1: \begin{abstract}

2:   We study a numerical method to compute probability

3:   density functions of solutions of stochastic differential

4:   equations. The method is sometimes called the numerical path integration

5:   method and has been shown to be fast and accurate in application

6:   oriented fields. In this paper we provide a rigorous analysis of

7:   the method that covers systems of equations with unbounded coefficients.

8: %  This method produces approximate probability density functions for the

9: %  solutions of the SDEs.

10: %, and enjoys several advantages compared to

11: %  path-wise approximation schemes.

12: Working in a natural space for densities, $L^1$, we obtain

13: stability, consistency, and new convergence results for the method, new

14: well-posedness and

15:   semigroup generation results for the related Fokker-Planck-Kolmogorov

16:   equation, and a new and rigorous connection to the corresponding probability

17:   density functions for both the approximate and the exact

18:   problems.

19: %Convergence of the path integration method corresponds to

20: %  convergence of probability densities of the path-wise  schemes

21: %  like Euler-Maruyama,

22: %  and our result gives convergence in $L^1$ uniformly in

23: %  (finite) time.

24: %For the Ornstein-Uhlenbeck process, we compute

25: %  explicitly the kernels and discuss under which assumptions there will be a

26: %  rate or merely convergence with no rate.

27: To prove the results we combine semigroup and PDE arguments in a new

28: way that should be of independent interest.

29: % On the technical side we

30: %  use the concept of dissipative operators, semigroup, PDE, and

31: %  numerical analysis techniques.

32: \end{abstract}