A mathematical preprint has derived explicit upper bounds for a tightly constrained kind of set family. The family consists of subsets of [n] with four conditions: it is nonempty, 3-wise intersecting, regular and increasing. In ordinary terms, 3-wise intersection requires every three selected members to share an element. The theorem says that any family meeting these conditions must obey a quantitative limit, expressed as log(2^n/|A|) ≥ (n/2)(|A|/(2^n−|A|))².
The notation compares |A|, the number of sets in the family, with 2^n, the number of all subsets of an n-element ground set. The left side is a logarithmic measure of the gap between the family and the full power set. Rather than reporting an observed average or a probability drawn from data, the statement applies to every finite family in the specified class.
The main inequality can be inverted. The preprint gives the explicit form |A| ≤ 2^n√(W(n)/n), where W is the principal Lambert branch, defined by W(x)e^{W(x)}=x for x≥0. This turns the implicit logarithmic constraint into a direct ceiling on the number of sets the family may contain.
For n≥3, the same result is recast as log_2|A| ≤ n−(1/2)log_2 n+(1/2)log_2 log n. The reformulation puts the conclusion on a base-2 logarithmic scale: it starts with n, subtracts half of log_2 n and adds a correction involving log_2 log n.
A theorem about structure, not a dataset
This is a methods paper, not an empirical study. Its mathematical population is the class of nonempty A⊆P_n that are 3-wise intersecting, regular and increasing. There are no sampled entities in the analysis; the conclusions range over any family satisfying the premises. The result therefore describes what these assumptions force, rather than what happened in a measured group.
The bounds are deterministic consequences of the stated conditions, so they do not come with sampling uncertainty to quantify. The note does not present a measured effect, a statistical test or a confidence interval; its conclusions are inequalities that hold for the mathematical class under consideration.
How the argument gets there
To make the problem tractable, the note identifies a set family with a Boolean function on {0,1}^n and, for one part of the argument, with a subset of F_2^n. The 3-wise intersection condition implies a sum-free property in F_2^n. From that property, the proof obtains the cubic Fourier identity ∑_{S⊆[n]} f̂(S)^3=0.
The argument combines that identity with the Fourier–Walsh expansion, Parseval's identity, coordinate influences and total influence. Regularity is used in a specific way: all coordinate influences of the Boolean function are equal, and every nonconstant Fourier coefficient is at most half of the relevant influence in absolute value.
Writing α for the density of A in the Boolean cube, one Fourier/Parseval step gives I[f] ≥ 2·n·α²/(1−α) ≥ 2·n·α². A second Parseval calculation gives I[f]≤2√(nα(1−α)). The preprint combines these inequalities to obtain the direct, weaker upper bound |A|≤2^n/(1+n^(1/3)).
A separate entropy refinement starts with a uniformly distributed member of A. Because the coordinate-inclusion probability is common across coordinates, the proof can call it θ and apply entropy subadditivity, obtaining log|A|≤n h(θ), where h is the relevant Bernoulli entropy. This route feeds into the main logarithmic inequality and its Lambert-W form.
The two bounds answer the same size question at different levels of explicitness. The Lambert-W version retains the sharper form of the main theorem; the influence argument yields a simpler expression involving n^(1/3), but the paper labels it weaker. Both are universal upper bounds for the same regular, increasing, 3-wise intersecting class.
Where the result can go next
The paper says both theorem statements extend to symmetric 3-wise intersecting families. The route is upward closure: for such a family, its upward closure is 3-wise intersecting, regular and increasing, bringing it within the scope of the two bounds.
The authors identify a main source of information loss in the proof: it discards the signs of Fourier coefficients and the way Fourier mass is distributed across levels. Retaining those details is the route the note points toward for seeking sharper explicit estimates.
These are results about a mathematical class itself, not about a measured population. The supplied metadata identifies the document as an arXiv preprint, version v1, dated 20 Aug 2026.
An author footnote reports support from the National Natural Science Foundation of China under grant 124B2019 and the Institute for Basic Science under grant IBS-R029-C4.
Paper data and sources
Original title: Quantitative bounds for regular $3$-wise intersecting families
Authors: Fan Chang
Journal/Repository: arXiv
Status: Preprint, not yet peer-reviewed
First online: 2026-08-20
DOI: Not available
Original paper · Full text