Definition. A group splits freely if
acts on a tree
without global fixed point and such that every edge stailizer is trivial. If
does not split freely, then
is called freely indecomposable.
Example. . Equivalently,
acts on
without global fixed points. So
splits freely.
If but
splits freely, then
for
.
Definition. The rank of is the minimal
such that
surjects
.
It is clear that .
Grushko’s Lemma. Suppose is surjective and
is minimal. If
, then
such that
for
.
Pf. Let be simplicial and let
be a graph of spaces with vertex spaces
and edge space a point. So
where
.
Let be a graph so that
and realize
as a simplicial map
. Let
. Because
is minimal,
is a forest, contained in
. The goal is to modify
by a homotopy to reduce the number of connected components of
.
Let be the component that contains
. Let
be some other component. Let
a path in
from
to
.
Look at . Because
is surjective, there exists
such that
. Therefore if
, then
is null-homotopic in
and
gives a path from
to
.
We can write as a concaternation as
such that for each
,
. By the Normal Form Theorem, there exists
such that
is null-homotopic in
.
We can now modify by a homotopy so that
. Therefore
and the number of components of
has gone down. By induction, we can choose
so that
is a tree. Now
factors through
. Then
and there is a unique vertex of
that maps to
. So every simple loop in
is either contained in
or
as required.
An immediate consequence is that .
Grushko’s Theorem. Let be finitely generated. Then
where each
is freely indecomposable and
is free. Furthermore, the integers
and
are unique and the
are unique up to conjugation and reordering.
Pf. Existence is an immediate corollary of the fact that rank is additive.
Suppose . Let
be the graph of groups. Let
be the Bass-Serre tree of
.
Consider the action of on
. Because
is freely indecomposable,
stabilize a vertex of
. Therefore
is conjugate into some
.
Now consider the action of on
.
is a graph of groups with underlying graph
, say, and
is a free factor in
. But there is a covering map
that induces a surjection
. Therefore,
. The other inequality can be obtained by switching
and
.

6 comments
Comments feed for this article
28 November 2011 at 9.27 am
Ronnie Brown
Are there any applications to geometric group theory of Higgins’ generalisation of Grushko’s theorem? This is proved in
Higgins, P.J._Notes on categories and groupoids_ , _Mathematical Studies_ , Volume 32. Van Nostrand Reinhold Co London (1971).; Reprints in _Theory and Applications of Categories_, No. 7 (2005) pp 1-195,
following his 1966 paper in J. Algebra. 4 (1966) 365–372.
12 December 2011 at 5.15 am
Henry Wilton
I don’t have a copy of Higgins’s book to hand. What’s the statement of the generalisation?
12 December 2011 at 5.53 am
ronniegpd
Here is Higgins’ Theorem.
Let
be groups with free decompositions
,
(
), and let
be a group morphism such that
for all
.
be a subgroup of
such that
. Then 
such that
for all
.
Let
has a decomposition
Higgins’ book is a free download.
See also
Braun, G. A proof of Higgins’s conjecture. Bull. Austral. Math. Soc 70~(2) (2004) 207–212.
13 December 2011 at 11.57 am
Henry Wilton
Thanks for posting this, Ronnie. I confess I don’t see why it is a generalisation of Grushko’s Theorem. That is, I don’t see how to deduce Grushko’s Theorem from Higgins’ Theorem.
13 December 2011 at 12.48 pm
Ronnie Brown
It seems best to refer to p. 124 of Higgins’ book, where Grushko’s theorem is given as a Corollary, with the proof on p. 125. Here is the link to the download.
http://www.tac.mta.ca/tac/reprints/articles/7/tr7abs.html
I think it is interesting that Philip was able by using groupoids to get this generalisation, which seems not to have been reached by the methods traditionally used in group theory to prove Grusko’s theorem.
Again, the statement of van Kampen’s theorem in this blog refers only to the version for the fundamental group, and not to the many base point version. I was kind of irritated in the 1960s that the standard version of this theorem could not even compute the fundamental group of the circle, a basic example, and saw this as an anomaly to be repaired somehow. This starting point eventually opened up for me large areas of research!
14 December 2011 at 5.02 am
Henry Wilton
Ah yes, that’s very nice! I hadn’t encountered that proof of Grushko’s Theorem before. As Grushko’s Theorem is very important in geometric group theory, I think you have answered your own question from the first comment.