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

Ramsey and Gallai-Ramsey numbers for linear forests and kipas111Supported by the National Natural Science Foundation of China (Nos. 12201375, 12061059).

Ping Li222Corresponding author: School of Mathematics and Statistics, Shaanxi Normal University, Xi’an, Shaanxi, China. [email protected],    Yaping Mao 333JSPS International Research Fellow: Faculty of Environment and Information Sciences, Yokohama National University, 79-2 Tokiwadai, Hodogaya-ku, Yokohama 240-8501, Japan. [email protected] ,   Ingo Schiermeyer 444Institut für Diskrete Mathematik und Algebra, Technische Universität Bergakademie Freiberg, 09596 Freiberg, Germany. [email protected] ,  Yifan Yao 555 School of Mathematics and Statistics, Qinghai Normal University, Xining, Qinghai 810008, China. [email protected]
Abstract

For two graphs G,HG,H, the Ramsey number r(G,H)r(G,H) is the minimum integer nn such that any red/blue edge-coloring of KnK_{n} contains either a red copy of GG or a blue copy of HH. For two graphs G,HG,H, the Gallai-Ramsey number grk(G:H)\operatorname{gr}_{k}(G:H) is defined as the minimum integer nn such that any kk-edge-coloring of KnK_{n} must contain either a rainbow copy of GG or a monochromatic copy of HH. In this paper, the classical Ramsey numbers of linear forest versus kipas are obtained. We obtain the exact values of grk(G:H)\operatorname{gr}_{k}(G:H), where HH is either a path or a kipas and G{K1,3,P4+,P5}G\in\{K_{1,3},P_{4}^{+},P_{5}\} and P4+P_{4}^{+} is the graph consisting of P4P_{4} with one extra edge incident with inner vertex.
Keywords: Ramsey number; Gallai-ramsey number; Path; Kipas

1 Introduction

All graphs considered in this paper are undirected, finite and simple. We refer to the book [1] for graph theoretical notation and terminology not described here. For a graph GG, let V(G)V(G), E(G)E(G), |G||G|, e(G)e(G) and δ(G)\delta(G) denote the set of vertices, the set of edges, the number of vertices, the number of edges and the minimum degree of GG, respectively. We also call |G||G| and e(G)e(G) the order and the size of GG, respectively. The neighborhood set of a vertex vV(G)v\in V(G) is NG(v)={uV(G)|uvE(G)}N_{G}(v)=\{u\in V(G)\,|\,uv\in E(G)\}, and its closed neighborhood set is NG[v]=NG(v){v}N_{G}[v]=N_{G}(v)\cup\{v\}. For two subsets XX and YY of V(G)V(G) we denote by EG[X,Y]E_{G}[X,Y] the set of edges of GG with one endpoint in XX and the other endpoint in YY. Let NG(v,S)=NG(v)SN_{G}(v,S)=N_{G}(v)\cap S. Let G¯\overline{G} be the complement of GG. For any integers m<nm<n, we use the convenient notation [m,n][m,n] for the set {m,m+1,,n1,n}\{m,m+1,\cdots,n-1,n\}, specially, we denote [n][n] as [1,n][1,n]. For a path Pn=v1v2vnP_{n}=v_{1}v_{2}\ldots v_{n}, we call vn2v_{\lceil\frac{n}{2}\rceil} and vn+12v_{\lceil\frac{n+1}{2}\rceil} center vertices of PnP_{n}. For a graph HH and a vertex vV(H)v\notin V(H), let vHv\vee H be a graph obtained from HH by connecting each vertex of HH and vv. If HH is a path of order nn, then vHv\vee H is called a kipas with center vv, and denote it by K^n\widehat{K}_{n}.

Definition 1.

For two graphs G,HG,H, the Ramsey number r(G,H)r(G,H) is the minimum integer nn such that any red/blue edge-coloring of KnK_{n} contains either a red copy of GG or a blue copy of HH.

Definition 2.

Let \mathcal{F} be a set of graphs. The Ramsey number r(G,)r(G,\mathcal{F}) is the minimum integer nn such that any red/blue edge-coloring of KnK_{n} contains either a red copy of GG or a blue copy of graph in \mathcal{F}.

For more details on Ramsey numbers, we refer to the survey paper [23] and some papers [2, 4, 5, 10].

1.1 Ramsey and Gallai-Ramsey numbers

The Gallai-kk-coloring (or Gallai-coloring for short) is a kk-edge-coloring of a complete graph without rainbow triangles. The following result gives the structure of Gallai-coloring.

Theorem 1.1.

[9, 13] In any rainbow triangle free edge-coloring of a complete graph, the vertex set can be partitioned into at least two parts such that between the parts there is a total of at most two colors, and between each pair of parts there is only one color on the edges.

Let gkq(p)g^{q}_{k}(p) be the smallest positive integer nn such that every Gallai-kk-coloring of KnK_{n} contains a KpK_{p} receiving at most qq distinct colors. In [6], Fox et al. studied the integer pp when nn is fixed, and got the following asymptotic result.

Theorem 1.2.

[6] Let kk and qq be fixed positive integers with qkq\leq k. Every Gallai-kk-coloring of KnK_{n} contains a set of order Ω(n(q2)/(k2)log2ck,qn)\Omega(n^{\binom{q}{2}/\binom{k}{2}}\log_{2}^{c_{k,q}}n) which uses at most qq colors, where ck,qc_{k,q} is only depending on kk and qq. Moreover, this bound is tight apart from the constant factor.

Note that gk1(p)g^{1}_{k}(p) is the minimum integer nn such that every Gallai-kk-coloring of KnK_{n} contains a monochromatic KpK_{p}. The gk1(p)g^{1}_{k}(p) is also denoted by grk(K3:Kp)\operatorname{gr}_{k}(K_{3}:K_{p}).

Definition 3.

For two graphs G,HG,H, the Gallai-Ramsey number grk(G:H)\operatorname{gr}_{k}(G:H) is defined as the minimum integer NN such that for any nNn\geq N and any exact kk-edge-coloring (using kk colors) of KnK_{n} must contain either a rainbow copy of GG or a monochromatic copy of HH.

Definition 4.

Define grk(G:H)\operatorname{gr}^{\prime}_{k}(G:H) as the minimum integer NN such that for all nNn\geq N, every edge-coloring of KnK_{n} using at most kk colors contains either a rainbow copy of GG or a monochromatic copy of HH.

It is obvious that grk(G:H)max{gri(G:H):1ik}\operatorname{gr}^{\prime}_{k}(G:H)\leq\max\{\operatorname{gr}_{i}(G:H):1\leq i\leq k\}. On the other hand, we have that grk(G:H)max{gri(G:H):1ik}\operatorname{gr}^{\prime}_{k}(G:H)\geq\max\{\operatorname{gr}_{i}(G:H):1\leq i\leq k\}. Otherwise, there exists an integer kk^{\prime} (1kk1\leq k^{\prime}\leq k) such that grk(G:H)>grk(G:H)\operatorname{gr}_{k^{\prime}}(G:H)>\operatorname{gr}^{\prime}_{k}(G:H). By the definition of grk(G:H)\operatorname{gr}_{k^{\prime}}(G:H), there is a kk^{\prime}-edge-coloring of Kgrk(G:H)1K_{\operatorname{gr}_{k^{\prime}}(G:H)-1} such that there is neither rainbow copy of GG nor monochromatic copy of HH. However, grk(G:H)1grk(G:H)\operatorname{gr}_{k^{\prime}}(G:H)-1\geq\operatorname{gr}^{\prime}_{k}(G:H), which implies that there is either a rainbow copy of GG or a monochromatic copy of HH in any kk^{\prime}-edge-coloring of Kgrk(G:H)1K_{\operatorname{gr}_{k^{\prime}}(G:H)-1}, a contradiction. Therefore, grk(G:H)=max{gri(G:H):1ik}\operatorname{gr}^{\prime}_{k}(G:H)=\max\{\operatorname{gr}_{i}(G:H):1\leq i\leq k\}, and we only need to talk about the parameter grk(G:H)\operatorname{gr}_{k}(G:H).

Fox et al. [6] proposed the following conjecture.

Conjecture 1.

[6] For integers k3k\geq 3 and p3p\geq 3,

grk(K3:Kp)={(r(Kp,Kp)1)k/2+1,if n is even,(p1)(r(Kp,Kp)1)(k1)/2+1,if n is odd.\operatorname{gr}_{k}(K_{3}:K_{p})=\left\{\begin{array}[]{ll}(r(K_{p},K_{p})-1)^{k/2}+1,&\mbox{if }n\mbox{ is even},\\ (p-1)(r(K_{p},K_{p})-1)^{(k-1)/2}+1,&\mbox{if }n\mbox{ is odd}.\\ \end{array}\right.

Chung and Graham [3] and Gyárfás et al. [12] proved the case p=3p=3, Liu et al. [17] proved the case p=4p=4, and Magnant and Schiermeyer [21] proved the case k=5k=5. For more details on the Gallai-Ramsey numbers, we refer to a dynamic survey [8] and the recent book [20].

1.2 Structural theorems and our methods

The Gallai-kk-coloring, a kk-edge-coloring of KnK_{n} without rainbow K3K_{3}, has an interesting structure. Fujita and Magnant [7] consider the structures of kk-edge-coloring of KnK_{n} without rainbow S3+S_{3}^{+}, where S3+S_{3}^{+} is a graph obtain from K3K_{3} by adding a pendent edge. Thomason and Wagner [25] got the structures of kk-edge-coloring of KnK_{n} without rainbow P5P_{5}. In addition, Gyárfás et al. [11] investigated the local Ramsey number of stars, which implies the structures of kk-edge-coloring of KnK_{n} without a rainbow star K1,3K_{1,3} or a rainbow tree P4+P_{4}^{+}, respectively, where P4+P_{4}^{+} is the graph consisting of P4P_{4} with one extra edge incident with a center vertex.

In this paper, we study the Gallai-Ramsey number grk(G:H)\operatorname{gr}_{k}(G:H), where G{K1,3,P5,P4+}G\in\{K_{1,3},P_{5},P_{4}^{+}\} and HH is a path or a kipas. The following are the structures of kk-edge-colored KnK_{n} without rainbow P5,K1,3P_{5},K_{1,3} and P4+P_{4}^{+}, respectively.

Given an edge coloring of KnK_{n}, let EjE_{j} be the set of edges colored jj and let VjV_{j} be the vertices at which an edge of color jj appears. The following are the structures of kk-edge-coloring of KnK_{n} without a rainbow P5P_{5}.

Theorem 1.3.

[25] Let k,nk,n be two integers with k4k\geq 4 and n5n\geq 5. Let KnK_{n} be a kk-edge-colored graph so that it contains no rainbow P5P_{5}. Then, after renumbering the colors, one of the following must hold:

  • (i)(i) color 1 is dominant, meaning that the sets VjV_{j}, j2j\geq 2, are disjoint;

  • (ii)(ii) KnaK_{n}-a is monochromatic for some vertex aa;

  • (iii)(iii) there are three vertices a,b,ca,b,c such that E2={ab}E_{2}=\{ab\}, E3={ac}E_{3}=\{ac\}, E4E_{4} contains bcbc plus perhaps some edges incident with aa, and every other edge is in E1E_{1};

  • (iv)(iv) there are four vertices a,b,c,da,b,c,d such that {ab}E2{ab,cd}\{ab\}\subseteq E_{2}\subseteq\{ab,cd\}, E3={ac,bd}E_{3}=\{ac,bd\}, E4={ad,bc}E_{4}=\{ad,bc\} and every other edge is in E1E_{1};

  • (v)(v) n=5n=5, V(Kn)={a,b,c,d,e}V(K_{n})=\{a,b,c,d,e\}, E1={ad,ae,bc}E_{1}=\{ad,ae,bc\}, E2={bd,be,ac}E_{2}=\{bd,be,ac\}, E3={cd,ce,ab}E_{3}=\{cd,ce,ab\} and E4={de}E_{4}=\{de\}.

Let G1(n)G_{1}(n) be a 33-edge-coloring of KnK_{n} such that V(Kn)V(K_{n}) can be partitioned into three parts V1,V2,V3V_{1},V_{2},V_{3} (at most one of them is empty) such that each edge of G[V1]G[V_{1}] is colored by either 11 or 33, each edge of G[V2]G[V_{2}] is colored by either 11 or 22, and each edge of G[V3]G[V_{3}] is colored by either 22 or 33. Moreover, each edge between V1,V2V_{1},V_{2} is colored by 11, each edge between V2,V3V_{2},V_{3} is colored by 22 and each edge between V1,V3V_{1},V_{3} is colored by 33.

Let G2(n)G_{2}(n) be a 44-edge-coloring of KnK_{n} such that there is exactly one edge, say xyxy, having color 22. All other edges incident with xx are colored by 33, and all other edges incident with yy are colored by 44. All edges not incident to vertices x,yx,y have color 11.

Let G3(n)G_{3}(n) be a 44-edge-coloring of KnK_{n} such that there exists a rainbow K3K_{3} (say the vertices in the triangle are a,b,ca,b,c, where abab has color 22, bcbc has color 33 and acac has color 44). In addition, any other edge has color 11.

Remark 1.1.

If Γ\Gamma is an edge-coloring of {(ii),(iii),(iv),(v),G2(n),G3(n)}\{(ii),(iii),(iv),(v),G_{2}(n),G_{3}(n)\} of KnK_{n}, then the graph induced by edges with color i2i\geq 2 is a star forest.

The following are the structures of kk-edge-coloring of KnK_{n} without a rainbow K1,3K_{1,3}.

Theorem 1.4 ([11]).

For positive integers k3k\geq 3 and nn, if GG is a kk-edge-coloring of KnK_{n} without rainbow K1,3K_{1,3}, then after renumbering the colors, one of the following holds.

(I1)(I1) k=3k=3 and GG1(n)G\cong G_{1}(n);

(I2)(I2) k4k\geq 4 and Item (i)(i) in Theorem 1.3 holds.

The following are the structures of kk-edge-coloring of KnK_{n} without rainbow P4+P_{4}^{+}.

Theorem 1.5 ([11]).

For positive integers k4k\geq 4 and nn, if GG is a kk-edge-coloring of KnK_{n} without rainbow P4+P_{4}^{+}, then after renumbering the colors, one of the following holds.

(J1)(J1) k=4k=4 and G{G2(n),G3(n)}G\in\{G_{2}(n),G_{3}(n)\};

(J2)(J2) k4k\geq 4 and Item (i)(i) in Theorem 1.3 holds.

Recall Theorem 1.1 and definition of grk(K3:H)\operatorname{gr}_{k}(K_{3}:H), the structure of Gallai-coloring is important for grk(K3:H)\operatorname{gr}_{k}(K_{3}:H). Analogously, the structures of kk-edge-colorings in Theorem 1.3 is important for grk(P5:H)\operatorname{gr}_{k}(P_{5}:H). However, there are five structures, and it is complicated if we consider the five structures at the same time. Observe that the structures (ii)(iv)(ii)-(iv) are simple since they contain a large monochromatic complete graph (the monochromatic complete graph can obtained by deleting at most three vertices), and structure (v)(v) is simple since there are only five vertices. Therefore, the structures (i)(i) is very important for grk(P5:H)\operatorname{gr}_{k}(P_{5}:H). Similarly, in Theorem 1.4, the structures G1(n)G_{1}(n) and (i)(i) are important; in Theorem 1.5, the structure (i)(i) is important and structures G2(n)G_{2}(n) and G3(n)G_{3}(n) are clear. By above discussion, G1(n)G_{1}(n) and (i)(i) are key structures in kk-edge-colored KnK_{n} without rainbow P5,K1,3P_{5},K_{1,3} or P4+P_{4}^{+}.

Firstly, we observe the structure (i)(i). Suppose k3k\geq 3. It is clear that KnK_{n} can be partitioned into k1k-1 parts V1,,Vk1V_{1},\cdots,V_{k-1} such that each edge between different parts is colored by 11, and each edge of G[Vi1]G[V_{i-1}] is colored by either 11 or ii. Therefore, G[Vi]G[V_{i}] contains all edges of color i+1i+1 and |Vi|2|V_{i}|\geq 2 for i[k1]i\in[k-1]. We denote the set of all such kk-edge-colorings on KnK_{n} by k(n)\mathcal{B}_{k}(n).

Definition 5.

Let bk(H)b_{k}(H) be the minimum number nn such that any edge-coloring Γk(n)\Gamma\in\mathcal{B}_{k}(n) of KnK_{n} contains a monochromatic HH.

For the structure G1(n)G_{1}(n), V(Kn)V(K_{n}) can be partitioned into three parts V1,V2,V3V_{1},V_{2},V_{3} as the definition, and at most one part is empty. If there is an empty part, then G1(n)3(n)G_{1}(n)\in\mathcal{B}_{3}(n). Thus, we only consider that V1,V2,V3V_{1},V_{2},V_{3} are all nonempty sets. We denote the set of all such 3-edge-colorings on KnK_{n} by 𝒯(n)\mathcal{T}(n).

Definition 6.

Let t(H)t(H) be the minimum number nn such that any edge-coloring Γ𝒯(n)\Gamma\in\mathcal{T}(n) of KnK_{n} contains a monochromatic HH.

Remark 1.2.

In order to determine the grk(G:H)\operatorname{gr}_{k}(G:H), where G{P5,K1,3,P4+}G\in\{P_{5},K_{1,3},P_{4}^{+}\}, the main work is to determine bk(H)b_{k}(H) or t(H)t(H). It is easy to verify whether KnK_{n} contains a monochromatic copy of HH under the other structures.

1.3 Progress and our results

Li et al. [18] got some exact values of grk(P5:Kt)\operatorname{gr}_{k}(P_{5}:K_{t}) for small tt and bounds for general tt. Recently, Wei et al. [26] got the exact values of grk(P5:H)(k3)\operatorname{gr}_{k}(P_{5}:H)\ (k\geq 3) for small tt and upper and lower bounds for general tt, where HH is either a general fan or a general wheel graph. The motivation of this paper is whether we can obtain exact values for grk(P5:H)(k4)\operatorname{gr}_{k}(P_{5}:H)\ (k\geq 4) such that HH is not just a graph but belongs to a class of graphs.

In this paper, we first study the new parameters bk(Pn),t(Pn),bk(K^n)b_{k}(P_{n}),t(P_{n}),b_{k}(\widehat{K}_{n}) and t(K^n)t(\widehat{K}_{n}), and further get the exact values of grk(G:Pn)\operatorname{gr}_{k}(G:P_{n}) and grk(G:K^n)\operatorname{gr}_{k}(G:\widehat{K}_{n}), where G{P5,K1,3,P4+}G\in\{P_{5},K_{1,3},P_{4}^{+}\}.

We call a linear forest a kk-linear forest if any component of the linear forest is a path of order at least kk. We first get a classical Ramsey number for K^n\widehat{K}_{n} and the set of 33-linear forests, and the result is useful and will be used in Section 5.

Theorem 1.6.

If 2mn22\leq m\leq\frac{n}{2} and \mathcal{L} is the set of linear forest LL of size at least mm, then

r(K^n,)=n+m/2.r(\widehat{K}_{n},\mathcal{L})=n+\lceil m/2\rceil.

In addition, if 6mn26\leq m\leq\frac{n}{2} and \mathcal{L}^{\prime} is the set of 33-linear forest LL of size at least mm, then

r(K^n,)=n+m/2.r(\widehat{K}_{n},\mathcal{L}^{\prime})=n+\lceil m/2\rceil.

The Gallai-Ramsey numbers are list as follows.

Theorem 1.7.

If nk4n\geq k\geq 4, then

grk(P5:Pn)={n+1,if2(k1)n4(k2)+1,3n32,ifn>4(k2)+1.\operatorname{gr}_{k}(P_{5}:P_{n})=\left\{\begin{array}[]{ll}n+1,&{\rm if}~{}2(k-1)\leq n\leq 4(k-2)+1,\\ \left\lceil\frac{3n-3}{2}\right\rceil,&{\rm if}~{}n>4(k-2)+1.\\ \end{array}\right.
Theorem 1.8.

If nk4n\geq k\geq 4, then

grk(P4+:Pn)={n+2,ifk=4and2(k1)n4(k2)+1,n,ifk5and2(k1)n4(k2)+1,3n32,otherwise.\operatorname{gr}_{k}(P_{4}^{+}:P_{n})=\left\{\begin{array}[]{ll}n+2,&{\rm if}~{}k=4~{}{\rm and}~{}2(k-1)\leq n\leq 4(k-2)+1,\\ n,&{\rm if}~{}k\geq 5~{}{\rm and}~{}2(k-1)\leq n\leq 4(k-2)+1,\\ \left\lceil\frac{3n-3}{2}\right\rceil,&{\rm otherwise}.\\ \end{array}\right.
Theorem 1.9.

(1)(1) If n2(k1)n\geq 2(k-1) and k4k\geq 4, then

grk(K1,3:Pn)={n,if2(k1)n4(k2)+1,3n32,ifn>4(k2)+1.\operatorname{gr}_{k}(K_{1,3}:P_{n})=\left\{\begin{array}[]{ll}n,&{\rm if}~{}2(k-1)\leq n\leq 4(k-2)+1,\\ \left\lceil\frac{3n-3}{2}\right\rceil,&{\rm if}~{}n>4(k-2)+1.\\ \end{array}\right.

(2)(2) If k=3k=3 and n4n\geq 4, then

grk(K1,3:Pn)={3n21,ifn is even,3n12,ifn is odd .\operatorname{gr}_{k}(K_{1,3}:P_{n})=\left\{\begin{array}[]{ll}\frac{3n}{2}-1,&{\rm if}~{}n\mbox{ is even},\\ \frac{3n-1}{2},&{\rm if}~{}n\mbox{ is odd }.\end{array}\right.
Remark 1.3.

We do not discuss grk(P5:Pn),grk(P4+:Pn)\operatorname{gr}_{k}(P_{5}:P_{n}),\operatorname{gr}_{k}(P_{4}^{+}:P_{n}) and grk(K1,3:Pn)\operatorname{gr}_{k}(K_{1,3}:P_{n}) when k4k\geq 4 and n<2(k1)n<2(k-1), since it is easily to obtain. In fact, if k4k\geq 4 and n<2(k1)n<2(k-1), then KnK_{n} must contain a monochromatic PnP_{n} when the edge-coloring Γ\Gamma of KnK_{n} belongs to k(n)\mathcal{B}_{k}(n). However, other edge-colorings G2(N),G3(N),(ii),(iii),(iv)G_{2}(N),G_{3}(N),(ii),(iii),(iv) and (v)(v) are simple.

Theorem 1.10.

For n5n\geq 5, we have that gr3(K1,3:K^n)=5n2\operatorname{gr}_{3}(K_{1,3}:\widehat{K}_{n})=\left\lfloor\frac{5n}{2}\right\rfloor if nn is odd, and 5n21gr3(K1,3:K^n)5n2\left\lfloor\frac{5n}{2}\right\rfloor-1\leq\operatorname{gr}_{3}(K_{1,3}:\widehat{K}_{n})\leq\left\lfloor\frac{5n}{2}\right\rfloor if nn is even.

2 Preliminaries

As we know, the study of Gallai-Ramsey numbers deeply depend on the classical Ramsey number. Therefore, we list some useful classical Ramsey numbers in this section, which can be found in the survey paper [23].

Theorem 2.1.

[10] If 2mn2\leq m\leq n, then r(Pn,Pm)=n+m/21r(P_{n},P_{m})=n+\left\lfloor m/2\right\rfloor-1.

The following theorem is a generalization of Theorem 2.1.

Theorem 2.2.

[5] Suppose L1,L2L_{1},L_{2} are 22-linear forests having j1,j2j_{1},j_{2} components of odd order, respectively. Then

r(L1,L2)=max{|L1|+|L2|j221,|L2|+|L1|j121}.r(L_{1},L_{2})=\max\left\{|L_{1}|+\left\lfloor\frac{|L_{2}|-j_{2}}{2}\right\rfloor-1,|L_{2}|+\left\lfloor\frac{|L_{1}|-j_{1}}{2}\right\rfloor-1\right\}.
Theorem 2.3.

[22] r(Pm,K1,n)=m+n1r(P_{m},K_{1,n})=m+n-1 if n1(modn1)n\equiv 1\pmod{n-1}, and r(Pm,K1,n)m+n2r(P_{m},K_{1,n})\leq m+n-2 otherwise.

Theorem 2.4.

[14] If m,n2m,n\geq 2, then

r(K1,n,K1,m)={m+n1, if m,n are even,m+n, otherwise.r(K_{1,n},K_{1,m})=\left\{\begin{array}[]{ll}m+n-1,&\mbox{ if }m,n\mbox{ are even},\\ m+n,&\mbox{ otherwise}.\\ \end{array}\right.

Li et al. [16] got a closing gap on path-kipas Ramsey numbers.

Theorem 2.5.

[16] For m2n1m\leq 2n-1 and m,n2m,n\geq 2, we have

r(Pn,K^m)=max{2n1,3m21,2m2+n2}.r(P_{n},\widehat{K}_{m})=\max\left\{2n-1,\left\lceil\frac{3m}{2}\right\rceil-1,2\left\lfloor\frac{m}{2}\right\rfloor+n-2\right\}.

Salman and Broersma [24] obtained the following result for path-kipas Ramsey numbers.

Theorem 2.6.

[24] If n4n\geq 4 and m2n2m\geq 2n-2, then r(Pn,K^m)m+n1r(P_{n},\widehat{K}_{m})\leq m+n-1.

Li et al. [15] obtained an exact formula for all star-kipas Ramsey numbers.

Theorem 2.7.

[15] Let n,m2n,m\geq 2 be two integers.

(1)(1) If m2nm\geq 2n, then

r(K1,n,K^m)={m+n1, if m,n are even,m+n, otherwise.r(K_{1,n},\widehat{K}_{m})=\left\{\begin{array}[]{ll}m+n-1,&\mbox{ if }m,n\mbox{ are even},\\ m+n,&\mbox{ otherwise}.\\ \end{array}\right.

(2)(2) If m2n1m\leq 2n-1, then

r(K1,n,K^m)={2n+m21, if m,m2 are even,2n+m2, otherwise.r(K_{1,n},\widehat{K}_{m})=\left\{\begin{array}[]{ll}2n+\lfloor\frac{m}{2}\rfloor-1,&\mbox{ if }m,\lfloor\frac{m}{2}\rfloor\mbox{ are even},\\ 2n+\lfloor\frac{m}{2}\rfloor,&\mbox{ otherwise}.\\ \end{array}\right.

The following is a result about decomposition of the complete graph.

Theorem 2.8.

[19] For n1n\geq 1, K2n+1K_{2n+1} can be decomposed into nn edge-disjoint Hamiltonian cycles; K2n+2K_{2n+2} can be decomposed into nn edge-disjoint Hamiltonian cycles and a perfect matching.

The following result will be used in the subsequent proofs, and we give a simple proof here, although it may be proved somewhere.

Lemma 2.1.

Suppose that GG is a complete rr-partite graph with parts V1,,VrV_{1},\ldots,V_{r} and VrV_{r} is one maximum part.

  1. 1.

    If i[r1]|Vi||Vr|\sum_{i\in[r-1]}|V_{i}|\geq|V_{r}|, then GG contains a Hamiltonian cycle.

  2. 2.

    If i[r1]|Vi|=|Vr|1\sum_{i\in[r-1]}|V_{i}|=|V_{r}|-1, GG contains a Hamiltonian path.

Proof.

The result is true for bipartite graphs. So, suppose r3r\geq 3. Without loss of generality, let |V1||V2||Vr||V_{1}|\leq|V_{2}|\leq\ldots\leq|V_{r}|. Then there exists an s[r]s\in[r] such that |Vs1|<|Vs|==|Vt||V_{s-1}|<|V_{s}|=\ldots=|V_{t}|. The proof proceeds by induction on |G||G|. If |G|=3|G|=3, then the result holds obviously. If |Vr|=1|V_{r}|=1, then GG is a complete graph and the result follows. So, suppose that |G|4|G|\geq 4 and |Vr|>1|V_{r}|>1. Choose a vertex viv_{i} from ViV_{i} for i[s,r]i\in[s,r]. Let Ui=ViU_{i}=V_{i} for i<si<s and Ui=ViviU_{i}=V_{i}-v_{i} for i[s,r]i\in[s,r], and let G=G{vs,,vr}G^{\prime}=G-\{v_{s},\ldots,v_{r}\}. Then UrU_{r} is a maximum part in GG^{\prime}. Since |Vi|2|V_{i}|\geq 2, each UiU_{i} is nonempty.

We prove the first statement. Since i[r1]|Vi||Vr|\sum_{i\in[r-1]}|V_{i}|\geq|V_{r}|, it follows that i[r1]|Ui||Ur|\sum_{i\in[r-1]}|U_{i}|\geq|U_{r}|. By induction, GG^{\prime} contains a Hamiltonian cycle CC. Note that vs,,vrv_{s},\ldots,v_{r} induces a complete graph. Since r3r\geq 3, there are three integer i,j,ki,j,k such that there are two adjacent edges xyxy and yzyz of GG with xVix\in V_{i}, yVjy\in V_{j} and zVkz\in V_{k}. If r=sr=s, then we can assume that r{i,j}r\notin\{i,j\}, and hence (C\{xy}){xvr,yvr}(C\backslash\{xy\})\cup\{xv_{r},yv_{r}\} is a Hamiltonian cycle of GG. If s<r1s<r-1 (resp. s=r1s=r-1), then there is a cycle C=vsvs+1vrvsC^{\prime}=v_{s}v_{s+1}\ldots v_{r}v_{s} in GG (resp. an edge vr1vrv_{r-1}v_{r} of GG). There is an integer of i,j,ki,j,k which is not in {r1,r}\{r-1,r\} (say i{r1,r}i\notin\{r-1,r\}), and one of r1,rr-1,r is not equal to jj (without loss of generality, suppose that jrj\neq r). Then (C\{xy})(C\{vrvr1}){xvr1,yvr}(C\backslash\{xy\})\cup(C^{\prime}\backslash\{v_{r}v_{r-1}\})\cup\{xv_{r-1},yv_{r}\} (resp. (C\{xy}){vr1vr,xvr1,yvr}(C\backslash\{xy\})\cup\{v_{r-1}v_{r},xv_{r-1},yv_{r}\}) is a Hamiltonian of GG.

Now we prove the second statement. In this case, s=rs=r and i[r1]|Ui|=|Ur|\sum_{i\in[r-1]}|U_{i}|=|U_{r}|. So, GG^{\prime} has a Hamiltonian cycle CC^{*} by the first result. Let wwE(C)ww^{\prime}\in E(C^{*}) and suppose that wVrw\notin V_{r} by symmetry. Then (C\{ww}){wvr}(C^{*}\backslash\{ww^{\prime}\})\cup\{wv_{r}\} is a Hamiltonian path of GG. ∎

3 Technique results and the proof of Theorem 1.6

In this section, we prove two critical lemmas, which will be used in Section 5.

Lemma 3.1.

Suppose that n4n\geq 4, a{1,2}a\in\{1,2\} and N=n+aN=n+a. If Γ\Gamma is a red/blue edge-coloring of KNK_{N} such that KNK_{N} does not contain a red copy of K^n\widehat{K}_{n}, then the following results hold.

(i)(i) If a=1a=1, then KNK_{N} contains either a blue copy of 2P22P_{2} or a blue copy of P3P_{3}.

(ii)(ii) If a=2a=2 and n5n\geq 5, then KNK_{N} contains either a blue copy of 2P32P_{3}, or a blue copy of P5P_{5}, or a blue copy of vertex-disjoint P4P_{4} and P2P_{2}.

Proof.

Suppose that R,BR,B are the graphs with the common vertex set V=V(KN)V=V(K_{N}) and their edges sets are red edges and blue edges, respectively. Note that RR does not contains a copy of K^n\widehat{K}_{n}. For convenience, we use P2P4P_{2}\cup P_{4} to denote vertex-disjoint P4P_{4} and P2P_{2} throughout this proof.

(i)(i) For a=1a=1, if δ(B)1\delta(B)\geq 1, then BB contains either a 2P22P_{2} or a P3P_{3}, since n4n\geq 4. If δ(B)=0\delta(B)=0, then there exists a vertex in BB such that dB(v)=0d_{B}(v)=0, and hence RvR-v does not contain a Hamiltonian path (otherwise there is a red K^n\widehat{K}_{n}, a contradiction). Suppose that PP is the maximum path of RvR-v with endpoints a,ba,b. Then V(R){v}V(P)V(R)-\{v\}-V(P)\neq\emptyset. By the maximality of PP, we have ua,ubE(B)ua,ub\in E(B) for each uV(R){v}V(P)u\in V(R)-\{v\}-V(P), and hence BB contains a P3P_{3}.

(ii)(ii) For a=2a=2, suppose to the contrary that BB does not contain 2P32P_{3}, P5P_{5} or P4P2P_{4}\cup P_{2}. We first assume that BB has at most one vertex of degree greater than one (if the vertex exists, denote it by ww). It is clear that |Bw|=n+1|B-w|=n+1 and Δ(Bw)1\Delta(B-w)\leq 1 if ww exists. If ww exists and BwB-w contains a perfect matching MM, then there is a blue P5P_{5} with center ww. Otherwise, there are two vertex u,vu,v of KNK_{N} such that uu is an isolated vertex of BvB-v and Δ(Bv)1\Delta(B-v)\leq 1. In fact, if ww exists and BwB-w does not contain a perfect matching MM, then there is an isolated vertex uu of BwB-w (we can see ww as vv) and Δ(Bv)1\Delta(B-v)\leq 1; if ww does not exist, then Δ(Bw)1\Delta(B-w)\leq 1, and for an edge xyxy of BB, xx is an isolated vertex of ByB-y (we can see x,yx,y as u,vu,v, respectively). Hence, the vertices u,vu,v are obtained. Let V=V{u,v}V^{\prime}=V-\{u,v\} and R=R{u,v}R^{\prime}=R-\{u,v\}. Then uVu\vee V^{\prime} is a red star. Since Δ(B{u,v})1\Delta(B-\{u,v\})\leq 1, it follows that δ(R)|V|2\delta(R^{\prime})\geq|V^{\prime}|-2. Since |V|=n5|V^{\prime}|=n\geq 5, by Dirac’s Theorem, RR^{\prime} contains a Hamiltonian path PP. Hence, uPu\vee P is a red K^n\widehat{K}_{n}, a contradiction.

Now we assume that there are at least two vertices, say z1,z2z_{1},z_{2}, of degree at least two in BB. Without loss of generality, we can choose z1,z2z_{1},z_{2} such that dB(z1)+dB(z2)d_{B}(z_{1})+d_{B}(z_{2}) is as large as possible. We use EuE_{u} to denote the graph induced by blue edges incident with uu. If NB(z1)NB(z2)=N_{B}(z_{1})\cap N_{B}(z_{2})=\emptyset, then Ez1Ez2E_{z_{1}}\cup E_{z_{2}} contains a 2P32P_{3}, unless Ez1Ez2E_{z_{1}}\cup E_{z_{2}} is a double star. If |NB(z1)NB(z2)|=1|N_{B}(z_{1})\cap N_{B}(z_{2})|=1, then Ez1Ez2E_{z_{1}}\cup E_{z_{2}} contains a P5P_{5}, unless Ez1Ez2E_{z_{1}}\cup E_{z_{2}} is a graph obtained form K1,kK_{1,k}, k2k\geq 2, by adding an edge of K1,k¯\overline{K_{1,k}} (the resulting graph is denoted by K1,k+K_{1,k}^{+}). If |NB(z1)NB(z2)|=2|N_{B}(z_{1})\cap N_{B}(z_{2})|=2, then Ez1Ez2E_{z_{1}}\cup E_{z_{2}} contains a P5P_{5}, unless Ez1Ez2E_{z_{1}}\cup E_{z_{2}} is either a K4K_{4}, or a K4K_{4}^{-}, or a C4C_{4}, where K4K_{4}^{-} is a graph obtained from K4K_{4} by deleting an edge. If |NB(z1)NB(z2)|=3|N_{B}(z_{1})\cap N_{B}(z_{2})|=3, then Ez1Ez2E_{z_{1}}\cup E_{z_{2}} contains a P5P_{5}. Recall that BB does not contain 2P3,P52P_{3},P_{5} or P2P4P_{2}\cup P_{4} by assumption. Hence, either Ez1Ez2E_{z_{1}}\cup E_{z_{2}} is a double star or Ez1Ez2{K1,k+,K4,K4,C4}E_{z_{1}}\cup E_{z_{2}}\in\{K_{1,k}^{+},K_{4},K_{4}^{-},C_{4}\}. For convenience, let H~\widetilde{H} denote the subgraph of BB induced by Ez1Ez2E_{z_{1}}\cup E_{z_{2}}. Then either H~\widetilde{H} is a double star or H~{K1,k+,K4,K4,C4}\widetilde{H}\in\{K_{1,k}^{+},K_{4},K_{4}^{-},C_{4}\}.

If H~{K4,K4,C4}\widetilde{H}\in\{K_{4},K_{4}^{-},C_{4}\}, then each edge between U=V(H~)U=V(\widetilde{H}) and VUV-U is red; otherwise there is a blue P5P_{5}, a contradiction. It is obvious that GUG-U is a red complete graph; otherwise there is a blue P2P4P_{2}\cup P_{4}, a contradiction. Choose a vertex uVUu\in V-U, R{u,z1}R-\{u,z_{1}\} contains a red PnP_{n} when n=5n=5. For n6n\geq 6, since Δ(B{u,z1})=2\Delta(B-\{u,z_{1}\})=2, it follows that δ(R{u,z1})n3n2\delta(R-\{u,z_{1}\})\geq n-3\geq\frac{n}{2}. By Dirac’s Theorem, R{u,z1}R-\{u,z_{1}\} contains a red PnP_{n}. Note that uu connects every other vertex by red an edge. For n5n\geq 5, there is a red K^n\widehat{K}_{n}, a contradiction.

If H~=K1,2+\widetilde{H}=K_{1,2}^{+} (in other words, H~\widetilde{H} is a triangle), then dB(z1)=dB(z2)=2d_{B}(z_{1})=d_{B}(z_{2})=2. without loss of generality, suppose that NB(z1)NB(z2)={z}N_{B}(z_{1})\cap N_{B}(z_{2})=\{z\}. By the maximality of dB(z1)+dB(z2)d_{B}(z_{1})+d_{B}(z_{2}), we have that dB(z)=2d_{B}(z)=2. Since BB does not contain 2P32P_{3}, Δ(B{z1,z2,z})1\Delta(B-\{z_{1},z_{2},z\})\leq 1. If n=5n=5, then there is a vertex uu of V{z1,z2,z}V-\{z_{1},z_{2},z\} such that dB(u)=0d_{B}(u)=0. Since all edges between H~\widetilde{H} and VuH~V-u-\widetilde{H} are red, it is clear that there is a red PnP_{n} between H~\widetilde{H} and VuH~V-u-\widetilde{H}. Hence, uPnu\vee P_{n} is a red K^n\widehat{K}_{n}, a contradiction. If n6n\geq 6, then choose a vertex uV{z1,z2,z}u\in V-\{z_{1},z_{2},z\} and let U=VNB(u)U=V-N_{B}(u). Since dB(u)1d_{B}(u)\leq 1, it follows that |U|n|U|\geq n, Δ(B[U])1\Delta(B[U])\leq 1 and uu connects every vertex of UU by an red edge. Then δ(R[U])n3n2\delta(R[U])\geq n-3\geq\frac{n}{2}. Hence, R[U]R[U] contains a Hamiltonian path PnP_{n}, and RR contains a K^n\widehat{K}_{n}, a contradiction.

If H~=K1,k+\widetilde{H}=K_{1,k}^{+}, where k3k\geq 3, then we can assume that the edges in H~\widetilde{H} are

z1x1,,z1xk1,z1z2,z2xk1.z_{1}x_{1},\ldots,z_{1}x_{k-1},z_{1}z_{2},z_{2}x_{k-1}.

It is obvious that Δ(Bz1)\Delta(B-z_{1}) contains only one edge z2xk1z_{2}x_{k-1}; otherwise there is either a blue P5P_{5} or a blue P2P4P_{2}\cup P_{4}, a contradiction. Therefore, δ(R{z1,x1})=n2n2\delta(R-\{z_{1},x_{1}\})=n-2\geq\frac{n}{2}. By Dirac’s Theorem, R{z1,x1}R-\{z_{1},x_{1}\} contains a Hamiltonian path PnP_{n}. Since x1x_{1} is an isolated vertex in BB, it follows that RR contains a K^n\widehat{K}_{n}, a contradiction.

If H~\widetilde{H} is a double star, then NB(z1)NB(z2)=N_{B}(z_{1})\cap N_{B}(z_{2})=\emptyset. If B{z1,z2}B-\{z_{1},z_{2}\} contains an edge ee, then Ez1Ez2E_{z_{1}}\cup E_{z_{2}} contains either a P5P_{5} or a P2P4P_{2}\cup P_{4}, unless Ez1Ez2E_{z_{1}}\cup E_{z_{2}} is a P4P_{4} and e=xye=xy (say the edges in Ez1Ez2E_{z_{1}}\cup E_{z_{2}} are xz1,z1z2xz_{1},z_{1}z_{2} and z2yz_{2}y). By the maximality of dB(z1)=dB(z2)d_{B}(z_{1})=d_{B}(z_{2}), we have that dB(x)=dB(y)=dB(z1)=dB(z2)=2d_{B}(x)=d_{B}(y)=d_{B}(z_{1})=d_{B}(z_{2})=2, and hence BB is a C4C_{4} with E(B)={z1z2,xz1,z2y,xy}E(B)=\{z_{1}z_{2},xz_{1},z_{2}y,xy\}. Recall that the case have discussed since B=Ez1EyB=E_{z_{1}}\cup E_{y} is a C4C_{4}. Therefore, assume that B{z1,z2}B-\{z_{1},z_{2}\} is the empty graph. Since |NB(z1){z2}||NB(z2){z1}|N+2=n|N_{B}(z_{1})-\{z_{2}\}|\cup|N_{B}(z_{2})-\{z_{1}\}|\leq N+2=n, either |NB(z1){z2}|n2|N_{B}(z_{1})-\{z_{2}\}|\leq\frac{n}{2} or |NB(z2){z1}|n2|N_{B}(z_{2})-\{z_{1}\}|\leq\frac{n}{2} (say |NB(z2){z1}|n2|N_{B}(z_{2})-\{z_{1}\}|\leq\frac{n}{2}). Then |Vz1NB[z2]|n+2(n2+2)=n22|V-z_{1}-N_{B}[z_{2}]|\geq n+2-(\frac{n}{2}+2)=\frac{n}{2}\geq 2. Choose two vertices x,yx,y of Vz1NB[z2]V-z_{1}-N_{B}[z_{2}]. It is obvious that xx connects every vertex of Vz1V-z_{1} by a red edge and yz2yz_{2} is a red edge. Note that R[V{z1,z2,x}]R[V-\{z_{1},z_{2},x\}] is a complete graph. Hence, there is a Hamiltonian path Q=Pn1Q=P_{n-1} of R[V{z1,z2,x}]R[V-\{z_{1},z_{2},x\}] with one endpoint yy, and hence Q{yz2}Q\cup\{yz_{2}\} is a red PnP_{n}. Now we obtain a red K^n=x(Q{yz2})\widehat{K}_{n}=x\vee(Q\cup\{yz_{2}\}), a contradiction. ∎

Let μL\mu_{L} be the number components of 33-linear forest LL. Note that e(L)=|L|μLe(L)=|L|-\mu_{L}.

Lemma 3.2.

Suppose that N=n+aN=n+a and 3an/43\leq a\leq\lfloor n/4\rfloor. For any red/blue edge-coloring of KNK_{N}, there is either a red copy of K^n\widehat{K}_{n} or a blue copy of 33-linear forest LL such that |L|μL2a|L|-\mu_{L}\geq 2a.

Proof.

Suppose that GG is a red/blue edge-colored KNK_{N} containing no red K^n\widehat{K}_{n}. Our aim is to obtain a blue 33-linear forest LL with |L|μL2a|L|-\mu_{L}\geq 2a in GG. Let G1G_{1} and G2G_{2} be graphs induced by red edges and blue edges, respectively. If G2G_{2} does not contain 33-linear forest, then Δ(G2)1\Delta(G_{2})\leq 1. We can choose a vertex uu of G1G_{1} and SV(G)NG[u]S\subseteq V(G)-N_{G}[u] such that |S|=n|S|=n. Since δ(G1[S])n2>n/2\delta(G_{1}[S])\geq n-2>n/2, G1[S]G_{1}[S] contains a PnP_{n}. Hence GG has a K^n\widehat{K}_{n} with center uu, a contradiction. Therefore, E(L)E(L)\neq\emptyset. We choose a 33-linear forest LL from G2G_{2} such that e(L)=|L|μLe(L)=|L|-\mu_{L} is maximum. Subjects to above, suppose that |L||L| is maximum.

Let U=V(G)V(L)U=V(G)-V(L). Then the following result holds.

Claim 1.

For each component DD of LL (say, D=PmD=P_{m}), the following results hold.

  • (i)(i) If m6m\geq 6, then EG[V(D),U]E(G1)E_{G}[V(D),U]\subseteq E(G_{1}).

  • (ii)(ii) If 3m53\leq m\leq 5, then either EG[U,V(D)]E(G1)E_{G}[U,V(D)]\subseteq E(G_{1}) or there is a vertex vv of DD such that EG[U,V(D)v]E(G1)E_{G}[U,V(D)-v]\subseteq E(G_{1}). Moveover, if vv exists, then vv is a center vertex of DD and G1[V(D){v}]G_{1}[V(D)-\{v\}] contains a Hamiltonian path.

Proof.

Suppose that D=v1v2vmD=v_{1}v_{2}\ldots v_{m}. We first prove the following result.

Fact 1.

For each uUu\in U and i[m1]i\in[m-1], v1u,vmuE(G2)v_{1}u,v_{m}u\not\in E(G_{2}) and at most one of uvi,uvi+1uv_{i},uv_{i+1} belongs to G2G_{2}.

Proof.

If v1uE(G2)v_{1}u\in E(G_{2}) or vmuE(G2)v_{m}u\in E(G_{2}), then we can find a 33-linear forest L{v1u}L\cup\{v_{1}u\} or L{vmu}L\cup\{v_{m}u\} which is larger than LL, a contradiction. For i[m1]i\in[m-1], if uvi,uvi+1E(G2)uv_{i},uv_{i+1}\in E(G_{2}), then we can get a larger 33-linear forest (Lvivi+1){uvi,uvi+1}(L\setminus v_{i}v_{i+1})\cup\{uv_{i},uv_{i+1}\} in G2G_{2}, a contradiction. ∎

(i)(i) Suppose to the contrary that there exists some i[2,m1]i\in[2,m-1] and uUu\in U such that viuE(G2)v_{i}u\in E(G_{2}). Without loss of generality, suppose that 2im/22\leq i\leq m/2. Then (D\{vivi+1}){viu}(D\backslash\{v_{i}v_{i+1}\})\cup\{v_{i}u\} is a new 33-linear forests of G2G_{2} with e(L)=e(L)e(L^{\prime})=e(L) and |L|=|L|+1|L^{\prime}|=|L|+1, which contradicts the choose of LL.

(ii)(ii) If EG[U,V(D)]E(G1)E_{G}[U,V(D)]\subseteq E(G_{1}), then the result holds. Thus, suppose that there exists a vertex viV(D)v_{i}\in V(D) and a vertex uUu^{\prime}\in U such that viuE(G2)v_{i}u^{\prime}\in E(G_{2}).

If m=3m=3, then by Fact 1, i=2i=2 and v2v_{2} is the unique vertex such that EG[V(D)v2,U]E(G1)E_{G}[V(D)-v_{2},U]\subseteq E(G_{1}). Moreover, we have v1v3E(G2)v_{1}v_{3}\not\in E(G_{2}); otherwise uv2v1v3u^{\prime}v_{2}v_{1}v_{3} is a path of G2G_{2} and we can find a 33-linear forest (LD)uv2v1v3(L-D)\cup u^{\prime}v_{2}v_{1}v_{3} larger than LL, a contradiction. Thus, v1v2v_{1}v_{2} is a Hamiltonian path of G1[V(Dv2)]G_{1}[V(D-v_{2})].

If m=4m=4, then by Fact 1, i{2,3}i\in\{2,3\}. Without loss of generality, assume that i=2i=2. If there is a vertex zz of UU with v3zE(G2)v_{3}z\in E(G_{2}), then zuz\neq u^{\prime} by Fact 1. Hence, EG[V(D)v2,U]E(G1)E_{G}[V(D)-v_{2},U]\subseteq E(G_{1}). By the same discussion as above, we get that v1v4,v1v3E(G1)v_{1}v_{4},v_{1}v_{3}\in E(G_{1}); otherwise there is a 33-linear forest larger than LL, a contradiction. So, v3v1v4v_{3}v_{1}v_{4} is a Hamiltonian path of G1[V(Dv2)]G_{1}[V(D-v_{2})].

If m=5m=5, then i{2,4}i\notin\{2,4\}. Otherwise, either L=(L\{v2v3}){v2u}L^{\prime}=(L\backslash\{v_{2}v_{3}\})\cup\{v_{2}u^{\prime}\} or L=(L\{v4v3}){v4u}L^{\prime}=(L\backslash\{v_{4}v_{3}\})\cup\{v_{4}u^{\prime}\} is a 33-linear forest with e(L)=e(L)e(L)=e(L^{\prime}) and |L|<|L||L|<|L^{\prime}|, a contradiction. Hence, i=3i=3 and EG[V(D)v3,U]E(G1)E_{G}[V(D)-v_{3},U]\subseteq E(G_{1}). It is obvious that v1v5,v2v5,v1v4E(G1)v_{1}v_{5},v_{2}v_{5},v_{1}v_{4}\in E(G_{1}); otherwise, we can find a larger 33-linear forest than LL, a contradiction. Therefore, v4v1v5v2v_{4}v_{1}v_{5}v_{2} is a Hamiltonian path of G1[V(Dv3)]G_{1}[V(D-v_{3})]. ∎

We call the unique vertex vv in Claim 1 the removable vertex of DD. Note that our aim is to prove that |L|μL2a|L|-\mu_{L}\geq 2a. For convenience, the proof is proceeded by contradiction, that is, |L|μL2a1|L|-\mu_{L}\leq 2a-1. In the following proof, in order to get a contradiction, we need to find a red K^n\widehat{K}_{n}.

If U=U=\emptyset, then |L|=N|L|=N and |L|μL2|L|3=2N3|L|-\mu_{L}\geq\frac{2|L|}{3}=\frac{2N}{3}. Since an4a\leq\frac{n}{4}, it follows that |L|μL2a|L|-\mu_{L}\geq 2a, the result follows. Thus, suppose that UU\neq\emptyset. Choose a vertex u0Uu_{0}\in U such that |NG2(u0,U)||N_{G_{2}}(u_{0},U)| is as small as possible. By the maximality of LL, we have that Δ(G2[U])1\Delta(G_{2}[U])\leq 1, and hence |NG2(u0,U)|1|N_{G_{2}}(u_{0},U)|\leq 1. Moreover, if |NG2(u0,U)|=1|N_{G_{2}}(u_{0},U)|=1, then G2[U]G_{2}[U] is a perfect matching of G[U]G[U]. Let U=Uu0NG2(u0,U)U^{\prime}=U-u_{0}-N_{G_{2}}(u_{0},U). Then |U||U|2|U^{\prime}|\geq|U|-2.

In order to get a contradiction, we need to find a red K^n\widehat{K}_{n} with center u0u_{0}, that is, to find a path PnP_{n} in G1[NG1(u0)]G_{1}[N_{G_{1}}(u_{0})].

Let 𝒟1\mathcal{D}_{1} be the set of components DD of LL such that EG[V(D),U]E(G1)E_{G}[V(D),U]\subseteq E(G_{1}), and let 𝒟2\mathcal{D}_{2} be the set of other components of LL. By Claim 1, for each D𝒟2D\in\mathcal{D}_{2}, we have 3|D|53\leq|D|\leq 5.

The following claim is immediate, which will be used later.

Claim 2.

If |NG2(u0,U)|=1|N_{G_{2}}(u_{0},U)|=1, then there is no P3P_{3} in 𝒟2\mathcal{D}_{2}.

Proof.

Assume, to the contrary, that 𝒟2\mathcal{D}_{2} contains a path P=vvv′′P=v^{\prime}vv^{\prime\prime}. Then there exists a vertex uUu\in U such that uvE(G2)uv\in E(G_{2}). Recall that |NG2(u0,U)|=1|N_{G_{2}}(u_{0},U)|=1 implying G2[U]G_{2}[U] is a perfect matching of G[U]G[U]. Thus, there exists a vertex uUu^{\prime}\in U such that uuE(G2)uu^{\prime}\in E(G_{2}). Let LL^{\prime} be a 33-linear forest obtained from LL by deleting PP and then adding vvuuv^{\prime}vuu^{\prime}. Then LL^{\prime} is a 33-linear forest larger than LL, a contradiction. ∎

From Claim 2, if |NG2(u0,U)|=1|N_{G_{2}}(u_{0},U)|=1, then for each D𝒟2D\in\mathcal{D}_{2}, |D|4|D|\geq 4.

Let A=D𝒟1V(D)A=\bigcup_{D\in\mathcal{D}_{1}}V(D) and B=D𝒟2V(D)B=\bigcup_{D\in\mathcal{D}_{2}}V(D). Let B0B_{0} be the set of removable vertices. Then by Claim 1,

NG1(u0)=A(BB0)U=V(G)B0NG2(u0,U)u0.N_{G_{1}}(u_{0})=A\cup(B-B_{0})\cup U^{\prime}=V(G)-B_{0}-N_{G_{2}}(u_{0},U)-u_{0}.

Since 2a1|L|μL2|L|32a-1\geq|L|-\mu_{L}\geq\frac{2|L|}{3}, it follows that

|L|3a2,|L|\leq 3a-2,

and hence

|U|N|L|2N3a=n2an2.|U^{\prime}|\geq N-|L|-2\geq N-3a=n-2a\geq\frac{n}{2}.

Since NG1(u0)=V(G)B0NG2(u0,U)u0N_{G_{1}}(u_{0})=V(G)-B_{0}-N_{G_{2}}(u_{0},U)-u_{0}, it suffices to find a path PnP_{n} in G1[V(G)B0NG2(u0,U)u0]G_{1}[V(G)-B_{0}-N_{G_{2}}(u_{0},U)-u_{0}] obtained from two vertex disjoint paths J,JJ,J^{\prime}, which will be defined below, by adding some edges to connect them.

We first construct JJ. Let |𝒟2|=t|\mathcal{D}_{2}|=t and L1,,LtL_{1},\ldots,L_{t} be all the paths in 𝒟2\mathcal{D}_{2}. Then

t|L|3a1<|U|.t\leq\frac{|L|}{3}\leq a-1<|U^{\prime}|.

Let fif_{i} be the removable vertex of LiL_{i}. By Claim 1, each G1[V(Li)]G_{1}[V(L_{i})] contains a path LiL^{\prime}_{i} such that V(Li)=V(Li)fiV(L^{\prime}_{i})=V(L_{i})-f_{i}. Then V(Li)NG1(u0)V(L^{\prime}_{i})\subseteq N_{G_{1}}(u_{0}) for each i[t]i\in[t] and i[t]V(Li)=BB0\bigcup_{i\in[t]}V(L^{\prime}_{i})=B-B_{0}. Let zi1,zi2z_{i}^{1},z_{i}^{2} be the endpoints of LiL^{\prime}_{i}. Since t<|U|t<|U^{\prime}|, we can choose a subset Y={y1,y2,,yt}Y=\{y_{1},y_{2},\ldots,y_{t}\} of UU^{\prime}. Let

E={y1z11,y2z12,y2z21,y3z22,,ytzt12,ytzt1}.E^{\prime}=\{y_{1}z_{1}^{1},y_{2}z_{1}^{2},y_{2}z_{2}^{1},y_{3}z_{2}^{2},\ldots,y_{t}z_{t-1}^{2},y_{t}z_{t}^{1}\}.

Then EE(G1)E^{\prime}\subseteq E(G_{1}) and J=E(i[t]Li)J=E^{\prime}\cup(\bigcup_{i\in[t]}L^{\prime}_{i}) is a path of G1G_{1} with endpoints y1y_{1} and zt2z^{2}_{t}. It is worth noting that if 𝒟2=\mathcal{D}_{2}=\emptyset, then let JJ be the empty graph.

Now we construct JJ^{\prime}. Let X=UYX=U^{\prime}-Y and =min{|X|,|A|}\ell=\min\{|X|,|A|\}. It follows from |U|>|Y||U^{\prime}|>|Y| that XX\neq\emptyset. Let AA^{\prime} be a subset of AA and XX^{\prime} be a subset of XX such that |A|=|X|=|A^{\prime}|=|X^{\prime}|=\ell. It is obvious that G1[AX]G_{1}[A^{\prime}\cup X^{\prime}] contains a red path JJ^{\prime} of order 22\ell. Moreover, one endpoint of JJ^{\prime} is in AA^{\prime} (say bb^{\prime}) and the other endpoint is in XX (say bb). Since XX\neq\emptyset, if =0\ell=0, then A=A=\emptyset, and let JJ^{\prime} be the empty graph.

We construct a red path PP as follows. If both J,JJ,J^{\prime} are not empty, then we let P=JJ{by1}P=J\cup J^{\prime}\cup\{b^{\prime}y_{1}\}; if JJ^{\prime} is the empty graph, then let P=JP=J; if JJ is the empty graph, then let P=JP=J^{\prime}. It is clear that PP is a red path of G1G_{1} with order 2+|BB0|+t2\ell+|B-B_{0}|+t and V(P)NG1(u0)V(P)\subseteq N_{G_{1}}(u_{0}). If |X||A||X|\leq|A|, then =|X|\ell=|X| and

2+|BB0|+t=2|U|t+|BB0|2|U|+tn2\ell+|B-B_{0}|+t=2|U^{\prime}|-t+|B-B_{0}|\geq 2|U^{\prime}|+t\geq n

(note that |BB0|2t|B-B_{0}|\geq 2t since LL is a 33-linear forest). Thus, u0Pu_{0}\vee P contains a red K^n\widehat{K}_{n}, a contradiction. Hence, we assume that |X|>|A||X|>|A| below (then =|A|\ell=|A|).

Refer to caption
Figure 1: The path PP^{*}.
Claim 3.

G1[NG1(u0)]G_{1}[N_{G_{1}}(u_{0})] contains a path of order N|B0|1|NG2(u0,U)|N-|B_{0}|-1-|N_{G_{2}}(u_{0},U)|.

Proof.

Let U′′=UYX=XXU^{\prime\prime}=U^{\prime}-Y-X^{\prime}=X-X^{\prime}. Since XXX^{\prime}\subseteq X and |X|==|A|<|X||X^{\prime}|=\ell=|A|<|X|, it follows that U′′U^{\prime\prime}\neq\emptyset. Let PP^{\prime} be a maximum path in G1[U′′]G_{1}[U^{\prime\prime}]. Since Δ(G2[U′′])1\Delta(G_{2}[U^{\prime\prime}])\leq 1, it follows that PP^{\prime} is a Hamiltonian path of G1[U′′]G_{1}[U^{\prime\prime}] (say the endpoints of PP^{\prime} are z,zz,z^{*}; if |U′′|=1|U^{\prime\prime}|=1, then z=zz=z^{*}), unless |U′′|=2|U^{\prime\prime}|=2 and the two vertices of U′′U^{\prime\prime} forms a blue edge.

If PP^{\prime} is a Hamiltonian path of G1[U′′]G_{1}[U^{\prime\prime}], then P=PP{zt2z}P^{*}=P^{\prime}\cup P\cup\{z^{2}_{t}z\} is a path of G1G_{1} and |P|=N|B0|1|NG2(u0,U)||P^{*}|=N-|B_{0}|-1-|N_{G_{2}}(u_{0},U)|. Otherwise, |U′′|=2|U^{\prime\prime}|=2 (say U′′={s1,s2}U^{\prime\prime}=\{s_{1},s_{2}\}) and s1s2E(G2)s_{1}s_{2}\in E(G_{2}). Thus, we assume that |P|=|U′′|1|P^{\prime}|=|U^{\prime\prime}|-1. Then U′′={s1,s2}U^{\prime\prime}=\{s_{1},s_{2}\}. Note that one endpoint of PP is zt2z_{t}^{2} and the other endpoint is w{b,y1}w\in\{b,y_{1}\} (if JJ^{\prime} is the empty graph, then w=y1w=y_{1}; otherwise w=bw=b). Since Δ(G2[U])1\Delta(G_{2}[U])\leq 1, we have that one of ws1,ws2ws_{1},ws_{2} is in E(G1)E(G_{1}) (by symmetry, suppose ws2E(G1)ws_{2}\in E(G_{1})). If JJ is not the empty graph, then P=P{ws2,zt2s1}P^{*}=P\cup\{ws_{2},z_{t}^{2}s_{1}\} is a path of G1G_{1}, and |P|=N|B0|1|NG2(u0,U)||P^{*}|=N-|B_{0}|-1-|N_{G_{2}}(u_{0},U)|; if JJ is the empty graph, then P=P{ws2,bs1}P^{*}=P\cup\{ws_{2},b^{\prime}s_{1}\} is a path of G1G_{1}, and |P|=N|B0|1|NG2(u0,U)||P^{*}|=N-|B_{0}|-1-|N_{G_{2}}(u_{0},U)|. ∎

By Claim 3, in order to get a PnP_{n} in G1[NG1(u0)]G_{1}[N_{G_{1}}(u_{0})], we only need to prove that

N|B0|1|NG2(u0,U)|n.N-|B_{0}|-1-|N_{G_{2}}(u_{0},U)|\geq n.

Suppose, to the contrary, that N|B0|1|NG2(u0,U)|<nN-|B_{0}|-1-|N_{G_{2}}(u_{0},U)|<n. Then

n+a=Nn+|B0|+|NG2(u0,U)|,n+a=N\leq n+|B_{0}|+|N_{G_{2}}(u_{0},U)|,

and hence a|B0|+|NG2(u0,U)|a\leq|B_{0}|+|N_{G_{2}}(u_{0},U)|. If |NG2(u0,U)|=0|N_{G_{2}}(u_{0},U)|=0, then a|B0|a\leq|B_{0}|. Recall that |L|3a2|L|\leq 3a-2. Then a|B0|μL3a23a\leq|B_{0}|\leq\mu_{L}\leq\frac{3a-2}{3}, a contradiction. If |NG2(u0,U)|=1|N_{G_{2}}(u_{0},U)|=1, then it follows from Claim 2 that each path of 𝒟2\mathcal{D}_{2} has order at least 44, and hence |B0||L|4|B_{0}|\leq\frac{|L|}{4}. Then a|B0|+1|L|4+13a+24a\leq|B_{0}|+1\leq\frac{|L|}{4}+1\leq\frac{3a+2}{4}, which contradicts that a3a\geq 3. Therefore, we get a PnP_{n} in NG1(u0)N_{G_{1}}(u_{0}), and hence there is a K^n\widehat{K}_{n} in G1G_{1}, a contradiction. ∎

Now we give the proof of Theorem 1.6.

Proof of Theorem 1.6: Let N=n+m/21N=n+\lceil m/2\rceil-1 and G=KNG=K_{N}. We partition V(G)V(G) into two parts A,BA,B such that |A|=n|A|=n and |B|=m/21|B|=\lceil m/2\rceil-1. Let Γ\Gamma be a red/blue edge-coloring of GG such that each edge of G[A]G[A] is red and the other edges are blue. It is easy to verify that there is no a red copy of K^n\widehat{K}_{n}. Suppose that LL is a blue linear forest with maximum number of edges. Since each edge of LL incidents to at least one vertex of BB and each vertex of BB incidents to at most two edges of LL, it follows that e(L)2|B|m1e(L)\leq 2|B|\leq m-1, and hence GG does not contain blue linear forests of size at least mm. Therefore, r(K^n,)n+m2r(\widehat{K}_{n},\mathcal{L})\geq n+\lceil\frac{m}{2}\rceil and r(K^n,)n+m2r(\widehat{K}_{n},\mathcal{L}^{\prime})\geq n+\lceil\frac{m}{2}\rceil. By Lemmas 3.1 and 3.2, we have r(K^n,)n+m2r(\widehat{K}_{n},\mathcal{L})\leq n+\lceil\frac{m}{2}\rceil and r(K^n,)n+m2r(\widehat{K}_{n},\mathcal{L}^{\prime})\leq n+\lceil\frac{m}{2}\rceil by assuming that a=m2a=\lceil\frac{m}{2}\rceil. Hence, r(K^n,)=n+m2r(\widehat{K}_{n},\mathcal{L})=n+\lceil\frac{m}{2}\rceil.

4 Proof of Theorems 1.7, 1.8 and 1.9

In order to prove Theorems 1.7, 1.8 and 1.9, we need the exact values of bk(Pn)b_{k}(P_{n}) and t(Pn)t(P_{n}) first. The following is the exact value of bk(Pn)b_{k}(P_{n}).

Lemma 4.1.

If k3k\geq 3 and n2(k1)n\geq 2(k-1), then

bk(Pn)={n,if2(k1)n4(k2)+1,3n32,ifn>4(k2)+1.b_{k}(P_{n})=\left\{\begin{array}[]{ll}n,&\mbox{if}~{}2(k-1)\leq n\leq 4(k-2)+1,\\ \left\lceil\frac{3n-3}{2}\right\rceil,&\mbox{if}~{}n>4(k-2)+1.\\ \end{array}\right.
Proof.

Let NN be a positive integer and GG be an edge-colored KNK_{N} such that the edge-coloring belongs to k(N)\mathcal{B}_{k}(N). Let V1,,Vk1V_{1},\ldots,V_{k-1} be all the k1k-1 parts. Without loss of generality, let 2|V1||Vk1|2\leq|V_{1}|\leq\ldots\leq|V_{k-1}|. Let U=i=1k2ViU=\bigcup_{i=1}^{k-2}V_{i}.

If 2(k1)n4(k2)+12(k-1)\leq n\leq 4(k-2)+1, then let N=nN=n. It is obvious that |U|2(k2)n2|U|\geq 2(k-2)\geq\left\lfloor\frac{n}{2}\right\rfloor. By Lemma 2.1, GG contains a Hamiltonian path with color 1. Thus, bk(Pn)=nb_{k}(P_{n})=n. Therefore, let n>4(k2)+1n>4(k-2)+1 in the following discussion.

We first prove the upper bounds. Let N=3n32N=\left\lceil\frac{3n-3}{2}\right\rceil. If |U|n2|U|\geq\left\lfloor\frac{n}{2}\right\rfloor, then by Lemma 2.1, GG contains a monochromatic path PnP_{n} with color 11. If |U|<n2|U|<\left\lfloor\frac{n}{2}\right\rfloor, then

|Vk1|N(n21)=n.\displaystyle|V_{k-1}|\geq N-(\left\lfloor\frac{n}{2}\right\rfloor-1)=n. (1)

Thus, by Theorem 2.1, there is an integer m2m\geq 2 such that |Vk1|=r(Pn,Pm)|V_{k-1}|=r(P_{n},P_{m}), and hence G[Vk1]G[V_{k-1}] contains either a monochromatic PmP_{m} with color 11 or a monochromatic PnP_{n} with color kk. If the latter holds, then the monochromatic PnP_{n} is obtained. Hence, suppose that the former holds. If mnm\geq n, then the result follows. Thus, let m<nm<n and G[Vi1]G[V_{i-1}] contains a monochromatic PmP_{m} with color 11. Then |Vk1|=r(Pn,Pm)=n+m21|V_{k-1}|=r(P_{n},P_{m})=n+\left\lfloor\frac{m}{2}\right\rfloor-1. Since UU\neq\emptyset and Vk1V(Pm)V_{k-1}-V(P_{m})\neq\emptyset, it follows that GG contains a monochromatic Pm+2P_{m+2} that is obtained from the PmP_{m} by adding two edges e1,e2e_{1},e_{2} between UU and Vk1V(Pn)V_{k-1}-V(P_{n}), where e1e_{1} joins one endpoint of PmP_{m} and a vertex of UU (say xx is the endpoint of e1e_{1} belonging to UU), and e2e_{2} joins xx and a vertex of Vk1V(Pm)V_{k-1}-V(P_{m}). This implies that GG contains a monochromatic PnP_{n} when n2m<nn-2\leq m<n. Hence, we assume that n3m2n-3\geq m\geq 2 below.

Case 1.

mm is odd.

Let W=Vk1V(Pm)W=V_{k-1}-V(P_{m}). If |W||U||W|\leq|U|, then G[WU]G[W\cup U] contains a monochromatic path P2|W|P_{2|W|} with color 11, and hence GG contains a monochromatic Pm+2|W|P_{m+2|W|} obtained form the monochromatic paths PmP_{m} and P2|W|P_{2|W|} by adding an edge joins one endpoint of PmP_{m} and the endpoint of P2|W|P_{2|W|} belonging to UU. Since m+2|W|=|Vk1|+|W|nm+2|W|=|V_{k-1}|+|W|\geq n, GG contains a monochromatic PnP_{n}. If |W|>|U||W|>|U|, then GG contains a monochromatic copy of P2|U|+mP_{2|U|+m}. It suffices to prove that 2|U|+mn2|U|+m\geq n. Since mm is odd, it follows that |Vk1|=n+m121|V_{k-1}|=n+\frac{m-1}{2}-1, and hence m=2|Vk1|2n+3m=2|V_{k-1}|-2n+3. So

2|U|+m=2|U|+2|Vk1|2n+3=2N2n+3=n.\displaystyle 2|U|+m=2|U|+2|V_{k-1}|-2n+3=2N-2n+3=n.
Case 2.

mm is even and m4m\geq 4.

Let m+2=m1+m2m+2=m_{1}+m_{2}, where m1,m23m_{1},m_{2}\geq 3 and m1,m2m_{1},m_{2} are odd. Since |Vk1|=n+m21=n+(m+2)221|V_{k-1}|=n+\frac{m}{2}-1=n+\frac{(m+2)-2}{2}-1 and n3mn-3\geq m, it follows from Theorem 2.2 that |Vk1|=r(Pn,Pm1Pm2)|V_{k-1}|=r(P_{n},P_{m_{1}}\cup P_{m_{2}}), and hence G[Vk1]G[V_{k-1}] contains either a monochromatic path PnP_{n} with color kk, or a monochromatic path Pm1Pm2P_{m_{1}}\cup P_{m_{2}} with color 11. If the former holds, then GG contains a monochromatic PnP_{n}. Hence, suppose that the latter holds. Assume that uUu\in U. Let U=U{u}U^{\prime}=U-\{u\} and W=Vk1V(Pm1Pm2)W=V_{k-1}-V(P_{m_{1}}\cup P_{m_{2}}). If |W||U||W|\leq|U^{\prime}|, then G[WU]G[W\cup U^{\prime}] contains a monochromatic path P2|W|P_{2|W|} with color 11. We can choose the two endpoints of P2|W|P_{2|W|} as xWx\in W and yUy\in U^{\prime}. Without loss of generality, suppose that the two endpoints of Pm1P_{m_{1}} are x1,y1x_{1}^{\prime},y_{1}^{\prime} and the two endpoints of Pm2P_{m_{2}} are x2,y2x_{2}^{\prime},y_{2}^{\prime}. Then Pm1Pm2P2|W|{uy1,ux2,y2y}P_{m_{1}}\cup P_{m_{2}}\cup P_{2|W|}\cup\{uy_{1}^{\prime},ux_{2}^{\prime},y_{2}^{\prime}y\} is a monochromatic P2|W|+(m1+m2)+1P_{2|W|+(m_{1}+m_{2})+1} with color 11. Note that

2|W|+(m1+m2)+1=|W|+|Vk1|+1>|Vk1|n2|W|+(m_{1}+m_{2})+1=|W|+|V_{k-1}|+1>|V_{k-1}|\geq n

by the inequality (1). It follows that GG contains a monochromatic copy of PnP_{n} with color 11. If |W|>|U||W|>|U^{\prime}|, then GG contains a monochromatic copy of Pm1+m2+1+2|U|P_{m_{1}+m_{2}+1+2|U^{\prime}|}. We need to prove that m1+m2+1+2|U|=m+2|U|+1nm_{1}+m_{2}+1+2|U^{\prime}|=m+2|U|+1\geq n. Since |Vk1|=n+m21|V_{k-1}|=n+\left\lfloor\frac{m}{2}\right\rfloor-1 and mm is even, it follows that m=2|Vk1|2n+2m=2|V_{k-1}|-2n+2, and hence

m+2|U|+1=2|Vk1|2n+2|U|+3=2(Nn)+3n.m+2|U|+1=2|V_{k-1}|-2n+2|U|+3=2(N-n)+3\geq n.
Case 3.

mm is even and 2m32\leq m\leq 3.

In this case, |Vk1|=n|V_{k-1}|=n and |U|=N|Vk1|=n121|U|=N-|V_{k-1}|=\lceil\frac{n-1}{2}\rceil-1. Since |Vk1|=r(Pn,P3)|V_{k-1}|=r(P_{n},P_{3}), if G[Vk1]G[V_{k-1}] does not contain a monochromatic copy of PnP_{n} with color kk, the monochromatic PnP_{n} is obtained. Otherwise, G[Vk1]G[V_{k-1}] contain monochromatic F=P3F=P_{3} with color 11. We can choose a subset WW of Vk1V(F)V_{k-1}-V(F) such that |W|=n121|W|=\lceil\frac{n-1}{2}\rceil-1, since |Vk1V(F)|=n3n121|V_{k-1}-V(F)|=n-3\geq\lceil\frac{n-1}{2}\rceil-1. Then G[WU]G[W\cup U] contains a monochromatic P2|W|P_{2|W|} with color 11, and there is a monochromatic P2|W|+3P_{2|W|+3} with color 11 that is obtained by connecting the endpoint of P2|W|P_{2|W|} belonging to UU and an endpoint of P3P_{3}. Since 2|W|+3n2|W|+3\geq n, there is a monochromatic PnP_{n} with color 11.

Now we prove the lower bounds. Let M=3n321M=\left\lceil\frac{3n-3}{2}\right\rceil-1 and G=KMG=K_{M}. We partition V(G)V(G) into k1k-1 parts V1,,Vk1V_{1},\ldots,V_{k-1} such that |V1|=|V2|==|Vk2|=2|V_{1}|=|V_{2}|=\ldots=|V_{k-2}|=2 and |Vk1|=3n122(k1)|V_{k-1}|=\left\lceil\frac{3n-1}{2}-2(k-1)\right\rceil. Let Vi={ui,vi}V_{i}=\{u_{i},v_{i}\} for i[k2]i\in[k-2], and let G[Vk1]=HH¯G[V_{k-1}]=H\cup\overline{H}, where HH is the disjoint union of H1=Kn4(k1)+12H_{1}=K_{\left\lceil\frac{n-4(k-1)+1}{2}\right\rceil} and H2=Kn1H_{2}=K_{n-1}. Choose a kk-edge-coloring Γk(M)\Gamma\in\mathcal{B}_{k}(M) such that uiviu_{i}v_{i} is colored by i+1i+1 for i[k2]i\in[k-2] and HH is a monochromatic graph with color kk, and the other edges are colored by 11. It is easy to verify that there is no monochromatic PnP_{n} with color i2i\geq 2. For color 11, it is clear that the edges with color 11 induces a spanning complete kk-partite graph with kk parts V1,V2,,Vk2,V(H1),V(H2)V_{1},V_{2},\ldots,V_{k-2},V(H_{1}),V(H_{2}). Let PP be a minimum monochromatic path. Since V(H2)V(H_{2}) is an independent set in the complete kk-partite graph, each edge of PP incidents with a vertex of V(G)V(H2)V(G)-V(H_{2}), and hence e(P)2|V(G)V(H2)|=2(Mn+1)n2.e(P)\leq 2|V(G)-V(H_{2})|=2(M-n+1)\leq n-2. Thus, |P|n1|P|\leq n-1, and there is no monochromatic PnP_{n} with color 11. Therefore, there is no monochromatic PnP_{n} in GG, which implies that bk(Pn)M+1=3n32b_{k}(P_{n})\geq M+1=\left\lceil\frac{3n-3}{2}\right\rceil. ∎

The following is the exact value of t(Pn)t(P_{n}).

Lemma 4.2.

If n3n\geq 3, then

t(Pn)={3n21,n is even,3n12,n is odd .t(P_{n})=\left\{\begin{array}[]{ll}\frac{3n}{2}-1,&n\mbox{ is even},\\ \frac{3n-1}{2},&n\mbox{ is odd }.\end{array}\right.
Proof.

We first prove the lower bounds. Suppose that M=3n22M=\frac{3n}{2}-2 if nn is even, and M=3n121M=\frac{3n-1}{2}-1 if nn is odd. Let Γ𝒯(M)\Gamma^{\prime}\in\mathcal{T}(M) be an edge-coloring of G=KMG^{\prime}=K_{M} and the three parts are V1,V2,V3V_{1},V_{2},V_{3}. If nn is even, then let |V1|=n2|V_{1}|=\frac{n}{2} and |V2|=|V3|=n21|V_{2}|=|V_{3}|=\frac{n}{2}-1. If nn is odd, then let |V1|=|V2|=|V3|=n12|V_{1}|=|V_{2}|=|V_{3}|=\frac{n-1}{2}. Moreover, let G[Vi]G^{\prime}[V_{i}] be a complete graph with color ii. Then KMK_{M} does not contain a monochromatic copy of PnP_{n}. Therefore, the lower bounds are obtained.

Now we prove the upper bounds. Suppose that N=3n21N=\frac{3n}{2}-1 if nn is even, and N=3n12N=\frac{3n-1}{2} if nn is odd. Let Γ𝒯(N)\Gamma\in\mathcal{T}(N) be an edge-coloring of KNK_{N} with the three parts V1,V2,V3V_{1},V_{2},V_{3}. Without loss of generality, suppose that |V1||V2||V3||V_{1}|\geq|V_{2}|\geq|V_{3}|. Note that V2,V3V_{2},V_{3}\neq\emptyset.

Let a1=n2|V2|a_{1}=n-2|V_{2}|. If a1<2a_{1}<2, then |V2|n12|V_{2}|\geq\lceil\frac{n-1}{2}\rceil. If |V1V2|n|V_{1}\cup V_{2}|\geq n, then there is a monochromatic PnP_{n} between V1V_{1} and V2V_{2}. Otherwise, suppose that |V1V2|n1|V_{1}\cup V_{2}|\leq n-1. Then |V1|n12|V_{1}|\leq\lfloor\frac{n-1}{2}\rfloor. Since |V1||V2||V3||V_{1}|\geq|V_{2}|\geq|V_{3}|, it follows that nn is odd and |V1|=|V2|=n12|V_{1}|=|V_{2}|=\frac{n-1}{2}. However, |V3|=N|V1V2|(3n1)/2(n1)=(n+1)/2>|V2||V_{3}|=N-|V_{1}\cup V_{2}|\geq(3n-1)/2-(n-1)=(n+1)/2>|V_{2}|, a contradiction. Therefore, suppose a12a_{1}\geq 2 below.

Let a2=|V1|a12+1a_{2}=|V_{1}|-\lfloor\frac{a_{1}}{2}\rfloor+1. Then a2a1a_{2}\geq a_{1}. Otherwise, if a2<a1a_{2}<a_{1}, then

|V1|\displaystyle|V_{1}| =a2+a121a1+a221a1+a1121\displaystyle=a_{2}+\left\lfloor\frac{a_{1}}{2}\right\rfloor-1\leq a_{1}+\left\lfloor\frac{a_{2}}{2}\right\rfloor-1\leq a_{1}+\left\lfloor\frac{a_{1}-1}{2}\right\rfloor-1
n2|V2|+n2|V2|121=n3|V2|+n32\displaystyle\leq n-2|V_{2}|+\left\lfloor\frac{n-2|V_{2}|-1}{2}\right\rfloor-1=n-3|V_{2}|+\left\lfloor\frac{n-3}{2}\right\rfloor
n(N|V1|)|V2|+n32\displaystyle\leq n-(N-|V_{1}|)-|V_{2}|+\left\lfloor\frac{n-3}{2}\right\rfloor
|V1||V2|<|V1|,\displaystyle\leq|V_{1}|-|V_{2}|<|V_{1}|,

a contradiction. Thus, 2a1a22\leq a_{1}\leq a_{2}. By Theorem 2.1, we have that |V1|=r(Pa1,Pa2)|V_{1}|=r(P_{a_{1}},P_{a_{2}}). Hence, G[V1]G[V_{1}] contains either a monochromatic copy of Pa1P_{a_{1}} with color 11 or a monochromatic copy of Pa2P_{a_{2}} with color 33.

Suppose that G[V1]G[V_{1}] contains a monochromatic copy of Pa1P_{a_{1}} with color 11 (say ww is an endpoint of Pa1P_{a_{1}}). Let U1=V1V(Pa1)U_{1}=V_{1}-V(P_{a_{1}}) and let =min{|U1|,|V2|}\ell=\min\{|U_{1}|,|V_{2}|\}. Then there is a monochromatic P2P_{2\ell} between U1U_{1} and V2V_{2} with color 11. Without loss of generality, suppose that w1w_{1} are endpoints of the path P2P_{2\ell} with w1V2w_{1}\in V_{2}. Then P2Pa1{ww1}P_{2\ell}\cup P_{a_{1}}\cup\{ww_{1}\} monochromatic path of order 2+a12\ell+a_{1} (see Figure 2 (1)). We only need to verify that 2+a1n2\ell+a_{1}\geq n below. If |U1||V2||U_{1}|\geq|V_{2}|, then 2+a1=2|V2|+a1=n2\ell+a_{1}=2|V_{2}|+a_{1}=n. If |U1|<|V2||U_{1}|<|V_{2}|, then

2+a1=2|U1|+a1=2|V1|a1=2(|V1|+|V2|)n22N3n=n.2\ell+a_{1}=2|U_{1}|+a_{1}=2|V_{1}|-a_{1}=2(|V_{1}|+|V_{2}|)-n\geq 2\left\lceil\frac{2N}{3}\right\rceil-n=n.
Refer to caption
Figure 2: Construct a monochromatic PnP_{n}.

Suppose that G[V1]G[V_{1}] contains a monochromatic Pa2P_{a_{2}} with color 33. Let a=min{a2,|V1||V3|}a^{\prime}=\min\{a_{2},|V_{1}|-|V_{3}|\}. Then we can choose a subpath PP of Pa2P_{a_{2}} with |P|=a|P|=a^{\prime} (say ww is an endpoint of PP). Let U1=V1V(P)U_{1}=V_{1}-V(P). Then |U1||V3||U_{1}|\geq|V_{3}|. Then there is a monochromatic P2|V3|P_{2|V_{3}|} between U1U_{1} and V3V_{3} with color 33. Without loss of generality, suppose that w1w_{1} is the endpoint of the path P2|V3|P_{2|V_{3}|} with w1V3w_{1}\in V_{3}. Then P2|V3|Pa2{ww1}P_{2|V_{3}|}\cup P_{a_{2}}\cup\{ww_{1}\} monochromatic path of order 2|V3|+a2|V_{3}|+a^{\prime} (see Figure 2 (2)). We only need to verify that 2|V3|+an2|V_{3}|+a^{\prime}\geq n below. Note that

a2=|V1|a12+1=|V1|n2+|V2|+1=(N|V3|)n2+1n|V3|.\displaystyle a_{2}=|V_{1}|-\left\lfloor\frac{a_{1}}{2}\right\rfloor+1=|V_{1}|-\left\lfloor\frac{n}{2}\right\rfloor+|V_{2}|+1=(N-|V_{3}|)-\left\lfloor\frac{n}{2}\right\rfloor+1\geq n-|V_{3}|.

If a=a2a^{\prime}=a_{2}, then 2|V3|+a=2|V3|+a2n2|V_{3}|+a^{\prime}=2|V_{3}|+a_{2}\geq n. If a=|V1||V3|a^{\prime}=|V_{1}|-|V_{3}|, then 2|V3|+a=|V1|+|V3|a2+|V3|n2|V_{3}|+a^{\prime}=|V_{1}|+|V_{3}|\geq a_{2}+|V_{3}|\geq n. ∎

Proof of Theorem 1.7: Let Γ\Gamma be an edge-coloring of KNK_{N} such that KNK_{N} does not contain rainbow P5P_{5}. From Theorem 1.3, one of (i),(ii)(i),(ii) holds if k4k\geq 4, and one of (iii)(v)(iii)-(v) holds if k=4k=4. Note that if N6N\geq 6, then KNK_{N} contains a monochromatic PN1P_{N-1} with color 11 whenever one of (ii)(v)(ii)-(v) holds.

For 2(k1)n4(k2)+12(k-1)\leq n\leq 4(k-2)+1, let N=n+1N=n+1. Since N6N\geq 6, KNK_{N} contains a monochromatic PnP_{n} with color 11 whenever one of (ii)(v)(ii)-(v) holds. By Lemma 4.1, bk(Pn)=nb_{k}(P_{n})=n. So, grk(P5:Pn)n+1\operatorname{gr}_{k}(P_{5}:P_{n})\leq n+1. Since we can choose a Γ\Gamma of (ii)(ii) such that KnK_{n} does not contain monochromatic PnP_{n}, grk(P5:Pn)=n+1\operatorname{gr}_{k}(P_{5}:P_{n})=n+1.

For n>4(k2)+1n>4(k-2)+1, let N=3n32N=\lceil\frac{3n-3}{2}\rceil. By Lemma 4.1, bk(Pn)=Nb_{k}(P_{n})=N. Since n10n\geq 10, it follows that Nn+1N\geq n+1, and hence KNK_{N} contains a monochromatic copy of PnP_{n} whenever Γ\Gamma satisfies one of (ii)(v)(ii)-(v). Therefore, grk(P5:Pn)=N=3n32\operatorname{gr}_{k}(P_{5}:P_{n})=N=\lceil\frac{3n-3}{2}\rceil.

Proof of Theorem 1.8: Let Γ\Gamma^{\prime} be an edge-coloring of KNK_{N^{\prime}} such that KNK_{N^{\prime}} does not contain rainbow P4+P_{4}^{+}. By Theorem 1.5, if k=4k=4, then either Γ{G2(N),G3(N)}\Gamma^{\prime}\in\{G_{2}(N^{\prime}),G_{3}(N^{\prime})\} or (i)(i) holds; if k5k\geq 5, then (i)(i) holds.

Suppose k5k\geq 5. Then gr4(P4+:Pn)=bk(n)\operatorname{gr}_{4}(P_{4}^{+}:P_{n})=b_{k}(n). By Lemma 4.1, the result holds. Thus, we talk about the case k=4k=4 below. If n6n\geq 6, then the maximum monochromatic path in KNK_{N^{\prime}} is PN2P_{N^{\prime}-2} when Γ=G2(N)\Gamma^{\prime}=G_{2}(N^{\prime}), and is PNP_{N^{\prime}} when Γ=G3(N)\Gamma^{\prime}=G_{3}(N^{\prime}). Thus, gr4(P4+:Pn)n+2\operatorname{gr}_{4}(P_{4}^{+}:P_{n})\geq n+2 when n6n\geq 6. If 2(k1)n4(k2)+12(k-1)\leq n\leq 4(k-2)+1, then bk(n)=nb_{k}(n)=n, and hence gr4(P4+:Pn)=n+2\operatorname{gr}_{4}(P_{4}^{+}:P_{n})=n+2. If n>4(k2)+1=9n>4(k-2)+1=9, then bk(Pn)=3n32n+2b_{k}(P_{n})=\lceil\frac{3n-3}{2}\rceil\geq n+2, and hence gr4(P4+:Pn)=3n32\operatorname{gr}_{4}(P_{4}^{+}:P_{n})=\lceil\frac{3n-3}{2}\rceil.

Proof of Theorem 1.9: Let Γ\Gamma^{*} be an edge-coloring of KNK_{N^{*}} such that KNK_{N^{*}} does not contain rainbow K1,3K_{1,3}. If k=3k=3, then either Γ𝒯(N)\Gamma^{*}\in\mathcal{T}(N^{*}) or Γ3(N)\Gamma^{*}\in\mathcal{B}_{3}(N^{*}). Therefore, gr3(K1,3:Pn)=max{t(Pn),b3(Pn)}=t(Pn)\operatorname{gr}_{3}(K_{1,3}:P_{n})=\max\{t(P_{n}),b_{3}(P_{n})\}=t(P_{n}). If k4k\geq 4, then Γk(N)\Gamma^{*}\in\mathcal{B}_{k}(N^{*}) and hence grk(K1,3:Pn)=bk(Pn)\operatorname{gr}_{k}(K_{1,3}:P_{n})=b_{k}(P_{n}).

5 Proof of Theorem 1.10

In order to prove Theorem 1.10, we need the exact values of b3(K^n)b_{3}(\widehat{K}_{n}) and a upper bound of t(K^n)t(\widehat{K}_{n}). The following is the exactly value of b3(K^n)b_{3}(\widehat{K}_{n}).

Lemma 5.1.

If n5n\geq 5 and nn is odd, then b3(K^n)=5n2b_{3}(\widehat{K}_{n})=\left\lfloor\frac{5n}{2}\right\rfloor; if n5n\geq 5 and nn is even, then 5n21b3(K^n)5n2\left\lfloor\frac{5n}{2}\right\rfloor-1\leq b_{3}(\widehat{K}_{n})\leq\left\lfloor\frac{5n}{2}\right\rfloor.

Proof.

The following claim indicates the lower bounds of b3(K^n)b_{3}(\widehat{K}_{n}).

Claim 4.

Let N=(5n)/21N^{\prime}=\lfloor(5n)/2\rfloor-1 if nn is odd, and let N=(5n)/22N^{\prime}=\lfloor(5n)/2\rfloor-2 if nn is even. There exists an 33-edge-coloring Γ3(n)\Gamma\in\mathcal{B}_{3}(n) of G=KNG=K_{N^{\prime}} such that GG does not contain monochromatic K^n\widehat{K}_{n}.

Proof.

We construct Γ\Gamma as follows.

  • Partition V(G)V(G) into four parts A,B1,B2,B3A,B_{1},B_{2},B_{3} such that |A|=n|A|=n, |B1|=|B2|=|B3|=n12|B_{1}|=|B_{2}|=|B_{3}|=\frac{n-1}{2} if nn is odd, and |B1|=n2|B_{1}|=\frac{n}{2} and |B2|=|B3|=n21|B_{2}|=|B_{3}|=\frac{n}{2}-1 if nn is even.

  • Let B=B1B2B3B=B_{1}\cup B_{2}\cup B_{3}. Each edge of G[A]G[A] is colored by 33, each edge of EG[A,B]E_{G}[A,B] is colored by 11, each edge between the three parts B1,B2,B3B_{1},B_{2},B_{3} is colored 22, and each edge in B1B_{1}, B2B_{2} and B3B_{3} are colored by 11.

Clearly, this edge-coloring Γ\Gamma belongs to 3(M)\mathcal{B}_{3}(M) and the two parts are AA and BB. It is easy to verify that there is no monochromatic K^n\widehat{K}_{n} with color 22 or 33. Now we prove that there is no monochromatic K^n\widehat{K}_{n} with color 11. Suppose to the contrary that there is a monochromatic K^n=vP\widehat{K}_{n}=v\vee P with color 11, where PP is a monochromatic PnP_{n} with color 11 and vv is the center of the K^n\widehat{K}_{n}. Let G1G_{1} be the subgraph of GG induced by all edges with color 11. Since G1[B]G_{1}[B] does not contain monochromatic PnP_{n}, vAv\notin A. Thus, vBiv\in B_{i} for some i[3]i\in[3]. Then PP is a subgraph G1[(Biv)A]G_{1}[(B_{i}-v)\cup A]. However, since G1[(Biv)A]G_{1}[(B_{i}-v)\cup A] is a complete bipartite graph with two parts AA and BivB_{i}-v, and since |Biv|n121|B_{i}-v|\leq\frac{n-1}{2}-1, it follows that the maximum path in G1[(Biv)A]G_{1}[(B_{i}-v)\cup A] is 2|Biv|+1n22|B_{i}-v|+1\leq n-2, a contradiction. ∎

Now we prove the upper bound. Suppose that N=5n2N=\left\lfloor\frac{5n}{2}\right\rfloor. Let GG be an edge-colored KNK_{N} and the edge-coloring belong to 3(N)\mathcal{B}_{3}(N). Then V(G)V(G) can be partitioned into two parts A,BA,B with 2|A||B|2\leq|A|\leq|B| such that each edge of G[A]G[A] is colored with 11 or 33, each edge of G[B]G[B] is colored with 11 or 22, and each edge between AA and BB is colored with 11.

Case 1 |A|<n2|A|<\frac{n}{2}.

In this case, |B|2n=r(Pn,K^n)|B|\geq 2n=r(P_{n},\widehat{K}_{n}) by Theorem 2.5. Then there is either a monochromatic copy of PnP_{n} with color 11 or a monochromatic copy of K^n\widehat{K}_{n} with color 22 in G[B]G[B], and hence there is either a monochromatic copy of K^n\widehat{K}_{n} with color 22, or a monochromatic copy of K^n\widehat{K}_{n} with color 11 which is obtained form the monochromatic copy of PnP_{n} with color 11 and a vertex vAv\in A by adding all possible edges between vv and V(Pn)V(P_{n}).

Case 2 n2|A|n\lceil\frac{n}{2}\rceil\leq|A|\leq n.

Clearly, |B|3n2r(K1,n/2,K^n)|B|\geq\left\lfloor\frac{3n}{2}\right\rfloor\geq r(K_{1,\left\lfloor n/2\right\rfloor},\widehat{K}_{n}). Thus, there is either a monochromatic copy of K1,n/2K_{1,\left\lfloor n/2\right\rfloor} with color 11 or a monochromatic copy of K^n\widehat{K}_{n} with color 22 in G[B]G[B]. If the latter holds, then the result follows. If the former holds, then let u0u_{0} be the center and UU be the set of leaves of the star K1,n/2K_{1,\left\lfloor n/2\right\rfloor}. Since |U|=n/2|U|=\left\lfloor n/2\right\rfloor, |A|n2|A|\geq\lceil\frac{n}{2}\rceil and each edge between UU and AA is colored by 11, it follows that there is a monochromatic path P=PnP=P_{n} with color 11 and V(P)UAV(P)\subseteq U\cup A, and hence u0Pu_{0}\vee P is a monochromatic copy of K^n\widehat{K}_{n} with color 11.

Case 3 |A|>n|A|>n.

Clearly, |B|>n|B|>n also holds. Suppose that there is no monochromatic K^n\widehat{K}_{n} with color 22 or 33. We will show that there is a monochromatic K^n\widehat{K}_{n} with color 11 below. Let |A|=n+a|A|=n+a and |B|=n+b|B|=n+b, where a,b1a,b\geq 1. Since a+b=n2a+b=\left\lfloor\frac{n}{2}\right\rfloor, it follows that a,b<n2a,b<\frac{n}{2}. Since |A||B||A|\leq|B|, we have that aba\leq b and an4a\leq\left\lfloor\frac{n}{4}\right\rfloor. By Theorem 2.7, |B|r(K1,b,K^n)|B|\geq r(K_{1,b},\widehat{K}_{n}) and hence there is a monochromatic copy of K1,bK_{1,b} with color 11 in G[B]G[B]. Let u0u_{0} be the center and let UU be the set of leaves in K1,bK_{1,b}. Then |U|=b|U|=b.

Refer to caption
Figure 3: The subsets UU^{\prime} and YY of BB.

By Lemma 3.1 and Lemma 3.2, one of the following statements holds.

  • a3a\geq 3 and there is a 33-linear forest LL in G[A]G[A] with color 11 such that |L|μL2a|L|-\mu_{L}\geq 2a, where μL\mu_{L} is the number of components of LL.

  • a=1a=1 and there is either a 2P22P_{2} or a P3P_{3} in G[A]G[A] with color 11.

  • a=2a=2 and there is either a P2P4P_{2}\cup P_{4} (the vertex disjoint union of P2P_{2} and P4P_{4}), or a P5P_{5}, or a 2P32P_{3} in G[A]G[A] with color 11.

In above cases, |L|μL2a|L|-\mu_{L}\geq 2a. If a3a\geq 3, then we can choose the 33-linear forest LL such that |L|μL=2a+ε|L|-\mu_{L}=2a+\varepsilon, where 0ε20\leq\varepsilon\leq 2.

Claim 5.

|L|3a+3|L|\leq 3a+3 when a3a\geq 3 and |L|2a+2|L|\leq 2a+2 when a{1,2}a\in\{1,2\}. Moreover, μL|U|+1\mu_{L}\leq|U|+1.

Proof.

If a3a\geq 3, then LL is a 33-linear forest and hence μL|L|/3\mu_{L}\leq|L|/3. Thus, |L|3a+3|L|\leq 3a+3. Since aba\leq b, it follows that

μL|L|3a+1b+1=|U|+1.\mu_{L}\leq\frac{|L|}{3}\leq a+1\leq b+1=|U|+1.

If a=1a=1 or a=2a=2, it is easy to verify that |L|2a+2|L|\leq 2a+2 and μL2\mu_{L}\leq 2. Hence, μL2|U|+1\mu_{L}\leq 2\leq|U|+1, since |U|=b1|U|=b\geq 1. ∎

By Claim 5, we can choose μL1\mu_{L}-1 vertices of UU (say y1,,yμL1y_{1},\ldots,y_{\mu_{L}-1}). Let U=UYU^{\prime}=U-Y, where Y={y1,,yμL1}Y=\{y_{1},\ldots,y_{\mu_{L}-1}\}.

Refer to caption
Figure 4: The construction of PP.
Claim 6.

|A||L|>|U||A|-|L|>|U^{\prime}|.

Proof.

If a3a\geq 3, then |L|3a+3|L|\leq 3a+3 and

|A||L|n+a(3a+3)=n2a3n2(n/2b)32b3.\displaystyle|A|-|L|\geq n+a-(3a+3)=n-2a-3\geq n-2\left(\lfloor n/2\rfloor-b\right)-3\geq 2b-3.

Since ba3b\geq a\geq 3, |A||L|b>|U||A|-|L|\geq b>|U^{\prime}|. If a{1,2}a\in\{1,2\}, then |L|2a+2|L|\leq 2a+2 and

|A||L|n+a(2a+2)=na2=n(n/2b)2b>|U|.|A|-|L|\geq n+a-(2a+2)=n-a-2=n-\left(\lfloor n/2\rfloor-b\right)-2\geq b>|U^{\prime}|.

By Claim 6, we can choose a subset AA^{\prime} of AV(L)A-V(L) such that |A|=|U||A^{\prime}|=|U^{\prime}|. We assume that LL consists of μL\mu_{L} paths D1,D2,,DμLD_{1},D_{2},\ldots,D_{\mu_{L}}, and the two endpoints of DiD_{i} are zi1z_{i}^{1} and zi2z_{i}^{2}. Observe that we can construct a monochromatic path PP^{\prime} that is obtained from LL and YY by connecting yiy_{i} and zi2,zi+11z_{i}^{2},z_{i+1}^{1}, where i[μL1]i\in[\mu_{L}-1]. On the other hand, since the edges between UU^{\prime} and AA^{\prime} are colors by 11, we can get a monochromatic P′′P^{\prime\prime} of order 2|U|2|U^{\prime}| with one endpoint in AA^{\prime} and the other endpoint in UU^{\prime} (say zz is the endpoint of P′′P^{\prime\prime} belonging to UU^{\prime}). Then there is a monochromatic path PP with color 11 that is obtained from PP^{\prime} and P′′P^{\prime\prime} by adding the edge zzμL2zz_{\mu_{L}}^{2}. Hence, u0Pu_{0}\vee P is monochromatic kipas (see Figure 4). Note that

|P|\displaystyle|P| =|L|+|Y|+|U|+|A|=|L|+|Y|+2(b|Y|)\displaystyle=|L|+|Y|+|U^{\prime}|+|A^{\prime}|=|L|+|Y|+2(b-|Y|)
=2b+|L||Y|=2b+|L|μL+1.\displaystyle=2b+|L|-|Y|=2b+|L|-\mu_{L}+1.

If a3a\geq 3, then |L|μL=2a+ε|L|-\mu_{L}=2a+\varepsilon. Hence, |P|=2b+2a+ε+1n|P|=2b+2a+\varepsilon+1\geq n, and u0Pu_{0}\vee P contains a monochromatic K^n\widehat{K}_{n} with color 11. If a{1,2}a\in\{1,2\}, it is easy to check that |L|μL=2a|L|-\mu_{L}=2a. Hence, |P|=2b+2a+1n|P|=2b+2a+1\geq n. ∎

The following is an upper bound of t(K^n)t(\widehat{K}_{n}).

Lemma 5.2.

If n5n\geq 5, then t(K^n)5n2t(\widehat{K}_{n})\leq\left\lfloor\frac{5n}{2}\right\rfloor if nn is odd, and t(K^n)5n21t(\widehat{K}_{n})\leq\left\lfloor\frac{5n}{2}\right\rfloor-1 if nn is even.

Proof.

Suppose that N=5n2N=\left\lfloor\frac{5n}{2}\right\rfloor if nn is odd, and N=5n21N=\left\lfloor\frac{5n}{2}\right\rfloor-1 if nn is even. Let G=KNG=K_{N}. Choose an edge-coloring of 𝒯(G)\mathcal{T}(G) with parts A,B,CA,B,C. Without loss of generality, let |A||B||C||A|\geq|B|\geq|C| and let |A|=a,|B|=b|A|=a,|B|=b and |C|=c|C|=c. Then a5n61a\geq\lceil\frac{5n}{6}\rceil-1.

If |A|3n2|A|\geq\left\lfloor\frac{3n}{2}\right\rfloor, then ar(Pn,Pn)a\geq r(P_{n},P_{n}), and hence there is a monochromatic copy of PnP_{n} in G[A]G[A] colored by 11 or 33. Since B,CB,C are both nonempty sets, it follows that there is a monochromatic copy of K^n\widehat{K}_{n} in GG. Thus, suppose that |A|<3n2|A|<\left\lfloor\frac{3n}{2}\right\rfloor. Then b+cnb+c\geq n and |B|=b(b+c)/2n2|B|=b\geq(b+c)/2\geq\frac{n}{2}. The proof will be given by contradiction, that is, GG does not contain a monochromatic copy of K^n\widehat{K}_{n}. Suppose that H1,H2H_{1},H_{2} are two subgraphs induced by edges in G[B]G[B] and G[C]G[C] with color 22, respectively. Let SS be a star in H2H_{2} such that e(S)e(S) is maximum. Let u0u_{0} be the center of SS. Let PP^{*} be a path in H1H_{1} such that e(P)e(P^{*}) is maximum. Let X=V(S){u0}X=V(S)-\{u_{0}\}, |X|=c1|X|=c_{1} and |P|=b1|P^{*}|=b_{1}. Then 0c1c10\leq c_{1}\leq c-1 and 1b1b1\leq b_{1}\leq b.

Claim 7.

G[B]G[B] and G[C]G[C] contain a monochromatic copies of K1,bb1K_{1,b-b_{1}} and K1,cc11K_{1,c-c_{1}-1} with color 1,31,3, respectively.

Proof.

If b1<bb_{1}<b and 1c1c31\leq c_{1}\leq c-3, then br(Pb1+1,K1,bb1)b\geq r(P_{b_{1}+1},K_{1,b-b_{1}}) by Theorem 2.3, and cr(K1,c1+1,K1,cc11)c\geq r(K_{1,c_{1}+1},K_{1,c-c_{1}-1}) by Theorem 2.4. For c1=0c_{1}=0, since K1,c1+1K_{1,c_{1}+1} is an edge, we also have that cr(K1,c1+1,K1,cc11)c\geq r(K_{1,c_{1}+1},K_{1,c-c_{1}-1}). Since G[B]G[B] does not contain a monochromatic copy of Pb1+1P_{b_{1}+1} and G[C]G[C] does not contain a monochromatic copy of K1,c1+1K_{1,c_{1}+1} with color 22 by the choice of PP^{*} and SS, it follows that G[B]G[B] and G[C]G[C] contain a monochromatic copies of K1,bb1K_{1,b-b_{1}} and K1,cc11K_{1,c-c_{1}-1} with color 1,31,3, respectively. If b1=bb_{1}=b or c2c1c1c-2\leq c_{1}\leq c-1, then G[B]G[B] and G[C]G[C] contain a monochromatic copies of K1,bb1K_{1,b-b_{1}} and K1,cc11K_{1,c-c_{1}-1} with color 1,31,3, respectively. So, G[B]G[B] and G[C]G[C] contain a monochromatic copies of K1,bb1K_{1,b-b_{1}} and K1,cc11K_{1,c-c_{1}-1} with color 1,31,3, respectively. ∎

We use SB,1S^{B,1} and SC,3S^{C,3} to denote monochromatic copies of K1,bb1,K1,cc11K_{1,b-b_{1}},K_{1,c-c_{1}-1} with color 1,31,3, respectively. Moreover, suppose that the center of SB,1S^{B,1} and SC,3S^{C,3} are u1,u2u_{1},u_{2}, respectively.

Claim 8.

bb1<n2b-b_{1}<\lfloor\frac{n}{2}\rfloor and cc11<n2c-c_{1}-1<\lfloor\frac{n}{2}\rfloor. Moreover, if bn2b\geq\frac{n}{2}, then c1<n2c_{1}<\lfloor\frac{n}{2}\rfloor.

Proof.

If bb1n2b-b_{1}\geq\lfloor\frac{n}{2}\rfloor, then since |A|n2|A|\geq\frac{n}{2}, there is a monochromatic PnP_{n} with color 11 between V(S)u1V(S)-u_{1} and AA, and hence GG contains a monochromatic copy of K^n\widehat{K}_{n} with color 11, a contradiction. As the same reason, we have cc11<n2c-c_{1}-1<\lfloor\frac{n}{2}\rfloor; otherwise, there is a monochromatic K^n\widehat{K}_{n} with color 33, a contradiction. Moreover, if bn2b\geq\frac{n}{2}, then c1<n2c_{1}<\lfloor\frac{n}{2}\rfloor; otherwise, there is a monochromatic K^n\widehat{K}_{n} with color 22, a contradiction. ∎

Let a1=n2b+2b1a_{1}=n-2b+2b_{1} and a2=n2c+2c1+2a_{2}=n-2c+2c_{1}+2. Then by Claim 8, a1,a22a_{1},a_{2}\geq 2. We have the following claim.

Claim 9.

There is no monochromatic Pa2P_{a_{2}} with color 33 in G[A]G[A].

Proof.

We first prove that an+cc110a-n+c-c_{1}-1\geq 0. If ana\geq n, then the result follows. If b<n2b<\frac{n}{2}, then b+c1b+c12b1n2b+c_{1}\leq b+c-1\leq 2b-1\leq n-2 and hence

an+cc11\displaystyle a-n+c-c_{1}-1 (5n21b)nc11\displaystyle\geq\left(\left\lfloor\frac{5n}{2}\right\rfloor-1-b\right)-n-c_{1}-1
=3n2(b+c1)2\displaystyle=\left\lfloor\frac{3n}{2}\right\rfloor-(b+c_{1})-2
3n2n0.\displaystyle\geq\left\lfloor\frac{3n}{2}\right\rfloor-n\geq 0.

For the case a<na<n and bn2b\geq\frac{n}{2}, by Claim 8, c1n21c_{1}\leq\lfloor\frac{n}{2}\rfloor-1. Moreover, n2b<n\frac{n}{2}\leq b<n since aba\geq b. Let b=nqb=n-q. Then q1q\geq 1. So,

an+cc11(5n/21b)nc11=n/2+qc120,a-n+c-c_{1}-1\geq(\lfloor 5n/2\rfloor-1-b)-n-c_{1}-1=\lfloor n/2\rfloor+q-c_{1}-2\geq 0,

as desired.

Assume, to the contrary, that G[A]G[A] contains a monochromatic P=Pa2P=P_{a_{2}} with color 33 (say one endpoint of PP is ww). Then

|AV(P)|an+2c2c12=(an+cc11)+(cc11)cc11.|A-V(P)|\geq a-n+2c-2c_{1}-2=(a-n+c-c_{1}-1)+(c-c_{1}-1)\geq c-c_{1}-1.

We can choose a subset AA^{\prime} of AV(P)A-V(P) with |A|=cc11|A^{\prime}|=c-c_{1}-1. Since |A|=|V(SC,3)u2||A^{\prime}|=|V(S^{C,3})-u_{2}|, there is a monochromatic PP^{\prime} of order 2|A|2|A^{\prime}| between AA^{\prime} and V(SC,3)u2V(S^{C,3})-u_{2} (say one endpoint of PP^{\prime} is ww^{\prime}). Then P′′=PP{ww}P^{\prime\prime}=P\cup P^{\prime}\cup\{ww^{\prime}\} is a monochromatic PnP_{n} with color 33, and hence u2P′′u_{2}\vee P^{\prime\prime} a monochromatic K^n\widehat{K}_{n} with color 33, a contradiction. ∎

Case 1 c1+bnc_{1}+b\geq n.

The following claim is useful for the calculation.

Claim 10.

b1+2c1n1b_{1}+2c_{1}\leq n-1.

Proof.

Suppose, to the contrary, that b1+2c1nb_{1}+2c_{1}\geq n. Then n2c1b1n-2c_{1}\leq b_{1}. We can choose a subpath PP^{\prime} of PP^{*} such that |P|=n2c1|P^{\prime}|=n-2c_{1} (say one endpoint of PP^{\prime} is aa). Since c1+bnc_{1}+b\geq n, it follows that |BV(P)|=b(n2c1)c1|B-V(P^{\prime})|=b-(n-2c_{1})\geq c_{1}. Choose a subset YY of BV(P)B-V(P^{\prime}) such that |Y|=|X|=c1|Y|=|X|=c_{1} (recall that XX is the set of leaves of the maximum star SS in H2H_{2}). Then there is a monochromatic path P′′P^{\prime\prime} with color 22 between XX and YY (say xXx\in X is an endpoint of P′′P^{\prime\prime}), and so u0(PP′′{ax})u_{0}\vee(P^{\prime}\cup P^{\prime\prime}\cup\{ax\}) is a monochromatic copy of K^n\widehat{K}_{n} with color 22, a contradiction. ∎

We have proved that G[A]G[A] does not contain a monochromatic copy of Pa2P_{a_{2}} with color 33 in Claim 9. In this case, we have the following claim.

Claim 11.

There is no monochromatic copy of Pa1P_{a_{1}} with color 11 in G[A]G[A].

Proof.

Assume, to the contrary, that G[A]G[A] contains a monochromatic copy of P=Pa1P=P_{a_{1}} with color 11 (say one endpoint of PP is ww). Then |AV(P)|an+2b2b1|A-V(P)|\geq a-n+2b-2b_{1}. We first prove that an+bb10a-n+b-b_{1}\geq 0. Since cc11<n2c-c_{1}-1<\lfloor\frac{n}{2}\rfloor by Claim 8, it follows that cn2+c1c\leq\lfloor\frac{n}{2}\rfloor+c_{1}, and hence

an+bb1\displaystyle a-n+b-b_{1} (5n21c)nb1=3n2(c+b1)1\displaystyle\geq\left(\left\lfloor\frac{5n}{2}\right\rfloor-1-c\right)-n-b_{1}=\left\lfloor\frac{3n}{2}\right\rfloor-(c+b_{1})-1
3n2(n2+c1+b1)1\displaystyle\geq\left\lfloor\frac{3n}{2}\right\rfloor-\left(\left\lfloor\frac{n}{2}\right\rfloor+c_{1}+b_{1}\right)-1
=n[(b1+2c1)c1]1\displaystyle=n-[(b_{1}+2c_{1})-c_{1}]-1
n[(n1)c1]1=c10,\displaystyle\geq n-[(n-1)-c_{1}]-1=c_{1}\geq 0,

as desired. Recall that

|AV(P)|an+2b2b1=(an+bb1)+(bb1)bb1.|A-V(P)|\geq a-n+2b-2b_{1}=(a-n+b-b_{1})+(b-b_{1})\geq b-b_{1}.

We can choose a subset AA^{\prime} of AA such that |A|=bb1|A^{\prime}|=b-b_{1}. Since |A|=|V(SB,1)u1||A^{\prime}|=|V(S^{B,1})-u_{1}|, there is a monochromatic PP^{\prime} of order 2|A|2|A^{\prime}| between AA^{\prime} and V(SB,1)u1V(S^{B,1})-u_{1} (say one endpoint of PP^{\prime} is ww^{\prime}). Then P′′=PP{ww}P^{\prime\prime}=P\cup P^{\prime}\cup\{ww^{\prime}\} is a monochromatic PnP_{n} with color 11, and hence u1P′′u_{1}\vee P^{\prime\prime} a monochromatic K^n\widehat{K}_{n} with color 11, a contradiction. ∎

By Claims 9 and 11, we have |A|=a<r(Pa1,Pa2)|A|=a<r(P_{a_{1}},P_{a_{2}}). However, if a1>a2a_{1}>a_{2}, then

r(Pa1,Pa2)\displaystyle r(P_{a_{1}},P_{a_{2}}) =a1+a221=3n2(b+c)b+2b1+c1\displaystyle=a_{1}+\left\lfloor\frac{a_{2}}{2}\right\rfloor-1=\left\lfloor\frac{3n}{2}\right\rfloor-(b+c)-b+2b_{1}+c_{1}
3n2(5n21a)b+2b1+c1\displaystyle\leq\left\lfloor\frac{3n}{2}\right\rfloor-\left(\left\lfloor\frac{5n}{2}\right\rfloor-1-a\right)-b+2b_{1}+c_{1}
=an(bb1)+(b1+c1)+1.\displaystyle=a-n-(b-b_{1})+(b_{1}+c_{1})+1.

Since bb1b\geq b_{1} and b1+2c1<nb_{1}+2c_{1}<n, we have r(Pa1,Pa2)ar(P_{a_{1}},P_{a_{2}})\leq a, a contradiction. If a1a2a_{1}\leq a_{2}, then

r(Pa1,Pa2)\displaystyle r(P_{a_{1}},P_{a_{2}}) =a2+a121=3n2(b+c)c+2c1+b1+1\displaystyle=a_{2}+\left\lfloor\frac{a_{1}}{2}\right\rfloor-1=\left\lfloor\frac{3n}{2}\right\rfloor-(b+c)-c+2c_{1}+b_{1}+1
3n2(5n21a)c+(2c1+b1)+1\displaystyle\leq\left\lfloor\frac{3n}{2}\right\rfloor-\left(\left\lfloor\frac{5n}{2}\right\rfloor-1-a\right)-c+(2c_{1}+b_{1})+1
(an)c+(n1)+2a,\displaystyle\leq(a-n)-c+(n-1)+2\leq a,

a contradiction.

Case 2 c1+b<nc_{1}+b<n.

By Claim 9, G[A]G[A] does not contain monochromatic Pa2P_{a_{2}} with color 33. Let

a=aa2+1=an+2c2c11.a^{\prime}=a-a_{2}+1=a-n+2c-2c_{1}-1.

Since ar(Pa2,K1,a)a\geq r(P_{a_{2}},K_{1,a^{\prime}}), it follows that G[A]G[A] contains a monochromatic copy of S=K1,aS^{*}=K_{1,a^{\prime}} with color 11. Note that

a(5n21bc)n+2c2c11=3n2b+c2c12.a^{\prime}\geq\left(\left\lfloor\frac{5n}{2}\right\rfloor-1-b-c\right)-n+2c-2c_{1}-1=\left\lfloor\frac{3n}{2}\right\rfloor-b+c-2c_{1}-2.

Since c1+b<nc_{1}+b<n, it follows that an2c1+c1n2a^{\prime}\geq\lfloor\frac{n}{2}\rfloor-c_{1}+c-1\geq\lfloor\frac{n}{2}\rfloor. Let uu^{*} be the center of SS^{*} and UU^{*} be the set of leaves of SS^{*}. Then |U|=an2|U^{*}|=a^{\prime}\geq\lfloor\frac{n}{2}\rfloor. Since bn2b\geq\frac{n}{2}, there is a monochromatic P=PnP=P_{n} with color 11 between AA and UU^{*}. Hence, uPu^{*}\vee P is a monochromatic copy of K^n\widehat{K}_{n} with color 11, a contradiction. ∎

Proof of Theorem 1.10: If n5n\geq 5, then the result follows immediately from Lemmas 5.1 and 5.2.

If n{2,3}n\in\{2,3\}, then let N=5n2N=\lfloor\frac{5n}{2}\rfloor. Let Γ3(N)\Gamma\in\mathcal{B}_{3}(N) be an edge-coloring of G=KNG=K_{N} with parts A1,A2A_{1},A_{2} and |A1||A2||A_{1}|\geq|A_{2}|. Then |A1|4|A_{1}|\geq 4 when n=3n=3 and |A1|=3|A_{1}|=3 when n=2n=2. Suppose that each edge of G[Ai]G[A_{i}] is colored by 11 or i+1i+1, and the edges between A1A_{1} and A2A_{2} are colored by 11. Let HH be the subgraph of G[A]G[A] induced by edges with color 11. For n=3n=3, if Δ(H)2\Delta(H)\geq 2 or Δ(H)=1\Delta(H)=1 and |A2|2|A_{2}|\geq 2, then there is a monochromatic copy of K^3\widehat{K}_{3} with color 11; if Δ(H)=0\Delta(H)=0 or Δ(H)=1\Delta(H)=1 and |A2|=1|A_{2}|=1, then there is a monochromatic copy of K^3\widehat{K}_{3} with color 22. For n=2n=2, if Δ(H)1\Delta(H)\geq 1, then there is a monochromatic copy of K^2\widehat{K}_{2} with color 11; if Δ(H)=0\Delta(H)=0, then there is a monochromatic copy of K^2\widehat{K}_{2} with color 22. Thus, we have that b3(K^n)Nb_{3}(\widehat{K}_{n})\leq N for n{2,3}n\in\{2,3\}. Let Γ1\Gamma_{1} (resp. Γ2\Gamma_{2}) be an edge-coloring of K4K_{4} (resp. K6K_{6}) such that the monochromatic graphs induced by color 1,21,2 and 33 are K2,2K_{2,2}, K2K_{2} and K2K_{2} (resp. K3,3K_{3,3}, K3K_{3} and K3K_{3}), respectively. Then Γ13(4)\Gamma_{1}\in\mathcal{B}_{3}(4) (resp. Γ23(6)\Gamma_{2}\in\mathcal{B}_{3}(6)) but there is not a monochromatic copy of K^2\widehat{K}_{2} (resp. K^3\widehat{K}_{3}). Therefore, we have b3(K^n)=5n2b_{3}(\widehat{K}_{n})=\lfloor\frac{5n}{2}\rfloor for n{2,3}n\in\{2,3\}.

If γ𝒯(N)\gamma\in\mathcal{T}(N) is an edge-coloring of KNK_{N} with the maximum part AA, then |A|n|A|\geq n, and hence G[A]G[A] contains a monochromatic copy of PnP_{n} with color 11 or 33, and so GG contains a monochromatic copy of K^n\widehat{K}_{n}, and hence gr3(K1,3:K^n)=5n2\operatorname{gr}_{3}(K_{1,3}:\widehat{K}_{n})=\lfloor\frac{5n}{2}\rfloor for n{2,3}n\in\{2,3\}.

6 Acknowledgements

We gratefully thank the anonymous reviewers for their careful reading of our manuscript and their valuable comments and suggestions. This paper is supported by the National Natural Science Foundation of China (Nos. 12201375, 12061059)

References

  • [1] J.A. Bondy, U.S.R. Murty, Graph Theory, GTM 244, Springer, 2008.
  • [2] V. Chvátal, F. Harary, Generalized Ramsey theory for graphs. III. Small off-diagonal numbers, Pacific J. Math. 41(2) (1972), 335–345.
  • [3] F.R.K. Chung, R.L. Graham, Edge-colored complete graphs with precisely colored subgraphs, Combinatorica 3 (1983), 315–324.
  • [4] P. Erdős, R.J. McEliece, H. Taylor, Ramsey bounds for graph products, Pacific J. Math. 37(1) (1971), 45–46.
  • [5] R.J. Faudree, R.H. Schelp, Ramsey numbers for all linear forests, Discrete Math. 16 (1976), 149–155.
  • [6] J. Fox, A. Grinshpun, J. Pach, The Erdös-Hajnal conjecture for rainbow triangles, J. Combin. Theory Ser. B 111 (2015), 75–125.
  • [7] S. Fujita, C. Magnant, Extensions of Gallai-Ramsey results, J. Graph Theory 70(4) (2012), 404–426.
  • [8] S. Fujita, C. Magnant, K. Ozeki, Rainbow generalizations of Ramsey theory-A dynamic survey, Theory Appl. Graphs. 0(1), Article 1, (2014)
  • [9] T. Gallai, Transitiv orientierbare Graphen, Acta Math. Acad. Sci. Hungar. 18 (1967), 25–66.
  • [10] L. Gerencsér, A. Gyárfás, On Ramsey-Type Problems, Annales Universitatis Scientiarum Budapestinensis, Eötvös Sect. Math. 10 (1967), 167–170.
  • [11] A. Gyárfás, J. Lehel, R.H. Schelp, Z.S. Tuza, Ramsey numbers for local colorings, Graphs Combin. 3 (1987), 267–277.
  • [12] A. Gyárfás, G.N. Sárközy, A. Sebö, S. Selkow, Ramsey-type results for Gallai colorings, J. Graph Theory, 64 (2010), 233–243.
  • [13] A. Gyárfás, G. Simonyi, Edge colorings of complete graphs without tricolored triangles. J. Graph Theory 46(3) (2004), 211–216.
  • [14] F. Harary, Recent results on generalized Ramsey theory for graphs, Graph Theory and Applications, Springer, Berlin, 125–138 (1972)
  • [15] B. Li, Y. Zhang, H. Broersma, An exact formula for all star-kipas Ramsey numbers, Graphs Combin. 33 (1), 141–148.
  • [16] B. Li, Y. Zhang, H. Broersma, P. Holub, Closing the gap on path-kipas Ramsey numbers, Electron. J. Combin. 22(3) (2015), 3–21.
  • [17] H. Liu, C. Magnant, A. Saito, I. Schiermeyer, Y.T. Shi, Gallai-Ramsey number for K4K_{4}, J. Graph Theory 94 (2020), 192–205.
  • [18] X. Li, L. Wang, X. Liu, Complete graphs and complete bipartite graphs without rainbow path, Discrete Math. 342 (2019), 2116–2126.
  • [19] E. Lucas, Récreations Mathématiqués, Vol. II. Gauthier-Villars, Paris, (1892)
  • [20] C. Magnant, P.S. Nowbandegani, Topics in Gallai-Ramsey theory, Springer Nature, Cham, Switzerland, 2020.
  • [21] C. Magnant, I. Schiermeyer, Gallai-Ramsey number for K5K_{5}, J. Graph Theory 101(3) (2022), 455–492.
  • [22] T.D. Parsons, Path-star Ramsey numbers, J. Combin. Theory, Ser. B 17(1) (1974), 51–58.
  • [23] S.P. Radziszowski, Small Ramsey numbers, Electron. J. Combin., Dynamic Survey, 1–30 (1994)
  • [24] A.N.M. Salman, H.J. Broersma, Path-kipas Ramsey numbers, Discrete Applied Mathematics 155 (2007), 1878–1884
  • [25] A. Thomason, P. Wagner, Complete graphs with no rainbow path. J. Graph Theory 54 (3) (2007), 261–266.
  • [26] M. Wei, C. He, Y. Mao, X. Zhou, Gallai-Ramsey numbers for rainbow P5P_{5} and monochromatic fans or wheels, Discrete Math. 345 (2022), 113092.