Matrix generators for Weil representations (Q7294265)

Let \(r\) be an odd prime, and let \( \mathbb{F} \) be a field containing a primitive \(r\)th root of unity. For all \( \ell\ge 1 \), there is an embedding of \( Sp_{2\ell} ( r ) \) into \( GL_{r^{\ell}} ( \mathbb{F} ) \) arising from the Weil representation of \( Sp_{2\ell} ( r ) \). The purpose of the paper under review is to provide explicit matrices that generate this copy of \( Sp_{2\ell} ( r ) \) in \( GL_{r^{\ell}} ( \mathbb{F} ) \).\N\NLet \(R = r_{+}^{1 + 2\ell}\) be an extraspecial \(r\)-group of exponent \(r\) such that \( | R | = r^{1 + 2\ell} \) for some \( \ell\ge 1 \). Suppose that \( R< GL ( W ) \), where \(W\) is an absolutely irreducible \( \mathbb{F} [ R ] \)-module of dimension \( r^{\ell} \). Let \( \theta \) be a primitive \(r\)th root of unity in \( \mathbb{F} \). The center \( Z ( R ) \) is cyclic, \( Z ( R ) = \langle \theta I_{W} \rangle \). The quotient \( R / Z ( R ) \) can be considered as a vector space of dimension \( 2\ell \) over the finite field \( \mathbb{F}_{r} \) of order \(r\). The group commutator induces a non-degenerate alternating bilinear form \( b\colon R / Z ( R )\times R / Z ( R ) \rightarrow \mathbb{F}_{r} \) on \( R / Z ( R ) \) with \( [ x, y ] = \theta^{b ( x Z ( R ), y Z ( R ) ) } \) for all \( x, y\in R \). Elements of \( GL ( W ) \) normalizing \(R\) leave invariant the bilinear form \(b\), so there is a homomorphism \( \pi\colon N_{GL ( W )} ( R )\rightarrow Sp_{2\ell} ( r ) \), where \( \pi ( g ) \) is defined as the action of \(g\) on \( R / Z ( R ) \) for \( g\in N_{GL ( W )} ( R ) \).\N\NThe normalizer \( N_{GL ( W )} ( R ) \) is a split extension of \( \ker\pi \), that is, the normalizer contains a subgroup \( G\cong Sp_{2\ell} ( r ) \) with \( G\cap\ker\pi = 1 \). It is this group \(G\) that defines the Weil representation of \( Sp_{2\ell} ( r ) \).\N\NTo describe the generators for \( G\cong Sp_{2\ell} ( r ) \) one has to describe the embedding \( R < GL ( W ) \). Let \(V\) be an \(r\)-dimensional \( \mathbb{F} \)-vector space with basis \( v_{0}, v_{1}, \dots, v_{r-1} \). Let \( A, B, C, E \) be linear maps \( V\rightarrow V \) defining by\N\begin{align*}\NAv_{\xi} &= \theta^{\xi}v_{\xi}, \qquad \qquad Bv_{\xi} = v_{\xi + 1},\\\NCv_{\xi} &= \sum_{0\le i < r} \theta^{i\xi}v_{i},\quad Ev_{\xi} = \theta^{\frac{\xi ( \xi - 1 )}{2}}v_{\xi}\N\end{align*}\Nfor all \( 0\le \xi < r \), where the indices are interpreted modulo \(r\). Let \( W = V\otimes\cdots\otimes V \), \(V\) appears \(\ell\) times in the tensor product. If \( R_{0} = \langle A, B\rangle\cong r_{+}^{1+2} \), then \( R_{0} \) is absolutely irreducible. Therefore, the tensor product \( R = R_{0}\otimes\cdots\otimes R_{0} \) (\(\ell\) times) is absolutely irreducible subgroup of \( GL ( W ) \). As a basis for \(W\) the vectors \( v_{\xi_{1}, \dots, \xi_{l}} = v_{\xi_{1}}\otimes\cdots\otimes v_{\xi_{l}} ( 0\le \xi_{1}, \dots, \xi_{\ell} < r) \) are taken. The following linear maps \( A_{t}, B_{t}, C_{t}, D_{st}, E_{t} \colon W\rightarrow W \) are defined for all \( 1\le t\le\ell \) and \( 1\le s < t\le \ell \) by\N\begin{align*}\NA_{t}v_{\xi_{1}, \dots, \xi_{\ell}} &= \theta^{\xi_{t}}v_{\xi_{1}, \dots, \xi_{\ell}}, \quad B_{t}v_{\xi_{1}, \dots, \xi_{\ell}} = v_{\xi_{1}, \dots, \xi_{t-1}, \xi_{t+1},\dots,\xi_{\ell}},\\\NC_{t}v_{\xi_{1}, \dots, \xi_{\ell}} &= \sum_{0\le i < r} \theta^{i\xi_{t}}v_{\xi_{1}, \dots, \xi_{t-1}, i , \xi_{t+1}, \dots, \xi_{\ell}}, \\\ND_{st}v_{\xi_{1}, \dots, \xi_{\ell}} &= \theta^{\xi_{s}\xi_{t}}v_{\xi_{1}, \dots, \xi_{\ell}}, \quad E_{t}v_{\xi_{1}, \dots, \xi_{\ell}} = \theta^{\frac{\xi_{t} ( \xi_{t} - 1)}{2}}v_{\xi_{1}, \dots, \xi_{\ell}} ,\N\end{align*}\Nwhere \( 0\le \xi_{1}, \dots , \xi_{\ell} < r \) and the indices are interpreted modulo \(r\). These \( A_{t}, B_{t}, C_{t}, D_{st}, E_{t} \) together with the group of scalar matrices in \( GL ( W ) \) generate the normalizer \( N_{GL ( W ) }( R _{1} ) \), where \( R_{1} \) is the subgroup of \( GL ( W ) \) generated by \( A_{t}, B_{t} \). Next define \( U\colon V\rightarrow V \) by \( U = A^{(r+1)/2}E \) and \( U_{t}\colon W\rightarrow W (1\le t\le\ell) \) by \( U_{t} = A_{t}^{(r+1)/2}E_{t} \) so that \[ U_{t}v_{\xi_{1}, \dots, \xi_{\ell}} = \theta^{\frac{\xi_{t} ( \xi_{t} + r) } {2}}v_{\xi_{1}, \dots, \xi_{\ell}} \] for all \( 0\le \xi_{1}, \dots, \xi_{\ell} < r \). Finally define a scalar multiplier \( \lambda = ( -r )^{\frac{r-1}{2}}\det (C)^{-1} \) so that one can change each \( C_{t} \) to a linear map of determinant one.\N\NThe main result of the paper (Theorem 1.3) states that if \(G\) is the subgroup of \( GL ( W ) \) generated by the linear maps \( \lambda C_{t}, D_{st}, U_{t} \), then \( G\cong Sp_{2\ell} ( r ) \), with an isomorphism given by restriction of the map \(\pi\) to \(G\).\N\NThe explicit matrices generating \( Sp_{2\ell} ( r ) \) in \( GL_{ r^{\ell} } ( \mathbb{F} ) \) have been calculated using Magma.