← gallery

Char

A glowing edge eats its way down a sheet of paper.

Laying a fresh sheet…

Touch the paper to light a ring. Drag to draw a line.Enter lights the centre.
front speed
measuring
front width w
0.00 mm
since lit
0.0 s
saltpetre
0.85
0.25010³10⁻⁵Jyvaskyla
G₂(r) · mm² / mm

1x.

A sheet of copier paper, soaked in saltpetre and dried, burns without a flame. A glowing edge creeps across it at about half a millimetre a second. Behind the edge the paper is char. Ahead of it the paper is still white. The edge is never quite straight, and the longer it runs, the rougher it is meant to get.

Around the year 2000, physicists in Jyväskylä, Finland, burned about a thousand sheets like this in the dark and filmed only the glowing edge. Markko Myllys, Juha Maunuksela, Mikko Alava and their colleagues found that it roughens by the Kardar-Parisi-Zhang law. That is the rule for any edge that grows fastest where it already sticks out. Their 70 gram copier paper burned at 0.64 mm a second. Two points on its edge 10 mm apart were usually about 2 mm out of line.

I built the fire from equations, not from pictures of fire. They are Nikolas Provatas and his colleagues', from 1995. Heat spreads through the sheet and leaks away into the room. Wherever there is fuel and it is hot enough, it burns and makes more heat. The fuel is the sheet itself. I made it the way paper is made, out of fibres. Each is a millimetre long, and I laid them down at random until they weighed what the paper's fibres weigh. I set one number by hand: the fire's clock, so the edge moves at their 0.64 mm a second. Nothing else was tuned.

It burns too smoothly. On my sheet, two points 10 mm apart on the edge are only about 0.2 mm out of line, a tenth of theirs. The fibres are rough enough on their own. I think the heat, which runs a little ahead of the edge, evens them out. Myllys's own simulations matched the real roughness, but they took the strength of the shaking from the real burns. So I think something in a real sheet shakes the edge much harder than its fibres do, and I don't know what it is. Their edges also lurched now and then, a stretch of it suddenly racing ahead. Mine barely does. Their guesses were uneven saltpetre and the air moving past the paper. I have written to Mikko Alava to ask.

My first test burns were short, 28 mm, because the computer was busy that morning. The next day I burned four sheets as wide as theirs, one for 300 mm and three for 205. The edge does keep getting rougher as it goes, but slowly. Its width grows by the slower of the two laws, the one for an edge that only smooths itself, not the KPZ law their paper followed. And the roughness stops growing beyond about 20 mm. At the end, two points 10 mm apart are about 0.4 mm out of line, a sixth of theirs.

The part of the law that pushes an edge on where it already sticks out is there. I measured it by lighting sheets along a slope, and it is even stronger than in their paper. But the fire shakes its edge so little that the edge never tilts enough for that push to matter. So the long burns ask the same question as the short ones: what shakes a real edge so hard?

Some of it I chose. The sheet stands in a steel frame lined with needles, as theirs did, and I follow the edge down, as their cameras did. The room is dark, as theirs was, so the burning edge is almost the only light. I drew the char curling up, its cracks and the smoke. They are not computed. The smoke rises where the paper burns fastest. The glow is drawn too: each bit of the edge shines by how fast it burns there, so the light gathers where the fibres are thickest. On the first day I blurred that light into soft beads as even as a string of lights, and the blur is gone. The bright points are still all one size, one cell of the sheet each. Real embers along fibres would be short streaks of different lengths.

On my computer the fire keeps real time at phone size. On a large screen it ran at about three quarters of real time while the machine was busy. The panel shows how fast it runs on yours.

Touch the paper to light it there, or drag to draw a line of fire. Rings spread, meet and close. When the edge reaches the bottom of the sheet, it goes out against the frame. Fast asks for ten times real time, and gets what your computer can give. A new sheet can have less saltpetre. With much less, the edge slows and frays. With too little, it goes out. The small plot beside the numbers shows how rough the edge is at each distance, with the Jyväskylä burns drawn for comparison.