See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/255965853

Maps preserving the fixed points of sum of operators

Article in Operators and Matrices · August 2013

DOI: 10.7153/oam-09-34 · Source: arXiv

CITATIONS READS

4 257

3 authors:

Ali Taghavi Roja Hosseinzadeh

University of Mazandaran University of Mazandaran
87 PUBLICATIONS 375 CITATIONS 23 PUBLICATIONS 73 CITATIONS

SEE PROFILE SEE PROFILE

Vahid Darvish
Nanjing University of Information Science & Technology
48 PUBLICATIONS 232 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Linear preservers of binary relations View project

All content following this page was uploaded by Roja Hosseinzadeh on 12 January 2023.

The user has requested enhancement of the downloaded file.

 Operators
and
Matrices
Volume 9, Number 3 (2015), 563–569 doi:10.7153/oam-09-34

MAPS PRESERVING THE FIXED POINTS OF SUM OF OPERATORS

A LI TAGHAVI , ROJA H OSSEINZADEH AND H AMID ROHI

(Communicated by C.-K. Li)

Abstract. Let B(X ) be the algebra of all bounded linear operators on a complex Banach space
X with dim X > 2 . In this paper, we characterize the maps on B(X ) which preserve the
fixed points of sum of operators. Moreover, if X is a finite dimensional Banach space, we
also characterize the maps on B(X ) which preserve the dimension of fixed points of sum of
operators.

1. Introduction

The study of maps on operator algebras preserving certain properties is a topic

which attracts much attention of many authors (see [1]–[18] and the references cited
there.) Some of these problems are concerned with preserving a certain property of
sum or products of operators (see [1]–[11] and [16]–[18]).
Let B(X ) denote the algebra of all bounded linear operators on a complex
Banach space X . Recall that x ∈ X is a fixed point of an operator A ∈ B(X ),
whenever we have Ax = x . For A ∈ B(X ) , denote by LatA and F(A) the lattice
of A, that is, the set of all invariant subspaces of A and the set of all fixed points of
A, respectively. Authors in [3] characterize the maps φ : B(X ) → B(X ) which
satisfy one of the following preserving properties: Lat(A + B) = Lat(φ (A) + φ (B)),
or Lat(AB) = Lat(φ (A)φ (B)), or Lat(AB + BA) = Lat(φ (A)φ (B) + φ (B)φ (A)), or
Lat(ABA) = Lat(φ (A)φ (B)φ (A)), or Lat(AB − BA) = Lat(φ (A)φ (B) − φ (B)φ (A)).
Since F(A) ∈ LatA, one can replace the lattice preserving property by the fixed
points preserving property. Moreover, one can consider the maps preserving the di-
mension of fixed points which is a very weaker condition. For usual product, in [16],
we characterized the surjective maps φ : B(X ) → B(X ) satisfying dim F(AB) =
dim F(φ (A)φ (B)), where dim F(T ) denotes the dimension of F(T ).
In this paper, we characterize the maps φ : B(X ) → B(X ) and φ : Mn →
Mn satisfying F(A + B) = F(φ (A) + φ (B)) and dim F(A + B) = dim F(φ (A) + φ (B)),
respectively.

Mathematics subject classification (2010): 46J10, 47B48.

Keywords and phrases: Preserver problem, operator algebra, fixed point.

c D l , Zagreb 563
Paper OaM-09-34
564 A. TAGHAVI , R. H OSSEINZADEH AND H. ROHI

2. Maps preserving the dimension of fixed points of sum of operators

Recall that two operators A and B are adjacent if A − B is of rank one.

L EMMA 2.1. Let n be an integer number such that n > 2 . Suppose that φ :
Mn −→ Mn is a map which satisfies

dim F(A + B) = dim F(φ (A) + φ (B)) (A, B ∈ Mn ).

Then the following statements are hold:

(i) φ is injective.
(ii) If φ is surjective, then φ preserves adjacency in both directions.

Proof. (i) From n = dim F(A + I − A) = dim F(φ (A) + φ (I − A)) we obtain

I = φ (A) + φ (I − A) (2.1)

for every A ∈ Mn . Let φ (A1 ) = φ (A2 ). By (2.1) we have

n = dim F(φ (A1 ) + I − φ (A2 ))

= dim F(φ (A1 ) + φ (I − A2 ))
= dim F(A1 + I − A2 ),

which implies that A1 + I − A2 = I and so A1 = A2 .

(ii) Let A and B be two matrices such that A − B is of rank one. Preserving
property of φ together with (2.1) implies that

n − 1 = dim ker(A − B) = dim F(A − B + I)

= dim F(φ (A) + φ (I − B))
= dim F(φ (A) + I − φ (B))
= dim ker(φ (A) − φ (B))

which implies that φ (A) − φ (B) is of rank one and so φ preserves adjacency. Since
φ −1 has the preserving property of φ , we can conclude that φ preserves adjacency in
both directions. 

T HEOREM 2.2. Let n be an integer number such that n > 2 . Suppose that φ :
Mn −→ Mn is a surjective map which satisfies

dim F(A + B) = dim F(φ (A) + φ (B)) (A, B ∈ Mn ).

Then there exists a matrix R ∈ Mn and invertible matrices U, S ∈ Mn such that φ (A) =
US−1Aσ S + R or φ (A) = US−1Atσ S + R, for every A ∈ Mn , where σ is an automor-
phism of C and Aσ = [σ (ai j )] for A = [ai j ].
 M APS PRESERVING THE FIXED POINTS OF SUM OF OPERATORS 565

Proof. By Lemma 2.1, φ is injective and so bijective and preserves adjacency in

both directions. By fundamental theorem of geometry of matrices [8], the forms of
bijective adjacency preserving map φ : Mn −→ Mn is φ (A) = TAσ S + R or φ (A) =
TAtσ S + R, where R is a matrix, T, S are invertible matrices, σ is an automorphism of
the underlying field and Aσ = [σ (ai j )] for A = [ai j ].
Let the first case occurs. From (2.1) we have φ (0) + φ (I) = I which implies that
R + T Iσ S + R = I . Since Iσ = I , we obtain T S = I − 2R. Setting U = I − 2R, we obtain
T = US−1 . Therefore, φ (A) = US−1Aσ S + R. Since

dim F(2R)) = dim F(2φ (0)) = dim F(0) = 0,

F(2R) = ker(I − 2R) = 0 and so I − 2R is invertible. In a similar way, we can obtain

the second case. 

3. Maps preserving the fixed points of sum of operators

We recall some notations. X ∗ denotes the dual space of X . For every nonzero
x ∈ X and f ∈ X ∗ , the symbol x ⊗ f stands for the rank one linear operator on X
defined by (x ⊗ f )y = f (y)x for every y ∈ X . Note that every rank one operator in
B(X ) can be written in this way. The rank one operator x ⊗ f is idempotent if and
only if f (x) = 1 . We denote by F1 (X ) and P1 (X ) the set of all rank one operators
and the set of all rank one idempotent operators on X , respectively.
Let x ⊗ f be a rank one operator. It is easy to check that x ⊗ f is an idempotent if
and only if F(x⊗ f ) = hxi (the linear subspace spanned by x ). If x⊗ f isn’t idempotent,
then F(x ⊗ f ) = {0} .
Let x, y ∈ X . We denote by Gcv{x, y} = {λ x + (1 − λ )y : λ ∈ C} the generalized
convex combination of x and y.
In order to prove the main results of this section, first we prove some auxiliary
lemmas. In the following lemmas assume that dim X > 3 .

L EMMA 3.1. [5] Let A, B ∈ B(X ) be non-scalar operators. Suppose that for
every such x ∈ X that x and Ax are linearly independent or that x = Ax , Bx ∈
Gcv{x, Ax} . Then B = λ I + (1 − λ )A for some λ ∈ C \ {1} .

L EMMA 3.2. Let A, B ∈ B(X ) be non-scalar operators. If F(A + P) = F(B +

P), for every P ∈ P1 (X ), B = λ I + (1 − λ )A, for some λ ∈ C \ {1} .

Proof. By Lemma 3.1, it is enough to consider the following two cases.

Case 1. Let x and Ax be linear independent. So there exists a linear functional f
such that f (x) = 1 and f (Ax) = 0 . Setting P = (x − Ax) ⊗ f yields that (A + P)x = x
which implies that (B + P)x = x and so Ax = Bx . Therefore, Bx ∈ Gcv{x, Ax} .
Case 2. Let x = Ax . There exists a vector z ∈ X such that x and z are linear
independent and so there exists a linear functional f such that f (x) = 0 and f (z) = 1 .
Setting P = z ⊗ f yields that (A + P)x = x which implies that (B + P)x = x and so
Ax = Bx . Therefore, Bx ∈ Gcv{x, Ax} . 
566 A. TAGHAVI , R. H OSSEINZADEH AND H. ROHI

L EMMA 3.3. Let A, B ∈ B(X ). If F(A+ R) = F(B+ R), for every R ∈ F1 (X ),

then A = B.

Proof. Let x ∈ X . If x and Ax is linear independent, then there exists a linear

functional f such that f (x) = f (Ax) = 1 , because dim X > 3 . Setting R = (x−Ax)⊗ f
yields that x ∈ F(A + R) which implies that x ∈ F(B + R) and so Ax = Bx .
Let Ax = ax for a nonzero complex number a . There exists a linear functional f
such that f (x) = 1 . Setting R = (1 − a)x ⊗ f yields that x ∈ F(A + R) which implies
that x ∈ F(B + R) and so Ax = Bx . The proof is complete. 

L EMMA 3.4. Let φ : B(X ) −→ B(X ) be a map which satisfies

F(A + B) = F(φ (A) + φ (B)) (A, B ∈ B(X )).

Then the following statements are hold:

(i) φ is injective.
(ii) φ (P) = UP + R for every rank one idempotent P, where U = I − 2φ (0) and
R = φ (0).

Proof. (i) From

X = F(A + I − A) = F(φ (A) + φ (I − A)),

we obtain
I = φ (A) + φ (I − A), (3.1)
for every A ∈ B(X ) . Let φ (A1 ) = φ (A2 ). By (3.1) we have

X = F(φ (A1 ) + I − φ (A2 )) = F(φ (A1 ) + φ (I − A2 )) = F(A1 + I − A2 ),

which implies that A1 + I − A2 = I and so A1 = A2 .

(ii) For every nonzero x ∈ X and nonzero f ∈ X ∗ , we have

ker( f ) = F(x ⊗ f + I) = F(φ (x ⊗ f ) + φ (I))

which implies that

φ (x ⊗ f ) + φ (I) = I (3.2)
or there exists a vector y ∈ X and a linear functional g ∈ X ∗ such that

φ (x ⊗ f ) + φ (I) = y ⊗ g + I. (3.3)

By (3.1) we have
φ (I) + φ (0) = I (3.4)
and so if (3.2) holds, then φ (x ⊗ f ) = φ (0). This is a contradiction, because φ is
injective. Thus (3.3) holds and so ker( f ) = F(y ⊗ g + I) = ker(g) which implies that
 M APS PRESERVING THE FIXED POINTS OF SUM OF OPERATORS 567

f and g are linear dependent. Without loss of generality, we can assume that f = g
and hence
φ (x ⊗ f ) + φ (I) = y ⊗ f + I. (3.5)
Let f (x) = 1 . By (3.4) and (3.5) we have

hxi = F(x ⊗ f ) = F(φ (x ⊗ f ) + φ (0))

= F(y ⊗ f + I − φ (I) + φ (0))
= F(y ⊗ f + 2φ (0)),

which implies that (y ⊗ f + 2φ (0))x = x and so y = (I − 2φ (0))x . This together with

(3.4) and (3.5) yields that φ (x ⊗ f ) = (I − 2φ (0))x ⊗ f + φ (0) which completes the
proof. 

T HEOREM 3.5. Let X be a complex Banach space with dim X > 2 . Suppose
that φ : B(X ) −→ B(X ) is a surjective map which satisfies

F(A + B) = F(φ (A) + φ (B)) (A, B ∈ B(X )).

Then φ (A) = UA + R for every A ∈ B(X ) , where U = I − 2φ (0) and R = φ (0).

Proof. If dim X = 2 , from Theorem 2.2 we can conclude that there exists a matrix
R ∈ M2 and invertible matrices U, S ∈ M2 such that φ (A) = US−1Aσ S + R or φ (A) =
US−1Atσ S + R, for every A ∈ M2 , where σ is an automorphism of C. Suppose the first
case occurs. This by assumption yields that

F(A + B) = F(U(S−1 (Aσ + Bσ )S) + 2R)

for every A, B ∈ M2 . It is easy to see that for arbitrary operators A, B ∈ B(X ),

F(S−1 AS) = S−1 (F(A)) and also F(A + B) = F(U(A + B) + 2R). Therefore, we obtain

SF(A + B) = F(Aσ + Bσ )

for every A, B ∈ M2 . Replacing A and B by

 
10
A1 = , B1 = 0
00

and then  
00
A2 = , B2 = 0
01
yields that S(x, 0) = (x, 0) and S(0, y) = (0, y), for all x, y ∈ C and hence S is the
identity operator. Thus we obtain

F(A + B) = F(Aσ + Bσ )
568 A. TAGHAVI , R. H OSSEINZADEH AND H. ROHI

for every A, B ∈ M2 . Let a ∈ C be nonzero and set

 
10
C= , D = 0.
a0
Replacing A and B by C and D yields that
{(x, ax); x ∈ C} = {(x, σ (a)x); x ∈ C}
which implies that σ (a) = a , for every a ∈ C and so σ is the identity automorphism.
Therefore, φ (A) = UA + R for every A ∈ M2 .
The second case can not occur. Otherwise, with a very similar way as above
discussion we obtain
F(A + B) = F(Atσ + Btσ )
for every A, B ∈ M2 . Again replacing A and B by C and D, yields that
{(x, ax); x ∈ C} = {(x, 0); x ∈ C}
which implies that a = 0 , a contradiction.
Now let dim X > 3 . Since F(2φ (0)) = F(0) = 0 , ker(I − 2φ (0)) = 0 and so
U = I − 2φ (0) is injective. Let ψ be a map on B(X ) such that U ψ = φ − R, where
R = φ (0). The injectivity of U yields that ψ is well-defined and that ψ satisfies
the preserving property of φ and also by Lemma 3.4, ψ (P) = P for every rank one
idempotent P. Therefore, without loss of generality, we can assume that U = I , R = 0
and so φ (P) = P for every rank one idempotent P.
We divide the proof into the following steps.
Step 1. φ (aI) = aI , for every a ∈ C.
Let A = aI , for a nonzero complex number a and set φ (A) = B. Let x be a
nonzero arbitrary vector of X . Then there exists a linear functional f such that
f (x) = 1 . Setting P = (x − Bx) ⊗ f yields that (B + P)x = x which implies that
(A + φ −1(P))x = x . So we obtain
(A + P)x = x ⇒ ax + x − Bx = x ⇒ Bx = ax.
Step 2. φ (A) = A for every rank one operator A.
Let x ∈ X and f ∈ X ∗ be nonzero. Since φ (I) = I , by (3.5), there exists a
y ∈ X such that
φ (x ⊗ f ) = y ⊗ f . (3.6)
On the other hand, by Lemma 3.2, there exists an λ ∈ C \ {1} such that
φ (x ⊗ f ) = λ I + (1 − λ )x ⊗ f , (3.7)
because by Step 1, φ (x ⊗ f ) is a non-scalar operator. From (3.6) and (3.7) we obtain
λ = 0 and y = x and so φ (x ⊗ f ) = x ⊗ f .
Step 3. φ (A) = A for every operator A ∈ B(X ).
Assertion follows from Step 2 and Lemma 3.3. 

Acknowledgements. The authors wish to thank the referee for many helpful com-
ments.
 M APS PRESERVING THE FIXED POINTS OF SUM OF OPERATORS 569

REFERENCES

[1] M.A. C HEBOTAR , W.-F. K E , P.-H. L EE , N.-C. W ONG, Mappings preserving zero products, Studia
Math. 155 (2003), 77–94.
[2] M. D OBOVI ŠEK , B. K UZMA , G. L E ŠNJAK , C. K. L I , T. P ETEK, Mappings that preserve pairs of
operators with zero triple Jordan Product, Linear Algebra Appl. 426 (2007), 255–279.
[3] G. D OLINAR , S. D U , J. H OU , P. L EGI ŠA, General preservers of invariant subspace lattices, Linear
Algebra Appl. 429 (2008), 100–109.
[4] L. FANG , G. J I , Linear maps preserving products of positive or Hermitian matrices, Linear Algebra
Appl. 419 (2006), 601–611.
[5] L. FANG , G. J I , Y. PANG, Maps preserving the idempotency of products of operators, Linear Algebra
Appl. 426 (2007), 40–52.
[6] J. H OU , Q. D I , Maps preserving numerical ranges of operator products, Proc. Amer. Math. Soc. 134
(2006), 1435–1446.
[7] J. H OU , C. K. L I , N. C. W ONG, Jordan isomorphisms and maps preserving spectra of certain oper-
ator products, Studia Math. 184 (2008), 31–47.
[8] L.K. H AU, A theorem on matrices over a field and its applications, Acta Math. Sinica 1 (1951),
109–163.
[9] J. H OU , L. Z HAO, Zero-product preserving additive maps on symmetric operator spaces and self-
adjoint operator spaces, Linear Algebra Appl. 399 (2005), 235–244.
[10] G. J I , Y. G AO, Maps preserving operator pairs whose products are projections, Linear Algebra Appl.
433 (2010), 1348–1364.
[11] C. K. L I , P. Š EMRL , N. S. S ZE, Maps preserving the nilpotency of products of operators, Linear
Algebra Appl. 424 (2007), 222–239.
[12] P. Š EMRL, Two characterizations of automorphisms on B(X ) , Studia Math. 105 (1993), 143–148.
[13] P. Š EMRL , N. S. S ZE, Non-linear commutativity preserving maps, Acta Scince Math. (Szeged) 71
(2005), 781–819.
[14] A. TAGHAVI , Additive mappings on C∗ -algebras Preserving absolute values, Linear and Multilinear
Algebra 60, 1 (2012), 33–38.
[15] A. TAGHAVI , R. H OSSEINZADEH, Linear maps preserving idempotent operators, Bull. Korean Math.
Soc. 47 (2010), 787–792.
[16] A. TAGHAVI , R. H OSSEINZADEH, Maps preserving the dimension of fixed points of products of op-
erators, Linear and Multilinear Algebra, accepted.
[17] M. WANG , L. FANG , G. J I , Y. PANG, Linear maps preserving idempotency of products or triple
Jordan products of operators, Linear Algebra Appl. 429 (2008), 181–189.
[18] L. Z HAO , J. H OU, Jordan zero-product preserving additive maps on operator algebras, J. Math. Anal.
Appl. 314 (2006), 689–700.

(Received May 21, 2014) Ali Taghavi

Department of Mathematics, Faculty of Mathematical Sciences
University of Mazandaran
P. O. Box 47416-1468 Babolsar, Iran
e-mail: Taghavi@umz.ac.ir
Roja Hosseinzadeh
Department of Mathematics, Faculty of Mathematical Sciences
University of Mazandaran
P. O. Box 47416-1468, Babolsar, Iran
e-mail: ro.hosseinzadeh@umz.ac.ir
Hamid Rohi
Department of Mathematics, Faculty of Mathematical Sciences
University of Mazandaran
P. O. Box 47416-1468, Babolsar, Iran

Operators and Matrices

www.ele-math.com
oam@ele-math.com

View publication stats