The problem is to show that if $e$ is an idempotent in a ring $R$, then $Re=S^{-1}R$ where $S=\{1,e,e^2,e^3,\dots\}=\{1,e\}$. In fact this doesn't even seem plausible to me, because $Re$ is "smaller" than $R$ (it's a subring of $R$) whereas $S^{-1}R$ is "larger" than $R$ (it's an extension of $R$).

  • $\begingroup$ What is $S^{-1}R$? What is $S^{-1}$? $\endgroup$ – YiFan Apr 25 at 22:22
  • $\begingroup$ $S^{-1}R$ is the localization of $R$ w.r.t. $S$. $\endgroup$ – user437309 Apr 25 at 22:24
  • 3
    $\begingroup$ The idea that $S^{-1}R$ is "larger" than $R$ is not in general correct, because $S^{-1}R$ may collapse some existing elements of $R$ together. $\endgroup$ – Ted Apr 26 at 5:59

I assume you're considering commutative rings for localizations to make sense.

The idea is that $S^{-1}R$ is a "generalization of a field of fractions". If $e^2=e$ in a field, then $e=0$ or $e=1$, and in the latter case $e=e^{-1}$, so it is not unexpected to have something like $Re=Re^{-1}\simeq RS^{-1}$.

Formally, you can prove that $Re$ is (naturally isomorphic to) $S^{-1}R$ using the universal property of localizations: There is a ring homomorphism $\phi\colon R\to Re$ taking $S$ to units, and which is universal (initial) with this property.

To prove this, just take $\phi\colon R\to Re$ as $\phi(r)=re$. Then $\phi$ is a ring homomorphism (because $e$ is idempotent), taking each element of $S$ to the unit $e$ of $Re$.

If $T$ is any other commutative ring and $\psi\colon R\to T$ is a ring homomorphism taking elements of $S$ to units of $T$, define $\psi'\colon Re\to T$ as the restriction of $\psi$ to $Re$. Then $\psi=\psi'\circ\phi$: Indeed, $\psi(e)^2=\psi(e)$, because $\psi$ is a ring homomorphism and $e^2=e$. Since $\psi(e)$ is a unit then $\psi(e)=1$. Given arbitrary $r\in R$, we multiply both sides of the equation "$1=\psi(e)$" by $\psi(r)$ to conclude $$\psi(r)=\psi(re)=\psi'(\phi(r))$$ so $\psi$ factors through $\phi$. This factor is unique since $\phi$ is obviously surjective, and this gives the universal property of $(Re,\phi)$ as the localization of $R$ wrt $S$.

  • $\begingroup$ $\psi(e)$ is an invertible element of $T$, but why must it be equal to $1$? And how do you use $\psi(e)^2=\psi(e)$? $\endgroup$ – user419669 May 1 at 17:05
  • $\begingroup$ @user419669 If $x$ is invertible and $x^2=x$, multiply both sides by $x^{-1}$ to get $x=1$.. $\endgroup$ – Luiz Cordeiro May 1 at 19:43

If $e\in\{0,1\}$ then the claim is clear. Otherwise $e$ is a zero divisor, and so the canonical map $$R\ \longrightarrow\ S^{-1}R,$$ is not injective; the localized ring $S^{-1}R$ is not "larger" than $R$. In particular, because $e(1-e)=0$, the entire ideal $(1-e)R$ is mapped to zero by this map.

In fact the equivalence relation on $R\times S$ that defines $S^{-1}R$, which is in general given by $$\frac{r_1}{s_1}\sim\frac{r_2}{s_2} \qquad\iff\qquad \exists t\in S:\ t(r_1s_2-r_2s_1)=0,\tag{1}$$ greatly simplifies here. Because $S=\{1,e\}$ the right hand side of $(1)$ is equivalent to $$e(r_1s_2-r_2s_1)=0,$$ where also $s_1,s_2\in\{1,e\}$. If $s_1=s_2$ then this is in turn equivalent to $$e(r_1-r_2)=0,$$ and if $s_1\neq s_2$ then without loss of generality $s_1=e$ and $s_2=1$, and we find that $$e(r_1s_2-r_2s_1)=e(r_1-r_2e)=e(r_1-r_2+(1-e)r_2)=e(r_1-r_2),$$ because $e(1-e)=0$. So in fact in this particular case the equivalence relation $(1)$ is equivalent to

$$\frac{r_1}{s_1}\sim\frac{r_2}{s_2} \qquad\iff\qquad er_1=er_2.$$ This shows that the map $$S^{-1}R\ \longrightarrow\ eR:\ \frac{r}{s}\ \longmapsto\ er,$$ is a bijection, and it is easily verified to be a ring homomorphism.


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.