Help with a proof on asymptotic behaviour of the counting function of Markov Numbers (In Zagier's article)

36 Views Asked by At

In this article (a pdf is easily findable online) Zagier gives an approximation of the number of Markoff number below a certain bound. What I struggle with is to show that if \begin{equation} E_{\alpha,\beta}(x) = \frac{3x^2}{\pi^2\alpha\beta} + O\left( \frac{x}{\alpha} \right) + O\left( \frac{x}{\beta} \log \frac{x}{\beta} \right) \end{equation} And \begin{equation} E_{f(p),f(q)}(f(x)) \leq M_m(x) \leq E_{f(p)-\frac{c}{r^2},f(q)- \frac{c}{r^2} }(f(x)) \end{equation} Then \begin{equation} M_m(x) = \frac{3f(x)^2}{\pi^2(f(p)+O(1/r^2)) (f(q)+O(1/r^2))}+ O\left( \frac{f(x)}{f(p)} \right) + O\left( \frac{f(x)}{f(q)} \log \frac{f(x)}{f(q)} \right) \end{equation}

Where $m=(p,q,r)$ is a Markoff triple, $f(x)=arccosh(3x/2)$ and c is a constant such that \begin{equation} f(r)<f(p)+f(q)<f(r)+\frac{c}{r^2} \end{equation}

More details can be found in the article but I believe that this is enough. Thanks a lot if you take time to help me.