Measuring the effect of spatial dimension on hydrodynamic turbulence using direct numerical simulation
Richard D.J.G. Ho, Daniel Clark, Andres Armua, Xichao Yang, Daniel J. Brener, Arjun Berera · 2025
Read Measuring the effect of spatial dimension on hydrodynamic turbulence using direct numerical simulation by Richard D.J.G. Ho, Daniel Clark, Andres Armua, Xichao Yang, Daniel J. Brener, Arjun Berera (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2512-00449.txt.
Introduction
Turbulence is usually studied in three dimensions, but its character changes with the number of dimensions available. Richard Ho and co-authors measure that effect with direct numerical simulation in 2025. They solve the incompressible Navier-Stokes equations with forcing at different spatial dimensions.
The team measures turbulent and chaotic properties, including Lyapunov exponents, which quantify how fast nearby flows diverge. Those exponents fell with dimension in their runs, turning negative in six dimensions at modest Reynolds numbers. Such results test basic theories of chaos and turbulence in a clean, controlled setting. Direct numerical simulation resolves every scale, which is why dimension studies use it. Lyapunov exponents turn the vague word chaos into a measurable number. For folks who study storms in air or sea, the gain is a clean lab. The spin of chaos can be weighed as the room it plays in gains new sides.
Catalog record
| CREATOR | Richard D.J.G. Ho, Daniel Clark, Andres Armua, Xichao Yang, Daniel J. Brener, Arjun Berera |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 19 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2512-00449
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Measuring the effect of spatial dimension on hydrodynamic turbulence using direct numerical simulation Preprint: APS/123-QED Richard D.J.G. Ho, Daniel Clark, Andres Armua, Xichao Yang, Daniel J. Brener, Arjun Berera Abstract We perform direct numerical simulation of the incompressible Navier-Stokes equation with forcing at different spatial dimensions and measure turbulent and chaotic properties. Lyapunov exponents, λ \lambda , decrease with dimension, and λ < 0 \lambda<0 for all simulations in six-dimensions up to R e = 40 Re=40 . These six-dimensional simulations display non-Gaussian statistics and other behavior similar to well developed turbulence despite their lack of chaos. Further, we find that small scale perturbations do not extend to the largest scales and that this terminal scale between correlation and decorrelation shrinks with dimension. We theorize that this change is related to the increased rate of vortex stretching. We find the interplay between turbulent and chaotic properties changes with increasing dimension. Introduction. —The Navier-Stokes equations (NSEs) in d -dimensions are a set of d non-linear partial differential equations which display sensitive dependence on initial conditions: i.e. chaos. Chaotic properties have emerged as an additional avenue for studying the NSEs, with the advantage of robustness of results at lower Reynolds numbers, Re , compared to spectral measures Ho et al. 2024 . Using direct numerical simulation (DNS), the advantage of using chaotic properties has already been shown in analysis of thin-layer turbulence, where the finite time Lyapunov exponent (FTLE) displays a discontinuous jump as the dimension is changed, with a transition between different dynamics Clark et al. 2021a . Outside DNS, numerical solutions to an eddy damped quasi-normal Markovian (EDQNM) approximation for the NSEs have found increased forward energy transfer at higher dimensions, with the enstrophy production term reaching a maximum near five dimensions Clark et al. 2021b . Remarkably, a transition was later found to a non-chaotic regime above d=5.88 Clark et al. 2022 , with no positive Lyapunov exponent measured even at very high Reynolds numbers. These results recaptured some of the same trends seen in 4d DNS Berera et al. 2020 . Analyzing a system’s physical properties whilst changing dimensionality has had a rich history of striking results. This was perhaps best demonstrated in the field of critical phenomena with the discovery of the Wilson-Fischer renormalization group fixed point Wilson and Fisher 1972 . This success in the early 1970s naturally prompted the application of similar methods with dimension as an analytic parameter in fluid turbulence Kraichnan 1959 ; Fournier et al. 1978 , though this program has proven to have had limited success. Difficulty in the theoretical analysis of the NSEs arises due to the turbulent regime being in the strong coupling limit, analogous to the problems encountered in quantum chromodynamics Cheung and Mangan 2020 . Even so, Liao developed an approach, very similar to the Wilson-Fisher theory for deriving critical phenomena, for the freely decaying isotropic and homogeneous Navier-Stokes turbulence Liao 1990 ; Liao 1991 . Liao derived to the one-loop level the Kolmogorov exponents for the energy spectrum of freely-decaying, fully-developed, near-incompressible turbulence, and found to first order an IR stable fixed-line with upper critical dimension of 6, which has also been found in other renormalization group approaches Verma 2024 . The results from EDQNM corroborate Liao’s work suggesting that interesting behavioral changes occur at six dimensions in the NSEs Clark et al. 2022 . Prior to Liao, Nelkin proposed, through a similar analogy to critical phenomena, the existence of a crossover dimension, above which the Kolmogorov scaling relations in K41 Kolmogorov 1962 become exact Nelkin 1974 ; Nelkin 1975 . Following Nelkin, Kraichnan derived that temporal fluctuations of a passive scalar field acting as a tracer for an incompressible d -dimensional turbulent flow were suppressed by increasing dimension Kraichnan 1974 . Around the same time, these ideas led to the use of dynamic renormalization group methods by Forster et al. Forster et al. 1976 ; Forster et al. 1977 and DeDomincis and Martin DeDominicis and Martin 1979 . Further analytical progress in this direction has proved challenging leaving numerical approaches, in particular DNS, as the optimal methodology with which to explore fluid turbulence in higher dimensions Gotoh et al. 2007 ; Yamamoto et al. 2012 . The insight we have added to these earlier studies has been to also utilize chaotic measures in studying the dimensional dependence of fluid turbulence Ho et al. 2024 . Very different from spectral measures, chaotic measures give a local measure of the flow and we have found them to be stable from fairly small box sizes Ho et al. 2020 . DNS results for spectral measures and scaling are prohibitively computational intensive at higher dimensions, whereas chaotic measures could provide reliable indicators and trends of possible changes in system behavior already at modest box sizes. In this Letter, we analyze the properties of the NSEs using DNS from 3 to 6 dimensions. When using N^{d} collocation points, computational cost rises faster than d , and high Re DNS of homogeneous isotropic turbulence (HIT) are prohibitively expensive. Since chaotic properties of turbulence have shown promise in analysis of turbulence at lower Re , this makes them the most suitable measure to identify changing behavior with dimension. Simulations. —The NSEs are simulated with a pseudospectral method using code modified from the one described in Yoffe 2013 ; Ho 2019 ; Berera et al. 2020 . Similar crosschecks and benchmarks were tested on this code as in these references just stated. The forcing employed is a delta-correlated in time stochastic force only applied to the low- k (large spatial scale) wavenumbers. To measure the chaotic properties of the system, we principally use the difference field \textbf{u}_{\Delta}=\textbf{u}_{1}-\textbf{u}_{2} , which has associated uncorrelated energy E_{\Delta}(k) , representing the error, and the finite time Lyapunov exponent (FTLE) Ott 2002 , \lambda , defined by \lambda_{i}(\Delta t)=\frac{1}{\Delta t}\text{ln}\left(\frac{|\textbf{u}_{\Delta}(\Delta t)|}{\delta_{0}}\right)\ , with the difference field allowed to evolve over a time interval \Delta t , and \textbf{u}_{1} rescaled to its original size of \delta_{0} by \textbf{u}_{1}=\textbf{u}_{2}+(\textbf{u}_{1}-\textbf{u}_{2})/\delta_{0} . The FTLE is then found after averaging over many iterations, i . The FTLE is normalized by the timescale T_{E_{0}}=E/\varepsilon , being the ratio of total energy to dissipation. The Reynolds number, Re=UL/\nu , where \nu is the viscosity and U^{2}=\frac{2}{d}\int_{0}^{\infty}\text{d}kE(k)\ ,\ L=\frac{\Gamma\left(\frac{d}{2}\right)\sqrt{\pi}}{\Gamma\left(\frac{d+1}{2}\right)U^{2}}\int_{0}^{\infty}\text{d}k\frac{E(k)}{k}\ , which depend on spatial dimension, d . FTLE were taken after a statistically steady state of E(k) was reached, which occurred much earlier for higher dimensions. The range of simulation parameters are summarized in Table 1 . d Re N^{d} \lambda T_{E0} 3 5.0 - 1790 128 - 1024 - 0.19 \rightarrow 16.73 4 22.4 - 184 32 - 256 - 0.02 \rightarrow 1.90 5 21.0 - 100 32 - 96 - 0.24 \rightarrow 0.72 6 21.2 - 40.4 32 - 48 - 0.39 \rightarrow - 0.16 Table 1: Range of simulation parameters Results. —Figure 1 shows the dependence of the FTLE with Reynolds number at different dimensions. The results here are similar to the ones seen in EDQNM simulations Clark et al. 2022 , where there was an approach to a scaling \lambda\sim Re^{-\alpha} with \alpha independent of dimension, but with a divergence at low Re indicating a crossover from negative to positive Lyapunov