arXiv is now an independent nonprofit! Learn more
License: arXiv.org perpetual non-exclusive license
arXiv:2608.19773v1 [math.CO] 20 Aug 2026

Non-persistence of equality between chromatic polynomials and list-color functions

Meiqiao Zhang Thanks: Email: meiqiaozhang95@163.com and meiqiaozhang@xmu.edu.cn. Affiliation: School of Mathematical Sciences, Xiamen University, China    Fengming Dong Thanks: Email: fengming.dong@nie.edu.sg and donggraph@163.com. Affiliation: National Institute of Education, Nanyang Technological University, Singapore
Abstract

For any graph GG, let P(G,k)P(G,k) and P(G,k)P_{\ell}(G,k) denote the chromatic polynomial and the list-color function of GG, respectively. It remains an open problem whether, for every graph GG and integer kk, the equality P(G,k)=P(G,k)>0P(G,k)=P_{\ell}(G,k)>0 implies that P(G,k+1)=P(G,k+1)P(G,k+1)=P_{\ell}(G,k+1) also holds. In this paper, we answer this question in the negative. For every integer k3k\geq 3, we construct an infinite family of graphs GG such that P(G,k)=P(G,k)>0P(G,k)=P_{\ell}(G,k)>0 while P(G,k+1)>P(G,k+1)P(G,k+1)>P_{\ell}(G,k+1). Moreover, using this infinite family of graphs as attachment gadgets, we further show that any graph HH with P(H,k)=P(H,k)>0P(H,k)=P_{\ell}(H,k)>0 can be developed into an infinite family of graphs HH^{\prime} with P(H,k)=P(H,k)>0P(H^{\prime},k)=P_{\ell}(H^{\prime},k)>0 and P(H,k+1)>P(H,k+1)P(H^{\prime},k+1)>P_{\ell}(H^{\prime},k+1).

1 Introduction

In this article, we consider simple graphs only. For any graph GG, let V(G)V(G) and E(G)E(G) be the vertex set and the edge set of GG, respectively. For any non-empty subset V0V_{0} of V(G)V(G), let G[V0]G[V_{0}] denote the subgraph of GG induced by V0V_{0}. Denote by \mathbb{N} the set of positive integers. For any k,rk,r\in\mathbb{N}, let [k]={1,,k}[k]=\{1,\dots,k\} and (k)r=k(k1)(kr+1)(k)_{r}=k(k-1)\cdots(k-r+1).

For any graph GG, a proper coloring of GG is a mapping θ:V(G)\theta:V(G)\rightarrow\mathbb{N}, such that θ(u)θ(v)\theta(u)\neq\theta(v) for all uvE(G)uv\in E(G). For any kk\in\mathbb{N}, a proper kk-coloring of GG is a proper coloring θ\theta with θ(v)[k]\theta(v)\in[k] for all vV(G)v\in V(G). Then the chromatic polynomial P(G,k)P(G,k) of GG is a polynomial that counts the number of proper kk-colorings of GG for each kk\in\mathbb{N}. We say GG is uniquely kk-colorable if V(G)V(G) has a unique partition into kk nonempty independent sets, which implies that P(G,k)=k!P(G,k)=k!. The chromatic polynomial was originally designed by Birkhoff in [5] as a tool to attack the Four-Color Conjecture, but later gained unique research significance because of its elegant properties. See [9, 10, 21, 22] for reference.

To generalize proper coloring, Vizing [24] and Erdős, Rubin and Taylor [13] independently introduced the notion of list-coloring. For any graph GG, a list assignment LL of GG is a mapping from V(G)V(G) to the power set of \mathbb{N}, and an LL-coloring of GG is a proper coloring θ\theta with θ(v)L(v)\theta(v)\in L(v) for all vV(G)v\in V(G). Denote the number of LL-colorings of GG by P(G,L)P(G,L). Then GG is called uniquely LL-colorable if P(G,L)=1P(G,L)=1.

We say LL is a kk-list assignment of GG if |L(v)|=k|L(v)|=k holds for all vV(G)v\in V(G). Then the list-color function P(G,k)P_{\ell}(G,k) of GG is defined to be the minimum value of P(G,L)P(G,L) over all kk-list assignments LL of GG for each kk\in\mathbb{N}. Observe that there is a trivial kk-list assignment LL^{*} of GG with L(v)=[k]L^{*}(v)=[k] for all vV(G)v\in V(G), which leads to the equality P(G,L)=P(G,k)P(G,L^{*})=P(G,k). Then by the definition of the list-color function, for each kk\in\mathbb{N},

P(G,k)P(G,k).\displaystyle P_{\ell}(G,k)\leq P(G,k). (1.1)

Note that the inequality of (1.1) can hold nontrivially, e.g., P(K2,4,2)=0<2=P(K2,4,2)P_{\ell}(K_{2,4},2)=0<2=P(K_{2,4},2). Consequently, Kostochka and Sidorenko [20] asked the question for which kk the equality of (1.1) holds. Very soon, Donner [12] answered this question by establishing a deletion-contraction formula for the list-color function.

Theorem 1 ([12]).

For any graph GG, P(G,k)=P(G,k)P(G,k)=P_{\ell}(G,k) if kk is sufficiently large.

In fact, Theorem 1 reveals that the list-color function of every graph tends to be the same as its chromatic polynomial. It further indicates that P(G,k)P_{\ell}(G,k) inherits all the nice properties of the chromatic polynomial when kk is sufficiently large. Following Theorem 1, the next question is naturally that for any graph GG, what is the minimum integer τ(G)\tau(G) such that P(G,k)=P(G,k)P(G,k)=P_{\ell}(G,k) whenever kτ(G)k\geq\tau(G). In 2009, Thomassen [23] gave the first answer regarding the order of the graph.

Theorem 2 ([23]).

For any graph GG, τ(G)|V(G)|10+1\tau(G)\leq|V(G)|^{10}+1.

Later in 2017, Wang, Qian and Yan [25] improved Theorem 2 by developing a list-coloring version of Whitney’s broken cycle theorem.

Theorem 3 ([25]).

For any graph GG, τ(G)|E(G)|1log(1+2)+1\tau(G)\leq\frac{|E(G)|-1}{\log(1+\sqrt{2})}+1.

Recently, by refining the techniques in [25], Dong and Zhang gave a better result in [11].

Theorem 4 ([11]).

For any graph GG, τ(G)|E(G)|1\tau(G)\leq|E(G)|-1.

By the definition of τ(G)\tau(G), the authors of [12, 23, 25, 11] were required to verify the equality for all relatively large integers kk, which is by no means trivial. This naturally raises the question of whether such a verification is always necessary. More specifically, the following question has recurred repeatedly over the past 10 years.

Question 5 ([7, 14, 8, 6, 15, 18, 4, 19]).

For all graphs GG and integers kk, does the equality P(G,k)=P(G,k)>0P(G,k)=P_{\ell}(G,k)>0 always imply P(G,k+1)=P(G,k+1)P(G,k+1)=P_{\ell}(G,k+1)?

Question 5 is trivially true for k=1k=1. For k=2k=2, all graphs GG satisfying P(G,2)=P(G,2)>0P(G,2)=P_{\ell}(G,2)>0 have been characterized in [3], and each such graph has the property that P(G,k)=P(G,k)P(G,k)=P_{\ell}(G,k) holds for all kk\in\mathbb{N}. For k3k\geq 3, the question remains open since 2016.

Also, a question parallel to Question 5 concerning the DP-color function and the chromatic polynomial was posed in [18] and subsequently resolved in [8, 6] by constructing an infinite family of counterexamples.

In this paper, we show that the answer to Question 5 is also negative by constructing an infinite family of graphs as counterexamples.

Theorem 6.

For each integer k3k\geq 3, there are infinitely many graphs GG such that P(G,k)=P(G,k)>0P(G,k)=P_{\ell}(G,k)>0 and P(G,k+1)>P(G,k+1)P(G,k+1)>P_{\ell}(G,k+1).

Moreover, using the graphs constructed in Theorem 6, we are able to provide an infinite family of counterexamples based on any graph GG with P(G,k)=P(G,k)>0P(G,k)=P_{\ell}(G,k)>0 when k3k\geq 3.

Theorem 7.

For any integer k3k\geq 3 and graph GG with P(G,k)=P(G,k)>0P(G,k)=P_{\ell}(G,k)>0, there exists an infinite set 𝒢np(G){\mathcal{G}}^{np}(G) such that for each graph G𝒢np(G)G^{\prime}\in{\mathcal{G}}^{np}(G), P(G,k)=P(G,k)>0P(G^{\prime},k)=P_{\ell}(G^{\prime},k)>0 but P(G,k+1)>P(G,k+1)P(G^{\prime},k+1)>P_{\ell}(G^{\prime},k+1).

Theorems 6 and 7 will be proven in Sections 2 and 3, respectively, with further discussion provided in Section 4.

2 Proof of Theorem 6

In this section, we prove Theorem 6 by providing for each integer k3k\geq 3 an infinite family of graphs Gk,tG_{k,t} with P(Gk,t,k)=P(Gk,t,k)>0P_{\ell}(G_{k,t},k)=P(G_{k,t},k)>0 and P(Gk,t,k+1)<P(Gk,t,k+1)P_{\ell}(G_{k,t},k+1)<P(G_{k,t},k+1).

Let kk be a fixed integer with k3k\geq 3, and let tt\in\mathbb{N}. Now we give the construction of Gk,tG_{k,t}. Let

V(Gk,t)=XkSYt,V(G_{k,t})=X_{k}\cup S\cup Y_{t},

where

{Xk={xi:i[k]},S={si:i[2]},Yt={yi,j:i[4],j[t]},\left\{\begin{aligned} &X_{k}=\{x_{i}:i\in[k]\},\\ &S=\{s_{i}:i\in[2]\},\\ &Y_{t}=\{y_{i,j}:i\in[4],\,j\in[t]\},\end{aligned}\right.

and

E(Gk,t)=\displaystyle E(G_{k,t})= {xixj:i,j[k],ij}{sixj:i[2],j[k],ij}\displaystyle\{x_{i}x_{j}:i,j\in[k],i\neq j\}\cup\{s_{i}x_{j}:i\in[2],j\in[k],i\neq j\}
{yi,jsr:i[4],j[t],r[2]}.\displaystyle\cup\{y_{i,j}s_{r}:i\in[4],j\in[t],r\in[2]\}.
x1x_{1}x2x_{2}x3x_{3}s1s_{1}s2s_{2}y1,1y_{1,1}y2,1y_{2,1}y3,1y_{3,1}y4,1y_{4,1}y1,2y_{1,2}y2,2y_{2,2}y3,2y_{3,2}y4,2y_{4,2}XkX_{k}SSYtY_{t}
Figure 1: The graph G3,2G_{3,2}.

An example for G3,2G_{3,2} is as shown in Figure 1. It is clear that |V(Gk,t)|=k+2+4t|V(G_{k,t})|=k+2+4t, and Gk,t[Xk]G_{k,t}[X_{k}] and all Gk,t[Xk{si}{xi}]G_{k,t}[X_{k}\cup\{s_{i}\}\setminus\{x_{i}\}] with i[2]i\in[2] are isomorphic to KkK_{k}.

In the following, we shall prove that each Gk,tG_{k,t} is a graph satisfying Theorem 6 whenever tt is sufficiently large. We first show that P(Gk,t,k)=P(Gk,t,k)P_{\ell}(G_{k,t},k)=P(G_{k,t},k).

Lemma 8.

For any k3k\geq 3 and tt\in\mathbb{N},

P(Gk,t,k)=P(Gk,t,k)=k!(k2)4t.\displaystyle P_{\ell}(G_{k,t},k)=P(G_{k,t},k)=k!(k-2)^{4t}. (2.1)
Proof.

We shall prove that P(Gk,t,k)=k!(k2)4tP(G_{k,t},k)=k!(k-2)^{4t} by counting the number of proper kk-colorings of Gk,tG_{k,t} in the order of x1,x2,,xk,s1,s2,y1,1,y1,2,,y4,tx_{1},x_{2},\dots,x_{k},s_{1},s_{2},y_{1,1},y_{1,2},\dots,y_{4,t}. It is easy to see that we have ki+1k-i+1 available colors for each xix_{i}. Since k3k\geq 3 and each sis_{i} is adjacent to k1k-1 vertices in XkX_{k}, we then have exactly one available color for each sis_{i}, which is the color we assign to xix_{i}. Moreover, in any proper kk-coloring of Gk,tG_{k,t}, the colors of all xix_{i} are pairwise distinct as all xix_{i} together forms a clique. Thus s1s_{1} and s2s_{2} have distinct colors, which leaves exactly k2k-2 available colors for each yi,jy_{i,j}. Hence P(Gk,t,k)=k!(k2)4tP(G_{k,t},k)=k!(k-2)^{4t}.

Now it remains to show that P(Gk,t,k)=k!(k2)4tP_{\ell}(G_{k,t},k)=k!(k-2)^{4t}. Since P(Gk,t,k)P(Gk,t,k)P_{\ell}(G_{k,t},k)\leq P(G_{k,t},k), we need only to prove that P(Gk,t,L)k!(k2)4tP(G_{k,t},L)\geq k!(k-2)^{4t} for all kk-list assignments LL of Gk,tG_{k,t}.

Let LL be any kk-list assignment of Gk,tG_{k,t}. We shall give a lower bound for the number of LL-colorings in the order of x1,x2,,xk,s1,s2,y1,1,y1,2,,y4,tx_{1},x_{2},\dots,x_{k},s_{1},s_{2},y_{1,1},y_{1,2},\dots,y_{4,t}. It is easy to see that we have at least ki+1k-i+1 available colors for each xix_{i}, at least one available color for each sis_{i}, and at least k2k-2 colors for each yi,jy_{i,j}, which implies P(Gk,t,L)k!(k2)4tP(G_{k,t},L)\geq k!(k-2)^{4t}. ∎

Next, we establish a lower bound for P(Gk,t,k+1)P(G_{k,t},k+1).

Lemma 9.

For any k3k\geq 3 and tt\in\mathbb{N},

P(Gk,t,k+1)(k+1)!k4t.\displaystyle P(G_{k,t},k+1)\geq(k+1)!k^{4t}. (2.2)
Proof.

We shall also prove by establishing a lower bound for the number of proper (k+1)(k+1)-colorings of Gk,tG_{k,t} in the order of s1,s2,x1,x2,,xk,y1,1,y1,2,,y4,ts_{1},s_{2},x_{1},x_{2},\dots,x_{k},y_{1,1},y_{1,2},\dots,y_{4,t}. We can first assign a same color to both s1s_{1} and s2s_{2}, for which we have k+1k+1 choices. Then we have ki+1k-i+1 available colors for each xix_{i}, and kk available colors for each yi,jy_{i,j}. Hence (2.2) holds. ∎

In the following, we shall determine an upper bound for P(Gk,t,k+1)P_{\ell}(G_{k,t},k+1) by finding a special (k+1k+1)-list assignment for Gk,tG_{k,t}.

Let C,B1,B2C,B_{1},B_{2} be pairwise disjoint sets where C={c1,c2,,ck1}C=\{c_{1},c_{2},\dots,c_{k-1}\}, B1={b1,1,b1,2}B_{1}=\{b_{1,1},b_{1,2}\} and B2={b2,1,b2,2}B_{2}=\{b_{2,1},b_{2,2}\}. Then let D1={b1,1,b2,1}D_{1}=\{b_{1,1},b_{2,1}\}, D2={b1,1,b2,2}D_{2}=\{b_{1,1},b_{2,2}\}, D3={b1,2,b2,1}D_{3}=\{b_{1,2},b_{2,1}\}, and D4={b1,2,b2,2}D_{4}=\{b_{1,2},b_{2,2}\}.

Now let LL be the (k+1)(k+1)-list assignment of Gk,tG_{k,t} such that

{L(xi)=CB1for alli[k],L(si)=CBifor alli[2],L(yi,j)=CDifor alli[4],j[t].\displaystyle\left\{\begin{aligned} &L(x_{i})=C\cup B_{1}~\text{for all}~i\in[k],\\ &L(s_{i})=C\cup B_{i}~\text{for all}~i\in[2],\\ &L(y_{i,j})=C\cup D_{i}~\text{for all}~i\in[4],\,j\in[t].\end{aligned}\right. (2.3)

Then it is easy to see that LL is a (k+1k+1)-list assignment, hence P(Gk,t,k+1)P(Gk,t,L).P_{\ell}(G_{k,t},k+1)\leq P(G_{k,t},L). We shall further determine P(Gk,t,L)P(G_{k,t},L) by the next two lemmas.

Lemma 10.

Let L|XkSL|_{X_{k}\cup S} be the (k+1)(k+1)-list assignment of Gk,t[XkS]G_{k,t}[X_{k}\cup S] which is the restriction of LL to XkSX_{k}\cup S. Then for each L|XkSL|_{X_{k}\cup S}-coloring θ\theta of Gk,t[XkS]G_{k,t}[X_{k}\cup S], the number of LL-colorings cc of Gk,tG_{k,t} with c|XkS=θc|_{X_{k}\cup S}=\theta is

{k4t,if θ(s1)=θ(s2),(k1)4t,if θ(s1)θ(s2)andθ(s1),θ(s2)C,(k(k1))2t,if θ(s1)θ(s2)and|{θ(s1),θ(s2)}C|=1,((k1)k2(k+1))t,if θ(s1)θ(s2)and|{θ(s1),θ(s2)}C|=0.\left\{\begin{aligned} &k^{4t},&&\qquad\text{if }\theta(s_{1})=\theta(s_{2}),\\ &(k-1)^{4t},&&\qquad\text{if }\theta(s_{1})\neq\theta(s_{2})~\text{and}~\theta(s_{1}),\theta(s_{2})\in C,\\ &\bigl(k(k-1)\bigr)^{2t},&&\qquad\text{if }\theta(s_{1})\neq\theta(s_{2})~\text{and}~|\{\theta(s_{1}),\theta(s_{2})\}\cap C|=1,\\ &\bigl((k-1)k^{2}(k+1)\bigr)^{t},&&\qquad\text{if }\theta(s_{1})\neq\theta(s_{2})~\text{and}~|\{\theta(s_{1}),\theta(s_{2})\}\cap C|=0.\end{aligned}\right.
Proof.

Let θ\theta be any L|XkSL|_{X_{k}\cup S}-coloring of Gk,t[XkS]G_{k,t}[X_{k}\cup S], and let Z={θ(si):i[2]}Z=\{\theta(s_{i}):i\in[2]\}. Then 1|Z|21\leq|Z|\leq 2. To extend θ\theta into an LL-coloring of Gk,tG_{k,t}, it remains to assign colors to the vertices in YtY_{t}. For each yi,jYty_{i,j}\in Y_{t}, we have |(CDi)Z||(C\cup D_{i})\setminus Z| available colors to choose. Then the number of LL-colorings of Gk,tG_{k,t} extending θ\theta is

(i=14|(CDi)Z|)t.\displaystyle\left(\prod_{i=1}^{4}\left|(C\cup D_{i})\setminus Z\right|\right)^{t}. (2.4)

If |Z|=1|Z|=1 (i.e., θ(s1)=θ(s2)\theta(s_{1})=\theta(s_{2})), then θ(s1)=θ(s2)=csC\theta(s_{1})=\theta(s_{2})=c_{s}\in C for some s[k1]s\in[k-1]. Thus |(CDi)Z|=k|(C\cup D_{i})\setminus Z|=k for all i[4]i\in[4], implying that

i=14|(CDi)Z|=k4.\displaystyle\prod_{i=1}^{4}|(C\cup D_{i})\setminus Z|=k^{4}. (2.5)

By (2.4), the result holds when θ(s1)=θ(s2)\theta(s_{1})=\theta(s_{2}).

Now consider the case |Z|=2|Z|=2 (i.e., θ(s1)θ(s2)\theta(s_{1})\neq\theta(s_{2}).) Let q=|ZC|q=|Z\cap C| and p=|Z|qp=|Z|-q. Then pp is the number of colors in Z(B1B2)Z\cap(B_{1}\cup B_{2}). Moreover, either (q,p)=(2,0)(q,p)=(2,0), or (q,p)=(1,1)(q,p)=(1,1), or (q,p)=(0,2)(q,p)=(0,2). We shall analyze each of the three cases below.

Case 1: (q,p)=(2,0)(q,p)=(2,0).

In this case, θ(s1)θ(s2)\theta(s_{1})\neq\theta(s_{2}) and θ(s1),θ(s2)C\theta(s_{1}),\theta(s_{2})\in C. Then |(CDi)Z|=k1\left|(C\cup D_{i})\setminus Z\right|=k-1 for all i[4]i\in[4], implying that

i=14|(CDi)Z|=(k1)4.\displaystyle\prod_{i=1}^{4}|(C\cup D_{i})\setminus Z|=(k-1)^{4}. (2.6)

Case 2: (q,p)=(1,1)(q,p)=(1,1).

In this case, θ(s1)θ(s2)\theta(s_{1})\neq\theta(s_{2}) and |{θ(s1),θ(s2)}C|=1|\{\theta(s_{1}),\theta(s_{2})\}\cap C|=1. Without loss of generality, assume that Z={c1,b1,1}Z=\{c_{1},b_{1,1}\}. Then,

{|(CD1)Z|=|(CD2)Z|=k1;|(CD3)Z|=|(CD4)Z|=k.\displaystyle\left\{\begin{array}[]{l}|(C\cup D_{1})\setminus Z|=|(C\cup D_{2})\setminus Z|=k-1;\\ |(C\cup D_{3})\setminus Z|=|(C\cup D_{4})\setminus Z|=k.\end{array}\right.

Thus

i=14|(CDi)Z|=k2(k1)2.\displaystyle\prod_{i=1}^{4}|(C\cup D_{i})\setminus Z|=k^{2}(k-1)^{2}. (2.9)

Case 3: (q,p)=(0,2)(q,p)=(0,2).

In this case, θ(s1)θ(s2)\theta(s_{1})\neq\theta(s_{2}) and |{θ(s1),θ(s2)}C|=0|\{\theta(s_{1}),\theta(s_{2})\}\cap C|=0. Without loss of generality, assume that Z={b1,1,b2,1}Z=\{b_{1,1},b_{2,1}\}. Then,

{|(CD1)Z|=k1;|(CD2)Z|=|(CD3)Z|=k;|(CD4)Z|=k+1.\displaystyle\left\{\begin{array}[]{l}|(C\cup D_{1})\setminus Z|=k-1;\\ |(C\cup D_{2})\setminus Z|=|(C\cup D_{3})\setminus Z|=k;\\ |(C\cup D_{4})\setminus Z|=k+1.\end{array}\right.

Thus,

i=14|(CDi)Z|=(k1)k2(k+1).\displaystyle\prod_{i=1}^{4}|(C\cup D_{i})\setminus Z|=(k-1)k^{2}(k+1). (2.13)

By (2.4) and the results in the three cases above, the lemma is proven. ∎

Lemma 11.

For any k3k\geq 3 and tt\in\mathbb{N}, let LL be the (k+1)(k+1)-list assignment of Gk,tG_{k,t} defined in (2.3). Then

P(Gk,t,L)=\displaystyle P(G_{k,t},L)= (k1)k!k4t+3(k2)(k1)!(k1)4t+1+(4(k1)k!CLOSE\displaystyle(k-1)k!k^{4t}+3(k-2)(k-1)!(k-1)^{4t+1}+\big(4(k-1)k!
OPEN+6(k1)(k1)!)(k(k1))2t+8k!((k1)k2(k+1))t.\displaystyle+6(k-1)(k-1)!\big)\bigl(k(k-1)\bigr)^{2t}+8k!\bigl((k-1)k^{2}(k+1)\bigr)^{t}. (2.14)
Proof.

We shall prove by analyzing all possible L|XkSL|_{X_{k}\cup S}-colorings of Gk,t[XkS]G_{k,t}[X_{k}\cup S] and combine them with Lemma 10.

We first count the number of L|XkSL|_{X_{k}\cup S}-colorings θ\theta of Gk,t[XkS]G_{k,t}[X_{k}\cup S] such that θ(s1)=θ(s2)\theta(s_{1})=\theta(s_{2}). Suppose θ(s1)=θ(s2)=c\theta(s_{1})=\theta(s_{2})=c. Then cCc\in C and cc can be any one of the k1k-1 colors in CC. According to the construction of Gk,tG_{k,t}, each vertex in XkX_{k} is adjacent to at least one vertex in SS, which implies that no vertex in XkX_{k} can receive color cc. Therefore, all vertices in XkX_{k} have all colors in (C{c})B1(C\setminus\{c\})\cup B_{1} to use, where |(C{c})B1|=k|(C\setminus\{c\})\cup B_{1}|=k. Since XkX_{k} is a clique of size kk, we have exactly k!k! ways to assign colors to XkX_{k}. Thus in this case, the number of L|XkSL|_{X_{k}\cup S}-colorings θ\theta of Gk,t[XkS]G_{k,t}[X_{k}\cup S] such that θ(s1)=θ(s2)\theta(s_{1})=\theta(s_{2}) is (k1)k!(k-1)k!. Then by Lemma 10, the number of LL-colorings θ\theta^{\prime} of Gk,tG_{k,t} such that θ(s1)=θ(s2)\theta^{\prime}(s_{1})=\theta^{\prime}(s_{2}) is (k1)k!k4t(k-1)k!k^{4t}.

For the remaining L|XkSL|_{X_{k}\cup S}-colorings θ\theta of Gk,t[XkS]G_{k,t}[X_{k}\cup S], θ(s1)θ(s2)\theta(s_{1})\neq\theta(s_{2}) and we shall discuss each of the three cases below.

Case 1: θ(s1)θ(s2)\theta(s_{1})\neq\theta(s_{2}) and θ(s1),θ(s2)C\theta(s_{1}),\theta(s_{2})\in C.

For this case, we count by coloring the vertices in the order s1,s2,x3,x4,,xk,x1,x2s_{1},s_{2},x_{3},x_{4},\dots,x_{k},x_{1},x_{2}. There are (k1)(k2)(k-1)(k-2) ways to color s1s_{1} and s2s_{2}, followed by (k1)(k2)2(k-1)(k-2)\cdots 2 ways to color x3,x4,,xkx_{3},x_{4},\dots,x_{k}, which leaves exactly one remaining color in CB1C\cup B_{1}, say qq. We may then assign to (x1,x2)(x_{1},x_{2}) any of the three ordered pairs (θ(s1),q)(\theta(s_{1}),q), (q,θ(s2))(q,\theta(s_{2})), and (θ(s1),θ(s2))(\theta(s_{1}),\theta(s_{2})). Hence, the number of such colorings is 3(k1)(k2)(k1)!3(k-1)(k-2)(k-1)!.

Case 2: θ(s1)θ(s2)\theta(s_{1})\neq\theta(s_{2}) and |{θ(s1),θ(s2)}C|=1|\{\theta(s_{1}),\theta(s_{2})\}\cap C|=1.

Suppose θ(s1)C\theta(s_{1})\in C and θ(s2)B2\theta(s_{2})\in B_{2} first. Then we count the colorings in the order s1,s2,x2,x3,,xk,x1s_{1},s_{2},x_{2},x_{3},\dots,x_{k},x_{1}. There are 2(k1)2(k-1) ways to color s1s_{1} and s2s_{2}, followed by k(k1)2k(k-1)\cdots 2 ways to color x2,x3,,xkx_{2},x_{3},\dots,x_{k}. This leaves exactly one remaining color in CB1C\cup B_{1}, say qq. We may then assign the color qq or θ(s1)\theta(s_{1}) to x1x_{1}. Hence, the number of such colorings is 4(k1)k!4(k-1)k!.

Then we suppose θ(s2)C\theta(s_{2})\in C and θ(s1)B1\theta(s_{1})\in B_{1}. We count the colorings in the order s1,s2,x3,x4,,xk,x1,x2s_{1},s_{2},x_{3},x_{4},\dots,x_{k},x_{1},x_{2}. There are 2(k1)2(k-1) ways to color s1s_{1} and s2s_{2}, followed by (k1)(k2)2(k-1)(k-2)\cdots 2 ways to color x3,x4,,xkx_{3},x_{4},\dots,x_{k}, which leaves exactly one remaining color in CB1C\cup B_{1}, say qq. We may then assign to (x1,x2)(x_{1},x_{2}) any of the three ordered pairs (θ(s1),q)(\theta(s_{1}),q), (q,θ(s2))(q,\theta(s_{2})), and (θ(s1),θ(s2))(\theta(s_{1}),\theta(s_{2})). Hence, the number of such colorings is 6(k1)(k1)!6(k-1)(k-1)!.

Case 3: θ(s1)θ(s2)\theta(s_{1})\neq\theta(s_{2}) and |{θ(s1),θ(s2)}C|=0|\{\theta(s_{1}),\theta(s_{2})\}\cap C|=0.

For this case, we count the colorings in the order s1,s2,x2,x3,,xk,x1s_{1},s_{2},x_{2},x_{3},\dots,x_{k},x_{1}. There are 2×22\times 2 ways to color s1s_{1} and s2s_{2}, followed by k(k1)2k(k-1)\cdots 2 ways to color x2,x3,,xkx_{2},x_{3},\dots,x_{k}, which leaves exactly one remaining color in CB1C\cup B_{1}, say qq. We may then assign the color qq or θ(s1)\theta(s_{1}) to x1x_{1}. Hence, the number of such colorings is 8k!8k!.

By Lemma 10 and the three cases above, we have 3(k1)(k2)(k1)!(k1)4t+(4(k1)k!+6(k1)(k1)!)(k(k1))2t+8k!((k1)k2(k+1))t3(k-1)(k-2)(k-1)!(k-1)^{4t}+\big(4(k-1)k!+6(k-1)(k-1)!\big)\bigl(k(k-1)\bigr)^{2t}+8k!\bigl((k-1)k^{2}(k+1)\bigr)^{t} LL-colorings θ\theta^{\prime} of Gk,tG_{k,t} such that θ(s1)θ(s2)\theta^{\prime}(s_{1})\neq\theta^{\prime}(s_{2}).

Hence the result is proven. ∎

Now we are able to show that P(Gk,t,k+1)<P(Gk,t,k+1)P_{\ell}(G_{k,t},k+1)<P(G_{k,t},k+1) when tt is sufficiently large.

Proposition 12.

For any integer k3k\geq 3, there is an integer t0=t0(k)t_{0}=t_{0}(k) such that P(Gk,t,k+1)<P(Gk,t,k+1)P_{\ell}(G_{k,t},k+1)<P(G_{k,t},k+1) holds for all tt0t\geq t_{0}.

Proof.

Let LL be the (k+1)(k+1)-list assignment of Gk,tG_{k,t} defined in (2.3). Then by Lemmas 9 and 11,

P(Gk,t,k+1)P(Gk,t,k+1)\displaystyle P(G_{k,t},k+1)-P_{\ell}(G_{k,t},k+1)
\displaystyle\geq P(Gk,t,k+1)P(Gk,t,L)\displaystyle P(G_{k,t},k+1)-P(G_{k,t},L)
\displaystyle\geq (k+1)!k4t(k1)k!k4t3(k2)(k1)!(k1)4t+1(4(k1)k!CLOSE\displaystyle(k+1)!k^{4t}-(k-1)k!k^{4t}-3(k-2)(k-1)!(k-1)^{4t+1}-\big(4(k-1)k!
OPEN+6(k1)(k1)!)(k(k1))2t8k!((k1)k2(k+1))t.\displaystyle+6(k-1)(k-1)!\big)\bigl(k(k-1)\bigr)^{2t}-8k!\bigl((k-1)k^{2}(k+1)\bigr)^{t}. (2.15)

Then we need only to show that for all sufficiently large integers tt,

(k+1)!k4t>\displaystyle(k+1)!k^{4t}> (k1)k!k4t3(k2)(k1)!(k1)4t+1+(4(k1)k!CLOSE\displaystyle(k-1)k!k^{4t}-3(k-2)(k-1)!(k-1)^{4t+1}+(4(k-1)k!
OPEN+6(k1)(k1)!)(k(k1))2t+8k!((k1)k2(k+1))t.\displaystyle+6(k-1)(k-1)!)\bigl(k(k-1)\bigr)^{2t}+8k!\bigl((k-1)k^{2}(k+1)\bigr)^{t}. (2.16)

Let r=11/k,s=11/k2,r=1-1/k,s=1-1/k^{2}, and

Fk(t)=3(k1)(k2)2kr4t+(2k+3)(k1)kr2t+4st.F_{k}(t)=\frac{3(k-1)(k-2)}{2k}r^{4t}+\frac{(2k+3)(k-1)}{k}r^{2t}+4s^{t}. (2.17)

Then it can be easily verified that (2) holds if and only if Fk(t)<1F_{k}(t)<1. Let t0t_{0} be the integer defined below:

t0=k2log(7k+12).t_{0}=\left\lceil k^{2}\log\left(\frac{7k+1}{2}\right)\right\rceil.

In the following, we shall prove that Fk(t)<1F_{k}(t)<1 whenever tt0t\geq t_{0}.

Since k3k\geq 3, we have

r2=(11k)2<11k2=s,r^{2}=\left(1-\frac{1}{k}\right)^{2}<1-\frac{1}{k^{2}}=s,

implying that r4t<r2t<str^{4t}<r^{2t}<s^{t}. Then for any integer tt0t\geq t_{0},

Fk(t)<\displaystyle F_{k}(t)< (3(k1)(k2)2k+(2k+3)(k1)k+4)st\displaystyle\left(\frac{3(k-1)(k-2)}{2k}+\frac{(2k+3)(k-1)}{k}+4\right)s^{t}
=\displaystyle= 7k+12(11k2)t\displaystyle\frac{7k+1}{2}(1-\frac{1}{k^{2}})^{t}
<\displaystyle< 7k+12exp(t/k2)\displaystyle\frac{7k+1}{2}\exp(-t/k^{2})
\displaystyle\leq 7k+12exp(log(7k+12))\displaystyle\frac{7k+1}{2}\exp\left(-\log\left(\frac{7k+1}{2}\right)\right)
=1.\displaystyle=1. (2.18)

Hence the result holds. ∎

By Lemma 8 and Proposition 12, Theorem 6 is proven.

3 Proof of Theorem 7

In this section, we prove Theorem 7 after introducing an elementary property of the list-color function.

Let rr\in\mathbb{N}. For any pair of vertex-disjoint graphs GG and HH that each contains a copy of KrK_{r}, let GrHG\cup_{r}H denote a graph obtained by identifying a copy of KrK_{r} in GG with a copy of KrK_{r} in HH. It is well known that for any r,kr,k\in\mathbb{N},

P(GrH,k)=P(G,k)P(H,k)(k)r.\displaystyle P(G\cup_{r}H,k)=\frac{P(G,k)P(H,k)}{(k)_{r}}. (3.1)

However, for the list-color function, the analogous identity does not hold in general. For example, P(C4,2)=2P_{\ell}(C_{4},2)=2 while P(C41C4,2)=0P_{\ell}(C_{4}\cup_{1}C_{4},2)=0.

On the other hand, for the DP-color function, it is known that

PDP(GrH,k)PDP(G,k)PDP(H,k)(k)rP_{DP}(G\cup_{r}H,k)\leq\frac{P_{DP}(G,k)P_{DP}(H,k)}{(k)_{r}}

holds for r=1,2r=1,2 and all kk\in\mathbb{N}, whereas a counterexample exists for r=3r=3 and k=4k=4 [4, 16]. This naturally raises the question of whether for the list-color function,

P(GrH,k)P(G,k)P(H,k)(k)r\displaystyle P_{\ell}(G\cup_{r}H,k)\leq\frac{P_{\ell}(G,k)P_{\ell}(H,k)}{(k)_{r}} (3.2)

holds for all r,kr,k\in\mathbb{N}.

In fact, (3.2) is not true even for r=2r=2. To see this, consider the graph K2,122K3K_{2,12}\cup_{2}K_{3}. It is clear that for any 3-list assignment LL of K2,122K3K_{2,12}\cup_{2}K_{3}, an L|V(K2,12)L|_{V(K_{2,12})}-coloring of K2,12K_{2,12} naturally extends to an LL-coloring of K2,122K3K_{2,12}\cup_{2}K_{3}, implying that P(K2,122K3,3)P(K2,12,3)P_{\ell}(K_{2,12}\cup_{2}K_{3},3)\geq P_{\ell}(K_{2,12},3). Moreover, results in [14] show that every 33-list assignment LL of K2,12K_{2,12} satisfying P(K2,12,L)=P(K2,12,3)P(K_{2,12},L)=P_{\ell}(K_{2,12},3) has the property that L(u)L(v)L(u)\neq L(v) for every uvE(K2,12)uv\in E(K_{2,12}). Then for any 3-list assignment LL of K2,122K3K_{2,12}\cup_{2}K_{3} such that P(K2,12,L|V(K2,12))=P(K2,12,3)P(K_{2,12},L|_{V(K_{2,12})})=P_{\ell}(K_{2,12},3), an L|V(K2,12)L|_{V(K_{2,12})}-coloring of K2,12K_{2,12} does not extend uniquely to LL-colorings of K2,122K3K_{2,12}\cup_{2}K_{3}, implying that P(K2,122K3,3)>P(K2,12,3)P_{\ell}(K_{2,12}\cup_{2}K_{3},3)>P_{\ell}(K_{2,12},3). Therefore

P(K2,122K3,3)>P(K2,12,3)=P(K2,12,3)P(K3,3)32,P_{\ell}(K_{2,12}\cup_{2}K_{3},3)>P_{\ell}(K_{2,12},3)=\frac{P_{\ell}(K_{2,12},3)P_{\ell}(K_{3},3)}{3\cdot 2},

which provides a counterexample to (3.2) when r=2r=2.

Thus, we restrict our attention to the case r=1r=1, as follows.

Lemma 13.

For any kk\in\mathbb{N} and any pair of vertex-disjoint graphs GG and HH,

P(G1H,k)P(G,k)P(H,k)k.\displaystyle P_{\ell}(G\cup_{1}H,k)\leq\frac{P_{\ell}(G,k)P_{\ell}(H,k)}{k}. (3.3)
Proof.

Suppose G1HG\cup_{1}H is obtained by identifying uu in V(G)V(G) and vv in V(H)V(H).

Let LGL_{G} and LHL_{H} be a kk-list assignment of GG and HH, respectively, such that

P(G,LG)=P(G,k)andP(H,LH)=P(H,k).P(G,L_{G})=P_{\ell}(G,k)\qquad\text{and}\qquad P(H,L_{H})=P_{\ell}(H,k).

Moreover, we assume without loss of generality that

LG(u)={1,,k}andLH(v)={1,,k}.L_{G}(u)=\{1,\dots,k\}\qquad\text{and}\qquad L_{H}(v)=\{1,\dots,k\}.

In the following, we shall construct a kk-list assignment LL of G1HG\cup_{1}H such that

P(G1H,L)P(G,LG)P(H,LH)k=P(G,k)P(H,k)k.\displaystyle P(G\cup_{1}H,L)\leq\frac{P(G,L_{G})P(H,L_{H})}{k}=\frac{P_{\ell}(G,k)P_{\ell}(H,k)}{k}. (3.4)

For i=1,2,,ki=1,2,\dots,k, let gig_{i} be the number of LGL_{G}-colorings θ\theta such that θ(u)=i\theta(u)=i, and let hih_{i} be the number of LHL_{H}-colorings φ\varphi such that φ(v)=i\varphi(v)=i. Then

i=1kgi=P(G,LG)andj=1khj=P(H,LH).\sum_{i=1}^{k}g_{i}=P(G,L_{G})\qquad\text{and}\qquad\sum_{j=1}^{k}h_{j}=P(H,L_{H}).

For each permutation σSk\sigma\in S_{k}, set Sσ=i=1kgihσ(i).S_{\sigma}=\sum_{i=1}^{k}g_{i}h_{\sigma(i)}. Then

1k!σSkSσ\displaystyle\frac{1}{k!}\sum_{\sigma\in S_{k}}S_{\sigma} =1k!σSk(i=1kgihσ(i))\displaystyle=\ \frac{1}{k!}\sum_{\sigma\in S_{k}}\left(\sum_{i=1}^{k}g_{i}h_{\sigma(i)}\right)
=i=1kgi(1k!σSkhσ(i))\displaystyle=\sum_{i=1}^{k}g_{i}\left(\frac{1}{k!}\sum_{\sigma\in S_{k}}h_{\sigma(i)}\right)
=i=1kgi((k1)!k!j=1khj)\displaystyle=\sum_{i=1}^{k}g_{i}\left(\frac{(k-1)!}{k!}\sum_{j=1}^{k}h_{j}\right)
=1k(i=1kgi)(j=1khj)\displaystyle=\frac{1}{k}\left(\sum_{i=1}^{k}g_{i}\right)\left(\sum_{j=1}^{k}h_{j}\right)
=P(G,LG)P(H,LH)k.\displaystyle=\frac{P(G,L_{G})P(H,L_{H})}{k}. (3.5)

Since |Sk|=k!|S_{k}|=k!, (3) indicates a permutation σ0Sk\sigma_{0}\in S_{k} such that

Sσ0P(G,LG)P(H,LH)k=P(G,k)P(H,k)k.\displaystyle S_{\sigma_{0}}\leq\frac{P(G,L_{G})P(H,L_{H})}{k}=\frac{P_{\ell}(G,k)P_{\ell}(H,k)}{k}. (3.6)

Let ff be the mapping from \mathbb{N} to \mathbb{N} such that f(σ0(i))=if(\sigma_{0}(i))=i for all i[k]i\in[k] and f(i)=if(i)=i for ik+1i\geq k+1.

Now we shall construct a special kk-list assignment LL of G1HG\cup_{1}H, where

{L(x)=LG(x)for allxV(G),L(y)={f(i):iLH(y)}for allyV(H){v}.\left\{\begin{aligned} &L(x)=L_{G}(x)~\text{for all}~x\in V(G),\\ &L(y)=\{f(i):i\in L_{H}(y)\}~\text{for all}~y\in V(H)\setminus\{v\}.\end{aligned}\right.

Then it is clear by (3.6) that

P(G1H,k)P(G1H,L)=i=1kgihσ0(i)=Sσ0P(G,k)P(H,k)k.\displaystyle P_{\ell}(G\cup_{1}H,k)\leq P(G\cup_{1}H,L)=\sum_{i=1}^{k}g_{i}h_{\sigma_{0}(i)}=S_{\sigma_{0}}\leq\frac{P_{\ell}(G,k)P_{\ell}(H,k)}{k}. (3.7)

Now we give the proof of Theorem 7.

Proof of Theorem 7. For any graph GG with P(G,k)=P(G,k)>0P(G,k)=P_{\ell}(G,k)>0, let Hk,tH_{k,t} be a graph obtained by identifying one vertex of GG and xkx_{k} of Gk,tG_{k,t}. We shall show that P(Hk,t,k)=P(Hk,t,k)>0P(H_{k,t},k)=P_{\ell}(H_{k,t},k)>0 and P(Hk,t,k+1)>P(Hk,t,k+1)P(H_{k,t},k+1)>P_{\ell}(H_{k,t},k+1) whenever tt is sufficiently large.

Recall that P(Gk,t,k+1)>P(Gk,t,k+1)P(G_{k,t},k+1)>P_{\ell}(G_{k,t},k+1) whenever tt is sufficiently large. Then by (3.1) and (3.3),

P(Hk,t,k+1)\displaystyle P_{\ell}(H_{k,t},k+1) P(G,k+1)P(Gk,t,k+1)k+1\displaystyle\leq\frac{P_{\ell}(G,k+1)P_{\ell}(G_{k,t},k+1)}{k+1}
<P(G,k+1)P(Gk,t,k+1)k+1\displaystyle<\frac{P(G,k+1)P(G_{k,t},k+1)}{k+1}
=P(Hk,t,k+1).\displaystyle=P(H_{k,t},k+1). (3.8)

Thus it remains to show that P(Hk,t,k)=P(Hk,t,k)>0P(H_{k,t},k)=P_{\ell}(H_{k,t},k)>0.

By Lemma 8 and (3.1), we have

P(Hk,t,k)=P(G,k)P(Gk,t,k)k=(k1)!(k2)4tP(G,k)>0.\displaystyle P(H_{k,t},k)=\frac{P(G,k)P(G_{k,t},k)}{k}=(k-1)!(k-2)^{4t}P(G,k)>0. (3.9)

We next consider P(Hk,t,k)P_{\ell}(H_{k,t},k). Let LL be a kk-list assignment of Hk,tH_{k,t} such that P(Hk,t,k)=P(Hk,t,L)P_{\ell}(H_{k,t},k)=P(H_{k,t},L). Then there are at least P(G,k)P_{\ell}(G,k) L|V(G)L|_{V(G)}-colorings of GG. Moreover, each of them can be extended to at least (k1)!(k2)4t(k-1)!(k-2)^{4t} LL-colorings of Hk,tH_{k,t} by considering in the order x1,,xk1,s1,s2,y1,1,,y4,tx_{1},\ldots,x_{k-1},s_{1},s_{2},y_{1,1},\dots,y_{4,t}. Thus

P(Hk,t,k)=P(Hk,t,L)(k1)!(k2)4tP(G,k)=(k1)!(k2)4tP(G,k).\displaystyle P_{\ell}(H_{k,t},k)=P(H_{k,t},L)\geq(k-1)!(k-2)^{4t}P_{\ell}(G,k)=(k-1)!(k-2)^{4t}P(G,k). (3.10)

Hence P(Hk,t,k)=P(Hk,t,k)>0P_{\ell}(H_{k,t},k)=P(H_{k,t},k)>0 follows from (3.9) and (3.10). ∎

4 Concluding Remarks

Although all counterexamples constructed in the proofs of Theorems 6 and 7 for each integer k3k\geq 3 contain copies of KkK_{k}, we shall show in this section that the presence of KkK_{k} is not essential.

s1s_{1}^{\prime}x2x_{2}^{\prime}x1x_{1}^{\prime}x3x_{3}^{\prime}s2s_{2}^{\prime}
Figure 2: The graph QQ.

Let QQ be the graph as shown in Figure 2. QQ was constructed in [2] as an example that is uniquely 3-colorable but contains no K3K_{3}. Therefore, P(Q,3)=3!=6P(Q,3)=3!=6. Moreover, P(Q,3)P_{\ell}(Q,3) can be determined by the following theorem.

Theorem 14 ([1]).

Let GG be a graph on a set V={v1,v2,,vn}V=\{v_{1},v_{2},\dots,v_{n}\} of n1n\geq 1 vertices. For 1in1\leq i\leq n, let LviL_{v_{i}} be a list of di+1d_{i}+1 colors where di0d_{i}\geq 0 is a given integer. Suppose that GG is uniquely LL-colorable and d1+d2++dn=md_{1}+d_{2}+\cdots+d_{n}=m, where mm is the size of GG. Then GG has ff-colorings for every list assignment ff of GG provided f(vi)=di+1f(v_{i})=d_{i}+1 for 1in1\leq i\leq n.

Proposition 15.

P(Q,3)=P(Q,3)=6P_{\ell}(Q,3)=P(Q,3)=6.

Proof.

Since P(Q,3)P(Q,3)=6P_{\ell}(Q,3)\leq P(Q,3)=6, we need only to show that QQ has at least 6 LL-colorings for all 3-list assignments LL of QQ.

Assume that u,vu,v are two vertices from different partition sets in the unique partition of V(Q)V(Q) into three independent sets. Let L0L_{0} be a list assignment of GG such that L0(u)={1},L0(v)={1,2}L_{0}(u)=\{1\},L_{0}(v)=\{1,2\}, and L0(w)={1,2,3}L_{0}(w)=\{1,2,3\} for all wV(Q){u,v}w\in V(Q)\setminus\{u,v\}. Then QQ is uniquely L0L_{0}-colorable as QQ is uniquely 3-colorable. Also, note that |V(Q)|=24|V(Q)|=24 and |E(Q)|=45|E(Q)|=45. Then L0L_{0} satisfies the requirements of Theorem 14 as 0+1+2×22=450+1+2\times 22=45, which implies that GG has LL-colorings for every list assignment LL of GG provided |L(u)|=1|L(u)|=1, |L(v)|=2|L(v)|=2, and |L(w)|=3|L(w)|=3 for all wV(Q){u,v}w\in V(Q)\setminus\{u,v\}.

Now we consider any 3-list assignment LL of QQ. Assume that L(u)={1,2,3}L(u)=\{1,2,3\} and L(v)={a,b,c}L(v)=\{a,b,c\}. For each i[3]i\in[3], let Li(u)={i},Li(v)={a,b}L_{i}(u)=\{i\},L_{i}(v)=\{a,b\}, and Li(w)=L(w)L_{i}(w)=L(w) for all wV(Q){u,v}w\in V(Q)\setminus\{u,v\}. Then by Theorem 14, QQ has an LiL_{i}-coloring θi\theta_{i}, which is also an LL-coloring. Further, let Li(u)={i},Li(v)=L(v){θi(v)}L_{i}^{\prime}(u)=\{i\},L_{i}^{\prime}(v)=L(v)\setminus\{\theta_{i}(v)\}, and Li(w)=L(w)L_{i}^{\prime}(w)=L(w) for all wV(Q){u,v}w\in V(Q)\setminus\{u,v\}. We again obtain by Theorem 14 an LiL_{i}^{\prime}-coloring θi\theta_{i}^{\prime} for each i[3]i\in[3], which is also an LL-coloring and is different from θi\theta_{i} as θi(v)θi(v)\theta_{i}(v)\neq\theta_{i}^{\prime}(v). Since θi(u)=θi(u)\theta_{i}(u)=\theta_{i}^{\prime}(u) are pairwise distinct for all i[3]i\in[3], we now have six LL-colorings of QQ, which completes the proof. ∎

For any pair of vertex-disjoint graphs GG and HH, let GHG\vee H be the join of GG and HH, i.e., V(GH)=V(G)V(H)V(G\vee H)=V(G)\cup V(H) and E(GH)=E(G)E(H){uv:uV(G),vV(H)}E(G\vee H)=E(G)\cup E(H)\cup\{uv:u\in V(G),v\in V(H)\}. Since QQ contains no K3K_{3}, QKk3Q\vee K_{k-3} contains no KkK_{k}. We shall show that QKk3Q\vee K_{k-3} is an ideal replacement for Gk,t[XkS]G_{k,t}[X_{k}\cup S] after proving the next proposition.

Lemma 16 ([17]).

For any graph GG and n,kNn,k\in N,

P(G,kn)P(Kn,k)P(GKn,k)P(GKn,k)=P(G,kn)P(Kn,k).P_{\ell}(G,k-n)P(K_{n},k)\leq P_{\ell}(G\vee K_{n},k)\leq P(G\vee K_{n},k)=P(G,k-n)P(K_{n},k).
Proposition 17.

For any k3k\geq 3, P(QKk3,k)=P(QKk3,k)=k!P_{\ell}(Q\vee K_{k-3},k)=P(Q\vee K_{k-3},k)=k!.

Proof.

By Lemma 16, we have

P(Q,3)P(Kk3,k)P(QKk3,k)P(QKk3,k)=P(Q,3)P(Kk3,k).P_{\ell}(Q,3)P(K_{k-3},k)\leq P_{\ell}(Q\vee K_{k-3},k)\leq P(Q\vee K_{k-3},k)=P(Q,3)P(K_{k-3},k).

Then by Proposition 15, the first equality holds.

The second equality is then obvious as QKk3Q\vee K_{k-3} is uniquely kk-colorable. ∎

Observe that Gk,t[XkS]G_{k,t}[X_{k}\cup S] is uniquely kk-colorable, where x1,x2,,xkx_{1},x_{2},\dots,x_{k} are in different color classes while xix_{i} and sis_{i} are in the same color class for i[2]i\in[2], and s1,s2s_{1},s_{2} are nonadjacent. We shall see that a similar structure also exists in QKk3Q\vee K_{k-3}.

Let x1,x2,x3,s1,s2x_{1}^{\prime},x_{2}^{\prime},x_{3}^{\prime},s_{1}^{\prime},s_{2}^{\prime} be the vertices in QQ as shown in Figure 2, and let V(Kk3)={x4,x5,,xk}V(K_{k-3})=\{x_{4}^{\prime},x_{5}^{\prime},\dots,x_{k}^{\prime}\}. Then QKk3Q\vee K_{k-3} is uniquely kk-colorable, where x1,x2,,xkx_{1}^{\prime},x_{2}^{\prime},\dots,x_{k}^{\prime} are in different color classes while xix_{i}^{\prime} and sis_{i}^{\prime} are in the same color class for i[2]i\in[2], and s1,s2s_{1}^{\prime},s_{2}^{\prime} are nonadjacent. Thus replacing Gk,t[XS]G_{k,t}[X\cup S] and x1,x2,,xk,s1,s2x_{1},x_{2},\dots,x_{k},s_{1},s_{2} in Gk,tG_{k,t} with QKk3Q\vee K_{k-3} and x1,x2,x3,,xk,s1,s2x_{1}^{\prime},x_{2}^{\prime},x_{3}^{\prime},\dots,x_{k}^{\prime},s_{1}^{\prime},s_{2}^{\prime} provides us another infinite family of graphs supporting Theorem 6 that contains no KkK_{k}. The proof can be carried out based on Proposition 17 by establishing a series of results analogous to Lemmas 8, 9, 10, and 11, which together yield Proposition 12.

However, it remains unclear for which graphs the agreement of the chromatic polynomial and the list-color function always persists.

Question 18.

Characterize the graphs GG such that for all integers kk, P(G,k)=P(G,k)>0P(G,k)=P_{\ell}(G,k)>0 implies P(G,k+1)=P(G,k+1).P(G,k+1)=P_{\ell}(G,k+1).

References

  • [1] S. Akbari, V.S. Mirrokni, B.S. Sadjad, A relation between choosability and uniquely list colorability, J. Combinatorial Theory Ser. B 96 (2006), 577–583.
  • [2] S. Akbari, V.S. Mirrokni, B.S. Sadjad, KrK_{r}-free uniquely vertex colorable graphs with minimum porrible edges, J. Combinatorial Theory Ser. B 82 (2001), 316–318.
  • [3] S. Allred and J.A. Mudrock, Enumerative chromatic choosability, https://arxiv.org/abs/2505.05662.
  • [4] J. Becker, J. Hewitt, H. Kaul, M. Maxfield, J.A. Mudrock, D. Spivey, S. Thomason and T. Wagstrom, The DP color function of joins and vertex-gluings of graphs, Discrete Math. 345 (2022), 113093.
  • [5] G.D. Birkhoff, A determinant formula for the number of ways of coloring a map, Annal. Math. 14 (1912), 42–46.
  • [6] M.V. Bui, H. Kaul, M. Maxfield, J.A. Mudrock, P. Shin and S. Thomason, Non-chromatic-adherence of the DP color function via generalized theta graphs, Graphs Combin. 39(42), 2023.
  • [7] Y. Chi, S. Lee, F. Morrissette, J.A. Mudrock, G. Nguyen and B. Whatley, Enumeratively Chromatic-Choosable Theta Graphs, https://arxiv.org/abs/2605.10861.
  • [8] S.L. Dahlberg, H. Kaul and J.A. Mudrock, An algebraic approach for counting DP-3-colorings of sparse graphs, Eur. J. Comb. 118 (2024), 103890.
  • [9] F.M. Dong and K.M. Koh, Foundations of the chromatic polynomial, in the Handbook on the Tutte Polynomial and Related Topics, Jo Ellis-Monaghan and Iain Moffatt (ed.), pp 232–266, CRC press, 2022.
  • [10] F.M. Dong, K.M. Koh and K.L. Teo, Chromatic Polynomials and Chromaticity of Graphs, World Scientific, Singapore, 2005.
  • [11] F.M. Dong and M.Q. Zhang, An improved lower bound of P(G,L)P(G,k)P(G,L)-P(G,k) for kk-assignments LL, J. Combinatorial Theory Ser. B 161 (2023), 109–119.
  • [12] Q. Donner, On the number of list-colorings, J. Graph Theory 16 (1992) 239–245.
  • [13] P. Erdős, A.L. Rubin and H. Taylor, Choosability in graphs, Congr. Numer. 26 (1979), 125–127.
  • [14] H. Kaul, A. Kumar, A. Liu, J.A. Mudrock, P. Rewers, P. Shin, M.S. Tanahara and K. To, Bounding the list color function threshold from above, Involve, a journal of mathematics 16(5) (2023), 849–882.
  • [15] H. Kaul, A. Kumar, J.A. Mudrock, P. Rewers, P. Shin and K. To, On the list color function threshold, J. Graph Theory 105 (2024), 386–397.
  • [16] H. Kaul, M. Maxfield, J.A. Mudrock and S. Thomason, The DP color function of clique-gluings of graphs, Enumerative Combinatorics and Applications 4:2 (2024), Article #S2R11.
  • [17] H. Kaul and J.A. Mudrock, Criticality, the list color function, and list coloring the Cartesian product of graphs, Journal of Combinatorics 12 (2021), 479–514.
  • [18] H. Kaul and J.A. Mudrock, On the chromatic polynomial and counting DP-colorings of graphs, Advances in Applied Math. 123 (2021), article 103121.
  • [19] R. Kirov and R. Naimi, List coloring and nn-monophilic graphs, Ars Comb. 124 (2016), 329–340.
  • [20] A.V. Kostochka and A.F. Sidorenko, Problem Session of the Prachatice Conference on Graph Theory. In Fourth Czechoslovak Symposium on Combinatorics, Graphs and Complexity, Ann. Discrete Math, vol. 51, p. 380, 1992.
  • [21] R.C. Read and W.T. Tutte, Chromatic polynomials, in Selected Topics in Graph Theory 3, Academic Press (1988), 15–42.
  • [22] G. Royle, Recent results on chromatic and flow roots of graphs and matroids, in: Surveys in combinatorics, London Math. Soc. Lecture Note Ser., vol. 365, Cambridge Univ. Press, Cambridge, 2009, pp. 289–327.
  • [23] C. Thomassen, The chromatic polynomial and list colorings, Journal of Combinatorial Theory Series B 99 (2009), 474–479.
  • [24] V.G. Vizing, Coloring the vertices of a graph in prescribed colors, Diskret. Analiz. no. 29, Metody Diskret. Anal. v Teorii Kodovi Skhem 101 (1976), 3–10.
  • [25] W. Wang, J. Qian and Z. Yan, When does the list-coloring function of a graph equal its chromatic polynomial, J. Comb. Theory, Ser. B 122 (2017) 543–549.