0

I am reading the Gortz's Algebraic Geometry, p.195, Lemma 7.42 and trying to understand some statements :

enter image description here

( Note (Errata) : In the proof, $A'^{r} \to M'$ is more correct instead of $A'^{n} \to M'$ and $ A_s^{r} \cong M_s$ is more correct instead of $A_s^{r}\cong M_s^{r}$. )

Throughout question, we will use the Corollary 7.41 (i) => (ii) in the above image.

First, for the second underlined statement, can we prove even "freeness of $M' = S^{-1}M$ of rank $r$", instead of the "local freeness of rank $r$"? Reason that I claim this is, since $M'$ is viewed as an $A'$-module and $A'$ is a semi-local ring, we may apply the answer of 'wxu' in next linked question : How can I find an element $x\not\in\mathfrak mM_{\mathfrak m}$ for every maximal ideal $\mathfrak m$ ? Is there something that I'm mistaken? If we can show the freeness of $M'$ (of rank $r$), then we can use A surjective homomorphism between finite free modules of the same rank to show that $A'^{r} \to M'$ is an isomorphism.

Or if not, is there any other method to show that $A'^{r}\to M'$ is isomorphism just from the local freeness of $M'$ of rank $r$?

Second. let's see the first underlined statement. Assume that we can prove that $M'$ is free over $A'$ of rank $r$. Then even if this statement is true, doesn't it guarantee the freeness of $M'/\tau M'$ over $A'/\tau $? So he detours through noting that "$A'/\tau$ is a product of fields" and shows the freeness of $M'/\tau M'$? (c.f. Every $A$-module is projective if and only if $A$ is a finite direct product of fields. )

Thrid, for the final underliend statement, at last, we have an isomorphism $(S^{-1}A)^{r} = A'^{r} \xrightarrow{\cong} M' = (S^{-1}M)$ Form this, how to show the existence of $s \in S$ and an isomorphism $A_s^{r} \xrightarrow{\cong} M_s$? Proposition 7.27 in the proof is as follows :

enter image description here

Thanks for reading. Can anyone help?

Plantation
  • 2,417

1 Answers1

1
  1. This is correct, but note that Gortz is basically redoing wxu’s answer. (You also need to see that $M’$ is projective, but it’s a tensor product of $M$ which is flat of finite presentation).

  2. Yes, but same thing: this is the first step of wxu’s approach anyway.

  3. Because $M$ is of finite presentation, you can multiply your isomorphism $(S^{-1}A)^r \rightarrow S^{-1}M$ by a fixed $s \in S$ so that it comes from a map $g: A^r \rightarrow M$. Because localization is exact, you know that $S^{-1}\ker{g}=S^{-1}\operatorname{coker}{g}=0$. Since $\operatorname{coker}{g}$ is finitely generated, there is some $t \in S$ such that $t\operatorname{coker}{g}=0$, thus $g_t: (A_t)^r \rightarrow M_t$ is onto; as $M_t$ has finite presentation, it follows that $\ker{g_t}$ is finitely generated and such that $S^{-1}\ker{g_t}=0$. Thus there is some $s \in S$ such that $s\ker{g_t}=0$ (over $A_t$). Thus the kernel and cokernel of $g: A_{st}^r \rightarrow M_{st}$ is zero, hence this map is an isomorphism.

Aphelli
  • 34,439
  • O.K. Thank you. For 3, can I ask question? :) 1) How can we choose $g:A^{r} \to M$ such that $S^{-1}g$ becomes the isomorphism $(S^{-1}A)^{r} \to S^{-1}M$, from the condition that $M$ is of finite presentation? 2) What the map $g_t : (A_t)^{r} \to M_t$ means exactly ? 3) Why $S^{-1}\operatorname{ker}g_t =0$? I guess that this can be deduced from $S^{-1}\operatorname{ker}g=0$. And how.. 4) Why the kernel and cokernel of $g_{st} : A_{st}^{r} \to M_{st}$ is zero? By the direct proof? – Plantation Jan 03 '23 at 11:16
  • 1
  • let $f$ be the original map $(A’)^r \rightarrow M’$. For each $1 \leq i \leq r$, $f(e_i)$ can be written as $m_i/s_i$ with $m_i \in M$, $s_i \in S$. So define $g: A^r \rightarrow M$ such that $g(e_i)=\prod_{j \neq i}{s_j} \cdot m_i$, then $g/(s_1 \cdot \ldots \cdot s_r)$ is exactly (after localization by $S$) $f$. I was going about this the wrong way, we don’t need any assumption on $M$ for this.
  • – Aphelli Jan 03 '23 at 11:26
  • 1
  • the localization of $g$ at $t$. 3) $\ker{g_t}$ is exactly the set of $m/t^r$ with $r \geq 0$ and $m \in \ker{g}$. 4) they’re localizations at $st$ of the kernel and cokernel of $g$, and $s,t$ have been constructed so that these localizations vanish.
  • – Aphelli Jan 03 '23 at 11:28
  • For $g_t$, it seems that more correct notation is $(A^{r})_t \to M_t$, instead of $(A_t)^{r} \to M_t$. Isn't that right? – Plantation Jan 03 '23 at 11:35
  • Uhm.. and for the comment for 1), what the notation $g/(s_1, \cdots , s_r)$ exactly means?~~ – Plantation Jan 03 '23 at 11:39
  • Anyway, I will try to fill in the details of the proof ! – Plantation Jan 03 '23 at 11:47
  • 1
    First comment: if you want. Second comment: just multiply $g$ by $\left(\prod_i{s_i}\right)^{-1}$. – Aphelli Jan 03 '23 at 12:14
  • C.f. After some investigation, I think that more correct statement(not involving the multiplicaion of $g$ by $(\prod_i s_i)^{-1}$ ) is, $f=(S^{-1}g) \circ \psi$, where $\psi : (A')^{r}=(S^{-1}A)^{r} \to S^{-1}(A^{r})$ is the canonical isomorphism. (We can provide its explicit form : $(a_1/s_1 , \cdots a_r/s_r) \mapsto (a_1s_2s_3\cdots s_r, \cdots ,a_rs_1s_2\cdots s_{r-1})/(s_1s_2 \cdots s_r)$ ) :) Anyway thank you. – Plantation Jan 05 '23 at 11:31