Convergence to a Lévy process in the Skorohod \(\mathcal{M}_1\) and \(\mathcal{M}_2\) topologies for nonuniformly hyperbolic systems, including billiards with cusps (Q1986583)

Sinai's work on dispersing billiards provided a piece of evidence that deterministic dynamical systems can exhibit a behavior that satisfies statistical limit laws. The current paper considers a class of billiards with cusps. The authors' results here generalize previous results by \textit{P. Jung} and \textit{H.-K. Zhang} [Ann. Henri Poincaré 19, No. 12, 3815--3853 (2018; Zbl 1407.37058)]. A motivating example uses a planar dispersing billiard with a cusp at a flat point. A billiard table \(Q\) in the plane has a boundary comprising a finite number of \(C^3\) curves \(\Gamma_i\) for \(i = 1, 2,\dots, n_0\) with \(n_0 \geq 3\) with a cusp formed by the two curves \(\Gamma_1\) and \(\Gamma_2\). In coordinates \((s,z)\), the cusp lies at \((0,0)\) and both \(\Gamma_1\) and \(\Gamma_2\) are tangent to the \(x\)-axis at \((0,0)\). Near \((0,0)\) \(\Gamma_1 = \{s, \beta^{-1} s^\beta\}\) and \(\Gamma_2 = \{s, -\beta^{-1} s^\beta\}\) where \(\beta > 2\). The phase space of the billiard map \(T\) is \(\Lambda = \partial Q \times\ [0, \pi \ ] \) with coordinates \((r, \theta)\) where \(r\) denotes the arc length along \(\partial Q\) and \(\theta\) is the angle between the tangent line of the boundary and the collision vector in the clockwise direction. There is a natural ergodic invariant probability measure on the phase space \(\Lambda:d\mu = (2 |\partial Q|^{-1}) \sin\theta dr d\theta\) where \(|\partial Q|\) is the length of the boundary of \(Q\). {P. Jung} and {H.-K. Zhang} investigated a class of billiards examples where the billiards with cusps had vanishing curvature at the cusp. They proved that the dynamics converge to a totally skewed \(\alpha\)-stable statistical limit law with \(\alpha \in (1,2)\). In this paper the authors extend the above results and show how to get from the stable law to the weak invariance principle that means weak convergence to the corresponding \(\alpha\)-stable Lévy process. Part of what makes this possible is the recognition of the appropriate topology for convergence. The authors use the \(\mathcal{M} _1\) and \(\mathcal{M}_2\) Skorohod topologies because the standard Skorohod topology is too strong for the class of function spaces having functions with the variety of jump discontinuities encountered here.