1: \begin{abstract}

2: 

3: The calculation of a segment of eigenvalues and their corresponding eigenvectors of a Hermitian matrix or

4: matrix pencil has many applications. 

5: A new density-matrix-based algorithm has been

6: proposed recently and a software package FEAST has been developed. The density-matrix approach

7: allows FEAST's implementation to exploit a key strength of modern computer architectures, namely,

8: multiple levels of parallelism. 

9: Consequently, the software package has been well received, especially

10: in the electronic structure community. Nevertheless, theoretical analysis of

11: FEAST has lagged. 

12: For instance, the FEAST algorithm has not been proven to converge.

13: This paper offers a detailed 

14: numerical analysis of FEAST. In particular, we show that the FEAST algorithm can be understood as 

15: an accelerated subspace iteration algorithm in conjunction with the Rayleigh-Ritz procedure.

16: The novelty of FEAST 

17: lies in its accelerator which is a rational matrix function that approximates the spectral projector

18: onto the eigenspace in question.

19: Analysis of the numerical nature of this approximate spectral projector and the resulting subspaces generated in the

20: FEAST algorithm establishes the algorithm's convergence. 

21: This paper shows that FEAST is resilient against rounding errors and establishes properties that can

22: be leveraged to enhance the algorithm's robustness.

23: Finally, we propose an extension of FEAST to handle non-Hermitian problems and

24: suggest some future research directions.

25: 

26: 

27: \end{abstract}

28: