The ubiquity of smooth Hilbert schemes (Q2205614)

Let \(k\) be a field. Work of \textit{F. S. Macaulay} [Proc. Lond. Math. Soc. (2) 26, 531--555 (1927; JFM 53.0104.01)] shows that the Hilbert scheme \(\text{Hilb}^{p(t)} (\mathbb P_k^n)\) parametrizing proper closed subschemes \(X \subset \mathbb P_k^n\) with Hilbert polynomial \(p(t)\) is non-empty precisely when \(p(t) = \sum_{j=1}^r \binom{t+b_j-j+1}{b_j}\) for uniquely determined integers \(b_1 \geq b_2 \geq \dots \geq b_r \geq 0\). Let \(\text{Hilb}^{p(t)} (\mathbb P_k^n)\) be the Hilbert scheme parametrizing closed subschemes \(X \subset \mathbb P_k^n\) with Hilbert polynomial \(p(t)\). It follows from work of Macaulay [loc. cit.] that the polynomials \(p(t) \in \mathbb Q [t]\) for which \(\text{Hilb}^{p(t)} (\mathbb P_k^n)\) is non-empty have unique expressions \(p(t) = \sum_{j=1}^r \binom{t+b_j-j+1}{b_j}\) for integers \(b_1 \geq b_2 \geq \dots \geq b_r \geq 0\). The author proves here that the lexicographic ideal is the unique saturated strongly stable ideal of codimension \(c\) with Hilbert polynomial \(p\) if and only if at least one of the following holds: (1) \(b_r>0\), (2) \(c \geq 2\) and \(r \leq 2\), (3) \(c=1\) and \(b_1=b_r\) or (4) \(c=1\) and \(r-s \leq 2\), where \(b_1=b_2=\dots=b_s > b_{s+1}\). When \(k\) is algebraically closed, the lex ideal is the unique saturated Borel-fixed ideal of codimension \(c\) with Hilbert polynomial \(p\) if and only if at least one of (1)-(4) hold. Using the result of \textit{A. Reeves} and \textit{M. Stillman} [J. Algebr. Geom. 6, No. 2, 235--246 (1997; Zbl 0924.14004)] that the scheme corresponding to the lex ideal is a smooth point of \(\text{Hilb}^{p(t)} (\mathbb P_k^n)\), it follows that \(\text{Hilb}^{p(t)} (\mathbb P_k^n)\) is nonsingular and irreducible if any of (1)-(4) hold, where \(n = c + \deg p(t)\). These include examples of \textit{G. Gotzmann} [Math. Z. 201, No. 1, 13--17 (1989; Zbl 0696.14001)]. The author uses the results above to study the geography of Hilbert schemes. He defines two operators \(A\) and \(\Phi\) on the set of Hilbert polynomials (\(A\) corresponds to adding a point, \(\Phi\) roughly corresponds to taking a projective cone). For each codimension \(c>0\), the graph with vertices \(\text{Hilb}^p (\mathbb P^n)\) such that \(\deg p + c = n\) and edges given by pairs \((\text{Hilb}^p (\mathbb P^n), \text{Hilb}^{A(p)} (\mathbb P^n))\) and \((\text{Hilb}^p (\mathbb P^n), \text{Hilb}^{\Phi (p)} (\mathbb P^{n+1}))\) is an infinite full binary tree with root \(\text{Hilb}^1 (\mathbb P^c)\) called the \textit{Hilbert tree} \(\mathcal H_c\).The \textit{Hilbert forest} is the disjoint union \(\mathcal H\) of Hilbert trees taken over all \(c>0\). Defining a probability distribution on \(\mathcal H\) for which all vertices at a fixed height are equally likely, the author proves that the probability of a vertex in \(\mathcal H\) corresponding to a smooth irreducible Hilbert scheme is greater than \(1/2\). The paper is well-written with helpful examples. Reviewers remark: Since the dimension $n$ of $\mathbb P^n$ varies in the Hilbert forest, one cannot draw similar conclusions about the geography of Hilbert schemes for subschemes of a fixed $\mathbb P^n$. Indeed, \textit{M. Martin-Deschamps} and \textit{D. Perrin} [Ann. Sci. Éc. Norm. Supér. (4) 29, No. 6, 757--785 (1996; Zbl 0892.14005)] have shown that for $d \geq 6$, there are onlyfinitely many arithmetic genera $g$ for which the open subset $H_{d,g} \subset \text{Hilb}^{dt+1-g}(\mathbb P_k^3)$ corresponding to locally Cohen-Macaulay curves is reduced or irreducible.