Smooth approximation of stochastic differential equations

Annals of Probability 44 (2016) 479-520.

David Kelly and Ian Melbourne


Abstract Consider an Itô process X satisfying the stochastic differential equation dX=a(X)dt+b(X)dW where a,b are smooth and W is a multidimensional Brownian motion. Suppose that Wn has smooth sample paths and that Wn converges weakly to W. A central question in stochastic analysis is to understand the limiting behaviour of solutions Xn to the ordinary differential equation dXn=a(Xn)dt+b(Xn)dWn.

The classical Wong-Zakai theorem gives sufficient conditions under which Xn converges weakly to X provided that the stochastic integral ∫ b(X)dW is given the Stratonovich interpretation. The sufficient conditions are automatic in one dimension, but in higher dimensions the correct interpretation of ∫ b(X)dW depends sensitively on how the smooth approximation Wn is chosen. In applications, a natural class of smooth approximations arise by setting $Wn(t)=n-1/20ntv o φs ds$ where φt is a flow (generated for instance by an ordinary differential equation) and v is a mean zero observable. Under mild conditions on φt we give a definitive answer to the interpretation question for the stochastic integral ∫ b(X)dW. Our theory applies to Anosov or Axiom A flows φt, as well as to a large class of nonuniformly hyperbolic flows (including the one defined by the well-known Lorenz equations) and our main results do not require any mixing assumptions on φt. The methods used in this paper are a combination of rough path theory and smooth ergodic theory.


pdf file

Typos, etc: The proofs for moments and iterated moments in Proposition 7.1 are correct for noninvertible maps but not for invertible maps since the martingales are not adapted. Also, the results for iterated moments are suboptimal.

Optimal results for moments for invertible maps were obtained in Proposition 2.6 of Demers, Melbourne & Nicol, 2020. Proposition 2.7 of that paper gives a correct but still suboptimal result for iterated moments for invertible maps.

Optimal results for iterated moments for noninvertible systems were obtained in Theorem 2.4 of Korepanov, Kosloff & Melbourne, 2022.

The optimal result for iterated moments for invertible systems is much more subtle and was eventually obtained by Nicholas Fleming-Vázquez in this paper: Functional correlation bounds and optimal iterated moment bounds for slowly-mixing nonuniformly hyperbolic maps. Commun. Math. Phys. 391 (2022) 173-198.

The indices β and γ in various formulas for modified drift (Theorem 1.2 for instance) have been interchanged. The correct formulas are given in Galton & Melbourne, 2022.

For some reason, the inducing step for flows in Theorem 6.1 is written inaccurately. In particular, the application of Eagleson/Zweimüller is done incorrectly. It should be done as in the similar step in Melbourne & Török, 2004.