1: \begin{abstract}

2: The optimal rate of convergence of estimators of the integrated

3: volatility, for a discontinuous It\^o semimartingale sampled at regularly

4: spaced times and over a fixed time interval, has been a long-standing

5: problem, at least when the jumps are not summable. In this paper, we

6: study this optimal rate, in the minimax sense and for appropriate

7: ``bounded'' nonparametric classes of semimartingales. We show that,

8: if the $r$th powers of the jumps are summable for some $r\in[0,2)$,

9: the minimax rate is equal to $\min(\sqrt{n},(n\log n)^{(2-r)/2})$, where

10: $n$ is the number of observations.

11: \end{abstract}