1: \begin{abstract}

2: GMRES is a popular Krylov subspace method for solving linear systems 

3: of equations involving a general non-Hermitian coefficient matrix.

4: The conventional bounds on GMRES convergence involve 

5: polynomial approximation problems in the complex plane.  

6: Three popular approaches pose this approximation problem on the spectrum, 

7: the field of values, or pseudospectra of the coefficient matrix. 

8: We analyze and compare these bounds,

9: illustrating with six examples the success and failure of each.

10: When the matrix departs from normality due only to a low-dimensional 

11: invariant subspace, we discuss how these bounds can be adapted 

12: to exploit this structure.

13: Since the Arnoldi process that underpins GMRES provides approximations 

14: to the pseudospectra, one can estimate the GMRES convergence bounds 

15: as an iteration proceeds.

16: \end{abstract}

17: