A closed subscheme $Z$ of a scheme $X$ is by definition an equivalence class of closed immersions. How can one view $Z$ as a subset of $X$ (on the level of topological spaces)? Can one define a closed subscheme structure on an arbitrary closed subset $Y$ of $X$?
-
1At the very least, a closed subscheme defined by a closed subset of $X$ wouldn't be unique. For instance, the $x$-axis in the affine plane might correspond to $\operatorname{Spec}k[x, y]/(x)$, or it might correspond to $\operatorname{Spec}k[x, y]/(x^2)$. – Arthur Jul 20 '17 at 07:41
1 Answers
0) Hartshorne's definition of closed subscheme, which you use, is surprisingly bad for a mathematician of his calibre.
(His definition of open subscheme is weird too: see here).
The correct definition, as given by Grothendieck, Mumford, Qing Liu, Görtz-Wedhorn, De Jong's Stacks Project, etc. is the following:
1) A closed subscheme of the scheme $(X,\mathcal O_X)$ is a scheme $(Z,\mathcal O_Z)$ such that $Z\subset X$ is a closed subspace (with inclusion denoted $j:Z\hookrightarrow X$) and such that the sheaf of local rings $\mathcal O_Z$ is the restriction to $Z$ of the sheaf $\mathcal O_X/\mathcal I$, where $\mathcal I$ is some quasi-coherent ideal sheaf $\mathcal I\subset \mathcal O_X$.
Such a closed subscheme comes equipped with an obvious closed immersion $(Z,\mathcal O_Z)\hookrightarrow (X,\mathcal O_X)$.
2) a) Given an arbitrary closed subset $Z\subset X$ of a scheme $(X,\mathcal O_X)$ there exist in general infinitely many closed subschemes $(Z,\mathcal O_Z)\subset (X,\mathcal O_X)$.
For example the affine line $X=\mathbb A_k^1$ has as closed subschemes with support the origin $\{0\}\subset X$ all the subschemes with structure sheaf $\mathcal O_X/\mathcal I^n$, where $\mathcal I$ is the sheaf of sections vanishing at $0$.
Of course the closed subscheme $(\{0\},\mathcal O_X/\mathcal I^n)$ is isomorphic to $\operatorname {Spec}( k[T]/(T^n))$.
b) However among the multitude of closed subschemes with $Z$ as supporting topological space there is a privileged one: the unique closed subscheme with reduced structural sheaf $\mathcal O_Z$, corresponding to the largest quasi-coherent ideal sheaf $\mathcal I\subset \mathcal O_X$ such that $\mathcal O_X/\mathcal I$ has support $Z$.
In the degenerate case where $Z=X$, that privileged closed subscheme is a closed subscheme called the reduced subscheme $X_{red}\subset X$ associated to $X$.
- 156,622
-
Tiny nitpick: I think you meant to reference G"ortz--Wedhorn. (I never met anyone called T"orsten!) – bertram Jul 20 '17 at 09:48
-
Oops, you're right: Torsten is Wedhorn's first name! Corrected. And thanks a lot for your vigilance, @bertram. – Georges Elencwajg Jul 20 '17 at 09:55
-
one small point here is that these authors have essentially defined a closed subscheme to be the same as a closed immersion, in that every closed immersion determines a closed subscheme according to their definition.
The difference is that Hartshorne was trying to collect together all the "closed subschmes" with the same underlying set, while the others made it so that $X_{red}$ would be the closed subscheme to identify with the closed subset.
– Victor Zhang Aug 18 '17 at 08:50