1: \begin{abstract}

2: Preconditioners are generally essential for fast convergence in the iterative solution

3: of linear systems of equations. However, the computation of a good preconditioner

4: can be expensive. So, while solving a sequence of many linear systems, it

5: is advantageous to recycle preconditioners, that is, update a previous

6: preconditioner and reuse the updated version. In this paper, we introduce a

7: simple and effective method for doing this.

8: We consider recycling preconditioners for both

9: the general case of sequences of linear systems

10: $\MA(\Vp_k) \Vx_k = \Vb_k$ as well as the

11: important special case

12: of the type $(s_k\ME + \MA)\Vx_k = \Vb_k$.

13: The right hand sides

14: may or may not change.

15: 

16: We update preconditioners by defining a map from a new matrix to a

17: previous matrix, for example, the first matrix in the sequence.  We then combine the preconditioner for this previous matrix with the map

18: to define the new preconditioner. This approach has several advantages.

19: {\em The update is independent from the original preconditioner, so

20: it can be applied to any preconditioner.}

21: The possibly high cost of an

22: initial preconditioner can be amortized over many linear solves.

23: The cost of updating the preconditioner is more or less constant and there is flexibility in balancing

24: the quality of the map with the computational cost.

25: 

26: In the numerical experiments section, we demonstrate good results for several

27: applications.

28: \end{abstract}

29: