1673
Christiaan Huygens
Horologium Oscillatorium, Paris
Having built the first pendulum clock in 1656–57, Huygens spent this book proving what a
swinging rod really does. His finding is the seed of everything on this bench: a bob on a
circular arc is not isochronous — big swings take longer than small ones. He cured it
by squeezing the cord between brass cycloidal cheeks, so the bob traces a cycloid and the
period stops caring about amplitude. A pendulum is never a clock until somebody fixes its arc.
1788
Joseph-Louis Lagrange
Mécanique analytique, Paris
Two scalars replace the whole diagram: kinetic energy T and potential energy V, combined as
T − V and fed through the Euler–Lagrange equation. That is exactly the method used in the box
below — for n rods it yields n coupled equations whose left-hand side is a mass matrix in the
cosines of the angle differences. No forces drawn, no strings cut, no free-body sketches;
a piece of algebra, then a computer integrates it.
1890
Henri Poincaré
Les Méthodes nouvelles de la mécanique céleste, Gauthier-Villars, 3 parts, 1889–93
Working on the three-body problem for the King Oscar II prize, Poincaré proved there is no
general closed-form integral to be found and drew the first tangle of intersecting asymptotic
curves — homoclinic intersections, the fingerprint of chaos, sixty years before the word.
The double pendulum is the same disease in a workshop: two degrees of freedom, one energy
integral, and nothing else to hold the trajectory still.
A very small difference that escapes observation produces a considerable one that cannot be
missed. — Leçons sur la mécanique céleste, Stockholm lectures 1909–10 (our translation)
1892
Aleksandr Lyapunov
The General Problem of the Stability of Motion, Kharkiv; French translation, Annales de la Faculté des Sciences de Toulouse, 1907
Lyapunov gave us the honest way to measure this bench: not “does the twin drift”, but how fast
does it drift, as an exponent. Two neighbouring trajectories separate roughly like
eλt, and the chart above is a straight line for exactly as long as that law holds.
On this rig λ is of order one per second — hence one thousandth of a degree becomes a visible
disagreement in about a dozen swings.
1963
Edward N. Lorenz
“Deterministic Nonperiodic Flow”, Journal of the Atmospheric Sciences 20 (1963) 130–141
A rerun started from a hand-typed 0.506 instead of 0.506127 gave a different weather, and the
twelve-printed-digit question became a science. Lorenz’s twelve equations are the meteorological
cousin of our three: deterministic, dissipative, and doomed to lose the argument with the
long-term forecast. The SPLIT button on this page is that keystroke, made visible.
1975
Tien-Yien Li & James A. Yorke
“Period Three Implies Chaos”, American Mathematical Monthly 82 (1975) 985–992
The paper that let the field keep its name: if a one-dimensional map has a cycle of period
three, it has cycles of every period, and an aperiodic orbit too. Every pendulum on this bench
is, seen once per drive period, such a map — which is why “it will settle” is never a safe
prediction, only a hopeful one.
Die Gleichung — as assembled on this rig
Generalised coordinates θi measured from the downward vertical; μk
is the mass carried outward of joint k, inclusive of it. The mass matrix is built and solved fresh at
every RK4 stage — four evaluations per step, 600 steps per second of bench time.
loading…