Euler scheme and tempered distributions (Q850027)

Consider smooth \(\mathbb R^d\)-valued diffusion processes \((X_t^x)_{t \in [0,1]}\) and its approximation by equidistant Euler method \((Y_{t,n}^x)_{n \geq 1}\) started at point \(x \in \mathbb R^d\) on \([0,1]\). If \(X\) is uniformly elliptic, the author proves that there is a constant \(C_1(f(x))\) such that \[ \mathbb E [ f(X_1^x) ] - \mathbb E [ f(Y_{1,n}^x) ] = \frac{C_1(f(x))}{n} + O \biggl(\frac{1}{n^2}\biggr) \] for all tempered distributions \(f\), i.p. for measurable functions \(f\) with polynomial or exponential growth, Dirac or any derivative of Dirac mass. A motivation with applications to option pricing, hedging, rates of prices, deltas and gammas in finance is given.