[Paper Review] Local Time in Parisian Walkways

Jirô Akahori|ArXiv.org|Apr 18, 2006

Public Spaces through Art1 references3 citations

TL;DR

This paper establishes a discrete Itô and Tanaka formula for Parisian walks—symmetric random walks on the complex lattice ℤ[ζ], where ζ is a primitive third root of unity. It shows that the graph distance from a fixed set P, minus the local time (number of exits from three specific regions), yields a simple symmetric random walk, providing a discrete analogue to planar Brownian motion with explicit martingale and stochastic calculus structure.

ABSTRACT

In the present paper, Ito formula and Tanaka formula for a special kind of symmetric random walk in the complex plain are studied. The random walk is called Parisian walk, and its local time is defined to be the number of exit from some regions.

Motivation & Objective

• To develop a stochastic calculus framework for symmetric random walks on the Eisenstein integers ℤ[ζ], termed Parisian walks.
• To define and analyze local time as the number of exits from three specific closed regions (A̅₁, A̅₃, A̅₅) in the lattice.
• To establish a discrete Itô formula for complex-valued functions on the walk's state space.
• To derive a Tanaka-type formula that decomposes the change in graph distance into martingale and local time components.
• To show that the difference between graph distance and local time is a simple symmetric random walk, establishing a key structural identity.

Proposed method

• Define Parisian walks as ℤ[ζ]-valued martingales with increments in {1, ζ, ζ²}, each with probability 1/3.
• Introduce local time Lₜ as the count of exits from the closures A̅₁, A̅₃, A̅₅ up to time t.
• Prove a martingale representation property for (Z, Z̄), showing every complex martingale is a stochastic integral w.r.t. Z and Z̄.
• Derive a discrete Itô formula (Proposition 3) expressing f(Zₜ₊₁) − f(Zₜ) in terms of ΔZₜ, ΔZ̄ₜ, and a discrete Laplacian-like term.
• Establish a Tanaka formula (Proposition 4) showing that the increment of the graph distance ‖Zₜ‖ equals a real part of a complex martingale term plus the increment of local time.
• Use the Tanaka formula to prove Theorem 1: ‖Zₜ‖ − Lₜ is a simple symmetric random walk on ℤ.

Experimental results

Research questions

• RQ1 How can a stochastic calculus framework be developed for symmetric random walks on the Eisenstein integers ℤ[ζ]?
• RQ2 What is the correct definition of local time for such walks, and how does it relate to geometric structure?
• RQ3 Can a discrete Itô formula be derived for complex-valued functions on ℤ[ζ] under the Parisian walk dynamics?
• RQ4 Does a Tanaka-type formula exist that decomposes the change in graph distance into martingale and local time components?
• RQ5 Is the difference between graph distance and local time a simple symmetric random walk, and if so, why?

Key findings

• The graph distance from the set P = {0, 1, 1+ζ} minus the local time Lₜ is a simple symmetric random walk on ℤ.
• The local time Lₜ counts the number of exits from the three closed regions A̅₁, A̅₃, A̅₅, which are defined via unions of lattice rays and boundary sets.
• A discrete Itô formula is derived for any complex-valued function f on ℤ[ζ], expressing f(Zₜ₊₁) − f(Zₜ) in terms of ΔZₜ, ΔZ̄ₜ, and a discrete Laplacian term.
• A Tanaka formula is established: ‖Zₜ₊₁‖ − ‖Zₜ‖ = (2/3)Re[(1−ζ²)(φₜΔZₜ + ψₜΔZ̄ₜ)] + (Lₜ₊₁ − Lₜ), where φₜ and ψₜ are indicator-based processes.
• The martingale representation property holds for (Z, Z̄), meaning every complex-valued Fₜ-martingale is a stochastic integral w.r.t. Z and Z̄.
• The discrete Itô and Tanaka formulas converge to their continuous counterparts under appropriate scaling, confirming that Parisian walks are the discrete analogue of planar Brownian motion.