2
$\begingroup$

Let $H_{P,n}$ be the Hilbert scheme of subschemes of $\mathbb{P}^n(\mathbb{C})$ with Hilbert polynomial $P\in\mathbb{Q}[t]$, and let $U_{P,n}\to H_{P,n}$ be the flat universal family. Are there $n,P$ and closed points $y_0,y_1$ in the same component of $H_{P,n}$ such that the fiber $U_{P,n,y_0}$ is equidimensional (meaning all components have the same dimension) and reduced, and the fiber $U_{P,n,y_1}$ is not equidimensional?

Comments: in simple examples (see e.g. Harris, Algebraic geometry, a first course, chapter 21, in particular the section on curves of degree 2), when a component merges with a higher dimensional component under a flat deformation, it still leaves behind a trace in the form of nilpotents in the structure sheaf of the limit. For example, one can deform the union of a line $l\subset\mathbb{P}^2(\mathbb{C})$ and a point $p\not\in l$ (with the reduced scheme structure) to a scheme $X$ such that the underlying topological space is the same as for $l$, but there are nilpotents in some $\mathcal{O}_{X,p_0},p_0\in X$. The question is whether or not this happens in general.

$\endgroup$
4
  • 2
    $\begingroup$ Consider a curve through $y_0,y_1$ in the Hilbert scheme. Take the generic point, which must have components of both dimensions, and delete the lower-dimensional component. By general results this smaller scheme extends to a flat proper family, which can only differ from the original family on the closure of that component. Two reduced equidimensional schemes which agree outside a lower-dimensional subsets are identical, so if the special fiber of the original family is reduced then the special fibers are equal, which is absurd (Hilbert polynomials). No? $\endgroup$
    – Will Sawin
    Commented Nov 10, 2021 at 17:31
  • 1
    $\begingroup$ In fact, even if the special fibers are just $S2$ this cannot happen (you do not need the special fibers to be reduced). You can read much more about this type of thing in EGA IV, Sections 5.9 and 5.10 on $Z$-purity and $Z$-closedness. $\endgroup$ Commented Nov 10, 2021 at 19:09
  • $\begingroup$ Will and Jason: thanks! Will: yes. If you post this as an answer and maybe add a few details (such as we need to delete all low dimensional components and then take the scheme theoretic closure of what remains ...), I'll accept it. Jason: am I right that "this cannot happen" refers to the question of the title:)? $\endgroup$
    – algori
    Commented Nov 11, 2021 at 2:05
  • $\begingroup$ @algori "... am I right .." Yes, that is what I was saying. $\endgroup$ Commented Nov 13, 2021 at 11:26

0

You must log in to answer this question.

Browse other questions tagged .