I am searching a module $M$ over a (von Neumann) regular ring $A$ ($\forall a\in A$, $\exists x\in A$: $axa=a$) with two properties:

(1) every finitely generated submodule of $M$ is projective (isomorphic to a direct summand of some $A^n$);

(2) at least one finitely generated [projective] submodule of $M$ is not a direct summand of $M$.

[Equivalent re-formulation: a Mittag-Leffler, not strictly Mittag-Leffler module $M$ over a regular ring $A$. Another re-formulation: a sectionally complemented modular lattice $L=L(A^4\oplus M)$, with a large quadruplicable element,such that not every element of $L$ is complemented in the ideal completion. These re-formulations are not needed in what follows]

It is well known that models of ZF (without choice) exist where even vector spaces give such examples: real numbers as vector space over the rational numbers (in models of ZF where all additive endomorphisms of the reals are linear); or: amorphous sets with a vector space structure over a finite field. These examples even have a property stronger than (2):

(2') there is no nonzero linear functional on $M$ (hence no nonzero finitely generated [projective] submodule is a direct summand)

Now it comes the tricky part. Suppose that, instead of needing models that negate AC, I were interested in some easier models, such that a Boolean valued universe would be sufficient (without the need to consider a symmetric sub-model). Then vector spaces in the ZFC models obtained form a Boolean valued universe are easily re-interpretable as coming from modules over [strongly] regular rings (whose Boolean algebra of [central] idempotents is the complete Boolean algebra of truth values for the Boolean valued universe).

The questions, returning to models of ZF without choice:

Is there an analogous reinterpretation (existing in every model of ZFC) for the case I am interested in (vector spaces in models of ZF without choice)?

If such a reinterpretation is possible, what should I study to understand it? [Note that I am not even able to have a intuitive mental picture of the topos of $G$-sets for a group $G$]

If such a reinterpretation is possible, is the syntactical form of the properties (1), (2), (2') such that their validity in the ZF model automatically implies validity in the reinterpretation?


I cannot give you any specific example, but at least over nontrivial countable regular rings such modules always exist.

We work over a countable regular ring $A$ which is not semisimple Artinian, and our modules are, for instance, right $A$-modules. In this case, $A$ cannot be injective as a module over itself.

The first ingredient is Lemma 9.10 from [1] which says that, if you take any module $F$ such that there exists a short exact sequence $0 \to A \to M \to F \to 0$ that does not split, then $M$ is not strict Mittag-Leffler. Since $A$ is not injective, we can always find such a module $F$.

The very same lemma also ensures that $M$ is Mittag-Leffler whenever $F$ is such. The question is: can we find $F$ which is Mittag-Leffler and at the same time $\mbox{Ext}^1(F,A)\neq 0$?

This is not easy since such $F$ cannot be countably presented: indeed, every countably presented (flat) Mittag-Leffler module is projective, hence strict Mittag-Leffler.

To find the right $F$ (or to be more precise, to prove its existence), you can use Theorem 6, in particular the moreover-clause statement (1), from [2]. In the setting of countable regular rings, it says that any module $N$ ($A$ in particular) which satisfies that $\mbox{Ext}^1(G,N) = 0$, for all Mittag-Leffler modules $G$, has to be injective.

Final remark: It is an open problem whether the statement (1) in Theorem 6 holds for uncountable rings. On the other hand, in your case, the assumption that $A$ is countable can be replaced by `there exists a countably presented module $C$ such that $\mbox{Ext}^1(C,A) \neq 0$'.

[1] L. Angeleri Hügel, D. Herbera: Mittag-Leffler conditions on modules, Indiana Univ. Math. J. 57 (2008), 2459-2518.

[2] S. Bazzoni, J. Šťovíček: Flat Mittag-Leffler modules over countable rings, Proc. Amer. Math. Soc. 140 (2012), 1527-1533.


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.