Quiver representations with an irreducible null cone. (Q619034): Difference between revisions

Let \(Q\) be a quiver and \(\mathbf d\) a dimension vector. Assume that there exists a representation \(T\) of \(Q\) with dimension vector \(\mathbf d\) such that \(\text{Ext}_Q^1(T,T)=0\). Such a representation, if exists, is uniquely determined by \(\mathbf d\). Let \(S_1,\dots,S_r\) be the pairwise-nonisomorphic simple objects in the category \(T^\perp\), which consists of the representations \(M\) of \(Q\) such that \(\Hom_Q(T,M)=0=\text{Ext}_Q^1(T,M)\). The author studies the following problem: under which conditions there exists a representation \(X\) of \(Q\) such that the following conditions are satisfied: {\parindent=6mm\begin{itemize}\item[---] \(\Hom_Q(X,S_i)\neq 0\) for each \(i=1,\dots,r\);\item[---] if \(N\) is a representation of \(Q\) of dimension vector \(N\cdot\mathbf d\) for some \(N\geq 1\) such that \(\Hom_Q(N,S_i)\neq 0\) for each \(i=1,\dots,r\), then \(N\) is a degeneration of \(X\oplus T^{N-1}\), i.e.\ there exists a representation \(Z\) of \(Q\) and an exact sequence of the form \(0\to Z\to Z\oplus X\oplus T^{N-1}\oplus N\to 0\). \end{itemize}} Among other things the author proves that such a representation exists if \(Q\) is a Dynkin quiver and \(\mathbf d\) is sufficiently big. The author's motivation for studying this problem is geometric. More precisely, for \(Q\) and \(\mathbf d\) as above one defines a variety called the variety of representations of \(Q\) with dimension vector \(\mathbf d\). It is an affine space equipped with an action of a product \(\text{GL}(\mathbf d)\) of general linear groups. The orbits with respect to this action correspond naturally to the isomorphism classes of the representations of \(Q\) with dimension vector \(\mathbf d\). Moreover, there exists a representation \(T\) of \(Q\) with dimension vector \(\mathbf d\) such that \(\text{Ext}_Q^1(T,T)=0\) if and only if there exists an open orbit in \(\text{GL}(\mathbf d)\) (and, if this is the case, this open orbit corresponds to the isomorphism class of \(T\)). Next, the set of the isomorphism classes of the representations \(X\) with dimension vector \(\mathbf d\) such that \(\Hom_Q(X,S_i)\neq 0\) for each \(i=1,\dots,r\) correspond to the set of the common zeros of the non-constant \(\text{GL}(\mathbf d)\)-semi-invariants. Finally, a representation \(N\) is a degeneration of a representation \(M\) if and only if the orbit corresponding to the isomorphism class of \(N\) is contained in the orbit corresponding to the isomorphism class of \(M\).