0
$\begingroup$

Given a commutative ring $ R $ and a multiplicatively closed subset $ S $ of $ R $, there are two ways to consturct $ S^{-1}R $:

  1. define an equivalence relation $ \sim $ on $ R\times S $ and then take $ S^{-1}R := (R\times S)/\sim $.

  2. Let $ R[S] $ be the free commutative algebra over $ R $ generated by $ S $ and $ i:S\to R[S] $ the canonical embedding and $ I $ the ideal of $ R[S] $ generated by $ i(s)s-1 $ for each $ s $ in $ S $ and then take $ S^{-1}R := R[S]/I $.

Suppose $ S $ dosen't contain zero divisors, if one works in the first way, it's easy to prove that the canonical map $ j:R\to R^{-1}S $ is injective. But if one works in the second way, then it's equivalent to state that $ I \cap R = \{ 0 \}$. Can we directly prove this without considering $ (R\times S)/\sim $ ?

$\endgroup$
7
  • $\begingroup$ The statement “there's no nonzero $r$ such that $\operatorname{expression} =r$” where the expression does not depend on $r$, is bizarre: why not write $\operatorname{expression}=0$ then? $\endgroup$
    – Gro-Tsen
    Commented Jan 4, 2022 at 14:39
  • $\begingroup$ It is bizarre indeed, also because the expression is generically nonzero. I think the question needs some rephrasing - as it is written, I cannot tell the correct form $\endgroup$ Commented Jan 4, 2022 at 14:44
  • $\begingroup$ I think your question is equivalent to whether $(a_1x_1-1,+\dots+,a_nx_n-1)\cap R=\{0\}$ (with $(a_1x_1-1,\dots,a_nx_n-1)$ denoting the ideal generated by $a_1x_1-1,\dots,a_nx_n-1$ in $R[x_1,\dots,x_n]$), am I right? This is true if $R$ is a field, and, by passing to the fraction field, also if $R$ is a domain. $\endgroup$ Commented Jan 4, 2022 at 15:36
  • 1
    $\begingroup$ if $r\in R$, $r\ne 0$, and $r=\sum (a_ix_i-i)f_i$, simply put $x_i=1/a_i$ (and multiply by a product of $a_i$ to a large power, so that you deal only with polynomial identity) to get 0 in RHS $\endgroup$ Commented Jan 4, 2022 at 16:54
  • 1
    $\begingroup$ Why would you want to prove a statement about localisations without using localisations? $\endgroup$ Commented Jan 4, 2022 at 17:02

2 Answers 2

3
$\begingroup$

Maybe these are the arguments you're looking for: First reduce to the case of $S$ being finitely generated as a multiplicative subset of $R$. Next, reduce to the case $S$ where $S$ is generated by only one element $s$. The morphism $$ R\rightarrow R[T]/(sT-1) $$ is injective for $s$ a nonzero divisor. As you said, it suffices to prove that $$ (sT-1)\cap R=\{0\}. $$ Since $s$ is a nonzero divisor, $$ \deg((sT-1)F)=\deg(sT-1)+\deg(F)= 1+\deg(F) $$ for all $F\in R[T]$. It follows that $(sT-1)F\in R$ only if $\deg(F)=-\infty$, i.e., $F=0$.

$\endgroup$
1
  • $\begingroup$ Thanks! Now I realize that what's exatly my question: The canonical map $ j:R\to S^{-1}R $ is characterized by the following property: If $ f:R\to T $ is a ring homomorphism that maps every element in $ S $ to a unit in $ T $, then there's a unique ring homorphism $g:S^{-1}R\to T$ such that $ f = g\circ j $. I'm wondering that how to prove $ j $ is injective if $ S $ contains no zero divisors? I know that one could prove this by define an equivalence relation on $ R\times S $, but could we prove this without considering any certain construction? $\endgroup$
    – Yu Li
    Commented Jan 7, 2022 at 5:11
2
$\begingroup$

Wlog assume $n=m$. Set $S:=\{a_1^{k_1}\cdots a_n^{k_n} \mid k_i\in\mathbb{N}\}$. Then $R[x_1,\ldots,x_n]/I$ is precisely the localisation $S^{-1} R$ that inverts $a_1, \ldots, a_n$ and your question is equivalent to asking whether or not the canonical map $R\to S^{-1} R$ is injective.

The kernel of the canonical map is well-known to be equal to $\{r\in R \mid \exists s\in S: sr=0\}$. In particular: It is zero iff none of the elements $a_1,\ldots,a_n$ are zero divisors of $R$.

$\endgroup$
3
  • $\begingroup$ Yes. Given a multiplicatively closed set $ S $ that contains $ 1 $, then there are two ways to construct $ S^{-1}R $. One is to define an equivalence relation on $ R\times S $ and then take quotient, and the other is $ R[S]/I $, where $ R[S] $ is the free commutative algebra over $ R $ generated by $ S $ and, let $ i:S\to R[S] $ the canonical embedding, $ I $ the ideal of $ R[S] $ generated by $ i(s)s-1 $ for each $ s $ in $ S $. It's very easy to prove that the canonical map $j:R\to S^{-1}R$ is injective while $ S $ doesn't contain zero divisor if one consider the first construction. $\endgroup$
    – Yu Li
    Commented Jan 4, 2022 at 17:29
  • $\begingroup$ I mean that could one directly prove that, if $ S $ doesn't contain zero divisors, then $ I \cap R = \{ 0 \} $, without considering $ (R\times S)/\sim $? $\endgroup$
    – Yu Li
    Commented Jan 4, 2022 at 17:41
  • $\begingroup$ Now I realize that what's exatly my question: The canonical map j:R→S−1R is characterized by the following property: If f:R→T is a ring homomorphism that maps every element in S to a unit in T, then there's a unique ring homorphism g:S−1R→T such that f=g∘j. I'm wondering that how to prove j is injective if S contains no zero divisors? I know that one could prove this by define an equivalence relation on R×S, but could we prove this without considering any certain construction? $\endgroup$
    – Yu Li
    Commented Jan 7, 2022 at 5:11

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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