The classical version of the van Kampen theorem is concerned about the fundamental group of a based space. In fact, it says that the functor $\pi_1$ preserves certain types of pushouts in $Top_*$.

There is also a generalization of the van Kampen theorem that holds for the fundamental groupoid of a space $X$, where in this case it states precisely that the fundamental groupoid functor, $\Pi$, preserves certain colimits in $Top$, namely, those that arise from "nice" open coverings of $X$.

The "groupoid" version of the van Kampen theorem seems to me more conceptual and more elegant than the classical version. Also, the groupoid version allows one to prove the classical version in a more or less easier way.

Although, apart from the conceptual advantages of the groupoid version of van Kampen theorem, I would like to know if we are able to do any interesting calculations using the fundamental groupoid version of the van Kampen. In fact, explicitly describing a groupoid as a colimit of "simpler" groupoids is something that it is not clear at all for me. I would like to know some concrete cases where is it possible to describe the fundamental groupoid of a space, using this generalization form of the van Kampen theorem, and, if possible, to calculate the fundamental group directly from our calculation of $\Pi(X).$

  • 6
    $\begingroup$ Have you searched in this site (and in math.stackexchange.com) for relevant questions and answers? I am sure this has been discussed before —in particular, by Ronnie Brown. $\endgroup$ Jul 12 '14 at 10:43
  • 5
    $\begingroup$ The fundamental groupoid functor preserves all homotopy colimits, and that's very nice. $\endgroup$ Jul 13 '14 at 10:33
  • $\begingroup$ This appears to be a duplicate question: mathoverflow.net/questions/40945/… $\endgroup$ Oct 17 '16 at 17:39
  • $\begingroup$ See also mathoverflow.net/questions/220561/… for an opinion of Grothendieck. Is it possible to construct examples where a groupoid $G$ is "computable" but some of its vertex groups are not computable? $\endgroup$ Aug 24 '18 at 10:49

I confess, in 1965 or so, I first thought that the version of the SvKT (Seifert-van Kampen Theorem) for the fundamental groupoid enabled us to get rid of base points. But then I wanted to calculate the fundamental group of the circle, and gradually realised that we needed $\pi_1(X,A)$, the fundamental groupoid on a set of $A$ of base points chosen according to the geometry. and for which covering space methods are not ideal. In fact one needs combinatorics and combinatorial group(oid) theory to calculate individual $\pi_1(X,a)$ from $\pi_1(X,A)$. See my book Topology and Groupoids and also the classic 1971 book (downloadable) by Philip Higgins, Categories and Groupoids.

Groupoids model homotopy 1-types. So one first determines the 1-type before calculating an individual fundamental group.

This idea is nicely modelled in higher dimensions: one can calculate some 2-types and higher types as "big" algebraic objects inside which are the homotopy groups which one might want. The methods are explained more in a talk given in Paris on June 5, 2014, at the IHP, available on my preprint page.

As Mariano points out, this has been discussed elsewhere on stackexchange and mathoverflow.

July 14: The most general version of the SvKT is given in [41] (downloadable) on my publication list,

R. Brown and A. Razak, `A van Kampen theorem for unions of non-connected spaces'', Archiv. Math. 42 (1984) 85-88.

in the form of a coequaliser statement when given an open cover and a set $A$ (of ``base points'' which meets each path component of each 1-,2-, and 3-fold intersections of the sets of the cover. The style of proof goes back to Crowell's original version, and has the advantage of generalising to higher dimensions. For example, in

[32] R. Brown and P.J. Higgins, ``Colimit theorems for relative homotopy groups'', J. Pure Appl. Algebra 22 (1981) 11-41.

The retraction from the version for the full fundamental groupoid, $A=X$, to this version is quite difficult to manage and is done in May's "Concise..." book, without the refinement on the conditions, but I feel is the wrong way to go, though it is quite elegant for the pushout version in dimension 1.

July 17, 2014: Actually I have missed out on three reasons why one is interested in the fundamental groupoid $\pi_1 X$.

  1. The notion of fibration of groupoids is relevant to topology particularly in construction of operations of groupoids on homotopy sets, and exact sequences. If $p: E \to B$ is a fibration of spaces then $\pi_1 p: \pi_1 E \to \pi_1 B$ is a fibration of groupoids. This is exploited in Chapter 7 of Topology and Groupoids. See for example arXiv:1207.6404 for other current uses of the algebra of groupoids, and in particular fibrations.

  2. Similarly if $p: E \to B$ is a covering map of spaces then on applying $\pi_1$ we get a covering morphism of groupoids. Thus a map is modelled by a morphism, and this often makes the theory easier to follow (IMHO!), particularly with regard to questions of lifting maps. See Chapter 10 of T&G.

  3. If $G$ is a (discrete) group acting on a space $X$ then it also acts on the fundamental groupoid $\pi_1 X$. So we have not only orbit spaces $X/G$ but also orbit groupoids $(\pi_1 X)/\!/G$. There is a canonical morphism $(\pi_1 X)/\!/G \to \pi_1(X/G)$ and there are useful conditions which ensure this is an isomorphism, e.g. $X$ is Hausdorff, has a universal cover, and the action is properly discontinuous. See Chapter 11 of T&G. One example given there is the cyclic group $Z_2$ acting on $X \times X$ whose orbit space is the symmetric square of $X$. Under useful conditions, its fundamental group is that of $X$ made abelian. It would be good to see lots more examples.

Aug 17, 2016: I can now refer to my answer to my own mathoverflow question on the relation of the van Kampen Theorem to the notion of descent.

Oct 17, 2016: I should add that a whole area of research into the use of strict higher groupoids in algebraic topology, one part of which is described in the book Nonabelian Algebraic Topology, (EMS 2011, 703 pages), arose out of seeking generalisations to higher dimensions of the use of the fundamental groupoid.

December 23, 2016 It may be useful to point out that a new volume by Bourbaki "Topologie alg'ebrique" Ch 1-4, (Springer) 2016, uses the fundamental groupoid extensively, and relates its use to descent theory. It does have results on orbit spaces, but no example of applications. It does not use the fundamental groupoid on a set of base points.

December 9, 2019 It should be remembered that homotopy groups of a pointed space were introduced at the 1932 ICM at Zurich, by E. Cech, but the idea was not welcomed because of their abelian nature, so that they did not seem satisfactory higher dimensional versions of the fundamental group. The idea was taken up by Hurewicz, and the fascination and utility of homotopy groups led to the idea of a nonabelian version being regarded as a mirage, though the early homotopy theorists were fascinated by the action of the fundamental group on the higher homotopy groups (a comment of J,H.C. Whitehead, 1958).

We now know that you can construct higher analogues of the fundamental groupoid for certain structured spaces, e.g. filtered spaces and $n$-cubes of pointed spaces. For the filtered case, see this 2011 book.

Note that the fascination and difficulty of the study of higher homotopy groups may, since these groups are defined only for pointed spaces, have deterred people from considering the possible uses of the many pointed case. Also the use of groupoids in algebraic topology has been in the past dismissed by many.

Nonetheless, there is the basic fact of life (to use a term of JHCW) that while "group objects in the category of groups are abelian groups", that is not so if the word "group" is replaced by "groupoid". In view of the importance of group theory in maths and science, it is reasonable to ask of the potential significance of that basic fact of life.

The questioner asks: " explicitly describing a groupoid as a colimit of "simpler" groupoids is something that it is not clear at all for me". This is answered in Appendix B of Nonabelian Algebraic Topology.

More information on the history of fundamental groupoids and higher homotopy groupoids is in the Open Access article available from its link Mathematics Intelligencer.

Can someone explain why the more general theorem in NOT usually given?


I'm very fond of Ronnie Brown's proof of the Jordan curve theorem using the groupoid version of the van Kampen theorem. I find it easier to understand and reproduce than other comparably elementary proofs, such as Munkres' argument using covering spaces (which can be found in his textbook Topology: a first course). I guess this application isn't really a calculation, as requested, but I hope you like it.

(I guess I should mention that a small gap in the proof is filled in in this short note.)

  • $\begingroup$ I really appreciate your answer. Thanks! $\endgroup$ Jul 15 '14 at 19:06
  • $\begingroup$ Of course, one of the things that's going on here is that covering spaces are very closely related to groupoids: the fundamental group lifts to a groupoid of paths between the preimages of the base point. Probably (though I haven't checked) the groupoid and covering space versions are more or less the same proof from different points of view. $\endgroup$
    – HJRW
    Oct 30 '18 at 13:23
  • $\begingroup$ @HJRW The van Kampen theorem for a set of base points calculates examples of unions where the subspaces may be bad locally and so have no reasonable covering spaces. Further, as Grothendieck writes in 1984 in "Esquisses ..." Section 2 : "...people still obstinately persist,when calculating with fundamental groups, in fixing a single base point, instead of cleverly choosing a whole packet of points which is invariant under the symmetries of the situation,.." Why throw away information? $\endgroup$ Jan 1 '19 at 22:40
  • $\begingroup$ @RonnieBrown : I would argue that (carefully) throwing away information is exactly what algebraic topology is all about. After all, that’s what an invariant is. Your point about pathological spaces is well taken. But, for non-pathological spaces, I still guess that the two proofs are essentially the same. $\endgroup$
    – HJRW
    Jan 2 '19 at 1:05
  • $\begingroup$ @HJRW I got into groupoids through writing a text on topology in the 1960s and getting dissatisfied with a van Kampen theorem that did not compute the fundamental group of the circle. It turned out that one needed 2 base points for this! Going down to 1 base point does not work for this example! After a 1967 meeting with George Mackey, I decided to write an account of covering spaces via groupoids, as in the first edition. Later, after work of Higgins and Taylor, I added orbit spaces. Your guess seems wrong to me! And higher dim groupoids can generalise deep work of JHCW! $\endgroup$ Jan 8 '19 at 17:59

The simplest natural example is to use the van Kampen theorem (in Ronnie Brown's version) to work out the fundamental group(oid) of the circle. The details are in his book Topology and Groupoids (http://groupoids.org.uk/topgpds.html). That source also discusses colimits of groupoids in this context. If you look around on Ronnie's home pages (http://groupoids.org.uk) you can find discussion of many of the points that you raise.


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.