Clique roots of K 4-free chordal graphs

The clique polynomial C(G, x) of a finite, simple and undirected graph G = (V,E) is defined as the ordinary generating function of the number of complete subgraphs of G. A real root of C(G, x) is called a clique root of the graph G. Hajiabolhasan and Mehrabadi showed that every simple graph G has at least a clique root in the interval [−1, 0). Moreover, they showed that the class of triangle-free graphs has only clique roots. In this paper, we extend their result by showing that the class of K4-free chordal graphs has also only clique roots. In particular, we show that this class always has a clique root −1. We conclude our paper with some interesting open questions and conjectures.


Introduction and Motivation
Throughout this paper graphs are finite, simple and undirected.For the terminology and notations which are not defined here, we refer the readers to the book [1].For a given graph G = (V, E), a complete subgraph of G on k vertices is called a k-clique.For a subset U ⊆ V (G), the subgraph induced on U will be denoted by G[U ].We recall that an edge which joins two vertices of a cycle but is not itself an edge of the cycle is called a chord of the cycle.A graph is chordal if each cycle of length at least four has a chord.We also recall that the clique polynomial [2] of the graph G, www.ejgta.orgClique roots of K 4 -free chordal graphs | Hossein Teimoori Faal denote it by C(G, x) is defined, as follows or equivalently in which c k (G) is the number of k-cliques in G and ω(G) is the size of the largest clique of G.
From now on, any real root of a clique polynomial C(G, x) will be called the clique root of G.
Hajiabolhasan and Mehrabadi [2] showed that the clique polynomial of every simple graph always has a clique root in the interval [−1, 0).They also showed that the class of triangle-free graphs has only clique roots.Polynomials with only real roots arise often in graph theory and other branches of mathematics.Such polynomials with positive coefficients have attracted the attention of many researchers because of their implication on unimodality and log-concavity.
Our main goal here is to contribute in this active line of research by extending the above result for the class of K 4 -free chordal graphs.More precisely, we will prove the following.
Theorem 1.1.The class of K 4 -free connected chordal graphs has only clique roots.In particular, this class always has a clique root −1.
We will also give the following immediate corollary of the above theorem, which is indeed a new algebraic proof of Turan's graph theorem for planar K 4 -free graphs.
Corollary 1.1.If G is a K 4 -free connected graph with n vertices and m edges, then we have m ≤ n 2 3 .

Chordal Graphs and Clique Polynomials
In this section, we investigate the important class of chordal graphs.This class of graphs is very important in computer science, specially from the computational complexity view point.Many hard problems in general graphs have easy solutions in the class of chordal graphs.As we will see, the clique polynomial of chordal graphs can give us important insights into the structure of these graphs.Definition 2.1.A graph is chordal if every cycle of length greater than three has a chord.A vertex of a graph is simplicial if its neighbors induces a clique in the graph.
One of the important properties of a chordal graph is that it always has a clique decomposition.For the sake completeness, here we quickly review the idea of decomposing a chordal graph into cliques.For detailed information, one can refer to [3].Definition 2.2.For given graphs G 1 and G 2 , we say that a graph G arises from G 1 and G 2 by pasting along S if we have G 1 ∪ G 2 = G and G 1 ∩ G 2 = S.In this case, the graphs G i are called the simplicial summands of G.
Remark 2.1.From the above definition, it is clear that a graph is chordal if it can be constructed recursively by pasting complete graphs along cliques.It is not hard to see that this process is independent of the order in which complete graphs paste to each other.Indeed, this recursive construction gives us a clique decomposition of chordal graphs which is essential to obtain their clique polynomials.
For simplicity of arguments, we use the notation G 1 ∪ S G 2 whenever G 1 and G 2 are pasted along S. The following lemma is key to obtain an explicit formula for the clique polynomial of chordal graphs.The proof is straight forward application of the inclusion-exclusion principle and left to the reader as a simple exercise.
Proposition 2.1.Let G 1 and G 2 be two simple graphs and G = G 1 ∪ Q G 2 be their pasting along an i -clique Q.Then, we have By the successive application of the formula (3) and the recursive construction of chordal graphs, we can obtain the following explicit formula for the clique polynomial of chordal graphs.
Theorem 2.1.Let G be a chordal graph defined as a pasting of the complete graphs j=1 are cliques of sizes l j 's, respectively.Then, we have As an immediate consequence of the above theorem, we have the following interesting result.
Corollary 2.1.Every chordal graph G without isolated vertices always has a clique root −1.The multiplicity of this root is equal to the size of the smallest clique in the pasting process of the recursive construction of G.
Remark 2.2.It is worth to note that the notation G 1 ∪ Q G 2 is also called the clique-sum of two graphs G 1 and G 2 .Indeed, based on the formula 3, the clique-sum of graphs preserve the property of having (at least) a clique root −1.Unfortunately, the (ordinary) sum of two graphs which is also called the (disjoint) union of two graphs has not this property.For instance, the graph G obtained by the sum of a 1-clique (an isolated vertex) and a 2-clique (an edge) with the clique polynomial C(G, x) = 1 + 3x + x 2 has no clique root −1.Therefore, the clique polynomial of a graph is not additive over it's connected components.Thus, from now on, we will concentrate on connected chordal graphs.

Proofs of the Main Results
Here, we first give a proof of the following proposition which is a weaker version of Theorem 1.1.From now on, we will assume that our graphs are connected.Before proceeding the proof, we need to have a quick review of breadth-first search (BFS, for short) form the theory of graph algorithms.
Recall that for a given graph G = (V, E) and a root (node) r ∈ V , the breadth-first search produces a search-tree T by exploring first the neighbors of r, then the neighbors of it's children.A tree obtained by running a breadth-first search is called a breadth-first search tree (BFS-tree, for short) rooted at r.It can be easily shown that when G is a connected graph, then the BFS-tree of G is indeed a spanning tree of G rooted at r.
A nice property of a breadth-first search tree is that it can give the distance from the root r to all other vertices.Therefore we just have to have a value l(v) to every vertex called level of v which corresponds to distT (r, v) (the length of the unique path form the root r to the vertex v in the BFS-tree T ).Moreover, for any non-tree edge e ∈ E(G) (an edge e which is not in E(T )), it's end vertices can only lie on the same level or the consecutive levels.
Proposition 3.1.The class of K 4 -free planar chordal graphs has only clique roots.In particular, −1 is always a clique root.
Proof.For a given K 4 -free planar graph G, by Euler Formula, we have where n, m and f are the number of vertices, edges and faces of a planar embedding of G, respectively.Moreover, if G is a chordal graph, then we observe that where t is the number of triangles ( triangular faces ) of G. Hence, form (5) and (6), we conclude that n − m + t = 1, By last identity and considering the fact that the clique polynomial of a That is, every K 4 -free planar chordal graph G always has −1 as a clique root.Therefore, we obtain the following multiplicative decomposition of C(G, x) The final step of the proof is to show that the quadratic polynomial: always has a real root.To this end, we actually prove that Q(G, x) is a clique polynomial of a triangle-free graph G which can be obtained from the original graph G based on the idea of the BFS-tree of G.
For a given K 4 -free chordal graph G, pick up an arbitrary vertex r ∈ V (G).Now, we construct a BFS-tree of G rooted at the vertex r.We will denote it by T G .Clearly, this tree has n vertices and n − 1 edges.Now in the graph G, delete all n − 1 edges of the tree T G and call the resulting graph Ĝ.This graph has clearly n vertices and m − (n − 1) edges.By the construction, it is clear that v is an isolated vertex of Ĝ. Next, we prove that Ĝ is a triangle -free graph.
Here is the argument.In contrary, let ∆ = abc be a triangle in Ĝ.Then, the end vertices of it's non-tree edges can only lie on the same level or the consecutive levels.Hence, at least two vertices of ∆ (let say e = ab) lies in the same level.For simplicity of arguments, we will assume that the vertex a appears in BFS-tree before the vertex b.For a given vertex v ∈ V , the unique (shortest) paths form the root r to the vertex v is denoted by path(r, v).Now, by the construction of the BFS-tree of G, there is a vertex c (in the same level as c) such that c a, c b ∈ E(T ).Now, let r be the last intersection point of the paths (starting from the root r) path(r, c ) and path(r, c).By the construction of BFS-tree and chordality of the graph G, since the cycle C based on the four vertices r , c, b, c is of length greater than 3, we conclude that c c ∈ E(G).Thus, it implies that the graph G contains a 4-clique abcc .This contradicts the fact that the graph G is K 4 -free.This completes our argument by the contradiction.
Thus, the triangle-free graph G obtained from Ĝ by deleting the vertex v has n − 1 vertices and m − n + 1 edges, as required.
Remark 3.1.The following figure illustrates the process of obtaining the triangle-free graph G from the original graph G for a sample Now, we are ready to give a proof of our main theorem.
Proof of Theorem 1.1.We first note that by Corollary 2.1, every connected chordal graph always has a clique root −1.Now, the rest of the proof is similar to the proof of Proposition 3.1.
Next we give an algebraic proof of the Turan's Graph Theorem [4] for K 4 -free graphs which is indeed Corollary 1.1.
Proof of Corollary 1.1.We first note that since we want to prove an upper bound for the maximum possible number of edges, without loss of generality we can assume that the graph G is chordal.By the proof of Theorem 1.1, we already know that if G is a given K 4 -free chordal graph with n vertices and m edges, then the following quadratic equation has only real zeros.Hence, it's discriminant is nonnegative and therefore we have the following inequality: which is equivalent to On the other hand, we have the inequality which is equivalent to the obvious inequality (n − 3) 2 ≥ 0. Thus, the inequalities (10) and (11) immediately imply the Turan's inequality for K 4 -free graphs.

Open problems and questions
We already showed that the class of K 4 -free connected chordal graphs has only clique roots.Now, one might ask whether the class of K 5 -free connected chordal graphs has the same property or not.Unfortunately, this is not true in general.For example, the connected graph K + 4 ( a complete graph K 4 plus one edge ) has only two clique roots.Indeed, we have Since the cubic polynomial φ(x) = 1 + 4x + 3x 2 + x 3 has the first derivative φ(x) = 3(x + 1) 2 + 1 which is always positive, by the first derivative criteria, φ(x) = 1 + 4x + 3x 2 + x 3 has exactly one real root.Thus, we come up with the following first open question.
Problem 1. Which subclasses of K 5 -free chordal graphs have only clique roots?
Recall that the class of 3 -trees are those graphs which can be constructed recursively by starting with a complete graph K 4 , and then repeatedly adding vertices in such a way that each added vertex has exactly three neighbors that form a clique (triangle).By the above definition, it is not hard to see that the class of 3-trees is a subclass of K 5 -free chordal graphs.Next, we come up with the following conjecture.We call an i-clique an isolated clique if the intersection of all neighborhoods of it's vertices is an empty set.In particular, an isolated 1-clique is called an isolated vertex.The interesting point is that we strongly believe that the following much stronger conjecture is true.
Conjecture 3. The class of K ω+3 -free graphs without isolated ω-cliques has only clique roots.

Figure 1 .
Figure 1.The construction of the triangle-free graph G.

Conjecture 1 .Conjecture 2 .Problem 2 .
The class of 3-trees has only clique roots.www.ejgta.orgCliqueroots of K 4 -free chordal graphs | Hossein Teimoori Faal Considering the fact that any connected chordal graph has a clique root −1 and the recursive definition of of chordal graphs, we made the following stronger conjecture.The class of K 5 -free connected chordal graphs with a clique root −1 of multiplicity 2 has only clique roots.Considering the above discussions, we also come up with the following open question.Which subclasses of K ω+3 -free connected graphs (ω > 0) have only clique roots?