MB04DY

Symplectic scaling of a Hamiltonian matrix

[Specification] [Arguments] [Method] [References] [Comments] [Example]

Purpose

  To perform a symplectic scaling on the Hamiltonian matrix

           ( A    G  )
       H = (       T ),                                          (1)
           ( Q   -A  )

  i.e., perform either the symplectic scaling transformation

                                -1
              ( A'   G'  )   ( D   0 ) ( A   G  ) ( D  0   )
       H' <-- (        T ) = (       ) (      T ) (     -1 ),    (2)
              ( Q'  -A'  )   ( 0   D ) ( Q  -A  ) ( 0  D   )

  where D is a diagonal scaling matrix, or the symplectic norm
  scaling transformation

               ( A''   G''  )    1  (   A   G/tau )
       H'' <-- (          T ) = --- (           T ),             (3)
               ( Q''  -A''  )   tau ( tau Q   -A  )

  where tau is a real scalar.  Note that if tau is not equal to 1,
  then (3) is NOT a similarity transformation.  The eigenvalues
  of H are then tau times the eigenvalues of H''.

  For symplectic scaling (2), D is chosen to give the rows and
  columns of A' approximately equal 1-norms and to give Q' and G'
  approximately equal norms.  (See METHOD below for details.) For
  norm scaling, tau = MAX(1, ||A||, ||G||, ||Q||) where ||.||
  denotes the 1-norm (column sum norm).

Specification
      SUBROUTINE MB04DY( JOBSCL, N, A, LDA, QG, LDQG, D, DWORK, INFO )
C     .. Scalar Arguments ..
      INTEGER           INFO, LDA, LDQG, N
      CHARACTER         JOBSCL
C     .. Array Arguments ..
      DOUBLE PRECISION  A(LDA,*), D(*), DWORK(*), QG(LDQG,*)

Arguments

Mode Parameters

  JOBSCL  CHARACTER*1
          Indicates which scaling strategy is used, as follows:
          = 'S'       :  do the symplectic scaling (2);
          = '1' or 'O':  do the 1-norm scaling (3);
          = 'N'       :  do nothing; set INFO and return.

Input/Output Parameters
  N       (input) INTEGER
          The order of the matrices A, G, and Q.  N >= 0.

  A       (input/output) DOUBLE PRECISION array, dimension (LDA,N)
          On input, if JOBSCL <> 'N', the leading N-by-N part of
          this array must contain the upper left block A of the
          Hamiltonian matrix H in (1).
          On output, if JOBSCL <> 'N', the leading N-by-N part of
          this array contains the leading N-by-N part of the scaled
          Hamiltonian matrix H' in (2) or H'' in (3), depending on
          the setting of JOBSCL.
          If JOBSCL = 'N', this array is not referenced.

  LDA     INTEGER
          The leading dimension of the array A.
          LDA >= MAX(1,N), if JOBSCL <> 'N';
          LDA >= 1,        if JOBSCL =  'N'.

  QG      (input/output) DOUBLE PRECISION array, dimension
          (LDQG,N+1)
          On input, if JOBSCL <> 'N', the leading N-by-N lower
          triangular part of this array must contain the lower
          triangle of the lower left symmetric block Q of the
          Hamiltonian matrix H in (1), and the N-by-N upper
          triangular part of the submatrix in the columns 2 to N+1
          of this array must contain the upper triangle of the upper
          right symmetric block G of H in (1).
          So, if i >= j, then Q(i,j) = Q(j,i) is stored in QG(i,j)
          and G(i,j) = G(j,i) is stored in QG(j,i+1).
          On output, if JOBSCL <> 'N', the leading N-by-N lower
          triangular part of this array contains the lower triangle
          of the lower left symmetric block Q' or Q'', and the
          N-by-N upper triangular part of the submatrix in the
          columns 2 to N+1 of this array contains the upper triangle
          of the upper right symmetric block G' or G'' of the scaled
          Hamiltonian matrix H' in (2) or H'' in (3), depending on
          the setting of JOBSCL.
          If JOBSCL = 'N', this array is not referenced.

  LDQG    INTEGER
          The leading dimension of the array QG.
          LDQG >= MAX(1,N), if JOBSCL <> 'N';
          LDQG >= 1,        if JOBSCL =  'N'.

  D       (output) DOUBLE PRECISION array, dimension (nd)
          If JOBSCL = 'S', then nd = N and D contains the diagonal
          elements of the diagonal scaling matrix in (2).
          If JOBSCL = '1' or 'O', then nd = 1 and D(1) is set to tau
          from (3). In this case, no other elements of D are
          referenced.
          If JOBSCL = 'N', this array is not referenced.

Workspace
  DWORK   DOUBLE PRECISION array, dimension (N)
          If JOBSCL = 'N', this array is not referenced.

Error Indicator
  INFO    INTEGER
          = 0:  successful exit;
          < 0:  if INFO = -i, then the i-th argument had an illegal
                value.

Method
  1. Symplectic scaling (JOBSCL = 'S'):

  First, LAPACK subroutine DGEBAL is used to equilibrate the 1-norms
  of the rows and columns of A using a diagonal scaling matrix D_A.
  Then, H is similarily transformed by the symplectic diagonal
  matrix D1 = diag(D_A,D_A**(-1)).  Next, the off-diagonal blocks of
  the resulting Hamiltonian matrix are equilibrated in the 1-norm
  using the symplectic diagonal matrix D2 of the form

              ( I/rho    0   )
         D2 = (              )
              (   0    rho*I )

  where rho is a real scalar. Thus, in (2), D = D1*D2.

  2. Norm scaling (JOBSCL = '1' or 'O'):

  The norm of the matrices A and G of (1) is reduced by setting
  A := A/tau  and  G := G/(tau**2) where tau is the power of the
  base of the arithmetic closest to MAX(1, ||A||, ||G||, ||Q||) and
  ||.|| denotes the 1-norm.

References
  [1] Benner, P., Byers, R., and Barth, E.
      Fortran 77 Subroutines for Computing the Eigenvalues of
      Hamiltonian Matrices. I: The Square-Reduced Method.
      ACM Trans. Math. Software, 26, 1, pp. 49-77, 2000.

Numerical Aspects
  For symplectic scaling, the complexity of the used algorithms is
  hard to estimate and depends upon how well the rows and columns of
  A in (1) are equilibrated.  In one sweep, each row/column of A is
  scaled once, i.e., the cost of one sweep is N**2 multiplications.
  Usually, 3-6 sweeps are enough to equilibrate the norms of the
  rows and columns of a matrix.  Roundoff errors are possible as
  LAPACK routine DGEBAL does NOT use powers of the machine base for
  scaling. The second stage (equilibrating ||G|| and ||Q||) requires
  N**2 multiplications.
  For norm scaling, 3*N**2 + O(N) multiplications are required and
  NO rounding errors occur as all multiplications are performed with
  powers of the machine base.

Further Comments
  None
Example

Program Text

*     MB04DY EXAMPLE PROGRAM TEXT.
*     Copyright (c) 2002-2017 NICONET e.V.
*
*     .. Parameters ..
      INTEGER          NIN, NOUT
      PARAMETER        ( NIN = 5, NOUT = 6 )
      INTEGER          NMAX
      PARAMETER        ( NMAX = 20 )
      INTEGER          LDA, LDQG
      PARAMETER        ( LDA = NMAX, LDQG = NMAX )
      INTEGER          LDWORK
      PARAMETER        ( LDWORK = NMAX )
*     .. Local Scalars ..
      INTEGER          I, INFO, J, N
      CHARACTER*1      JOBSCL
*     .. Local Arrays ..
      DOUBLE PRECISION A(LDA,NMAX), D(NMAX), DWORK(LDWORK),
     $                 QG(LDQG,NMAX+1)
*     .. External Functions ..
      LOGICAL          LSAME
      EXTERNAL         LSAME
*     .. External Subroutines ..
      EXTERNAL         MB04DY
*     .. Executable Statements ..
*
      WRITE ( NOUT, FMT = 99999 )
*     Skip the heading in the data file and read the data.
      READ ( NIN, FMT = '()' )
      READ ( NIN, FMT = * ) N, JOBSCL
      IF ( N.LT.0 .OR. N.GT.NMAX ) THEN
         WRITE ( NOUT, FMT = 99998 ) N
      ELSE
         READ ( NIN, FMT = * ) ( ( A(I,J),    J = 1,N ), I = 1,N )
         READ ( NIN, FMT = * ) ( ( QG(J,I+1), I = J,N ), J = 1,N )
         READ ( NIN, FMT = * ) ( ( QG(I,J),   I = J,N ), J = 1,N )
*        Scale the Hamiltonian matrix.
         CALL MB04DY( JOBSCL, N, A, LDA, QG, LDQG, D, DWORK, INFO )
         IF ( INFO.NE.0 ) THEN
            WRITE ( NOUT, FMT = 99997 ) INFO
         ELSE
*           Show the scaled Hamiltonian matrix.
            WRITE ( NOUT, FMT = 99996 )
            DO 10 I = 1, N
              WRITE ( NOUT, FMT = 99993 )  ( A(I,J),    J = 1,N ),
     $           ( QG(J,I+1), J = 1,I-1 ), ( QG(I,J+1), J = I,N )
10          CONTINUE
            DO 20 I = 1, N
               WRITE ( NOUT, FMT = 99993 ) (  QG(I,J), J = 1,I-1 ),
     $               ( QG(J,I), J = I,N ), ( -A(J,I),  J = 1,N )
20          CONTINUE
*           Show the scaling factors.
            IF ( LSAME( JOBSCL, 'S' ) ) THEN
               WRITE ( NOUT, FMT = 99995 )
               WRITE ( NOUT, FMT = 99993 ) ( D(I), I = 1,N )
            ELSE IF ( LSAME( JOBSCL, '1' ) .OR. LSAME( JOBSCL, 'O' ) )
     $            THEN
               WRITE ( NOUT, FMT = 99994 )
               WRITE ( NOUT, FMT = 99993 ) D(1)
            END IF
         ENDIF
      END IF
      STOP
*
99999 FORMAT (' MB04DY EXAMPLE PROGRAM RESULTS',/1X)
99998 FORMAT (/' N is out of range.',/' N = ',I5)
99997 FORMAT (' INFO on exit from MB04DY = ',I2)
99996 FORMAT (/' The scaled Hamiltonian is ')
99995 format (/' The scaling factors are ')
99994 format (/' The scaling factor tau is ')
99993 FORMAT (1X,8(F10.4))
      END
Program Data
MB04DY EXAMPLE PROGRAM DATA
3 S
-0.4   0.05  0.0007   
-4.7   0.8   0.025   
81.0  29.0  -0.9        
0.0034  0.0014  0.00077 -0.005  0.0004  0.003
-18.0 -12.0  43.0  99.0  420.0  -200.0
Program Results
 MB04DY EXAMPLE PROGRAM RESULTS


 The scaled Hamiltonian is 
    -0.4000    0.4000    0.3584  418.4403   21.5374    0.1851
    -0.5875    0.8000    1.6000   21.5374   -9.6149    0.0120
     0.1582    0.4531   -0.9000    0.1851    0.0120    0.0014
    -0.0001   -0.0008    0.1789    0.4000    0.5875   -0.1582
    -0.0008    0.0515   13.9783   -0.4000   -0.8000   -0.4531
     0.1789   13.9783 -426.0056   -0.3584   -1.6000    0.9000

 The scaling factors are 
     0.0029    0.0228    1.4595

Return to Supporting Routines index