betaFeedback

WorksPhysics and Astronomy

Variational determination of minimal absorbing zones in incompressible shear flows

Abstract

The dynamical analysis of shear flows remains challenging, as turbulence generation and evolution are not fully understood. Here, a lesser-explored feature of incompressible shear flows-the absorbing zone-is investigated. This region in the infinite dimensional state space is shown to act as an attractor for all trajectories: any solution initialised outside eventually enters and remains inside. Consequently, the absorbing zone must contain all possible attractors, both chaotic and non-chaotic. Existence is established through the Reynolds-Orr identity, which indicates that nonlinear terms do not directly influence the temporal evolution of kinetic energy. The zone is constructed around a so-called shift flow, and multiple such regions may exist, even when the laminar state is linearly unstable. Gradient based optimisation is employed to identify the absorbing zone of minimal radius. Assuming a chaotic trajectory explores state space randomly, it is hypothesised that the centroid of this minimal zone approximates the turbulent mean flow. This central state is computed for plane Poiseuille and Couette flows and compared with established turbulent mean profiles. Although the proposed hypothesis is found to be an oversimplification-yielding profiles that qualitatively resemble but do not quantitatively reproduce the turbulent means-the methodology provides novel insights into shear flow dynamics. Furthermore, it offers a promising foundation for refining estimates of global stability thresholds. With continued development, this framework may facilitate the direct computation of turbulent mean states without reliance on empirical turbulence models or time-dependent numerical simulations.

Introduction

The Navier–Stokes equations, cornerstone of fluid dynamics, represent a quintessential example of a non-linear system capable of exhibiting remarkably complex behaviour [4, 7], including the transition from laminar to turbulent flow and fully developed turbulence itself. Understanding the intricate phase space structure of such systems is paramount for prediction and control. Over the past decades, a dynamical systems approach has provided profound insights, revealing that the seemingly chaotic nature of turbulence may be underpinned by a framework of special invariant solutions of the Navier–Stokes equations. Among these, edge states have emerged as crucial entities in comprehending subcritical transition phenomena, prevalent in many shear flows. These are unstable solutions, whose stable manifolds form the boundary (the ”edge”) separating initial conditions that relaminarise from those that transition to turbulence [6, 2]. Identifying and characterising these states, which can range from steady solutions to time-periodic or even chaotic saddles, provides a pathway to understanding the minimal seeds [16, 5] for turbulence and the geometry of the laminar-turbulent boundary [11].

Further into the turbulent regime, the chaotic dynamics are thought to be organised around a skeleton of unstable periodic orbits (UPOs). Although trajectories in a turbulent flow are not themselves periodic, they may closely shadow various UPOs for finite times. These UPOs are exact, albeit unstable, solutions of the Navier-Stokes equations, and their properties are believed to encode essential statistical and structural features of turbulent flows [9]. The systematic computation and analysis of UPOs thus offer a promising avenue for a more fundamental description of turbulence, moving beyond purely statistical characterisations towards a dynamical understanding. In the last years, the computational efficiency of UPOs has been significantly improved [1].

A rarely studied feature of shear-flow dynamics is the absorbing zone introduced by Olivier Dauchot & Paul Manneville [13]. Using the Reynolds–Orr identity, one can show that there exists a region in the infinite-dimensional state space such that any trajectory initiated outside this region eventually enters it and remains there. Consequently, this zone contains all possible chaotic and non-chaotic attractors, periodic orbits, edge state solutions. The main objective of this study is to characterise and analyse this region in the case of fundamental flow configurations. In the simplest setting the absorbing zone is a hypersphere, specified by its centre and radius. The centre is termed the shift flow, while the radius is measured by the perturbation kinetic energy. A schematic illustration of this concept is provided in Figure 1.

The sketch of the minimal absorbing zone around the optimal shift flow

Figure 1. The sketch of the minimal absorbing zone around the optimal shift flow

Assuming that the chaotic solution of a turbulent flow exhibits a random wandering behaviour in the state space, the time-averaged (mean) flow can be approximated by the centroid of the smallest absorbing zone. Below the energy-stability threshold, denoted ReERe_E, the centre of the minimal absorbing zone coincides with the laminar solution and its radius collapses to zero; no attractor other than the laminar state can therefore exist. However, as the Reynolds number exceeds ReERe_E, the size of the minimal absorbing zone increases, and additional attractors may emerge. Identifying minimal absorbing zones thus offers new insights into shear-flow behaviour. By employing variational equations, the shift flow can be optimised to minimise the size of the absorbing zone.

Variational methods have proved highly successful across physics, and fluid mechanics is no exception. Recently, Sanders et al. [17] presented a Hamiltonian formulation of the Navier–Stokes equations and provided a comprehensive literature review. Another variational approach focuses on maximising the rate of turbulent energy dissipation. The principle was described in detail by Kerswell [10] and later solved for plane Couette flow [14] and pipe flow [15]. Although this approach identifies the maximum dissipation—a key flow property—the resulting velocity profiles depart markedly from those observed in fully developed turbulence.

In the present work we pursue an alternative route to predicting turbulent mean flows without time-dependent simulation. Conceptually akin to a Reynolds-averaged Navier–Stokes (RANS) model, our formulation is derived directly from the Navier–Stokes equations and requires no empirical closure.

First, the governing equations are derived in § 2. Next, a solution procedure for one-dimensional mean flows, based on Chebyshev collocation, is described in Section 3. The method is then validated on two canonical configurations—Couette and Poiseuille flow—and the results are reported in § 4. The minimal absorbing-zone radius is quantified as a function of Reynolds number, and the corresponding optimal shift flows are compared with mean-flow profiles from direct numerical simulations available in the literature. Finally, the main conclusions are summarised in § 5.

Theory

The non-dimensional Navier–Stokes equations, written with the Einstein summation convention, are

∂ui∂t=−uj∂ui∂xj−∂p∂xi+1Re∂2ui∂xj∂xj+fi.(1)\frac{\partial u_i}{\partial t}=-u_j\frac{\partial u_i}{\partial x_j}-\frac{\partial p}{\partial x_i}+\frac{1}{Re}\frac{\partial^2u_i}{\partial x_j\partial x_j}+f_i. \tag*{(1)}
∂ui∂xi=0.(2)\frac{\partial u_i}{\partial x_i}=0. \tag*{(2)}

where uiu_i and pp are the velocity and the pressure fields, fif_i is an arbitrary momentum source. Re=L0U0/νRe=L_0U_0/\nu is the Reynolds number based on the characteristic length L0L_0, velocity U0U_0 and kinematic viscosity ν\nu.

Throughout this study the boundary conditions and momentum source are assumed to be time-independent. Accordingly, a stationary shift flow UiU_i and the associated pressure PP are introduced, both of which satisfy the prescribed boundary conditions and uiu_i becomes the remaining time-dependent fluctuations around this shift flow. Such a decomposition resembles stability analysis, where UiU_i usually denotes the base flow being time independent and fulfilling the governing equation. However, the shift flow in this analysis does not necessarily fulfil the governing equation, it is merely a state-space point complying with the boundary conditions. After the decomposition, the following equation can be obtained

∂ui∂t=−Uj∂Ui∂xj−Uj∂ui∂xj−uj∂Ui∂xj−uj∂ui∂xj−∂p∂xi+1Re∂2Ui∂xj∂xj+1Re∂2ui∂xj∂xj+fi.(3)\frac{\partial u_i}{\partial t}=-U_j\frac{\partial U_i}{\partial x_j}-U_j\frac{\partial u_i}{\partial x_j}-u_j\frac{\partial U_i}{\partial x_j}-u_j\frac{\partial u_i}{\partial x_j}-\frac{\partial p}{\partial x_i}+\frac{1}{Re}\frac{\partial^2U_i}{\partial x_j\partial x_j}+\frac{1}{Re}\frac{\partial^2u_i}{\partial x_j\partial x_j}+f_i. \tag*{(3)}

subject to the incompressibility constraint ∂ui∂xi=0\frac{\partial u_i}{\partial x_i}=0.

The non-dimensional kinetic energy of the fluctuations is defined as

e=12∫Vuiui dV.(4)e=\frac{1}{2}\int_{\mathcal{V}}u_i u_i\,\mathrm{d}\mathcal{V}. \tag*{(4)}

where V\mathcal{V} denotes the flow domain. The derivative of kinetic energy with respect to time is:

dedt=∫V(−uiUj∂Ui∂xj−uiuj∂Ui∂xj+ui1Re∂2Ui∂xj∂xj+ui1Re∂2ui∂xj∂xj+uifi) dV.(5)\frac{\mathrm{d}e}{\mathrm{d}t}=\int_{\mathcal{V}}\left(-u_iU_j\frac{\partial U_i}{\partial x_j}-u_i u_j\frac{\partial U_i}{\partial x_j}+u_i\frac{1}{Re}\frac{\partial^2U_i}{\partial x_j\partial x_j}+u_i\frac{1}{Re}\frac{\partial^2u_i}{\partial x_j\partial x_j}+u_i f_i\right)\,\mathrm{d}\mathcal{V}. \tag*{(5)}

here certain terms have been eliminated by applying the Gauss divergence theorem and assuming periodic, far-field vanishing, or wall-bounded velocity fluctuations. These steps are analogous to those used in the proof of the well-known Reynolds-Orr identity. A significant consequence of this simplification is that the non-linear terms (uj∂ui/∂xju_j \partial u_i/\partial x_j) do not affect the time derivative of the kinetic energy. This observation is also a contributing factor to the Sommerfeld paradox. Furthermore, this simplification substantially facilitates the proof of the existence of an absorbing zone.

Such a zone attracts all of the trajectories, meaning that the flow state can be outside of the zone temporarily, as it evolves in time it will eventually enter the zone and be unable to escape. Consequently, the zone contains all possible attractors. To prove the existence of such an absorbing zone, it is necessary to demonstrate that outside this zone, the kinetic energy decreases for any given state. In this context, the kinetic energy is similar to a Lyapunov functional.

Infinite-amplitude perturbations

The maximum possible growth of kinetic energy associated with perturbations of arbitrarily large amplitude about the shift flow is examined first. In this limit, it is convenient to investigate the temporal growth rate of the kinetic energy defined as

μe=1ededt.(6)\mu_e = \frac{1}{e}\frac{\mathrm{d}e}{\mathrm{d}t}. \tag*{(6)}

It can be shown that, as the perturbation energy tends to infinity, the terms in (5) that scale linearly with uiu_i vanish, yielding

μe→μe,∞=1e∫V(−uiuj∂Ui∂xj+ui1Re∂2ui∂xj∂xj) dVase→∞.(7)\mu_e \to\mu_{e,\infty} = \frac{1}{e}\int_{\mathcal{V}}\left(-u_i u_j \frac{\partial U_i}{\partial x_j} + u_i \frac{1}{Re}\frac{\partial^2 u_i}{\partial x_j\partial x_j}\right)\,\mathrm{d}\mathcal{V}\quad\text{as}\quad e\to\infty. \tag*{(7)}

For such states the growth rate is independent of the perturbation amplitude; consequently, the maximum achievable growth depends solely on the spatial structure of the velocity field. The maximum of μe,∞\mu_{e,\infty} is obtained by solving the associated Euler–Lagrange equations,

−uj(∂Ui∂xj+∂Uj∂xi)+2Re∂2ui∂xj∂xj−∂q∂xi=μe,∞ui,(8)-u_j\left(\frac{\partial U_i}{\partial x_j}+\frac{\partial U_j}{\partial x_i}\right)+\frac{2}{Re}\frac{\partial^2u_i}{\partial x_j\partial x_j}-\frac{\partial q}{\partial x_i}=\mu_{e,\infty}u_i, \tag*{(8)}

where the space-dependent field qq acts as a Lagrange multiplier enforcing the divergence-free constraint on uiu_i.

Equations (8) and (2) form an eigenvalue problem in which μe,∞\mu_{e,\infty} is the eigenvalue. The largest eigenvalue represents the maximum possible growth rate, and the corresponding eigenfunction u~i\tilde{u}_i is referred to as the critical infinite perturbation. The existence of an absorbing zone requires this eigenvalue to be negative, indicating that trajectories initialised sufficiently far from the base flow are attracted towards UiU_i. As a trajectory approaches UiU_i, its energy decreases; therefore, the linear terms neglected for large perturbations become significant, and the maximum possible growth rate increases.

Finite-amplitude perturbations

The critical energy level at which the maximal possible growth rate of the perturbation energy first becomes zero when the energy is decreased from infinity is now determined. This threshold defines the boundary of the absorbing zone. Outside this zone every trajectory must converge toward it because the growth rate is negative there. The radius of the absorbing zone is therefore obtained by evaluating the maximum of de/dt\mathrm{d}e/\mathrm{d}t at successive energy levels while the energy is reduced and by identifying the critical level at which this maximum changes sign. The constrained maximum is determined by another Euler–Lagrange equation. The admissible velocity fields are required to be divergence-free and to possess a prescribed energy level β\beta:

β−12∫Vuiui dV=0.(9)\beta-\frac{1}{2}\int_{\mathcal{V}}u_i u_i\,\mathrm{d}V=0. \tag*{(9)}

This constraint is enforced through the Lagrange multiplier μL\mu_L, while the divergence-free condition is imposed by the field qq as in the previous subsection. The resulting Euler–Lagrange equations read

−uj(∂Ui∂xj+∂Uj∂xi)+2Re∂2ui∂xj∂xj−∂q∂xi−μLui=Uj∂Ui∂xj−1Re∂2Ui∂xj∂xj−fi,(10)-u_j\left(\frac{\partial U_i}{\partial x_j}+\frac{\partial U_j}{\partial x_i}\right)+\frac{2}{Re}\frac{\partial^2u_i}{\partial x_j\partial x_j}-\frac{\partial q}{\partial x_i}-\mu_Lu_i=U_j\frac{\partial U_i}{\partial x_j}-\frac{1}{Re}\frac{\partial^2U_i}{\partial x_j\partial x_j}-f_i, \tag*{(10)}

Equations (2), (9), and (10) must be solved simultaneously to obtain candidate critical perturbations. The true critical state is the one whose energy β\beta is maximal and for which the time derivative of the kinetic energy vanishes. Multiple solutions may exist for a given energy level because equation (9) is non-linear.

For practical implementation it is convenient to treat the Lagrange multiplier μL\mu_L as a free parameter and to solve only the linear equations (2) and (10). The energy level β\beta then emerges as part of the solution instead of being prescribed a priori. Among the resulting states, the one satisfying de/dt=0\mathrm{d}e/\mathrm{d}t=0 and possessing the largest β\beta is selected.

Figure 2 illustrates this procedure by plotting the energy level of the solutions and the corresponding time derivative of the kinetic energy as functions of μL\mu_L. In the example shown, three values of μL\mu_L yield de/dt=0\mathrm{d}e/\mathrm{d}t=0. The largest energy level belongs to the largest μL\mu_L among these values; this β\beta determines the size of the absorbing zone eAZe_{\mathrm{AZ}} around the base flow, and the associated velocity field is termed the critical finite perturbation. The other two solutions correspond to local extrema only.

Figure 2. Time derivative of the kinetic energy (a,b) and energy level (c,d) of the solutions of (2), (9), and (10) as functions of the Lagrange multiplier μL\mu_L. Solutions for which de/dt=0\mathrm{d}e/\mathrm{d}t = 0 are marked by purple crosses. Panels (b) and (d) provide zoomed views of panels (a) and (c), respectively, focusing on the neighbourhood of the critical value μL=0.024\mu_L = 0.024. The results correspond to Poiseuille flow at Re=200Re = 200 with the optimal shift flow.

It is worth mentioning that the eigenvalues μe,∞\mu_{e,\infty} obtained from (8) and (2) are also eigenvalues of the operator that appears on the left-hand side of (10) together with (2). If μL\mu_L is set equal to one of these eigenvalues, the finite-energy problem (10) has no solution, whereas the energy level of the solution diverges as μL\mu_L approaches an eigenvalue. In figure 2 the eigenvalues are indicated by vertical dashed red lines, and the associated singularities are clearly visible.

The procedure for computing the absorbing zone around a given base flow UiU_i is summarised below.

  1. Solve (8) and (2) to obtain the largest eigenvalue μe,∞\mu_{e,\infty}. If μe,∞>0\mu_{e,\infty}>0, the absorbing zone does not exist.

  2. If μe,∞<0\mu_{e,\infty}<0, solve (10) and (2) for various values of μL\mu_L and adjust μL\mu_L until de/dt=0\mathrm{d}e/\mathrm{d}t=0 meaning that

∫V(−uiUj∂Ui∂xj−uiuj∂Ui∂xj+1Reui∂2Ui∂xj∂xj+1Reui∂2ui∂xj∂xj+uifi) dV=0.(11)\int_{\mathcal{V}}\left(-u_iU_j\frac{\partial U_i}{\partial x_j}-u_i u_j\frac{\partial U_i}{\partial x_j}+\frac{1}{Re}u_i\frac{\partial^2U_i}{\partial x_j\partial x_j}+\frac{1}{Re}u_i\frac{\partial^2u_i}{\partial x_j\partial x_j}+u_i f_i\right)\,\mathrm{d}\mathcal{V}=0. \tag*{(11)}
  1. Among the (possibly several) solutions that satisfy (11), select the one with the highest energy; this energy is denoted by eAZe_{\mathrm{AZ}}.

Minimising the absorbing zone

For a given flow configuration several admissible shift flows UiU_i may exist, each of which possesses its own absorbing zone. According to the present theory, the shift flow that generates the smallest absorbing zone must be selected, because it is expected to provide the best approximation to the turbulent mean flow. The selection can be formulated as the optimisation problem

eAZ,opt=min⁡UieAZ[Ui](12)e_{AZ,\mathrm{opt}} = \min_{U_i} e_{AZ}[U_i] \tag*{(12)}

The functional derivative of eAZe_{AZ} with respect to UiU_i cannot be written in explicit form; therefore, the corresponding Euler-Lagrange equation is not solved. Instead, the direct optimisation of eAZe_{AZ} is proposed using gradient-based methods.

The convergence of such algorithms can be accelerated by exploiting the sensitivity of eAZe_{AZ} to variations of the shift flow,

δeAZδUi=ujδujδUi(13)\frac{\delta e_{AZ}}{\delta U_i} = u_j \frac{\delta u_j}{\delta U_i} \tag*{(13)}

where δujδUi\frac{\delta u_j}{\delta U_i} is the sensitivity of the critical finite perturbation with respect to the shift flow modification. It is obtained by perturbing equations (10), (2) and (11) leading to the following equation system.

\begin{bmatrix} A_{11} & -\left(\frac{\partial U_1}{\partial x_2}+\frac{\partial U_2}{\partial x_1}\right) & -\left(\frac{\partial U_1}{\partial x_3}+\frac{\partial U_3}{\partial x_1}\right) & -\frac{\partial}{\partial x_1} & -u_1 \\ -\left(\frac{\partial U_2}{\partial x_1}+\frac{\partial U_1}{\partial x_2}\right) & A_{22} & -\left(\frac{\partial U_2}{\partial x_3}+\frac{\partial U_3}{\partial x_2}\right) & -\frac{\partial}{\partial x_2} & -u_2 \\ -\left(\frac{\partial U_3}{\partial x_1}+\frac{\partial U_1}{\partial x_3}\right) & -\left(\frac{\partial U_3}{\partial x_2}+\frac{\partial U_2}{\partial x_3}\right) & A_{33} & -\frac{\partial}{\partial x_3} & -u_3 \\ \frac{\partial}{\partial x_1} & \frac{\partial}{\partial x_2} & \frac{\partial}{\partial x_3} & 0 & 0 \\ A_{51} & A_{52} & A_{53} & 0 & 0 \end{bmatrix} \begin{bmatrix} \delta u_1 \\ \delta u_2 \\ \delta u_3 \\ \delta q \\ \delta\mu_L \end{bmatrix} = \begin{bmatrix} B_{11} & u_2\frac{\partial}{\partial x_1}+\frac{\partial U_1}{\partial x_2} & u_3\frac{\partial}{\partial x_1}+\frac{\partial U_1}{\partial x_3} \\ u_1\frac{\partial}{\partial x_2}+\frac{\partial U_2}{\partial x_1} & B_{22} & u_3\frac{\partial}{\partial x_2}+\frac{\partial U_2}{\partial x_3} \\ u_1\frac{\partial}{\partial x_3}+\frac{\partial U_3}{\partial x_1} & u_2\frac{\partial}{\partial x_3}+\frac{\partial U_3}{\partial x_2} & B_{33} \\ 0 & 0 & 0 \\ B_{51} & B_{52} & B_{53} \end{bmatrix} \begin{bmatrix} \delta U_1 \\ \delta U_2 \\ \delta U_3 \end{bmatrix}. \tag*{(14)}\end{bmatrix}

The operators appearing above are defined by

Aii=−2∂Ui∂xi+2Re∂∂xj∂xj−μL(15)A_{ii}=-2\frac{\partial U_i}{\partial x_i}+\frac{2}{Re}\frac{\partial}{\partial x_j\partial x_j}-\mu_L \tag*{(15)}
A5i=−Uj∂Ui∂xj−uj(∂Ui∂xj+∂Uj∂xi)+1Re∂2Ui∂xj∂xj+2Re∂2ui∂xj∂xj+fi(16)A_{5i}=-U_j\frac{\partial U_i}{\partial x_j}-u_j\left(\frac{\partial U_i}{\partial x_j}+\frac{\partial U_j}{\partial x_i}\right)+\frac{1}{Re}\frac{\partial^2 U_i}{\partial x_j\partial x_j}+\frac{2}{Re}\frac{\partial^2 u_i}{\partial x_j\partial x_j}+f_i \tag*{(16)}
B11=uj∂∂xj+u1∂∂x1+∂U1∂x1+Uj∂∂xj−1Re∂2∂xj∂xj(17)B_{11}=u_j\frac{\partial}{\partial x_j}+u_1\frac{\partial}{\partial x_1}+\frac{\partial U_1}{\partial x_1}+U_j\frac{\partial}{\partial x_j}-\frac{1}{Re}\frac{\partial^2}{\partial x_j\partial x_j} \tag*{(17)}
B22=uj∂∂xj+u2∂∂x2+∂U2∂x2+Uj∂∂xj−1Re∂2∂xj∂xj(18)B_{22}=u_j\frac{\partial}{\partial x_j}+u_2\frac{\partial}{\partial x_2}+\frac{\partial U_2}{\partial x_2}+U_j\frac{\partial}{\partial x_j}-\frac{1}{Re}\frac{\partial^2}{\partial x_j\partial x_j} \tag*{(18)}
B33=uj∂∂xj+u3∂∂x3+∂U3∂x3+Uj∂∂xj−1Re∂2∂xj∂xj(19)B_{33}=u_j\frac{\partial}{\partial x_j}+u_3\frac{\partial}{\partial x_3}+\frac{\partial U_3}{\partial x_3}+U_j\frac{\partial}{\partial x_j}-\frac{1}{Re}\frac{\partial^2}{\partial x_j\partial x_j} \tag*{(19)}
B5i=−uj∂Uj∂xi−Uj∂ui∂xj+uj∂ui∂xj−1Re∂2ui∂xj∂xj.(20)B_{5i}=-u_j\frac{\partial U_j}{\partial x_i}-U_j\frac{\partial u_i}{\partial x_j}+u_j\frac{\partial u_i}{\partial x_j}-\frac{1}{Re}\frac{\partial^2u_i}{\partial x_j\partial x_j}. \tag*{(20)}

Calculating δuj/δUi\delta u_j/\delta U_i requires inverting the operator on the left-hand side of equation (14), a task that is computationally infeasible in most cases. However, this difficulty can be overcome by directly evaluating equation (13), where the computation can be rearranged as solving a linear equation.

Furthermore, because the optimisation is constrained by μe,∞<0\mu_{e,\infty}<0, the corresponding sensitivity can be used to accelerate convergence. The variation of the leading eigenvalue is

δμe,∞δUi=2u~j∂u~i∂xj(21)\frac{\delta\mu_{e,\infty}}{\delta U_i}=2\widetilde{u}_j\frac{\partial\widetilde{u}_i}{\partial x_j} \tag*{(21)}

where u~i\widetilde{u}_i is the eigenvector associated with the largest eigenvalue, namely the critical infinite perturbation. The fact that the eigenvectors are real-valued—owing to the symmetry of the linear operator in the eigenvalue problem (8) and (2)—was invoked in the derivation of this expression.

Using these gradients, the shift flow can be optimised to minimise the absorbing zone.

Solution method for one-dimensional shift flow

The optimisation procedure is first applied to plane Couette and plane Poiseuille flows, where the base and optimal shift flows reduce to one-dimensional wall-normal profiles. Throughout, x1x_1, x2x_2, and x3x_3 denote the streamwise, wall-normal, and spanwise coordinates, respectively; the shift flows therefore depend solely on x2x_2.

In this case, the eigenvalue problem, (2) and (8) can be simplified by assuming a complex wave form solution

ui(x1,x2,x3)=u^i(x2)ei(k1x1+k3x3)(22)u_i(x_1,x_2,x_3)=\hat{u}_i(x_2)e^{\mathrm{i}(k_1x_1+k_3x_3)} \tag*{(22)}
p(x1,x2,x3)=p^(x2)ei(k1x1+k3x3).(23)p(x_1,x_2,x_3)=\hat{p}(x_2)e^{\mathrm{i}(k_1x_1+k_3x_3)} . \tag*{(23)}

with streamwise and spanwise wavenumbers k1k_1 and k3k_3.

The substitution yields

\begin{bmatrix} \frac{2}{Re}L-\frac{\mathrm{d}U_1}{\mathrm{d}x_2} & 0 & -\mathrm{i}k_1 & 0 \\ -\frac{\mathrm{d}U_1}{\mathrm{d}x_2} & \frac{2}{Re}L & 0 & -D \\ 0 & 0 & \frac{2}{Re}L & -\mathrm{i}k_3 \\ \mathrm{i}k_1 & D & \mathrm{i}k_3 & 0 \end{bmatrix} \begin{bmatrix} \hat{u}_1\\ \hat{u}_2\\ \hat{u}_3\\ \hat{q}_m \end{bmatrix} = \tilde{\mu}_{e,\infty} \begin{bmatrix} I & 0 & 0 & 0 \\ 0 & I & 0 & 0 \\ 0 & 0 & I & 0 \\ 0 & 0 & 0 & 0 \end{bmatrix} \begin{bmatrix} \hat{u}_1\\ \hat{u}_2\\ \hat{u}_3\\ \hat{q}_m \end{bmatrix}. \tag*{(24)}\end{bmatrix}

where L=D2−(k12+k32)L=D^2-(k_1^2+k_3^2) is the Laplace operator and D=ddx2D=\frac{\mathrm{d}}{\mathrm{d}x_2} is the differential operator with respect to x2x_2, II is the identity operator.

The true critical eigenvalue, μe,∞\mu_{e,\infty} is the largest one among all possible wavenumber pairs:

μe,∞=max⁡k1,k3μ~e,∞.(25)\mu_{e,\infty}=\max_{k_1,k_3}\tilde{\mu}_{e,\infty}. \tag*{(25)}

The maximum is consistently found to occur at k1=0k_1=0. This finding is analogous to the behaviour of the laminar base flow profile in Couette flow, as proven by Joseph & Carmi [8].

The equation (10), the condition for maximal growth in the case of finite perturbations, simplifies similarly. It is also a linear equation; therefore, the complex Fourier transformation can be applied. Because the right-hand side contains only non-oscillatory terms (k1=k3=0k_1=k_3=0), the left-hand side forces all oscillatory u^i\hat{u}_i to vanish, meaning that the critical finite perturbation field depends only on x2x_2. The continuity equation then implies u2=u3=0u_2=u_3=0, so that

2Re∂2u1∂x2∂x2+μLu1=−1Re∂2U1∂x2∂x2−f1(26)\frac{2}{Re}\frac{\partial^2u_1}{\partial x_2\partial x_2}+\mu_Lu_1=-\frac{1}{Re}\frac{\partial^2U_1}{\partial x_2\partial x_2}-f_1 \tag*{(26)}

When the shift flow coincides with the laminar base flow, the right-hand side of the governing equation vanishes; consequently, the left-hand side must also be zero. All requirements for an absorbing zone are therefore satisfied except the inequality μe,∞<0\mu_{e,\infty}<0. This condition holds whenever the Reynolds number lies below the global-stability limit obtained from the classical Reynolds–Orr energy method, Re<ReERe<Re_E. In that regime the minimal absorbing zone collapses to a single point—the laminar base-flow state—which is consistent with the base flow being a global attractor.

The equation (14), which expresses the sensitivity, also reduces in one dimension to

\begin{bmatrix} \frac{2}{Re}D^2+\mu_L & -u_1 \\ \frac{1}{Re}(D^2U_1+2D^2u_1)+f_1 & 0 \end{bmatrix} \begin{bmatrix} \delta u_1\\ \delta\mu_L \end{bmatrix} = \begin{bmatrix} -\frac{1}{Re}D^2\\ -\frac{1}{Re}D^2u_1 \end{bmatrix} [\delta U_1]. \tag*{(27)}\end{bmatrix}

Two configurations are examined: (i) plane Couette flow between moving walls, and (ii) plane Poiseuille flow driven by a constant pressure gradient between stationary walls. The Reynolds number is defined as:

Re=U0hν(28)Re = \frac{U_0 h}{\nu} \tag*{(28)}

where hh is the half-gap. For Couette flow U0U_0 is half the wall-speed difference; for Poiseuille flow it is the centreline velocity of the laminar base flow. The forcing term f1f_1 vanishes for Couette flow and is chosen to yield the prescribed ReRe for Poiseuille flow.

In the numerical study ReRe was varied from 30 to 500 in increments of 2.5 for plane Couette flow, and from 100 to 4000 in increments of 10 for plane Poiseuille flow. At each step the optimiser is initialised with the previously converged solution. In the first step, the laminar base flow was used, however in the case of Couette flow a minor perturbation was necessary. Optimisation was carried out in MATLAB 2024b using the built-in fmincon routine with the interior-point algorithm. Spectral-collocation grids containing 71, 81, and 91 points were tested and produced indistinguishable results; the data presented below correspond to the 91-point grid.

Results

The dependence of the minimal absorbing zone on the Reynolds number is examined first. Figure 3 displays the kinetic energy of the critical finite perturbation normalised by the kinetic energy of the laminar base flow. Below the classical energy-stability limits, ReE=20.6625Re_E = 20.6625 for plane-Couette flow and ReE=49.6035Re_E = 49.6035 for plane-Poiseuille flow, the size of the region is zero. Immediately above these thresholds the zone expands rapidly, after which its growth reduces to an approximately linear trend in both configurations. (Unfortunately, for larger Reynolds numbers very sharp velocity profiles are obtained as optimal shift flow, which makes it difficult to analyse the further trend.)

Figure 3. Kinetic energy of the critical finite perturbation, normalised by the kinetic energy of the laminar base flow, for (a) plane Couette and (b) plane Poiseuille flow.

Even at modest Reynolds numbers the energy of the critical finite perturbation remains comparable to that of the base flow. For Poiseuille flow the ratio stays below unity and grows only slowly, whereas in Couette flow it exceeds unity for Re≳270Re \gtrsim270 and increases more steeply.

The optimal shift-flow profiles and their corresponding critical finite perturbations are depicted in Figure 4. For Couette flow, the velocity gradient near the wall increases sharply while the slope near the channel centre approaches zero. This behaviour reflects the requirement that, at high perturbation amplitudes, the kinetic energy must decrease therefore viscous dissipation must dominate production. Since production is proportional to the product of fluctuation velocity and shift-flow shear, the shear is concentrated where the fluctuations must vanish due to boundary condition: close to the walls.

Figure 4. Optimal shift-flow profiles (a,c) and their corresponding critical finite perturbations (b,d). Panels (a–b): Couette flow; panels (c–d): Poiseuille flow.

In Poiseuille flow the wall gradient grows only moderately, whereas the bulk velocity is markedly reduced. This is a consequence of the chosen non-dimensionalisation; a Reynolds number based on bulk velocity would reveal a more Couette-like trend.

The critical finite perturbation in Couette flow changes little in shape as ReRe increases—only its amplitude grows. In contrast, the Poiseuille counterpart undergoes a pronounced shape change, presumably owing to the imposed pressure gradient.

The results are also compared with recent DNS simulations from the literature [3] at Re=500Re = 500 for Couette flow, as shown in Figure 4(a). The turbulent mean profile is significantly less sharp close to the wall and does not satisfy the condition μe,∞<0\mu_{e,\infty} < 0. It closely resembles the optimal shift flow profile at a much lower Reynolds number, Re=60Re = 60, although it still does not match it perfectly.

Figure 5 compares the dimensionless optimal shift-flow profiles with the classical Spalding fit for turbulent boundary layers, using viscous scaling. Above Re≈200Re \approx200 the optimal profiles become nearly Reynolds-number-independent and almost identical between the two configurations, yet they differ substantially from the turbulent law of the wall. Therefore, we can conclude that the approach to obtain the minimal absorbing zone cannot predict the turbulent mean profile well.

Figure 5. Velocity profiles non-dimensionalised by the friction velocity versus wall distance for (a) Couette and (b) Poiseuille flow, compared with the Spalding profile (κ=0.41\kappa= 0.41, C=5.2C = 5.2).

Interestingly, the optimal shift flow profiles resembles the optimal dissipation-maximising profile obtained by Plasting & Kerswell [14] (figure 6 in the cited paper) at Re=73 000Re = 73\,000 for the Couette flow, although the latter exhibits a slightly lower friction velocity. There U+≈4U^{+} \approx4 at y+≈80y^{+} \approx80, while in our case U+≈5.6U^{+} \approx5.6 at the same location. The source of this coincidence between two quite different optimisation criteria remains unclear.

Two main factors are believed to be the main source of error in approximating the turbulent mean profile. First, probably the usage of kinetic energy to prove the existence of the absorbing zone simplifies the problem significantly but is not optimal. A more general energy functional—albeit harder to employ owing to the loss of the Reynolds–Orr identity—might yield a smaller zone that coincides more closely with the turbulent part of the state space. Hopefully, such an approach could lead to the turbulent mean flow without actually averaging a time-dependent DNS solution. Second, the assumption that the turbulent attractor explores state space randomly and uniformly is likely oversimplified; in reality the dynamics may linger near certain invariant solutions (e.g. unstable periodic orbits), so that the centroid of the minimal absorbing zone does not represent the mean flow.

Although minimising the absorbing zone does not predict the turbulent mean profile, it offers a route to proving global stability of the laminar base flow. Any larger domain which includes the whole minimal absorbing zone is also an absorbing zone. Therefore, if the laminar base flow solution is inside the minimal absorbing zone, we can also define an absorbing zone around the laminar base flow, as visualised in Figure 6. The size of this zone can be characterised by the kinetic energy:

The sketch of the absorbing zones around the optimal shift flow and the laminar base flow

Figure 6. The sketch of the absorbing zones around the optimal shift flow and the laminar base flow

eAZ,BF=eAZ,opt+ed.(29)e_{\mathrm{AZ,BF}} = e_{\mathrm{AZ,opt}} + e_d. \tag*{(29)}

where

ed=12∫V(UiBF−UiSF)(UiBF−UiSF) dV,(30)e_{\mathrm{d}} = \frac{1}{2}\int_{\mathcal{V}}\left(U_i^{\mathrm{BF}}-U_i^{\mathrm{SF}}\right)\left(U_i^{\mathrm{BF}}-U_i^{\mathrm{SF}}\right)\,\mathrm{d}V, \tag*{(30)}

is the kinetic energy of the difference between the laminar base flow UiBFU_i^{\mathrm{BF}} and the optimal shift flow UiSFU_i^{\mathrm{SF}}. These energies, normalised by the laminar kinetic energy EBFE_{\mathrm{BF}}, are plotted in Figure 7. First, it can be seen that ed≤eAZ,BFe_{\mathrm{d}} \le e_{\mathrm{AZ,BF}} in both flow configurations; therefore, the minimal absorbing zone includes the laminar state, as expected, since in both configurations, the base flow is linearly stable. Consequently, there exists an absorbing zone around the laminar base state.

Figure 7. The kinetic energy of critical finite perturbations of the optimal shift flow (eAZ,opte_{\mathrm{AZ,opt}}), the size of the absorbing zone around the laminar base flow (eAZ,BFe_{\mathrm{AZ,BF}}), and the energy of the difference between the base flow and the optimal shift flow (ede_{\mathrm{d}}) normalised by the energy of the laminar base flow (EBFE_{\mathrm{BF}}). (a) Couette flow, (b) Poiseuille flow.

This absorbing zone can be utilised to prove the global stability of the base flow. In the case of a linearly stable base state, there exists a region of attraction of the laminar state. If the provable region of attraction is large enough and this minimal zone is fully inside, it proves the global stability of the flow. Recently, for finite-dimensional systems, Nagy [12] showed an efficient way to construct such a conditional Lyapunov function to calculate the size of the region of attraction and showed possible ways of extension to infinite-dimensional systems. However, it must be mentioned that even in these preliminary results, the normalised allowable kinetic energy is of the order of 10−210^{-2}–10−310^{-3}. Since the size of the minimal absorbing zone increases rapidly after the well-known stability limit ReERe_E and is of the order of 1, further improvement is still necessary to significantly increase the provable global stability limit of shear flows.

A further attempt is made to slightly improve the global stability results. In this case, the size of the absorbing zone around the laminar base flow (eAZ,BFe_{\mathrm{AZ,BF}}) is minimised instead of only the size of the absorbing zone. This needs only a minor, trivial modification in the calculation of the cost function and the gradient. In the case of Couette flow, the results were identical to the previous optimisation procedure; therefore, these results are not shown. At the same time, a non-negligible improvement is made in the case of Poiseuille flow. The observed difference between the flows is attributed to the presence of forcing, which is zero in the case of Couette flow. The result of the new optimisation procedure is presented in Figure 8. Compared with the previous optimisation, the absorbing zone around the laminar state is ≈5%\approx5\% smaller. In contrast, the absorbing zone around the shift flow grew by ≈10%\approx10\%, and the size difference between the shift flow and the laminar base flow decreased by ≈35%\approx35\%.

Figure 8. Poiseuille flow: comparison of the original optimisation (dashed, pale curves) with direct minimisation of eAZ,BFe_{\mathrm{AZ,BF}} (solid curves). Quantities as in Figure 7.

Conclusion

In this study, a variational method was introduced and applied to determine the minimal absorbing zone for incompressible shear flows, exploiting the kinetic-energy balance and the Reynolds-Orr identity. The primary hypothesis was that the centroid of this minimal zone—the optimal shift flow—could provide a direct approximation of the turbulent mean flow without requiring time-dependent simulations or empirical closures.

The procedure was successfully implemented for plane Couette and plane Poiseuille flows, producing optimised shift–flow profiles and corresponding absorbing–zone sizes over a range of Reynolds numbers. However, comparison with direct numerical simulation data and canonical turbulent profiles revealed substantial quantitative discrepancies. The optimal shift flows systematically exhibit steeper near-wall gradients than their turbulent counterparts, indicating that the hypothesis, in its present form, is not supported.

Two plausible sources of this discrepancy are suggested. First, reliance on the standard kinetic energy, while analytically convenient (it excludes explicit contributions from the nonlinear advection terms), may be overly restrictive for delineating the true boundary of the absorbing zone. Employing a more general energy functional could alleviate this limitation, albeit at the cost of a much more challenging optimisation problem. Second, the assumption that a chaotic turbulent trajectory explores the state space randomly and uniformly is likely an oversimplification; the dynamics may instead stay near invariant solutions, such as unstable periodic orbits. Therefore, the centroid of the minimal absorbing zone need not coincide with the turbulent mean state.

Despite the failure of the primary hypothesis, the methodology retains significant value for global-stability analysis. The computed minimal absorbing zone constitutes a rigorous, provable subset of state space that must contain all attractors, including the laminar solution. An absorbing region centred on the laminar base flow was also constructed. If this region can be shown to lie entirely within the laminar state’s region of attraction, global stability follows. Although the present bounds remain too large to markedly extend existing global-stability limits, the approach is novel and promising. Future work based on more general Lyapunov functionals may yield smaller, tighter absorbing zones, potentially opening a new route to estimating turbulent mean states and deepening the understanding of hydrodynamic stability.

Funding. This paper was supported by the János Bolyai Research Scholarship of the Hungarian Academy of Sciences.

Declaration of interests. The author reports no conflict of interest.

Author ORCIDs. P. T. Nagy, https://orcid.org/0000-0002-8024-3824

References

  1. [1]Ashtari, Omid & Schneider, Tobias M. 2023 Identifying invariant solutions of wall-bounded three-dimensional shear flows using robust adjoint-based variational techniques. Journal of Fluid Mechanics 977, A7.arxiv.org/abs/2306.00165
  2. [2]Avila, Marc, Barkley, Dwight & Hof, Björn 2023 Transition to turbulence in pipe flow. Annual Review of Fluid Mechanics 55 (Volume 55, 2023), 575–602.DOI
  3. [3]Cavalieri, André V. G. & Noqueira, Petrônio A. S. 2022 Reduced-order Galerkin models of plane Couette flow. Phys. Rev. Fluids 7, L102601.arxiv.org/abs/2207.06808
  4. [4]Doering, Charles R. 2009 The 3D Navier-Stokes Problem. Annual Review of Fluid Mechanics 41 (Volume 41, 2009), 109–128.DOI
  5. [5]Duguet, Yohann, Monokrousos, Antonios, Brandt, Luca & Henningson, Dan S. 2013 Minimal transition thresholds in plane Couette flow. Physics of Fluids 25 (8).DOI
  6. [6]Eckhardt, Bruno, Schneider, Tobias M., Hof, Björn & Westerweel, Jerry 2007 Turbulence transition in pipe flow. Annual Review of Fluid Mechanics 39 (Volume 39, 2007), 447–468.DOI
  7. [7]Jiménez, Javier 2018 Coherent structures in wall-bounded turbulence. Journal of Fluid Mechanics 842, P1.
  8. [8]Joseph, D. D. & Carmi, S 1969 Stability of Poiseuille flow in pipes, annuli, and channels. Quarterly of Applied Mathematics 26 (4), 575–599.DOI
  9. [9]Kawahara, Genta, Uhlmann, Markus & van Veen, Lennaert 2012 The significance of simple invariant solutions in turbulent flows. Annual Review of Fluid Mechanics 44 (Volume 44, 2012), 203–225.DOI
  10. [10]Kerswell, R. R. 1999 Variational principle for the Navier-Stokes equations. Phys. Rev. E 59, 5482–5494.pubmed.ncbi.nlm.nih.gov/11969527
  11. [11]Kerswell, R R, Pringle, C C T & Willis, A P 2014 An optimization approach for analysing nonlinear stability with transition to turbulence in fluids as an exemplar. Reports on Progress in Physics 77 (8), 085901.
  12. [12]Nagy, Péter Tamás 2025 Constructing quadratic Lyapunov functions for conditionally stable fluid dynamics systems. Journal of Fluid Mechanics 1014, A25.
  13. [13]Olivier Dauchot & Paul Manneville 1997 Local versus global concepts in hydrodynamic stability theory. J. Phys. II France 7 (2), 371–389.
  14. [14]Plasting, S. C. & Kerswell, R. R. 2003 Improved upper bound on the energy dissipation rate in plane Couette flow: the full solution to Busse’s problem and the Constantin–Doering–Hopf problem with one-dimensional background field. Journal of Fluid Mechanics 477, 363–379.
  15. [15]Plasting, S. C. & Kerswell, R. R. 2004 A friction factor bound for transitional pipe flow. Physics of Fluids 17 (1), 011706, arXiv: https://pubs.aip.org/aip/pof/article-pdf/doi/10.1063/1.1828013/14691402/011706_1_online.pdf.
  16. [16]Pringle, Chris C. T., Willis, Ashley P. & Kerswell, Rich R. 2012 Minimal seeds for shear flow turbulence: using nonlinear transient growth to touch the edge of chaos. Journal of Fluid Mechanics 702, 415–443.
  17. [17]Sanders, John W., DeVoria, A.C., Washuta, Nathan J., Elamin, Gafar A., Skenes, Kevin L. & Berlinghieri, Joel C. 2024 A canonical Hamiltonian formulation of the Navier–Stokes problem. Journal of Fluid Mechanics 984, A27.

Paper details

vlnc:W_gmvzn84r
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents