← gallery

Drip

A drop holds the window upside down. Open the valve and watch it let go.

drag up to open · down to close · up/down arrows nudge · shift finer100 mm · f/16

what the picture can say about itself

flow
0.2609 cm³/s
last interval
waiting
bobs
waiting
clock
1×

full photograph

100600600
Tₙ₊₁ against Tₙ · ms

A tap that won't quite close lets water out one drop at a time. Open it a little and the drops come faster. At some settings they stop keeping time: long, short, long, short, or no pattern at all. Open it a little more and the steady beat comes back. Physicists have timed dripping taps since the 1980s, because a leaky tap is one of the simplest places to hear chaos.

I copied a real tap, as closely as I could. In 1999 Tomoo Katsuyama and his colleague Nagata ran distilled water through a glass tube 5 mm wide. They timed every drop with a laser beam 15 cm below it, to a tenth of a millisecond. As they opened the valve, steady dripping and irregular dripping took turns, again and again.

I made the water follow a model from the same year, by Fuchikami, Ishioka and Kiyono. It slices the hanging water into thin discs. Gravity pulls each disc down. Surface tension pulls along the real surface between them, and viscosity slows them. New water comes in at the top as fast as the valve allows. When the neck gets thin enough, the drop lets go. Nothing tells it when.

Before anything else I checked it against their numbers. Their first drop let go 0.21 seconds into their run, carrying 3.85 of their units of water and leaving 1.28 behind. Mine lets go a millisecond earlier, with 3.85, and leaves 1.27. At one flow they found the hanging drop bobbing seven times before each fall, once every 0.36 seconds. Mine does the same. At a faster flow they found five bobs every 0.28 seconds, and so do I.

Two things came out differently, and I have kept them. Between those two flows they found a steady six bobs every 0.32 seconds. Mine also bobs six times, but each fall comes somewhere between 0.32 and 0.34 seconds after the last, and it never quite settles.

The other difference is bigger. On the real tap, the steady stretches came back again and again as the valve opened, and each one's longest interval was 48.5 milliseconds shorter than the one before. In the model each new stretch is a drop that bobs once fewer, and its interval is about 39 milliseconds shorter, here and in the model's own paper. So the model and its authors agree with each other, and both fall short of the tap by about a fifth. I don't know why. One guess is that the water wetted the end of the glass out to its outer edge, 7 mm across, which makes bigger drops. I tried that, but the second, smaller drops grew big enough to count as drops of their own, and I couldn't read a steady beat out of it. Only someone who watched that tap could say.

Some of it I chose. The glass tube is theirs. The room, the window 1.5 m behind the tube and the lens are mine: a 100 mm macro at f/16, close enough that each drop is a lens too, holding the window upside down. Once a drop lets go, the model stops following it, as their paper does. From then on I draw it as Lord Rayleigh's wobbling drop, falling freely. After each drop the tip also sheds specks about a fifth of a millimetre across. The paper saw them too and could not tell whether they are real, so I leave them out. The tick you can turn on is my choice of sound, but it comes when the drop crosses the laser's height.

The model is slow to compute. On my computer, in Chrome, the middle of the valve runs at about nine tenths of real time. In some hours the same sums take four times longer, for a reason I haven't found, and it falls to a quarter. The panel shows how fast it runs on yours.

Drag up to open the valve, and down to close it. Slow shows the neck and the second, smaller bead, as a fast camera would. The square beside the numbers plots each interval against the next, as Katsuyama and Nagata did. A steady beat is one point, long and short make two, and chaos draws a curve. When no one touches it, the valve wanders by itself across the middle of its range.