The Gold Standard of Fractals | Julia Sets, Critical Orbits and the Parameter That Changes the Picture

Fix the quadratic mapping f_c(z)=z²+c. There are two inputs to keep distinct: the parameter c that selects the function, and the initial point z₀ whose orbit the function generates. Mixing them up changes the mathematical question even when the computer’s update line looks similar.

In the Mandelbrot set experiment, vary c across the parameter plane while fixing the starting orbit at z₀=0. Test whether this special orbit remains bounded. Each plotted position stands for a different function f_c. Thus the familiar cardioid-and-bulbs picture is a catalogue of parameters, not a picture of all bounded initial points for one function.

In a filled Julia set experiment, choose one c and hold it fixed. Vary z₀ over the dynamical plane and ask which starting points have bounded orbits under that single function. Its ordinary Julia set is the boundary of that filled set. These objects are related but not the same plotted data.

Every exercise in this article labels both c and z₀. A record that merely says ‘the point 1 escaped’ is ambiguous unless it specifies whether 1 was the parameter or the initial orbit point.

Contents · Next · Fractals root

For fixed c, define K_c as all starting points z₀ in the complex plane whose iterates under f_c stay bounded. Symbolically, K_c={z₀: the sequence z₀, f_c(z₀), f_c²(z₀), … remains bounded}. The boundary J_c=∂K_c is the Julia set in this quadratic setting.

Filled and boundary are not synonyms. In the simplest example, c=0, the filled set includes an entire disk while its Julia set contains only the circle around that disk. A picture that colours all bounded starting points black depicts the filled Julia set, not just the boundary, unless special rendering steps isolate the edge.

An orbit may be bounded yet fail to lie on the boundary: for c=0, the starting point z₀=0 is inside the filled disk. Conversely, a starting point outside the filled set can escape in only a few iterations. The boundary separates these behaviours but can be much more geometrically elaborate for other parameters.

Connectedness is a topological property: a set is connected when it cannot be partitioned into two separated, nonempty open parts in the relative topology. We will use exact examples and the known quadratic theorem rather than attempting to infer topology from a blurry pixel map.

Contents · Previous · Next · Fractals root

Set c=0. The iteration is zₙ₊₁=zₙ², and repeated substitution gives zₙ=z₀^(2ⁿ). Consequently |zₙ|=|z₀|^(2ⁿ). Every statement about boundedness now follows from whether the initial magnitude is below, on, or above one.

If |z₀|<1, powers drive the magnitude toward zero. If |z₀|=1, each magnitude remains one. If |z₀|>1, repeated squaring sends the magnitude without bound. Therefore K₀={z:|z|≤1}, the closed unit disk, and J₀={z:|z|=1}, its unit-circle boundary.

Take z₀=0.8 on the real axis. The next magnitudes are 0.64, 0.4096 and approximately 0.1678. Compare z₀=1.2: the sequence starts 1.44, 2.0736, 4.2998 and grows. Both calculations use c=0; only the initial point differs.

This example has no need for an uncertain escape-time picture. We derived the entire set analytically. It also proves that ‘c=0 belongs to Mandelbrot’ does not mean every z₀ is bounded when c=0.

Contents · Previous · Next · Fractals root

Now set c=−2 and consider f(z)=z²−2. For a real starting value x between −2 and 2, write x=2cosθ for a suitable real angle θ. Then x²−2=4cos²θ−2=2cos(2θ). After n steps, x_n=2cos(2ⁿθ), which remains between −2 and 2.

Any real starting x with |x|>2 escapes. For example x=3 maps to 7, then to 47 and onward. The complete complex-dynamics result is stronger: for c=−2, the filled Julia set is exactly the real interval K_−2=[−2,2], and its boundary is that same interval in the complex plane.

The boundary of a line interval regarded as a subset of the plane is indeed the entire interval, because it has empty two-dimensional interior. This can surprise learners who are picturing the two endpoints as the only boundary. They are the boundary of an interval considered within the real line, a different ambient-space question.

Do not generalise this exceptional case to all negative real c. Other parameters produce highly intricate boundaries. The purpose of comparing c=0 and c=−2 is to show how the same quadratic formula, with a new parameter, changes the geometry.

Contents · Previous · Next · Fractals root

The critical point of z²+c is z=0 because the complex derivative f′_c(z)=2z vanishes there. In the Mandelbrot definition we repeatedly iterate that critical point. For c=−2, the special sequence is 0→−2→2→2→…, clearly bounded. Hence c=−2 belongs to the Mandelbrot set.

Yet at the same fixed c=−2, the dynamical-plane point z₀=3 escapes immediately through 3→7→47→…. This is not a contradiction. The Mandelbrot question concerned one distinguished starting point, while the Julia question varies all possible starting points for that function.

Similarly c=0 lies in Mandelbrot because the critical orbit stays at zero, while z₀=2 escapes under f₀(z)=z². A common plotting mistake is to vary c and z₀ together as though the question were about z₀=c. That procedure defines a different experiment.

An accurate visualization legend therefore states ‘parameter c varied, z₀=0’ or ‘parameter c fixed, starting z₀ varied’. A label saying only ‘Mandelbrot/Julia fractal’ is too vague for a reproducible numerical explanation.

Contents · Previous · Next · Fractals root

Let c=−1. The critical orbit is 0→−1→0→−1→…, a bounded two-cycle. Therefore −1 is in the Mandelbrot set. The associated filled Julia set K_−1 is connected, and its boundary is the familiar connected basilica-type Julia set.

The displayed basilica shape cannot be proved merely by writing the two-cycle, but the theorem relating quadratic critical-orbit boundedness to connectedness gives the legitimate connection. Our direct two-cycle calculation establishes the premise: the critical orbit is bounded.

Take a different initial point, z₀=2, under the same map: 2→3→8→63→… escapes. Thus connected K_−1 still does not contain every starting point. Connectedness refers to the geometry of the bounded-start set; it does not imply universal boundedness under f_−1.

A student should separate three claims: the critical orbit is bounded; K_−1 is connected by the quadratic theorem; and a particular other z₀ may or may not belong to K_−1. They are connected reasoning steps, not interchangeable definitions.

Contents · Previous · Next · Fractals root

Choose c=1. Starting at the critical point zero gives 0→1→2→5→26→…, so strict radius-two escape occurs when the sequence reaches five. The parameter is outside the Mandelbrot set.

For quadratic polynomials z²+c, the connectedness criterion states that the filled Julia set is connected precisely when the critical orbit of zero is bounded. When that orbit escapes, the corresponding Julia set is totally disconnected in the usual quadratic setting. Thus c=1 yields a disconnected Cantor-like Julia set, not a connected unit-circle-type boundary.

The word ‘Cantor-like’ describes topology and the complex dynamics, not the exact middle-thirds Cantor construction from elementary set theory. Its dimension and detailed geometry depend on the chosen parameter; do not simply assign log2/log3 because the set is disconnected.

The computation of the critical orbit is short, while the theorem connecting it to the full Julia-set structure is deeper. A good article identifies where a calculation ends and where a mathematical theorem supplies the subsequent conclusion.

Contents · Previous · Next · Fractals root

For the family f_c(z)=z²+c, the celebrated result is: the filled Julia set K_c is connected if and only if the orbit of the critical point 0 is bounded. One can equivalently discuss connectedness of the Julia set in the quadratic setting under the standard theorem.

This is the bridge between the Mandelbrot parameter plane and the topology of Julia sets. The Mandelbrot set collects exactly those c for which the critical orbit remains bounded, so its parameter picture organises which fixed-parameter dynamical sets are connected.

The theorem is not a rule that every orbit of every point stays bounded whenever c belongs to M. Nor is it a claim that all connected Julia sets are smooth. Many connected Julia boundaries are geometrically complicated, and c=−2 gives the unexpected line-interval example.

A finite plot can suggest that a set is broken into islands, but isolated dark pixel clusters might be sampling artifacts. Connectedness is a claim about the exact infinite set; rigorous topology or certified methods are needed for proof beyond the trusted theorem and checkable premises.

Contents · Previous · Next · Fractals root

An escape-time renderer for a fixed-parameter Julia set uses z₀ as the point corresponding to a pixel. Set z=z₀, then repeat z←z²+c while checking a justified escape radius. For |c|≤2 a suitable radius of two works for many standard quadratic implementations; more generally a radius larger than max(2,|c|) supplies a safe escape threshold.

An observed orbit leaving the chosen valid escape bound certifies that its starting z₀ does not belong to K_c. A run that stays below the bound for N steps says only ‘not escaped by N’, unless a separate analytic interior criterion is available.

Give the two example programs different names: Mandelbrot(c) iterates z₀=0, while Julia(c,z₀) iterates a variable z₀. Reusing the same loop body is fine; silently changing which argument varies is not.

Keep the distinction between the filled Julia set and the Julia boundary in saved results. A dark pixel can indicate a candidate bounded interior point, not necessarily a point on the boundary. To draw J_c, a renderer needs a boundary-focused method or sufficiently careful edge extraction, not merely a change of title.

Contents · Previous · Next · Fractals root

An image is sampled on a finite coordinate grid, and each point is assigned a result after a finite iteration budget. It therefore cannot generally establish exact membership or connectedness merely through a palette. Near the Julia boundary, escape can be delayed and numerical precision can affect when the program detects it.

At a large zoom depth, the pixel centres may become closer than the resolution supported by a selected floating-point representation. Then two nominally different starting points can round to the same value, giving apparent repeated details or blocky structures unrelated to the exact set.

Antialiasing can make an image visually smoother by averaging multiple samples, but a smooth edge is not proof that the topological boundary is smooth. Likewise, dark areas may indicate analytic interiors or simply points not observed to escape.

For an educational computation, record the fixed parameter, spatial window, pixel resolution, escape threshold, numerical precision and maximum iteration count. If the goal is a theorem rather than an illustration, rely on the exact critical-orbit reasoning and the connectedness criterion rather than pixel appearance.

Contents · Previous · Next · Fractals root

Exercise A: With c=0 and z₀=0.5, the orbit tends to zero because |z₀|<1. With c=0 and z₀=2, the orbit diverges. Both outcomes occur under the same function and show why K₀ is the disk rather than the whole plane.

Exercise B: With c=−2 and z₀=1, the orbit is 1→−1→−1→…, so this starting point belongs to K_−2. With c=−2 and z₀=3, the orbit escapes. Check why the interval [−2,2] is not merely a drawing convenience.

Exercise C: For c=−1, show the critical orbit cycles between zero and −1 and use the theorem to conclude K_−1 is connected. For c=1, show the critical orbit escapes and conclude the corresponding Julia set is disconnected.

Exercise D: A screenshot labelled ‘Julia set for c=1’ is generated using z₀=0 at every pixel while varying c across pixels. What does it actually show? It approximates the Mandelbrot parameter experiment, not a Julia set for fixed c=1.

Contents · Previous · Next · Fractals root

First write f_c(z)=z²+c. Decide whether you are studying parameters or starting points. For Mandelbrot, vary c and fix z₀=0; for a filled Julia set, fix c and vary z₀. Make both settings visible in the article title, code and figure legend.

Second, use exact examples as regression tests. For c=0, K is the unit disk; for c=−2, K is the real interval [−2,2]; for c=−1, the critical orbit is a bounded two-cycle; for c=1, the critical orbit escapes. These cases test both orbit arithmetic and the meaning of the rendered object.

Third, distinguish computational evidence from mathematical proof. Detecting escape is decisive once the threshold is justified, while finite non-escape needs a cautious status. A connectedness conclusion relies on the quadratic critical-orbit theorem, not on how many pixels look joined.

Continue with The Gold Standard of Fractals, Mandelbrot escape radius and finite iterations, and Topology: connected versus path-connected to see how dynamical evidence and topological properties differ.

Contents · Previous · Fractals root