The fundamental localic groupoid of a topos (Q1184023): Difference between revisions

The notion of a fundamental group of a Grothendieck topos is defined in different ways by various authors, beginning with \textit{A. Grothendieck} [Revètements étales et groupe fondamental, SGA 1, Lect. Notes Math. 224 (1971; Zbl 0234.14002)], who defined it as a pro-group in the pointed case. The present paper deals with the unpointed case, as also do \textit{I. Moerdijk} and \textit{G. Wraith} [Trans. Am. Math. Soc. 295, 849-859 (1986; Zbl 0592.18003)] and the reviewer [``Classifying toposes and fundamental localic groupoids'', in: (\textit{R. A. G. Seely} (ed.)) Category theory 1991, Can. Math. Soc. Conf. Proc. 13, 75-96 (1992)]. In this paper, to a connected Grothendieck topos \({\mathcal E}\) a pro- discrete localic groupoid \(\pi\) is associated that plays the role of the fundamental group in that the classifying topos \(B\pi\) classifies torsors in \({\mathcal E}\) for discrete groups. The basic approach is to define \(\pi\) as some kind of 2-limit of the projection from a 2-category \(\text{Tors}({\mathcal E})\) to Groups. It is then shown that this 2-diagram of groups has an equivalent 1-diagram of groupoids with an equivalent limit. This gives the fundamental group of \({\mathcal E}\) as a pro-groupoid. It is then shown that such a pro-groupoid is represented by a pro- discrete localic groupoid which classifies torsors and gives a universal torsor, an analogue of the universal covering space. As compared with the other two treatments mentioned earlier, the present one does not assume the topos \({\mathcal E}\) to be locally connected and is thus led to make an explicit use of the uniform structure on the fundamental group. On the other hand, and unlike the alternative treatments mentioned above, the present one is highly dependent on the base topos \({\mathcal S}\) taken to be ``the'' topos of sets and functions. In addition, it differs from the reviewer's treatment [loc. cit.] first in that, in the latter, the fundamental group arises from a Čech-style definition of topos cohomology and thus is based on coverings rather than on torsors, and secondly in that the localic groupoid \(\pi\) is given by the reviewer as a direct application of the Joyal-Tierney representation theorem [\textit{A. Joyal} and \textit{M. Tierney}, ``An extension of the Galois theory of Grothendieck'', Mem. Am. Math. Soc. 309 (1984; Zbl 0541.18002)] rather than by some technical results specifically proven for this purpose. Other than that, the two definitions must be equivalent, at least over Sets, since both represent cohomology. In contrast to the previous two treatments, Moerdijk and Wraith [loc. cit.] define the fundamental groupoid of \({\mathcal E}\) as the topos analogue of the original notion of fundamental group and, as such, depends on an appropriate notion of path. In the connected locally connected case and in the presence of a universal covering, the paths definition given by Moerdijk and Wraith has been shown by the reviewer [``Universal covering localic toposes'', C. R. Math. Acad. Sci., Soc. R. Can. 14, No. 6, 245-250 (1992)] to be equivalent to her coverings definition [loc. cit.]. The paper being reviewed is not too easy to read, partly on account of its technical aspects. However, precisely some of the 2-categorical technical results, notably Theorem 3.10, should prove useful for other purposes. A highly positive aspect of this paper are its examples. A couple of these examples show, on the one hand, the necessity of considering localic groupoids even in the locally connected case, and on the other, that one cannot expect too good a behaviour if one goes beyond the locally connected case.