Entropy decrease and emergence of order in collective dynamics

Eitan Tadmor · 2025

Read Entropy decrease and emergence of order in collective dynamics by Eitan Tadmor (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-2506-17635.txt.

Introduction

Eitan Tadmor studies how groups of self-moving agents settle into orderly flocking. His 2025 paper works with the fluid-style equations for crowds whose members align their velocities with neighbors. Such systems are known to flock when their solutions stay smooth, but proving smoothness for all time was open in higher dimensions.

Tadmor proceeds in two steps. First he examines how entropy falls and the system drifts toward a simple mono-kinetic state, where all local agents share one velocity. Then he proves lasting smoothness for a wide class of starting states that meet a threshold condition tied to that entropy drop. The proof works in any number of dimensions.

Catalog record

CREATOREitan Tadmor
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE62 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2506-17635

Browse the full library · About the Tower of Babel

Full text

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

Entropy decrease and emergence of order in collective dynamics Thanks: I thank Thomas Chen for helpful discussions. Research was supported by ONR grant N00014-2412659, NSF grant DMS-2508407 and by the Fondation Sciences Mathématiques de Paris (FSMP) while being hosted by LJLL at Sorbonne University. Eitan Tadmor Abstract . We study the hydrodynamic description of collective dynamics driven by velocity alignment . It is known that such Euler alignment systems must flock towards a limiting “flocking” velocity, provided their solutions remain globally smooth. To address this question of global existence we proceed in two steps. (i) Entropy and closure. The system lacks a closure, reflecting lack of detailed energy balance in collective dynamics. We discuss the decrease of entropy and the asymptotic behavior towards a mono-kinetic closure; and (ii) Mono-kinetic closure. We prove that global regularity persists for all time for a large class of initial conditions satisfying a critical threshold condition, which is intimately linked to the decrease of entropy. The result applies in any number of spatial dimensions, thus addressing the open question of existence beyond two dimensions. 1991 Mathematics Subject Classification 92D25, 35Q31, 76N10 1. Introduction and statement of main results We are concerned with the existence and large-time behavior of global smooth solutions for the multi-dimensional Euler alignment system (1.1) \displaystyle\quad\left\{\begin{aligned} &\rho_{t}+\nabla_{\mathbf{x}}\cdot(\rho{\mathbf{u}})=0,\\ &(\rho{\mathbf{u}})_{t}+\nabla_{\mathbf{x}}\cdot(\rho{\mathbf{u}}\otimes{\mathbf{u}}+{\mathbb{P}})=\tau\int\phi({\mathbf{x}},{\mathbf{y}})\big({\mathbf{u}}(t,{\mathbf{y}})-{\mathbf{u}}(t,{\mathbf{x}})\big)\rho(t,{\mathbf{x}})\rho(t,{\mathbf{y}}){\textnormal{d}}{\mathbf{y}}.\end{aligned}\right. A solution pair of density-velocity, (\rho,{\mathbf{u}}):\mathbb{R}_{+}\times\Omega\rightarrow\mathbb{R}_{+}\times\mathbb{R}^{n} , is sought subject to compactly supported initial data \big(\rho(0,{\mathbf{x}}),{\mathbf{u}}(0,{\mathbf{x}})\big)=\big(\rho_{0}({\mathbf{x}}),{\mathbf{u}}_{0}({\mathbf{x}})\big) , either in the whole space, \Omega=\mathbb{R}^{n} , or in the n -dimensional torus \Omega=\mathbb{T}^{n} . System ( 1.1 ) is the large-crowd hydrodynamic description of ‘social agents’ identified by positions and velocities, \{{\mathbf{x}}_{i}(t),{\mathbf{v}}_{i}(t)\}_{i=1}^{N},\ N\gg 1 , governed by the Cucker-Smale alignment model [ CS2007 ] \displaystyle\left\{\begin{array}[]{rr}\begin{aligned} &\dot{{\mathbf{x}}}_{i}={\mathbf{v}}_{i}\\ &\dot{{\mathbf{v}}}_{i}=\frac{\tau}{N}\sum_{j=1}^{N}\phi({\mathbf{x}}_{i},{\mathbf{x}}_{j})({\mathbf{v}}_{j}-{\mathbf{v}}_{i}).\end{aligned}\end{array}\right. Alignment dynamics is a canonical model governing emergence phenomena in collective dynamics of flocking, swarming etc. The dynamics is driven by a non-negative symmetric communication kernel , \phi({\mathbf{x}},{\mathbf{y}})=\phi({\mathbf{y}},{\mathbf{x}})\geqslant 0 , with amplitude \tau>0 . We have two main examples of symmetric kernels in mind — the canonical Cucker-Smale class of metric kernels , see [ CS2007 ] (1.4) \phi({\mathbf{x}},{\mathbf{y}})=\phi(|{\mathbf{x}}-{\mathbf{y}}|), and the class of topologically-based kernels [ ST2020b ] \phi({\mathbf{x}},{\mathbf{y}})=\phi_{1}(|{\mathbf{x}}-{\mathbf{y}}|)\phi_{2}(d_{\rho}({\mathbf{x}},{\mathbf{y}})),\qquad d_{\rho}({\mathbf{x}},{\mathbf{y}}):=\int\limits_{{\mathcal{C}}({\mathbf{x}},{\mathbf{y}})}\!\!\!\rho(t,{\mathbf{z}}){\textnormal{d}}{\mathbf{z}}, reflecting the dependence on the mass in an intermediate ‘domain of communication’, {\mathcal{C}}({\mathbf{x}},{\mathbf{y}}) , enclosed between {\mathbf{x}} and {\mathbf{y}} . The Euler alignment system ( 1.1 ) involves the pressure {\mathbb{P}} — a symmetric positive definite tensor which should encode the thermodynamics of large-crowd collective dynamics. But ( 1.1 ) is not a closed system: it lacks closure of {\mathbb{P}} in terms of the macroscopic variables \rho and {\mathbf{u}} . This reflects the fact that unlike physical particles, the social agents engaged in collective dynamics form a thermo-dynamically open system which is far from equilibrium and does not admit a universal closure. Indeed, most of the relevant literature assumes a mono-kinetic closure, {\mathbb{P}}\equiv 0 . Accordingly, our study of solutions of ( 1.1 ) proceeds in two main stages: (i) We investigate a rather general class of so-called entropic pressures introduced in [ Tad2023 ] and conclude that their strong solutions must approach mono-kinetic closure. Indeed, strong Euler alignment solutions with isentropic closure experience a uniform entropy decrease to -\infty , and if we reject such a scenario then we must impose mono-kinetic closure , {\mathbb{P}}\equiv 0 . (ii) We consider the pressure-less or mono-kinetic Euler alignment, proving existence of global smooth solutions under certain threshold conditions . This addresses the open question of existence for n\geqslant 3 , extending the known results for dimensions n=1,2 [ TT2014 , CCTT2016 , HT2017 ] . 1.1. The road to mono-kinetic closure To investigate this mono-kinetic assumption we let \rho E denote the (total) energy associated with the pressure in ( 1.1 ) (here and below, |{\mathbf{w}}| denotes the \ell_{2} -norm of {\mathbf{w}} and |{\mathbf{w}}|_{\infty} denote its L^{\infty} -norm), (1.5) \rho E:=\tfrac{1}{2}\rho|{\mathbf{u}}|^{2}+\rho e,\qquad\rho e:=\tfrac{1}{2}\textnormal{trace}({\mathbb{P}}), Since the social agents engaged in collective dynamics are often driven by energy received from the “outside”, the detailed energy balance associated with ( 1.1 ) may be less relevant, [ VZ2012 , §1.1] . Instead, lack of thermal equilibrium in the form of closure equalities , can be relaxed to certain inequalities , which are compatible with the decay of energy fluctuations . We impose the notion of an entropic pressure in which {\mathbb{P}} , augmented with arbitrary “heat-flux” vector function {\mathbf{q}} , are required to satisfy the inequality, (1.6) \begin{split}(\rho E)_{t}+\nabla_{\mathbf{x}}\cdot&(\rho E{\mathbf{u}}+{\mathbb{P}}{\mathbf{u}}+{\mathbf{q}})\\ &\leqslant-\tau\int\phi({\mathbf{x}},{\mathbf{y}})\big(2\rho E(t,{\mathbf{x}})-\rho{\mathbf{u}}(t,{\mathbf{x}})\cdot{\mathbf{u}}(t,{\mathbf{y}})\big)\rho(t,{\mathbf{y}}){\textnormal{d}}{\mathbf{y}}.\end{split} This can be expressed in an equivalent form 1 1 1 The formal manipulation ( 1.1 ) {}_{2}\cdot{\mathbf{u}}- ( 1.1 ) {}_{1}\times\frac{|{\mathbf{u}}|^{2}}{2} yields \big(\tfrac{1}{2}\rho|{\mathbf{u}}|^{2}\big)_{t}+\nabla_{\mathbf{x}}\cdot(\tfrac{1}{2}\rho|{\mathbf{u}}|^{2}{\mathbf{u}}+{\mathbb{P}}{\mathbf{u}})-\textnormal{trace}({\mathbb{P}}\nabla{\mathbf{u}})=-\tau\int\phi({\mathbf{x}},{\mathbf{y}})\rho{\mathbf{u}}(t,{\mathbf{x}})\cdot\big({\mathbf{u}}(t,{\mathbf{x}})-{\mathbf{u}}(t,{\mathbf{y}})\big)\rho(t,{\mathbf{y}}){\textnormal{d}}{\mathbf{y}}. Combining this equality with ( 1.7 ) is equivalent with ( 1.6 ). in term of an entropic inequality for the internal energy, \rho e , stating [ Tad2023 , Definition 1.1] , (1.7) (\rho e)_{t}+\nabla_{\mathbf{x}}\cdot(\rho e{\mathbf{u}}+{\mathbf{q}})+\textnormal{trace}\big({\mathbb{P}}\nabla{\mathbf{u}}\big)\leqslant-2\tau\int\phi({\mathbf{x}},{\mathbf{y}})\rho e(t,{\mathbf{x}})\rho(t,{\mathbf{y}}){\textnormal{d}}{\mathbf{y}}. The entropic inequalities ( 1.6 ) or ( 1.7 ) are flexible enough to cover pressure tensors derived from an underlying kinetic formulation discussed in § 2 , as well as the mono-kinetic closure {\mathbb{P}}\equiv 0 . They imply depletion of fluctuations and in particular, as we shall see later on, that the pressure in entropic alignment dynamics approaches the mono-kinetic closure. In § 3 we discuss several results which quantify this statement of “approach towards mono-kinetic closure”. To state one of these results, we need the following notations. Notations . We let \nabla_{\!{}_{S}}{\mathbf{u}} denote the symmetric gradient, \displaystyle(\nabla_{\!{}_{S}}{\mathbf{u}})_{ij}=\tfrac{1}{2}\Big(\frac{\partial u_{i}}{\partial x_{j}}+\frac{\partial u_{j}}{\partial x_{i}}\Big) . We order the real eigenvalues of \nabla_{\!{}_{S}}{\mathbf{u}} : \lambda_{-}:=\lambda_{1}\leqslant\lambda_{2}<\ldots\leqslant\lambda_{n}=:\lambda_{+} , and in general, given a closed set X\subset\mathbb{R} we let X_{+} and X_{-} denote its largest, respectively smallest elements. We use \eta_{c} to denote different positive constants. Finally, we use \phi*\rho to denote the average density, or thickness , \phi*\rho(t,{\mathbf{x}}):=\int\phi({\mathbf{x}},{\mathbf{y}})\rho(t,{\mathbf{y}}){\textnormal{d}}{\mathbf{y}}; of course, in case of metric kernel, this coincides with the usual notion of convolution. In § 3 we prove the following. Theorem 1.1 ( Decay towards mono-kinetic closure). Let (\rho,{\mathbf{u}},{\mathbb{P}}) be a strong entropic solution of ( 1.1 ) and assume the following threshold condition holds: there exists a constant, \eta_{c}>0 , such that (1.8) \lambda_{-}(\nabla_{\!{}_{S}}{\mathbf{u}})(t,\cdot)+\tau\phi*\rho(t,\cdot)\geqslant\eta_{c}>0. Then \displaystyle\int\|{\mathbb{P}}(t,{\mathbf{x}})\|{\textnormal{d}}{\mathbf{x}}\leqslant e^{-\eta_{c}t}\int\|{\mathbb{P}}_{0}({\mathbf{x}})\|{\textnormal{d}}{\mathbf{x}} . 1.2. Communications kernels and thickness Different classes of communication kernels \phi({\mathbf{x}},{\mathbf{y}}) are treated in the literature, classified according to their short- and long-range behavior where {\mathbf{x}}\approx{\mathbf{y}} , and, respectively, |{\mathbf{x}}-{\mathbf{y}}|\gg 1 . The class of singular kernels \phi({\mathbf{x}},{\mathbf{y}})\sim|{\mathbf{x}}-{\mathbf{y}}|^{-\alpha} was treated in [ FP2024 ] where it was shown that the enstrophy associated with strongly singular kernels, \alpha\in[n,n+2) , enforced mono-kinetic closure. Existence and flocking behavior of the 1D mono-kinetic case with strongly singular kernels was studied in [ DKRT2018 , ST2017a , ST2017b , ST2018 ] . In this paper we restrict attention to the case of bounded kernels, \phi(\cdot,\cdot)\leqslant\phi_{+} . The results can be extended to the case of integrable kernel with weak singularity, \phi({\mathbf{x}},{\mathbf{y}})\sim|{\mathbf{x}}-{\mathbf{y}}|^{-\alpha} with \alpha<n . Alignment with such weakly singular kernels were studied in [ MMPZ2019 ] . Next, we make the distinction between long-range and short-range kernels. Assume that \phi(\cdot,\cdot) admits a metric lower envelope, \phi({\mathbf{x}},{\mathbf{y}})\gtrsim\varphi(|{\mathbf{x}}-{\mathbf{y}}|) , with decreasing radial \varphi . Heavy-tailed kernels is the subclass of long-range kernels that naturally arise in connection with the unconditional flocking behavior of ( 1.1 ), [ HT2008 , HL2009 , CFTV2010 , CFRT2010 ] \int^{\infty}\varphi(r){\textnormal{d}}r=\infty. Heavy-tails rule out short-range kernels

Read the complete text