1: \begin{abstract}

2: The concept of qualification for spectral regularization methods

3: (SRM) for inverse ill-posed problems is strongly associated to the

4: optimal order of convergence of the regularization error

5: (\cite{bookEHN}, \cite{Mathe2004}, \cite{ref:Mathe-Pereverzev-2003},

6: \cite{Vainikko-1982}). In this article, the definition of

7: qualification is extended and three different levels are introduced:

8: weak, strong and optimal.

9:  It is shown that the weak qualification extends the

10: definition introduced by Math\'{e} and Pereverzev

11: (\cite{ref:Mathe-Pereverzev-2003}), mainly in the sense that the

12: functions associated to orders of convergence and source sets need

13: not be the same.  It is shown that certain methods possessing

14: infinite classical qualification, e.g. truncated singular value

15: decomposition (TSVD), Landweber's method and Showalter's method,

16: also have generalized qualification leading to an optimal order of

17: convergence of the regularization error.

18: %

19: Sufficient conditions for a SRM to have weak qualification are

20: provided and necessary and sufficient conditions for a given order

21: of convergence to be strong or optimal qualification are found.

22: Examples of all three qualification levels are provided and the

23: relationships between them as well as with the classical concept

24: of qualification  and the qualification introduced in

25: \cite{ref:Mathe-Pereverzev-2003} are shown. In particular, SRMs

26: having extended qualification in each one of the three levels and

27: having zero or infinite classical qualification are presented.

28: Finally several implications of this theory in the context of

29: orders of convergence, converse results and maximal source sets

30: for inverse ill-posed problems, are shown.

31: \end{abstract}

32: