Semi-abelian categories
Posted by Andreas Holmstrom on November 1, 2009
The usual setting for doing homological algebra is abelian categories. However, many of the things one can do in abelian categories also make sense in more general settings. For example, the category of groups is not abelian, but one can still make sense of exact sequences, diagram lemmas, and so on.
A more general framework for doing homological algebra, which I first learnt about from Julia Goedecke, is given by the notion of semi-abelian categories. Some examples of semi-abelian categories are: groups, compact Hausdorff spaces, crossed modules, Lie algebras, any abelian category, and any category of algebras over a reduced operad (although I am not sure what it means for an operad to be reduced).
A very nice introduction and survey of semi-abelian categories can be found in the recent article of Hartl and Loiseau, on the arXiv. Other references include the nLab page and the thesis of Van der Linden.
The category of monoids is unfortunately not semi-abelian, but there was an interesting discussion on Math Overflow recently about making sense of homological algebra in the category of commutative monoids, which is interesting when trying to do algebraic geometry over the field with one element.
Motivic stuff 
James Griffin said
Hi Andreas,
According to Berger and Moerdijk (link) p8, a reduced operad P over a symmetric monoidal category E is an operad such that P(0) is the unit of E.
However this would mean that the category of unital associative algebras is semi-abelian and I don’t think that it is, so perhaps there’s another definition of reduced where P(0) should be the 0 object of E, where E is a symmetric monoidal semi-abelian category. This means that the category of non-unital associative algebras is semi-abelian, which is good because it is. Kernels are ideals, which are ofcourse also associative algebras themselves.
Obvious questions now arise, such as what are the homotopy kernels, the loop spaces etc. and what do they mean? Actually it turns out not much, but it’s still nice to know that they’re there.
See you at Geometry Tea,
James
Andreas Holmstrom said
Thanks!!