Anisotropy of emergent large-scale dynamics in forced stratified shear flows

Through three-dimensional direct numerical simulations, this study demonstrates that forced, stably stratified shear flows evolve into a statistically steady state where the shear layer converges to a finite depth and self-organizes into large-scale anisotropic structures, ultimately tuning the gradient Richardson number to a value below 0.2.

The Gist

The Big Picture: Stirring a Cup of Coffee in a Giant Ocean

Imagine you have a cup of coffee with a layer of cream on top. If you stir it, the cream and coffee mix. But in the real world, like in the ocean or the atmosphere, fluids are often "stratified." This means they are layered by density (like heavy cold water at the bottom and light warm water on top).

Usually, if you stir a layered fluid, the layers mix for a while, and then the turbulence dies out because the layers try to settle back down. But this paper asks a different question: What happens if you keep stirring forever?

The researchers simulated a scenario where a "force" (like wind or tides) constantly tries to keep the fluid moving and prevents it from settling. They wanted to see what kind of patterns emerge when you keep this mixing going for a very long time, and how big the "stirring spoon" needs to be to see the whole picture.

The Experiment: A Digital Ocean Tank

The scientists built a giant, invisible digital tank using a supercomputer.

• The Setup: They created a flow where fast-moving fluid slides over slow-moving fluid (like wind over water). This naturally creates instability, causing the layers to roll up into giant waves (called Kelvin-Helmholtz billows, which look like ocean waves breaking).
• The Twist: In a normal experiment, these waves would eventually crash, mix, and stop. In this study, they used a "magic reset button" (mathematical forcing) that constantly nudged the fluid back to its original shape. This kept the turbulence alive forever, allowing them to study a "steady state" of chaos.
• The Variable: They ran this simulation in tanks of different sizes. They started with a small tank and kept making it wider and wider to see if the size of the tank changed the results.

Key Findings: The "Goldilocks" Zone and the Giant Imprint

Here are the three main discoveries, explained simply:

1. The Mixing Zone Has a "Sweet Spot" Size

When they started stirring, the layers of fluid got thicker and thicker as they mixed. You might think, "If I keep stirring, the layers will just get infinitely thick."

• The Discovery: No! The layers stopped growing once they reached a specific size (about 16 times the original thickness).
• The Analogy: Imagine a child trying to build a sandcastle. No matter how much sand they pile on, the castle stops growing at a certain height because the sand keeps sliding off. The fluid "tuned" itself. It found a balance where the force trying to mix it was exactly equal to the force of the layers trying to stay separate.
• The Catch: To see this "sweet spot," the digital tank had to be huge. If the tank was too small, the mixing zone kept growing because it didn't have enough room to settle. The tank needed to be at least 96 times wider than the original layer to get the right answer.

2. The Flow is "Anisotropic" (It Has a Preferred Direction)

This is a fancy word meaning "not the same in all directions."

• The Discovery: The turbulence didn't form random blobs. It formed giant, organized structures.
  • Up and Down: The mixing happened in a layer about 16 units thick.
  • Side to Side (Spanwise): The patterns stretched about 50 units wide.
  • Front to Back (Streamwise): The patterns stretched a massive 115 units long!
• The Analogy: Think of a river. The water might be deep (vertical), but the ripples and currents often stretch out for miles along the river (horizontal). The fluid here created "super-highways" of turbulence that were much longer than they were wide or deep.
• Why it matters: If you only look at a small patch of the river, you might think the water is just churning randomly. But if you zoom out, you see a massive, organized flow pattern. The researchers found that you need a very wide view (a huge computer domain) to see these giant "highways."

3. The "Ghost" of the First Instability

Why did the fluid organize itself into these specific giant shapes?

• The Discovery: The researchers found that the giant patterns (about 115 units long) were almost exactly the same size as the first wave that formed when the fluid started moving.
• The Analogy: Imagine you drop a pebble in a pond. It makes a small ripple. If you keep the water moving, that initial ripple leaves a "fingerprint" or an "imprint" on the water. Even though the water is now a chaotic mess of turbulence, the giant structures that form later are actually just a "ghost" of that very first ripple, stretched out and amplified.
• The Insight: Even in a highly turbulent, chaotic flow, the system remembers the very first instability that started it all.

Why Should We Care?

This isn't just about math; it's about understanding our planet.

• Climate and Weather: The ocean and atmosphere are full of these stratified shear flows. They control how heat, carbon, and pollution move around the globe.
• The Problem: If scientists use computer models that are too small (like looking at the river through a straw), they miss these giant, organized structures. They might calculate the mixing wrong, leading to bad predictions about climate change or weather patterns.
• The Lesson: To get the right answer, we need to simulate these flows in "extra-large" domains. We need to zoom out far enough to see the giant patterns, not just the local chaos.

Summary in One Sentence

The researchers found that when you constantly stir a layered fluid, it doesn't just mix randomly; it organizes itself into a specific, stable thickness and forms giant, elongated patterns that are "imprinted" by the very first wave of instability, but you need a massive view to see the whole picture.

Technical Summary

1. Problem Statement

Stably stratified shear flows are ubiquitous in geophysical and environmental contexts (e.g., oceanic currents, atmospheric layers). While the initial transition to turbulence via primary instabilities (like Kelvin-Helmholtz Instability, KHI) is well-studied, the nature of the statistically steady state in continuously forced flows remains an open question.

Previous studies often relied on initial value problems where turbulence decays, or used computational domains that were too small to capture large-scale emergent structures. The central questions addressed in this work are:

• Does a continuously forced stratified shear flow converge to a statistically steady state with a finite, characteristic shear layer depth independent of domain size?
• What are the unconstrained, emergent length scales of such flows?
• How does the horizontal extent of the computational domain affect the statistics of mixing, dissipation, and large-scale self-organization?

2. Methodology

The authors employed Direct Numerical Simulations (DNS) using the GPU-accelerated spectral element solver NekRS.

• Governing Equations: The incompressible Oberbeck-Boussinesq approximation with a linear equation of state.
• Forcing Mechanism: Unlike decaying turbulence studies, this setup uses a continuous volumetric forcing ($f_u, f_b$) that relaxes the instantaneous velocity and buoyancy profiles back toward their initial base states ($u_{x,0} = \tanh(z)$ and $b_0 = \tanh(R_0 z)$) over a response time $t_r$. This mimics geophysical forcing (e.g., wind, tides) and sustains turbulence indefinitely.
• Parameters:
  • Prandtl number ($Pr$) = 1.
  • Initial bulk Reynolds number ($Re_0$) = 50.
  • Initial bulk Richardson number ($Ri_0$) = 1/80 (0.0125).
  • Response time ($t_r$) = 100.
  • These parameters ensure the flow is initially unstable to primary KHI but "weakly stratified."
• Domain Variation: The study systematically varied the horizontal domain extent ($L_h = L_x = L_y$) from 16 to 512 (in units of initial shear layer half-depth $d_0$). Additional non-square domains ($2048 \times 512$ and $512 \times 2048$) were used to isolate streamwise and spanwise effects.
• Resolution: Spectral element methods with polynomial order $N=6$ and non-uniform vertical element distribution to resolve the Kolmogorov scale efficiently.

3. Key Contributions

1. Convergence of Shear Layer Depth: The study demonstrates that despite continuous forcing and mixing, the turbulent shear layer depth converges to a finite, characteristic value ($d \approx 8$, total depth $\Lambda_z \approx 16$) once the domain is sufficiently large ($L_h \gtrsim 96$).
2. Discovery of Strong Anisotropy: The emergent large-scale dynamics exhibit extreme anisotropy. The characteristic length scales follow a hierarchy: $\Lambda_z (\approx 16) < \Lambda_y (\approx 50) < \Lambda_x (\approx 115)$.
3. Domain Size Sensitivity: The paper establishes that standard computational domains (often limited to a few wavelengths of the primary instability) are insufficient for capturing converged statistics of mixing and dissipation. Convergent statistics require domains roughly 100 times larger than the initial shear layer depth.
4. Dynamical "Tuning" Hypothesis: The authors propose that forced stratified shear flows self-organize ("tune") to a quasi-equilibrium state where the gradient Richardson number ($Ri_g$) stabilizes around 0.2 (specifically $\lesssim 0.25$), balancing the forcing against turbulent mixing.

4. Key Results

A. Vertical Convergence and Mixing

• Shear Layer Deepening: The shear layer and buoyancy interface deepen significantly from their initial values ($d_0=1$) to a steady state of $d \approx 8$. This deepening increases the effective Reynolds number to $Re \approx 400$ and the bulk Richardson number to $Ri \approx 0.1$.
• Critical Domain Size: Vertical profiles of buoyancy, shear, and mixing statistics (flux, dissipation) only converge when $L_h \gtrsim 96$. Smaller domains artificially constrain the vertical stirring and mixing.
• Flux Coefficient: The bulk mixing coefficient ($\bar{\Gamma}_\chi$) within the mixing zone is approximately 0.13, slightly lower than the canonical Osborn value of 0.2. The authors attribute this to the specific nature of the forcing, which perturbs the energy balance between kinetic and potential energy reservoirs.

B. Emergent Large-Scale Structures

• Streamwise Scale ($\Lambda_x$): Spectral analysis reveals a pronounced peak in the streamwise energy spectrum at a wavelength $\Lambda_x \approx 115$. This scale is remarkably close to the theoretical wavelength of the most unstable KHI mode calculated based on the emergent (deepened) shear layer profile ($\lambda_{KHI} \approx 15 \times d \approx 120$).
• Spanwise Scale ($\Lambda_y$): A characteristic spanwise scale of $\Lambda_y \approx 50$ is observed, primarily in the streamwise velocity component. Other fields ($u_y, u_z, b$) show no distinct spanwise peak, indicating a lack of preferred spanwise organization compared to the streamwise direction.
• Anisotropy: The flow exhibits a clear hierarchy of scales ($\Lambda_z < \Lambda_y < \Lambda_x$), spanning nearly an order of magnitude.

C. Physical Interpretation

• Self-Organization: The flow "tunes" itself to a state where $Ri_g \lesssim 0.2$. This is consistent with the Miles-Howard criterion for linear stability and suggests the flow adjusts to the maximum stratification that still permits vigorous turbulence.
• Vestigial Imprint: The authors conjecture that the emergent large-scale streamwise structures are a "vestigial imprint" of the primary linear instability (KHI) acting on the converged, deepened shear layer, rather than the initial thin layer.
• Time Scales: A hierarchy of time scales is identified ($t_{adv} < t_K < t_S < t_N < t_\nu < t_b < t_r$), confirming the flow is shear-dominated and weakly stratified.

5. Significance and Implications

• Computational Modeling: The findings challenge the standard practice of using small periodic domains in DNS of stratified shear flows. To obtain reliable, converged statistics for mixing and dissipation, domains must be scaled to the emergent turbulent depth (approx. 100x the initial depth), not the initial conditions.
• Geophysical Relevance: The identified large-scale anisotropy (streamwise scales up to 100x the shear layer depth) suggests that in geophysical flows (e.g., estuarine outflows, oceanic mixed layers), emergent structures could be significantly affected by rotation (Coriolis effects) and other large-scale processes previously ignored in idealized models.
• Mixing Parameterization: The observation that flows self-tune to a specific $Ri_g$ range supports the development of parameterizations for turbulent mixing in climate and ocean models, suggesting a universal equilibrium state for forced stratified turbulence.

In conclusion, this paper provides a rigorous demonstration that forced stratified shear flows possess intrinsic, anisotropic large-scale structures and a self-regulating mechanism that sets the depth of the turbulent mixing zone, fundamentally altering how such flows should be modeled and understood.