Stability of nonlinear dissipative systems with applications in fluid dynamics

Javier Gonzalez-Conde, Daniel Isla, Sergiy Zhuk, Mikel Sanz · 2026

Read Stability of nonlinear dissipative systems with applications in fluid dynamics by Javier Gonzalez-Conde, Daniel Isla, Sergiy Zhuk, Mikel Sanz (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-2603-26627.txt.

Introduction

Gonzalez-Conde and co-authors derive a stability test for nonlinear dissipative systems with quadratic nonlinearities. Such equations drive physics, engineering, and finance, yet exact solutions are rare and numeric ones grow costly. Turbulent behaviour can hinge on tiny changes in starting data, unless stability holds. By tracking solution norms in Sobolev spaces, the team links the linear damping, the quadratic term, and the forcing in one clear inequality. Meeting it keeps perturbations bounded for a wide set of starts, and fluid models show the test at work.

Catalog record

CREATORJavier Gonzalez-Conde, Daniel Isla, Sergiy Zhuk, Mikel Sanz
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE29815
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2603.26627

Browse the full library · About the Tower of Babel

Full text

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

Stability of nonlinear dissipative systems with applications in fluid dynamics Javier Gonzalez-Conde, Daniel Isla, Sergiy Zhuk, Mikel Sanz Abstract Nonlinear partial differential equations are central to physics, engineering, and finance. Except in a limited number of integrable cases, their solution generally requires numerical methods whose cost becomes prohibitive in high-dimensional regimes or at fine resolution. Nonlinear phenomena such as turbulence are notoriously difficult to predict because of their extreme sensitivity to small variations in initial conditions, except when certain stability conditions are fulfilled. Indeed, stability allows us to achieve reliable approximate dynamics, since it determines whether small perturbations remain bounded or are amplified, potentially leading to markedly different long-term behavior. Here, we investigate the stability of dissipative partial differential equations with second-order nonlinearities. By analyzing the time evolution of solution norms in Sobolev spaces, we establish a sufficient condition for stability that links the characteristics of the linear dissipative operator, the quadratic nonlinear term, and the external forcing. The resulting criterion is expressed as an explicit inequality that guarantees stability for a wide range of initial conditions. As an illustration, we apply the framework to fluid-dynamical models governed by nonlinear partial differential equations. In particular, for the Burgers equation, the condition admits a natural interpretation in terms of the Reynolds number, thereby directly linking the stability threshold to the competition between viscous dissipation and inertial advection. We further demonstrate the scope of the approach by extending the analysis to the KPP–Fisher and Kuramoto–Sivashinsky equations. Introduction– Obtaining sufficiently accurate solutions for the evolution of nonlinear dynamical systems is typically computationally expensive. However, given a nonlinear differential problem, a numerical scheme, a set of computational resources, and a simulation time, the ability to predict whether the simulation results will achieve the required level of accuracy is of fundamental importance. Khalil 2002 ; Strogatz 2018 ; Slotine and Li 1991 ; Hirsch et al. 2003 . This is where the concept of stability in partial differential equations (PDEs) becomes crucial. It ensures that nearby trajectories remain close under small perturbations—that is, such perturbations are not amplified—thereby reducing the required computational resources Markus and Yamabe 1960 ; Oliveira 2022 ; Grigorian 2021 , as proven in Brudno’s theorem Brudno 1983 . In engineering, stability underpins the design of reliable control systems, ensuring, for example, that autopilots and robotic arms perform correctly under a broad range of operating conditions Åström and Murray 2021 ; Ogata 2010 . In biology, it provides a framework for modeling population dynamics that settle into steady patterns over time Hofbauer and Sigmund 1998 ; Murray 2007 ; May 2001 . In economics, it supports models in which markets or broader systems return to equilibrium despite shocks or fluctuations Fisher 1983 ; Benhabib and Farmer 1994 . Likewise, in fluid-flow dynamics, stability is essential for understanding how velocity fields evolve and whether disturbances in flows, such as turbulence, dissipate or persist over time Drazin and Reid 2004 ; Jameson and Vassberg 2001 ; Macián-Pérez et al. 2020 ; Ghafori 2020 and to be able to provide reliable state estimates and predict the future states Zhuk et al. 2023 ; Yeo et al. 2024 . Figure 1: Stability ensures that if the initial profile is slightly altered, then the resulting evolution remains close to the reference solution for all later times, guaranteeing that numerical solutions behave consistently with the true physical problem over time. Without stability, even accurate or consistent schemes can produce meaningless results. In this work, we investigate the stability properties of dissipative second-order nonlinear PDEs. In particular, we derive a sufficient stability condition expressed in terms of the parameters of the PDE. This condition is obtained by analyzing the evolution of the Sobolev norm associated with two nearby initial conditions in the Sobolev space H^{2}(\Omega) . We then examine how this stability criterion can be applied to relevant PDEs in fluid dynamics, establishing connections with key physical quantities such as the Reynolds number. Theoretical framework — Let us start by considering the family of “generalized” advection–diffusion nonlinear-quadratic PDEs with the form \displaystyle\partial_{t}u(x,t)=\mathcal{L}[u]+\mathcal{Q}[u]+f(x,t) (1) where d is the order of the differential operator, the term \mathcal{L}[u]=~\sum_{l=0}^{d}\frac{1}{L^{l}}a_{l}\partial_{x}^{(l)}u(x,t) represents the linear operator, the nonlinearity is introduced via \mathcal{Q}[u]=~\sum_{0\leq k\leq l}^{d}\frac{1}{L^{k+l}}b_{kl}\partial_{x}^{(k)}u(x,t)\partial_{x}^{(l)}u(x,t), and f(x,t) is the forcing term that feeds energy into the system. Here for simplicity, we have rescaled our physical coordinate \hat{x}\in[0,L] to the dimensionless coordinate x=\frac{\hat{x}}{L}\in\Omega\mathrel{\mathop{\mathchar 58\relax}}=[0,1] . Time remains dimensional and denoted by t\in[0,\infty) . From now on, we presume that the solutions to Eq. ( 1 ) belong to the Hilbert–Sobolev space H^{d+2}(\Omega) , and we proceed to study their stability in the Hilbert–Sobolev space H^{2}(\Omega) . Additionally, we assume u(0,t)=0 and periodic boundary conditions for all the spatial derivatives, i.e. \partial_{x}^{(k)}u(x,t)\rvert_{x=0}=\partial_{x}^{(k)}u(x,t)\rvert_{x=L}\ \ \forall k\in\{0,\ldots,d\}. Stability of trajectories — To analyze the robustness of solutions to Eq. ( 1 ) under small perturbations, we adopt a standard notion of stability formulated in the Sobolev space H^{2}(\Omega) . Informally, stability means that if the initial profile is slightly altered (in the H^{2} -norm), then the resulting evolution remains uniformly close to the reference solution for all later times, see Fig. 1 . More precisely, a solution u(x,t) of Eq. ( 1 ) is said to be trajectory-based stable in the Lyapunov sense if, for every \varepsilon>0 , there exists a \delta_{\epsilon}>0 such that whenever the initial condition u(x,0) is perturbed by less than \delta_{\epsilon} , i.e. \|u(x,0)-\tilde{u}(x,0)\|_{H^{2}}<\delta_{\epsilon} , the perturbed solution \tilde{u}(x,t) satisfies that \|u(x,t)-\tilde{u}(x,t)\|_{H^{2}}<\varepsilon\quad\forall t\geq 0. Additionally, if on top of trajectory-based stable in the Lyapunov sense, we also have that \lim_{t\rightarrow\infty}\|u(x,t)-\tilde{u}(x,t)\|_{H^{2}}=0 , then the solution is trajectory-based asymptotically stable . Theorem 1.— Let u(x,t) be a solution to Eq. ( 1 ) such that \mathcal{L}[\cdot] is self-adjoint and negative definite. Then, if y_{0}<\frac{\beta}{2\mu} and \mathcal{R}\mathrel{\mathop{\mathchar 58\relax}}=\frac{f_{\text{max}}+\mu y_{0}^{2}}{\beta y_{0}}<1 , where \mu\mathrel{\mathop{\mathchar 58\relax}}=\mathcal{C}\sum_{0\leq k\leq l}^{d}\frac{1}{L^{k+l}}|b_{kl}|>0 , y_{0}\mathrel{\mathop{\mathchar 58\relax}}=\|u(0)\|_{H^{2}(\Omega)} , f_{\text{max}}=\max_{t}\|f(t)\|_{H^{2}(\Omega)} and \beta\mathrel{\mathop{\mathchar 58\relax}}=-\ \text{sup}_{k\neq 0}\lambda_{k} , \lambda_{k}\in\sigma(\mathcal{L}[\cdot]) , the solution u(x,t) is trajectory-based asymptotically stable . Before outlining the proof of Theorem 1, we first establish sufficient conditions to find the regime where the H^{2} -norm of the solution u(x,t) is upper bounded by a positive quantity, y(t) , that decreases monotonically in time. To this end, we study the temporal evolution of \|u(t)\|_{H^{2}(\Omega)} , and in particular of its square, i.e. \frac{d}{dt}\lVert u\rVert_{H^{2}}^{2}=2\left\langle u,\partial_{t}u\right\rangle_{H^{2}} . After substituting the governing equation, it reads \displaystyle\frac{d}{dt}\lVert u\rVert_{H^{2}}^{2}\leq 2\langle u,\mathcal{L}[u]\rangle_{H^{2}}+2\left|\langle u,\mathcal{Q}[u]\rangle_{H^{2}}\right|+2\langle u,f\rangle_{H^{2}}. (2) We now examine each term in Eq. ( 2 ) individually. Firstly, the linear operator \mathcal{L} , whose contribution can be expressed as \langle u,\mathcal{L}[u]\rangle_{H^{2}}=\sum_{l=0}^{d}\frac{1}{L^{l}}a_{l}\left\langle u,\partial^{(l)}u\right\rangle_{H^{2}} To facilitate the spectral analysis, we introduce the eigenfunctions of \mathcal{L}[\cdot] , which take the form u_{k}(x)=~C_{k}\exp\left(i{2\pi k}x\right),\ k\in\mathbb{Z} , with corresponding eigenvalues \lambda_{k}=\sum_{l=0}^{d}a_{l}\left(\frac{2\pi ik}{L}\right)^{l}. The solution u(x,t) to Eq. ( 1 ) may then be expanded in the basis of orthonormal eigenfunctions \{u_{k}(x)\}_{\ k\in\mathbb{Z}} , as u(x,t)=~\sum_{k\neq 0}c_{k}(t)\,u_{k}(x) , where we have ignored the identically zero eigenfunction. Thus, as \mathcal{L}[\cdot] is self-adjoint and negative definite, we have \beta>0 and therefore \langle u,\mathcal{L}[u]\rangle_{H^{2}}=\langle u,\sum_{k}\lambda_{k}c_{k}u_{k}\rangle_{H^{2}}\leq~-\beta\|u\|^{2}_{H^{2}}. Next, we analyze how to bound the quadratic term. By using Cauchy–Schwarz inequality, we can bound |\langle u,\mathcal{Q}[u]\rangle_{H^{2}}|\leq\sum_{0\leq k\leq l}^{d}\frac{1}{L^{k+l}}|b_{kl}|\|u\|_{H^{2}}\|\partial_{x}^{(k)}u\,\partial_{x}^{(l)}u\|_{H^{2}}. We now use the fact that the Sobolev space H^{2}(\Omega) , is a Banach algebra with respect to the

Read the complete text