1: \begin{abstract}

2: The numerical approximation of solutions to

3: stochastic partial differential equations

4: with additive spatial white noise

5: on bounded domains in $\mathbb{R}^d$ is considered.

6: The differential operator is given by the

7: fractional power~$L^\beta$, $\beta\in(0,1)$,

8: of an integer order elliptic differential operator $L$

9: and is therefore non-local.

10: %

11: Its inverse

12: $L^{-\beta}$

13: is represented

14: by a Bochner integral

15: from the Dunford--Taylor functional calculus.

16: %

17: By applying a quadrature formula

18: to this integral representation,

19: the inverse fractional power operator $L^{-\beta}$

20: is approximated by a weighted sum of

21: non-fractional

22: resolvents

23: $(I + t_j^2 L)^{-1}$

24: at certain quadrature nodes $t_j>0$.

25: %

26: The resolvents are then discretized

27: in space by a

28: standard finite element method.

29: 

30: This approach is

31: combined with an

32: approximation of the

33: white noise,

34: which is based only

35: on the mass matrix

36: of the finite element discretization.

37: In this way,

38: an efficient

39: numerical algorithm

40: for computing samples

41: of the approximate solution

42: is obtained.

43: %

44: For the resulting approximation,

45: the strong mean-square error is analyzed

46: and an explicit rate of convergence

47: is derived.

48: Numerical experiments for $L=\kappa^2-\Delta$,

49: $\kappa > 0$, with homogeneous Dirichlet boundary conditions

50: on the unit cube $(0,1)^d$

51: in $d=1,2,3$ spatial dimensions

52: for

53: varying $\beta\in(0,1)$

54: attest the theoretical results.

55: \end{abstract}