Llamando a $P_k = \left(\begin{array}{c}a_k\\ b_k\end{array}\right)$
$$ V = \frac{1}{\sqrt 2}\left( \begin{array}{cc} 1 & i \\ -1 & i \\ \end{array} \right),\ \ \ \Lambda_k = \left( \begin{array}{cc} k^2-2i & 0 \\ 0 & k^2+2i \\ \end{array} \right) $$
$$ M_k = \bar V^{\dagger}\cdot\Lambda_k\cdot V $$
tenemos
$$ \left(V\cdot P_{n+1}\right) = M_k\cdot\left( V\cdot P_n\right) $$
o llamando a
$$ R_n = V\cdot P_n $$
$$ R_{n+1} = M_n\cdot R_n $$
y así
$$ R_n = \prod_{k=1}^n M_k\cdot R_0 $$
pero
$$ \prod_{k=1}^n M_k = \left(\prod_{k=1}^n \rho_k\right)\left(\begin{array}{cc}e^{i\sum_{k=1}^n \phi_k}& 0 \\ 0 & e^{-i\sum_{k=1}^n\phi_k} \end{array}\right) $$
entonces
$$ V\cdot \left(\begin{array}{c}a_k\\ b_k\end{array}\right) = \left(\prod_{k=1}^n \rho_k\right)\left(\begin{array}{cc}e^{i\sum_{k=1}^n \phi_k}& 0 \\ 0 & e^{-i\sum_{k=1}^n\phi_k} \end{array}\right)\cdot V\cdot \left(\begin{array}{c}1\\ 0\end{array}\right) $$
así que
$$ a_n= \frac{1}{2} e^{-i \Phi_n } \left(e^{2 i \Phi }+1\right) \left(\prod_{k=1}^n \rho_k\right) ,b_n= -\frac{1}{2} i e^{-i \Phi_n } \left(e^{2 i \Phi }-1\right) \left(\prod_{k=1}^n \rho_k\right) $$
y
$$ \frac{b_n}{a_n} = \tan{\Phi_n} $$
con
$$ \Phi_n = \sum_{k=1}^n\arctan\left(\frac{2}{k^2}\right) $$
de ahí
$$ \lim_{n\to\infty}\frac{b_n}{a_n} = -1 $$