REVIEW 1 major objections 3 minor 18 references
Erd\H{o}s--Ko--Rado and Hilton--Milner Theorems in the Partition Lattice
T0 review · 1 major / 3 minor · reviewed 2026-08-07 · deepseek-v4-flash
Pith's one-line read In the partition lattice, pairwise-intersecting families of rank-k flats have size at most the largest full edge-star whenever the ground set has at least 8k elements, and the only families reaching that size are full edge-stars.
desk verdict Substantial new EKR results for the partition lattice, but the central linear-range proof has a repairable gap in the sparse-family bound. 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 main instrument is rank spread approximation and peeling on the lattice of flats, where the spread exponent is the closure rank of a flat rather than the cardinality of its atom set. The peeling procedure repeatedly replaces a covering subfamily by a smaller centre while preserving t-intersection, then bounds each peeled layer using the number of rank-k extensions of a rank-i flat, denoted E_i. The linear t=1 argument adds two further mechanisms: an occupancy model whose log-concave distribution and exponential tilt control the tail of dense flats with many atoms, and a switching lemma that, for any flat F avoiding an atom a, compares the number of full-star members avoiding F with E_1 times eta raised to the atomic weight of F, where eta=(N-3k+1)/(N-3k+2) and N=n+1.
What would settle it
Take N=n+1 and compute the number of (N-k)-block partitions of an N-set that contain a fixed edge and avoid all edges of a given flat F, for N just below 3k (say N=3k-1), and compare it with E_1 times $eta^{{omega(F)}}$; a value smaller than the bound in Lemma 5.10 would invalidate the switching step. Separately, any intersecting family in the range 2k <= N <= 8k-1 with more than binomial(n-1,k-1) members would falsify the full conjecture that the theorem approximates.
Extended reading notes
Core claim
The central discovery is that the correct spread parameter in the partition lattice is the closure rank of a flat, not the number of atoms: the three edges of a triangle have cardinality three but closure rank two. With this rank-sensitive spreadness, a peeling argument bounds sparse families, an occupancy-tail estimate controls dense families of high atomic weight, and a singleton-switching comparison near a full edge-star yields the linear bound n+1 >= 8k for t=1. For general t, a peeling decomposition with a dimensionless series gives an explicit quadratic threshold n+1-k >= c_t times the number of rank-two extensions, with equality only for full t-stars. The Hilton–Milner theorem gives an explicit O($k^{6}$) condition under which a nontrivial intersecting family without a common atom has size at most HM(n,k), with equality only for a family built from a fixed atom, an exceptional clique, and two exceptional clique flats.
Load-bearing premise
The switching step assumes that after contracting a fixed atom, a star-avoiding flat F becomes a graph of maximum degree at most k, and that each added edge destroys at most the proportion eta=(N-3k+1)/(N-3k+2) of the remaining star members; if this retention factor is ever smaller than eta, the strict inequality |A|<E_1 can fail.
Editorial extensions
If this is right
- If Theorem 1.2 is correct, the Czabarka partition-EKR inequality holds for all n >= 8k-1, so any counterexample to the full conjecture would have to lie in the remaining constant-factor window between 2k and 8k-1.
- The t-intersection theorem gives a fully explicit, checkable condition under which the full t-star is the unique extremal family, reducing verification for any fixed t to a finite substitution.
- The Hilton–Milner theorem identifies the unique largest non-star intersecting family up to isomorphism for n = O(k^6), giving a concrete structural description rather than a size bound alone.
- Together, these results convert qualitative 'for fixed k and n sufficiently large' statements into explicit polynomial and linear thresholds that can be applied directly.
Reading between the lines
- The switching lemma's N >= 3k barrier suggests that within this proof framework the constant 8 is not fundamental: sharper estimates may pull it down, but the singleton-switching comparison can at best reach N=3k, so reaching the conjectured N>=2k+1 range would require a genuinely new final step such as a global closure-compatible matching or compression.
- A testable extension is whether the same rank-sensitive peeling yields an Erdős–Ko–Rado theorem for other graphic matroids, where the rank spread exponent and the switching count would need to be re-derived from the graph's cycle matroid structure.
- The paper's compatible-split construction for t>=2 shows that the naive endpoint n=2k-t+1 fails, indicating that the true threshold n0(k,t) is governed by a separate obstruction; a natural follow-up is to check whether the quadratic condition in Theorem 1.3 can be replaced by a linear one for each fixed t.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper studies t-intersecting families of rank-k flats in the graphic matroid M_n=M(K_{n+1}), equivalently families of partitions of an (n+1)-set into n+1-k blocks with pairwise meet rank at least t. The main results are: Theorem 1.2, an Erdős–Ko–Rado bound |A|≤binom(n-1,k-1) for intersecting families in the explicit linear range n+1≥8k, with equality only for a full edge-star; Theorem 1.3, a t-intersection EKR bound under the explicit quadratic condition n+1-k≥c_t(m_k+1), with equality only for a full t-star; and Theorem 1.5, a Hilton–Milner theorem under an explicit O(k^6) threshold with a unique extremal family up to isomorphism. The proofs combine a rank-based spread/peeling framework, an atomic-weight truncation with an occupancy tail, and a singleton-switching comparison for the linear t=1 case.
Significance. If correct, the paper gives a substantial advance: it replaces the previously known eventual (non-explicit) EKR result for partition pairs with an explicit linear-range theorem, provides fully explicit quadratic thresholds for general t-intersection, and gives an explicit Hilton–Milner theorem with extremal structure. The rank-sensitive spread formalism is a natural and interesting adaptation of the Kupavskii–Zakharov peeling framework, and the paper is unusually explicit about all constants, proving the required one-variable estimates in an appendix. The conjectures and the lower-bound constructions in Remark 1.4 and Corollary 6.2 give falsifiable targets for the conjectured sharp range. The central t=1 proof currently contains a local but load-bearing gap concerning the definition of r0, so the paper as written does not yet establish Theorem 1.2; the gap appears repairable within the manuscript's scope.
major comments (1)
- [Section 5, proof of Theorem 1.2 and Corollary 5.3] The proof sets r0 := ceil(7b/10) and then asserts that a sparse family's maximum atomic weight μ satisfies μ+1 ≤ r0 ≤ 7b/10. The second inequality is false whenever 7b/10 is not an integer. Since Corollary 5.3 is stated under the hypothesis μ+1 ≤ 7b/10, it is not applicable as written. This is load-bearing: the displayed sparse bound |C|/E1 < 22/23 is exactly what yields the strict inequality |A| < E1 in the no-common-atom case of Theorem 1.2. Moreover, the proof of Corollary 5.3 does not close with the weaker bound r ≤ ceil(7b/10): for instance, with N=88, k=11, b=77, r0=54, one has 2r0 k/(N b) ≈ 0.1753 > 7/40, so the factor estimate in Corollary 5.3 fails. The argument can likely be repaired by redefining r0 as floor(7b/10) or by proving a version of Corollary 5.3 with the ceiling and then re-closing the constants, but the density estimates in Lemma 5.6 and Corollary 5.9 must be rechecked under either modification. As written, the central proof of Theorem 1.2 is incomplete.
minor comments (3)
- [Section 5, Lemma 5.1] The double-counting proof of (5.1) is correct, but the phrase 'inserting it into one of the b blocks' can easily be misread as an overcount; readers should be told explicitly that the target partition is on the remaining m-1 elements and that the insertion block is the original block of v.
- [Section 3, Lemma 3.3] In the proof of part (2), the sentence 'The flat X∨e has rank rk(X)+1 and lies below S' is correct only because every atom e not below X is a flat of rank one; this could be stated explicitly for clarity.
- [Section 4.1, Remark 1.4 lower-bound construction] The construction proves that no threshold below 2k-t+1 can work for all larger n; the wording 'at least 2k−t+1' is slightly imprecise and could be phrased as 'the threshold must be at least 2k−t+1' with the definition of n0(k,t) made explicit before the remark.
Circularity Check
No significant circularity: the main theorems are derived from lemmas re-proved in the paper and explicit rational estimates; the only flagged issue is a non-circular arithmetic gap involving r0.
full rationale
I find no circularity. The central claims (Theorems 1.2, 1.3, and 1.5) are proved through a chain of lemmas that are derived inside the paper rather than assumed as black boxes with the same conclusion. The rank-spread peeling machinery is formalized and proved in Section 3, including the maximal-link lemma, the partition-lattice peeling lemma, and the uniform-parameter version. The occupancy model, exponential tilting, modal estimates, and dense-tail bounds in Section 5 and Appendix A are all proved from first principles with explicit constants. The singleton-switching bound in Lemma 5.10 is a problem-specific counting argument, and the final inequality |A| < E1 is obtained by combining the sparse and dense estimates, not by invoking the theorem being proved. No fitted parameter is renamed as a prediction, and no load-bearing conclusion is imported from the authors' own prior work; in fact the authors cite no prior papers of their own. External references such as [4], [10], [12], [14], and [15] are used for context, comparison, or framework, but the deterministic ingredients actually used are re-proved in the manuscript. The constants 8, c_t, and 3(m^3 + binom(m,2)) are thresholds obtained by closing inequalities in Lemmas A.2 through A.6, so they are not fitted to the target results. Remark 6.1 explicitly describes the N >= 3k barrier as a limitation of the method, which is transparent and not circular. I also flag a non-circular correctness issue in the proof of Theorem 1.2: r0 is defined as ceil(7b/10), but the proof asserts the chain mu+1 <= r0 <= 7b/10, and the second inequality is false when 7b/10 is not an integer. This affects the application of Corollary 5.3 as written, but it is an arithmetic gap in closing an estimate, not a reduction of the theorem to its own inputs, so it does not raise the circularity score.
Assumptions & free parameters
assumptions (2)
- standard math Flats of the graphic matroid M(K_{n+1}) are in bijection with partitions of an (n+1)-set, with rank equal to n+1 minus the number of blocks.
- standard math The sequence 1/(j+1)! is log-concave and convolution preserves log-concavity.
Cite this review
Pith. "Pith review of Erd\H{o}s--Ko--Rado and Hilton--Milner Theorems in the Partition Lattice." pith.science (2026). https://pith.science/paper/UURFLIQP
@misc{pith2026260805951,
author = {Pith},
title = {Pith review of: Erd\Hos--Ko--Rado and Hilton--Milner Theorems in the Partition Lattice},
year = {2026},
howpublished = {\url{https://pith.science/paper/UURFLIQP}},
note = {Machine review of arXiv:2608.05951}
}
abstract
Let $M_n=M(K_{n+1})$ be the graphic matroid of the complete graph, and let $\mathcal{F}_k(M_n)$ be its rank-$k$ flats. We study families $\mathcal{A}\subseteq\mathcal{F}_k(M_n)$ satisfying $\mathrm{rk}(A\wedge B)\ge t$ for all $A,B\in\mathcal{A}$. For $t=1$, this problem is exactly equivalent to Czabarka's partition-EKR conjecture, first introduced in print by P.~L. Erd\H{o}s and L.~A. Sz\'ekely~\cite{ErdosSzekelyHigher}. We prove the corresponding Erd\H{o}s--Ko--Rado theorem in the explicit linear range $n+1\ge8k$, giving a constant-factor advance toward the conjectured sharp range $n\ge2k$. For every fixed $t$, we further prove an Erd\H{o}s--Ko--Rado theorem under an explicit condition of order $O_t(k^2)$ on the block number $n+1-k$, with equality only for a full $t$-star. We also determine the largest nontrivial intersecting families under an explicit $O(k^6)$ threshold and characterize the unique extremal family up to isomorphism.
Reference graph
Works this paper leans on
-
[10]
Kupavskii, Erd˝ os–Ko–Rado type results for partitions via spread approximations,European J
A. Kupavskii, Erd˝ os–Ko–Rado type results for partitions via spread approximations,European J. Combin.132, Part B (2026), Article 104288
work page 2026
- [15]
-
[1]
R. Alweiss, S. Lovett, K. Wu, and J. Zhang, Improved bounds for the sunflower lemma,Ann. of Math. (2)194(2021), 795–815
work page 2021
-
[2]
R. Ahlswede and L. H. Khachatrian, The complete intersection theorem for systems of finite sets,European J. Combin.18(1997), 125–136
work page 1997
-
[3]
P. Erd˝ os, C. Ko, and R. Rado, Intersection theorems for systems of finite sets,Quart. J. Math. Oxford Ser. (2)12(1961), 313–320
work page 1961
-
[4]
P. L. Erd˝ os and L. A. Sz´ ekely, Erd˝ os–Ko–Rado theorems of higher order, inNumbers, Informa- tion and Complexity, Kluwer Academic Publishers, Boston, 2000, 117–124
work page 2000
-
[5]
K. Frankston, J. Kahn, B. Narayanan, and J. Park, Thresholds versus fractional expectation- thresholds,Ann. of Math. (2)194(2021), 475–495
work page 2021
-
[6]
A. J. W. Hilton and E. C. Milner, Some intersection theorems for systems of finite sets,Quart. J. Math. Oxford Ser. (2)18(1967), 369–384
work page 1967
Show all 18 references
-
[7]
S. G. Hoggar, Chromatic polynomials and logarithmic concavity,J. Combin. Theory Ser. B16 (1974), 248–254
1974
-
[8]
Meagher and L
K. Meagher and L. Moura, Erd˝ os–Ko–Rado theorems for uniform set-partition systems,Electron. J. Combin.12(2005), Research Paper 40, 12 pp
2005
-
[9]
Meagher, M
K. Meagher, M. N. Shirazi, and B. Stevens, An extension of the Erd˝ os–Ko–Rado theorem to uniform set partitions,Ars Math. Contemp.23(2023), Paper No. 4.02
2023
-
[11]
C. Y. Ku and K. B. Wong, An analogue of the Hilton–Milner theorem for set partitions,J. Combin. Theory Ser. A120(2013), 1508–1520
2013
-
[12]
Kupavskii and D
A. Kupavskii and D. Zakharov, Spread approximations for forbidden intersections problems, Adv. Math.445(2024), Article 109653
2024
-
[13]
Rao, Coding for sunflowers,Discrete Anal.(2020), Paper No
A. Rao, Coding for sunflowers,Discrete Anal.(2020), Paper No. 2, 8 pp
2020
-
[14]
Ihringer and A
F. Ihringer and A. Kupavskii, Structure of t-intersecting families of vector spaces, preprint, arXiv:2605.02698 (2026)
2026 arXiv
-
[16]
Wen and B
J. Wen and B. Lv, A unified approach to cross-intersection problems with applications to Hilton–Milner type theorems and stability, preprint, arXiv:2607.03315 (2026)
2026 arXiv
-
[17]
Oxley,Matroid Theory, 2nd ed., Oxford Graduate Texts in Mathematics 21, Oxford University Press, Oxford, 2011
J. Oxley,Matroid Theory, 2nd ed., Oxford Graduate Texts in Mathematics 21, Oxford University Press, Oxford, 2011
2011
-
[18]
Whitney, On the abstract properties of linear dependence,Amer
H. Whitney, On the abstract properties of linear dependence,Amer. J. Math.57(1935), 509–533. 26 A Technical estimates for the EKR bounds This appendix contains only the one-variable inequalities and finite rational estimates used in the EKR proofs. All combinatorial reductions...
1935
Reviewed August 7, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.