Extended Self-Similarity in Turbulence

• ESS is a statistical scaling concept that reveals power-law relationships between structure functions, enabling robust exponent extraction even outside classical inertial ranges.
• It has been successfully applied across diverse systems such as atmospheric flows, wall-bounded turbulence, optical speckle, and plasmas, highlighting its cross-disciplinary impact.
• The ESS workflow involves data normalization, computation of structure functions, and log–log regression to extract normalized exponents, effectively mitigating finite-size and noise effects.

Extended Self-Similarity (ESS) is a statistical scaling concept originating in turbulence research that uncovers power-law relationships between structure functions (or related statistical moments) of different orders, even in regimes where the traditional self-similar, scale-invariant, or inertial-range power-laws fail or are absent. ESS has since found applications across a wide domain of classical and quantum turbulent, disordered, and complex systems, including atmospheric flows, Rayleigh-Bénard convection, wall-bounded turbulence, optical speckle, plasmas, and active matter.

1. Mathematical Foundations and Core Definitions

The central object of ESS analysis is the $p$th-order structure function of a fluctuating field $u(x)$ (velocity, temperature, density, intensity, etc.), typically defined as

$S_p(r) = \langle |u(x + r) - u(x)|^p \rangle$

with $\langle \cdot \rangle$ denoting an ensemble or spatial/temporal average and $r$ a spatial, temporal, or otherwise relevant separation.

In scale-invariant (classically self-similar) regimes, conventional theory (e.g., Kolmogorov 1941 for turbulence) predicts

$S_p(r) \sim r^{\zeta_p}$

with scaling exponents $\zeta_p$ (e.g., $\zeta_p = p/3$ for K41 inertial-range velocity increments).

ESS reframes this relationship by positing that, even when $S_p(r)$ vs. $r$ displays no clear power-law, plotting $S_p(r)$ against a reference structure function $S_q(r)$ (as $r$ is varied) often yields a robust extended power-law:

$S_p(r) \sim [ S_q(r) ]^{\beta_{p,q}}$

where $\beta_{p,q} = \zeta_p / \zeta_q$. When $q=3$ (chosen due to the exact four-fifths law for $S_3$ in isotropic turbulence), the ESS exponents are directly interpretable as normalized scaling exponents.

A closely related formulation appears in wall-bounded turbulence for moment-generating functions (MGFs) of velocity fluctuations, $M(q, z) = \langle e^{q u_z^+} \rangle$, which obey

$M(q, z) \sim [ M(q_0, z) ]^{\xi(q, q_0)}$

with $\xi(q, q_0) = \tau(q) / \tau(q_0)$, and where $\tau(q)$ is the MGF scaling exponent (Yang et al., 2016).

2. ESS in Turbulent Flows and Wall-Bounded Turbulence

ESS was first introduced to analyze fully developed isotropic turbulence, where Benzi et al. demonstrated that plotting $S_n(r)$ against $S_3(r)$ uncovers extended, often superior, scaling regions compared to plots versus $r$ (McComb et al., 2013). The phenomenon was quickly found to provide robust power-law behavior for exponents $\zeta_n / \zeta_3$ even outside the strict inertial range. Typical empirical findings for velocity structure functions are:

$n$	Empirical $\zeta_n^{\text{ESS}}$ (Taylor $\text{Re}_\lambda \sim 200$–800)
2	0.70–0.71
4	1.27–1.30
6	1.75–1.80

ESS analysis in high-Reynolds-number wall turbulence revealed universal scaling behaviors between energy-containing and dissipative ranges, both for longitudinal and transverse structure functions (Krug et al., 2017). In the energy-containing range, the relative exponents $D_p/D_m = \zeta_p / \zeta_m$ (where $D_p$ is the log-law slope for $S_p$) collapse across canonical flows (channel, pipe, Taylor-Couette) and exhibit universality, extending even to cases with anisotropic eddies, as streamwise and spanwise scaling differences are neutralized in the ESS ratio (Krug et al., 2017).

Furthermore, moment-generating functions $M(q, z)$ in wall-bounded turbulence were shown to satisfy ESS scaling over a far broader $z^+$ range ($30 \leq z^+ \leq \delta$), encompassing the buffer, logarithmic and part of the outer layer (Yang et al., 2016). This expansion is crucial for precisely extracting scaling exponents in high-Re cases.

3. Algorithmic Application and Quantitative Extraction

The standard workflow for ESS involves:

1. Data acquisition and normalization: Time series or snapshots of the fluctuating field are collected and, if necessary, normalized.
2. Structure function computation: For each order $p$ and range of lag values $r$, $S_p(r)$ is calculated.
3. ESS plotting: For a suite of $p$, log–log plots of $S_p(r)$ vs.\ $S_q(r)$ (with $q$ typically 1 or 3) are generated.
4. Exponent extraction: Linear regression over the largest region exhibiting geometric linearity is used to compute $\beta_{p,q}$.

This approach generalizes both to time (e.g., wind increments over lags in atmospheric data (Kiliyanpilakkil et al., 2015)) and to scalar structure functions (e.g., temperature increments in Rayleigh-Bénard convection (Krug et al., 2018)).

Bootstrapped uncertainty estimates and surrogate-data analyses form standard tools for underpinning statistical claims, with surrogate series consistently yielding the “expected” monofractal exponents (e.g., $\zeta_{p,3}^* \approx p/3$), confirming the robustness of ESS to experimental noise and finite-size effects (Kiliyanpilakkil et al., 2015).

4. Universality, System Independence, and Generalizations

ESS has been verified in a spectrum of systems, with notable results demonstrating universality of the exponent ratios:

• In two-dimensional complex plasmas, ESS was clearly observed for velocity structure functions of both active Janus-particle suspensions and externally driven passive suspensions (Nosenko, 30 Apr 2025). Here, ESS persisted even when traditional distance-based power-law scaling was barely discernible, highlighting its diagnostic value in systems lacking a classical cascade.
• For multimode optical fiber speckles, the ESS exponents for intensity fluctuations matched the Kolmogorov prediction to within a few percent despite the underlying process being strictly linear (deterministic modal mixing). This demonstrates that nonlinearity, while essential for turbulent cascades, is not a prerequisite for ESS—the crucial ingredient is multiscale mode-mixing (Wu et al., 25 Jan 2026).
• Strong agreement in ESS exponents is observed for both scalar (temperature) and velocity structure functions across wall-bounded and Rayleigh-Bénard flows, confirming universality up to high orders and independence from whether the underlying field is an active or passive scalar (Krug et al., 2018).

Empirical findings consistently show sub-Gaussian (multiscaling) corrections to the exponent ratios, signaling the presence of intermittency across both classical turbulence and non-traditional systems.

5. Theoretical Models and Physical Interpretation

The physical mechanisms behind ESS differ by context but share the essential property of hierarchical multiscale organization.

• In wall-bounded turbulent flows, ESS can be derived within the attached-eddy hypothesis. Pure self-similarity of eddy contributions (variance) is not strictly required: the normalization inherent in ESS ratios cancels certain scale-dependent terms, revealing scale-invariant relationships even where strict log-laws break down outside the canonical region (Krug et al., 2017).
• The two-component (“hierarchical random additive process”) model for wall-turbulence moment-generating functions splits the signal into large- and small-scale components. Only small scales produce a pure power-law in $z^+$, but both components individually obey ESS over a broad range, with the theoretical construction reproducing all observed scaling behaviors (Yang et al., 2016).
• In complex plasmas and active matter, even when no inertial-range cascade exists, ESS persists as a statistical property that captures the hierarchical ordering among fluctuations. Similarly, in Bose-Einstein condensate simulations, density structure functions display ESS scaling in both space and time, suggesting nonlinear mode-mixing creates effective cascades akin to turbulence (Zhao, 2024).

The physical interpretation is that ESS exposes an underlying, generalized self-similarity—often hierarchical or multifractal in nature—that is otherwise obscured by boundary effects, dissipation, anisotropy, or lack of a clear separation of scales.

6. Critique, Limitations, and Controversies

Despite its utility, ESS has limitations:

• Extrapolation into Dissipative/Viscous Ranges: ESS scaling often extends into scales where no traditional inertial-range power law is expected, raising questions over its physical interpretation. Barenblatt et al. argue that small-$r$ regularity can generate ESS-like plateaus even in monofractal (or trivial) processes (McComb et al., 2013).
• Masking Finite-Size Corrections: ESS, by construction, may conflate finite-Reynolds-number corrections with genuine anomalous scaling (intermittency). The “ratio method” (plotting $|S_n(r)/S_3(r)|$ vs $r$) was proposed as a cross-check: in isotropic turbulence, it recovers $\zeta_2 \to 2/3$ in the infinite Reynolds number limit, while ESS-derived $\zeta_2$ is observed to increase with $R_\lambda$ (McComb et al., 2013).
• Ambiguity of Deviations: ESS alone cannot distinguish whether observed deviations from theoretical exponent values arise from multifractality, finite statistical ensemble, sub-dominant contributions, or system-specific physics. Complementary analyses are necessary for robust physical interpretation.

7. Broader Impact and Current Research Directions

ESS is now recognized as a diagnostic tool applicable to a wide array of physical systems—classical and quantum turbulence, plasmas, geophysical flows, nonlinear optics, and more. Its invariance to certain system details, e.g., dimension (1D BEC, 2D plasmas, 3D turbulence), points to a deeper statistical principle governing multiscale organization in driven systems.

Recent work has focused on:

• Sensitive benchmarking of turbulence models and closures in numerical simulations (e.g., atmospheric boundary layers and mesoscale wind models) (Kiliyanpilakkil et al., 2015).
• Probing the breakdown or emergence of cascade mechanisms, such as the laminar-to-turbulent BL transitions in Rayleigh-Bénard convection, via the onset of ESS scaling (Krug et al., 2018).
• Establishing the presence of ESS in purely linear systems, challenging the notion that intermittency or nonlinearity are necessary prerequisites (Wu et al., 25 Jan 2026).
• Leveraging ESS to guide theoretical development in settings with obscure or absent inertial ranges, including high-Prandtl-number convection, stratified/rotating flows, and quantum fluids (Zhao, 2024).

The trend in contemporary research is to use ESS not only as a tool for improved exponent extraction, but as a window into the universality and statistical structure underlying complex dynamics far beyond classical fluid turbulence.