Linear response of turbulence

H. E. Cekli and G. Bertens and W van de Water · 2025

Read Linear response of turbulence by H. E. Cekli and G. Bertens and W van de Water (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-2508-05362.txt.

Introduction

Cekli and co-authors test how turbulence in a wind tunnel answers a poke, in a 2025 paper. A classic theory says the answer fades at a set pace as the flow grows stronger, the poke shrinks in scale, or time passes. The team built two ways to poke the flow.

An active grid made both the churn and the poke in the first setup. The answer faded with time but far slower than theory said, and flow strength did not matter. In the second setup, a loudspeaker jet poked at one set scale. Fading sped up for finer pokes, roughly as theory says, yet flow strength again played no role.

Catalog record

CREATORH. E. Cekli and G. Bertens and W van de Water
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE29 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2508-05362

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Linear response of turbulence H. E. Cekli and G. Bertens and W van de Water Abstract We study the response of wind tunnel turbulence to perturbations using an active grid. We compare our findings to Kraichnan’s linear response result R ⁡ ( k , τ d ) = exp ⁡ ( − k 2 ​ τ d 2 ​ u 2 ) R(k,\tau_{d})=\exp(-k^{2}\>\tau_{d}^{2}\>u^{2}) which predicts a decay of the response with increasing turbulence intensity u u , wave number of the perturbation k k and delay time τ d \tau_{d} since the perturbation was applied ( Kraichnan 1964 ) . In our experiments we used two different mechanisms to create a perturbation of a pre–existing and well–developed turbulent flow. In the first case we combine both the turbulence generation and the additional random perturbation in the same active grid. We find that the reponse decays with increasing τ d \tau_{d} , but much slower than predicted, while the decay was virtually independent of the turbulence intensity. In the second type of experiments, we perturb an active grid–generated turbulent flow using a loudspeaker–driven synthetic jet. The perturbation is at a single wave number k k , placed well inside the inertial range. The response was found to decay with increasing time delay τ d \tau_{d} , while the decay is faster for larger wave numbers, roughly as predicted by Kraichnan’s model. However, also in this case the decay rate is independent of the turbulence intensity. I Introduction The fluctuation-dissipation theorem of statistical physics expresses that in thermal equilibrium the linear response of a system is proportional to the magnitude of its intrinsic thermal fluctuations. It is tantalizing to apply this concept to turbulence, as it would link the response of a turbulent flow to forcing to the size of its spontaneous velocity fluctuations. The problem is that turbulence is not linear and it is not in thermal equilibrium. Nevertheless, the idea of linear response is central to Kraichnan’s direct interaction approximation, and there have been several experimental and numerical attempts to measure the linear response of turbulence. Understanding the response of turbulence to perturbation is the central issue in the field of turbulence control. This is an important technological application which aims at reducing turbulent drag or delay separation. A large body of work exists where the methods of linear control theory are aplied to the turbulent boundary layer ( Kim 2003 ) . The linear regime dictates very small perturbation magnitudes, which presents an experimental and numerical challenge. A more philosophical problem is the definition of the response function in an experiment. In numerical simulations, the unperturbed state can be specified completely in a perfectly reproducible way. Averages can then be performed over an ensemble of unperturbed states. This is impossible in an experiment where averages have to be done over the perturbed state instead. The response of a laminar boundary layer was first studied by Schubauer and Skramstad 1947 and that of a turbulent boundary layer by Hussain and Reynolds 1970 . In both experiments the perturbation was at a single wave number and frequency. The response of homogeneous grid–generated turbulence was measured by Kellogg and Corrsin 1980 and Itsweire and van Atta 1984 . The experimental results were compared to Kraichan’s simple advection model ( Kraichnan 1964 ) in which the forced small scales are advected randomly by the large scales, for which Gaussian statistics is assumed. This model provides an attractive organizing principle for experiments and simulations, with the linear response function: R(k,\tau_{d})=e^{-\frac{1}{2}\>k^{2}\>\tau_{d}^{2}\>u^{2}}, (1) which decays very rapidly with increasing large–scale turbulence intensity u , increasing wave number k of the perturbation, and increasing elapsed time \tau_{d} since the perturbation was applied. In order to appreciate the response function (Eq. 1 ), let us briefly sketch a derivation ( Kraichnan 1964 ) . In the random advection problem, the small–scale velocity field v is advected by a large–scale Gaussian field u and forced by F , there is no other dynamics. Thus, ignoring pressure and viscosity, the Fourier amplitudes \widetilde{v}_{i} satisfy \frac{\partial\widetilde{v}_{i}}{\partial t}-\imath k_{j}\>u_{j}\widetilde{v}_{i}=\widetilde{F}_{i}. If the forcing is delta–correlated in time, this equation yields the linear response (Eq. 1 ) \left\langle\widetilde{v}_{i}(\mbox{\boldmath$k$},t+\tau_{d})\>\widetilde{F}_{i}^{*}(\mbox{\boldmath$k$},t)\right\rangle=\mbox{e}^{-\frac{1}{2}\>k^{2}\>\tau_{d}^{2}\>u^{2}}\left\langle\widetilde{F}_{i}(\mbox{\boldmath$k$},t)\>\widetilde{F}_{i}^{*}(\mbox{\boldmath$k$},t)\right\rangle. Central to the derivation of (Eq. 1 ) is a separation of the scales of the fluctuations v and the force, and those of the random field u . Therefore, the force has to act at inertial–range scales. An experimental attempt to measure the response of homogeneous turbulence was made by Kellogg and Corrsin 1980 . They disturbed a standard grid–generated, decaying and nearly–isotropic turbulent flow in a wind tunnel using a separate extra grid to impose the perturbation. The first grid produced the turbulence while the second grid, which consisted of a one-dimensional mesh of parallel wires, imposed the perturbation at wave number k , with k determined by the separation of the wires. Similar experiments were done by Itsweire and van Atta 1984 . Both articles document the distortion of the spectrum due to the perturbation. Kellogg and Corrsin 1980 found a near–perfect exponential decay with distance of the oscillation amplitude in the two–point correlation function. In these experiments, the time delay \tau_{d} since application of the perturbation is set by the separation x_{1} to the second grid as \tau_{d}=x_{1}/U , where U is the mean velocity. Surprisingly, although (Eq. 1 ) implies an exponential decay with increasing \tau_{d}^{2} , rather than with increasing \tau_{d} , Kellogg and Corrsin 1980 conclude good agreement of their response function with (Eq. 1 ). Kellog and Corrsin also consider a time–dependent perturbation using an oscillating wire and suggest modulation using a kind of active grid. No response function for these alternative excitations were reported. The present paper was inspired by their suggestion. Camussi et al. 1997 studied the evolution of a perturbation generated by a small pulsed jet in the presence of a turbulent jet flow. They measured the response function as the Fourier transform of the velocity, conditionally averaged on the pulsed perturbation. Agreement with (Eq. 1 ) was found. Similarly to the present experiment, the time delay was set by the spatial separation x_{1} of the probe and perturbation. However, with x_{1} also the turbulence intensity changed, so that \tau_{d} and u varied simultaneously in these experiments. Evidence for the response function (Eq. 1 ) was found in direct numerical simulations of turbulence by Carini and Quadrio 2009 . By forcing with white noise, the response function is computed simultaneously at all wave numbers. A similar appoach was used by Luchini et al. 2006 to compute the response of a turbulent channel flow. Using the axisymmetry of a turbulent von Kármán flow, Monchaux et al. 2008 identified fluctuation–dissipation relations in their analysis of the fluctuating velocity fields measured in their experiment. Let us now sketch how we will confront (Eq. 1 ) with experiments. Our experiments are done in a recirculating wind tunnel, in which relatively small turbulent velocity fluctuations (u\approx 1\>\text{ms}^{-1}) are carried by a mean wind (U\approx 10\>\text{ ms}^{-1}) . We will impose perturbations on this turbulent flow as temporal modulations at a fixed location, and study their fate at various downstream locations, with x_{1} the distance to the source of the perturbation.

Read the complete text