• 検索結果がありません。

The Lipschitz operators algebras Lα(K, A) and lα(K, A) are developed here further, and we study their amenability and weak amenability of these algebras

N/A
N/A
Protected

Academic year: 2022

シェア "The Lipschitz operators algebras Lα(K, A) and lα(K, A) are developed here further, and we study their amenability and weak amenability of these algebras"

Copied!
10
0
0

読み込み中.... (全文を見る)

全文

(1)

Tomus 45 (2009), 213–222

ON THE LIPSCHITZ OPERATOR ALGEBRAS

A. Ebadian and A. A. Shokri

Abstract. In a recent paper by H. X. Cao, J. H. Zhang and Z. B. Xu an α-Lipschitz operator from a compact metric space into a Banach spaceA is defined and characterized in a natural way in the sence thatF :KA is aα-Lipschitz operator if and only if for eachσXthe mappingσF is aα-Lipschitz function. The Lipschitz operators algebras Lα(K, A) and lα(K, A) are developed here further, and we study their amenability and weak amenability of these algebras. Moreover, we prove an interesting result thatLα(K, A) andlα(K, A) are isometrically isomorphic toLα(K) ˇ⊗Aand lα(K) ˇ⊗Arespectively. Also we study homomorphisms on theLαA(X, B).

1. Introduction

Let (K, d) be compact metric space with at least two elements and (X,k · k) be a Banach space over the scalar fieldF(= R or C). For a constantα >0 and an operator T:KX, set

(1) Lα(T) := sup

s6=t

kT(t)−T(s)k d(s, t)α , which is called the Lipschitz constant ofT. Define

Tα(x, y) =T(x)−T(y)

d(x, y)α , x6=y Lα(K, X) ={T:KX:Lα(T)<∞}

and

lα(K, X) ={T:KX:kTα(x, y)k →0 as d(x, y)→0}.

The elements ofLα(K, X) andlα(K, X) are called big and little Lipschitz operators, respectively [1].

Let C(K, X) be the set of all continuous operators fromK intoX and for each TC(K, X), define

kTk= sup

x∈K

kT(x)k. ForS,T inC(K, X) andλin F, define

(S+T)(x) =S(x) +T(x), (λT)(x) =λT(x), (x∈X).

2000Mathematics Subject Classification: primary 47B48; secondary 46J10.

Key words and phrases: Lipschitz algebras, amenability, homomorphism.

Received October 7, 2008. Editor A. Pultr.

(2)

It is easy to see that (C(K, X), k · k) becomes a Banach space over F and Lα(K, X) is a linear subspace ofC(K, X). For each elementT ofLα(K, X), define kTkα=Lα(T) +kTk.

In their papers [3, 4], Cao, Zhang and Xu proved that (Lα(K, X),k · kα) is a Banach space over F andlα(K, X) is a closed linear subspace of (Lα(K, X),k · kα).

Now, let (A,k · k) be a unital Banach algebra with unite. In this paper, we show that (Lα(K, A),k·kα) is a Banach algebra under pointwise and scalar multiplication andlα(K, A) is a closed linear subalgebra of (Lα(K, A),k · kα) and study many aspects of these algebras. The spaces Lα(K, A) and lα(K, A) are called big and little Lipschitz operators algebras. Note that Lipschitz operators algebras are, in fact, extensions of Lipschitz algebras. Sherbert [11, 12], Weaver [13, 14], Honary and Mahyar [7], Johnson [8, 9], Alimohammadi and Ebadian [1], Ebadian [6], Bade, Curtis and Dales [2], studied some properties of Lipschitz algebras. We will study (weak) amenability of Lipschitz operators algebras. Also we study homomorphisms on theLαA(X, B).

2. Characterizations of Lipschitz operators algebras

In this section, let (K, d) be a compact metric space which has at least two elements and (A,k · k) to denote a unital Banach algebra over the scalar field F (= R or C).

Theorem 2.1. (Lα(K, A),k · kα) is a Banach algebra overFand lα(K, A) is a closed linear subspace of (Lα(K, A),k · kα).

Proof. As we have already Lα(K, A) is a Banach space andlα(K, A) is a closed linear subspace if it. Now let T,SLα(K, A), and define

(T S)(t) =T(t)S(t) (t∈K). Then

kT Skα=kT Sk+Lα(T S)

≤ kTkkSk+ sup

t6=s

kT(t)S(t)−T(s)S(s)k d(t, s)α

≤ kTkkSk+kTkLα(S) +kSkLα(T)

≤ kTk+Lα(T)

kSk+Lα(S)

=kTkαkSkα.

So that we see that (Lα(K, A),k · kα) is a Banach algebra andlα(K, A) is a closed

linear subspace of (Lα(K, A),k · kα).

Theorem 2.2. Let (K, d)be a compact metric space. Then Lα(K, A)is uniformly dense in C(K, A).

Proof. Let fC(K, A). Then for every σA we haveσfC(K), so that there isgLα(K) such thatkg−σfk< ε. We define, the mapη: C→A by

(3)

η(λ) =λ·e. It is easy to see that ηgLα(K, A), and for everyσA, we have

|σ g(x)·ef(x)

|=|g(x)−(σ◦f)(x)|< ε , (x∈K).

Therefore |σ(η◦gf)(x)| < ε for every σA andxK. This implies that k(η◦gf)(x)k< εfor everyxK. Therefore,kη◦gfk< εand the proof is

complete.

Remark 2.3. LetA, Bbe unital Banach algebras over F. Then the injective tensor A⊗Bˇ is a unital Banach algebra under normk · k, [10].

Theorem 2.4. Lα(K, A) ={F:KA|σFLα(K,C),(∀σ∈A)}

Proof. Use the principle of Uniform Boundedness.

Lemma 2.5. Let (E1,k · k1),(E2,k · k2)be Banach spaces. Then for GE1⊗Eˇ 2

kGkε= sup

k id⊗φ

(G)k1 : φE2, kφk ≤1 .

Proof. See [10].

Theorem 2.6. Let(K, d)be a compact metric space andAbe a unital commutative Banach algebra. ThenLα(K, A)is isometrically isomorphic toLα(K) ˇ⊗A.

Proof. It is straightforward to prove that the mappingV:Lα(K)×A→Lα(K, A) defined by

V(f, a) =f a (f ∈Lα(K), a∈A), (f a)(x) :=f(x)a (x∈K),

is bilinear. Therefore there exists a unique linear mapT:Lα(K) ˇ⊗A→Lα(K, A) such thatT(f⊗a) =V(f, a) =f a, [10]. For everyGLα(K) ˇ⊗A, there ism∈N, fjLα(K) andajA(1≤jm) such thatG=Pm

j=1fjaj, so we have kGkε= sup

φ∈A,kφk≤1

k(id⊗φ)(G)kα= sup

φ∈A,kφk≤1

(id⊗φ)Xm

j=1

fjaj

= sup

φ∈A,kφk≤1

m

X

j=1

fjφ(aj)

α= sup

φ∈A,kφk≤1

h sup

x∈K

m

X

j=1

fj(x)φ(aj)

+ sup

x6=y

|Pm

j=1fj(x)φ(aj)−Pm

j=1fj(y)φ(aj)|

dα(x, y)

i

= sup

φ∈A,kφk≤1

h sup

x∈K

φXm

j=1

fj(x)aj

+ sup

x6=y

φ Pm

j=1(fj(x)ajfj(y)aj) dα(x, y)

i

(4)

≤ sup

φ∈A,kφk≤1

h sup

x∈K

kφk

m

X

j=1

fj(x)aj

+ sup

x6=y

kφkkPm

j=1fj(x)aj−Pm

j=1fj(y)ajk dα(x, y)

i

≤sup

x∈K

m

X

j=1

fj(x)aj

+ sup

x6=y

kPm

j=1fj(x)aj−Pm

j=1fj(y)ajk dα(x, y)

=

m

X

j=1

fjaj

+pα

Xm

j=1

fjaj

=

m

X

j=1

fjaj

α=

TXm

j=1

fjaj

α

=kT Gkα =⇒ kGkε≤ kT Gkα.

Now let γ >0 be arbitrary, such thatkT Gkα> γ. ThenkPm

j=1fjajkα> γ, and so we have

m

X

j=1

fjaj

+pα

Xm

j=1

fjaj

> γ

⇒sup

x∈K

m

X

j=1

fj(x)aj + sup

x6=y

kPm

j=1fj(x)aj−Pm

j=1fj(y)ajk dα(x, y) > γ

⇒ sup

φ∈A,kφk≤1

h sup

x∈K

m

X

j=1

fj(x)φ(aj)

+ sup

x6=y

|Pm

j=1fj(x)φ(aj)−Pm

j=1fj(y)φ(aj)|

dα(x, y)

i

> γ

⇒ sup

φ∈A,kφk≤1

h

m

X

j=1

fjφ(aj) +pα

Xm

j=1

fjφ(aj)i

> γ

⇒ sup

φ∈A,kφk≤1

hk(id⊗φ)Xm

j=1

fjaj

+pα

(id⊗φ)Xm

j=1

fjaji

> γ

⇒ sup

φ∈A,kφk≤1

(id⊗φ)Xm

j=1

fjaj α> γ

m

X

j=1

fjaj

ε> γ ⇒ kGkε> γ .

Sinceγ >0 is arbitrary, then we havekT Gkα≤ kGkε. ThereforekT Gkα=kGkε, and this implies thatT is a linear isometry map. So T one-one and continuous map. Now, we show that T is a onto map. For this, we show that the range of T, RT is a closed and dense subset of Lα(K, A). It is easy to see that RT

is closed. Let fLα(K, A) and γ > 0. There exist a1, . . . , anA such that X :=f(K)⊂Sn

i=1B(ai, γ). SetUj=f−1(B(aj, γ)) wherej= 1, . . . , n. Then there

(5)

exist f1, . . . , fnLα(K, A) and σAsuch that supp(fj)⊂Uj forj = 1, . . . , n andσ◦(f1+. . .+fn) = 1. For everyxK we have,

f(x)− (σ◦f1)a1+· · ·+ (σ◦fn)an

(x)

=

f(x) (σ◦f1)(x) +· · ·+ (σ◦fn)(x)

− (σ◦f1)(x)a1+· · ·+ (σ◦fn)(x)an

=

(σ◦f1)(x) f(x)−a1

+· · ·+ (σ◦fn)(x) f(x)−an

n

X

i=1

(σ◦fi)(x)

f(x)−ai

< γ , since suppfjUj. Therefore,

f− (σ◦f1)a1+· · ·+ (σ◦fn)an α< γ . This implies that

f

n

X

i=1

T(σ◦fiai) α

< γ .

We conclude that ¯RT =Lα(K, A). SoRT =Lα(K, A), sinceRT is closed. Hence T is a onto map. Also by product• onLα(K) ˇ⊗A

(f⊗a)•(g⊗b) =f gab (f, g∈Lα(K), a, b∈A),

clearlyT is homomorphism.

Furthermore T is open map, for this purpose, let τ and τ0 be topologies on Lα(K) ˇ⊗AandLα(K, A) respectively. LetUτ, we show thatT(U)∈τ0. Letpbe a limit point inLα(K, A)\T(U). Then there exists a sequence{pn}inLα(K, A)\T(U) converges to p. SinceT is onto, there is a sequence{qn}and qinLα(K) ˇ⊗Asuch that T(qn) = pn and T q = p. Therefore T(qn) converges to p in Lα(K). Since qnLα(K) ˇ⊗A, we can findm∈N,fj(n)Lα(K) anda(n)jAsuch that whenever 1≤jmwe have

(1) T(qn) =

m

X

j=1

fj(n)a(n)j .

Also, sinceqLα(K) ˇ⊗A there existr∈N,giLα(K) andbiAsuch that

(2) p=T(q) =

r

X

i=1

gibi.

Since kT(qn)−pkα → 0 as n → ∞, for every positive number γ there exists a positive integerN such that

(3)

m

X

j=1

fj(n)a(n)j

r

X

i=1

gibi α< γ ,

(6)

whennN. By applying (3), we have sup

(x∈K)

m

X

j=1

fj(n)(x)a(n)j

r

X

i=1

gi(x)bi + sup

(x6=y)

1 d(x, y)α

×

m

X

j=1

fj(n)(x)a(n)j

r

X

i=1

gi(x)bi

m

X

j=1

fj(n)(y)a(n)j +

r

X

i=1

gi(y)bi < γ . Therefore ifσA withkσk ≤1 then

sup

(x∈K)

m

X

j=1

fj(n)(x)σ(a(n)j )−

r

X

i=1

gi(x)σ(bi) + sup

(x6=y)

1 d(x, y)α

×

m

X

j=1

fj(n)(x)σ(a(n)j )−

r

X

i=1

gi(x)σ(bi)

m

X

j=1

fj(n)(y)σ(a(n)j ) +

r

X

i=1

gi(y)σ(bi) < γ . This implies that

(4)

m

X

j=1

fj(n)σ(a(n)j )−

r

X

i=1

giσ(bi) α< γ Now by using (4), for everyφLα(K) withkφkα≤1 we have,

φXm

j=1

fj(n)σ(a(n)j )−

r

X

i=1

giσ(bi) < γ , hence

(5)

m

X

j=1

φ(fj(n))σ(a(n)j )−

r

X

i=1

φ(gi)σ(bi) < γ , By (5), we conclude

(6) sup

m

X

j=1

φ(fj(n))σ(a(n)j )−

r

X

i=1

φ(gi)σ(bi)

< γ , kσk ≤1, kφkα≤1. Thereforekqnqkγand henceqnqorqnT−1pinLα(K) ˇ⊗A. This show that pT(U)c.

Remark 2.7. By using the above theorem we can prove thatlα(K, A)∼=lα(K) ˇ⊗A.

3. (Weak) Amenability ofLα(K, A)

Let Abe a Banach algebra and X be a BanachA-module over F. The linear map D: AX is called an X-derivation onA, ifD(ab) = D(a)·b+a·D(b), for everya, bA. The set of all continuesX-derivations onA is a vector space over F which is denoted by Z1(A, X). For each xX the map δx: AX, defined byδx(a) =a·xx·a, is a continuesX-derivation onA. TheX-derivation D:AX is called an inner derivation onA if there exists anxX such that D=δx. The set of all innerX-derivations onA is a linear subspace ofZ1(A, X)

(7)

which is denoted by B1(A, X). The quotient space Z1(A, X)/B1(A, X) is deno- ted byH1(A, X) and is called the first cohomology group ofAwith coefficients inX. Definition 3.1. The Banach algebra A over F is called amenable if for every Banach A-moduleX over F,H1(A, X) ={0}. The Banach algebra Aover F is called weakly amenable ifH1(A, A) ={0}.

The notion of amenability of Banach algebras were first introduced by B. E.

Johnson in 1972 [8]. Bade, Curtis and Dales [2], studied the (weak) amenability of Lipschitz algebras in 1987 [2]. In this section, we study the (weak) amenability of Lα(K, A).

For every Banach algebraB, let ΦB be the space of maximal ideal ofB.

Definition 3.2. LetA be a commutative Banach algebra and letφ∈ΦA∪ {0}.

The non-zero linear functionalD onA is called point derivation atφif D(ab) =φ(a)D(b) +φ(b)D(a), (a, b∈A).

Lemma 3.3. For each non-isolated point xK and σA, if the map φ:Lα(K, A)→Cis given by

φ(f) = (σ◦f)(x), fLα(K, A) then φ∈ΦLα(K,A).

Proof. Obvious.

Let (K, d) be a fixed non-empty compact metric space, set

∆ =

(x, y)∈K×K:x=y , W =K×K−∆.

We now examine the amenability and weak amenability of Lipschitz operators algebras Lα(K, A) andlα(K, A).

Theorem 3.4. Let (K, d)be an infinite compact metric space and takeα∈(0,1].

ThenLα(K, A) is not weakly amenable.

Proof. Letxbe a non-isolated point inK. We define Wx:=

{(xn, yn)}n=1: (xn, yn)∈W, (xn, yn)→(x, x) asn→ ∞ . For the net w={(xn, yn)}n=1 in WxandσA, we put

w(f) =(σ◦f)(xn)−(σ◦f)(yn)

d(xn, yn)α , fLα(K, A) thenkw(f)k≤ kσk kfkα. Hence,wis continues. Now set

Dw(f) = LIM w(f)

, fLα(K, A) ,

(8)

where LIM(·) is Banach limit [12]. We show that the linear mapDw is a non-zero point derivation atφ, whichφis given by Lemma 6. We have,

Dw(f g) = LIM(w(f g))

= LIM(σ◦f g)(xn)−(σ◦f g)(yn) d(xn, yn)α

= LIM 1

d(xn, yn)α

σf(xn)g(xn)−f(xn)g(yn)

= LIM 1

d(xn, yn)α

σf(xn) g(xn)−g(yn)

+g(yn) f(xn)−f(yn)

= (σ◦f)(x) LIM w(g)

+ (σ◦g)(x) LIM w(g)

=φ(f)Dw(g) +φ(g)Dw(f)

Therefore, by the continuityf,g and properties of Banach limit we concludeDw is a non-zero, continues point derivation at φonLα(K, A), an so by [5], Lα(K, A)

is not weakly amenable.

Corollary 3.5. Lα(K, A)is not amenable.

Theorem 3.6. LetK⊆C be an infinite compact set, and takeα∈(0,1). Then lα(K, A)is not amenable.

Proof. Letx0K. We define

Mx0 :={f ∈lα(K, A) : (σ◦f)(x0) = 0 ∀σ∈A}. IfσA, then for eachfMx2

0 we have (σ◦f)(x)

d(x, x0) −→0 as d(x, x0)−→0. For β ∈ (α,2α), set fβ(x) := η d(x, x0)β

, xK where, the map η: C → A defined byη(λ) =λ·e. ThenfβMx0 and{fβ+Mx20 : β ∈(α,2α)} is a linearly independent set in MMx20

x0

becausex0 is non-isolated inK. ThereforeMx20 has infinite codimension in Mx0, and so Mx0 6= Mx20 then by [5] Mx0 has not a bounded approximate identity, and sinceMx0 is closed ideal inlα(K, A), thenlα(K, A) is

not amenable.

Theorem 3.7. Let(K, d)be a compact metric space andAbe a unital commutative Banach algebra. If 12 < α <1, thenlα(T, A)is not weakly amenable, whereT is unit circle in complex plane.

Proof. By Remark 2.7, we have lα(T, A)∼=lα(T) ˇ⊗A. Since by [5],lα(T) is not weakly amenable, hencelα(T, A) is not weakly amenable.

Corollary 3.8. LetA be a finite-dimensional weakly amenable Banach algebra. If 0< α < 12, thenlα(K, A)is weakly amenable.

Proof. By [10],lα(K) ˆ⊗Ais weakly amenable. Now by [10], we have lα(K) ˆ⊗A∼= lα(K) ˇ⊗Aand this implies that lα(K) ˇ⊗Ais weakly amenable and so lα(K, A) is

weakly amenable.

(9)

4. Homomorphisms on theLαA(X, B)

Definition 4.1. Let (X, d) be a compact metric space inC,α∈(0,1], (B,k · k) be a commutative Banach algebra with unite, andB be the dual space ofB, define

A(X, B) =

fC(X, B) : Λf is analytic in interior ofX,Λ∈B LαA(X, B) =

fLα(X, B) : Λ◦f is analytic in interior ofX, Λ∈B lαA(X, B) =

flα(X, B) : Λ◦f is analytic in interior of X, Λ∈B In this case, we have

LαA(X, B) =Lα(X, B)∩A(X, B) and

lαA(X, B) =lα(X, B)∩A(X, B). SoLαA(X, B)∼=LαA(X) ˇ⊗B andlAα(X, B)∼=lAα(X) ˇ⊗B.

Theorem 4.2. Every characterχonLαA(X, B)(andlAα(X, B)) is of formχ=ψ◦δz

for some character ψ onB and some zX.

Proof. SinceLαA(X, B)∼=LαA(X) ˇ⊗B, letj:Lα(X)→LαA(X, B),h7→he, be the canonical embedding. Then there iszX such thatχj is the evaluation in z, that isχj=δz where δz(ϕ) =ϕ(z). Consider the ideal

I:=

fLαA(X, B) :f(z) = 0 .

We will show thatI is contained in the kernel ofχ. GivenfIwe define, ϕ(ω) :=

(ωz if ω6=z;

0 if ω=z.

and

g(ω) :=

f(ω)

ωz if ω6=z;

f0(z) if ω=z.

Sincef has a Taylor series expansion f(ω) =

X

n=1

f(n)(z)

n! (ω−z)n

around z, it is easy to see that Λg is holomorphic (Λ∈ B), and hence gLαA(X, B). We have

χ(f) =χ j(ϕ)g

= (χ◦j)(ϕ)χ(g) =δz(ϕ)χ(g) =ϕ(z)χ(g) = 0.

The evaluation δz is an epimorphism and since kerδz=I ⊂kerχ, we obtain the desired factorization χ=ψδz for some characterψ onB.

(10)

References

[1] Alimohammadi, D., Ebadian, A.,Headberg’s theorem in real Lipschitz algebras, Indian J.

Pure Appl. Math.32(2001), 1479–1493.

[2] Bade, W. G., Curtis, P. C., Dales, H. G.,Amenability and weak amenability for Berurling and Lipschitz algebras, Proc. London Math. Soc.55(3) (1987), 359–377.

[3] Cao, H. X., Xu, Z. B.,Some properties of Lipschitz-αoperators, Acta Math. Sin. (Engl. Ser.) 45(2) (2002), 279–286.

[4] Cao, H. X., Zhang, J. H., Xu, Z. B.,Characterizations and extentions of Lipschitz-αoperators, Acta Math. Sin. (Engl. Ser.)22(3) (2006), 671–678.

[5] Dales, H. G.,Banach Algebras and Automatic Continuty, Clarendon Press, Oxford, 2000.

[6] Ebadian, A.,Prime ideals in Lipschitz algebras of finite differentable function, Honam Math.

J.22(2000), 21–30.

[7] Honary, T. G, Mahyar, H.,Approximation in Lipschitz algebras, Quaest. Math.23(2000), 13–19.

[8] Johnson, B. E.,Cohomology in Banach algebras, Mem. Amer. Math. Soc.127(1972).

[9] Johnson, B. E.,Lipschitz spaces, Pacific J. Math.51(1975), 177–186.

[10] Runde, V.,Lectures on Amenability, Springer, 2001.

[11] Sherbert, D. R.,Banach algebras of Lipschitz functions, Pacific J. Math.3(1963), 1387–1399.

[12] Sherbert, D. R.,The structure of ideals and point derivations in Banach algebras of Lipschitz functions, Trans. Amer. Math. Soc.111(1964), 240–272.

[13] Weaver, N.,Subalgebras of little Lipschitz algebras, Pacific J. Math.173(1996), 283–293.

[14] Weaver, N.,Lipschitz Algebras, World Scientific Publishing Co., Inc., River Edge, NJ, 1999.

Faculty of Basic Science, Science and Research Brunch Islamic Azad University (IAU), Tehran, Iran

E-mail:[email protected]

Department of Mathematics, Faculty of Science Urmia University, Urmia, Iran

E-mail:[email protected]

参照

関連したドキュメント