Chapter 01 · Self-organized criticality
The sandpile
Drop sand on a table one grain at a time. Most grains do nothing. Then one of them, no different from the others, moves half the pile.
In 1987 three physicists, Per Bak, Chao Tang and Kurt Wiesenfeld, wrote down a toy version of this and found something odd in it. Their pile did not need anyone to set it up carefully. Left alone, it drove itself into a state balanced on the edge of collapse and stayed there. They called this self-organized criticality, and for a decade afterwards people saw it everywhere.
The toy is simple enough to fit in a sentence, so let's build it.
The rule
Take a grid. Each square holds some grains of sand: zero, one, two or three. When a square gets a fourth grain, it is too tall and it topples. It gives one grain to each of its four neighbours and is left with nothing.
That is the whole rule. A toppling can make a neighbour reach four, which topples too, which can tip its own neighbours, and so on. That chain is an avalanche. Grains pushed off the edge of the grid fall off the table and are gone, which is what eventually stops every avalanche.
What the pile does on its own
At first nothing much happens. The table is nearly empty, a fourth grain almost never lands on a square that already has three, and the few avalanches are tiny. Watch the grains per square number climb.
It doesn't climb forever. It levels off at about 2.1 and stays there, wobbling. (On an infinite grid the exact value is 2.125.) The pile has found a balance: on average, the grains arriving from above match the grains falling off the edges. Nobody chose 2.1. You can't tune it, and you can't start the pile somewhere else and have it stay there. Pile on extra sand and the avalanches carry it off. Take sand away and the rain fills it back in.
Once the pile is in this state, look at the avalanches. Most are a single toppling. Some are a dozen. Every so often one tears across the whole table. If you press +20,000 grains a few times, the chart below starts to fill in.
No typical size
On these axes the points fall roughly along a straight line that slopes downward. That shape is a power law. In plain terms, it means that every time you make an avalanche ten times bigger, it becomes a fixed number of times rarer, and that holds whether you start at size 1 or size 1,000. There is no scale at which the pattern changes, until you hit the edge of the table.
This is stranger than it sounds. Most things we measure have a typical size. Adult heights cluster around an average, and a person three times the average height simply does not exist. The sandpile has no average avalanche in any useful sense. The tiny ones and the catastrophic ones come out of the same process, and you cannot tell them apart in advance by looking at the grain that started them.
The grain that triggers the big avalanche is not special. The pile is.
Physicists had seen power laws like this before, at critical points: water at the exact temperature and pressure where liquid and gas stop being different, or a magnet at the exact temperature where it loses its magnetism. There, fluctuations of every size appear at once. But you only get there by tuning a dial to precisely the right setting. Be a fraction of a degree off and the effect disappears.
The sandpile reaches its critical state without anyone touching a dial. Slow driving (one grain at a time) plus a way for sand to leave (the edges) is enough. That is the "self-organized" part, and it is why the idea spread so fast. If critical states could build themselves, maybe that explained why power laws keep showing up in nature, in earthquake sizes, forest fires and solar flares.
Order doesn't matter
The model has a hidden property that makes it unusually easy to reason about. In 1990 Deepak Dhar showed that the final state does not depend on the order in which squares topple. If two squares are both over the limit, you can topple either one first and you always end up in the same place. Dhar called the model abelian, a word mathematicians use for operations whose order doesn't matter.
That means we can do something dramatic. Instead of raining grains one at a time, put all of them on one square in the middle and let it collapse however it likes.
The result is not a smooth mound. It's an intricate, symmetric pattern of nested triangles and patches, and it looks more like a snowflake or a Persian rug than a heap of sand. Make the pile bigger and the same motifs return at larger scales. Nothing in the rule mentions triangles, so every bit of that geometry comes from the toppling itself.
Mathematicians are still working on why these patterns look the way they do. Some of the shapes were explained only in the last decade or so, when Lionel Levine, Wesley Pegden and Charles Smart found a surprising connection to the geometry of circles packed together. It is a lot of structure to get from "four grains and you fall over."
The honest caveats
Real sand turned out to be a disappointing sandpile. When experimenters poured actual sand, its grains had momentum. Once a slide started it tended to keep going, so piles alternated between building up and big collapses instead of producing avalanches of every size. The model ignores this because its grains have no momentum at all.
A better match came, oddly, from rice. In 1996 a group in Oslo slowly fed rice into a narrow pile between two glass plates and found scale-free avalanches of the kind the model predicts, but only with the most elongated grains. Rounder rice did not show them. The shape of the grains, which let them jam and catch on one another, turned out to matter.
The broader claims also outran the evidence for a while. Self-organized criticality was proposed as an explanation for earthquakes, extinctions, traffic, wars and stock markets. A straight line on a log–log plot is easy to see and hard to prove, and many claimed power laws later looked like other distributions once the data were examined carefully. The idea survives as a useful lens, not as a theory of everything.
What survives without caveats is the core lesson of the toy. A system made of identical parts, each obeying a dull local rule, can organise itself into a state where its own future is radically uncertain. Small causes have effects of every size. No grain is in charge of that, and no grain can see it.
Try this
Press Clear in Figure 1 and pour every grain into a single square by holding down on it. You'll grow a small version of Figure 3, with the same geometry. Then press Rain and watch the random sand slowly erase it.
Next: The flock · All chapters