Saramsha

No. 057  ·  12 November 2026

The Branch That Gets Ahead

Thousands of specks drift in at random from every direction and freeze the instant they touch something already frozen. There is no plan and no target shape — only one feedback loop: a branch that gets a little ahead of the others feels more of the incoming drift, so it grows faster, and gets further ahead still. Click anywhere on the dark field to strike there, and drag the slider to see the same rule draw a lightning bolt or a scorch mark in wood.

Click or tap the field to start a new discharge there.

About this piece

Georg Lichtenberg dusted a charged plate with resin powder

In 1777 the German physicist Georg Christoph Lichtenberg charged a plate with an electrophorus, dusted it with a mix of sulfur and lead powder, and watched the powder settle into branching, tree-like patterns radiating out from the point of discharge. He had found a way to make an electric charge's own shape visible, and the pattern is still named after him. The same shape turns up wherever a conductive path forces its way through something that mostly isn't: a lightning bolt threading the air between a cloud and the ground, metal deposited branch by branch in electroplating, and the scorched fractal patterns a few people burn into wood with a high-voltage transformer — a practice that has killed several of them, since the current that draws the pattern is also happy to pass through a person holding the wood.

What all of those share is a feedback loop, not a blueprint. Wherever the discharge has already pushed a little further, the electric field right at its tip is stronger than anywhere else nearby — field lines crowd together around a sharp point the same way they do around a lightning rod. A stronger field pulls in the next bit of charge fastest, so the tip that is already ahead gets there first, extends a little further, and becomes an even stronger tip next time. Growth concentrates on whoever is winning, which is exactly why the result is a sparse branching tree rather than a solid blob: the inside of a concave gap between two branches sees a weak field and is almost never where the next step lands.

The walkers running here

Solving the real electric field at every instant is possible but slow, so this page uses a trick discovered by Thomas Witten and Leonard Sander in 1981, called diffusion-limited aggregation: release a speck far away, let it wander in a random direction at every step, and freeze it the moment it bumps into the growing cluster. That sounds like nothing to do with electric fields, but it isn't a hack — the odds that a random walker from far away is the one that happens to touch a particular point on the cluster's boundary first obey the exact same equation (Laplace's equation) as the electric potential around a real charged conductor. Build enough of these chance arrivals and the random walkers' cluster and a true electrical discharge grow into the same shape, down to the branching statistics. Simulations consistently measure its fractal dimension at about 1.71 in two dimensions — rougher than a smooth curve (dimension 1) but nowhere near solid (dimension 2) — and this page measures that dimension live, by counting how many boxes of several different sizes are needed to cover whatever has grown so far.

The "how much the leader wins" slider changes one thing: instead of freezing the instant a single walker arrives, a site has to be touched several times before it joins the cluster. At the low end that barely matters, so every tip gets roughly its fair share of arrivals and the figure fills in bushy, like the fine coral branching of a real Lichtenberg burn. Push it up and a site that falls slightly behind almost never catches the arrivals it needs before a luckier neighbour pulls ahead of it — so the growth collapses onto one or two channels that keep winning, the way a real lightning strike usually shows one dominant stroke with only a few forks rather than a dense tree.

The field is a 240×240 lattice; a fixed pool of walkers is released on a shrinking-then-growing ring around the struck point and respawned whenever one wanders too far to matter, which is the standard way to keep diffusion-limited aggregation fast enough to run in a browser. The dimension reading uses six box sizes from 2 to 21 cells and a least-squares fit of their counts on a log-log scale, recomputed every half-second. ?warm= fast-forwards the growth so a screenshot shows an established figure rather than a single bright point.