Fractal

A fractal is a geometric structure that exhibits self-similarity across scales: any finite region of the structure, when magnified, resembles the whole. This is not merely visual resemblance. It is statistical self-similarity: the same statistical properties — dimension, roughness, correlation functions — persist across magnifications. Fractals are the spatial expression of scale invariance: where scale-invariant systems show power laws in time or probability, fractals show power laws in space.

The concept was formalized by Benoit Mandelbrot, who observed that natural objects — coastlines, clouds, mountain ranges, river networks — resist description by classical Euclidean geometry. A coastline does not have a well-defined length: the measured length increases as the measuring stick shrinks, following a power law with exponent related to the fractal dimension. This is not a measurement error. It is a genuine property of the object: the coastline is more complex at smaller scales, and that complexity never bottoms out at any finite resolution.

Fractals arise in systems where the same generative process operates at multiple scales. Diffusion-limited aggregation produces fractal clusters because particles stick at the growing tips, and the tips have the same geometry at all scales. Turbulent energy dissipation is spatially fractal because the cascade concentrates energy into ever-smaller regions with the same statistical geometry. Neural dendritic trees are fractal because growth rules that optimize coverage operate recursively at branchings of all sizes.

The connection to scale invariance is deep but often overstated. Not all scale-invariant systems are spatially fractal, and not all fractals are dynamically scale-invariant. A static coastline is fractal but not scale-invariant in its dynamics — it does not evolve self-similarly in time. A stock market is scale-invariant in its return distribution but not spatially fractal. The two concepts share a mathematical signature — the power law — but describe different domains of application. Conflating them is a category error that has plagued both physics and popular science.