An underexpanded jet — gas escaping a nozzle at higher pressure than the surrounding atmosphere — creates a pattern of shock cells, expansion fans, and Mach disks. The velocity distribution inside this structure is not Maxwellian. Shock interactions create populations of molecules moving at distinctly different speeds, superimposed in the same spatial location.
The paper uses coherent Rayleigh-Brillouin scattering to resolve this structure in single 200-nanosecond laser shots. Each shot captures the velocity distribution at multiple spatial points simultaneously, mapping velocity, density, and local velocity gradients across the jet. The technique is optical and non-invasive — no probes disturbing the flow.
Note (2026-11-13; read against the paper's full text): the paper does not say that the averaged spectrum looks Maxwellian; it says the 20-shot averaged spectra give a poor representation of the jet dynamics, and that single-shot spectra deviate substantially from standard lineshapes, with non-Maxwellian distributions in many of them, particularly in regions near shock waves. I did not find the essay's bimodal distributions or asymmetric tails in the paper. The velocity gradients come from velocities measured at several separated locations within one laser pulse, not from a gradient at a point. The paper did not extract temperatures because the non-Maxwellian spectra make a translational temperature highly uncertain, and it reports good agreement overall between the averaged measurements and a simulation.
The critical observation: individual single-shot spectra frequently deviate from the averaged lineshapes. The averaged spectrum looks Maxwellian (or near-Maxwellian), but the instantaneous measurements reveal non-Maxwellian structure — bimodal velocity distributions, asymmetric tails, features that wash out in time-averaging. The deviations are not noise; they're the nonequilibrium physics that averaging destroys.
This matters for turbulence. Local velocity gradients extracted from single shots provide direct measurements of the quantities that turbulence models need but usually have to infer. The gradient at a point, at an instant, without spatial or temporal smoothing.
The averaged measurement tells you about the flow. The single-shot measurement tells you about the flow's fluctuations. The latter is the harder measurement and the more informative one — the fluctuations are where the physics lives that the mean field misses.