1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
|
<HTML>
<HEAD><TITLE>MB03ZA - SLICOT Library Routine Documentation</TITLE>
</HEAD>
<BODY>
<H2><A Name="MB03ZA">MB03ZA</A></H2>
<H3>
Reordering a selected cluster of eigenvalues of a given matrix pair in periodic Schur form
</H3>
<A HREF ="#Specification"><B>[Specification]</B></A>
<A HREF ="#Arguments"><B>[Arguments]</B></A>
<A HREF ="#Method"><B>[Method]</B></A>
<A HREF ="#References"><B>[References]</B></A>
<A HREF ="#Comments"><B>[Comments]</B></A>
<A HREF ="#Example"><B>[Example]</B></A>
<P>
<B><FONT SIZE="+1">Purpose</FONT></B>
<PRE>
1. To compute, for a given matrix pair (A,B) in periodic Schur
form, orthogonal matrices Ur and Vr so that
T [ A11 A12 ] T [ B11 B12 ]
Vr * A * Ur = [ ], Ur * B * Vr = [ ], (1)
[ 0 A22 ] [ 0 B22 ]
is in periodic Schur form, and the eigenvalues of A11*B11
form a selected cluster of eigenvalues.
2. To compute an orthogonal matrix W so that
T [ 0 -A11 ] [ R11 R12 ]
W * [ ] * W = [ ], (2)
[ B11 0 ] [ 0 R22 ]
where the eigenvalues of R11 and -R22 coincide and have
positive real part.
Optionally, the matrix C is overwritten by Ur'*C*Vr.
All eigenvalues of A11*B11 must either be complex or real and
negative.
</PRE>
<A name="Specification"><B><FONT SIZE="+1">Specification</FONT></B></A>
<PRE>
SUBROUTINE MB03ZA( COMPC, COMPU, COMPV, COMPW, WHICH, SELECT, N,
$ A, LDA, B, LDB, C, LDC, U1, LDU1, U2, LDU2, V1,
$ LDV1, V2, LDV2, W, LDW, WR, WI, M, DWORK,
$ LDWORK, INFO )
C .. Scalar Arguments ..
CHARACTER COMPC, COMPU, COMPV, COMPW, WHICH
INTEGER INFO, LDA, LDB, LDC, LDU1, LDU2, LDV1, LDV2,
$ LDW, LDWORK, M, N
C .. Array Arguments ..
LOGICAL SELECT(*)
DOUBLE PRECISION A(LDA,*), B(LDB,*), C(LDC,*), DWORK(*),
$ U1(LDU1,*), U2(LDU2,*), V1(LDV1,*), V2(LDV2,*),
$ W(LDW,*), WI(*), WR(*)
</PRE>
<A name="Arguments"><B><FONT SIZE="+1">Arguments</FONT></B></A>
<P>
<B>Mode Parameters</B>
<PRE>
COMPC CHARACTER*1
= 'U': update the matrix C;
= 'N': do not update C.
COMPU CHARACTER*1
= 'U': update the matrices U1 and U2;
= 'N': do not update U1 and U2.
See the description of U1 and U2.
COMPV CHARACTER*1
= 'U': update the matrices V1 and V2;
= 'N': do not update V1 and V2.
See the description of V1 and V2.
COMPW CHARACTER*1
Indicates whether or not the user wishes to accumulate
the matrix W as follows:
= 'N': the matrix W is not required;
= 'I': W is initialized to the unit matrix and the
orthogonal transformation matrix W is returned;
= 'V': W must contain an orthogonal matrix Q on entry,
and the product Q*W is returned.
WHICH CHARACTER*1
= 'A': select all eigenvalues, this effectively means
that Ur and Vr are identity matrices and A11 = A,
B11 = B;
= 'S': select a cluster of eigenvalues specified by
SELECT.
SELECT LOGICAL array, dimension (N)
If WHICH = 'S', then SELECT specifies the eigenvalues of
A*B in the selected cluster. To select a real eigenvalue
w(j), SELECT(j) must be set to .TRUE.. To select a complex
conjugate pair of eigenvalues w(j) and w(j+1),
corresponding to a 2-by-2 diagonal block in A, both
SELECT(j) and SELECT(j+1) must be set to .TRUE.; a complex
conjugate pair of eigenvalues must be either both included
in the cluster or both excluded.
</PRE>
<B>Input/Output Parameters</B>
<PRE>
N (input) INTEGER
The order of the matrix A. N >= 0.
A (input/output) DOUBLE PRECISION array, dimension (LDA,N)
On entry, the leading N-by-N part of this array must
contain the upper quasi-triangular matrix A of the matrix
pair (A,B) in periodic Schur form.
On exit, the leading M-by-M part of this array contains
the matrix R22 in (2).
LDA INTEGER
The leading dimension of the array A. LDA >= MAX(1,N).
B (input/output) DOUBLE PRECISION array, dimension (LDB,N)
On entry, the leading N-by-N part of this array must
contain the upper triangular matrix B of the matrix pair
(A,B) in periodic Schur form.
On exit, the leading N-by-N part of this array is
overwritten.
LDB INTEGER
The leading dimension of the array B. LDB >= MAX(1,N).
C (input/output) DOUBLE PRECISION array, dimension (LDC,N)
On entry, if COMPC = 'U', the leading N-by-N part of this
array must contain a general matrix C.
On exit, if COMPC = 'U', the leading N-by-N part of this
array contains the updated matrix Ur'*C*Vr.
If COMPC = 'N' or WHICH = 'A', this array is not
referenced.
LDC INTEGER
The leading dimension of the array C. LDC >= 1.
LDC >= N, if COMPC = 'U' and WHICH = 'S'.
U1 (input/output) DOUBLE PRECISION array, dimension (LDU1,N)
On entry, if COMPU = 'U' and WHICH = 'S', the leading
N-by-N part of this array must contain U1, the (1,1)
block of an orthogonal symplectic matrix
U = [ U1, U2; -U2, U1 ].
On exit, if COMPU = 'U' and WHICH = 'S', the leading
N-by-N part of this array contains U1*Ur.
If COMPU = 'N' or WHICH = 'A', this array is not
referenced.
LDU1 INTEGER
The leading dimension of the array U1. LDU1 >= 1.
LDU1 >= N, if COMPU = 'U' and WHICH = 'S'.
U2 (input/output) DOUBLE PRECISION array, dimension (LDU2,N)
On entry, if COMPU = 'U' and WHICH = 'S', the leading
N-by-N part of this array must contain U2, the (1,2)
block of an orthogonal symplectic matrix
U = [ U1, U2; -U2, U1 ].
On exit, if COMPU = 'U' and WHICH = 'S', the leading
N-by-N part of this array contains U2*Ur.
If COMPU = 'N' or WHICH = 'A', this array is not
referenced.
LDU2 INTEGER
The leading dimension of the array U2. LDU2 >= 1.
LDU2 >= N, if COMPU = 'U' and WHICH = 'S'.
V1 (input/output) DOUBLE PRECISION array, dimension (LDV1,N)
On entry, if COMPV = 'U' and WHICH = 'S', the leading
N-by-N part of this array must contain V1, the (1,1)
block of an orthogonal symplectic matrix
V = [ V1, V2; -V2, V1 ].
On exit, if COMPV = 'U' and WHICH = 'S', the leading
N-by-N part of this array contains V1*Vr.
If COMPV = 'N' or WHICH = 'A', this array is not
referenced.
LDV1 INTEGER
The leading dimension of the array V1. LDV1 >= 1.
LDV1 >= N, if COMPV = 'U' and WHICH = 'S'.
V2 (input/output) DOUBLE PRECISION array, dimension (LDV2,N)
On entry, if COMPV = 'U' and WHICH = 'S', the leading
N-by-N part of this array must contain V2, the (1,2)
block of an orthogonal symplectic matrix
V = [ V1, V2; -V2, V1 ].
On exit, if COMPV = 'U' and WHICH = 'S', the leading
N-by-N part of this array contains V2*Vr.
If COMPV = 'N' or WHICH = 'A', this array is not
referenced.
LDV2 INTEGER
The leading dimension of the array V2. LDV2 >= 1.
LDV2 >= N, if COMPV = 'U' and WHICH = 'S'.
W (input/output) DOUBLE PRECISION array, dimension (LDW,2*M)
On entry, if COMPW = 'V', then the leading 2*M-by-2*M part
of this array must contain a matrix W.
If COMPW = 'I', then W need not be set on entry, W is set
to the identity matrix.
On exit, if COMPW = 'I' or 'V' the leading 2*M-by-2*M part
of this array is post-multiplied by the transformation
matrix that produced (2).
If COMPW = 'N', this array is not referenced.
LDW INTEGER
The leading dimension of the array W. LDW >= 1.
LDW >= 2*M, if COMPW = 'I' or COMPW = 'V'.
WR (output) DOUBLE PRECISION array, dimension (M)
WI (output) DOUBLE PRECISION array, dimension (M)
The real and imaginary parts, respectively, of the
eigenvalues of R11. The eigenvalues are stored in the same
order as on the diagonal of R22, with
WR(i) = -R22(i,i) and, if R22(i:i+1,i:i+1) is a 2-by-2
diagonal block, WI(i) > 0 and WI(i+1) = -WI(i).
In exact arithmetic, these eigenvalue are the positive
square roots of the selected eigenvalues of the product
A*B. However, if an eigenvalue is sufficiently
ill-conditioned, then its value may differ significantly.
M (output) INTEGER
The number of selected eigenvalues.
</PRE>
<B>Workspace</B>
<PRE>
DWORK DOUBLE PRECISION array, dimension (LDWORK)
On exit, if INFO = -28, DWORK(1) returns the minimum
value of LDWORK.
LDWORK INTEGER
The length of the array DWORK.
LDWORK >= MAX( 1, 4*N, 8*M ).
</PRE>
<B>Error Indicator</B>
<PRE>
INFO INTEGER
= 0: successful exit;
< 0: if INFO = -i, the i-th argument had an illegal
value;
= 1: reordering of the product A*B in Step 1 failed
because some eigenvalues are too close to separate;
= 2: reordering of some submatrix in Step 2 failed
because some eigenvalues are too close to separate;
= 3: the QR algorithm failed to compute the Schur form
of some submatrix in Step 2;
= 4: the condition that all eigenvalues of A11*B11 must
either be complex or real and negative is
numerically violated.
</PRE>
<A name="Method"><B><FONT SIZE="+1">Method</FONT></B></A>
<PRE>
Step 1 is performed using a reordering technique analogous to the
LAPACK routine DTGSEN for reordering matrix pencils [1,2]. Step 2
is an implementation of Algorithm 2 in [3]. It requires O(M*N*N)
floating point operations.
</PRE>
<A name="References"><B><FONT SIZE="+1">References</FONT></B></A>
<PRE>
[1] Kagstrom, B.
A direct method for reordering eigenvalues in the generalized
real Schur form of a regular matrix pair (A,B), in M.S. Moonen
et al (eds), Linear Algebra for Large Scale and Real-Time
Applications, Kluwer Academic Publ., 1993, pp. 195-218.
[2] Kagstrom, B. and Poromaa P.:
Computing eigenspaces with specified eigenvalues of a regular
matrix pair (A, B) and condition estimation: Theory,
algorithms and software, Numer. Algorithms, 1996, vol. 12,
pp. 369-407.
[3] Benner, P., Mehrmann, V., and Xu, H.
A new method for computing the stable invariant subspace of a
real Hamiltonian matrix, J. Comput. Appl. Math., 86,
pp. 17-43, 1997.
</PRE>
<A name="Comments"><B><FONT SIZE="+1">Further Comments</FONT></B></A>
<PRE>
None
</PRE>
<A name="Example"><B><FONT SIZE="+1">Example</FONT></B></A>
<P>
<B>Program Text</B>
<PRE>
None
</PRE>
<B>Program Data</B>
<PRE>
None
</PRE>
<B>Program Results</B>
<PRE>
None
</PRE>
<HR>
<A HREF=support.html><B>Return to Supporting Routines index</B></A></BODY>
</HTML>
|
