Fusion Rules in Navier-Stokes Turbulence: First Experimental Tests

Adrienne L. Fairhall, Brindesh Dhruva, Victor S. L’vov, Itamar Procaccia, Katepalli R. SreenivasanPhysical 1997 · DOI · arXiv

Abstract

We present the first experimental tests of the recently derived fusion rules for Navier-Stokes turbulence. These rules address the asymptotic properties of many-point correlation functions as spatial coordinates coalesce, forming a critical component of the nonperturbative statistical theory of turbulence. By analyzing high-Reynolds-number atmospheric data, we test these rules within the inertial range and find strong agreement between experimental observations and theoretical predictions. Notably, for inertial-range separations, we identify a simple linear relationship for the Laplacian of velocity fluctuations conditioned on velocity increments.

Introduction: The Closure Problem and Fusion Rules

Fluid turbulence is often cited as the last great unsolved problem of classical physics. At the heart of this challenge lies the closure problem: in the Navier-Stokes (N-S) equations, the evolution of any $n$-point correlation function depends on $(n+1)$-point correlations. This creates an infinite hierarchy, forcing researchers to rely on ad-hoc assumptions or phenomenological models to "close" the equations and make them solvable.

Recently, a non-perturbative statistical theory introduced the concept of fusion rules. These rules provide a rigorous mathematical description of how many-point correlation functions behave when a subset of coordinates approaches one another, or "fuses." If validated, these rules offer a path toward understanding the analytical structure of turbulence without the need for arbitrary closures. This paper details the first experimental validation of these rules using high-Reynolds-number atmospheric flow.

Experimental Methodology

The study utilized a massive dataset comprising 10 million samples of atmospheric wind speed. The data was collected using a single hot-wire probe mounted 35 meters high on a meteorological tower at Brookhaven National Laboratory.

To translate the time-series signal from the stationary probe into a spatial field, the researchers employed Taylor’s hypothesis, which assumes the turbulent field is "frozen" as it sweeps past the sensor. The resulting dataset reached a Taylor microscale Reynolds number ($R_\lambda$) of approximately 9540, representing highly developed turbulence. The analysis focused on equal-time $n$-tensor correlation functions ($F_n$) and structure functions ($S_n$) across three decades of the inertial range—the scale where energy is transferred from large eddies to smaller ones without being dissipated by viscosity.

Key Insights and Theoretical Validation

1. Factorization of Correlation Functions

The core prediction of the fusion rules is that many-point correlations factorize under specific geometric conditions. When a subset of $p$ points fuses within a larger set of $n$ points, the correlation function splits into a product of a $p$-order function and a remaining $(n-p)$-order function. The experimental data confirmed this factorization, demonstrating that the statistical structure of turbulence simplifies in predictable ways as scales diminish.

2. Geometric Scaling ($r/R$)

For the specific case where a single pair of points fuses ($p=1$), the theory predicts a distinct scaling relationship based on the ratio of the small separation ($r$) to the large separation ($R$). The experimental results showed clean scaling in the inertial range, matching the fusion rule predictions and confirming that the relative geometry of points is a fundamental determinant of turbulent statistics.

3. The Discovery of a Linear Law

One of the most surprising outcomes of the study was the discovery of a linear relationship between the conditional average of the Laplacian of velocity fluctuations and the velocity increment. While similar linear laws had been observed in "passive scalars" (such as the diffusion of smoke in air), finding this relationship in the actual velocity field of Navier-Stokes turbulence was unexpected. The data showed a near-perfect collapse onto a straight line with a slope of unity across multiple large-scale separations, suggesting a deeper, previously unrecognized simplicity in the statistical governing of flow.

Universal Constants and Statistical Independence

The researchers also examined the coefficient $C_n$, which relates the dissipative term $J_n$ to the structure functions $S_n$. The analysis found that $C_n$ is approximately 1 and, crucially, remains independent of $n$. This independence is a significant finding for the statistical theory of turbulence, as it suggests universal properties that transcend the specific order of the correlation being studied.

Limitations and Future Directions

While the study successfully validated the fusion rules in the inertial range, the resolution of the dissipative range (where viscosity dominates) remained a limitation. The Kolmogorov microscale was 0.57 mm, and the sampling frequency was insufficient to fully resolve the tiniest scales. Consequently, the second set of fusion rules—those applying when separations are smaller than the dissipative scale—could not be tested.

Future work will focus on acquiring higher-resolution data at high Reynolds numbers to probe these sub-dissipative scales. Understanding these "ultraviolet" properties is essential for a complete nonperturbative description of turbulence and remains the next frontier in validating the statistical theory of Navier-Stokes flows.

Conclusion

By providing the first experimental evidence for fusion rules, this research moves the statistical theory of turbulence from theoretical abstraction to empirical reality. The confirmation of factorization and the discovery of the linear law for the conditional Laplacian provide new, rigorous constraints for any future theories attempting to solve the Navier-Stokes equations.

Summary updated

Abstract

We present the first experimental tests of the recently derived fusion rules for Navier-Stokes turbulence. The fusion rules address the asymptotic properties of many-point correlation functions as some of the coordinates coalesce, and form an important ingredient of the nonperturbative statistical theory of turbulence. Here we test the fusion rules when the spatial separations lie within the inertial range, and find good agreement between experiment and theory. For inertial-range separations and for velocity increments which are not too large, a simple linear relation appears to exist for the Laplacian of the velocity fluctuation conditioned on velocity increments.
BibTeX
@inproceedings{fairhall1997fusion,
  author = {Adrienne L. Fairhall and Brindesh Dhruva and Victor S. L’vov and Itamar Procaccia and Katepalli R. Sreenivasan},
  title = {{Fusion Rules in Navier-Stokes Turbulence: First Experimental Tests}},
  booktitle = {Physical},
  year = {1997},
  doi = {10.1103/physrevlett.79.3174},
}
Figure