№ 116 · physics

The three-body problem: why three is harder than two

Two bodies pulling on each other by gravity is a solved problem. They trace ellipses, and you can write down where they will be a thousand years from now. Add a third body and that formula is gone. Not because nobody has been clever enough yet: the rules that made two bodies easy do not carry over to three.

Where it came from

Newton solved the two-body problem and used it for planets and comets. Then he turned to the Moon, and the Moon would not behave. It is pulled by the Earth and by the Sun, and the Sun's pull turns the Moon's ellipse around: the point where the Moon comes closest to the Earth swings round by about 40 degrees a year. This was not idle curiosity. Sailors found their longitude at sea with tables of where the Moon would be, so the tables had to be right. The Moon was the first three-body problem anyone needed solved.

Why two is easy

With two bodies, the point between them that balances their masses (the centre of mass) just drifts at a steady speed. Ride along with it, and what is left is one body pulled toward one fixed point. Several things then never change: the energy; the angular momentum, which keeps the motion in a flat plane; and a third, less famous quantity that fixes which way the ellipse's long axis points. Lock that many things in place and a bound orbit has no room to wander. It must close on itself and repeat, forever.

Why three is not

With three bodies, each one pulls on the other two, and every pull changes as all three move. The same bookkeeping still holds: total momentum, angular momentum and energy are conserved. But three bodies moving in space have nine directions to move in, and those few locks are not enough to pin the motion down. In 1887 Heinrich Bruns proved that no further conserved quantity of a simple algebraic kind exists. Henri Poincaré then showed the same for a simplified version with one tiny body, and went further.

Poincaré's 1890 memoir on the problem won King Oscar II's prize. Along the way he had at first assumed something about how paths near an unstable orbit behave, and used it to argue that the solar system is stable. With help from Lars Phragmén he found that the assumption was false. The paths cross and fold over each other in a tangle of endless detail. This was the discovery of what we now call chaos: a rule with no randomness in it that still turns a tiny change in the start into a large change later.

"Unsolvable" needs one footnote. In 1912 Karl Sundman found a series that does converge to the general solution, as long as all three bodies never collide at once. It converges so slowly that it has not been of much practical use. So there is no usable formula. There is a recipe that works in principle and has been of little use in practice.

Interactive Pick a start, set how hard to nudge the copy, and press run. Filled dots are the original; rings are the nudged copy.

Three point masses pulling on each other by Newton's law of gravity (G = 1), stepped forward with a fourth-order Runge–Kutta integrator whose step shrinks when two bodies come close. The copy starts identical except that the first body's speed in the x direction is raised by the nudge. Figure-eight start from Šuvakov (2013), Table I, row M8. The equal-mass triangle fails Routh's stability test, so it breaks up even with no nudge: rounding error alone is enough. Our drawing, not a figure from any paper.

What does exist

Special arrangements can be solved exactly. In 1767 Leonhard Euler found motions with all three bodies in a straight line. In 1772 Joseph-Louis Lagrange found motions with the three at the corners of an equilateral triangle. In both, each body traces a rescaled copy of a two-body orbit. Whether Lagrange's triangle survives a nudge depends on the masses. Edward Routh's 1877 test says it is stable only when the square of the total mass is at least 27 times the sum of the products of the masses taken in pairs. That holds when one mass dominates. The Sun and Jupiter pass it, and the Trojan asteroids sit at the triangle points of that pair. Three equal masses fail it badly: 9 against 81. In the figure above, the triangle falls apart.

In 1993 Chris Moore found something new by computer: three equal masses chasing each other around a single figure eight. Alain Chenciner and Richard Montgomery later proved that it really exists. It is stable, so a nudged copy keeps close company with the original. Searches like this are still finding orbits. One 2013 search reported 13 new distinct ones, and a follow-up found eleven more near the figure eight.

Most starts look nothing like these. In one famous numerical experiment, masses of 3, 4 and 5 are placed at the corners of a triangle with sides 3, 4 and 5. They do not settle into any pattern: two of them pair off and the third is thrown out at high speed.

In one breath

Two bodies keep enough quantities fixed that a bound orbit must close. Three do not, so in general there is no formula, only computation. Computation works: it is how spacecraft are sent to the Moon and Mars. But where the motion is chaotic, a tiny error in the start grows until the prediction is worthless, and only a more precise start buys you more time.

Where this comes from

  1. An introduction to the classical three-body problem: From periodic solutions to instabilities and chaos linked only, not reproduced
    Govind S. Krishnaswami and Himalaya Senapati · arXiv:1901.07289 · 2019
    arxiv.org/abs/1901.07289
  2. The three-body problem linked only, not reproduced
    Z. E. Musielak and B. Quarles · arXiv:1508.02312 · 2015
    arxiv.org/abs/1508.02312
  3. Three Classes of Newtonian Three-Body Planar Periodic Orbits linked only, not reproduced
    Milovan Šuvakov and V. Dmitrašinović · arXiv:1303.0181 · 2013
    arxiv.org/abs/1303.0181
  4. Numerical Search for Periodic Solutions in the Vicinity of the Figure-Eight Orbit: Slaloming around Singularities on the Shape Sphere linked only, not reproduced
    Milovan Šuvakov · arXiv:1312.7002 · 2013
    arxiv.org/abs/1312.7002