Abstract
A statistical mechanics theory for a fluid stratified in density is presented. The predicted statistical equilibrium state is the most probable outcome of turbulent stirring. The slow temporal evolution of the vertical density profile is related to the presence of irreversible mixing, which alters the global distribution of density levels. We propose a model in which the vertical density profile evolves through a sequence of statistical equilibrium states. The theory is then tested with laboratory experiments in a two-layer stably stratified fluid forced from below by an oscillating grid. Quantitative measurements of density fluctuations across the interface are made by planar laser induced fluorescence. These fluctuations are splitted in a "wave" part and a "turbulent" part. The wave part of the density field is well described by a previous theory due to Phillips. We argue that statistical mechanics predictions apply for the turbulent part of the density field sufficiently close to the interface. However inside the mixed layer density fluctuations are instead controlled by a balance between eddy flux downward and dissipation by cascade to small scales. We report exponential tails for the density pdf in this region.
Introduction
In stratified fluids—such as the Earth's oceans where warm water sits atop cold, salty water—turbulence often creates "staircase" structures: regions of uniform density separated by very thin, sharp interfaces. Understanding what determines the shape and thickness of these interfaces is a classic problem in fluid dynamics. Traditionally, researchers have relied on "eddy diffusivity" models, which treat turbulence as an enhanced version of molecular diffusion. However, these empirical models often fail to explain the persistence and specific geometry of these boundaries. This paper provides a fresh perspective by applying the tools of statistical mechanics to show that sharp density interfaces are the "most probable" state resulting from a competition between turbulent mixing and buoyancy-driven sorting.
Theoretical Framework
The authors extend the Miller-Robert-Sommeria (MRS) theory, originally used to describe two-dimensional vortices, to the case of a density-stratified fluid. By treating the fluid as a system seeking to maximize its "mixing entropy" under the constraints of energy and density conservation, the theory predicts a specific equilibrium state.
Interestingly, for a system with two distinct density levels, the mathematical formulation mirrors Fermi-Dirac statistics used in quantum mechanics. The theory predicts that the mean vertical density profile should follow a hyperbolic tangent ($\tanh$) shape. This suggests the interface is a dynamic equilibrium: turbulence attempts to smear the density gradient (increasing entropy), while buoyancy pulls heavy fluid down and pushes light fluid up (minimizing potential energy).
Experimental Methodology
To validate the theory, the researchers conducted laboratory experiments in a tank filled with two layers (salty water on the bottom, fresh water/ethanol on top). Turbulence was generated from the bottom by an oscillating grid.
To measure density fluctuations with high precision, the team used Planar Laser Induced Fluorescence (PLIF). By matching the refractive index of both layers, they could shine a laser sheet through the tank without optical distortion. They analyzed data across seven distinct runs (EXP1 to EXP7), varying the density jump and the resulting Richardson numbers ($Ri$) from 1 to 150. A specialized "sorting algorithm" was developed to define the interface height even during wave breaking, allowing the researchers to separate large-scale "wave" motions from small-scale "turbulent" fluctuations.
Key Results and Insights
Shape and Thickness Scaling
The experimental density profiles collapsed onto a single $\tanh$ curve when rescaled by the predicted interface thickness, confirming the statistical mechanics prediction. A major finding was the scaling rule for interface thickness ($\Delta h$): it was found to be inversely proportional to the Richardson number ($Ri$), specifically following the relation $\Delta h \approx 3H/Ri_H$. This provides a simple, predictive rule for oceanographers and limnologists.
Wave Spectra and Mixing Efficiency
The frequency spectrum of the interface waves followed a $-3$ power law ($\omega^{-3}$), matching the 1977 theory by Phillips. This suggests the waves are limited by a "criticality" where they break if they become too steep. Furthermore, by assuming an approximate "equipartition" between kinetic energy and available potential energy, the authors derived a mixing efficiency of 0.25—a value frequently observed in oceanographic studies but rarely derived from first principles.
Non-Gaussian Fluctuations
While the interface itself follows statistical equilibrium, the "mixed layer" below it behaves differently. Near the interface, the density distribution is "double-peaked" (representing the two primary fluid densities). However, deep in the mixed layer, the density fluctuations show exponential tails. This indicates that rare, high-intensity "filaments" of light fluid are being pulled deep into the turbulent zone, governed by a balance between downward eddy flux and dissipation by cascade to small scales.
Conclusion
This research demonstrates that the tools of statistical mechanics can successfully predict the properties of density interfaces in turbulent fluids. While the core theory assumes density levels are conserved, the authors propose that real-world systems move through a "sequence of equilibria" as molecular diffusion eventually blurs the density levels. Future work may address higher Richardson numbers where the interface might become intermittent, or explore the effects of asymmetric forcing where both sides of the interface are turbulent.
Summary updated