Documentation

FormalConjectures.ErdosProblems.«533»

Erdős Problem 533 #

References:

theorem Erdos533.erdos_533 :
False ∀ (δ : ), 0 < δ∃ (c : ), 0 < c ∀ᶠ (n : ) in Filter.atTop, ∀ (G : SimpleGraph (Fin n)), G.CliqueFree 5δ * n ^ 2 G.edgeFinset.card∃ (S : Finset (Fin n)), c * n S.card G.CliqueFreeOn (↑S) 3

Let $\delta > 0$. If $n$ is sufficiently large and $G$ is a graph on $n$ vertices with no $K_5$ and at least $\delta n^2$ edges, must $G$ contain a set of $\gg_\delta n$ vertices spanning no triangle?

Equivalently, writing $\mathrm{RT}_3(n, K_5, m)$ for the maximum number of edges of a $K_5$-free graph on $n$ vertices in which every triangle-free vertex set has fewer than $m$ vertices (the triangle Ramsey–Turán number), is $$\delta_3(5) = \lim_{\epsilon \to 0} \lim_{n \to \infty} \frac{\mathrm{RT}_3(n, K_5, \epsilon n)}{n^2} = 0?$$

This is a problem of Erdős, Hajnal, Simonovits, Sós, and Szemerédi [EHSSS94], who proved $\delta_3(5) \leq 1/12$ and the analogous $\delta_3(4) = 0$, and observed $\delta_3(7) \geq 1/4$ via a construction of Erdős and Rogers [ErRo62].

The answer is no: Balogh and Lenz [BaLe13] disproved it by showing $\delta_3(5) > 0$, and the exact value $\delta_3(5) = 1/12$ was determined by the matching lower-bound construction of Liu, Reiher, Sharifzadeh, and Staden [LRSS21] (see erdos_533.variants.lrss_lower).

theorem Erdos533.erdos_533.variants.ehsss_upper (ε : ) ( : 0 < ε) :
∃ (δ : ), 0 < δ ∀ᶠ (n : ) in Filter.atTop, ∀ (G : SimpleGraph (Fin n)), G.CliqueFree 5(∀ (S : Finset (Fin n)), G.CliqueFreeOn (↑S) 3S.card δ * n)G.edgeFinset.card (1 / 12 + ε) * n ^ 2

The upper bound $\delta_3(5) \leq 1/12$ of Erdős, Hajnal, Simonovits, Sós, and Szemerédi [EHSSS94]: for every $\epsilon > 0$ there is a $\delta > 0$ such that for all sufficiently large $n$, every $K_5$-free graph $G$ on $n$ vertices in which every triangle-free vertex set has at most $\delta n$ vertices has at most $(1/12 + \epsilon)n^2$ edges.

theorem Erdos533.erdos_533.variants.lrss_lower (ε δ : ) ( : 0 < ε) ( : 0 < δ) :
∀ᶠ (n : ) in Filter.atTop, ∃ (G : SimpleGraph (Fin n)), G.CliqueFree 5 (∀ (S : Finset (Fin n)), G.CliqueFreeOn (↑S) 3S.card δ * n) (1 / 12 - ε) * n ^ 2 G.edgeFinset.card

The matching lower bound $\delta_3(5) \geq 1/12$, from the construction of Liu, Reiher, Sharifzadeh, and Staden [LRSS21] (improving the earlier $\delta_3(5) > 0$ of Balogh and Lenz [BaLe13]): for every $\epsilon, \delta > 0$ and all sufficiently large $n$ there is a $K_5$-free graph $G$ on $n$ vertices in which every triangle-free vertex set has at most $\delta n$ vertices, yet which has at least $(1/12 - \epsilon)n^2$ edges. In particular this refutes erdos_533.

theorem Erdos533.erdos_533.variants.delta_four_eq_zero (δ : ) :
0 < δ∃ (c : ), 0 < c ∀ᶠ (n : ) in Filter.atTop, ∀ (G : SimpleGraph (Fin n)), G.CliqueFree 4δ * n ^ 2 G.edgeFinset.card∃ (S : Finset (Fin n)), c * n S.card G.CliqueFreeOn (↑S) 3

The contrasting positive result $\delta_3(4) = 0$ of Erdős, Hajnal, Simonovits, Sós, and Szemerédi [EHSSS94]: the $K_4$ analogue of erdos_533 is true. For every $\delta > 0$ there is a $c > 0$ such that for all sufficiently large $n$, every $K_4$-free graph $G$ on $n$ vertices with at least $\delta n^2$ edges contains a triangle-free vertex set of size at least $c n$.

theorem Erdos533.erdos_533.variants.delta_seven_ge_quarter (ε δ : ) ( : 0 < ε) ( : 0 < δ) :
∀ᶠ (n : ) in Filter.atTop, ∃ (G : SimpleGraph (Fin n)), G.CliqueFree 7 (∀ (S : Finset (Fin n)), G.CliqueFreeOn (↑S) 3S.card δ * n) (1 / 4 - ε) * n ^ 2 G.edgeFinset.card

The observation $\delta_3(7) \geq 1/4$ of Erdős, Hajnal, Simonovits, Sós, and Szemerédi [EHSSS94], via a construction of Erdős and Rogers [ErRo62]: for every $\epsilon, \delta > 0$ and all sufficiently large $n$ there is a $K_7$-free graph $G$ on $n$ vertices in which every triangle-free vertex set has at most $\delta n$ vertices, yet which has at least $(1/4 - \epsilon)n^2$ edges.

Sanity check for the triangle-free-set condition: in the empty graph every vertex set spans no triangle, since the empty graph has no $3$-clique.