Generating invariant rings of quivers in arbitrary characteristic (Q1682049): Difference between revisions

Throughout the review \(K\) denotes an algebraically closed field of arbitrary characteristic, and \(Q\) denotes a quiver with finite sets of vertices \(Q_0\) and of arrows \(Q_1\), and with maps \(t: Q_1\rightarrow Q_0\) assigning to every arrow its tail and \(h: Q_1\rightarrow Q_0\) assigning to every arrow its head. Every representation \(V\) of \(Q\) over \(K\) consists of a family of finite dimensional \(K\)-vector spaces \(V(i)\) determined by \(i\in Q_0\), together with a family of \(K\)-linear maps \(V(a): V(ta)\rightarrow V(ha)\) determined by \(a\in Q_1\). The dimension vector of \(V\) is \(\alpha: Q_0\rightarrow\mathbb{N}\), where \(\mathbb{N}\) means nonnegative integers and \(\alpha (i)\) is the dimension of \(V(i)\) for every \(i\in Q_0\). For a given dimension vector \(\alpha\in\mathbb{N}^{Q_0}\), fix a \(K\)-vector space \(V(i)\) of dimension \(\alpha (i)\) at every \(i\in Q_0\), and define the representation space to be \(\mathrm{Rep}(Q,\alpha)=\prod_{a\in Q_1}\mathrm{Hom}(V(ta),V(ha))\). Let \(K[\mathrm{Rep}(Q,\alpha)]\) denote the ring of polynomial functions on \(\mathrm{Rep}(Q,\alpha)\). Consider the group \(\mathrm{GL}(\alpha)=\prod_{i\in Q_0}\mathrm{GL}(V(i))\) and its subgroup \(\mathrm{SL}(\alpha)=\prod_{i\in Q_0}\mathrm{SL}(V(i))\). Then \(\mathrm{GL}(\alpha)\) acts on \(\mathrm{Rep}(Q,\alpha)\) via \((A(i)\mid i\in Q_0)\cdot (V(a)\mid a\in V_1)=(A(ha)V(a)A(ta)^{-1}\mid a\in Q_1)\), and on \(K[\mathrm{Rep}(Q,\alpha)]\) via \(A\cdot f(V)=f(A^{-1}\cdot V)\). Consider the rings of invariants \(\mathrm{I}(Q,\alpha)=K[\mathrm{Rep}(Q,\alpha)]^{\mathrm{GL}(\alpha)}\) and of semi-invariants \(\mathrm{SI}(Q,\alpha)=K[\mathrm{Rep}(Q,\alpha)]^{\mathrm{SL}(\alpha)}\). In the case when \(Q\) is the \(m\)-loop quiver with dimension vector \(\alpha=n\in\mathbb{N}\), the ring of invariants \(S(n,m):=K[{\mathrm{Mat}_{nn}}^m]^{\mathrm{GL}(n)}\) is known as the ring of matrix invariants. In the case when \(Q\) is the \(m\)-Kronecker quiver with dimension vector \(\alpha=(n,n)\in\mathbb{N}^2\), the ring of semi-invariants \(R(n,m):=K[{\mathrm{Mat}_{nn}}^m]^{\mathrm{SL}(n)\times\mathrm{SL}(n)}\) is known as the ring of matrix semi-invariants. In the paper under review, the authors prove that: The ring \(S(n,m)\) is generated by invariants of degree no greater that \((m+1)n^4\) (Theorem 1.4); The ring \(\mathrm{I}(Q,\alpha)\) is generated by invariants of degree \((M+1)N^4\), where \(M=|Q_1|\) and \(N=\sum_{i\in Q_0}\alpha (i)\) (Corollary 1.6); the ring \(R(n,m)\) is generated by invariants of degree no greater than \(mn^4\) (Theorem 1.7). A multiplicative character of \(\mathrm{GL}(\alpha)\) is of the form \(\chi_{\sigma}: (A(i)\mid i\in Q_0)\mapsto\prod_{i\in Q_0}\mathrm{det}(A(i))^{\sigma(i)}\), where \(\sigma: Q_0\rightarrow\mathbb{Z}\) is the weight of \(\chi_{\sigma}\). For a given weight \(\sigma\in\mathbb{Z}^{Q_0}\), define \(\mathrm{SI}(Q,\alpha)_{\sigma}=\{f\in K[\mathrm{Rep}(Q,\alpha)]\mid A\cdot f=\chi_{\sigma}(A)f\text{ for every }A\in\mathrm{GL}(\alpha)\}\). Then \(\mathrm{SI}(Q,\alpha)=\bigoplus_{\sigma}\mathrm{SI}(Q,\alpha)_{\sigma}\). The necessary condition for \(\mathrm{SI}(Q,\alpha)_{\sigma}\) to be nonzero is \(\sigma\cdot\alpha=\sum_{i\in Q_0}\sigma (i)\alpha (i)=0\). For given \(\alpha\in\mathbb{N}^{Q_0}\) and \(\sigma\in\mathbb{Z}^{Q_0}\) such that \(\sigma\cdot\alpha =0\), write \(\sigma_+(i)=\mathrm{max}\{\sigma (i),0\}\) and \(\sigma_-(i)=\mathrm{max}\{0,-\sigma (i)\}\), and define \(|\sigma|_{\alpha}=\sum_{i\in Q_0}\sigma_+(i)\alpha (i)=\sum_{i\in Q_0}\sigma_-(i)\alpha (i)\). In the paper under review, for a quiver without oriented cycles \(Q\) the authors prove that: The ring \(\mathrm{SI}(Q,\alpha,\sigma):=\bigoplus_{d\geq 0}\mathrm{SI}(Q,\alpha)_{d\sigma}\) is generated by invariants of degree no grater that \(mn^3\), where \(n=|\sigma|_{\alpha}\) and \(m=\sum_{i,j\in Q_0}\sigma_+(i)b_{ij}\sigma_-(i)\) and where \(b_{ij}\) is the number of paths from \(i\) to \(j\) (Corollary 1.8); The ring \(\mathrm{SI}(Q,\alpha)\) is generated by semi-invariants of weights \(\sigma\) with \(|\sigma|_{\alpha}\leq 3rn^2{||\alpha||_1}^{4n}/128(n-1)^{4n-4}\), where \(r\) is the Krull dimension of \(\mathrm{SI}(Q,\alpha)\), \(n=|Q_0|\) and \(||\alpha||_1=\sum_{i\in Q_0}\alpha (i)\) (Theorem 1.9).