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

ON GENERALIZED LIST 𝒒\operatorname{\mathcal{G}}-FREE COLOURINGS OF GRAPHS

Yaser Rowshan1 1Y. Rowshan, Department of Mathematics, Institute for Advanced Studies in Basic Sciences (IASBS), Zanjan 45137-66731, Iran. [email protected],Β Β Β [email protected].
Abstract.

For given graph HH and graphical property PP, the conditional chromatic number χ​(H,P)\chi(H,P) of HH, is the smallest number kk, so that V​(H)V(H) can be decomposed into sets V1,V2,…,VkV_{1},V_{2},\ldots,V_{k}, in which H​[Vi]H[V_{i}] satisfies the property PP, for each 1≀i≀k1\leq i\leq k. When property PP be that each color class contains no copy of GG, we write Ο‡G​(H)\chi_{G}(H) instead of χ​(G,P)\chi(G,P), which is called the GG-free chromatic number. Due to this, we say HH has a kk-GG-free coloring if there is a map c:V​(H)⟢{1,…,k}c:V(H)\longrightarrow\{1,\ldots,k\}, so that each of the color classes of cc be GG-free. Assume that for each vertex vv of a graph HH is assigned a set L​(V)L(V) of colors, called a color list. Set g​(L)={g​(v):v∈V​(H)}g(L)=\{g(v):v\in V(H)\}, that is the set of colors chosen for the vertices of HH under gg. An LL-coloring gg is called GG-free, so that:

  • β€’

    g​(v)∈L​(v)g(v)\in L(v), for any v∈V​(H)v\in V(H).

  • β€’

    H​[Vi]H[V_{i}] is GG-free for each i=1,2,…,Li=1,2,\ldots,L.

If there exists an LL-coloring of HH, then HH is called LL-GG-free-colorable. A graph HH is said to be kk-GG-free-choosable if there exists an LL-coloring for any list-assignment LL satisfying |L​(V)|β‰₯k|L(V)|\geq k for each v∈V​(H)v\in V(H), and H​[Vi]H[V_{i}] be GG-free for each i=1,2,…,Li=1,2,\ldots,L. Let graph HH and a collection of graphs 𝒒\operatorname{\mathcal{G}} are given, the χ𝒒L​(H)\chi_{\operatorname{\mathcal{G}}}^{L}(H) of HH is the last integer kk, so that HH is kk-𝒒\operatorname{\mathcal{G}}-free-choosable i.e. H​[Vi]H[V_{i}] is 𝒒\operatorname{\mathcal{G}}-free for each i=1,2,…,ki=1,2,\ldots,k i.e. contains no copy of any member of 𝒒\operatorname{\mathcal{G}}. In this article, we determin some upper bounds for Ο‡GL​(H)\chi_{G}^{L}(H), in term of the Δ​(H),|V​(H)|\Delta(H),|V(H)| and δ​(G)\delta(G). In particular, we show that Ο‡GL​(H)=Ο‡G​(H)\chi_{G}^{L}(H)=\chi_{G}(H) for some graph HH and GG, Ο‡GL​(HβŠ•Hβ€²)≀χGL​(H)+Ο‡GL​(Hβ€²)\chi_{G}^{L}(H\oplus H^{\prime})\leq\chi_{G}^{L}(H)+\chi_{G}^{L}(H^{\prime}) for each GG, HH, and Hβ€²H^{\prime}. Also, we show that χ𝒒​(HβŠ•Kn)=χ𝒒L​(HβŠ•Kn)\chi_{\operatorname{\mathcal{G}}}(H\oplus K_{n})=\chi^{L}_{\operatorname{\mathcal{G}}}(H\oplus K_{n}), where 𝒒\operatorname{\mathcal{G}} is a collection of all dd-regular graphs, and some nn.

Key words and phrases:
𝒒\operatorname{\mathcal{G}}-free Subset, kk-choosable, dd-regular graphs, Conditional Coloring, Vertex Arboricity, LL-GG-free-colorable.
2010 Mathematics Subject Classification:
05C15.

1. Introduction

All graphs GG considered here are undirected, simple, and finite graphs. For given graph H=(V​(H),E​(H))H=(V(H),E(H)), the maximum degree of HH is denoted by Δ​(H)\Delta(H), and the minimum degree of HH is denoted by δ​(H)\delta(H). If vv be a vertex of HH, the degree and neighbors of vv are denoted by degH⁑(v)\deg_{H}{(v)} (deg⁑(v)\deg{(v)}) and NH​(v)N_{H}(v), respectively. The join of two graphs GG and HH is denoted by GβŠ•HG\oplus H and obtained from GG and HH by joining each vertex of GG to all vertices of HH. We use χ​(H)\chi(H) to denote the chromatic number of HH. For given graph HH, let each vertex vv of HH be assigned a set L​(V)L(V) of colors, called a color list. An LL-coloring of HH is a vertex-coloring cc so that:

  • β€’

    For any v∈V​(H)v\in V(H), c​(v)∈L​(v)c(v)\in L(v).

  • β€’

    c​(v)β‰ c​(vβ€²)c(v)\neq c(v^{\prime}) for each v​vβ€²βˆˆE​(H)vv^{\prime}\in E(H).

If there exists an LL-coloring of HH, then HH is called LL-colorable. A graph HH is said to be kk-choosable if there exists LL-coloring for any list-assignment LL satisfying |L​(V)|β‰₯k|L(V)|\geq k for each v∈V​(H)v\in V(H). The choice number Ο‡L​(H)\chi_{L}(H) of HH is the minimum integer kk so that HH is kk-choosable. Note that χ​(H)≀χL​(H)\chi(H)\leq\chi_{L}(H) for any graph HH, however, equality does not necessarily hold. The following particular case has been proved by Ohba [9]:

Theorem A.

[9] If |V​(H)|≀χ​(H)+2​χ​(H)|V(H)|\leq\chi(H)+\sqrt{2\chi(H)}, then:

Ο‡L​(H)=χ​(H).\chi_{L}(H)=\chi(H).
Theorem B.

[9] If |V​(H)|+|V​(G)|≀χ​(H)+χ​(G)+2​(χ​(H)+χ​(G))|V(H)|+|V(G)|\leq\chi(H)+\chi(G)+\sqrt{2(\chi(H)+\chi(G))}, then:

Ο‡L​(HβŠ•G)=χ​(H)+χ​(G).\chi_{L}(H\oplus G)=\chi(H)+\chi(G).

One can refer to [5, 4, 11, 7] and [14] and their references for further studies about LL-coloring of graph.

1.1. Conditional Coloring

For given graph HH and graphical property PP, the conditional chromatic number χ​(H,P)\chi(H,P) of HH, is the smallest number kk, so that V​(H)V(H) can be decomposed into sets V1,V2​…,VkV_{1},V_{2}\ldots,V_{k}, in which H​[Vi]H[V_{i}] satisfies the property PP, for each 1≀i≀k1\leq i\leq k. This extension of graph coloring was stated by Harary in 1985Β [2]. A particular state, when PP is the property of being acyclic, χ​(H,P)\chi(H,P) is said the vertex arboricity of HH. The vertex arboricity of HH is shown α​(H)\alpha(H) and is defined as the last number of subsets in a partition of the vertex set of HH, so that any subset induces an acyclic subgraph. When PP is the property that each color class contains no copy of GG, we write Ο‡G​(H)\chi_{G}(H) instead of χ​(G,P)\chi(G,P), which is called the GG-free chromatic number. Due to this, we say a graph HH has a kk-GG-free coloring if there exists a map g:V​(H)⟢{1,…,k}g:V(H)\longrightarrow\{1,\ldots,k\}, so that each of the color classes of gg be GG-free.

We use Ξ±L​(H)\alpha_{L}(H) to denote the list vertex arboricity of HH, which is the least integer kk, such that there exists an acyclic LL-coloring for each list assignment LL of HH, in which k≀|L​(v)|k\leq|L(v)|. So, α​(H)≀αL​(H)\alpha(H)\leq\alpha_{L}(H) for any graph HH. It has been proved that for each graph HH, α​(H)β‰€βŒˆΞ”β€‹(H)+12βŒ‰\alpha(H)\leq\lceil\frac{\Delta(H)+1}{2}\rceilΒ [1], and Ξ±L​(H)β‰€βŒˆΞ”β€‹(H)+12βŒ‰\alpha_{L}(H)\leq\lceil\frac{\Delta(H)+1}{2}\rceilΒ [12]. When HH is not a complete graph or a cycle of odd order, then α​(H)β‰€βŒˆΞ”2βŒ‰\alpha(H)\leq\lceil\frac{\Delta}{2}\rceil[6]. Also, it has been shown that α​(H)≀3,Ξ±L​(H)≀3\alpha(H)\leq 3,\alpha_{L}(H)\leq 3 if HH be a planar graph, [1, 3, 12]. As a generalized result of Theorem A and B, Lingyan Zhe, and Baoyindureng Wu have proven the following theorem [13].

Theorem C.

[13] If |V​(H)|≀2​α​(H)+2​α​(H)βˆ’1|V(H)|\leq 2\alpha(H)+\sqrt{2\alpha(H)}-1, then:

Ξ±L​(H)=α​(H).\alpha_{L}(H)=\alpha(H).

Assume that each vertex v∈V​(H)v\in V(H) is assigned a set L​(V)L(V) of colors, told a color list. Set c​(L)={c​(v):v∈V​(H)}c(L)=\{c(v):v\in V(H)\}. An LL-coloring cc is called GG-free, so that:

  • β€’

    c​(v)∈L​(v)c(v)\in L(v) for each v∈V​(H)v\in V(H).

  • β€’

    H​[Vi]H[V_{i}] is GG-free for each i=1,2,…,Li=1,2,\ldots,L.

If there exists an LL-coloring of HH, then HH is said to be LL-GG-free-colorable. A graph HH is said kk-GG-free-choosable if there exists an LL-coloring for any list-assignment LL satisfying |L​(V)|β‰₯k|L(V)|\geq k for each v∈V​(H)v\in V(H), and H​[Vi]H[V_{i}] be GG-free for each i=1,2,…,Li=1,2,\ldots,L. For given two graphs HH and GG, the Ο‡GL​(H)\chi_{G}^{L}(H) of HH is the minimum integer kk, if HH be kk-GG-free-choosable.

In this article, we shall use Ohba’s idea to show similar results in terms of GG-free coloring, and list GG-free coloring of graphs. In particular, in this article, we prove the subsequent results.

Theorem 1.

Suppose that HH, Hβ€²H^{\prime}, and GG are three graphs, where δ​(G)=Ξ΄\delta(G)=\delta, HH is a kk-GG-free choosable, and Hβ€²H^{\prime} is a kβ€²k^{\prime}-GG-free choosable. Suppose that SS and Sβ€²S^{\prime} be the maximum subsets of V​(H)V(H) and V​(Hβ€²)V(H^{\prime}), respectively, so that H​[S]H[S] is GG-free and H′​[Sβ€²]H^{\prime}[S^{\prime}] is GG-free. In this case, if either (|Sβ€²|βˆ’1)​(|V​(H)|+|Sβ€²|)≀|Sβ€²|​δ​(k+1)(|S^{\prime}|-1)(|V(H)|+|S^{\prime}|)\leq|S^{\prime}|\delta(k+1) or (|S|βˆ’1)​(|V​(Hβ€²)|+|S|)≀|S|​δ​(kβ€²+1)(|S|-1)(|V(H^{\prime})|+|S|)\leq|S|\delta(k^{\prime}+1), then HβŠ•Hβ€²H\oplus H^{\prime} is a (k+kβ€²)(k+k^{\prime})-GG-free-choosable, that is:

Ο‡GL​(HβŠ•Hβ€²)≀χGL​(H)+Ο‡GL​(Hβ€²)=k+kβ€².\chi_{G}^{L}(H\oplus H^{\prime})\leq\chi_{G}^{L}(H)+\chi_{G}^{L}(H^{\prime})=k+k^{\prime}.
Theorem 2.

If |V​(H)|≀δ​χG​(H)+δ​χG​(H)βˆ’(Ξ΄βˆ’1)|V(H)|\leq\delta\chi_{G}(H)+\sqrt{\delta\chi_{G}(H)}-(\delta-1), then:

Ο‡GL​(H)=Ο‡G​(H).\chi_{G}^{L}(H)=\chi_{G}(H).
Theorem 3.

For each two connected graphs HH and GG, we have:

Ο‡GL​(H)β‰€βŒˆΞ”β€‹(H)δ​(G)βŒ‰+1.\chi^{L}_{{}_{G}}(H)\leq\lceil\frac{\Delta(H)}{\delta(G)}\rceil+1.
Theorem 4.

Assume that ℝ\operatorname{\mathbb{R}} be a collection of all dd-regular graphs. For each arbitrary graph HH, there exists a non-negative integer nβ€²n^{\prime}, so that χℝ​(HβŠ•Kn)=χℝL​(HβŠ•Kn)\chi_{\operatorname{\mathbb{R}}}(H\oplus K_{n})=\chi^{L}_{\operatorname{\mathbb{R}}}(H\oplus K_{n}), where nn is integer so that for which n′≀nn^{\prime}\leq n.

2. Proof of the Main results

Before proving the main theorems, we need some basic results. Suppose that HH and GG are two graphs, and let LL be a list-assignment color to V​(H)V(H). Assume that S=v1,v2,…,vmβŠ†V​(H)S={v_{1},v_{2},\ldots,v_{m}}\subseteq V(H). Set L​(S)=βˆͺi=1i=mL​(vi)L(S)=\cup_{i=1}^{i=m}L(v_{i}). Now, we have the following lemma.

Lemma 5.

Suppose that HH and GG be two graphs, where δ​(G)=Ξ΄\delta(G)=\delta. Let HH is not LL-GG-free colorable. Hence there is a subset of V​(H)V(H) say SS, such that:

|S|>δ​|L​(S)|.|S|>\delta|L(S)|.
Proof.

By contradiction, suppose that |S|≀δ​|L​(S)||S|\leq\delta|L(S)| for each subset SS of V​(H)V(H). Now, define the bipartite graph BB with bipartition (Y,Yβ€²)(Y,Y^{\prime}), where Y=V​(H)Y=V(H), and Yβ€²Y^{\prime} is the Ξ΄\delta copies of L​(V​(H))L(V(H)). Also, for any member of YY say yy, and each member of Yβ€²Y^{\prime} say yβ€²y^{\prime}, y​yβ€²βˆˆE​(B)yy^{\prime}\in E(B) if yβ€²βˆˆL​(y)y^{\prime}\in L(y). It is clear to say that NB​(W)=δ​|L​(W)|N_{B}(W)=\delta|L(W)|, for each WβŠ†V​(H)W\subseteq V(H). Now, let SβŠ†V​(H)S\subseteq V(H). By the assumption, NB​(S)=δ​|L​(S)|β‰₯|S|N_{B}(S)=\delta|L(S)|\geq|S|, therefore, by Hall’s theorem, there exists a matching MM that saturates V​(H)V(H). Consider yy and color yy with the color matched by it in MM. As Yβ€²Y^{\prime} is a Ξ΄\delta copy of L​(V​(H))L(V(H)). Any member yβ€²y^{\prime} of L​(V​(H))L(V(H)) appears in at most Ξ΄\delta times as an end vertex of some edges of the MM. Which means that, any color of L​(H)L(H) is assigned in at most Ξ΄\delta vertices of V​(H)V(H). Hence, we achieve a LL-GG-free coloring of V​(H)V(H), which is impossible. Therefore the proof is complete. ∎

To prove the following lemma, we use Ohba’s notion.

Lemma 6.

Suppose that HH, Hβ€²H^{\prime}, and GG are three graphs, where δ​(G)=Ξ΄\delta(G)=\delta, |V​(Hβ€²)|=n|V(H^{\prime})|=n, and Hβ€²H^{\prime} is GG-free, i.e G⊈Hβ€²G\nsubseteq H^{\prime}. If HH be kk-GG-free choosable, and (nβˆ’1)​(|V​(H)|+n)≀n​δ​(k+1)(n-1)(|V(H)|+n)\leq n\delta(k+1), then HβŠ•Hβ€²H\oplus H^{\prime} is (k+1)(k+1)-GG-free choosable, that is:

Ο‡GL​(HβŠ•Hβ€²)≀χGL​(H)+1=k+1.\chi_{G}^{L}(H\oplus H^{\prime})\leq\chi_{G}^{L}(H)+1=k+1.
Proof.

The proof proceeds by induction on |V​(Hβ€²)||V(H^{\prime})|. Suppose that LL is a (k+1)(k+1)-list assignment of HβŠ•Hβ€²H\oplus H^{\prime}. For the induction basis, first assume that n=1n=1, and V​(Hβ€²)={vβ€²}V(H^{\prime})=\{v^{\prime}\}. We color vβ€²v^{\prime} by one of the members of L​(vβ€²)L(v^{\prime}), say cβ€²c^{\prime}. For each member of V​(H)V(H) say vv, assume that L​(v)βˆ–{cβ€²}=L′​(v)L(v)\setminus\{c^{\prime}\}=L^{\prime}(v). As for any member of V​(H)V(H) say vv, |L′​(v)|β‰₯|L​(v)|βˆ’1=k|L^{\prime}(v)|\geq|L(v)|-1=k, and HH is kk-GG-free choosable. Thus HβŠ•Hβ€²H\oplus H^{\prime} is (k+1)(k+1)-GG-free choosable. So, assume that nβ‰₯2n\geq 2. Suppose that Vβ€²=V​(Hβ€²)={v1β€²,v2β€²,…,vnβ€²}V^{\prime}=V(H^{\prime})=\{v^{\prime}_{1},v^{\prime}_{2},\ldots,v^{\prime}_{n}\}. Now, by considering β‹‚vβ€²βˆˆVβ€²L​(vβ€²)\bigcap\limits_{v^{\prime}\in V^{\prime}}L(v^{\prime}), we have two cases as follow:

Case 1: β‹‚vβ€²βˆˆVβ€²L​(vβ€²)β‰ βˆ…\bigcap\limits_{v^{\prime}\in V^{\prime}}L(v^{\prime})\neq\emptyset.

Set cβ€²βˆˆβ‹‚vβ€²βˆˆVβ€²L​(vβ€²)c^{\prime}\in\bigcap\limits_{v^{\prime}\in V^{\prime}}L(v^{\prime}) and assign cβ€²c^{\prime} to each member of V​(Hβ€²)V(H^{\prime}). Therefore, as G⊈Hβ€²G\nsubseteq H^{\prime}, one can check that HβŠ•H′​[Vcβ€²]H\oplus H^{\prime}[V_{c^{\prime}}] is GG-free. Now, for each member of V​(H)V(H) say vv, assume that L​(v)βˆ–{cβ€²}=L′​(v)L(v)\setminus\{c^{\prime}\}=L^{\prime}(v). As for each member of V​(H)V(H) say vv, |L′​(v)|β‰₯|L​(v)|βˆ’1=k|L^{\prime}(v)|\geq|L(v)|-1=k, and HH is kk-GG-free choosable, so HβŠ•Hβ€²H\oplus H^{\prime} is a (k+1)(k+1)-GG-free choosable.

Case 2: β‹‚vβ€²βˆˆVβ€²L​(vβ€²)=βˆ…\bigcap\limits_{v^{\prime}\in V^{\prime}}L(v^{\prime})=\emptyset.

By contradiction, assume that HβŠ•Hβ€²H\oplus H^{\prime} has no LL-GG-free coloring. By Lemma 5, there must exist SβŠ†V​(HβŠ•Hβ€²)S\subseteq V(H\oplus H^{\prime}) with |S|>δ​|L​(S)||S|>\delta|L(S)|. Suppose that SS be the maximal subset of V​(HβŠ•Hβ€²)V(H\oplus H^{\prime}) so that |S|>δ​|L​(S)||S|>\delta|L(S)|.

Suppose that Vβ€²βŠ†SV^{\prime}\subseteq S. As β‹‚vβ€²βˆˆVβ€²L​(vβ€²)=βˆ…\bigcap\limits_{v^{\prime}\in V^{\prime}}L(v^{\prime})=\emptyset, any color in L​(Vβ€²)L(V^{\prime}) appears in the lists of at most nβˆ’1n-1 vertices of Vβ€²V^{\prime}, therefore one can say that:

(1) |L​(S)|β‰₯|L​(Vβ€²)|β‰₯nnβˆ’1​(k+1).|L(S)|\geq|L(V^{\prime})|\geq\frac{n}{n-1}(k+1).

On the other hand, by the assumption, as (nβˆ’1)​(|V​(H)|+n)≀n​δ​(k+1)(n-1)(|V(H)|+n)\leq n\delta(k+1), it can be check that:

(2) |S|≀|V​(HβŠ•Hβ€²)|≀δ​nnβˆ’1​(k+1).|S|\leq|V(H\oplus H^{\prime})|\leq\delta\frac{n}{n-1}(k+1).

Therefore, by Equations 1, 2, we have |S|≀δ​|L​(S)||S|\leq\delta|L(S)|, a contradiction. So, we may suppose that Vβ€²βŠˆSV^{\prime}\nsubseteq S, that is |Vβ€²βˆ–S|β‰ 0|V^{\prime}\setminus S|\neq 0.

Set Vβ€²β€²=V​(HβŠ•Hβ€²)βˆ–SV^{\prime\prime}=V(H\oplus H^{\prime})\setminus S. Assume that L​(vβ€²β€²)βˆ–L​(S)=L′​(vβ€²β€²)L(v^{\prime\prime})\setminus L(S)=L^{\prime}(v^{\prime\prime}) for each vβ€²β€²βˆˆVβ€²β€²v^{\prime\prime}\in V^{\prime\prime}. By considering H​[S]H[S] and H​[Vβ€²β€²]H[V^{\prime\prime}], we have two claims as follow:

Claim 7.

H​[S]H[S] has a LL-GG-free coloring.

Proof of the Claim7.

As (nβˆ’1)​(|V​(H)|+n)≀n​δ​(k+1)(n-1)(|V(H)|+n)\leq n\delta(k+1), so (nβˆ’2)​(|V​(H)|+nβˆ’1)≀(nβˆ’1)​δ​(k+1)(n-2)(|V(H)|+n-1)\leq(n-1)\delta(k+1), because this operation is an increasing operation. Therefore, by the induction hypothesis, HβŠ•Hβ€²βˆ–{vβ€²}H\oplus H^{\prime}\setminus\{v^{\prime}\} is (k+1)(k+1)-GG-free list colorable, for any vβ€²βˆˆV​(Hβ€²)v^{\prime}\in V(H^{\prime}). Since H​[S]βŠ†HβŠ•Hβ€²βˆ–{vβ€²}H[S]\subseteq H\oplus H^{\prime}\setminus\{v^{\prime}\} it can be said H​[S]H[S] has an LL-GG-free coloring. ∎

Claim 8.

H​[Vβ€²β€²]H[V^{\prime\prime}] has an Lβ€²L^{\prime}-GG-free coloring.

Proof of the Claim8.

By contradiction, let H​[Vβ€²β€²]H[V^{\prime\prime}] is not Lβ€²L^{\prime}-GG-free colorable. Therefore, by Lemma 5, there exists a set Sβ€²S^{\prime} of Vβ€²β€²V^{\prime\prime}, so that |Sβ€²|>δ​|L′​(Sβ€²)||S^{\prime}|>\delta|L^{\prime}(S^{\prime})|. In the other hand, |L​(SβˆͺSβ€²)|=|L​(S)|+|L′​(Sβ€²)|<1δ​|S+Sβ€²||L(S\cup S^{\prime})|=|L(S)|+|L^{\prime}(S^{\prime})|<\frac{1}{\delta}|S+S^{\prime}|, therefore, δ​|L​(SβˆͺSβ€²)|<|S+Sβ€²|\delta|L(S\cup S^{\prime})|<|S+S^{\prime}| which contradicts to the maximality of SS. Hence, |Sβ€²|≀δ​|L′​(Sβ€²)||S^{\prime}|\leq\delta|L^{\prime}(S^{\prime})| for each Sβ€²βŠ†Vβ€²β€²S^{\prime}\subseteq V^{\prime\prime}, which means that H​[Vβ€²β€²]H[V^{\prime\prime}] has an Lβ€²L^{\prime}-GG-free coloring. ∎

Therefore, by Claim 7, and Claim 8, we can say that HβŠ•Hβ€²H\oplus H^{\prime} has a LL-GG-free colorable, a contradiction. Hence:

Ο‡GL​(HβŠ•Hβ€²)≀χGL​(H)+1.\chi_{G}^{L}(H\oplus H^{\prime})\leq\chi_{G}^{L}(H)+1.

Which means that the proof is complete. ∎

In the following, by using Lemma 6, we prove the first main result, namely Theorem 1.

Proof of the Theorem 1.

Without loss of generality (W.l.g), suppose that:

(|Sβ€²|βˆ’1)​(|V​(H)|+|Sβ€²|)≀|Sβ€²|​δ​(k+1).(|S^{\prime}|-1)(|V(H)|+|S^{\prime}|)\leq|S^{\prime}|\delta(k+1).

Where Sβ€²S^{\prime} is the maximum subset of V​(Hβ€²)V(H^{\prime}) so that G⊈H′​[Sβ€²]G\nsubseteq H^{\prime}[S^{\prime}]. Therefore, as Hβ€²H^{\prime} is kβ€²k^{\prime}-GG-free choosable and Ο‡G​(Hβ€²)≀χGL​(Hβ€²)\chi_{G}(H^{\prime})\leq\chi_{G}^{L}(H^{\prime}), so Ο‡G​(Hβ€²)≀kβ€²\chi_{G}(H^{\prime})\leq k^{\prime}. For i=1,2,…,kβ€²i=1,2,\ldots,k^{\prime}, set Viβ€²βŠ†V​(Hβ€²)V^{\prime}_{i}\subseteq V(H^{\prime}) where V1β€²V^{\prime}_{1} is the maximum subset of V​(Hβ€²)V(H^{\prime}) so that G⊈H′​[V1]G\nsubseteq H^{\prime}[V_{1}] and for each iβ‰₯2i\geq 2, Viβ€²V^{\prime}_{i} be the maximum subset of V​(Hβ€²)βˆ–(βˆͺj=1j=iβˆ’1Vjβ€²)V(H^{\prime})\setminus(\cup_{j=1}^{j=i-1}V^{\prime}_{j}), and (Hβ€²βˆ–(βˆͺj=1j=iβˆ’1Vjβ€²))​[Viβ€²](H^{\prime}\setminus(\cup_{j=1}^{j=i-1}V^{\prime}_{j}))[V^{\prime}_{i}] be GG-free. It can be said that |V1β€²|=|Sβ€²||V^{\prime}_{1}|=|S^{\prime}|, and |Viβ€²|≀|Viβˆ’1β€²||V^{\prime}_{i}|\leq|V^{\prime}_{i-1}| for each iβ‰₯2i\geq 2. So, as (|Sβ€²|βˆ’1)​(|V​(H)|+|Sβ€²|)≀|Sβ€²|​δ​(k+1)(|S^{\prime}|-1)(|V(H)|+|S^{\prime}|)\leq|S^{\prime}|\delta(k+1), Sβ€²S^{\prime} is the maximum subset of V​(Hβ€²)V(H^{\prime}), G⊈H′​[Sβ€²]G\nsubseteq H^{\prime}[S^{\prime}] and |V1|=|Sβ€²||V_{1}|=|S^{\prime}|, then by Lemma 6 we have the following:

(3) Ο‡GL​(HβŠ•H′​[V1β€²])≀χGL​(H)+1.\chi_{G}^{L}(H\oplus H^{\prime}[V^{\prime}_{1}])\leq\chi_{G}^{L}(H)+1.

For each iβ‰₯1i\geq 1, set Hi=Hiβˆ’1βŠ•H′​[Viβ€²]H_{i}=H_{i-1}\oplus H^{\prime}[V^{\prime}_{i}], where H0=HH_{0}=H. Therefore, it can be say that the following claim is true:

Claim 9.

For each iβ‰₯1i\geq 1, HβŠ•H′​[βˆͺj=1j=iVjβ€²]βŠ†HiH\oplus H^{\prime}[\cup_{j=1}^{j=i}V^{\prime}_{j}]\subseteq H_{i}.

Proof of Claim 9.

It is clear that V​(HβŠ•H′​[βˆͺj=1j=iVjβ€²])=V​(Hi)V(H\oplus H^{\prime}[\cup_{j=1}^{j=i}V^{\prime}_{j}])=V(H_{i}), also one can say that E​(HβŠ•H′​[βˆͺj=1j=iVjβ€²])=E​(Hi)βˆ–Eβ€²E(H\oplus H^{\prime}[\cup_{j=1}^{j=i}V^{\prime}_{j}])=E(H_{i})\setminus E^{\prime}, where Eβ€²={vj​vjβ€²:j≀jβ€²,vj∈Vjβ€²,vjβ€²βˆˆVj′′​f​o​r​e​a​c​h​j≀j′≀i}E^{\prime}=\{v_{j}v_{j^{\prime}}~{}:j\leq j^{\prime},~{}~{}v_{j}\in V^{\prime}_{j},~{}v_{j^{\prime}}\in V^{\prime}_{j^{\prime}}~{}~{}for~{}each~{}~{}j\leq j^{\prime}\leq i\}, which means that the claim is true. ∎

Also, since (|Sβ€²|βˆ’1)​(|V​(H)|+|Sβ€²|)≀|Sβ€²|​δ​(k+1)(|S^{\prime}|-1)(|V(H)|+|S^{\prime}|)\leq|S^{\prime}|\delta(k+1), |V1β€²|=|Sβ€²||V^{\prime}_{1}|=|S^{\prime}|, and |Viβ€²|≀|Viβˆ’1β€²||V^{\prime}_{i}|\leq|V^{\prime}_{i-1}| for each iβ‰₯2i\geq 2, it can be said that the following claim is true:

Claim 10.

For each i∈{1,2,…,kβ€²}i\in\{1,2,\ldots,k^{\prime}\}, we have:

(|Viβ€²|βˆ’1)​(|V​(H)|+|Viβ€²|)≀|Viβ€²|​δ​(k+1).(|V_{i}^{\prime}|-1)(|V(H)|+|V_{i}^{\prime}|)\leq|V_{i}^{\prime}|\delta(k+1).
Proof of Claim 10.

For i=1i=1, since |V1β€²|=|Sβ€²||V^{\prime}_{1}|=|S^{\prime}|, and (|Sβ€²|βˆ’1)​(|V​(H)|+|Sβ€²|)≀|Sβ€²|​δ​(k+1)(|S^{\prime}|-1)(|V(H)|+|S^{\prime}|)\leq|S^{\prime}|\delta(k+1), the proof is complete. So, suppose that 2≀j≀kβ€²2\leq j\leq k^{\prime}, also for each 2≀j≀kβ€²2\leq j\leq k^{\prime} suppose that |V1β€²|βˆ’|Vjβ€²|=tj|V^{\prime}_{1}|-|V^{\prime}_{j}|=t_{j}. Therefore, we need to show that (|Vjβ€²|βˆ’1)​(|V​(H)|+|Vjβ€²|)≀|Vjβ€²|​δ​(k+1)(|V^{\prime}_{j}|-1)(|V(H)|+|V^{\prime}_{j}|)\leq|V^{\prime}_{j}|\delta(k+1). As |V1β€²|βˆ’|Vjβ€²|=tj|V^{\prime}_{1}|-|V^{\prime}_{j}|=t_{j}, so:

(|Vjβ€²|βˆ’1)​(|V​(H)|+|Vjβ€²|)=(|V1β€²|βˆ’1βˆ’tj)​(|V​(H)|+|V1β€²|βˆ’tj)(|V^{\prime}_{j}|-1)(|V(H)|+|V^{\prime}_{j}|)=(|V^{\prime}_{1}|-1-t_{j})(|V(H)|+|V^{\prime}_{1}|-t_{j})
=(|V1β€²|βˆ’1)(|V(H)|+|V1β€²|)βˆ’(tj(|V1β€²|βˆ’1)βˆ’tj(|V(H)|+|V1β€²|βˆ’tj)=(|V^{\prime}_{1}|-1)(|V(H)|+|V^{\prime}_{1}|)-(t_{j}(|V^{\prime}_{1}|-1)-t_{j}(|V(H)|+|V^{\prime}_{1}|-t_{j})
≀(|V1β€²|)Ξ΄(k+1)βˆ’(tj(|V1β€²|βˆ’1)βˆ’tj(|V(H)|+|V1β€²|βˆ’tj)\leq(|V^{\prime}_{1}|)\delta(k+1)-(t_{j}(|V^{\prime}_{1}|-1)-t_{j}(|V(H)|+|V^{\prime}_{1}|-t_{j})
=(|Vjβ€²|)Ξ΄(k+1)+tjΞ΄(k+1)βˆ’(tj(|V1β€²|βˆ’1)βˆ’tj(|V(H)|+|V1β€²|βˆ’tj)=(|V^{\prime}_{j}|)\delta(k+1)+t_{j}\delta(k+1)-(t_{j}(|V^{\prime}_{1}|-1)-t_{j}(|V(H)|+|V^{\prime}_{1}|-t_{j})
=(|Vjβ€²|)​δ​(k+1)+tj​(δ​(k+1)βˆ’|V​(H)|βˆ’2​|V1β€²|+1+tj)=(|V^{\prime}_{j}|)\delta(k+1)+t_{j}(\delta(k+1)-|V(H)|-2|V^{\prime}_{1}|+1+t_{j})
=(|Vjβ€²|)​δ​(k+1)+tj​(k​δ+Ξ΄βˆ’|V​(H)|βˆ’|V1β€²|βˆ’|Vjβ€²|+1)=(|V^{\prime}_{j}|)\delta(k+1)+t_{j}(k\delta+\delta-|V(H)|-|V^{\prime}_{1}|-|V^{\prime}_{j}|+1)

So for each j∈{2,3,…,kβ€²}j\in\{2,3,\ldots,k^{\prime}\} we have the following equation:

(4) (|Vjβ€²|βˆ’1)​(|V​(H)|+|Vjβ€²|)≀(|Vjβ€²|)​δ​(k+1)+tj​(k​δ+Ξ΄βˆ’|V​(H)|βˆ’|V1β€²|βˆ’|Vjβ€²|+1)(|V^{\prime}_{j}|-1)(|V(H)|+|V^{\prime}_{j}|)\leq(|V^{\prime}_{j}|)\delta(k+1)+t_{j}(k\delta+\delta-|V(H)|-|V^{\prime}_{1}|-|V^{\prime}_{j}|+1)

Now by Equation 4, it is sufficient to show that for each j∈{2,3,…,kβ€²}j\in\{2,3,\ldots,k^{\prime}\} we have the followings:

tj​(k​δ+Ξ΄βˆ’|V​(H)|βˆ’|V1β€²|βˆ’|Vjβ€²|+1)≀0.t_{j}(k\delta+\delta-|V(H)|-|V^{\prime}_{1}|-|V^{\prime}_{j}|+1)\leq 0.

It is easy to say that |V​(H)|β‰₯(kβˆ’1)​δ+1|V(H)|\geq(k-1)\delta+1, also one can say that |V1β€²|+|Vjβ€²|β‰₯2​δ|V^{\prime}_{1}|+|V^{\prime}_{j}|\geq 2\delta, which means that:

tj​(k​δ+Ξ΄βˆ’|V​(H)|βˆ’|V1β€²|βˆ’|Vjβ€²|+1)≀0.t_{j}(k\delta+\delta-|V(H)|-|V^{\prime}_{1}|-|V^{\prime}_{j}|+1)\leq 0.

Hence, for each i∈{1,2,…,kβ€²}i\in\{1,2,\ldots,k^{\prime}\}:

(5) (|Viβ€²|βˆ’1)​(|V​(H)|+|Viβ€²|)≀|Viβ€²|​δ​(k+1).(|V_{i}^{\prime}|-1)(|V(H)|+|V_{i}^{\prime}|)\leq|V_{i}^{\prime}|\delta(k+1).

Therefore, Equation 5 shows that the proof of the claim is complete. ∎

Hence, by Claim 10 and by Lemma 6, for each i∈{1,,…,kβ€²}i\in\{1,,\ldots,k^{\prime}\} we can say that:

(6) Ο‡GL​(HβŠ•H′​[Viβ€²])≀χGL​(H)+1≀k+1.\chi_{G}^{L}(H\oplus H^{\prime}[V^{\prime}_{i}])\leq\chi_{G}^{L}(H)+1\leq k+1.

So, regarding Equation 6, and Ο‡G​(H′​[βˆͺj=1j=iVjβ€²])=i\chi_{G}(H^{\prime}[\cup_{j=1}^{j=i}V^{\prime}_{j}])=i, by assigning ii new separate and unique colors to each Viβ€²V^{\prime}_{i}, it can concluded that Ο‡GL​(HβŠ•H′​[βˆͺj=1j=iVjβ€²])≀k+i\chi_{G}^{L}(H\oplus H^{\prime}[\cup_{j=1}^{j=i}V^{\prime}_{j}])\leq k+i, for each i∈{1,2,…,kβ€²}i\in\{1,2,\ldots,k^{\prime}\}. Also, since HβŠ•Hβ€²β‰…HβŠ•H′​[βˆͺi=1i=kβ€²Viβ€²]H\oplus H^{\prime}\cong H\oplus H^{\prime}[\cup_{i=1}^{i=k^{\prime}}V^{\prime}_{i}], so:

(7) Ο‡GL​(HβŠ•Hβ€²)=Ο‡GL​(HβŠ•H′​[βˆͺi=1i=kβ€²Viβ€²])≀k+kβ€²=Ο‡GL​(H)+Ο‡GL​(Hβ€²).\chi_{G}^{L}(H\oplus H^{\prime})=\chi_{G}^{L}(H\oplus H^{\prime}[\cup_{i=1}^{i=k^{\prime}}V^{\prime}_{i}])\leq k+k^{\prime}=\chi_{G}^{L}(H)+\chi_{G}^{L}(H^{\prime}).

Equation 7 means that the proof is complete. ∎

Before proving Theorem 2, we need two lemmas, which we state and prove in the following. In the following lemma, we determine an upper bond for Ο‡GL​(H)\chi_{G}^{L}(H), for each graph HH and GG.

Lemma 11.

Suppose that HH and GG be two graphs, where δ​(G)=Ξ΄\delta(G)=\delta, |V​(H)|=n|V(H)|=n, then:

Ο‡GL​(H)β‰€βŒˆnΞ΄βŒ‰\chi_{G}^{L}(H)\leq\lceil\frac{n}{\delta}\rceil
Proof.

The proof proceeds by induction on |V​(H)||V(H)|. It is clear that Ο‡GL​(H)β‰€βŒˆnΞ΄βŒ‰\chi_{G}^{L}(H)\leq\lceil\frac{n}{\delta}\rceil when |V​(H)|∈{1,2,…,Ξ΄}|V(H)|\in\{1,2,\ldots,\delta\}. Hence, assume that |V​(H)|β‰₯Ξ΄+1|V(H)|\geq\delta+1 and suppose that LL is a ⌈nΞ΄βŒ‰\lceil\frac{n}{\delta}\rceil-list apportion of HH. If there exist v1,v2,…,vΞ΄βˆ’1v_{1},v_{2},\ldots,v_{\delta-1}, and vΞ΄v_{\delta} in HH so that L​(v1)∩L​(v2)βˆ©β€¦βˆ©L​(vΞ΄)β‰ βˆ…L(v_{1})\cap L(v_{2})\cap\ldots\cap L(v_{\delta})\neq\emptyset, then we catch a color c∈L​(v1)∩L​(v2)βˆ©β€¦βˆ©L​(vΞ΄)c\in L(v_{1})\cap L(v_{2})\cap\ldots\cap L(v_{\delta}) and allocate cc to v1,v2,…,vΞ΄βˆ’1v_{1},v_{2},\ldots,v_{\delta-1}, and vΞ΄v_{\delta} and set L​(v)βˆ–{c}=L​(v)L(v)\setminus\{c\}=L(v) for any v∈V​(H)βˆ–{v1,v2,…,vΞ΄}v\in V(H)\setminus\{v_{1},v_{2},\ldots,v_{\delta}\}. Regarding |L​(v)|β‰₯|L​(v)|βˆ’1β‰₯⌈nβˆ’Ξ΄Ξ΄βŒ‰|L(v)|\geq|L(v)|-1\geq\lceil\frac{n-\delta}{\delta}\rceil, and by the induction hypothesis, Hβˆ–{v1,v2,…,vΞ΄}H\setminus\{v_{1},v_{2},\ldots,v_{\delta}\} has an LL-GG-free coloring. Therefore, HH is LL-GG-free colorable. Otherwise, if for any Ξ΄\delta vertices v1,v2,…,vΞ΄βˆ’1v_{1},v_{2},\ldots,v_{\delta-1}, and vΞ΄v_{\delta} of HH, L​(v1)∩L​(v2)βˆ©β€¦βˆ©L​(vΞ΄)=βˆ…L(v_{1})\cap L(v_{2})\cap\ldots\cap L(v_{\delta})=\emptyset, then it is easy to say that HH is LL-GG-free colorable, by considering ⌈nΞ΄βŒ‰\lceil\frac{n}{\delta}\rceil class of size at most Ξ΄\delta. ∎

Suppose that HH and GG are two graphs, where δ​(G)=Ξ΄\delta(G)=\delta, |V​(H)|=n|V(H)|=n, and we may suppose that g:V​(H)β†’{1,2,…,Ο‡G​(H)}g:V(H)\rightarrow\{1,2,\ldots,\chi_{G}(H)\} is a Ο‡G​(H)\chi_{G}(H)-coloring of HH, so that for each i∈[Ο‡G​(H)]i\in[\chi_{G}(H)], G⊈H​[Vi]G\nsubseteq H[V_{i}] and Vi={v∈V​(H):g​(v)=i}V_{i}=\{v\in V(H):~{}~{}g(v)=i\}. For each ii, suppose that |Vi|=ni|V_{i}|=n_{i}, and w.l.g assume that n1β‰₯n2β‰₯…β‰₯nΟ‡G​(H)n_{1}\geq n_{2}\geq\ldots\geq n_{\chi_{G}(H)}. It is easy to say that niβ‰₯Ξ΄n_{i}\geq\delta for each i∈[Ο‡G​(H)βˆ’1]i\in[\chi_{G}(H)-1]. Now by considering nin_{i}, we have the following lemma.

Lemma 12.

If (n1βˆ’1)​|V​(H)|≀n1​δ​χG​(H)(n_{1}-1)|V(H)|\leq n_{1}\delta\chi_{G}(H). Then:

Ο‡GL​(H)=Ο‡G​(H).\chi_{G}^{L}(H)=\chi_{G}(H).
Proof.

Since Ο‡G​(H)≀χGL​(H)\chi_{G}(H)\leq\chi_{G}^{L}(H), it do to prove that HH is Ο‡G​(H)\chi_{G}(H)-GG-free choosable(list colorable). The proof proceeds by induction on Ο‡G​(H)\chi_{G}(H). If HH be GG-free, then Ο‡GL​(H)=Ο‡G​(H)=1\chi_{G}^{L}(H)=\chi_{G}(H)=1, so suppose that HH has at least one copy of GG as a subgraph, that is 2≀χG​(H)2\leq\chi_{G}(H). Since n1β‰₯Ξ΄n_{1}\geq\delta, then by lemma assumption one can say that:

(8) |V​(H)|≀n1​δn1βˆ’1​χG​(H).|V(H)|\leq\frac{n_{1}\delta}{n_{1}-1}\chi_{G}(H).

If n1=Ξ΄n_{1}=\delta, then as n1n_{1} is maximal, so Ο‡G​(H)=⌈|V​(H)|Ξ΄βŒ‰\chi_{G}(H)=\lceil\frac{|V(H)|}{\delta}\rceil, hence by Lemma 11, Ο‡GL​(H)=Ο‡G​(H)\chi_{G}^{L}(H)=\chi_{G}(H). Now, suppose that n1β‰₯Ξ΄+1n_{1}\geq\delta+1. Hence:

|V​(H)βˆ–V1|=|V​(H)|βˆ’n1|V(H)\setminus V_{1}|=|V(H)|-n_{1}
≀n1​δn1βˆ’1​χG​(H)βˆ’n1\leq\frac{n_{1}\delta}{n_{1}-1}\chi_{G}(H)-n_{1}
=n1n1βˆ’1​(δ​χG​(H)βˆ’n1+1)=\frac{n_{1}}{n_{1}-1}(\delta\chi_{G}(H)-n_{1}+1)
≀n1n1βˆ’1​(δ​χG​(H)βˆ’Ξ΄)\leq\frac{n_{1}}{n_{1}-1}(\delta\chi_{G}(H)-\delta)
≀n2​δn2βˆ’1​(Ο‡G​(H)βˆ’1)\leq\frac{n_{2}\delta}{n_{2}-1}(\chi_{G}(H)-1)

So, this inequality and induction hypothesis implies that Hβˆ–V1H\setminus V_{1} is (Ο‡G​(H)βˆ’1)(\chi_{G}(H)-1)-GG-free choosable, because V2V_{2} is the maximal color class of Hβˆ–V1H\setminus V_{1}. Now, regarding Equation 8 and the fact that Hβˆ–V1H\setminus V_{1} is (Ο‡G​(H)βˆ’1)(\chi_{G}(H)-1)-GG-free list colorable, Lemma 6 implies that HH is Ο‡G​(H)\chi_{G}(H)-GG-free choosable, which means that the proof is complete. ∎

In the following, we prove Theorem 2 and Theorem 3 by using Lemma 12.

Theorem 13.

(Theorem2) Suppose that HH and GG are two graphs, where δ​(G)=Ξ΄\delta(G)=\delta. If we have, |V​(H)|≀δ​χG​(H)+δ​χG​(H)βˆ’(Ξ΄βˆ’1)|V(H)|\leq\delta\chi_{G}(H)+\sqrt{\delta\chi_{G}(H)}-(\delta-1), then:

Ο‡GL​(H)=Ο‡G​(H).\chi_{G}^{L}(H)=\chi_{G}(H).
Proof.

Assume that g:V​(H)β†’{1,2,…,Ο‡G​(H)}g:V(H)\rightarrow\{1,2,\ldots,\chi_{G}(H)\} is the Ο‡G​(H)\chi_{G}(H)-coloring of HH, so that for each i∈[Ο‡G​(H)]i\in[\chi_{G}(H)], G⊈H​[Vi]G\nsubseteq H[V_{i}] and Vi={v∈V​(H):g​(v)=i}V_{i}=\{v\in V(H):~{}~{}g(v)=i\}. For each ii, suppose that |Vi|=ni|V_{i}|=n_{i}, and w.l.g assume that n1β‰₯n2β‰₯…β‰₯nΟ‡G​(H)n_{1}\geq n_{2}\geq\ldots\geq n_{\chi_{G}(H)}. It is easy to say that niβ‰₯Ξ΄n_{i}\geq\delta for each i∈[Ο‡G​(H)βˆ’1]i\in[\chi_{G}(H)-1], and nΟ‡G​(H)β‰₯1n_{\chi_{G}(H)}\geq 1. It can be checked that:

(9) n1≀|V​(H)|βˆ’(δ​χG​(H)βˆ’(2β€‹Ξ΄βˆ’1)).n_{1}\leq|V(H)|-(\delta\chi_{G}(H)-(2\delta-1)).

Otherwise one can check that βˆ‘i=1i=Ο‡G​(H)niβ‰₯|V​(H)|+1\sum_{i=1}^{i=\chi_{G}(H)}n_{i}\geq|V(H)|+1, a contradiction. Therefore, by assumption lemma and by Equation 9 we have the next equation:

(10) n1≀δ​χG​(H)+δ​χG​(H)βˆ’(Ξ΄βˆ’1)βˆ’(δ​χG​(H)βˆ’(2β€‹Ξ΄βˆ’1))=δ​χG​(H)+Ξ΄.n_{1}\leq\delta\chi_{G}(H)+\sqrt{\delta\chi_{G}(H)}-(\delta-1)-(\delta\chi_{G}(H)-(2\delta-1))=\sqrt{\delta\chi_{G}(H)}+\delta.

Hence, by Equation 10 we have the following:

n1n1βˆ’1​δ​χG​(H)β‰₯δ​χG​(H)+δδ​χG​(H)+(Ξ΄βˆ’1)×δ​χG​(H)\frac{n_{1}}{n_{1}-1}\delta\chi_{G}(H)\geq\frac{\sqrt{\delta\chi_{G}(H)}+\delta}{\sqrt{\delta\chi_{G}(H)}+(\delta-1)}\times\delta\chi_{G}(H)
=δ​χG​(H)+δ​χG​(H)δ​χG​(H)+(Ξ΄βˆ’1)=\delta\chi_{G}(H)+\frac{\delta\chi_{G}(H)}{\sqrt{\delta\chi_{G}(H)}+(\delta-1)}
>δ​χG​(H)+δ​χG​(H)βˆ’(Ξ΄βˆ’1)δ​χG​(H)+(Ξ΄βˆ’1)>\delta\chi_{G}(H)+\frac{\delta\chi_{G}(H)-(\delta-1)}{\sqrt{\delta\chi_{G}(H)}+(\delta-1)}
=δ​χG​(H)+δ​χG​(H)βˆ’(Ξ΄βˆ’1)β‰₯|V​(H)|=\delta\chi_{G}(H)+\sqrt{\delta\chi_{G}(H)}-(\delta-1)\geq|V(H)|

So, by Lemma 12, Ο‡GL​(H)=Ο‡G​(H)\chi_{G}^{L}(H)=\chi_{G}(H), which means that the proof is complete. ∎

Now by Theorem 13, one can say that the following results are valid. For this, first we need the following definition.

Definition 14.

𝒒\operatorname{\mathcal{G}}-free graph: Suppose that 𝒒\operatorname{\mathcal{G}} is a collection of some graphs, a given graph HH is called 𝒒\operatorname{\mathcal{G}}-free, if HH contains no copy of GG as a subgraph for each Gβˆˆπ’’G\in\operatorname{\mathcal{G}}. When 𝒒={G}\operatorname{\mathcal{G}}=\{G\}, then HH is GG-free if G⊈HG\nsubseteq H. In this attention, we say a graph HH has a kk-𝒒\operatorname{\mathcal{G}}-free coloring if there exists a map g:V​(H)⟢{1,…,k}g:V(H)\longrightarrow\{1,\ldots,k\} so that each of the color classes of gg is 𝒒\operatorname{\mathcal{G}}-free i.e. contain no copy of each member of 𝒒\operatorname{\mathcal{G}} as a subgraph. For given graphs HH and 𝒒\operatorname{\mathcal{G}}, the χ𝒒​(H)\chi_{\operatorname{\mathcal{G}}}(H) of HH is the minimum integer kk so that HH is kk-𝒒\operatorname{\mathcal{G}}-free coloring.

Corollary 15.

Suppose that 𝒒\operatorname{\mathcal{G}} be a collection of all dd-regular graphs, where for each Gβˆˆπ’’G\in\operatorname{\mathcal{G}} we have:

|V​(H)|≀d​χG​(H)+d​χG​(H)βˆ’(dβˆ’1).|V(H)|\leq d\chi_{G}(H)+\sqrt{d\chi_{G}(H)}-(d-1).

Then:

χ𝒒L​(H)=χ𝒒​(H).\chi_{\operatorname{\mathcal{G}}}^{L}(H)=\chi_{\operatorname{\mathcal{G}}}(H).
Proof.

Consider Gβˆˆπ’’G\in\operatorname{\mathcal{G}}, so that for which Ο‡G′​(H)≀χG​(H)\chi_{G^{\prime}}(H)\leq\chi_{G}(H) for each Gβ€²βˆˆπ’’G^{\prime}\in\operatorname{\mathcal{G}}. Therefore, one can say that χ𝒒​(H)≀χG​(H)\chi_{\operatorname{\mathcal{G}}}(H)\leq\chi_{G}(H), also by assumption and by Theorem 13 we have Ο‡G​(H)=Ο‡GL​(H)\chi_{G}(H)=\chi^{L}_{G}(H), that is χ𝒒​(H)≀χGL​(H)\chi_{\operatorname{\mathcal{G}}}(H)\leq\chi^{L}_{G}(H). So as Ο‡GL​(H)≀χ𝒒L​(H)\chi^{L}_{G}(H)\leq\chi^{L}_{\operatorname{\mathcal{G}}}(H) we have χ𝒒​(H)≀χ𝒒L​(H)\chi_{\operatorname{\mathcal{G}}}(H)\leq\chi^{L}_{\operatorname{\mathcal{G}}}(H). Now we have to show that χ𝒒L​(H)≀χ𝒒​(H)\chi^{L}_{\operatorname{\mathcal{G}}}(H)\leq\chi_{\operatorname{\mathcal{G}}}(H). Consider Gβ€²β€²βˆˆπ’’G^{\prime\prime}\in\operatorname{\mathcal{G}}, so that for which χ𝒒L​(H)≀χGβ€²β€²L​(H)\chi^{L}_{\operatorname{\mathcal{G}}}(H)\leq\chi^{L}_{G^{\prime\prime}}(H). Therefore, by assumption and by Theorem 13, we have Ο‡G′′​(H)=Ο‡Gβ€²β€²L​(H)\chi_{G^{\prime\prime}}(H)=\chi^{L}_{G^{\prime\prime}}(H), and as Ο‡G′′​(H)≀χ𝒒​(H)\chi_{G^{\prime\prime}}(H)\leq\chi_{\operatorname{\mathcal{G}}}(H), it can be said that χ𝒒L​(H)≀χ𝒒​(H)\chi^{L}_{\operatorname{\mathcal{G}}}(H)\leq\chi_{\operatorname{\mathcal{G}}}(H), which means that the proof is complete. ∎

In corollary 15, if we take d=2d=2 hence 𝒒={Cn,nβ‰₯3}\operatorname{\mathcal{G}}=\{C_{n},n\geq 3\} and for any arbitrary graph for HH, then it is easy to say that we get Theorem C. Also, for d=1d=1, we have 𝒒={K2}\operatorname{\mathcal{G}}=\{K_{2}\}. Hence, as |V​(H)|≀χ​(H)+χ​(H)≀χ​(H)+2​χ​(H)|V(H)|\leq\chi(H)+\sqrt{\chi(H)}\leq\chi(H)+\sqrt{2\chi(H)}, therefore by corollary 15, if we take d=1d=1, then we get Theorem A.

Theorem 16.

(Theorem 3) For each two connected graphs HH and GG, we have:

Ο‡GL​(H)β‰€βŒˆΞ”β€‹(H)δ​(G)βŒ‰+1.\chi^{L}_{{}_{G}}(H)\leq\lceil\frac{\Delta(H)}{\delta(G)}\rceil+1.
Proof.

Suppose that HH and GG are two connected graphs that satisfy the theorem conditions. Assume that LL is a kk-list assignment of HH, where k=βŒˆΞ”β€‹(H)δ​(G)βŒ‰+1k=\lceil\frac{\Delta(H)}{\delta(G)}\rceil+1. Suppose that HH is not GG-free. Otherwise, it is easy to say that Ο‡GL​(H)=1β‰€βŒˆΞ”β€‹(H)δ​(G)βŒ‰+1\chi^{L}_{{}_{G}}(H)=1\leq\lceil\frac{\Delta(H)}{\delta(G)}\rceil+1. Hence assume that GβŠ†HG\subseteq H, that is 2≀χG​(H)2\leq\chi_{{}_{G}}(H). Suppose that v1,v2,…,vnv_{1},v_{2},\ldots,v_{n} be an ordering of V​(H)V(H) so that for each ii, vertex viv_{i}, have at most Δ​(H)\Delta(H) low-indexed neighbors in HH. Now we color the vertices of HH from their lists by this order. Assume that v1,v2,…,viv_{1},v_{2},\ldots,v_{i} has been colored and every subset of v1,v2,…,viv_{1},v_{2},\ldots,v_{i} which colored with the same color be GG-free. Since vi+1v_{i+1} has at most Δ​(H)+1βˆ’Ξ΄β€‹(G)\Delta(H)+1-\delta(G) neighbors in v1,v2,…,viv_{1},v_{2},\ldots,v_{i}, there are at most βŒˆΞ”β€‹(H)δ​(G)βŒ‰\lceil\frac{\Delta(H)}{\delta(G)}\rceil of this GG containing at least Ξ΄\delta neighbors of vi+1v_{i+1}. Therefore, it can be said that for vi+1v_{i+1}, there exists at least one color available in L​(vi+1)L(v_{i+1}) which appears at most once among the low-indexed neighbor of vi+1v_{i+1}. We assign such color to vi+1v_{i+1}. Therefore, we gain a GG-free coloring of H​[v1,v2,…,vi+1]H[{v_{1},v_{2},\ldots,v_{i+1}}]. By continuing this method until i+1=ni+1=n, we obtain a GG-free LL-coloring of HH. ∎

2.1. Proof of Theorem 4

Before proving Theorem 4, we need to some results. Suppose that ℝ\operatorname{\mathbb{R}} is a collection of all dd-regular graphs and let HH be an arbitrary graph. Hence, we have the following lemma:

Lemma 17.

Suppose that nn be a positive integer, and HH be an arbitrary graph. So:

χℝ​(HβŠ•Kn)β‰₯χℝ​(H)+nd.\chi_{\operatorname{\mathbb{R}}}(H\oplus K_{n})\geq\frac{\chi_{\operatorname{\mathbb{R}}}(H)+n}{d}.
Proof.

Suppose that t=χℝ​(HβŠ•Kn)t=\chi_{\operatorname{\mathbb{R}}}(H\oplus K_{n}) and let gg be a tt-𝒒\operatorname{\mathcal{G}}-free coloring of HβŠ•KnH\oplus K_{n}. For each i∈{1,2,…,χℝ​(H)}i\in\{1,2,\ldots,\chi_{\operatorname{\mathbb{R}}}(H)\}, let:

Vi={v∈V​(HβŠ•Kn):g​(v)=i}.V_{i}=\{v\in V(H\oplus K_{n})~{}:~{}g(v)=i\}.

Since H​[Vi]H[V_{i}] is a ℝ\operatorname{\mathbb{R}}-free, and ℝ\operatorname{\mathbb{R}} is a collection of all dd-regular graphs, so Kd+1⊈H​[Vi]K_{d+1}\nsubseteq H[V_{i}] for each ii. Therefore, each ViV_{i} contains at most dd vertices of V​(Kn)V(K_{n}). Now, if tj=|{i:1≀i≀t​a​n​d​|Vi∩V​(Kn)|=j}|t_{j}=|\{i~{}:~{}1\leq i\leq t~{}~{}and~{}~{}|V_{i}\cap V(K_{n})|=j\}|, and kjβ‰ 0k_{j}\neq 0, then j≀dj\leq d. Therefore, t=t0+t1+…+tdt=t_{0}+t_{1}+\ldots+t_{d}, and it is easy to say that n=βˆ‘m=1m=dm​tmn=\sum_{m=1}^{m=d}mt_{m}, also one can say that χℝ​(H)β‰€βˆ‘m=0m=dβˆ’1tm\chi_{\operatorname{\mathbb{R}}}(H)\leq\sum_{m=0}^{m=d-1}t_{m}. Hence,

d​t=d​(βˆ‘m=0m=dtm)=d​t0+βˆ‘m=1m=dβˆ’1(dβˆ’m)​tm+βˆ‘m=1m=dm​tmdt=d(\sum_{m=0}^{m=d}t_{m})=dt_{0}+\sum_{m=1}^{m=d-1}(d-m)t_{m}+\sum_{m=1}^{m=d}mt_{m}
β‰₯(dβˆ’1)​t0+βˆ‘m=0m=dβˆ’1tm+βˆ‘m=1m=dm​tm\geq(d-1)t_{0}+\sum_{m=0}^{m=d-1}t_{m}+\sum_{m=1}^{m=d}mt_{m}
β‰₯χℝ​(H)+n\geq\chi_{\operatorname{\mathbb{R}}}(H)+n

Therefore, t=χℝ​(HβŠ•Kn)β‰₯χℝ​(H)+ndt=\chi_{\operatorname{\mathbb{R}}}(H\oplus K_{n})\geq\frac{\chi_{\operatorname{\mathbb{R}}}(H)+n}{d}, which means that proof is complete. ∎

Theorem 18.

Assume that ℝ\operatorname{\mathbb{R}} is a collection of all dd-regular graphs where dβ‰₯1d\geq 1. For each arbitrary graph HH, there exists a non-negative integer nβ€²n^{\prime}, so that χℝ​(HβŠ•Kn)=χℝL​(HβŠ•Kn)\chi_{\operatorname{\mathbb{R}}}(H\oplus K_{n})=\chi^{L}_{\operatorname{\mathbb{R}}}(H\oplus K_{n}), for any integer nn with nβ‰₯nβ€²n\geq n^{\prime}.

Proof.

Suppose that m=|V​(H)|m=|V(H)|, and g:V​(H)β†’{1,2,…,χℝ​(HβŠ•Kn)}g:V(H)\rightarrow\{1,2,\ldots,\chi_{\operatorname{\mathbb{R}}}(H\oplus K_{n})\} be an optimal ℝ\operatorname{\mathbb{R}}-free coloring of HβŠ•KnH\oplus K_{n}, with a color class of sizes n1β‰₯n2β‰₯…β‰₯nχℝ​(HβŠ•Kn)n_{1}\geq n_{2}\geq\ldots\geq n_{\chi_{\operatorname{\mathbb{R}}}(H\oplus K_{n})}. One can check that d≀n1≀m+dd\leq n_{1}\leq m+d for sufficiently large nn. By Lemma 17, χℝ​(HβŠ•Kn)β‰₯χℝ​(H)+nd\chi_{\operatorname{\mathbb{R}}}(H\oplus K_{n})\geq\frac{\chi_{\operatorname{\mathbb{R}}}(H)+n}{d}. Observe that m+n≀d​(m+d)m+dβˆ’1​d​n1n1+1βˆ’d≀d​n1n1+1βˆ’d​χℝ​(HβŠ•Kn)m+n\leq\frac{d(m+d)}{m+d-1}\frac{dn_{1}}{n_{1}+1-d}\leq\frac{dn_{1}}{n_{1}+1-d}\chi_{\operatorname{\mathbb{R}}}(H\oplus K_{n}) for sufficiently large nn. Therefore,

(n1+1βˆ’d)​(m+n)≀d​n1​χℝ​(HβŠ•Kn)(n_{1}+1-d)(m+n)\leq dn_{1}\chi_{\operatorname{\mathbb{R}}}(H\oplus K_{n})

Hence, by Lemma 12, χℝ​(HβŠ•Kn)=χℝL​(HβŠ•Kn)\chi_{\operatorname{\mathbb{R}}}(H\oplus K_{n})=\chi^{L}_{\operatorname{\mathbb{R}}}(H\oplus K_{n}), which means that the proof is complete. ∎

The following theorem has been proved by Ohba [9]:

Theorem 19.

[9] Let HH be a graph. Hence there is a positive integer nβ€²n^{\prime} so that χ​(HβŠ•Kn)=Ο‡L​(HβŠ•Kn)\chi(H\oplus K_{n})=\chi_{L}(H\oplus K_{n}), for each integer nn with n′≀nn^{\prime}\leq n.

As a generalized result of Theorem 19, Lingyan Zhe, and Baoyindureng Wu have proven the following theorem [13].

Theorem 20.

[13] Let HH be a graph. Hence there is a positive integer nβ€²n^{\prime} so that α​(HβŠ•Kn)=Ξ±L​(HβŠ•Kn)\alpha(H\oplus K_{n})=\alpha_{L}(H\oplus K_{n}), for each integer nn with n′≀nn^{\prime}\leq n.

In Theorem 18, if we take d=1d=1 hence ℝ={K2}\operatorname{\mathbb{R}}=\{K_{2}\}, then we get Theorem 19. Also, by setting d=2d=2, we have ℝ={Cn,nβ‰₯3}\operatorname{\mathbb{R}}=\{C_{n},n\geq 3\} and for any arbitrary graph for HH, then it is easy to say that we get Theorem 20.

2.2. Some research problems related to the contents of this paper.

In this section, we propose some research problems related to the contents of this paper. The following conjecture has been proposed by Ohba [9]:

Conjecture 1.

[9] If |V​(H)|≀2​χ​(H)+1|V(H)|\leq 2\chi(H)+1, then:

Ο‡L​(H)=χ​(H).\chi_{L}(H)=\chi(H).

Jonathan A. Noel et.al in [8] have proven the conjecture1. The first problem concerns Theorem 2 and Conjecture 1, as we address below:

Problem 20.1.

For each two connected graphs HH and GG, find a constant MM, so that if |V​(H)|≀M​χG​(H)|V(H)|\leq M\chi_{G}(H), Then:

Ο‡GL​(H)=Ο‡G​(H).\chi^{L}_{G}(H)=\chi_{G}(H).

Y.Rowshan and A.Taherkhani in [10] have proven the following theorem.

Theorem D.

Suppose that HH and GG be two connected graphs while HH satisfies the followinf items:

  • β€’

    If GG is regular, then H≇GH\ncong G.

  • β€’

    If Gβ‰…Kδ​(G)+1G\cong K_{\delta(G)+1}, then HH is not Kk​δ​(G)+1K_{k\delta(G)+1}.

  • β€’

    If G≅K2G\cong K_{2}, then HH is neither a complete graph nor an odd cycle.

Then:

Ο‡G​(H)β‰€βŒˆΞ”β€‹(H)δ​(G)βŒ‰.\chi_{{}_{G}}(H)\leq\lceil\frac{\Delta(H)}{\delta(G)}\rceil.

And one of those color classes is a maximum induced GG-free subgraph in HH.

The second problem concerns Theorem 3 and D, as we address below:

Problem 20.2.

Suppose that HH and GG be two connected graphs while HH satisfies the following items:

  • β€’

    If GG is regular, then H≇GH\ncong G.

  • β€’

    If Gβ‰…Kδ​(G)+1G\cong K_{\delta(G)+1}, then HH is not Kk​δ​(G)+1K_{k\delta(G)+1}.

  • β€’

    If G≅K2G\cong K_{2}, then HH is neither a complete graph nor an odd cycle.

Then:

Ο‡GL​(H)β‰€βŒˆΞ”β€‹(H)δ​(G)βŒ‰.\chi^{L}_{{}_{G}}(H)\leq\lceil\frac{\Delta(H)}{\delta(G)}\rceil.

And one of those color classes is a maximum induced GG-free subgraph in HH.

References

  • [1] Gary Chartrand, HudsonΒ V. Kronk, and CurtissΒ E. Wall. The point-arboricity of a graph. Israel J. Math., 6:169–175, 1968.
  • [2] Frank Harary. Conditional colorability in graphs. In Graphs and applications (Boulder, Colo., 1982), Wiley-Intersci. Publ., pages 127–136. Wiley, New York, 1985.
  • [3] Stephen Hedetniemi. On partitioning planar graphs. Canad. Math. Bull., 11:203–211, 1968.
  • [4] HalΒ A Kierstead, Andrew Salmon, and Ran Wang. On the choice number of complete multipartite graphs with part size four. European Journal of Combinatorics, 58:1–16, 2016.
  • [5] Jakub Kozik, Piotr Micek, and Xuding Zhu. Towards an on-line version of ohba’s conjecture. European Journal of Combinatorics, 36:110–121, 2014.
  • [6] HudsonΒ V. Kronk and John Mitchem. Critical point-arboritic graphs. J. London Math. Soc. (2), 9:459–466, 1974/75.
  • [7] JeffreyΒ Allen Mudrock. On the list coloring problem and its equitable variants. PhD thesis, Illinois Institute of Technology, 2018.
  • [8] JonathanΒ A Noel, BruceΒ A Reed, and Hehui Wu. A proof of a conjecture of ohba. Journal of Graph Theory, 79(2):86–102, 2015.
  • [9] Kyoji Ohba. On chromatic-choosable graphs. Journal of Graph Theory, 40(2):130–135, 2002.
  • [10] Yaser Rowshan and Ali Taherkhani. A catlin-type theorem for graph partitioning avoiding prescribed subgraphs. arXiv preprint arXiv:2002.04702, 2020.
  • [11] Rongxing Xu, Yeong-Nan Yeh, and Xuding Zhu. List colouring of graphs and generalized dyck paths. Discrete Mathematics, 341(3):810–819, 2018.
  • [12] Nini Xue and Baoyindureng Wu. List point arboricity of graphs. Discrete Mathematics, Algorithms and Applications, 4(02):1250027, 2012.
  • [13] Lingyan Zhen and Baoyindureng Wu. List point arboricity of dense graphs. Graphs and Combinatorics, 25(1):123–128, 2009.
  • [14] Jialu Zhu and Xuding Zhu. Chromatic Ξ»\lambda-choosable and Ξ»\lambda-paintable graphs. Journal of Graph Theory, 98(4):642–652, 2021.