On $Q$-Polynomial distance-regular graphs with a linear dependency involving a $3$-clique

Abstract.

Let $\Gamma$ denote a distance-regular graph with diameter $D\geq 2$. Let $E$ denote a primitive idempotent of $\Gamma$ with respect to which $\Gamma$ is $Q$-polynomial. Assume that there exists a $3$-clique $x,y,z$ such that $E\hat{x},E\hat{y},E\hat{z}$ are linearly dependent. In this paper, we classify all the $Q$-polynomial distance-regular graphs $\Gamma$ with the above property. We describe these graphs from multiple points of view.

1. Introduction

This paper is about a certain kind of finite undirected graph, said to be distance-regular [1, § 4.1(A)], [5, § 2]. There is a well known property for a distance-regular graph, called the $Q$-polynomial property [1, § 4.1(E)], [5, § 11]. In this paper, we classify a certain type of $Q$-polynomial distance-regular graph. In our treatment, the following concepts will be relevant. We will consider the concepts of classical parameters [5, § 18], negative type [9], the cosine sequence [5, § 4], and regular near $2D$-gons [1, § 6.4]. Before we state our main results, we give some background about these concepts.

For the rest of this section, let $\Gamma$ denote a distance-regular graph with diameter $D\geq 2$. The concept of classical parameters was introduced in [1, § 6.1]. If $\Gamma$ has classical parameters, then the intersection numbers of $\Gamma$ are given by attractive formulas in terms of four parameters $(D,b,\alpha,\sigma)$ [1, § 6.1(1a,1b)]. See [1, § 6], [2, § 3.1.1], [5, § 18] for some results about classical parameters.

Assume that $\Gamma$ has classical parameters $(D,b,\alpha,\sigma)$. It is known that $b$ is an integer not equal to $0$ or $-1$ (cf. [1, Proposition 6.2.1]). The graph $\Gamma$ is said to have negative type whenever $b<-1$ [9, § 1]. Distance-regular graphs with negative type are investigated in [2, § 5.2], [9], [10].

Associated with each eigenvalue of $\Gamma$, there is a sequence of scalars $\{\sigma_{i}\}_{i=0}^{D}$ called the cosine sequence. For $0\leq i\leq D$, the scalar $\sigma_{i}$ can be interpreted as an angle cosine, see (5) below. It is known that the cosine sequence satisfies a $3$-term recurrence, see [5, § 14] and (2), (3), (4) below. See [1, § 8.1], [2, § 2.5], [5, § 4] for some results about the cosine sequence.

A regular near $2D$-gon is a type of distance-regular graph with an attractive geometric structure [1, § 6.4]. The regular near $2D$-gons were first introduced in [4]. An example of the regular near $2D$-gons are the dual polar graphs [1, § 9.4]. See [3] for a detailed study of the regular near $2D$-gons that are $Q$-polynomial. Other studies of the regular near $2D$-gons can be found in [1, § 6.6], [2, § 9.6], [8].

Paul Terwilliger has stated the following problem (cf. [7, Problem 1]).

Problem 1.1.

Let $\Gamma$ denote a distance-regular graph with diameter $D\geq 2$. Let $E$ denote a primitive idempotent of $\Gamma$ with respect to which $\Gamma$ is $Q$-polynomial. Assume that there exists a $3$-clique $x,y,z$ such that $E\hat{x},E\hat{y},E\hat{z}$ are linearly dependent. Investigate the combinatorial meaning of this condition.

In this paper, we investigate the $Q$-polynomial distance-regular graphs with the given property. The following is our main result.

Theorem 1.2.

Let $\Gamma$ denote a distance-regular graph with diameter $D\geq 2$. Let $E$ denote a primitive idempotent of $\Gamma$ with respect to which $\Gamma$ is $Q$-polynomial. Then the following are equivalent:

1. There exists a $3$-clique $x,y,z$ such that $E\hat{x},E\hat{y},E\hat{z}$ are linearly dependent;

2. The graph $\Gamma$ has classical parameters $(D,b,\alpha,\sigma)=(D,-2,\alpha,2+\alpha-\alpha[_{1}^{D}])$ and $E$ is for the eigenvalue $\frac{b_{1}}{b}-1$;

3. The graph $\Gamma$ is a regular near $2D$-gon of order $(2,t)$ and $E$ is for the eigenvalue $-t-1$;

4. The intersection number $a_{1}=1$ and for every $3$-clique $x,y,z$ we have $E\hat{x}+E\hat{y}+E\hat{z}=0$;

5. The graph $\Gamma$ is listed below and $E$ is for the minimal eigenvalue of $\Gamma$:

   • The unique regular near $4$-gon of order $(2,1)$,

   • The unique regular near $4$-gon of order $(2,2)$,

   • The unique regular near $6$-gon of order $(2,8)$,

   • The unique regular near $6$-gon of order $(2,11)$,

   • The unique regular near $6$-gon of order $(2,14)$,

   • The dual polar graph $A_{2D-1}(2)$;

6. The cosine sequence $\{\sigma_{i}\}_{i=0}^{D}$ for $E$ satisfies $\sigma_{i}=(-\frac{1}{2})^{i}$, where $0\leq i\leq D$.

This paper is organized as follows. In Section 2, we give some basic facts that will be used to state and prove our main results. In Section 3, first we state and prove a sequence of lemmas and propositions. Next we use these results to prove Theorem 1.2.

2. Preliminaries

From now on, let $\Gamma$ be a connected graph with vertex set $X$ and diameter $D\geq 2$. For $x,y\in X$, let $d(x,y)$ denote the path-length distance between $x$ and $y$. Pick $x,y\in X$ and write $d(x,y)=i$. Let $b_{i}$ denote the number of neighbors of $x$ at distance $i+1$ from $y$, $a_{i}$ denote the number of neighbors of $x$ at distance $i$ from $y$, and $c_{i}$ denote the number of neighbors of $x$ at distance $i-1$ from $y$. The graph $\Gamma$ is called distance-regular whenever $a_{i}$, $b_{i}$, and $c_{i}$ are independent of $x,y$ and depend only on $i$. For the rest of this paper, assume that $\Gamma$ is distance-regular. Note that $\Gamma$ is regular with valency $k=b_{0}$, and that

(1) $$k=a_{i}+b_{i}+c_{i}\;\;\;\;(0\leq i\leq D),$$

where $b_{D}=0$ and $c_{0}=0$. The sequence

$$\{b_{0},b_{1},\ldots,b_{D-1};c_{1},c_{2},\ldots,c_{D}\}$$

is called the intersection array of $\Gamma$. Pick $x\in X$. For $0\leq i\leq D$, let $k_{i}$ denote the number of vertices in $X$ at distance $i$ from $x$. Note that $k_{0}=1$ and $k_{1}=k$. By a routine counting argument, we find $k_{i}c_{i}=k_{i-1}b_{i-1}$ for $1\leq i\leq D$. It follows that $k_{i}$ is independent of choice of $x$.

Let $V=\mathbb{R}^{|X|}$ denote the vector space over $\mathbb{R}$, consisting of the column vectors with coordinates indexed by $X$ and all entries in $\mathbb{R}$. We endow $V$ with a bilinear form $\langle~{},~{}\rangle$ that satisfies $\langle u,v\rangle=u^{t}v$ for $u,v\in V$, where $t$ denotes transpose. We abbreviate $\langle u,u\rangle$ by $\|u\|^{2}$. We note that $\|u\|^{2}\geq 0$, with equality if and only if $u=0$. For $x\in X$, let $\hat{x}$ denote a vector in $V$ that has $x$-coordinate $1$ and all other coordinates $0$. Observe that the vectors $\{\hat{x}\mid x\in X\}$ form an orthonormal basis for $V$.

Let $Mat_{X}(\mathbb{R})$ denote the $\mathbb{R}$-algebra consisting of the matrices with rows and columns indexed by $X$ and all entries in $\mathbb{R}$. Let $A\in Mat_{X}(\mathbb{R})$ denote the adjacency matrix of $\Gamma$. Then the Bose-Mesner algebra of $\Gamma$ is the subalgebra of $Mat_{X}(\mathbb{R})$ generated by $A$. The Bose-Mesner algebra of $\Gamma$ has a basis $\{E_{i}\}_{i=0}^{D}$ such that $E_{0}=|X|^{-1}J$, $E_{i}E_{j}=\delta_{i,j}E_{i}$ ($0\leq i,j\leq D$), and $\sum_{i=0}^{D}E_{i}=I$, where $I$ is the identity matrix, and $J$ is the all one matrix (cf. [5, p. 4]). Following [5], we call $\{E_{i}\}_{i=0}^{D}$ the primitive idempotents of $\Gamma$. The primitive idempotent $E_{0}$ is called trivial. For $B,C\in Mat_{X}(\mathbb{R})$, define the matrix $B\circ C\in Mat_{X}(\mathbb{R})$ with entries

$$(B\circ C)_{y,z}=B_{y,z}C_{y,z}\;\;\;\;(y,z\in X).$$

The operation $\circ$ is called entrywise multiplication. Recall that the Bose-Mesner algebra of $\Gamma$ is closed under entrywise multiplication (see [5, § 5]). By [5, Eqn.(8)], there exist real numbers $q_{i,j}^{h}$ ($0\leq h,i,j\leq D$) such that

$$E_{i}\circ E_{j}=\frac{1}{|X|}\sum_{h=0}^{D}q_{i,j}^{h}E_{h}\;\;\;\;\;(0\leq i,j\leq D).$$

The parameters $q_{i,j}^{h}$ are called the Krein parameters. These parameters are nonnegative and this property is called Krein condition [5, § 5]. Because $\{E_{i}\}_{i=0}^{D}$ is a basis for the Bose-Mesner algebra of $\Gamma$, there exist real numbers $\{\theta_{i}\}_{i=0}^{D}$ such that

$$A=\sum_{i=0}^{D}\theta_{i}E_{i}.$$

The scalars $\theta_{i}$ $(0\leq i\leq D)$ are mutually distinct (see [1, § 4.1(B)]). For $0\leq i\leq D$, we have $AE_{i}=\theta_{i}E_{i}$. Therefore $\theta_{i}$ is an eigenvalue of $A$, and $E_{i}V$ is the corresponding eigenspace. By the eigenvalues of $\Gamma$, we mean the scalars $\theta_{i}$ $(0\leq i\leq D)$. Let $m_{i}$ denote the dimension of $E_{i}V$. Then $\|E_{i}\hat{x}\|^{2}=|X|^{-1}m_{i}$ for all $x\in X$ (cf. [5, Lemma 4.1(ii)]).

Let $E$ denote a primitive idempotent of $\Gamma$, and let $\theta$ denote the corresponding eigenvalue. We define a sequence of scalars $\{\sigma_{i}\}_{i=0}^{D}$ such that

(2) $$\sigma_{0}=1,\qquad\sigma_{1}=\frac{\theta}{k},$$

(3) $$\theta\sigma_{i}=c_{i}\sigma_{i-1}+a_{i}\sigma_{i}+b_{i}\sigma_{i+1}\;\;\;\;1\leq i\leq D-1.$$

By [5, Lemma 4.8], we have

(4) $$\;\;c_{D}\sigma_{D-1}+a_{D}\sigma_{D}=\theta\sigma_{D}.$$

The sequence $\{\sigma_{i}\}_{i=0}^{D}$ is called the cosine sequence for $E$ (or $\theta$). This name is motivated by the following result. Pick $x,y\in X$ and write $d(x,y)=i$. By [5, Lemma 4.1(iii)] we have

(5) $$\sigma_{i}=\frac{\langle E\hat{x},E\hat{y}\rangle}{\|E\hat{x}\|\|E\hat{y}\|}.$$

Observe that $\sigma_{i}$ is the cosine of the angle between $E\hat{x}$ and $E\hat{y}$.

The graph $\Gamma$ is called $Q$-polynomial (with respect to the given ordering $\{E_{i}\}_{i=0}^{D}$ of the primitive idempotents) whenever the following holds for $0\leq h,i,j\leq D$ (cf. [5, Definition 11.1]):

1. $q_{i,j}^{h}=0$ if one of $h,i,j$ is greater than the sum of the other two,

2. $q_{i,j}^{h}\neq 0$ if one of $h,i,j$ is equal to the sum of the other two.

In this case, we say that $\Gamma$ is $Q$-polynomial with respect to $E$, where $E=E_{1}$. Recall that if $\Gamma$ is $Q$-polynomial with respect to $E$, then the elements of $\{\sigma_{i}\}_{i=0}^{D}$ are mutually distinct (cf. [1, Proposition 8.1.3]).

Recall that our distance-regular graph $\Gamma$ is said to have classical parameters $(D,b,\alpha,\sigma)$ whenever the intersection array satisfies

(6) $$c_{i}=[^{i}_{1}](1+\alpha[^{i-1}_{1}])\;\;\;\;\;(0\leq i\leq D),$$

(7) $$b_{i}=([^{D}_{1}]-[^{i}_{1}])(\sigma-\alpha[^{i}_{1}])\;\;\;\;\;(0\leq i\leq D),$$

where $[^{i}_{1}]=[^{i}_{1}]_{b}=1+b+\cdots+b^{i-1}$ for $1\leq i\leq D$. Note that by a convention in [1, § 6.1(2)], we have $[^{0}_{1}]=0$. Recall that if $\Gamma$ has classical parameters, then $\Gamma$ is $Q$-polynomial with respect to the following ordering of the eigenvalues of $\Gamma$ (cf. [5, Theorem 18.2, Lemma 18.3]):

(8) $$\theta_{i}=\frac{b_{i}}{b^{i}}-[^{i}_{1}]\;\;\;\;\;(0\leq i\leq D).$$

3. Main result

In this section, we concentrate on Problem 1.1.

Lemma 3.1.

Let $\Gamma$ denote a distance-regular graph with diameter $D\geq 2$. Let $E$ denote a nontrivial primitive idempotent of $\Gamma$. Let $x,y,z$ denote a $3$-clique in $\Gamma$. Then the following items hold.

1. The matrix of inner products of $E\hat{x},E\hat{y},E\hat{z}$ is $|X|^{-1}mC$, where $m$ is the rank of $E$ and

   $$C=\left[\begin{array}[]{ccc}\sigma_{0}&\sigma_{1}&\sigma_{1}\\ \sigma_{1}&\sigma_{0}&\sigma_{1}\\ \sigma_{1}&\sigma_{1}&\sigma_{0}\\ \end{array}\right].$$

2. The eigenvalues of $C$ are $1-\sigma_{1}$, $1-\sigma_{1}$, and $1+2\sigma_{1}$.

3. We have $1>\sigma_{1}\geq-\frac{1}{2}$.

4. $\sigma_{1}=-\frac{1}{2}$ if and only if $E\hat{x},E\hat{y},E\hat{z}$ are linearly dependent. In this case, $E\hat{x}+E\hat{y}+E\hat{z}=0$.

Proof.

(i): By (5).
(ii): Use linear algebra and recall that $\sigma_{0}=1$.
(iii): By construction $C$ is positive semidefinite, so its eigenvalues are nonnegative. We have $\sigma_{1}\neq 1$ since $E$ is nontrivial.
(iv): By linear algebra, we have $\sigma_{1}=-\frac{1}{2}$ if and only if $0$ is an eigenvalue of $C$ if and only if $C$ is singular if and only if $E\hat{x},E\hat{y},E\hat{z}$ are linearly dependent. Assume that this is the case. Then

$$\|E\hat{x}+E\hat{y}+E\hat{z}\|^{2}=3|X|^{-1}m(1+2\sigma_{1})=0.$$

It follows that $E\hat{x}+E\hat{y}+E\hat{z}=0$. ∎

Lemma 3.2.

Let $\Gamma$ denote a distance-regular graph with diameter $D\geq 2$. Let $E$ denote a primitive idempotent of $\Gamma$ with respect to which $\Gamma$ is $Q$-polynomial. Assume that there exists a $3$-clique $x,y,z$ such that $E\hat{x},E\hat{y},E\hat{z}$ are linearly dependent. Then:

1. For $E$ the corresponding eigenvalue $\theta$ is equal to $-\frac{k}{2}$, and this is the minimal eigenvalue of $\Gamma$.

2. $E\hat{x}+E\hat{y}+E\hat{z}=0$ for all $3$-cliques $x,y,z$.

3. $\sigma_{i}=(-\frac{1}{2})^{i}$ for $0\leq i\leq D$.

4. $a_{i}=c_{i}$ for $0\leq i\leq D$.

5. $a_{1}=1$.

Proof.

(i): We have $\theta=-\frac{k}{2}$ because $\theta=k\sigma_{1}$ and $\sigma_{1}=-\frac{1}{2}$ by Lemma 3.1(iv). Moreover, $-\frac{k}{2}$ is the minimal eigenvalue of $\Gamma$ by Lemma 3.1(iii).
(ii): Immediate from Lemma 3.1(iv).
(iii): We use induction on $i$. The result holds for $i=0$ since $\sigma_{0}=1$. The result holds for $i=1$ by (i) and (2). For the rest of this proof, assume that $i\geq 2$. By induction, we assume that $\sigma_{j}=(-\frac{1}{2})^{j}$ for $0\leq j\leq i-1$. We show that $\sigma_{i}=(-\frac{1}{2})^{i}$. We have $E\hat{x}+E\hat{y}+E\hat{z}=0$ by (ii). Let $w\in X$ be a vertex at distance $i$ from $x$ and $i-1$ from $y$. Define $r=d(z,w)$. We show that $r=i$. By the triangle inequality, $r=i-1$ or $r=i$. Taking the inner product of $E\hat{w}$ and $E\hat{x}+E\hat{y}+E\hat{z}$, we find $\sigma_{i-1}+\sigma_{i}+\sigma_{r}=0$ by (5). Suppose that $r=i-1$. Then $0=2\sigma_{i-1}+\sigma_{i}$. This implies that $\sigma_{i}=-2\sigma_{i-1}=\sigma_{i-2}$, for a contradiction. Therefore $r=i$. We have $0=\sigma_{i-1}+2\sigma_{i}$. This implies that $\sigma_{i}=-\frac{1}{2}\sigma_{i-1}=(-\frac{1}{2})^{i}$ as desired.
(iv): If we substitute the data of (i) and (iii) in (3), then we have $k=4c_{i}-2a_{i}+b_{i}$, where $1\leq i\leq D-1$. This implies that $a_{i}=c_{i}$ using (1). Moreover, we have $4c_{D}-2a_{D}=k$ by (4) and the mentioned substitution. By this and $a_{D}+c_{D}=k$ we obtain $a_{D}=c_{D}$.
(v): Immediate from (iv) since $c_{1}=1$. ∎

Definition 3.3.

Let $\Gamma$ be a distance-regular graph with diameter $D\geq 2$. For $x\in X$, let $\Gamma(x)$ denote the set of neighbors of $x$ in $\Gamma$. The induced subgraph on $\Gamma(x)$ is called the first subconstituent or local graph of $\Gamma$ with respect to $x$. If $\Gamma(x)$ is a disjoint union of cliques, then each clique has size $a_{1}+1$ and there are $\frac{k}{a_{1}+1}$ cliques. $\Gamma$ is said to be locally a disjoint union of cliques whenever $\Gamma(x)$ is a disjoint union of cliques for all $x\in X$. Assume that $\Gamma$ is locally a disjoint union of cliques. Then $\Gamma$ is said to have order $(s,t)$, where $s=a_{1}+1$ and $t+1=\frac{k}{a_{1}+1}$. If $s\geq 2$, then $\Gamma$ is called thick (cf. [2, p. 35]).

Definition 3.4.

Let $\Gamma$ be a distance-regular graph with diameter $D\geq 2$. Then $\Gamma$ is called regular near $2D$-gon whenever $\Gamma$ is locally a disjoint union of cliques and $a_{i}=a_{1}c_{i}$ for $1\leq i\leq D$ (see [2, p. 35]).

Proposition 3.5.

Let $\Gamma$ denote a distance-regular graph with diameter $D\geq 2$. Let $E$ denote a primitive idempotent of $\Gamma$ with respect to which $\Gamma$ is $Q$-polynomial. Assume that there exists a $3$-clique $x,y,z$ such that $E\hat{x},E\hat{y},E\hat{z}$ are linearly dependent. Then $\Gamma$ is regular near $2D$-gon.

Proof.

We have $a_{1}=1$ by Lemma 3.2(v). Therefore $\Gamma$ is locally a disjoint union of $2$-cliques. Moreover, $a_{i}=c_{i}$ for $0\leq i\leq D$ by Lemma 3.2(iv). This completes the proof. ∎

Definition 3.6.

(cf. [6]) Let $\Gamma$ be a distance-regular graph with diameter $D\geq 2$. For $2\leq i\leq D$, a kite of length $i$ is $4$-tuple $xyzw$ of vertices of $\Gamma$ such that $x,y,z$ are mutually adjacent and $w$ is at distance $d(x,w)=i$, $d(y,w)=i-1$, and $d(z,w)=i-1$.

Remark 3.7.

Let $\Gamma$ be a distance-regular graph with diameter $D\geq 2$. If $\Gamma$ is kite-free, then it has no kite of length $2$ and therefore $\Gamma$ is locally a disjoint union of cliques. Let $\Gamma$ be regular near $2D$-gon. Then every $3$-clique lies in a unique maximal clique in $\Gamma$. Furthermore, for a given $x\in X$ and maximal clique $C$ of $\Gamma$, there is a unique vertex $y\in C$ that is closest to $x$ (cf. [1, § 6.4]). This implies that $\Gamma$ is kite-free. It follows that the distance-regular graph $\Gamma$ is regular near $2D$-gon if and only if $\Gamma$ is kite-free and $a_{i}=a_{1}c_{i}$ for $1\leq i\leq D$.

Lemma 3.8.

Let $\Gamma$ denote a distance-regular graph with diameter $D\geq 2$. Let $E$ denote a primitive idempotent of $\Gamma$ with respect to which $\Gamma$ is $Q$-polynomial. Assume that there exists a $3$-clique $x,y,z$ such that $E\hat{x},E\hat{y},E\hat{z}$ are linearly dependent. Then $\Gamma$ has classical parameters $(D,b,\alpha,\sigma)$, where

$$b=-2,\qquad\qquad\sigma=2+\alpha-\alpha[_{1}^{D}].$$

Moreover, $\alpha=-1-c_{2}$.

Proof.

Using Lemma 3.2(iii) and from [1, Theorem 8.4.1], $\Gamma$ has classical parameters $(D,b,\alpha,\sigma)$ with $b=-2$. This implies that $\alpha=-1-c_{2}$ by (6). We have $b_{0}=k$ and therefore $k=\sigma[_{1}^{D}]$ by (7). Moreover, $b_{1}=k-2$ by Lemma 3.2(v). By substituting $k=\sigma[_{1}^{D}]$ in $b_{1}=k-2$ and using (7), we have $\sigma=2+\alpha-\alpha[_{1}^{D}]$. This completes the proof. ∎

Lemma 3.9.

Let $\Gamma$ denote a distance-regular graph with diameter $D\geq 2$. Let $E$ denote a primitive idempotent of $\Gamma$ with respect to which $\Gamma$ is $Q$-polynomial. Assume that there exists a $3$-clique $x,y,z$ such that $E\hat{x},E\hat{y},E\hat{z}$ are linearly dependent. Then $1\leq c_{2}\leq 5$.

Proof.

By construction $c_{2}\geq 1$. We show that $c_{2}\leq 5$. Pick $x,y\in X$ with $d(x,y)=2$. Note that $|\Gamma(x)\cap\Gamma(y)|=c_{2}$. Also note that two distinct vertices in $\Gamma(x)\cap\Gamma(y)$ are at distance $2$, because $a_{1}=1$ by Lemma 3.2(v). Define

$$u=E\hat{x}+E\hat{y},$$

and

$$v=\sum_{z\in\Gamma(x)\cap\Gamma(y)}E\hat{z}.$$

By the Cauchy-Schwarz inequality,

(9) $$\langle u,v\rangle^{2}\leq\langle u,u\rangle\langle v,v\rangle.$$

Using the data in (5) and Lemma 3.2, we obtain

(10) $$\langle u,v\rangle=-c_{2}m|X|^{-1},$$

(11) $$\langle u,u\rangle=\frac{5m}{2}|X|^{-1},$$

(12) $$\langle v,v\rangle=\frac{(c_{2}^{2}+3c_{2})m}{4}|X|^{-1},$$

where $m$ denotes the rank of $E$. Evaluating (9) using (10), (11), and (12), we obtain $c_{2}(5-c_{2})\geq 0$. By this, we have $c_{2}\leq 5$. This completes the proof. ∎

Proposition 3.10.

Let $\Gamma$ denote a distance-regular graph with $D=2$. Let $E$ denote a primitive idempotent of $\Gamma$ with respect to which $\Gamma$ is $Q$-polynomial. Assume that there exists a $3$-clique $x,y,z$ such that $E\hat{x},E\hat{y},E\hat{z}$ are linearly dependent. Then $\Gamma$ is isomorphic to one of the following graphs.

• The unique regular near $4$-gon of order $(2,1)$,

• The unique regular near $4$-gon of order $(2,2)$,

• The unique regular near $4$-gon of order $(2,4)$.

Proof.

The graph $\Gamma$ is regular near $4$-gon by Proposition 3.5. Moreover, $a_{1}=1$ by Lemma 3.2(v). The regular near $4$-gons with $a_{1}=1$ are classified in [1, p. 30(Examples)]. The result follows from that classification. ∎

Proposition 3.11.

Let $\Gamma$ denote a distance-regular graph with diameter $D=3$. Let $E$ denote a primitive idempotent of $\Gamma$ with respect to which $\Gamma$ is $Q$-polynomial. Assume that there exists a $3$-clique $x,y,z$ such that $E\hat{x},E\hat{y},E\hat{z}$ are linearly dependent. Then $\Gamma$ is isomorphic to one of the following graphs.

• The unique regular near $6$-gon of order $(2,8)$,

• The unique regular near $6$-gon of order $(2,11)$,

• The unique regular near $6$-gon of order $(2,14)$,

• The dual polar graph $A_{5}(2)$.

Proof.

By Lemma 3.9 we have $1\leq c_{2}\leq 5$. For each choice of $c_{2}$, we compute the intersection array using Lemma 3.8 and (6),(7). The results are in the following table.

$c_{2}$	Intersection array
$1$	$\{18,16,16;1,1,9\}$
$2$	$\{24,22,20;1,2,12\}$
$3$	$\{30,28,24;1,3,15\}$
$4$	$\{36,34,28;1,4,18\}$
$5$	$\{42,40,32;1,5,21\}$

Assume that $c_{2}=1$. Then $\Gamma$ exists and is unique by [1, p. 427]. Assume that $c_{2}=2$. Then $\Gamma$ exists and is unique by [1, p. 427]. Assume that $c_{2}=3$. Then $\Gamma$ exists and is unique by [1, p. 428]. Assume that $c_{2}=4$. Then $\Gamma$ does not exist. Indeed the intersection array is not feasible for the following reason. By Lemma 3.2(i), $\theta=-18$ is an eigenvalue of $\Gamma$ and by the Bigg’s formula [2, Theorem 2.8] the multiplicity of $\theta$ is not integer. Using Lemma 3.2(iii) and the intersection array of $\Gamma$, the multiplicity of $\theta$ is

$$\frac{\sum_{i=0}^{3}k_{i}}{\sum_{i=0}^{3}k_{i}\sigma_{i}^{2}}=$$

$$\frac{819}{1+36(\frac{1}{4})+306(\frac{1}{16})+476(\frac{1}{64})}=22.4.$$

Assume that $c_{2}=5$. Then $\Gamma$ exists and is unique by [1, p. 428]. This completes the proof. ∎

Proposition 3.12.

Let $\Gamma$ denote a distance-regular graph with diameter $D\geq 4$. Let $E$ denote a primitive idempotent of $\Gamma$ with respect to which $\Gamma$ is $Q$-polynomial. Assume that there exists a $3$-clique $x,y,z$ such that $E\hat{x},E\hat{y},E\hat{z}$ are linearly dependent, and $c_{2}=1$. Then $\Gamma$ does not exist.

Proof.

By Lemma 3.8 with $c_{2}=1$, we find that $\Gamma$ has classical parameters $(D,-2,-2,2[_{1}^{D}])$. The graph $\Gamma$ does not exist by [3, Corollary 5.4]. ∎

Proposition 3.13.

Let $\Gamma$ denote a distance-regular graph with diameter $D\geq 4$ and $c_{2}\geq 2$. Let $E$ denote a primitive idempotent of $\Gamma$ with respect to which $\Gamma$ is $Q$-polynomial. Assume that there exists a $3$-clique $x,y,z$ such that $E\hat{x},E\hat{y},E\hat{z}$ are linearly dependent. Then $c_{2}=5$ and $\Gamma$ is the dual polar graph $A_{2D-1}(2)$.

Proof.

The graph $\Gamma$ has classical parameters $(D,b,\alpha,\sigma)$, where $b=-2$, by Lemma 3.8. Therefore $\Gamma$ is the dual polar graph $A_{2D-1}(2)$ by [10, Theorem B] because $a_{1}=1$ by Lemma 3.2(v). Moreover, $c_{2}=5$ by Lemma 3.8 because $\alpha=-6$ by [2, Tbl. 1]. This completes the proof. ∎

Now we can prove Theorem 1.2.

Proof of Theorem 1.2. (i) $\Rightarrow$ (ii): By Lemma 3.8, $\Gamma$ has classical parameters $(D,b,\alpha,\sigma)=(D,-2,\alpha,2+\alpha-\alpha[_{1}^{D}])$. Moreover, the corresponding eigenvalue for $E$ is equal to $-\frac{k}{2}$ by Lemma 3.2(i). We have $-\frac{k}{2}=\frac{b_{1}}{b}-1$ by Lemma 3.2(v) and $b=-2$. Therefore, $E$ is for the eigenvalue $\frac{b_{1}}{b}-1$.
(ii) $\Rightarrow$ (iii): Using (1) along with (6), (7), we find that $a_{i}=c_{i}$ for $1\leq i\leq D$. In particular $a_{1}=c_{1}=1$, so $\Gamma$ has no $2$-kites. By these comments and Remark 3.7, we see that $\Gamma$ is a regular near $2D$-gon. With reference to Definition 3.3, we see that $\Gamma$ has order $(s,t)$, where $s=a_{1}+1=2$ and $t=\frac{k}{a_{1}+1}-1=\frac{k}{2}-1$. Note that $\frac{b_{1}}{b}-1=-\frac{k}{2}=-t-1$. The result follows.
(iii) $\Rightarrow$ (iv): The intersection number $a_{1}=1$ because $\Gamma$ is regular near $2D$-gon of order $(2,t)$. Moreover, the corresponding eigenvalue for $E$ is equal to $-\frac{k}{2}$ since $t=\frac{k}{2}-1$. This implies that $\sigma_{1}=-\frac{1}{2}$ by (2) and the result follows by Lemma 3.1(iv) and Lemma 3.2(ii).
(iv) $\Rightarrow$ (v): We refer to the table in Remark 3.14. First assume that $D=2$. Then $c_{2}=2,3,5$ and $\Gamma$ is the unique regular near $4$-gon of order $(2,t)$, where $t=1,2,4$, by Proposition 3.10. Moreover, the intersection array of the unique regular near $4$-gon of order $(2,4)$ is the same as intersection array of the dual polar graph $A_{3}(2)$ (cf. [1, Theorem 9.4.3]). Next assume that $D=3$. Then $c_{2}=1,2,3,5$ and $\Gamma$ is the unique regular near $6$-gon of order $(2,t)$, where $t=8,11,14$, or the dual polar graph $A_{5}(2)$ by Proposition 3.11. Next assume that $D\geq 4$. Then by Proposition 3.12 and Proposition 3.13, $c_{2}=5$ and $\Gamma$ is the dual polar graph $A_{2D-1}(2)$.
(v) $\Rightarrow$ (vi): It is easily checked that for each of the cases listed in (v), the cosine sequence of $E$ satisfies $\sigma_{i}=(-\frac{1}{2})^{i}$ for $0\leq i\leq D$.
(vi) $\Rightarrow$ (i): We have $-\frac{k}{2}\sigma_{1}=\sigma_{0}+a_{1}\sigma_{1}+(k-a_{1}-1)\sigma_{2}$ by (3) and using (1). This implies that $\frac{k}{4}=1-\frac{a_{1}}{2}+\frac{k-a_{1}-1}{4}$ and therefore $a_{1}=1$. Thus the result follows by Lemma 3.1(iv). This completes the proof.

Remark 3.14.

In the following tables, we bring out some properties of the distance-regular graphs listed in item (v) of Theorem 1.2.

Name of graph $D$ $\{b_{i}\}_{i=0}^{D-1}$ $\{c_{i}\}_{i=1}^{D}$ Regular near $4$-gon of order $(2,1)$ $2$ $4,2$ $1,2$ Regular near $4$-gon of order $(2,2)$ $2$ $6,4$ $1,3$ Regular near $6$-gon of order $(2,8)$ $3$ $18,16,16$ $1,1,9$ Regular near $6$-gon of order $(2,11)$ $3$ $24,22,20$ $1,2,12$ Regular near $6$-gon of order $(2,14)$ $3$ $30,28,24$ $1,3,15$ Dual polar graph $A_{2D-1}(2)$ $D$ see (7) see (6)

$t$ from Theorem Minimal Classical Reference 1.2(iii) eigenvalue parameters $1$ $-2$ $(2,-2,-3,-4)$ [1, p. 30 (Examples)] $2$ $-3$ $(2,-2,-4,-6)$ [1, p. 30 (Examples)] $8$ $-9$ $(3,-2,-2,6)$ [1, p. 427] $11$ $-12$ $(3,-2,-3,8)$ [1, p. 427] $14$ $-15$ $(3,-2,-4,10)$ [1, p. 428] $3[^{D}_{1}]^{2}-2[^{D}_{1}]-1$ $2[^{D}_{1}]-3[^{D}_{1}]^{2}$ $(D,-2,-6,6[^{D}_{1}]-4)$ [1, Thm. 9.4.3]

Acknowledgements

Mojtaba Jazaeri is indebted to Paul Terwilliger for his supervision in obtaining and editing the results of this paper.