Efficiently Computing Eigenvalues of a Sparse Banded Matrix

Click For Summary
SUMMARY

The discussion focuses on efficiently computing eigenvalues and eigenvectors of a sparse penta-diagonal matrix representing a Hamiltonian. The user initially utilized the LAPACK routine dsyev for symmetric matrices but switched to dsbev for symmetric band matrices without significant improvement due to the distance of the outer bands. Recommendations include exploring the ARPACK library, although its documentation poses challenges. The consensus is that LAPACK is suitable for banded matrices, but may not fully leverage sparsity in this specific case.

PREREQUISITES
  • Understanding of Hamiltonian mechanics and matrix representation
  • Familiarity with LAPACK routines, specifically dsyev and dsbev
  • Knowledge of sparse matrix concepts and their computational advantages
  • Basic proficiency in using ARPACK for eigenvalue problems
NEXT STEPS
  • Research the implementation and usage of ARPACK for sparse eigenvalue problems
  • Explore advanced LAPACK functionalities for banded matrices
  • Study optimization techniques for sparse matrix computations
  • Investigate alternative libraries for eigenvalue computations, such as Eigen or SLEPc
USEFUL FOR

Researchers, physicists, and computational mathematicians working with Hamiltonians and seeking efficient algorithms for eigenvalue problems in sparse banded matrices.

Jimmy and Bimmy
Messages
3
Reaction score
0
I have a Hamiltonian represented by a penta-diagonal matrix

The first bands are directly adjascent to the diagonals. The other two bands are N places above and below the diagonal.

Can anyone recommend an efficient algorithm or routine for finding the eigenvalues and eigenvectors?
 
Physics news on Phys.org
DrClaude said:
Thanks for the reply.

I have been using dsyev from the lapack routine (which is for symmetric matrices). I switched to dsbev (which is for symmetric band matrices), but didn't see much improvement. This is because, even though I only have 5 bands (diagonal + 2 upper + 2 lower), the outer 2 bands are a good distance away, and lapack doesn't take advantage of spare matrices.

Someone recommended arpack, but the documentation can be hard to follow.
 

Similar threads

  • · Replies 33 ·
2
Replies
33
Views
2K
  • · Replies 22 ·
Replies
22
Views
4K
  • · Replies 5 ·
Replies
5
Views
2K
  • · Replies 4 ·
Replies
4
Views
3K
  • · Replies 1 ·
Replies
1
Views
2K
  • · Replies 2 ·
Replies
2
Views
2K
  • · Replies 1 ·
Replies
1
Views
3K
  • · Replies 5 ·
Replies
5
Views
2K
  • · Replies 11 ·
Replies
11
Views
2K
  • · Replies 4 ·
Replies
4
Views
5K