This paper was converted on www.awesomepapers.org from LaTeX by an anonymous user.
Want to know more? Visit the Converter page.


Ramsey-Turán numbers for intersecting odd cliques

Guantao Chen  and Min Liu
Department of Mathematics and Statistics, Georgia State University, Atlanta, GA 30303
College of Economics, Northwest University of Politics and Law, Xi’an, Shaanxi, China 710063
Supported in part by NSF grant DMS-1855716Supported in part by the Research Foundation of College of Economics, Northwest University of Politics and Law (No. 19XYKY07), and by China Scholarship Council (No. 201808610115).
Abstract

Given a graph HH and a function f:++f:\mathbb{Z}^{+}\longrightarrow\mathbb{Z}^{+}, the Ramsey-Turán number of HH and ff, denoted by RT(n,H,f(n))RT(n,H,f(n)), is the maximum number of edges a graph GG on nn vertices can have, which does not contain HH as a subgraph and also does not contain a set of f(n)f(n) independent vertices. Let rr be a positive integer. In 1969, Erdős and Sós proved that RT(n,K_2r+1,o(n))=n22(11r)+o(n2)RT(n,K_{\_}{2r+1},o(n))=\frac{n^{2}}{2}(1-\frac{1}{r})+o(n^{2}). Let F_k(2r+1)F_{\_}k(2r+1) denote the graph consisting of kk copies of complete graphs K_2r+1K_{\_}{2r+1} sharing exactly one vertex. In this paper, we show that RT(n,F_k(2r+1),o(n))=n22(11r)+o(n2)RT(n,F_{\_}k(2r+1),o(n))=\frac{n^{2}}{2}(1-\frac{1}{r})+o(n^{2}), which is of the same magnitude with RT(n,K_2r+1,o(n))RT(n,K_{\_}{2r+1},o(n)).

Keywords:    Turán number, Ramsey-Turán number, odd cliques, independent set


1 Introduction

Let HH be a graph. A graph GG is called HH-free if GG does not contain a copy of HH as a subgraph. For a positive integer nn, let ex(n,H)ex(n,H) denote the maximum number of edges of an HH-free graph of order nn, and call such a graph an extremal graph for HH. For any positive integer rr, let K_rK_{\_}r denote the nn-vertex complete graph. Turán’s classical result states that the balanced complete nn-vertex rr-partite graph, denoted by T_r(n)T_{\_}r(n), is the unique K_r+1K_{\_}{r+1}-extremal graph of order nn. A graph on 2k+12k+1 vertices consisting of kk triangles which intersect in exactly one common vertex is called a kk-fan and denoted by F_kF_{\_}k. Erdős, Füredi, Gould, and Gunderson [2] determined ex(n,F_k)ex(n,F_{\_}k) for n50k2n\geq 50k^{2}. A graph on (r1)k+1(r-1)k+1 vertices consisting of kk copies of K_rK_{\_}r sharing exactly one common vertex is called a (k,r)(k,r)-fan and denoted by F_k(r)F_{\_}k(r). Chen, Gould, Pfender, and Wei [1] determined ex(n,F_k(r))ex(n,F_{\_}k(r)) for n16k3r8n\geq 16k^{3}r^{8}.

Notice that the independence number α(T_r(n))\alpha(T_{\_}r(n)) is at least nr\frac{n}{r}. Given a graph HH and a function f:++f:\ \mathbb{Z}^{+}\longrightarrow\mathbb{Z}^{+}, the Ramsey-Turán number for a large positive integer nn with regard to HH and f(n)f(n), denoted by RT(n,H,f(n))RT(n,H,f(n)), is the maximum number of edges of an nn-vertex HH-free graph GG with α(G)f(n)\alpha(G)\leq f(n). Erdős and Sós in 1969 [3] studied of Ramsey-Turán numbers and showed that RT(n,K_2r+1,o(n))=n22(11r)+o(n2)RT(n,K_{\_}{2r+1},o(n))=\frac{n^{2}}{2}(1-\frac{1}{r})+o(n^{2}), which is approximately the value of ex(n,K_r+1)ex(n,K_{\_}{r+1}). The study of Ramsey-Turán numbers has drawn a great deal of attention over the last 40 years; see the survey by Simonovits and Sós [7] and [4, 5, 6] for more results on various Ramsey-Turán problems. Inspired by the study of extremal numbers ex(n,F_k)ex(n,F_{\_}k) and ex(n,F_k(r))ex(n,F_{\_}k(r)) in [2] and [1], respectively, we consider Ramsey-Turán numbers for F_k(r)F_{\_}k(r) and obtained the following result.

Theorem 1.

For any two fixed positive integers kk and rr,

RT(n,F_k(2r+1),o(n))=n22(11r)+o(n2).{\displaystyle RT(n,F_{\_}{k}(2r+1),o(n))=\frac{n^{2}}{2}\left(1-\frac{1}{r}\right)+o(n^{2}).}

In the next section we will first give a short proof of the case r=1r=1, i.e., RT(n,F_k(3),o(n))=o(n2)RT(n,F_{\_}{k}(3),o(n))=o(n^{2}), and then prove the general case. We now introduce notations and terminology that will be used in the proof.

All graphs considered in this paper are simple graphs. Let G=(V(G),E(G))G=(V(G),E(G)) be a graph with vertex set V(G)V(G) and edge set E(G)E(G). Denote by |G||G| and G||G|| for |V(G)||V(G)| and |E(G)||E(G)|, respectively; and call them the order and the size of GG, respectively. Denote by G(n)G(n) a graph of order nn and denote by G(n,m)G(n,m) a graph of order nn and size mm. For a vertex xV(G)x\in V(G), the neighborhood of xx in GG, denoted by N_G(x)N_{\_}G(x), is the set of vertices adjacent to xx, i.e., N_G(x)={yV(G):xyE(G)}N_{\_}{G}(x)=\{y\in V(G):xy\in E(G)\}. We use N(x)N(x) for N_G(x)N_{\_}G(x) when the referred graph GG is clear. The degree of xx in GG, denoted by d_G(x)d_{\_}G(x), or d(x)d(x), is the cardinality of N_G(x)N_{\_}{G}(x), i.e., d_G(x)=|N_G(x)|d_{\_}G(x)=|N_{\_}G(x)|. Denote by δ(G)\delta(G) the minimum degree in GG. For a subset XV(G)X\subseteq V(G), let G[X]G[X] denote the subgraph of GG induced by XX, i.e., a subgraph of GG with vertex set XX and the set of all edges with two ends in XX. For a vertex set X={x_1,x_2,,x_r}X=\{x_{\_}1,x_{\_}2,\dots,x_{\_}r\}, denote by G[x_1,x_2,,x_r]G[x_{\_}1,x_{\_}2,\dots,x_{\_}r] for G[{x_1,x_2,,x_r}]G[\{x_{\_}1,x_{\_}2,\dots,x_{\_}r\}].

The matching number of a graph GG, denoted by ν(G)\nu(G), is the maximum number of edges in a matching in GG. Let α(G)\alpha(G) denote the independence number of GG. Clearly, if GG is an nn-vertex graph, then α(G)n2ν(G)\alpha(G)\geq n-2\nu(G).

2 Proof of Theorem 1

2.1 Prilimilary

We first consider triangles sharing a common vertex and prove the following simple fact.

Lemma 1.

For any fixed positive integer kk, RT(n,F_k(3),o(n))=o(n2)RT(n,F_{\_}{k}(3),o(n))=o(n^{2}).

Proof.

Let GG be an nn-vertex F_k(3)F_{\_}{k}(3)-free graph with independence number α(G)ϵn\alpha(G)\leq\epsilon n for some small positive real number ϵ\epsilon. For any xV(G)x\in V(G), since GG does not contain kk edge-disjoint triangles intersecting in the vertex xx, the neighborhood of xx contains at most k1k-1 independent edges, i.e., ν(G[N(x)])k1\nu(G[N(x)])\leq k-1. Thus, α(G[N(x)])d(x)2(k1)\alpha(G[N(x)])\geq d(x)-2(k-1). Since α(G[N(x)])α(G)ϵn\alpha(G[N(x)])\leq\alpha(G)\leq\epsilon n, we have d(x)ϵn+2(k1)d(x)\leq\epsilon n+2(k-1). Counting the total degree of GG, we have the following inequality: |E(G)|12(ϵ+2(k1)/n)n2|E(G)|\leq\frac{1}{2}(\epsilon+2(k-1)/n)n^{2}. Since lim_n12(ϵ+2(k1)/n)=12ϵ\lim_{\_}{n\rightarrow\infty}\frac{1}{2}(\epsilon+2(k-1)/n)=\frac{1}{2}\epsilon, we complete the proof of Lemma 1. ∎

The following result from [3] is needed in our proof and for completeness we give its proof here.

Lemma 2 (Erdős and Sós).

Let β\beta and ϵ\epsilon be two positive numbers with 0<β<120<\beta<\frac{1}{2} and 0<ϵ<10<\epsilon<1. Then for every positive number cc with 0<c<βϵ0<c<\sqrt{\beta\epsilon} and any graph G(n)G(n) with |E(G)|βn2(1+ϵ)|E(G)|\geq\beta n^{2}(1+\epsilon), there is a subgraph G(m)G(n)G(m)\subseteq G(n) with m>cnm>cn and δ(G(m))>2β(1+ϵ/2)m\delta(G(m))>2\beta(1+\epsilon/2)m.

Proof.

To avoid cumbersome notation, we assume without loss of generality that cncn is an integer. Suppose on the contrary a such graph G(m)G(m) does not exist. Denote the graph G(n)G(n) by GG with |E(G)|βn2(1+ϵ)|E(G)|\geq\beta n^{2}(1+\epsilon). Then the vertices of GG can be written in a sequence x_1,x_2,,x_nx_{\_}1,x_{\_}2,\dots,x_{\_}n so that for every i(1c)ni\leq(1-c)n the degree of x_ix_{\_}i in G[x_i,x_i+1,,x_n]G[x_{\_}i,x_{\_}{i+1},\dots,x_{\_}n] is less than 2β(1+ϵ/2)(ni+1)2\beta(1+\epsilon/2)(n-i+1).

βn2(1+ϵ)|E(G)|\displaystyle\beta n^{2}(1+\epsilon)\leq|E(G)| \displaystyle\leq 2β(1+ϵ2)(n+(n1)++(cn+1))+(cn2)\displaystyle 2\beta\left(1+\frac{\epsilon}{2}\right)(n+(n-1)+\dots+(cn+1))+{cn\choose 2}
<\displaystyle< 2β(1+ϵ2)(n22)+c2n22\displaystyle 2\beta\left(1+\frac{\epsilon}{2}\right)\left(\frac{n^{2}}{2}\right)+\frac{c^{2}n^{2}}{2}
<\displaystyle< βn2(1+ϵ) (since c<βϵ),\displaystyle\beta n^{2}(1+\epsilon)\qquad\mbox{ (since $c<\sqrt{\beta\epsilon}$), }

which gives a contradiction. ∎

Lemma 3.

Let rr be a positive integer. For any ϵ>0\epsilon>0, there exists positive integer N=N(ϵ)N=N(\epsilon) such that for a clique DD of a graph G(n)G(n) with n>Nn>N and |D|r|D|\leq r, if d(v)(11r+ϵ3)nd(v)\geq\left(1-\frac{1}{r}+\frac{\epsilon}{3}\right)n for each vDv\in D, then there exists a clique DD^{*} that contains DD as a proper subset.

Proof.

Let mm denote the number of vertices in V(G)DV(G)-D that are adjacent to all vertices in DD. It is sufficient to show that m>0m>0. By counting the edges between DD and GDG-D, we have the following inequality.

|D|m+(|D|1)(n|D|m)_vD(d(v)(|D|1))|D|((11r+ϵ3)n(|D|1))\displaystyle|D|m+(|D|-1)(n-|D|-m)\geq\sum_{\_}{v\in D}(d(v)-(|D|-1))\geq|D|\left(\left(1-\frac{1}{r}+\frac{\epsilon}{3}\right)n-(|D|-1)\right)

Thus, m|D|(11r+ϵ3)n(|D|1)n=n(1rϵ3)|D|n>0m\geq|D|\left(1-\frac{1}{r}+\frac{\epsilon}{3}\right)n-(|D|-1)n=n-\left(\frac{1}{r}-\frac{\epsilon}{3}\right)|D|n>0, where in the last inequality we used |D|r|D|\leq r. ∎

Lemma 4.

Let rr be a positive integer. For any ϵ>0\epsilon>0, there exist δ=ϵ4\delta=\frac{\epsilon}{4} and N=N(ϵ)N=N(\epsilon) such that for any clique DD of a graph G(n)G(n) with n>Nn>N and α(G(n))δn\alpha(G(n))\leq\delta n, if |D|2r|D|\leq 2r and d(v)(11r+ϵ3)nd(v)\geq(1-\frac{1}{r}+\frac{\epsilon}{3})n for each vDv\in D, then there exists a clique DD^{*} with |D|=|D|+1|D^{*}|=|D|+1 such that |DD||D|1|D^{*}\cap D|\geq|D|-1.

Proof.

We set |D|=s2r|D|=s\leq 2r and G=G(n)G=G(n). If there is a vertex vV(G)Dv\in V(G)-D that is adjacent to all vertices in DD, then D{v}D\cup\{v\} is the desired clique. So, we assume that d_D(v)s1d_{\_}D(v)\leq s-1 for all vV(G)Dv\in V(G)-D. We divide V(GD)V(G-D) into two vertex-disjoint sets UU and WW such that

U\displaystyle U =\displaystyle= {vV(G)D|d_D(v)s2}, and\displaystyle\{v\in V(G)-D\ |\ d_{\_}D(v)\leq s-2\},\mbox{ and}
W\displaystyle W =\displaystyle= {vV(G)D)|d_D(v)=s1}.\displaystyle\{v\in V(G)-D)\ |\ d_{\_}D(v)=s-1\}.

Computing the edges between DD and V(G)DV(G)-D, we get the following inequality.

(s2)|U|+(s1)|W|_vDd_GD(v)s((11r+ϵ3)n(s1))\displaystyle(s-2)|U|+(s-1)|W|\geq\sum_{\_}{v\in D}d_{\_}{G-D}(v)\geq s\left(\left(1-\frac{1}{r}+\frac{\epsilon}{3}\right)n-(s-1)\right)

By assuming nN>3ϵn\geq N>\frac{3}{\epsilon} and applying the inequality s2rs\leq 2r, we get

|W|(2sr+sϵ3sn)n>(sϵ3sn)n>0.|W|\geq\left(2-\frac{s}{r}+s\frac{\epsilon}{3}-\frac{s}{n}\right)n>\left(s\frac{\epsilon}{3}-\frac{s}{n}\right)n>0.

Let D={v_1,v_2,,v_s}D=\{v_{\_}1,v_{\_}2,\dots,v_{\_}s\} and divide WW into ss vertex-disjoint sets as W_1,W_2,W_sW_{\_}1,W_{\_}2,\dots W_{\_}s such that W_i={wW|v_iwE(G)}W_{\_}i=\{w\in W\ |\ v_{\_}iw\notin E(G)\} for each 1is1\leq i\leq s. We claim W_iW_{\_}i is an independent set. Otherwise, let xx and yy be two adjacent vertices in W_iW_{\_}i. Then, D=D\{v_i}{x,y}D^{*}=D\backslash\{v_{\_}i\}\cup\{x,y\} is the desired clique. Hence,

α(G)max{|W_i|,1is}>(ϵ31n)n>δn,\alpha(G)\geq max\{|W_{\_}i|,{1\leq i\leq s}\}>\left(\frac{\epsilon}{3}-\frac{1}{n}\right)n>\delta n,

giving a contradiction. ∎

2.2 Proof of Theorem 1

Recall the graph F_k(2r+1)F_{\_}{k}(2r+1) is the union of kk copies of K_2r+1K_{\_}{2r+1} sharing a common vertex. We call the common vertex the center of F_k(2r+1)F_{\_}{k}(2r+1). For convention, if k=0k=0, F_k(2r+1)F_{\_}{k}(2r+1) is a single vertex graph, which is the center. For any mm nonnegative integers k_1k_{\_}1, k_2k_{\_}2, \dots, k_mk_{\_}m, let F_k_1,k_2,,k_m(2r+1)F_{\_}{k_{\_}1,k_{\_}2,\dots,k_{\_}m}(2r+1) be the graph consisting of a clique BB with mm vertices and mm vertex-disjoint copies of F_k_i(2r+1)F_{\_}{k_{\_}i}(2r+1) with its center in BB. In the above definition, we call BB the base of F_k_1,k_2,,k_m(2r+1)F_{\_}{k_{\_}1,k_{\_}2,\dots,k_{\_}m}(2r+1).

We now show RT(n,F_k(2r+1),o(n))=n22(11r)+o(n2)RT(n,F_{\_}{k}(2r+1),o(n))=\frac{n^{2}}{2}(1-\frac{1}{r})+o(n^{2}), where r2r\geq 2, as the theorem holds when r=1r=1 by Lemma 1. Since RT(n,K_2r+1,o(n))=n22(11r)+o(n2)RT(n,K_{\_}{2r+1},o(n))=\frac{n^{2}}{2}(1-\frac{1}{r})+o(n^{2}), we only need to show that RT(n,F_k(2r+1),o(n))n22(11r)+o(n2)RT(n,F_{\_}{k}(2r+1),o(n))\leq\frac{n^{2}}{2}(1-\frac{1}{r})+o(n^{2}). Let ϵ>0\epsilon>0 be a small positive real number. We will show that there exists a positive number δ:=δ(ϵ)\delta:=\delta(\epsilon) and a large positive integer N:=N(ϵ)N:=N(\epsilon) such that all graphs G(n)G(n) with n>Nn>N, if |E(G)|(11r+ϵ)n2/2|E(G)|\geq(1-\frac{1}{r}+\epsilon)n^{2}/2 and α(G(n))δn\alpha(G(n))\leq\delta n, then there exists a F_k(2r+1)F_{\_}k(2r+1) as a subgraph of G(n)G(n). In the proof, we will take δ=210ϵ2\delta=\frac{\sqrt{2}}{10}\epsilon^{2} and we will not specify the value of N(ϵ)N(\epsilon) by only assuming nn is large enough to ensure our counting work.

Since |E(G)|(11r+ϵ)n2/2|E(G)|\geq\left(1-\frac{1}{r}+\epsilon\right)n^{2}/2, by Lemma 2, GG contains a subgraph G_1G_{\_}1 of order n_1ϵ/2nn_{\_}1\geq\sqrt{\epsilon/2}n such that the minimum degree δ(G_1)(11rϵ2)n_1\delta(G_{\_}1)\geq(1-\frac{1}{r}-\frac{\epsilon}{2})n_{\_}1. Since α(G_1)α(G)210ϵ2nϵ5n_1\alpha(G_{\_}1)\leq\alpha(G)\leq\frac{\sqrt{2}}{10}\epsilon^{2}n\leq\frac{\epsilon}{5}n_{\_}1. So, we assume without loss of generality that δ(G)(11rϵ2)n\delta(G)\geq(1-\frac{1}{r}-\frac{\epsilon}{2})n and α(G)ϵ5n\alpha(G)\leq\frac{\sqrt{\epsilon}}{5}n.

For each integer mm with 1mr+11\leq m\leq r+1, let _m\mathcal{F}_{\_}{m} be the set of mm-tuples (k_1,k_2,,k_m)(k_{\_}1,k_{\_}2,\dots,k_{\_}m) of mm nonnegative integers k_1k_{\_}1, k_2k_{\_}2, \dots, k_mk_{\_}m listed with non-increasing order such that GG contains a copy of F_k_1,k_2,,k_m(2r+1)F_{\_}{k_{\_}1,k_{\_}2,\dots,k_{\_}m}(2r+1). By Lemma 3, GG contains a copy of K_mK_{\_}{m} for every 1mr+11\leq m\leq r+1. Thus, _m\mathcal{F}_{\_}m\neq\emptyset. We also notice that if there exists an mm-tuple (k_1,k_2,,k_m)_m(k_{\_}1,k_{\_}2,\dots,k_{\_}m)\in\mathcal{F}_{\_}m with k_1kk_{\_}1\geq k, then GG contains a copy of F_k(2r+1)F_{\_}k(2r+1). So, we assume k_1k1k_{\_}1\leq k-1 for all (k_1,k_2,,k_m)_m(k_{\_}1,k_{\_}2,\dots,k_{\_}m)\in\mathcal{F}_{\_}m. Consequently, |_m|mk|\mathcal{F}_{\_}m|\leq m^{k}.

For each _m\mathcal{F}_{\_}m, we define a lexicographic order \prec such that (k_1,k_2,,k_m)(_1,_2,,_m)(k_{\_}1,k_{\_}2,\dots,k_{\_}m)\prec(\ell_{\_}1,\ell_{\_}2,\dots,\ell_{\_}m) if there exist ss with 1sm1\leq s\leq m such that k_i=_ik_{\_}i=\ell_{\_}i for i<si<s and k_i<_ik_{\_}i<\ell_{\_}i. Clearly, (0,0,,0)(0,0,\dots,0) is the minimum element in _m\mathcal{F}_{\_}m following this order.

Let (k_1,k_2,,k_r+1)(k_{\_}1,k_{\_}2,\dots,k_{\_}{r+1}) be the maximum element in _r+1\mathcal{F}_{\_}{r+1}. We will lead a contradiction by showing that either GG contains a copy of F_k(2r+1)F_{\_}{k}(2r+1) or (k_1,k_2,,k_r+1)(k_{\_}1,k_{\_}2,\dots,k_{\_}{r+1}) is not maximum in _r+1\mathcal{F}_{\_}{r+1}. Let FF be a copy of F_k_1,k_2,,k_r+1(2r+1)F_{\_}{k_{\_}1,k_{\_}2,\dots,k_{\_}{r+1}}(2r+1) in GG and BB be the base. Denote BB by {v_1,v_2,,v_r+1}\{v_{\_}1,v_{\_}2,\dots,v_{\_}{r+1}\} and assume v_iv_{\_}i is the center of the corresponding F_k_i(2r+1)F_{\_}{k_{\_}i}(2r+1) in FF. Let HH be obtained from G_1G_{\_}1 by deleting all vertices in V(F)BV(F)-B, i.e., H=G(V(F)B)H=G-(V(F)-B). Since |V(F)|2r(k1)(r+1)+(r+1)|V(F)|\leq 2r(k-1)(r+1)+(r+1) and δ(G)(11r+ϵ2)n\delta(G)\geq(1-\frac{1}{r}+\frac{\epsilon}{2})n, we may assume the minimum degree δ(H)(11rϵ3)|H|\delta(H)\geq(1-\frac{1}{r}-\frac{\epsilon}{3})|H| and α(H)ϵ4|H|\alpha(H)\leq\frac{\epsilon}{4}|H| provided nn is large enough.

Starting with clique B_0=BB_{\_}0=B in HH, we apply Lemma 4 repeatedly rr times, we get a sequence of cliques B_1B_{\_}1, B_2B_{\_}2, \dots, B_rB_{\_}r such that |B_i|=|B_i1|+1|B_{\_}i|=|B_{\_}{i-1}|+1 and |B_iB_i1||B_i1|1|B_{\_}i\cap B_{\_}{i-1}|\geq|B_{\_}{i-1}|-1. At the end, we have |B_r|=|B_0|+r=(r+1)+r=2r+1|B_{\_}r|=|B_{\_}0|+r=(r+1)+r=2r+1 and |B_rB_0||B_0|r1|B_{\_}r\cap B_{\_}0|\geq|B_{\_}0|-r\geq 1. Let s1s\geq 1 be the smallest integer such that v_sB_rB_0v_{\_}s\in B_{\_}r\cap B_{\_}0. Then, we get subgraph FF^{*} which is a copy of F_k_1,k_2,,k_s+1(2r+1)F_{\_}{k_{\_}1,k_{\_}2,\dots,k_{\_}s+1}(2r+1) with base B={v_1,v_2,,v_s}B^{*}=\{v_{\_}1,v_{\_}2,\dots,v_{\_}s\}. Let H=G(V(F)B)H^{*}=G-(V(F^{*})-B^{*}).

Applying Lemma 3 repeatedy (r+1s)(r+1-s) times, we get a clique BBB^{\prime}\supseteq B^{*} with r+1r+1 vertices in HH^{*}. Combining BB^{\prime} with FF^{*}, we get a copy of F_k_1,,k_s+1,0,,0(2r+1)F_{\_}{k_{\_}1,\dots,k_{\_}s+1,0,\dots,0}(2r+1) with base BB^{\prime}. Let _1,_2,,_s\ell_{\_}1,\ell_{\_}2,\dots,\ell_{\_}s be an new list of k_1,k_2,,k_s+1k_{\_}1,k_{\_}2,\dots,k_{\_}s+1 in non-increasing order. Then, (r+1)(r+1)-tuple (_1,_2,,_s,0,,0)_r+1(\ell_{\_}1,\ell_{\_}2,\dots,\ell_{\_}s,0,\dots,0)\in\mathcal{F}_{\_}{r+1}. But, (k_1,k_2,,k_r+1)(_1,_2,,_s,0,,0)(k_{\_}1,k_{\_}2,\dots,k_{\_}{r+1})\prec(\ell_{\_}1,\ell_{\_}2,\dots,\ell_{\_}s,0,\dots,0), giving a contradiction to the maximality of (k_1,k_2,,k_r+1)(k_{\_}1,k_{\_}2,\dots,k_{\_}{r+1}).

3 Acknowledgement

We would like to thank Yan Cao for the helpful communication.

References

  • [1] Guantao Chen, Ronald J. Gould, Florian Pfender, and Bing Wei. Extremal graphs for intersecting cliques. Journal of Combinatorial Theory, Series B, 89(2):159 – 171, 2003.
  • [2] P. Erdős, Z. Füredi, R. J. Gould, and D. S. Gunderson. Extremal graphs for intersecting triangles. J. Combin. Theory Ser. B, 64(1):89–100, 1995.
  • [3] P. Erdős and Vera T. Sós. Some remarks on Ramsey’s and Turán’s theorem. In Combinatorial theory and its applications, II (Proc. Colloq., Balatonfüred, 1969), pages 395–404, 1970.
  • [4] P. Erdős, A. Hajnal, Vera T. Sós, and et al. More results on ramsey-turan type problems, 1983.
  • [5] Jacob Fox, Po-Shen Loh, and Yufei Zhao. The critical window for the classical ramsey-turán problem. Combinatorica, 35:435–476, 2015.
  • [6] Jaehoon. Kim, Younjin. Kim, and Hong. Liu. Two conjectures in ramsey–turán theory. SIAM Journal on Discrete Mathematics, 33(1):564–586, 2019.
  • [7] Miklós Simonovits and Vera T. Sós. Ramsey-turán theory. Discrete Mathematics, 229(1-3):293–340, 2001.