Abstract: We introduce new notions of approximate amenability for a Banach algebra A. A Banach algebra A is n-approximately weakly amenable, for n ∈ N, if every continuous derivation from A into the n-th dual space A(n) is approximately inner. First we examine the relation between m-approximately weak amenability and n-approximately weak amenability for distinct m,n ∈ N. Then we investigate (2n+1)-approximately weak amenability of module extension Banach algebras. Finally, we give an example of a Banach algebra that is 1-approximately weakly amenable but not 3-approximately weakly amenable.
INTRODUCTION
Let A be a Banach algebra and X a Banach A-bimodule. A derivation from A into X is a bounded linear map satisfying:
D(ab) = a.D(b)+D(a).b (a, b ∈ A) |
For each x ∈ X we denote by adx the derivation D(a) = ax-xa for all a ∈ A, which is called inner derivation. We denote by Z1(A, X) the space of all derivations from A into X and by B1(A,X) the space of all inner derivations from A into X. The first cohomology group of A and X which is denoted by H1(A,X), is the quotient space Z1(A, X)/B1(A, X). A Banach algebra A is amenable if H1(A, X*) = 0 for each A-bimodule X (X* is the dual space of X which is an A-bimodule as usual). The concept of amenability for a Banach algebra A, introduced by Johnson (1972). The Banach algebra A is weakly amenable if H1(A, A*) = 0. Ghahramani and Loy (2004) and Dales et al. (1998) introduced several modifications of this notion. We recall the definitions in definitions 1 and 2, below:
Definition 1: A Banach algebra A is called approximately amenable
if for each A-bimodule X and for each derivation D: A→X* there is
a net (xα) ⊆ X such that
Definition 2: A Banach algebra A is called n-weakly amenable if H1(A,A(n)) = 0, where, A(n) is the n-th dual space of A.
Dales et al. (1998) investigated the relation between m-weak amenability and n-weak amenability for distinct m,n ∈ N. They obtained important results on Banach algebras and they characterized large classes of them. Ghahramani and Loy (2004) extensively studied approximate amenability of Banach algebras and they opened a new research field on amenability.
Many other researchers have followed these studies and worked on this topic. For example, Dales et al. (2006) investigated this topic on Banach sequence algebras. Lashkarizadeh and Samea (2005) studied the approximate amenability for large classes of semiegroup algebras. Ghahramani and Loy (2004) developed valuable results and gave new proofs for the characterization of amenability for Beurling algebras. Also, Choi et al. (2008) developed the recent research and as a result, they solved Johnsons (1972) problem which states that for any locally compact group G, the group algebra L1(G) is n-weakly amenable for each n ∈ N.
The contribution of this study is defining a new notion of amenability which helps to characterize Banach algebras. Such as amenability, n-weak amenability and approximate amenability, present definition determines differences between two Banach algebras.
In this study, we compose two definitions 1 and 2 together and define n-approximately weak amenability and we determine the relations between m-approximately weak amenability and n-approximately weak amenability for distinct m, n ∈ N (N denotes the set of all positive integers). Then we investigate (2n+1)-approximate weak amenability of module extension Banach algebras. Finally, we give a counter example which shows that approximately weak amenability does not imply 3-approximately weak amenability.
n-APPROXIMATE WEAK AMENABILITY
Definition 3: Let A be a Banach algebra and n ∈ N. Then A is n-approximately weakly amenable, for n ∈ N, if every continuous derivation from A into the n-th dual space A(n) is approximately inner. We say that A is approximately weakly amenable if A is 1-approximately weakly amenable.
Proposition 1: Let A be a Banach algebra and n ∈ N. Assume that A is (n+2)-approximately weakly amenable. Then A is n-approximately weakly amenable.
Proof: Let D: A→A(n) be a bounded derivation.
Then D: A→A(n+2) is a bounded derivation, so there exists
a net (Λα) ⊆ A(n+2) such that
Proposition 2: Let A be a Banach algebra and n ∈ N. Suppose that A is (2n-1)-approximately weakly amenable. Then A2 is dense in A.
Proof: By proposition 1, it is sufficient to prove the proposition
for case n = 1. Let Φ ∈ A* and Φ|A2 = 0. Then
D: A→A* with D(a) = Φ(a)Φ is a bounded derivation. Thus
there exists a net (Φα) ⊆ A* such that
Then for a ∈ A we have:
Thus Φ = 0 and so A2 is dense in A.
Let A be a non-unital Banach algebra. We denote by A# the unitization of A which is A# = A ⊕ C with the product:
It is obvious that A# is a Banach algebra as well. We denote by e* the bounded linear functional on A# which is zero on A and e*(1) = 1. By these notations we have the following identifications:
The module actions of A# on A#(2n) are same az the multiplication on A# and so A(2n) is a submodule of A#(2n). The module actions of A# on A#(2n+1) are as follows:
and so A(2n+1) is not a submodule of A#(2n+1).
Theorem 1: Let A be a non-unital Banach algebra and n ∈ N.
• | If A# is (2n)-approximately weakly amenable, then A is (2n)-approximately weakly amenable |
• | If A is (2n-1)-approximately weakly amenable, then A# is (2n-1)-approximately weakly amenable |
• | Assume that A is commutative. Then A# is n-approximately weakly amenable if and only if A is n-approximately weakly amenable |
Proof
• | Every derivation D:A→A(2n) can be extended to a derivation D0:A#→A(2n) ⊕ Ce = A#(2n) with D0(1) = 0. Thus D0 is approximately inner and so D is approximately inner |
• | Let D:A→A(2n-1) ⊕ Ce* be a bounded derivation.
Then it is easy to see that D is of the form D(a) = D0(a)+Φ(a)e*
where, D0 ∈ Z1(A,A(2n-1)) and
Φ ∈ A*. Thus there exists a net (Φα) ⊆ A(2n-1)
such that |
and so Φ|A2 = 0. By proposition 2, Φ = 0 and so D = D0. Thus, D is approximately inner | |
• | Since, A is commutative, n-approximate weak amenability and n-weak amenability are the same (note that every inner derivation is zero in commutative case). Thus this is immediate by proposition 1.4 of Dales et al. (1998) |
The following example shows that (2n+1)-weak amenability and (2n+1)-approximately weak amenability are two different notions of amenability.
Example 1: Let Mk denote the algebra of kxk matrices over C (the space of complex numberc), with norm:
It is easy to see that Mk(n) ≅Mk for each n ∈ N. Set An = (M2n)# and
Where:
By example 6.2 of Ghahramani and Loy (2004) and proposition 1.2 of Dales et al. (1998), A is (2n+1)-approximately weakly amenable (approximately amenable) but it is not (2n+1)-weakly amenable.
We denote by ο and ◊ the first and second Arens product on A**, respectively. The Banach algebra A is called Arens regular if ο = ◊ (Palmer, 1994).
Proposition 3: Let A be an Arens regular Banach algebra and let every D ∈ Z1(A**, A**) be approximately inner. Then A is 2-approximately weakly amenable.
Proof: Let D ∈ Z1(A, A**), then by theorem 1.9 of Dales
et al. (1998) there exists
So that, for a ∈ A, we have
Theorem 2: Let A be a Banach algebra such that A(2n–2) is Arens regular for each n ∈ N, (A(0) = A). If every D ∈ Z1(A(2n), A(2n)) is approximately inner, then A is 2n-approximately weakly amenable for each n ∈ N.
Proof: The case n = 1 is proposition 3. Assume that every Banach algebra
with the stated properties is 2k-approximately weakly amenable. We show that
A is (2k+2)-approximately weakly amenable. Let D ∈ Z1(A, A(2k+2))
and let P:A(2k+4)→A(2k+2) be the canonical projection.
By propositions 1.7 and 1.8 of Dales et al. (1998),
we have D** ∈ Z1(A**, A(2k+4)) (D** is the second
dual of D) and P is an A**-bimodule homomorphism. Let
Therefore,
It has been discussed by Dales et al. (1998) that any commutative, weakly amenable Banach algebra (or equivalently commutative, approximately weakly amenable Banach algebra) is n-weak amenable for each n ∈ N.
The next theorem is the partial result for general case of the earlier fact in special case. First, we note that (A(2n+2), ο) is the second dual of (A(2n), ο) for n ∈ N. Also, (A(2k), ο) is a subalgebra of (A(2n), ο) for k, n ∈ N and k≤n.
Theorem 3: Let A be an approximately weakly amenable Banach algebra such that A is an ideal in (A**, ο). Then A is (2n+1)-approximately weakly amenable for each n ∈ N.
Proof: Let An be the linear span of
Since, Ψ.a and b.Ψ ∈ A. Thus D(ab) = 0 and so D|A2n = 0. Since, A2n is dense in A, by proposition 2, we have D = 0. Therefore, A is (2n+1)-approximately weakly amenable.
Corollary 1: Let A be a Banach algebra such that A is an ideal in (A**, ο). Then the following are equivalent:
• | A is approximately weakly amenable |
• | A is (2n+1)-approximately weakly amenable for some n ∈ N |
• | A is (2n+1)-approximately weakly amenable for each n ∈ N |
Dales et al. (1998) proved that every C*-algebra is n-weakly amenable for each n ∈ N, so obviously every C*-algebra is n-approximately weakly amenable for each n ∈ N.
(2n+1)-APPROXIMATELY WEAK AMENABILITY OF A ⊕ X
Let A be a Banach algebra and X be a Banach A-bimodule. Then the module extension Banach algebra corresponding to A and X is A ⊕ X, the l1-direct sum of A and X with the algebra product defined as follows:
We investigate (2n+1)-approximately weak amenability of module extension Banach algebra A ⊕ X. As it has been discussed by Zhang (2002), according to Dales et al. (1998), X** is a Banach A**-bimodule, where, A** is equipped with the first Arens product. The module actions are defined as follows:
For x ∈ X, f ∈ X*, Φ ∈ X** and u ∈ A**, define Φf, fx ∈ A* and uf ∈ X** by
Then, for Φ ∈ X** and u ∈ A**, define uΦ, Φu ∈ X** by
These give the left and right A**-module actions on X**. Also, the definition for uf with u ∈ A** and f ∈ X* gives a left Banach A**-module action on X*. When u = a ∈ A, all the above A**-module actions agree with the A-module actions on the corresponding dual modules X* and X**.
Viewing A(2n) as a new A and X(2n) as a new X, the preceding procedure will successively define X(2n+2) as a Banach A(2n+2)-bimodule. The first Arens product is consistently assumed on each A(2n). Now suppose that the bimodule action of A(2n) on X(2n) has been defined, where, n≥1. Then in a natural way, X(2n+k), k≥1, is a Banach A(2n)-bimodule with the module multiplications uΛ and Λu ∈ X(2n+k), for Λ ∈ X(2n+k) and u ∈ A(2n), defined by:
If u = a ∈ A, these module actions coincide with A-module actions on X(2n+k). Then, for F ∈ X(2n+1) and Φ ∈ X(2n+2), define FΦ, ΦF ∈ A(2n+1) by:
For a Banach space Y and an element y ∈ Y denote by
By using the canonical image of F or Φ in the appropriate 2l-th dual space of the space that it belongs to, we can then signify a meaning for FΦ and ΦF for every F ∈ X(2n+1) and Φ ∈ X(2m); they are elements of A(2k+1), where, k = max {m-1, n}. Now for μ ∈ A(2n+2) and F ∈ X(2n+1), we define μF ∈ X(2n+1) by:
This actually defines a left Banach A(2n+2)-module action on X(2n+1).
Finally, for μ ∈ A(2n+2) and Φ ∈ X(2n+2), define μΦ, Φμ ∈ X(2n+2) by:
These actually define the A(2n+2)-module actions on X(2n+2).
It is easy to see that (A ⊕ X)(2n+1) can be identified with A(2n+1) ⊕ 4 X(2n+1), the l∞-direct sum of A(2n+1) and X(2n+1). Also, the (A ⊕ X)-bimodule actions on A(2n+1) ⊕ 4X(2n+1) are as follows:
where, a ∈ A, x ∈ X, F ∈ A(2n+1) and G ∈ X(2n+1).
It is assume that A is a Banach algebra, X is a Banach A-bimodule and A ⊕ X is their corresponding module extension Banach algebra.
Lemma 1: Let T:X→A(2n+1) be a continuous A-bimodule
homomorphism. Then
for all a ∈ A and x ∈ X.
Proof: It is routinely checked that
Thus
Therefore,
Lemma 2: Let D: A→X(2n+1) be a bounded derivation
and D(2n+1) be the (2n+1)-th dual operator of it. Then
Proof: By lemma 3.4 of Zhang (2002),
that:
Thus,
Therefore,
and let
Then for a ∈ A and x ∈ X we have
Thus
Lemma 3: Let D: A→A(2n+1), n≥0, be a bounded
derivation. Then
Proof: It is routine to check that
and so
Since,
and so
Lemma 4: Let T: X→X(2n+1), n≥0, be a continuous
A-bimodule homomorphism, satisfying xT(y)+T(x)y = 0 for all x,y ∈ X.
Then
Proof: It is routine to check that
Since,
Therefore, T = 0.
Theorem 4: For n≥0, the module extension Banach algebra A ⊕ X is (2n+1)-approximately weakly amenable if and only if the following conditions hold:
(i) | A is (2n+1)-approximately weakly amenable; |
(ii) | Every derivation from A into X(2n+1) is approximately inner; |
(iii) | For each continuous A-bimodule homomorphism T: X→A(2n+1),
n≥0, there is a net (Fα) ⊆ X(2n+1)
such that |
(iv) | The only A-bimodule homomorphism T: X→X(2n+1), n≥0 for which xT(y)+T(x)y = 0, x,y ∈ X in A(2n+1) is T = 0. |
Proof: Let A ⊕ X be (2n+1)-approximately weakly amenable. Then by lemmas 2 and 3, A is (2n+1)-approximately weakly amenable and every derivation from A into X(2n+1) is approximately inner. Furthermore, lemma 1 gives condition (iii) and lemma 4 gives condition (iv). For the converse, let:
be the canonical projections and let
is a bounded derivation and there is a net of bounded derivations
is inner. We have
Let:
Then, for all α,
and
Clearly
for a ∈ A and
Thus
Therefore,
So,
is approximately inner.
A COUNTER EXAMPLE
Here, motivated by Zhang (2002), we consider the case that the module action on one side of X is trivial. Then we give a counter example which shows the converse of proposition 1 is not correct.
We denote by X0 and Y0 the A-bimodules X with trivial right module action and Y with trivial left module action, respectively. By proposition 2, in case X = X0 it is easy to see that conditions (iii) and (iv) of Theorem 4 are reduced as follow:
(iii)0 for each continuous A-bimodule homomorphism T: X0→A(2n+1),
there is a net (Fα) ⊆ X0(2n+1)
such that
(iv)0 AX0 is dense in X0.
Proposition 4: Let A be a (2n+1)-approximately weakly amenable Banach algebra with bounded approximate identity and let AA(2n) = A(2n). Then A ⊕ X0 is (2n+1)-approximately weakly amenable if and only if AX0 is dense in X0.
Proof: By proposition 1.5 of Johnson (1972), condition
(ii) in theorem 4 always hold. Let T: X0→A(2n+1)
be a continuous A-bimodule homomorphism. Then
For n = 0, an easy application of Cohen Factorization Theorem (Bonsall and Duncan, 1973) implies that:
Corollary 2: Let A be an approximately weakly amenable Banach algebra with bounded approximate identity. Then A ⊕ X0 is approximately weakly amenable if and only if AX0 is dense in X0.
A dual result to corollary 2 is as follows:
Corollary 3: Let A be an approximately weakly amenable Banach algebra with bounded approximate identity. Then A ⊕ 0Y is approximately weakly amenable if and only if A0Y is dense in 0Y.
If X and Y are Banach A-bimodules, we denote by X ⊕ 1Y the l1-direct sum of X and Y.
Proposition 5: Suppose that A ⊕ Y and A ⊕ Y are approximately weakly amenable. Then the following are equivalent:
(i) | A ⊕ (X ⊕ 1Y) is approximately weakly amenable; |
(ii) | There is no nonzero, continuous A-bimodule homomorphism T: X→Y*; |
(iii) | There is no nonzero, continuous A-bimodule homomorphism S: Y→X*; |
Proof: The proof is quite similar to the proof of lemma 7.1 of Zhang (2002) and so we omit its proof.
Corollary 4: The algebra A ⊕ (X ⊕ 1Y) is approximately weakly amenable if and only if both A ⊕ X and A ⊕ Y are approximately weakly amenable and condition (ii) or (iii) in proposition 5 holds.
Proof: This is an immediate consequence of theorem 4 and proposition 5.
Let H be an infinite dimensional Hilbert space. We denote by B(H) and K(H) the space of all bounded linear operators and compact operators on H, respectively. By lemma 7.4 of Zhang (2002), there is an element a0 ∈ B(H)\K(H) such that a0 is not right invertible in B(H) and a0K(H) is dense in K(H). For such element a0 ∈ B(H)\K(H), a0B(H) is a proper right ideal of B(H). Thus cl(a0B(H)), the closure of a0B(H), is also a proper right ideal of B(H). So there is 0≠Λ ∈ B(H)* such that Λ.a0 = 0. Then ΛB(H)≠{0} is a right B(H)-submodule of B(H)*. Set
Example 2: By above notations, B(H) ⊕ (X0 ⊕ 1 0Y) is (approximately) weakly amenable but not 3-approximately weakly amenable.
Proof: By example 7.5 (Zhang, 2002), B(H) ⊕ (X0 ⊕ 1 0Y) is weakly amenable. Also, it is shown that B(H) ⊕ (X0 ⊕ 1 0Y) fails condition (iv) theorem 2.1 (Zhang, 2002) for m = 1, which is condition (iv) of present theorem 4 for n = 1. Thus B(H) ⊕ (X0 ⊕ 1 0Y) is not 3-approximately weakly amenable.
ACKNOWLEDGMENT
The authors would like to thank the Persian Gulf University Research Council for their financial support.