I have been experimenting with recursive formulas of the form:
\begin{equation} \forall c \in \mathbb{C} , z_{n+1} = z_{n}^\alpha + c \tag{1} \end{equation}
as well as:
\begin{equation} \forall c \in \mathbb{C}, z_{n+1} = \overline{z_{n}}^\alpha + c \tag{2} \end{equation}
where $z_0 = c$, $\alpha \in \mathbb{Z}$ and $|\alpha| > 1$.
I made the following observations:
- In case $(1)$, the resulting structure has $\alpha-1$ symmetries when $\alpha \geq 2$ and $|\alpha|+1$ symmetries when $\alpha \leq -2$.
- In case $(2)$, the resulting structure has $\alpha+1$ symmetries when $\alpha \geq 2$ and $|\alpha|-1$ symmetries when $\alpha \leq -2$.
If you'd like to experiment with the software I used to gain more insights, I made it publicly available: https://github.com/AidanRocke/TensorFlow-Fractals
Here are a few visualisations of case $(1)$ for $\alpha \in {2,4,-2,-4}$:
So far I don't have an explanation for all four observations but I have a stability argument for case $(1)$ where $\alpha > 1$:
\begin{equation} z_{n+1} \sim z_0^{\alpha(n+1)} \end{equation}
which may be deduced by multiplying and dividing all terms in the series expansion by $z_0^{\alpha(n+1)}$. As a result, near the boundary of the circumscribing disk with radius $R$ the most stable(and therefore most distant) points are near the roots of unity:
\begin{equation} \mathcal{U}_n = \{e^{\frac{2ik\pi}{\alpha}}: k \in [0,\alpha-1]\} \end{equation}
I think this argument is sufficient but feel free to correct me if I'm wrong. As for the three other cases, I don't have a good explanation yet.



