WorksPhysics and Astronomy
Elastic turbulence in straight confined geometries
Abstract
Elastic turbulence (ET) in straight geometries has been extensively studied in idealised configurations such as periodic shear and planar channel flow, yet experiments are typically restricted to confined systems like ducts and pipes. While chaotic elastic fluctuations are observed in ducts, their onset in pipes is debated due to geometric constraints. Here we report sustained chaotic dynamics in both geometries and show they correspond to the same ET state identified in idealised setups. Our results establish ET as a geometry-independent attractor and point to a distinct transition mechanism in pipes.
The velocity field is governed by the Navier–Stokes equations together with the incompressibility constraint. The dynamics of the fluid microstructure are described by the Oldroyd-B model [24]. The flow thus evolves according to
where is the fluid density, the solvent viscosity, the viscosity contribution associated with the microstructure, and its relaxation time. These equations are solved numerically using our extensively validated solver Fujin [25] (https://www.oist.jp/research/research-units/cffu/fujin), in square duct and circular pipe geometries. Further details on the numerical methods, grid resolutions, domain sizes, and flow initialisation are provided in the End Matter.
The flow is characterised by the Reynolds number (), quantifying the relative importance of inertial and viscous effects, and the Deborah number (), comparing the micro-structural relaxation time with the flow time scale. These are defined as
where the velocity scale is the mean flow speed (maintained constant throughout simulations) and the length scale is a macroscopic characteristic of the cross-section (the half-edge length for the duct, and the radius for the pipe). All simulations are performed at and , with a fixed viscosity ratio . Although the chosen value of is finite and larger than unity, the resulting elasticity number has been demonstrated to characterise pure ET in planar channels [8] and Taylor-Couette flows [20].
At steady state, both duct and pipe flows exhibit sustained velocity fluctuations characterised by the coexistence of large-scale motions and localised dynamics. The two configurations, however, differ markedly. Similarly to what is reported in channels [8], coherent structures are observed at the duct centre and finer-scale motions at the walls, reminiscent of the centre-mode [9, 11] and of the unsaturated polymer diffusive instability (PDI) [6, 26, 27, 28] (visualisations in the End Matter). Superimposed on these, we nevertheless identify streak-like structures at intermediate distances from the centre, meandering between the corners and the face centres (Figure 1, top), never reported in prior studies to the best of our knowledge and specific to the duct geometry. In the pipe, centre-localised structures akin to the centre-mode [13, 15] are observed along with large-scale motions localised at the wall, evocative of the saturated PDI (Figure 1, bottom, and visualisations in the End Matter). Remarkably, velocity fluctuations in the pipe do not comply with the axial symmetry of the geometry (also visible for the -criterion in the End Matter), hinting towards the attainment of chaotic dynamics rather than a pure and saturated centre-mode state.
Figure 1. Side and back views (left/right) of the streamwise velocity fluctuations iso-surfaces in the duct (top) and pipe (bottom) geometries. Levels are set at in the duct, and in the pipe. The flow goes from left to right in the side views, and enters the page in the back views.
Differences between the two setups extend also to the mean flow (Figure 2). In the duct, the face-centred streamwise velocity profile remains close to the laminar parabola. In the pipe, streamwise-averages of the streamwise velocity at specific time instants do not comply with axial symmetry and attain their maxima away from the centre. The time-and azimuthally-averaged velocity profile, , thus exhibits a plug-like shape, with depleted shear () in the central region relative to laminar flow. This reflects a strong modulation of the mean flow by coherent structures, which are significantly more energetic in the pipe, where fluctuation levels exceed those in the duct by orders of magnitude (Figure 4). Such divergence might appear consistent with the curvature effect introduced previously, becoming relevant once streamlines are curved by the developing centre-mode and not hindered by flat boundaries. Despite the macroscopic differences above, we now demonstrate that the underlying turbulent state is identical in the two systems and corresponds to ET.
Figure 2. Mean profiles of the streamwise velocity for the duct (left) and the pipe (right), normalised with the centreline velocity of the laminar solution, . Laminar profiles are shown as dashed lines for comparison. Profiles are extracted through the cross-section centre, orthogonal to the faces in the duct.

Figure 4. Wavenumber and frequency spectra (top/bottom) of the fluctuating energy at the centreline of their respective geometry. Wavenumber spectra are computed along the streamwise direction. In the bottom panel inset we show extracts from the time-histories of the cross-stream energy , measured with a centreline probe.
Streamwise and wall-normal velocity fluctuations are fully de-correlated (Figure 3) and thus give an effectively null contribution to the Reynolds stresses:
Figure 3. Joint probability density functions of the normalised streamwise () and wall-normal () velocity fluctuations at from the wall in the duct (orange) and the pipe (blue). Normalisation relies on the mean turbulent kinetic energy in the considered geometry, and isolines are drawn at the function maxima.
a defining feature of ET [8]. Chaotic motions can therefore develop decoupled from the mean flow, allowing for both the near-laminar profile of the duct and the plug-like profile of the pipe, which we impute to the more intense viscoelastic stresses developed in that case.
Velocity fluctuations permeate all flow scales and carry the spectral signature of ET: the steep power-law exponent for the spatial decay of the fluctuating energy spectrum (Figure 4, top). As demonstrated in former work [21], spatial and temporal scalings (Figure 4, bottom) differ due to the breakdown of Taylor’s hypothesis. The sustained and developed nature of the ET state attained is confirmed by the time-histories of the kinetic energy associated with cross-stream velocity fluctuations at the centreline (inset of Figure 4, bottom), oscillating about a finite value. In the pipe, these fluctuations necessarily break axial symmetry and therefore cannot be attributed to the centre-mode alone, demonstrating the presence of fully developed chaotic dynamics in addition to it. The time-history reported for the duct covers the whole range of accumulated statistics after initial transients elapsed. Convergence is assessed by continuing the simulation on a coarser grid for further , leading to a slight depletion of due to enhanced diffusivity but otherwise no appreciable variation of the flow state attained. For the pipe, instead, changing the grid (hence the stress diffusivity) appears to alter flow structures and leads to a more pronounced plug-like velocity profile, while turbulent statistics remain unchanged. Reported measurements are thus accumulated on a fine grid over , at the limit of what is numerically feasible with our tools. Full details on the above variations, along with a further pipe simulation extending for on a coarser grid, are discussed in the End Matter.
The ultimate confirmation that the chaotic dynamics described here are those of ET is provided by the scaling of the structure functions of the symmetric three-point velocity increment, , where and angled brackets denote ensemble-averaging. This definition removes leading-order analytic contributions, so that at sub-viscous scales , with predicted for ET [22], in contrast to in inertial turbulence (IT). Our measurements (Figure 5) follow the ET scaling in both duct and pipe. At high orders, the exponent saturation below is consistent with the appearance of intermittency [22, 29]. Finally, the good agreement of our measurements with the established fingerprints of ET verified above confirms the sub-dominant role of inertial effects relative to the elastic ones in the small, finite- regime considered.
Figure 5. Scaling exponents of the th-order structure functions at sub-viscous scales, at the duct and pipe centrelines. Theoretical predictions without intermittency are reported as dashed/dash-dotted lines for ET/IT, respectively.
The results presented here, limited to the rheological model considered, demonstrate that elastic turbulence can be sustained in confined geometries even in the absence of curvature in the base flow, and that the resulting chaotic state coincides with the same ET regime previously identified in tri-periodic [22, 30, 31, 32], planar [7, 8, 26, 33, 34, 35], and free-shear configurations [23, 36, 37]. ET therefore constitutes a unique dynamical attractor independent of flow geometry. This has direct implications for viscoelastic pipe flow. As discussed above, a pipe cannot reproduce the transverse modulation identified in channel flow [12] if a non-zero streamwise velocity at the symmetry element or one of its defining properties, since the pipe’s centreline – unlike the channel’s centre-plane – cannot support such a disturbance away from full axial symmetry. The attainment of ET in pipes thus implies the existence of an alternative route to chaos, possibly mediated by a secondary instability distinct from the one identified in channels, or by a mechanism akin to a bypass transition [38, 39, 40]. Clarifying the nature of this transition remains an open problem.
Despite convergence towards the same chaotic attractor, the flow retains clear signatures of the underlying geometry. In the duct, we observe novel streak-like structures at intermediate locations and unsaturated PDI manifestations at the walls [6, 26]. In the pipe, by contrast, the PDI manifests in its saturated form. Coherent structures are overall more energetic in the pipe, consistent with enhanced mean-flow modulation and the emergence of a time-averaged, plug-like velocity profile. These findings raise questions for future work. First, what mechanism underlies the transition to ET in pipe flow in the absence of the “channel route”? Second, what is the role of coherent structures within ET (e.g., the emerging centre-mode and PDI manifestations), and how do they couple to the background chaotic fluctuations and to the mean flow? Addressing these questions will be essential to develop a unified understanding of elasticity-driven turbulence in complex flows.
Acknowledgments
The research was supported by the Okinawa Institute of Science and Technology Graduate University (OIST) with subsidy funding to M.E.R. from the Cabinet Office, Government of Japan. The authors acknowledge the computer time provided by the Scientific Computing and Data Analysis section of the Core Facilities at OIST, and the computational resources offered by the HPCi System Research Project with grant hp260009.
Data Availability Statement
All information needed to evaluate the conclusions of this Letter is present in the main text and/or the End Matter. Data required to reproduce the figures are available on the website of the Complex Fluids and Flows Unit at OIST(https://www.oist.jp/research/research-units/cffu/publications/publication-data). Further material is available from the authors upon reasonable request.
References
End Matter
Computational details
Our in-house solver Fujin [25] (https://www.oist.jp/research/research-units/cffu/fujin) uses a staggered, uniform Cartesian grid and discretises the governing equations (Eq. (1)-(3)) in space with a second-order central finite-difference scheme. Time advancement is performed using a second-order Adams–Bashforth method, combined with a fractional-step approach [41]. At each time step, the pressure field enforcing incompressibility is obtained via an efficient spectral Poisson solver, while domain decomposition and parallelisation are handled through the library together with the message passing interface (MPI). The evolution equation for the conformation tensor requires particular attention: we adopt a logarithmic formulation [42, 43] to mitigate the well-known high-Deborah-number instability, and employ a high-order weighted essentially non-oscillatory (WENO) scheme [44] for the discretisation of the upper-convected derivative on the left-hand side. This strategy avoids the need for an explicit stress-diffusion term, and thus no additional boundary conditions are required for [45]. For the velocity field, no-slip and no-penetration conditions are imposed at the walls, while periodic boundary conditions are applied in the homogeneous streamwise direction. The curved pipe wall is obtained with an Immersed Boundary Method (IBM) [46].
Denoting by either the half-edge of the square duct cross-section or the pipe radius, both geometries extend in the streamwise direction and are discretised on a fine grid with 512 points streamwise and 1024 points in each cross-stream direction. When a coarser grid is mentioned, 128 points streamwise and 256 points in each cross-stream direction are employed instead. Turbulent statistics reported in the main text are extracted from two orthogonal planes normal to the cross-section, and do not exhibit significant variations with the grid, as here exemplified for the fluctuating energy spectra (Figure 6). The chosen domain size and resolution, in line with previous literature [7, 8], are sufficient to accommodate and resolve the relevant dynamics of the flow.

Figure 6. Wavenumber/frequency spectra of the fluctuating energy from our duct and pipe simulations on a coarser grid (main axes/inset, respectively), at the centreline of the geometry. Wavenumber spectra are computed along the streamwise direction. The same scaling behaviours reported on the finer grid (Figure 4) are achieved, even though the scaling range in the pipe appears shrunk, likely due to the loss of finer-scale motions.
Simulations are initialised from fully developed inertial turbulence fields, after which transients are allowed to elapse until a statistically stationary ET state is reached. All reported measurements are performed in this sustained regime.
Further flow visualisations
We consider volumetric visualisations of the -criterion iso-surfaces (Figure 7). These highlight fine-scale activity at the duct walls, and larger-scale motions at the pipe wall, where the breakdown of axial symmetry is also evident.

Figure 7. Side and back views of iso-surfaces (left/right) in the duct (top) and pipe (bottom) geometries. is defined as the second invariant of the velocity gradient [47], and its levels are at in the duct, and in pipe. The flow goes left to right in the side views, and enters the page in the back views.
Finally, slicing the duct along its centreline (Figure 8, top) reveals wall-localised structures reminiscent of the unsaturated PDI [6, 26], together with large-scale coherent fluctuations at the middle suggestive of the centre-mode [13, 15]. Coherent centreline motions are also clearly observed in the pipe (Figure 8, bottom), along with large-scale wall motions evocative of the saturated PDI.

Figure 8. Longitudinal slices along the centreline of the wall-normal () and transverse () velocity fluctuations. We adopt a blue to red colormap ranging linearly between for in the duct and between for both in the pipe. The flow goes from left to right.
Increasing diffusivity in the pipe
Running our pipe simulation on a coarser grid (i.e., increasing the stress diffusivity) leads to a flow state characterised by the same turbulent statistics of the fine one (signature of ET), but a pronounced plug-like velocity profile and modified flow structures. To better characterise this behaviour, we observe the time-history of the first normal-stress difference () at the pipe centreline throughout the whole extent of the coarser simulation (Figure 9). After the initial transient, the flow momentarily settles into a state characterised by the coexistence of centre- and wall-localised structures: this is the final, sustained state achieved by our simulation on the fine grid. On the coarser grid, instead, centreline motions eventually die out while wall-motions intensify, leading to a final state characterised by significantly higher stretching of the microstructure.

Figure 9. Time-history of the first normal-stress difference from our simulation on a coarser grid, transiently visiting the same state achieved on the fine one before reaching a final state characterised by dominant wall-motions and enhanced stretching of the microstructure.
References
- [6]M. Beneitez, J. Page, and R. R. Kerswell, Polymer diffusive instability leading to elastic turbulence in plane Couette flow, Phys. Rev. Fluids 8, L101901 (2023).DOI
- [7]M. Lellep, M. Linkmann, and A. Morozov, Purely elastic turbulence in pressure-driven channel flows, PNAS 121, e2318851121 (2024).DOI
- [8]G. Foggi Rota, C. Amor, S. Le Clainche, and M. E. Rosti, Unified view of elastic and elasto-inertial turbulence in channel flows at low and moderate Reynolds numbers, Phys. Rev. Fluids 9, L122602 (2024).
- [9]J. Page, Y. Dubief, and R. R. Kerswell, Exact Traveling Wave Solutions in Viscoelastic Channel Flow, Phys. Rev. Lett. 125, 154501 (2020).DOI
- [11]G. Buza, M. Beneitez, J. Page, and R. R. Kerswell, Finite-amplitude elastic waves in viscoelastic channel flow from large to zero Reynolds number, J. Fluid Mech. 951, A3 (2022).DOI
- [12]M. Lellep, M. Linkmann, and A. Morozov, Linear stability analysis of purely elastic traveling-wave solutions in pressure-driven channel flows, J. Fluid Mech. 959, R1 (2023).DOI
- [13]P. Garg, I. Chaudhary, M. Khalid, V. Shankar, and G. Subramanian, Viscoelastic Pipe Flow is Linearly Unstable, Phys. Rev. Lett. 121, 024502 (2018).DOI
- [15]I. Chaudhary, P. Garg, G. Subramanian, and V. Shankar, Linear instability of viscoelastic pipe flow, J. Fluid Mech. 908, A11 (2021).DOI
- [20]Y. B. Zhang, L. Li, Y. Fan, J. Su, H. D. Xi, and C. Sun, Experimental evidence for the continuous transition between elastic and elasto-inertial turbulence, PNAS 122, e2505007122 (2025).
- [21]G. Foggi Rota, R. K. Singh, A. Chiarini, C. Amor, G. Soligo, D. Mitra, and M. E. Rosti, The broken link between space and time in elastic turbulence, Int. J. Multiph. Flow 197, 105630 (2026).DOI
- [22]R. K. Singh, P. Perlekar, D. Mitra, and M. E. Rosti, Intermittency in the not-so-smooth elastic turbulence, Nat. Commun. 15, 4070 (2024).DOI
- [23]G. Soligo and M. E. Rosti, Non-Newtonian turbulent jets at low-Reynolds number, Int. J. Multiph. Flow 167, 104546 (2023).
- [24]J. G. Oldroyd, On the formulation of rheological equations of state, Proc. R. Soc. Lond. A 200, 523 (1950).DOI
- [25]M. E. Rosti, Simulating laminar and turbulent multiphase flows with Fujin, Fluid Dyn. Res. 58, 021401 (2026).DOI
- [26]M. Beneitez, J. Page, Y. Dubief, and R. R. Kerswell, Transition route to elastic and elasto-inertial turbulence in polymer channel flows, Phys. Rev. Fluids 9, 123302 (2024).arxiv.org/abs/2408.11508
- [27]P. S. D. Surya Phani Tej, P. Kumar Mohanty, and V. Shankar, Hoop stress and polymer diffusive instabilities in viscoelastic Taylor–Couette flow, Proc. R. Soc. A 481, 20250455 (2025).
- [28]S. Pandey and V. Shankar, Effect of anisotropic mobility on the diffusive instability in viscoelastic shear flows, Phys. Rev. Fluids 10, 053301 (2025).
- [29]P. Garg and M. E. Rosti, Elastic Turbulence Hides in the Small Scales of Inertial Polymeric Turbulence, Phys. Rev. Lett. 135, 074001 (2025).arxiv.org/abs/2507.16789
- [30]S. Berti and G. Boffetta, Elastic waves and transition to elastic turbulence in a two-dimensional viscoelastic Kolmogorov flow, Phys. Rev. E 82, 036314 (2010).DOI
- [31]V. Dzanic, C. S. From, and E. Sauret, The effect of periodicity in the elastic turbulence regime, J. Fluid Mech. 937, A31 (2022).DOI
- [32]T. Lewy and R. R. Kerswell, Revisiting two-dimensional viscoelastic Kolmogorov flow: A centre-mode-driven transition, Journal of Fluid Mechanics 1007, A55 (2025).
- [33]B. Qin and P. E. Arratia, Characterizing elastic turbulence in channel flows at low Reynolds number, Phys. Rev. Fluids 2, 083302 (2017).arxiv.org/abs/1609.08532
- [34]A. Varshney and V. Steinberg, Elastic Alfven waves in elastic turbulence, Nat. Commun. 10, 652 (2019).
- [35]M. Khalid, V. Shankar, and G. Subramanian, Continuous Pathway between the Elasto-Inertial and Elastic Turbulent States in Viscoelastic Channel Flow, Phys. Rev. Lett. 127, 134502 (2021).DOI
- [36]S. Yamani, B. Keshavarz, Y. Raj, T. A. Zaki, G. H. McKinley, and I. Bischofberger, Spectral Universality of Elastoinertial Turbulence, Phys. Rev. Lett. 127, 074501 (2021).DOI
- [37]S. Yamani, Y. Raj, T. A. Zaki, G. H. McKinley, and I. Bischofberger, Spatiotemporal signatures of elastoinertial turbulence in viscoelastic planar jets, Phys. Rev. Fluids 8, 064610 (2023).DOI
- [38]N. Itoh, Nonlinear stability of parallel flows with subcritical Reynolds numbers. Part 2. Stability of pipe Poiseuille flow to finite axisymmetric disturbances, Journal of Fluid Mechanics 82, 469 (1977).DOI
- [39]R. R. Kerswill, Recent progress in understanding the transition to turbulence in a pipe, Nonlinearity 18, R17 (2005).
- [40]M. Avila, D. Barkley, and B. Hof, Transition to Turbulence in Pipe Flow, Annual Review of Fluid Mechanics 55, 575 (2023).DOI
- [41]J. Kim and P. Moin, Application of a fractional-step method to incompressible Navier-Stokes equations, J. Comput. Phys. 59, 308 (1985).DOI
- [42]R. Fattal and R. Kupferman, Time-dependent simulation of viscoelastic flows at high Weissenberg number using the log-conformation representation, J. Nonnewton. Fluid. Mech. 126, 23 (2005).DOI
- [43]F. De Vita, M. E. Rosti, D. Izbassarov, L. Duffo, O. Tammisola, S. Hormozi, and L. Brandt, Elastoviscoplastic flows in porous media, J. Nonnewton. Fluid. Mech. 258, 10 (2018).
- [44]K. Sugiyama, S. Ii, S. Takeuchi, S. Takagi, and Y. Matsumoto, A full Eulerian finite difference approach for solving fluid–structure coupling problems, J. Comput. Phys. 230, 596 (2011).DOI
- [45]M. Beneitez, J. Page, Y. Dubief, and R. R. Kerswell, Multistability of elasto-inertial two-dimensional channel flow, J. Fluid Mech. 981, A30 (2024).arxiv.org/abs/2308.11554
- [46]N. Hori, M. E. Rosti, and S. Takagi, An Eulerian-based immersed boundary method for particle suspensions with implicit lubrication model, Computers & Fluids 236, 105278 (2022).arxiv.org/abs/2201.03708
- [47]J. Jeong and F. Hussain, On the identification of a vortex, J. Fluid Mech. 285, 69 (1995).



