8
$\begingroup$

For two unital rings $R,S$, let $R\sim S$ denote that $R$ and $S$ are Morita equivalent, i.e. the categories $\mathrm{Mod}_R$ and $\mathrm{Mod}_S$ are equivalent. Examples include:

  • Matrix rings: $R\sim M_n(R)$ for any $n\geq 1$,
  • Full corners: $R\sim eRe$ whenever $e$ is a full idempotent in $R$, i.e. $e=e^2$ and $ReR=R$.

The former is an example of the latter, since $R\simeq \varepsilon_{11}M_n(R)\varepsilon_{11}$ where $\varepsilon_{11}\in M_n(R)$ is the idempotent with $1$ in the $(1,1)$ entry and $0$'s everywhere else.

It turns out these are essentially the only ways to construct Morita equivalent rings.

Theorem. $R\sim S$ if and only if $S\simeq eM_n(R)e$ for some $n\geq 1$ and some full idempotent $e\in M_n(R)$.


If $R$ is a unital ring, a proper ideal $P\subseteq R$ is prime if $IJ\subseteq P$ implies $I\subseteq P$ or $J\subseteq P$ for all ideals $I,J$. Let $\mathrm{Spec}(R)$ denote the set of prime ideals of $R$, called the prime spectrum of $R$, with the usual Zariski topology: the closed sets are

$$V(I):=\{ P\in\mathrm{Spec}(R) : P\supseteq I\}$$

for $I$ an ideal of $R$.


My question is: if $R$ and $S$ are Morita equivalent rings, then is $\mathrm{Spec}(R)$ homeomorphic to $\mathrm{Spec}(S)$? I'm also wondering if Morita equivalence determines the poset of ideals. By the above Theorem, we just have to investigate the cases where $S=M_n(R)$ or $S=eRe$.

It's not hard to see that $I\mapsto M_n(I)$ is an order-preserving bijection between ideals of $R$ and ideals of $M_n(R)$, which is multiplicative in the sense that $M_n(IJ)=M_n(I)M_n(J)$, and this multiplicativity implies that it takes prime ideals to prime ideals. It's also easily seen to be continuous when restricted to $\mathrm{Spec}(R)$. So we can conclude that

$$\mathrm{Spec}(R) \simeq \mathrm{Spec}(M_n(R)).$$

Now how about full corners? If $eRe$ is not full then it's easy to see that it's false: for example $S=\mathbf{Z}$ is a corner of $R=\mathbf{Z}\oplus \mathbf{Z}$ where $e=(1,0)\in R$, and yet $\mathrm{Spec}(\mathbf{Z})\not\simeq \mathrm{Spec}(\mathbf{Z}\oplus\mathbf{Z})$. This corner is not full though.

Here's my idea: we have a map $$\Phi:\{\text{ideals of } R\} \rightarrow \{\text{ideals of } eRe\}$$ $$I\longmapsto eIe$$ which is surjective --- a right inverse is given by $\Psi(J):=RJR$.

  • Is $\Phi$ injective if $eRe$ is a full corner?
  • Does $\Phi$ preserve prime ideals? Would it be Zariski-continuous?

EDIT: Clearly order-preserving maps are Zariski-continuous, so $\Phi$ is definitely continuous.

$\endgroup$
5
$\begingroup$

I figured it out --- all my dreams were true.


$\Phi$ is injective

Since $e$ is full we have $ReR=R$. Now for any ideal $I$ of $R$ we have

$$I=RIR=(ReR)I(ReR) = (Re)I(eR) = R(eIe)R$$

thus observing that $eIe$ completely determines $I$. So $I\mapsto eIe$ is injective.


$\Phi$ is multiplicative

We want to show that $eIJe =(eIe)(eJe)$. But this is again fullness of $e$.

$$eIJe = (eI)R(Je) = (eI)(ReR)(Je) = eIeJe = (eIe)(eJe).$$

We conclude $\Phi$ restricts to a bijection $\mathrm{Spec}(R)\rightarrow \mathrm{Spec}(eRe)$. But $\Phi$ and $\Phi^{-1}$ are order-preserving, in particular they are Zariski-continuous. So $\Phi$ gives a nice homeomorphism

$$\mathrm{Spec}(R)\simeq \mathrm{Spec}(eRe).$$

$\endgroup$
  • $\begingroup$ A natural followup question is to write down a definition of the prime spectrum solely in terms of $\text{Mod}(R)$. I think such a definition is known but I'm not personally familiar with it. $\endgroup$ – Qiaochu Yuan Oct 7 '15 at 21:46

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.