Take the 2-minute tour ×
MathOverflow is a question and answer site for professional mathematicians. It's 100% free, no registration required.

Let $M$ be an $R$-module,where $R$ is a hereditary (or cohomological dimension less or equal to 1).Take $E(R)$ to be injective hull of $R$, then we have the essential extension $i:R^I\rightarrowtail E(R)^I$ (product $I$ times) and we also have $p:R^I\twoheadrightarrow M$ is epimorpshim. Then I take the push forward of these two morphisms $i$ and $p$, denote the push out by $(N,f,g)$, where $f:M\to N$ and $g:E(R)^I\twoheadrightarrow N$.

It is clear that $N$ is an injective module (because it is image of $E(R)^I$ and $R$ is hereditary) and $f:M\to N$ is injective morphism. However, $N$ is not necessarily injective hull of $M$ because in general, essential extension does not commutes with colimit.

My question is: can we give some conditions to $R$ or other extra conditions to make N is injective hull of $M$?

In general, I know it is not true, but it seems that it gives a approximation of "functor" $M \to E(M)$

share|improve this question

1 Answer 1

First, I want to point out that in general there is no surjection from a direct product of copies of $R$ to an arbitrary module $M$. For example, if $R=\mathbb{Z}$ and $M=\mathbb{Z}^{(\omega)}$ (a countable direct sum of copies of $\mathbb{Z}$), then there is no surjection $\mathbb{Z}^I\to M$ according to the paper "Extension of a theorem on direct products of slender modules" by John D. O'Neill. [There are probably much simpler examples, but this will do.]

Thus, in general, there is no map $p$, and thus no pushout $N$.

Second, maps defined from $R^{I}$ are notoriously difficult to understand, and often depend on cardinality considerations for the set $I$. In particular, when $R=\mathbb{Z}$, things get really weird when $|I|$ is measurable.

That all said, assume $p$ does exist. Since, as you pointed out, $N$ is an injective module, we just need to examine when $N$ is an essential extension of $M$ (viewing $M$ as a submodule). One obvious situation when this holds is when $R$ itself is a (right) self-injective ring, for in that case $N=M$ is already injective. Of course, this case is somewhat trivial since self-injective hereditary rings are already semisimple.

More generally, write

$$N=M\oplus E(R)^I/\langle (p(x),-i(x))\ :\ x\in R^I \rangle.$$

For each $e\in E(R)^I\setminus\{0\}$, let $X_e:=\{r\in R\ :\ er\in R^I\}$. The set $E_e=\{er\ :\ r\in X_eR\}$ is a submodule of $R^I$. Given any element $\overline{(m,e)}\in N\setminus M$, we need $\overline{(m,e)}R\cap M\neq (0)$. Thus, we need some element $r\in X_e$ such that $mr\neq -p(er)$. In other words, we do not want $p|_{E_e}$ to extend to a map $eR\to M$. Thus, whatever condition you impose will need to restrict what types of maps $p$ are available.

share|improve this answer

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

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