The Recurrence Fermi Didn't Expect

A computer experiment that was supposed to be boring

In 1955, Enrico Fermi, John Pasta, Stanislaw Ulam and Mary Tsingou ran one of the first computer experiments in physics on the MANIAC I: a chain of masses connected by springs with a small nonlinear correction added to Hooke's law, set into motion in a single low-frequency mode. The expectation, based on statistical mechanics, was that the nonlinearity would slowly spread the energy evenly across all the chain's vibrational modes — a process called thermalisation, or equipartition. Instead, the energy sloshed between a handful of low modes and, after a long but finite time, very nearly returned to the original single-mode state. This near-recurrence, wholly unexpected, is now called the Fermi-Pasta-Ulam-Tsingou (FPUT) problem, and resolving why it happens took physicists over a decade.

The chain and its nonlinearity

The FPUT chain is N masses connected by nearest-neighbour springs, where the force between adjacent masses includes the usual linear term plus a small quadratic or cubic correction:

m · d²x_i/dt² = k(x_{i+1} − 2x_i + x_{i−1}) + α[(x_{i+1}−x_i)² − (x_i−x_{i−1})²]   (α-model, quadratic)
      or with a cubic term β[(x_{i+1}−x_i)³ − (x_i−x_{i−1})³]                              (β-model)

With α = β = 0 this is just a linear chain of coupled harmonic oscillators, whose normal modes never exchange energy at all — a linear chain started in one mode stays in that mode forever. It is only the nonlinear term that lets modes couple and exchange energy, which is precisely why the expectation of eventual full thermalisation seemed so natural, and why its failure to appear (at least on the observed timescale) was so surprising.

The soliton resolution

The explanation, found by Zabusky and Kruskal in 1965, was that in the continuum limit the FPUT chain's equations approximate the Korteweg-de Vries (KdV) equation, famous for supporting solitons: stable, localised wave pulses that pass through each other in collisions and retain their shape and speed afterward, a signature of a special, nearly integrable structure hiding underneath the apparent nonlinear chaos. Because the underlying dynamics are close to an integrable system (one with as many conserved quantities as degrees of freedom), energy trajectories in mode space stay confined near quasi-periodic orbits instead of spreading ergodically — which is exactly the near-recurrence Fermi's team observed rather than genuine thermalisation.

The Toda lattice: the exactly solvable cousin

Around the same period, Morikazu Toda found a specific nonlinear spring force, exponential rather than polynomial, for which the chain is exactly integrable — it has as many conserved quantities as particles, and its solutions can be written in closed form in terms of solitons that never decay. The Toda lattice sits as a landmark alongside FPUT: FPUT's polynomial chain is not exactly integrable, but for weak enough nonlinearity it behaves almost as if it were, over exponentially long times, which is the modern understanding of why the recurrence appears at all rather than immediate thermalisation.

Wave chaos versus true equipartition

Push the nonlinearity strength α or β high enough, or wait long enough at moderate strength, and the FPUT chain eventually does depart from its quasi-periodic recurrences and settles into genuine equipartition — a threshold behaviour tied to the onset of large-scale chaos in the mode-amplitude phase space, marked by the modes' effective Lyapunov exponents turning positive. This transition from ordered energy sloshing to true wave chaos, as a single parameter is turned up, is one of the cleanest laboratory demonstrations available of how deterministic nonlinear systems cross from near-integrable behaviour into full thermalisation.

Frequently asked questions

Try it live

Everything above runs in your browser — open Chaotic Waves and change the parameters while it is running. Nothing is installed, nothing is uploaded, the whole model lives in one tab.