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

net/publication/281110636

Mapping preserving numerical range of operator products on $C^*$-algebras

Article in Bulletin of the Korean Mathematical Society · June 2015

CITATIONS READS

0 241

1 author:

Mabrouk Mohamed Ben Salah

Umm Al-Qura University
29 PUBLICATIONS 137 CITATIONS

SEE PROFILE

All content following this page was uploaded by Mabrouk Mohamed Ben Salah on 20 August 2015.

The user has requested enhancement of the downloaded file.

Bull. Korean Math. Soc.

MAPPING PRESERVING NUMERICAL RANGE OF

OPERATOR PRODUCTS ON C ∗ -ALGEBRAS

Mohamed Mabrouk

Abstract. Let A and B be two unital C ∗ -algebras. Denote by W (a)

the numerical range of an element a ∈ A. We show that the condition
W (ax) = W (bx), ∀x ∈ A implies that a = b. Using this, among other
results, it is proved that if ϕ : A −→ B is a surjective map such that
W (ϕ(a)ϕ(b)ϕ(c)) = W (abc) for all a, b and c ∈ A, then ϕ(1) ∈ Z(B) and
the map ψ = ϕ(1)2 ϕ is multiplicative.

1. Introduction
Let A be a C ∗ -algebra with unit 1 and let S(A) be the state space of A, i.e.,
′
S(A) = {φ ∈ A′ : φ ≥ 0, φ(1) = 1} (here A is the topological dual of A). For
each a ∈ A, the algebraic numerical range V (a) and numerical radius v(a) are
defined by
V (a) = {f (a) : f ∈ S(A)} and v(a) = sup |z|.
z∈V (a)
By the Gelefand-Naimark theorem, every C ∗ -algebra may be viewed as a closed
∗-subalgebra of B(H) where B(H) denotes the algebra of all bounded linear
operators acting on a Hilbert space H. It is well known that V (a) is the closure
of W (a) and v(a) = w(a) = sup |λ|, where W (a) = {(at, t) : t ∈ H, ∥t∥ = 1}
λ∈W (a)
and (, ) denotes the inner product. Here W (a) is called the usual numerical
range of the operator a.
In the last few decades, there has been a considerable interest in the problem
of characterization of maps that preserves the numerical range or the numerical
radius, see for instance the papers [3, 11, 12, 14] and the references therein.
Notice that, based on the aforesaid, preserving the usual numerical range W
implies the preservation of the spacial numerical range V . Therefore, we will
concentrate our study henceforth on W . Recently, Hou and Di described in [8]
surjective maps on the algebra B(H) which preserves the numerical range of
the product. Namely, they characterized surjective mappings which satisfy one
of the following conditions
(1a) W (ϕ(a)ϕ(b)) = W (ab),
(1b) W (ϕ(a)∗ ϕ(b)) = W (a∗ b),

2000 Mathematics Subject Classification. 15A86; 46L05; 47A12; 47B49.

Key words and phrases. C ∗ -algebras, Numerical range, Preserving the numerical range.
1
2 M. MABROUK

(1c) W (ϕ(a)ϕ(b)ϕ(a)) = W (aba),

for every a and b in B(H). In this paper, we extend these results by completely
describing additive and surjective maps ϕ : A −→ B between C ∗ -algebras
satisfying (1a) or (1b) for every a, b ∈ A. Concerning the condition (1c), we
consider a more general case. More precisely, we show that if ϕ is surjective
and satisfy W (ϕ(a)ϕ(b)ϕ(c)) = W (abc), ∀a, b and c ∈ A (without the additivity
assumption), then the map ψ = ϕ(1)2 ϕ is multiplicative and preserves the set
of self-adjoint elements. It is worth noticing that our proofs differ from those
of [6] and [8] since we do not assume that A contains rank one operators. At
last, observe that the proof we put forward is much simpler.
The outline of the paper is as follows. Firstly, we show that if a and b in
A are such that W (ax) = W (bx), for every x ∈ A, then the two operators a
and b coincide. This result is used several times in our proofs. Namely, it helps
us to show that if ϕ is additive and satisfies (1a) or (1b), then ϕ(1) ∈ Z(B),
where Z(B) stands for the center of B, and ϕ(1)ϕ is a Jordan ∗-isomorphism.
This characterization also allows us to show that if a map ϕ is surjective and
satisfies W (abc) = W (ϕ(a)ϕ(b)ϕ(c)) whenever a, b and c are in A, then the
map ϕ(1)2 ϕ is multiplicative and therefore ϕ has standard forms when A and
B are the algebras of all bounded linear operators acting on a Hilbert space.

2. Preliminaries
In this section, we collect some properties of the numerical range needed in
the sequel. Let two unital C ∗ -algebras A ⊂ B(H) and B ⊂ B(K) be given.
By Sp(a) (resp. r(a)) we denote the spectrum (resp. the spectral radius) of an
element a ∈ A. Since it does not lead to misunderstanding, we shall denote the
norms in both algebras by the same symbols ∥.∥. We denote by H(A) the set
of self adjoint elements defined by H(A) = {a ∈ A : a = a∗ }. It is well-known
that a ∈ H(A) if and only if W (a) ⊂ R. Further, an element a ∈ A is positive
if and only if W (a) ⊂ R+ (or equivalently a = a∗ and Sp(a) ⊂ R+ ), where
R+ denotes the set of positive real numbers. In the case where A = C(K) for
some Hausdorff compact space K we have W (a) ⊂ V (a) = co(a(K)) for each
a ∈ C(K), see [16, Theorem 6.]. Here co stands for closed convex hull. We
summarize some other basic properties of the numerical range on the following
lemma. One may see [1, 7] for more information.
Lemma 2.1.
(i) ∥a∥ = w(a) = r(a) for every a ∈ A such that aa∗ = a∗ a.
(ii) W (a) = {λ} ⇐⇒ a = λ1, for every a ∈ A and λ ∈ C.
Finally, recall that a linear map ψ : A → B is called unital if ψ(1) = 1,
and it is said to be a Jordan homomorphism if ψ(a2 ) = ψ(a)2 for all a ∈ A.
Equivalently, the map ψ is a Jordan homomorphism if and only if ψ(ab + ba) =
ψ(a)ψ(b) + ψ(b)ψ(a) for all a and b in A. We also recall that the map ψ is
∗
said to be self-adjoint provided that ψ(a∗ ) = ψ(a) for all a ∈ A. Self-adjoint
MAPPING PRESERVING NUMERICAL RANGE OF OPERATOR PRODUCTS ON C ∗ -ALGEBRAS
3

Jordan homomorphisms are called Jordan ∗-homomorphisms, and by a Jordan

∗-isomorphism, we mean a bijective ∗-homomorphism.

3. Main results and proofs

We start with the following introductory results, which may be of inde-
pendent interest. We give a characterization of elements a, b ∈ A satisfying
W (ax) = W (bx), ∀x ∈ A or w(ax) = w(bx), ∀x ∈ H(A). It is worth observ-
ing that the authors in [2] have recently considered the question whether the
equality Sp(ax) = Sp(bx) for every x ∈ A, where a, b ∈ A are fixed elements,
implies a = b. An affirmative answer has been obtained for some classes of
algebras, including C ∗ -algebras.
We begin with the following Proposition which gives necessary conditions
which ensure that a = b if w(ax) = w(bx), ∀x ∈ H(A). The argument of
the proof is borrowed from [10, Lemma 3.4] by slight some modifications. We
present it here for the sake of completeness.

Proposition 3.1. Let A be an unital C ∗ -algebra and a, b ∈ A be be two positive

elements such that ab = ba. If w(ax) = w(bx), for every x ∈ H(A), then a = b.

Proof. Let B be the unital C ∗ -algebra B generated by a and b. Since, ab = ba,

this algebra is commutative. Henceforth, without loss of generality we may
suppose that A is a commutative C ∗ -algebra. On the other hand, it is well
known that every positive element in a C ∗ -algebra has unique square root,
then to prove that a = b, it suffices to show that a2 = b2 . Suppose to the
contrary that a2 ̸= b2 . Since a2 − b2 is self-adjoint, there exists a non-zero
β ∈ Sp(a2 − b2 ). We may assume that β > 0 (otherwise, we could replace
1
a2 − b2 by b2 − a2 ). Let α = sup Sp(a2 − b2 ) > 0, and consider the real
2
valued continuous function f defined on the spectrum of a2 − b2 such that
f (2α) = 1, 0 ≤ f (λ) ≤ 1, ∀λ ∈ Sp(a2 − b2 ) and f (λ) = 0 ⇐⇒ λ ≤ α. Put
x1 = f (a2 − b2 ) and g(λ) = λf (λ)2 , ∀λ ∈ Sp(a2 − b2 ). So, using functional
calculus (see [5, Theorem 2.9]) and the fact that x1 (a2 − b2 )x1 is self adjoint,
we get w(x1 (a2 − b2 )x1 ) = r(x1 (a2 − b2 )x1 ) = supλ∈Sp(a2 −b2 ) |g(λ)| = 2α. In
addition by using the same argument, it is easily shown that x1 (a2 − b2 )x1 ≥
αx21 . Now for t ∈ H such that ∥t∥ = 1, let us define the positive linear
form φt by φt (a) = (at, t), ∀a ∈ A. Since ∥b∥x21 − x1 b2 x1 is positive, we have
∥b∥φt (x21 ) ≥ φt (x1 b2 x1 ) ≥ 0. On account of x1 (a2 −b2 )x1 ≥ αx21 , it follows that
α
φt (x1 a2 x1 ) ≥ (1+ )φt (x1 b2 x1 ). Since 0 ≤ φt (x1 a2 x1 ) ≤ w(x1 a2 x1 ), we infer
∥b∥ √
α α
that w(x1 a2 x1 ) ≥ (1+ )w(x1 b2 x1 ). Accordingly ∥ax1 ∥ ≥ 1 + ∥bx1 ∥ >
∥b∥ ∥b∥
∥bx1 ∥. This obviously contradicts the hypothesis of the Proposition, since
x1 ∈ H(A) and by Lemma 2.1, we have ∥ax1 ∥ = w(ax1 ) = w(bx1 ) = ∥bx1 ∥.
Thus a = b as desired. 
4 M. MABROUK

The next two Propositions are crucial for the rest of the paper. They give
a characterization of elements a, b ∈ A satisfying W (ax) = W (bx), for every
x ∈ A (or H(A)).
Proposition 3.2. Let A be an unital C ∗ -algebra and a, b ∈ A. If W (ax) =
W (bx), ∀x ∈ A, then a = b.
Proof. Firstly, assume that a = a∗ . Since W (b) = W (a) ⊂ R, then b∗ = b. On
the other hand, by observing that W (a2 ) = W (ba) and the fact that a2 is self
adjoint, we infer that (ba)∗ = ba. By taking into account that (ba)∗ = a∗ b∗ and
that a and b are self adjoint, we get ab = ba. We prove now that a = b. Let B
be the C ∗ -algebra B generated by a and b. Since a and b are self adjoint and
satisfy ab = ba, this algebra is commutative. Hence, it can be identified with
C(K), the algebra of all continuous functions on a compact Hausdorff space K.
Observe also that W (ax) = W (bx), ∀x ∈ B. We claim that the two functions
a and b have the same sign (both positive, or both negative on K). Indeed,
assuming that there exists t1 ∈ K such that a(t1 ) > 0 and b(t1 ) < 0. Therefore
there exists an open set U1 such that a(t) > 0 and b(t) < 0, for all t ∈ U1 .
By Urysohn’s lemma, one can find a continuous function c1 : K −→ [0, 1]
satisfying c(t1 ) = 1 and supp(c) ⊂ U1 . On the other hand W (ac) ⊂ co(ac(K))
and W (bc) ⊂ co(bc(K)). Then, we get W (ac) ⊂ [0, +∞) and W (bc) ⊂ (−∞, 0].
Observe that the case where W (ac) = W (bc) = {0}, does not occur since
ac ̸= 0. Therefore W (ac) ̸= W (bc), contrary to our assumption. Thus a
and b have the same sign as suggested above. Consequently, without loss of
generality, we can suppose that a, b are positive on U . From the condition
W (ax) = W (bx), ∀x ∈ B, we infer that w(ax) = w(bx), ∀x ∈ B. It follows from
Proposition 3.1 that a = b. We return now to the general case; i.e., if a ∈ A
is arbitrary. Observe that, by assumption, we have, W (aa∗ x) = W (ba∗ x) and
W (ab∗ x) = W (bb∗ x), ∀x ∈ A. Based on the aforesaid, we infer that aa∗ = ba∗
and ab∗ = bb∗ . Accordingly, (a − b)(a∗ − b∗ ) = 0, which implies that a = b.
This ends the proof. 
Proposition 3.3. If A is a unital C ∗ -algebra and a and b ∈ H(A). If W (ax) =
W (xb), ∀x ∈ H(A) or W (ax) = W (bx), ∀x ∈ H(A) , then a = b.
Proof. We give the proof for the condition W (ax) = W (xb), ∀x ∈ H(A). For
the other condition the proof is similar. By using a similar reasoning as above,
we can easily prove that ab = ba. Considering the commutative C ∗ -algebras B
generated by a, b. and by taking into account that W (ax) = W (bx), ∀x ∈ H(B).
By a similar reasoning as in the proof of the above proposition, we can show that
a and b are either both positive or negative. We infer by means of Proposition
3.1 that a = b. 
Remark 3.4. The results of Proposition 3.2 and 3.3 are still valid if we replace
the numerical range by its closure. That is if two elements a and b satisfy
V (ax) = V (bx) for every x ∈ A (or in H(A)), then a similar argument can be
used to show that a = b.
MAPPING PRESERVING NUMERICAL RANGE OF OPERATOR PRODUCTS ON C ∗ -ALGEBRAS
5

At this juncture, we are in a position to characterize surjective mapping satis-

fying
(2a) W (ϕ(a)ϕ(b)ϕ(c)) = W (abc) for all a, b and c ∈ A
(2b) W (ϕ(a)ϕ(b)ϕ(c)) = W (abc) for all a, b and c ∈ H(A).
Theorem 3.5. Let A and B be two unital C ∗ -algebras. Let ϕ : A −→ B
be a surjective mapping satisfying (2a). Then ϕ(1) ∈ Z(B), ϕ(1)3 = 1, and
ϕ satisfies ϕ(ab) = ϕ(1)2 ϕ(a)ϕ(b) for all a and b ∈ A. In particular, the
mapping ψ = ϕ(1)2 ϕ is multiplicative and preserves self-adjoint elements (i.e
ψ(a) ∈ H(B) whenever a ∈ H(A)).
Proof. Set u = ϕ(1). Take a = b = c = 1 in (2a), we obtain W (u3 ) = W (1) =
{1}. Thus u3 = 1 and hence u is invertible. Given a, b ∈ A such that ϕ(a) =
ϕ(b). By (2a), we have W (ac) = W (uϕ(a)ϕ(c)) = W (uϕ(b)ϕ(c)) = W (bc), for
every c ∈ A. By Proposition 3.2, we infer that a = b and ϕ is bijective as
desired. Also, we have W (uϕ(a)ϕ(b)) = W (1ab) = W (a1b) = W (ϕ(a)uϕ(b)).
Thus we get W (uϕ(a)ϕ(b)) = W (ϕ(a)uϕ(b)), ∀b ∈ A. Since ϕ is a bijection and
on account of Proposition 3.2, we get uϕ(a) = ϕ(a)u, ∀a ∈ A. That is to say
that u ∈ Z(B). To end the proof, observe that
W (ϕ(a)ϕ(b)ϕ(c)) = W (abc) = W (1(ab)c)
= W (uϕ(ab)ϕ(c)), ∀c ∈ A

By recalling that ϕ is bijective, again Proposition 3.2 implies that ϕ(ab) =

u−1 ϕ(a)ϕ(b) = u2 ϕ(a)ϕ(b). Now, put ψ = u2 ϕ. We have
ψ(a)ψ(b) = u2 ϕ(a)u2 ϕ(b) = uϕ(a)ϕ(b) = u2 ϕ(ab) = ψ(ab).
Finally, observe that ϕ preserves self-adjoint elements because for every self
adjoint element a ∈ A, we have W (ψ(a)) = W (ϕ(1)2 ϕ(a)) = W (a) ⊂ R. The
proof is thus complete. 
Corollary 3.6. Let A and B be two unital C ∗ -algebras. Let ϕ : A −→ B be a
surjective mapping satisfying
(3) W (ϕ(a1 )ϕ(a2 ) . . . ϕ(ap )) = W (a1 a2 . . . ap ) for all a1 a2 . . . ap ∈ A
for some integer p ∈ N with p ≥ 3. Then ϕ(1) ∈ Z(B), ϕ(1)p = 1 and
ϕ(1)p−1 ϕ(a1 a2 . . . ap ) = ϕ(a1 )ϕ(a2 ) . . . ϕ(ap ) for all a1 a2 . . . ap ∈ A.
Proof. It is obvious that ϕ(1)p = 1. If ϕ(a) = ϕ(b) for some a, b ∈ A. Since
W (ϕ(a)ϕ(x)ϕ(1)p−2 ) = W (ϕ(b)ϕ(x)ϕ(1)p−2 ), by (3), it yields that W (ax) =
W (bx) for every x ∈ A. By Proposition 3.2, we get a = b. Hence ϕ is a
bijection. Define ψ = ϕ(1)p−1 ϕ. We see that ψ(1) = 1 and
ψ(a)ψ(b)ψ(c) = ϕ(1)p−1 ϕ(a)ϕ(1)p−1 ϕ(b)ϕ(1)p−1 ϕ(c) = ϕ(1)p−3 ϕ(a)ϕ(b)ϕ(c).
From (3), we can deduce that W (ψ(a)ψ(b)ψ(c)) = W (abc), ∀a, b, c ∈ A. By
Theorem 3.5, the results follows. 
6 M. MABROUK

Based, on Theorem 3.5, we know that if ϕ satisfy (2a), then the mapping
ψ = ϕ(1)2 ϕ is multiplicative. The question of when a multiplicative map is
additive was attacked by several authors. For instance, if ψ is a bijective
map on a standard operator algebra, Molnàr showed in [13] that if ϕ satisfies
ψ(ABA) = ψ(A)ψ(B)ψ(A), then ψ is additive. Hence, based on the aforesaid,
when the algebras A and B are the algebras of all bounded linear operators
acting on some Hilbert spaces, the Theorem 3.5 can be refined as follows.
Corollary 3.7. Let H and K be complex Hilbert spaces and let ϕ : B(H) −→
B(K) be a surjective map (without the assumption of additivity). Then ϕ sat-
isfies Eq. (2a) if and only if there is a unitary operator U : H −→ K such that
ϕ is of the form ϕ(A) = εU AU ∗ for all A ∈ B(H), where ε3 = 1.
Proof. Checking the ‘if’ part is straightforward, and we therefore will only deal
with the ‘only if’ part. Assume that ϕ satisfies (2a). By Theorem 3.5, we
have that ϕ(1) ∈ Z(B(K)) and ϕ(1)3 = 1. Since the algebra B(K) has a
trivial center, then u = ϕ(1) = ε.1, where ε is a complex number such that
ε3 = 1. Also according to Theorem 3.5, the map ψ = u2 ϕ, is multiplicative
and ψ(1) = 1. Consequently, by [13] it is additive. Finally, we have shown that
ψ is an algebra isomorphism which preserves self-adjoint elements. Thus, by
[3] ψ takes the following form: ψ(A) = U AU ∗ for all A ∈ B(H) where U is
unitary. 
For mapping ϕ : H(A) −→ H(B) satisfying (2b), we have a similar result
which follows.
Theorem 3.8. Let A and B be two unital C ∗ -algebras. Let ϕ : H(A) −→ H(B)
be a surjective mapping satisfying (2b). Then ϕ(1) ∈ Z(H(B)), ϕ(1)3 = 1, and
ϕ = ϕ(1)ψ, where ψ is multiplicative.
Proof. The proof is similar to that of Theorem 3.5 by invoking Proposition 3.3.
The details are omitted. 
As a special case of Theorem 3.8 we derive the following result
Theorem 3.9. Consider the case where A = B(H) and B = B(K) for some
complex Hilbert spaces H and K. Let ϕ : H(A) −→ H(B) be a surjective map.
Then, ϕ satisfies (2b) if and only if there exists a unitary or conjugate unitary
operator U and ε ∈ C such that ϕ(A) = εU AU ∗ for all A ∈ H(A) and ε3 = 1.
Proof. The sufficiency is easy to see. Indeed, this follows from the well-known
fact that if U is a unitary or conjugate unitary operator , then W (U AU ∗ ) =
W (A) for every A ∈ A. Conversely, suppose that ϕ satisfies Equation (2b), for
every A ∈ H(A). Theorem 3.8, implies that ψ = ϕ(1)2 ϕ is multiplicative and
ψ(1) = 1. Therefore, by [15, Theorem 2.1] there exists a unitary or conjugate
unitary operator U such that ψ(A) = U AU ∗ for all A ∈ H(A). To end the
proof observe that ϕ = uψ = ϕ(1)ψ and in particular ϕ is linear. Since, by
Theorem 3.8 we have ϕ(1) ∈ Z(H(B)), we infer that ϕ(1) ∈ Z(B). Therefore,
MAPPING PRESERVING NUMERICAL RANGE OF OPERATOR PRODUCTS ON C ∗ -ALGEBRAS
7

ϕ(1) = ε.1, where ε is a complex number such that ε3 = 1. Thus, completing

the proof. 

We give now the following theorem which characterizes surjective maps sat-
isfying (1a) in the case of C ∗ -algebras. This result has been also proved in
[6, Theorem 2.1.] for the Hilbert space operators case but without the extra
condition that ϕ is additive. It would be interesting to remove the additive
assumption in the Theorem 3.10 bellow. We are not able to do that at present.
Theorem 3.10. Let A and B be two unital C ∗ -algebra and ϕ : A −→ B be a
surjective and additive map satisfying (1a). Then, ϕ is a Jordan ∗-isomorphism
followed by a left multiplication by a fixed element u ∈ Z(B) with u2 = 1, where
Z(B) stands for the center of B.
Proof. Firstly, we prove that ϕ is bijective. It suffice to show that it is injec-
tive. Let a, b ∈ A such that ϕ(a) = ϕ(b). By (1a), we have W (ϕ(a)ϕ(c)) =
W (ac) = W (ϕ(b)ϕ(c)) = W (bc), ∀c ∈ A. By Proposition 3.2, yields that a = b.
Hence we have proved that ϕ is injective. Take a = b = 1 in (1a), we obtain
W (ϕ(1)2 ) = W (1) = {1}. Whence ϕ(1) is invertible and ϕ(1)2 = 1. We show
now that ϕ is linear. Let λ ∈ C. By (1a) we have
W (λϕ(a)ϕ(b)) = λW (ϕ(a)ϕ(b)) = λW (ab) = W (λa)b) = W (ϕ(λa)ϕ(b)), ∀a, b ∈ A.
Whence, by Proposition 3.2, yields that ϕ(λa) = λϕ(a), ∀a ∈ A. Since ϕ is
additive, we infer that ϕ is a linear bijection. Now, take a, b ∈ A such that
ab = 0. By (1a), yields that ϕ(a)ϕ(b) = 0. Hence [4, Lemma 4.4], implies that
ϕ(1)ϕ(a) = ϕ(a)ϕ(1), ∀a ∈ A. Together with the bijectivity of ϕ, this implies
that ϕ(1) ∈ Z(B).
Finally, we show that ϕ has the asserted form. Set ψ = uϕ. It suffices to
show that ψ is C ∗ -isomorphism. It is obvious that ψ(1) = u2 = 1 and
W (ψ(a)ψ(b)) = W (ab), ∀a, b ∈ A. Thus, we have proved that ψ is a linear
isomorphism satisfying W (ψ(a)) = W (a), for every a ∈ A. By [14, Theorem
3.1], the result follows.


Finally, we turn to the second type of preserver problems involving involution.

We have the following result.
Theorem 3.11. Let A and B be two unital C ∗ -algebra and ϕ : A −→ B
be a surjective and additive map satisfying (1b). Then, ϕ(1) is unitary and
ϕ = ϕ(1)ψ, where ψ is a Jordan ∗-isomorphism.
Proof. Firstly, observe that by (1b), we have ∥ϕ(a)∥2 = ∥ϕ(a)ϕ(a)∗ ∥ = w(ϕ(a)ϕ(a)∗ ) =
w(aa∗ ) = ∥aa∗ ∥2 , ∀a ∈ A. Taking the square root, we obtain ∥ϕ(a)∥ = ∥a∥,
which yields that ϕ is an isometry and hence a bijection. Now, let λ ∈ C and
a ∈ A. For all b ∈ A, we have
W ((λϕ(a))∗ ϕ(b)) = λW (ϕ(a)∗ ϕ(b)) = λW (a∗ b) = W ((λa)∗ b) = W ((ϕ(λa))∗ ϕ(b))
 8 M. MABROUK

Using Proposition 3.2, we infer that (λϕ(a))∗ = (ϕ(λa))∗ . Accordingly, λϕ(a) =

ϕ(λa). In consequence of this, ϕ is a linear bijection. Thus, we have proved
that ϕ is a linear isomorphism between two C ∗ -algebras which are isometric.
By [9, Theorem 7.], ϕ is a Jordan ∗-isomorphism followed by left multiplication
by the fixed unitary operator ϕ(1). 
Acknowledgements. The author wishes to express his thanks to the referee
for carefully reading the paper and suggestions.

References
1. F. F Bonsall and J. Duncan, Numerical ranges of operators on normed spaces and ele-
ments of normed algebras, Cambridge Univ. Press, London, 1971.
2. M. Bresar and S. Spela, Determining elements in Banach algebras through spectral prop-
erties, Journal of Mathematical Analysis and Applications. 393 (2012), 144–150.
3. J. T. Chan, Numerical radius preserving operators on B(H), Proc. Amer. Math. Soc.
123 (1995), no. 5, 1437–1439.
4. M. A. Chebotar, W. F. Ke, P. K. Lee, and N. C. Wong, Mappings preserving zero
products, Studia Math 155 (2003), no. 1, 77–94.
5. J. B. Conway, A course in functional analysis, 2nd ed., Springer, 1997.
6. H. L. Gau and C. K. Li, C*-isomorphisms, Jordan isomorphisms, and numerical range
preserving maps, Proc. Amer. Math. Soc. 135 (2007), no. 9, 2907–2914.
7. P.R. Halmos, A Hilbert space problem book 2nd ed., Springer, New York, 1982.
8. J. Hou and Q. Di, Maps preserving numerical ranges of operator products, Proc. Amer.
Math. Soc. 134 (2006), no. 5, 1435–1446.
9. R. V. Kadisson, Isometries of operator algebras, Annals Of Mathematics 54 (1951), no. 2,
325–338.
10. E. C. Lance, Unitary operators on Hilbert C*-Modules, Bull. London Math. Soc. 4 (1994),
no. 363–366.
11. C. K. Li, A survey on linear preservers of numerical ranges and radii, Taiwanese Journal
of Mathematics 5 (2001), no. 3, pp–477.
12. C. K. Li and E. Poon, Maps preserving the joint numerical radius distance of operators,
Lin. Alg. Appl. 437 (2012), no. 5, 1194–1204.
13. L. Molnär, On isomorphisms of standard operator algebras, Studia Math 142 (2000), 3.
14. V. Pellegrini, Numerical range preserving operators on a Banach algebra, Studia Math
54 (1975), no. 2, 143–147.
15. A. Runling and H. Jinchuan, Additivity of Jordan multiplicative maps on Jordan operator
algebras, Taiwanese journal of mathematics 10 (2006), no. 1, pp–45.
16. J. G. Stampfli and J. P. Williams, Growth conditions and the numerical range in a
Banach algebra, Tohoku Math. J. 20 (1968), no. 4, 417–424.

Mohamed Mabrouk
Department of Mathematics, College of Applied Sciences, P. O. Box 715, Makkah
21955, KSA & Department of Mathematics, Faculty of Sciences, Gabès – Riadh city
– Zirig 6072 – Tunisia
E-mail address: [email protected]

View publication stats