Rates of convergence to the local time of a diffusion (Q1264268)

The author considers the approximation of the local time \(L_{t}\) at level 0 of a one-dimensional diffusion process \(X\), by normalized sums \(U_{t}^{n}\) of functions of values \(X_{i/n}\), for \(i\leq nt\), as \(n\rightarrow\infty\). Under quite general conditions, it is proved that the processes \(\frac{u_{n}}{n}U^{n}\) converge in probability to the process \(cL\), were \(c\) is a real number, as soon as \(u_{n}/n\rightarrow 0\) and \(u_{n}\rightarrow\infty\) [see also \textit{J.-M. Azaïs}, Ann. Inst. Henri Poincaré, Probab. Stat. 25, No. 2, 175-194 (1989; Zbl 0674.60032), \textit{A. N. Borodin}, Probab. Theory Relat. Fields 72, 231-250 and 251-277 (1986; Zbl 0572.60078 and Zbl 0572.60079) and Russ. Math. Surv. 44, No. 2, 1-51 (1989); translation from Usp. Mat. Nauk 44, No. 2(266), 7-48 (1989; Zbl 0697.60080) and \textit{D. Florens-Zmirou}, J. Appl. Probab. 30, No. 4, 790-804 (1993; Zbl 0796.62070)]. Next, the author restricts to the case \(u_{n}=n^{\alpha}\), for some \(\alpha\in (0,1)\), and then a functional central limit theorem giving a mixed normal limiting value to the sequence of processes \(n^{\alpha}(U_{t}^{n}-L_{t})\), for a suitable value of \(\alpha\) is given. It is shown how to reduce the proofs to the case where \(X\) is a standard Brownian motion. Then the processes \(n^{\alpha} (U_{t}^{n}-L_{t})\) are written as sums of a sequence of martingales and a sequence of processes which go to 0. The problem is then reduced to a central limit theorem for a suitable martingale.