Scilab Function
Last update : 31/10/2006

dnaupd - Interface for the Implicitly Restarted Arnoldi Iteration, to compute approximations to a few eigenpairs of a real linear operator

Calling Sequence

[IDO,RESID,V,IPARAM,IPNTR,WORKD,WORKL,INFO] = dsaupd(ID0,BMAT,N,WHICH,NEV,TOL,RESID,NCV,V,IPARAM,IPNTR,WORKD,WORKL,INFO)

Parameters

Description

Reverse communication interface for the Implicitly Restarted Arnoldi iteration. This subroutine computes approximations to a few eigenpairs of a linear operator "OP" with respect to a semi-inner product defined by a symmetric positive semi-definite real matrix B. B may be the identity matrix. NOTE: If the linear operator "OP" is real and symmetric with respect to the real positive semi-definite symmetric matrix B, i.e. B*OP = (OP`)*B, then subroutine dsaupd should be used instead.

The computed approximate eigenvalues are called Ritz values and the corresponding approximate eigenvectors are called Ritz vectors.

dnaupd is usually called iteratively to solve one of the following problems:

Mode 1: A*x = lambda*x. OP = A , B = I .

Mode 2: A*x = lambda*M*x, M symmetric positive definite OP = inv[M]*A, B = M . (If M can be factored see remark 3 below)

Mode 3: A*x = lambda*M*x, M symmetric positive semi-definite. OP = Real_Part{ inv[A - sigma*M]*M }, B = M . shift-and-invert mode (in real arithmetic) If OP*x = amu*x , then amu = 1/2 * [ 1/(lambda-sigma) + 1/(lambda-conjg(sigma))] . Note: If sigma is real, i.e. imaginary part of sigma is zero; Real_Part{ inv[A - sigma*M]*M } == inv[A - sigma*M]*M amu == 1/(lambda-sigma) .

Mode 4: A*x = lambda*M*x, M symmetric semi-definite OP = Imaginary_Part{ inv[A - sigma*M]*M } , B = M . shift-and-invert mode (in real arithmetic) If OP*x = amu*x , then amu = 1/2i * [ 1/(lambda-sigma) - 1/(lambda-conjg(sigma)) ] .

Both mode 3 and 4 give the same enhancement to eigenvalues close to the (complex) shift sigma. However, as lambda goes to infinity, the operator OP in mode 4 dampens the eigenvalues more strongly than does OP defined in mode 3.

NOTE: The action of w <- inv[A - sigma*M]*v or w <- inv[M]*v should be accomplished either by a direct method using a sparse matrix factorization and solving [A - sigma*M]*w = v or M*w = v , or through an iterative method for solving these systems. If an iterative method is used, the convergence test must be more stringent than the accuracy requirements for the eigenvalue approximations.

Remarks

1. The computed Ritz values are approximate eigenvalues of OP. The selection of WHICH should be made with this in mind when Mode = 3 and 4. After convergence, approximate eigenvalues of the original problem may be obtained with the ARPACK subroutine dneupd. 2. If a basis for the invariant subspace corresponding to the converged Ritz values is needed, the user must call dneupd immediately following completion of dnaupd. This is new starting with release 2 of ARPACK. 3. If M can be factored into a Cholesky factorization M = LL` then Mode = 2 should not be selected. Instead one should use Mode = 1 with OP = inv(L)*A*inv(L`). Appropriate triangular linear systems should be solved with L and L` rather than computing inverses. After convergence, an approximate eigenvector z of the original problem is recovered by solving L`z = x where x is a Ritz vector of OP. 4. At present there is no a-priori analysis to guide the selection of NCV relative to NEV. The only formal requrement is that NCV > NEV + 2. However, it is recommended that NCV >= 2*NEV+1. If many problems of the same type are to be solved, one should experiment with increasing NCV while keeping NEV fixed for a given test problem. This will usually decrease the required number of OP*x operations but it also increases the work and storage required to maintain the orthogonal basis vectors. The optimal "cross-over" with respect to CPU time is problem dependent and must be determined empirically. See Chapter 8 of Reference 2 for further information. 5. When IPARAM(1) = 0, and IDO = 3, the user needs to provide the NP = IPARAM(8) real and imaginary parts of the shifts in locations


   real part                  imaginary part
   -----------------------    -----------------------
1  WORKL(IPNTR(14))           WORKL(IPNTR(14)+NP)
2  WORKL(IPNTR(14)+1)         WORKL(IPNTR(14)+NP+1)

NP  WORKL(IPNTR(14)+NP-1)     WORKL(IPNTR(14)+2*NP-1).

		

Only complex conjugate pairs of shifts may be applied and the pairs must be placed in consecutive locations. The real part of the eigenvalues of the current upper Hessenberg matrix are located in WORKL(IPNTR(6)) through WORKL(IPNTR(6)+NCV-1) and the imaginary part in WORKL(IPNTR(7)) through WORKL(IPNTR(7)+NCV-1). They are ordered according to the order defined by WHICH. The complex conjugate pairs are kept together and the associated Ritz estimates are located in WORKL(IPNTR(8)), WORKL(IPNTR(8)+1), ... , WORKL(IPNTR(8)+NCV-1).

See Also

dsaupd ,  

Authors

Danny Sorensen, Richard Lehoucq, Phuong Vu CRPC / Rice University Applied Mathematics Rice University Houston, Texas

Bibliography

1. D.C. Sorensen, "Implicit Application of Polynomial Filters in a k-Step Arnoldi Method", SIAM J. Matr. Anal. Apps., 13 (1992), pp 357-385.

2. R.B. Lehoucq, "Analysis and Implementation of an Implicitly Restarted Arnoldi Iteration", Rice University Technical Report TR95-13, Department of Computational and Applied Mathematics.

3. B.N. Parlett, "The Symmetric Eigenvalue Problem". Prentice-Hall, 1980.

4. B.N. Parlett, B. Nour-Omid, "Towards a Black Box Lanczos Program", Computer Physics Communications, 53 (1989), pp 169-179.

5. B. Nour-Omid, B.N. Parlett, T. Ericson, P.S. Jensen, "How to Implement the Spectral Transformation", Math. Comp., 48 (1987), pp 663-673.

6. R.G. Grimes, J.G. Lewis and H.D. Simon, "A Shifted Block Lanczos Algorithm for Solving Sparse Symmetric Generalized Eigenproblems", SIAM J. Matr. Anal. Apps., January (1993).

7. L. Reichel, W.B. Gragg, "Algorithm 686: FORTRAN Subroutines for Updating the QR decomposition", ACM TOMS, December 1990, Volume 16 Number 4, pp 369-377.

8. R.B. Lehoucq, D.C. Sorensen, "Implementation of Some Spectral Transformations in a k-Step Arnoldi Method". In Preparation.

Used Function

Based on ARPACK routine dsaupd