Spectral Erdős–Ko–Rado problem for uniform hypergraphs

For integers n≥kn\ge k and 1≤t<k1\le t<k, determine

max⁡{ρ(F):F⊆([n]k), ∣F∩F′∣≥t for all F,F′∈F},\max\left\{\rho(\mathcal F):\mathcal F\subseteq\binom{[n]}k,\ |F\cap F'|\ge t\text{ for all }F,F'\in\mathcal F\right\},

where ρ(F)\rho(\mathcal F) is the spectral radius of the adjacency tensor of the kk-uniform hypergraph with edge set F\mathcal F. Characterize all extremal families attaining this maximum. In particular, determine for which parameters the complete tt-star

A0={F∈([n]k):[t]⊆F}\mathcal A_0=\{F\in\binom{[n]}k:[t]\subseteq F\}

is spectrally extremal, and whether the competing extremal families are the Frankl families.

References

Primary source

arXiv

Progress summary

Refreshed
Claimed progress

A new paper proves only a substantial portion of the conjectured spectral pattern, so the problem remains open outside its stated parameter ranges.

The problem asks whether the asymptotic spectral Erdős–Ko–Rado theorem extends to all parameter ranges. Existing work proves the star is optimal only when the ground set is sufficiently large, while smaller or exceptional regimes may have different extremal families.

Known results

  • Keevash, Lenz, and Mubayi: for fixed kk, tt, and α>1\alpha>1, sufficiently large nn forces the complete tt-star to uniquely maximize λα\lambda_\alpha.
  • The same work gives stability: sufficiently large spectral radius forces a family to be a tt-star.
  • A separate asymptotic result gives the t=1t=1 case for intersecting uniform families, with unique star-type extremality for sufficiently large nn.

August 2026 partial theorem

On August 25, 2026, a report on A Sharp Spectral Erdős--Ko--Rado Theorem for Uniform Hypergraphs stated that the paper characterizes competition between the first two Frankl families for 2≤t<k2\leq t<k, gives a sufficient threshold for unique tt-star extremality, and determines all t=1t=1 extremal structures for n≥2kn\geq2k. The result is explicitly restricted to these ranges and does not settle every n,k,tn,k,t case.

Current status (as of August 2026): A claimed partial advance covers the stated parameter ranges, but the full problem remains open for the omitted n,k,tn,k,t regimes.

Sources

Solutions 0

No solutions have been posted yet.