Julia escape-potential landscape (3D)

Height is the square root of the escape potential. The filled Julia set is the flat floor; its boundary is the shoreline.

Switch to 2D

Presets

Playback

Constant c

Landscape

Camera

Layers

Quality

Model check

The science

What you see

The stand, posts and map board are Blender models. The landscape is the height field sqrt(G(z)) over a square window of the z plane: the filled Julia set is the flat dark floor (G = 0), its boundary the shoreline, and the land rises outward like ln|z| far away. Contour lines are curves of constant G. Click the map board (the Mandelbrot set) to choose c. The square root only exaggerates the thin filaments; the data are G itself.

What is modelled

For a complex constant c, iterate z → z² + c from every starting point z0 in the picture. If |z| never exceeds the escape radius the point belongs to the filled Julia set Kc (drawn dark); otherwise the number of steps before escaping gives the colour. The Julia set Jc is the boundary of Kc. The escape radius must be at least max(2, |c|); this page uses 256 so the smooth colouring is accurate.

Smooth colouring and the escape potential

Plain step counts give bands. The smooth value ν = n + 1 − log2(ln|zn| / ln R) removes them. The same data give the Green's function (escape potential) G(z) = lim ln|zn| / 2n, which is zero on Kc, positive outside and satisfies G(z²+c) = 2 G(z). The contour colouring mode draws curves of constant G; in the 3D twin G is the height of a landscape.

Connectedness and the Mandelbrot set

Fatou and Julia: Jc is connected exactly when the critical orbit of 0 stays bounded, i.e. when c lies in the Mandelbrot set M. For c outside M the Julia set is a Cantor dust of zero area. The map in the first graph is M; click it to choose c, and the page reports which case you are in.

Grid

The landscape is computed on the CPU in 64-bit arithmetic on a 176 x 176 grid (100 x 100 in Low quality), so thin filaments narrower than a grid cell are under-resolved; the GPU shader in the 2D twin gives the sharp picture.

Honest simplifications

  • A finite iteration cap treats very slowly escaping points as members, so the filled set and the Mandelbrot area are slightly overestimated; raising the cap sharpens thin filaments.
  • The Mandelbrot map is drawn at low resolution with 120 iterations.
  • Colours carry no physical meaning.

Numerical checks on this page

the functional equation G(z²+c) = 2G(z); the far-field value G(100) against ln 100; the z → −z symmetry on the actual grid; the filled fraction for c = 0 against the unit disc; and connected versus dust through critical-orbit membership and filled area.

Loading model...
Loading Blender model...
Slice through the landscape along the real axis
Area below each potential level
Critical orbit of 0
Made with 🟠 Blender · Three.js