Using Hartshorne's definition of coherence, we should still have that an $\mathcal{O}_X$-module $\mathscr{F}$ is coherent if and only if for every affine open $U=\operatorname{Spec}(A)$ of $X$, there exists a finitely generated $A$-module $M$ such that $\mathscr{F}|_U=\widetilde{M}$, $\textbf{even for $X$ not noetherian}$, because if $(f_1,\ldots,f_r)=A$ and each $M_{f_i}$ is a finite $A_{f_i}$-module, then $M$ is a finite $A$-module even if $A$ is not noetherian (see for example Stacks Project).

But then why does the equivalence of categories between finitely generated $A$-modules and coherent $\mathcal{O}_{\operatorname{Spec}(A)}$-modules fail without the noetherian assumption? The functor $M\mapsto \widetilde{M}$ is fully faithful and essentially surjective.


The stacks project tag you link to is irrelevant to this question. Of course if you simply take the Hartshorne definition of a coherent sheaf even in the non-Noetherian case, you will get the equivalence with finitely generated $A$-modules (by definition).

The main point is that if $A$ is not noetherian, then the subcategory of $QCoh(Spec(A))$ (the category of quasi-coherent sheaves) given by the sheaves $\tilde{M}$, where $M$ is finitely generated, is not abelian. The problem is that kernels will not be finitely generated.

Once you know this you understand that for the correct definition of a coherent sheaf you need to fix this, which leads you to the definition that you find on Wikipedia for example. Note that this definition might exclude $\tilde{A}$ as a coherent sheaf.

Edit: A ring $A$ for which the structure sheaf is a coherent sheaf on $Spec(A)$ is called a coherent ring. See here.

There are very important rings that are not noetherian but coherent. For example the ring of holomorphic functions on a Stein manifold, which leads to Oka's theorem saying that the structure sheaf of a complex manifold is coherent (note that this is a very hard theorem).

Moreover I think that the ring of smooth functions on a real manifold is not coherent. Unfortunately I cannot find a reference. If anyone finds a reference (or a proof) for this, I would be happy to see it.

  • $\begingroup$ I don't think that $\mathcal{O}_{\mathrm{Spec}(A)}=\tilde{A}$ is excluded as a coherent sheaf, regardless which definition you take. This sheaf has global and local trivial finite presentation. $\endgroup$ – user213008 Jan 4 '18 at 12:35
  • $\begingroup$ For $\tilde{A}$ to be coherent you also need that finitely generated ideals of $A$ are finitely presented. This is not always true. Of course in my answer one gets the impression that $\tilde{A}$ is always excluded in the non-noetherian case. This is not true, and I will edit my answer $\endgroup$ – Rieux Jan 4 '18 at 13:11
  • $\begingroup$ Thanks, you are right. $\endgroup$ – user213008 Jan 4 '18 at 13:53

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.