The optimal chromatic bound for even-hole-free graphs without induced seven-vertex paths Shenwei Huang, Yidong Zhou, Yeonsu Chang Abstract The class of even-hole-free graphs has been extensively studied on its own and on its relation to perfect graphs. In this paper, we study the χ \chi -boundedness of even-hole-free graphs which itself is an important topic in graph theory. In particular, we prove that every even-hole-free graph G G without induced 7-vertex paths satisfies χ ⁡ ( G ) ≤ ⌈ 5 4 ​ ω ​ ( G ) ⌉ \chi(G)\leq\left\lceil\frac{5}{4}\omega(G)\right\rceil , where χ ⁡ ( G ) \chi(G) and ω ⁡ ( G ) \omega(G) denote the chromatic number and clique number of G G , respectively. This bound is optimal. Our result strictly extends the result of Karthick and Maffary [ 10 ] on even-hole-free graphs without induced 6-vertex paths, and implies that even-hole-free graphs without induced 7-vertex paths satisfy Reed’s Conjecture. Our proof relies on a heavy structural analysis on a maximal substructure called a nice blowup of a five-cycle and can be viewed for graphs in which all holes are of length five (graphs with all holes having the same length gain increasing interest in recent years [ 6 ] ). Our result gives a partial answer to a conjecture of Wang and Wu [ 22 ] on graphs in which all holes are of length 5. One of the key technical ingredients is a technical lemma proved via clique cutset argument combined with the idea of Infinite Descent Method (often used in number theory). 1 Introduction For a list of graphs (H_{1},\ldots,H_{k}) , we say that a graph G is (H_{1},\ldots,H_{k}) -free if G has no graph H_{i} as an induced subgraph for each 1\leq i\leq k . Let \omega(G) denote the clique number of a graph G . For a positive integer k , a proper k -coloring of a graph G is a function f\mathrel{\mathop{\mathchar 58\relax}}V(G)\rightarrow[k] such that f(u)\neq f(v) for every edge uv\in E(G) . In such a case, we say that G is k -colorable . The chromatic number of G , denoted by \chi(G) , is the minimum positive integer k such that G is k -colorable. For a graph class \mathcal{G} , we say that \mathcal{G} is \chi -bounded if there exists a function f\mathrel{\mathop{\mathchar 58\relax}}\mathbb{N}\rightarrow\mathbb{N} such that every graph G\in\mathcal{G} has \chi(G)\leq f(\omega(G)) . The function f is called a \chi -bounding function for \mathcal{G} . The study of \chi -bounding functions for graph classes is an important area in graph theory in the past two decades, see for example the survey of Scott and Seymour [ 19 ] and Scott’s survey [ 16 ] in Proceedings of International Congress Mathematics 2022. The study can be roughly divided into two lines: graphs with a single forbidden induced subgraph and graphs with forbidden holes. The research on \chi -boundedness of graphs with a single forbidden induced subgraph is largely surrounded by Gyárfás-Sumner Conjecture. The conjecture is notoriously hard despite some partial progress (see [ 18 ] for instance). In this paper, we focus on graphs with forbidden holes. A hole in G is an induced cycle of order at least four. Scott and Seymour [ 17 ] proved that the class of odd-hole-free graphs, graphs with no holes of odd length, is \chi -bounded with f=2^{2^{n}} being a \chi -bounding function. It is still open whether this double-exponential bound can be improved to a single-exponential bound. In sharp contrast, the class of even-hole-free graphs, graphs with no holes of even length, admits a linear \chi -bounding function f(n)=2n-1 . This follows from a stronger result by Chudnovsky and Seymour [ 4 ] that every even-hole-free graph has a vertex whose neighborhood can be covered by two cliques. It is not known whether this bound is optimal. In fact, several known results on subclasses of even-hole-free graphs suggest that this bound may not be optimal. (Pan) A pan is a hole with a pendent edge. It is shown in [ 1 ] that every (pan, even-hole)-free graph G has \chi(G)\leq\left\lceil\frac{3}{2}\omega(G)\right\rceil . (Cap) A cap is a hole with an additional vertex adjacent to an edge on the hole. It is shown in [ 2 ] that every (cap, even-hole)-free graph G has \chi(G)\leq\left\lceil\frac{5}{4}\omega(G)\right\rceil . ( P_{6} ) The path on t vertices is denoted by P_{t} . It is shown in [ 10 ] that every ( P_{6} , even-hole)-free graph G has \chi(G)\leq\left\lceil\frac{5}{4}\omega(G)\right\rceil . (Diamond) A diamond is obtained from the complete graph on four vertices by removing an edge. It is shown in [ 11 ] that every (diamond, even-hole)-free graph G has \chi(G)\leq\omega(G)+1 . For more background and results on even-hole-free graphs, we refer the readers to a comprehensive survey by Vušković [ 21 ] . The main result of this paper is the following optimal \chi -bounding function for the class of even-hole-free graphs without an induced 7-vertex path. Theorem 1.1 . For every ( P_{7} , even-hole)-free graph G , \chi(G)\leq\left\lceil\frac{5}{4}\omega(G)\right\rceil . Remark. This bound is optimal as seen by equal-size blowups of a five-cycle (the precise definition of equal-size blowups is given in Section 2 ). Reed’s Conjecture [ 15 ] states that every graph G satisfies \chi(G)\leq\left\lceil\frac{\Delta(G)+\omega(G)+1}{2}\right\rceil where \Delta(G) is the maximum degree of G . Our result implies by a result in [ 9 ] that the class of ( P_{7} , even-hole)-free graphs satisfies Reed’s Conjecture. For an integer \ell\geq 4 , a graph is \ell -holed if all holes in this graph have length \ell . Such graphs were studied [ 23 ] in the context of algebraic combinatorics and commutative algebra. A structure theorem for \ell -holed graphs when \ell\geq 7 is proved in [ 6 ] . Using this structure theorem, Wang and Wu [ 22 ] showed that f(n)=\left\lceil\frac{\ell}{\ell-1}n\right\rceil is the optimal \chi -bounding function for the class of \ell -holed graphs when \ell\geq 7 is odd. They further conjecture that this bound also holds for \ell=5 . Our result is essentially for the class of 5-holed graphs without an induced P_{7} , and proves their conjecture for this graph class. One difficulty in our proof is that there are no structure theorems for 5-holed graphs at hand and we need to do heavy structural analysis along the way. Proof sketch. Since the proof is long, let us sketch the proof here. For integer s\geq 3 , we denote by C_{s} the cycle on s vertices. Let G be a (P_{7},C_{4},C_{6}) -free (which is the same as ( P_{7} , even-hole)-free) graph. The overall strategy is to consider two cases depending on whether G contains an induced C_{7} . In the case that G contains an induced C_{7} , we show, based on a structural result of Penev [ 14 ] , that Theorem 1.1 holds (this case is easy). Afterwards, we can assume that G is C_{7} -free. In this case, we can assume that G contains an induced C_{5} (this case is difficult and most of our proofs are devoted to this case) using a classical result on chordal graphs by Dirac (Lemma 2.4 ). Then we define and analyze the so-called nice blowup H of an induced C_{5} that mimics the structure of an induced C_{5} . We then partition V(G)\setminus V(H) into subsets A_{0},A_{1},A_{2},A_{3},A_{5} in terms of their neighborhoods on H . After proving structural properties of those subsets, we show that Theorem 1.1 holds when A_{1}\neq\emptyset (Section 11 ), A_{1}=\emptyset and A_{2}\neq\emptyset (Section 12 ), and A_{1}=A_{2}=\emptyset but A_{3}\neq\emptyset (Section 13 ). These three sections together complete the proof of Theorem 1.1 . Technical highlights. To carry out this strategy, we use technical arguments based on several reducible structures in the study of \chi -bounding functions-clique cutsets, small vertices (Lemma 2.11 ) and (p,q) -good subgraphs (Lemma 2.10 ). It is easy to show that it suffices to prove Theorem 1.1 for graphs G without those structures (Lemma 2.12 ). The first argument is called clique cutset argument , which aims to show that the neighborhood of certain subsets is a clique. In some cases, a direct analysis does not work and we need to combine new technical ideas such as nice partitions defined in Section 10 . The idea is similar to Infinite Descent Method in number theory: if the desired property does not hold, one can first show the existence of a nice partition and then argue that a given nice partition can be made “smaller” in the sense defined in Section 10 , as long as G has no clique cutsets. Since the process cannot continue forever, G contains a clique cutset, which is a contradiction. The second argument is called small vertex argument . The idea is to show that the neighborhood of a vertex v in V(G)\setminus K is a clique where K is a clique of G . This implies that K is “large”, meaning that |K|>\left\lceil\frac{\omega(G)}{4}\right\rceil . Applying this argument to (minimal) simplicial vertices allows us to conclude in many cases that there are four pairwise disjoint “large” cliques whose union is still a clique. This gives a clique of size larger than \omega(G) , which is a contradiction. The good subgraph argument is to find a (p,q) -good subgraph. The (p,q) -good subgraphs were already implicit in [ 10 ] . We explicitly formulate this notion and use (4,5) -good subgraphs to prove our main theorem. This argument is technically involved and used in Sections 3 and 13 . The final argument is called pre-coloring argument . For graphs G that cannot be handled by the previous three arguments, we need to give an explicit proper coloring using \left\lceil\frac{5}{4}\omega(G)\right\rceil colors. The idea is to color a subgraph of G first and then color the rest of the graph by reducing the problem (based on structural insight) to color hyperholes (Lemma 2.1 ). To be able to apply Lemma 2.1 , we need to carefully choose the subgraph to be pre-colored so that the calculations can go through. Paper organization. The paper is organized as follows. In Section 2 , we define terminology and notations used in this paper and prove some general lemmas. In Section 3 , we handle the case that G contains an induced C_{7} . In Section 4 , we define nice blowups of C_{5} and their attachments A_{0},A_{1},A_{2},A_{3},A_{5} and prove some basic properties about those sets. In Section 5 , we show that A_{0}=\emptyset . In Sections 6 - 8 , we obtain further properties of A_{1},A_{2},A_{3} . In Section 9 , we prove a useful lemma about components of non-neighbors of a C_{5} . In Section 10 , we prove a key technical lemma. In Section 11 , we prove Theorem 1.1 when A_{1} is not empty. In Section 12 , we prove Theorem 1.1 when A_{1} is empty and A_{2} is not empty. In Section 13 , we prove Theorem 1.1 when A_{1},A_{2} are empty and A_{3} is not empty. In Section 14 , we give some concluding remarks. 2 Preliminaries All graphs in this paper are simple and finite. Let G be a graph. We denote by V(G) and E(G) the vertex set and the edge set of G , respectively. For an edge \{u,v\} of G , we shortly say uv . For S\subseteq V(G) , let G[S] denote the subgraph of G induced by S , and let G-S=G[V(G)\setminus S] . For convenience, we identify a subset S\subseteq V(G) as G[S] . A graph H is an induced subgraph of G if there exists a subset S\subseteq V(G) such that G[S] is isomorphic to H . For two disjoint vertex sets S_{1} and S_{2} in G , S_{1} is complete to S_{2} if every vertex in S_{1} is adjacent to every vertex in S_{2} , and S_{1} is anticomplete to S_{2} if there are no edges between S_{1} and S_{2} . If S_{1}=\{v\} , we simply say that v is complete or anticomplete to S_{2} instead of \{v\} is complete or anticomplete to S_{2} . A vertex v mixes on a subset S if v is neither complete nor anticomplete to S . We say v mixes on an edge xy if v mixes on \{x,y\} . Let v be a vertex of G and H be a subgraph of V(G) . The neighbors of v in G that are contained in H is denoted by N_{H}(v) . So N_{G}(v) is the set of all neighbors of v in G . The degree of v in G , denoted by d_{G}(v) , is the number of neighbors of v in G . Let N_{G}[v]=N_{G}(v)\cup\{v\} . For a non-empty vertex subset S of G , we define N_{H}(S)=\left(\bigcup_{v\in S}N_{H}(v)\right)\setminus S . If the context is clear, we shall omit the subscript G , writing N_{G}(v) as N(v) and so on. We say that v\in V(H) is universal in H if N_{H}[v]=V(H) . A set S\subseteq V(H) is universal in H if each vertex in S is universal in H . For a positive integer n , let [n] denote the set of positive integers at most n . For two integers i,j\in[n] , the operations + and - are the same as in the finite cyclic group \mathbb{Z}_{n} . For a positive integer n , let P_{n} denote the path on n vertices where V(P_{n})=\{v_{1},\ldots,v_{n}\} and E(P_{n})=\{v_{i}v_{i+1}\mathrel{\mathop{\mathchar 58\relax}}i\in[n-1]\} . The internal of P is V(P)\setminus\{v_{1},v_{n}\} . Let C_{n} denote the cycle on n vertices where V(C_{n})=\{v_{1},\ldots,v_{n}\} and E(C_{n})=\{v_{i}v_{i+1}\mathrel{\mathop{\mathchar 58\relax}}i\in[n]\} . For clarity, we write P_{n} as v_{1}-v_{2}-\cdots-v_{n} and write C_{n} as v_{1}-v_{2}-\cdots-v_{n}-v_{1} . For v_{i} and v_{j} in P_{n} with i\left\lceil\frac{\omega}{4}\right\rceil and |L_{i}|\geq 3 . Proof. By symmetry, it suffices to prove the lemma for i=2,3,4,5 . (L_{2}) Let v\in L_{1} . Then N(v)\setminus L_{2}\subseteq L_{1}\cup L_{7}\cup L_{8}\cup L_{9} is a clique. Since G has no small vertices, |L_{2}|\geq|N(v)\cap L_{2}|>\left\lceil\frac{\omega}{4}\right\rceil . (L_{3}) Let v\in L_{2} . Then N(v)\setminus L_{3}\subseteq L_{1}\cup L_{8}\cup L_{9} is a clique. Since G has no small vertices, |L_{3}|\geq|N(v)\cap L_{3}|>\left\lceil\frac{\omega}{4}\right\rceil . (L_{4}) Let v\in L_{5} . Then N(v)\setminus L_{4}\subseteq L_{6}\cup L_{10} is a clique. Since G has no small vertices, |L_{4}|\geq|N(v)\cap L_{4}|>\left\lceil\frac{\omega}{4}\right\rceil . (L_{5}) Let v\in L_{4} . Then N(v)\setminus L_{5}\subseteq L_{3}\cup L_{9}\cup L_{10}\cup L_{12} is a clique. Since G has no small vertices, |L_{5}|\geq|N(v)\cap L_{5}|>\left\lceil\frac{\omega}{4}\right\rceil Since \omega\geq 2 , |L_{i}|\geq 2 for i\in[7] . Since L_{j}\neq\emptyset for some j\in\{8,9,10,11,12\} , \omega\geq 5 and so |L_{i}|\geq 3 for i\in[7] . ∎ We show that G is not basic by finding a (4,5) -good subgraph, which contradicts our assumption on G . By Lemma 3.2 , one can choose three vertices from each L_{i} for i\in[7] when constructing (4,5) -good subgraphs. We first give two lemmas (Lemmas 3.3 and 3.4 ) and then consider all possible cases depending on the number of non-empty L_{j} s for j\in\{8,9,10\} (Lemmas 3.5 - 3.8 ). Lemma 3.3 . If L_{10}=L_{11}=L_{12}=\emptyset , then G has a (4,5) -good subgraph. Proof. Suppose not. Choose two vertices in L_{i} for each i\in[7] and let H be the union of all those chosen vertices. By Lemma 2.1 , \chi(H)\leq 5 . It is easy to see that every maximal clique of G contains 4 vertices of H . Therefore, H is a (4,5) -good subgraph. This contradicts that G is basic. ∎ Lemma 3.4 . If L_{10}=\emptyset and L_{11},L_{12}\neq\emptyset , G has a (4,5) -good subgraph. Proof. If not, we give a (4,5) -good subgraph with a 5-coloring as follows (see Figure 2 ). 1,2 L_{1} 3,4 L_{2} 1,5 L_{3} 3 L_{4} 2,4,5 L_{5} 1 L_{6} 3,5 L_{7} L_{8} L_{9} 4 L_{11} 2 L_{12} Figure 2 : A (4,5) -good subgraph in Lemma 3.4 . Choose two vertices in L_{1} and color them with colors 1 and 2, two vertices in L_{2} and color them with colors 3 and 4, two vertices in L_{3} and color them with colors 1 and 5, one vertex in L_{4} and color them with color 3, three vertices in L_{5} and color them with colors 2, 4 and 5, one vertex in L_{6} and color them with color 1, two vertices in L_{7} and color them with colors 3 and 5, one vertex in L_{11} and color it with color 4, one vertex in L_{12} and color it with color 2. ∎ Lemma 3.5 . If L_{8},L_{9},L_{10}\neq\emptyset , then G has a (4,5) -good subgraph. Proof. We give a (4,5) -good subgraph with a 5-coloring as follows (see Figure 3 ). 1 L_{1} 2 L_{2} 3 L_{3} 4 L_{4} 2,3 L_{5} 5 L_{6} 2 L_{7} 4 L_{8} 5 L_{9} 1 L_{10} L_{11} L_{12} Figure 3 : A (4,5) -good subgraph in Lemma 3.5 . Choose one vertex in L_{1} and color it with color 1, one vertex in L_{2} and color it with color 2, one vertex in L_{3} and color it with color 3, one vertex in L_{4} and color it with color 4, two vertices in L_{5} and color them with colors 2 and 3, one vertex in L_{6} and color it with color 5, one vertex in L_{7} and color it with color 2, one vertex in L_{8} and color it with color 4, one vertex in L_{9} and color it with color 5, one vertex in L_{10} and color it with color 1. ∎ Lemma 3.6 . If exactly two of L_{8} , L_{9} and L_{10} are not empty, then G has a (4,5) -good subgraph. Proof. Suppose first that L_{8},L_{10}\neq\emptyset (the case L_{9},L_{10}\neq\emptyset is symmetric). We give a (4,5) -good subgraph with a 5-coloring as follows (see Figure 4(a) ). Choose two vertices in L_{1} and color them with colors 1 and 2, one vertex in L_{2} and color it with color 3, two vertices in L_{3} and color them with colors 2 and 5, one vertex in L_{4} and color it with color 4, two vertices in L_{5} and color them with colors 3 and 5, one vertex in L_{6} and color it with color 2, two vertices in L_{7} and color them with colors 3 and 5, one vertex in L_{8} and color it with color 4, one vertex in L_{10} and color it with color 1. Suppose now that L_{8},L_{9}\neq\emptyset and so L_{10}=\emptyset . By Lemma 3.3 , we may assume by symmetry that L_{12}\neq\emptyset . We give a (4,5) -good subgraph with a 5-coloring as follows (see Figure 4(b) . Choose one vertex in L_{1} and color it with color 1, one vertex in L_{2} and color it with color 2, one vertex in L_{3} and color it with color 3, one vertex in L_{4} and color it with color 4, three vertices in L_{5} and color them with colors 1, 2 and 3, one vertex in L_{6} and color it with color 5, two vertices in L_{7} and color them with colors 2 and 3, one vertex in L_{8} and color it with color 4, one vertex in L_{9} and color it with color 5, one vertex in L_{12} and color it with color 1. ∎ Lemma 3.7 . If exactly one of L_{8} , L_{9} and L_{10} is not empty, then G has a (4,5) -good subgraph. Proof. Suppose first that L_{9}\neq\emptyset (the case that L_{8}\neq\emptyset is symmetric). By Lemmas 3.3 and 3.4 , exactly one of L_{11} and L_{12} is not empty. If L_{11}\neq\emptyset , we give a (4,5) -good subgraph with a 5-coloring as follows (see Figure 4(c) ). Choose two vertices in L_{1} and color them with colors 1 and 2, one vertex in L_{2} and color it with color 3, two vertices in L_{3} and color them with colors 2 and 5, one vertex in L_{4} and color it with color 3, three vertices in L_{5} and color them with colors 1, 2 and 5, one vertex in L_{6} and color it with color 4, two vertices in L_{7} and color them with colors 3 and 5, one vertex in L_{9} and color it with color 4, one vertex in L_{11} and color it with color 1. If L_{12}\neq\emptyset , we give a (4,5) -good subgraph with a 5-coloring as follows (see Figure 4(d) ). Choose one vertex in L_{1} and color it with color 1, two vertices in L_{2} and color them with colors 2 and 3, one vertex in L_{3} and color it with color 5, one vertex in L_{4} and color it with color 2, three vertices in L_{5} and color them with colors 1, 3 and 5, two vertices in L_{6} and color them with colors 2 and 4, two vertices in L_{7} and color them with colors 3 and 5, one vertex in L_{9} and color it with color 4, one vertex in L_{12} and color it with color 1. Suppose now that L_{10}\neq\emptyset . By Lemma 3.3 and by symmetry, L_{11}\neq\emptyset or L_{12}\neq\emptyset . By symmetry, we assume that L_{11}\neq\emptyset . We give a (4,5) -good subgraph with a 5-coloring as follows (see Figure 4(e) ). Choose two vertices in L_{1} and color them with colors 1 and 2, two vertices in L_{2} and color them with colors 3 and 5, two vertices in L_{3} and color them with colors 1 and 5, one vertex in L_{4} and color it with color 4, three vertices in L_{5} and color them with colors 1, 2 and 5, two vertices in L_{7} and color them with colors 4 and 5, one vertex in L_{10} and color it with color 3, one vertex in L_{11} and color it with color 2. ∎ Lemma 3.8 . If L_{8}=L_{9}=L_{10}=\emptyset , then G has a (4,5) -good subgraph. Proof. If L_{11} and L_{12} are not empty, we are done by Lemma 3.4 . By symmetry, we assume that L_{12}\neq\emptyset and L_{11}=\emptyset . We give a (4,5) -good subgraph with a 5-coloring as follows (see Figure 4(f) ). Choose one vertex in L_{1} and color it with color 4, three vertices in L_{2} and color them with colors 1, 2 and 5, one vertex in L_{3} and color it with color 3, two vertices in L_{4} and color them with colors 2 and 4, three vertices in L_{5} and color them with colors 1, 3 and 5, one vertex in L_{6} and color it with color 4, three vertices in L_{7} and color them with colors 1, 2 and 3, one vertex in L_{12} and color it with color 5. ∎ This completes the proof for special blowups of M . Non-empty blowups of M_{1} and M_{2} can be handled in a similar way. We leave the routine work to the readers. 1,2 L_{1} 3 L_{2} 2,5 L_{3} 4 L_{4} 3,5 L_{5} 2 L_{6} 3,5 L_{7} 4 L_{8} 1 L_{10} L_{11} L_{12} (a) A (4,5) -good subgraph in Lemma 3.6 . 1 L_{1} 2 L_{2} 3 L_{3} 4 L_{4} 1,2,3 L_{5} 5 L_{6} 2,3 L_{7} 4 L_{8} 5 L_{9} L_{11} 1 L_{12} (b) A (4,5) -good subgraph in Lemma 3.6 . 1,2 L_{1} 3 L_{2} 2,5 L_{3} 3 L_{4} 1,2,5 L_{5} 4 L_{6} 3,5 L_{7} 4 L_{9} 1 L_{11} (c) A (4,5) -good subgraph in Lemma 3.7 . 1 L_{1} 2,3 L_{2} 5 L_{3} 2 L_{4} 1,3,5 L_{5} 2,4 L_{6} 3,5 L_{7} 4 L_{9} 1 L_{12} (d) A (4,5) -good subgraph in Lemma 3.7 . 1,2 L_{1} 3,5 L_{2} 1,5 L_{3} 4 L_{4} 1,2,5 L_{5} L_{6} 4,5 L_{7} 3 L_{10} 2 L_{11} L_{12} (e) A (4,5) -good subgraph in Lemma 3.7 . 4 L_{1} 1,2,5 L_{2} 3 L_{3} 2,4 L_{4} 1,3,5 L_{5} 4 L_{6} 1,2,3 L_{7} 5 L_{12} (f) A (4,5) -good subgraph in Lemma 3.8 . Figure 4 : (4,5) -good subgraphs in Lemmas 3.6 - 3.8 . 4 Nice blowups of C_{5} From now on, we may assume that our graph G is (P_{7},C_{4},C_{6},C_{7}) -free. If G is C_{5} -free, then G is chordal and so \chi(G)=\omega(G) . So we assume that G contains an induced C_{5} . The indices below are modulo 5. A tuple (B_{1},B_{2},B_{3},B_{4},B_{5}) is called a nice blowup of C_{5} in a graph G if • each B_{i} is a clique in G , and distinct B_{i} and B_{j} are disjoint, • for every i\in[5] and v\in B_{i} , v has a neighbor in B_{i-1} and a neighbor in B_{i+1} , • for every i\in[5] , B_{i} is anticomplete to B_{i+2} , • for every i\in[5] , a\in B_{i} , b,c\in B_{i+1} , d\in B_{i+2} with b\neq c , G[\{a,b,c,d\}] is not isomorphic to P_{4} . We call the last condition of nice blowups the P_{4} -structure condition. Observe that a nice blowup of C_{5} is (P_{7},C_{4},C_{6}) -free. Next we give some basic properties of nice blowups of C_{5} . Lemma 4.1 . Let G be a (P_{7},C_{4},C_{6}) -free graph, and let H=(B_{1},B_{2},B_{3},B_{4},B_{5}) be a nice blowup of C_{5} in G . (1) One can order vertices in B_{i} as b_{i}^{1},b_{i}^{2},\ldots,b_{i}^{|B_{i}|} so that N_{B_{i-1}\cup B_{i+1}}(b_{i}^{1})\subseteq N_{B_{i-1}\cup B_{i+1}}(b_{i}^{2})\subseteq\cdots\subseteq N_{B_{i-1}\cup B_{i+1}}(b_{i}^{|B_{i}|}) . (2) For each i\in[5] , there is a vertex in B_{i} complete to B_{i-1}\cup B_{i+1} . In particular, each vertex in B_{i-1} and each vertex in B_{i+1} have a common neighbor in B_{i} . Proof. We prove the properties one by one. (1) It suffices to show that for every two distinct vertices x,y\in B_{i} , we have N_{B_{i-1}\cup B_{i+1}}(x)\subseteq N_{B_{i-1}\cup B_{i+1}}(y) or N_{B_{i-1}\cup B_{i+1}}(y)\subseteq N_{B_{i-1}\cup B_{i+1}}(x) . By Lemma 2.2 , we may assume that N_{B_{i-1}}(x)\subseteq N_{B_{i-1}}(y) . If N_{B_{i-1}}(x)=N_{B_{i-1}}(y) , we are done by Lemma 2.2 . So we may assume that exists a vertex z\in B_{i-1} that is adjacent to y but not adjacent to x . Then the P_{4} -structure condition implies that N_{B_{i+1}}(x)\subseteq N_{B_{i+1}}(y) . So N_{B_{i-1}\cup B_{i+1}}(x)\subseteq N_{B_{i-1}\cup B_{i+1}}(y) . (2) If b_{i}^{|B_{i}|} is not adjacent to a vertex v\in B_{j} for some j\in\{i-1,i+1\} , then v is anticomplete to B_{i} by (2). This contradicts the definition of nice blowup of C_{5} . So b_{i}^{|B_{i}|} is a desired vertex. ∎ Let H=(B_{1},B_{2},B_{3},B_{4},B_{5}) be a nice blowup of C_{5} in a graph G . For each i\in[5] , we fix a vertex q_{i}\in B_{i} that is complete to B_{i-1}\cup B_{i+1} . We point out that nice blowups of C_{5} coincide with 5-rings defined in [ 12 ] . It was proved in [ 12 ] (Lemma 6.7) that any nice blowup H of C_{5} has \chi(H)\leq\left\lceil\frac{5}{4}\omega(H)\right\rceil . So we can assume that V(G)\setminus V(H) is not empty. For every vertex v\in V(G)\setminus V(H) , we write \operatorname{supp}(v)=\{i\in[5]\mathrel{\mathop{\mathchar 58\relax}}B_{i}\cap N_{G}(v)\neq\emptyset\} . Next, we investigate how the vertices outside H attach to H . Lemma 4.2 . For v\in V(G)\setminus V(H) , \operatorname{supp}(v) consists of consecutive integers modulo 5 that has size in \{0,1,2,3,5\} . Proof. Suppose that \operatorname{supp}(v) does not consist of consecutive integers modulo 5 . Then there exists i\in[5] where N_{B_{i}}(v) and N_{B_{i+2}}(v) are non-empty and N_{B_{i+1}}(v) is empty. Let u\in N_{B_{i}}(v) and w\in N_{B_{i+2}}(v) . Then v-u-q_{i+1}-w-v is an induced C_{4} , a contradiction. By the same argument, \operatorname{supp}(v) cannot have size 4 . ∎ Definition 4.3 (Attachment of H ) . We divide the vertices in V(G)\setminus V(H) as follows. Let i\in[5] . • A_{0} is the set of vertices v in V(G)\setminus V(H) where \operatorname{supp}(v)=\emptyset . • A_{1}(i) is the set of vertices v in V(G)\setminus V(H) where \operatorname{supp}(v)=\{i\} . • A_{2}(i) is the set of vertices v in V(G)\setminus V(H) where \operatorname{supp}(v)=\{i,i+1\} . • A_{3}(i) is the set of vertices v in V(G)\setminus V(H) where \operatorname{supp}(v)=\{i-1,i,i+1\} . • A_{5} is the set of vertices v in V(G)\setminus V(H) where \operatorname{supp}(v)=[5] . For each j\in\{1,2,3\} , let A_{j}=\bigcup_{i\in[5]}A_{j}(i) . By Lemma 4.2 , the union of above sets is V(G)\setminus V(H) . For convenience, we write A_{x}(i\pm y) for the union of A_{x}(i+y) and A_{x}(i-y) . Let A^{\prime}_{3}(i)=A_{3}(i)\cup B_{i} for i\in[5] . Let G be a graph and P=v_{1}-v_{2}-\cdots-v_{7} be a 7-vertex path (not necessarily induced) in G . We say that P is bad if v_{1}-v_{2}-\cdots-v_{6} is an induced P_{6} and v_{3}-v_{4}-\cdots-v_{7} is an induced P_{5} . The following observation encodes (P_{7},C_{6},C_{7}) -freeness by bad P_{7} s. Lemma 4.4 . A (P_{7},C_{6},C_{7}) -free graph does not have a bad P_{7} . Proof. Suppose that P=v_{1}-v_{2}-\cdots-v_{7} is a bad P_{7} . If v_{7}v_{2}\in E(G) , then v_{2}-v_{3}-v_{4}-v_{5}-v_{6}-v_{7}-v_{2} is an induced C_{6} . So we may assume that v_{7}v_{2}\notin E(G) . Now V(P) induces either a C_{7} or a P_{7} depending on whether v_{7}v_{1}\in E(G) . ∎ Next we define the notions of induced sequences and induced cyclic sequences which will be useful for finding forbidden induced paths and cycles. Definition 4.5 (Induced sequences and induced cyclic sequences) . Let A_{1},\ldots,A_{m} be disjoint sets of vertices in G where m\in\mathbb{N} . We say that a sequence A_{1}-A_{2}-\cdots-A_{m} is an induced sequence if • for all i\in[m] , A_{i} is connected, • for all i\in[m-1] , there is an edge between A_{i} and A_{i+1} , and • for i,j\in[m] with i\left\lceil\frac{\omega}{4}\right\rceil . Proof. Suppose not. By symmetry, we may assume that there are two non-adjacent vertices s,s^{\prime} in N(v)\setminus B_{i-1} . Note that N(v)\setminus B_{i-1}\subseteq B_{i+1}\cup B_{i}\cup A_{3}(i)\cup A_{5} . By Lemmas 4.14 , 8.1 (2) and 8.3 , we may assume s\in A_{3}(i) and s^{\prime}\in B_{i+1} . Suppose first that s is simplicial in A_{3}(i) . By the choice of v , s has a neighbor s^{\prime\prime}\in B_{i-1}\cup B_{i+1} with s^{\prime\prime}v\notin E(G) . This contradicts Lemmas 2.2 and 4.11 . So s is not simplicial in A_{3}(i) . Then s has a neighbor s^{\prime\prime}\in A_{3}(i) with s^{\prime\prime}v\notin E(G) . Now s^{\prime\prime}-s-v is an induced P_{3} in A_{3}(i) , which contradicts Lemma 8.4 . ∎ Lemma 8.15 (Small vertex argument for non-neighbors of A_{3}(i) in B_{i} ) . Suppose that A_{3}(i)\neq\emptyset and B_{i} is anticomplete to A_{1}\cup A_{2} . If w\in B_{i}\setminus N(A_{3}(i)) has minimal neighborhood in H , then N(w)\setminus B_{i-1} and N(w)\setminus B_{i+1} are cliques, and so |N_{B_{i-1}}(w)|,|N_{B_{i+1}}(w)|>\left\lceil\frac{\omega}{4}\right\rceil . Proof. Suppose for a contradiction that N(w)\setminus B_{i+1} is not a clique. By Lemma 8.11 and A_{3}(i)\neq\emptyset , N(w)\setminus B_{i+1}\subseteq B_{i-1}\cup B_{i}\cup A_{5} . Let s\in B_{i} and s^{\prime}\in B_{i-1} be two non-adjacent neighbors of w . By Lemma 8.1 (3), s\in B_{i}\setminus N(A_{3}(i)) . By the minimality of w , s has a neighbor s^{\prime\prime}\in B_{i-1}\cup B_{i+1} not adjacent to w . It follows that s^{\prime\prime}-s-w-s^{\prime} is a P_{4} -structure if s^{\prime\prime}\in B_{i+1} or s^{\prime\prime}-s-w-s^{\prime}-s^{\prime\prime} is an induced C_{4} if s^{\prime\prime}\in B_{i-1} . ∎ 9 Non-neighbors of C_{5} For an induced C_{5}=v_{1}-v_{2}-v_{3}-v_{4}-v_{5}-v_{1} and X\subseteq V(C) , we denote by S(X) the set of vertices v\in V(G)\setminus V(C) with N_{C}(v)=X . We set S_{0}=S(\emptyset) and S_{5}=S(V(C)) . For convenience, we omit the set notation \{\} when we write X in S(X) . For instance, we write S(\{v_{1},v_{2}\}) as S(v_{1},v_{2}) etc. The following lemma will be used to prove certain subgraphs are chordal in Section 11 . Lemma 9.1 . Let C=v_{1}-v_{2}-v_{3}-v_{4}-v_{5}-v_{1} be an induced C_{5} of G and K be a non-clique component of S_{0} . Then the following holds for neighbors of K . (1) Every vertex in N(K)\cap S(v_{i-2},v_{i+2}) is complete to K . (2) N(K)\cap S(v_{i-1},v_{i},v_{i+1}) or N(K)\cap S(v_{i+1},v_{i+2},v_{i+3}) is empty. (3) If p,q\in N(K) are not adjacent, then p\in S(v_{i-1},v_{i},v_{i+1}) and q\in S(v_{i-2},v_{i+2}) for some i\in\mathbb{Z}_{5} . (4) N(K) cannot contain 4 neighbors p,p^{\prime},q,q^{\prime} such that p\in S(v_{i-1},v_{i},v_{i+1}) , q\in S(v_{i-2},v_{i+2}) , p^{\prime}\in S(v_{i},v_{i+1},v_{i+2}) and q^{\prime}\in S(v_{i-2},v_{i-1}) . Proof. We prove the properties one by one. (1) Let c\in N(K)\cap S(v_{i-2},v_{i+2}) be adjacent to b\in K . If c has a non-neighbor a\in K , then a-b-c-v_{i-2}-v_{i-1}-v_{i}-v_{i+1} is an induced P_{7} . This proves (1). (2) Suppose for a contradiction that a\in N(K)\cap S(v_{i-1},v_{i},v_{i+1}) and b\in N(K)\cap S(v_{i+1},v_{i+2},v_{i+3}) . Since a-v_{i-1}-v_{i-2}-b-a is not an induced C_{4} , ab\notin E(G) . Let P be a shortest path between a and b with internal in K . Since G is C_{4} -free, P is an induced path on at least 4 vertices. Then a-v_{i-1}-v_{i-2}-b-P-a is an induced cycle of length at least 6. This proves (2). (3) Since G has no clique cutsets, N(K) contains two non-adjacent vertices p and q . Note first that if a vertex a\in N(K) has at most 2 neighbors on C , then a is complete to K for otherwise there is an induced P_{7} . Since G is C_{4} -free and K is not a clique, we may assume that p\in S(v_{i-1},v_{i},v_{i+1})\cup S_{5} . Suppose first that p\in S_{5} . Since S_{5} is a clique complete to S(v_{j-1},v_{j},v_{j+1}) for each j\in[5] , we have q has at two neighbors on C . It follows that p and q have a common neighbor in both K and C , a contradiction. So p\in S(v_{i-1},v_{i},v_{i+1}) . Since S_{5} is complete to p , we have q has at most 3 neighbors on C . By (2), q\notin S(v_{i+1},v_{i+2},v_{i+3}) . If q\in S(v_{i-1},v_{i},v_{i+1}) , p-v_{i-1}-q-v_{i+1} is an induced C_{4} . If q\in S(v_{i},v_{i+1},v_{i+2}) , then p-v_{i-1}-v_{i-2}-v_{i+2}-q-K-p is an induced cyclic sequence. So q has at most two neighbors on C . It follows that p and q have a common neighbor u\in K . Since G is C_{4} -free, p and q have no common neighbors on C . If q\in S(v_{i+2}) , then p-v_{i-1}-v_{i-2}-v_{i+2}-q-N_{K}(p)-p is an induced C_{6} . So q\in S(v_{i-2},v_{i+2}) . This proves (3). (4) Suppose not. It follows from (3) that p-p^{\prime}-q-q^{\prime}-p is an induced C_{4} . This proves (4). ∎ 10 A key technical lemma In this section, we prove a key technical lemma. Let K be a connected subgraph of G and a,b\in N(K) be non-adjacent. We say that a 6-tuple (a,b,U,P_{a},P_{b},Q) is nice if the following hold: • U,P_{a},P_{b},Q form a partition of K where U,P_{a},P_{b}\neq\emptyset . • a is complete to P_{a} and anticomplete to P_{b} , and b is complete to P_{b} and anticomplete to P_{a} . • P_{a},P_{b},Q are pairwise anticomplete. • U is complete to P_{a}\cup P_{b}\cup\{a,b\} . Note that in a nice 6-tuple, the adjacency between Q and \{a,b\}\cup U can be arbitrary. Lemma 10.1 . Let G be a (P_{7},C_{4},C_{6},C_{7}) -free graph with no clique cutsets. Let K be a non-empty connected subgraph of G with the following properties. (1) There is a subset X\subseteq V(G) such that each vertex in N(K)\setminus X is universal in N(K) . (2) K\cup X contains no induced P_{4}=a-b-c-d with d\in X , c\in K and a,b\in K\cup(N(K)\cap X) . (3) There is a subset Y\subseteq V(G) anticomplete to K such that any two non-adjacent vertices in N(K)\cap X have non-empty, disjoint and complete neighborhoods in Y . Then N(K)\cap X contains two non-adjacent vertices that are complete to K . In particular, K is a clique. Proof. By (1) and G has no clique cutsets, N(K)\cap X contains two non-adjacent vertices. (i) If N(a)\cap K\nsubseteq N(b)\cap K and N(b)\cap K\nsubseteq N(a)\cap K where a,b\in N(K)\cap X are not adjacent, then there is a nice 6-tuple. Let U be the set of common neighbors of a and b in K , P_{a} be the set of private neighbors of a in K , P_{b} be the set of private neighbors of b in K , and Q be the set of common non-neighbors of a and b in K . By assumption, P_{a} and P_{b} are not empty. We show that (a,b,U,P_{a},P_{b},Q) is nice. Let a^{\prime}\in Y be a neighbor of a and Let b^{\prime}\in Y be a neighbor of b . By (3), a^{\prime}b^{\prime}\in E(G) . If u\in U is not adjacent to p_{a}\in P_{a} , then p_{a}-a-u-b is a forbidden P_{4} in (2). So U is complete to P_{a} . By symmetry, U is complete to P_{b} . If p_{a}\in P_{a} is adjacent to p_{b}\in P_{b} , then p_{a}-a-a^{\prime}-b^{\prime}-b-p_{b}-p_{a} is an induced C_{6} . So P_{a} and P_{b} are anticomplete. If q\in Q has a neighbor p_{a}\in P_{a} , then q-p_{a}-a-a^{\prime}-b^{\prime}-b-p_{b} is a bad P_{7} , since the only possible chord is qp_{b} . So Q is anticomplete to P_{a} and P_{b} . By connectivity of K , we have U\neq\emptyset . This completes the proof of (i). (ii) For any non-adjacent vertices a and b in N(K) , N(a)\cap K\subseteq N(b)\cap K or N(b)\cap K\subseteq N(a)\cap K . Suppose not. By (i), nice 6-tuples exist. We choose a nice 6-tuple (a,b,U,P_{a},P_{b},Q) so that |P_{a}|+|P_{b}| is minimum. We shall deduce a contradiction by applying the clique cutset argument to P_{a} . Let p_{a}\in P_{a} and p_{b}\in P_{b} . Let a^{\prime}\in Y be a neighbor of a and b^{\prime}\in Y be a neighbor of b . By (3), a^{\prime}b^{\prime}\in E(G) . Note that N(P_{a})\subseteq(N(K)\setminus X)\cup(X\setminus\{b\})\cup U by the definition of 6-nice tuple. Since all vertices in N(K)\setminus X are common neighbors of a and b and G is C_{4} -free, U is a clique complete to N(K)\setminus X . If u\in U is not adjacent to v\in N(P_{a})\cap X , then v-p_{a}-u-b is a forbidden P_{4} in (2). So U is complete to N(P_{a})\cap X and thus each vertex in U is universal in N(P_{a}) . Next we show that N(P_{a})\cap X is complete to a . Suppose by contradiction that c\in N(P_{a})\cap X is not adjacent to a . We may assume that p_{a}c\in E(G) . Note that p_{b}-b-b^{\prime}-a^{\prime}-a-p_{a} is an induced P_{6} . Since p_{b}-b-b^{\prime}-a^{\prime}-a-p_{a}-c is not a bad P_{7} , cb^{\prime}\in E(G) . Since c-U-b-b^{\prime}-c is not an induced C_{4} , cb\in E(G) . Then b-c-p_{a}-a contradicts (2). This shows that N(P_{a})\cap X is complete to a . Since N(P_{a}) is not a clique, there exist vertices c,c^{\prime}\in N(P_{a})\cap X such that c and c^{\prime} are not adjacent. It follows that ca,c^{\prime}a\in E(G) . Since G is C_{4} -free, b is not adjacent to c or c^{\prime} , say c . Applying to P_{b} the statement N(P_{a})\cap X is complete to a , we have c is anticomplete to P_{b} . If there is an edge st with s\in N(c)\cap P_{a} and t\in P_{a}\setminus N(c) , then t-s-c-N_{Y}(c)-b^{\prime}-b-p_{b} is an induced P_{7} . So N(c)\cap P_{a} is anticomplete to P_{a}\setminus N(c) and thus (c,b,U,P_{a}\cap N(c),P_{b},Q\cup(P_{a}\setminus N(c))) is a nice 6-tuple. By the choice of (a,b,U,P_{a},P_{b},Q) , c is complete to P_{a} . But now c^{\prime} is a neighbor of P_{a} and should be adjacent to c by the argument showing N(P_{a})\cap X is complete to a . This completes the proof of (ii). (iii) Let \{a,b\} in N(K)\cap X such that a and b have maximum number of common neighbors in K . Then K\subseteq N(a)\cap N(b) . By (ii), we may assume that N(b)\cap K\subseteq N(a)\cap K . Let U be the set of common neighbors in K , P be the private neighbors of a in K and Q be the set of common non-neighbors of a and b in K . Suppose that P\cup Q is not empty. Let T be a non-empty component of P\cup Q . Note that N(T)\subseteq(N(K)\setminus X)\cup(X\setminus\{b\})\cup U . Since all vertices in N(K)\setminus X are common neighbors of a and b by (1), U is a clique complete to N(K)\setminus X . Suppose that u\in U is not adjacent to c\in N(T)\cap X . By (2), c and u have a common neighbor d\in T . If d\in P , then then bc\in E(G) by (ii) and so c-d-u-b-c is an induced C_{4} . If d\in Q , then ca,cb\in E(G) by (ii). It follows that c-a-u-b-c is an induced C_{4} . So U is complete to N(T)\cap X and so each vertex in N(T)\cap U is universal in N(T) . Since N(T) is not a clique, P\cup Q has two non-adjacent neighbors c,c^{\prime} in N(T)\cap X . Since b is anticomplete to T , we have b\neq c,c^{\prime} . Since T is connected, G is C_{6} -free and c,c^{\prime} have disjoint and complete neighborhoods in Y , we have that c,c^{\prime} have a common neighbor in T . Then \{c,c^{\prime}\} contradicts the choice of \{a,b\} . This completes the proof of the lemma. ∎ 11 A_{1} is not empty In this section, we color G when A_{1}\neq\emptyset . Let C=v_{1}-v_{2}-v_{3}-v_{4}-v_{5}-v_{1} be an induced C_{5} of G such that |S_{5}| (see Section 9 ) is minimum among all induced C_{5} s of G . Let H be a maximal nice blowup of C . From now on, we fix the choice of C and H . We first prove some general properties and then consider two cases depending the number of non-empty A_{1}(i) s (Subsections 11.1 and 11.2 ). Lemma 11.1 . A_{5} is complete to any edge xy with x\in A_{1}(i) and y\in A_{2}(i+2) for each i\in[5] . Proof. Suppose not. Let xy be an edge with x\in A_{1}(i) and y\in A_{2}(i+2) such that some u\in A_{5} is not complete to \{x,y\} . Let x^{\prime}\in N_{B_{i}}(x) and y^{\prime}\in N_{B_{i+2}}(y) . Then C^{\prime}=x^{\prime}-q_{i+1}-y^{\prime}-y-x-x^{\prime} is an induced C_{5} . Let S_{5}^{\prime} be the set of vertices that are complete to V(C^{\prime}) . By definition of S^{\prime}_{5} , S^{\prime}_{5}\subseteq A^{\prime}_{3}(i+1)\cup A_{5} . By Lemma 4.20 , S^{\prime}_{5}\subseteq A_{5} . Since A_{5}\subseteq S_{5} and u\notin S^{\prime}_{5} , |S^{\prime}_{5}|<|S_{5}| which contradicts the choice of C . ∎ Lemma 11.2 . For each i\in\mathbb{Z}_{5} , either A_{1}(i) is anticomplete to A_{2}(i+2) or A_{1}(i+1) is anticomplete to A_{2}(i-2) . Proof. Suppose not. Let x\in A_{1}(i) be adjacent to w\in A_{2}(i+2) and y\in A_{1}(i+1) be adjacent to z\in A_{2}(i-2) . It follows from anticomplete pair properties that xw and yz induce a 2P_{2} . By Lemma 4.9 , w and z have common neighbor a in B_{i-2} . Then B_{i}-x-w-a-z-y-B_{i+1}-B_{i} is an induced cyclic sequence. ∎ By Lemma 11.2 and Proposition 6.1 , there are at most 2 non-empty A_{1}(i) s. If there are two such A_{1}(i) s, their support are of distance 2 on the maximal nice blowup H . Lemma 11.3 . A_{1}(i) is chordal for each i\in[5] . Proof. Suppose that C^{\prime}=v_{1}-v_{2}-v_{3}-v_{4}-v_{5} is an induced C_{5} in A_{1}(i) . Let N=B_{i}\cup A_{1}(i)\cup A_{2}(i-1)\cup A_{2}(i)\cup A_{2}(i+2)\cup A_{3}(i-1)\cup A_{3}(i)\cup A_{3}(i+1)\cup A_{5}. Note that N(C^{\prime})\subseteq N . Let K be the component of G-(N(C^{\prime})\cup C) containing B_{i-2}\cup B_{i-1}\cup B_{i+1}\cup B_{i+2} . The notations S(v_{3},v_{4}) and S(v_{5},v_{1},v_{2}) below are the same as defined in Section 9 . Since K is not a clique, K has two non-adjacent neighbors p and q by Lemma 9.1 . By symmetry, we may assume that p\in S(v_{5},v_{1},v_{2}) and q\in S(v_{3},v_{4}) . Since q is complete to K by Lemma 9.1 and each vertex in N(C^{\prime})\setminus A_{5} has a non-neighbor in K , we have q\in A_{5} . If p\in N(C^{\prime})\setminus(A_{1}(i)\cup A_{2}(i+2)\cup A_{5}) , then one can extend the induced P_{4}=v_{3}-v_{4}-v_{5}-p to an induced P_{7} by using vertices in B_{i-2}\cup B_{i-1}\cup B_{i+1}\cup B_{i+2} . So p\in(A_{1}(i)\cup A_{2}(i+2)\cup A_{5}) . Since q\in A_{5} is adjacent to any vertex in A_{5} or any vertex in A_{2}(i+2) that has a neighbor in A_{1}(i) and pq\notin E(G) , we have p\in A_{1}(i) . Let S be the set of vertices in K that is a neighbor of any such vertex p\in S(v_{5},v_{1},v_{2}) . Since A_{1}(i) is anticomplete to V(H)\setminus B_{i} , K\setminus S contains V(H)\setminus B_{i} . Let L be a component of K\setminus S containing V(H)\setminus B_{i} . We show that N(L) is a clique. By Lemma 9.1 (3), N(L)\setminus S is a clique. Note that any vertex N(L)\cap S is complete to L for otherwise there is an induced P_{7} in G . Since any neighbor of a vertex in A_{1}(i) which is not in A_{5} has a non-neighbor in L , N(L)\cap S\subseteq A_{5} and so N(L)\cap S is a clique. Next we show that N(L)\cap S is complete to N(L)\setminus S . Suppose by contradiction that a\in N(L)\cap S and b\in N(L)\setminus S are not adjacent. Since a is complete to L , we can take a common neighbor c\in L of a and b . By definition of S , a has a neighbor r\in S(v_{5},v_{1},v_{2}) . Since r-a-c-b-r is not an induced C_{4} , br\notin E(G) . If b is adjacent to both v_{2} and v_{5} , then r-v_{2}-b-v_{5}-r is an induced C_{4} . So we may assume that bv_{2}\notin E(G) by symmetry. Then v_{4}-v_{3}-v_{2}-r-a-c-b is a bad P_{7} . This proves that N(L)\cap S is complete to N(L)\setminus S and so N(L) is a clique, a contradiction. ∎ Lemma 11.4 . A_{2}(i-2) is chordal for each i\in[5] . Proof. Suppose that C^{\prime}=v_{1}-v_{2}-v_{3}-v_{4}-v_{5} is an induced C_{5} in A_{2}(i-2) . Let N=B_{i-2}\cup B_{i-1}\cup A_{1}(i-2)\cup A_{1}(i-1)\cup A_{1}(i+1)\cup A_{2}(i-2)\cup A_{3}(i-2)\cup A_{3}(i-1)\cup A_{5}. Note that N(C^{\prime})\subseteq N . Let K be the component of G-(N(C^{\prime})\cup V(C^{\prime})) containing B_{i}\cup B_{i+1}\cup B_{i+2} . Since K is not a clique, K has two non-adjacent neighbors p and q by Lemma 9.1 . The notations S(v_{3},v_{4}) and S(v_{5},v_{1},v_{2}) below are the same as defined in Section 9 . By symmetry, we may assume that p\in S(v_{5},v_{1},v_{2}) and q\in S(v_{3},v_{4}) . Since q is complete to K by Lemma 9.1 and each vertex in N(C^{\prime})\setminus A_{5} has a non-neighbor in K , we have q\in A_{5} . By the choice of C=C_{5} and q\in A_{5} , we have p\in A_{1}(i-2)\cup A_{1}(i-1)\cup A_{2}(i-2) . Let S be the set of vertices in K that is a neighbor of any such vertex p\in S(v_{5},v_{1},v_{2}) . Since A_{1}(i-2)\cup A_{1}(i-1)\cup A_{2}(i-2) is anticomplete to B_{i}\cup B_{i+1}\cup B_{i+2} , we have B_{i}\cup B_{i+1}\cup B_{i+2}\subseteq K\setminus S . Let L be the component of K\setminus S containing B_{i}\cup B_{i+1}\cup B_{i+2} . We show that N(L) is a clique. By Lemma 9.1 , N(L)\setminus S is a clique. Note that any vertex N(L)\cap S is complete to L for otherwise there is an induced P_{7} in G . Since any vertex in H\cup A_{1}\cup A_{2} has a non-neighbor in B_{i}\cup B_{i+1}\cup B_{i+2} , (N(L)\cap S)\cap(H\cup A_{1}\cup A_{2})=\emptyset . Since any vertex in A_{3} has a non-neighbor in B_{i}\cup B_{i+1}\cup B_{i+2} , N(L)\cap S contains no vertex from A_{3} . So N(L)\cap S\subseteq A_{5} and so is a clique. Next we show that N(L)\cap S is complete to N(L)\setminus S . Suppose by contradiction that a\in N(L)\cap S and b\in N(L)\setminus S are not adjacent. Since a is complete to L , we can take a common neighbor c\in L of a and b . By definition of S , a has a neighbor r\in S(v_{5},v_{1},v_{2}) . Since r-a-c-b-r is not an induced C_{4} , br\notin E(G) . If b is adjacent to both v_{2} and v_{5} , then r-v_{2}-b-v_{5}-r is an induced C_{4} . So we may assume that bv_{2}\notin E(G) by symmetry. Then v_{4}-v_{3}-v_{2}-r-a-c-b is a bad P_{7} . This proves that N(L)\cap S is complete to N(L)\setminus S and so N(L) is a clique, a contradiction. ∎ Lemma 11.5 . For any two non-adjacent vertices u,v\in A_{3}(i)\cup B_{i} , u and v cannot have a common neighbor in A_{1}(i) . Proof. Suppose not. Let x\in A_{1}(i) be a common neighbor of u and v . Since A_{1}(i) is anticomplete to B_{i-1}\cup B_{i+1} and G is C_{4} -free, u and v have disjoint neighborhoods in each of B_{i-1} and B_{i+1} . This contradicts Lemma 4.12 . ∎ Lemma 11.6 . Any component of A_{1}(i)\setminus N(A_{2}(i+2)) is complete to a pair of non-adjacent vertices in N(A_{2}(i+2))\cap A_{1}(i) and so is a clique. Proof. Let S=N(A_{2}(i+2))\cap A_{1}(i) . Let K be a component of A_{1}(i)\setminus S . Let N=B_{i}\cup A_{2}(i-1)\cup A_{2}(i)\cup A_{3}(i-1)\cup A_{3}(i)\cup A_{3}(i+1)\cup A_{5} . Note that N(K)\subseteq N\cup S . By Proposition 6.1 , N(K)\setminus S is a clique. Next we show that N(K)\setminus S is complete to N(K)\cap S . Suppose by contradiction that s\in N(K)\cap S is not adjacent to t\in N(K)\setminus S . Let r\in A_{2}(i+2) be a neighbor of s . If t\in B_{i} , then t-K-s-r-B_{i-2}-B_{i-1}-t is an induced cyclic sequence. So t\notin B_{i} . By Lemma 6.2 , t\notin A_{2}(i-1)\cup A_{2}(i) . By Lemma 6.3 , t\notin A_{3}(i-1)\cup A_{3}(i)\cup A_{3}(i+1) . By the choice of C , t\notin A_{5} . This proves that N(K)\setminus S is complete to N(K)\cap S . Let a,b\in N(K)\cap S be two non-adjacent vertices. If a,b have a common neighbor in K , then a,b have disjoint neighborhoods in A_{2}(i+2) which are complete to each other by G is C_{6} -free. Suppose that a,b do not have a common neighbor in K . Then each of a and b mixes on K . If a mixes on uv\in K , then Since u-v-a-(N(a)\cap A_{2}(i+2))-q_{i+2}-q_{i+1}-B_{i} is not a bad P_{7} , a is complete to B_{i} . By symmetry, b is complete to B_{i} . So a,b have a common neighbor in B_{i} , which implies that a,b have disjoint and complete neighborhoods in A_{2}(i+2) . If there is an induced P_{4}=a-b-c-d such that d\in N(K)\cap S , c\in K and a,b\in(N(K)\cap S)\cup K , then a-b-c-d-(N(d)\cap A_{2}(i+2))-q_{i+2}-q_{i+1} is an induced P_{7} . Now all three conditions of Lemma 10.1 are satisfied for K where X=S and Y=A_{2}(i+2) , and so we are done. ∎ Lemma 11.7 . A_{5} is complete to A_{1} . Proof. We show that A_{5} is complete to A_{1}(i) . Let S=N(A_{2}(i+2))\cap A_{1}(i) . By Lemma 11.1 , A_{5} is complete to S . By Lemma 11.6 , each component of A_{1}(i)\setminus S is complete to a pair of non-adjacent vertices in S . Since G is C_{4} -free, A_{5} is complete to that component. ∎ Lemma 11.8 (Simplicial vertices in A_{1}(i) ) . Let K be a component of A_{1}(i) and s be a (K,B_{i}) -minimal simplicial vertex. Then N(s)\setminus A_{2}(i+2) is a clique and so N(s)\cap A_{2}(i+2) is a clique of size >\left\lceil\frac{\omega}{4}\right\rceil complete to B_{i+2}\cup B_{i-2} . Proof. By Proposition 6.1 applying to \{s\} , N(s)\setminus(K\cup A_{2}(i+2)) is a clique. Next we investigate the neighbor of s in K and the neighbor of s in N(s)\setminus(A_{2}(i+2)\cup K)\subseteq B_{i}\cup A_{1}(i)\cup A_{2}(i-1)\cup A_{2}(i)\cup A_{3}(i-1)\cup A_{3}(i)\cup A_{3}(i+1)\cup A_{5} . By Lemma 6.4 (5), K has neighbors in at most one of A_{3}(i-1) , A_{3}(i) and A_{3}(i+1) . If K is anticomplete to A_{3}(i-1)\cup A_{3}(i+1) , we may assume by Lemma 6.4 (1) that K is anticomplete to A_{2}(i) . So N(s)\setminus A_{2}(i+2)\subseteq B_{i}\cup A_{2}(i-1)\cup A_{3}(i)\cup A_{5}\cup K . If K has a neighbor in A_{3}(i-1) , then K is anticomplete to A_{2}(i) by Lemmma 6.4 (2). Then N(s)\setminus A_{2}(i+2)\subseteq B_{i}\cup A_{2}(i-1)\cup A_{3}(i-1)\cup A_{5}\cup K . In either case, we have N(s)\setminus A_{2}(i+2)\subseteq B_{i}\cup A_{2}(i-1)\cup A_{3}(j)\cup A_{5}\cup K for j\in\{i-1,i\} . Recall that N(s)\setminus(A_{2}(i+2)\cup K) is a clique. If N(s)\cap B_{i} is not complete to N(s)\cap K , there is an induced P_{4} with one end in B_{i} and the other three vertices in A_{1}(i) by Lemma 2.7 . This contradicts Lemma 6.5 . So N(s)\cap B_{i} is complete to N(s)\cap K . By Lemma 6.2 , N(s)\cap A_{2}(i-1) is complete to N(s)\cap K . By Lemma 6.3 , N(s)\cap A_{3}(j) is complete to N(s)\cap K for j\in\{i-1,i\} . By Lemma 11.7 , N(s)\cap K is complete to N(s)\cap A_{5} . This completes the proof. ∎ Let A_{2}^{1}(i-2) be the set of vertices in A_{2}(i-2) that have a neighbor in A_{1}(i-1) and A_{2}^{1}(i+1) be the set of vertices in A_{2}(i+1) that have a neighbor in A_{1}(i+1) . Lemma 11.9 . Each vertex in A^{1}_{2}(i-2) is a universal vertex in A_{2}(i-2) . By symmetry, each vertex in A^{1}_{2}(i+1) is a universal vertex in A_{2}(i+1) . Proof. Suppose not. Let a\in A^{1}_{2}(i-2) have a non-neighbor d\in A_{2}(i-2) . By Lemma 6.6 , a is complete to B_{i-2} and thus a and d have a common neighbor v\in B_{i-2} . By definition of A_{2}^{1}(i-2) , a has a neighbor in a component K of A_{1}(i-1) . By Lemma 6.2 , a is complete to K . By Proposition 6.1 , K has a vertex b that is adjacent to some vertex c\in A_{2}(i+1) . Then d-v-a-b-c-q_{i+1}-q_{i} is an induced P_{7} , a contradiction. ∎ 11.1 Two non-empty A_{1}(i) The global assumption in this subsection is that A_{1}(i-1) and A_{1}(i+1) are not empty. By Proposition 6.1 , there is an edge between A_{1}(i-1) and A_{2}(i+1) and there is an edge between A_{1}(i+1) and A_{2}(i-2) . Let x\in A_{1}(i-1) , y\in A_{1}(i+1) , z\in A_{2}(i+1) and w\in A_{2}(i-2) with xz,yw\in E(G) . By Lemma 4.9 , z is complete to B_{i+1}\cup B_{i+2} and w is complete to B_{i-1}\cup B_{i-2} . By Lemma 4.8 , B_{i-1} is complete to B_{i-2} and B_{i+1} is complete to B_{i+2} . Recall that \omega=\omega(G) . Lemma 11.10 . Suppose that x^{\prime}\in A_{1}(i-1) has a neighbor z^{\prime}\in A_{2}(i+1) and y^{\prime}\in A_{1}(i+1) has a neighbor w^{\prime}\in A_{2}(i-2) . Then exactly one of x^{\prime}w^{\prime} and y^{\prime}z^{\prime} is an edge. Proof. By Lemmas 4.15 and 4.18 , x^{\prime}y^{\prime},z^{\prime}w^{\prime}\notin E(G) . By Lemma 4.9 , z^{\prime} is complete to B_{i+1} and w^{\prime} is complete to B_{i-1} . If x^{\prime}w^{\prime},y^{\prime}z^{\prime}\notin E(G) , then x^{\prime}-N_{B_{i-1}}(x^{\prime})-w^{\prime}-y^{\prime}-N_{B_{i+1}}(y^{\prime})-z^{\prime}-x^{\prime} is an induced C_{6} . If x^{\prime}w^{\prime},y^{\prime}z^{\prime}\in E(G) , then x^{\prime}-w^{\prime}-y^{\prime}-z^{\prime}-x^{\prime} is an induced C_{4} . So exactly one of x^{\prime}w^{\prime} and y^{\prime}z^{\prime} is an edge. ∎ Lemma 11.11 . If xw\in E(G) , then x is complete to B_{i-1} and B_{i} is complete to B_{i+1} . By symmetry, if yz\in E(G) , then y is complete to B_{i+1} and B_{i} is complete to B_{i-1} . Proof. By Lemma 11.10 , exactly one of xw and yz is an edge. By symmetry, we may assume that xw\in E(G) and yz\notin E(G) . If x has a non-neighbor s\in B_{i-1} , then x-w-s-q_{i}-q_{i+1}-z-x is an induced C_{6} . So x is complete to B_{i-1} . If c\in B_{i+1} is adjacent to b\in B_{i} but not adjacent to a\in B_{i} , then a-b-c-(N(c)\cap B_{i+2})-q_{i-2}-w-x is an induced P_{7} . So B_{i} is complete to B_{i+1} . ∎ Lemma 11.12 . A_{3}(i-1)\cup A_{3}(i+1)=\emptyset . Proof. By Lemma 11.10 , we may assume that xw\in E(G) . By Lemma 11.11 , B_{i+1} is complete to B_{i}\cup B_{i+2} . It follows from Lemma 4.10 that A_{3}(i+1)=\emptyset . Suppose that A_{3}(i-1) contains a vertex t . By Lemma 4.10 (1), let b\in B_{i-1} be a non-neighbor of t . Since B_{i-1} is complete to B_{i-2} , we can take a common neighbor c\in B_{i-2} of t and b . Let d\in N_{B_{i}}(t) and so bd\notin E(G) by Lemma 4.10 (3). By Lemma 11.11 , xb\in E(G) . If tx\in E(G) , then t-c-b-x-t is an induced C_{4} . So tx\notin E(G) . By Lemma 4.18 , tz\notin E(G) . Then t-c-b-x-z-q_{i+1}-d-t is an induced C_{7} , a contradiction. ∎ Lemma 11.13 . A_{2}=A_{2}(i-2)\cup A_{2}(i+1) . Proof. By Lemma 11.10 , we may assume by symmetry that xw\in E(G) . We show that A_{2}(i-1)\cup A_{2}(i+2)\cup A_{2}(i)=\emptyset . If A_{2}(i-1) contains a vertex t , then x-w-q_{i-2}-q_{i+2}-q_{i+1}-(N(t)\cap B_{i})-t is a bad P_{7} . If A_{2}(i+2) contains a vertex t , then t-(N(t)\cap B_{i-2})-w-x-z-q_{i+1}-q_{i} is an induced P_{7} . This shows that A_{2}(i-1)\cup A_{2}(i+2)=\emptyset . It remains to show that A_{2}(i)=\emptyset . Suppose by contradiction that A_{2}(i) contains a vertex a . If ay\in E , (N(a)\cap B_{i})-a-y-w-x-z-q_{i+2} is an induced P_{7} . So y is anticomplete to A_{2}(i) . If a has a non-neighbor b\in B_{i+1} , then a-(N(a)\cap B_{i})-b-z-x-w-q_{i-2} is an induced P_{7} . So A_{2}(i) is complete to B_{i+1} . If a is adjacent to some vertex y^{\prime}\in A_{1}(i+1)\setminus\{y\} , then y^{\prime}-a-(N(a)\cap B_{i})-q_{i-1}-x-z-q_{i+2} is a bad P_{7} . So A_{2}(i) is anticomplete to A_{1}=A_{1}(i-1)\cup A_{1}(i+1) . By Lemmas 4.17 , 4.18 , 4.20 , 4.8 , 7.1 and 11.12 , any component of A_{2}(i) have a neighbor in A_{3}(i) . Let K be the component of A_{2}(i) containing a . By Lemmas 4.15 , 4.17 and 4.18 , N(K)\subseteq H\cup A_{3}(i)\cup A_{5} . Note that any vertex in N(K)\cap A_{5} is universal in N(K) and N(K)\cap H is a clique by Lemma 4.8 . Next we show that N(K)\cap H is complete to N(K)\cap A_{3}(i) . Suppose first that a\in A_{2}(i) is adjacent to t\in A_{3}(i) . If a has a neighbor b\in B_{i}\setminus N(t) , then b-a-t-N_{B_{i-1}}(t)-q_{i-2}-q_{i+2}-z is an induced P_{7} . So a is anticomplete to B_{i}\setminus N(t) , i.e., N(a)\cap B_{i}\subseteq N(t)\cap B_{i} . By Lemma 4.8 , t is complete to B_{i+1} . This shows that N(K)\cap A_{3}(i) is complete to B_{i+1} . Now suppose that t\in A_{3}(i)\cap N(K) is not adjacent to s\in N(K)\cap B_{i} . By Lemma 4.10 and the fact that B_{i} is complete to B_{i+1} , s and t have disjoint neighborhoods in B_{i-1} . Let P be a shortest path from s to t such that the internal of P is contained in K . Since N(a)\cap B_{i}\subseteq N(t)\cap B_{i} , P has at least 4 vertices. Then s-P-t-N_{B_{i-1}}(t)-q_{i-2}-q_{i+2} is an induced path of order at least 7 . This proves that N(K)\cap H is complete to N(K)\cap A_{3}(i) . Since N(K) is not a clique, N(K)\cap A_{3}(i) contains two non-adjacent vertices t and t^{\prime} . The previous argument shows that t and t^{\prime} are complete to B_{i+1} , and so t and t^{\prime} have disjoint neighborhoods in B_{i-1} . Moreover, t and t^{\prime} have a common neighbor a\in K by Lemma 7.2 . If x is not adjacent to t and t^{\prime} , then t^{\prime}-a-t-(N(t)\cap B_{i-1})-x-z-q_{i+2} is an induced P_{7} . So we may assume by symmetry that xt\in E . Then x-t-q_{i+1}-z-x is an induced C_{4} . ∎ Lemma 11.14 . The following holds for A_{3}(i) . (1) A_{3}(i) is anticomplete to A_{1}(i-1)\cup A_{1}(i+1) . (2) If xw\in E(G) , then A_{3}(i) is complete to B_{i+1} . By symmetry, if yz\in E(G) , then A_{3}(i) is complete to B_{i-1} . (3) Suppose that A_{3}(i)\neq\emptyset . If xw\in E(G) , then |B_{i-1}|>\frac{\omega}{2} . By symmetry, if yz\in E(G) , then |B_{i+1}|>\frac{\omega}{2} . Proof. We prove one by one. (1) Suppose that p\in A_{1}(i-1)\cup A_{1}(i+1) is adjacent to t\in A_{3}(i) . By Lemma 11.10 , we may assume by symmetry that xw\in E(G) and so B_{i+1} is complete to B_{i} by Lemma 11.11 . By Lemma 4.10 (3), B_{i}\setminus N(t) is anticomplete to B_{i-1}\cap N(t) . By Lemma 8.1 (1), q_{i}\in B_{i}\cap N(t) and q_{i-1}\in B_{i-1}\setminus N(t) . Then p-t-q_{i}-q_{i-1}-q_{i-2}-q_{i+2}-z is a bad P_{7} . (2) If t\in A_{3}(i) is not adjacent to a\in B_{i+1} , then t-(N(t)\cap B_{i})-a-q_{i+2}-q_{i-2}-w-x is an induced P_{7} . (3) By Lemma 11.13 and (1), A^{\prime}_{3}(i) is anticomplete to A_{1}\cup A_{2} . By Lemmas 8.14 and 8.15 , there exist vertices v\in A_{3}(i) and u\in B_{i}\setminus N(v) such that |N(v)\cap B_{i-1}|,|N(u)\cap B_{i-1}|>\left\lceil\frac{\omega}{4}\right\rceil . Since B_{i} is complete to B_{i+1} by Lemma 11.11 , N(v)\cap B_{i-1} and N(u)\cap B_{i-1} are disjoint. So |B_{i-1}|>\frac{\omega}{2} . ∎ Recall that A_{2}^{1}(i-2) is the set of vertices in A_{2}(i-2) that have a neighbor in A_{1}(i-1) and A_{2}^{1}(i+1) is the set of vertices in A_{2}(i+1) that have a neighbor in A_{1}(i+1) . Let A_{2}^{0}(i-2)=A_{2}(i-2)\setminus A_{2}^{1}(i-2) , X=A^{0}_{2}(i-2)\cap N(A_{1}(i+1)) and Y=A^{0}_{2}(i-2)\setminus X . Let A_{2}^{0}(i+1)=A_{2}(i+1)\setminus A_{2}^{1}(i+1) , X^{\prime}=A^{0}_{2}(i+1)\cap N(A_{1}(i-1)) and Y^{\prime}=A^{0}_{2}(i+1)\setminus X^{\prime} . Lemma 11.15 . The following properties hold for A_{3}(i-2)\cup A_{3}(i+2) . (1) A_{3}(i-2) is complete to A_{2}^{1}(i-2)\cup X . By symmetry, A_{3}(i+2) is complete to A_{2}^{1}(i+1)\cup X^{\prime} . (2) A_{3}(i-2) is complete to B_{i-1} . By symmetry, A_{3}(i+2) is complete to B_{i+1} . (3) A_{3}(i-2) is anticomplete to A_{1}(i-1) . By symmetry, A_{3}(i+2) is complete to A_{1}(i+1) . Proof. We prove one by one. (1) Suppose by contradiction that t\in A_{3}(i-2) and a\in A_{2}^{1}(i-2)\cup X are not adjacent. Suppose first that a\in A_{2}^{1}(i-2) . By definition of A_{2}^{1}(i-2) , a has a neighbor in a component K of A_{1}(i-1) . By Lemma 6.2 , a is complete to K . By Proposition 6.1 , K has a vertex b that is adjacent to some vertex c\in A_{2}(i+1) . Then t-N_{B_{i-2}}(t)-a-b-c-q_{i+1}-q_{i} is an induced sequence. Now suppose that a\in X . By definition of X , a has a neighbor b\in A_{1}(i+1) . Then a-B_{i-1}-t-(N(t)\cap B_{i+2})-(N(b)\cap B_{i+1})-b-a is an induced cyclic sequence. This proves (1). (2) By definition, w\in A_{2}^{1}(i-2)\cup X . By (1), w is complete to A_{3}(i-2) . By Lemma 4.8 , (2) holds. (3) By Lemma 6.3 , t cannot mix on any component K of A_{1}(i-1) . By Proposition 6.1 , K has a neighbor p^{\prime}\in A_{2}(i+1) . Let p\in K be a neighbor of p^{\prime} . If tp\in E(G) , then t-p-p^{\prime}-N_{B_{i+2}}(t)-t is an induced C_{4} . So t is not adjacent to p and thus anticomplete to K . ∎ Lemma 11.16 . G has no induced P_{4}=a-b-c-d contained in A_{2}(i-2)\cup B_{i-1}\cup B_{i-2}\cup A_{3}(i-2) such that a has a neighbor in B_{i+2} that is not a neighbor of b . By symmetry, G has no induced P_{4}=a-b-c-d contained in A_{2}(i+1)\cup B_{i+1}\cup B_{i+2}\cup A_{3}(i+2) such that a has a neighbor in B_{i-2} that is not a neighbor of b . Proof. Suppose not. Let a^{\prime}\in B_{i+2} be a neighbor of a not adjacent to b . Since G is C_{4} -free, a^{\prime}c\notin E(G) . Note that a\in B_{i-2}\cup A_{3}(i-2) . If d\in B_{i-2}\cup A_{3}(i-2) , then a^{\prime}d\notin E(G) since B_{i-1} is complete to B_{i-2}\cup A_{3}(i-2) by Lemma 11.15 (2). If d\in A_{2}(i-2)\cup B_{i-1} , then a^{\prime}d\notin E(G) by definition. Then d-c-b-a-a^{\prime}-q_{i+1}-q_{i} is an induced P_{7} . ∎ Lemma 11.17 . Each vertex in B_{i-1} is either complete or anticomplete to any component K of A_{1}(i-1) . By symmetry, each vertex in B_{i+1} is either complete or anticomplete to any component of A_{1}(i+1) . Proof. If c\in B_{i-1} mixes on an edge ab\in K , then a-b-c-(N(c)\cap B_{i-2})-q_{i+2}-(N(y)\cap B_{i+1})-y is an induced P_{7} . ∎ Lemma 11.18 . Let K be a component of A_{1}(i-1) . Then any simplicial s vertex of K has N(s)\setminus A_{2}(i+1) is a clique and so |N(s)\cap A_{2}(i+1)|>\left\lceil\frac{\omega}{4}\right\rceil . By symmetry, any simplicial s vertex of A_{1}(i+1) has N(s)\setminus A_{2}(i-2) is a clique and so |N(s)\cap A_{2}(i-2)|>\left\lceil\frac{\omega}{4}\right\rceil . Proof. By Proposition 6.1 applying to \{s\} , N(s)\setminus(K\cup A_{2}(i+1)) is a clique. Next, we investigate the neighbor of s in K and the neighbor of s in N(s)\setminus A_{2}(i+1) . By Lemma 11.15 (3) and Lemma 11.14 (1), N(s)\setminus A_{2}(i+1)\subseteq B_{i-1}\cup A^{1}_{2}(i-2)\cup A_{5}\cup K . By Lemma 6.2 , N(s)\cap K is complete to N(s)\cap A^{1}_{2}(i-2) . By Lemma 11.17 , N(s)\cap K is complete to N(s)\cap B_{i-1} . By Lemma 11.7 , N(s)\cap K is complete to N(s)\cap A_{5} . This completes the proof. ∎ Lemma 11.19 . A_{2}(i-2)\cup B_{i-1}\cup B_{i-2}\cup A_{3}(i-2) is chordal. By symmetry, A_{2}(i+1)\cup B_{i+1}\cup B_{i+2}\cup A_{3}(i+2) is chordal. Proof. We first show that A_{2}(i-2)\cup B_{i-1}\cup B_{i-2} is chordal. Since B_{i-1}\cup B_{i-2} is a clique, any induced C_{5} contains at most two vertices from B_{i-1}\cup B_{i-2} and so any induced C_{5} in B_{i-1}\cup B_{i-2}\cup A_{2}(i-2) contains an induced P_{4}=a-b-c-d where a,b,c\in A_{2}(i-2) and d\in B_{i-1}\cup B_{i-2} by Lemma 11.4 . This contradicts Lemma 7.3 . So A_{2}(i-2)\cup B_{i-1}\cup B_{i-2} is chordal. Suppose that A_{3}(i-2)\cup B_{i-2}\cup A_{2}(i-2) contains an induced C_{5}=v_{1}-v_{2}-v_{3}-v_{4}-v_{5}-v_{1} . Suppose first that v_{1}\in A_{2}(i-2) . If v_{2}\in B_{i-2}\cup A_{3}(i-2) , then v_{4}-v_{5}-v_{1}-v_{2}-N_{B_{i+2}}(v_{2})-q_{i+1}-q_{i} is a bad P_{7} . So v_{2},v_{5}\in A_{2}(i-2) . By the same argument, v_{3},v_{4}\in A_{2}(i-2) . This contradicts Lemma 11.4 . So we may assume that C_{5} is contained in B_{i-2}\cup A_{3}(i-2) . Since B_{i-2} is a clique and A_{3}(i-2) is P_{4} -free, we may assume that v_{1},v_{2}\in B_{i-2} and v_{3},v_{4},v_{5}\in A_{3}(i-2) . But this contradicts Lemma 8.3 . So A_{3}(i-2)\cup B_{i-2}\cup A_{2}(i-2) is chordal. If A_{2}(i-2)\cup B_{i-1}\cup B_{i-2}\cup A_{3}(i-2) contains an induced C_{5}=v_{1}-v_{2}-v_{3}-v_{4}-v_{5} , then we may assume that v_{1}\in B_{i-1} and v_{2}\in A_{3}(i-2) . But then v_{4}-v_{5}-v_{1}-v_{2}-N_{B_{i+2}}(v_{2})-N_{B_{i+1}}(y)-y is a bad P_{7} . ∎ Lemma 11.20 . If A_{3}(i)=\emptyset , then |B_{i-1}|,|B_{i+1}|>\left\lceil\frac{\omega}{4}\right\rceil . Proof. Let b\in B_{i} such that |N(b)\cap(B_{i-1}\cup B_{i+1})| is minimum. By Lemma 11.12 and the assumption, A_{3}(j)=\emptyset for j\in\{i-1,i,i+1\} . Then N(b)=(N(b)\cap(B_{i}\cup B_{i-1}))\cup B_{i+1} . By the choice of b , N(b)\cap(B_{i}\cup B_{i-1}) is a clique. So d(b)\leq(\omega-1)+|N(b)\cap B_{i+1}| . Since G has no small vertices, |N(b)\cap B_{i+1}|>\left\lceil\frac{\omega}{4}\right\rceil . ∎ Lemma 11.21 . A^{1}_{2}(i-2)\cup B_{i-1}\cup B_{i-2}\cup A_{5} is a clique. By symmetry, A^{1}_{2}(i+1)\cup B_{i+1}\cup B_{i+2}\cup A_{5} is a clique. Proof. By Lemma 6.6 , A^{1}_{2}(i-2) is complete to B_{i-2} . Suppose that p\in A^{1}_{2}(i-2) is not adjacent to q\in B_{i-1}\cup A_{5} . Suppose first that q\in A_{5} . By definition of A_{2}^{1}(i-2) , p is complete to a component K of A_{1}(i-1) . By Proposition 6.1 , K has a vertex a that has a neighbor in A_{2}(i+1) . By the choice of C=C_{5} , qa\in E(G) . Then p-(N(p)\cap B_{i-2})-q-a-p is an induced C_{4} . So q\in B_{i-1} . If q is complete to K , then q-(N(p)\cap K)-p-(N(q)\cap B_{i-2})-q is an induced C_{4} . If q is anticomplete to K , then q-q_{i-2}-p-a-(N(a)\cap A_{2}(i+1))-q_{i+1}-q_{i}-q is an induced C_{7} . By Lemma 11.9 and the fact that B_{i-2} is complete to B_{i-1} , A^{1}_{2}(i-2)\cup B_{i-1}\cup B_{i-2}\cup A_{5} is a clique. ∎ Lemma 11.22 . If each vertex in A_{2}(i-2)\setminus Y is universal in A_{2}(i-2) , then any simplicial vertex of A_{2}(i-2)\cup B_{i-2}\cup B_{i-1}\cup A_{3}(i-2) is not in Y . By symmetry, if each vertex in A_{2}(i+1)\setminus Y^{\prime} is universal in A_{2}(i+1) , then any simplicial vertex of A_{2}(i+1)\cup B_{i+2}\cup B_{i+1}\cup A_{3}(i+2) is not in Y^{\prime} . Proof. Let S=A_{2}(i-2)\cup B_{i-2}\cup B_{i-1}\cup A_{3}(i-2) . Suppose by contradiction that there is a simplicial vertex u of S in Y . Let K be the component of Y containing u . We choose a vertex s\in N_{S}[u]\cap K simplicial in S such that |N(s)\cap A_{5}| is minimum. By definition of Y , K is anticomplete to A_{1} . By Lemma 11.12 , N(K)\subseteq A_{5}\cup B_{i-1}\cup B_{i-2}\cup(A_{2}(i-2)\setminus Y)\cup A_{3}(i-2) . By Lemma 11.21 , B_{i-2}\cup B_{i-1}\cup A^{1}_{2}(i-2)\cup A_{5} is a clique. By the choice of C , X is complete A_{5} . By Lemma 4.9 , X is complete to B_{i-2}\cup B_{i-1} . By the assumption that each vertex in A_{2}(i-2)\setminus Y is universal in A_{2}(i-2) , N(K)\setminus A_{3}(i-2) is a clique. By Lemma 11.15 (1) and (2), A_{3}(i-2) is complete to B_{i-1}\cup(A_{2}(i-2)\setminus Y)\cup A_{5} . It follows that each vertex in N(K)\setminus A^{\prime}_{3}(i-2) is universal in N(K) . By Lemma 11.15 (2), any two non-adjacent vertices in N(K)\cap A^{\prime}_{3}(i-2) has disjoint neighborhoods in B_{i+2} . If there is an induced P_{4}=a-b-c-d with a,b\in K\cup(N(K)\cap A^{\prime}_{3}(i-2)) , c\in K and d\in N(K)\cap A^{\prime}_{3}(i-2)) , then a-b-c-d-N_{B_{i+2}}(d)-q_{i+1}-q_{i} is an induced P_{7} . So all three conditions of Lemma 10.1 are satisfied for K with X=A^{\prime}_{3}(i-2) and Y=B_{i+2} , and so K is clique. We claim that s is simplicial in G . Suppose not. Since s is simplicial in S , there exist vertices p\in N(s)\cap Y and q\in A_{5} that are not adjacent by Lemmas 11.21 and the choice of C . By the choice of s and G is C_{4} -free, p is not simplicial in S and so has a neighbor a\in S\setminus N(s) . Since G is C_{4} -free, aq\notin E(G) . Since q\in A_{5} is complete to S\setminus Y , we have a\in Y . Now the induced path s-p-a contradicts that K is a clique. ∎ Lemma 11.23 . If u\in A_{3}(i-2)\cup B_{i-2} is simplicial in A_{2}(i-2)\cup B_{i-2}\cup B_{i-1}\cup A_{3}(i-2) , then N(u)\setminus(B_{i+2}\cup A_{3}(i+2)) is a clique and so |N(u)\cap(B_{i+2}\cup A_{3}(i+2))|>\left\lceil\frac{\omega}{4}\right\rceil . By symmetry, if u\in A_{3}(i+2)\cup B_{i+2} is simplicial in A_{2}(i+1)\cup B_{i+2}\cup B_{i+1}\cup A_{3}(i+2) . Then N(u)\setminus(B_{i-2}\cup A_{3}(i-2)) is a clique and so |N(u)\cap(B_{i-2}\cup A_{3}(i-2))|>\left\lceil\frac{\omega}{4}\right\rceil . Proof. Let S=A_{2}(i-2)\cup B_{i-2}\cup B_{i-1}\cup A_{3}(i-2) . Suppose that N(u)\setminus(B_{i+2}\cup A_{3}(i+2)) is not a clique. Since u is simplicial in S , there exist vertices p\in Y and q\in A_{5} with pq\notin E(G) . Let K be the component of Y\setminus N(q) containing p . Since N(K) is not a clique, N(K) contains two non-adjacent vertices a,b . Note that a,b\in A_{5}\cup B_{i-1}\cup B_{i-2}\cup A_{3}(i-2)\cup(A_{2}(i-2)\setminus Y)\cup(N(q)\cap Y) are adjacent to q . So a and b do not have a common neighbor in K . We show that if a\notin A_{5} , then au\in E(G) . Suppose not. Since q is adjacent to both u and a , u and a have no common neighbors in K . Then there exists an induced path P=a-\cdots-u with at least 4 vertices such that P\setminus\{a,u\}\subseteq K . This contradicts Lemma 11.16 . If a,b\notin A_{5} , u is not adjacent to at least one of a,b since u is simplicial in S , a contradiction. Therefore, a\in A_{5} and b\in Y with bu\in E(G) . Since a,b have no common neighbors in K , there exist c,d\in K such that a-c-d-b is an induced P_{4} and so a-c-d-b-q-a is an induced C_{5} . Since u is simplicial in S , u is not adjacent to c . Then u is not adjacent to d for otherwise a-c-d-u-a is an induced C_{4} . So c-d-b-u contradicts Lemma 11.16 . ∎ Next, we consider two cases. The first case is that some vertex in X is not universal in A_{2}(i-2) or some vertex in X^{\prime} is not universal in A_{2}(i+1) , and the second case is that every vertex in X is universal in A_{2}(i-2) and every vertex in X^{\prime} is universal in A_{2}(i+1) . 11.1.1 X or X^{\prime} is not universal By symmetry, we may assume that a\in X^{\prime} is not universal in A_{2}(i+1) . By definition of X^{\prime} , a has a neighbor b\in A_{1}(i-1) and ay\notin E(G) . By applying Lemma 11.10 to edges ab and yw , we have bw\in E(G) (at this point, we break the symmetry). For consistency, we rename a as z and rename b as x . So the global assumption in this subsection is that z has a non-neighbor z^{\prime}\in A_{2}(i+1) . It follows from Lemma 11.11 that G contains the graph in Figure 5 as an induced subgraph. B_{i} B_{i-1} B_{i+1} B_{i-2} B_{i+2} x w y z Figure 5 : A basic graph arising from two non-empty A_{1}(j) s. A double line means complete. Lemma 11.24 . A_{3}(i)=\emptyset . Proof. If a\in B_{i-1} is adjacent to b\in B_{i} but not adjacent to c\in B_{i} , then z^{\prime}-(N(z^{\prime})\cap B_{i+2})-z-x-a-b-c is an induced P_{7} . So B_{i-1} is complete to B_{i} . By Lemma 11.11 , B_{i+1} is complete to B_{i} . By Lemma 4.10 (3), A_{3}(i)=\emptyset . ∎ Lemma 11.25 . A_{3}(i+2)=\emptyset . Proof. Suppose by contradiction that A_{3}(i+2) contains a vertex t . Since B_{i+1} is complete to B_{i+2} , B_{i+2}\setminus N(t) is anticomplete to N(t)\cap B_{i-2} by Lemma 4.10 . By Lemmas 4.16 and 4.20 , tx,tw\notin E(G) . Case 1. z^{\prime} has a neighbor b in B_{i+2}\setminus N(t) . If tz\notin E(G) , then z^{\prime}-b-z-x-w-(N(t)\cap B_{i-2})-t is a bad P_{7} . So tz\in E(G) . If tz^{\prime}\in E(G) , then z^{\prime}-b-t-z^{\prime} is an induced C_{4} . So tz^{\prime}\notin E(G) and thus z^{\prime}-b-z-t-(N(t)\cap B_{i-2})-q_{i-1}-q_{i} is an induced P_{7} . Case 2. z^{\prime} has no neighbors in B_{i+2}\setminus N(t) . By Case 1 , A_{2}(i+1)\setminus N(z) is anticomplete to A_{1}(i-1) . By Lemma 11.9 , A_{2}(i+1)\setminus N(z) is anticomplete to A_{1} . Let K_{z^{\prime}} be the component of A_{2}(i+1)\setminus N(z) containing z^{\prime} . By Lemma 4.18 , K_{z^{\prime}} is anticomplete to A_{2}\setminus N_{A_{2}(i+1)}(z) . By Lemmas 4.18 , 4.20 , 11.12 and 11.24 , N(K_{z^{\prime}})\cap A_{3}\subseteq A_{3}(i+2) and so N(K_{z^{\prime}})\subseteq B_{i+1}\cup N_{A_{2}(i+1)}(z)\cup A^{\prime}_{3}(i+2)\cup A_{5} . Next we show that N(K_{z^{\prime}})\cap A_{5} is complete to N(K_{z^{\prime}})\cap N_{A_{2}(i+1)}(z) . For contradiction, let p\in N(K_{z^{\prime}})\cap N_{A_{2}(i+1)}(z) and q\in N(K_{z^{\prime}})\cap A_{5} be non-adjacent. By the choice of C , q is complete to \{x,z\} . If p and q have a common neighbor c\in K_{z^{\prime}} , then c-p-z-q-c is an induced C_{4} . So p and q have disjoint neighborhoods in K_{z^{\prime}} . This implies that p mixes on an edge ab\in K_{z^{\prime}} . Then xp\in E(G) for otherwise a-b-p-z-x-q_{i-1}-q_{i} is an induced P_{7} . By Lemma 4.9 , p is complete to B_{i+2} . Then p-x-q-q_{i+2}-p is an induced C_{4} . By Lemma 4.14 , each vertex in N(K_{z^{\prime}})\cap A_{5} is universal in N(K_{z^{\prime}}) . Next we show that N(K_{z^{\prime}})\cap B_{i+1} is complete to N(K_{z^{\prime}})\cap N_{A_{2}(i+1)}(z) . For contradiction, let p\in N(K_{z^{\prime}})\cap N_{A_{2}(i+1)}(z) and q\in N(K_{z^{\prime}})\cap B_{i+1} be non-adjacent. By definition, pz\in E(G) and qz\in E(G) . Since G is C_{4} -free, p and q do not have a common neighbor in K_{z^{\prime}} . So p mixes on K_{z^{\prime}} and thus xp\in E(G) . By Lemma 4.9 , pq\in E(G) , a contradiction. By Lemma 4.8 , each vertex in N(K_{z^{\prime}})\cap B_{i+1} is universal in N(K_{z^{\prime}}) . We show that N(K_{z^{\prime}})\cap N_{A_{2}(i+1)}(z) is a clique complete to A^{\prime}_{3}(i+2) . Suppose that p,q\in N_{A_{2}(i+1)}(z) are not adjacent. Since G is C_{4} -free, p and q have disjoint neighborhoods in K_{z^{\prime}} . This implies that both p and q mix on K_{z^{\prime}} and so xp,xq\in E(G) . Then p-x-q-q_{i+2}-p is an induced C_{4} . This shows that N(K_{z^{\prime}})\cap N_{A_{2}(i+1)}(z) is a clique. Suppose that p\in N(K_{z^{\prime}})\cap N_{A_{2}(i+1)}(z) and q\in N(K_{z^{\prime}})\cap A^{\prime}_{3}(i+2) be non-adjacent. Recall that qx\notin E(G) . If qz\notin E(G) , then q-K_{z^{\prime}}-\{p,z\}-x-w-B_{i-2}-q is an induced cyclic sequence. So qz\in E(G) . Since pz\in E(G) and G is C_{4} -free, p and q do not have common neighbors in K_{z^{\prime}} . So p mixes on K_{z^{\prime}} and so xp\in E(G) . Then q-K_{z^{\prime}}-p-x-w-B_{i-2}-q is an induced cyclic sequence. So each vertex in N(K_{z^{\prime}})\cap N_{A_{2}(i+1)}(z) is universal in N(K_{z^{\prime}}) . Since B_{i+1} is complete to A^{\prime}_{3}(i+2) by Lemma 11.15 (2), each pair of non-adjacent vertices in N(K_{z^{\prime}})\cap A^{\prime}_{3}(i+2) has disjoint neighborhoods in B_{i-2} . If there is an induced P_{4}=a-b-c-d with a,b\in K_{z^{\prime}}\cup N(K_{z^{\prime}}) , c\in K_{z^{\prime}} and d\in A^{\prime}_{3}(i+2) , then a-b-c-d-N_{B_{i-2}}(d)-q_{i-1}-q_{i} is an induced P_{7} . So all three conditions of Lemma 10.1 are satisfied for K_{z^{\prime}} with X=A^{\prime}_{3}(i+2) and Y=B_{i-2} . It follows that K_{z^{\prime}} is complete to two non-adjacent vertices p,q\in A^{\prime}_{3}(i+2) . If zq\notin E(G) , then p-z^{\prime}-q-(N(q)\cap B_{i-2})-w-x-z is a bad P_{7} . So zq\in E(G) . By symmetry, zp\in E(G) . Then p-z^{\prime}-q-z-p is an induced C_{4} . ∎ Lemma 11.26 . Each vertex in A_{1}(i+1) has a neighbor in A_{2}(i-2) . Proof. Let S\subseteq A_{1}(i+1) be the set of vertices that have a neighbor in A_{2}(i-2) and T=A_{1}(i+1)\setminus S . The lemma is equivalent to show T=\emptyset . Suppose not. Let K be a connected component of T . By Lemmas 11.24 and 11.25 , N(K)=(N(K)\cap B_{i+1})\cup(N(K)\cap A_{1}(i+1))\cup(N(K)\cap A_{2}(i+1))\cup(N(K)\cap A_{5}). By Proposition 6.1 , N(K)\setminus A_{1}(i+1) is a clique. Observe that any vertex p\in S is not adjacent to z by Lemma 11.9 and the assumption that z is not universal in A_{2}(i+1) . If a vertex p\in N(K)\cap A_{1}(i+1) is mixed on ab\in K , then a-b-p-(N(p)\cap A_{2}(i-2))-q_{i-2}-q_{i+2}-z is a bad P_{7} . So any vertex in N(K)\cap A_{1}(i+1) is complete to K . We show that N(K)\cap A_{1}(i+1) is complete to N(K)\setminus A_{1}(i+1) . Suppose by contradiction that p\in N(K)\cap A_{1}(i+1) and q\in N(K)\setminus A_{1}(i+1) with pq\notin E(G) . By definition of p , p has a neighbor w^{\prime}\in A_{2}(i-2) . Let u\in K be a common neighbor of p and q . By Lemma 11.7 , q\notin A_{5} . If q\in B_{i+1} , then p-u-q-q_{i}-q_{i-1}-w^{\prime}-p is an induced C_{6} . If q\in A_{2}(i+1) , then p-u-q-N_{B_{i+2}}(q)-q_{i-2}-w^{\prime}-p is an induced C_{6} . Therefore, each vertex in N(K)\setminus A_{1}(i+1) is universal in N(K) . Since N(K) is not clique, there are two non-adjacent vertices p,q\in N(K)\cap A_{1}(i+1) . So p and q have a common neighbor u\in K and thus p and q have disjoint neighborhoods in A_{2}(i-2) . Since pz\notin E(G) , q-u-p-(N(p)\cap A_{2}(i-2))-q_{i-2}-q_{i+2}-z is a bad P_{7} . ∎ Lemma 11.27 ( P_{4} -freeness of A_{1}(i+1) ) . The following properties hold for A_{1}(i+1) . (1) x is complete to N(A_{1}(i+1))\cap A_{2}(i-2) . (2) For any vertex w^{\prime}\in A_{2}(i-2) and any component K of A_{1}(i+1) , w is either complete or anticomplete to K . (3) A_{1}(i+1) is P_{4} -free. Proof. We prove these properties one by one. (1) Let w^{\prime}\in N(A_{1}(i+1))\cap A_{2}(i-2) and y^{\prime}\in A_{1}(i+1) be a neighbor of w^{\prime} . By Lemma 11.9 , zy^{\prime}\notin E(G) . By Lemma 11.10 applying to xz and y^{\prime}w^{\prime} , xw^{\prime}\in E(G) . (2) Suppose by contradiction that there exists an edge ab\in E(K) such that w^{\prime} is adjacent to b but not to a . By (1), xw^{\prime}\in E(G) and thus a-b-w^{\prime}-x-z-(N(z^{\prime})\cap B_{i+2})-z^{\prime} is a bad P_{7} . (3) It suffices to show that each component K of A_{1}(i+1) is P_{4} -free. By Lemma 11.26 and (2), there exists a vertex w^{\prime}\in A_{2}(i-2) that is complete to K . If there is an induced P_{5}=a-b-c-d-e with a,b,c,d\in K and e\in B_{i+1} , then a-b-c-d-e-q_{i}-q_{i-1} is an induced P_{7} . By Lemma 2.8 , K is P_{4} -free. ∎ Lemma 11.28 . A^{0}_{2}(i-2) is anticomplete to A_{1}(i+1) . Proof. Suppose that a\in A^{0}_{2}(i-2) is adjacent to b\in A_{1}(i+1) . By definition of A^{0}_{2}(i-2) , ax\notin E(G) . Applying Lemma 11.10 to the edges xz and ab , we have bz\in E(G) . By Lemma 11.9 , z is universal in A_{2}(i+1) . This contradicts the assumption that z is not universal in A_{2}(i+1) . ∎ Lemma 11.29 . A_{2}^{0}(i-2)=\emptyset . Proof. Suppose not. Let L be a component of A_{2}^{0}(i-2) . By Lemma 11.27 (1), L is anticomplete to A_{1}(i+1) . By Lemma 11.28 , N(L)\subseteq A_{5}\cup B_{i-1}\cup B_{i-2}\cup A^{1}_{2}(i-2)\cup A_{3}(i-2) . By Lemma 11.21 , N(L)\setminus A_{3}(i-2) is a clique. Since G has no clique cutsets, L has neighbors in A_{3}(i-2) and so A_{3}(i-2)\neq\emptyset . Let S=B_{i-1}\cup B_{i-2}\cup A_{3}(i-2)\cup A_{2}(i-2) . Since A_{3}(i-2)\neq\emptyset , S is not a clique and so S has two non-adjacent simplicial vertices u,u^{\prime} by Lemma 11.19 . Since A^{1}_{2}(i-2)\cup B_{i-1}\cup B_{i-2} is a clique, we may assume that u\in A^{0}_{2}(i-2)\cup A_{3}(i-2) . By Lemma 11.28 , X=\emptyset and so A_{2}(i-2)\setminus Y=A_{2}^{1}(i-2) . By Lemma 11.9 , each vertex in A_{2}^{1}(i-2) is universal in A_{2}(i-2) . By Lemma 11.22 , u\in A_{3}(i-2) . By Lemma 11.15 (1) and (2) and Lemma 11.22 , u^{\prime}\in A_{3}(i-2)\cup B_{i-2} . By Lemma 11.25 , A_{3}(i+2)=\emptyset . It follows from Lemma 11.23 that |N(u)\cap B_{i+2}|>\left\lceil\frac{\omega}{4}\right\rceil and |N(u^{\prime})\cap B_{i+2}|>\left\lceil\frac{\omega}{4}\right\rceil . Since A_{3}(i-2)\cup B_{i-2} is complete to B_{i-1} , u and u^{\prime} have disjoint neighborhoods in B_{i+2} . By Lemma 11.18 , |A_{2}(i+1)|>\left\lceil\frac{\omega}{4}\right\rceil . By Lemmas 11.24 and Lemma 11.20 , |B_{i+1}|>\left\lceil\frac{\omega}{4}\right\rceil . So B_{i+1}\cup B_{i+2}\cup A_{2}(i+1) contains a clique of size larger than \omega , a contradiction. ∎ Lemma 11.30 . There is a vertex b\in B_{i-2} anticomplete to A_{3}(i-2) such that |N(b)\cap B_{i+2}|>\left\lceil\frac{\omega}{4}\right\rceil . This implies that A_{3}(i-2)=\emptyset . Proof. By Lemma 4.13 , there is a vertex b\in B_{i-2} that is anticomplete to A_{3}(i-2) . By Lemmas 11.21 and 11.29 , we have N(b)\setminus B_{i+2} is a clique. Since G has no small vertices, |N(b)\cap B_{i+2}|>\left\lceil\frac{\omega}{4}\right\rceil . By Lemma 11.18 , A_{2}(i+1) contains a clique M of size larger than \left\lceil\frac{\omega}{4}\right\rceil complete to B_{i+1}\cup B_{i+2} . Suppose that A_{3}(i-2)\neq\emptyset . By Lemma 8.5 , let t be a simplicial vertex of A_{3}(i-2) . By Lemma 8.3 , N(t)\cap A_{3}(i-2) is complete to N(t)\cap B_{i-2} . By Lemma 11.15 (1), (2) and Lemma 11.21 , N(t)\setminus B_{i+2} is a clique. Since G has no small vertices, |N(t)\cap B_{i+2}|>\left\lceil\frac{\omega}{4}\right\rceil . Since N(t)\cap B_{i+2} and N(b)\cap B_{i+2} are disjoint, B_{i+1}\cup B_{i+2}\cup M is a clique of size larger than \omega , a contradiction. ∎ Lemma 11.31 . A_{1}(i-1) is a clique. Proof. Let K be the component of A_{1}(i-1) containing x . By Lemmas 11.11 and 11.17 , K is complete to B_{i-1} . If there is simplicial vertex of A_{1}(i-1) not adjacent to some simplicial vertex of K , then A_{2}(i+1)\cup B_{i+1}\cup B_{i+2} contains a clique larger than \omega by Lemmas 11.18 and 11.30 . This implies that A_{1}(i-1) is a clique. ∎ Lemma 11.32 . A_{2}^{1}(i+1) is anticomplete to A_{1}(i-1) . Proof. Suppose by contradiction that v\in A_{2}^{1}(i+1) has a neighbor v^{\prime}\in A_{1}(i-1) . By definition, v has a neighbor a\in A_{1}(i+1) . By Lemma 11.26 , a has a neighbor a^{\prime}\in A_{2}(i-2) . By applying Lemma 11.10 to aa^{\prime} and vv^{\prime} , a^{\prime}v^{\prime}\notin E(G) . By Lemma 11.27 (3), xa^{\prime}\in E(G) . Since A_{1}(i-1) is a clique by Lemma 11.31 , a^{\prime} mixes on xv^{\prime} which contradicts Lemma 6.2 . ∎ Recall that A_{2}^{1}(i+1) is the set of vertices in A_{2}(i+1) that have a neighbor in A_{1}(i+1) , A_{2}^{0}(i+1)=A_{2}(i+1)\setminus A_{2}^{1}(i+1) , X^{\prime}=A^{0}_{2}(i+1)\cap N(A_{1}(i-1)) and Y^{\prime}=A^{0}_{2}(i+1)\setminus X^{\prime} . Lemma 11.33 . Let s\in A_{2}(i+1)\cup B_{i+1}\cup B_{i+2} be a simplicial vertex of A_{2}(i+1)\cup B_{i+1}\cup B_{i+2} . If s\in X^{\prime} , then N(s)\setminus A_{1}(i-1) is a clique and so |N(s)\cap A_{1}(i-1)|>\left\lceil\frac{\omega}{4}\right\rceil . Proof. Suppose not. By Lemma 11.32 , s is anticomplete to A_{1}(i+1) and thus N(s)\setminus A_{1}(i-1)\subseteq B_{i+1}\cup B_{i+2}\cup A_{2}(i+1)\cup A_{5} . By Lemma 11.21 and 11.1 , there exist vertices b\in N(s)\cap Y^{\prime} and q\in N(s)\cap A_{5} that are not adjacent. Let K be the component of Y^{\prime}\setminus N(q) containing b . Note that N(K)\subseteq A_{2}(i+1)\cup B_{i+1}\cup B_{i+2}\cup A_{5} . By Lemmas 11.9 and 11.21 , N(K)\cap(A_{2}^{1}(i+1)\cup B_{i+1}\cup B_{i+2}\cup A_{5}) is a clique and complete to N(K)\cap X^{\prime} . Since N(K) is not a clique, K has two non-adjacent neighbors p_{1},p_{2} . Suppose first that p_{1}\in X^{\prime} . Then p_{2}\in X^{\prime}\cup Y^{\prime} . If p_{1},p_{2} have a common neighbor u\in K , then p_{1}-u-p_{2}-q-p_{1} is an induced C_{4} . So p_{1},p_{2} have no common neighbors in K . Let P be a shortest path from p_{1} to p_{2} whose internal is contained in K . So |P|\geq 4 . Then p_{2}-P-p_{1}-r-N_{B_{i-1}}(r)-q_{i} is an induced path with at least 7 vertices, where r is a neighbor of p_{1}\in A_{1}(i-1) . This proves that each vertex in N(K)\cap X^{\prime} is universal in N(K) . So p_{1},p_{2}\notin X^{\prime} and s is adjacent to p_{1} and p_{2} . Therefore, we may assume that p_{1}\in Y^{\prime} . Since A_{2}^{1}(i+1) is universal in A_{2}(i+1) , p_{2}\in Y^{\prime}\cup B_{i+1}\cup B_{i+2}\cup A_{5} . Since s is simplicial in A_{2}(i+1)\cup B_{i+1}\cup B_{i+2} , we have p_{2}\in A_{5} . Since q is a common neighbor of p_{1} and p_{2} , p_{1} and p_{2} have no common neighbors in K . Since s is simplicial in A_{2}(i+1)\cup B_{i+1}\cup B_{i+2} , N(s)\cap K\subseteq N(p_{1})\cap K and so s and p_{2} have no common neighbors in K . Let P=s-v_{1}-\cdots v_{k}-p_{2} be a shortest path between s and p_{2} with v_{i}\in K for all i\in[k] . If k=2 , then s-v_{1}-v_{2}-p_{2}-s is an induced C_{4} . If k\geq 3 , then v_{3}-v_{2}-v_{1}-s-r-N_{B_{i-1}}(r)-q_{i} is an induced path at least 7 vertices, where r\in A_{1}(i-1) is a neighbor of s . ∎ Lemma 11.34 . \chi(G)\leq\left\lceil\frac{5}{4}\omega\right\rceil . Proof. By Lemma 11.18 , A_{2}(i+1) contains a clique M complete to B_{i+1}\cup B_{i+2} with |M|>\left\lceil\frac{\omega}{4}\right\rceil . Since A_{5}\cup M\cup B_{i+1}\cup B_{i+2} is a clique, |A_{5}|<\left\lceil\frac{\omega}{4}\right\rceil . (1) A_{1}(i-1) is a clique and complete to B_{i-1}\cup A_{2}(i-2) . By Lemma 6.2 and A_{2}(i-2)=A_{2}^{1}(i-2) , A_{2}(i-2) is complete to A_{1}(i-1) . By x is complete to B_{i-1} and Lemma 11.17 , A_{1}(i-1) is complete to B_{i-1} . This proves (1). Let s\in A_{2}(i+1)\cup B_{i+1}\cup B_{i+2} be a simplicial vertex in A_{2}(i+1)\cup B_{i+1}\cup B_{i+2} . If s\in Y^{\prime} , then N(s)\setminus A_{5} is a clique and so |A_{5}|>\left\lceil\frac{\omega}{4}\right\rceil , a contradiction. So s\notin Y^{\prime} . (2) A_{2}(i+1)\cup B_{i+1}\cup B_{i+2} is a clique. Suppose not. A_{2}(i+1)\cup B_{i+1}\cup B_{i+2} contains two non-adjacent simplicial vertices s,s^{\prime} by Lemma 11.19 . Since A_{2}^{1}(i+1)\cup B_{i+1}\cup B_{i+2} is a clique complete to X^{\prime} , we have s,s^{\prime}\in X^{\prime} . By Lemma 11.33 , |N(s)\cap A_{1}(i-1)|,|N(s^{\prime})\cap A_{1}(i-1)|>\left\lceil\frac{\omega}{4}\right\rceil . Since N(s)\cap A_{1}(i-1) and N(s^{\prime})\cap A_{1}(i-1) are disjoint and complete, then B_{i-1}\cup A_{1}(i-1)\cup A_{2}(i-2) is a clique of size larger than \omega by (1). This proves (2). By (2), Y^{\prime}=\emptyset . By Lemmas 11.1 , 11.21 and 11.7 , each vertex in A_{5} is universal in G . Since G is basic, A_{5}=\emptyset . (3) A_{1}(i+1) is a clique. Suppose not. Let s,s^{\prime}\in A_{1}(i+1) be two non-adjacent simplicial vertices of A_{1}(i+1) . If N(s)\cap A_{2}(i-2) and N(s^{\prime})\cap A_{2}(i-2) are disjoint, then A_{2}(i-2)\cup B_{i-1}\cup A_{1}(i-1) is a clique of size larger than \omega by Lemmas 11.18 , 11.33 and 11.20 . So s and s^{\prime} have a common neighbor in A_{2}(i-2) and so s and s^{\prime} have disjoint neighborhoods in B_{i+1}\cup A_{2}^{1}(i+1) . By Lemma 11.27 (2) and A_{2}^{0}(i-2)=\emptyset , |N(s)\cap(B_{i+1}\cup A_{2}^{1}(i+1))|>\left\lceil\frac{\omega}{4}\right\rceil and |N(s^{\prime})\cap(B_{i+1}\cup A_{2}^{1}(i+1))|>\left\lceil\frac{\omega}{4}\right\rceil . Then M\cup A_{2}^{1}(i+1)\cup B_{i+1}\cup B_{i+2} is a clique of size larger than \omega . This proves (3). (4) For every vertex v\in A_{1}(i+1) , |N(v)\cap B_{i+1}|>\left\lceil\frac{\omega}{4}\right\rceil . If v has no neighbors in A_{2}^{1}(i+1) , then N(v)\cap B_{i+1}=N(v)\cap(B_{i+1}\cup A_{2}^{1}(i+1)) and so the statement holds since |N(v)\cap(B_{i+1}\cup A_{2}^{1}(i+1))|>\left\lceil\frac{\omega}{4}\right\rceil . So we may assume that v has a neighbor c\in A_{2}^{1}(i+1) . If v has a non-neighbor v^{\prime}\in B_{i+1} , then v^{\prime}-c-v-N_{A_{2}(i-2)}(v)-q_{i-1}-q_{i}-v^{\prime} is an induced C_{6} . Then v is complete to B_{i+1} and so |N(v)\cap B_{i+1}|>\left\lceil\frac{\omega}{4}\right\rceil . This proves (4). Next we give a \left\lceil\frac{5}{4}\omega\right\rceil -coloring of G . Let x=|B_{i-1}| , y=|B_{i-2}| , and z=|A_{2}(i-2)| . Our strategy is to use p=x+y+z-3\left\lceil\frac{\omega}{4}\right\rceil colors to color some vertices of G first and then argue that the remaining vertices can be colored with f=\left\lceil\frac{5}{4}\omega\right\rceil-p colors. Note that \displaystyle f \displaystyle=\left\lceil\frac{5}{4}\omega\right\rceil-(x+y+z-3\left\lceil\frac{\omega}{4}\right\rceil) \displaystyle=\omega+(4\left\lceil\frac{\omega}{4}\right\rceil-(x+y+z)) \displaystyle\geq\omega, since x+y+z\leq\omega . Step 1: Precolor a subgraph. Next we define vertices to be colored with p colors. Let S_{i-1}\subseteq B_{i-1} be the set of x-\left\lceil\frac{\omega}{4}\right\rceil vertices, S_{i-2}\subseteq B_{i-2} be the set of y-\left\lceil\frac{\omega}{4}\right\rceil vertices, S_{z}\subseteq A_{2}(i-2) be the set of z-\left\lceil\frac{\omega}{4}\right\rceil vertices, and S_{i+1}\subseteq B_{i+1} be the set of p vertices with largest neighborhoods in A_{1}(i+1) . Since (x+y+z)-3\left\lceil\frac{\omega}{4}\right\rceil\leq\left\lceil\frac{\omega}{4}\right\rceil<|B_{i+1}| , the choice of S_{i} is possible. By (4), S_{i+1} is complete to A_{1}(i+1) . We color S_{i+1} with all colors in [p] , and color all vertices in S_{i-1}\cup S_{i-2}\cup S_{z} with all colors in [p] . Since B_{i-1} is anticomplete to B_{i-1}\cup B_{i-2}\cup A_{2}(i-2) , this coloring is proper. So it suffices to show that G_{1}=G-(S_{i-1}\cup S_{i-2}\cup S_{i+1}\cup S_{z}) is f -colorable. Step 2: Reduce by degeneracy. Let v be a vertex in B_{i}\cup B_{i+2}\cup A_{1}(i+1) . Note that |N_{G_{1}}(v)|<\left\lceil\frac{\omega}{4}\right\rceil+(\omega-\left\lceil\frac{\omega}{4}\right\rceil)\leq f . It follows that G_{1} is f -colorable if and only if G_{1}-(B_{i}\cup B_{i+2}\cup A_{1}(i+1)) is f -colorable. Since G_{1}-(B_{i}\cup B_{i+2}\cup A_{1}(i+1)) is chordal, \chi(G_{1}-(B_{i}\cup B_{i+2}\cup A_{1}(i+1)))\leq\omega\leq f . ∎ 11.1.2 X and X^{\prime} are universal Recall that A_{2}^{1}(i-2) is the set of vertices in A_{2}(i-2) that have a neighbor in A_{1}(i-1) and A_{2}^{0}(i-2)=A_{2}(i-2)\setminus A_{2}^{1}(i-2) , X=A^{0}_{2}(i-2)\cap N(A_{1}(i+1)) and Y=A^{0}_{2}(i-2)\setminus X . Similarly. A_{2}^{1}(i+1) is the set of vertices in A_{2}(i+1) that have a neighbor in A_{1}(i+1) and A_{2}^{0}(i+1)=A_{2}(i+1)\setminus A_{2}^{1}(i+1) , and X^{\prime}=A^{0}_{2}(i+1)\cap N(A_{1}(i-1)) and Y^{\prime}=A^{0}_{2}(i+1)\setminus X^{\prime} . The global assumption in this subsection is that each vertex in X is universal in A_{2}(i-2) and each vertex in X^{\prime} is universal in A_{2}(i+1) . Lemma 11.35 . A_{3}(i-2)=A_{3}(i+2)=\emptyset . Proof. Suppose by symmetry that A_{3}(i-2)\neq\emptyset . By Lemma 11.12 , A_{3}(i-1)=\emptyset . Since A_{3}(i-2)\neq\emptyset , S=A_{2}(i-2)\cup B_{i-1}\cup B_{i-2}\cup A_{3}(i-2) is not clique. By Lemma 11.19 , S contains two non-adjacent simplicial vertices u and u^{\prime} . Since each vertex in X is universal in A_{2}(i-2) , A^{1}_{2}(i-2)\cup X\cup B_{i-2}\cup B_{i-1} is a clique. So we may assume that u\in Y\cup A_{3}(i-2) . By Lemma 11.22 , u\in A_{3}(i-2) . By Lemma 11.15 (1) and (2) and Lemma 11.22 , u^{\prime}\in A_{3}(i-2)\cup B_{i-2} . Since u\in A_{3}(i-2) and uu^{\prime}\notin E(G) , N(u)\cap A_{3}^{\prime}(i+2)\subseteq B_{i+2} and N(u^{\prime})\cap A_{3}^{\prime}(i+2)\subseteq B_{i+2} by Lemma 8.11 . By Lemma 11.23 , |N_{B_{i+2}}(u)|,|N_{B_{i+2}}(u^{\prime})|>\left\lceil\frac{\omega}{4}\right\rceil . Since A_{3}(i-2)\cup B_{i-2} is complete to B_{i-1} , u and u^{\prime} have disjoint neighborhoods in B_{i+2} . By Lemma 11.18 , |X^{\prime}|>\left\lceil\frac{\omega}{4}\right\rceil . If A_{3}(i)=\emptyset , then |B_{i+1}|>\left\lceil\frac{\omega}{4}\right\rceil by Lemma 11.20 . If A_{3}(i)\neq\emptyset , then |B_{i+1}|>\left\lceil\frac{\omega}{4}\right\rceil by Lemmas 8.15 and 11.13 . So B_{i+1}\cup B_{i+2}\cup A_{2}(i+1) contains a clique of size larger than \omega , a contradiction. ∎ Lemma 11.36 . Y=Y^{\prime}=\emptyset . Proof. If Y\neq\emptyset , then N(Y)\subseteq A^{1}_{2}(i-2)\cup X\cup B_{i-1}\cup B_{i-2}\cup A_{5} by Lemmas 11.12 and 11.35 , and so is a clique. This contradicts that G has no clique cutsets. ∎ Lemma 11.37 . A_{3}(i)=\emptyset . Proof. Suppose that t\in A_{3}(i) . If a\in A_{1}(i+1) is adjacent to b\in A_{2}(i+1) , then t-q_{i}-q_{i-1}-q_{i-2}-(N(b)\cap B_{i+2})-b-a is an induced P_{7} . Then A_{1}(i+1) is anticomplete to A_{2}(i+1) and so A_{2}^{1}(i+1)=\emptyset . By Lemma 11.35 , A_{3}(i+2)=\emptyset . By Lemma 11.36 , Y^{\prime}=\emptyset . By Lemma 11.15 (2) and Lemma 8.1 (1), q_{i+2} is anticomplete to A_{3}(i-2) . It follows that N(q_{i+2})\setminus B_{i-2}\subseteq B_{i+1}\cup B_{i+2}\cup X^{\prime}\cup A_{5} is a clique. Since G has no small vertices, |B_{i-2}|>\left\lceil\frac{\omega}{4}\right\rceil . By Lemma 11.18 , A_{2}^{1}(i-2)\cup X contains a clique L of size larger than \left\lceil\frac{\omega}{4}\right\rceil . By Lemma 11.14 (3), then |B_{i-1}|>\frac{\omega}{2} . By Lemmas 11.15 (1) and (2), B_{i-2}\cup B_{i-1}\cup L contains a clique of size larger than \omega , a contradiction. ∎ We say that an edge ab\in E(G) with a\in A_{1}(i-1) and b\in A_{2}(i-2) is a vertical edge if a has a neighbor in A_{2}(i+1) and b has a neighbor in A_{1}(i+1) . A vertical edge between A_{1}(i+1) and A_{2}(i+1) is defined in a similar way. Lemma 11.38 . Vertical edges between A_{1}(i-1) and A_{2}(i-2) and vertical edges between A_{1}(i+1) and A_{2}(i+1) cannot exist at the same time. Proof. Suppose not. Without loss of generality, xw\in E(G) and zy\notin E(G) . By Lemmas 11.37 , 11.12 and 11.35 , A_{3}=\emptyset . By Lemma 11.36 , A_{2}(i-2)\cup B_{i-1}\cup B_{i-2}\cup A_{5} is a clique and A_{2}(i+1)\cup B_{i+1}\cup B_{i+2}\cup A_{5} is a clique. Since q_{i-2} is not a small vertex, |B_{i+2}|>\left\lceil\frac{\omega}{4}\right\rceil . By symmetry, |B_{i-2}|>\left\lceil\frac{\omega}{4}\right\rceil . Next we show that A_{1}(j) is a clique for j\in\{i-1,i+1\} . Let K be the component of A_{1}(i-1) containing x . By Lemmas 11.11 and 11.17 , K is complete to B_{i-1} . If there is simplicial vertex of A_{1}(i-1) not adjacent to some simplicial vertex of K , then A_{2}(i+1)\cup B_{i+1}\cup B_{i+2} contains a clique larger than \omega by Lemma 11.18 . This implies that A_{1}(i-1) is a clique. By symmetry A_{1}(i+1) is a clique. Suppose that v\in A_{1}(i+1) has a neighbor a\in A^{1}_{2}(i-2) and a neighbor b\in X . Let c\in A_{1}(i-1) be a neighbor of a that has a neighbor d\in A_{2}(i+1) . By Lemma 11.10 , v is adjacent to and non-adjacent to d , a contradiction. So no vertex in A_{1}(i+1) can have a neighbor in both A^{1}_{2}(i-2) and X . Next we show that X=\emptyset . Suppose not. Let a\in X and b\in A_{1}(i+1) be a neighbor of a . Then b\neq y and bw\notin E(G) . Since A_{1}(i+1) is a clique, b-y-w-a-b is an induced C_{4} . This proves that X=\emptyset . So A_{2}(i-2)=A^{1}_{2}(i-2) or equivalently each vertex in A_{2}(i-2) has a neighbor in A_{1}(i-1) . By symmetry, each vertex in A_{2}(i+1) has a neighbor in A_{1}(i+1) . By Lemmas 6.2 , A_{2}(i-2) is complete to A_{1}(i-1) and A_{2}(i+1) is complete to A_{1}(i+1) . This contradicts that zy\notin E(G) . ∎ Lemma 11.39 . \chi(G)\leq\left\lceil\frac{5}{4}\omega\right\rceil . Proof. By Lemmas 11.37 , 11.12 and 11.35 , A_{3}=\emptyset . By Lemma 11.36 , Y=Y^{\prime}=\emptyset . Then each vertex in A_{5} is universal in G by Lemma 11.7 . So A_{5}=\emptyset by Lemma 2.11 . Since q_{i-2} is not a small vertex, |B_{i+2}|>\left\lceil\frac{\omega}{4}\right\rceil . By symmetry, |B_{i-2}|>\left\lceil\frac{\omega}{4}\right\rceil . It follows from Lemma 11.18 that A_{1}(i-1) and A_{1}(i+1) are cliques. Suppose that X contains a vertex a and X^{\prime} contains a vertex b . Let c\in A_{1}(i+1) be a neighbor of a and d\in A_{1}(i-1) be a neighbor of b . Then ac and bd contradict Lemma 11.10 . So we may assume that X=\emptyset . Equivalently, each vertex in A_{2}(i-2) has a neighbor in A_{1}(i-1) . By Lemma 6.2 , A_{2}(i-2) is complete to A_{1}(i-1) . Let v\in A_{2}(i-2) be a vertex such that |N(v)\cap A_{1}(i+1)| is maximum. If A^{1}_{2}(i+1) contains a vertex u , we show that \{v,u,q_{i}\} is a (1,1) -good subgraph. It suffices to show that every maximum clique M of G contains a vertex in \{v,u,q_{i}\} . Since v is universal in B_{i-2}\cup B_{i-1}\cup A_{1}(i-1)\cup A_{2}(i-2) and u is universal in B_{i+2}\cup B_{i+1}\cup A_{1}(i+1)\cup A_{2}(i+1) , M contains v or u if M\subseteq B_{i-2}\cup B_{i-1}\cup A_{1}(i-1)\cup A_{2}(i-2) or M\subseteq B_{i+2}\cup B_{i+1}\cup A_{1}(i+1)\cup A_{2}(i+1) . For the same reason, M contains q_{i} if M\subseteq B_{i}\cup B_{j} for j\in\{i-1,i+1\} . The remaining cases are the following. \bullet M\subseteq B_{i-2}\cup B_{i+2} . Then |B_{i-2}|\geq\frac{\omega}{2} or |B_{i+2}|\geq\frac{\omega}{2} . If |B_{i-2}|\geq\frac{\omega}{2} , then B_{i-2}\cup B_{i-1}\cup A_{2}(i-2) is a clique of size larger than \omega . If |B_{i+2}|\geq\frac{\omega}{2} , then B_{i+2}\cup B_{i+1}\cup A_{2}(i+1) is a clique of size larger than \omega . \bullet M\subseteq A_{2}(i-2)\cup A_{1}(i+1) . Then v\in M by the choice of v . \bullet M\subseteq A_{2}(i+1)\cup A_{1}(i-1) . Then |A_{2}(i+1)|\geq\frac{\omega}{2} or |A_{1}(i-1)|\geq\frac{\omega}{2} . If |A_{2}(i+1)|\geq\frac{\omega}{2} , then B_{i+2}\cup B_{i+1}\cup A_{2}(i+1) is a clique of size larger than \omega . If |A_{1}(i-1)|\geq\frac{\omega}{2} , then A_{1}(i-1)\cup B_{i-1}\cup A_{2}(i-2) is a clique of size larger than \omega . So A_{2}^{1}(i+1)=\emptyset and thus A_{1}(i+1) is anticomplete to A_{2}(i+1) . Next we give a \left\lceil\frac{5}{4}\omega\right\rceil -coloring of G . Let x=|B_{i-1}| , y=|B_{i-2}| , and z=|A_{2}(i-2)| . Our strategy is to use p=x+y+z-3\left\lceil\frac{\omega}{4}\right\rceil colors to color some vertices of G first and then argue that the remaining vertices can be colored with f=\left\lceil\frac{5}{4}\omega\right\rceil-p colors. Note that \displaystyle f \displaystyle=\left\lceil\frac{5}{4}\omega\right\rceil-(x+y+z-3\left\lceil\frac{\omega}{4}\right\rceil) \displaystyle=\omega+(4\left\lceil\frac{\omega}{4}\right\rceil-(x+y+z)) \displaystyle\geq\omega, since x+y+z\leq\omega . Step 1: Precolor a subgraph. Next we define vertices to be colored with p colors. Let S_{i-1}\subseteq B_{i-1} be the set of x-\left\lceil\frac{\omega}{4}\right\rceil vertices, S_{i-2}\subseteq B_{i-2} be the set of y-\left\lceil\frac{\omega}{4}\right\rceil vertices, S_{z}\subseteq A_{2}(i-2) be the set of z-\left\lceil\frac{\omega}{4}\right\rceil vertices, and S_{i+1}\subseteq B_{i+1} be the set of p vertices with largest neighborhoods in A_{1}(i+1) . Since (x+y+z)-3\left\lceil\frac{\omega}{4}\right\rceil\leq\left\lceil\frac{\omega}{4}\right\rceil<|B_{i+1}| , the choice of S_{i+1} is possible. Since A_{1}(i+1) is anticomplete to A_{2}(i+1) , every vertex in A_{1}(i+1) has at least \left\lceil\frac{\omega}{4}\right\rceil+1 neighbors in B_{i+1} and so S_{i+1} is complete to A_{1}(i+1) . We color S_{i+1} with all colors in [p] , and color all vertices in S_{i-1}\cup S_{i-2}\cup S_{z} with all colors in [p] . Since B_{i+1} is anticomplete to B_{i-1}\cup B_{i-2}\cup A_{2}(i-2) , this coloring is proper. So it suffices to show that G_{1}=G-(S_{i-1}\cup S_{i-2}\cup S_{i+1}\cup S_{z}) is f -colorable. Step 2: Reduce by degeneracy. Let v be a vertex in B_{i}\cup B_{i+2}\cup A_{1}(i+1) . Note that |N_{G_{1}}(v)|<\left\lceil\frac{\omega}{4}\right\rceil+(\omega-\left\lceil\frac{\omega}{4}\right\rceil)\leq f . It follows that G_{1} is f -colorable if and only if G_{1}-(B_{i}\cup B_{i+2}\cup A_{1}(i+1)) is f -colorable. Since G_{1}-(B_{i}\cup B_{i+2}\cup A_{1}(i+1)) is chordal, \chi(G_{1}-(B_{i}\cup B_{i+2}\cup A_{1}(i+1)))\leq\omega\leq f . ∎ 11.2 One non-empty A_{1}(i) The global assumption in this subsection is that A_{1}=A_{1}(i)\neq\emptyset . Suppose that x\in A_{1}(i) . By Proposition 6.1 , the component of A_{1}(i) containing x has a neighbor y\in A_{2}(i+1) . We may assume that xy\in E(G) by renaming the vertex. Lemma 11.40 . B_{i-2} and B_{i+2} are complete. Proof. By Lemma 4.9 , y is complete to B_{i-2}\cup B_{i+2} . We are done by Lemma 4.8 . ∎ Lemma 11.41 . A_{2}(i-2)=\emptyset . By symmetry, A_{2}(i+1)=\emptyset . Proof. Suppose by contradiction that A_{2}(i-2)\neq\emptyset . Let K be a component of A_{2}(i-2) . By Lemma 4.15 (2) and A_{1}=A_{1}(i) , K has no neighbor in A_{1} . By Lemma 4.8 , N(K)\cap(H\cup A_{5}) is a clique. Since G has no clique cutsets, K has a neighbor in A_{3} . By Lemmas 7.1 and 4.18 , N(K)\cap A_{3}\subseteq A_{3}(i-2) or N(K)\cap A_{3}\subseteq A_{3}(i-1) . Suppose first that N(K)\cap A_{3}\subseteq A_{3}(i-2) . Then N(K)\subseteq B_{i-2}\cup B_{i-1}\cup A_{3}(i-2)\cup A_{5} . Each vertex in N(K)\cap A_{5} is universal in N(K) . By Lemma 4.8 , N(K)\cap B_{i-1} and N(K)\cap A^{\prime}_{3}(i-2) are complete. Since G has no clique cutsets, there exist a,b\in N(K)\cap A^{\prime}_{3}(i-2) . By Lemma 7.1 , a,b have a common neighbor in B_{i-1} and so have disjoint neighborhoods in B_{i+2} . Then a-K-b-(N(b)\cap B_{i+2})-B_{i+1}-(N(x)\cap B_{i})-x is an induced sequence, a contradiction. The argument for the case N(K)\cap A_{3}\subseteq A_{3}(i-1) is symmetric. ∎ Lemma 11.42 . A_{3}(i)=\emptyset . Proof. Suppose not. Let t\in A_{3}(i) and b\in B_{i} be a non-neighbor of t . By Lemma 4.12 , we may assume that b and t have a common neighbor c\in B_{i-1} and have disjoint neighborhoods in B_{i+1} . Since t-c-b-(N(b)\cap B_{i+1})-q_{i+2}-y-x is not a bad P_{7} , bx\in E(G) . By symmetry, tx\in E(G) . But this contradicts Lemma 11.5 . ∎ Let S=A_{2}(i-1)\cap N(A_{1}(i)) . Lemma 11.43 . For any vertex in S , it has a neighbor in A_{1}(i)\cap N(A_{2}(i+2)) . Proof. Let s\in S and a\in A_{1}(i) be a neighbor of s . Suppose by contradiction that a\notin A_{1}(i)\cap N(A_{2}(i+2)) . Let b\in B_{i-1} be a neighbor of s . By Lemma 11.6 , a has a neighbor a^{\prime}\in A_{1}(i)\cap N(A_{2}(i+2)) . Since sa^{\prime}\notin E(G) , b-s-a-a^{\prime}-(N(a^{\prime})\cap A_{2}(i+2))-q_{i+2}-q_{i+1} is an induced P_{7} . ∎ Lemma 11.44 . For any component K of A_{2}(i-1)\setminus S , K has two non-adjacent neighbors in A^{\prime}_{3}(i-1) complete to K . Proof. Note that N(K)\subseteq S\cup A_{5}\cup B_{i}\cup A^{\prime}_{3}(i-1) by Lemma 11.42 . We first show that each vertex in N(K)\cap S is universal in N(K) . Suppose by contradiction that s\in N(K)\cap S is not adjacent to t\in N(K) . By Lemma 11.43 , let s^{\prime}\in A_{1}(i)\cap N(A_{2}(i+2)) be a neighbor of s . If s mixes on ab\in K , then a-b-s-s^{\prime}-(N(s^{\prime})\cap A_{2}(i+2))-q_{i+2}-q_{i+1} is an induced P_{7} . So s is complete to K and thus s and t have a common neighbor u\in K . Since G is C_{4} -free, ts^{\prime}\notin E(G) and thus t\notin A_{5} . So t\in S\cup B_{i}\cup B_{i-1}\cup A_{3}(i-1) and then t-u-s-s^{\prime}-(N(s^{\prime})\cap A_{2}(i+2))-q_{i+2}-q_{i+1} is a bad P_{7} . This shows that each vertex in N(K)\cap S is universal in N(K) . This implies that each vertex in N(K)\cap A_{5} is universal in N(K) . If s\in N(K)\cap B_{i} is not adjacent to t\in N(K)\cap A^{\prime}_{3}(i-1) , then t-K-s-q_{i+1}-q_{i+2}-(N(t)\cap B_{i-2})-t is an induced cyclic sequence. So each vertex in N(K)\cap B_{i} is universal in N(K) . Therefore, each vertex in N(K)\setminus A^{\prime}_{3}(i-1) is universal in N(K) . By Lemma 7.2 , each pair of non-adjacent vertices in N(K)\setminus A^{\prime}_{3}(i-1) has a common neighbor in K and so has disjoint neighborhoods in B_{i-2} . If K\cup A^{\prime}_{3}(i-1) contains an induced P_{4}=a-b-c-d with d\in N(K)\cap A^{\prime}_{3}(i-1) and c\in K and a,b\in(N(K)\cap A^{\prime}_{3}(i-1))\cup K , then a-b-c-d-(N(d)\cap B_{i-2})-q_{i+2}-q_{i+1} is an induced P_{7} . So K\cup A^{\prime}_{3}(i-1) contains no such induced P_{4} s. Therefore, all three conditions of Lemma 10.1 are satisfied for K where X=A^{\prime}_{3}(i-1) and Y=B_{i-2} . This completes the proof. ∎ Lemma 11.45 . One of A_{2}(i-1) and A_{2}(i) is empty. Proof. Suppose not. Let v\in A_{2}(i-1) and w\in A_{2}(i) . By Lemma 6.4 (1), A_{1}(i) is anticomplete to A_{2}(i)\cup A_{2}(i-1) . By Lemma 11.44 , v has two non-adjacent neighbors a,b in B_{i-1}\cup A_{3}(i-1) . Since G is C_{4} -free, a,b do not have a common neighbor in B_{i-2} . Then a-v-b-(N(b)\cap B_{i-2})-q_{i+2}-(N(w)\cap B_{i+1})-w is an induced P_{7} . ∎ Lemma 11.46 . Either A_{3}(i+2)=\emptyset or A_{3}(i-2)=\emptyset . Proof. Suppose not. Let s\in A_{3}(i-2) and t\in A_{3}(i+2) . Let a\in B_{i-2} be a non-neighbor of s and b\in B_{i-2} be a non-neighbor of t . By Lemmas 4.12 and 11.40 , there exist vertices c,d\in B_{i-1} such that a-c-d-s is an induced P_{4} and there exist vertices p,q\in B_{i+1} such that b-p-q-t is an induced P_{4} . Since t-q-p-b-a-c-d is not an induced P_{7} , at\in E(G) . Then s-d-c-a-t-q-p is an induced P_{7} . ∎ Let T=N(A_{1}(i))\cap A_{2}(i+2) and T_{0}=A_{2}(i+2)\setminus T . Lemma 11.47 . The following holds for A_{3}(i-2) . (1) A_{3}(i-2) is complete to T . (2) A_{3}(i-2) is complete to B_{i+2} . This implies by C_{4} -freeness that T or A^{\prime}_{3}(i-2) is a clique. Proof. We prove one by one. (1) Let t\in A_{3}(i-2) be non-adjacent to y^{\prime}\in T . By definition, y^{\prime} has a neighbor x^{\prime}\in A_{1}(i) . Since B_{i-2} is complete to B_{i+2} by Lemma 11.40 , t has a non-neighbor in B_{i-1} by Lemma 4.12 and so t mixes on an edge ab\in B_{i-1} . Then a-b-t-(N(t)\cap B_{i+2})-y^{\prime}-x^{\prime}-(N(x^{\prime})\cap B_{i}) is a bad P_{7} . (2) This follows from (1) and Lemma 4.8 . ∎ Next we explore the internal structure of A_{2}(i+2) . Lemma 11.48 . There exists j\in\{i-2,i+2\} such that every component of T_{0} has two non-adjacent neighbors in either T or A^{\prime}_{3}(j) but not in both and it is complete to two non-adjacent vertices in T or A^{\prime}_{3}(j) . Proof. Let K be a component of T_{0} . By symmetry, we may assume that A_{3}(i+2)=\emptyset by Lemma 11.46 . Note that N(K)\subseteq A_{3}(i-2)\cup B_{i-2}\cup T\cup B_{i+2}\cup A_{5} . By Lemma 11.47 , each vertex in N(K)\cap(A_{5}\cup B_{i+2}) is universal in N(K) , and either each vertex in N(K)\cap T or each vertex in N(K)\cap A^{\prime}_{3}(i-2) is universal in N(K) . Suppose first that T is a clique and so each vertex in N(K)\setminus A^{\prime}_{3}(i-1) is universal in N(K) . By Lemma 11.47 (2), each pair of non-adjacent vertices in N(K)\cap A^{\prime}_{3}(i-2) has disjoint neighborhoods in B_{i-1} . If there is an induced P_{4}=a-b-c-d with d\in N(K)\cap A^{\prime}_{3}(i-1) , c\in K and a,b\in(N(K)\cap A^{\prime}_{3}(i-1))\cup K , then a-b-c-d-(N(d)\cap B_{i-1})-q_{i}-q_{i+1} is an induced P_{7} , a contradiction. So all three conditions of Lemma 10.1 are satisfied for K where X=A^{\prime}_{3}(i-2) and Y=B_{i-1} . Therefore, the lemma follows. Now suppose that A^{\prime}_{3}(i-1) is a clique. Then each vertex in N(K)\setminus T is universal in N(K) . For any pair of non-adjacent vertices in N(K)\cap T , they have a common neighbor q_{i+2} and so have disjoint complete neighborhoods in A_{1}(i) by (C_{4},C_{6},C_{7}) -freeness of G . If there is an induced P_{4}=a-b-c-d with d\in N(K)\cap T , c\in K and a,b\in(N(K)\cap T)\cup K , then a-b-c-d-(N(d)\cap A_{1}(i))-q_{i}-q_{i+1} is an induced P_{7} , a contradiction. So all three conditions of Lemma 10.1 are satisfied for K where X=T and Y=A_{1}(i) . Therefore, the lemma follows. ∎ Lemma 11.49 . Each vertex in A_{5} is a universal vertex in G . Proof. We first show that A_{5} is complete to A_{2}(i+2) . By the choice of C , A_{5} is complete to T . By Lemma 11.48 , every component K of T_{0} has two non-adjacent vertices \{a,b\} complete to K where a,b\in T or a,b\in A^{\prime}_{3}(i-2) . Since A_{5} is complete to T\cup A^{\prime}_{3}(i-2) , A_{5} is complete to K by C_{4} -free. Next we show that A_{5} is complete to A_{2}(i)\cup A_{2}(i-1) . Suppose first that s\in A_{2}(i-1) has no neighbor in A_{1}(i) . By Lemma 11.44 , s has two non-adjacent vertices \{a,b\} with a,b\in A^{\prime}_{3}(i-1) . Since G is C_{4} -free, A_{5} is complete to s . Now suppose that s has a neighbor in A_{1}(i) . By Lemma 11.43 , s has a neighbor s^{\prime}\in A_{1}(i) which has a neighbor in A_{2}(i+2) . Then A_{5} is complete to s^{\prime} . Since s-(N(s)\cap B_{i-1})-A_{5}-s^{\prime}-s is not an induced C_{4} , A_{5} is complete to s . This proves that A_{5} is complete to A_{2}(i-1)\cup A_{2}(i) . The lemma now follows from Lemma 11.7 . ∎ So we may assume that A_{5}=\emptyset . Lemma 11.50 . There exists a vertex v\in B_{i-1}\setminus N(A_{3}(i-1)) such that |N(v)\cap B_{i-2}|>\left\lceil\frac{\omega}{4}\right\rceil . By symmetry, There exists a vertex u\in B_{i+1}\setminus N(A_{3}(i+1)) such that |N(u)\cap B_{i+2}|>\left\lceil\frac{\omega}{4}\right\rceil . Proof. If A_{3}(i-1)\neq\emptyset , let v be a vertex in B_{i-1}\setminus N(A_{3}(i-1)) with minimal neighborhood in B_{i} . If A_{3}(i-1)=\emptyset , then let v=q_{i-1} if B_{i-1} is complete to B_{i} , and let v\in B_{i-1} be a vertex with minimal neighborhood in B_{i} if B_{i-1} is not complete to B_{i} . We show that v is anticomplete to A_{3}(i-2) and N(v)\cap(B_{i-1}\cup B_{i}) is a clique. Suppose first that A_{3}(i-1)\neq\emptyset . By Lemma 8.11 , v is anticomplete to A_{3}(i-2) . By Lemma 8.1 (3) and the choice of v , N(v)\cap(B_{i-1}\cup B_{i}) is a clique. If v=q_{i-1} , then the claim holds by the choice of v . If A_{3}(i-1)=\emptyset and B_{i-1} is not complete to B_{i} , then the claim holds by Lemma 8.10 and the choice of v . So N(v)\setminus B_{i-2}\subseteq B_{i-1}\cup B_{i}\cup S\cup(A_{2}(i-1)\setminus S) . Suppose that N(v)\setminus B_{i-2} contains two non-adjacent vertices s and s^{\prime} . Suppose first that s\in S . By Lemma 11.43 , we can choose p\in A_{1}(i)\cap N(A_{2}(i+2)) to be a neighbor of s . Since s^{\prime}-v-s-p-s^{\prime} is not an induced C_{4} , ps^{\prime}\notin E(G) . Then s^{\prime}-v-s-p-(N(p)\cap A_{2}(i+2))-q_{i+2}-q_{i+1} is a bad P_{7} . So s\notin S . Since N(v)\cap(B_{i-1}\cup B_{i}) is a clique, we may assume that s\in A_{2}(i-1)\setminus S . By Lemma 11.44 , s has two non-adjacent neighbors a,b in A^{\prime}_{3}(i-1) . By symmetry, we may assume that a\in A_{3}(i-1) . Since s is a common neighbor of a,b , we have a,b have disjoint neighborhoods in B_{i-2} . Since b-s-a-N(a)\cap B_{i-2}-q_{i+2}-q_{i+1}-B_{i} is not a bad P_{7} , we have a is complete to B_{i} . It follows that a,s^{\prime} and a,v have a common neighbor in B_{i} and so N(a)\cap B_{i-2} is anticomplete to s^{\prime} and v . Since va\notin E(G) by the choice of v , s^{\prime}-v-s-a is an induced P_{4} . Then s^{\prime}-v-s-a-(N(a)\cap B_{i-2})-q_{i+2}-q_{i+1} is an induced P_{7} or C_{7} . So N(v)\setminus B_{i-2} is a clique. Since G has no small vertices, |N(v)\cap B_{i-2}|>\left\lceil\frac{\omega}{4}\right\rceil . ∎ Lemma 11.51 . T_{0} is a clique. Proof. Suppose not. Let K and K^{\prime} be two components of T_{0} . Let v\in K and v^{\prime}\in K^{\prime} . By Lemma 11.48 , K has two non-adjacent neighbors a,b complete to K and K^{\prime} has two non-adjacent neighbors a^{\prime},b^{\prime} complete to K^{\prime} . Suppose first that a,b\in A^{\prime}_{3}(i-2) . Then a^{\prime},b^{\prime}\in A^{\prime}_{3}(i-2) . Since G is C_{4} -free, v^{\prime} is not adjacent to either a or b . If v^{\prime}a\in E(G) , then v^{\prime}-a-v-b-(N(b)\cap B_{i-1})-q_{i}-q_{i+1} is an induced P_{7} . So v^{\prime}a,v^{\prime}b\notin E(G) . By symmetry, va^{\prime},vb^{\prime}\notin E(G) . So a,a^{\prime},b,b^{\prime} are pairwise distinct. If aa^{\prime}\in E(G) , then a^{\prime}b\notin E(G) or a^{\prime}-a-v-b-a^{\prime} is an induced C_{4} , and so a^{\prime}-a-v-b-N_{B_{i-1}}(b)-q_{i}-q_{i+1} is an induced P_{7} . So \{a,a^{\prime},b,b^{\prime}\} is a stable set. Let c,c^{\prime}\in B_{i-1} be a neighbor of a,a^{\prime} , respectively. Then b-v-a-c-c^{\prime}-a^{\prime}-v^{\prime} is an induced P_{7} . Suppose now that a,b\in T . Then a^{\prime},b^{\prime}\in T . Since G is C_{4} -free, v^{\prime} is not adjacent to either a or b . If v^{\prime}a\in E(G) , then v^{\prime}-a-v-b-(N(b)\cap A_{1}(i))-B_{i}-q_{i+1} is an induced P_{7} . So v^{\prime}a,v^{\prime}b\notin E(G) . By symmetry, va^{\prime},vb^{\prime}\notin E(G) . So a,a^{\prime},b,b^{\prime} are pairwise distinct. If aa^{\prime}\in E(G) , then a^{\prime}b\notin E(G) or a^{\prime}-a-v-b-a^{\prime} is an induced C_{4} , and so a^{\prime}-a-v-b-N_{A_{1}(i)}(b)-B_{i}-q_{i+1} is an induced sequence. So \{a,a^{\prime},b,b^{\prime}\} is a stable set. Let c,c^{\prime}\in A_{1}(i) be a neighbor of a,a^{\prime} , respectively. Then b-v-a-c-c^{\prime}-a^{\prime}-v^{\prime} is an induced P_{7} . ∎ Lemma 11.52 . Either T is a clique or A_{3}(i-2)\cup A_{3}(i+2)=\emptyset . Proof. Suppose that T contains two non-adjacent vertices a and b . If t\in A_{3}(i-2) , then t has a non-neighbor s\in B_{i-2} and it follows from Lemma 11.47 that a-s-b-t-a is an induced C_{4} . ∎ Lemma 11.53 . For any non-empty component K of A_{3}(i-2) and any simplicial vertex t\in K , N(t)\setminus B_{i-1} is a clique and so |N(t)\cap B_{i-1}|>\left\lceil\frac{\omega}{4}\right\rceil . Proof. Note that N(t)\setminus B_{i-1}\subseteq K\cup B_{i-2}\cup B_{i+2}\cup T\cup T_{0} . Suppose that N(t)\setminus B_{i-1} contains two non-adjacent vertices v and u . Suppose first that v,u\notin T_{0} . Since B_{i-2}\cup B_{i+2}\cup T is a clique by Lemma 11.52 and N_{K}(t) is a clique, we may assume that v\in K and u\in B_{i-2}\cup B_{i+2}\cup T . Since B_{i+2}\cup T is complete to A_{3}(i-2) by Lemma 11.47 , u\in B_{i-2} . But this contradicts Lemma 8.3 . So we may assume that v\in T_{0} . By Lemma 11.48 , v has two non-adjacent neighbors a,b\in T or A^{\prime}_{3}(i-2) . By Lemma 11.52 , a,b\in A^{\prime}_{3}(i-2) . We may assume that a\in A_{3}(i-2) and t\neq b . If tb\notin E(G) , then u-t-v-b-(N(b)\cap B_{i-1})-q_{i}-q_{i+1} is an induced P_{7} . So tb\in E(G) and so t\neq a . By symmetry, ta\in E(G) . Since t is simplicial in K , b\in B_{i-2} and so b mixes on at\in E(K) . This contradicts Lemma 8.3 . ∎ Lemma 11.54 . |B_{i}|>\left\lceil\frac{\omega}{4}\right\rceil . Proof. By Lemma 11.45 , we may assume that A_{2}(i-1)=\emptyset . We choose a vertex v\in B_{i-1} anticomplete to A_{3}(i-2)\cup A_{3}(i-1) such that |N(v)\cap B_{i-2}| is minimum. Then N(v)\setminus B_{i}\subseteq B_{i-1}\cup B_{i-2} . Suppose that N(v)\setminus B_{i} contains two non-adjacent vertices s,t with s\in B_{i-1} and t\in B_{i-2} . By the choice of v , s has a neighbor r in A_{3}(i-1) or A_{3}(i-2) . By Lemma 8.1 (3), r\in A_{3}(i-2) . By Lemma 11.46 , A_{3}(i+2)=\emptyset . If A_{2}(i)=\emptyset , we are done by symmetry. So A_{2}(i) contains a vertex w . If w has a neighbor w^{\prime}\in A_{1}(i) , then v-s-r-q_{i+2}-N_{B_{i+1}}(w)-w-w^{\prime} is an induced P_{7} . So w is anticomplete to A_{1}(i) . By Lemma 11.44 , w has two non-adjacent neighbors a,b\in A^{\prime}_{3}(i+1) . Then v-s-r-N_{B_{i+2}}(a)-a-w-b is an induced P_{7} . This proves that N(v)\setminus B_{i} is a clique and the lemma follows. ∎ Lemma 11.55 . For any two non-adjacent vertices t_{1},t_{2}\in T , N(t_{1})\cap A_{1}(i) and N(t_{2})\cap A_{1}(i) are disjoint and (N(t_{1})\cap A_{1}(i))\cup(N(t_{2})\cap A_{1}(i))\cup B_{i} is a clique. Proof. Since q_{i+2} is a common neighbor of t_{1} and t_{2} and G is C_{4} -free, N(t_{1})\cap A_{1}(i) and N(t_{2})\cap A_{1}(i) are disjoint. If s_{j}\in N(t_{j})\cap A_{1}(i) are not adjacent, then t_{1}-s_{1}-B_{i}-s_{2}-t_{2}-q_{i+2}-t_{1} is an induced cyclic sequence. So N(t_{1})\cap A_{1}(i) and N(t_{2})\cap A_{1}(i) are complete. Since G is C_{4} -free, N(t_{1})\cap A_{1}(i) and N(t_{2})\cap A_{1}(i) are clique. Now suppose b\in B_{i} has a neighbor s_{1}\in N(t_{1})\cap A_{1}(i) . If b has a non-neighbor s_{2}\in N(t_{2})\cap A_{1}(i) , then b-s_{1}-s_{2}-t_{2}-q_{i+2}-q_{i+1}-b is an induced C_{6} . So b is complete to N(t_{2})\cap A_{1}(i) . It follows that b is complete to N(t_{1})\cap A_{1}(i) . So each vertex in B_{i} is either complete or anticomplete to (N(t_{1})\cap A_{1}(i))\cup(N(t_{2})\cap A_{1}(i)) . If there is a vertex b\in B_{i} anticomplete to (N(t_{1})\cap A_{1}(i))\cup(N(t_{2})\cap A_{1}(i)) , then s_{1}-s_{2}-t_{2}-q_{i-2}-q_{i-1}-b-q_{i+1} is an induced P_{7} . This proves that B_{i} is complete to (N(t_{1})\cap A_{1}(i))\cup(N(t_{2})\cap A_{1}(i)) . ∎ Lemma 11.56 . T is a blowup of P_{3} . This implies that \alpha(T)\leq 2 . Proof. If T is a clique, then there is nothing to prove. So T is not a clique and so A_{3}(i-2)\cup A_{3}(i+2)=\emptyset by Lemma 11.52 . (1) Each vertex in B_{i-2}\cup B_{i+2} is universal in T\cup T_{0}\cup B_{i-2}\cup B_{i+2} . By Lemmas 4.8 and 4.9 , B_{i+2}\cup B_{i-2} is a clique complete to T . If T_{0} is empty, we are done. So T_{0} is not empty. By Lemma 11.48 , there are two non-adjacent vertices a,b\in T complete to T_{0} . Since G is C_{4} -free, B_{i-2}\cup B_{i+2} is universal in T\cup T_{0}\cup B_{i-2}\cup B_{i+2} . This proves (1). (2) T cannot contain three pairwise non-adjacent vertices simplicial in T\cup T_{0}\cup B_{i-2}\cup B_{i+2} . Suppose that T contains three pairwise non-adjacent vertices v_{1},v_{2},v_{3} simplicial in T\cup T_{0}\cup B_{i-2}\cup B_{i+2} . For j=1,2,3 , we have |N(v_{j})\cap A_{1}(i)|>\left\lceil\frac{\omega}{4}\right\rceil . It follows from Lemmas 11.55 and 11.54 that (N(v_{1})\cap A_{1}(i))\cup(N(v_{2})\cap A_{1}(i))\cup(N(v_{3})\cap A_{1}(i))\cup B_{i} is a clique of size larger than \omega , a contradiction. This proves (2). (3) Let T^{\prime}\subseteq T be cutset of G[T] such that each vertex in T^{\prime} is universal in A_{2}(i+2)\cup B_{i-2}\cup B_{i+2} . Then for each component T^{\prime} -component K of G[T] , there is a vertex v_{K} in K\setminus T^{\prime} that is simplicial in K\cup T_{0}\cup B_{i-2}\cup B_{i+2} . If K\cup T_{0}\cup B_{i-2}\cup B_{i+2}\cup T^{\prime} is a clique, there is nothing to prove. If K\cup T_{0}\cup B_{i-2}\cup B_{i+2} is not a clique, then K\cup T_{0}\cup B_{i-2}\cup B_{i+2} contains two non-adjacent simplicial vertices s and s^{\prime} by Lemma 11.4 . Since each vertex in B_{i-2}\cup B_{i+2}\cup T^{\prime} is universal in K\cup T_{0}\cup B_{i-2}\cup B_{i+2} , we have s,s^{\prime}\in(K\setminus T^{\prime})\cup T_{0} . Since T_{0} is a clique by Lemma 11.51 , one of s,s^{\prime} is in K\setminus T^{\prime} . This proves (3). By (2) and (3), T has at most two components. Next we show the following. (4) Let L_{1},L_{2}\subseteq T be two connected subgraphs such that L_{1} and L_{2} are disjoint and anticomplete, and each vertex in T\setminus(L_{1}\cup L_{2}) is universal in A_{2}(i+2)\cup B_{i-2}\cup B_{i+2} . Then L_{1} and L_{2} are cliques. Let T^{\prime}=T\setminus(L_{1}\cup L_{2}) . Suppose first that T_{0}=\emptyset . By the assumption of (4), every simplicial vertex of L_{i} is simplicial in A_{2}(i+2)\cup B_{i-2}\cup B_{i+2} . By (2), L_{1} and L_{2} are cliques. So we assume that T_{0}\neq\emptyset . By Lemma 11.48 , there exist two non-adjacent vertices a,b\in T complete to T_{0} . Since each vertex in T^{\prime} is universal in A_{2}(i+2)\cup B_{i-2}\cup B_{i+2} , a,b\in L_{1}\cup L_{2} . Suppose first that a,b belong to the same L_{i} , say L_{1} . Since L_{1}\cup T^{\prime}\cup T_{0} is not a clique, it contains two non-adjacent vertices s,s^{\prime} simplicial in L_{1}\cup T^{\prime}\cup T_{0} by Lemmas 11.4 and 2.4 . By our assumption on T^{\prime} , s,s^{\prime}\in L_{1} . By L_{1} is anticomplete to L_{2} and (1), s,s^{\prime} are simplicial in A_{2}(i+2)\cup B_{i-2}\cup B_{i+2} . By (3), L_{2} contains a vertex t simplicial in A_{2}(i+2)\cup B_{i-2}\cup B_{i+2} . Then s,s^{\prime},t contradicts (2). So we assume that a\in L_{1} and b\in L_{2} . Let u\in L_{1} be a simplicial vertex of L_{1} . Choose a simplicial vertex s\in N_{L_{1}}[u] with |N(s)\cap T_{0}| is minimum. Suppose that s is not simplicial in T^{\prime}\cup L_{1}\cup T_{0} . Then s has two non-adjacent vertices p\in L_{1} and q\in T_{0} . If p is not simplicial in L_{1} , then p has a neighbor r\in L_{1} not adjacent to s . It follows that r-p-s-q-b-(N(b)\cap A_{1}(i))-B_{i} is an induced sequence. So p is simplicial in L_{1} . By the choice of s , p has a neighbor q^{\prime}\in T_{0} not adjacent to s . It follows that p-q^{\prime}-q-s-p is an induced C_{4} . So s is simplicial in T^{\prime}\cup T_{0}\cup L_{1} . By Lemma 2.6 , (2) and (3), L_{1} and L_{2} are cliques. This proves (4). If T has two components, then each component is a clique by applying (4) with L_{1} and L_{2} being the two components of T . So T is connected. Let x,y\in T with xy\notin E(G) . If p\in N(x)\cap A_{1}(i) and q\in N(y)\cap A_{1}(i) are not adjacent, then p-B_{i}-q-y-B_{i+2}-x-p is an induced cyclic sequence. If there is an induced P_{5}=a-b-c-d-e with a,b,c,d\in T and e\in A_{1}(i) , then a-b-c-d-e-N_{B_{i}}(e)-q_{i+1} is an induced P_{7} . By Lemma 4.9 , q_{i+2} is complete to T . So all three conditions of Lemma 2.8 are satisfied for X=T , Y=A_{1}(i) and u=q_{i+2} . By Lemma 2.8 , T is P_{4} -free and so T can be partitioned into two subsets T_{1} and T_{2} such that T_{1} and T_{2} are complete. Since G is C_{4} -free, we may assume that T_{1} is a clique. We choose such a partition (T_{1},T_{2}) so that |T_{1}| is maximum. It follows that T_{2} is disconnected. Since A_{3}(i-2)\cup A_{3}(i+2)=\emptyset and T_{1} is a universal clique in T , T_{1} is complete to T_{0} . It follows that each component of T_{2} has a vertex simplicial in A_{2}(i+2)\cup B_{i-2}\cup B_{i+2} . By (2), T_{2} has exactly two components. By applying (4) to the two components of T_{2} , each component of T_{2} is a clique. So T is a blowup of P_{3} . ∎ Lemma 11.57 . T is complete to T_{0} . Proof. If T_{0}=\emptyset , there is nothing to prove. Let us assume that T_{0} is not empty. By Lemmas 11.51 and 11.48 , T_{0} is complete to two non-adjacent vertices a,b where a,b\in A^{\prime}_{3}(i-2) or a,b\in T . If a,b\in A^{\prime}_{3}(i-2) , then T\cup T_{0} is complete to \{a,b\} by Lemma 11.47 and so T\cup T_{0} is a clique. So a,b\in T . Suppose for a contradiction that c\in T is not adjacent to v\in T_{0} . Then c\neq a,b . Since G is C_{4} -free, c is not adjacent to a or b . If c is adjacent to a , then c-a-v-b-(N(b)\cap A_{1}(i))-B_{i}-q_{i+1} is an induced sequence. So c is anticomplete to \{a,b\} . Then \{a,b,c\} is an independent set in T , contradicting Lemma 11.56 . ∎ Lemma 11.58 . T_{0} is complete to B_{i-2}\cup B_{i+2} . Proof. If T_{0} is empty, there is nothing to prove. Let v\in T_{0} . By Lemma 11.48 , there are two vertices a,b complete to T_{0} where a,b\in A^{\prime}_{3}(i-2) or a,b\in T . If a,b\in T , then B_{i-2}\cup B_{i+2}\cup T_{0} is complete to \{a,b\} and so we are done. So assume that a,b\in A^{\prime}_{3}(i-2) . By Lemmas 11.40 and 11.47 , T_{0} is complete to B_{i+2} . Suppose for a contradiction that u\in B_{i-2} is not adjacent to v . Then u\neq a,b . Since G is C_{4} -free, u is not adjacent to a or b . If u is adjacent to a , then u-a-v-b-(N(b)\cap B_{i-1})-q_{i}-q_{i+1} is an induced P_{7} . So u is anticomplete to \{a,b\} . In particular, a,b\in A_{3}(i-2) . Choose a vertex w\in B_{i-2} anticomplete to A_{3}(i-2) . By Lemma 8.11 , w is anticomplete to A_{3}(i-1) . By Lemma 11.46 and A_{3}(i-2)\neq\emptyset , we have A_{3}(i+2)=\emptyset . By Lemma 11.41 , w has no neighbor in A_{2}\setminus A_{2}(i+2) . Then N(w)\setminus B_{i-1}\subseteq B_{i-2}\cup B_{i+2}\cup T\cup T_{0} . As T_{0}\cup B_{i+2} is complete to \{a,b\} , T_{0}\cup B_{i+2} is a clique. Suppose that v^{\prime},u\in N(w)\setminus B_{i-1} are not adjacent. By Lemma 11.57 , u\in B_{i-2} and v^{\prime}\in T_{0} and so u-w-v^{\prime}-a-N_{B_{i-1}}(a)-q_{i}-q_{i+1} is an induced P_{7} . This proves that N(w)\setminus B_{i-1} is a clique and so |N(w)\cap B_{i-1}|>\left\lceil\frac{\omega}{4}\right\rceil . Since a,b\in A_{3}(i-2) are not adjacent, there exist two non-adjacent vertices t_{1},t_{2}\in A_{3}(i-2) such that |N(t_{j})\cap B_{i-1}|>\left\lceil\frac{\omega}{4}\right\rceil for j=1,2 by Lemma 11.53 . By Lemmas 11.54 and 8.10 , B_{i}\cup(N(t_{1})\cap B_{i-1})\cup(N(t_{2})\cap B_{i-1})\cup(N(w)\cap B_{i-1}) is a clique of size larger than \omega . ∎ Lemma 11.59 . A_{1}(i)\setminus N(A_{2}(i+2))=\emptyset . In other words, each vertex in A_{1}(i) has a neighbor in A_{2}(i+2) . Proof. Suppose not. Let X=N(A_{2}(i+2))\cap A_{1}(i) . We first show that A_{1}(i)\setminus X is a clique. Suppose for a contradiction that K_{1},K_{2} are two components of A_{1}(i)\setminus X . By Lemma 11.6 , there are two non-adjacent vertices a_{1},b_{1}\in X complete to K_{1} and two non-adjacent vertices a_{2},b_{2}\in X complete to K_{2} . Let v_{1} be a vertex in K_{1} . Since v_{1}\notin K_{2} , v_{1} is not complete to \{a_{2},b_{2}\} . If v_{1} is adjacent to a_{2} , then v_{1}-a_{2}-K_{2}-b_{2}-(N(b_{2})\cap T)-q_{i+2}-q_{i+1} is an induced P_{7} . This proves that K_{1} is anticomplete to \{a_{2},b_{2}\} . By symmetry, K_{2} is anticomplete to \{a_{1},b_{1}\} . It follows that a_{1},b_{1},a_{2},b_{2} are four pairwise distinct vertices. Let b_{1}^{\prime}\in N(b_{1})\cap T and b_{2}^{\prime}\in N(b_{2})\cap T . Next, we show that \{a_{1},b_{1},a_{2},b_{2}\} is a stable set. Suppose for a contradiction that a_{1}a_{2}\in E(G) . Since a_{1} is anticomplete to K_{2} and a_{2} is anticomplete to K_{1} , we have a_{1}b_{2},a_{2}b_{1}\notin E(G) . Since a_{1}-a_{2}-K_{2}-b_{2}-N_{T}(b_{2})-q_{i+2}-q_{i+1} is not an induced P_{7} , a_{1} is complete to N(b_{2})\cap T . By symmetry, a_{2} is complete to N(b_{1})\cap T . Then a_{1}b_{2}^{\prime},a_{2}b_{1}^{\prime}\in E(G) . Since a_{1},b_{1} have a common neighbor in K_{1} , the neighborhoods of a_{1} and b_{1} in T are disjoint and complete cliques and so b_{1}^{\prime}b_{2}^{\prime}\in E(G) . It follows that a_{1}-a_{2}-b_{1}^{\prime}-b_{2}^{\prime}-a_{1} is an induced C_{4} . This proves that \{a_{1},b_{1},a_{2},b_{2}\} is stable. If b_{1}^{\prime}=b_{2}^{\prime} , then a_{1},a_{2} are not adjacent to b_{1}^{\prime} since G is C_{4} -free. Then a_{1}-K_{1}-b_{1}-b_{1}^{\prime}-b_{2}-K_{2}-a_{2} is an induced P_{7} . Therefore, each pair of vertices from \{a_{1},b_{1},a_{2},b_{2}\} has no common neighbors in T . Then a_{1}-K_{1}-b_{1}-\{b_{1}^{\prime},q_{i+2},b_{2}^{\prime}\}-b_{2}-K_{2}-a_{2} is an induced sequence. This proves that A_{1}(i)\setminus X is connected. By Lemma 11.6 , A_{1}(i)\setminus X is a clique. Let u\in A_{1}(i)\setminus X . By Lemma 11.6 , there are two non-adjacent vertices a,b\in X complete to A_{1}(i)\setminus X . If v\in B_{i} that is not adjacent to a , then b-u-a-(N(a)\cap T)-q_{i+2}-q_{i+1}-v is a bad P_{7} . So a,b are complete to B_{i}\cup(A_{1}(i)\setminus X) and B_{i}\cup(A_{1}(i)\setminus X) is a clique. Suppose that v\in N(a)\cap X is not adjacent to v^{\prime}\in A_{1}(i)\setminus X . It follows that v^{\prime}b\in E(G) and so vb\notin E(G) . Since v and b have a common neighbor in B_{i} , v-a-v^{\prime}-b-(N(b)\cap T)-q_{i+2}-q_{i+1} is an induced P_{7} . So N(a)\cap X is complete to A_{1}(i)\setminus X . By symmetry, \{v,b\} is complete to B_{i} . So N(a)\cap X is complete to B_{i}\cup(A_{1}(i)\setminus X) . We choose a non-adjacent pair \{a_{0},b_{0}\} of vertices in X complete to A_{1}(i)\setminus X such that |N(a_{0})\cap X|+|N(b_{0})\cap X| is minimum. Next we show that N(a_{0})\setminus T is a clique. If a_{0} has a neighbor t\in A_{3}(i-1) , then a_{0}-t-(N(t)\cap B_{i-2})-(N(a_{0})\cap T)-a_{0} is an induced C_{4} . So a_{0} is anticomplete to A_{3}(i-1)\cup A_{3}(i+1) . By Lemma 11.45 , we may assume that A_{2}(i)=\emptyset and so N(a)\setminus T\subseteq B_{i}\cup A_{1}(i)\cup A_{2}(i-1) . Let s,s^{\prime} be two non-adjacent neighbors of a_{0} in B_{i}\cup A_{1}(i)\cup A_{2}(i-1) . By Proposition 6.1 , N(a_{0})\cap(B_{i}\cup A_{2}(i-1)) is a clique and so we may assume that s\in A_{1}(i) . By Lemma 6.2 , s^{\prime}\notin A_{2}(i-1) and so s^{\prime}\in B_{i}\cup A_{1}(i) . Since B_{i}\cup(A_{1}(i)\setminus X) is a clique, we may assume that s\in X . Since s is complete to B_{i}\cup(A_{1}(i)\setminus X) , we have s^{\prime}\in X . Since G is C_{4} -free, b_{0} is not adjacent to s or s^{\prime} , say sb_{0}\notin E(G) . By the choice of \{a_{0},b_{0}\} , s has a neighbor s^{\prime\prime}\in X with a_{0}s^{\prime\prime}\notin E(G) . Applying the claim N(a)\cap X is complete to A_{1}(i)\setminus X to the pair \{s,b_{0}\} , we have s^{\prime\prime} is complete to A_{1}(i)\setminus X . So s^{\prime\prime}-s-a_{0}-s^{\prime} is induced P_{4} , which contradicts Lemma 2.8 . So N(a_{0})\setminus T is a clique and thus |N(a_{0})\cap T|>\left\lceil\frac{\omega}{4}\right\rceil . By symmetry, |N(b_{0})\cap T|>\left\lceil\frac{\omega}{4}\right\rceil . Since a_{0},b_{0} have a common neighbor in A_{1}(i)\setminus X , N(a_{0})\cap T and N(b_{0})\cap T are disjoint and complete cliques. By Lemma 11.50 , (N(a_{0})\cap T)\cup(N(b_{0})\cap T)\cup B_{i+2}\cup B_{i-2} is a clique of size larger than \omega , a contradiction. ∎ Lemma 11.60 . A_{1}(i) is P_{4} -free. Proof. Suppose by contradiction that a-b-c-d is an induced P_{4} in A_{1}(i) . By Lemma 6.5 , N(a)\cap B_{i},N(d)\cap B_{i}\subseteq N(b)\cap B_{i}=N(c)\cap B_{i} . Suppose first that a and d have no common neighbors in B_{i} . Since G is (C_{6},C_{7}) -free, a and d have neighborhood containment in A_{2}(i+2) , say N(a)\cap T\subseteq N(d)\cap T . By Lemma 11.59 , there is a vertex s\in N(a)\cap A_{2}(i+2) . Then sd\in E(G) . Since b,d has a common neighbor u\in B_{i} , sb\notin E(G) and then u-b-a-s-q_{i+2}-q_{i+1}-u is an induced C_{6} . So a,d have a common neighbor s\in B_{i} and this vertex is complete to a-b-c-d . But this contradicts Lemma 2.8 . ∎ Lemma 11.61 . A_{1}(i) is anticomplete to A_{3}(i-1)\cup A_{3}(i+1) . This implies that A_{1}(i) is anticomplete to A_{3} . Proof. Let t\in A_{3}(i-1) be adjacent to a\in A_{1}(i) . Let b\in B_{i-1} be a non-neighbor of t . By Lemma 11.59 , a has a neighbor c\in T . Then b-q_{i-1}-t-a-c-q_{i+2}-q_{i+1} is an induced P_{7} . ∎ Lemma 11.62 . For a component K of A_{3}(i-1) , let s be a (K,B_{i}) -minimal simplicial vertex. Then N(s)\setminus B_{i-2} is a clique and so |N(s)\cap B_{i-2}|>\left\lceil\frac{\omega}{4}\right\rceil . Proof. By Lemma 11.61 , N(s)\setminus B_{i-2}\subseteq B_{i}\cup B_{i-1}\cup A_{3}(i-1)\cup A_{2}(i-1) . Suppose that v,v^{\prime}\in N(s)\setminus B_{i-2} are not adjacent. Suppose first that v,v^{\prime}\notin A_{2}(i-1) . By Lemmas 8.1 and 8.3 , v\in K and v^{\prime}\in B_{i} . By Lemma 2.7 , there is an induced P_{4}=a-b-s-c with a,b\in K and c\in B_{i} . Then a-b-s-c-q_{i+1}-q_{i+2}-y is an induced P_{7} . Now we may assume that v\in A_{2}(i-1) and v^{\prime}\in B_{i}\cup B_{i-1}\cup A_{3}(i-1)\cup A_{2}(i-1) . Suppose first that v has a neighbor u\in A_{1}(i) . By Lemma 11.59 , u has a neighbor w\in T . By Lemma 11.61 , su\notin E(G) . It follows that v^{\prime}-s-v-u-w-q_{i+2}-q_{i+1} is an induced P_{7} . So v has no neighbors in A_{1}(i) . By Lemma 11.44 , v has non-adjacent vertices t_{1},t_{2}\in A^{\prime}_{3}(i-1) . Since t_{1}-v-t_{2}-(N(t)\cap B_{i-2})-q_{i+2}-q_{i+1}-B_{i} is not a bad P_{7} , t_{2} is complete to B_{i} . By symmetry, t_{1} is complete to B_{i} . We may assume that s\neq t_{1} . Note that both pairs \{v^{\prime},t_{1}\} and \{s,t_{1}\} have a common neighbor in B_{i} and so have disjoint neighborhoods in B_{i-2} . If st_{1}\notin E(G) , then v^{\prime}-s-v-t_{1}-(N(t_{1})\cap B_{i-2})-q_{i+2}-q_{i+1} is an induced P_{7} . So st_{1}\in E(G) . Since t_{2}t_{1}\notin E(G) , we have s\neq t_{2} . By symmetry, st_{2}\in E(G) . Since s is simplicial in K , we may assume that t_{1}\in B_{i-1} and t_{2}\in A_{3}(i-1) . Now t_{1}\in B_{i-1} mixes on an edge in A_{3}(i-1) , which contradicts Lemma 8.3 . This proves that N(s)\setminus B_{i-2} is a clique and so |N(s)\cap B_{i-2}|>\left\lceil\frac{\omega}{4}\right\rceil . ∎ Lemma 11.63 . A_{3}(i-1) is anticomplete to A_{2}(i-1) . By symmetry, A_{3}(i+1) is anticomplete to A_{2}(i) . Proof. Suppose for a contradiction that t\in A_{3}(i-1) is adjacent to v\in A_{2}(i-1) . We first show that t is complete to B_{i} . Suppose first that v has a neighbor u\in A_{1}(i) . By Lemma 11.61 , ut\notin E(G) . Since u-v-t-(N(t)\cap B_{i-2})-q_{i+2}-q_{i+1}-B_{i} is not a bad P_{7} , t is complete to B_{i} . Now suppose that v has no neighbors in A_{1}(i) . By Lemma 11.44 , v has two non-adjacent neighbors t_{1},t_{2}\in A_{3}^{\prime}(i-1) . Since t_{1}-v-t_{2}-(N(t_{2})\cap B_{i-2})-q_{i+2}-q_{i+1}-B_{i} is not a bad P_{7} , we have t_{2} is complete to B_{i} . By symmetry, t_{1} is complete to B_{i} . Since G is C_{4} -free, v is complete to B_{i} . By Lemma 4.8 , t is complete to B_{i} . Let K be the component of A_{3}(i-1) containing t and s\in K be a (K,B_{i}) -minimal simplicial vertex. By Lemma 11.62 , |N(s)\cap B_{i-2}|>\left\lceil\frac{\omega}{4}\right\rceil . By Lemma 11.50 , there exists w\in B_{i-1} be anticomplete to A_{3}(i-1) with |N(w)\cap B_{i-2}|>\left\lceil\frac{\omega}{4}\right\rceil . If N(w)\cap B_{i-2} and N(s)\cap B_{i-2} are disjoint, then |B_{i-2}|>\frac{\omega}{2} . By Lemma 11.8 , |T|>\left\lceil\frac{\omega}{4}\right\rceil . By Lemma 11.50 , B_{i-2}\cup B_{i+2}\cup T is a clique of size larger than \omega . So s and w have a common neighbor in B_{i-2} and thus they have disjoint neighborhoods in B_{i} . Recall that t is complete to B_{i} . So s\neq t and thus sv\notin E(G) . Let P=v_{0}-v_{1}-\cdots-v_{k}-v be a shortest path such that v_{0}=s and v_{i}\in K for each 1\leq i\leq k . We show that each v_{i} with 1\leq i\leq k is complete to B_{i} . We have proved that v_{k} is complete to B_{i} . Suppose now that v_{i} is complete to B_{i} for 2\leq i\leq k . If v_{i-1} has a non-neighbor b\in B_{i} , then v_{i-2}-v_{i-1}-v_{i}-b-q_{i+1}-q_{i+2}-y is an induced P_{7} . So v_{i-1} is complete to B_{i} . Therefore, v_{1} is complete to B_{i} . By Lemma 8.2 (3), N(s)\cap B_{i-2}\subseteq N(v_{1})\cap B_{i-2} . By Lemma 4.10 , w is anticomplete to N(v_{1})\cap B_{i-2} and so anticomplete to N(s)\cap B_{i-2} . This implies that N(w)\cap B_{i-2} and N(s)\cap B_{i-2} are disjoint, a contradiction. ∎ Lemma 11.64 . A_{2}(i-1) is a clique complete to B_{i-1}\cup B_{i} . By symmetry, A_{2}(i) is a clique complete to B_{i+1}\cup B_{i} . Proof. By Lemmas 11.63 and Lemma 11.44 , each vertex in A_{2}(i-1) has a neighbor in A_{1}(i) . Suppose that v\in A_{2}(i-1) mixes on an edge ab\in B_{i-1} . Let u\in A_{1}(i) be a neighbor of v and w\in A_{2}(i+2) be a neighbor of u . Then a-b-v-u-w-q_{i+2}-q_{i+1} is an induced P_{7} . So v is complete to B_{i-1} . Suppose that v is not adjacent to v^{\prime}\in A_{2}(i-1)\cup B_{i} . Then v^{\prime}-q_{i-1}-v-u-w-q_{i+2}-q_{i+1} is a bad P_{7} . This proves the lemma. ∎ Lemma 11.65 . If A_{2}(i)\cup A_{2}(i-1)\neq\emptyset , then A_{1}(i) is a clique. Proof. By symmetry, assume that A_{2}(i) contains a vertex b . By Lemmas 11.63 and 11.44 , b has a neighbor a\in A_{1}(i) . By Lemma 6.2 , b is complete to the component K of A_{1}(i) containing a . Suppose that K is not a clique. By Lemmas 11.3 , 2.6 and 11.8 , there exist two non-adjacent simplicial vertices s_{1},s_{2} of K with |N(s_{j})\cap T|>\left\lceil\frac{\omega}{4}\right\rceil for j=1,2 . Since s_{1} and s_{2} have a common b\in A_{2}(i) , N(s_{1})\cap T and N(s_{2})\cap T are disjoint and complete. By Lemma 11.50 , T\cup B_{i-2}\cup B_{i+2} is a clique of size larger than \omega . So K is a clique. If v\in K is not adjacent to u\in B_{i} , then u-b-v-N_{T}(v)-q_{i-2}-q_{i-1}-u is an induced C_{6} by Lemma 11.64 . So K is complete to B_{i} . If A_{1}(i)\neq K , then there exist two non-adjacent simplicial vertices s_{2}\in A_{1}(i)\setminus K and s_{1}\in K with |N(s_{j})\cap T|>\left\lceil\frac{\omega}{4}\right\rceil for j=1,2 by Lemma 11.8 . Since s_{1} is complete to K , s_{1} and s_{2} hava a common neighbor in B_{i} and so N(s_{1})\cap T and N(s_{2})\cap T are disjoint and complete. Then T\cup B_{i-2}\cup B_{i+2} is a clique of size larger than \omega . ∎ 11.2.1 T is not a clique In this subsection, T is not a clique. Let t_{1},t_{2}\in T be two non-adjacent simplicial vertices of T . Lemma 11.66 . For j=1,2 , N(t_{j})\setminus A_{1}(i) is a clique and so K_{t_{j}}=N(t_{j})\cap A_{1}(i) has size >\left\lceil\frac{\omega}{4}\right\rceil . Moreover, K_{t_{1}} and K_{t_{2}} are disjoint and complete. Proof. By Lemma 11.52 , N(t_{j})\setminus A_{1}(i)\subseteq B_{i-2}\cup B_{i+2}\cup A_{2}(i+2) . By Lemmas 11.51 , 11.57 , 11.58 and 4.9 , N(t_{j})\setminus A_{1}(i) is a clique. ∎ Lemma 11.67 . A_{1}(i) is a clique. Proof. Suppose not. By Lemma 11.65 , A_{2}(i-1)=A_{2}(i)=\emptyset . Let a^{\prime}_{1},a^{\prime}_{2} be two non-adjacent simplicial vertices of A_{1}(i) . Choose a_{j}\in N_{A_{1}(i)}[a^{\prime}_{j}] simplicial in A_{1}(i) with minimal neighborhood in B_{i} and with minimal neighborhood in T for j=1,2 . By Lemma 11.8 , N(a_{j})\setminus T is a clique and so |N(a_{j})\cap T|>\left\lceil\frac{\omega}{4}\right\rceil . If a_{1} and a_{2} have a common neighbor in B_{i} , then N(a_{1})\cap T and N(a_{2})\cap T are disjoint and complete sine G is (C_{4},C_{6}) -free, and so (N(a_{1})\cap T)\cup(N(a_{2})\cap T)\cup B_{i-2}\cup B_{i+2} is a clique of size larger than \omega . So a_{1} and a_{2} have disjoint neighborhoods in B_{i} . Next we show that N(a_{j})\setminus B_{i} is a clique. Note that N(a_{j})\setminus B_{i}\subseteq A_{1}(i)\cup T . Suppose that s,s^{\prime} are two non-adjacent vertices in N(a_{j})\setminus B_{i} . Then s\in A_{1}(i) and s^{\prime}\in T . Suppose first that s is simplicial in A_{1}(i) . By the minimality of a_{1} , s has a neighbor w\in T\cup B_{i} that is not adjacent to a_{1} . If w\in B_{i} , then w-s-a_{1}-s^{\prime}-q_{i+2}-q_{i+1}-w is an induced C_{6} . So w\in T and N(s)\cap B_{i}\subseteq N(a_{1})\cap B_{i} . Then s-a_{1}-s^{\prime}-q_{i+2}-q_{i+1}-N_{B_{i}}(a_{2})-q_{i-1} is an induced P_{7} . So s is no simplicial in A_{1}(i) . Let w be a neighbor of s in A_{1}(i) that is not adjacent to a_{1} . Then w-s-a_{1}-s^{\prime}-q_{i+2}-q_{i+1}-N_{B_{i}}(a_{2}) is a bad P_{7} . So N(a_{1})\setminus B_{i} is a clique and so |N(a_{1})\cap B_{i}|>\left\lceil\frac{\omega}{4}\right\rceil . By symmetry, |N(a_{2})\cap B_{i}|>\left\lceil\frac{\omega}{4}\right\rceil . By Lemmas 11.55 and 11.66 , (N(a_{1})\cap B_{i})\cup(N(a_{2})\cap B_{i})\cup K_{t_{1}}\cup K_{t_{2}} is a clique of size larger than \omega . ∎ Lemma 11.68 . A_{3}(i+1)=A_{3}(i-1)=\emptyset Proof. By symmetry, we prove that A_{3}(i+1)=\emptyset . Suppose not. By Lemma 8.14 , there exists a vertex v\in A_{3}(i+1) such that |N(v)\cap B_{i}|,|N(v)\cap B_{i+2}|>\left\lceil\frac{\omega}{4}\right\rceil . Let w\in B_{i+1}\setminus N(A_{3}(i+1)) with minimal neighborhood in B_{i}\cup B_{i+2} . Suppose first that w and v have disjoint neighborhoods in B_{i+2} . By Lemma 11.50 , |N(w)\cap B_{i+2}|>\left\lceil\frac{\omega}{4}\right\rceil and so (N(v)\cap B_{i+2})\cup(N(w)\cap B_{i+2})\cup B_{i-2}\cup T is a clique of size larger than \omega . So w and v have disjoint neighborhoods in B_{i} and so B_{i} is not complete to B_{i+1} . By Lemma 11.64 , A_{2}(i)=\emptyset . By Lemma 8.15 , |N(w)\cap B_{i}|>\left\lceil\frac{\omega}{4}\right\rceil . Then (N(v)\cap B_{i})\cup(N(w)\cap B_{i})\cup K_{t_{1}}\cup K_{t_{2}} is a clique of size larger than \omega by Lemma 8.10 . ∎ Next, we give a lemma for reducing vertices of small degrees when we color G . This lemma will also be used in Section 12 . Lemma 11.69 . Let X_{1},X_{2},X_{3},X^{*} be four disjoint cliques of G with |X_{1}|>\left\lceil\frac{\omega}{4}\right\rceil , |X_{2}|\leq\frac{3}{4}\omega and |X_{3}|=\left\lceil\frac{\omega}{4}\right\rceil such that • X_{1}\cup X^{*} is anticomplete to X_{3} , • every two vertices in X_{1}\cup X^{*} have comparable neighborhoods in X_{2} , • for every v\in X^{*} and every vertex u\in X_{1} , N_{X_{2}}(v)\subseteq N_{X_{2}}(u) , • P\subseteq X_{1} is the set of p vertices with largest neighborhoods in X_{2} with p\in\left[\left\lceil\frac{\omega}{4}\right\rceil\right] , and • F is an induced subgraph of G with F\cap P=\emptyset . If a vertex v\in F\cap X_{2} satisfies N_{F}(v)\subseteq X_{1}\cup X_{2}\cup X_{3}\cup X^{*} and N(v)\cap(X_{1}\cup X_{2}\cup X^{*}) is a clique of G , then d_{F}(v)<\left\lceil\frac{5}{4}\omega\right\rceil-p . Proof. If N_{F_{1}}(v)\cap X_{1}\neq\emptyset , then (N_{F_{1}}(v)\cap(X_{1}\cup X_{2}\cup X^{*}))\cup P is also a clique of G by the definition of P . This implies that |N_{F_{1}}(v)\cap(X_{1}\cup X_{2}\cup X^{*})|\leq\omega-p or |N_{F_{1}}(v)\cap(X_{1}\cup X_{2}\cup X^{*})|=|N_{F_{1}}(v)\cap X_{2}|<\frac{3}{4}\omega . It follows that \displaystyle|N_{F_{1}}(v)| \displaystyle=|N_{F_{1}}(v)\cap X_{3}|+|N_{F_{1}}(v)\cap(X_{1}\cup X_{2}\cup X^{*})| \displaystyle<\left\lceil\frac{\omega}{4}\right\rceil+\max\left\{\omega-p,\frac{3}{4}\omega\right\} \displaystyle\leq\left\lceil\frac{5}{4}\omega\right\rceil-p. This completes the proof. ∎ Lemma 11.70 . \chi(G)\leq\left\lceil\frac{5}{4}\omega\right\rceil . Proof. By Lemmas 11.45 and 11.41 , we may assume that A_{2}=A_{2}(i)\cup A_{2}(i+2) . By Lemmas 11.42 11.68 and 11.52 , A_{3}=\emptyset . By Lemma 11.56 , T can be partitioned into three cliques T_{1},T_{2},T_{3} such that T_{2} is complete to T_{1}\cup T_{3} and T_{1} and T_{3} are anticomplete where T_{1},T_{3}\neq\emptyset and T_{2} may be empty. Let L=N(T_{1})\cap A_{1}(i) , R=N(T_{3})\cap A_{1}(i) and M=A_{1}(i)\setminus(L\cup R) . Note that L and R are disjoint. We show that T_{2} is complete to L\cup R . Suppose that v\in T_{2} has a non-neighbor u\in L\cup R , say u\in L . If v is anticomplete to L\cup R , then v has a neighbor v^{\prime}\in M and so v-v^{\prime}-u-N_{T_{1}}(u)-v is an induced C_{4} . So v has a neighbor w\in L\cup R . If w\in R , then w-u-N_{T_{1}}(u)-v-w is an induced C_{4} . So v is anticomplete to R . By symmetry, v is anticomplete to L , a contradiction. Next we give a \left\lceil\frac{5\omega}{4}\right\rceil -coloring of G . Let x=|B_{i-2}| and y=|B_{i+2}| . Our strategy is to use p=x+y-2\left\lceil\frac{\omega}{4}\right\rceil colors to color some vertices of G first and then argue that the remaining vertices can be colored with f=\left\lceil\frac{5}{4}\omega\right\rceil-p colors. Note that \displaystyle f \displaystyle=\left\lceil\frac{5}{4}\omega\right\rceil-(x+y-2\left\lceil\frac{\omega}{4}\right\rceil) \displaystyle=\omega+(3\left\lceil\frac{\omega}{4}\right\rceil-(x+y)) \displaystyle\geq\omega, since x+y\leq\frac{3}{4}\omega . Step 1: Precolor a subgraph. Next we define vertices to be colored with p colors. Let S_{i-2}\subseteq B_{i-2} be the set of x-\left\lceil\frac{\omega}{4}\right\rceil vertices, S_{i+2}\subseteq B_{i+2} be the set of y-\left\lceil\frac{\omega}{4}\right\rceil vertices, and S_{i}\subseteq B_{i} be the set of p vertices with largest neighborhoods in H . Since (x+y)-2\left\lceil\frac{\omega}{4}\right\rceil\leq\left\lceil\frac{\omega}{4}\right\rceil<|B_{i}| , the choice of S_{i} is possible. We color S_{i} with all colors in [p] , and color all vertices in S_{i-2}\cup S_{i+2} with all colors in [p] . Since B_{i} is anticomplete to B_{i-2}\cup B_{i+2} , this coloring is proper. So it suffices to show that G_{1}=G-(S_{i}\cup S_{i-2}\cup S_{i+2}) is f -colorable. Step 2: Reduce by degeneracy. Next, we use Lemma 11.69 to further reduce the problem to color a subgraph of G_{1} . Let v\in B_{i-1} with minimal neighborhood in B_{i-2}\cup B_{i} . By Lemma 11.69 with (X_{1},X_{2},X_{3},X^{*})=(B_{i},B_{i-1},B_{i-2}\setminus S_{i-2},\emptyset) , P=S_{i} , and F=G_{1} , we have d_{G_{1}}(v)\left\lceil\frac{\omega}{4}\right\rceil . Proof. By Lemma 11.46 , we have A_{3}(i+2)=\emptyset . By Lemma 8.11 , w is anticomplete to A_{3}(i-1) . By Lemma 11.41 , w has no neighbor in A_{2}\setminus A_{2}(i+2) . By Lemmas 11.57 , 11.58 , 11.51 and the assumption that T is a clique, N(w)\setminus B_{i-1}=B_{i-2}\cup B_{i+2}\cup T\cup T_{0} is a clique. ∎ Lemma 11.72 . A_{3}(i-2) is a clique. By symmetry, A_{3}(i+2) is a clique. Proof. Suppose not. Let s,s^{\prime}\in A_{3}(i-2) be two non-adjacent vertices simplicial in A_{3}(i-2) . By Lemma 11.53 , |N(s)\cap B_{i-1}|,|N(s^{\prime})\cap B_{i-1}|>\left\lceil\frac{\omega}{4}\right\rceil . Let w\in B_{i-2} be anticomplete to A_{3}(i-2) . By Lemma 11.71 , |N(w)\cap B_{i-1}|>\left\lceil\frac{\omega}{4}\right\rceil . It follows from Lemmas 8.10 and 11.54 that B_{i}\cup(N(s)\cap B_{i-1})\cup(N(s^{\prime})\cap B_{i-1})\cup(N(w)\cap B_{i-1}) is a clique of size larger than \omega . ∎ Lemma 11.73 . If A_{3}(i-2)\cup A_{3}(i+2)\neq\emptyset , then A_{3}(i-1)\cup A_{3}(i+1)=\emptyset . Proof. By symmetry, we assume that A_{3}(i-2)\neq\emptyset . Let t\in A_{3}(i-2) be a vertex simplicial in A_{3}(i-2) and b_{i-2}\in B_{i-2} be a vertex anticomplete to A_{3}(i-2) . By Lemma 8.11 , b_{i-2} is anticomplete to A_{3}(i-1) . By Lemma 11.53 , |N(t)\cap B_{i-1}|>\left\lceil\frac{\omega}{4}\right\rceil . By Lemma 11.71 , |N(b_{i-2})\cap B_{i-1}|>\left\lceil\frac{\omega}{4}\right\rceil . Let s\in A_{3}(i-1) and b_{i-1}\in B_{i-1} be anticomplete to A_{3}(i-1) . If A_{2}(i-1) contains a vertex a , then s-(N(s)\cap B_{i-1})-a-(N(a)\cap A_{1}(i))-T-q_{i+2}-q_{i+1} is an induced sequence. So A_{2}(i-1)=\emptyset . By Lemmas 8.14 and 8.15 , there are v\in A_{3}(i-1) and w\in B_{i-1}\setminus N(A_{3}(i-1)) such that N(v)\cap B_{i},N(v)\cap B_{i-2},N(w)\cap B_{i},N(w)\cap B_{i-2} are cliques of size larger than \left\lceil\frac{\omega}{4}\right\rceil . If v and w have disjoint neighborhoods in B_{i} , then (N(v)\cap B_{i})\cup(N(w)\cap B_{i})\cup(N(t)\cap B_{i-1})\cup(N(b_{i-2})\cap B_{i-1}) is a clique of size larger than \omega by Lemma 8.10 . So v and w have disjoint neighborhoods in B_{i-2} . Then (N(v)\cap B_{i-2})\cup(N(w)\cap B_{i-2})\cup(N(t)\cap B_{i-1})\cup(N(b_{i-2})\cap B_{i-1}) is a clique of size larger than \omega by Lemma 8.13 . ∎ Lemma 11.74 . If A_{3}(i\pm 2)\neq\emptyset , then A_{1}(i) is a clique. Proof. By symmetry, assume that A_{3}(i-2)\neq\emptyset . Suppose that a\in A_{3}(i-2) is simplicial in A_{3}(i-2) . Let b\in B_{i-2} anticomplete to A_{3}(i-2) . If c\in B_{i} mixes on an edge st\in A_{1}(i) , then a-q_{i+2}-b-q_{i-1}-c-s-t is an induced P_{7} . So each vertex in B_{i} is complete or anticomplete to any component of A_{1}(i) . If a component K of A_{1}(i) is not a clique, then |T|>\frac{\omega}{2} by Lemma 11.8 and so B_{i-2}\cup B_{i+2}\cup T is a clique of size larger than \omega . So any component of A_{1}(i) is a clique. Suppose that A_{1}(i) has two components L_{1} and L_{2} . Note that K_{a}=N(a)\cap B_{i-1} and K_{b}=N(b)\cap B_{i-1} are disjoint and have size >\left\lceil\frac{\omega}{4}\right\rceil by Lemmas 11.71 and 11.53 . If u\in L_{j} has a neighbor w\in A_{2}(i) , then u-w-(N(w)\cap B_{i+1})-q_{i+2}-b-K_{b}-K_{a} is an induced P_{7} . So L_{j} is anticomplete to A_{2}(i) for j=1,2 . It follows that N(L_{j})\subseteq B_{i}\cup A_{2}(i-1)\cup T . Let s_{j}\in L_{j} such that |N(s_{j})\cap T| is minimum. Since L_{j} and T are cliques, N(s_{j})\cap(L_{j}\cup T) is a clique. By Proposition 6.1 , K_{s_{j}}=N(s_{j})\cap(B_{i}\cup A_{2}(i-1))=N(s_{j})\setminus(L_{j}\cup T) is a clique and so |K_{s_{j}}|>\left\lceil\frac{\omega}{4}\right\rceil . If K_{s_{1}} and K_{s_{2}} are not disjoint, then |T|>\frac{\omega}{2} and so T\cup B_{i-2}\cup B_{i+2} is a clique of size larger than \omega . So K_{s_{1}} and K_{s_{2}} are disjoint. Since G is (C_{6},C_{7}) -free, K_{s_{1}} and K_{s_{2}} are complete. By Lemma 8.10 and 11.64 , K_{s_{1}}\cup K_{s_{2}}\cup K_{a}\cup K_{b} is a clique of size larger than \omega . This proves that A_{1}(i) is a clique. ∎ Lemma 11.75 . A_{3}(i-1) or A_{3}(i+1) is empty. Proof. Suppose not. By Lemmas 11.62 and 11.50 , there exist s\in A_{3}(i-1) and a\in B_{i-1}\setminus N(s) such that |N(a)\cap B_{i-2}|>\left\lceil\frac{\omega}{4}\right\rceil and |N(s)\cap B_{i-2}|>\left\lceil\frac{\omega}{4}\right\rceil , and there exist t\in A_{3}(i+1) and b\in B_{i+1}\setminus N(t) such that |N(b)\cap B_{i+2}|>\left\lceil\frac{\omega}{4}\right\rceil and |N(t)\cap B_{i+2}|>\left\lceil\frac{\omega}{4}\right\rceil . By Lemma 8.9 , we may assume that N(a)\cap B_{i-2} and N(s)\cap B_{i-2} are disjoint, and so \omega(B_{i-2}\cup B_{i+2}\cup T)>\omega , a contradiction. ∎ Lemma 11.76 . A_{1}(i) is a blowup of P_{3} . Proof. By Lemma 11.65 , we may assume that A_{2}(i-1)=A_{2}(i)=\emptyset and A_{1}(i) is not a clique. (1) Any two non-adjacent (A_{1}(i),B_{i}) -minimal simplicial vertices a,b have a common neighbor in T . Suppose not. By Lemma 11.8 , |N(a)\cap T|,|N(b)\cap T|>\left\lceil\frac{\omega}{4}\right\rceil . It follows that \omega(T\cup B_{i-2}\cup B_{i+2})>\omega , a contradiction. This proves (1). (2) If there are two non-adjacent simplicial vertices a,b of A_{1}(i) which have disjoint neighborhoods in B_{i} , then |N(a)\cap B_{i}|,|N(b)\cap B_{i}|>\left\lceil\frac{\omega}{4}\right\rceil . Moreover, (N(a)\cap B_{i})\cup(N(b)\cap B_{i}) is complete to B_{i-1}\cup B_{i+1} . It suffices to show that N(a)\setminus B_{i} and N(b)\setminus B_{i} are cliques. Note that N(a)\setminus B_{i}\subseteq A_{1}(i)\cup T . Since a is simplicial in A_{1}(i) and T is a clique, it suffices to show that N(a)\cap A_{1}(i) is complete to N(a)\cap T . Suppose that s\in N(a)\cap A_{1}(i) is not adjacent to s^{\prime}\in N(a)\cap T . If s has a neighbor p\in B_{i} not adjacent to a , then p-s-a-s^{\prime}-q_{i+2}-q_{i+1}-p is an induced C_{6} . So N(s)\cap B_{i}\subseteq N(a)\cap B_{i} . Then s-a-s^{\prime}-q_{i+2}-q_{i+1}-N_{B_{i}}(b)-q_{i-1} is an induced P_{7} . This proves that N(a)\setminus B_{i} is a clique. By symmetry, N(b)\setminus B_{i} is a clique. We first show that each vertex v\in B_{i-1} is either complete or anticomplete to (N(a)\cap B_{i})\cup(N(b)\cap B_{i}) . Suppose that v has a neighbor a^{\prime}\in N(a)\cap B_{i} . If v has a non-neighbor b^{\prime}\in N(b)\cap B_{i} , then v-a^{\prime}-b^{\prime}-b-N_{T}(b)-q_{i-2}-v is an induced C_{6} . So v is complete to N(b)\cap B_{i} . It follows that v is complete to N(a)\cap B_{i} . Now suppose that v is anticomplete to (N(a)\cap B_{i})\cup(N(b)\cap B_{i}) . Let c\in B_{i} be a neighbor of v and d\in N(a)\cap B_{i} . Then v-c-d-a-N_{T}(a)-q_{i-2}-v is an induced C_{6} . This proves (2). (3) |B_{i-1}|>\left\lceil\frac{\omega}{4}\right\rceil or |B_{i+1}|>\left\lceil\frac{\omega}{4}\right\rceil . By Lemma 11.71 , we may assume that A_{3}(i-2)=A_{3}(i+2)=\emptyset . By Lemma 11.75 , we may assume that A_{3}(i-1)=\emptyset . For any vertex v\in B_{i-2} , N(v)\setminus B_{i-1}=A_{2}(i+2)\cup B_{i+2}\cup B_{i-2} is a clique since T is a clique. It follows that |B_{i-1}|>\left\lceil\frac{\omega}{4}\right\rceil . This proves (3). By (1), (2), (3) and Lemma 2.6 , A_{1}(i) cannot have three pairwise non-adjacent simplicial vertices. This implies that A_{1}(i) has at most two components and if A_{1}(i) has two components, then each component is a clique. So assume that A_{1}(i) is connected. Since A_{1}(i) is P_{4} -free, A_{1}(i) can be partitioned into two non-empty subsets X_{1} and X_{2} such that X_{1} and X_{2} are complete. Since G is C_{4} -free, X_{1} is a clique. Choose such a partition (X_{1},X_{2}) such that |X_{1}| is maximum. It follows that X_{2} is disconnected. Since each vertex simplicial in X_{2} is simplicial in A_{1}(i) , X_{2} has exactly two clique components. So A_{1}(i) is a blowup of P_{3} . ∎ Lemma 11.77 . A_{3}(i-1) is a clique and for every vertex a\in A_{3}(i-1) , N_{B_{i}}(a) is anticomplete to B_{i-1}\setminus N(a) . Proof. Suppose first that A_{3}(i-1) is not a clique. By Lemma 11.73 , A_{3}(i-2)=A_{3}(i+2)=\emptyset . If A_{2}(i-1) contains a vertex p , let s\in A_{3}(i-1) and then s-(N(s)\cap B_{i-1})-p-(N(p)\cap A_{1}(i))-T-q_{i+2}-q_{i+1} is an induced sequence. So A_{2}(i-1)=\emptyset . By Lemmas 2.6 , 8.14 and 8.15 , there exist three pairwise non-adjacent vertices v_{1},v_{2}\in A_{3}(i-1) and w\in B_{i-1}\setminus N(A_{3}(i-1)) such that N(v_{1})\cap B_{i},N(v_{1})\cap B_{i-2},N(v_{2})\cap B_{i},N(v_{2})\cap B_{i-2},N(w)\cap B_{i},N(w)\cap B_{i-2} are cliques of size larger than \left\lceil\frac{\omega}{4}\right\rceil . If any two of v_{1},v_{2},w have disjoint neighborhoods in B_{i-2} , then B_{i-2}\cup B_{i+2}\cup T is a clique of size larger than \omega . So v_{1},v_{2},w have pairwise comparable neighborhoods in B_{i-2} and so have disjoint neighborhoods in B_{i} . Let v be a vertex in B_{i+2} . Since N(v)\subseteq B_{i+1}\cup(B_{i-2}\cup B_{i+2}\cup A_{2}(i+2)) and B_{i-2}\cup B_{i+2}\cup A_{2}(i+2) is a clique, |N(v)\cap B_{i+1}|>\left\lceil\frac{\omega}{4}\right\rceil . By Lemmas 8.10 , (N(v_{1})\cap B_{i})\cup(N(v_{2})\cap B_{i})\cup(N(w)\cap B_{i})\cup B_{i+1} is a clique of size larger than \omega . So A_{3}(i-1) is a clique. Let u be a vertex in A_{3}(i-1) with minimal neighborhood in H . Since A_{3}(i-1) is a clique, u is simplicial in A_{3}(i-1) . By Lemma 8.14 , |N(u)\cap B_{i-2}|>\left\lceil\frac{\omega}{4}\right\rceil . By Lemma 8.15 , there exists w\in B_{i-1}\setminus N(A_{3}(i-1)) such that |N(w)\cap B_{i-2}|>\left\lceil\frac{\omega}{4}\right\rceil . If u and w have disjoint neighborhoods in B_{i-2} , then B_{i-2}\cup B_{i+2}\cup A_{2}(i+2) is a clique of size larger than \omega . So w and v have a common neighbor in B_{i-2} . By the minimality of u , every vertex a in A_{3}(i-1) and w have a common neighbor in B_{i-2} , and so N_{B_{i}}(a) is anticomplete to B_{i-1}\setminus N(a) . ∎ Lemma 11.78 . If A_{3}(i\pm 1)\neq\emptyset , then A_{1}(i) is a clique. Proof. Suppose for a contradiction that A_{1}(i) is not a clique. By symmetry, A_{3}(i-1)\neq\emptyset . Let v\in A_{3}(i-1) and w\in B_{i-1}\setminus N(A_{3}(i-1)) . By Lemma 11.77 , v and w have disjoint neighborhoods in B_{i} . By Lemma 11.8 , A_{1}(i) has two non-adjacent vertices s_{1},s_{2} such that |N(s_{j})\cap T|>\left\lceil\frac{\omega}{4}\right\rceil for j=1,2 . If s_{1} and s_{2} have disjoint neighborhoods in T , then \omega(T\cup B_{i-2}\cup B_{i+2})>\omega . So s_{1} and s_{2} have disjoint neighborhoods in B_{i} . If there is an induced P_{4}=a-b-c-d such that a\in A^{\prime}_{3}(i-1) , b,c\in B_{i} and d\in A_{1}(i) , then a-b-c-d-N_{T}(d)-N_{B_{i-2}}(a)-a is an induced C_{6} . It follows that N(v),N(w)\subseteq N(s_{j}) for j=1,2 . This contradicts that s_{1} and s_{2} have disjoint neighborhoods in B_{i} . ∎ Lemma 11.79 . If A_{3}\neq\emptyset , \chi(G)\leq\left\lceil\frac{5}{4}\omega\right\rceil . Proof. By Lemma 11.41 , we may assume that A_{2}=A_{2}(i-1)\cup A_{2}(i)\cup A_{2}(i+2) , where A_{2}(i-1) and A_{2}(i) may be empty. By Lemmas 11.74 and 11.78 , A_{1}(i) is a clique. If A_{3}(i-2)\neq\emptyset , then by Lemmas 11.73 , 11.42 and 11.46 , A_{3}=A_{3}(i-2) and by Lemma 11.72 , A_{3}(i-2) is a clique. If A_{3}(i-2)=A_{3}(i+2)=\emptyset , then by Lemma 11.75 , we may assume that A_{3}(i+1)=\emptyset . Moreover, by Lemma 11.77 , A_{3}(i-1) is a clique. Next we give a \left\lceil\frac{5}{4}\omega\right\rceil -coloring of G . Let x=\begin{cases}\,|A_{3}(i-2)\cup N_{B_{i-2}}(A_{3}(i-2))|,&\text{if }\,A_{3}(i-2)\neq\emptyset,\\ \,|B_{i-2}|,&\text{if }\,A_{3}(i-2)=\emptyset,\end{cases} y=|B_{i+2}| , and z=|T| . Our strategy is to use p=x+y+z-3\left\lceil\frac{\omega}{4}\right\rceil colors to color some vertices of G first and then argue that the remaining vertices can be colored with f=\left\lceil\frac{5}{4}\omega\right\rceil-p colors. Note that \displaystyle f \displaystyle=\left\lceil\frac{5}{4}\omega\right\rceil-(x+y+z-3\left\lceil\frac{\omega}{4}\right\rceil) \displaystyle=\omega+(4\left\lceil\frac{\omega}{4}\right\rceil-(x+y+z)) \displaystyle\geq\omega, since x+y+z\leq\omega . Step 1: Precolor a subgraph. Next we define vertices to be colored with p colors. Let S_{i-2}\subseteq\begin{cases}\,A_{3}(i-2)\cup N_{B_{i-2}}(A_{3}(i-2)),&\text{if }\,A_{3}(i-2)\neq\emptyset,\\ \,B_{i-2},&\text{if }\,A_{3}(i-2)=\emptyset,\end{cases} be the set of x-\left\lceil\frac{\omega}{4}\right\rceil vertices, S_{i+2}\subseteq B_{i+2} be the set of y-\left\lceil\frac{\omega}{4}\right\rceil vertices, S_{T}\subseteq T be the set of z-\left\lceil\frac{\omega}{4}\right\rceil vertices, and S_{i}\subseteq B_{i} be the set of p vertices with largest neighborhoods in H\cup A_{1}(i) if A_{3}(i-2)\neq\emptyset or with largest neighborhoods in A_{3}(i-1)\cup B_{i+1}\cup A_{1}(i) if A_{3}(i-2)=\emptyset . Since (x+y+z)-3\left\lceil\frac{\omega}{4}\right\rceil\leq\left\lceil\frac{\omega}{4}\right\rceil<|B_{i}| , the choice of S_{i} is possible. We color S_{i} with all colors in [p] , and color all vertices in S_{i-2}\cup S_{i+2}\cup S_{T} with all colors in [p] . Since B_{i} is anticomplete to A^{\prime}_{3}(i-2)\cup B_{i+2}\cup T , this coloring is proper. So it suffices to show that G_{1}=G-(S_{i}\cup S_{i-2}\cup S_{i+2}\cup S_{T}) is f -colorable. Step 2: Reduce by degeneracy. Next, we use Lemma 11.69 to further reduce the problem to color a subgraph of G_{1} . Suppose first that A_{3}(i-2)\neq\emptyset . Let R_{i-1} be the set of vertices in B_{i-1} that is anticomplete to B_{i-2}\setminus N(A_{3}(i-2)) . Let v\in R_{i-1} with minimal neighborhood in B_{i-2}\cup B_{i} among R_{i-1} . Since v is anticomplete to B_{i-2}\setminus N(A_{3}(i-2)) , v is also minimal among B_{i-1} . By Lemma 11.69 with (X_{1},X_{2},X_{3},X^{*})=\left(B_{i},B_{i-1},\left(A_{3}(i-2)\cup N_{B_{i-2}}\left(A_{3}(i-2)\right)\right)\setminus S_{i-2},A_{2}(i-1)\right) , P=S_{i} , and F=G_{1} , we have d_{G_{1}}(v)\left\lceil\frac{\omega}{4}\right\rceil . Proof. Let w\in B_{i+1}\setminus N(A_{3}(i+1)) with minimal neighborhood in B_{i+2} . By Lemma 12.7 , N(w)\setminus B_{i}\subseteq B_{i+1}\cup B_{i+2} . Suppose that N(w)\setminus B_{i} has two non-adjacent vertices s,s^{\prime} with s\in B_{i+1} and s^{\prime}\in B_{i+2} . If s\in B_{i+1}\setminus N(A_{3}(i+1)) , then by the minimality of w , we have s has a neighbor s^{\prime\prime}\in B_{i+2} with s^{\prime\prime}w\notin E(G) . Then s-w-s^{\prime}-s^{\prime\prime}-s is an induced C_{4} . So we may assume that s\in N_{B_{i+1}}(A_{3}(i+1)) . Then N_{A_{3}(i+1)}(s)-s-w-s^{\prime}-a-N_{B_{i-1}}(a)-N_{B_{i-1}}(b) is an induced P_{7} . ∎ Lemma 12.9 . For every special pair \{a_{0},b_{0}\} of A_{2}(i+2) , each vertex in (N(a_{0})\cup N(b_{0}))\cap A^{\prime}_{3}(i-2) is complete to A_{2}(i+2) . Proof. Suppose not. By symmetry, we may assume a_{0} has a neighbor a^{\prime}\in A^{\prime}_{3}(i-2) that has a non-neighbor a^{\prime\prime}\in A_{2}(i+2) . By definition of a_{0},b_{0} and C_{4} -freeness of G , a^{\prime}b_{0}\notin E(G) . Then a^{\prime}-a_{0}-a^{\prime\prime}-b_{0}-N_{B_{i-1}}(b_{0})-q_{i}-q_{i+1} is an induced P_{7} . ∎ Let w be a vertex in B_{i-2}\setminus N(A_{3}(i-2)) with minimal neighborhood in H among all vertices in B_{i-2}\setminus N(A_{3}(i-2)) . Lemma 12.10 . N(w)\setminus B_{i-1} is a clique and so |N_{B_{i-1}}(w)|>\left\lceil\frac{\omega}{4}\right\rceil . Proof. Suppose not. Let s,s^{\prime} be two non-adjacent vertices in N(w)\setminus B_{i-1} . By Lemmas 12.7 and 8.11 , N(w)\setminus B_{i-1}\subseteq B_{i-2}\cup B_{i+2}\cup A_{2}(i+2) . Since B_{i+2}\cup A_{2}(i+2) is a clique, we may assume that s\in B_{i-2} and s^{\prime}\in B_{i+2}\cup A_{2}(i+2) . If s^{\prime}\in A_{2}(i+2) , then s-w-s^{\prime}-b-N_{B_{i-1}}(b)-q_{i}-q_{i+1} is an induced P_{7} . So we may assume that s^{\prime}\in B_{i+2} . Suppose first that s\in B_{i-2}\setminus N(A_{3}(i-2)) . By the minimality of w , s has a neighbor s^{\prime\prime} in B_{i+2}\cup B_{i-1} that is not adjacent to w . If s^{\prime\prime}\in B_{i+2} , then s^{\prime\prime}-s-w-s^{\prime}-s^{\prime\prime} is an induced C_{4} . If s^{\prime\prime}\in B_{i-1} , then s^{\prime\prime}-s-w-s^{\prime} is a P_{4} -structure. So we may assume that s has a neighbor s^{\prime\prime}\in A_{3}(i-2) . Since s^{\prime\prime}-s-w-s^{\prime}-q_{i+1}-q_{i}-B_{i-1} is not a bad P_{7} , we have w is complete to B_{i-1} . So w and s^{\prime\prime} have a common neighbor in B_{i-1} , and hence w,s^{\prime\prime} have disjoint neighborhoods in B_{i+2} . By Lemma 8.1 (2), w-s^{\prime}-N_{B_{i+2}}(s^{\prime\prime})-s-w is an induced C_{4} . ∎ By Lemma 2.4 , we can assume that a,b are simplicial vertices of the set that consists of all vertices in A^{\prime}_{3}(i-2) complete to A_{2}(i+2) . Lemma 12.11 . a\in B_{i-2} . Proof. Suppose not. By Lemmas 12.10 and 12.4 , |N_{B_{i-1}}(w)|,|N_{B_{i-1}}(a)|,|N_{B_{i-1}}(b)|>\left\lceil\frac{\omega}{4}\right\rceil . Since a,b,w are pairwise non-adjacent and they have a common neighbor in B_{i+2} , N_{B_{i-1}}(w) , N_{B_{i-1}}(a) , N_{B_{i-1}}(b) are pairwise disjoint. By Lemma 8.10 , N_{B_{i-1}}(w) , N_{B_{i-1}}(a) , N_{B_{i-1}}(b) are complete to B_{i} . By Lemma 12.8 , N_{B_{i-1}}(w)\cup N_{B_{i-1}}(a)\cup N_{B_{i-1}}(b)\cup B_{i} is a clique of size larger than \omega , a contradiction. ∎ Lemma 12.12 . Let v be a (A_{3}(i-2),B_{i+2}\cup A_{2}(i+2)) -minimal simplicial vertex. Then N(v)\setminus B_{i-1} is a clique and so |N_{B_{i-1}}(v)|>\left\lceil\frac{\omega}{4}\right\rceil . Proof. Suppose not. Let s,s^{\prime} be two non-adjacent neighbors of v in N(v)\setminus B_{i-1} . Note that N(v)\setminus B_{i-1}\subseteq B_{i-2}\cup B_{i+2}\cup A_{2}(i+2)\cup A_{3}(i-2) . Since v is simplicial in A_{3}(i-2) and by Lemma 8.3 , we have at most one of s,s^{\prime}\in A^{\prime}_{3}(i-2) . Since B_{i+2}\cup A_{2}(i+2) is a clique, we may assume that s\in A^{\prime}_{3}(i-2) and s^{\prime}\in B_{i+2}\cup A_{2}(i+2) . Assume first that s\in B_{i-2} . By Lemma 8.1 (2), s^{\prime}\in A_{2}(i+2) . By Lemma 12.11 , as\in E(G) and so sb\notin E(G) by G is C_{4} -free. Then s-a-s^{\prime}-b-N_{B_{i-1}}(b)-q_{i}-q_{i+1} is an induced P_{7} . So we may assume that s\in A_{3}(i-2) . If s is simplicial in A_{3}(i-2) , then s has a neighbor s^{\prime\prime}\in A_{2}(i+2)\cup B_{i+2} that is not adjacent to v by the minimality of v , and so s-s^{\prime\prime}-s^{\prime}-v-s is an induced C_{4} . So s is not simplicial in A_{3}(i-2) and thus s has a neighbor s^{\prime\prime}\in A_{3}(i-2) that is not adjacent to v . It follows that v-s-s^{\prime\prime} is an induced P_{3} in A_{3}(i-2) . By Lemma 8.4 , s^{\prime}\in A_{2}(i+2) . Since s is not complete to A_{2}(i+2) , v\neq a,b by Lemma 12.9 . If v is complete to A_{2}(i+2) , then v is adjacent to a and b by Lemma 12.9 . Then a\in B_{i-2} mixes on vb\in A_{3}(i-2) , which contradicts Lemma 8.3 . So v is not complete to A_{2}(i+2) and so av\notin E(G) or bv\notin E(G) by Lemma 12.1 . By symmetry, we assume that av\notin E(G) . Then s-v-s^{\prime}-a-N_{B_{i-1}}(a)-q_{i}-q_{i+1} is an induced P_{7} . ∎ Lemma 12.13 . A_{3}(i-2)\cup A_{2}(i+2)\cup B_{i+2} is a clique. Proof. We first show that A_{3}(i-2) is a clique. Suppose not. By Lemma 2.6 , there are two non-adjacent (A_{3}(i-2),A_{2}(i+2)\cup B_{i+2}) -minimal simplicial vertices v_{1},v_{2} . If b is adjacent to both v_{1} and v_{2} , then v_{1} and v_{2} complete to A_{2}(i+2) by Lemma 12.9 , which contradicts the choice of b . So we may assume that v_{1} is not adjacent to b . For convenience, let v=v_{1} . Suppose first that av\notin E(G) . Since a,b,v have a common neighbor in B_{i+2} , a,b,v have pairwise disjoint neighborhoods in B_{i-1} . By Lemma 8.10 , N_{B_{i-1}}(v)\cup N_{B_{i-1}}(a)\cup N_{B_{i-1}}(b)\cup B_{i} is a clique. By Lemmas 12.12 , 12.4 and 12.8 , |N_{B_{i-1}}(v)\cup N_{B_{i-1}}(a)\cup N_{B_{i-1}}(b)\cup B_{i}|>\omega , a contradiction. So av\in E(G) . By Lemma 12.9 , v is complete to A_{2}(i+2) and so \{v,b\} is a special pair for A_{2}(i+2) . By Lemma 8.3 , v and b are simplicial in A^{\prime}_{3}(i-2) . The fact that v,b\in A_{3}(i-2) contradicts Lemma 12.11 . This proves that A_{3}(i-2) is a clique. By Lemma 12.9 , A_{3}(i-2) is complete to A_{2}(i+2) . Since a\in B_{i-2} , a is anticomplete to A_{3}(i-2) by Lemma 8.3 . For any vertex v\in A_{3}(i-2) , \{a,v\} is a special pair for A_{2}(i+2) and so v is complete to B_{i+2} by Lemma 12.1 . This proves the lemma. ∎ Lemma 12.14 . B_{i-2} is complete to B_{i+2}\cup A_{2}(i+2) . Proof. Since a\in B_{i-2} , B_{i-2} is complete to A_{2}(i+2) by Lemma 12.9 . By Lemma 4.8 , B_{i-2} is complete to B_{i+2} . ∎ Lemma 12.15 . A_{3}(i-1)=\emptyset . Proof. Suppose not. Let u be a (A_{3}(i-1),H) -minimal simplicial vertex. By Lemma 8.14 , |N_{B_{i}}(u)|,|N_{B_{i-2}}(u)|>\left\lceil\frac{\omega}{4}\right\rceil . Let w^{\prime}\in B_{i-1}\setminus N(A_{3}(i-1)) be a vertex with minimal neighborhood in H . By Lemma 8.15 , |N_{B_{i}}(w^{\prime})|>\left\lceil\frac{\omega}{4}\right\rceil and |N_{B_{i-2}}(w^{\prime})|>\left\lceil\frac{\omega}{4}\right\rceil . If N_{B_{i}}(w^{\prime}) and N_{B_{i}}(u) are disjoint, then |B_{i}|>2\cdot\left\lceil\frac{\omega}{4}\right\rceil and so N_{B_{i-1}}(w)\cup N_{B_{i-1}}(b)\cup B_{i} is a clique of size larger than \omega by Lemmas 12.10 and 12.4 (recall that w is the vertex defined before Lemma 12.10 ). So we may assume that N_{B_{i}}(w^{\prime}) and N_{B_{i}}(u) have a common vertex and so N_{B_{i-2}}(w^{\prime}) and N_{B_{i-2}}(u) are disjoint. By Lemma 8.13 , N_{B_{i-1}}(b)\cup N_{B_{i-1}}(w)\cup N_{B_{i-2}}(u)\cup N_{B_{i-2}}(w^{\prime}) is a clique of size larger than \omega , a contradiction. ∎ Lemma 12.16 . A_{3}(i+1) is a clique. Proof. Suppose not. By Lemma 2.6 , there are two non-adjacent (A_{3}(i+1),H) -minimal simplicial vertices v_{1},v_{2} . By Lemma 8.14 , |N_{B_{i}}(v_{1})|,|N_{B_{i+2}}(v_{1})|,|N_{B_{i}}(v_{2})|,|N_{B_{i+2}}(v_{2})|>\left\lceil\frac{\omega}{4}\right\rceil . If N_{B_{i}}(v_{1}) and N_{B_{i}}(v_{2}) are disjoint, then N_{B_{i}}(v_{1})\cup N_{B_{i}}(v_{2})\cup N_{B_{i-1}}(a)\cup N_{B_{i-1}}(b) is a clique of size larger than \omega by Lemmas 12.4 and 8.10 . So we may assume that N_{B_{i}}(v_{1}) and N_{B_{i}}(v_{2}) have a common vertex and so N_{B_{i+2}}(v_{1}) and N_{B_{i+2}}(v_{2}) are disjoint. Let w^{\prime} be a vertex in B_{i+1}\setminus N(A_{3}(i+1)) that has minimal neighborhood in H . By Lemma 8.15 , |N_{B_{i+2}}(w^{\prime})|>\left\lceil\frac{\omega}{4}\right\rceil . Since B_{i} is complete to B_{i+1} , each of w^{\prime},v_{1} and w^{\prime},v_{2} have a common neighbor in B_{i} . So N_{B_{i+2}}(w^{\prime}) , N_{B_{i+2}}(v_{1}) and N_{B_{i+2}}(v_{2}) are pairwise disjoint cliques that are complete to A_{3}(i-2)\cup N_{B_{i-2}}(A_{3}(i-2)) by Lemma 8.10 . Let v be a vertex in B_{i-1} with minimal neighborhood in B_{i}\cup B_{i-2} . Since A_{3}(i-2)\cup B_{i+2} is a clique by Lemma 12.13 , N_{B_{i-1}}(A_{3}(i-2)) is anticomplete to B_{i-2}\setminus N(A_{3}(i-2)) . It follows that N(v)\setminus(A_{3}(i-2)\cup N_{B_{i-2}}(A_{3}(i-2)))\subseteq B_{i}\cup B_{i-1} . By the minimality of v , N(v)\setminus(A_{3}(i-2)\cup N_{B_{i-2}}(A_{3}(i-2))) is a clique. So |A_{3}(i-2)\cup N_{B_{i-2}}(A_{3}(i-2))|>\left\lceil\frac{\omega}{4}\right\rceil . Hence, (A_{3}(i-2)\cup N_{B_{i-2}}(A_{3}(i-2)))\cup N_{B_{i+2}}(w^{\prime})\cup N_{B_{i+2}}(v_{1})\cup N_{B_{i+2}}(v_{2}) is a clique of size larger than \omega , a contradiction. ∎ Next, we collect some information about the size of certain sets. Lemma 12.17 . The following statements hold (1) |A_{3}(i+1)\cup N_{B_{i+1}}(A_{3}(i+1))|,|A_{3}(i-2)\cup N_{B_{i-2}}(A_{3}(i-2))|,|N_{B_{i+2}}(A_{3}(i+1))|,|B_{i+2}\cap N(B_{i+1}\setminus N(A_{3}(i+1)))|>\left\lceil\frac{\omega}{4}\right\rceil . (2) |B_{i-1}|,|B_{i+1}|,|A_{3}(i+1)\cup(B_{i+1}\cap N(A_{3}(i+1)))|<\frac{3}{4}\omega . (3) |A_{3}(i-2)\cup(B_{i-2}\cap N(A_{3}(i-2)))|,|B_{i}|,|B_{i+1}\cap N(A_{3}(i+1))|<\frac{\omega}{2} . Proof. By Lemmas 8.15 and 8.14 , |N_{B_{i+2}}(A_{3}(i+1))| , |B_{i+2}\cap N(B_{i+1}\setminus N(A_{3}(i+1)))|>\left\lceil\frac{\omega}{4}\right\rceil . Let u\in B_{i-1} be a vertex with minimal neighborhood in B_{i}\cup B_{i-2} . Since N_{B_{i-1}}(A_{3}(i-2)) is anticomplete to B_{i-2}\setminus N(A_{3}(i-2)) , u is anticomplete to B_{i-2}\setminus N(A_{3}(i-2)) . By the minimality of u , N(u)\setminus(A_{3}(i-2)\cup N_{B_{i-2}}(A_{3}(i-2)))\subseteq B_{i}\cup B_{i-1} is a clique and so |N_{B_{i-2}}(A_{3}(i-2))\cup A_{3}(i-2)|\geq|N(u)\cap(N_{B_{i-2}}(A_{3}(i-2))\cup A_{3}(i-2))|>\left\lceil\frac{\omega}{4}\right\rceil . Moreover, N(u)\cap(N_{B_{i-2}}(A_{3}(i-2))\cup A_{3}(i-2)) is complete to B_{i-1} and so |B_{i-1}|<\frac{3}{4}\omega . Let v\in B_{i+2} be a vertex with minimal neighborhood in B_{i+1}\cup B_{i-2} . Since N_{B_{i+2}}(A_{3}(i+1)) is anticomplete to B_{i+1}\setminus N(A_{3}(i+1)) , v is anticomplete to B_{i+1}\setminus N(A_{3}(i+1)) . By the minimality of v , N(v)\setminus(A_{3}(i+1)\cup N_{B_{i+1}}(A_{3}(i+1)))\subseteq B_{i}\cup B_{i-1} is a clique and so |A_{3}(i+1)\cup N_{B_{i+1}}(A_{3}(i+1))|>\left\lceil\frac{\omega}{4}\right\rceil . Let w\in A_{3}(i+1)\cup N_{B_{i+1}}(A_{3}(i+1)) be a vertex with minimal neighborhood in B_{i}\cup B_{i+2} . Since N_{B_{i+1}}(A_{3}(i+1)) is complete to N_{B_{i+2}}(B_{i+1}\setminus N(A_{3}(i+1))) , w\in A_{3}(i+1) . Since A_{3}(i+1) is a clique by Lemma 12.16 , w is simplicial in A_{3}(i+1) and so |N_{B_{i+2}}(w)|>\left\lceil\frac{\omega}{4}\right\rceil by Lemma 8.14 . By the minimality of w , N_{B_{i+2}}(w) is complete to A_{3}(i+1)\cup N_{B_{i+1}}(A_{3}(i+1)) and so |A_{3}(i+1)\cup N_{B_{i+1}}(A_{3}(i+1))|<\frac{3}{4}\omega . Let x\in B_{i+1} be a vertex with minimal neighborhood in B_{i}\cup B_{i+2} . Since N_{B_{i+1}}(A_{3}(i+1)) is complete to N_{B_{i+2}}(A_{3}(i+1)) , x\in B_{i+1}\setminus N(A_{3}(i+1)) . By Lemma 8.15 , |N_{B_{i+2}}(x)|>\left\lceil\frac{\omega}{4}\right\rceil and so by the minimality of x , N_{B_{i+2}}(x) is complete to B_{i+1} . It follows that |B_{i+1}|<\frac{3}{4}\omega . Since |B_{i+2}|>2\cdot\left\lceil\frac{\omega}{4}\right\rceil , and A_{3}(i-2)\cup N_{B_{i-2}}(A_{3}(i-2)) and N_{B_{i+1}}(A_{3}(i+1)) are complete to B_{i+2} , |A_{3}(i-2)\cup N_{B_{i-2}}(A_{3}(i-2))| and |N_{B_{i+1}}(A_{3}(i+1))|<\frac{\omega}{2} . Since B_{i} is complete to two disjoint cliques of size larger than \left\lceil\frac{\omega}{4}\right\rceil in B_{i-1} by Lemma 8.10 , 8.15 and 8.14 , |B_{i}|<\frac{\omega}{2} . ∎ We are now ready to give a desired coloring of G by considering whether A_{3}(i+1) is empty or not. Lemma 12.18 . If A_{3}(i+1)\neq\emptyset , \chi(G)\leq\left\lceil\frac{5}{4}\omega\right\rceil . Proof. Let \displaystyle x \displaystyle=|A_{3}(i-2)\cup(B_{i-2}\cap N(A_{3}(i-2)))|, \displaystyle y \displaystyle=|B_{i+2}\cap N(B_{i+1}\setminus N(A_{3}(i+1)))|, \displaystyle z \displaystyle=|B_{i+2}\cap N(A_{3}(i+1))|. By Lemma 12.14 , x+y+z\leq\omega . Our strategy is to use p=x+y+z-3\left\lceil\frac{\omega}{4}\right\rceil colors to color some vertices of G first and then argue that the remaining vertices can be colored with f=\left\lceil\frac{5}{4}\omega\right\rceil-p colors. Note that \displaystyle f \displaystyle=\left\lceil\frac{5}{4}\omega\right\rceil-(x+y+z-3\left\lceil\frac{\omega}{4}\right\rceil) \displaystyle=4\left\lceil\frac{\omega}{4}\right\rceil+(\omega-(x+y+z)) \displaystyle\geq\omega, since x+y+z\leq\omega . Step 1: Precolor a subgraph. Next we define vertices to be colored with p colors. Let S_{i}\subseteq B_{i} be the set of p vertices with largest neighborhoods in H , S_{i-2}\subseteq A_{3}(i-2)\cup(B_{i-2}\cap N(A_{3}(i-2))) be the set of x-\left\lceil\frac{\omega}{4}\right\rceil vertices with largest neighborhoods in H , S^{y}_{i+2}\subseteq B_{i+2}\cap N(B_{i+1}\setminus N(A_{3}(i+1))) be the set of y-\left\lceil\frac{\omega}{4}\right\rceil vertices with largest neighborhoods in H , and S^{z}_{i+2}\subseteq B_{i+2}\cap N(A_{3}(i+1)) be the set of z-\left\lceil\frac{\omega}{4}\right\rceil vertices with largest neighborhoods in H . Since (x+y+z)-3\left\lceil\frac{\omega}{4}\right\rceil\leq\left\lceil\frac{\omega}{4}\right\rceil<|B_{i}| , the choice of S_{i} is possible. We color S_{i} with all colors in [p] , and color all vertices in S_{i-2}\cup S^{y}_{i+2}\cup S^{z}_{i+2} with all colors in [p] . Since B_{i} is anticomplete to A^{\prime}_{3}(i-2)\cup B_{i+2} , this coloring is proper. So it suffices to show that G_{1}=G-(S_{i}\cup S_{i-2}\cup S^{y}_{i+2}\cup S^{z}_{i+2}) is f -colorable. Step 2: Reduce by degeneracy. Next, we use Lemma 11.69 to further reduce the problem to color a subgraph of G_{1} . Let R_{i-1} be the set of vertices in B_{i-1} that is anticomplete to B_{i-2}\setminus N(A_{3}(i-2)) . Let v\in R_{i-1} with minimal neighborhood in B_{i-2}\cup B_{i} among R_{i-1} . Since v is anticomplete to B_{i-2}\setminus N(A_{3}(i-2)) , v is minimal among B_{i-1} . By Lemma 11.69 with (X_{1},X_{2},X_{3},X^{*})=\left(B_{i},B_{i-1},\left(A_{3}(i-2)\cup N_{B_{i-2}}\left(A_{3}(i-2)\right)\right)\setminus S_{i-2},A_{2}(i-1)\right) , P=S_{i} , and F=G_{1} , we have d_{G_{1}}(v)\left\lceil\frac{\omega}{4}\right\rceil and so |B_{i+1}|>\left\lceil\frac{\omega}{4}\right\rceil . By symmetry, we have |B_{i}|>\left\lceil\frac{\omega}{4}\right\rceil . We first show that there are no three pairwise non-adjacent simplicial vertices of A_{3}(i-2) . Suppose for a contradiction that v_{1}^{\prime},v_{2}^{\prime},v_{3}^{\prime} are three pairwise non-adjacent simplicial vertices of A_{3}(i-2) . For j\in[3] , let v_{j}\in N_{A_{3}(i-2)}[v^{\prime}_{j}] be a vertex simplicial in A_{3}(i-2) with minimal neighborhood in B_{i-1}\cup B_{i+2} . By Lemma 2.6 , v_{1},v_{2},v_{3} are pairwise non-adjacent. By Lemmas 8.10 , 8.15 and 8.14 , we have that no three of w,v_{1},v_{2},v_{3} have pairwise disjoint neighborhoods in B_{i+2} or in B_{i-1} . In other word, no three of w,v_{1},v_{2},v_{3} have pairwise neighborhood containment in B_{i+2} or in B_{i-1} . So, we may assume that v_{2} and v_{3} have disjoint neighborhoods in B_{i+2} . Then v_{1} has neighborhood containment with either both v_{2} and v_{3} in B_{i+2} , or exactly one of v_{2} and v_{3} in B_{i+2} . Suppose first that v_{1} has neighborhood containment with both v_{2} and v_{3} in B_{i+2} . As w has neighborhood containment with at least one of v_{2} and v_{3} in B_{i+2} , w and v_{1} have neighborhood containment in B_{i+2} . Then w , v_{1} and one of v_{2} and v_{3} have pairwise neighborhood containment in B_{i+2} , a contradiction. So, we may assume that v_{1} has neighborhood containment with exactly one of v_{2} and v_{3} in B_{i+2} . By symmetry, we may assume that v_{1} has neighborhood containment with v_{2} in B_{i+2} . Then v_{1} and v_{2} have disjoint neighborhood in B_{i-1} and v_{3} have neighborhood containment with both v_{1} and v_{2} in B_{i-1} . By same argument, as w has neighborhood containment with at least one of v_{1} and v_{2} in B_{i-1} , w , v_{3} and one of v_{1} and v_{2} have pairwise neighborhood containment in B_{i-1} , a contradiction. Therefore, there are no three pairwise non-adjacent simplicial vertices of A_{3}(i-2) . This immediately implies that A_{3}(i-2) has at most two components and if A_{3}(i-2) has two components, then A_{3}(i-2) is a disjoint union of two cliques. So we assume that A_{3}(i-2) is connected. By Lemma 8.5 , A_{3}(i-2) can be partitioned into two complete subsets X and Y . Since G is C_{4} -free, we may assume that X is a clique. Choose such a partition (X,Y) such that |X| is maximum. It follows that Y is disconnected. Since any v\in Y simplicial in Y is also simplicial in A_{3}(i-2) , it follows that Y has two component and each component is a clique. So A_{3}(i) is a blowup of P_{3} . This completes the proof of (v). This completes the proof of Lemma 13.1 . ∎ Next we color H\cup A_{3} in terms of the outcomes of Lemma 13.1 . For a subset S\subseteq B_{i} and v\in S , v is minimal if N_{H}(v) is minimal over all vertices in S , and v is maximal if N_{H}(v) is maximal over all vertices in S . 13.1 Distance 2 In this subsection, we handle the first outcome of Lemma 13.1 . Lemma 13.2 . Let G=H\cup K\cup J where K is a non-empty component of A_{3}(i-1) and J is a non-empty component of A_{3}(i+1) . Then \chi(G)\leq\left\lceil\frac{5}{4}\omega\right\rceil . Proof. By Lemma 13.1 , both K and J are cliques. Moreover, B_{i-1}\setminus N(K) is anticomplete to N_{B_{i}}(K) and B_{i+1}\setminus N(J) is anticomplete to N_{B_{i+2}}(J) . Let z\in B_{i-1}\setminus N(K) be a minimal vertex in B_{i-1}\setminus N(K) and let w\in B_{i+1}\setminus N(J) be a minimal vertex in B_{i+1}\setminus N(J) . Then N_{B_{i}}(K) and N_{B_{i}}(z) are disjoint and N_{B_{i+2}}(J) and N_{B_{i+2}}(w) are disjoint. By Lemma 8.15 and Lemma 8.14 , |B_{i-2}| , |N_{B_{i}}(K)| , |N_{B_{i}}(z)| , |N_{B_{i+2}}(J)| , |N_{B_{i+2}}(w)|>\left\lceil\frac{\omega}{4}\right\rceil . By Lemma 8.1 (4), there is a vertex in N_{B_{i-2}}(K) complete to B_{i-1}\cup B_{i+2} . We set this vertex to be q_{i-2} . Let v_{K}\in K that has maximum neighborhood in H . Let v_{K}^{\prime\prime}\in K that has minimal neighborhood in H and let v_{J}^{\prime\prime}\in J that has minimal neighborhood in H . We set F=(F_{1},F_{2},F_{3},F_{4},F_{5},O) as follows: • F_{1}=\{q_{i-2}\} , • F_{2}=\begin{cases}\{q_{i-1},q_{i-1}^{\prime},a\}\quad&\text{if }\,\,|N_{B_{i-1}}(K)|\geq 2\\ \{q_{i-1},a,a^{\prime}\}\quad&\text{if }\,\,|N_{B_{i-1}}(K)|=1,\end{cases} where a,a^{\prime}\in B_{i-1}\setminus N(K) are the first and second maximal in B_{i-1}\setminus N(K) , respectively, • F_{3}=\{q_{i},q_{i}^{\prime},b\} , where b\in N_{B_{i}}(v_{K}^{\prime\prime}) , • F_{4}=\{q_{i+1},c\} , where c\in B_{i+1}\setminus N(J) is a maximal vertex in B_{i+1}\setminus N(J) , • F_{5}=\{q_{i+2},d,d^{\prime}\} , where d,d^{\prime}\in N_{B_{i+2}}(v_{J}^{\prime\prime}) , • O=\begin{cases}\{v_{K}\}\quad&\text{if }\,\,|N_{B_{i-1}}(K)|\geq 2\\ \{v_{K},v_{K}^{\prime}\}\quad&\text{if }\,\,|N_{B_{i-1}}(K)|=1,\end{cases} where v_{K}^{\prime}\in K\setminus\{v_{K}\} that has maximal neighborhood in H . We can assign color 1 to a , b and d , color 2 to q_{i-1} and q_{i+1} , color 3 to v_{K} , q_{i} and q_{i+2} , color 4 to q_{i-2} and q_{i}^{\prime} , and color 5 to v_{K}^{\prime} , q_{i-1}^{\prime} , a^{\prime} , c and d^{\prime} . So F is 5-colorable (see Figure 6 ). B_{i-2} N_{B_{i-2}}(K) q_{i-2} B_{i-1} N_{B_{i-1}}(K) q_{i-1} q_{i-1}^{\prime} a a^{\prime} B_{i} N_{B_{i}}(v_{K}^{\prime\prime}) N_{B_{i}}(z) b q_{i} q_{i}^{\prime} B_{i+1} N_{B_{i+1}}(J) q_{i+1} c B_{i+2} N_{B_{i+2}}(J) N_{B_{i+2}}(w) d d^{\prime} q_{i+2} K v_{K} v_{K}^{\prime} J Figure 6 : Illustration of F and a 5-coloring of F . For convenience, all vertices of color 1 are shown in pink, all vertices of color 2 in green, all vertices of color 3 in blue, all vertices of color 4 in red and all vertices of color 5 in yellow. Note that q_{i-1}^{\prime},v_{K}^{\prime} and a^{\prime} may not exist. Next, we show that every maximal clique of size \omega-j with j\in\{0,1,2,3\} of G contains at least 4-j vertices from F . Let M be a maximal clique of G . We consider the following cases depending on all possible locations of M . (i) M\subseteq K\cup N_{H}(K) We first assume that |N_{B_{i-1}}(K)|\geq 2 . Then q_{i-1} , q_{i-1}^{\prime} and v_{K} are universal in K\cup N_{H}(K) by Lemmas 8.1 (2) and 8.3 . If q_{i-2} has non-neighbor x\in K , then x-v_{K}-q_{i-2}-q_{i+2}-w-q_{i}-J is an induced P_{7} . So, q_{i-2} is complete to K . Note that b is complete to K because b\in N_{B_{i}}(v_{K}^{\prime\prime}) . Therefore, M contains either \{q_{i-2},q_{i-1},q_{i-1}^{\prime},v_{K}\} or \{q_{i-1},q_{i-1}^{\prime},v_{K},b\} . We now assume that |N_{B_{i-1}}(K)|=1 . Recall that q_{i-1} and v_{K} are universal in K\cup N_{H}(K) , and q_{i-2} and b is complete to K . If v_{K}^{\prime}\in M , then we are done. So we may assume that v_{K}^{\prime}\notin M . It follows that |M\cap(K\cup N_{B_{i-1}}(K))|=2 . As |M\cap\{q_{i-2},q_{i-1},v_{K},b\}|=3 , we have |M|=\omega and so either |M\cap N_{B_{i-2}}(K)|=\omega-2 or |M\cap N_{B_{i}}(K)|=\omega-2 . If |M\cap N_{B_{i-2}}(K)|=\omega-2 , then N_{B_{i-2}}(K)\cup N_{B_{i+2}}(J)\cup N_{B_{i+2}}(w) is a clique of size larger than \omega by Lemma 8.10 . If |M\cap N_{B_{i}}(K)|=\omega-2 , then N_{B_{i}}(K)\cup N_{B_{i}}(z)\cup B_{i+1} is a clique of size larger than \omega . Therefore, v_{K}^{\prime}\in M . (ii) M\subseteq J\cup N_{H}(J) We first assume that M\subseteq J\cup N_{B_{i}}(J)\cup N_{B_{i+1}}(J) . By Lemma 8.10 , q_{i} , q_{i}^{\prime} and b are universal in J\cup N_{B_{i}}(J)\cup N_{B_{i+1}}(J) . So, q_{i},q_{i}^{\prime},q_{i+1},b\in M . We now assume that M\subseteq J\cup N_{B_{i+1}}(J)\cup N_{B_{i+2}}(J) . Note that d is complete to J because d\in N_{B_{i+2}}(v_{J}^{\prime\prime}) . So, q_{i+1},d\in M . Hence, we may assume that |M|\geq\omega-1 and it follows that either |M\cap(J\cup N_{B_{i+1}}(J))|\geq\frac{\omega-1}{2} or |M\cap N_{B_{i+2}}(J)|\geq\frac{\omega-1}{2} . Then either N_{B_{i}}(K)\cup N_{B_{i}}(z)\cup J\cup N_{B_{i+1}}(J) or N_{B_{i+2}}(J)\cup N_{B_{i+2}}(w)\cup B_{i-2} is a clique of size larger than \omega by Lemma 8.10 , a contraction. Therefore, |M|\leq\omega-2 and the claim holds. (iii) M\subseteq B_{i-2}\cup B_{i-1} Note that q_{i-2} and q_{i-1} are universal in B_{i-2}\cup B_{i-1} . Suppose first that a\in M . If |N_{B_{i-1}}(K)|\geq 2 , then q_{i-1}^{\prime}\in M because N_{B_{i-2}}(a)\subseteq N_{B_{i-2}}(q_{i-1}^{\prime}) . So we may assume that |N_{B_{i-1}}(K)|=1 . This implies that |M|=\omega and a^{\prime}\notin M . Then |M\cap B_{i-2}|=\omega-2 and it follows that N_{B_{i+2}}(J)\cup N_{B_{i+2}}(w)\cup B_{i-2} is a clique of size larger than \omega , a contradiction. So a^{\prime}\in M and we are done. Hence, we may assume that a\notin M . As q_{i-2},q_{i-1}\in M , we have |M|\geq\omega-1 . Then either |M\cap B_{i-2}|\geq\frac{\omega-1}{2} or |M\cap N_{B_{i-1}}(K)|\geq\frac{\omega-1}{2} . If |M\cap B_{i-2}|\geq\frac{\omega-1}{2} , then N_{B_{i+2}}(J)\cup N_{B_{i+2}}(w)\cup B_{i-2} is a clique of size larger than \omega . If |M\cap N_{B_{i-1}}(K)|\geq\frac{\omega-1}{2} , then N_{B_{i-1}}(K)\cup N_{B_{i}}(K)\cup N_{B_{i}}(z) is a clique of size larger than \omega . Therefore, |M|\leq\omega-2 and the claim holds. (iv) M\subseteq B_{i-1}\cup B_{i} Note that q_{i-1} , q_{i} and q_{i}^{\prime} are universal in B_{i-1}\cup B_{i} . So, q_{i-1},q_{i},q_{i}^{\prime}\in M . Since b is universal in N_{B_{i-1}}(K)\cup B_{i} , we have either a\in M or b\in M . In both case, M contains at least 4 vertices from F . (v) M\subseteq B_{i}\cup B_{i+1} By Lemma 8.10 , q_{i} , q_{i}^{\prime} and b are universal in B_{i}\cup B_{i+1} . So q_{i},q_{i}^{\prime},q_{i+1},b\in M . (vi) M\subseteq B_{i+1}\cup B_{i+2} Note that q_{i+1} and q_{i+2} are universal in B_{i+1}\cup B_{i+2} . So, q_{i+1},q_{i+2}\in M . Since d and d^{\prime} are universal in N_{B_{i+1}}(J)\cup B_{i+2} , we have either c\in M or d,d^{\prime}\in M . If d,d^{\prime}\in M , then we are done. So we may assume that c\in M . As q_{i+1},q_{i+2},c\in M , we have |M|=\omega . It follows that either |M\cap B_{i+1}|\geq\frac{\omega}{2} or |M\cap N_{B_{i+2}}(w)|\geq\frac{\omega}{2} . Then either N_{B_{i}}(K)\cup N_{B_{i}}(z)\cup B_{i+1} or N_{B_{i+2}}(J)\cup N_{B_{i+2}}(w)\cup B_{i-2} is a clique of size larger than \omega by Lemma 8.10 , a contradiction. Therefore, |M|\leq\omega-1 and the claim holds. (vii) M\subseteq B_{i-2}\cup B_{i+2} By Lemma 8.10 , d and d^{\prime} are universal in B_{i+2}\cup B_{i-2} . So q_{i+2},d,d^{\prime},q_{i-2}\in M . Therefore, F is a (4,5) -good subgraph. ∎ 13.2 Distance 1 In this subsection, we handle the second outcome of Lemma 13.1 . Lemma 13.3 . Let G=H\cup K\cup J where K is a non-empty component of A_{3}(i-2) and J is a non-empty component of A_{3}(i+2) . Then \chi(G)\leq\left\lceil\frac{5}{4}\omega\right\rceil . Proof. By Lemma 13.1 , both K and J are cliques. Moreover, B_{i-2}\setminus N(K) is anticomplete to N_{B_{i-1}}(K) and B_{i+2}\setminus N(J) is anticomplete to N_{B_{i+1}}(J) . Let z_{K}\in B_{i-2}\setminus N(K) be a minimal vertex in B_{i-2}\setminus N(K) and let z_{J}\in B_{i+2}\setminus N(J) be a minimal vertex in B_{i+2}\setminus N(J) . Then N_{B_{i-1}}(K) and N_{B_{i-1}}(z_{K}) are disjoint and N_{B_{i+1}}(J) and N_{B_{i+1}}(z_{J}) are disjoint. . By Lemmas 8.15 and 8.14 , |N_{B_{i-1}}(K)| , |N_{B_{i-1}}(z_{K})| , |N_{B_{i+1}}(J)| , |N_{B_{i+1}}(z_{J})|>\left\lceil\frac{\omega}{4}\right\rceil . By Lemmas 8.11 and 8.14 , we have |N_{B_{i-2}}(K)|\geq|N_{B_{i-2}}(J)|>\left\lceil\frac{\omega}{4}\right\rceil and |N_{B_{i+2}}(J)|\geq|N_{B_{i+2}}(K)|>\left\lceil\frac{\omega}{4}\right\rceil . If there is an edge uv with u\in B_{i-2}\setminus N(K) and v\in B_{i+2}\setminus N(J) , then K-N_{B_{i-1}}(K)-N_{B_{i-1}}(z_{K})-u-v-N_{B_{i+1}}(z_{J})-N_{B_{i+1}}(J) is an induced P_{7} . So B_{i-2}\setminus N(K) is anticomplete to B_{i+2}\setminus N(J) . Let u_{K}\in K be the vertex that has minimum neighborhood in H . Let p\in N_{B_{i+2}}(u_{K}) . We show that p is complete to B_{i-2}\cup B_{i+1} . Suppose not. By Lemma 8.1 (4), there is a vertex x in N_{B_{i+2}}(K) that is complete to B_{i-2}\cup B_{i+1} . By assumption x\notin N_{B_{i+2}}(u_{K}) . If p has non-neighbor y\in B_{i+1} , then u_{K}-p-x-y is a P_{4} -structure. So, p is complete to B_{i+1} . Hence, we may assume that p has a non-neighbor y^{\prime}\in B_{i-2} . Since p is complete to N_{B_{i-2}}(K) , y^{\prime}\in B_{i-2}\setminus N(K) . Then u_{K}-p-x-y^{\prime}-N_{B_{i-1}}(y^{\prime})-N_{B_{i-1}}(u_{K})-u_{K} is an induced C_{6} . So, p is complete to B_{i-2} . Therefore, p is complete to K\cup B_{i-2}\cup B_{i+1} . We set this vertex to be q_{i+2} . By symmetry, there is a vertex in N_{B_{i-2}}(J) complete to J\cup B_{i-1}\cup B_{i+2} . We set this vertex to be q_{i-2} . For X\in\{K,J\} , let v_{X},v_{X}^{\prime}\in X be the vertices with the largest and second-largest neighborhoods in H , respectively. Let w_{K},w_{K}^{\prime}\in B_{i-2}\setminus N(K) be the first and second maximal in B_{i-2}\setminus N(K) , respectively. Let w_{J},w_{J}^{\prime}\in B_{i+2}\setminus N(J) be the first and second maximal in B_{i+2}\setminus N(J) , respectively. Note that v_{K}^{\prime},v_{J}^{\prime},w_{K}^{\prime} and w_{J}^{\prime} may not exist. Let a,b\in N_{B_{i-1}}(v_{K}^{\prime\prime}) where v_{K}^{\prime\prime}\in K that has a minimum neighborhood in H and let c,d\in N_{B_{i+1}}(v_{J}^{\prime\prime}) where v_{J}^{\prime\prime}\in J that has a minimum neighborhood in H . We set F=(F_{1},F_{2},F_{3},F_{4},F_{5},O) as follows: • F_{1}=\{q_{i-2},w_{K},w_{K}^{\prime}\} , • F_{2}=\{q_{i-1},a,b\} , • F_{3}=\{q_{i}\} , • F_{4}=\{q_{i+1},c,d\} , • F_{5}=\{q_{i+2},w_{J},w_{J}^{\prime}\} , • O=\{v_{K},v_{K}^{\prime},v_{J},v_{J}^{\prime}\} . We can assign color 1 to q_{i-1} , q_{i+1} , v_{K} and v_{J} , color 2 to q_{i} , v_{K}^{\prime} , v_{J}^{\prime} , w_{K} and w_{J} , color 3 to q_{i-2} and c , color 4 to q_{i+2} and a , and color 5 to w_{K}^{\prime} , w_{J}^{\prime} , b and d . So F is 5-colorable (see Figure 7 ). B_{i-2} N_{B_{i-2}}(K) q_{i-2} w_{K} w_{K}^{\prime} B_{i-1} N_{B_{i-1}}(K) N_{B_{i-1}}(z_{K}) a b q_{i-1} B_{i} q_{i} B_{i+1} N_{B_{i+1}}(J) N_{B_{i+1}}(z_{J}) c d q_{i+1} B_{i+2} N_{B_{i+2}}(J) q_{i+2} w_{J}^{\prime} w_{J} K v_{K} v_{K}^{\prime} J v_{J} v_{J}^{\prime} Figure 7 : Illustration of F and a 5-coloring of F . For convenience, all vertices of color 1 are shown in pink, all vertices of color 2 in green, all vertices of color 3 in blue, all vertices of color 4 in red and all vertices of color 5 in yellow. Note that v_{K}^{\prime} , v_{J}^{\prime} , w_{K}^{\prime} and w_{J}^{\prime} may not exist. Next, we show that every maximal clique of size \omega-j with j\in\{0,1,2,3\} of G contains at least 4-j vertices from F . Let M be a maximal clique of G . We consider the following cases depending on all possible locations of M . Since K and J play symmetric roles, it suffices to consider K\cup N_{H}(K) , B_{i-2}\cup B_{i+2} , B_{i-2}\cup B_{i-1} and B_{i-1}\cup B_{i} . (i) M\subseteq K\cup N_{H}(K) (or J\cup N_{H}(J) ) Note that q_{i-2} and v_{K} are universal in K\cup N_{H}(K) . We first assume that M\subseteq K\cup N_{B_{i-2}}(K)\cup N_{B_{i-1}}(K) . Since a,b\in N_{B_{i-1}}(v_{K}^{\prime\prime}) , a and b are complete to K . By Lemma 8.1 , a and b are complete to N_{B_{i-2}}(K) . Hence, q_{i-2},v_{K},a,b\in M . We now assume that M\subseteq K\cup N_{B_{i-2}}(K)\cup N_{B_{i+2}}(K) . Note that q_{i+2} is universal in K\cup N_{B_{i-2}}(K)\cup N_{B_{i+2}}(K) . So, q_{i-2},q_{i+2},v_{K}\in M . Suppose that v_{K}^{\prime}\notin M . This implies that |M|=\omega and we have either |M\cap N_{B_{i-2}}(K)|\geq\frac{\omega-1}{2} or |M\cap N_{B_{i+2}}(K)|\geq\frac{\omega-1}{2} . It follows that either N_{B_{i-2}}(K)\cup N_{B_{i-1}}(K)\cup N_{B_{i-1}}(z_{K}) or N_{B_{i+2}}(K)\cup N_{B_{i+1}}(J)\cup N_{B_{i+1}}(z_{J}) is a clique of size larger than \omega , a contradiction. (ii) M\subseteq B_{i-2}\cup B_{i-1} (or B_{i+1}\cup B_{i+2} ) Note that q_{i-2} and q_{i-1} are universal in B_{i-2}\cup B_{i-1} . So, q_{i-2},q_{i-1}\in M . Since a and b are universal in N_{B_{i-2}}(K)\cup B_{i-1} , we have either a,b\in M or w_{K}\in M . If a,b\in M , then we are done. So, we may assume that w_{K}\in M . Suppose that w_{K}^{\prime}\notin M . This implies that |M|=\omega . Then (M\setminus\{w_{K}\})\cup\{a,b\} is a clique of size larger than \omega , a contradiction. (iii) M\subseteq B_{i-1}\cup B_{i} (or B_{i}\cup B_{i+1} ) Note that q_{i-1} and q_{i} are universal in B_{i-1}\cup B_{i} . By Lemma 8.10 , a and b are complete to B_{i} and so a and b are universal in B_{i-1}\cup B_{i} . Therefore, q_{i-1},q_{i},a,b\in M . (vii) M\subseteq B_{i-2}\cup B_{i+2} Note that q_{i-2} and q_{i+2} are universal in B_{i-2}\cup B_{i+2} . So, q_{i-2},q_{i+2}\in M . Since B_{i-2}\setminus N(K) is anticomplete to B_{i+2}\setminus N(J) , either M\subseteq N_{B_{i-2}}(K)\cup B_{i+2} or M\subseteq N_{B_{i+2}}(J)\cup B_{i-2} . By symmetric, we may assume that M\subseteq N_{B_{i-2}}(K)\cup B_{i+2} . If w_{J},w_{J}^{\prime}\in M , then we are done. Let j=|M\cap\{w_{J},w_{J}^{\prime}\} . Then, we have |M|=\omega-1+j and it follows that either |M\cap N_{B_{i-2}}(K)|\geq\frac{\omega-1}{2} or |M\cap N_{B_{i+2}}(J)|\geq\frac{\omega-1}{2} . Then either N_{B_{i-2}}(K)\cup N_{B_{i-1}}(K)\cup N_{B_{i-1}}(z_{K}) or N_{B_{i+2}}(J)\cup N_{B_{i+1}}(J)\cup N_{B_{i+1}}(z_{J}) is a clique of size larger than \omega , a contradiction. Therefore, F is a (4,5) -good subgraph. ∎ 13.3 Distance 0 In this subsection, we handle the third outcome of Lemma 13.1 . We assume that A_{3}=A_{3}(i)\neq\emptyset has at most two components. We first present several simple lemmas. Lemma 13.4 . There exists a clique of size larger than \left\lceil\frac{\omega}{4}\right\rceil in B_{i-2} complete to B_{i+2} . By symmetry, there exists a clique of size larger than \left\lceil\frac{\omega}{4}\right\rceil in B_{i+2} complete to B_{i-2} . Proof. Let v be a vertex in B_{i+2} with minimal neighborhood in H . By the minimality of v , N(v)\setminus B_{i-2} is a clique and so N_{B_{i-2}}(v) is a clique of size larger than \left\lceil\frac{\omega}{4}\right\rceil complete to B_{i+2} . ∎ Lemma 13.5 . If B_{i-1} is complete to B_{i-2} and B_{i+1} is complete to B_{i+2} , then B_{i-2} and B_{i+2} are complete. Proof. Let z\in B_{i-2} be a minimal vertex in B_{i-2} and z^{\prime}\in B_{i+2}\setminus N(z) be a maximal vertex in B_{i+2}\setminus N(z) . Let M\subseteq B_{i-2}\cup B_{i+2} be a maximum clique of G . We show that M contains one of z and z^{\prime} . Suppose for a contradiction that z,z^{\prime}\notin M . If M\cap B_{i+2}\subseteq N_{B_{i+2}}(z) , then M\cup\{z\} is also a clique, a contradiction. So (M\cap B_{i+2})\setminus N(z)\neq\emptyset and let a\in(M\cap B_{i+2})\setminus N(z) . If M\cap B_{i-2}\subseteq N_{B_{i-2}}(z^{\prime}) , then M\cup\{z^{\prime}\} is also a clique, a contradiction. So (M\cap B_{i-2})\setminus N(z^{\prime})\neq\emptyset and let b\in(M\cap B_{i-2})\setminus N(z^{\prime}) . It follows that ab\in E(G) and z^{\prime}b\notin E(G) , which contradicts the choice of z^{\prime} . Then \{z,z^{\prime},q_{i}\} is a good stable set of G . ∎ Lemma 13.6 . If |N_{B_{i}}(A_{3}(i))|=1 , then every vertex in (B_{i-1}\cup B_{i+1})\setminus N(A_{3}(i)) has a neighbor in B_{i}\setminus N(A_{3}(i)) . Proof. Let K be a non-empty component of A_{3}(i) and J be the possible second component of A_{3}(i) . By Lemma 4.10 , we may assume that B_{i}\setminus N(K) is anticomplete to N_{B_{i+1}}(K) . Suppose that this lemma is not true. We first assume that there exists x\in B_{i+1}\setminus N(A_{3}(i)) anticomplete to B_{i}\setminus\{q_{i}\} . Then we may assume that x is minimal in B_{i+1}\setminus N(A_{3}(i)) . By Lemma 8.10 , N_{B_{i+1}}(K) is complete to B_{i+2} . If B_{i}\setminus N(J) is anticomplete to N_{B_{i+1}}(J) , then N_{B_{i+1}}(J) is complete to B_{i+2} . If B_{i}\setminus N(J) is anticomplete to N_{B_{i-1}}(J) , then by Lemma 8.1 (4), for any vertex u\in N_{B_{i+1}}(J) and any vertex v\in B_{i+1}\setminus N(J) , N(v)\cap H\subseteq N(u)\cap H . So x is also minimal in B_{i+1} . Hence, B_{i+1}\cup N_{B_{i+2}}(x) is a clique and so d(x)\leq\omega . We now assume that there exists x\in B_{i-1}\setminus N(A_{3}(i)) anticomplete to B_{i}\setminus N(A_{3}(i)) . Let y\in B_{i-1}\setminus N(A_{3}(i)) be a minimal vertex in B_{i-1}\setminus N(A_{3}(i)) . It suffices to show that y is minimal in B_{i-1} . By assumption, y is anticomplete to B_{i}\setminus N(A_{3}(i)) . Suppose that there exist a\in B_{i-2} and b\in N_{B_{i-1}}(A_{3}(i)) such that ab\notin E(G) and ay\in E(G) . Let c\in A_{3}(i)\cap N(b) . Then a-y-b-c contradicts Lemma 4.11 . This completes the proof. ∎ Lemma 13.7 . Let J_{1} be a non-empty component of A_{3}(i) and J_{2} be the possible second component of A_{3}(i) . Suppose that B_{i}\setminus N(J_{s}) is anticomplete to B_{i+1}\cap N(J_{s}) and B_{i-1}\setminus N(J_{s})\neq\emptyset for s=1,2 . If every vertex in (B_{i-1}\setminus N(A_{3}(i)))\cup(B_{i+1}\setminus N(A_{3}(i))) has a neighbor in B_{i}\setminus N(A_{3}(i)) , then B_{j} is complete to B_{j+1} for j\neq i . Proof. Note that q_{i+1}\in B_{i+1}\setminus N(A_{3}(i)) . By assumption and Lemma 8.10 , B_{i+1} is complete to B_{i+2} . Next we show that B_{i-1} is complete to B_{i-2}\cup B_{i} . Suppose first that B_{i-1}\setminus N(A_{3}(i))\neq\emptyset . By assumption and Lemma 8.7 , B_{i-1}\setminus N(K) is complete to B_{i-2}\cup B_{i} . By Lemma 8.1 (4), N_{B_{i-1}}(K) is complete to B_{i-2}\cup B_{i} . Hence, B_{i-1} is complete to B_{i-2}\cup B_{i} . Now suppose that B_{i-1}\setminus N(A_{3}(i))=\emptyset . It follows that J_{2}\neq\emptyset , J_{1} and J_{2} have disjoint neighborhoods in B_{i-1} and B_{i-1}=N_{B_{i-1}}(J_{1})\cup N_{B_{i-1}}(J_{2}) . By assumption and Lemma 8.7 , B_{i-1} is complete to B_{i-2}\cup B_{i} . By Lemma 13.5 , B_{i-2} and B_{i+2} are complete. ∎ We first deal with the following special case. Lemma 13.8 (Coloring H\cup A_{3}(i) with four complete pairs) . Let K and J be two components of A_{3}(i) ( It is also possible that J=\emptyset ) and G=H\cup K\cup J . If B_{j} and B_{j+1} complete for j\neq i , then \chi(G)\leq\left\lceil\frac{5}{4}\omega\right\rceil . Proof. We assume that \omega(J)\leq\omega(K) . By assumption and Lemma 4.10 , we may assume that B_{i}\setminus N(K) is anticomplete to N_{B_{i+1}}(K) , and B_{i}\setminus N(J) is anticomplete to N_{B_{i+1}}(J) . By Lemmas 8.14 and 13.4 , we have |B_{j}|>\left\lceil\frac{\omega}{4}\right\rceil for j\in\{i-2,i-1,i+2\} . By Lemma 8.6 , there exist vertices w_{K}\in K universal in K and w_{j}\in J universal in J such that N_{H}(K)=N_{H}(w_{K}) and N_{H}(J)=N_{H}(w_{J}) . It follows from Lemma 8.14 that |N_{B_{j}}(K)|>\left\lceil\frac{\omega}{4}\right\rceil and |N_{B_{j}}(J)|>\left\lceil\frac{\omega}{4}\right\rceil for j\in\{i-1,i+1\} . Let z\in B_{i}\setminus(N(K)\cup N(J)) be a minimal vertex in B_{i}\setminus(N(K)\cup N(J)) . Then, N_{B_{i+1}}(z)\subseteq B_{i+1}\setminus(N(K)\cup N(J)) . Since every vertex in N_{B_{i}}(K)\cup N_{B_{i}}(J) has a neighbor in N_{B_{i+1}}(K)\cup N_{B_{i+1}}(J) , we obtain z is also minimal in B_{i} . By applying a small vertex argument, |B_{i+1}\setminus(N(K)\cup N(J))|>\left\lceil\frac{\omega}{4}\right\rceil . If N_{B_{i+1}}(K) and N_{B_{i+1}}(J) are disjoint, then B_{i+1}\cup B_{i+2} is a clique of size larger than \omega . So N_{B_{i+1}}(K) and N_{B_{i+1}}(J) are comparable. We now determine a maximum clique in G . By Lemma 8.1 (1), q_{i}\in N_{B_{i}}(K)\cap N_{B_{i}}(J) and q_{i+1}\in B_{i+1}\setminus(N(K)\cup N(J)) . Let p\in B_{i}\setminus(N(K)\cup N(J)) be a maximal vertex in B_{i}\setminus(N(K)\cup N(J)) and let r\in N_{B_{i+1}}(K)\cap N_{B_{i+1}}(J) . Then, • If B_{i-2}\cup B_{i-1} is not a maximum clique, then \{q_{i},q_{i+2}\} is a good stable set. • If B_{i-1}\cup B_{i} is not a maximum clique, then \{q_{i-2},q_{i+1},w_{K},w_{J}\} is a good stable set. • If B_{i+1}\cup B_{i+2} is not a maximum clique, then \{q_{i-2},q_{i}\} is a good stable set. So B_{i-2}\cup B_{i-1} , B_{i-1}\cup B_{i} and B_{i+1}\cup B_{i+2} are maximum clique in G . We now develop the coloring strategy. By Lemma 8.5 , K and J are P_{4} -free and so we have \chi(K)=\omega(K) and \chi(J)=\omega(J) . Let \omega(K)=\alpha\omega and |B_{i-1}|=\beta\omega . By Lemma 8.8 , |B_{i}\setminus N(K)|\geq\alpha\omega , and so we can assume that |B_{i}\setminus N(K)|=(\alpha+\gamma)\omega for some \gamma\geq 0 . Observe that \gamma\omega\in\mathbb{N} because \gamma\omega=|B_{i}\setminus N(K)|-\omega(K) . Since B_{i-2}\cup B_{i-1} and B_{i-1}\cup B_{i} are maximum clique, this implies that |B_{i-2}|=(1-\beta)\omega and |N_{B_{i}}(K)|=(1-\alpha-\beta-\gamma)\omega . As B_{i-2}\cup B_{i+2} may not be a maximum clique, let |B_{i+2}|=(\beta-\delta)\omega with \delta\geq 0 . Since B_{i+1}\cup B_{i+2} is a maximum clique, |B_{i+1}|=(1-\beta+\delta)\omega . If \omega(K)\geq\frac{3}{4}\omega , then |B_{i}\setminus N(K)|\geq\frac{3}{4}\omega and this implies that B_{i-1}\cup B_{i} is a clique of size larger than \omega . So we may assume that \omega(K)<\frac{3}{4}\omega . (i) \frac{1}{4}\omega<\omega(K)<\frac{3}{4}\omega We now partition K into three parts (X,Y,Z) of color sizes x\omega , y\omega , and \left\lceil\frac{\omega}{4}\right\rceil where x\omega,y\omega\leq\left\lceil\frac{\omega}{4}\right\rceil . Since x\omega+y\omega=\alpha\omega-\left\lceil\frac{\omega}{4}\right\rceil<\frac{1}{2}\omega , this implies that we can choose x\omega and y\omega that satisfies x\omega,y\omega\leq\left\lceil\frac{\omega}{4}\right\rceil . We then color some vertices in G using x\omega+y\omega+2\cdot\left\lceil\frac{\omega}{4}\right\rceil colors as follows. • Assign the colors a_{1},\ldots,a_{x\omega} to the vertices of X , and x\omega vertices in B_{i-2} . • Assign the colors b_{1},\ldots,b_{y\omega} to the vertices of Y , and y\omega vertices in B_{i+2} . • Assign the colors c_{1},\ldots,c_{\left\lceil\frac{\omega}{4}\right\rceil} to the vertices of Z , \left\lceil\frac{\omega}{4}\right\rceil vertices in B_{i+1}\setminus(N(K)\cup N(J)) , and \left\lceil\frac{\omega}{4}\right\rceil vertices in B_{i-2} . • Assign colors d_{1},\ldots,d_{\left\lceil\frac{\omega}{4}\right\rceil} to \left\lceil\frac{\omega}{4}\right\rceil vertices in N_{B_{i+1}}(K) , and \left\lceil\frac{\omega}{4}\right\rceil vertices in B_{i-2} . • The vertices of J are colored starting from c_{1} and proceeding sequentially through c_{1},\ldots,c_{\left\lceil\frac{\omega}{4}\right\rceil},a_{1},\ldots,a_{x\omega},b_{1},\ldots,b_{y\omega} using exactly \omega(J) colors. • Assign the colors d_{1},\ldots,d_{\left\lceil\frac{\omega}{4}\right\rceil},a_{1},\ldots,a_{x\omega},b_{1},\ldots,b_{y\omega} to \alpha\omega vertices in B_{i}\setminus N(K) in such a way that non-neighbors of J in B_{i}\setminus N(K) are colored first, in order. |B_{i-2}|=(1-\beta)\omega B_{i-2} a_{1},\ldots,a_{x\omega} c_{1},\ldots,c_{\left\lceil\frac{\omega}{4}\right\rceil} d_{1},\ldots,d_{\left\lceil\frac{\omega}{4}\right\rceil} |B_{i-1}|=\beta\omega B_{i-1} |B_{i}|=(1-\beta)\omega B_{i} b_{1},\ldots,b_{y\omega} a_{1},\ldots,a_{x\omega} d_{1},\ldots,d_{\left\lceil\frac{\omega}{4}\right\rceil} |B_{i+1}|=(1-\beta+\delta)\omega B_{i+1} d_{1},\ldots,d_{\left\lceil\frac{\omega}{4}\right\rceil} c_{1},\ldots,c_{\left\lceil\frac{\omega}{4}\right\rceil} |B_{i+2}|=(\beta-\delta)\omega B_{i+2} b_{1},\ldots,b_{y\omega} K a_{1},\ldots,a_{x\omega} b_{1},\ldots,b_{y\omega} c_{1},\ldots,c_{\left\lceil\frac{\omega}{4}\right\rceil} J c_{1},\ldots,c_{\left\lceil\frac{\omega}{4}\right\rceil}, a_{1},\ldots,a_{x\omega}, b_{1},\ldots Figure 8 : The coloring strategy for Case (i). For each j\in\{i,i+1\} , the black line in B_{j} distinguishes N_{B_{j}}(K) from B_{j}\setminus N(K) . Two black line between B_{j} and B_{j+1} mean that the two sets are complete to each other. The remaining graph consists of the red-boxed parts. The coloring strategy is illustrated in the above figure. By Lemma 4.13 , N_{B_{i}}(K)\subseteq N_{B_{i}}(J) or N_{B_{i}}(J)\subseteq N_{B_{i}}(K) . By Lemma 8.8 and \omega(K)\geq\omega(J) , the number of non-neighbors of J in B_{i}\setminus N(K) is at least \omega(J) and so the coloring used on J\cup B_{i} is proper. The remaining graph consists of the red-boxed parts in the figure, and it suffices to show that the remaining graph G^{\prime} is \left(\left\lceil\frac{5}{4}\omega\right\rceil-x\omega-y\omega-2\cdot\left\lceil\frac{\omega}{4}\right\rceil\right) -colorable. Since |B_{i+1}\setminus(N(K)\cup N(J))|,|N_{B_{i+1}}(K)\cap N_{B_{i+1}}(J)|>\left\lceil\frac{\omega}{4}\right\rceil , the each red-boxed parts in B_{i+1} have non-negative size. Since |B_{i+2}|>\left\lceil\frac{\omega}{4}\right\rceil , the red-boxed part in B_{i+2} also has non-negative size. Note that the red-boxed part in B_{i-2} could be empty. Then we only need to consider the worst case that B_{i} is complete to B_{i+1} in G^{\prime} and so G^{\prime} is a hyperhole C_{5}[s_{1},s_{2},s_{3},s_{4},s_{5}] , where • s_{1}=(1-\beta)\omega-x\omega-2\cdot\left\lceil\frac{\omega}{4}\right\rceil , • s_{2}=\beta\omega , • s_{3}=(1-\beta)\omega-x\omega-y\omega-\left\lceil\frac{\omega}{4}\right\rceil , • s_{4}=(1-\beta+\delta)\omega-2\cdot\left\lceil\frac{\omega}{4}\right\rceil , • s_{5}=(\beta-\delta)\omega-y\omega . We first show that \omega(G^{\prime})\leq\left\lceil\frac{5}{4}\omega\right\rceil-x\omega-y\omega-2\cdot\left\lceil\frac{\omega}{4}\right\rceil . It is equivalent to show that for every j\in[5] , s_{j}+s_{j+1}\leq\left\lceil\frac{5}{4}\omega\right\rceil-x\omega-y\omega-2\cdot\left\lceil\frac{\omega}{4}\right\rceil where the indices are taken modulo 5 (so that s_{6}=s_{1} ). For j=1,2,4,5 , the above inequality is equivalent to the following respective inequality y\omega\leq\left\lceil\frac{\omega}{4}\right\rceil\,,\,\omega\leq\left\lceil\frac{5}{4}\omega\right\rceil-\left\lceil\frac{\omega}{4}\right\rceil\,,\,x\omega\leq\left\lceil\frac{\omega}{4}\right\rceil\,,\,\omega\leq\left\lceil\frac{5}{4}\omega\right\rceil+\delta\omega, and in each case the inequality holds clearly. When j=3 , the inequality yields 2\omega\leq 2\beta\omega+\left\lceil\frac{\omega}{4}\right\rceil+\left\lceil\frac{5}{4}\omega\right\rceil-\delta\omega . Since (\beta-\delta)\omega=|B_{i+2}|>\left\lceil\frac{\omega}{4}\right\rceil and \beta\omega=|B_{i-1}|>\left\lceil\frac{\omega}{4}\right\rceil , this inequality also holds for j=3 . Hence, the inequality holds for all j\in[5] . Now we show that \left\lceil\frac{|V(G^{\prime})|}{\alpha(G^{\prime})}\right\rceil\leq\left\lceil\frac{5}{4}\omega\right\rceil-x\omega-y\omega-2\cdot\left\lceil\frac{\omega}{4}\right\rceil . Observing that \alpha(G^{\prime})=2 , it suffices to show |V(G^{\prime})|\leq 2\cdot\left\lceil\frac{5}{4}\omega\right\rceil-2x\omega-2y\omega-4\cdot\left\lceil\frac{\omega}{4}\right\rceil. Since |V(G^{\prime})|=3\omega-\beta\omega-2x\omega-2y\omega-5\cdot\left\lceil\frac{\omega}{4}\right\rceil , we obtain 3\omega\leq\beta\omega+\left\lceil\frac{\omega}{4}\right\rceil+2\cdot\left\lceil\frac{5}{4}\omega\right\rceil and this inequality is satisfied because \beta\omega\geq\left\lceil\frac{\omega}{4}\right\rceil+1 . By Lemma 2.1 , \chi(G^{\prime})=\max\{\omega(G^{\prime}),\left\lceil\frac{|V(G^{\prime})|}{\alpha(G^{\prime})}\right\rceil\}\leq\left\lceil\frac{5}{4}\omega\right\rceil-x\omega-y\omega-2\cdot\left\lceil\frac{\omega}{4}\right\rceil . (ii) \omega(K)\leq\frac{1}{4}\omega Recall that z\in B_{i}\setminus(N(K)\cup N(J)) be a minimal vertex in B_{i}\setminus(N(K)\cup N(J)) and |N_{B_{i+1}}(z)|>\left\lceil\frac{\omega}{4}\right\rceil . As z is minimal in B_{i} , N_{B_{i+1}}(z) is complete to B_{i} . Since N_{B_{i}}(K)\cup N_{B_{i+1}}(K)\cup N_{B_{i+1}}(z) is a clique, |N_{B_{i}}(K)|<\frac{\omega}{2} . Let x\omega=\min\{|B_{i}\setminus N(K)|,\left\lceil\frac{1}{4}\omega\right\rceil\} . We then color some vertices of G using \alpha\omega+x\omega+\left\lceil\frac{1}{4}\omega\right\rceil colors as follows. Recall that |B_{i}\setminus N(K)|=(\alpha+\gamma)\omega . • Assign the colors a_{1},\ldots,a_{\alpha\omega} to the vertices of K , \alpha\omega vertices in B_{i+1}\setminus(N(K)\cup N(J)) , and \alpha\omega vertices in B_{i-2} . • Assign the colors a_{1},\ldots,a_{\omega(J)} to the vertices of J . • Assign the colors b_{1},\ldots,b_{x\omega} to the vertices of B_{i}\setminus N(K) , x\omega vertices in N_{B_{i+1}}(K) , and x\omega vertices in B_{i-2} . • Assign colors c_{1},\ldots,c_{\alpha\omega} to \alpha\omega vertices in B_{i} excluding the vertices that have already been colored with b_{1},\ldots,b_{x\omega} , and \alpha\omega vertices in B_{i-2} . • Assign colors d_{1},\ldots,d_{\left\lceil\frac{\omega}{4}\right\rceil-\alpha\omega} to \left\lceil\frac{\omega}{4}\right\rceil-\alpha\omega vertices in B_{i+1}\setminus(N(K)\cup N(J)) , and \left\lceil\frac{\omega}{4}\right\rceil-\alpha\omega vertices in B_{i-1} . |B_{i-2}|=(1-\beta)\omega B_{i-2} a_{1},\ldots,a_{\alpha\omega} b_{1},\ldots,b_{x\omega} c_{1},\ldots,c_{\alpha\omega} |B_{i-1}|=\beta\omega B_{i-1} d_{1},\ldots,d_{\left\lceil\frac{\omega}{4}\right\rceil-\alpha\omega} |B_{i}|=(1-\beta)\omega B_{i} c_{1},\ldots,c_{\alpha\omega} b_{1},\ldots,b_{x\omega} |B_{i+1}|=(1-\beta+\delta)\omega B_{i+1} b_{1},\ldots,b_{x\omega} a_{1},\ldots,a_{\alpha\omega} d_{1},\ldots,d_{\left\lceil\frac{\omega}{4}\right\rceil-\alpha\omega} |B_{i+2}|=(\beta-\delta)\omega B_{i+2} K a_{1},\ldots,a_{\alpha\omega} J a_{1},\ldots,a_{\omega(J)} Figure 9 : The coloring strategy for Case (ii). For each j\in\{i,i+1\} , the black line in B_{j} distinguishes N_{B_{j}}(K) from B_{j}\setminus N(K) . Two black line between B_{j} and B_{j+1} mean that the two sets are complete to each other. The remaining graph consists of the red-boxed parts. The coloring strategy is illustrated in the above figure. The remaining graph G^{\prime} consists of the red-boxed parts in the figure, and it suffices to show that G^{\prime} is \left(\left\lceil\frac{5}{4}\omega\right\rceil-\alpha\omega-z\omega-\left\lceil\frac{1}{4}\omega\right\rceil\right) -colorable. Since |B_{i+1}\setminus(N(K)\cup N(J))|>\left\lceil\frac{\omega}{4}\right\rceil , |N_{B_{i+1}}(K)|>\left\lceil\frac{\omega}{4}\right\rceil and x\omega\leq\left\lceil\frac{\omega}{4}\right\rceil , the each red-boxed parts in B_{i+1} have non-negative size. Since |B_{i-1}|>\left\lceil\frac{\omega}{4}\right\rceil , the red-boxed parts in B_{i-1} has non-negative size. Note that the red-boxed part in B_{i-2} and B_{i} could be empty. Then we only need to consider the worst case that B_{i} is complete to B_{i+1} in G^{\prime} and so G^{\prime} is a hyperhole C_{5}[s_{1},s_{2},s_{3},s_{4},s_{5}] , where • s_{1}=(1-\beta)\omega-2\alpha\omega-x\omega , • s_{2}=\beta\omega-\left\lceil\frac{1}{4}\omega\right\rceil+\alpha\omega , • s_{3}=(1-\beta)\omega-\alpha\omega-x\omega , • s_{4}=(1-\beta+\delta)\omega-x\omega-\left\lceil\frac{1}{4}\omega\right\rceil , • s_{5}=(\beta-\delta)\omega . We first show that \omega(G^{\prime})\leq\left\lceil\frac{5}{4}\omega\right\rceil-\alpha\omega-x\omega-\left\lceil\frac{1}{4}\omega\right\rceil . It suffices to show that for every j\in[5] , s_{j}+s_{j+1}\leq\left\lceil\frac{5}{4}\omega\right\rceil-\alpha\omega-x\omega-\left\lceil\frac{1}{4}\omega\right\rceil where the indices are taken modulo 5 (so that s_{6}=s_{1} ). From the j=1,2,4,5 , the above inequality is equivalent to the following respective inequality \omega\leq\left\lceil\frac{5}{4}\omega\right\rceil\,,\,\alpha\omega\leq\left\lceil\frac{1}{4}\omega\right\rceil\,,\,\alpha\omega\leq\left\lceil\frac{1}{4}\omega\right\rceil\,,\,0\leq\alpha\omega+\delta\omega, and in each case the inequality holds. When j=3 , the inequality yields 2\omega\leq 2\beta\omega+x\omega+\left\lceil\frac{5}{4}\omega\right\rceil-\delta\omega . Since \beta\omega-\delta\omega=|B_{i+2}|>\left\lceil\frac{1}{4}\omega\right\rceil and \beta\omega+x\omega=\omega-|N_{B_{i}}(K)|>\frac{\omega}{2} , this inequality also holds for j=3 . Hence, the inequality holds for all j\in[5] . Now we show that \left\lceil\frac{|V(G^{\prime})|}{\alpha(G^{\prime})}\right\rceil\leq\left\lceil\frac{5}{4}\omega\right\rceil-\alpha\omega-x\omega-\left\lceil\frac{1}{4}\omega\right\rceil . Observing that \alpha(G^{\prime})=2 , it suffices to show |V(G^{\prime})|\leq 2\cdot\left\lceil\frac{5}{4}\omega\right\rceil-2\alpha\omega-2x\omega-2\cdot\left\lceil\frac{1}{4}\omega\right\rceil. Since |V(G^{\prime})|=3\omega-\beta\omega-2\alpha\omega-3x\omega-2\cdot\left\lceil\frac{1}{4}\omega\right\rceil , the inequality is equivalent to 3\omega\leq\beta\omega+x\omega+2\cdot\left\lceil\frac{5}{4}\omega\right\rceil , which holds as \beta\omega+x\omega>\frac{1}{2}\omega . By Lemma 2.1 , \chi(G^{\prime})=\max\{\omega(G^{\prime}),\left\lceil\frac{|V(G^{\prime})|}{\alpha(G^{\prime})}\right\rceil\}\leq\left\lceil\frac{5}{4}\omega\right\rceil-\alpha\omega-x\omega-\left\lceil\frac{1}{4}\omega\right\rceil . ∎ We then consider cases depending on the number of components of A_{3}(i) and the number of neighbors of each component in B_{i} . Lemma 13.9 (One component with one neighbor in B_{i} ) . If A_{3}(i) is connected and |N(A_{3}(i))\cap B_{i}|=1 , then \chi(G)\leq\left\lceil\frac{5}{4}\omega\right\rceil . Proof. For convenience, let K=A_{3}(i) . By Lemma 8.6 , let w\in K be a universal vertex in K such that N_{H}(w)=N_{H}(K) . By Lemma 4.10 , we may assume that B_{i}\setminus N(K) is anticomplete to N_{B_{i+1}}(K) . So N_{B_{i+1}}(K)=\{q_{i}\} . By Lemma 8.10 , N_{B_{i+1}}(K) is complete to B_{i+2} . Let X_{i+1}=B_{i+1}\setminus N(K) . By Lemma 13.6 , each vertex in X_{i+1} has a neighbor in B_{i}\setminus\{q_{i}\} . By Lemma 8.10 , X_{i+1} is complete to B_{i+2} . It follows that B_{i+1} is complete to B_{i+2} . Suppose that B_{i-1}\setminus N(K)\neq\emptyset and let X_{i-1}=B_{i-1}\setminus N(K) . By Lemma 13.6 , each vertex in X_{i-1} has a neighbor in B_{i}\setminus\{q_{i}\} . By Lemmas 13.7 and 13.8 , we have \chi(G)\leq\left\lceil\frac{5}{4}\omega\right\rceil . So we may assume that X_{i-1}=\emptyset . Let z\in B_{i}\setminus\{q_{i}\} be a minimal vertex in B_{i}\setminus\{q_{i}\} . By Lemma 8.15 , |N_{B_{i+1}}(z)|>\left\lceil\frac{\omega}{4}\right\rceil . By Lemma 13.4 , |B_{j}|>\left\lceil\frac{\omega}{4}\right\rceil for j\in\{i-2,i+2\} . This implies that \omega\geq 5 . By Lemma 13.1 (3), K is a blowup of P_{3} and so \alpha(K)\leq 2 . Case 1: K contains two non-adjacent vertices. Since K is a P_{4} -free, let u and v be two simplicial vertices in K . Let u^{\prime}\in N_{G}[u]\cap K be a vertex minimizing |N_{H}(u^{\prime})| and v^{\prime}\in N_{G}[v]\cap K be a vertex minimizing |N_{H}(v^{\prime})| . By Lemma 2.6 , u^{\prime}v^{\prime}\notin E(G) . By Lemma 8.14 , we have |N_{B_{j}}(u^{\prime})|,|N_{B_{j}}(v^{\prime})|>\left\lceil\frac{\omega}{4}\right\rceil for each j\in\{i-1,i+1\} . If N_{B_{i+1}}(u^{\prime}) and N_{B_{i+1}}(v^{\prime}) are disjoint, then B_{i+1}\cup B_{i+2} is a clique of size larger than \omega . So N_{B_{i+1}}(u^{\prime}) and N_{B_{i+1}}(v^{\prime}) are comparable. By Lemma 4.12 , N_{B_{i-1}}(u^{\prime}) and N_{B_{i-1}}(v^{\prime}) are disjoint. Since N_{B_{i+1}}(K) and N_{B_{i+1}}(z) are disjoint, we have N_{B_{i-1}}(u^{\prime})\cup N_{B_{i-1}}(v^{\prime})\subseteq N_{B_{i-1}}(z) and it follows from the choice of z that N_{B_{i-1}}(u^{\prime})\cup N_{B_{i-1}}(v^{\prime}) is complete to B_{i} . By Lemma 8.10 , N_{B_{i-1}}(u^{\prime})\cup N_{B_{i-1}}(v^{\prime}) is complete to B_{i-2} . Note that B_{i-1}\cup\{q_{i}\} and B_{i+1}\cup\{q_{i}\} are not maximum clique, since B_{i-1}\cup\{q_{i},w\} and B_{i+1}\cup B_{i+2} are cliques. By the definition of z , there is a vertex in N_{B_{i+1}}(z) complete to B_{i}\cup B_{i+2} . We set this vertex to be q_{i+1} . Let a\in N_{B_{i-1}}(u^{\prime}) and let b\in N_{B_{i-1}}(v^{\prime}) . Let a^{\prime}\in K be a vertex non-adjacent to a maximizing |N_{H}(a^{\prime})| and let b^{\prime}\in K be a vertex non-adjacent to b maximizing |N_{H}(b^{\prime})| . If a^{\prime}u^{\prime}\in E(G) , then a^{\prime}a\in E(G) by the choice of u , a contradiction. So a^{\prime}u^{\prime}\notin E(G) and thus a^{\prime}v^{\prime}\in E(G) by \alpha(K)\leq 2 . By symmetry, b^{\prime}u^{\prime}\in E(G) . Let c\in N_{B_{i+1}}(u^{\prime})\cap N_{B_{i+1}}(v^{\prime}) . Then c is complete to K . Let q_{j}^{\prime}\in B_{j} be the second largest vertex in B_{j} for j\in\{i-2,i+2\} . By Lemma 13.4 , we have q^{\prime}_{i+2} is complete to B_{i-2} and q^{\prime}_{i-2} is complete to B_{i+2} . Let q^{\prime}_{i},q^{\prime\prime}_{i} be the second and third largest vertices in B_{i} , respectively. We set F=(F_{1},F_{2},F_{3},F_{4},F_{5},O) as follows: • F_{1}=\{q_{i-2},q_{i-2}^{\prime}\} , • F_{2}=\{a,b\} , • F_{3}=\{q_{i},q_{i}^{\prime},q_{i}^{\prime\prime}\} , • F_{4}=\{c,q_{i+1}\} , • F_{5}=\{q_{i+2},q_{i+2}^{\prime}\} , • O=\{w,a^{\prime},b^{\prime}\} . We can assign color 1 to q_{i-2} and q_{i} , color 2 to w , q_{i}^{\prime} and q_{i+2} , color 3 to a , a^{\prime} and q_{i+2}^{\prime} , color 4 to b , b^{\prime} and q_{i+1} , and color 5 to q_{i-2}^{\prime} , q_{i}^{\prime\prime} and c . So F is 5-colorable. Next, we show that every maximal clique of size \omega-j with j\in\{0,1,2,3\} of G contains at least 4-j vertices from F . Let M be a maximal clique of G . We consider the following cases depending on all possible locations of M . If M is contained in B_{i+1}\cup B_{i+2} or B_{i-2}\cup B_{i+2} , M contains four vertices of F . (i) M\subseteq K\cup N_{B_{i}}(K)\cup N_{B_{j}}(K) for every j\in\{i-1,i+1\} Note that w and q_{i} are universal vertex in K\cup N_{H}(K) . We first assume that M is contained in K\cup N_{B_{i-1}}(K)\cup\{q_{i}\} . By definition of u^{\prime} and v^{\prime} , M contains at least one of a and b . If a,b\in M , then we are done. So we may assume that M contains exactly one of a and b . By symmetry, we may assume that a\in M and this implies that M\cap N_{B_{i-1}}(v^{\prime})=\emptyset . By definition of b^{\prime} , we have b^{\prime}\in M and so M contains \{a,b^{\prime},q_{i},w\} . So we may assume that M\subseteq K\cup N_{B_{i+1}}(K)\cup\{q_{i}\} . Note that q_{i},w,c are universal in K\cup N_{B_{i+1}}(K) . Suppose that a^{\prime},b^{\prime}\notin M . By definition of a^{\prime} and b^{\prime} , M\cap K is complete to \{a,b\} and it follows that M\cap K is complete to N_{B_{i-1}}(u^{\prime})\cup N_{B_{i-1}}(v^{\prime}) . As (M\cap K)\cup N_{B_{i-1}}(u^{\prime})\cup N_{B_{i-1}}(v^{\prime})\cup\{q_{i}\} is a clique and |N_{B_{i-1}}(u^{\prime})\cup N_{B_{i-1}}(v^{\prime})|>2\cdot\left\lceil\frac{\omega}{4}\right\rceil , we obtain |M\cap K|<\omega-2\cdot\left\lceil\frac{\omega}{4}\right\rceil-1 . So, |M\cap N_{B_{i+1}}(K)|>2\cdot\left\lceil\frac{\omega}{4}\right\rceil and B_{i+1}\cup B_{i+2} is a clique of size larger than \omega . Hence, M\cap\{a^{\prime},b^{\prime}\}\neq\emptyset and we are done. (ii) M\subseteq B_{i-2}\cup B_{i-1} . Since a,b,q_{i-2} are universal in B_{i-2}\cup B_{i-1} , a,b,q_{i-2}\in M . Suppose that |M|=\omega and q^{\prime}_{i-2}\notin M . Then M\cap B_{i-2}=\{q_{i-2}\} and so |M\cap B_{i-1}|=\omega-1 . Then (M\cap B_{i-1})\cup\{q_{i},w\} is a clique of size \omega+1 , a contradiction. (iii) M\subseteq B_{i-1}\cup B_{i} Note that a,b,q_{i} are universal in B_{i-1}\cup B_{i} . If M=B_{i-1}\cup\{q_{i}\} , then M is not a maximum clique and a,b,q_{i}\in M . So we may assume that M\cap(B_{i}\setminus\{q_{i}\})\neq\emptyset and this implies that q_{i}^{\prime}\in M . Hence, M contains \{a,b,q_{i},q_{i}^{\prime}\} . (iv) M\subseteq B_{i}\cup B_{i+1} Note that q_{i} and q_{i+1} are universal in B_{i}\cup B_{i+1} . If M=B_{i+1}\cup\{q_{i}\} , then M is not a maximum clique and c,q_{i},q_{i+1}\in M . So we may assume that M\cap(B_{i}\setminus\{q_{i}\})\neq\emptyset . Then q_{i}^{\prime}\in M and it follows that M\cap N_{B_{i+1}}(K)=\emptyset . Thus, we can assume that |M|=\omega . If q_{i}^{\prime\prime}\notin M , then |M\cap B_{i+1}|=\omega-2 and it follows that B_{i+1}\cup B_{i+2} is a clique of size larger than \omega . So, q_{i}^{\prime\prime}\in M and the claim holds. Therefore, F is a (4,5) -good subgraph. From now on, we may assume that K is a clique. Case 2: K is a clique. Suppose that |K|=1 . For any x\in N_{B_{i+1}}(K) , we have d_{G}(x)\leq(\omega-1)+2\leq\left\lceil\frac{5}{4}\omega\right\rceil-1 since \omega\geq 5 . So we may assume that |K|\geq 2 . By Lemma 8.8 , we have |B_{i}|\geq 3 . Let w^{\prime}\in K be a vertex minimizing |N_{H}(w^{\prime})| . By Lemma 8.1 (4), there exists a vertex in N_{B_{i-1}}(w^{\prime}) that is complete to B_{i-2}\cup B_{i} . We set this vertex to be q_{i-1} . Moreover, q_{i-1} is complete to K . Let a\in N_{B_{i+1}}(w^{\prime}) . Then a is complete to K . Let v\in K be the second largest vertex in K (It is possible that v=w^{\prime} ). Note that q_{i+1}\in N_{B_{i+1}}(z) . Let q_{j}^{\prime}\in B_{j} be the second largest vertex in B_{j} for j\in\{i-2,i+2\} . By Lemma 13.4 , we have q^{\prime}_{i+2} is complete to B_{i-2} . Let q^{\prime}_{i},q^{\prime\prime}_{i} be the second and third largest vertices in B_{i} , respectively. Since |B_{i-2}|\geq 3 , the third largest vertex q_{i-2}^{\prime\prime} in B_{i-2} exists. By Lemma 13.4 , we have q_{i-2}^{\prime} and q_{i-2}^{\prime\prime} are complete to B_{i+2} . We set F=(F_{1},F_{2},F_{3},F_{4},F_{5},O) as follows: • F_{1}=\{q_{i-2},q_{i-2}^{\prime},q_{i-2}^{\prime\prime}\} , • F_{2}=\{q_{i-1}\} , • F_{3}=\{q_{i},q_{i}^{\prime},q_{i}^{\prime\prime}\} , • F_{4}=\{a,q_{i+1}\} , • F_{5}=\{q_{i+2},q_{i+2}^{\prime}\} , • O=\{w,v\} . We can assign color 1 to q_{i-2} and q_{i} , color 2 to w , q_{i}^{\prime} and q_{i+2} , color 3 to q_{i-2}^{\prime} , v and q_{i+1} , color 4 to q_{i-1} and q_{i+2}^{\prime} , and color 5 to q_{i-2}^{\prime\prime} , q_{i}^{\prime\prime} and a . So F is 5-colorable. Next, we show that every maximal clique of size \omega-j with j\in\{0,1,2,3\} of G contains at least 4-j vertices from F . Let M be a maximal clique of G . We consider the following cases depending on all possible locations of M . If M is one of B_{i+1}\cup B_{i+2} and B_{i-2}\cup B_{i+2} , M contains four vertices of F . (i) M\subseteq K\cup N_{B_{i}}(K)\cup N_{B_{j}}(K) for every j\in\{i-1,i+1\} We first assume that M is contained in K\cup(N_{B_{i-1}}(K))\cup\{q_{i}\} . Note that q_{i-1} , q_{i} and w are universal in K\cup(N_{B_{i-1}}(K))\cup\{q_{i}\} . So q_{i-1},q_{i},w\in M and we can assume that |M|=\omega . Suppose that v\notin M . Then there exists a vertex x\in B_{i-1}\cap M that is non-adjacent to v . If x has a non-neighbor x^{\prime}\in B_{i}\setminus\{q_{i}\} , then v-w-x-q_{i-2}-q_{i+2}-q_{i+1}-x^{\prime} is an induced P_{7} . By Lemma 8.1 (4), B_{i-1}\cap M is complete to B_{i} . Since |B_{i-1}\cap M|=\omega-2 and |B_{i}|\geq 3 , (B_{i-1}\cap M)\cup B_{i} is a clique of size larger than \omega . So v\in M and M contains \{q_{i-1},q_{i},w,v\} . We then assume that M\subseteq K\cup(N_{B_{i+1}}(K))\cup\{q_{i}\} . Note that q_{i},w,a are universal in K\cup N_{B_{i+1}}(K) . Suppose that v\notin M and so we can assume that |M|=\omega . It follows that |B_{i+1}\cap M|=\omega-2 . Then B_{i+1}\cup B_{i+2} is a clique of size larger than \omega . Therefore, v\in M and we are done. (ii) M\subseteq B_{i-2}\cup B_{i-1} Note that q_{i-2} and q_{i-1} are universal in B_{i-2}\cup B_{i-1} . Define q to be q_{i-2}^{\prime\prime} if |M|=\omega and to be q_{i-2}^{\prime} if |M|=\omega-1 . We show that q\in M , which completes the proof. Suppose for a contradiction that q\notin M . Then there exists a vertex x\in B_{i-1}\cap M such that xq\notin E(G) . If x has non-neighbor x^{\prime}\in B_{i}\setminus\{q_{i}\} , then q-q_{i-2}-x-w-N_{B_{i+1}}(K)-q_{i+1}-x^{\prime} is an induced P_{7} . So x is complete to B_{i} . For any y\in B_{i-1}\cap M with yq\in E(G) , we have N_{B_{i-2}\cup B_{i}}(x)\subseteq N_{B_{i-2}\cup B_{i}}(y) and so B_{i-1}\cap M is complete to B_{i} . This implies that (B_{i-1}\cup M)\cup B_{i} is a clique of size larger than \omega . This completes the claim. (iii) M\subseteq B_{i-1}\cup B_{i} Note that q_{i-1} and q_{i} are universal in B_{i-1}\cup B_{i} . Since B_{i-1}\cup\{q_{i},w\} is a clique, M\cap(B_{i}\setminus\{q_{i}\})\neq\emptyset and this implies that q_{i}^{\prime}\in M . Suppose that q_{i}^{\prime\prime}\notin M . Then there exists a vertex x\in M\cap B_{i-1} such that xq_{i}^{\prime\prime}\notin E(G) . If x has a non-neighbor x^{\prime}\in B_{i-2} , then x^{\prime}-q_{i-2}-x-w-N_{B_{i+1}}(K)-q_{i+1}-q_{i}^{\prime\prime} is an induced P_{7} . So, x is complete to B_{i-2} . For any y\in B_{i-1}\cap M with yq_{i}^{\prime\prime}\in E(G) , we have N_{B_{i-2}\cup B_{i}}(x)\subseteq N_{B_{i-2}\cup B_{i}}(y) and so B_{i-1}\cap M is complete to B_{i-2} . This implies that (B_{i-1}\cup M)\cup B_{i-2} is a clique of size larger than \omega . Therefore, q_{i}^{\prime\prime}\in M and so M contains \{q_{i-1},q_{i},q_{i}^{\prime},q_{i}^{\prime\prime}\} . (iv) M\subseteq B_{i}\cup B_{i+1} Note that q_{i} and q_{i+1} are universal in B_{i}\cup B_{i+1} . If M=B_{i+1}\cup\{q_{i}\} , then M is not a maximum clique and a,q_{i},q_{i+1}\in M . So we may assume that M\cap(B_{i}\setminus\{q_{i}\})\neq\emptyset . Then q_{i}^{\prime}\in M and it follows that M\cap N_{B_{i+1}}(K)=\emptyset . We can assume that |M|=\omega . If q_{i}^{\prime\prime}\notin M , then |M\cap B_{i+1}|=\omega-2 and it follows that B_{i+1}\cup B_{i+2} is a clique of size larger than \omega . So q_{i}^{\prime\prime}\in M and the claim holds. Therefore, F is a (4,5) -good subgraph. ∎ Next we deal with the case that A_{3}(i) is connected with at least two neighbors in B_{i} . The lemma below also works for the case that A_{3}(i) has two components with each component having at least two neighbors in B_{i} . Lemma 13.10 (One or two components with two neighbors in B_{i} ) . If each component of A_{3}(i) has at least two neighbors in B_{i} , then G has a (4,5) -good subgraph. Proof. Let K be a non-empty component of A_{3}(i) and J be a possible second component of A_{3}(i) . By Lemma 13.1 , K can be partitioned into three disjoint cliques L_{1},L_{2},L_{3} such that L_{2} is complete to L_{1}\cup L_{3} and L_{1} is anticomplete to L_{3} . We may assume that L_{1} is non-empty if L_{3} is non-empty. By Lemma 13.1 , L_{3}=\emptyset if J\neq\emptyset . Let w\in K be universal in K with maximum neighborhood in H (such a vertex exists by Lemma 8.6 ). By Lemma 8.14 , |N_{B_{i-1}}(K)|>\left\lceil\frac{\omega}{4}\right\rceil and so |N_{B_{i-1}}(K)|\geq 2 . It follows that \{w\}\cup N_{B_{i-1}}(K)\cup N_{B_{i}}(K) is a clique and so \omega\geq 5 . By Lemma 13.4 , we have |B_{i-2}|,|B_{i+2}|>\left\lceil\frac{\omega}{4}\right\rceil and so |B_{i-2}|,|B_{i+2}|\geq 3 . By Lemma 4.1 , we can order vertices in B_{j} as b_{j}^{1},b_{j}^{2},\ldots,b_{j}^{|B_{i}|}=q_{j} so that N_{B_{j-1}\cup B_{j+1}}(b_{j}^{1})\subseteq N_{B_{j-1}\cup B_{j+1}}(b_{j}^{2})\subseteq\cdots\subseteq N_{B_{j-1}\cup B_{j+1}}(b_{j}^{|B_{j}|})=B_{j-1}\cup B_{j+1} for every j\in[5] . Let q_{j}^{\prime}=b_{j}^{|B_{j}|-1} and q_{j}^{\prime\prime}=b_{j}^{|B_{j}|-2} for each j . By Lemma 8.5 , K and J are P_{4} -free. By Lemma 4.10 , we may assume that B_{i}\setminus N(w) is anticomplete to N_{B_{i+1}}(w) and q_{i}\in N_{B_{i}}(w) . By Lemma 4.10 , N_{B_{i-1}\cup B_{i+1}}(a)\subset N_{B_{i-1}\cup B_{i+1}}(b) for any a\in B_{i}\setminus N(K) and any b\in N_{B_{i}}(K) . This implies that q^{\prime}_{i}\in N_{B_{i}}(K) . By symmetry, q_{i},q^{\prime}_{i}\in N_{B_{i}}(J) if J\neq\emptyset . Suppose that q^{\prime}_{i} has a non-neighbor in B_{i+1}\setminus N(K) . Let a be a minimal vertex over all non-neighbors of q^{\prime}_{i} in B_{i+1}\setminus N(K) . By Lemma 8.1 (2), a is minimal over B_{i+1} and a is anticomplete to J . It follows that N_{B_{i}}(a)=\{q_{i}\} and N(a)\setminus B_{i} is a clique. So d(a)\leq\omega , a contradiction. Therefore, q^{\prime}_{i} is complete to B_{i+1}\setminus N(K) and so q^{\prime}_{i} complete to B_{i+1} by Lemma 8.1 (2). Let z\in J be universal in J with maximum neighborhood in H (such a vertex exists by Lemma 8.6 ) if J\neq\emptyset . By Lemma 8.14 , |N_{B_{i+1}}(K)|,|N_{B_{i+1}}(J)|>\left\lceil\frac{\omega}{4}\right\rceil . By Lemma 4.13 , N_{B_{i}}(K)\subseteq N_{B_{i}}(J) or N_{B_{i}}(J)\subseteq N_{B_{i}}(K) . Let w_{0}\in L_{1}\cup L_{2} be a vertex with minimal neighborhood on H and z_{0}\in J be a vertex with minimal neighborhood in H . Let a\in N_{B_{i-1}}(w_{0}) with maximum neighborhood in H and let a^{\prime}\in N_{B_{i-1}}(z_{0}) with maximum neighborhood in H . By Lemma 8.1 (4), a and a^{\prime} are complete to B_{i-2}\cup B_{i} . If J\neq\emptyset , then K and J are cliques by Lemma 13.1 and so w_{0} and z_{0} are minimum and a is complete to K and a^{\prime} is complete to J . Next we deduce some structural information about J and its neighbors on H when J\neq\emptyset . We partition B_{i-1}\setminus N(K) into two subsets X and Y where every vertex in X has a neighbor in B_{i}\setminus N(K) and Y is anticomplete to B_{i}\setminus N(K) . By Lemma 8.7 , X is complete to B_{i-2}\cup B_{i} . Let x\in B_{i}\setminus(N(K)\cup N(J)) be a minimal vertex in B_{i}\setminus(N(K)\cup N(J)) . Case 1: J\neq\emptyset and B_{i}\setminus N(J) is anticomplete to N_{B_{i+1}}(J) Then N_{B_{i+1}}(x) is disjoint from both N_{B_{i+1}}(J) and N_{B_{i+1}}(K) . By Lemma 8.15 , |N_{B_{i+1}}(x)|>\left\lceil\frac{\omega}{4}\right\rceil . Assume that N_{B_{i+1}}(J) and N_{B_{i+1}}(K) are disjoint. By Lemma 8.10 , N_{B_{i+1}}(K)\cup N_{B_{i+1}}(J)\cup N_{B_{i+1}}(x) is complete to B_{i+2} and so N_{B_{i+1}}(K)\cup N_{B_{i+1}}(J)\cup N_{B_{i+1}}(x)\cup B_{i+2} is a clique of size larger than \omega . Hence, we may assume that N_{B_{i+1}}(J) and N_{B_{i+1}}(K) are comparable. By symmetry, we may assume that N_{B_{i+1}}(J)\subseteq N_{B_{i+1}}(K) . By Lemma 8.1 (2), N_{B_{i}}(K)=N_{B_{i}}(J) . Case 2: J\neq\emptyset and B_{i}\setminus N(J) is anticomplete to N_{B_{i-1}}(J) . By symmetry, we may assume that N_{B_{i+1}}(K) and N_{B_{i+1}}(J) are disjoint. So, by Lemma 4.12 , N_{B_{i-1}}(K) and N_{B_{i-1}}(J) are comparable. By Lemma 8.1 (4), there exists a vertex in N_{B_{i-1}}(K) complete to B_{i-2}\cup B_{i} . By assumption of J , N_{B_{i-1}}(J)\subsetneq N_{B_{i-1}}(K) . By Lemma 8.1 (2), N_{B_{i}}(K)\subseteq N_{B_{i}}(J) . Since N_{B_{i-1}}(x) and N_{B_{i-1}}(J) are disjoint, N_{B_{i+1}}(x) and N_{B_{i+1}}(J) are comparable and so q_{i+1}\in N_{B_{i+1}}(J) . Let b\in B_{i}\setminus N(K) be a maximal vertex in B_{i}\setminus N(K) . If N_{B_{i}}(K)=N_{B_{i}}(J) , then b\in B_{i}\setminus N(J) . If N_{B_{i}}(K)\neq N_{B_{i}}(J) , then b\in N_{B_{i}}(J)\setminus N_{B_{i}}(K) because N_{B_{i}}(J)\setminus N_{B_{i}}(K) is complete to N_{B_{i-1}}(J) . We show that X=\emptyset . If not, let x^{\prime} be a vertex in B_{i}\setminus N(J) with maximum neighborhood in H . Since N_{B_{i+1}}(x^{\prime}) and N_{B_{i+1}}(K) are disjoint, N_{B_{i-1}}(x^{\prime}) and N_{B_{i-1}}(K) are comparable. Since X\subseteq N_{B_{i-1}}(x^{\prime}) , N_{B_{i-1}}(K)\subseteq N_{B_{i-1}}(x^{\prime}) and so N_{B_{i-1}}(J)\subseteq N_{B_{i-1}}(x^{\prime}) . This contradicts the assumption of this case. Next, we will use the structure information in Cases 1 and 2 to define good subgraphs. Since |B_{i-2}|,|B_{i+2}|\geq 3 , q^{\prime}_{j} and q^{\prime\prime}_{j} exist for j\in\{i-2,i+2\} . We set F=(F_{1},F_{2},F_{3},F_{4},F_{5},O) as follows: • F_{5}=\{q_{i+2},q_{i+2}^{\prime}\} , • F_{4}=\{q_{i+1},c\} where c is a vertex in N_{B_{i+1}}(z_{0}) if Case 1 and c\in N_{B_{i+1}}(w_{0}) is a vertex with maximum neighborhood in H if Case 2 or J=\emptyset , • F_{3}=\{q_{i},q_{i}^{\prime},b\} where b\in B_{i}\setminus N(K) is a maximal vertex in B_{i}\setminus N(K) , • F_{2} is defined as follows. Let d\in Y be a vertex with maximum neighborhood in B_{i-2}\cup B_{i} if Y\neq\emptyset . – If J=\emptyset or Case 2 , then F_{2}=\begin{cases}\{a\},&Y=\emptyset\\ \{a,d\},&Y\neq\emptyset\end{cases} – If Case 1 , then F_{2}=\{a,a^{\prime}\} . • F_{1} is defined as follows. Let e\in B_{i-2}\setminus N(d) with maximal neighborhood in B_{i-2}\setminus N(d) if Y\neq\emptyset . – If J=\emptyset or Case 2 , then F_{1}=\begin{cases}\{q_{i-2},q_{i-2}^{\prime},q_{i-2}^{\prime\prime}\},&Y=\emptyset\\ \{q_{i-2},q_{i-2}^{\prime},e\},&Y\neq\emptyset\end{cases} – If Case 1 , then F_{1}=\{q_{i-2},q_{i-2}^{\prime}\} . • – O=\{w,v_{k}\} if J=\emptyset , where v_{k} is a vertex in L_{3} with maximal neighborhood in H . Note that v_{k} may not exist. – O=\{w,z\} if Case 1 or Case 2 with N_{B_{i}}(K)\neq N_{B_{i}}(J) – O=\{w,z,z^{\prime}\} if Case 2 with N_{B_{i}}(K)=N_{B_{i}}(J) where z^{\prime}\in J such that z^{\prime} is second maximal in H . We now show that F is 5-colorable. We first consider Case 1 . In this case, we can assign color 1 to q_{i-2},q_{i} , color 2 to a^{\prime},q_{i+1},w , color 3 to q_{i-2}^{\prime} , b , c , color 4 to a , q_{i+2} , z , and color 5 to q_{i}^{\prime} and q_{i+2}^{\prime} . In Case 2 , we assign color 1 to q_{i-2},q_{i} , color 2 to a^{\prime} , q_{i+1} , w , q_{i-2}^{\prime\prime} , d,e and color 3 to z^{\prime} , q_{i-2}^{\prime} , b,c , color 4 to a , q_{i+2},z , and color 5 to q_{i}^{\prime} and q_{i+2}^{\prime} . If J=\emptyset , we assign color 1 to q_{i-2},q_{i} , color 2 to q_{i-2}^{\prime},q_{i+1},w , color 3 to q_{i-2}^{\prime\prime},b,c,d,e , color 4 to a,q_{i+2} , and color 5 to q_{i}^{\prime},q_{i+2}^{\prime} . If v_{k} does not exist, the coloring is proper. So assume that v_{k} exists. By the definition of L_{1} and L_{3} , w_{0}\in L_{1} and so either a or c is not adjacent to v_{k} . We assign to v_{k} the color of a non-neighbor of v_{k} in \{a,c\} . Therefore, F is 5-colorable. Next, we show that every maximal clique of size \omega-j with j\in\{0,1,2,3\} of G contains at least 4-j vertices from F . Let M be a maximal clique of G . We consider the following cases depending on all possible locations of M . (i) M\subseteq K\cup N_{B_{i}}(K)\cup N_{B_{j}}(K) for every j\in\{i-1,i+1\} Since \{w\}\cup N_{B_{i}}(w) dominates K\cup N_{H}(K) , every maximal clique in K\cup N_{B_{i}}(K)\cup N_{B_{j}}(K) contains \{w,q_{i},q_{i}^{\prime}\} . We first consider Case 2 or J=\emptyset . If J\neq\emptyset , every maximum clique in K\cup N_{B_{i}}(K)\cup N_{B_{j}}(K) contains \{a,c\}\cap B_{j} by choice of a and c . So we may assume that J=\emptyset . We now consider Case 1 . Suppose that c has a neighbor w_{1}\in K and a non-neighbor w_{2}\in K . Then w_{2}-w_{1}-c-q_{i+1}-(B_{i}\setminus N(K))-a^{\prime}-q_{i-2} is an induced P_{7} . So c is complete to K . Since a is complete to L_{1}\cup L_{2} by choice of a , every maximum clique in K\cup N_{B_{i-1}}(K)\cup N_{B_{i}}(K) contains exactly one of \{a,v_{k}\} . This completes the proof of this case. (ii) M\subseteq J\cup N_{B_{i}}(J)\cup N_{B_{j}}(J) for every j\in\{i-1,i+1\} In this case, J\neq\emptyset . We first consider the case that N_{B_{i}}(K)\neq N_{B_{i}}(J) . Then, we have N_{B_{i}}(K)\subsetneq N_{B_{i}}(J) and since N_{B_{i}}(J)\setminus N_{B_{i}}(K) is complete to N_{B_{i-1}}(J) , we have b\in N_{B_{i}}(J)\setminus N_{B_{i}}(K) . So, since \{z\}\cup N_{B_{i}}(z) dominates J\cup N_{H}(J) , every maximal clique in J\cup N_{B_{i}}(J)\cup N_{B_{j}}(J) contains \{z,q_{i},q_{i}^{\prime},b\} . We now consider the case that N_{B_{i}}(K)=N_{B_{i}}(J) . Note that q_{i},q_{i}^{\prime},z\in M . In Case 1 , every maximum clique in J\cup N_{B_{i}}(J)\cup N_{B_{j}}(J) contains \{a^{\prime},c\}\cap B_{j} by choice of a^{\prime} and c . In Case 2 , suppose that z^{\prime}\notin M . Then we have |M|=\omega and |M\setminus J|=\omega-1 . Hence, either (M\setminus J)\cup N_{B_{i+1}}(K) or (M\setminus J)\cup\{a,w\} is a clique of size larger than \omega and so z^{\prime}\in M . This completes the proof of this case. (iii) B_{i-2}\cup B_{i-1} Since q_{i-2} and a are universal in B_{i-2}\cup B_{i-1} , q_{i-2},a\in M . So we may assume that |M|\geq\omega-1 . We first consider Case 1 . Since a^{\prime} is universal in B_{i-2}\cup B_{i-1} , a^{\prime}\in M . Next, we show that q_{i-2}^{\prime}\in M , which will complete the proof for this case. Suppose for a contradiction that q_{i-2}^{\prime}\notin M . Suppose first that B_{i-1}\setminus(N(K)\cup N(J))\neq\emptyset . Let v\in B_{i-1}\setminus(N(K)\cup N(J)) be a vertex with minimal neighborhood in B_{i-2}\cup B_{i} . By Lemma 8.1 (4), v is a vertex in B_{i-1} with minimal neighborhood in B_{i-2}\cup B_{i} . It follows that N(v)\setminus B_{i-2} is a clique. Thus, q_{i-2}^{\prime} is complete to B_{i-1} and so q_{i-2}^{\prime}\in M , a contradiction. So B_{i-1}\setminus(N(K)\cup N(J))=\emptyset . By Lemma 8.1 (2), q_{i},q_{i}^{\prime} are complete to B_{i-1} . Then |M\cap B_{i-2}|=1 and |M\cap B_{i-1}|=\omega-1 . So (M\setminus\{q_{i-2}\})\cup\{q_{i},q_{i}^{\prime}\} is a clique of size larger than \omega . Next we consider Case 2 or J=\emptyset . We first consider the case Y=\emptyset . Define q to be q_{i-2}^{\prime\prime} if |M|=\omega and to be q_{i-2}^{\prime} if |M|=\omega-1 . We show that q\in M , which completes the proof. Suppose for a contradiction that q\notin M . Then there exists a vertex u\in M\cap B_{i-1} such that uq\notin E(G) . By Lemma 8.7 , u\in N_{B_{i-1}}(K) . If u has non-neighbor v in B_{i}\setminus N(K) , then q-q_{i-2}-u-w-N_{B_{i+1}}(K)-q_{i+1}-v is an induced P_{7} . So u is complete to B_{i}\setminus N(K) . For any x\in M\cap N_{B_{i-1}}(K) with xq\in E(G) , we have N_{B_{i-2}\cup B_{i}}(u)\subseteq N_{B_{i-2}\cup B_{i}}(x) and so M\cap N_{B_{i-1}}(K) is complete to B_{i} . By Lemma 8.7 and Y=\emptyset , M\cap(B_{i-1}\setminus N(K)) is complete to B_{i} . This implies that (M\cap B_{i-1})\cup\{q_{i},q^{\prime}_{i},b\} is a clique of size \omega+1 . We now consider the case Y\neq\emptyset . Let y\in Y be minimal in Y . Since Y is anticomplete to B_{i}\setminus N(K) , y is also minimal in B_{i-1} . So N(y)\setminus B_{i-2} is a clique and thus |N_{B_{i-2}}(y)|>\left\lceil\frac{1}{4}\omega\right\rceil . In particular, |N_{B_{i-2}}(y)|\geq 2 and so q^{\prime}_{i-2}\in N_{B_{i-2}}(y) . Therefore, q_{i-2},q_{i-2}^{\prime},a\in M . If M\cap Y\neq\emptyset , then d\in M , since d is maximal in Y . Assume that M\cap Y=\emptyset . If M\cap B_{i-2}\subseteq N(d) , then M\cup\{d\} is a clique, a contradiction. So (M\cap B_{i-2})\setminus N(d)\neq\emptyset . By the choice of e , we have e\in M . This completes the proof of this case. (iv) B_{i-1}\cup B_{i} We first consider Case 1 . Since a,a^{\prime},q_{i} are universal in B_{i-1}\cup B_{i} , a,a^{\prime},q_{i}\in M . Suppose first that B_{i-1}\setminus(N(K)\cup N(J))\neq\emptyset . Let v\in B_{i-1}\setminus(N(K)\cup N(J)) be a vertex with minimal neighborhood in B_{i-2}\cup B_{i} . By Lemma 8.1 (4), v is a vertex in B_{i-1} with minimal neighborhood in B_{i-2}\cup B_{i} . It follows that N(v)\setminus B_{i} is a clique. Thus, q_{i}^{\prime} is complete to B_{i-1} and so q_{i}^{\prime}\in M . So B_{i-1}\setminus(N(K)\cup N(J))=\emptyset . By Lemma 8.1 (2), q_{i}^{\prime} is complete to N_{B_{i-1}}(K)\cup N_{B_{i-1}}(J)=B_{i-1} and so q_{i}^{\prime}\in M . We then consider Case 2 or J=\emptyset . We first consider the case Y=\emptyset . Assume that X\neq\emptyset . By Lemma 8.1 (4) and Lemma 8.7 , M=B_{i-1}\cup B_{i} contains \{a,q_{i},q_{i}^{\prime},b\} . Next we assume that X=\emptyset . Note that a,q_{i},q_{i}^{\prime} are universal in B_{i-1}\cup B_{i} and so a,q_{i},q_{i}^{\prime}\in M . If M\cap(B_{i}\setminus N(K))=\emptyset , then M\cup\{w\} is a clique, a contradiction. So M\cap(B_{i}\setminus N(K))\neq\emptyset and this implies that b\in M . We now consider the case Y\neq\emptyset . Let y\in Y be minimal in Y . By Y is anticomplete to B_{i}\setminus N(K) and P_{4} -structure, y is also minimal in B_{i-1} . So N(y)\setminus B_{i} is a clique and so |N(y)\cap N_{B_{i}}(K)|>\left\lceil\frac{1}{4}\omega\right\rceil . So q_{i},q_{i}^{\prime} are universal in B_{i-1}\cup B_{i} . By the choice of a , we have a,q_{i},q^{\prime}_{i}\in M . We assume that X\neq\emptyset . Recall that N_{B_{i-1}}(K) is complete to B_{i}\setminus N(K) and so X\cup N_{B_{i-1}}(K)\cup B_{i} is a clique. If M\cap Y=\emptyset , then M contains \{a,q_{i},q_{i}^{\prime},b\} . If M\cap Y\neq\emptyset , then d\in M by the choice of d . In this case, every maximal clique in B_{i-1}\cup B_{i} contains one of \{a,q_{i},q_{i}^{\prime},b\} and \{a,q_{i},q_{i}^{\prime},d\} . So X=\emptyset . Since \{w\}\cup N_{B_{i-1}}(w)\cup N_{B_{i}}(w) is a clique, M\cap Y\neq\emptyset or M\cap(B_{i}\setminus N(K))\neq\emptyset . If M\cap(B_{i}\setminus N(K))\neq\emptyset , then b\in M by the choice of b . If M\cap Y\neq\emptyset , d\in M by the choice of d . (v) M\subseteq B_{i}\cup B_{i+1} Note that q_{i} and q_{i+1} are universal in B_{i}\cup B_{i+1} and so q_{i},q_{i+1}\in M . Since B_{i}\setminus N(K) is anticomplete to N_{B_{i+1}}(K) , it follows that M contains no vertex from B_{i}\setminus N(K) or N_{B_{i+1}}(K) . Since q^{\prime}_{i} is complete to B_{i+1} , q^{\prime}_{i}\in M . If M contains no vertices from N_{B_{i+1}}(K) , then M contains a vertex in B_{i}\setminus N(K) (for otherwise \{c\}\cup M is a clique, which contradicts the maximality of M ). By the choice of b , b\in M . If M contains no vertices from B_{i}\setminus N(K) , then c\in M since c is universal in (B_{i}\cup B_{i+1})\setminus(B_{i}\setminus N(K)) . So |M\cap F|=4 . (vi) M\subseteq B_{i+1}\cup B_{i+2} Since q_{i+1},q_{i+2},c are universal in B_{i+1}\cup B_{i+2} , q_{i+1},q_{i+2},c\in M . Suppose that |M|=\omega and q^{\prime}_{i+2}\notin M . It follows that M\cap B_{i+2}=\{q_{i+2}\} and M=B_{i+1}\cup\{q_{i+2}\} . So |B_{i+1}|=\omega-1 . Since B_{i+1} is complete to \{q_{i},q_{i}^{\prime}\} , |B_{i+1}\cup\{q_{i},q^{\prime}_{i}\}|\geq\omega+1 , a contradiction. (vii) M\subseteq B_{i-2}\cup B_{i+2} By Lemma 13.4 , M contains \{q_{i-2},q_{i-2}^{\prime},q_{i+2},q_{i+2}^{\prime}\} . Therefore, F is a (4,5) -good subgraph. ∎ Finally, we deal with the case that A_{3}(i) has two components and one of the components has only one neighbor in B_{i} . Theorem 13.11 (Two components with one having one neighbor in B_{i} ) . Let K and J be two non-empty clique components of A_{3}(i) . If |N_{B_{i}}(K)|=1 , then \chi(G)\leq\left\lceil\frac{5}{4}\omega\right\rceil . Proof. Note that N_{B_{i}}(K)=\{q_{i}\} . By Lemma 4.10 , we may assume that B_{i}\setminus N(K) is anticomplete to N_{B_{i+1}}(K) . By Lemma 13.4 , we have |B_{i-2}|,|B_{i+2}|>\left\lceil\frac{\omega}{4}\right\rceil . Let w_{K} and w_{J} be a maximum vertex in K and J , respectively. Let w_{K}^{\prime} and w_{J}^{\prime} be a minimum vertex in K and J , respectively. It follows that |N_{B_{j}}(w_{K}^{\prime})|>\left\lceil\frac{\omega}{4}\right\rceil and |N_{B_{j}}(w_{J}^{\prime})|>\left\lceil\frac{\omega}{4}\right\rceil for j\in\{i-1,i+1\} . Let z\in B_{i}\setminus\{q_{i}\} be a minimal vertex in B_{i}\setminus\{q_{i}\} . Since N_{B_{i}}(J) is complete to N_{B_{i\pm 1}}(J) but B_{i}\setminus N(J) is not, it follows that z\in B_{i}\setminus N(J) . By assumption, N_{B_{i+1}}(K) and N_{B_{i+1}}(z) are disjoint. As z is minimal in B_{i} , we have N_{B_{i-1}}(z)\cup B_{i} is a clique and so |N_{B_{i+1}}(z)|>\left\lceil\frac{\omega}{4}\right\rceil by Lemma 8.15 . By Lemma 8.10 , N_{B_{i+1}}(z) and N_{B_{i+1}}(K) are complete to B_{i+2} . This implies that \omega\geq 6 . If N_{B_{i+1}}(J) and N_{B_{i+1}}(K)\cup N_{B_{i+1}}(z) are disjoint, then N_{B_{i+1}}(K)\cup N_{B_{i+1}}(J)\cup N_{B_{i+1}}(z)\cup B_{i+2} is a clique of size larger than \omega by 8.10 . Therefore, N_{B_{i+1}}(J) is comparable with N_{B_{i+1}}(K) or N_{B_{i+1}}(z) . Case 1: N_{B_{i+1}}(K)\cup N_{B_{i+1}}(z)\subseteq N_{B_{i+1}}(J) . Then N_{B_{i-1}}(J) and N_{B_{i-1}}(K)\cup N_{B_{i-1}}(z) are disjoint. Since z\in B_{i}\setminus N(J) , B_{i}\setminus N(J) is anticomplete to N_{B_{i-1}}(J) . If |N_{B_{i}}(J)|\geq 2 , then there is a vertex x\in N_{B_{i}}(J)\setminus N_{B_{i}}(K) that is complete to N_{B_{i+1}}(K) , a contradiction. So N_{B_{i}}(J)=N_{B_{i}}(K)=\{q_{i}\} . Suppose that B_{i+1}\setminus N(J)\neq\emptyset . By Lemma 4.12 , B_{i}\setminus\{q_{i}\} is anticomplete to B_{i+1}\setminus N(J) . Let x\in B_{i+1}\setminus N(J) be a minimal vertex in B_{i+1}\setminus N(J) . If B_{i+1}\cup N_{B_{i+2}}(x) contains two non-adjacent vertices u and v , then u\in N_{B_{i+1}}(J)\setminus(N(K)\cup N(z)) and v\in B_{i+2} , then w_{J}-u-x-v contradicts Lemma 4.11 . Since N_{B_{i}}(x)=\{q_{i}\} , d(x)\leq\omega , a contradiction. This show that N_{B_{i+1}}(J)=B_{i+1} . By Lemma 8.10 , N_{B_{i-1}}(K) and N_{B_{i-1}}(J) are complete to B_{i-2} . Let X_{i-1}=B_{i-1}\setminus(N(K)\cup N(J)) . By Lemma 13.6 , every vertex in X_{i-1} has a neighbor in B_{i}\setminus\{q_{i}\} . By Lemma 8.7 , X_{i-1} is complete to B_{i-2} and it follows that B_{i-1} is complete to B_{i-2} . Recall that N_{B_{i+1}}(K) and N_{B_{i+1}}(z) are complete to B_{i+2} . If y\in B_{i+1}=N_{B_{i+1}}(J) has a non-neighbor y^{\prime}\in B_{i+2} , then y^{\prime}-q_{i+2}-y-J-N_{B_{i-1}}(J)-N_{B_{i-1}}(K)-K is an induced sequence. So B_{i+1} is complete to B_{i+2} . By Lemma 13.5 , B_{i+2} is complete to B_{i-2} . Since |B_{i-2}|,|B_{i+2}|<\frac{\omega}{2} , B_{i-2}\cup B_{i+2} is not a maximum clique. By Lemma 8.7 , X_{i-1} is complete to B_{i} . We now show that K is complete to N_{B_{i-1}}(K) . If x\in N_{B_{i-1}}(K) has a non-neighbor x^{\prime}\in K , then x^{\prime}-w_{K}-x-q_{i-2}-q_{i+2}-N_{B_{i+1}}(z)-w_{J} is an induced P_{7} . So K is complete to N_{B_{i-1}}(K) and K\cup N_{B_{i-1}}(K)\cup\{q_{i}\} is a clique. Suppose that J\cup N_{B_{i-1}}(J)\cup N_{B_{i}}(J) contains a maximum clique. It follows that |N_{B_{i-1}}(J)|\geq\frac{\omega-1}{2} or |J|\geq\frac{\omega-1}{2} . If |N_{B_{i-1}}(J)|\geq\frac{\omega-1}{2} , then B_{i-2}\cup B_{i-1} is a clique of size larger than \omega . So |J|\geq\frac{\omega-1}{2} . As N_{B_{i+1}}(K)\cup N_{B_{i+1}}(z)\subseteq N_{B_{i+1}}(J) , J is complete to N_{B_{i+1}}(K) and N_{B_{i+1}}(z) . Since |N_{B_{i+1}}(K)|,|N_{B_{i+1}}(z)|>\left\lceil\frac{\omega}{4}\right\rceil , we obtain N_{B_{i+1}}(K)\cup N_{B_{i+1}}(z)\cup J is a clique of size larger than \omega . This proves that J\cup N_{B_{i-1}}(J)\cup N_{B_{i}}(J) does not contain a maximum clique. If X_{i-1}\neq\emptyset , then \{x_{i-1},w_{K},q_{i+1}\} is a good stable set where x_{i-1}\in X_{i-1} (rely on J is complete to N_{B_{i+1}}(K) and N_{B_{i+1}}(z) ). So X_{i-1}=\emptyset . Note that B_{i-1}\cup\{q_{i}\} and B_{i+1}\cup\{q_{i}\} are not maximum clique. Let q^{\prime}_{i-1}\in B_{i-1} be a common neighbor of w_{K}^{\prime} and z . It follows that q^{\prime}_{i-1} is complete to B_{i} and K . Let a\in N_{B_{i-1}}(w_{J}^{\prime}) . Then a is complete to J . Let b\in N_{B_{i+1}}(w_{K}^{\prime}) . Then b is complete to K . Let v_{K}\in K be the second largest vertex in K if |K|\geq 2 . Let q^{\prime}_{j}\in B_{j} be the second largest vertex in B_{j} for j\in\{i-2,i+2\} . Let q^{\prime}_{i},q^{\prime\prime}_{i} be the second and third largest vertices in B_{i} , respectively. We set F=(F_{1},F_{2},F_{3},F_{4},F_{5},O) as follows: • F_{1}=\{q_{i-2},q_{i-2}^{\prime}\} , • F_{2}=\{q^{\prime}_{i-1},a\} , • F_{3}=\{q_{i},q_{i}^{\prime},q_{i}^{\prime\prime}\} , • F_{4}=\{b,q_{i+1}\} , • F_{5}=\{q_{i+2},q_{i+2}^{\prime}\} , • O=\{w_{K},v_{K},w_{J}\} . We can assign color 1 to q_{i-2} and q_{i} , color 2 to w_{K} , w_{J} , q_{i}^{\prime} and q_{i+2} , color 3 to a , q_{i}^{\prime\prime} and b , color 4 to v_{K} , q_{i-2}^{\prime} and q_{i+1} , and color 5 to q_{i-1}^{\prime} and q_{i+2}^{\prime} . So F is 5-colorable. Next, we show that every maximal clique of size \omega-j with j\in\{0,1,2,3\} of G contains at least 4-j vertices from F . Let M be a maximal clique of G . We consider the following cases depending on all possible locations of M . If M is one of B_{i-2}\cup B_{i-1} , B_{i+1}\cup B_{i+2} and B_{i-2}\cup B_{i+2} , M contains four vertices of F . (i) M\subseteq K\cup N_{H}(K) We first assume that M is contained in K\cup N_{B_{i-1}}(K)\cup\{q_{i}\} . As K\cup N_{B_{i-1}}(K)\cup\{q_{i}\} is a clique, M contains \{w_{K},q_{i-1}^{\prime},q_{i}\} . If v_{K}\notin M , then |N_{B_{i-1}}(K)|\geq\omega-2 and so B_{i-1} is a clique of size larger than \omega as |N_{B_{i-1}}(J)|\geq 3 . So we may assume that M\subseteq K\cup N_{B_{i+1}}(K)\cup\{q_{i}\} . Note that q_{i},w_{K},b are universal in K\cup N_{B_{i+1}}(K) . So we can assume that |M|=\omega . If v_{K}\notin M , then |N_{B_{i+1}}(K)|\geq\omega-2 and so B_{i+1}\cup B_{i+2} is a clique of size larger than \omega as |N_{B_{i+1}}(z)|\geq 3 and |B_{i+2}|\geq 3 . Therefore, M contains \{w_{K},v_{K},q_{i},b\} . (ii) M\subseteq J\cup N_{H}(J) Suppose first that M\cap B_{i+1}=\emptyset . Since w_{J},q_{i},a are universal in J\cup N_{B_{i-1}}(J)\cup\{q_{i}\} , w_{J},q_{i},a\in M . Since J\cup N_{B_{i-1}}(J)\cup\{q_{i}\} does not contain a maximum clique of G , we are done in this case. So M\subseteq J\cup N_{B_{i+1}}(J)\cup\{q_{i}\} . As N_{B_{i+1}}(K)\cup N_{B_{i+1}}(z) is complete to J , M contains \{w_{J},b,q_{i},q_{i+1}\} . (iii) M\subseteq B_{i-1}\cup B_{i} ( M\subseteq B_{i}\cup B_{i+1} is symmetric) Since B_{i-1}\cup\{q_{i}\} is not a maximum clique, M\cap(B_{i}\setminus\{q_{i}\})\neq\emptyset and so q_{i}^{\prime}\in M . Then, M\cap N_{B_{i-1}}(J)=\emptyset . Since q_{i},q^{\prime}_{i-1} are universal in B_{i-1}\cup B_{i} , q^{\prime}_{i-1},q_{i}\in M . So we can assume that |M|=\omega . If q^{\prime\prime}_{i} is not in M , then |M\cap(B_{i-1}\setminus N(J))|\geq\omega-2 . Since |B_{i-2}|\geq 3 , B_{i-1}\cup B_{i-2} is a clique of size larger than \omega . So M contains \{q_{i-1},q_{i},q^{\prime}_{i},q^{\prime\prime}_{i}\} . Therefore, F is a (4,5) -good subgraph. Hence, we may assume that N_{B_{i+1}}(J) is comparable with exactly one of N_{B_{i+1}}(K) and N_{B_{i+1}}(z) , and is disjoint from the other. Case 2: N_{B_{i+1}}(J) and N_{B_{i+1}}(K) are comparable. Then N_{B_{i+1}}(J) and N_{B_{i+1}}(z) are disjoint. If |N_{B_{i}}(J)|\geq 2 , then there is a vertex x\in N_{B_{i}}(J)\setminus N_{B_{i}}(K) that is complete to N_{B_{i+1}}(J)\cap N_{B_{i+1}}(K) , a contradiction. So we obtain |N_{B_{i}}(J)|=1 and this implies that N_{B_{i}}(J)=N_{B_{i}}(K)=\{q_{i}\} . If B_{i}\setminus N(J) is anticomplete to N_{B_{i-1}}(J) , then z-q_{i-1}-N_{B_{i-1}}(J)-J-(N_{B_{i+1}}(K)\cap N_{B_{i+1}}(J))-N_{B_{i+1}}(z)-z is an induced cyclic sequence. So B_{i}\setminus N(J) is anticomplete to N_{B_{i+1}}(J) . By Lemmas 13.6 , 13.7 and 13.8 , \chi(G)\leq\left\lceil\frac{5}{4}\omega\right\rceil . Case 3: N_{B_{i+1}}(J) and N_{B_{i+1}}(z) are comparable. Then N_{B_{i+1}}(J) and N_{B_{i+1}}(K) are disjoint and it follows that N_{B_{i-1}}(J)\cup N_{B_{i-1}}(z)\subseteq N_{B_{i-1}}(K) . So |N_{B_{i-1}}(K)|>2\cdot\left\lceil\frac{\omega}{4}\right\rceil . Since z\in B_{i}\setminus N(J) , we have B_{i}\setminus N(J) is anticomplete to N_{B_{i-1}}(J) . If |N_{B_{i}}(J)|=1 , then we argue symmetrically to Case 1 . So we may assume that |N_{B_{i}}(J)|\geq 2 . Let Y\subseteq B_{i+1}\setminus N(K) be the set of vertices that are anticomplete to B_{i}\setminus\{q_{i}\} . We show that Y is complete to B_{i+2} . Suppose not. Choose a vertex y\in Y that has a non-neighbor in B_{i+2} with N_{B_{i}\cup B_{i+2}}(y) is minimum. By Lemma 8.10 , N_{B_{i+1}}(K) and N_{B_{i+1}}(J) are complete to B_{i+2} and so y is anticomplete to K\cup J . By Lemmas 8.1 (2) and 8.10 , y is minimum in B_{i+1} . It follows that d(y)\leq\omega , a contradiction. So Y is complete to B_{i+2} . By Lemma 8.10 , B_{i+1} is complete to B_{i+2} . We then show that N_{B_{i-1}}(K) is complete to B_{i-2} . By Lemma 8.10 , N_{B_{i-1}}(J) and N_{B_{i-1}}(z) is complete to B_{i-2} . If x\in N_{B_{i-1}}(K) has a non-neighbor x^{\prime}\in B_{i-2} , then x^{\prime}-q_{i-2}-x-K-N_{B_{i+1}}(K)-N_{B_{i+1}}(J)-J is an induced sequence. So N_{B_{i-1}}(K) is complete to B_{i-2} . We now show that K is complete to N_{B_{i-1}}(K) . As N_{B_{i-1}}(J)\cup N_{B_{i-1}}(z)\subseteq N_{B_{i-1}}(K) , N_{B_{i-1}}(J)\cup N_{B_{i-1}}(z) is complete to K . If x\in N_{B_{i-1}}(K) has a non-neighbor x^{\prime}\in K , then x^{\prime}-w_{K}-x-q_{i-2}-q_{i+2}-N_{B_{i+1}}(J)-J is an induced sequence. So K is complete to N_{B_{i-1}}(K) . We then show that J is complete to N_{B_{i+1}}(J) . If y\in N_{B_{i+1}}(J) has a non-neighbor y^{\prime}\in J , then y^{\prime}-w_{J}-y-N_{B_{i+1}}(K)-K-N_{B_{i-1}}(z)-q_{i-2} is an induced sequence. So J is complete to N_{B_{i+1}}(J) . We now show that K\cup N_{B_{i+1}}(K)\cup\{q_{i}\} does not contain a maximum clique. Let M be a maximum clique in K\cup N_{B_{i+1}}(K)\cup\{q_{i}\} . Since K\cup N_{B_{i-1}}(K)\cup\{q_{i}\} is a clique and |N_{B_{i-1}}(K)|>2\cdot\left\lceil\frac{\omega}{4}\right\rceil , we have |K|<\omega-2\cdot\left\lceil\frac{\omega}{4}\right\rceil-1 . Hence, we obtain |M\cap N_{B_{i+1}}(K)|>2\cdot\left\lceil\frac{1}{4}\omega\right\rceil . Then, N_{B_{i+1}}(K)\cup N_{B_{i+1}}(J)\cup B_{i+2} is a clique of size larger than \omega and the claim holds. By the definition of z and Lemma 8.10 , there is a vertex in N_{B_{i-1}}(z) complete to B_{i-2}\cup B_{i} . We set this vertex to be q_{i-1} . Then q_{i-1}\in N_{B_{i-1}}(K) . Let a\in N_{B_{i-1}}(w_{J}^{\prime}) . Then a is complete to J . Let q_{i}^{\prime}\in N_{B_{i}}(J)\setminus\{q_{i}\} be maximal over N_{B_{i}}(J)\setminus\{q_{i}\} . Let b\in B_{i}\setminus N(J) be a maximal vertex over B_{i}\setminus N(J) . Let c\in N_{B_{i+1}}(w_{K}^{\prime}) . Then c is complete to K . Let q_{j}^{\prime}\in B_{j} be the second largest vertex in B_{j} for j\in\{i-2,i+2\} . By Lemma 13.4 , we have q^{\prime}_{i+2} is complete to B_{i-2} and q^{\prime}_{i-2} is complete to B_{i+2} . If q^{\prime}_{i-2} has a non-neighbor in B_{i-1}\setminus N(K) , then a minimal vertex x in B_{i-1}\setminus N(K) has N(x)\setminus B_{i-2} is a clique and N_{B_{i-2}}(x)=\{q_{i-2}\} . So d(x)\leq\omega , a contradiction. So q^{\prime}_{i-2} is complete to B_{i-1} . We set F=(F_{1},F_{2},F_{3},F_{4},F_{5},O) as follows: • F_{1}=\{q_{i-2},q_{i-2}^{\prime}\} , • F_{2}=\{q_{i-1},a\} , • F_{3}=\{q_{i},q_{i}^{\prime},b\} , • F_{4}=\{c,q_{i+1}\} . • F_{5}=\{q_{i+2},q_{i+2}^{\prime}\} , • O=\{w_{K},w_{J}\} . We can assign color 1 to q_{i} and q_{i+2} , color 2 to q_{i-2} , w_{K} and w_{J} , color 3 to q_{i-2}^{\prime} , q_{i}^{\prime} and c , color 4 to a , b and q_{i+2}^{\prime} , and color 5 to q_{i-1} and q_{i+1} . So F is 5 -colorable. Next, we show that every maximal clique of size \omega-j with j\in\{0,1,2,3\} of G contains at least 4-j vertices from F . Let M be a maximal clique of G . We consider the following cases depending on all possible locations of M . If M is contained in one of B_{i-2}\cup B_{i-1} , B_{i+1}\cup B_{i+2} and B_{i-2}\cup B_{i+2} , M contains four vertices of F . (i) M\subseteq K\cup N_{H}(K) Since M\subseteq K\cup N_{B_{i-1}}(K)\cup\{q_{i}\} is a clique, M contains \{w_{K},a,q_{i-1},q_{i}\} . Hence, we can assume that M\subseteq K\cup N_{B_{i+1}}(K)\cup\{q_{i}\} . Since K\cup N_{B_{i+1}}(K)\cup\{q_{i}\} does not contain a maximum clique and w_{K},q_{i},c are universal in K\cup N_{B_{i+1}}(K) , we obtain w_{K},q_{i},c\in M . (ii) M\subseteq J\cup N_{H}(J) Note that q_{i},q_{i}^{\prime},w_{J} are universal in J\cup N_{H}(J) . We first assume that M\subseteq J\cup N_{B_{i}}(J)\cup N_{B_{i+1}}(J) . Since N_{B_{i+1}}(J) is complete to J\cup N_{B_{i}}(J) , q_{i+1} is universal in J\cup N_{B_{i}}(J)\cup N_{B_{i+1}}(J) . So w_{J},q_{i},q_{i}^{\prime},q_{i+1}\in M . Now, we may assume that M\subseteq J\cup N_{B_{i-1}}(J)\cup N_{B_{i}}(J) . Since a is complete to J , we obtain w_{J},a,q_{i},q_{i}^{\prime}\in M . (iii) M\subseteq B_{i-1}\cup B_{i} Note that q_{i-1} and q_{i} are universal in B_{i-1}\cup B_{i} . If M=B_{i-1}\cup\{q_{i}\} , then M is not a maximum clique and q_{i-1},q_{i},a\in M . So we may assume that M\cap(B_{i}\setminus\{q_{i}\})\neq\emptyset . We first assume that M\cap(B_{i}\setminus N(J))=\emptyset and this implies that M\cap(N_{B_{i}}(J)\setminus\{q_{i}\})\neq\emptyset and so q_{i}^{\prime}\in M . Since a is universal in B_{i-1}\cup N_{B_{i}}(J) , we have a\in M . Hence, we may assume that M\cap(B_{i}\setminus N(J))\neq\emptyset and this implies that b\in M . Since q^{\prime}_{i} has a neighbor in N_{B_{i-1}}(J) , N_{B_{i-1}}(b)\subseteq N_{B_{i-1}}(q^{\prime}_{i}) and so q^{\prime}_{i}\in M . (iv) M\subseteq B_{i}\cup B_{i+1} Note that q_{i} and q_{i+1} are universal in B_{i}\cup B_{i+1} . If M=B_{i+1}\cup\{q_{i}\} , then M is not a maximum clique and q_{i},q_{i+1},c\in M . So we may assume that M\cap(B_{i}\setminus\{q_{i}\})\neq\emptyset . We first assume that M\cap(B_{i}\setminus N(J))=\emptyset and this implies that M\cap(N_{B_{i}}(J)\setminus\{q_{i}\})\neq\emptyset and q_{i}^{\prime}\in M . So M\cap N_{B_{i+1}}(K)=\emptyset and we can assume that |M|=\omega . As N_{B_{i}}(J)\cup N_{B_{i-1}}(J)\cup N_{B_{i-1}}(z) is a clique and |N_{B_{i-1}}(J)\cup N_{B_{i-1}}(z)|>2\cdot\left\lceil\frac{\omega}{4}\right\rceil , we obtain |M\cap(B_{i+1}\setminus N(K))|>2\cdot\left\lceil\frac{\omega}{4}\right\rceil . Then B_{i+1}\cup B_{i+2} is a clique of size larger than \omega . Hence, M is not a maximum clique and q_{i},q_{i}^{\prime},q_{i+1}\in M . Now we may assume that M\cap(B_{i}\setminus N(J))\neq\emptyset and so b\in M . Since N_{B_{i-1}}(b)\subseteq N_{B_{i-1}}(q^{\prime}_{i}) , we have q_{i}^{\prime}\in M . Hence, M contains \{q_{i},q_{i}^{\prime},q_{i+1},b\} . Therefore, F is a (4,5) -good subgraph. ∎ 14 Concluding Remarks In this paper, we prove that f(n)=\left\lceil\frac{5}{4}n\right\rceil is the optimal \chi -bounding function for the class of ( P_{7} , even-hole)-free graphs, which generalizes the result of Karthick and Maffary [ 10 ] on ( P_{6} , even-hole)-free graphs. We feel that f(n)=2n-1 may not be the optimal \chi -bounding function for even-hole-free graphs, as this comes merely from the existence of f -small vertices. Equal-size blowups of C_{5} suggest that f(n)=\left\lceil\frac{5}{4}n\right\rceil may be the optimal \chi -bounding function for the class of even-hole-free graphs and we conjecture this should be the case. Conjecture 14.1 . For every even-hole-free graph G , \chi(G)\leq\left\lceil\frac{5}{4}\omega(G)\right\rceil . Chen, Xu and Xu [ 2 ] shows that the conjecture is true for (cap, even-hole)-free graphs. Theorem 1.1 shows that the conjecture is true for ( P_{7} , even-hole)-free graphs. Recently, there has been a series of work on tree-width and tree-independence number of even-hole-free graphs (see [ 3 , 5 ] for instance). The techniques developed there may offer new insights and tools to attack this conjecture.