I found the following result on some lecture notes, but there was no proof... I am not even sure whether it is correct or not... I tried to prove it, but I'm having some troubles:
Let $f$ be a hyperbolic transformation with fixed points $x$ and $y$, and let $g$ be a Moebius transformation mapping $x$ to $y$. Then the following hold:
1) $fgf^{-1}g^{-1}$ is hyperbolic
2) $fgfg^{-1}$ is parabolic
Here is what I tried:
Firstly, I assumed $x=0$ and $y=\infty$ (indeed, the thesis is invantiant by conjugation, since the trace is). Hence, I wrote $f$ and $g$ as $f(z)=\lambda z$ and $g(z)=\frac{\gamma}{z}+\delta$. Now, an explicit computation gives me:
1) $(fgf^{-1}g^{-1})(z)=...=\lambda^2 z+\lambda\delta(1-\lambda)$
Now I should compute the trace to show that $fgf^{-1}g^{-1}$ is hyperbolic when $f$ is. Is that correct? And how can I do this computation?
2) $(fgfg^{-1})(z)=...=z+\delta(\lambda-1)$
Now, if $\delta(\lambda-1)\neq 0$, clearly the transformation is parabolic, with only fixed point $\infty$. We need to check that $\delta(\lambda-1)= 0$ cannot hold.
If $\gamma=1$ we have $f(z)=z$, which contradicts the hypothesis that f is hyperbolic.
If $\delta=1$, we have $g(z)=\frac{\gamma}{z}$, which is a mobius transformation sending $0$ to $\infty$, and I cannot find an absurd in this. It seems to me that $fgfg^{-1}$ can also be the identity (hence, not parabolic). Does it make sense? Any help?