Matrix: Sinclair/3Dspectralwave2

Description: 3-D spectral-element elastic wave modelling in freq. domain, C. Sinclair, Univ. Adelaide

Sinclair/3Dspectralwave2 graph
(undirected graph drawing)


Sinclair/3Dspectralwave2

  • Home page of the UF Sparse Matrix Collection
  • Matrix group: Sinclair
  • Click here for a description of the Sinclair group.
  • Click here for a list of all matrices
  • Click here for a list of all matrix groups
  • download as a MATLAB mat-file, file size: 73 MB. Use UFget(1857) or UFget('Sinclair/3Dspectralwave2') in MATLAB.
  • download in Matrix Market format, file size: 59 MB.
  • download in Rutherford/Boeing format, file size: 52 MB.

    Matrix properties
    number of rows292,008
    number of columns292,008
    nonzeros12,935,272
    structural full rank?yes
    structural rank292,008
    # of blocks from dmperm1
    # strongly connected comp.1
    explicit zero entries1,387,472
    nonzero pattern symmetrysymmetric
    numeric value symmetrysymmetric
    typecomplex
    structureHermitian
    Cholesky candidate?yes
    positive definite?no

    authorC. Sinclair
    editorT. Davis
    date2007
    kindmaterials problem
    2D/3D problem?yes

    Additional fieldssize and type
    bsparse 292008-by-1
    shiftsparse 292008-by-292008

    Notes:

    The A matrix is produced using 3-D spectral-element elastic wave modelling in
    the frequency domain.The medium is homogeneous and isotropic with elastic    
    coefficients: c11 = 6.30, c44 = 1.00. The B matrix contains only one non-zero
    entry, representing a real y-directed source, placed approximately in the    
    centre.  The model size in elements is 10x10x10. Each element is 1m x1m x 1m.
    Each element is a 4x4x4 Gauss-Lobbato-Legendre mesh, so the height, width and
    depth of the system is 31 nodes. There are 3 unknown complex components at   
    each node - the x, y and z displacements. The A matrix therefore has         
    dimension 89373 x 89373.  ((10 x 4) - (10 - 1))^3 * 3 = 89373.  The solution 
    will consist of x-z planes.  Note that A is complex and b is sparse and real 
    (b has a single nonzero).                                                    
                                                                                 
    The A matrix was provided with a nonzero imaginary part, but was otherwise   
    complex Hermitian.  To save space in the Matrix Market and Rutherford/Boeing 
    formats, the A matrix here has had this imaginary diagonal removed.  The     
    shift can be found in the aux.shift auxiliary matrix.  To reproduce the      
    original A matrix, use A = Problem.A + Problem.aux.shift ;                   
    

    Ordering statistics:result
    nnz(chol(P*(A+A'+s*I)*P')) with AMD2,070,437,023
    Cholesky flop count4.2e+13
    nnz(L+U), no partial pivoting, with AMD4.140582e+09
    nnz(V) for QR, upper bound nnz(L) for LU, with COLAMD3.742234e+09
    nnz(R) for QR, upper bound nnz(U) for LU, with COLAMD7.912859e+09

    Note that all matrix statistics (except nonzero pattern symmetry) exclude the 1387472 explicit zero entries.

    Maintained by Tim Davis, last updated 01-Apr-2011.
    Matrix pictures by cspy, a MATLAB function in the CSparse package.
    Matrix graphs by Yifan Hu, AT&T Labs Visualization Group.