Veronese quotient models of \(\overline M_{0,n}\) and conformal blocks (Q2443024)

This paper continues the recent study of compactifications of the moduli space \(M_{0,n}\) of \(n\) distinct points on the projective line, up to automorphism, obtained through geometric invariant theory (GIT), and its relation to morphisms associated to the determinant of the so-called Verlinde, or conformal block, vector bundles on the Deligne-Mumford compactification \(\overline{M}_{0,n}\) arising from the WZW model. This subject was launched by the seminal paper of \textit{N. Fakhruddin} [Contemp. Math. 564, 145--176 (2012; Zbl 1244.14007)], which (1) proved that all conformal block vector bundles in genus zero are globally generated, so that in particular their determinants induce morphisms to projective space, (2) provided a recursive formula for computing the isomorphism class of their determinant line bundle, and its intersection numbers with boundary 1-strata (i.e., F-curves), (3) observed that in special cases these classes coincide with the pullback of the distinguished polarization on the GIT quotients \[ \overline{M}_{0,n} \rightarrow (\mathbb{P}^1)^n/\!/\mathrm{SL}_2, \] and (4) asked how much of the geometry of the nef cone (the subject of the famous F-conjecture) is controlled by conformal blocks. These particular GIT quotients have a long history, and in fact were introduced in Mumford's original book as an elementary example of GIT. \textit{M. M. Kapranov}, in his [``Chow quotients of Grassmannians'', Adv. Sov. Math. 16(2), 29--110 (1993; Zbl 0811.14043)], introduced the above-mentioned morphism, and the computation of the polarization pullback was performed by \textit{V. Alexeev} and \textit{D. Swinarski} in [``Nef divisors on \(\overline{M}_{0,n}\) from GIT'', in: C. Faber (ed.) et al., Geometry and arithmetic. Based on the conference, Island of Schiermonnikoog, Netherlands, 2010. Zürich: European Mathematical Society. 1--21 (2012; Zbl 1317.14103)]. Building on this work, and Fakhruddin's observation about the coincidence with conformal block determinants, the reviewer introduced in [J. Algebr. Geom. 22, No. 4, 773--793 (2013; Zbl 1279.14029)] a generalization of these GIT quotients which allow the points to lie on a rational normal curve, not just a \(\mathbb{P}^1\). Later with \textit{A. Gibney} in [Adv. Math. 231, No. 2, 798--814 (2012; Zbl 1316.14051)] we showed that the pullback GIT polarizations obtained from these are exactly the level one type A conformal block divisor classes, and with \textit{D. Jensen} and \textit{H.-B. Moon} in [Adv. Math. 248, 242--278 (2013; Zbl 1345.14034)] we beefed up the GIT construction by remembering the rational normal curve, and its degenerations, rather than simply the configuration of points. A tantalizing question left open in the latter paper is whether these more complicated GIT polarizations still link up with conformal block divisors. That is the primary subject of the present paper. The main technical obstruction was to determine the isomorphism class of these GIT polarization pullbacks. In the case the reviewer studied with Gibney, this was possible due to a convenient use of Gale duality to reduce to the Alexeev-Swinarksi case. The present paper, however, accomplishes an impressive feat of writing an explicit formula for the intersection numbers between these more general GIT quotient polarizations and all F-curves (so that, by inverting an intersection matrix one obtains the desired class). This is a remarkable formula and remarkable accomplishment; it is the technical workhorse underlying most of the present paper. They achieve their formula through a clever use of push-pull and the projection formula and some explicit intersection theory machinery. With this formula in hand, they are able to look for a connection to conformal block divisors. Interestingly, they develop a case where the GIT polarization is \textit{not} simply proportional to the conformal block class, as it had been in all previously known cases, but the two classes lie on the same face of the nef cone of \(\overline{M}_{0,n}\) and hence determine the same alternative compactifcation via their induced morphism. A concluding section of the paper illustrates numerical/computation evidence that there are surely more tight connections between these GIT quotients and conformal block divisors, and thanks to their intersection formula one now has the tools to explore these further. Additionally, in the paper the authors provide an alternate modular interpretation to some compactifications, which is partially revealed through the GIT perspective, and they conclude by stating a conjecture that all positive weight type A determinants induce morphisms that preserve the interior \(M_{0,n}\) and discuss consequences of this for these morphisms, in essence that they would then be combinatorially determined by their recursive nature. It would be interesting if any of the general conformal blocks machinery, such as the projectively flat connection, could help resolve this conjecture. Additionally, the conformal blocks determinants have very recently been studied from the perspective of Gromov-Witten invariants of Grassmannians (see [\textit{P. Belkale}, \textit{A. Gibney} and \textit{S. Mukhopadhyay}, ``Quantum cohomology and conformal blocks on \(\overline{M}_{0,n}\)'', Preprint, \url{arXiv:1308.4906}]). It would be interesting to reanalyze the results in the present paper from this perspective, perhaps via a link between these GIT quotients and Grassmannians extending Kapranov's original work on the \(\mathbb{P}^1\) quotients.