[Paper Review] Asymptotic expansion of the density for hypoelliptic rough differential equation

Yuzuru Inahama, Nobuaki Naganuma|arXiv (Cornell University)|Feb 14, 2019

Stochastic processes and financial applicationsEconomics, Econometrics and Finance22 references3 citations

TL;DR

This paper establishes a full short-time asymptotic expansion of the transition density for hypoelliptic rough differential equations driven by fractional Brownian motion with Hurst parameter $ H \in (1/4, 1/2] $. Using Watanabe's distributional Malliavin calculus under Hörmander's condition, it derives a fractional version of Ben Arous' off-diagonal asymptotics, extending prior results by removing ellipticity assumptions and generalizing to the rough path framework.

ABSTRACT

We study a rough differential equation driven by fractional Brownian motion with Hurst parameter $H$ $(1/4

Motivation & Objective

• To extend short-time density asymptotics for rough differential equations beyond the elliptic case to hypoelliptic settings under Hörmander's condition.
• To establish a full asymptotic expansion of the transition density for fractional Brownian motion-driven SDEs with $ H \in (1/4, 1/2] $.
• To generalize Ben Arous' off-diagonal asymptotics to the fractional Brownian motion setting using Malliavin calculus.
• To refine and extend previous results in [Ina16b] by removing ellipticity assumptions and working under the more general Hörmander condition.
• To provide a rigorous asymptotic expansion framework compatible with rough path theory and distributional Malliavin calculus.

Proposed method

• Utilizes Watanabe's distributional Malliavin calculus to analyze the density of solutions to rough differential equations.
• Applies the Lyons-Itô map and its Taylor-like expansion in both deterministic and probabilistic settings to control pathwise behavior.
• Employs Kusuoka-Stroock-type estimates and uniform non-degeneracy of the Malliavin covariance matrix for the scaled-shifted RDE.
• Implements Young integration and Besov rough path techniques, particularly for the third-level iterated integral.
• Establishes moment estimates for the inverse of the smallest eigenvalue of the Malliavin matrix under scaling.
• Relies on assumptions (A1) and (A2): Hörmander's bracket-generating condition and a unique minimizer for the energy functional.

Experimental results

Research questions

• RQ1 Can a full short-time asymptotic expansion of the transition density be derived for hypoelliptic rough differential equations driven by fractional Brownian motion with $ H \in (1/4, 1/2] $?
• RQ2 How does the asymptotic density expansion behave under Hörmander's condition rather than ellipticity?
• RQ3 To what extent can Ben Arous' off-diagonal asymptotics for diffusions be generalized to the fractional Brownian motion setting?
• RQ4 What are the necessary and sufficient conditions on the vector fields and the path to ensure smooth density and its asymptotic expansion?
• RQ5 How can Malliavin calculus be adapted to handle the non-Markovian and rough nature of fractional Brownian motion in the short-time regime?

Key findings

• The paper establishes a full short-time asymptotic expansion of the density $ p_t(a,a') $ as $ t \searrow 0 $ for $ a \neq a' $, valid under Hörmander’s condition and the unique minimizer assumption.
• The main result (Theorem 2.3) recovers Ben Arous' classical result for $ H = 1/2 $, confirming consistency with the standard diffusion case.
• The expansion is derived using Watanabe’s distributional Malliavin calculus, avoiding the need for ellipticity and instead relying on Hörmander’s bracket-generating condition.
• A Kusuoka-Stroock-type estimate (Proposition 5.3) is proven for the Malliavin matrix, ensuring uniform non-degeneracy under scaling.
• The uniform non-degeneracy of the scaled-shifted RDE (Proposition 5.4) is established, crucial for controlling the density near the diagonal.
• The proof introduces novel techniques in Besov rough path theory, particularly for the third-level iterated integral, extending previous results in [Ina16b].