1: \begin{abstract}

2: This paper discusses the computation of derivatives for optimization

3: problems governed by linear hyperbolic systems of partial differential

4: equations (PDEs) that are discretized by the discontinuous Galerkin

5: (dG) method. An efficient and accurate computation of these

6: derivatives is important, for instance, in inverse problems

7: %

8: and optimal control problems.

9: %

10: This computation is usually based on an adjoint PDE system, and the

11: question addressed in this paper is how the discretization of this

12: adjoint system should relate to the dG discretization of the

13: hyperbolic state equation. Adjoint-based derivatives can either be computed

14: before or after discretization; these two options are often referred

15: to as the optimize-then-discretize and discretize-then-optimize

16: approaches. We discuss the relation between these two options for dG

17: discretizations in space and Runge--Kutta time integration.  The

18: influence of different dG formulations and of numerical quadrature is

19: discussed. Discretely exact discretizations for several hyperbolic

20: optimization problems are derived, including the advection equation,

21: Maxwell's equations and the coupled elastic-acoustic wave equation.

22: We find that the discrete adjoint equation inherits a natural dG

23: discretization from the discretization of the state equation and

24: that the expressions for the discretely exact gradient often have to

25: take into account contributions from element faces.  For the coupled

26: elastic-acoustic wave equation, the correctness and accuracy of our

27: derivative expressions are illustrated by comparisons with finite

28: difference gradients. The results show that a straightforward

29: discretization of the continuous gradient differs from the discretely

30: exact gradient, and thus is not consistent with the discretized

31: objective. This inconsistency may cause difficulties in the

32: convergence of gradient based algorithms for solving optimization

33: problems. 

34: 

35: 

36: %

37: %

38: %

39: 

40: 

41: \end{abstract}

42: