6
$\begingroup$

I am try to prove that $X(C) = \{\text{nilpotent elements in C}\}$ is not a scheme.

I have proved that it is local, so what makes it is not a scheme must be the open cover stuff. And I think it is covered by $X_n(C) = \{x|\text{$x^n=0$ in C}\}$. So I think it is covered by an infinite union of affine schemes. Therefore, I think the problem is that the cover is not open.

But even if I could prove this obvious cover is not open. It is still not sufficient to prove that $X$ has no open affine cover. And I have no idea for finding some contradiction assuming $X$ has an open affine cover.

So how to show that $X$ has no open affine cover? Thanks for any help.

$\endgroup$
2
  • 1
    $\begingroup$ What is the space $C$? $\endgroup$
    – Joppy
    Oct 29, 2017 at 4:19
  • $\begingroup$ @Joppy $X$ is an $R$-functor for a ring $R$, and it is applied on $C$ where $C$ is an $R$-algebra. $\endgroup$
    – Y.X.
    Oct 29, 2017 at 4:21

1 Answer 1

Reset to default
6
$\begingroup$

Here's one way to prove it. Take your favorite field $k$ and let's base-change to $k$: supposing $X$ is a scheme, consider the scheme $Y=X\times \operatorname{Spec} k$ as a scheme over $\operatorname{Spec} k$. For a $k$-algebra $C$, then, $Y(C)$ is the set of nilpotent elements of $C$. As you say, this is covered by the affine schemes $Y_n=\operatorname{Spec} k[t]/(t^n)$ of elements $x$ such that $x^n=0$. Now the key observation is that each $Y_n$ has only one point, and so when you consider $Y$ as the nested union of the $Y_n$, this means $Y$ has only one point as well. In particular, $Y$ must be affine, so it must be $\operatorname{Spec} A$ for some $k$-algebra $A$. Then $A$ would be the universal $k$-algebra with a nilpotent element: there is an element $a\in A$ such that for any nilpotent element $x$ in any $k$-algebra $C$, there is a unique homomorphism $A\to C$ sending $a$ to $x$. This is impossible, since we must have $a^n=0$ for some particular $n$ and then $a$ cannot be mapped to $x$ if $x^n\neq 0$ but $x^N=0$ for some $N>n$ (e.g., $x=t$ in $k[t]/(t^N)$).

$\endgroup$
8
  • $\begingroup$ Maybe I have some misunderstandness. But may I please ask if this $Y$ is equivalent to taking $R$ to be $k$, so if $Y$ is an $R$-scheme for all $R$, then it must be a $k$ scheme? If so, I think if we take an $k$-algebra $C$ such like $k[x]/\langle x^2\rangle$. Then $0$ is not the only nilpotent so $Y(C)$ is not a single point. $\endgroup$
    – Y.X.
    Oct 29, 2017 at 8:16
  • $\begingroup$ Yes, $Y$ is equivalent to taking $R$ to be $k$. When I refer to "points", I mean in the sense of points of Spec of a ring being prime ideals, not in the sense of the functor of points. $\endgroup$
    – Eric Wofsey
    Oct 29, 2017 at 8:23
  • $\begingroup$ It seems that you are talking about some intrinsic notion of a point of a functor without mention of an $k$-algebra $C$. It seems unfamiliar to me. If possible, may I please ask what is the "points" here refers to, and how does it related to "prime ideals"？ $\endgroup$
    – Y.X.
    Oct 29, 2017 at 8:36
  • $\begingroup$ Oh, it seems you have never seen that definition of schemes at all then. Instead of defining schemes as functors, you can define them as certain locally ringed spaces. If $A$ is a commutative ring, then $\operatorname{Spec} A$ is the set of all prime ideals of $A$ equipped with a certain topology and sheaf of rings. General schemes are then locally ringed spaces that are locally isomorphic to spaces of the form $\operatorname{Spec} A$. $\endgroup$
    – Eric Wofsey
    Oct 29, 2017 at 8:39
  • 4
    $\begingroup$ That today students may learn that affine schemes are a type of functors but are not told that these schemes have a certain relationship with prime ideals makes me feel very old... $\endgroup$
    – Georges Elencwajg
    Oct 29, 2017 at 12:32

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.