1: \begin{abstract}

2: 	We analyze a goal-oriented adaptive

3: 	algorithm that aims to efficiently

4: 	compute the quantity of interest \(G(u^\star)\) with a

5: 	linear goal functional \(G\) and the solution

6: 	\(u^\star\) to a general second-order nonsymmetric

7: 	linear elliptic partial differential equation. The

8: 	current state of the analysis of iterative algebraic

9: 	solvers for nonsymmetric systems lacks the contraction

10: 	property in the norms that are prescribed by the

11: 	functional ana\-lytic setting. This seemingly prevents

12: 	their application in the optimality analysis

13: 	of goal-oriented adaptivity. As a remedy, this paper

14: 	proposes a

15: 	goal-oriented adaptive iteratively symmetrized finite

16: 	element method (GOAISFEM). It employs a nested loop

17: 	with

18: 	a contractive symmetrization procedure, e.g., the

19: 	Zarantonello iteration, and a contractive algebraic

20: 	solver, e.g., an optimal multigrid solver.

21: 	The various

22: 	iterative procedures require well-designed stopping

23: 	criteria such that the adaptive algorithm can

24: 	effectively steer the local mesh refinement and the

25: 	computation of the

26: 	inexact discrete approximations. The main results

27: 	consist of full linear convergence of the proposed

28: 	adaptive algorithm and the proof of optimal

29: 	convergence rates with respect to both degrees of

30: 	freedom and total computational cost (i.e., optimal

31: 	complexity). Numerical experiments confirm the

32: 	theoretical results and investigate the selection of the

33: 	parameters.

34: \end{abstract}

35: