Erdős–Simonovits conjecture for the asymptotic number of triangle- and four-cycle-free edges

Determine whether the limit of the normalized extremal function for graphs containing neither a triangle nor a four-cycle exists with value \(1/(2\sqrt{2})\), namely, whether \(\lim_{n\to\infty}\mathrm{ex}_2(n,\{C_3,C_4\})/n^{3/2}=1/(2\sqrt{2})\).

Background

The paper places this question in the general context of extremal graph theory, where one first seeks the polynomial order of an extremal function and then asks whether a precise leading asymptotic constant exists. It states that ex2(n,{C3,C4})=Θ(n3/2)\mathrm{ex}_2(n,\{C_3,C_4\})=\Theta(n^{3/2}) is known, but that the sharper asymptotic constant remains conjectural.

The problem concerns the maximum number of edges in an nn-vertex graph containing no triangle C3C_3 and no four-cycle C4C_4. The conjectured limit would identify the exact leading constant in its n3/2n^{3/2}-scale growth.

References

For example, it is known that \mathrm{ex}2(n,{C_3,C_4})=\Theta(n{3/2}), while a conjecture of Erd\H{o}s from 1975 asserts that \lim{n\to\infty}\frac{\mathrm{ex}2(n,{C_3,C_4})}{n{3/2}=\frac{1}{2\sqrt{2}, where C\ell denotes a cycle of length \ell.

Extremal problems for cancellative and locally thin hypergraphs  (2609.08858 - Liu et al., 8 Sep 2026) in Introduction, paragraph beginning “A sharper problem is to determine whether”