NAG Library Routine Document

F08YUF (ZTGSEN)

+− Contents

1 Purpose

2 Specification

3 Description

4 References

5 Parameters

6 Error Indicators and Warnings

7 Accuracy

8 Further Comments

+− 9 Example

9.1 Program Text

9.2 Program Data

9.3 Program Results

1 Purpose

F08YUF (ZTGSEN) reorders the generalized Schur factorization of a complex matrix pair in generalized Schur form, so that a selected cluster of eigenvalues appears in the leading elements on the diagonal of the generalized Schur form. The routine also, optionally, computes the reciprocal condition numbers of the cluster of eigenvalues and/or corresponding deflating subspaces.

2 Specification

SUBROUTINE F08YUF (

IJOB, WANTQ, WANTZ, SELECT, N, A, LDA, B, LDB, ALPHA, BETA, Q, LDQ, Z, LDZ, M, PL, PR, DIF, WORK, LWORK, IWORK, LIWORK, INFO)

INTEGER	IJOB, N, LDA, LDB, LDQ, LDZ, M, LWORK, IWORK(max(1,LIWORK)), LIWORK, INFO
REAL (KIND=nag_wp)	PL, PR, DIF(*)
COMPLEX (KIND=nag_wp)	A(LDA,), B(LDB,), ALPHA(N), BETA(N), Q(LDQ,), Z(LDZ,), WORK(max(1,LWORK))
LOGICAL	WANTQ, WANTZ, SELECT(N)

The routine may be called by its LAPACK name ztgsen.

3 Description

F08YUF (ZTGSEN) factorizes the generalized complex

n

n

matrix pair

(S, T)

in generalized Schur form, using a unitary equivalence transformation as

S = \hat{Q} \hat{S} {\hat{Z}}^{H}, T = \hat{Q} \hat{T} {\hat{Z}}^{H},

where

(\hat{S}, \hat{T})

are also in generalized Schur form and have the selected eigenvalues as the leading diagonal elements. The leading columns of

Q

and

Z

are the generalized Schur vectors corresponding to the selected eigenvalues and form orthonormal subspaces for the left and right eigenspaces (deflating subspaces) of the pair

(S, T)

The pair

(S, T)

are in generalized Schur form if

S

and

T

are upper triangular as returned, for example, by F08XNF (ZGGES), or F08XSF (ZHGEQZ) with

JOB ='S'

. The diagonal elements define the generalized eigenvalues

(α_{i}, β_{i})

, for

i = 1, 2, \dots, n

, of the pair

(S, T)

. The eigenvalues are given by

λ_{i} = α_{i} / β_{i},

but are returned as the pair

(α_{i}, β_{i})

in order to avoid possible overflow in computing

λ_{i}

. Optionally, the routine returns reciprocals of condition number estimates for the selected eigenvalue cluster,

p

and

q

, the right and left projection norms, and of deflating subspaces,

{Dif}_{u}

and

{Dif}_{l}

. For more information see Sections 2.4.8 and 4.11 of Anderson et al. (1999).

S

and

T

are the result of a generalized Schur factorization of a matrix pair

(A, B)

A = Q S Z^{H}, B = Q T Z^{H}

then, optionally, the matrices

Q

and

Z

can be updated as

Q \hat{Q}

and

Z \hat{Z}

. Note that the condition numbers of the pair

(S, T)

are the same as those of the pair

(A, B)

4 References

Anderson E, Bai Z, Bischof C, Blackford S, Demmel J, Dongarra J J, Du Croz J J, Greenbaum A, Hammarling S, McKenney A and Sorensen D (1999) LAPACK Users' Guide (3rd Edition) SIAM, Philadelphia http://www.netlib.org/lapack/lug

5 Parameters

1: IJOB – INTEGERInput

On entry: specifies whether condition numbers are required for the cluster of eigenvalues (

p

and

q

) or the deflating subspaces (

{Dif}_{u}

and

{Dif}_{l}

$IJOB = 0$: Only reorder with respect to SELECT. No extras.
$IJOB = 1$: Reciprocal of norms of ‘projections’ onto left and right eigenspaces with respect to the selected cluster ( $p$ and $q$ ).
$IJOB = 2$: The upper bounds on ${Dif}_{u}$ and ${Dif}_{l}$ . $F$ -norm-based estimate ( $DIF (1 : 2)$ ).
$IJOB = 3$: Estimate of ${Dif}_{u}$ and ${Dif}_{l}$ . $1$ -norm-based estimate ( $DIF (1 : 2)$ ). About five times as expensive as $IJOB = 2$ .
$IJOB = 4$: Compute PL, PR and DIF as in $IJOB = 0$ , $1$ and $2$ . Economic version to get it all.
$IJOB = 5$: Compute PL, PR and DIF as in $IJOB = 0$ , $1$ and $3$ .

Constraint:

0 \leq IJOB \leq 5

2: WANTQ – LOGICALInput

On entry: if

WANTQ = .TRUE.

, update the left transformation matrix

Q

WANTQ = .FALSE.

, do not update

Q

3: WANTZ – LOGICALInput

On entry: if

WANTZ = .TRUE.

, update the right transformation matrix

Z

WANTZ = .FALSE.

, do not update

Z

4: SELECT(N) – LOGICAL arrayInput

On entry: specifies the eigenvalues in the selected cluster. To select an eigenvalue

λ_{j}

SELECT (j)

must be set to .TRUE..

5: N – INTEGERInput

On entry:

n

, the order of the matrices

S

and

T

Constraint:

N \geq 0

6: A(LDA, $*$ ) – COMPLEX (KIND=nag_wp) arrayInput/Output

Note: the second dimension of the array A must be at least

\max (1, N)

On entry: the matrix

S

in the pair

(S, T)

On exit: the updated matrix

\hat{S}

7: LDA – INTEGERInput

On entry: the first dimension of the array A as declared in the (sub)program from which F08YUF (ZTGSEN) is called.

Constraint:

LDA \geq \max (1, N)

8: B(LDB, $*$ ) – COMPLEX (KIND=nag_wp) arrayInput/Output

Note: the second dimension of the array B must be at least

\max (1, N)

On entry: the matrix

T

, in the pair

(S, T)

On exit: the updated matrix

\hat{T}

9: LDB – INTEGERInput

On entry: the first dimension of the array B as declared in the (sub)program from which F08YUF (ZTGSEN) is called.

Constraint:

LDB \geq \max (1, N)

10: ALPHA(N) – COMPLEX (KIND=nag_wp) arrayOutput

11: BETA(N) – COMPLEX (KIND=nag_wp) arrayOutput

On exit: ALPHA and BETA contain diagonal elements of

\hat{S}

and

\hat{T}

, respectively, when the pair

(S, T)

has been reduced to generalized Schur form.

ALPHA (i) / BETA (i)

, for

i = 1, 2, \dots, N

, are the eigenvalues.

12: Q(LDQ, $*$ ) – COMPLEX (KIND=nag_wp) arrayInput/Output

Note: the second dimension of the array Q must be at least

\max (1, N)

WANTQ = .TRUE.

, and at least

1

otherwise.

On entry: if

WANTQ = .TRUE.

, the

n

n

matrix

Q

On exit: if

WANTQ = .TRUE.

, the updated matrix

Q \hat{Q}

WANTQ = .FALSE.

, Q is not referenced.

13: LDQ – INTEGERInput

On entry: the first dimension of the array Q as declared in the (sub)program from which F08YUF (ZTGSEN) is called.

Constraints:

if $WANTQ = .TRUE.$ , $LDQ \geq \max (1, N)$ ;
otherwise $LDQ \geq 1$ .

14: Z(LDZ, $*$ ) – COMPLEX (KIND=nag_wp) arrayInput/Output

Note: the second dimension of the array Z must be at least

\max (1, N)

WANTZ = .TRUE.

, and at least

1

otherwise.

On entry: if

WANTZ = .TRUE.

, the

n

n

matrix

Z

On exit: if

WANTZ = .TRUE.

, the updated matrix

Z \hat{Z}

WANTZ = .FALSE.

, Z is not referenced.

15: LDZ – INTEGERInput

On entry: the first dimension of the array Z as declared in the (sub)program from which F08YUF (ZTGSEN) is called.

Constraints:

if $WANTZ = .TRUE.$ , $LDZ \geq \max (1, N)$ ;
otherwise $LDZ \geq 1$ .

16: M – INTEGEROutput

On exit: the dimension of the specified pair of left and right eigenspaces (deflating subspaces).

Constraint:

0 \leq M \leq N

17: PL – REAL (KIND=nag_wp)Output

18: PR – REAL (KIND=nag_wp)Output

On exit: if

IJOB = 1

4

5

, PL and PR are lower bounds on the reciprocal of the norm of ‘projections’

p

and

q

onto left and right eigenspace with respect to the selected cluster.

0 < PL

PR \leq 1

M = 0

M = N

PL = PR = 1

IJOB = 0

2

3

, PL and PR are not referenced.

19: DIF( $*$ ) – REAL (KIND=nag_wp) arrayOutput

Note: the dimension of the array DIF must be at least

2

On exit: if

IJOB \geq 2

DIF (1 : 2)

store the estimates of

{Dif}_{u}

and

{Dif}_{l}

IJOB = 2

4

DIF (1 : 2)

are

F

-norm-based upper bounds on

{Dif}_{u}

and

{Dif}_{l}

IJOB = 3

5

DIF (1 : 2)

are

1

-norm-based estimates of

{Dif}_{u}

and

{Dif}_{l}

M = 0

n

DIF (1 : 2)

= {‖(A, B)‖}_{F}

IJOB = 0

1

, DIF is not referenced.

20: WORK( $\max (1, LWORK)$ ) – COMPLEX (KIND=nag_wp) arrayWorkspace

On exit: if

INFO = 0

, the real part of

WORK (1)

contains the minimum value of LWORK required for optimal performance.

21: LWORK – INTEGERInput

On entry: the dimension of the array WORK as declared in the (sub)program from which F08YUF (ZTGSEN) is called.

LWORK = - 1

, a workspace query is assumed; the routine only calculates the minimum sizes of the WORK and IWORK arrays, returns these values as the first entries of the WORK and IWORK arrays, and no error message related to LWORK or LIWORK is issued.

Constraints:

LWORK \neq - 1

if $IJOB = 1$ , $2$ or $4$ , $LWORK \geq \max (1, 2 \times M \times (N - M))$ ;
if $IJOB = 3$ or $5$ , $LWORK \geq \max (1, 4 \times M \times (N - M))$ ;
otherwise $LWORK \geq 1$ .

22: IWORK( $\max (1, LIWORK)$ ) – INTEGER arrayWorkspace

On exit: if

INFO = 0

IWORK (1)

returns the minimum LIWORK.

23: LIWORK – INTEGERInput

On entry: the dimension of the array IWORK as declared in the (sub)program from which F08YUF (ZTGSEN) is called.

LIWORK = - 1

Constraints:

LIWORK \neq - 1

if $IJOB = 1$ , $2$ or $4$ , $LIWORK \geq N + 2$ ;
if $IJOB = 3$ or $5$ , $LIWORK \geq \max (N + 2, 2 \times M \times (N - M))$ ;
otherwise $LIWORK \geq 1$ .

24: INFO – INTEGEROutput

On exit:

INFO = 0

unless the routine detects an error (see Section 6).

6 Error Indicators and Warnings

Errors or warnings detected by the routine:

$INFO < 0$: If $INFO = - i$ , argument $i$ had an illegal value. An explanatory message is output, and execution of the program is terminated.

$INFO = 1$: Reordering of $(S, T)$ failed because the transformed matrix pair $(\hat{S}, \hat{T})$ would be too far from generalized Schur form; the problem is very ill-conditioned. $(S, T)$ may have been partially reordered. If requested, $0$ is returned in $DIF (1 : 2)$ , PL and PR.

7 Accuracy

The computed generalized Schur form is nearly the exact generalized Schur form for nearby matrices

(S + E)

and

(T + F)

, where

{‖E‖}_{2} = O ε {‖S‖}_{2} and {‖F‖}_{2} = O ε {‖T‖}_{2},

and

ε

is the machine precision. See Section 4.11 of Anderson et al. (1999) for further details of error bounds for the generalized nonsymmetric eigenproblem, and for information on the condition numbers returned.

8 Further Comments

The real analogue of this routine is F08YGF (DTGSEN).

9 Example

This example reorders the generalized Schur factors

S

and

T

and update the matrices

Q

and

Z

given by

S = (\begin{array}{r} 4.0 + 4.0 i & 1.0 + 1.0 i & 1.0 + 1.0 i & 2.0 - 1.0 i \\ 0 & 2.0 + 1.0 i & 1.0 + 1.0 i & 1.0 + 1.0 i \\ 0 & 0 & 2.0 - 1.0 i & 1.0 + 1.0 i \\ 0 & 0 & 0 & 6.0 - 2.0 i \end{array}),

T = (\begin{array}{r} 2.0 & 1.0 + 1.0 i & 1.0 + 1.0 i & 3.0 - 1.0 i \\ 0 & 1.0 & 2.0 + 1.0 i & 1.0 + 1.0 i \\ 0 & 0 & 1.0 & 1.0 + 1.0 i \\ 0 & 0 & 0 & 2.0 \end{array}),

Q = (\begin{array}{r} 1.0 & 0 & 0 & 0 \\ 0 & 1.0 & 0 & 0 \\ 0 & 0 & 1.0 & 0 \\ 0 & 0 & 0 & 1.0 \end{array}) and Z = (\begin{array}{r} 1.0 & 0 & 0 & 0 \\ 0 & 1.0 & 0 & 0 \\ 0 & 0 & 1.0 & 0 \\ 0 & 0 & 0 & 1.0 \end{array}),

selecting the second and third generalized eigenvalues to be moved to the leading positions. Bases for the left and right deflating subspaces, and estimates of the condition numbers for the eigenvalues and Frobenius norm based bounds on the condition numbers for the deflating subspaces are also output.

NAG Library Routine DocumentF08YUF (ZTGSEN)

+− Contents

1 Purpose

2 Specification

3 Description

4 References

5 Parameters

6 Error Indicators and Warnings

7 Accuracy

8 Further Comments

9 Example

9.1 Program Text

9.2 Program Data

9.3 Program Results

NAG Library Routine Document

F08YUF (ZTGSEN)