Sah–Sawhney–Stoner–Zhao conjecture on antiferromagnetic homomorphism maxima

Let HH be a dd-regular graph, and let GG be a graph possibly with loops. Regard GG as its symmetric adjacency matrix, and suppose that it has at most one positive eigenvalue. Sah–Sawhney–Stoner–Zhao conjecture. Then

hom⁡(H,G)1/v(H)≤hom⁡(Kd,d,G)1/(2d).\hom(H,G)^{1/v(H)} \leq \hom(K_{d,d},G)^{1/(2d)}.

This would generalize the Kahn–Zhao theorem and the corresponding result for complete target graphs, asserting that complete bipartite graphs maximize normalized homomorphism counts from regular graphs under the stated spectral condition. The conjecture is presented as open in the paper.

References

Primary source

Joonkyung Lee, Jaeseong Oh and Jaehyeon Seo, “Counting homomorphisms in antiferromagnetic graphs via Lorentzian polynomials”, arXiv:2506.13659 (2025).

Progress summary

Refreshed
Claimed progress

A 2025 paper proves the conjecture in several structured cases, but the general question remains open.

The conjecture asks whether, for every dd-regular graph HH and every graph GG with at most one positive eigenvalue, the normalized homomorphism count is maximized by Kd,dK_{d,d}. The antiferomagnetic extension was formulated by Sah, Sawhney, Stoner, and Zhao and remains unresolved in general.

Known results

  • Zhao, 2018: the complete-target case, including some looped targets, was known; the general extension was stated as a conjecture.
  • Sah, Sawhney, Stoner, and Zhao, 2018: the complete-bipartite extremal bound was proved for triangle-free source graphs, including proper colorings.
  • The 2025 paper confirms the relevant inequality for semiproper colorings, namely complete targets possibly with loops.

2025 partial progress

The paper Counting homomorphisms in antiferromagnetic graphs via ... proves hom⁡(H,G)2≤hom⁡(H×K2,G)\hom(H,G)^2\leq\hom(H\times K_2,G) when HH is formed by blowing up vertices of a bipartite graph into cliques, with further cases for G=KqG=K_q. These results are presented as progress toward the conjecture, not a complete proof. A 2026 clique-minimization theorem concerns a different inequality and does not settle this problem.

Current status (as of September 2026): The conjecture is open in general, with partial results for structured source graphs and complete or looped complete targets.

Sources

Solutions 0

No solutions have been posted yet.