Suppose $a,b \in GL(n,\mathbb{C})$, and $\langle a,b\rangle$ is a free group of rank $2$.

Is there a way to choose a $c$ to guarantee that $\langle a,b,c\rangle$ is a free group of rank $3$?


Yes. Note first that it suffices to answer the question in $\textit{SL}(2,\mathbb{C})$. In particular, if $A,B\in\textit{GL}(2,\mathbb{C})$, let $\hat{A},\hat{B}$ be scalar multiples of $A,B$ that lie in $\textit{SL}(2,\mathbb{C})$, and suppose that there exists a $C\in\textit{SL}(2,\mathbb{C})$ so that $\langle\hat{A},\hat{B},C\rangle$ is a free group of rank $3$. Then for any word $w(x_1,x_2,x_3)$ in the free group $\langle x_1,x_2,x_3\rangle$, if $w(A,B,C) = I$, then $w(\hat{A},\hat{B},C)$ would have to be a scalar matrix, and hence would lie in the center of $\langle\hat{A},\hat{B},C\rangle$, which is impossible. We conclude that $w(A,B,C) \ne I$ for all $w$, and hence $\langle A,B,C\rangle$ is free as well.

So suppose $A,B\in\textit{SL}(2,\mathbb{C})$ and $\langle A,B\rangle$ is free of rank $2$. Consider the group $\textit{SL}(2,R)$, where $R$ is the ring $\mathbb{C}[t_1,t_2,t_3,t_4]/(t_1t_4-t_2t_3-1)$, and let $$ T \;=\; \begin{bmatrix}t_1 & t_2 \\ t_3 & t_4\end{bmatrix}. $$ Note that $T$ is invertible over $R$, with $$ T^{-1} \;=\; \begin{bmatrix}t_4 & -t_2 \\ -t_3 & t_1\end{bmatrix}, $$ and hence $T \in \textit{SL}(2,R)$. Then $\langle A,B,T\rangle$ must be free, since any element of $\langle A,B\rangle$ can be substituted for $T$, and the free group $\langle A,B\rangle$ has no non-trivial laws.

Now, if $w = w(x_1,x_2,x_3)$ is any non-trivial word in the free group $\langle x_1,x_2,x_3\rangle$, the equation $$ w(A,B,T) \;=\; I $$ is equivalent to a nontrivial system of polynomial equations in $t_1,t_2,t_3,t_4$, which define a proper subvariety $V_w$ of $\textit{SL}(2,\mathbb{C})$. But $\textit{SL}(2,\mathbb{C})$ cannot be expressed as a countable union of proper subvarieties, and hence there exists a matrix $C\in\textit{SL}(2,\mathbb{C})$ that does not lie in any $V_w$. Then $\langle A,B,C\rangle$ is free.

Practically speaking, the way to choose $C$ is to choose a generic matrix, i.e. a matrix whose entries do not satisfy any polynomial equations over the field generated by the entries of $A$ and $B$.

  • $\begingroup$ thanks for your answer. So i guess the same argument also works when you have $n$ free generators and you want to choose a $(n+1)$'th element such that they form a free group of rank $n+1$ $\endgroup$ – FunctionOfX Jun 11 '15 at 14:47

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.