Smooth approximation of stochastic differential equations (Q272965)

This paper considers approximations of stochastic differential equations (SDE). For a \(d\)-dimensional Itô process \(X\) expressed by the SDE NEWLINE\[NEWLINE dX_t = a(X) dt + b(X) dW_t NEWLINE\]NEWLINE with smooth functions \(a\), \(b\) and a \(d\)-dimensional Brownian motion \(W\), the authors consider a sequence \((X^n)_{n \geq 1}\) of ordinary differential equations NEWLINE\[NEWLINE dX^n_t = a(X^n) dt + b(X^n) dW^n_t, NEWLINE\]NEWLINE where \((W^n)_{n \geq 1}\) is a sequence of \(e\)-dimensional processes with smooth sample paths. Assuming the weak invariance principle (WIP) for the sequence \(W^n\), i.e., \(W^n \to_w W\) in \(C([0,T],\mathbb{R}^e)\), they ask whether \(X^n \to_w X\) in \(C([0,T],\mathbb{R}^d)\). The Wong-Zakai theorem gives general conditions which are satisfied for \(d=e=1\). The paper answers the question of the correct interpretation of the stochastic integral \(\int b(X) * dW\) in order to ensure \(X^n \to_w X\). For that, they first prove the discrete time iterated WIP by using a standard method of passing from invertible maps to noninvertible maps. Then they derive the continuous time WIP by methods of rough path theory and smooth ergodic theory. Moreover, they present a result on cohomological invariance of weak limits of iterated processes and give several generalizations of their main results.