Representation of Global Viscosity Solutions for Tonelli Hamiltonians
CHARFI Skander · 2025
Read Representation of Global Viscosity Solutions for Tonelli Hamiltonians by CHARFI Skander (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-2503-16035.txt.
Introduction
Skander Charfi stood for the eternal answers of a Hamilton-Jacobi equation in a 2025 analysis paper. For time-dependent Tonelli Hamiltonians, a answer defined for all past and future times is a rare global object. The Lax-Oleinik semigroup that usually builds answers fails to converge in this non-self-running setting, as Fathi and Mather showed. Charfi studies the semigroup's action on its non-wandering set, the states it returns to forever.
On that set the semigroup acts as an isometry, and the set equals exactly the global viscosity answers. A broad Peierls barrier and static classes inside the Mather set then yield a formula writing each global answer as an infimum over those classes. Fathi's self-running convergence theorem falls out as a corollary.
Catalog record
| CREATOR | CHARFI Skander |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 150 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2503-16035
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Representation of Global Viscosity Solutions for Tonelli Hamiltonians CHARFI Skander Abstract We consider the Lax-Oleinik operator 𝒯 \mathcal{T} associated with the non-stationary Hamilton-Jacobi equation ∂ t u + H ( t , x , ∂ x u ) = α 0 \partial_{t}u+H(t,x,\partial_{x}u)=\alpha_{0} for a Tonelli Hamiltonian H H and its Mañé critical value α 0 \alpha_{0} . It is known from the work of A. Fathi and J.N. Mather [ FM00 ] that the convergence of this semigroup fails in the non-autonomous framework. In this context, we study the action of 𝒯 \mathcal{T} on its non-wandering set Ω ( 𝒯 ) \Omega(\mathcal{T}) . First, we show that 𝒯 \mathcal{T} acts as an isometry on this set, and then we characterize Ω ( 𝒯 ) \Omega(\mathcal{T}) as the set of global viscosity solutions of the Hamilton-Jacobi equation, i.e. solutions that are defined for all real times. Next, we introduce a generalized Peierls barrier k ¯ \underline{k} and a set of generalized static classes 𝕄 ¯ \underline{\mathbb{M}} within the Mather set. Using these, we represent elements u u of Ω ( 𝒯 ) \Omega(\mathcal{T}) as u ( x ) = inf y ∈ 𝕄 ¯ { u ( y ) + k ¯ ( y , x ) } u(x)=\inf_{y\in\underline{\mathbb{M}}}\{u(y)+\underline{k}(y,x)\} We apply this representation formula to prove Fathi’s convergence theorem for autonomous systems and provide a representation formula for n n -periodic viscosity solutions. Additionally, we establish that the dynamics of non-wandering viscosity solutions are governed by the Lagrangian flow on the Mather set. Specifically, we show that if the Mather set consists solely of N N -periodic orbits for some integer N N , then all non-wandering viscosity solutions are N N -periodic. Furthermore, we show that if the restriction of the Lagrangian flow to the Mather set is uniformly recurrent for a time sequence p n p_{n} , then all non-wandering viscosity solutions are uniformly recurrent for the same time sequence p n p_{n} . 1 Introduction 1.1 Overview Let M be a connected closed manifold. Denote by TM its tangent bundle and by T^{*}M its cotangent bundle with their respective projections \pi_{TM}:TM\to M and \pi_{T^{*}M}:T^{*}M\to M . Unless it is ambiguous, these projections will be simply denoted by \pi . Denote by \mathcal{C}(M,\mathbb{R}) the set of continuous scalar maps on M endowed with its usual infinite norm \|.\|_{\infty} given by ||u||_{\infty}=\sup_{x\in M}|u(x)| . The set \mathbb{T}^{1}=\mathbb{R}/\mathbb{Z} refers to the circle. Let H(t,x,p):\mathbb{T}^{1}\times T^{*}M\to\mathbb{R} be a 1 -time periodic Tonelli Hamiltonian (See Definition 2.1 ) and let \alpha_{0} be the critical Mañé value associated to H (See Definition 2.3 ). We consider the Hamilton-Jacobi equation \partial_{t}u+H(t,x,\partial_{x}u)=\alpha_{0} (1.1) This equation has a well posed Cauchy problem in the viscosity sense (see [ Lio82 ] ), [ CL83 ] ). And A.Fathi [ Fat97 ] developed the weak-KAM theory which states that in the Tonelli framework, these solutions are generated by the Lax-Oleinik operator \mathcal{T}^{s,t} defined on \mathcal{C}(M,\mathbb{R}) (see Definition 2.3 ). More specifically, for any real time s\in\mathbb{R} and any u_{0}\in\mathcal{C}(M,\mathbb{R}) , there exists a unique viscosity solution u(t,x):[s,+\infty)\times M\to\mathbb{R} of ( 2.8 ) with u(s,\cdot)=u_{0} defined by u(t,x)=\mathcal{T}^{s,t}u_{0}(x) . The weak-KAM theory, initially developed in the autonomous framework by A.Fathi, focused on weak-KAM solutions, which are fixed points of the operator \mathcal{T}^{0,1} . These solutions correspond to viscosity solutions that are one-periodic in time. Furthermore, Fathi explicited examples of weak-KAM solutions using the Peierls barrier h^{\infty}(x,\cdot) (see Definition 5.1 ) and established various properties of weak-KAM solutions, such as their calibration on the Mather set (see Definitions 2.15 and 2.21 ) and domination by the barrier h^{\infty} on the Mather set (see Definition 1.4 ). These were foundational for subsequent results on weak-KAM solutions, including their C^{1,1} -regularity when restricted to the Aubry set, uniqueness on the Mather set, and representation through the Peierls barrier (see [ Fat08 ] ). The representation formula provides a description of any weak-KAM solutions u using the Peierls barrier h^{\infty} as follows u(x)=\inf_{y\in\mathbb{M}}\{\psi(y)+h^{\infty}(y,\cdot)\} (1.2) where \mathbb{M} is a subset of the Mather set called the set of static classes (See Definition 6.3 ) and the map \psi:\mathcal{M}\to\mathbb{R} is dominated by the Peierls barrier h^{\infty} . This formula was later extended for non-autonomous Tonelli Hamiltonians by G. Contreras, R. Iturriaga, and H. Sánchez-Morgado in [ CISM13 ] . In the non-autonomous framework, P. Bernard and J.-M. Roquejoffre [ BR04 ] demonstrated that the calibration on the Mather set applies to non-wandering viscosity solutions of the Hamilton-Jacobi equation 1.1 . This crucial result implies that non-wandering viscosity solutions share similar properties with weak-KAM solutions, such as uniqueness on the Mather set (see Theorem 4.1 ) and C^{1,1} -regularity when restricted to the Aubry set. As a result, it is reasonable to consider the set \Omega(\mathcal{T}^{0,1}) of non-wandering solutions in the non-autonomous framework, rather than focusing solely on the set \fix(\mathcal{T}^{0,1}) of weak-KAM solutions. This paper aims to study the action of the Lax-Oleinik operator \mathcal{T}^{0,1} on \Omega(\mathcal{T}^{0,1}) , which reveals to be the set of all global viscosity solutions, and to generalize the representation formula using a generalized Peierls barrier. 1.2 Main Results We set \mathcal{T}=\mathcal{T}^{0,1} and \mathcal{T}^{\tau}=\mathcal{T}^{0,\tau} . In the first part, we study the action of the Lax-Oleinik operator on its non-wandering set, \Omega(\mathcal{T}) . We begin by noting that \mathcal{T} is non-expanding on \mathcal{C}(M,\mathbb{R}) , which has several implications for the behavior of viscosity solutions. One key implication is that for a viscosity solutions, being non-wandering is equivalent to recurrent (see Proposition 3.11 ). Additionally, we establish that \mathcal{T} acts as an isometry on the non-wandering set \Omega(\mathcal{T}) , as demonstrated by the following proposition Proposition 1.1 . 1. The restriction of \mathcal{T} to the non-wandering set \Omega(\mathcal{T}) is an isometric bijection, i.e \mathcal{T} is invertible and for all v and w in \Omega(\mathcal{T}) , \|\mathcal{T}v-\mathcal{T}w\|_{\infty}=\|v-w\|_{\infty} . 2. More generally, if we denote by \Omega_{\tau}(\mathcal{T})=\mathcal{T}^{\tau}\big(\Omega(\mathcal{T})\big) , then for all times s<t , the operator \mathcal{T}^{s,t}:\Omega_{s}(\mathcal{T})\to\Omega_{t}(\mathcal{T}) is an isometric bijection. We denote its inverse map by \mathcal{T}^{t,s} . 3. For all times s , t and \tau in \mathbb{R} , \mathcal{T}^{s,t}=\mathcal{T}^{\tau,t}\circ\mathcal{T}^{s,\tau} . This property, demonstrated in Subsection 3.3 , arises from the fact that the Lax-Oleinik operator \mathcal{T} is both surjective and non-expanding on the \omega -limit sets of scalar maps, which are compact. It follows from a classical result that any surjective, non-expanding map on a compact set is a bijective isometry. It is a classical result that surjective, non-expanding maps on compact sets are bijective isometries. From this proposition, we establish that all non-wandering elements of \Omega(\mathcal{T}) can be associated with global viscosity solutions of the form u(t,x)=\mathcal{T}^{0,t}u(x):\mathbb{R}\times M\to\mathbb{R} , where the initial data is u(0,\cdot)=u . This reveals to be a characterization of non-wandering viscosity solutions. Indeed, if we denote by \mathcal{V}(\mathcal{T}) the set of the initial datum of global viscosity solutions i.e. defined for all real times, then we have the following result. Theorem 1.2 . A viscosity solution of the Hamilton-Jacobi equation ( 1.1 ) is global if and only if it is non-wandering. In other words, the following equality holds \Omega(\mathcal{T})=\mathcal{V}(\mathcal{T}) (1.3) The idea of the proof is to first show that non-wandering viscosity solutions correspond to bounded global viscosity solutions.