Irreducibility and singularities of some nested Hilbert schemes (Q2168818)

Let \(S\) be a smooth complex algebraic surface. Given positive integers \(n_1<n_2<\cdots <n_k\), let \(S^{[n_1, n_2, \dots, n_k]}\) denote the nested Hilbert scheme parameterizing nested 0-dimensional sub-schemes of \(S\): \(\xi_{n_1}\subseteq \xi_{n_2}\subseteq \cdots \xi_{n_k}\) of length \(n_i\). The nested Hilbert schemes are natural analogues for the Hilbert schemes \(S^{[n]}\) of points, and some of them have played an important role in the study of syzygies. The present well-written paper gives a quite comprehensive study of \(S^{[n, n+1, n+2]}\). The first main result is a new proof of the irreducibility of \(S^{[n, n+1, n+2]}\) due to \textit{N. Addington} [Algebr. Geom. 3, No. 2, 223--260 (2016; Zbl 1372.14009)]. The idea here is to realize \(S^{[n, n+1, n+2]}\) as \(\mathbb{P}(\mathscr{I}_{Z_{[n, n+1]}})\), where \(Z_{[n, n+1]}\) is the subscheme of \(S\times S^{[n, n+1]}\) parameterizing triples \((p, \xi_{n}, \xi_{n+1})\) with \(p\in \text{Supp}(\xi_{n+1})\), and use a criterion of \textit{G. Ellingsrud} and \textit{S. A. Strømme} [Trans. Amer. Math. Soc. 350, No. 6, 2547--2552 (1998; Zbl 0893.14001)]. Along the way, with more care the authors establish an estimate on codimension of certain strata of \(Z_{[n, n+1]}\) in \(S\times S^{[n, n+1]}\), which is quadratic in the minimal number \(i\) of generators for the localized ideal. While a linear estimate is sufficient in the criterion mentioned above, the quadratic one is of great interest on its own. Via forgetful and residual point maps, the irreducibility of \(S^{[n, n+2]}, S^{[1, n, n+1, n+2]}, S^{[1, n+1, n+2]}\) and \(S^{[1, n, n+2]}\) are deduced from that of \(S^{[n, n+1, n+2]}\). The second one is an explicit construction of a family of nested subschemes, indicating that \(S^{[1, 2, \dots, 23]}\) is reducible. As a corollary, \(S^{[n_1, n_2, \dots, n_k]}\) is reducible whenever \(k\ge 23\). The third one is that \(S^{[n, n+1, n+2]}\) is a local complete intersection and has klt singularities. The proof involves showing that a two-step blowup gives a (small) resolution of singularities of \(S^{[n, n+1, n+2]}\). In the end, the Picard group and the canonical divisor of \(S^{[n, n+1, n+2]}\) are computed in case \(S\) is regular.