REGULARITY OF BOUNDED TRI-LINEAR MAPS AND THE FOURTH ADJOINT OF A TRI-DERIVATION
Ali Ebadian, KAZEM HAGHNEJAD AZAR, Abotaleb Sheikhali
Depatment of Mathematics, Payame Noor University (PNU), Tehran, Iran.
Department of Mathematics, University of Mohghegh Ardabili, Ardabil, Iran.
Depatment of Mathematics, Payame Noor University (PNU), Tehran, Iran.
[email protected]
[email protected]
[email protected]
Abstract.
In this Article, we give a simple criterion for the regularity of a tri-linear mapping. We provide if f:X×Y×Z⟶W is a bounded tri-linear mapping and h:W⟶S is a bounded linear mapping, then f is regular if and only if hof is regular. We also shall give some necessary and sufficient conditions such the fourth adjoint D∗∗∗ of a tri-derivation D is again tri-derivation.
Key words and phrases:
Arens regular, Banach module, Bi-derivation, Derivation, Fourth adjoint, Regular, Tri-derivation, Tri-linear.
2010 Mathematics Subject Classification:
46H25
1. Introduction and Preliminaries
Richard Arens showed in [3] that a bounded bilinear map m:X×Y⟶Z on normed spaces, has two natural different extensions m∗∗∗, mr∗∗∗r from X∗∗×Y∗∗ into Z∗∗. When these extensions are equal, m is said to be Arens regular. A Banach algebra A is said to be Arens regular, if its product π(a,b)=ab considered as a bilinear mapping π:A×A⟶A isَ Arens regular. The first and second Arens products of A∗∗ by symbols □ and ◊ respectively defined by
[TABLE]
Arens regularity of π is equivalent to the following
[TABLE]
given by {aα} and {bβ} are bounded nets in A and a∗∗=w∗−αlimaα and b∗∗=w∗−βlimbβ are elements of A∗∗. Some characterizations for the Arens regularity of bounded bilinear map m and Banach algebra A are proved in [1, 2, 3, 4, 5, 11, 14, 16] and [17].
Suppose X,Y,Z,W and S be normed spaces and f:X×Y×Z⟶W be a bounded tri-linear mapping. In this paper we first define regularity of f map and showing that f is regular if and only if f∗∗∗r∗(X∗∗,W∗,Z)⊆Y∗ and f∗∗∗∗∗(W∗,X∗∗,Y∗∗)⊆Z∗. Also we show that be a bounded tri-linear maps f:X×Y×Z⟶W and h:W⟶S, f is regular if and only if hof is regular.
The natural extensions of f are as follows:
- (1)
f∗:W∗×X×Y⟶Z∗, given by ⟨f∗(w∗,x,y),z⟩=⟨w∗,f(x,y,z)⟩ where x∈X,y∈Y,z∈Z,w∗∈W∗ (f∗ is said the adjoint of f and is a bounded tri-linear map).
2. (2)
f∗∗=(f∗)∗:Z∗∗×W∗×X⟶Y∗, given by ⟨f∗∗(z∗∗,w∗,x),y⟩=⟨z∗∗,f∗(w∗,x,y) where x∈X,y∈Y,z∗∗∈Z∗∗,w∗∈W∗.
3. (3)
f∗∗∗=(f∗∗)∗:Y∗∗×Z∗∗×W∗⟶X∗, given by ⟨f∗∗∗(y∗∗,z∗∗,w∗),x⟩=⟨y∗∗,f∗∗(z∗∗,w∗,x)⟩ where x∈X,y∗∗∈Y∗∗,z∗∗∈Z∗∗,w∗∈W∗.
4. (4)
f∗∗∗∗=(f∗∗∗)∗:X∗∗×Y∗∗×Z∗∗⟶W∗∗, given by ⟨f∗∗∗∗(x∗∗,y∗∗,z∗∗),w∗⟩=⟨x∗∗,f∗∗∗(y∗∗,z∗∗,w∗)⟩ where x∗∗∈X∗∗,y∗∗∈Y∗∗,z∗∗∈Z∗∗,w∗∈W∗.
Now let fr:Z×Y×X⟶W be the flip of f defined by fr(z,y,x)=f(x,y,z), for every x∈X,y∈Y and z∈Z. Then fr is a bounded tri-linear map and it may extends as above to fr∗∗∗∗:Z∗∗×Y∗∗×X∗∗⟶W∗∗. When f∗∗∗∗ and fr∗∗∗∗r are equal, then f is called regular.
Suppose A be a Banach algebra and π1:A×X⟶X is a bounded bilinear map. The pair (π1,X) is said to be a left Banach A−module when π1(π1(a,b),x)=π1(a,π1(b,x)), for each a,b∈A and x∈X. A right Banach A−module may is defined similarly. Let π2:X×A⟶X be a bounded bilinear map. The pair (X,π2) is said to be a right Banach A−module if π2(x,π2(a,b))=π2(π2(x,a),b). A triple (π1,X,π2) is said to be a Banach A−module if (X,π1) and (X,π2) are left and right Banach A−modules, respectively, and π1(a,π2(x,b))=π2(π1(a,x),b).
Let (π1,X,π2) be a Banach A−module. Then (π2r∗r,X∗,π1∗) is the dual Banach A−module of (π1,X,π2).
A bounded linear mapping D1:A⟶X∗ is said to be a derivation if for each a,b∈A
[TABLE]
A bounded bilinear map D2:A×A⟶X(or X∗) is called a bi-derivation, if for each a,b,c and d∈A
[TABLE]
Let D1:A⟶A∗ be a derivation. Dales, Rodriguez and Velasco, in [8] showed that D1∗∗:(A∗∗,□)⟶A∗∗∗ is a derivation if and only if πr∗∗∗∗(D1∗∗(A∗∗),A∗∗)⊆A∗.
In [15], S. Mohamadzadeh and H.R.E Vishki extends this and showed that second adjont D1∗∗:(A∗∗,□)⟶A∗∗∗ is a derivation if and only if π2∗∗∗∗(D1∗∗(A∗∗),X∗∗)⊆A∗ and which D1∗∗:(A∗∗,◊)⟶A∗∗∗ is a derivation if and only if π1r∗∗∗∗(D1∗∗(A∗∗),X∗∗)⊆A∗.
A. Erfanian Attar et al, provide condition such that the third adjoint D2∗∗∗ of a bi-derivation D2:A×A⟶X (or X∗) is again a bi-derivation, see [10]. For a Banach A−module (π1,X,π2), the fourth adjoint D∗∗∗∗ of a tri-derivation D:A×A×A⟶X∗ is trivially a tri-linear extension of D. A problem which is of interest is under what conditions we need that D∗∗∗∗ is again a tri-derivation. The section 4 we will extend above mentioned result.
A bounded trilinear mapping f:X×Y×Z⟶W is said to factor if it is surjective, that is f(X×Y×Z)=W.
Throughout the article, we usually identify a normed space with its canonical image in its second dual.
2. regularity of bounded tri-linear maps
Theorem 2.1**.**
Let f:X×Y×Z⟶W be a bounded tri-linear map. Then f is regular if and only if
[TABLE]
where {xα},{yβ} and {zγ} are nets in X,Y and Z which converge to x∗∗∈X∗∗,y∗∗∈Y∗∗ and z∗∗∈Z∗∗ in the w∗−topologies, respectively.
Proof.
For every w∗∈W∗ we have
⟨f∗∗∗∗(x∗∗,y∗∗,z∗∗),w∗⟩=⟨x∗∗,f∗∗∗(y∗∗,z∗∗,w∗)⟩=αlim⟨f∗∗∗(y∗∗,z∗∗,w∗),xα⟩
=αlim⟨y∗∗,f∗∗(z∗∗,w∗,xα)⟩=αlimβlim⟨f∗∗(z∗∗,w∗,xα),yβ⟩
=αlimβlim⟨z∗∗,f∗(w∗,xα,yβ)⟩=αlimβlimγlim⟨f∗(w∗,xα,yβ),zγ⟩
=αlimβlimγlim⟨f(xα,yβ,zγ),w∗⟩.
Therefore f∗∗∗∗(x∗∗,y∗∗,z∗∗)=w∗−αlimw∗−βlimw∗−γlimf(xα,yβ,zγ). In the other hands
⟨fr∗∗∗∗r(x∗∗,y∗∗,z∗∗),w∗⟩=⟨fr∗∗∗∗(z∗∗,y∗∗,x∗∗),w∗⟩=⟨z∗∗,fr∗∗∗(y∗∗,x∗∗,w∗)⟩
=γlim⟨fr∗∗∗(y∗∗,x∗∗,w∗),zγ⟩=γlim⟨y∗∗,fr∗∗(x∗∗,w∗,zγ)
=γlimβlim⟨fr∗∗(x∗∗,w∗,zγ),yβ⟩=γlimβlim⟨x∗∗,fr∗(w∗,zγ,yβ)⟩
=γlimβlimαlim⟨fr∗(w∗,zγ,yβ),xα⟩=γlimβlimαlim⟨w∗,fr(zγ,yβ,xα)⟩
=γlimβlimαlim⟨f(xα,yβ,zγ),w∗⟩.
Therefore fr∗∗∗∗r(x∗∗,y∗∗,z∗∗)=w∗−γlimw∗−βlimw∗−αlimf(xα,yβ,zγ), and proof follows.
∎
In the following theorem, we provide a criterion concerning to the regularity of a bounded tri-linear map.
Theorem 2.2**.**
For a bounded tri-linear map f:X×Y×Z⟶W the following statements are equivalent:
- (1)
f* is regular.*
2. (2)
f∗∗∗∗∗=fr∗∗∗∗∗∗∗r.**
3. (3)
f∗∗∗r∗(X∗∗,W∗,Z)⊆Y∗* and f∗∗∗∗∗(W∗,X∗∗,Y∗∗)⊆Z∗.*
Proof.
(1) ⇒ (2), if f is regular, then f∗∗∗∗=fr∗∗∗∗r. For every x∗∗∈X∗∗,y∗∗∈Y∗∗,z∗∗∈Z∗∗ and w∗∗∗∈W∗∗∗ we have
⟨f∗∗∗∗∗(w∗∗∗,x∗∗,y∗∗),z∗∗⟩=⟨w∗∗∗,f∗∗∗∗(x∗∗,y∗∗,z∗∗)⟩=⟨w∗∗∗,fr∗∗∗∗r(x∗∗,y∗∗,z∗∗)⟩
=⟨w∗∗∗,fr∗∗∗∗(z∗∗,y∗∗,x∗∗)⟩=⟨fr∗∗∗∗∗(w∗∗∗,z∗∗,y∗∗),x∗∗⟩
=⟨fr∗∗∗∗∗∗(x∗∗,w∗∗∗,z∗∗),y∗∗⟩=⟨fr∗∗∗∗∗∗∗(y∗∗,x∗∗,w∗∗∗),z∗∗⟩=⟨fr∗∗∗∗∗∗∗r(w∗∗∗,x∗∗,y∗∗),z∗∗⟩
as claimed.
(2) ⇒ (1), let f∗∗∗∗∗=fr∗∗∗∗∗∗∗r, then for every w∗∈W∗,
⟨fr∗∗∗∗r(x∗∗,y∗∗,z∗∗),w∗⟩=⟨fr∗∗∗∗(z∗∗,y∗∗,x∗∗),w∗⟩=⟨fr∗∗∗∗∗(w∗,z∗∗,y∗∗),x∗∗⟩
=⟨fr∗∗∗∗∗∗(x∗∗,w∗,z∗∗),y∗∗⟩=⟨fr∗∗∗∗∗∗∗(y∗∗,x∗∗,w∗),z∗∗⟩
=⟨fr∗∗∗∗∗∗∗r(w∗,x∗∗,y∗∗),z∗∗⟩=⟨f∗∗∗∗∗(w∗,x∗∗,y∗∗),z∗∗⟩=⟨f∗∗∗∗(x∗∗,y∗∗,z∗∗),w∗⟩.
It follows that f is regular.
(1) ⇒ (3), let f is regular and x∗∗∈X∗∗,y∗∗∈Y∗∗,z∈Z,w∗∈W∗. Then we have
⟨f∗∗∗r∗(x∗∗,w∗,z),y∗∗⟩=⟨x∗∗,f∗∗∗r(w∗,z,y∗∗)⟩=⟨x∗∗,f∗∗∗(y∗∗,z,w∗)⟩
=⟨f∗∗∗∗(x∗∗,y∗∗,z),w∗⟩=⟨fr∗∗∗∗r(x∗∗,y∗∗,z),w∗⟩
=⟨fr∗∗∗∗(z,y∗∗,x∗∗),w∗⟩=⟨fr∗∗∗(y∗∗,x∗∗,w∗),z⟩=⟨fr∗∗(x∗∗,w∗,z),y∗∗⟩.
Therefore f∗∗∗r∗(x∗∗,w∗,z)=fr∗∗(x∗∗,w∗,z)∈Y∗. So f∗∗∗r∗(X∗∗,W∗,Z)⊆Y∗. A similar argument shows that f∗∗∗∗∗(w∗,x∗∗,y∗∗)=fr∗∗∗r(w∗,x∗∗,y∗∗)∈Z∗. Thus f∗∗∗∗∗(W∗,X∗∗,Y∗∗)⊆Z∗, as claimed.
(3) ⇒ (1), let {xα},{yβ} and {zγ} are nets in X,Y and Z which converge to x∗∗,y∗∗ and z∗∗ in the w∗−topologies, respectively. For every w∗∈W∗ we have
⟨fr∗∗∗∗r(x∗∗,y∗∗,z∗∗),w∗⟩=γlimβlimαlim⟨f(xα,yβ,zγ),w∗⟩=γlimβlimαlim⟨f∗(w∗,xα,yβ),zγ⟩
=γlimβlimαlim⟨f∗∗(zγ,w∗,xα),yβ⟩=γlimβlimαlim⟨f∗∗∗(yβ,zγ,w∗),xα⟩
=γlimβlim⟨x∗∗,f∗∗∗(yβ,zγ,w∗)=γlimβlim⟨x∗∗,f∗∗∗r(w∗,zγ,yβ)⟩
=γlimβlim⟨f∗∗∗r∗(x∗∗,w∗,zγ),yβ⟩=γlim⟨f∗∗∗r∗(x∗∗,w∗,zγ),y∗∗⟩
=γlim⟨x∗∗,f∗∗∗r(w∗,zγ,y∗∗)⟩=γlim⟨x∗∗,f∗∗∗(y∗∗,zγ,w∗)⟩
=γlim⟨f∗∗∗∗(x∗∗,y∗∗,zγ),w∗⟩=γlim⟨f∗∗∗∗∗(w∗,x∗∗,y∗∗),zγ⟩
=f∗∗∗∗∗(w∗,x∗∗,y∗∗),z∗∗⟩=⟨f∗∗∗∗(x∗∗,y∗∗,z∗∗),w∗⟩.
It follows that f is regular and this completes the proof.
∎
Corollary 2.3**.**
For a bounded tri-linear map f:X×Y×Z⟶W the following statements are equivalent:
- (1)
f* is regular.*
2. (2)
fr∗∗∗∗∗r=f∗∗∗∗∗∗∗.
3. (3)
fr∗∗∗r∗(Z∗∗,W∗,X)⊆Y∗* and f∗∗∗∗∗(W∗,Z∗∗,Y∗∗)⊆X∗.*
Proof.
The mapping f is regular if and only if fr is regular. Therefore by Theorem 2.2, the desired result is obtained.
∎
Corollary 2.4**.**
For a bounded tri-linear map f:X×Y×Z⟶W, if from X,Y or Z at least two reflexive then f is regular.
Proof.
Without having to enter the whole argument, let Y and Z are reflexive. Since Y is reflexive, Y∗=Y∗∗∗. Therefore
[TABLE]
In the other hands, because Z is the reflexive space, thus
[TABLE]
Now Using (2-1), (2-2) and Theorem 2.2, the result holds.
∎
Corollary 2.5**.**
Let bounded tri-linear map f:X×Y×Z⟶W is regular. Then
- (1)
If f∗∗∗r∗(X∗∗,W∗,Z) factors, then Y is reflexive space.
2. (2)
If f∗∗∗∗∗(W∗,X∗∗,Y∗∗) factors, then Z is reflexive space.
3. (3)
If f∗∗∗∗r∗(W∗,Z,Y) factors, then X is reflexive space.
Proof.
(1) Let f be regular. It follows that f∗∗∗r∗(X∗∗,W∗,Z)⊆Y∗. In the other hands, f∗∗∗r∗(X∗∗,W∗,Z) is factor. So for each y∗∗∗∈Y∗∗∗ there exist x∗∗∈X∗∗,w∗∈W∗ and z∈Z such that f∗∗∗r∗(x∗∗,w∗,z)=y∗∗∗. Therefore Y∗∗∗⊆Y∗.
(2) The proof similar to (1).
(3) Enough show that, if f is regular, then f∗∗∗∗r∗(W∗,Z,Y)⊆X∗. For every x∗∗∈X∗∗,y∈Y,z∈Z and w∗∈W∗ we have
⟨f∗∗∗∗r∗(w∗,z,y),x∗∗⟩=⟨w∗,f∗∗∗∗r(z,y,x∗∗)⟩=⟨w∗,f∗∗∗∗(x∗∗,y,z)⟩
=⟨fr∗∗∗∗r(x∗∗,y,z),w∗⟩=⟨fr∗∗∗∗(z,y,x∗∗),w∗⟩
=⟨z,fr∗∗∗(y,x∗∗,w∗)⟩=⟨fr∗∗(x∗∗,w∗,z),y⟩
=⟨fr∗(w∗,z,y),x∗∗⟩.
Therefore f∗∗∗∗r∗(w∗,z,y)=fr∗(w∗,z,y)∈X∗. The rest of proof has similar argument such as (1).
∎
The next corollary, let X,Y and Z be Banach spaces, IX:X⟶X is identity map on X and IY,IZ are identity on Y,Z respectively.
Corollary 2.6**.**
If IX, IY and IZ are weakly compact identity mapping, then all of them and all of their adjoints are regular.
Example 2.7**.**
- (1)
Let G be a compact group. Let 1<p,q<∞ and p1+q1=1+r1. Then by [12, Sections 2.4 and 2.5], we conclude that L1(G)⋆Lp(G)⊂Lp(G) and Lp(G)⋆Lq(G)⊂Lr(G) where (g⋆h)(x)=∫Gg(y)h(y−1x)dy for x∈G. The Banach spaces Lp(G) and Lq(G) are reflexive, thus by corollary 2.4 we conclude that the bounded tri-linear mapping
[TABLE]
defined by f(k,g,h)=(k⋆g)⋆h, is regular for every k∈L1(G),g∈Lp(G) and h∈Lq(G).
2. (2)
Let G be a locally compact group. We know from [18] that L1(G) is regular if and only if it is reflexive or equivalent to G being finite. It follows that for every finite locally compact group G, by corollary 2.4, the bounded tri-linear mapping
[TABLE]
defined by f(k,g,h)=k⋆g⋆h, is regular for every k,g and h∈L1(G).
3. (3)
C∗−algebras are standard examples of Banach algebras that are Arens regular (see[6]). We know that a C∗−algebra is reflexive if and only if it is of finite dimension. Since if A be a finite dimension C∗-algebra, then we conclude that the bounded tri-linear mapping f:A×A×A⟶A is regular by corollary 2.4.
4. (4)
Let G be a locally compact group and let M(G) be measure algebra of G, see [12, Section 2.5]. The convolution for μ1,μ2∈M(G) defined by
[TABLE]
We have
[TABLE]
for μ1,μ2 and μ3∈M(G). Therefore convolution is associative. Now we define the bounded tri-linear mapping
[TABLE]
by f(μ1,μ2,μ3)=∫ψd(μ1∗μ2∗μ3). If G is finite, then f is regular.
3. Some results for regularity
Dales, Rodriguez-Palacios and Velasco, in the theorem of their paper, [[8], Theorem 4.1], for a bonded bilinear map m:X×Y⟶Z have shown that, mr∗r∗∗∗=m∗∗∗r∗r if and only if both m and mr∗ are Arens regular. The We study some theorems from this paper for general case as following.
Remark 3.1*.*
In the next theorem, fn is n−th adjoint of f for each n∈N.
Theorem 3.2**.**
If f and frn are reular, then f4rnr=frnr4.
Proof.
Since f is regular, so f4r=fr4. Therefore f4rn=fr(n+4). In the other hands, regularity of frn follows that fr(n+4)=frnr4r. Thus frnr4r=f4rn and this completes the proof.
∎
In general the converse of preceding theorem not holds as following theorem.
Theorem 3.3**.**
Let f:X×Y×Z⟶W be a bounded tri-linear mapping. Then
- (1)
f∗∗∗∗r∗∗r=fr∗∗r∗∗∗∗* if and only if both f and fr∗∗ are regular.*
2. (2)
f∗∗∗∗r∗∗∗r=fr∗∗∗r∗∗∗∗* if and only if both f and fr∗∗∗ are regular.*
Proof.
We prove only (1), the other part has the same argument. If both f and fr∗∗ are regular, then by applying Theorem 3.2, for n=2, f∗∗∗∗r∗∗r=fr∗∗r∗∗∗∗.
Conversely, suppose that f∗∗∗∗r∗∗r=fr∗∗r∗∗∗∗. First we show that f is regular. Let {zγ} is net in Z which converge to z∗∗∈Z∗∗ in the w∗−topologies. Then for every x∗∗∈X∗∗,y∗∗∈Y∗∗ and w∗∈W∗ we have
⟨f∗∗∗∗(x∗∗,y∗∗,z∗∗),w∗⟩=⟨f∗∗∗∗r(z∗∗,y∗∗,x∗∗),w∗⟩=⟨f∗∗∗∗r∗(w∗,z∗∗,y∗∗),x∗∗⟩
=⟨f∗∗∗∗r∗∗(x∗∗,w∗,z∗∗),y∗∗⟩=⟨f∗∗∗∗r∗∗r(z∗∗,w∗,x∗∗),y∗∗⟩
=⟨fr∗∗r∗∗∗∗(z∗∗,w∗,x∗∗),y∗∗⟩=⟨z∗∗,fr∗∗r∗∗∗(w∗,x∗∗,y∗∗)⟩
=γlim⟨fr∗∗r∗∗∗(w∗,x∗∗,y∗∗),zγ⟩=γlim⟨w∗,fr∗∗r∗∗(x∗∗,y∗∗,zγ)⟩
=γlim⟨fr∗∗r∗(y∗∗,zγ,w∗),x∗∗⟩=γlim⟨y∗∗,fr∗∗r(zγ,w∗,x∗∗)⟩
=γlim⟨y∗∗,fr∗∗(x∗∗,w∗,zγ)⟩=γlim⟨fr∗∗∗(y∗∗,x∗∗,w∗),zγ⟩
=⟨z∗∗,fr∗∗∗(y∗∗,x∗∗,w∗)⟩=⟨fr∗∗∗∗(z∗∗,y∗∗,x∗∗),w∗⟩
=⟨fr∗∗∗∗r(x∗∗,y∗∗,z∗∗),w∗⟩.
Therefore f is regular. Now we show that fr∗∗ is regular. Let {xα∗∗} is net in X∗∗ which converge to x∗∗∗∗∈X∗∗∗∗ in the w∗−topologies. Then for every y∗∗∈Y∗∗,z∗∗∈Z∗∗ and w∗∗∗∈W∗∗∗ we have
⟨fr∗∗r∗∗∗∗r(x∗∗∗∗,w∗∗∗,z∗∗),y∗∗⟩=⟨fr∗∗r∗∗∗∗(z∗∗,w∗∗∗,x∗∗∗∗),y∗∗⟩
=⟨f∗∗∗∗r∗∗r(z∗∗,w∗∗∗,x∗∗∗∗),y∗∗⟩=⟨f∗∗∗∗r∗∗(x∗∗∗∗,w∗∗∗,z∗∗),y∗∗⟩⟩
=⟨x∗∗∗∗,f∗∗∗∗r∗(w∗∗∗,z∗∗,y∗∗)=αlim⟨f∗∗∗∗r∗(w∗∗∗,z∗∗,y∗∗),xα∗∗⟩
=αlim⟨w∗∗∗,f∗∗∗∗r(z∗∗,y∗∗,xα∗∗)⟩=αlim⟨w∗∗∗,f∗∗∗∗(xα∗∗,y∗∗,z∗∗)⟩
=αlim⟨w∗∗∗,fr∗∗∗∗r(xα∗∗,y∗∗,z∗∗)⟩=αlim⟨w∗∗∗,fr∗∗∗∗(z∗∗,y∗∗,xα∗∗)⟩
=αlim⟨fr∗∗∗∗∗(w∗∗∗,z∗∗,y∗∗),xα∗∗⟩=⟨x∗∗∗∗,fr∗∗∗∗∗(w∗∗∗,z∗∗,y∗∗)⟩
=⟨fr∗∗∗∗∗∗(x∗∗∗∗,w∗∗∗,z∗∗),y∗∗⟩.
It follows that fr∗∗ is regular and this completes the proof.
∎
Arens has shown [3] that bounded bilinear map m is regular if and only if for each z∗∈Z∗, the bilinear form z∗om is regular. In the next theorem we give an important characterization of regularity bounded tri-linear mappings.
Lemma 3.4**.**
Suppose X,Y,Z,W and S be normed spaces and f:X×Y×Z⟶W be a bounded tri-linear mapping and h:W⟶S be a bounded linear mapping. Then we have
- (1)
h∗∗of∗∗∗∗=(hof)∗∗∗∗.
2. (2)
h∗∗ofr∗∗∗∗r=(hof)r∗∗∗∗r.
Proof.
(1) Let {xα},{yβ} and {zγ} be nets in X,Y and Z which converge to x∗∗∈X∗∗,y∗∗∈Y∗∗ and z∗∗∈Z∗∗ in the w∗−topologies, respectively. For each s∗∈S∗ we have
⟨h∗∗of∗∗∗∗(x∗∗,y∗∗,z∗∗),s∗⟩=⟨h∗∗(f∗∗∗∗(x∗∗,y∗∗,z∗∗)),s∗⟩=⟨f∗∗∗∗(x∗∗,y∗∗,z∗∗),h∗(s∗)⟩
=⟨x∗∗,f∗∗∗(y∗∗,z∗∗,h∗(s∗))⟩=αlim⟨f∗∗∗(y∗∗,z∗∗,h∗(s∗)),xα⟩
=αlim⟨y∗∗,f∗∗(z∗∗,h∗(s∗),xα)⟩=αlimβlim⟨f∗∗(z∗∗,h∗(s∗),xα),yβ⟩
=αlimβlim⟨z∗∗,f∗(h∗(s∗),xα,yβ)⟩=αlimβlimγlim⟨f∗(h∗(s∗),xα,yβ),zγ⟩
=αlimβlimγlim⟨h∗(s∗),f(xα,yβ,zγ)⟩=αlimβlimγlim⟨s∗,h(f(xα,yβ,zγ))⟩
=αlimβlimγlim⟨s∗,hof(xα,yβ,zγ)⟩=αlimβlimγlim⟨(hof)∗(s∗,xα,yβ),zγ⟩
=αlimβlim⟨z∗∗,(hof)∗(s∗,xα,yβ)⟩=αlimβlim⟨(hof)∗∗(z∗∗,s∗,xα),yβ⟩
=αlim⟨y∗∗,(hof)∗∗(z∗∗,s∗,xα)⟩=αlim⟨(hof)∗∗∗(y∗∗,z∗∗,s∗),xα)⟩
=⟨x∗∗,(hof)∗∗∗(y∗∗,z∗∗,s∗)⟩=⟨(hof)∗∗∗∗(x∗∗,y∗∗,z∗∗),s∗⟩.
Hence h∗∗of∗∗∗∗(x∗∗,y∗∗,z∗∗)=(hof)∗∗∗∗(x∗∗,y∗∗,z∗∗). A similar argument applies for (2).
∎
Theorem 3.5**.**
Let f:X×Y×Z⟶W be a bounded tri-linear mapping and h:W⟶S be a bounded linear mapping. Then f is regular if and only if hof is regular.
Proof.
Let f is regular. Then for every x∗∗∈X∗∗,y∗∗∈Y∗∗,z∗∗∈Z∗∗ and s∗∈S∗ we have
⟨h∗∗(fr∗∗∗∗r(x∗∗,y∗∗,z∗∗)),s∗⟩=⟨fr∗∗∗∗r(x∗∗,y∗∗,z∗∗),h∗(s∗)⟩
=⟨f∗∗∗∗(x∗∗,y∗∗,z∗∗),h∗(s∗)⟩=⟨h∗∗(f∗∗∗∗(x∗∗,y∗∗,z∗∗)),s∗⟩.
Therefore h∗∗ofr∗∗∗∗r(x∗∗,y∗∗,z∗∗)=h∗∗of∗∗∗∗(x∗∗,y∗∗,z∗∗) and by applying Lemma 3.4, we implies that
[TABLE]
Thus hof is regular.
For the converse, suppose that hof is regular. By contradiction, let f be not regular. There exist x∗∗∈X∗∗,y∗∗∈Y∗∗ and z∗∗∈Z∗∗ such that f∗∗∗∗(x∗∗,y∗∗,z∗∗)=fr∗∗∗∗r(x∗∗,y∗∗,z∗∗). Therefore we have
(hof)∗∗∗∗(x∗∗,y∗∗,z∗∗)=w∗−αlimw∗−βlimw∗−γlim(hof)(xα,yβ,zγ)
=αlimβlimγlim⟨f(xα,yβ,zγ),h⟩=⟨f∗∗∗∗(x∗∗,y∗∗,z∗∗),h⟩
=⟨fr∗∗∗∗r(x∗∗,y∗∗,z∗∗),h⟩=γlimβlimαlim⟨f(xα,yβ,zγ),h⟩
=w∗−γlimw∗−βlimw∗−αlim(hof)(xα,yβ,zγ)=(hof)r∗∗∗∗r(x∗∗,y∗∗,z∗∗).
It follows that (hof)∗∗∗∗(x∗∗,y∗∗,z∗∗)=(hof)r∗∗∗∗r(x∗∗,y∗∗,z∗∗).
∎
Another interesting case of regularity is in the following.
Theorem 3.6**.**
Let f:X×Y×Z⟶W be a bounded tri-linear mapping, m:X×Y⟶Z be a bounded bilinear mapping, T:X×Y⟶W defined by T(x,y)=f(x,y,m(x,y)) for each x∈X,y∈Y and m∗∗∗ is factors. Then T is regular if and only if f is regular.
Proof.
Let T is regular. The mapping m∗∗∗:X∗∗×Y∗∗⟶Z∗∗ is onto, so for each z∗∗∈Z∗∗ There must then exist x∗∗∈X∗∗ and y∗∗∈Y∗∗ such that m∗∗∗(x∗∗,y∗∗)=z∗∗. Let {xα} and {yβ} are nets in X and Y which converge to x∗∗ and y∗∗ in the w∗−topologies, respectively. For every w∗∈W∗ we have
⟨f∗∗∗∗(x∗∗,y∗∗,z∗∗),w∗⟩=⟨f∗∗∗∗(x∗∗,y∗∗,m∗∗∗(x∗∗,y∗∗)),w∗⟩
=⟨x∗∗,f∗∗∗(y∗∗,m∗∗∗(x∗∗,y∗∗),w∗)⟩=αlim⟨f∗∗∗(y∗∗,m∗∗∗(x∗∗,y∗∗),w∗),xα⟩
=αlim⟨y∗∗,f∗∗(m∗∗∗(x∗∗,y∗∗),w∗,xα)⟩=αlimβlim⟨f∗∗(m∗∗∗(x∗∗,y∗∗),w∗,xα),yβ⟩
=αlimβlim⟨(m∗∗∗(x∗∗,y∗∗),f∗(w∗,xα,yβ)⟩=αlimβlim⟨x∗∗,m∗∗(y∗∗,f∗(w∗,xα,yβ))⟩
=αlimβlimαlim⟨m∗∗(y∗∗,f∗(w∗,xα,yβ)),xα⟩=αlimβlimαlim⟨y∗∗,m∗(f∗(w∗,xα,yβ),xα)⟩
=αlimβlimαlimβlim⟨m∗(f∗(w∗,xα,yβ),xα),yβ⟩=αlimβlimαlimβlim⟨f∗(w∗,xα,yβ),m(xα,yβ)⟩
=αlimβlimαlimβlim⟨w∗,f(xα,yβ,m(xα,yβ))⟩=αlimβlimαlimβlim⟨w∗,T(xα,yβ)⟩
=αlimβlimαlimβlim⟨T∗(w∗,xα),yβ⟩=αlimβlimαlim⟨y∗∗,T∗(w∗,xα)⟩
=αlimβlimαlim⟨T∗∗(y∗∗,w∗),xα⟩=αlimβlim⟨x∗∗,T∗∗(y∗∗,w∗)⟩
=αlimβlim⟨T∗∗∗(x∗∗,y∗∗),w∗⟩=αlimβlim⟨Tr∗∗∗r(x∗∗,y∗∗),w∗⟩
=αlimβlim⟨Tr∗∗∗(y∗∗,x∗∗),w∗⟩=αlimβlim⟨y∗∗,Tr∗∗(x∗∗,w∗)⟩
=αlimβlimβlim⟨Tr∗∗(x∗∗,w∗),yβ⟩=αlimβlimβlim⟨x∗∗,Tr∗(w∗,yβ)⟩
=αlimβlimβlimαlim⟨Tr∗(w∗,yβ),xα⟩=αlimβlimβlimαlim⟨w∗,Tr(yβ,xα)⟩
=αlimβlimβlimαlim⟨w∗,T(xα,yβ)⟩=αlimβlimβlimαlim⟨w∗,f(xα,yβ,m(xα,yβ))⟩
=αlimβlimβlimαlim⟨w∗,fr(m(xα,yβ),yβ,xα)⟩=αlimβlimβlimαlim⟨fr∗(w∗,m(xα,yβ),yβ),xα⟩
=αlimβlimβlim⟨x∗∗,fr∗(w∗,m(xα,yβ),yβ)⟩=αlimβlimβlim⟨fr∗∗(x∗∗,w∗,m(xα,yβ)),yβ⟩
=αlimβlim⟨y∗∗,fr∗∗(x∗∗,w∗,m(xα,yβ)⟩=αlimβlim⟨fr∗∗∗(y∗∗,x∗∗,w∗),m(xα,yβ)⟩
=αlimβlim⟨m∗(fr∗∗∗(y∗∗,x∗∗,w∗),xα),yβ⟩=αlim⟨y∗∗,m∗(fr∗∗∗(y∗∗,x∗∗,w∗),xα)⟩
=αlim⟨m∗∗(y∗∗,fr∗∗∗(y∗∗,x∗∗,w∗)),xα⟩=⟨x∗∗,m∗∗(y∗∗,fr∗∗∗(y∗∗,x∗∗,w∗))⟩
=⟨m∗∗∗(x∗∗,y∗∗),fr∗∗∗(y∗∗,x∗∗,w∗)⟩=⟨fr∗∗∗∗(m∗∗∗(x∗∗,y∗∗),y∗∗,x∗∗),w∗⟩
=⟨fr∗∗∗∗r(x∗∗,y∗∗,m∗∗∗(x∗∗,y∗∗)),w∗⟩=⟨fr∗∗∗∗r(x∗∗,y∗∗,z∗∗),w∗⟩.
Therefore f is regular.
For the converse, suppose that f is regular. For every w∗∈W∗ we have
⟨T∗∗∗(x∗∗,y∗∗),w∗⟩=⟨x∗∗,T∗∗(y∗∗,w∗)⟩=αlim⟨T∗∗(y∗∗,w∗),xα⟩
=αlim⟨y∗∗,T∗(w∗,xα)⟩=αlimβlim⟨T∗(w∗,xα),yβ⟩
=αlimβlim⟨w∗,T(xα,yβ)⟩=αlimβlim⟨w∗,f(xα,yβ,m(xα,yβ))⟩
=αlimβlim⟨f∗(w∗,xα,yβ),m(xα,yβ)⟩=αlimβlim⟨m∗(f∗(w∗,xα,yβ),xα),yβ⟩
=αlimβlim⟨m∗∗(yβ,f∗(w∗,xα,yβ)),xα⟩=αlimβlim⟨m∗∗∗(xα,yβ),f∗(w∗,xα,yβ)⟩
=αlimβlim⟨f∗∗(m∗∗∗(xα,yβ),w∗,xα),yβ⟩=αlim⟨y∗∗,f∗∗(m∗∗∗(xα,yβ),w∗,xα)⟩
=αlim⟨f∗∗∗(y∗∗,m∗∗∗(xα,yβ),w∗),xα⟩=⟨x∗∗,f∗∗∗(y∗∗,m∗∗∗(xα,yβ),w∗)⟩
=⟨f∗∗∗∗(x∗∗,y∗∗,m∗∗∗(xα,yβ)),w∗⟩=⟨fr∗∗∗∗r(x∗∗,y∗∗,m∗∗∗(xα,yβ)),w∗⟩
=⟨fr∗∗∗∗(m∗∗∗(xα,yβ),y∗∗,x∗∗),w∗⟩=⟨m∗∗∗(xα,yβ),fr∗∗∗(y∗∗,x∗∗,w∗)⟩
=⟨xα,m∗∗(yβ,fr∗∗∗(y∗∗,x∗∗,w∗))⟩=⟨yβ,m∗(fr∗∗∗(y∗∗,x∗∗,w∗),xα)⟩
=⟨fr∗∗∗(y∗∗,x∗∗,w∗),m(xα,yβ)⟩=⟨y∗∗,fr∗∗(x∗∗,w∗,m(xα,yβ))⟩
=βlim⟨fr∗∗(x∗∗,w∗,m(xα,yβ)),yβ⟩=βlim⟨x∗∗,fr∗(w∗,m(xα,yβ),yβ)⟩
=βlimαlim⟨fr∗(w∗,m(xα,yβ),yβ),xα⟩=βlimαlim⟨w∗,fr(m(xα,yβ),yβ,xα)⟩
=βlimαlim⟨w∗,f(xα,yβ,m(xα,yβ))⟩=βlimαlim⟨w∗,T(xα,yβ)⟩
=βlimαlim⟨w∗,Tr(yβ,xα)⟩=βlimαlim⟨Tr∗(w∗,yβ),xα⟩
=βlim⟨x∗∗,Tr∗(w∗,yβ)⟩=βlim⟨Tr∗∗(x∗∗,w∗),yβ⟩
=⟨y∗∗,Tr∗∗(x∗∗,w∗)⟩=⟨Tr∗∗∗(y∗∗,x∗∗),w∗⟩=⟨Tr∗∗∗r(x∗∗,y∗∗),w∗⟩.
Therefore T∗∗∗=Tr∗∗∗r , as claimed.
∎
Theorem 3.7**.**
Let X,Y,Z,W and S be Banach spaces, f:X×Y×Z⟶W be a bounded tri-linear mapping and x∈X,y∈Y,z∈Z arbitrary. Then
- (1)
Let g1:S×Y×Z⟶W be a bounded tri-linear mapping and let h1:X⟶S be a bounded linear mapping such that f(x,y,z)=g1(h1(x),y,z). If h1 is weakly compact, then f∗∗∗∗r∗(W∗∗∗,Z∗∗,Y∗∗)⊆X∗.
2. (2)
Let g2:X×S×Z⟶W be a bounded tri-linear mapping and let h2:Y⟶S be a bounded linear mapping such that f(x,y,z)=g2(x,h2(y),z). If h2 is weakly compact, then f∗∗∗r∗(X∗∗,W∗,Z∗∗)⊆Y∗.
3. (3)
Let g3:X×Y×S⟶W be a bounded tri-linear mapping and let h3:Z⟶S be a bounded linear mapping such that f(x,y,z)=g3(x,y,h3(z)). If h3 is weakly compact, then f∗∗∗∗∗(W∗∗∗,X∗∗,Y∗∗)⊆Z∗.
Proof.
We prove only (1), the other parts have the same argument. For every x∈X,y∈Y,z∈Z and w∗∈W∗ we have
[TABLE]
Therefore f∗(w∗,x,y)=g1∗(w∗,h1(x),y), and implies that for every z∗∗∈Z∗∗,
[TABLE]
So f∗∗(z∗∗,w∗,x)=g1∗∗(z∗∗,w∗,h1(x)) and implies that for every y∗∗∈Y∗∗,
⟨f∗∗∗(y∗∗,z∗∗,w∗),x⟩=⟨y∗∗,f∗∗(z∗∗,w∗,x)⟩=⟨y∗∗,g1∗∗(z∗∗,w∗,h1(x))⟩
=⟨g1∗∗∗(y∗∗,z∗∗,w∗),h1(x)⟩=⟨h1∗(g1∗∗∗(y∗∗,z∗∗,w∗)),x⟩.
Thus f∗∗∗(y∗∗,z∗∗,w∗)=h1∗(g1∗∗∗(y∗∗,z∗∗,w∗)) and implies that for every x∗∗∈X∗∗,
⟨f∗∗∗∗(x∗∗,y∗∗,z∗∗),w∗⟩=⟨x∗∗,f∗∗∗(y∗∗,z∗∗,w∗)⟩=⟨x∗∗,h1∗(g1∗∗∗(y∗∗,z∗∗,w∗))⟩
=⟨h1∗∗(x∗∗),(g1∗∗∗(y∗∗,z∗∗,w∗)⟩=⟨g1∗∗∗∗(h1∗∗(x∗∗),y∗∗,z∗∗),w∗⟩.
Therefore for every w∗∗∗∈W∗∗∗ we have
⟨f∗∗∗∗r∗(w∗∗∗,z∗∗,y∗∗),x∗∗⟩=⟨w∗∗∗,f∗∗∗∗r(z∗∗,y∗∗,x∗∗)⟩=⟨w∗∗∗,f∗∗∗∗(x∗∗,y∗∗,z∗∗)⟩
=⟨w∗∗∗,g1∗∗∗∗(h1∗∗(x∗∗),y∗∗,z∗∗)⟩=⟨w∗∗∗,g1∗∗∗∗r(z∗∗,y∗∗,h1∗∗(x∗∗))⟩
=⟨g1∗∗∗∗r∗(w∗∗∗,z∗∗,y∗∗),h1∗∗(x∗∗)⟩=⟨h1∗∗∗(g1∗∗∗∗r∗(w∗∗∗,z∗∗,y∗∗)),x∗∗⟩.
Therefore f∗∗∗∗r∗(w∗∗∗,z∗∗,y∗∗)=h1∗∗∗(g1∗∗∗∗r∗(w∗∗∗,z∗∗,y∗∗)). The weak compactness of h1 implies that of h1∗, from which we have h1∗∗∗(S∗∗∗)⊆X∗. Thus h1∗∗∗(g1∗∗∗∗r∗(w∗∗∗,z∗∗,y∗∗))∈X∗ and this completes the proof.
∎
This theorem, combined with Theorem 2.2, yields the next result.
Corollary 3.8**.**
With the assumptions Theorem 3.7, if h2 and h3 are weakly compact, then f is regular.
Proof.
Both h2 and h3 are weakly compact, so by Theorem 3.7 we have
[TABLE]
In particular
[TABLE]
Now by Theorem 2.2, f is regular.
∎
The converse of previous result is not true in general sense as following corollary.
Corollary 3.9**.**
With the assumptions Theorem 3.7, if f is regular and both g2∗∗∗r∗ and g3∗∗∗∗∗ are factors, then h2 and h3 are weakly compact.
Proof.
Since f∗∗∗r∗(X∗∗,W∗,Z∗∗)=h2∗∗∗(g2∗∗∗r∗(X∗∗,W∗,Z∗∗)), thus h2∗∗∗(g2∗∗∗r∗(X∗∗,W∗,Z∗∗))⊆Y∗. In the other hands g2∗∗∗r∗ is factors, so implies that h2∗∗∗(S∗∗∗)⊆Y∗. Therefore h2∗ is weakly compact and implies that h2 is weakly compact. The other part has the same argument for h3.
∎
4. The fourth adjoint of a tri-derivation
Definition 4.1**.**
Let (π1,X,π2) be a Banach A−module. A bounded tri-linear mapping D:A×A×A⟶X is
said to be a tri-derivation when
- (1)
D(π(a,d),b,c)=π2(D(a,b,c),d)+π1(a,D(d,b,c)),
2. (2)
D(a,π(b,d),c)=π2(D(a,b,c),d)+π1(b,D(a,d,c)),
3. (3)
D(a,b,π(c,d))=π2(D(a,b,c),d)+π1(c,D(a,b,d)),
for each a,b,c,d∈A. If (π1,X,π2) is a Banach A−module, then (π2r∗r,X∗,π1∗) is the dual Banach A−module of (π1,X,π2). Therefore a bounded tri-linear mapping D:A×A×A⟶X∗ is a tri-derivation when
- (1)
D(π(a,d),b,c)=π1∗(D(a,b,c),d)+π2r∗r(a,D(d,b,c)),
2. (2)
D(a,π(b,d),c)=π1∗(D(a,b,c),d)+π2r∗r(b,D(a,d,c)),
3. (3)
D(a,b,π(c,d))=π1∗(D(a,b,c),d)+π2r∗r(c,D(a,b,d)).
It can also be written, a bounded tri-linear mapping D:A×A×A⟶A is said to be a tri-derivation when
- (1)
D(π(a,d),b,c)=π(D(a,b,c),d)+π(a,D(d,b,c)),
2. (2)
D(a,π(b,d),c)=π(D(a,b,c),d)+π(b,D(a,d,c)),
3. (3)
D(a,b,π(c,d))=π(D(a,b,c),d)+π(c,D(a,b,d)).
Example 4.2**.**
Let A be a Banach algebra, for any a,b∈A the symbol [a,b]=ab−ba stands for multiplicative commutator of a and b. Let Mn×n(C) be the Banach algebra of all n×n matrix and A={(x0y0)∈Mn×n(C)∣x,y∈C}. Then A is Banach algebra with the norm
[TABLE]
We define D:A×A×A⟶A to be the bounded tri-linear map given by
[TABLE]
Then for a=(x10y10),b=(x20y20),c=(x30y30) and d=(x40y40)∈A we have
D(π(a,d),b,c)=D((x1x40x1y40),(x20y20),(x30y30))=[(0010),(x1x2x3x40x1x2x4y30)]
=(00−x1x2x3x40)=(00−x1x2x30)(x40y40)+(x10y10)(00−x2x3x40)
=((0000)−(00x1x2x30))(x40y40)+(x10y10)((0000)−(00x2x3x40))
=((0010)(x1x2x30x1x2y30)−(x1x2x30x1x2y30)(0010))(x40y40)
+(x10y10)((0010)(x2x3x40x2x4y30)−(x2x3x40x2x4y30)(0010))
=[(0010),(x1x2x30x1x2y30)](x40y40)+(x10y10)[(0010),(x2x3x40x2x4y30)]
=[(0010),(x10y10)(x20y20)(x30y30)](x40y40)
+(x10y10)[(0010),(x40y40)(x20y20)(x30y30)]
=D((x10y10),(x20y20),(x30y30))(x40y40)
+(x10y10)D((x40y40),(x20y20),(x30y30))
=π(D(a,b,c),d)+π(a,D(d,b,c)).
Similarly, we have D(a,π(b,d),c)=π(D(a,b,c),d)+π(b,D(a,d,c)) and D(a,b,π(c,d))=π(D(a,b,c),d)+π(c,D(a,b,d)). Thus D is tri-derivation.
In the section, we provide a necessary and sufficient condition such that the fourth adjoint D∗∗∗∗ of a tri-derivation D:A×A×A⟶X is again a tri-derivation. For the fourth adjoint D∗∗∗∗ of a tri-derivation D:A×A×A⟶X, we are faced with the case eight:
[TABLE]
[TABLE]
[TABLE]
[TABLE]
[TABLE]
[TABLE]
[TABLE]
[TABLE]
In the follwing, we prove the state of case 1. The remaining state are proved in the same way.
Theorem 4.3**.**
Let (π1,X,π2) be a Banach A−module and D:A×A×A⟶X be a tri-derivation. Then D∗∗∗∗:(A∗∗,□)×(A∗∗,□)×(A∗∗,□)⟶X∗∗ is a tri-derivation if and only if
- (1)
π2∗∗r∗(D∗∗∗∗(A,A,A∗∗),X∗)⊆A∗,
2. (2)
π2∗∗∗∗(X∗,D∗∗∗∗(A,A∗∗,A∗∗))⊆A∗,
3. (3)
D∗∗∗∗r∗(π1∗∗∗∗(X∗,A∗∗),A∗∗,A∗∗)⊆A∗,
4. (4)
D∗∗∗∗∗∗(A∗∗,π1∗∗∗∗(X∗,A∗∗),A)⊆A∗,
5. (5)
D∗∗∗∗∗∗∗(A∗∗,A∗∗,π1∗∗∗∗(X∗,A∗∗))⊆A∗.
Proof.
Let D:A×A×A⟶X be a tri-derivation and (1),(2),(3),(4),(5) holds.
If {aα},{bβ},{cγ} and {dτ} be bounded nets in A , converging in w∗−topology
to a∗∗,b∗∗,c∗∗ and d∗∗∈A∗∗ respectively, in this case using (2), we conclude that
[TABLE]
Thus for every x∗∈X∗ we have
⟨D∗∗∗∗(π∗∗∗(a∗∗,d∗∗),b∗∗,c∗∗),x∗⟩=⟨π∗∗∗(a∗∗,d∗∗),D∗∗∗(b∗∗,c∗∗,x∗)⟩
=⟨a∗∗,π∗∗(d∗∗,D∗∗∗(b∗∗,c∗∗,x∗))⟩=αlim⟨π∗∗(d∗∗,D∗∗∗(b∗∗,c∗∗,x∗)),aα⟩
=αlim⟨d∗∗,π∗(D∗∗∗(b∗∗,c∗∗,x∗),aα)⟩=αlimτlim⟨π∗(D∗∗∗(b∗∗,c∗∗,x∗),aα),dτ⟩
=αlimτlim⟨D∗∗∗(b∗∗,c∗∗,x∗),π(aα,dτ)⟩=αlimτlim⟨b∗∗,D∗∗(c∗∗,x∗,π(aα,dτ))⟩
=αlimτlimβlim⟨D∗∗(c∗∗,x∗,π(aα,dτ)),bβ⟩=αlimτlimβlim⟨c∗∗,D∗(x∗,π(aα,dτ),bβ)⟩
=αlimτlimβlimγlim⟨D∗(x∗,π(aα,dτ),bβ),cγ⟩=αlimτlimβlimγlim⟨x∗,D(π(aα,dτ),bβ,cγ)⟩
=αlimτlimβlimγlim⟨x∗,π2(D(aα,bβ,cγ),dτ)+π1(aα,D(dτ,bβ,cγ))⟩
=αlimτlimβlimγlim⟨x∗,π2(D(aα,bβ,cγ),dτ)⟩+αlimτlimβlimγlim⟨x∗,π1(aα,D(dτ,bβ,cγ))⟩
=⟨x∗,π2∗∗∗(D∗∗∗∗(a∗∗,b∗∗,c∗∗),d∗∗)⟩+⟨x∗,π1∗∗∗(a∗∗,D∗∗∗∗(d∗∗,b∗∗,c∗∗))⟩
=⟨π2∗∗∗(D∗∗∗∗(a∗∗,b∗∗,c∗∗),d∗∗)+π1∗∗∗(a∗∗,D∗∗∗∗(d∗∗,b∗∗,c∗∗)),x∗⟩.
Therefore
[TABLE]
Applying (1) and (3) respectively, we can deduce that
[TABLE]
[TABLE]
So, by proof very similar we can deduce that
[TABLE]
Applying (4) and (5), we can deduce that
[TABLE]
Thus we deduce that
[TABLE]
By comparing equations (4-1),(4-2) and (4-3) follows that D∗∗∗∗:(A∗∗,□)×(A∗∗,□)×(A∗∗,□)⟶X∗∗ is a tri-derivation.
For the converse, let D and D∗∗∗∗:(A∗∗,□)×(A∗∗,□)×(A∗∗,□)⟶X∗∗ are tri-derivation. We show that (1),(2),(3),(4),(5) holds. We shall only prove (2) the others have similar argument. Fourth adjoint D∗∗∗∗ is tri-derivation, thus we have
[TABLE]
In the other hands, the mapping D is tri-derivation, thus we have
[TABLE]
Therefore follows that π2∗∗∗(D∗∗∗∗(a,b∗∗,c∗∗),d∗∗)=w∗−τlimw∗−βlimw∗−γlimπ2(D(a,bβ,cγ),dτ). So, for every d∗∗∈A∗∗ we have
⟨π2∗∗∗∗(x∗,D∗∗∗∗(a,b∗∗,c∗∗)),d∗∗⟩=⟨x∗,π2∗∗∗(D∗∗∗∗(a,b∗∗,c∗∗),d∗∗)⟩
=τlimβlimγlim⟨x∗,π2(D(a,bβ,cγ),dτ)⟩=τlimβlimγlim⟨x∗,π2r(dτ,D(a,bβ,cγ))⟩
=τlimβlimγlim⟨π2r∗(x∗,dτ),D(a,bβ,cγ)⟩=τlimβlimγlim⟨D∗(π2r∗(x∗,dτ),a,bβ),cγ⟩
=τlimβlim⟨c∗∗,D∗(π2r∗(x∗,dτ),a,bβ)⟩=τlimβlim⟨D∗∗(c∗∗,π2r∗(x∗,dτ),a),bβ⟩
=τlim⟨b∗∗,D∗∗(c∗∗,π2r∗(x∗,dτ),a)⟩=τlim⟨D∗∗∗(b∗∗,c∗∗,π2r∗(x∗,dτ)),a⟩
=τlim⟨D∗∗∗∗(a,b∗∗,c∗∗),π2r∗(x∗,dτ)⟩=τlim⟨D∗∗∗∗(a,b∗∗,c∗∗),π2r∗r(dτ,x∗)⟩
=τlim⟨π2r∗r∗(D∗∗∗∗(a,b∗∗,c∗∗),dτ),x∗⟩=τlim⟨π2r∗r∗∗(x∗,D∗∗∗∗(a,b∗∗,c∗∗)),dτ⟩
=⟨π2r∗r∗∗(x∗,D∗∗∗∗(a,b∗∗,c∗∗)),d∗∗⟩.
As π2r∗r∗∗(x∗,D∗∗∗∗(a,b∗∗,c∗∗)) always lies in A∗, we have reached (2).
∎
For case 2, forth adjoint D∗∗∗∗ of tri-derivation D:A×A×A⟶X is a tri-derivation if and only if
- (1)
π2∗∗r∗(D∗∗∗∗(A∗∗,A∗∗,A∗∗),X∗)⊆A∗,
2. (2)
D∗∗∗∗r∗(π1∗∗∗∗(X∗,A∗∗),A∗∗,A∗∗)⊆A∗,
3. (3)
D∗∗∗∗∗∗(A∗∗,π1∗∗∗∗(X∗,A∗∗),A)⊆A∗,
4. (4)
D∗∗∗∗∗∗∗(A∗∗,A∗∗,π1∗∗∗∗(X∗,A∗∗))⊆A∗.
For case 3, Forth adjoint D∗∗∗∗ of tri-derivation D:A×A×A⟶X is a tri-derivation if and only if
- (1)
π2∗∗∗∗(X∗,D∗∗∗∗(A,A∗∗,A∗∗))⊆A∗,
2. (2)
D∗∗∗∗∗∗(A∗∗,π1∗∗∗∗(X∗,A∗∗),A)⊆A∗,
3. (3)
D∗∗∗∗∗∗∗(A∗∗,A∗∗,π1∗∗∗∗(X∗,A∗∗))⊆A∗.
For case 4, Forth adjoint D∗∗∗∗ of tri-derivation D:A×A×A⟶X is a tri-derivation if and only if
- (1)
π2∗∗r∗(D∗∗∗∗(A,A,A∗∗),X∗)⊆A∗,
2. (2)
π2∗∗∗∗(X∗,D∗∗∗∗(A,A∗∗,A∗∗))⊆A∗,
3. (3)
D∗∗∗∗r∗(π1∗∗∗∗(X∗,A∗∗),A∗∗,A∗∗)⊆A∗,
4. (4)
D∗∗∗∗∗(π1∗∗∗∗(X∗,A∗∗),A,A)⊆A∗,
5. (5)
D∗∗∗∗∗∗(A∗∗,π1∗∗∗∗(X∗,A∗∗),A)⊆A∗,
6. (6)
D∗∗∗∗∗∗∗(A∗∗,A∗∗,π1∗∗∗∗(X∗,A∗∗))⊆A∗.
For case 5, Forth adjoint D∗∗∗∗ of tri-derivation D:A×A×A⟶X is a tri-derivation if and only if
- (1)
π2∗∗r∗(D∗∗∗∗(A∗∗,A∗∗,A∗∗),X∗)⊆A∗,
2. (2)
π2∗∗∗∗(X∗,D∗∗∗∗(A,A∗∗,A∗∗))⊆A∗,
3. (3)
D∗∗∗∗r∗(π1∗∗∗∗(X∗,A∗∗),A∗∗,A∗∗)⊆A∗,
4. (4)
D∗∗∗∗∗∗(A∗∗,π1∗∗∗∗(X∗,A∗∗),A)⊆A∗,
5. (5)
D∗∗∗∗∗∗∗(A∗∗,A∗∗,π1∗∗∗∗(X∗,A∗∗))⊆A∗.
For case 6, Forth adjoint D∗∗∗∗ of tri-derivation D:A×A×A⟶X is a tri-derivation if and only if
- (1)
π2∗∗r∗(D∗∗∗∗(A∗∗,A∗∗,A∗∗),X∗)⊆A∗,
2. (2)
D∗∗∗∗r∗(π1∗∗∗∗(X∗,A∗∗),A∗∗,A∗∗)⊆A∗,
3. (3)
D∗∗∗∗∗(π1∗∗∗∗(X∗,A∗∗),A,A)⊆A∗,
4. (4)
D∗∗∗∗∗∗(A∗∗,π1∗∗∗∗(X∗,A∗∗),A)⊆A∗,
5. (5)
D∗∗∗∗∗∗∗(A∗∗,A∗∗,π1∗∗∗∗(X∗,A∗∗))⊆A∗.
For case 7, Forth adjoint D∗∗∗∗ of tri-derivation D:A×A×A⟶X is a tri-derivation if and only if
- (1)
π2∗∗∗∗(X∗,D∗∗∗∗(A,A∗∗,A∗∗))⊆A∗,
2. (2)
π2∗∗r∗(D∗∗∗∗(A,A,A∗∗),X∗)⊆A∗,
3. (3)
D∗∗∗∗∗(π1∗∗∗∗(X∗,A∗∗),A,A)⊆A∗,
4. (4)
D∗∗∗∗∗∗(A∗∗,π1∗∗∗∗(X∗,A∗∗),A)⊆A∗,
5. (5)
D∗∗∗∗∗∗∗(A∗∗,A∗∗,π1∗∗∗∗(X∗,A∗∗))⊆A∗.
For case 8, Forth adjoint D∗∗∗∗ of tri-derivation D:A×A×A⟶X is a tri-derivation if and only if
- (1)
π2∗∗∗∗(X∗,D∗∗∗∗(A,A∗∗,A∗∗))⊆A∗,
2. (2)
π2∗∗r∗(D∗∗∗∗(A∗∗,A∗∗,A∗∗),X∗)⊆A∗,
3. (3)
D∗∗∗∗r∗(π1∗∗∗∗(X∗,A∗∗),A∗∗,A∗∗)⊆A∗,
4. (4)
D∗∗∗∗∗(π1∗∗∗∗(X∗,A∗∗),A,A)⊆A∗,
5. (5)
D∗∗∗∗∗∗(A∗∗,π1∗∗∗∗(X∗,A∗∗),A)⊆A∗,
6. (6)
D∗∗∗∗∗∗∗(A∗∗,A∗∗,π1∗∗∗∗(X∗,A∗∗))⊆A∗.
Remark 4.4*.*
For adjoint Dr∗∗∗∗r of tri-derivation D:A×A×A⟶X we can use the same argument.