Relations from quotient of free product

127 Views Asked by At

This question arose from an exercise which asks you show that the fiber coproduct exists in the category of groups. I was eventually able to (mostly) solve the problem by “gluing” the images of elements under the two homomorphisms(i.e letting $\phi(h)\psi^{-1}(h)=1$ where $\phi : H \rightarrow G$ and $\psi : H \rightarrow G’$ are homomorphisms.)

I looked further into this concept (amalgamation) on the Wikipedia page and it says that in order to obtain the relation $\phi(h)\psi^{-1}(h)=1$ on the free product $G * G’$ you must quotient out by the smallest normal subgroup (i.e. the intersection of all normal subgroups) containing the words $\phi(h)\psi^{-1}(h)$. So my questions are: why isn’t it sufficient to simply adjust the definition of the binary operation so that whenever we have a product ($a_1a’_1…a’_n\phi(h)$)($\psi(h)b_1…b_kb’_k$) it is equivalent to $a_1a’_1…a_nb_1…b_kb’_k$ and how do we know that the normal subgroup isn’t “too big”(that is, it establishes an unwanted relation)?

I am familiar with relations established by quotients of free groups, however I don’t quite see why this works.

1

There are 1 best solutions below

0
On BEST ANSWER

Perhaps it helps to look more closely at the meaning of "the normal subgroup generated by" some subset.

Let $\Gamma$ be a group and $S \subset \Gamma$ a subset. Let $N$ be the smallest normal subgroup of $\Gamma$ that contain $S$, meaning the intersection of all normal subgroups of $\Gamma$ that contain $S$ (one should check that this intersection is indeed a normal subgroup that contains $S$).

One can write out a formula for a subset of $\Gamma$ that is a generating set for $N$, namely: $$S^\Gamma = \{g^{-1} s g \mid s \in S, g \in \Gamma\} $$ In all likelihood this is going to be an infinite set.

Let's prove that $N = \langle S^\Gamma \rangle$. First, using that $S \subset N$ together with normality of $N$ it follows that $S^\Gamma \subset N$, and therefore $\langle S^\Gamma \rangle < N$. In the other direction, the subset $S^\Gamma$ is clearly invariant under conjugation by elements of $\Gamma$: for each element $g^{-1} s g \in S^\Gamma$ and each $g' \in \Gamma$ we have $$g'{}^{-1}(g^{-1}sg) g' = (gg')^{-1} s (gg') \in S^\Gamma $$ It follows that the subgroup $\langle S^\Gamma \rangle$ is also invariant under conjugation by elements of $\Gamma$, which is just a long-winded way of saying that $\langle S^\Gamma \rangle$ is normal in $\Gamma$. Since $N$ is the intersection of all normal subgroups of $\Gamma$ that contain $S$, it follows that $N < \langle S^\Gamma \rangle$.

With that settled, let $\Gamma = G * G'$, let $S = \{\phi(h) \psi^{-1}(h) \mid h \in H\}$, and let $N$ be the subgroup of $\Gamma$ normally generated by $S$. Let's convince ourselves that $N$ does not "establish an unwanted relation". We have proved that $N$ is generated (in the ordinary sense) by $S^\Gamma$. Therefore every element of $N$ can be written as a product of elements of the form $g^{-1} s g$ for some $s \in S$, $g \in G * G'$. Each such element, and any product of such elements, is clearly a "wanted" relation!