Chapter 14 Parameters of Thin Modules, I
Friday, February 19, 1993
Summary.
Definition 14.1 Assume \(\Gamma = (X, E)\) is distance-regular with respect to every vertex \(x\in X\).
Notation: Let \(x\in X\). The data of the trivial \(T(x)\)-module.
\[\begin{array}{|c|c|c|} \hline & \text{Case DR} & \text{Case DBR} \\ \hline \text{valency} k_x & k & \begin{cases} k^+ & \text{ if } x\in X^+\\ k^- & \text{ if } x\in X^-\end{cases}\\ x\text{-diameter } D_x & D & \begin{cases} D^+ & \text{ if } x\in X^+\\ D^- & \text{ if } x\in X^-\end{cases}\\ \text{measure $m_x$} & m & \begin{cases} m^+ & \text{ if } x\in X^+\\ m^- & \text{ if } x\in X^-\end{cases}\\ \text{int. number }c_i(x) & c_i & \begin{cases} c_i^+ & \text{ if } x\in X^+\\ c_i^- & \text{ if } x\in X^-\end{cases}\\ \text{int. number }b_i(x) & b_i & \begin{cases} b_i^+ & \text{ if } x\in X^+\\ b_i^- & \text{ if } x\in X^-\end{cases}\\ \text{int. number }a_i(x) & a_i & 0\\ \hline \end{array}\]
Call \(m\), \(m^{\pm 1}\) the measure of \(\Gamma\).
Assume \(\Gamma = (X, E)\) is distance-regular.
To what extent do \(a_i\)’s, \(b_i\)’s and \(c_i\)’s determine the structure of irreducible \(T(x)\)-modules? In general, the following hold.
Lemma 14.1 Assume \(\Gamma = (X, E)\) is distance-regular. Pick \(x\in X\). Let \(W\) be a thin irreducible \(T(x)\)-module with endpoint \(r\), diameter \(d\) and measure \(m_W\).
Moreover, \(f_W\in \mathbb{R}[\lambda]\), and
\((iii)\) For all eigenvalues \(\theta_i\) of \(\Gamma\), \(\lambda - \theta_i\) is a factor of \(f_W\) whenever, \(E_iW = 0\).
In particular, \(2r-D+d\geq 0\).
Proof. Let \(\theta_0, \ldots, \theta_D\) denote distinct eigenvalues of \(\Gamma\). Then \(m(\theta_i) \neq 0\) \((0\leq i\leq D)\) by Proposition 13.1.
There exists a unique \(f_W\in \mathbb{C}[\lambda]\) with \(\deg f_W\leq D\) such that \[f_W(\theta_i) = \frac{m_W(\theta_i)}{m(\theta_i)} \quad (0\leq i\leq D)\] by polynomial interpolation.
\(f_W\in \mathbb{R}[\lambda]\) since \[\theta_0, \ldots, \theta_D\in \mathbb{R} \quad \text{and}\quad f_W(\theta_0), \ldots, f_W(\theta_D) \in \mathbb{R}.\]
\((ii)\) Without loss of generality, we may assume \(r < D/2\), else trivial.
Pick \(0\neq w \in E^*_r(x)W\). \[w = \sum_{y\in W, \partial(x,y) = r}\alpha_y\hat{y} \quad \text{ for some } \; \alpha_y\in \mathbb{C}.\] Pick \(y\in X\) such that \(\alpha_y\neq 0\).
Set \(W'\) be the trivial \(T(y)\)-module. (\(\langle w, \hat{y}\rangle \neq 0, \text{ as } W\not\bot W'\).) \[r' = 0, \quad m' = m, \quad \Delta = r.\]
Apply Theorem 12.1, we have \[\begin{align} \deg p & \leq \Delta - r' + r = 2r, \quad p\neq 0\\ \deg p' & \leq \Delta - r + r' = 0 , \quad p'\neq 0. \end{align}\] \[m_W(\theta)\overline{p'(\theta)} = m(\theta)p(\theta) \quad (\text{ for all } \theta \in \mathbb{R}).\] So, \[\deg p/\bar{p}' \leq 2r,\] and \(p/\bar{p}'\) satisfies the conditions of \(f_W\). \[\left(\frac{p(\theta)}{\bar{p}'(\theta)} = \frac{m_W(\theta)}{m(\theta)}\right)\]
\((iii)\) \[E_iW = 0 \Rightarrow m_W(\theta_i) = 0 \Rightarrow f_W(\theta_i) = 0.\] that is, \(E_iW = 0\). Hence \(\theta_i\) is a root of \(f_W(\lambda) = 0\). So, \[2r \geq \deg f_W \geq |\{\theta_i\mid E_iW = 0\}| = D-d.\] Hence, \[2r-D + d \geq 0.\] This proves the assertions.
Lemma 14.2 Let \(\Gamma = (X, E)\) be any distance-regular graph with valency \(k\), diameter \(D\) \((D\geq 2)\), measure \(m\), and eigenvalues \[k = \theta_0 > \theta_1 > \cdots > \theta_D.\] Pick \(x\in X\). Let \(W\) be a thin irreducible \(T(x)\)-module with endpoint \(r = 1\), diameter \(d\) and measure \(m_W = mf_W\). Then one fo the following cases \((i)-(iv)\) occurs.
\[\begin{array}{|c|c|c|c|} \hline \text{Case} & d & f_W(\lambda) & a_0(W) \\ \hline (i) & D-2 & \frac{(\lambda - k)(\lambda - \theta_1)}{k(\theta_1 + 1)} & -\frac{b_1}{\theta_1 + 1} -1\\ (ii) & D-2 & \frac{(\lambda - k)(\lambda - \theta_D)}{k(\theta_D + 1)} & -\frac{b_1}{\theta_D + 1} -1\\ (iii) & D-1 & \frac{k - \lambda}{k} & -1\\ (iv) & D-1 & \frac{(\lambda - k)(\lambda - \beta)}{k(\beta + 1)} & -\frac{b_1}{\beta + 1} -1\\ \hline \end{array}\] for some \(\beta\in \mathbb{R}\) with \(\beta\in (-\infty, \theta_D) \cup (\theta_1, \infty)\). Moreover, the isomorphism class of \(W\) is determined by \(a_0(W)\).
Note. By \((iii)\), the possible “shapes” of a thin irreducible \(T(x)\)-modules are: \[\begin{align} r = 0 & \quad d = D,\\ r = 1 & \quad d = D-1,\\ r = 1 & \quad d = D-2. \end{align}\]