A Note On Drazin Inverses: HEN ING

Download as pdf or txt
Download as pdf or txt
You are on page 1of 11

Pacific Journal of

Mathematics

A NOTE ON DRAZIN INVERSES


C HEN F. K ING

Vol. 70, No. 2

October 1977

PACIFIC JOURNAL OF MATHEMATICS


Vol. 70, No 2, 1977

A NOTE ON DRAZIN INVERSES


CHEN F. KING

D is the Drazin inverse of T if TD = DT, D = TD2, and


k+
T = T D for some k.
k

In recent years, there has been a great deal of interest in generalized


inverses of matrices ([2], [4], [5]) and many of the concepts can be
formulated in Banach space. In particular, if X is a Banach space and
B(X) denotes the algebra of bounded operators on X, then we make the
following definitions:
1. An operator S in B(X) is called a generalized
inverse of T if TST = T and STS = S.
DEFINITION

DEFINITION 2. An operator T in B(X) is called generalized


Fredholm if both the range R(T) and the null space N(T) are closed
complemented subspaces of X.

Let an operator D in B(X) be the Drazin inverse of T. Then


k
k+
T = T D for some nonnegative integer fc.
DEFINITION 3. The smallest k for which the latter is valid is called
the index of T.
In fact, if an operator T in B(X) has a Drazin inverse then it has
only one ([2], Theorem 1).
REMARKS. (1) It is well known and easy to prove that T is a
generalized Fredholm operator if and only if it has a generalized
inverse. Some properties of the operator thus defined are obtained in
[1] but generally there remain unsatisfactory features. For example, in
Banach space there is no obvious way of defining a unique generalized
inverse and there is no useful relation between the spectrum of an
operator and of any of its generalized inverse.
(2) The Drazin inverse was introduced in [2] in a very general
context and avoids the two defects mentioned above. Note also that if
the index is equal to 1, then D is a generalized inverse of T.

We will now proceed to obtain some properties of operators with a


Drazin inverse including an exact characterization of such operators. In
order to simplify the proof of Theorem 1, we prove the following lemma:
383

384

CHEN F. KING

LEMMA 1. Let T be an operator in B(X).


Then T has a generalized inverse S such that TS = ST if and only if X can be written
X = R(T)N(T).

Proof Let X = R(T)($N(T)


onto R(T) along N(T). Let

and let P be the projection from X

0 = T\R(T)
then N(Q) = (0) and Q is bounded with closed range.
bounded inverse on R(T). We define

Hence, Q has a

S = Q- P.
It is easy to see that S is a commuting generalized inverse of .
Conversely, if T has a commuting generalized inverse 5 then TS is a
projection from X onto R(T). Let

where Xx = N(5). For each x E X,, TSx = 0 and


x = Sx = TSx = 0;
this implies x N ( ) .
x = 0 and

On the other hand, for each xEN(T)

this says x E X,. Consequently, N() = Xi.


In fact, T5 = S implies N() = N(S) and R(T) = (S).

then

Thus,

THEOREM 1. Suppose T is an operator in B(X) with generalized


inverse S such that TS = S. Then the nonzero points in p(), the
resolvent set of T are precisely the reciprocals of the nonzero points in p (S).

Proof. By Lemma 1, X can be decomposed into

Assume ^O in p() then

A NOTE ON DRAZIN INVERSES

385

( - iy\ - i)s = TS,


which yields

-T(T- /)"1 (s-j

TS\ = TS.

Since TS is the identity on R(T), for each x E R(T),

- \( - iy1 is -jx = .
This implies (5 - (1/)/) has a bounded inverse on R(T) for all ^ 0 in
P(T).
On the other hand, for each x E N(T)

or

Thus (S A"1/) also has a bounded inverse on N(T)for all A/ 0 in p(T).


Because ( S - - 1 ) ( ) = ( S - - 1 I ) ( S ) C U ( S ) = ( T ) and
( S - - 7 ) N ( ) = (S--7)N(S)CN(S)=(), so 1/ 6p(S).
The converse statement is established with T replaced by 5 and S by
T. The proof is complete.
REMARK. The commutativity condition in Theorem 1 is essential,
for consider the shift operator S: (JC1? X2, JC3, ) (0, xu x2, ) in I2. Then
SS*S = S and S*SS* = 5* so that S* is a generalized inverse of 5. But

THEOREM

and index k.
with Tk.

2. Let T be an operator in B(X) with Drazin inverse D


Then Dk is a generalized inverse of Tk and Dk commutes

Proof. Obviously Dk and Tk commute.


DkTkDk

= D2kTk

= (D2T)k

Then
= Dk

386

CHEN F. KING

and

_ rk + l krk-1

Tk+D

If D is the Drazin inverse of T with index k, then


R(Tk)N(Tk).

COROLLARY.

X =

THEOREM 3. // T in B(X) has a Drazin inverse D and is a


nonzero point in p(T), then A"1 belongs to p(D).

Proof (TDf = TDTD = TD, so TD is a projection. It is easy to


verify that R(D)=R(TD) and N(D) = N(TD). Hence R(D) and
N(D) are closed complemented in X.
Since
2

D(T D)D

= T D =TD = D

and
(T2D)D(T2D) = TD3 = TD2 = T2D,
this shows that T2D is a commuting generalized inverse of D.
Lemma 1,

Then, by

The rest of the proof is analogous to the first part of Theorem 1 since
TD is identity and zero on R(D) and N(D) respectively.
Recall the definition of ascent a(T) and descent d(T) for operator T
2
in B(S): an operator has finite ascent if the chain
N(T)CN(T )C
N(Ty)C''becomes constant after a finite number of steps. The
k
k+1
smallest integer k such that N(T ) = N(T ) is then defined to be
2
a(T). The descent is defined similary for the chain R(T)D R(T )D
3
R(T )D - - . If T has finite ascent and descent, then they are equal ([6],
Theorem 5.41-E).

A NOTE ON DRAZIN INVERSES

387

4. An operator T in B(X) has a Drazin inverse if and


only if it has finite ascent and descent. In such a case, the index of T is
equal to the common value of a{T) and d(T).
THEOREM

Proof of sufficiency. Let k = a(T)= d(T) be finite. Then ([61,


Theorem 5.41-G) T is completely reduced by the pair of closed
complemented subspaces R(Tk) and N(Tk) of X and
X =

R(Tk)N(Tk).

Let P be the projection from X onto R(Tk) along N(Tk).

Then

PTk = TkP.

(1)

For each x in X, x can be written as x = y + z where y E R(Tk) and


z EN(Tk).
TkPx = Tkp(y + z) = kPy = ky
P k x = PTk(y + z) = P k y = Tky.
Since N(Tk) = N(Tn) and R(Tk) = R(Tn) for all n ^ fc, we have X =
) for all n^k.
This implies

(2)

PT = TP.

From (1), we have


(TP)Tk = k + 1 P = (PT)Tk.
Thus, P and commute on R(Tk).
PTJC = PT(y

Again, for each x = y + z in X,

+ 2) = PTy = Py = TPx.

Therefore PT = TP on X.
(3) Define Q = TR(Tk).
Q is a closed operator follows from the
fact that O is bounded with closed domain. To show Q has a bounded
inverse on R(Tk) we need only to prove that Q maps R(Tk) in a
k
one one manner onto itself. Because T maps R(T ) onto itself, so does
k
O. If Ox = 0 with x E jR(T ) then
k

k+]

0 = Ox = OT y = T y

for some

yE

388

CHEN F. KING
k+1

This implies yN(T ) = N(T ),

thus x = T y = 0. We define
D = Q-'P.

(4) Now, we must show that D, defined as above, is a Drazin


inverse of , which is unique by ([2], Theorem 1). For every x = y + z
k
k
in X with y G R(T ) and z G N( ) then
TDx = TQ^P(y + z) = TQ^Py = y
Dx = QPT(y + z) = QTP(y + z)=QTy

= y,

so that D = D.
D 2 x = Q'PTQP(y

+ z) = Q- ! P 2 JC = DJC.

Thus, D = TD 2 .
Finally, (TD) 2 = TDTD = TD = P. Hence / - TD is a projection
from X onto N( fc ) along i?( k ). For any x in X
(I-TD)x

N(Tk).

This implies Tk(I - TD)x = 0 and then we have

(5) It remains only to show that k is the smallest positive integer


such that Tk = Tk+D. Suppose there is a positive integer m < k such
that
m __ rm+r\

then
Tm(I-TD)x =0
m

VxEX,
k

hence ( / - D)x EN(T ).


But (I - D)x G N(T ), this contradicts the
hypothesis that k is the smallest common value of a(T) and d(T).
Proof of necessity. In Theorem 3 we have proved that if D is the
Drazin inverse of T with index k then T2D is a commuting generalized
inverse of D and X = i? (D) 0 N(D). The proof will be complete if we
can show that JR(D) = R(Tk) and N(D)=N(Tk).
k
If y G R(T ) then there is some x G X such that

A NOTE ON DRAZIN INVERSES


k

k+

k+

y = T x = T Dx = DT x

389

ER(D).

Conversely, if y G R(D) then there is some x G X such that


= TD2x = T
k

D3JC

This shows that R(D) = R(T ).


k
N(T ).
Conclusion is that

= TkDfc+1JC G

Similarly, we can show that N(D)

X
This implies Tk(I- TD)x = 0 and then we have

(6) It remains only to prove that k is the smallest positive integer


such that Tk = Tk+D. Suppose there is a positive integer m < k such
that

then
Tm(I-TD)x=0

JCGX,

hence (/ - TD)x G N(T ).


But (/ - D)x E N(T ), which contradicts
the hypothesis that k is the smallest common value of a(T) and d(T).
The proof of the necessary part is included in Theorem 1.
The operator T can be written as

since T and p commute, then for each x E. X


T(I-p)kx = Tk(I-p)x=0.
This shows that T(I - p) is nilpotent of order k. As mentioned earlier
T2D = TP is a commuting generalized inverse of D, so that TP has index
0 or 1 (it is zero when T is invertible). The following theorem is proved
by Greville ([4], Theorem 9.3) in finite dimensional space. It can be
extended to the general case without changing the proof. We merely
state:

390

CHEN F. KING

THEOREM

5. The decomposition (*) is the only decomposition of T

of the form
T = A + B,
where A has index 0 or 1, B is nilpotent of order k and AB = BA = 0.
The author takes pleasure in thanking Dr. S.
R. Caradus for his valuable comments on this paper.
ACKNOWLEDGEMENT.

REFERENCES
1. S. R. Caradus, Generalized Fredholm Operators, to be published.
2. M. P. Drazin, Pseudo-inverse in associative rings and semigroups, Amer. Math. Monthly, 65
(1958), 506-514.
3. S. Goldberg, Unbounded Linear Operators, McGraw-Hill, N.Y., 1966.
4. T. N. E. Greville, Spectral Generalized Inverse of Square Matrices, MRC Technical Summary
Report No. 823, Math. Research Center, U.S. Army, University of Wisconsin, Madison.
5. C. R. Rao, and S. K. Mitra, The Generalized Inverse of Matrices and its Applications, Wiley,
N.Y., 1971.
6. A. E. Taylor, Functional Analysis, John Wiley and Sons, N.Y., 1967.
Received December 4, 1973.
QUEEN'S UNIVERSITY, KINGSTON, ONTARIO

PACIFIC JOURNAL OF MATHEMATICS


EDITORS
RICHARD ARENS (Managing
University of California
Los Angeles, CA 90024
R. A.

Editor)

J. DUGUNDJI
Department of Mathematics
University of Southern California
Los Angeles, CA 90007

BEAUMONT

R. FINN AND J. M I L G R A M

University of Washington
Seattle, WA 98105
C. C.

Stanford University
Stanford, CA 94305

MOORE

University of California
Berkeley, CA 94720
ASSOCIATE
E. F. BECKENBACH

B. H. NEUMANN

EDITORS
F. W O L F

K. YOSHIDA

SUPPORTING INSTITUTIONS
UNIVERSITY OF BRITISH COLUMBIA
CALIFORNIA INSTITUTE OF TECHNOLOGY
UNIVERSITY OF CALIFORNIA
MONTANA STATE UNIVERSITY
UNIVERSITY OF NEVADA
NEW MEXICO STATE UNIVERSITY
OREGON STATE UNIVERSITY
UNIVERSITY OF OREGON
OSAKA UNIVERSITY

UNIVERSITY OF SOUTHERN CALIFORNIA


STANFORD UNIVERSITY
UNIVERSITY OF HAWAII
UNIVERSITY OF TOKYO
UNIVERSITY OF UTAH
WASHINGTON STATE UNIVERSITY
UNIVERSITY OF WASHINGTON
*
*
*
AMERICAN MATHEMATICAL SOCIETY

The Supporting Institutions listed above contribute to the cost of publication of this Journal, but they are
not owners or publishers and have no responsibility for its contents or policies.
Mathematical papers intended for publication in the Pacific Journal of Mathematics should be in typed
form or offset-reproduced (not dittoed), double spaced with large margins. Underline Greek letters in red,
German in green, and script in blue. The first paragraph or two must be capable of being used separately as a
synopsis of the entire paper. Items of the bibliography should not be cited there unless absolutely necessary, in
which case they must be identified by author and Journal, rather than by item number. Manuscripts, in
duplicate, may be sent to any one of the four editors. Please classify according to the scheme of Math. Reviews,
Index to Vol. 39. All other communications should be addressed to the managing editor, or Elaine Barth,
University of California, Los Angeles, California, 90024.
100 reprints are provided free for each article, only if page charges have been substantially paid.
Additional copies may be obtained at cost in multiples of 50.
The Pacific Journal of Mathematics is issued monthly as of January 1966. Regular subscription rate: $ 72.00
a year (6 Vols., 12 issues). Special rate: $36.00 a year to individual members of supporting institutions.
Subscriptions, orders for back numbers, and changes of address should be sent to Pacific Journal of
Mathematics, 103 Highland Boulevard, Berkeley, California, 94708.
PUBLISHED BY PACIFIC JOURNAL OF MATHEMATICS, A NON-PROFIT CORPORAON
Printed at Jerusalem Academic Press, POB 2390, Jerusalem, Israel.
Copyright 1977 Pacific Journal of Mathematics
All Rights Reserved

Pacific Journal of Mathematics


Vol. 70, No. 2

October, 1977

B. Arazi, A generalization of the Chinese remainder theorem . . . . . . . . . . . . . .


Thomas E. Armstrong, Polyhedrality of infinite dimensional cubes . . . . . . . . .
Yoav Benyamini, Mary Ellen Rudin and Michael L. Wage, Continuous
images of weakly compact subsets of Banach spaces . . . . . . . . . . . . . . . . .
John Thomas Burns, Curvature functions on Lorentz 2-manifolds . . . . . . . . . .
Dennis F. De Riggi and Nelson Groh Markley, Shear distality and
equicontinuity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Claes Fernstrm, Rational approximation and the growth of analytic
capacity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Pl Fischer, On some new generalizations of Shannons inequality . . . . . . . . .
Che-Kao Fong, Quasi-affine transforms of subnormal operators . . . . . . . . . . .
Stanley P. Gudder and W. Scruggs, Unbounded representations of
-algebras . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Chen F. King, A note on Drazin inverses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Ronald Fred Levy, Countable spaces without points of first countability . . . .
Eva Lowen-Colebunders, Completeness properties for convergence
spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Calvin Cooper Moore, Square integrable primary representations . . . . . . . . .
Stanisaw G. Mrwka and Jung-Hsien Tsai, On preservation of
E-compactness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Yoshiomi Nakagami, Essential spectrum 0() of a dual action on a von
Neumann algebra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
L. Alayne Parson, Normal congruence subgroups of the Hecke groups
G(2(1/2) ) and G(3(1/2) ) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Louis Jackson Ratliff, Jr., On the prime divisors of zero in form rings . . . . . .
Caroline Series, Ergodic actions of product groups . . . . . . . . . . . . . . . . . . . . . . .
Robert O. Stanton, Infinite decomposition bases . . . . . . . . . . . . . . . . . . . . . . . . . .
David A. Stegenga, Sums of invariant subspaces . . . . . . . . . . . . . . . . . . . . . . . . .

289
297
309
325
337
347
351
361
369
383
391
401
413
429
437
481
489
519
549
567

You might also like