The purpose of this note is to prove the conjecture in [M], viz. that if F → E is a connected and locally connected morphism between (Grothendieck) toposes then F → E is an open surjection. In [M], the following partial results were proved (cf. also Remark (ii) below): (1) If F → E is a connected and locally connected, then F → E is a stable surjection (between toposes). (2) If Y → X is a connected and locally connected map between locales, then Y I → X is an open surjection (between locales). Here, as in [M], I denotes the unit interval [0, 1], viewed as a locale or as a topos (viz. the topos of sheaves of sets on [0, 1]). I would like to gratefully acknowledge that my attention was drawn again to this issue in relation to my work with Bunge on the paths-fundamental group of a topos [BM], and with Kock on etendues. Indeed, it is the stronger result to be proved here which is needed in these applications. (Without going into details, the point is that this result implies that for a localic groupoid G with connected and locally connected source and target maps, one can now immediately obtain a groupoid representation for the path-space of the classifying topos of G, by the equivalence B(G) ' B(G); I assume here that G is etale complete, of course.) Consider for a topos E its “canonical” connected and locally connected cover XE → E by a locale, where XE = E [En(G)] is the topos of E-internal sheaves on the locale En(G) of infinite-to-one partial enumerations of a generating object G in E (see [JM]). Recall that this construction is “natural in E” in the sense that it is the pullback of the generic case where E is the object classifier S[U ] and G is the universal object U . I will show:
Read full abstract