Exact coherent structures as building blocks of turbulence on large domains

Dmitriy Zhigunov, Jacob Page · 2026

Read Exact coherent structures as building blocks of turbulence on large domains by Dmitriy Zhigunov, Jacob Page (2026) 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-2601-21970.txt.

Introduction

Dmitriy Zhigunov and Jacob Page study turbulence through its hidden skeleton. Exact unstable solutions of the Navier-Stokes equations are thought to underpin how turbulence behaves. These exact coherent structures are usually computed only in tiny domains, the smallest boxes that can sustain turbulent motion.

Real flows, though, sprawl across large domains with many interacting substructures. The team's 2026 preprint extends the dynamical systems approach to such extended flows. They show how new, simple invariant solutions can be built by combining smaller pieces in space. If large-scale turbulence really is assembled from these blocks, the picture could guide both theory and smarter flow control in engineering.

Catalog record

CREATORDmitriy Zhigunov, Jacob Page
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE19 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2601-21970

Browse the full library · About the Tower of Babel

Full text

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

Exact coherent structures as building blocks of turbulence on large domains Preprint: APS/123-QED Dmitriy Zhigunov, Jacob Page Abstract Exact unstable solutions of the Navier-Stokes equations are thought to underpin the dynamics of turbulence, but are usually computed in minimal computational domains. Here, we extend this dynamical systems approach to spatially extended turbulent flows featuring multiple interacting ‘substructures’, and show how new simple invariant solutions can be constructed by spatial tiling of exact solutions from small-box calculations. Candidate solutions are found via gradient-based optimization of a scalar loss function which targets autorecurrence in spatially-masked regions of the flow. We apply these ideas to a vertically-extended Kolmogorov flow, where we first identify large numbers of relative periodic orbits (RPOs) which are combinations of high-dissipation, small-box solutions with laminar patches. We then show that vertically-stacked combinations of pairs of distinct small-box RPOs can form robust guesses for dynamically-relevant two-tori in the larger domain. Finally, we show how our optimization procedure can identify ‘turbulent’ trajectories which locally shadow a small-box RPO for multiple periods in a subdomain. These small-box combinations are possible as the flow spends prolonged periods in a regime where it can be effectively considered as a pair of weakly-coupled small-box systems, due to shielding effects associated with higher-dissipation flow structures. The dynamical systems view of turbulence considers a realization of a particular flow as a trajectory in a high-dimensional state space, pinballing between exact, unstable solutions of the Navier Stokes equations 16 . This picture has been realised experimentally in both 2D 20 and 3D 6 flows. Early applications of these ideas explained many aspects of the subcritical transition to turbulence in parallel shear flows 11 ; 22 , resulting in intensive focus on the identification and computation of exact solutions in fully-developed turbulence 15 ; 12 ; 4 . The exact solutions, which are typically relative periodic orbits (RPOs) due to the presence of continuous symmetries in many of the canonical fluid systems, each contain a closed cycle of physical processes helping to sustain the fully turbulent state 21 ; 14 . Computing RPOs is notoriously challenging, particularly in more strongly turbulent systems, and has restricted much of the work to date to weak turbulence in ‘minimal’ computational domains (e.g. see the review 13 ). While recent approaches based on optimization have produced enough solutions to reconstruct turbulent statistics at modest Reynolds numbers 18 , Re , there are to our knowledge no examples of exact solutions featuring multiple spatially localized substructures typically associated with turbulence in larger domains. There have been notable examples of spatially localized exact solutions obtained in extended domains for various parallel shear flows. This includes the identification of spanwise-localized roll/streak solutions (and associated homoclinic snaking) obtained by 19 and streamwise-localized RPOs computed at transitional Re in a pipe 1 . Solutions which localize in the cross shear direction have also been found in planar configurations 8 ; 10 and appear to continue indefinitely up in Re . A slightly different perspective was adopted by 2 , who sought new large-box exact solutions (in a reduced system of equations) which are spanwise-spatially-modulated small-box states. However, these ‘substructures’ are all dominated by a single cross-shear lengthscale, while here we explicitly seek new solutions with multiple substructures. The method is inspired by the ideas introduced by Cvitanović and coauthors 17 ; 7 describing spatiotemporal chaos as a space-time tapestry composed from smaller solutions. In this Letter we construct spatially extended solutions of the Navier-Stokes equations from combinations of small-box exact solutions. Our starting point is a large library of RPOs assembled in two-dimensional Kolmogorov flow on a 2\pi\times 2\pi domain 5 which we seek to combine in a larger 2\pi\times 4\pi box. The new solutions include (i) exact RPOs which are spatially-tiled combinations of a small box RPO with the laminar solution; (ii) unstable two-tori which are constructed from pairs of non-trivial small-box RPOs ; and (iii) turbulent trajectories which locally shadow a small box solution in a subregion of the flow domain. Computation of the new states is performed with a combination of gradient-based optimization and classical Newton-Krylov methods, and opens up a promising avenue to seek multiscale solutions in three-dimensional, wall bounded flows. Figure 1 : Cross-stream coupling in the tall-box Kolmogorov flow. (a) Cross-coupling terms C_{12} and C_{21} (see equation 3 ) for a turbulent 2\pi\times 4\pi trajectory of total length T=10^{5} (contours; levels evenly spaced logarithmically from 0.4 to 228.0). Also shown are the coupling coefficients for the RPO-laminar tiles (green curves/points) and the RPO-RPO tiles (blue curves). (b, c, d) Examples of weakly coupled, two-way coupled and one-way coupled vorticity fields respectively, corresponding to the magenta symbols in (a). Background filled contours are the explicit coupling terms {\bf u}_{1}\cdot\bm{\nabla}\omega (left) and {\bf u}_{2}\cdot\bm{\nabla}\omega (right) with contours in ranges (a) [-24.2,24.2] , (b) [-17.5,17.5] , and (c) [-25.0,25.0] . Line contours show the vorticity, evenly spaced -11.4<\omega<11.4 , in each half domain which is used to compute the induced velocities \{\mathbf{u}_{i}\} . We consider two-dimensional Kolmogorov flow driven by a monochromatic force in the horizontal, \mathbf{f}^{*}:=\hat{\mathbf{x}}\,\chi^{*}\sin(2\pi ny^{*}/L^{*}) (an asterisk indicates a dimensional variable). The forcing amplitude, \chi^{*} , and fundamental wavenumber associated with the box width, k^{*}:=2\pi/L^{*} , are used to non-dimensionalize the system. The out-of-plane vorticity evolves according to \partial_{t}\omega+\mathbf{u}\cdot\bm{\nabla}\omega+\frac{1}{Re}\Delta\omega-n\cos(ny), (1) where the velocity and vorticity are related by the definition \omega:=(\bm{\nabla}\times\mathbf{u})\cdot\hat{\mathbf{z}} , while the Reynolds number based on our non-dimensionalization is Re:=(1/\nu)\sqrt{\chi^{*}/k^{*3}} . We fix the number of forcing waves per box width at n=4 but contrary to previous studies 4 ; 18 consider ‘tall’ boxes of size (L_{x},L_{y})=(2\pi,4\pi) (so that there are eight forcing waves in the box). The Reynolds number is held fixed at Re=40 throughout this paper. Equation ( 1 ) is equivariant under continuous translations in x , under discrete shift-reflects over half-wavelengths of the forcing in y and under rotation through \pi 4 . We seek relative periodic/quasi-periodic solutions in the 4\pi\times 2\pi box which are constructed from RPOs found in the ‘standard’ 2\pi\times 2\pi geometry. Our expectation that this is possible is due to the fact that the tall box can often be considered as a pair of weakly interacting small-boxes. To quantify the interaction, consider a partitioning of the vorticity in the vertical coordinate, \omega(x,y)=\omega(x,y)\,\mathbf{1}_{[0,2\pi)}(y)+\omega(x,y)\,\mathbf{1}_{[2\pi,4\pi)}(y):=\omega_{1}+\omega_{2} , with matching conditions across the interfaces [\![\omega]\!]=0 and [\![\partial_{y}\omega]\!]=0 . The masked vorticities \omega_{1} and \omega_{2} each evolve according to \partial_{t}\omega_{i}+(\mathbf{u}_{1}+\mathbf{u}_{2})\cdot\bm{\nabla}\omega_{i}=\frac{1}{Re}\Delta\omega_{i}-n\cos(ny), (2) on their respective domains. Here the velocity has been decomposed into components induced by the vorticity on each half domain, with \mathbf{u}_{i}:=(-\hat{\mathbf{x}}\partial_{y}+\hat{\mathbf{y}}\partial_{x})\Delta^{-1}\omega_{i} . We then quantify coupling between the vorticity fields by the action of the induced velocity from one field onto another, defining a coupling coefficient j\to i : c_{ij}:=\|\mathbf{u}_{j}(\omega_{j})\cdot\bm{\nabla}\omega_{i}\|, (3) where \|\bullet\|:=[(8\pi^{2})^{-1}\iint|\bullet|^{2}d^{2}\mathbf{x}]^{1/2} and the definition ( 3 ) allows for examination of ‘self interaction’ when i=j . It is helpful to consider the normalized version of ( 3 ): C_{ij}:=c_{ij}/(c_{11}+c_{22}+c_{12}+c_{21}) with C_{11}+C_{22}+C_{12}+C_{21}=1 by construction. The cross-coupling terms, C_{12} and C_{21} , were computed for snapshots on a long turbulent trajectory (with discrete shifts applied to \omega in y such that the C_{12}+C_{21} was minimized) and a joint probability density function (pdf) of these variables is reported in figure 1 (a). These results imply that the

Read the complete text