When the averages stop agreeing
An optical study of a model colloidal fluid found that the system behaved consistently at equilibrium but became uneven and time-dependent in a metastable state. In the equilibrium sample, averages taken over time matched averages taken across measured regions. In the metastable sample, they did not, and the particle dynamics changed during the longer time-average measurement. The change appeared before crystallization began.
The study set out to measure density-fluctuation probability distributions directly in a hard-sphere-like colloidal system. It used those distributions to assess ergodicity, the practical expectation that a time average and an average across an ensemble of measured states should agree, and to quantify how that relationship changed with metastability and waiting time.
A direct comparison
The particles had hard-sphere-like interactions. The reported freezing and melting volume fractions were 0.494 and 0.545. The main comparison used an equilibrium sample at volume fraction 0.460 and a metastable sample at 0.528. At least seven measurements were made at a single volume fraction, with additional comparisons across volume fraction and waiting time.
Multispeckle dynamic light scattering provided the spatial readout. It measured up to 6,241 individual intensity autocorrelation functions from separated sample volumes measuring 7 by 7 by 20 micrometres. The functions track how the scattered-light signal changes over a correlation time. The test was simple: for an ergodic sample, the normalized coherent intensity autocorrelation from time averaging should equal the corresponding ensemble-averaged function.
The fluid becomes patchy
At 0.460, the two averaging procedures agreed, which the authors describe as substantial confirmation of the ergodic hypothesis. The equilibrium frequency and relaxation-time distributions were narrow and symmetric. The relaxation-time distribution was well described by a Gaussian, and spatial fluctuations were minor. The measured regions therefore showed much the same relaxation pattern.
The picture changed at 0.528. Time and ensemble averages disagreed, and the longer time-average measurement showed particle dynamics evolving during the measurement. The distribution of intermediate-scattering-function frequencies was broad and asymmetric. The collective mean-squared displacement changed regime at approximately 0.2 seconds, showed power-law regimes below 0.03 seconds and above 1 second, and the dynamic four-point susceptibility, a measure of spatially varying relaxation, peaked at about 0.8 seconds. Maps showed solid-like regions encircled by fluid.
A signal before crystallization
Looking across volume fractions, the study found a broader departure from the equilibrium pattern. The first four central moments, which describe the shape and spread of a distribution, changed from a narrow, symmetric form in equilibrium to a broad, asymmetric and partly bimodal form out of equilibrium. The higher moments showed a jump at the freezing point of 0.494, and the proportion of non-Gaussian fluctuations increased with volume fraction. The study describes this shift as an ergodicity-breaking transition in the tested system.
Waiting time added another layer. During the induction phase before crystallization, the third and fourth moments rose sharply until crystallization onset at approximately 22 hours. That pattern indicates that the metastable fluid's fluctuation distribution was changing rather than remaining stationary during the waiting period. The study quantified crystallinity from a normalized fitted Bragg-peak area, allowing the fluctuation record to be compared with the onset of crystallization.
A result with a narrow scope
The authors interpret the results as time-dependent probability distributions, broken ergodicity and heterogeneous dynamics. The evidence comes from one hard-sphere-like colloidal model and the reported light-scattering setup, so it does not establish that the same pattern occurs in other physical systems. The observed associations also do not show that metastability or waiting time alone caused the changes. The report gives trends and fitted guide lines rather than formal uncertainty estimates, and it does not quantify the size of the time-versus-ensemble disagreement.
Paper data and sources
Original title: Ergodicity, ergodicity breaking, and temporal evolution of probability density in non-equilibrium: new perspectives by multispeckle DLS
Authors: Hans Joachim Schöpe
Journal/Repository: arXiv
Status: Preprint, not yet peer-reviewed
First online: 2026-09-28
DOI: Not available
Original paper · Full text