1: \begin{abstract}

2: Consider solving large sparse range symmetric 

3: singular linear systems $A\vector{x} = \vector{b}$ which arise, for instance, in the discretization 

4: of convection diffusion equations with periodic boundary conditions, and partial differential equations for electromagnetic fields using the edge-based finite element method.

5: 

6: In theory, the Generalized Minimal Residual (GMRES) method converges to the least squares solution for

7: inconsistent systems if the coefficient matrix $A$ is range symmetric, 

8: i.e. $ {\rm R}(A)= {\rm R}(A^{ \rm T } )$, where 

9: $ {\rm R}(A)$ is the range space of $A$.

10: 

11: We derived the necessary and sufficient conditions for GMRES to determine a least squares solution of inconsistent

12: and consistent range symmetric systems assuming exact arithmetic except for the computation of the elements

13: of the Hessenberg matrix.

14: 

15: In practice, GMRES may not converge due to numerical instability. In order to improve the convergence, 

16: we propose using the pseudoinverse

17: for the solution of the severely ill-conditioned Hessenberg systems in GMRES.

18: Numerical experiments on 

19: %% 22.04.06semi-definite 

20: inconsistent systems indicate that the method is effective and robust.

21: Finally, we further improve the convergence of the method by reorthogonalizing the Modified Gram-Schmidt procedure. 

22: \end{abstract}

23: