[go: nahoru, domu]

nLab regular category (Rev #21)

A regular category is a finitely complete category which admits a good notion of image factorization. A primary raison d’etre behind regular categories CC is to have a decently behaved calculus of relations in CC. Regular categories are also the natural setting for regular logic.

Definition

A category CC is regular if

  • It is finitely complete;

  • The kernel pair p 1,p 2:edp_1, p_2: e \,\rightrightarrows\, d of any f:dcf: d \to c admits a coequalizer; and

  • The pullback of a regular epi along any map is a regular epi.

Here the “kernel pair” is the parallel pair of projection maps coming out of a pullback ee of the diagram

dfcfdd \stackrel{f}{\to} c \stackrel{f}{\leftarrow} d

The kernel pair is always an congruence on dd in CC; informally, ker(f)\ker(f) is the subobject of d×dd \times d consisting of pairs of elements which have the same value under ff (sometimes called the ‘kernel’ of a function in Set\Set). The coequalizer above is supposed to be the “object of equivalence classes” of ker(f)\ker(f) as an internal equivalence relation.

A map which is the coequalizer of a parallel pair of morphisms is called a regular epimorphism. In fact, in any category satisfying the first two conditions above, every coequalizer is the coequalizer of its kernel pair. (See for instance Lemma 5.6.6 in Practical Foundations.)

The last condition may equivalently be stated in the form “coequalizers of kernel pairs are stable under pullback”. However, it is not generally true in a regular category that the pullback of a general coequalizer diagram

edce \stackrel{\to}{\to} d \to c

along a morphism ccc' \to c is again a coequalizer diagram.

To form the image factorization of a map f:dcf: d \to c, let q:dkq: d \to k be the coequalizer of the kernel pair of ff. Since ff coequalizes its kernel pair, there is a unique map i:kci: k \to c such that f=iqf = i q. It may be shown from the regular category axioms that ii is monic and in fact represents the image of ff, i.e., the smallest subobject through which ff factors. Moreover, the classes of regular epis and of (all) monomorphisms form a factorization system.

In fact, a regular category can alternately be defined as a finitely complete category with pullback-stable image factorizations. See familial regularity and exactness for a generalization of this approach to include coherent categories as well.

Examples

  • Set is a regular category. In fact, any topos is regular. More generally, a locally cartesian closed category with coequalizers is regular, and so any quasitopos is regular.

  • The category of models of any finitary algebraic theory (i.e., Lawvere theory) TT is regular. This applies in particular to the category Ab of abelian groups.

  • Any abelian category is regular.

  • If CC is regular, then so is C DC^D for any category DD.

Examples of categories which are not regular include Cat, Pos, and Top.

Remarks

  1. As exactness properties go, the ones possessed by general regular categories are fairly moderate; the main condition is of course stability of regular epis under pullback. In the first place, the focus is just on certain coequalizers; finite coproducts aren’t even mentioned. Some of that imbalance is redressed by the notion of lextensive category, where coproducts are also stable under pullback (and are also disjoint).

Further desirable exactness properties can be phrased in the language of Galois connections. For each object dd, consider the following relation coeqcoeq between the class of parallel pairs f,g:edf, g: e \stackrel{\to}{\to} d) and maps h:dch: d \to c:

(f,g);hcoeqiffhf=hg\langle (f, g); h \rangle \in coeq \qquad iff \qquad h f = h g

[to be continued]

Making categories regular

Any category CC with finite limits has a reg/lex completion C reg/lexC_{reg/lex} with the following properties:

  • There is a full and faithful functor CC reg/lexC\hookrightarrow C_{reg/lex}
  • Each object of CC becomes projective in C reg/lexC_{reg/lex}
  • Each left-exact functor CDC\to D, where DD is regular, extends to an essentially unique regular functor C reg/lexDC_{reg/lex}\to D.

In particular, the reg/lex completion is a left adjoint to the forgetful functor from regular categories to lex categories (categories with finite limits). The reg/lex completion can be obtained by “formally adding images” for all morphisms in CC, or by “closing up” CC under images in its presheaf category [C op,Set][C^{op},Set]. In general, even if CC is regular, C reg/lexC_{reg/lex} is larger than CC (that is, it is a free cocompletion rather than merely a completion), although if CC satisfies the axiom of choice (in the sense that all regular epimorphisms are split), then CC reg/lexC\simeq C_{reg/lex}.

See regular and exact completions.

H'm, so what is Set reg/lexSet_{reg/lex} like if the axiom of choice fails in SetSet? Some category like SetSet, but larger, in which every set becomes projective (but perhaps not every object in the larger category). Of course, I'd have to check that the statements above that this conclusion relies don't themselves use the axiom of choice in SetSet! —Toby

Mike Shulman: I believe that Set reg/lexSet_{reg/lex} admits the following simple description: it is the category of setoids (X,)(X,\equiv) where a morphism (X, X)(Y, Y)(X,\equiv_X)\to (Y,\equiv_Y) is a function f:XYf\colon X\to Y such that xxx\equiv x' implies f(x)f(x)f(x)\equiv f(x'), and two such morphisms ff and gg are set equal if f(x)g(x)f(x)\equiv g(x) for all xXx\in X. In this case Set reg/lexSet_{reg/lex} is already exact, and therefore it is the same as Set ex/lexSet_{ex/lex}. The monics in Set reg/lexSet_{reg/lex} are morphisms ff such that f(x)f(x)f(x)\equiv f(x') implies xxx\equiv x', and the regular epics are the morphisms f:(X, X)(Y, Y)f\colon (X,\equiv_X)\to (Y,\equiv_Y) such that there exists a function g:YXg\colon Y\to X such that f(g(y))yf(g(y))\equiv y for all yYy\in Y. Evidently every set, considered as a setoid in the obvious way, is then regular projective. And of course, if SetSet satisfies AC, then every setoid is isomorphic to a set in Set reg/lexSet_{reg/lex}.

Toby: This is very interesting! I should check that your construction satisfies the universal property, but for the moment I'll just believe it. Then this construction mimics the construction of sets as presets with equivalence relations. If you adopt the axiom of choice, then that construction stabilises; but otherwise, it seems that you can continue it over and over again! (Perhaps we can get a family of weaker and weaker versions of AC by postulating that the construction stabilises at some point.)

Mike Shulman: Yes, your question started me thinking in that direction too. This category Set ex/lexSet_{ex/lex} of setoids is interesting in some other ways. It’s not only exact, I’m pretty sure that it’s a Π-pretopos (assuming that SetSet is). However, there’s no reason for it to be a topos even if SetSet is; in fact it may not even be well-powered (relative to SetSet)! In general the subobject lattice Sub Set ex/lex(X)Sub_{Set_{ex/lex}}(X) of a set (considered as a setoid) is the preorder reflection of Set/XSet/X. If SetSet satisfies AC, then of course any YXY\to X is isomorphic in this poset reflection to its image, but in the complete absence of AC I don’t see any reason for this preordered class to be small. So “Set ex/lexSet_{ex/lex} is a topos” would also be an interesting weaker form of AC.

Additionally, in Set ex/lexSet_{ex/lex} the terminal object 11 is a generator which is nondegenerate, projective, and indecomposable, but not (in general) a strong generator. Thus, it is not well-pointed in the correct sense for a pretopos. In fact, I believe saying that 11 is a strong generator in Set ex/lexSet_{ex/lex} is equivalent to AC in SetSet.

Maybe we should move this discussion to a page like exact completion of Set?.

Regular categories of the form C reg/lexC_{reg/lex} for a lex category CC can be characterized as those regular categories in which every object admits both a regular epi from a projective object and a monomorphism into a projective object, and the projective objects are closed under finite limits. In this case CC can be recovered as the subcategory of projective objects. In fact, the construction of C reg/lexC_{reg/lex} can be extended to categories having only weak finite limits, and the regular categories of the form C reg/lexC_{reg/lex} for a “weakly lex” category CC are those satisfying the first two conditions but not the third.

When the reg/lex completion is followed by the ex/reg completion which completes a regular category into an exact one, the result is unsurprisingly the ex/lex completion.

Regular categories were introduced by Barr in Exact categories, by M. Barr, Lecture Notes in Mathematics, Springer-Verlag 1971 and by Grillet in the same volume of Lecture Notes in Mathematics.

Some of the history is provided in Topos Theory, by P. Johnstone, 1977.

An application of the regularity condition is found in the paper Tensor envelopes of regular categories, by F. Knop. arXiv:math/0610552v2

Knop’s condition for regularity is slightly different from that presented here; he works with categories that when augmented by an absolutely initial object are regular in the terminology here. In the paper, Knop generalizes a construction of Deligne by showing how to construct a symmetric pseudo-abelian tensor category out of a regular category through the calculus of relations.

Revision on November 5, 2009 at 19:15:53 by Mike Shulman See the history of this page for a list of all contributions to it.