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

Structural properties of Toeplitz graphs

Seyed Ahmad Mojallal [email protected] Ji-Hwan Jung [email protected] Gi-Sang Cheon [email protected] Suh-Ryung Kim [email protected] Bumtle Kang [email protected] Department of Mathematics and Statistics, University of Regina, Regina, Saskatchewan, S4S0A2, Canada Applied Algebra and Optimization Research Center, Department of Mathematics, Sungkyunkwan University, Suwon 1641916419, Republic of Korea Department of Mathematics Education, Seoul National University, Seoul 0882608826, Republic of Korea Center for Educational Research, Seoul National University, Seoul 0882608826, Republic of Korea
Abstract

In this paper, we study structural properties of Toeplitz graphs. We characterize KqK_{q}-free Toeplitz graphs for an integer q3q\geq 3 and give equivalent conditions for a Toeplitz graph Gnt1,t2,,tkG_{n}\langle t_{1},t_{2},\ldots,t_{k}\rangle with t1<<tkt_{1}<\cdots<t_{k} and ntk1+tkn\geq t_{k-1}+t_{k} being chordal and equivalent conditions for a Toeplitz graph Gnt1,t2G_{n}\langle t_{1},t_{2}\rangle being perfect. Then we compute the edge clique cover number and the vertex clique cover number of a chordal Toeplitz graph. Finally, we characterize the degree sequence (d1,d2,,dn)(d_{1},d_{2},\ldots,d_{n}) of a Toeplitz graph with nn vertices and show that a Toeplitz graph is a regular graph if and only if it is a circulant graph.

keywords:
Toeplitz graphs , Chordal Toeplitz graphs , Perfect Toeplitz graphs , Regular Toeplitz graphs , Circulant graphs
MSC:
05C70 , 05C69
journal: Discrete Mathematics

1 Introduction

An n×nn\times n matrix T=(tij)1i,jnT=(t_{ij})_{1\leq i,j\leq n} is called a Toeplitz matrix if ti,j=ti+1,j+1t_{i,j}=t_{i+1,j+1} for each {i,j}[n1]\{i,j\}\subset[n-1] where [m][m] denotes the set {1,2,,m}\{1,2,\ldots,m\} for a positive integer mm. Toeplitz matrices are precisely those matrices that are constant along all diagonals parallel to the main diagonal, and thus a Toeplitz matrix is determined by its first row and column. Toeplitz matrices occur in a large variety of areas in pure and applied mathematics. For example, they often appear when differential or integral equations are discretized and arise in physical data-processing applications and in the theories of orthogonal polynomials, stationary processes, and moment problems; see Heinig and Rost [8]. Other references on Toeplitz matrices are Gohberg [7] and lohvidov [13].

A Toeplitz graph is defined to be a simple, undirected graph whose adjacency matrix is a (0,1)(0,1)-symmetric Toeplitz matrix. Any Toeplitz matrix mentioned in this paper has the main diagonal entries 0. One can see that the first row of a symmetric Toeplitz matrix determines a unique Toeplitz graph. In this vein, we denote a Toeplitz graph with nn vertices by Gnt1,t2,,tkG_{n}\langle t_{1},t_{2},\ldots,t_{k}\rangle if the 11’s in the first row of its adjacency matrix are placed at positions 1+t11+t_{1}, 1+t21+t_{2}, …, 1+tk1+t_{k} with 1t1<t2<<tk<n1\leq t_{1}<t_{2}<\ldots<t_{k}<n. In addition, we label the vertices of Gnt1,t2,,tkG_{n}\langle t_{1},t_{2},\ldots,t_{k}\rangle with 1,,n1,\ldots,n so that the iith row of its adjacency matrix corresponds to the vertex labeled ii. See Figure 1 for an illustration.

(0110110110110110110110110)\left(\begin{array}[]{ccccc}0&1&1&0&1\\ 1&0&1&1&0\\ 1&1&0&1&1\\ 0&1&1&0&1\\ 1&0&1&1&0\end{array}\right)
Refer to caption1122334455G51,2,4G_{5}\langle 1,2,4\rangle
Figure 1: A (0,1)(0,1)-symmetric Toeplitz matrix and its Toeplitz graph

For V=V=\mathbb{N} and k<k<\infty, infinite Toeplitz graphs Gt1,t2,,tkG_{\infty}\langle t_{1},t_{2},\ldots,t_{k}\rangle are defined the same way. Both types may be studied as special subgraphs of integer distance graphs [1, 14, 15].

Toeplitz graphs have been introduced by G. Sierksma and first been investigated with respect to hamiltonicity by van Dal et al. [24] (see also Heuberger [9], Malik and Qureshi [19], Malik and Zamfirescu [20] for more recent works). Infinite, bipartite Toeplitz graphs have been fully characterized in terms of bases and circuits by Euler et al. [6] (with results on the finite case presented in Euler [3]). Colouring aspects are especially treated in Heuberger [10], Kemnitz and Marangio [15], Nicoloso and Pietropaoli [23]. Infinite, planar Toeplitz graphs have been fully characterized in Euler [4] providing, in particular, a complete description of the class of 3-colourable such graphs. Finite planar Toeplitz graphs are studied in [5].

A hole is a chordless cycle of length at least 44 as an induced subgraph, while an anti-hole is the complement of a hole. An odd hole (respectively odd anti-hole) is a hole (respectively anti-hole) with an odd number of vertices. A chordal graph is a simple graph without holes. A graph G=(V,E)G=(V,E) is an interval graph if it captures the intersection relation for some set of intervals on the real line. Formally, GG is an interval graph provided that one can assign to each vVv\in V an interval IvI_{v} such that IuIvI_{u}\cap I_{v} is nonempty precisely when uvEuv\in E. Three independent vertices form an asteroidal triple in a graph GG if, for each two, there exists a path containing those two but no neighbor of the third. It is well-known that a graph GG is an interval graph if and only if it is chordal and has no asteroidal triple [17].

A clique is a complete subgraph or a subset of vertices of an undirected graph such that every two distinct vertices in the clique are adjacent. A clique cover of GG is a set of cliques of GG such that every vertex is in at least one of them. The clique cover number is the minimum size of a clique cover, and is denoted by θv(G)\theta_{v}(G). An edge clique cover of GG is a set of cliques of GG, which together contain each edge of GG at least once. The smallest cardinality of any edge clique cover of GG is called the edge clique cover number of GG, and is denoted by θE(G)\theta_{E}(G). Those numbers exist as the vertex set (resp. the edge set) of GG forms a clique cover (resp. an edge clique cover) for GG.

The chromatic number of a graph GG, denoted by χ(G)\chi(G), is the smallest number of colors needed to color the vertices of GG so that no two adjacent vertices share the same color.

The clique cover number of GG equals the chromatic number of its complement G¯\overline{G}, that is,

θv(G)=χ(G¯).\theta_{v}(G)=\chi(\overline{G}). (1)

A perfect graph is a graph GG such that for every induced subgraph HH of GG, the clique number equals the chromatic number, i.e., ω(H)=χ(H)\omega(H)=\chi(H). A graph for which ω(G)=χ(G)\omega(G)=\chi(G) (without any requirement that this condition also hold on induced subgraphs) is called a weakly perfect graph. All perfect graphs are therefore weakly perfect by definition.

A circulant matrix CnC_{n} is an n×nn\times n Toeplitz matrix in the following form:

Cn=[c0cn1c2c1c1c0cn1c2c1c0cn2cn1cn1cn2c1c0].C_{n}=\left[\begin{array}[]{ccccc}c_{0}&c_{n-1}&\ldots&c_{2}&c_{1}\\ c_{1}&c_{0}&c_{n-1}&&c_{2}\\ \vdots&c_{1}&c_{0}&\ddots&\vdots\\ c_{n-2}&&\ddots&\ddots&c_{n-1}\\ c_{n-1}&c_{n-2}&\ldots&c_{1}&c_{0}\\ \end{array}\right].

A graph is said to be a circulant graph if it is isomorphic to a Toeplitz graph whose adjacency matrix is a (0,1)(0,1)-symmetric circulant matrix CnC_{n} where

ci=cni{0,1}c_{i}=c_{n-i}\in\{0,1\} (2)

for each i{1,,n/2}i\in\{1,\ldots,\left\lfloor n/2\right\rfloor\} and c0=0c_{0}=0. Circulant graphs are well-studied (see [12, 16, 21, 22, 25] for references). The family of circulant graphs is an important subclass of Toeplitz graphs. Circulant graphs have various applications in the design of interconnection networks in parallel and distributed computing.

Refer to captionToeplitz graphsregular graphscirculant graphsperfect graphschordal graphsinterval graphsABC
Figure 2: Where do Toeplitz graphs stand?

The paper is organized as follows. In Section 2, we characterize KqK_{q}-free Toeplitz graphs for an integer q3q\geq 3, where KqK_{q} denotes the complete graph with qq vertices, and then we compute the edge clique cover number and the vertex clique cover number of a Toeplitz graph. In section 3, we study holes in Toeplitz graphs and give a condition for a Toeplitz graph not having holes, which leads to a characterization of chordal Toeplitz graphs. Then we give equivalent conditions for a Toeplitz graph G=Gnt1,t2G=G_{n}\langle t_{1},t_{2}\rangle being perfect. In Section 4, we characterize the degree sequence (d1,d2,,dn)(d_{1},d_{2},\ldots,d_{n}) of a Toeplitz graph with nn vertices and show that a Toeplitz graph is a regular graph if and only if it is a circulant graph. In Section 5, we propose open problems.

Our results are summarized in Figure 2. The grey region AA represents the set of Toeplitz graphs G=Gnt1,t2,,tkG=G_{n}\langle t_{1},t_{2},\ldots,t_{k}\rangle with t1<<tkt_{1}<\cdots<t_{k} and ntk1+tkn\geq t_{k-1}+t_{k} being chordal for a fixed positive integer nn (Theorem 3.3); the region BB represents the set of odd-hole-free Toeplitz graphs G=Gnt1,t2G=G_{n}\langle t_{1},t_{2}\rangle (Theorem 3.7); the region CC represents the set of circulant graphs (Theorem 4.6).

2 Cliques in Toeplitz graphs

In this section, we give an upper bound for the clique number of Toeplitz graphs and characterize KqK_{q}-free Toeplitz graphs for an integer q3q\geq 3, and then we compute the edge clique cover number and the vertex clique cover number of a Toeplitz graph Gnt,2t,,ktG_{n}\langle t,2t,\ldots,kt\rangle. The following two results immediately follow from the definition of Toeplitz graph.

Proposition 2.1.

For a positive integer kk, let G=Gnt1,t2,,tkG=G_{n}\langle t_{1},t_{2},\ldots,t_{k}\rangle. Then for each i,j[n]i,j\in[n], ii and jj are adjacent if and only if |ij|{t1,,tk}|i-j|\in\{t_{1},\ldots,t_{k}\}.

Lemma 2.2.

For positive integers tt and kk, let G=Gnt,2t,,ktG=G_{n}\langle t,2t,\ldots,kt\rangle. Then uu and vv are connected in GG if and only if vuv-u is a multiple of tt.

Proposition 2.3.

For positive integers tt and kk, let G=Gnt,2t,,ktG=G_{n}\langle t,2t,\ldots,kt\rangle. Then GG has tt components. In particular, if H1,,HtH_{1},\ldots,H_{t} are the components of GG, then HiH_{i} is isomorphic to the graph G(ni)/t+11,2,,kG_{\lfloor(n-i)/t\rfloor+1}\langle 1,2,\ldots,k\rangle and the vertex set of HiH_{i} is {s[n]si(modt)}\{s\in[n]\mid s\equiv i\pmod{t}\} for each i[t]i\in[t].

Proof.

For each i[t]i\in[t], we let Vi={i+sts=0,1,,(ni)/t}V_{i}=\{i+st\mid s=0,1,\ldots,\left\lfloor(n-i)/t\right\rfloor\}. Then |Vi|=(ni)/t+1|V_{i}|=\lfloor(n-i)/t\rfloor+1 for each i[t]i\in[t]. By Lemma 2.2, Hi:=G[Vi]H_{i}:=G[V_{i}] is a component of GG for each i[t]i\in[t]. Fix i[t]i\in[t] and let ff be a function from V(G(ni)/t+11,2,,k)V(G_{\lfloor(n-i)/t\rfloor+1}\langle 1,2,\ldots,k\rangle) to ViV_{i} defined by f(s+1)=i+stf(s+1)=i+st for each s{0,1,,(ni)/t}s\in\{0,1,\ldots,\lfloor(n-i)/t\rfloor\}. It is easy to see that ff is a bijection. Then uu and vv are adjacent in G(ni)/t+11,2,,kG_{\lfloor(n-i)/t\rfloor+1}\langle 1,2,\ldots,k\rangle if and only if |uv|{1,,k}|u-v|\in\{1,\ldots,k\}, which is equivalent to |(i+(u1)t)(i+(v1)t)|{t,,kt}|(i+(u-1)t)-(i+(v-1)t)|\in\{t,\ldots,kt\}, that is, f(u)f(u) and f(v)f(v) are adjacent in HiH_{i}. ∎

Lemma 2.4.

Let G=Gnt1,t2,,tkG=G_{n}\langle t_{1},t_{2},\ldots,t_{k}\rangle. Then there is a maximum clique of GG that contains the vertex 11.

Proof.

Let S1={i1,i2,,i}S_{1}=\{i_{1},i_{2},\ldots,i_{\ell}\} be a maximum clique in GG and let i1=minS1i_{1}=\min S_{1}. Then, by Proposition 2.1, |iuiv|{t1,,tk}|i_{u}-i_{v}|\in\{t_{1},\ldots,t_{k}\} for each {u,v}[l]\{u,v\}\subset[l]. If i1=1i_{1}=1, then we are done. If i1>1i_{1}>1, then the vertices 1,i2i1+1,i3i1+1,,ii1+11,i_{2}-i_{1}+1,i_{3}-i_{1}+1,\ldots,i_{\ell}-i_{1}+1 form a clique of size \ell in GG by Proposition 2.1. ∎

In the following, we present a condition for a Toeplitz graph being KqK_{q}-free.

Theorem 2.5.

Let G=Gnt1,t2,,tkG=G_{n}\langle t_{1},t_{2},\ldots,t_{k}\rangle. Then GG is KqK_{q}-free if and only if for any subset S[k]S\subseteq[k] with size q1q-1, there is a pair of distinct integers a,bSa,b\in S such that |tatb|{t1,,tk}|t_{a}-t_{b}|\notin\{t_{1},\ldots,t_{k}\}.

Proof.

Let N(1)N(1) be the set of neighbors of 11 in GG. Then N(1)={t1+1,t2+1,,tk+1}N(1)=\{t_{1}+1,t_{2}+1,\ldots,t_{k}+1\}. Now GG is KqK_{q}-free if and only if 11 does not belong to a clique of size qq (by Lemma 2.4), equivalently, any subset of N(1)N(1) with size q1q-1 contains a pair ti+1t_{i}+1, tj+1t_{j}+1 such that |titj|=|(ti+1)(tj+1)|{t1,,tk}|t_{i}-t_{j}|=|(t_{i}+1)-(t_{j}+1)|\notin\{t_{1},\ldots,t_{k}\}.

If any subset S[k]S\subseteq[k] with size q1q-1 contains a pair of distinct integers a,bSa,b\in S such that |tatb|{t1,,tk}|t_{a}-t_{b}|\notin\{t_{1},\ldots,t_{k}\}, then any subset of N(1)N(1) with size q1q-1 contains a pair ti+1t_{i}+1, tj+1t_{j}+1 such that |titj|{t1,,tk}|t_{i}-t_{j}|\notin\{t_{1},\ldots,t_{k}\}. Suppose that any subset of N(1)N(1) with size q1q-1 contains a pair ti+1t_{i}+1, tj+1t_{j}+1 such that |titj|{t1,,tk}|t_{i}-t_{j}|\notin\{t_{1},\ldots,t_{k}\}. Let SS be a subset of [k][k] with size q1q-1. Then {ti+1iS}\{t_{i}+1\mid i\in S\} is a subset of N(1)N(1) with size q1q-1, so there is a pair ti+1t_{i}+1, tj+1t_{j}+1 such that |titj|{t1,,tk}|t_{i}-t_{j}|\notin\{t_{1},\ldots,t_{k}\} by the last equivalence above. Thus there is a pair of distinct integers a,bSa,b\in S such that |tatb|{t1,,tk}|t_{a}-t_{b}|\notin\{t_{1},\ldots,t_{k}\}.∎

Corollary 2.6.

Let G=Gnt1,t2,,tkG=G_{n}\langle t_{1},t_{2},\ldots,t_{k}\rangle. Then GG is triangle-free if and only if |titj|{t1,t2,,tk}|t_{i}-t_{j}|\notin\{t_{1},t_{2},\ldots,t_{k}\} for any pair i,j[k]i,j\in[k].

For a Toeplitz graph G:=Gnt1,,tkG:=G_{n}\langle t_{1},\ldots,t_{k}\rangle, we denote B(G)={t1,,tk}B(G)=\{t_{1},\ldots,t_{k}\}.

Lemma 2.7.

Let G=Gnt1,t2,,tkG=G_{n}\langle t_{1},t_{2},\ldots,t_{k}\rangle. Then |titj|B(G)|t_{i}-t_{j}|\in B(G) for every {ti,tj}B(G)\{t_{i},t_{j}\}\subset B(G) with iji\neq j if and only if ti=it1t_{i}=it_{1} for each i[k]i\in[k].

Proof.

We can easily check the ‘if’ part. To show the ‘only if’ part, suppose that |titj|B(G)|t_{i}-t_{j}|\in B(G) for every {ti,tj}B(G)\{t_{i},t_{j}\}\subset B(G) with iji\neq j. Then t2t1B(G)t_{2}-t_{1}\in B(G). Since t2t1<t2t_{2}-t_{1}<t_{2}, t2t1=t1t_{2}-t_{1}=t_{1}. Therefore t2=2t1t_{2}=2t_{1}. By the supposition again, t3t1B(G)t_{3}-t_{1}\in B(G). Since t3t1<t3t_{3}-t_{1}<t_{3}, t3t1=t2t_{3}-t_{1}=t_{2} or t3t1=t1t_{3}-t_{1}=t_{1}. If t3t1=t1t_{3}-t_{1}=t_{1}, then t3=t2t_{3}=t_{2}, a contradiction. Thus t3t1=t2t_{3}-t_{1}=t_{2} and consequently t3=3t1t_{3}=3t_{1}. By repeating this procedure, we conclude that ti=it1t_{i}=it_{1} for each i[k]i\in[k] and thus B(G)={t1,2t1,,kt1}B(G)=\{t_{1},2t_{1},\ldots,kt_{1}\}. ∎

We denote the degree of the vertex ii by deg(i)\deg(i) in a Toeplitz graph.

Theorem 2.8.

Let G=Gnt1,t2,,tkG=G_{n}\langle t_{1},t_{2},\ldots,t_{k}\rangle. Then ω(G)k+1\omega(G)\leq k+1. Furthermore, the equality holds if and only if ti=it1t_{i}=it_{1} for each i[k]i\in[k].

Proof.

By Lemma 2.4, there is a maximum clique that contains the vertex 11. Then the clique is contained in the closed neighborhood of 11. Since deg(1)=k\deg(1)=k, the inequality holds. Now we show the equality part. If B(G)={t1,2t1,,kt1}B(G)=\{t_{1},2t_{1},\ldots,kt_{1}\}, then {1,1+t1,,1+kt1}\{1,1+t_{1},\ldots,1+kt_{1}\} is a clique of size k+1k+1 in GG and so ω(G)=k+1\omega(G)=k+1.

Suppose that ω(G)=k+1\omega(G)=k+1. Then, by Lemma 2.4, there is a maximum clique of size k+1k+1 containing 11. Therefore {1,1+t1,,1+tk}\{1,1+t_{1},\ldots,1+t_{k}\} is a clique. Now, for each i,j[k]i,j\in[k], 1+ti1+t_{i} and 1+tj1+t_{j} are adjacent since they belong to the same clique. Then |(1+tj)(1+ti)|=|tjti|B(G)|(1+t_{j})-(1+t_{i})|=|t_{j}-t_{i}|\in B(G) by Proposition 2.1, so B(G)={t1,2t1,,kt1}B(G)=\{t_{1},2t_{1},\ldots,kt_{1}\} by Lemma 2.7.∎

Theorem 2.9.

Let G=Gnt,2t,,ktG=G_{n}\langle t,2t,\ldots,kt\rangle. Then θE(G)=max{t,nkt}\theta_{E}(G)=\max\{t,n-kt\}. Moreover, for H:=Gns1,s2,,skH:=G_{n}\langle s_{1},s_{2},\ldots,s_{k}\rangle with s1=ts_{1}=t and sk=kts_{k}=kt, θE(G)θE(H)\theta_{E}(G)\leq\theta_{E}(H).

Proof.

Suppose that nkttn-kt\leq t. Then n(k+1)tn\leq(k+1)t. By Proposition 2.3, GG has tt components and so θE(G)t\theta_{E}(G)\geq t. Again, by Proposition 2.3, each component is isomorphic to G(ni)/t+11,2,,kG_{\lfloor(n-i)/t\rfloor+1}\langle 1,2,\ldots,k\rangle for each i[t]i\in[t]. Since n(k+1)tn\leq(k+1)t, the number of vertices in a component is at most k+1k+1 and so G(ni)/t+11,2,,kG_{\lfloor(n-i)/t\rfloor+1}\langle 1,2,\ldots,k\rangle for each i[t]i\in[t] is a complete graph by Proposition 2.1. Thus θE(G)=t\theta_{E}(G)=t.

Now, suppose that t<nktt<n-kt. Then n>(k+1)tn>(k+1)t. Let Ci={i,i+t,,i+kt}C_{i}=\{i,i+t,\ldots,i+kt\} for each i{1,,nkt}i\in\{1,\ldots,n-kt\}. Then, for each i{1,,nkt}i\in\{1,\ldots,n-kt\}, an element in CiC_{i} is at least i1i\geq 1 and at most i+kt(nkt)+kt=ni+kt\leq(n-kt)+kt=n, so CiC_{i} is a vertex set of GG . In addition, CiC_{i} is a clique for each i{1,,nkt}i\in\{1,\ldots,n-kt\} by definition. Take an edge uvE(G)uv\in E(G) such that u<vu<v. If kt<vkt<v, then Cvkt={vkt,v(k1)t,,v}C_{v-kt}=\{v-kt,v-(k-1)t,\ldots,v\} contains uu and vv. Suppose that vktv\leq kt. Then there exists an integer rr such that rv(modt)r\equiv v\pmod{t} with 1rt1\leq r\leq t. Obviously, rvr\leq v and vrv-r is a multiple of tt. Since vktv\leq kt, vrktv-r\leq kt and so vr{t,2t,,kt}v-r\in\{t,2t,\ldots,kt\}. Thus Cr={r,r+t,,r+kt}C_{r}=\{r,r+t,\ldots,r+kt\} contains vv. Thus {Ci1inkt}\{C_{i}\mid 1\leq i\leq n-kt\} is an edge-clique cover of GG and so θE(G)nkt\theta_{E}(G)\leq n-kt.

Now, let F={ii+kt1inkt}F=\{i\ i+kt\mid 1\leq i\leq n-kt\}. Take edges ii+kti\ i+kt and jj+ktj\ j+kt for some 1i<jnkt1\leq i<j\leq n-kt. Since j+kti>ktj+kt-i>kt, ii and j+ktj+kt are not adjacent by Proposition 2.1 and so ii+kti\ i+kt and jj+ktj\ j+kt do not belong to the same clique. Thus θE(G)|F|=nkt\theta_{E}(G)\geq|F|=n-kt and hence θE(Gn)=nkt\theta_{E}(G_{n})=n-kt.

Let H:=Hns1,s2,,skH:=H_{n}\langle s_{1},s_{2},\ldots,s_{k}\rangle be a Toeplitz graph with s1=ts_{1}=t and sk=kts_{k}=kt. Note that G=HG=H if k2k\leq 2, so θE(G)=θE(H)\theta_{E}(G)=\theta_{E}(H). We assume that k3k\geq 3. Suppose that kt<n(k+1)tkt<n\leq(k+1)t. Let I={ii+s11is1}I=\{i\ i+s_{1}\mid 1\leq i\leq s_{1}\}. Since k3k\geq 3, 2s1=2t<kt<n2s_{1}=2t<kt<n and so, by Proposition 2.1, IE(H)I\subset E(H). Take edges ii+s1i\ i+s_{1} and jj+s1j\ j+s_{1} for some 1i<js11\leq i<j\leq s_{1}. Since ji<s1j-i<s_{1}, ii and jj are not adjacent by Proposition 2.1 and so ii+s1i\ i+s_{1} and jj+s1j\ j+s_{1} do not belong to the same clique. Thus we have shown that θE(H)|I|=s1=t=θE(G)\theta_{E}(H)\geq|I|=s_{1}=t=\theta_{E}(G) if kt<n(k+1)tkt<n\leq(k+1)t. Suppose that n>(k+1)tn>(k+1)t. Since sk=kts_{k}=kt, FE(H)F\subset E(H) by Proposition 2.1 and so θE(H)|F|=nkt\theta_{E}(H)\geq|F|=n-kt. Therefore θE(G)θE(H)\theta_{E}(G)\leq\theta_{E}(H) if (k+1)t<n(k+1)t<n. Thus the “moreover” part is true. ∎

In the following, we compute the vertex clique cover number of a Toeplitz graph Gnt,2t,,ktG_{n}\langle t,2t,\ldots,kt\rangle.

Theorem 2.10.

Let G=Gnt,2t,,ktG=G_{n}\langle t,2t,\ldots,kt\rangle for n>(2k1)tn>(2k-1)t. Then

θv(G)=sn/tk+1+(ts)n/tk+1\theta_{v}(G)=s\left\lceil\frac{\lceil n/t\rceil}{k+1}\right\rceil+(t-s)\left\lceil\frac{\lfloor n/t\rfloor}{k+1}\right\rceil

where ss is the positive integer such that sn(modt)s\equiv n\pmod{t} and 1st1\leq s\leq t.

Proof.

Since sn(modt)s\equiv n\pmod{t} and 1st1\leq s\leq t, it follows from Proposition 2.3 that GG has tt components, each of the first ss components H1,,HsH_{1},\ldots,H_{s} is isomorphic to Gn/t1,2,,kG_{\lceil n/t\rceil}\langle 1,2,\ldots,k\rangle, and each of the other tst-s components Hs+1,,HtH_{s+1},\ldots,H_{t} is isomorphic to Gn/t1,2,,kG_{\lfloor n/t\rfloor}\langle 1,2,\ldots,k\rangle.

Since HiGn/t1,2,,kH_{i}\simeq G_{\lceil n/t\rceil}\langle 1,2,\ldots,k\rangle for each i[s]i\in[s] and k+1k+1 consecutive vertices of {1,2,,nt}\{1,2,\ldots,\lceil\frac{n}{t}\rceil\} form a clique in Gn/t1,2,,kG_{\lceil n/t\rceil}\langle 1,2,\ldots,k\rangle, we have θv(Hi)n/tk+1\theta_{v}(H_{i})\leq\left\lceil\frac{\lceil n/t\rceil}{k+1}\right\rceil for each i[s]i\in[s]. Now, by Theorem 2.8, ω(Hi)=k+1\omega(H_{i})=k+1 for each i[s]i\in[s], so we can conclude that θv(Hi)=n/tk+1\theta_{v}(H_{i})=\left\lceil\frac{\lceil n/t\rceil}{k+1}\right\rceil for each i[s]i\in[s]. Similarly, we can show that θv(Hi)=n/tk+1\theta_{v}(H_{i})=\left\lceil\frac{\lfloor n/t\rfloor}{k+1}\right\rceil for each i{s+1,,t}i\in\{s+1,\ldots,t\}. Therefore

θv(G)=i=1tθv(Hi)=i=1sθv(Hi)+i=s+1tθv(Hi)=sn/tk+1+(ts)n/tk+1.\theta_{v}(G)=\sum_{i=1}^{t}\theta_{v}(H_{i})=\sum_{i=1}^{s}\theta_{v}(H_{i})+\sum_{i=s+1}^{t}\theta_{v}(H_{i})=s\left\lceil\frac{\lceil n/t\rceil}{k+1}\right\rceil+(t-s)\left\lceil\frac{\lfloor n/t\rfloor}{k+1}\right\rceil.\qed

3 Chordal Toeplitz graphs and Perfect Toeplitz graphs

In this section, we study holes in Toeplitz graphs and give a condition for a Toeplitz graph not having holes, which leads to a characterization of chordal Toeplitz graphs. Then we give equivalent conditions for a Toeplitz graph G=Gnt1,t2G=G_{n}\langle t_{1},t_{2}\rangle being perfect. By Theorem 2.8, we know that, for G=Gnt1,t2,,tkG=G_{n}\langle t_{1},t_{2},\ldots,t_{k}\rangle, ω(G)=k+1\omega(G)=k+1 if and only if ti=it1t_{i}=it_{1} for each i[k]i\in[k]. Yet, as long as ntk1+tkn\geq t_{k-1}+t_{k}, we add more equivalent statements such as GG is chordal.

Proposition 3.1.

For positive integers kk and tt, Gnt,2t,,ktG_{n}\langle t,2t,\ldots,kt\rangle is chordal.

Proof.

By Proposition 2.3, it suffices to show that G:=Gm1,2,,kG:=G_{m}\langle 1,2,\ldots,k\rangle is chordal for each integer mm, m>km>k. To reach a contradiction, suppose that GG contains a hole C:=v1v2vv1C:=v_{1}v_{2}\ldots v_{\ell}v_{1} for some integer 4\ell\geq 4. We identify v+1v_{\ell+1} with v1v_{1}. Since the sequence CC cannot strictly increase or decrease, either vi1<viv_{i-1}<v_{i} and vi>vi+1v_{i}>v_{i+1} or vi1>viv_{i-1}>v_{i} and vi<vi+1v_{i}<v_{i+1} for some ii, 2i2\leq i\leq\ell. Assume the former. Then vivi1=|vivi1|=av_{i}-v_{i-1}=|v_{i}-v_{i-1}|=a and vivi+1=|vivi+1|=bv_{i}-v_{i+1}=|v_{i}-v_{i+1}|=b for some a,b{1,2,,k}a,b\in\{1,2,\ldots,k\} by Proposition 2.1. Since 4\ell\geq 4, aba\neq b and so |vi1vi+1|=|ab|{1,2,,k1}|v_{i-1}-v_{i+1}|=|a-b|\in\{1,2,\ldots,k-1\}. Therefore vi1vi+1v_{i-1}v_{i+1} is a chord of CC and we reach a contradiction. We can similarly show that CC also has a chord in the latter case. Thus GG is chordal. ∎

For a path P=v1v2vkP=v_{1}v_{2}\cdots v_{k}, we denote by P1P^{-1} the path vkvk1v2v1v_{k}v_{k-1}\cdots v_{2}v_{1}.

Lemma 3.2.

Let G=Gnt1,t2G=G_{n}\langle t_{1},t_{2}\rangle be a Toeplitz graph with nt1+t2n\geq t_{1}+t_{2}. If t22t1t_{2}\neq 2t_{1}, then GG has a hole of length (t1+t2)/gcd(t1,t2)(t_{1}+t_{2})/\gcd(t_{1},t_{2}).

Proof.

For each i{1,,t1}i\in\{1,\ldots,t_{1}\}, let PiP_{i} be the path such that

Pi=ii+t1i+2t1i+(ni)/t1t1.P_{i}=i\ i+t_{1}\ i+2t_{1}\cdots i+\left\lfloor(n-i)/t_{1}\right\rfloor t_{1}.

Then P1,,Pt1P_{1},\ldots,P_{t_{1}} are t1t_{1} disjoint paths which contain all the vertices of GG.

Now we construct a cycle in the following way. We start from the vertex 11 and consider the edge 1 1+t21\ 1+t_{2}. Then 1+t21+t_{2} is on the path PjP_{j} for j1+t2(modt1)j\equiv 1+t_{2}\pmod{t_{1}}. We denote by P1P_{1}^{\prime} the (1+t2,j)(1+t_{2},j)-section of Pj1{P_{j}}^{-1}. Since jt1j\leq t_{1}, j+t2t1+t2nj+t_{2}\leq t_{1}+t_{2}\leq n and so j+t2j+t_{2} is a vertex of GG. Then we take the edge jj+t2j\ j+t_{2} and the path PkP_{k} where kj+t21+2t2(modt1)k\equiv j+t_{2}\equiv 1+2t_{2}\pmod{t_{1}}. We denote by P2P_{2}^{\prime} the (j+t2,k)(j+t_{2},k)-section of Pk1{P_{k}}^{-1}. Then 11 P1P_{1}^{\prime} P2P_{2}^{\prime} is a (1,k)(1,k)-path. Noting that 11+xt2(modt1)1\equiv 1+xt_{2}\pmod{t_{1}} has a solution s:=t1/ds:=t_{1}/d where d=gcd(t1,t2)d=\gcd(t_{1},t_{2}), we may conclude that 11 P1P_{1}^{\prime} P2P_{2}^{\prime} \cdots PsP_{s}^{\prime} is a cycle in GG. By construction, the length of this cycle is the minimum of sum of two positive integers xx and yy satisfying yt2=xt1yt_{2}=xt_{1}, that is, min{x+yyt2xt1=0,x,y+}=(t1+t2)/d\min\{x+y\mid yt_{2}-xt_{1}=0,x,y\in\mathbb{Z}^{+}\}=(t_{1}+t_{2})/d.

To show that the cycle has no chord, take two vertices uu and vv with u<vu<v on the cycle. Then u=1+u1t2v1t1u=1+u_{1}t_{2}-v_{1}t_{1} and v=1+u2t2v2t1v=1+u_{2}t_{2}-v_{2}t_{1} for some {u1,u2}{1,,t1/d}\{u_{1},u_{2}\}\subset\{1,\ldots,t_{1}/d\} and {v1,v2}{1,,t2/d}\{v_{1},v_{2}\}\subset\{1,\ldots,t_{2}/d\}. If uu and vv are adjacent, vuv-u is either t1t_{1} or t2t_{2}. Suppose vu=t1v-u=t_{1}. Then (u2u1)t2(v2v1+1)t1=0(u_{2}-u_{1})t_{2}-(v_{2}-v_{1}+1)t_{1}=0. However, |u2u1|<t1/d|u_{2}-u_{1}|<t_{1}/d and so u2=u1u_{2}=u_{1} and |v2v1|=1|v_{2}-v_{1}|=1, which implies that uu and vv are consecutive on the cycle. Similarly one can show that if vu=t2v-u=t_{2}, then uu and vv are also consecutive. Thus the cycle has no chord. Since t22t1t_{2}\neq 2t_{1}, (t1+t2)/d>3(t_{1}+t_{2})/d>3 and so the cycle is a hole. ∎

The following theorem characterizes the chordal Toeplitz graphs G=Gnt1,t2,,tkG=G_{n}\langle t_{1},t_{2},\ldots,t_{k}\rangle with ntk1+tkn\geq t_{k-1}+t_{k}.

Theorem 3.3.

Let G=Gnt1,t2,,tkG=G_{n}\langle t_{1},t_{2},\ldots,t_{k}\rangle be a Toeplitz graph. If ntk1+tkn\geq t_{k-1}+t_{k}, then the following statements are equivalent.

  • (i)

    GG is interval.

  • (ii)

    GG is chordal.

  • (iii)

    ti=it1t_{i}=it_{1} for each i[k]i\in[k].

  • (iv)

    ω(G)=k+1\omega(G)=k+1.

Proof.

(i)(ii)\mbox{(i)}\Rightarrow\mbox{(ii)} is obvious. By Theorem 2.8, (iii)(iv)\mbox{(iii)}\Leftrightarrow\mbox{(iv)}. To complete the proof, we shall show that (ii)(iii)\mbox{(ii)}\Rightarrow\mbox{(iii)} and (iii)(i)\mbox{(iii)}\Rightarrow\mbox{(i)}. To show (ii)(iii)\mbox{(ii)}\Rightarrow\mbox{(iii)}, we denote by C4(ti,tj)C_{4}(t_{i},t_{j}) the 44-cycle

1(1+ti)(1+ti+tj)(1+tj) 11\ (1+t_{i})\ (1+t_{i}+t_{j})\ (1+t_{j})\ 1

for each {i,j}[k]\{i,j\}\subset[k] with i<ji<j and ti+tj<nt_{i}+t_{j}<n. Since GG is chordal, C4(ti,tj)C_{4}(t_{i},t_{j}) has a chord for each {i,j}[k]\{i,j\}\subset[k] with i<ji<j and ti+tj<nt_{i}+t_{j}<n. Thus, for each {i,j}[k]\{i,j\}\subset[k] with i<ji<j and tj+ti<nt_{j}+t_{i}<n,

tjti{t1,t2,,tk}ortj+ti{t1,t2,,tk}.t_{j}-t_{i}\in\{t_{1},t_{2},\ldots,t_{k}\}\quad\mbox{or}\quad t_{j}+t_{i}\in\{t_{1},t_{2},\ldots,t_{k}\}. (3)

Suppose that k=2k=2. If t22t1t_{2}\neq 2t_{1}, then GG has a hole of length (t1+t2)/gcd(t1,t2)(t_{1}+t_{2})/\gcd(t_{1},t_{2}) by Lemma 3.2 and we reach a contradiction. Thus t2=2t1t_{2}=2t_{1}. Now we suppose that k=3k=3. Since t1+t2<t1+t3<t2+t3nt_{1}+t_{2}<t_{1}+t_{3}<t_{2}+t_{3}\leq n, C4(t1,t2)C_{4}(t_{1},t_{2}) and C4(t1,t3)C_{4}(t_{1},t_{3}) exist. Therefore we have t2t1=t1t_{2}-t_{1}=t_{1} or t1+t2=t3t_{1}+t_{2}=t_{3} from C4(t1,t2)C_{4}(t_{1},t_{2}) and t3t1{t1,t2}t_{3}-t_{1}\in\{t_{1},t_{2}\} from C4(t1,t3)C_{4}(t_{1},t_{3}). If t2t1=t1t_{2}-t_{1}=t_{1}, then t3t1=t2t_{3}-t_{1}=t_{2}, so t2=2t1t_{2}=2t_{1} and t3=3t1t_{3}=3t_{1}. Assume that t1+t2=t3t_{1}+t_{2}=t_{3}. To reach a contradiction, suppose that t22t1t_{2}\neq 2t_{1}. By definition, H=Gt1+t2t1,t2H=G_{t_{1}+t_{2}}\langle t_{1},t_{2}\rangle is a subgraph of GG and HH contains a hole CC by Lemma 3.2. Since GG is chordal, CC has a chord in GG. Then the difference of two ends of a chord is t3t_{3}, for otherwise the chord also exists in HH. However, since CC is a subgraph of HH, the difference of any pair of vertices on CC is at most t1+t21=t31t_{1}+t_{2}-1=t_{3}-1 and we reach a contradiction. Thus t2=2t1t_{2}=2t_{1} and t3=t1+t2=3t1t_{3}=t_{1}+t_{2}=3t_{1}.

Now suppose k4k\geq 4. We consider the cycle C4(tj,tk)C_{4}(t_{j},t_{k}) for each j[k2]j\in[k-2]. Then for each j[k2]j\in[k-2], tj+tk<nt_{j}+t_{k}<n and tj+tk{t1,,tk}t_{j}+t_{k}\notin\{t_{1},\ldots,t_{k}\} and so tktj{t1,,tk}t_{k}-t_{j}\in\{t_{1},\ldots,t_{k}\} for each j[k2]j\in[k-2] by (3). To reach a contradiction, suppose that tktk1Bt_{k}-t_{k-1}\notin B. Then tktjtk1t_{k}-t_{j}\neq t_{k-1} for any j[k2]j\in[k-2] and so {tkt1,tkt2,,tktk2}={t1,,tk2}\{t_{k}-t_{1},t_{k}-t_{2},\ldots,t_{k}-t_{k-2}\}=\{t_{1},\ldots,t_{k-2}\}. Since tkt1t_{k}-t_{1} is the largest element in the set, we have tkt1=tk2t_{k}-t_{1}=t_{k-2}. Then

tk=tk2+t1<tk1+tk2<tk+tk1n,t_{k}=t_{k-2}+t_{1}<t_{k-1}+t_{k-2}<t_{k}+t_{k-1}\leq n,

so tk1tk2Bt_{k-1}-t_{k-2}\in B by (3). However, tk1tk2<tktk2=t1t_{k-1}-t_{k-2}<t_{k}-t_{k-2}=t_{1} and we reach a contradiction. Therefore tktk1Bt_{k}-t_{k-1}\in B and so {tkt1,tkt2,,tktk1}={t1,,tk1}\{t_{k}-t_{1},t_{k}-t_{2},\ldots,t_{k}-t_{k-1}\}=\{t_{1},\ldots,t_{k-1}\}. Thus

tk=tj+tkjfor each j[k1].t_{k}=t_{j}+t_{k-j}\quad\mbox{for each }j\in[k-1]. (4)

For each j{2,,k2}j\in\{2,\ldots,k-2\},

tk1tjtk1t2<tkt2=tk2t_{k-1}-t_{j}\leq t_{k-1}-t_{2}<t_{k}-t_{2}=t_{k-2}

by (4). Yet, since tj+tk1>tkt_{j}+t_{k-1}>t_{k} for each j{2,,k2}j\in\{2,\ldots,k-2\} by (4), tk1tj{t1,,tk}t_{k-1}-t_{j}\in\{t_{1},\ldots,t_{k}\} for each j{2,,k2}j\in\{2,\ldots,k-2\}. Therefore

tk1tjtk3for each j{2,,k2}.t_{k-1}-t_{j}\leq t_{k-3}\quad\mbox{for each }j\in\{2,\ldots,k-2\}. (5)

In addition, since tk1tk2<tktk2=t2t_{k-1}-t_{k-2}<t_{k}-t_{k-2}=t_{2} by (4), we have tk1tk2=t1t_{k-1}-t_{k-2}=t_{1}. Thus, by (5),

tk1=tj+tkj1for each j{1,,k2}.t_{k-1}=t_{j}+t_{k-j-1}\quad\mbox{for each }j\in\{1,\ldots,k-2\}. (6)

By subtracting (6) from (4) for each j{1,,k2}j\in\{1,\ldots,k-2\}, we have

tktk1=tkjtkj1for each j{1,,k2}.t_{k}-t_{k-1}=t_{k-j}-t_{k-j-1}\quad\mbox{for each }j\in\{1,\ldots,k-2\}. (7)

By (4), tktk1=t1t_{k}-t_{k-1}=t_{1}. Thus, by (7),

tj=tj1+t1=tj2+2t1==t1+(j1)t1=jt1t_{j}=t_{j-1}+t_{1}=t_{j-2}+2t_{1}=\cdots=t_{1}+(j-1)t_{1}=jt_{1}

for each j[k]j\in[k].

Now we show (iii)(i)\mbox{(iii)}\Rightarrow\mbox{(i)}. Suppose that G=Gnt1,2t1,,kt1G=G_{n}\langle t_{1},2t_{1},\ldots,kt_{1}\rangle.

By Proposition 2.3, each component of GG is Gni1,2,,kG_{n_{i}}\langle 1,2,\ldots,k\rangle for some ninn_{i}\leq n. Therefore it suffices to show that Gni1,2,,kG_{n_{i}}\langle 1,2,\ldots,k\rangle is an interval graph for any nin_{i}. To each vertex u[ni]u\in[n_{i}], we assign an interval [u,u+k][u,u+k]. We note that [u,u+k][v,v+k][u,u+k]\cap[v,v+k]\neq\emptyset if and only if vu[k]v-u\in[k], that is, uu and vv are adjacent in Gni1,2,,kG_{n_{i}}\langle 1,2,\ldots,k\rangle. Thus Gni1,2,,kG_{n_{i}}\langle 1,2,\ldots,k\rangle is an interval graph and hence we may conclude GG is an interval graph. ∎

In the rest of this section, we characterize perfect Toeplitz graphs in the following by utilizing the results we have shown. We first introduce classes of well-known perfect graphs.

Theorem 3.4.

[11] Chordal graphs, cographs and bipartite graphs are perfect.

The following corollary is an immediate consequence of Proposition 3.1 and Theorem 3.4.

Corollary 3.5.

Let G=Gnt,2t,,ktG=G_{n}\langle t,2t,\ldots,kt\rangle. Then χ(G)=k+1\chi(G)=k+1.

Proof.

By Proposition 3.1, G=Gnt,2t,,ktG=G_{n}\langle t,2t,\ldots,kt\rangle is chordal, and thus, by Theorem 3.4, it is a perfect graph. Then, by Theorem 3.3, χ(G)=ω(G)=k+1.\chi(G)=\omega(G)=k+1.

Next, we characterize perfect Toeplitz graphs Gnt1,t2G_{n}\langle t_{1},t_{2}\rangle. To do so, we introduce the following Theorem. A graph GG is called a Berge graph if it contains neither an odd hole nor an odd anti-hole as an induced subgraph.

Theorem 3.6.

(Strong Perfect Graph Theorem [2]) A graph is perfect if and only if it is Berge.

Theorem 3.7.

Let G=Gnt1,t2G=G_{n}\langle t_{1},t_{2}\rangle with nt1+t2n\geq t_{1}+t_{2} and d=gcd(t1,t2)d=\gcd(t_{1},t_{2}). Then the following statements are equivalent.

  • (i)

    (t1+t2)/d(t_{1}+t_{2})/d is even or (t1+t2)/d=3(t_{1}+t_{2})/d=3.

  • (ii)

    GG is an odd-hole-free graph.

  • (iii)

    GG is a perfect graph.

  • (iv)

    GG is a weakly perfect graph.

Proof.

Let k1=t1/dk_{1}=t_{1}/d and k2=t2/dk_{2}=t_{2}/d. We first show (ii)(i)\mbox{(ii)}\Rightarrow\mbox{(i)}. Since GG has a hole of length k1+k2k_{1}+k_{2} by Lemma 3.2, GG is an odd-hole-free graph only if k1+k2k_{1}+k_{2} is even or k1+k2=3k_{1}+k_{2}=3.

Now we show (i)(ii)\mbox{(i)}\Rightarrow\mbox{(ii)}. Suppose that k1+k2=3k_{1}+k_{2}=3. Since t1<t2t_{1}<t_{2}, k1=1k_{1}=1 and k2=2k_{2}=2 and so t2=2t1t_{2}=2t_{1}. Therefore GG is chordal by Proposition 3.1 and thus GG is odd-hole-free by Theorem 3.4. Suppose that k1+k2k_{1}+k_{2} is even. We prove by contradiction. Suppose that GG has an odd-hole CC of length \ell for some positive odd integer 5\ell\geq 5. Let C=v1v2vv1C=v_{1}v_{2}\ldots v_{\ell}v_{1}. Then, by Proposition 2.1, |vi+1vi|{t1,t2}|v_{i+1}-v_{i}|\in\{t_{1},t_{2}\} for each i[]i\in[\ell] (we identify v1v_{1} with v+1v_{\ell+1}). For each j{1,2}j\in\{1,2\}, let aja_{j} be the number of indices ii such that vi+1vi=tjv_{i+1}-v_{i}=t_{j}, and let bjb_{j} be the number of indices ii such that vivi+1=tjv_{i}-v_{i+1}=t_{j} for each i[]i\in[\ell]. Then the length of CC is a1+b1+a2+b2a_{1}+b_{1}+a_{2}+b_{2}. Since v1=v+1v_{1}=v_{\ell+1},

0\displaystyle 0 =(v+1v)+(vv1)++(v2v1)+(v1v+1)\displaystyle=(v_{\ell+1}-v_{\ell})+(v_{\ell}-v_{\ell-1})+\cdots+(v_{2}-v_{1})+(v_{1}-v_{\ell+1})
=a1t1+a2t2b1t1b2t2,\displaystyle=a_{1}t_{1}+a_{2}t_{2}-b_{1}t_{1}-b_{2}t_{2},

or (a1b1)t1=(b2a2)t2(a_{1}-b_{1})t_{1}=(b_{2}-a_{2})t_{2}. If a2=b2a_{2}=b_{2}, then b1=a1b_{1}=a_{1} and so the length of CC is 2(a1+a2)2(a_{1}+a_{2}), which is a contradiction. Thus a2b2a_{2}\neq b_{2}. Then

k2k1=t2t1=a1b1b2a2.\frac{k_{2}}{k_{1}}=\frac{t_{2}}{t_{1}}=\frac{a_{1}-b_{1}}{b_{2}-a_{2}}.

Therefore αk2=a1b1\alpha k_{2}=a_{1}-b_{1} and αk1=b2a2\alpha k_{1}=b_{2}-a_{2} for some integer α\alpha. Then the length of CC is 2b1+2a2+α(k1+k2)2b_{1}+2a_{2}+\alpha(k_{1}+k_{2}). Since k1+k2k_{1}+k_{2} is even by the supposition, we reach a contradiction. Thus GG is odd-hole-free. Hence we have shown that (i)(ii)\mbox{(i)}\Leftrightarrow\mbox{(ii)}.

By Theorem 3.6, (iii)(ii)\mbox{(iii)}\Rightarrow\mbox{(ii)}. Next, we will show (i)(iii)\mbox{(i)}\Rightarrow\mbox{(iii)}. Suppose that k1+k2k_{1}+k_{2} is even or k1+k2=3k_{1}+k_{2}=3. If k2=2k1k_{2}=2k_{1}, then GG is chordal by Proposition 3.1 and thus GG is a perfect graph by Theorem 3.4. Suppose that k22k1k_{2}\neq 2k_{1}. Then k1+k23k_{1}+k_{2}\neq 3, so k1+k2k_{1}+k_{2} is even. In addition, GG is not chordal by Theorem 3.3, so ω(G)2\omega(G)\leq 2 by Theorem 2.8. Since we have shown (i)(ii)\mbox{(i)}\Rightarrow\mbox{(ii)}, GG is an odd-hole-free graph. By Theorem 3.6, it remains to show that GG does not contain an odd anti-hole. Since the complement of a cycle C5C_{5} is C5C_{5} again, GG does not contain an anti-hole on 55 vertices. Note that any odd anti-hole with at least 77 vertices contains a triangle. Yet, since ω(G)2\omega(G)\leq 2, GG does not contain a triangle. Therefore GG does not contain any odd anti-hole with at least 77 vertices and so we have shown that GG is odd anti-hole-free.

Obviously (iii)(iv)\mbox{(iii)}\Rightarrow\mbox{(iv)}. To complete the proof, we will show (iv)(ii)\mbox{(iv)}\Rightarrow\mbox{(ii)}. Suppose that GG is a weakly perfect graph. Then χ(G)=ω(G)\chi(G)=\omega(G) by definition. By Theorem 2.8, ω(G)3\omega(G)\leq 3 and the equality holds if and only if GG is chordal. If ω(G)=3\omega(G)=3, then GG is chordal and so, by Theorem 3.4, GG is a perfect graph. Since we have shown (iii)(ii)\mbox{(iii)}\Rightarrow\mbox{(ii)}, GG is odd-hole-free if ω(G)=3\omega(G)=3. Suppose that ω(G)=2\omega(G)=2. Then χ(G)=2\chi(G)=2, so GG is a bipartite graph, which implies that GG is odd-hole-free. ∎

Theorem 3.8.

(Weak Perfect Graph Theorem [18]) A graph is perfect if and only if its complement is perfect.

By definition, it is easy to check that the complement of a Toeplitz graph Gns1,s2G_{n}\langle s_{1},s_{2}\rangle is Gnt1,t2,,tn3G_{n}\langle t_{1},t_{2},\ldots,t_{n-3}\rangle where {t1,,tn3}=[n1]{s1,s2}\{t_{1},\ldots,t_{n-3}\}=[n-1]\setminus\{s_{1},s_{2}\}. Thus, by Theorems 3.7 and 3.8, the following corollary is true.

Corollary 3.9.

Let G=Gnt1,,tn3G=G_{n}\langle t_{1},\ldots,t_{n-3}\rangle and {s1,s2}=[n1]{t1,,tn3}\{s_{1},s_{2}\}=[n-1]\setminus\{t_{1},\ldots,t_{n-3}\} with s1+s2ns_{1}+s_{2}\leq n. Then GG is a perfect graph if and only if (s1+s2)/gcd(s1,s2)(s_{1}+s_{2})/\gcd(s_{1},s_{2}) is even or (s1+s2)/gcd(s1,s2)=3(s_{1}+s_{2})/\gcd(s_{1},s_{2})=3.

4 Degree sequence of Toeplitz graphs

The degree sequence of a graph is defined to be a monotonic nonincreasing sequence of the vertex degrees of its graph vertices. In this section, we characterize the degree sequence (d1,d2,,dn)(d_{1},d_{2},\ldots,d_{n}) of a Toeplitz graph with nn vertices and show that a Toeplitz graph is a regular graph if and only if it is a circulant graph.

We recall that B(G)={t1,,tk}B(G)=\{t_{1},\ldots,t_{k}\} for G:=Gnt1,,tkG:=G_{n}\langle t_{1},\ldots,t_{k}\rangle. For a Toeplitz graph G=Gnt1,,tkG=G_{n}\langle t_{1},\ldots,t_{k}\rangle, we denote by G(i)\ell_{G}(i) the number of elements in {t1,,tk}\{t_{1},\ldots,t_{k}\} which are less than ii. Then it is easy to see that

G(j+1)G(j)=1jB(G)\ell_{G}(j+1)-\ell_{G}(j)=1\Leftrightarrow j\in B(G) (8)

for any j[n1]j\in[n-1].

Theorem 4.1.

Let G=Gnt1,,tkG=G_{n}\langle t_{1},\ldots,t_{k}\rangle. Then for each i[n]i\in[n],

deg(i)=G(i)+G(ni+1).\deg(i)=\ell_{G}(i)+\ell_{G}(n-i+1).
Proof.

Take a vertex i[n]i\in[n]. Then

deg(i)\displaystyle\deg(i) =|{it1,,itk,i+t1,,i+tk}[n]|\displaystyle=|\{i-t_{1},\ldots,i-t_{k},i+t_{1},\ldots,i+t_{k}\}\cap[n]|
=|{it1,,itk}[n]|+|{i+t1,,i+tk}[n]|.\displaystyle=|\{i-t_{1},\ldots,i-t_{k}\}\cap[n]|+|\{i+t_{1},\ldots,i+t_{k}\}\cap[n]|.

By definition of G(i)\ell_{G}(i),

{it1,,itk}[n]={it1,,itG(i)}\{i-t_{1},\ldots,i-t_{k}\}\cap[n]=\{i-t_{1},\ldots,i-t_{\ell_{G}(i)}\}

and

{i+t1,,i+tk}[n]={i+t1,,i+tG(ni+1)},\{i+t_{1},\ldots,i+t_{k}\}\cap[n]=\{i+t_{1},\ldots,i+t_{\ell_{G}(n-i+1)}\},

where the right sides of the first and second equalities are \emptyset if G(i)=0\ell_{G}(i)=0 or G(ni+1)=0\ell_{G}(n-i+1)=0, respectively. ∎

Corollary 4.2.

For a Toeplitz graph on nn vertices, the following are true:

  • (a)

    For each j[n]j\in[n], deg(j)=deg(nj+1)\deg(j)=\deg(n-j+1).

  • (b)

    If nn is odd, then deg(n+12)\deg\left({\frac{n+1}{2}}\right) is even.

Lemma 4.3.

The difference of degrees of two consecutive vertices of a Toeplitz graph is at most 11. Moreover, for G:=Gnt1,,tkG:=G_{n}\langle t_{1},\ldots,t_{k}\rangle and each vertex j[n1]j\in[n-1], the following are true:

  • (a)

    deg(j)=deg(j+1)\deg(j)=\deg(j+1) if and only if {j,nj}B(G)\{j,n-j\}\subseteq B(G) or {j,nj}[n1]\B(G)\{j,n-j\}\subseteq[n-1]\backslash B(G).

  • (b)

    deg(j)+1=deg(j+1)\deg(j)+1=\deg(j+1) if and only if jB(G)j\in B(G) and njB(G)n-j\notin B(G).

  • (c)

    deg(j)1=deg(j+1)\deg(j)-1=\deg(j+1) if and only if njB(G)n-j\in B(G) and jB(G)j\notin B(G).

Proof.

Let G=Gnt1,,tkG=G_{n}\langle t_{1},\ldots,t_{k}\rangle. Take j[n1]j\in[n-1]. By definition,

0G(j+1)G(j)1and1G(nj)G(nj+1)0.0\leq\ell_{G}(j+1)-\ell_{G}(j)\leq 1\ \ \mbox{and}\ \ -1\leq\ell_{G}(n-j)-\ell_{G}(n-j+1)\leq 0.

Therefore

1deg(j+1)deg(j)1-1\leq\deg(j+1)-\deg(j)\leq 1

by Theorem 4.1. To show the ‘moreover’ part, suppose that deg(j+1)=deg(j)\deg(j+1)=\deg(j). Then either (i) G(j+1)G(j)=1\ell_{G}(j+1)-\ell_{G}(j)=1 and G(nj)G(nj+1)=1\ell_{G}(n-j)-\ell_{G}(n-j+1)=-1 or (ii) G(j+1)G(j)=G(nj)G(nj+1)=0\ell_{G}(j+1)-\ell_{G}(j)=\ell_{G}(n-j)-\ell_{G}(n-j+1)=0. Thus, by (8), {j,nj}B(G)\{j,n-j\}\subseteq B(G) in the case (i) and {j,nj}[n1]\B(G)\{j,n-j\}\subseteq[n-1]\backslash B(G) in the case (ii). One may check (b) and (c) in a similar manner as above. ∎

From Corollary 4.2, we know that the degree sequence of a Toeplitz graph of order nn has the property that each term appears an even number of times, except the degree of (n+1)/2(n+1)/2 which happens to be even when nn is odd. In addition, the terms form consecutive integers with possible repetition by Lemma 4.3. However, this necessary condition cannot be sufficient. To see why, we take the sequence 𝐝=(4,3,3,2,2,1,1){\bf d}=(4,3,3,2,2,1,1), which satisfies the necessary condition. Suppose that 𝐝{\bf d} is the degree sequence of a Toeplitz graph GG. Then deg(1)=1\deg(1)=1, deg(2)=2\deg(2)=2, deg(3)=3\deg(3)=3. By Lemma 4.3(b), 1B(G)1\in B(G) and 2B(G)2\in B(G). Then 11 is adjacent to 22 and 33 and we reach a contradiction.111The authors thank Homoon Ryu for finding this counterexample. This counterexample is a special case of (2m,2m1,2m1,2m2,2m2,,2,2,1,1)(2m,2m-1,2m-1,2m-2,2m-2,\ldots,2,2,1,1) for some m2m\geq 2, which cannot be the degree sequence of any Toeplitz graph of order 4m14m-1.

The following theorem gives a necessary and sufficient condition for a monotone nonincreasing sequence 𝐝n=(d1,d2,,dn){\bf d}_{n}=(d_{1},d_{2},\ldots,d_{n}) of nonnegative integers being the degree sequence of a Toeplitz graph.

Theorem 4.4.

A monotone nonincreasing sequence 𝐝n=(d1,d2,,dn){\bf d}_{n}=(d_{1},d_{2},\ldots,d_{n}) of nonnegative integers is the degree sequence of a Toeplitz graph if and only if there is a permutation π\pi on [n][n] such that

  • (a)

    |dπ(i+1)dπ(i)|1|d_{\pi(i+1)}-d_{\pi(i)}|\leq 1 for each i[n1]i\in[n-1];

  • (b)

    dπ(i)=dπ(ni+1)d_{\pi(i)}=d_{\pi(n-i+1)} for each i[n]i\in[n];

  • (c)

    sdπ(1)n1ss\leq d_{\pi(1)}\leq n-1-s;

  • (d)

    dπ(1)d_{\pi(1)} and ss have the same parity if nn is odd,

where ss is the number of i{1,,(n1)/2}i\in\{1,\ldots,\left\lfloor(n-1)/2\right\rfloor\} for which dπ(i+1)dπ(i)0d_{\pi(i+1)}-d_{\pi(i)}\neq 0.

Proof.

For notational convenience, we let m=(n1)/2m=\left\lfloor(n-1)/2\right\rfloor. We first show the ‘only if’ part. Suppose that 𝐝n=(d1,d2,,dn){\bf d}_{n}=(d_{1},d_{2},\ldots,d_{n}) is the degree sequence of a Toeplitz graph G:=Gnt1,,tkG:=G_{n}\langle t_{1},\ldots,t_{k}\rangle. Let π\pi be the permutation on [n][n] such that dπ(i)=deg(i)d_{\pi(i)}=\deg(i) for each i[n]i\in[n]. Then the conditions (a) and (b) immediately come from Corollary 4.2 and Lemma 4.3. Obviously, dπ(1)=deg(1)=kd_{\pi(1)}=\deg(1)=k. Now, for each i[m]i\in[m] such that deg(i+1)deg(i)0\deg(i+1)-\deg(i)\neq 0, exactly one of ii or nin-i belongs to B(G)B(G) and the other belongs to [n1]B(G)[n-1]\setminus B(G) by Lemma 4.3 (b) and (c). Therefore s|B(G)|=ks\leq|B(G)|=k and s|[n1]B(G)|=n1ks\leq|[n-1]\setminus B(G)|=n-1-k, so the sequence with the permutation π\pi satisfies the condition (c). Let

B+={i[m]deg(i+1)deg(i)=1}B^{+}=\{i\in[m]\mid\deg(i+1)-\deg(i)=1\}

and

B={i[m]deg(i+1)deg(i)=1}.B^{-}=\{i\in[m]\mid\deg(i+1)-\deg(i)=-1\}.

Then

dπ(m+1)dπ(1)=(dπ(m+1)dπ(m))++(dπ(2)dπ(1))=|B+||B|.d_{\pi(m+1)}-d_{\pi(1)}=\left(d_{\pi(m+1)}-d_{\pi(m)}\right)+\cdots+\left(d_{\pi(2)}-d_{\pi(1)}\right)=|B^{+}|-|B^{-}|.

Since the parities of s=|B+|+|B|s=|B^{+}|+|B^{-}| and |B+||B||B^{+}|-|B^{-}| are the same, dπ(m+1)d_{\pi(m+1)} is odd and we reach a contradiction. Thus the sequence with the permutation π\pi satisfies the condition (d).

Now we show the ‘if’ part. Suppose that there exists a permutation π\pi on [n][n] satisfying the conditions (a), (b), (c) and (d). For notational convenience, let k=dπ(1)k=d_{\pi(1)}. We will construct a Toeplitz graph G:=Gnt1,,tkG:=G_{n}\langle t_{1},\ldots,t_{k}\rangle such that deg(i)=dπ(i)\deg(i)=d_{\pi(i)} for each i[n]i\in[n] as follows.

Let

B1={idπ(i+1)dπ(i)=1,i[m]},B_{1}=\{i\mid d_{\pi(i+1)}-d_{\pi(i)}=1,i\in[m]\},
B2={idπ(i+1)dπ(i)=1,i[m]},B_{2}=\{i\mid d_{\pi(i+1)}-d_{\pi(i)}=-1,i\in[m]\},

and

B3={idπ(i+1)dπ(i)=0,i[m]}.B_{3}=\{i\mid d_{\pi(i+1)}-d_{\pi(i)}=0,i\in[m]\}.

Then |B1|+|B2|+|B3|=m|B_{1}|+|B_{2}|+|B_{3}|=m by condition (a) and they are mutually disjoint sets. By definition, |B1|+|B2|=s|B_{1}|+|B_{2}|=s and so |B3|=ms|B_{3}|=m-s. In addition, since dπ(1)n1sd_{\pi(1)}\leq n-1-s by the condition (c), dπ(1)sn12sd_{\pi(1)}-s\leq n-1-2s. Thus

dπ(1)s2n12s2|B3|,\left\lfloor\frac{d_{\pi(1)}-s}{2}\right\rfloor\leq\left\lfloor\frac{n-1-2s}{2}\right\rfloor\leq|B_{3}|,

and so we may choose (dπ(1)s)/2\left\lfloor(d_{\pi(1)}-s)/2\right\rfloor elements from B3B_{3}. We denote by B3B_{3}^{*} the set of those (dπ(1)s)/2\left\lfloor(d_{\pi(1)}-s)/2\right\rfloor elements. Now we define {t1,,tk}\{t_{1},\ldots,t_{k}\} by

{B1{niiB2}B3{niiB3}if dπ(1)s is even;B1{niiB2}B3{niiB3}{n2}otherwise.\begin{cases}B_{1}\cup\{n-i\mid i\in B_{2}\}\cup B_{3}^{*}\cup\{n-i\mid i\in B_{3}^{*}\}&\mbox{if $d_{\pi(1)}-s$ is even;}\\ B_{1}\cup\{n-i\mid i\in B_{2}\}\cup B_{3}^{*}\cup\{n-i\mid i\in B_{3}^{*}\}\cup\{\frac{n}{2}\}&\mbox{otherwise.}\end{cases}

Then the set {t1,,tk}\{t_{1},\ldots,t_{k}\} is well-defined because n2\frac{n}{2} is integer when dπ(1)sd_{\pi(1)}-s is odd, by the condition (d), and it is straightforward to check that the set defined above has kk element.

Let G=Gnt1,,tkG=G_{n}\langle t_{1},\ldots,t_{k}\rangle. Then

degG(1)=k=dπ(1).\deg_{G}(1)=k=d_{\pi(1)}. (9)

Take i[m]i\in[m]. By the definition of B(G)B(G), iB1i\in B_{1} if iB(G)i\in B(G) and niB(G)n-i\notin B(G); iB2i\in B_{2} if iB(G)i\notin B(G) and niB(G)n-i\in B(G); iB3i\in B_{3}^{*} if either iB(G)i\in B(G) and niB(G)n-i\in B(G) or iB(G)i\notin B(G) and niB(G)n-i\notin B(G). Thus deg(i)=dπ(i)\deg(i)=d_{\pi(i)} by Lemma 4.3. By the way, for each i[m]i\in[m], dπ(i)=dπ(ni+1)d_{\pi(i)}=d_{\pi(n-i+1)} by the condition (b), so deg(ni+1)=deg(i)=dπ(i)=dπ(ni+1)\deg(n-i+1)=\deg(i)=d_{\pi(i)}=d_{\pi(n-i+1)}. Hence dπ(i)=deg(i)d_{\pi(i)}=\deg(i) for each i[n]i\in[n] and 𝐝n{\bf d}_{n} is the degree sequence of GG.∎

In the rest of this section, we will show that a necessary and sufficient condition for a Toeplitz graph being regular is its being a circulant graph. To do so, we need the following lemma.

Lemma 4.5.

Let G=Gnt1,t2,,tkG=G_{n}\langle t_{1},t_{2},\ldots,t_{k}\rangle. Then GG is a circulant graph if and only if tk+1i=ntit_{k+1-i}=n-t_{i} for each i[k]i\in[k].

Proof.

Let (aij)1i,jn(a_{ij})_{1\leq i,j\leq n} be the adjacency matrix of GG. To show the ‘only if’ part, suppose that GG is a circulant graph.

Take i[n1]i\in[n-1]. Suppose that iB(G)i\in B(G) (recall B(G)={t1,,tk}B(G)=\{t_{1},\ldots,t_{k}\}). Then ai+1,1=1a_{i+1,1}=1 by definition. By (2), ani+1,1=ai+1,1=1a_{n-i+1,1}=a_{i+1,1}=1. Therefore niB(G)n-i\in B(G). Thus

B(G)={ntk,,nt1}.B(G)=\{n-t_{k},\ldots,n-t_{1}\}.

Since t1<t2<<tkt_{1}<t_{2}<\cdots<t_{k}, tk+1j=ntjt_{k+1-j}=n-t_{j} for each j[k]j\in[k].

Now we show the ‘if’ part. Suppose that tk+1i=ntit_{k+1-i}=n-t_{i} for each i[n1]i\in[n-1]. Then, for each s[n1]s\in[n-1], as+1,1=1a_{s+1,1}=1 if and only if sB(G)s\in B(G) if and only if nsB(G)n-s\in B(G) if and only if ans+1,1=1a_{n-s+1,1}=1. Thus (aij)1i,jn(a_{ij})_{1\leq i,j\leq n} is a circulant matrix. ∎

Let GG be a circulant graph isomorphic to Gnt1,,tkG_{n}\langle t_{1},\ldots,t_{k}\rangle. It is well-known that if tin/2t_{i}\neq n/2 for each i[k]i\in[k], then kk is even and GG is a regular graph; if nn is even and tj=n/2t_{j}=n/2 for some j[k]j\in[k], then kk is odd and GG is a regular graph (see [22]). Therefore any circulant graph is a regular Toeplitz graph. In the following, we show that its converse is also true.

Theorem 4.6.

A Toeplitz graph is regular if and only if it is a circulant graph.

Proof.

By the argument above, it is sufficient to show the ‘only if’ part. Let G=Gnt1,,tkG=G_{n}\langle t_{1},\ldots,t_{k}\rangle be a regular Toeplitz graph. Then deg(i)=deg(i+1)\deg(i)=\deg(i+1) for each i{1,,n/21}i\in\{1,\ldots,\left\lceil n/2\right\rceil-1\}. Now, by Lemma 4.3(a), either {i,ni}B(G)\{i,n-i\}\subset B(G) or {i,ni}[n1]\B(G)\{i,n-i\}\subset[n-1]\backslash B(G) for each i[n1]i\in[n-1]. Since B(G)B(G)\neq\emptyset,

B(G)={t1,,tk}={ntk,,nt1}.B(G)=\{t_{1},\ldots,t_{k}\}=\{n-t_{k},\ldots,n-t_{1}\}.

Thus tk+1i=ntit_{k+1-i}=n-t_{i} for each i[k]i\in[k] and so, by Lemma 4.5, GG is a circulant graph. ∎

5 Closing remarks

Theorem 3.3 gives equivalent conditions for a Toeplitz graph Gnt1,t2,,tkG_{n}\langle t_{1},t_{2},\ldots,t_{k}\rangle with t1<<tkt_{1}<\cdots<t_{k} and ntk1+tkn\geq t_{k-1}+t_{k} being chordal. In the proof of Theorem 3.3, we did not use the condition ntk1+tkn\geq t_{k-1}+t_{k} when we showed (iii)(iv)\mbox{(iii)}\Leftrightarrow\mbox{(iv)}, (iii)(i)\mbox{(iii)}\Rightarrow\mbox{(i)}, and (i)(ii)\mbox{(i)}\Rightarrow\mbox{(ii)}.

Question 1. Does the statement (ii) imply the statement (i) in Theorem 3.3 without the condition ntk1+tkn\geq t_{k-1}+t_{k}?

Theorem 3.7 gives equivalent conditions for a Toeplitz graph Gnt1,t2G_{n}\langle t_{1},t_{2}\rangle being perfect for any positive integers nn and t1<t2t_{1}<t_{2}.

Question 2. Is there any meaningful characterization for (weakly) perfect Toeplitz graphs Gnt1,t2,,tkG_{n}\langle t_{1},t_{2},\ldots,t_{k}\rangle with k3k\geq 3?

Acknowledgements

The authors thank anonymous referees for his/her helpful comments to improve the paper (in particular, the proof of Lemma 3.2 and Section 4). This work was supported by the National Research Foundation of Korea(NRF) grant funded by the Korea government(MSIP) (2016R1A5A1008055, NRF-2017R1E1A1A03070489, 2021R1C1C2014187), the Ministry of Education (NRF-2016R1A6A3A11930452).

References

  • [1] J. Chen, G. Chang, K. Huang, Integral distance graphs, J. Graph Theory 25 (1997) 287–294.
  • [2] M. Chudnovsky, N. Robertson, P. Seymour, R. Thomas, The strong perfect graph theorem, Annals of Mathematics 164 (2006) 51–229.
  • [3] R. Euler, Characterizing bipartite Toeplitz graphs, Theor. Comput. Sci 263 (2001) 47–58.
  • [4] R. Euler, Coloring planar Toeplitz graphs and the stable set polytope, Discret. Math. 276 (2004) 183–200.
  • [5] R. Euler, T. Zamfirescu, On planar Toeplitz graphs, Graphs and Combinatorics 29 (2013) 1311–1327.
  • [6] R. Euler, H. Le Verge, T. Zamfirescu, A characterization of infinite, bipartite Toeplitz graphs, In: Ku Tung-Hsin (ed.) Combinatorics and Graph Theory’95, vol. 1, pp. 119–130. Academia Sinica, World Scientific, Singapore (1995).
  • [7] I. Gohberg (ed.), Toeplitz centennial, Birkhäuser, Boston, 1982.
  • [8] G. Heinig, K. Rost, Algebraic methods for Toeplitz-like matrices and operators, Birkhäuser, Boston, 1984.
  • [9] C. Heuberger, On hamiltonian Toeplitz graphs, Discret. Math. 245 (2002) 107–125.
  • [10] C. Heuberger, On planarity and colorability of circulant graphs, Discret. Math. 268 (2003) 153–169.
  • [11] S. Hougardy, Classes of perfect graphs, Discret. Math. 306 (2006) 2529–2571.
  • [12] A. Ilić, M. Bašić, On the chromatic number of integral circulant graphs, Computers and Mathematics with Applications 60 (2010) 144–150.
  • [13] I. S. Iohvidov, Hankel and Toeplitz matrices and forms algebraic theory, Birkhäuser, Boston, 1982.
  • [14] A. Kemnitz, H. Kolberg, Coloring of integer distance graphs, Discret. Math. 191 (1998) 113–123.
  • [15] A. Kemnitz, M. Marangio, Chromatic numbers of integer distance graphs, Discret. Math. 233 (2001) 239–246.
  • [16] M. Klin, V. Liskovets, R. Pöschel, Analytical enumeration of circulant graphs with prime-squared number of vertices, Sém. Lotharing Combin. 36(B36d) (1996) 1–36.
  • [17] C. G. Lekkerkerker and J. Ch. Boland, Representation of a finite graph by a set of intervals on the real line, Fund. Math. 51 (1962) 45–64.
  • [18] L. Lovász, A characterization of perfect graphs, J. Combin. Theory 13 (1972) 95–98.
  • [19] S. Malik, A. M. Qureshi, Hamiltonian cycles in directed Toeplitz graphs, Ars Combin. 109 (2013) 511–526.
  • [20] S. Malik, T. Zamfirescu, Hamiltonian connectedness in directed Toeplitz graphs, Bull. Math. Soc. Sci. Math. Roum. 53(2) (2010) 145–156.
  • [21] M. Muzychuk, A solution of the isomorphism problem for circulant graphs, Proc. London Math. Soc. 88 (2004) 1–41.
  • [22] P. T. Meijer, Connectivities and diameters of circulant graphs, Master Thesis, Simon Fraster University, 1987.
  • [23] S. Nicoloso, U. Pietropaoli, On the chromatic number of Toeplitz graphs, Discret. Appl. Math. 164 (2014) 286–296.
  • [24] R. van Dal, G. Tijssen, Z. Tuza, J. van der Veen, C. H. Zamfirescu, T. Zamfirescu, Hamiltonian properties of Toeplitz graphs, Discret. Math. 159 (1996) 69–81.
  • [25] A. Zhou, X. D. Zhang, Enumeration of circulant graphs with order nn and degree 4 or 5 [Chinese], Dianzi Keji Daxue Xuebao 25 (1996) 272–276.