Quivers, curves, and the tropical vertex (Q973681)

The paper surveys the relationship between three mathematical structures: (i) Euler characteristics of the moduli space of quiver representations, (ii) factorizations of commutators in the tropical group. (iii) counts of rational curves on toric surfaces. We explain the structures (i)-(iii). For (i), consider the \(m\)-Kronecker quiver \(Q_m\) with \(2\) vertices and \(m\) edges. Each representation of \(Q_m\) has two underlying vector spaces \((V_1, V_2)\) and write \[ d_1=\dim V_1, \,\, d_2=\dim V_2. \] Stability conditions are necessary to construct moduli spaces. A stability condition is a pair \(\omega=(w_1, w_2)\) necessary to write down the slope of a representation \[ \mu=\frac{w_1\dim V_1+w_2\dim V_2}{\dim V_1+\dim V_2}. \] Semistability is defined with respect to slope. Oftentimes, additional structure is necessary: the choice of a line \[ L_1\subset V_1 \text{ or } L_2\subset V_2, \] called back or front framing. The stability condition \((1, 0)\) plays an important role. We consider the moduli spaces \[ \mathcal M^{(1, 0), B}(d_1, d_2),\,\,\, \mathcal M^{(1, 0), F}(d_1, d_2) \] of \((1, 0)\)-semistable representations of \(Q_m\) of dimensions \((d_1, d_2)\), endowed with either back or front framings. For each pair \((a, b)\) of positive integers, one forms the generating series of Euler characteristics \[ B_{a, b}=1+\sum_{k>0} \chi(\mathcal M^{(1, 0), B}(ak, bk))\cdot (tx)^{ak}\cdot (ty)^{bk} \] \[ F(a, b)=1+\sum_{k>0} \chi(\mathcal M^{(1, 0), F}(ak, bk))\cdot (tx)^{ak}\cdot (ty)^{bk}. \] For (ii), consider pairs \((a, b)\in \mathbb Z^2\) and functions of the form \[ f=1+tx^ay^b\cdot g(x^ay^b, t),\,\, g(z, t)\in \mathbb C[z][[t]]. \] An automorphism \[ \theta_{(a, b), f}\in \text{Aut}_{\mathbb C[[t]]}(\mathbb C^{\star}\times \mathbb C^{\star}\times \mathbb C[[t]]) \] of the torus over \(\mathbb C[[t]]\) can be defined by sending the torus coordinates \((x, y)\) to \[ \theta_{(a, b), f}(x)=xf^{-b},\,\,\, \theta_{(a, b), f}(y)=yf^a. \] Special examples include the automorphisms \[ S_{\ell_1}=\theta_{(1, 0), (1+tx)^{\ell_1}} \text{ and } T_{\ell_2}=\theta_{(0, 1), (1+ty)^{\ell_2}}. \] Their commutator can be written in the form \[ S_{\ell_1}\circ T_{\ell_2} \circ S_{\ell_1}^{-1} \circ T_{\ell_2}^{-1}=\prod\theta_{(a,b), f_{a,b}} \] where the (suitably interpreted infinite) product on the right is ordered by the slopes \((a, b)\). On the curve counting side (iii), fix first integers \((a, b)\) and \(\ell_1, \ell_2\) as above. Consider the weighted projective plane \[ X_{a,b}=(\mathbb C^3-\{0\})/\mathbb C^{\star}, \] the quotient of \(\mathbb C^3\setminus \{0\}\) via the action \[ t\cdot (z_1, z_2, z_3)=(t^a z_1, t^b z_2, z_3). \] Let \[ D_1, D_2, D_{out} \] denote the toric divisors corresponding to \[ z_1=0,\,\, z_2=0\,\,\, z_3=0, \] and let \(X_{a,b}^{\circ}\) be the surface obtained by removing the three torus fixed points \[ [1, 0, 0], \, [0, 1, 0], \, [0, 0, 1]. \] The divisors \(D_1, D_2, D_{out}\) restrict to \(D_1^{\circ}, D_2^{\circ}, D_{out}^{\circ}\) on \(X_{a, b}^{\circ}\). For a suitable blowup \[ \widetilde {X^{\circ}_{a, b}}\to X_{a,b}^{\circ} \] along \(\ell_1\) points of \(D_1^{\circ}\) and \(\ell_2\) points \(D_2^{\circ}\), one counts curves in a fixed homology class \(\beta_k[\mathbf P_a, \mathbf P_b]\) with full contact order along the divisor \(D_{out}^{\circ}\) at an uspecified point. The cohomology class \(\beta_k[\mathbf P_a, \mathbf P_b]\) is defined by picking partitions \((\mathbf P_a, \mathbf P_b)\) of size \(ak\) and \(bk\) to indicate the order of contact at the \(\ell_1\) and \(\ell_2\) blown-up points. As a result, one can define the relative invariants \[ N_{a, b}[\mathbf P_a, \mathbf P_b]=\int_{[\mathfrak M(\widetilde X^{\circ}_{a, b}/D_{out}^{\circ})]^{vir}} 1. \] The main result connecting (i) and (ii) is provided by \textit{M. Reineke}'s theorem [``Moduli of representations of quivers'', in: Trends in representation theory of algebras and related topics, EMS Ser. Congr. Rep., Eur. Math. Soc., Zürich, 589--637 (2008; Zbl 1206.16009); \url{arXiv:0802.2147}] \[ f_{a, b}=B_{a, b}^{\frac{m}{a}}=F_{a, b}^{\frac{m}{b}}. \] The connection between (ii) and (iii) is provided by the equality \[ \log f_{a, b}=\sum_k k c_k(\ell_1, \ell_2)\cdot (tx)^{ak} \cdot (ty)^{bk} \] where \[ c_k(\ell_1, \ell_2)=\sum_{|\mathbf P_a|=ak, |\mathbf P_b|=bk} N_{a,b}[\mathbf P_a, \mathbf P_b], \] with \(\mathbf P_a, \mathbf P_b\) partitions of length \(\ell_1, \ell_2\). This connection was proved by \textit{M. Gross}, \textit{R. Pandharipande} and \textit{B. Siebert} [``The tropical vertex'', Duke Math. J. 153, No. 2, 297--362 (2010; Zbl 1205.14069)]. The current paper surveys these relationships, offers some open questions and directions of research. Furthermore, new results about the rays and symmetries of scattering diagrams of commutators are proven.