1STREVC(1)                LAPACK routine (version 3.1)                STREVC(1)
2
3
4

NAME

6       STREVC  -  some  or all of the right and/or left eigenvectors of a real
7       upper quasi-triangular matrix T
8

SYNOPSIS

10       SUBROUTINE STREVC( SIDE, HOWMNY, SELECT, N, T, LDT, VL, LDVL, VR, LDVR,
11                          MM, M, WORK, INFO )
12
13           CHARACTER      HOWMNY, SIDE
14
15           INTEGER        INFO, LDT, LDVL, LDVR, M, MM, N
16
17           LOGICAL        SELECT( * )
18
19           REAL           T( LDT, * ), VL( LDVL, * ), VR( LDVR, * ), WORK( * )
20

PURPOSE

22       STREVC  computes some or all of the right and/or left eigenvectors of a
23       real upper quasi-triangular matrix T.  Matrices of this type  are  pro‐
24       duced  by  the  Schur  factorization  of  a  real  general matrix:  A =
25       Q*T*Q**T, as computed by SHSEQR.
26
27       The right eigenvector x and the left eigenvector y of  T  corresponding
28       to an eigenvalue w are defined by:
29
30          T*x = w*x,     (y**H)*T = w*(y**H)
31
32       where y**H denotes the conjugate transpose of y.
33       The  eigenvalues  are  not input to this routine, but are read directly
34       from the diagonal blocks of T.
35
36       This routine returns the matrices X and/or Y of right and  left  eigen‐
37       vectors  of  T,  or  the  products  Q*X and/or Q*Y, where Q is an input
38       matrix.  If Q is the orthogonal factor that reduces a matrix A to Schur
39       form  T,  then Q*X and Q*Y are the matrices of right and left eigenvec‐
40       tors of A.
41
42

ARGUMENTS

44       SIDE    (input) CHARACTER*1
45               = 'R':  compute right eigenvectors only;
46               = 'L':  compute left eigenvectors only;
47               = 'B':  compute both right and left eigenvectors.
48
49       HOWMNY  (input) CHARACTER*1
50               = 'A':  compute all right and/or left eigenvectors;
51               = 'B':  compute all right and/or left eigenvectors,  backtrans‐
52               formed  by  the  matrices  in  VR  and/or  VL;  = 'S':  compute
53               selected right and/or left eigenvectors, as  indicated  by  the
54               logical array SELECT.
55
56       SELECT  (input/output) LOGICAL array, dimension (N)
57               If  HOWMNY  = 'S', SELECT specifies the eigenvectors to be com‐
58               puted.  If w(j) is a real eigenvalue,  the  corresponding  real
59               eigenvector  is  computed  if SELECT(j) is .TRUE..  If w(j) and
60               w(j+1) are the real and imaginary parts of a complex  eigenval‐
61               ue, the corresponding complex eigenvector is computed if either
62               SELECT(j) or SELECT(j+1) is .TRUE., and on  exit  SELECT(j)  is
63               set  to  .TRUE.  and  SELECT(j+1)  is  set to Not referenced if
64               HOWMNY = 'A' or 'B'.
65
66       N       (input) INTEGER
67               The order of the matrix T. N >= 0.
68
69       T       (input) REAL array, dimension (LDT,N)
70               The upper quasi-triangular matrix T in Schur canonical form.
71
72       LDT     (input) INTEGER
73               The leading dimension of the array T. LDT >= max(1,N).
74
75       VL      (input/output) REAL array, dimension (LDVL,MM)
76               On entry, if SIDE = 'L' or 'B' and HOWMNY = 'B', VL  must  con‐
77               tain  an  N-by-N  matrix  Q (usually the orthogonal matrix Q of
78               Schur vectors returned by SHSEQR).  On exit, if SIDE =  'L'  or
79               'B',  VL contains: if HOWMNY = 'A', the matrix Y of left eigen‐
80               vectors of T; if HOWMNY = 'B', the matrix Q*Y; if HOWMNY = 'S',
81               the left eigenvectors of T specified by SELECT, stored consecu‐
82               tively in the columns of VL, in the same order as their  eigen‐
83               values.   A  complex eigenvector corresponding to a complex ei‐
84               genvalue is stored in two consecutive columns, the first  hold‐
85               ing the real part, and the second the imaginary part.  Not ref‐
86               erenced if SIDE = 'R'.
87
88       LDVL    (input) INTEGER
89               The leading dimension of the array VL.  LDVL >= 1, and if  SIDE
90               = 'L' or 'B', LDVL >= N.
91
92       VR      (input/output) REAL array, dimension (LDVR,MM)
93               On  entry,  if SIDE = 'R' or 'B' and HOWMNY = 'B', VR must con‐
94               tain an N-by-N matrix Q (usually the  orthogonal  matrix  Q  of
95               Schur  vectors  returned by SHSEQR).  On exit, if SIDE = 'R' or
96               'B', VR contains: if HOWMNY = 'A', the matrix X of right eigen‐
97               vectors of T; if HOWMNY = 'B', the matrix Q*X; if HOWMNY = 'S',
98               the right eigenvectors of T specified by SELECT, stored consec‐
99               utively in the columns of VR, in the same order as their eigen‐
100               values.  A complex eigenvector corresponding to a  complex  ei‐
101               genvalue  is stored in two consecutive columns, the first hold‐
102               ing the real part and the second the imaginary part.  Not  ref‐
103               erenced if SIDE = 'L'.
104
105       LDVR    (input) INTEGER
106               The  leading dimension of the array VR.  LDVR >= 1, and if SIDE
107               = 'R' or 'B', LDVR >= N.
108
109       MM      (input) INTEGER
110               The number of columns in the arrays VL and/or VR. MM >= M.
111
112       M       (output) INTEGER
113               The number of columns in the arrays VL and/or VR actually  used
114               to store the eigenvectors.  If HOWMNY = 'A' or 'B', M is set to
115               N.  Each selected real eigenvector occupies one column and each
116               selected complex eigenvector occupies two columns.
117
118       WORK    (workspace) REAL array, dimension (3*N)
119
120       INFO    (output) INTEGER
121               = 0:  successful exit
122               < 0:  if INFO = -i, the i-th argument had an illegal value
123

FURTHER DETAILS

125       The algorithm used in this program is basically backward (forward) sub‐
126       stitution, with scaling to make the the code  robust  against  possible
127       overflow.
128
129       Each eigenvector is normalized so that the element of largest magnitude
130       has magnitude 1; here the magnitude of a complex number (x,y) is  taken
131       to be |x| + |y|.
132
133
134
135
136 LAPACK routine (version 3.1)    November 2006                       STREVC(1)
Impressum