The Doeblin-Lenstra conjecture for a complex continued fraction algorithm (Q2407608)

The transformation of Tanaka \(T\) is determined as follows: \(T(0)=0\) and \[ T(z):=\frac{1}{z}-\left[\frac{1}{z}\right]_T, \] where \[ [z]_T:=\left\lfloor \lambda+\frac{1}{2}\right\rfloor \alpha +\left\lfloor \mu+\frac{1}{2} \right\rfloor \bar{\alpha}\quad \text{for} \quad z=\lambda \alpha+\mu \bar{\alpha} \quad \text{with} \quad \alpha:=1+i. \] Each \(z \in X:=\{z=\lambda\alpha+\mu\bar{\alpha} : -\tfrac{1}{2} \leq \alpha,\mu < \tfrac{1}{2} \}\) has a unique complex continued fraction expansion \[ z=a_0+\frac{1}{a_1}{+}\frac{1}{a_2}{+}\frac{1}{a_3+T^3 z}=[a_0,a_1,a_2,a_3+T^3 z] \qquad \text{with} \quad a_n:=\left\lfloor \frac{1}{T^n z} \right\rfloor_T \] as long as \(T^n (z) \neq 0\). Let be given \(z=[a_0,a_1,a_2,\dots]\) and define sequences of Gaussian integers \(p_n\) and \(q_n\) by \[ p_{-1}=\alpha,\; p_0=0 \quad \text{and} \quad p_{n+1}=a_{n+1} p_n+p_{n-1} \quad \text{for} \; n\geq 1, \] \[ q_{-1}=0,\; q_0=\alpha \quad \text{and} \quad q_{n+1}=a_{n+1} q_n+q_{n-1} \quad \text{for} \; n\geq 1, \] From the representation \[ z=\frac{p_n+p_{n-1}T^n z}{q_n+q_{n-1}T^n z} \] follows that the so-called convergents \(\frac{p_n}{q_n}\) converge to \(z\) as \(n \to \infty\). The approximation coefficients \[ \theta_n:=\theta_n(z):=|q_n|^2 \left| z-\frac{p_n}{q_n}\right| \] are examined in this paper. The following main result is proved in the article. Theorem. For \(x >0\) define \[ l(x):=\lim_{N \to \infty} \frac{1}{N} \sharp \{n \leq N: \theta_n(z)\leq x\}. \] Then, for almost all \(z\), the distribution function \(l(x)\) exist and is given by \[ l(x)=\lim_{N \to \infty} \frac{1}{N} \sum_{n \leq N} \chi_{A_x}(\mathbf{T}^n(z,0))= \] \[ =\mu(A_x)=\frac{1}{G} \iint_{A_x} \frac{d \lambda(u,v)}{|1+uv|^4}, \] where \(\mathbf{T}\) is defined in paper, \(\lambda\) is the Lebesgue measure, \(\chi_{A_x}\) is characteristic function of the set \[ A_x:= \left\{ (u,v)\in X \times Y: \left|\frac{1}{u}+v \right|\geq \frac{1}{x} \right\}, \] and the normalizing constant \[ G:=4\pi \sum_{n=0}^\infty \frac{(-1)^n}{(2n+1)^2} \] equals \(4\pi\) times the Catalan constant.