# z→z²+cz⁵

mick asked a math.stackexchange.com question about the filled-in Julia sets of the polynomial:

fc(z)=z2+cz5f_c(z) = z^2 + c z^5

# 1 Critical Points

fc(z)=2z+5cz4=0f_c'(z) = 2 z + 5 c z^4 = 0

has 4 solutions

z0=0z0=25c3\begin{aligned} z_0 = 0 \\ z_0 = \sqrt[3]{-\frac{2}{5c}} \end{aligned}

which is where iteration should start from to determine the number of components that the filled-in Julia set has (1 if all critical points remain bounded, infinitely many if a critical point escapes to infinity).

z0=0z_0 = 0 implies zn=0z_n = 0 for all n0n \ge 0, which is not interesting.

# 2 Boundedness

Suppose |zn|<r|z_n| < r. If that implies |zn+1|<r|z_{n+1}| < r, then |zm|<r|z_m| < r for all n>mn > m (by induction).

We want |zn+1|=|zn2+czn5|<|zn|+|c||zn|5<r2+|c|r5<r|z_{n+1}| = |z_{n}^2 + c z_{n}^5| < |z_{n}| + |c| |z_n|^5 < r^2 + |c| r^5 < r whence |c|<1rr4|c| < \frac{1 - r}{r^4}

Suppose |z0|=25|c|3<r|z_0| = \sqrt[3]{\frac{2}{5|c|}} < r whence 25r3<|c|\frac{2}{5r^3} < |c|

Maximizing rr such that 25r3<|c|<1rr4\frac{2}{5r^3} < |c| < \frac{1 - r}{r^4} gives r<57,r < \frac{5}{7}, that is, if |zn|<57|z_{n}| < \frac{5}{7} for any nn then it’ll remain bounded.

Additionally, if |c|>25(57)3=686625=1.0976|c| > \frac{2}{5\left(\frac{5}{7}\right)^3} = \frac{686}{625} = 1.0976 all the critical points are certainly bounded, because |z0|<57|z_0| < \frac{5}{7}.

Note: this does not imply it converges to 00, but the technique of the next section could maybe be adapted. I think boundedness is enough but I’m not 100% sure.

# 3 Escape To Infinity

Assume |c|686625|c| \le \frac{686}{625}.

Suppose |zn|>(1+ϵ)R|z_n| > (1 + \epsilon) R with some ϵ>0\epsilon > 0. If that implies |zn+1|>(1+ϵ)2R|z_{n+1}| > (1 + \epsilon)^2 R, then |zn+m|>(1+ϵ)m+1R|z_{n + m}| > (1 + \epsilon)^{m+1} R \to \infty as mm \to \infty.

Concretely, let R=1|c|3R = \frac{1}{\sqrt[3]{|c|}} and ϵ=415\epsilon = \frac{4}{15}.

Writing |c|=1bk|c| = \frac{1}{b^k} for some kk and b>1b > 1, then R=bk/3R = b^{k/3}. Now

|zn+1|=|zn2+czn5|>||zn|2|c||zn|4|>|(1+ϵ)2R2|c|(1+ϵ)5R5|=(1+ϵ)2|b2k/3(1+ϵ)3b5k/3bk|=(1+ϵ)2((1+ϵ)31)b2k/3=(1+ϵ)2R×((1+ϵ)31)R\begin{aligned} |z_{n+1}| &= |z_n^2 + c z_n^5| \\ &> \left| |z_n|^2 - |c| |z_n|^4 \right| \\ &> \left| (1 + \epsilon)^2 R^2 - |c| (1 + \epsilon)^5 R^5 \right| \\ &= (1 + \epsilon)^2 \left| b^{2k/3} - (1+\epsilon)^3 \frac{b^{5k/3}}{b^k} \right| \\ &= (1 + \epsilon)^2 ((1 + \epsilon)^3 - 1) b^{2k/3} \\ &= (1 + \epsilon)^2 R \times ((1 + \epsilon)^3 - 1) R \end{aligned}

In the second part of the product, substitute ϵ=415\epsilon = \frac{4}{15}, and having assumed |c|686625|c| \le \frac{686}{625} we know R6256863R \ge \sqrt[3]{\frac{625}{686}}, that is,

|zn+1|>(1+ϵ)2R×348433756256863=(1+ϵ)2R×1.000744019>(1+ϵ)2R\begin{aligned} |z_{n+1}| &> (1 + \epsilon)^2 R \times \frac{3484}{3375} \sqrt[3]{\frac{625}{686}} \\ &= (1 + \epsilon)^2 R \times 1.000744019\ldots \\ &> (1 + \epsilon)^2 R \end{aligned}

Therefore, if |c|686625=1.0976|c| \le \frac{686}{625} = 1.0976 then |zn|>1915|c|3|z_n| > \frac{19}{15 \sqrt[3]{|c|}} implies that that the iteration escapes to infinity.

Moreover, if 0<|c|<0.010 < |c| < 0.01, then |z1|>1915|c|3|z_1| > \frac{19}{15 \sqrt[3]{|c|}}, that is this region of the cc plane always escapes to infinity (for z0z_0 a non-zero critical point). The bound 0.01 could be increased with more effort.

# 4 Special c

The right-most tip of the escaping region (see visualisation below) is pre-periodic, with positive real cc satisfying fc(fc(z0(c)))=fc(z0(c))f_c(f_c(z_0(c))) = f_c(z_0(c)) for the real critical point and fc(fc(fc(z0(c))))=fc(z0(c))f_c(f_c(f_c(z_0(c)))) = f_c(z_0(c)) for the complex critical points (which means all the critical points are bounded and the filled-in Julia set has a single component).

Using SageMath, the exact value can be found as the cube of a root of a quintic with algebraic coefficients:

f(c, z) = z^2 + c * z^5
for cp in solve(diff(f(c, z), z) == 0, z):
  try:
    g(c) = (f(c, f(c, z)) - f(c, z)).substitute(cp).expand()
    h(x) = (g(x^3) * x^7).canonicalize_radical()
    P = sum([ QQbar(H[0]) * polygen(QQbar)^H[1] for H in h(x).coefficients() ])
    c0 = QQbar.polynomial_root(P, CIF(RIF(0.6, 0.7), RIF(-0.1, 0.1)))^3
    z0 = QQbar(- (2 / (5 * c0))^(1/3))
    z1 = QQbar((- 2 / (5 * c0))^(1/3))
    print(all([f(c0, z0) == f(c0, f(c0, z0)), f(c0, z1) == f(c0, f(c0, f(c0, z1)))]))
    print(h(x))
    print(P)
    print(CIF(c0))
  except:
    pass

(download: z2cz5.sage) which outputs:

True
-3/1953125*5^(2/3)*(15625*5^(2/3)*2^(2/3)*x^5 - 18750*2^(1/3)*x^3 - 648*2^(1/3))
-0.3257301139913888?*x^5 + 0.1061001071608432?*x^3 + 0.003666819703478739?
0.2294487977025897?

that is, the special cc is c=0.229448797702589c = 0.229448797702589\ldots.

I used SageMath version 9.5, Release Date: 2022-01-30 as found in Debian Bookworm (current testing distribution) on 2023-03-09.

# 5 Visualisation

Source code for Fragmentarium (FragM fork version Fragmentarium 2.5.7.221026 Digilantism with include path Examples/Claude): z2cz5.frag

  • green: critical point proven to escape to infinity (center) or remain bounded (outside)
  • red: no decision reached after 1000 iterations
  • light: critical point escapes to infinity (filled-in Julia set has infinitely-many components)
  • dark: critical point remains bounded (filled-in Julia set has a single component)
  • blue coordinate grid at 1, 0.1, 0.01, … spacing

Wide view:

whole set

Zoomed view:

zoomed view

Close up view of special cc:

close up of special c

Animation of filled-in Julia set near special cc for real c±110c \pm \frac{1}{10}:

filled-in Julia sets near special c