GIT compactifications of \(M_{0,n}\) and flips (Q2437522)

The authors consider a VGIT problem in order to construct and describe different modular compactifications of the moduli space of rational pointed curves \( M_{0,n}\) along with maps among them. Following \textit{Y. Hu} and \textit{S. Keel}'s [Mich. Math. J. 48, 331--348 (2000; Zbl 1077.14554)] ideas this is indeed intimately related to the understanding of the birational geometry of \(\overline M_{0,n}\). Let \(\mathrm{Chow}(1,d,\mathbb P^d)\) be the irreducible component parametrizing rational normal curves and their limit cycles inside the Chow variety of degree \(d\) curves in \(\mathbb P^d\). The compactifications described by the authors are constructed via GIT quotients of the natural action of \(\mathrm{SL}(d+1)\) in the locus \(U_{d,n}\) described as follows \[ U_{d,n}:=\{(X,p_1,\dots,p_n)\in \mathrm{Chow}(1,d,\mathbb P^d)\times {(\mathbb P^d)}^n| p_i\in X, \forall i\}. \] The GIT quotients \(U_{d,n}/\!/\mathrm{SL}(d+1)\) depend on a linearization \(L\in \mathbb Q_{>0}^{n+1}\), viewed as the assignment of rational weights to the curve and each of the marked points. A GIT stability analysis then reveals that there is a convex cone with cross-section \(\Delta^o\subset \mathbb Q^{n+1}\) such that, for each \(L\in \Delta^o\), \(U_{d,n}/\!/_L\mathrm{SL}(d+1)\) is a compactification of \(M_{0,n}\) endowed with a birational map from \(\overline M_{0,n}\) preserving \(M_{0,n}\). This VGIT set-up implies that, for each \(d\), there is a finite wall and chamber decomposition of \(\Delta^o\), corresponding to different compactifications of \(M_{0,n}\), along with different birational maps among them. The authors study the situation in this case, obtaining a description of interesting flips between these spaces. The authors then relate the compactifications obtained via this GIT analysis with the other known compactifications of \(M_{0,n}\). They show that all Hassett's spaces in genus \(0\) can be obtained as quotients \(U_{d,n}/\!/\mathrm{SL}(d+1)\). Moreover, for each open chamber in \(\Delta^o\), the associated GIT quotient is a modular compactification of \(M_{0,n}\) in the sense of \textit{D. I. Smyth} [Invent. Math. 192, No. 2, 459--503 (2013; Zbl 1277.14023)], whereas the walls correspond to weakly modular compactifications in the sense \textit{J. Alper}, \textit{D. I. Smyth} and \textit{F. van der Wyck} [``Weakly proper moduli stacks of curves'', Preprint, \url{arXiv:1012.0538}]. Finally, the authors observe that their constructions yield common generalizations for Kapranov's constructions of \(\overline M_{0,n}\).