1: \begin{abstract}

2: We propose an unconditionally robust and highly effective preconditioner for general symmetric positive definite (SPD)

3: matrices based on structured

4: incomplete factorization (SIF), called enhanced SIF (eSIF) preconditioner. The

5: original SIF strategy proposed recently derives a structured preconditioner by applying block diagonal

6: preprocessing to the matrix and then compressing

7: appropriate scaled off-diagonal blocks. Here, we use an enhanced

8: scaling-and-compression strategy to design the new eSIF preconditioner. Some

9: subtle modifications are made, such as the use of two-sided block triangular

10: preprocessing. A practical multilevel eSIF scheme is then designed. We give

11: rigorous analysis for both the enhanced scaling-and-compression strategy and

12: the multilevel eSIF preconditioner. The new eSIF framework has some

13: significant advantages and overcomes some major limitations of the SIF strategy. (i) With the same

14: tolerance for compressing the off-diagonal blocks,

15: the eSIF preconditioner can approximate the original matrix to a much higher

16: accuracy. (ii) The new preconditioner leads to much more significant

17: reductions of condition numbers due to an accelerated magnification effect for

18: the decay in the singular values of the scaled off-diagonal blocks. (iii) With

19: the new preconditioner, the eigenvalues of the preconditioned matrix are much

20: better clustered around $1$. (iv) The multilevel eSIF preconditioner is

21: further unconditionally robust or is guaranteed to be positive definite

22: without the need of extra stabilization, while the multilevel SIF

23: preconditioner has a strict requirement in order to preserve positive

24: definiteness. Comprehensive numerical tests are used to show the advantages of

25: the eSIF preconditioner in accelerating the convergence of iterative solutions.

26: \end{abstract}

27: