathematics . Cambridge University Press, New York, 2016

2016
[15]

M´ esz´ aros

A. M´ esz´ aros. The local weak limit ofk-dimensional hypertrees. Trans. Amer. Math. Soc., 375(9):6127–6154, 2022

2022
[16]

M´ esz´ aros

A. M´ esz´ aros. The 2-torsion of determinantal hypertrees is not cohen-lenstra. arXiv preprint arXiv:2404.02308, 2024

2024 arXiv
[17]

M´ esz´ aros

A. M´ esz´ aros. Bounds on the mod 2 homology of random 2-dimensional determinantal hypertrees. Combinatorica, 45(2):17, 2025

2025
[18]

M´ esz´ aros

A. M´ esz´ aros. The homology torsion growth of determinantal hypertrees. arXiv preprint arXiv:2506.14694, 2025

2025 arXiv
[19]

M´ esz´ aros

A. M´ esz´ aros. Using dense graph limit theory to count cocycles of random simplicial complexes. arXiv preprint arXiv:2509.06559 , 2025

2025 arXiv
[20]

V. B. Mnukhin and I. Siemons. On modular homology in projective space. Journal of Pure and Applied Algebra , 151(1):51–65, 2000

2000
[21]

M´ esz´ aros

A. M´ esz´ aros. Coboundary expansion for the union of determinantal hypertrees.Ran- dom Structures & Algorithms , 65(4):896–914, 2024

2024
[22]

Nachmias and Y

A. Nachmias and Y. Peres. The local limit of uniform spanning trees. Probab. Theory Related Fields, 182(3-4):1133–1161, 2022

2022
[23]

R. P. Stanley. The sperner property. In Algebraic Combinatorics: Walks, Trees, Tableaux, and More, pages 31–41. Springer, 2013

2013
[24]

Tessler and E

R. Tessler and E. Tzalik. Topological expansion for posets and the homological k- connectivity of random q-complexes. arXiv preprint arXiv:2306.14317 , 2023

2023 arXiv
[25]

Vander Werf

A. Vander Werf. Simplex links in determinantal hypertrees. Journal of Applied and Computational Topology, 8(2):401–426, 2024

2024
[26]

Zaslavsky

T. Zaslavsky. Signed graphs. Discrete Applied Mathematics , 4(1):47–74, 1982. 23

1982

Pith tools

Reviewed August 4, 2026 · model on record in the stance chip above.

[1] [1]

R. M. Adin. Counting colorful multi-dimensional trees. Combinatorica, 12(3):247– 260, 1992

1992

[2] [2]

Benjamini and O

I. Benjamini and O. Schramm. Recurrence of distributional limits of finite planar graphs. Electron. J. Probab., 6:no. 23, 13, 2001

2001

[3] [3]

N. Curien. Random graphs, the local convergence point of view. Lec- ture notes, https: // www. dropbox. com/ s/ llq1oreh7m25dad/ cours-RG-online. pdf? dl= 0, 2024

2024

[4] [4]

A. M. Duval, C. J. Klivans, and J. L. Martin. Cellular spanning trees and laplacians of cubical complexes. Advances in Applied Mathematics , 46(1-4):247–274, 2011

2011

[5] [5]

A. M. Duval, C. J. Klivans, and J. L. Martin. Simplicial and cellular trees , pages 713–752. Springer International Publishing, Cham, 2016. 22

2016

[6] [6]

G. R. Grimmett. Random labelled trees and their branching networks. Journal of the Australian Mathematical Society , 30(2):229–237, 1980

1980

[7] [7]

D. A. Harville. Matrix algebra from a statistician ’s perspective. Springer-Verlag, New York, 1997

1997

[8] [8]

R. A. Horn and C. R. Johnson. Matrix analysis . Cambridge University Press, Cam- bridge, second edition, 2013

2013