0

Let $f: X \to Y$ a morphism between $S$-schemes where $Y$ is a separated $S$-scheme. Therefore the diagonal morphism $Y \to Y \times_S Y$ is a closed immersion.

Let $\Gamma_f = (id,f): X \to X \times_S Y$ the graph morphism.

Why and how to see that $\Gamma_f $ is a closed immersion.

user267839
  • 7,293
  • 1
    Can you realise this situation as a fibre product (where the bottom horizontal line is the diagonal morphism and the top one if $\Gamma_f$)? – User3773 Apr 13 '18 at 12:02
  • 1
    Why and how to see that $\Gamma_f $ is a closed immersion. I suppose that you are mean

    $$ \require{AMScd} \begin{CD} X @>\Gamma_f >> Y\times_S X \ @VVfV @VV(id,f)V \ Y @>\Delta>> X \times_S Y \end{CD} $$

    Is it that what you mean? If yes, why does if prove the claim? Is $(id,f)$ closed immersion?

    – user267839 Apr 13 '18 at 21:40
  • 2
    Yes, $\Delta$ is a closed embedding by assumption and since closed embeddings are stable under base change it follows that $\Gamma_f$ is also a closed embedding. To see that this diagram is actually a fibre product I guess you can work locally on affine opens where it should be easier and then glue everything together again (set-theoretically is quite intuitive I guess). – User3773 Apr 16 '18 at 16:31

1 Answers1

2

Let the structure maps of $X$ and $Y$ as $S$-schemes be $u$ and $v$, respectively, so that $u=v\circ f$.

As mentioned in the comments, we want to check $$\require{AMScd} \begin{CD} X @>{\Gamma_f}>> X\times_SY\\ @V{f}VV @V{f\times Y}VV \\ Y @>{\Delta_Y}>> Y\times_SY \end{CD}$$ is a fibered product. We may check this locally [the commuting diagram produces a morphism of schemes $X\to Y\times_{Y\times_SY}(X\times_SY)$, and we may check this is an isomorphism locally], so let $U=\mathrm{Spec} R\subset S$ be an affine open set, $W=\mathrm{Spec} B\subset v^{-1}(U)\subset Y$, and $V=\mathrm{Spec} A\subset f^{-1}(W)\subset X$. Now the diagram solely consists of affine schemes: $$\require{AMScd} \begin{CD} V @>{\Gamma_{f|_V}}>> V\times_UW\\ @V{f|_V}VV @V{f|_V\times W}VV \\ W @>{\Delta_Y}>> W\times_UW \end{CD}$$

Suppose $f\colon V\to W$ is given by $\varphi\colon B\to A$. Then in the category of commutative rings the diagram becomes:

$$\require{AMScd} \begin{CD} A @<{\gamma}<< A\otimes_RB\\ @A{\varphi}AA @A{\varphi\otimes B}AA \\ B @<{\delta}<< B\otimes_RB \end{CD}$$ where $\gamma(a\otimes b):=a\varphi(b)$ and $\delta(b_1\otimes b_2)=b_1b_2$. Now following the diagram gives $$\varphi\circ\delta(b_1\otimes b_2)=\varphi(b_1b_2)=\varphi(b_1)\cdot\varphi(b_2)=\gamma(\varphi(b_1)\otimes b_2)=\gamma\circ(\varphi\otimes B)(b_1\otimes b_2),$$ so it indeed commutes.

To show this is a pullback diagram: We want to show $A\cong B\otimes_{B\otimes_RB}(A\otimes_RB)$. We have maps:

$$A\to B\otimes_{B\otimes_RB}(A\otimes_RB):a\mapsto1\otimes(a\otimes1)$$ and $$B\otimes_{B\otimes_RB}(A\otimes_RB)\to A:b_1\otimes(a\otimes b_2)\mapsto \varphi(b_1b_2)a.$$ There are a couple of things to check here:

  • That the second map is well-defined: Check that $B\times(A\otimes_RB)\to A:(b_1,a\otimes b_2)\mapsto \varphi(b_1b_2)a$ is a well-defined $B\otimes_RB$-bilinear map.
  • That these maps are inverses of each other. One composition is obvious: $a\mapsto 1\otimes(a\otimes 1)\mapsto a$. The other direction is: \begin{align*}1\otimes(\varphi(b_1b_2)a\otimes 1)&=1\otimes\big((b_1b_2\otimes1)\cdot(a\otimes 1)\big)\\ &=\big((b_1b_2\otimes1)\cdot1\big)\otimes(a\otimes 1)\\ &=b_1b_2\otimes(a\otimes 1)\\ &=\big((1\otimes b_2)\cdot b_1\big)\otimes(a\otimes 1)\\ &=b_1\otimes\big((1\otimes b_2)\cdot(a\otimes 1)\big)\\ &=b_1\otimes(a\otimes b_2). \end{align*}
Kenta S
  • 16,151
  • 15
  • 26
  • 53
  • I think the answer is rather incomplete (or at least requires some clarifications): Firstly, why it suffice to work locally here? Secondly, passing to diagram for commutative rings you have shown that the diagram commutes, but not that it's a fibered product / pullback square. – user267839 Feb 25 '22 at 15:28
  • My bad, I agree. I will try to edit it within the next few hours. – Kenta S Feb 25 '22 at 17:35