Weak KAM theory for action minimizing random walks

DOI10.1007/S00526-021-02025-2MaRDI QIDQ2048897zbMATH OpenOpenAlexWikidataFDO

Publication date 24 August 2021

Published in Calculus of Variations and Partial Differential Equations (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/2010.12193

hyperbolic scaling limit KAM theory random walks on a grid

Random walks on graphs (05C81) Random iteration (37H12) Perturbations of finite-dimensional Hamiltonian systems, normal forms, small divisors, KAM theory, Arnol'd diffusion (37J40) Action-minimizing orbits and measures for finite-dimensional Hamiltonian and Lagrangian systems; variational principles; degree-theoretic methods (37J51) Sums of independent random variables; random walks (60G50)

Abstract: We introduce a class of controlled random walks on a grid in

m a t h b b T^{d}

and investigate global properties of action minimizing random walks for a certain action functional together with Hamilton-Jacobi equations on the grid. This yields an analogue of weak KAM theory, which recovers a part of original weak KAM theory through the hyperbolic scaling limit.

Review text

The author describes a class of random walks on a grid on the \(d\)-dimensional torus \(\mathbb{T}^d\) and investigates global properties of action-minimizing random walks with an action functional plus the Hamiltonian-Jacobi equations on the grid. The goal of the author is to establish an analog of weak KAM theory in this setting. He shows that this recovers part of the original weak KAM theory in the hyperbolic scaling limit. The setting is as follows. \(G_h\) is the grid with small unit length \(0 < h << 1\) in \(\mathbb{T}^d\). So \(G_h = h \mathbb{Z}^d \cap [0,1]^d\) where \(h^{-1}\) is an even integer and 0 and 1 are identified. If \(\{e_i\}\) is the standard basis of \(\mathbb{R}^d\) and \(B = \{ \pm {e_i} \}\), the jump process is defined for \(x \in G_h\) to one of the points \(x + h \omega\) where \(\omega \in B\) and the transition probability is \(\rho(\omega)\). With iteration the jump process gives rise to a random walk with paths \(\gamma^0 = x \in G_h\), and \(\gamma^{k+1} = \gamma^k + h \omega\) on the \(k^{th}\) iteration. The author associates \(k\) with time so that the jump process occurs within a small unit time \(0 < \tau <<1\) where \(\tau^{-1}\) is an even integer. Inhomogeneous time 1-periodic transition probabilities are then introduced via a given function: \(\xi: G_h \times \{\tau k\} \rightarrow \mathbb{R}^d\) with \(k\) and integer and \(\xi(\cdot, t+1) = \xi(\cdot, t)\). The transition probabilities are then \(\rho^{\pm} (x,t,\omega) = \frac{1}{2d} \pm \frac{\tau}{2h} \xi(x,t) \cdot \omega\) For \(\rho^+\) this is the transition probability from \((x,t)\) to \((x + h \omega, t + \tau)\), and for \(\rho^-\) from \((x,t)\) to \((x + h\omega, t - \tau)\). The author regards each function \(\xi\) as a control and creates cost functionals \(\mathcal{L}\) that depend on a Tonelli Hamiltonian \(L : \mathbb{T}^d \times \mathbb{T} \times \mathbb{R}^d \rightarrow \mathbb {R}\), and given functions \(v^\pm: G_h \times \{\tau k\} \rightarrow \mathbb{R}\) with \(v^\pm(\cdot, t+1) = v^\pm (\cdot, t)\). The author finds families of \(v^\pm\) via equations of Hamilton-Jacobi type on the grid and then looks at properties of the maximizers and minimizers of the \(\mathcal{L}\) functionals. He also looks at the asymptotic behavior of the random walks in the context of weak KAM theory and Aubry-Mather theory. An essential tool in his work is the value function described in [\textit{K. Soga}, Math. Comput. 89 (323), 1135--159 (2020)].

Recommendations

Cites work

Cited in

(7)

This page was built for publication: Weak KAM theory for action minimizing random walks

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2048897)