Optimal total variation bounds for stochastic differential delay equations with small noises (Q6123412)

Letting \(\varepsilon\in(0,1)\), the authors consider solutions to stochastic differential delay equations of the form \[ X_{\varepsilon,t}=\varphi(0)+\int_{0}^tb(X_{\varepsilon,s},X_{\varepsilon,s-\tau})\text{d}s+\varepsilon\int_{0}^t\sigma(X_{\varepsilon,s},X_{\varepsilon,s-\tau})\text{d}B_s\,, \] for \(t\in[0,T]\) and with \(X_{\varepsilon,t}=\varphi(t)\) for \(t\in[-\tau,0]\), where \(\varphi:[-\tau,0]\to\mathbb{R}\) is a bounded deterministic function, \((B_t)_{t\in[0,T]}\) is a standard Brownian motion, and \(b,\sigma:\mathbb{R}^2\to\mathbb{R}\) are twice differentiable deterministic functions with bounded partial derivatives. Letting \(\tilde{X}_{\varepsilon,t}\) denote a suitably normalised version of \(X_{\varepsilon,t}\), the authors establish upper bounds on the total variation distance between \(\tilde{X}_{\varepsilon,t}\) and an appropriate Gaussian random variable \(Y_t\). These bounds converge to zero as \(\varepsilon\to0\) at rate \(O(\varepsilon)\). The authors further show that this rate is optimal. The proofs make use of tools from Malliavin calculus.