1: \begin{abstract}

2: 

3: 

4: 

5: We consider  $N\times N$ Hermitian Wigner random matrices 

6: $H$ where the probability density

7: for each matrix element is given by the density $\nu(x)= e^{- U(x)}$. 

8: We prove that the eigenvalue statistics in the bulk is given by  

9: Dyson sine kernel provided that $U \in C^6(\RR)$ 

10: with at most polynomially growing derivatives and

11: $\nu(x) \le C\, e^{ - C |x|}$ for $x$ large. 

12: The proof is based upon an approximate time reversal of 

13: the Dyson Brownian motion combined with the convergence of

14: the eigenvalue density to the Wigner semicircle law on

15: short scales.

16: 

17: 

18: 

19: 

20: \end{abstract}

21: