In Peter Freyd's Abelian Categories, it is mentioned in passing that the category of abelian groups (more generally, $R$-modules for a ring $R$) satisfy the axiom AB5:

For each linearly ordered family $\{S_i\}_I$ in the lattice of subobjects of an object $S$, and $T$ is any subobject of $S$, then we have $$T \cap \bigcup S_i = \bigcup (T \cap S_i).$$

Isn't this completely obvious? Let $r \in RHS$. Then $r \in LHS$. Conversely if $r \in LHS$ then $r \in RHS$. (The only non-triviality is that the union of a family of submodules is a module in the first place, and this holds because the family is linearly ordered.)

I am confused because here https://ncatlab.org/nlab/show/Grothendieck+category#Kiersz they link to a 9 page paper that proves that the category of $R$-modules satisfies what they say is an equivalent condition to AB5 ('small filtered colimits are exact').

  • 1
    $\begingroup$ This is obvious for the case at hand because $\cap $ and $\bigcup$ for goups or $R$-,odules happen to be simply the set-theoretic intersection and union of he underlying set (and same underlying set means same module as long as we speak about submodules of a fixed module) $\endgroup$ – Hagen von Eitzen Dec 15 '18 at 4:38
  • $\begingroup$ Thank you! May i ask then why in that link I gave (ncatlab.org/nlab/show/Grothendieck+category#Kiersz), under the section of examples they say "for $R$ a commutative ring, its category of modules $R$-Mod is a Grothendieck category. (see e.g Kiersz 06, prop. 4 for the proof that filtered colimits here are exact.)"? I mean to ask, is it really the case that their version of AB5 is that much harder to prove than the one I gave? $\endgroup$ – SSF Dec 15 '18 at 4:41

Yes, you are correct. The formulation of AB5 used on the nlab page you linked is substantially different from Freyd's (though it turns out to be equivalent) and so it takes a bit more work to prove it for modules over a ring. Note that the 9-page paper you refer to also starts by introducing all the concepts needed to formulate the statement from scratch. The actual proof of the final result (Proposition 4) is quite short and the important tools it uses from the previous 8 pages are just basic background on how to concretely identify all of the relevant category-theoretic notions in the case of modules over a ring (which is comparable to how you tacitly used the fact that the lattice of subobjects of a module can be identified with its usual lattice of submodules).


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.