If you set out to measure the coastline of Britain with a ruler a kilometre long, you’d get one number. Switch to a metre ruler and you’d get a bigger number, because you’d trace tighter curves. A centimetre ruler would give an even bigger number, tracing around rocks. A millimetre ruler would account for pebbles. At every scale you’d find more detail, and the measured length would keep growing. This is not a limitation of your ruler. It’s a feature of coastlines — they don’t have a single well-defined length at all.
That observation, which the mathematician Benoît Mandelbrot made in a famous 1967 paper titled “How Long Is the Coast of Britain?”, is the starting point for fractal geometry. It opened a line of research showing that the smooth curves and shapes of classical geometry don’t actually describe most of the natural world. Nature is rough. And mathematics needed new tools to describe the roughness.
The idea in one picture: the Koch snowflake
The easiest fractal to meet is the Koch snowflake, invented in 1904 by the Swedish mathematician Helge von Koch. Start with an equilateral triangle. On each side, replace the middle third with a triangular bump. You now have a six-pointed star. Repeat: on each of the new sides, replace the middle third with a bump. Keep going.
After infinitely many iterations, you have a curve with strange properties.
- Infinite perimeter. Each iteration multiplies the total length by 4/3. So after steps the perimeter is times the original. As , the perimeter goes to infinity.
- Finite area. The area added at each step forms a geometric series that converges. The final snowflake fits comfortably inside a bounded region.
- No tangent lines. At every point on the curve, no matter how much you zoom in, the curve keeps wiggling. There is no smooth direction, so no derivative, so no tangent line.
A curve with infinite length enclosing finite area, continuous but nowhere differentiable, would have been called a pathological monster in the nineteenth century. And indeed, when Weierstrass constructed his nowhere-differentiable function in 1872, many mathematicians found it distasteful. Charles Hermite wrote to a colleague that he “turned away with fear and horror from this lamentable scourge of functions.”
A century later, Mandelbrot realised that the “lamentable scourge” is the rule, not the exception.
Fractional dimensions
What makes a fractal a fractal, formally, is a number called the Hausdorff dimension — a way of measuring dimensionality that doesn’t have to be an integer.
A line has dimension 1: if you scale it by a factor of 2, it covers twice as much length. A square has dimension 2: scale by 2 and it covers four times the area. A cube has dimension 3: scale by 2 and the volume goes up by eight. In general, a -dimensional object scaled by has its measure multiplied by .
Now look at the Koch snowflake. Each side is made of four copies of itself, each scaled by 1/3. For a -dimensional object, that would require , which gives
The Koch curve is somewhere between a line and a square. It has too much wiggle to be one-dimensional but not enough filling to be two-dimensional. The number 1.26 is its dimension — an honest measurement of how much the curve actually occupies.
This is what Mandelbrot called a fractal: a set whose Hausdorff dimension exceeds its topological dimension. The coastline of Britain has empirical fractal dimension around 1.25. The human circulatory system has dimension around 2.7. A cloud’s boundary has dimension around 1.35.
The Mandelbrot set
No discussion of fractals is complete without the most famous one. The Mandelbrot set is defined with a startlingly short rule: for each complex number , iterate
If the sequence stays bounded, is in the Mandelbrot set. If it escapes to infinity, is outside. Colouring points by how fast they escape produces the iconic image most people have seen on posters.
What makes the Mandelbrot set extraordinary is that this trivial rule — square and add — produces infinite structure. Zoom into any edge and you find spirals, islands, miniature copies of the whole set, and regions whose complexity doesn’t diminish with magnification. The boundary has Hausdorff dimension exactly 2 (shown by Mitsuhiro Shishikura in 1991) — as “large” as any planar region — despite being the boundary of a set with finite area.
The Mandelbrot set is also the most compressed demonstration of how simple rules produce unbounded complexity. There is no shorter description of that image than the one-line iteration. You cannot hand-draw it. You have to compute it.
Where fractals show up in nature
Once you know what to look for, fractal structures are everywhere.
Trees and lungs. A tree branches, each branch branches again, each sub-branch branches again. Your lungs do the same in reverse — the bronchial tree bifurcates into smaller and smaller passages until you reach alveoli. Both follow fractal scaling laws because both need to pack a large surface area (for sunlight or gas exchange) into a compact volume.
Mountains and rivers. Mountain outlines have roughly constant fractal dimension regardless of zoom level. Computer graphics exploits this to generate realistic landscapes — procedurally roughened terrain with the right dimension looks like a mountain range without anyone drawing one.
Blood vessels and neurons. The vascular system and neural dendrites both branch fractally to maximise reach per unit of biological material.
Financial markets. Mandelbrot himself spent decades arguing that price fluctuations in markets have fractal properties, with heavy tails and scale-invariant volatility. Modern quantitative finance has absorbed many of these ideas even when it doesn’t name them.
Antennas. Fractal antenna design uses self-similar shapes to receive a wide range of frequencies in a compact device. If you have a Wi-Fi-capable phone, there’s a good chance it contains a fractal antenna.
What fractals teach
There are two lessons worth taking from the fractal revolution.
The first is that smoothness is a modelling assumption, not a property of nature. Classical geometry — points, lines, circles, planes — was so successful for so long that we forgot it was a choice. The real world is often rough, and when it’s rough, smoothing it out loses information.
The second is that complexity and rule-simplicity can happily coexist. The Mandelbrot set is defined by a one-line rule and contains structure no human could invent. This is now one of the core intuitions of modern science: biology, cosmology, and artificial intelligence all rely on the idea that simple iterative processes can produce arbitrarily complex patterns.
A fractal is, at heart, a reminder that “looks complicated” and “is complicated to describe” are not the same thing. Sometimes the most intricate objects in nature have the shortest possible definitions.
Frequently asked
What is a fractal, precisely?
A fractal is a set whose Hausdorff dimension exceeds its topological dimension, and which exhibits self-similarity at different scales. Informally: a shape that stays rough no matter how much you zoom in.
Are fractals useful outside of pretty pictures?
Very much so. Fractal geometry models coastlines, cloud boundaries, blood vessels, stock-price volatility, and antenna design. Many real-world objects are better described as fractals than as smooth curves.