On the wall-normal velocity variance in canonical wall-bounded turbulence

This study investigates wall-normal velocity variance across various canonical wall-bounded turbulent flows, revealing that deviations from a universal constant are primarily driven by local shear stress variations and low-wavenumber "inactive" motions, while a semi-empirical fit successfully predicts variance trends that align with the attached eddy hypothesis in the high-Reynolds-number limit.

The Gist

Imagine you are standing on the bank of a fast-flowing river. You know the water is moving, but if you look closely, you see it's not just a smooth sheet of water. It's a chaotic dance of swirling eddies, tiny ripples, and massive currents all happening at once. This is turbulence.

Scientists have been trying to write a "rulebook" for how this turbulence behaves, especially near the riverbank (or in engineering terms, near a wall like a pipe or an airplane wing). This paper is a new chapter in that rulebook, focusing specifically on how the water moves up and down (wall-normal velocity) as it swirls.

Here is the story of what the researchers found, explained simply:

1. The Old Rulebook Was a Bit Too Simple

For decades, scientists relied on a famous idea called the Attached Eddy Hypothesis. Think of this hypothesis like a theory that says: "All the swirling water near the bank is made of tiny, self-similar whirlpools that are just scaled-down versions of bigger ones."

According to this old rule, if you measure how much the water jiggles up and down, it should be a constant number relative to how fast the water is flowing at the surface. It's like saying, "No matter how big the river is, the up-and-down wiggles are always exactly 1.5 times the speed of the current."

The Problem: When scientists looked at real data from super-computer simulations, the numbers didn't match perfectly. The "wiggle" amount changed depending on:

• How fast the river was flowing (Reynolds number).
• Whether the river was in a pipe, a channel, or an open flat area.

2. The New Discovery: It's About the "Local Boss"

The researchers realized the old rulebook was looking at the wrong boss. It was looking at the surface speed (the speed right at the wall) to predict the wiggles everywhere else.

The Analogy: Imagine a company. The old rule said, "The energy of a junior employee depends on the CEO's salary."
The new discovery says, "No! The energy of a junior employee depends on their immediate manager's salary."

In the river, the "immediate manager" is the local shear stress. As you move away from the wall, the force driving the turbulence changes. The researchers found that if you measure the "wiggle" against the local force at that specific height (rather than the surface force), the data from pipes, channels, and open rivers all line up perfectly.

3. The "Active" vs. "Inactive" Dancers

The paper also explains why the old rule worked so well for some flows but not others. They use a concept of "Active" and "Inactive" motions.

• Active Motions: These are the dancers who are actually doing the work. They are the swirls that push against the wall and create friction. These follow the "Local Manager" rule perfectly.
• Inactive Motions: These are the "spectators." They are huge, lazy swirls that float high above the wall. They don't really push against the wall, but they still take up space and add a little bit of extra "wiggle" to the water.

The Twist: In open rivers (Zero-Pressure-Gradient flows), these "spectator" swirls are a bit more energetic and larger than in pipes. This extra energy makes the open river wiggle slightly more than a pipe, even when you account for the local manager's salary.

4. The New "Perfect" Formula

The team created a new semi-empirical formula (a mix of theory and real-world fitting) that accounts for:

1. The Local Force: Using the local stress instead of the surface stress.
2. The "Spectators": Acknowledging that the big, lazy swirls add a tiny bit of extra energy, which changes slightly depending on the type of flow.

The Result:
They calculated that in the limit of a super-fast, massive river (infinite Reynolds number), the "wiggle" factor settles at a value between 1.45 and 1.65.

This is a huge step forward because:

• It explains why different flows (pipes vs. open rivers) look different.
• It predicts that there isn't just one magic number for all turbulence, but a small range, because the "spectator" swirls are never exactly the same in every situation.

Summary in a Nutshell

Think of turbulence as a crowded dance floor.

• Old View: Everyone dances to the rhythm of the DJ at the front of the room (Surface Speed).
• New View: Everyone dances to the rhythm of the person standing right next to them (Local Stress).
• The Catch: In some rooms (like open rivers), there are giant, lazy people in the back (Inactive motions) who sway a bit more than in other rooms (pipes), making the whole dance floor wiggle slightly differently.

This paper gives us a better map to predict how that dance floor will move, which is crucial for designing better airplanes, more efficient pipelines, and understanding weather patterns.

Technical Summary

1. Problem Statement

The study addresses three fundamental unresolved questions regarding the wall-normal velocity variance ($w'^2$) in canonical wall-bounded turbulent flows (channel, pipe, and zero-pressure-gradient (ZPG) boundary layers):

1. Reynolds Number Dependence: What is the correct formulation for the dependence of $w'^2$ on the Reynolds number, particularly within the logarithmic region?
2. Flow Configuration Differences: Why do ZPG boundary layers exhibit a higher $w'^2$ (normalized by surface friction velocity $U_\tau$) and a peak location farther from the wall compared to enclosed flows (channels and pipes)?
3. Universality of $B_3$: What is the asymptotic value of the constant $B_3$ in the Attached Eddy Hypothesis (AEH) relation $w'^2/U_\tau^2 = B_3$ at high Reynolds numbers, and is this value universal across different flow configurations?

Previous literature has reported a wide range of $B_3$ values (approx. 1.5 to 1.85) and struggled to reconcile differences between flow types, often attributing them to "enclosed" vs. "non-enclosed" flow effects without a clear mechanistic explanation.

2. Methodology

The authors analyzed Direct Numerical Simulation (DNS) datasets spanning a decade of friction Reynolds numbers ($Re_\tau$) for three canonical flows:

• ZPG Flat Plate Boundary Layer (Sillero et al., 2013)
• Channel Flow (Lee & Moser, 2015)
• Pipe Flow (Yao et al., 2023)

Key Analytical Approaches:

• Spanwise Wavenumber Spectra ($E_{ww}(k_y)$): Unlike streamwise spectra which often lack a distinct peak, the spanwise spectra exhibit a pronounced peak at intermediate scales. The authors analyzed the properties of this peak (wavenumber $k_{peak}$, energy density $E_{peak}$) and the low-wavenumber limit.
• Generalized Scaling Parameters: Instead of relying solely on the surface friction velocity ($U_\tau$) and wall distance ($z$), the study introduced:
  • Local Shear Velocity ($u_{\tau z}$): Defined by the total local shear stress (turbulent + viscous), $u_{\tau z} = \sqrt{-\overline{u'w'} + \nu \partial U/\partial z}$.
  • Dissipation-based Length Scale ($\ell_\epsilon$): Defined as $\ell_\epsilon = u_{\tau z}^3 / \epsilon$, where $\epsilon$ is the turbulent dissipation rate. This scale bridges the gap between the production region and the inertial subrange.
• Scale Decomposition: The wall-normal variance was decomposed into "large-scale" (below the spectral peak) and "small-scale" (above the spectral peak) contributions to isolate the effects of active and inactive motions.
• Semi-Empirical Modeling: The authors fitted the spectral data to a von Kármán model and derived a semi-empirical expression for the variance based on the inertial subrange behavior and Reynolds number corrections.

3. Key Contributions

1. Identification of Local Stress Dominance: The study demonstrates that $w'^2$ is primarily governed by the local shear stress ($u_{\tau z}$) rather than the surface friction velocity ($U_\tau$). This explains why ZPG flows (where stress is nearly constant) differ from pressure-driven flows (where stress decays linearly).
2. Refined Reynolds Number Correction: A new semi-empirical formulation is proposed that accounts for finite Reynolds number effects using a local Reynolds number based on dissipation ($Re_\epsilon = \ell_\epsilon / \eta$).
3. Mechanism for Flow Differences: The paper attributes the higher variance in ZPG boundary layers to the specific shape of the mean stress profile and the contribution of low-wavenumber "inactive" motions, which are more pronounced in non-enclosed flows.
4. Re-evaluation of Townsend's AEH: The work validates the core of Townsend's Attached Eddy Hypothesis (that wall-normal motions are "active" and tied to local stress) while identifying a limitation: "inactive" motions (large-scale structures not contributing to local shear stress) introduce small but non-negligible discrepancies that prevent a single universal constant $B_3$.

4. Key Results

Spectral Properties:

• The spanwise spectral peak ($k_{peak}$ and $E_{peak}$) collapses universally when normalized by the local parameters $u_{\tau z}$ and $\ell_\epsilon$, rather than $U_\tau$ and $z$.
• This collapse holds from the viscous sublayer up to $z/\delta \approx 0.5$.
• At the lowest wavenumbers, ZPG flows show significantly more energy relative to the local scaling compared to enclosed flows, indicating a stronger contribution from large-scale "inactive" motions.

Variance Formulation:
The authors propose a semi-empirical equation for the wall-normal variance:
$$\frac{w'^2}{u_{\tau z}^2} = B_3 - f_3(Re_\epsilon)$$
Where:

• $B_3 \approx 1.55 \pm 0.1$ (the high-Reynolds-number limit).
• $f_3(Re_\epsilon) \approx 1.5 C_w Re_\epsilon^{-2/3} + 18 Re_\epsilon^{-2}$ accounts for finite Reynolds number effects (viscous cutoff and inertial subrange transition).
• In the logarithmic region, this can be expressed in terms of $z^+$ as $f_3(z^+) \approx 1.55 (z^+)^{-1/2} + 71 (z^+)^{-3/2}$.

Comparison of Flows:

• When variances are normalized by $u_{\tau z}$, channel, pipe, and ZPG flows show much better agreement.
• The remaining differences (e.g., the higher peak in ZPG flows) are attributed to the "inactive" low-wavenumber contributions, which vary with flow configuration (enclosed vs. non-enclosed) and wall-normal position.
• Plane Couette flow data, when adjusted for stress, aligns well with other enclosed flows, confirming that the stress profile is the primary driver of variance differences.

The Value of $B_3$:

• The extrapolated high-Reynolds-number limit is $B_3 \approx 1.55$.
• This value is consistent with atmospheric surface layer measurements but lower than some previous experimental estimates (which ranged up to 1.85).
• The paper argues that a single universal $B_3$ is precluded because the "inactive" contribution ($A_{ia}$) varies slightly between flow types, leading to a range of possible values (approx. 1.2 to 1.63 in the idealized model) rather than a fixed constant.

5. Significance

This paper provides a critical refinement to the understanding of wall-bounded turbulence statistics:

• Theoretical Advancement: It bridges the gap between the idealized Attached Eddy Hypothesis (which assumes constant stress) and real-world flows with decaying stress profiles by introducing local scaling ($u_{\tau z}$).
• Practical Application: The derived semi-empirical formula allows for more accurate prediction of wall-normal velocity variance in engineering applications (e.g., atmospheric boundary layers, pipe transport) by accounting for local stress and Reynolds number effects.
• Resolution of Discrepancies: It resolves long-standing debates about why different canonical flows yield different variance profiles, attributing the differences to the interplay between local stress profiles and the varying footprint of inactive large-scale motions.
• Limitation of Universality: It suggests that while the "active" component of turbulence is universal, the "inactive" component introduces a small but fundamental non-universality in second-order statistics, challenging the search for a single, exact universal constant for $B_3$.