009 / CHAOS
Where will the ball stop?
A steel ball on a string swings over three magnets and finally comes to rest above one of them. Which one depends only on where it starts. Nothing is random, yet it is almost impossible to predict. Try to guess, then look at the map of all outcomes.
Theory behind the experimentWhy a ball over three magnets is unpredictable, where the fractal map comes from, and what we checked ourselvesPlain explanation, equations and how the model works
In plain words
The string pulls the ball back to the centre, each magnet pulls it towards itself, and friction slowly drains the energy. The ball loops over the magnets, and when little energy is left one of them catches it. Nothing is random: from the same point the ball always reaches the same magnet.
But while it still has plenty of energy, the ball keeps arriving at forks where a tiny difference decides which way it turns. Every fork multiplies the gap between two nearby starts. After a few forks a millimetre turns into different magnets.
Colour every starting point by the magnet the ball ends up at and you get a map. Around the magnets there are large calm patches: from there the ball simply rolls to the nearest one. Further out the colours weave into stripes, and zooming into a boundary reveals ever thinner stripes.
The chaos is temporary: sooner or later friction stops every ball. This is called doubly transient chaos, and the magnetic pendulum is its textbook example.
Try it yourself
- Play “Guess the magnet”. Starts near a magnet are easy; try picking points in the colourful zone.
- In the laboratory turn on “Two balls 1 mm apart” and tap a colour boundary. The trails coincide at first, then split.
- Set the starting uncertainty to 1 mm and move the point around. In a patch near a magnet H is zero, in the colourful zone it is close to one.
- Shrink the uncertainty. In some places the outcome quickly becomes certain, in others all three colours survive even at hundredths of a millimetre.
- Lower the friction to 0.05: the colourful zone grows and starts to look like noise. Zoom in and ordered stripes appear inside it.
- Turn on the settle-time layer. The longest flights start right on the colour boundaries.
- Try two magnets, then four, or drag one of the three. The pattern rearranges, but the colourful fractal zone stays.
Physics and equations
The ball moves in the horizontal plane, the small-angle approximation. The equation of motion in dimensionless units:
The first term pulls the pendulum back to the centre, the second is viscous friction, and the sum is the attraction of three magnets at , a height below the bob. Friction steadily lowers the energy:
So every trajectory finally stops in one of the energy minima, above a magnet. The set of starts that lead to magnet is its basin of attraction. The boundary between basins is fractal: it holds trajectories that take a long time to choose a magnet, and near them small deviations grow exponentially.
Chen, Nishikawa and Motter showed that in such systems the boundary is fractal at all scales, but its dimension drops as the resolution increases. They called such sets slim fractals. We saw this ourselves: beyond about a hundredfold zoom, new stripes get thinner much faster than the scale grows.
The Wada property. It is sometimes said that a third colour always squeezes in between any two colours of such a map. We tested this numerically, following the idea of the basin-merging method by Daza and co-authors. In the outer colourful zone all three colours almost always meet nearby. On the boundary between the central patches, however, a noticeable part of the boundary is two-coloured at every resolution. So for the map as a whole the claim does not hold up.
How the model works
Parameters: , , magnets on a circle of radius 1, default friction as in the video. Fourth-order Runge–Kutta with step . The ball counts as caught when
If this has not happened within 130 time units, the outcome is marked undecided rather than given the colour of the nearest magnet.
Trajectories and clouds of starts run in your browser in double precision with the same algorithm as the Python version used for the video. The site tests compare outcomes and settle times for 455 starts: at three friction values, with two, four and five magnets, and with moved magnets.
The map at the video parameters was also computed in Python in double precision. Maps at other friction values and zoomed views are computed by your graphics card in single precision. At they agree with the exact calculation at about 99 % of points, with the differences concentrated at the finest boundaries.
The outcome uncertainty of a cloud of starts:
where is the share of balls that reached magnet . It measures our ignorance of the exact start, not randomness in the motion itself.
Simplifications: the magnetic force is that of a point charge, a real dipole magnet falls off faster; the rise of the ball on its string is ignored. The millimetre scale is nominal: the magnets are taken to be about 5 cm apart.
Sources and further reading
- A. E. Motter, M. Gruiz, G. Károlyi, T. Tél. Doubly transient chaos: generic form of chaos in autonomous dissipative systems. Phys. Rev. Lett. 111, 194101 (2013). doi ↗
- X. Chen, T. Nishikawa, A. E. Motter. Slim fractals: the geometry of doubly transient chaos. Phys. Rev. X 7, 021040 (2017). doi ↗
- A. Daza, A. Wagemakers, M. A. F. Sanjuán. Ascertaining when a basin is Wada: the merging method. Sci. Rep. 8, 9954 (2018). doi ↗