Tensor product for g-fusion frame in Hilbert C
∗−modules
Fakhr-dine Nhari1, Mohamed Rossafi2, Youssef Aribou3
1Laboratory Analysis, Geometry and Applications Department of Mathematics, Faculty of Sci- ences, University of Ibn Tofail, Kenitra, Morocco
2Department of Mathematics, Faculty of Sciences, Dhar El Mahraz University Sidi Mohamed Ben Abdellah, P. O. Box 1796 Fez Atlas, Morocco
3Laayoune Higher School of Technology, Ibn Zohr University, Laayoune, Morocco
E-mail: 1[email protected], 2[email protected], 3[email protected]
Abstract.
In this paper, we stady the tensor product ofg−fusion frame in HilbertC∗−modules and we give the frame operator for a pair ofg−fusion Bessel sequences in tensor product of Hilbert C∗−modules.
Keywords: G-fusion Frame, tensor product, Hilbert C∗−module.
1. Introduction and Preliminaries
In the study of vector spaces one of the most important concepts is that of a basis, allowing each element in the space to be written as a linear combination of the elements in the basis.
However, the conditions to a basis are very restrictive: linear independence between the ele- ments. This makes it hard or even impossible to find bases satisfying extra conditions, and this is the reason that one might look for a more flexible substitute. Frames are such tools.
A frame for a vector space equipped with inner product also allows each element in the space to be written as a linear combination of the elements in the frame, but linear independence between the frame elements is not required.
Frames for Hilbert spaces were introduced by Duffin and schaefer [5] in 1952 to study some deep problems in nonharmonic fourier series by abstracting the fondamental notion of Gabor [6] for signal processing.
136
Tensor product for g-fusion frame in HilbertC∗−modules 137
Many generalizations of the concept of frame have been defined in HilbertC∗-modules [7,9, 11,12,13,14,15]
Tensor products have important applications, for example tensor products are useful in the approximation of multi-variate functions of combinations of univariate ones.
In this paper we consider the tensor product of HilbertC∗−modules and we generalize some of known results about frames to generalized-fusion frames in HilbertC∗−modules.
The paper is organized as follows, we continue this introductory section we briefly recall the definitions and basic properties of C∗−algebra and Hilbert C∗−modules. In section 2, we introduce the concept ofg−fusion frame, and gives an equivalent definition. In section 3, we introduce the concept of the tensor product of g−fusion frame and gives some properties.
Finally in section 4, we discuss the frame operator for a pair ofg−fusion Bessel sequences in tensor product of HilbertC∗−modules.
In the following we briefly recall the definitions and basic properties ofC∗-algebra, Hilbert A-modules. Our reference forC∗-algebras is [4,3]. For aC∗-algebraAifa∈ A is positive we writea≥0 andA+ denotes the set of positive elements ofA.
Definition 1.1. [3]. IfAis a Banach algebra, an involution is a mapa→a∗ ofAinto itself such that for allaandb inAand all scalarsαthe following conditions hold:
(1) (a∗)∗=a.
(2) (ab)∗=b∗a∗.
(3) (αa+b)∗= ¯αa∗+b∗.
Definition 1.2. [3]. AC∗-algebraAis a Banach algebra with involution such that : ka∗ak=kak2
for everyainA.
Example 1.3. B=B(H) the algebra of bounded operators on a Hilbert space, is aC∗-algebra, where for each operatorA,A∗ is the adjoint of A.
Definition 1.4. [8]. LetA be a unital C∗-algebra andH be a left A-module, such that the linear structures ofA and U are compatible. H is a pre-Hilbert A-module if H is equipped with anA-valued inner producth., .i:H×H → A, such that is sesquilinear, positive definite and respects the module action. In the other words,
(i) hx, xi ≥0 for allx∈H andhx, xi= 0 if and only if x= 0.
(ii) hax+y, zi=ahx, zi+hy, zifor alla∈ Aandx, y, z∈H.
(iii) hx, yi=hy, xi∗ for allx, y∈H.
136137
138 Fakhr-dine Nhari, Mohamed Rossafi and Youssef Aribou
For x ∈ H, we define ||x|| = ||hx, xi||12. If H is complete with ||.||, it is called a Hilbert A-module or a HilbertC∗-module overA.
Throughout this paper I and J are finite or countably infinite index sets, we fix the nota- tionsAand B for a given unitalC∗−algebras, H and K are the countably generated Hilbert A−module andB−module, respectively. Let{Hi}i∈I and{Kj}j∈J are the sequences of closed orthogonally complemented submodules of H and K, respectively. {Wi}i∈I and {Vj}j∈J are the sequences of HilbertC∗−modules. End∗A(H, Wi) is a set of all adjointable operator from H toWi. In particularEnd∗A(H) denote the set of all adjointable operators onH. PHi denote the orthogonal projection onto the closed submodule orthogonally complementedHi ofH.
Define the HilbertA−module
l2({Wi}i∈I) ={{xi}i∈I :xi ∈Wi,kX
i∈I
hxi, xiik<∞}
withA−valued inner producth{xi}i∈I,{yi}i∈Ii=P
i∈Ihxi, yii, where{xi}i∈I,{y}i∈I ∈l2({Wi}i∈I) Lemma 1.5. [2]. Let H and K two Hilbert A-modules and T ∈ End∗A(H, K). Then the following statements are equivalent:
(i) T is surjective.
(ii) T∗is bounded below with respect to norm, i.e., there ism >0such thatkT∗xk ≥mkxk for allx∈K.
(iii) T∗ is bounded below with respect to the inner product, i.e., there is m0 >0 such that hT∗x, T∗xi ≥m0hx, xifor all x∈K.
Lemma 1.6. [1]. LetU andH two HilbertA-modules andT ∈End∗A(U, H). Then:
(i) IfT is injective andT has closed range, then the adjointable mapT∗T is invertible and k(T∗T)−1k−1≤T∗T ≤ kTk2.
(ii) IfT is surjective, then the adjointable mapT T∗ is invertible and k(T T∗)−1k−1≤T T∗≤ kTk2.
Lemma 1.7. [2] Let H be a Hilbert A-module over aC∗-algebraA, and T ∈End∗A(H) such thatT∗=T. The following statements are equivalent:
(i) T is surjective.
(ii) There arem, M >0 such thatmkxk ≤ kT xk ≤Mkxk, for allx∈H.
(iii) There arem0, M0 >0 such that m0hx, xi ≤ hT x, T xi ≤M0hx, xifor allx∈H.
137138138
Tensor product for g-fusion frame in HilbertC∗−modules 139
2. g−fusion frame in HilbertC∗−modules We begin this section with the following lemma:
Lemma 2.1. Let {Hi}i∈I be a sequence of orthogonally complemented closed submodules of H andT ∈ End∗A(H) invertible, if T∗T Hi ⊆Hi for each i∈I, then{T Hi}i∈I is a sequence of orthogonally complemented closed submodules andPHiT∗=PHiT∗PT Hi.
Proof. Firstly for each i∈I, T :Hi →T Hi is invertible, so each T Hi is a closed submodule of H. We show that H =T Hi⊕T(Hi⊥). SinceH =T H, then for each x∈H, there exists y ∈ H sutch that x = T y. On the other hand y = u+v, for some u ∈ Hi and v ∈ Hi⊥. Hencex=T u+T v, whereT u∈T HiandT v∈T(Hi⊥), plainlyT Hi∩T(Hi⊥) = (0), therefore H = T Hi ⊕T(Hi⊥). Hence for every y ∈ Hi, z ∈ Hi⊥ we have T∗T y ∈ Hi and therefore hT y, T zi = hT∗T y, zi = 0, so T(Hi⊥) ⊂ (T Hi)⊥ and consequently T(Hi⊥) = (T Hi)⊥ witch implies thatT Hi is orthogonally complemented.
Letx∈Hwe havex=PT Hix+y, for somey∈(T Hi)⊥, thenT∗x=T∗PT Hix+T∗y. Letv∈Hi
thenhT∗y, vi=hy, T vi= 0 thenT∗y ∈Hi⊥ and we havePHiT∗x=PHiT∗PT Hix+PHiT∗y, thenPHiT∗x=PHiT∗PT Hixthus implies that for eachi∈Iwe havePHiT∗=PHiT∗PT Hi.
Definition 2.2. Let{Hi}i∈I be a sequence of closed orthogonally complemented submodules ofH, {vi}i∈I be a familly of positive weights inA, i.e., eachvi is a positive invertible element from the center of the C∗−algebra A and Λi ∈ End∗A(H, Wi) for all i ∈ I. We say that Λ ={Hi,Λi, vi}i∈I is ag−fusion frame forH if and only if there exists two constants 0< A≤ B <∞such that
(2.1) Ahx, xi ≤X
i∈I
vi2hΛiPHix,ΛiPHixi ≤Bhx, xi, ∀x∈H.
The constantsAandBare called the lower and upper bounds ofg−fusion frame, respactively.
IfA =B then Λ is called tight g-fusion frame and ifA =B = 1 then we say Λ is a Parseval g−fusion frame. If Λ satisfies the inequality
X
i∈I
vi2hΛiPHix,ΛiPHixi ≤Bhx, xi, ∀x∈H.
then it is called ag−fusion Bessel sequence with boundB in H.
Lemma 2.3. letΛ = (Hi,Λi, vi)i∈I be ag−fusion Bessel sequence forH with boundB. Then for each sequence{xi}i∈I ∈l2({Wi}i∈I), the seriesP
i∈IviPHiΛ∗ixiis converge unconditionally.
138139
140 Fakhr-dine Nhari, Mohamed Rossafi and Youssef Aribou
Proof. letJ be a finite subset ofI, then
||X
i∈J
viPHiΛ∗ixi||= sup
||y||=1
||hX
i∈J
viPHiΛ∗ixi, yi||
≤ ||X
i∈J
hxi, xii||12 sup
||y||=1
||X
i∈J
vi2hΛiPHiy,ΛiPHiyi||12
≤√ B||X
i∈J
hxi, xii||12. And it follows thatP
j∈IvjPWjΛ∗jfj is unconditionally convergent inH. Now, we can define the synthesis operator by lemma2.3
Definition 2.4. let Λ = (Hi,Λi, vi)i∈I be ag−fusion Bessel sequence forH. Then the operator TΛ:l2({Wi}i∈I)→H defined by
TΛ({xi}i∈I) =X
i∈I
viPWiΛ∗ixi, ∀{xi}i∈I ∈l2({Wi}i∈I).
Is called synthesis operator. We say the adjoint TΛ∗ of the synthesis operator the analysis operator and it is defined byTΛ∗:H →l2({Wi}i∈I) such that
TΛ∗(x) ={viΛiPHi(x)}i∈I, ∀x∈H.
The operatorSΛ:H →H defined by SΛx=TΛTΛ∗x=X
j∈I
v2iPHiΛ∗iΛiPHi(x), ∀x∈H.
Is calledg−fusion frame operator. It can be easily verify that
(2.2) hSΛx, xi=X
i∈I
vi2hΛiPHi(x),ΛiPHi(x)i, ∀x∈H.
Furthermore, if Λ is ag−fusion frame with boundsAandB, then Ahx, xi ≤ hSΛx, xi ≤Bhx, xi, ∀x∈H.
It easy to see that the operator SΛ is bounded, self-adjoint, positive, now we proof the inversibility ofSΛ. Letx∈H we have
||TΛ∗(x)||=||{viΛiPWi(x)}i∈I||=||X
i∈I
v2ihΛiPHi(x),ΛiPHi(x)i||12. Since Λ isg−fusion frame then
√A||hx, xi||12 ≤ ||TΛ∗x||.
Then
√
A||x|| ≤ ||TΛ∗x||.
Frome lemma 1.5, TΛ is surjective and by lemma 1.6, TΛTΛ∗ = SΛ is invertible. We now, AIH≤SΛ≤BIH and this givesB−1IH ≤SΛ−1≤A−1IH
139 140
Tensor product for g-fusion frame in HilbertC∗−modules 141
Theorem 2.5. Let H be a Hilbert A−module overC∗−algebra. ThenΛ = (Hi,Λi, vi)i∈I is a g−fusion frame forH if and only if there exist two constants0< A≤B <∞such that for all x∈H
A||hx, xi|| ≤ ||X
i∈I
v2ihΛiPHix,ΛiPHixi|| ≤B||hx, xi||.
Proof. Suppose Λ isg−fusion frame forH, then for allx∈H, A||hx, xi|| ≤ ||X
i∈I
vi2hΛiPHix,ΛiPHixi|| ≤B||hx, xi||
Conversely, for eachx∈H we have
||X
i∈I
vi2hΛiPHix,ΛiPHixi||=||X
i∈I
hviΛiPHix, viΛiPHixi||
=||h{viΛiPHix}i∈I,{viΛiPHix}i∈Ii||
=||{viΛiPHix}i∈I||2.
We define the operatorL:H → ⊕i∈IWi byL(x) ={viΛiPHix}i∈I, then
||L(x)||2=||(viΛiPHix)i∈I||2≤B||x||2. LisA−linear bounded operator, then there existC >0 sutch that
hL(x), L(x)i ≤Chx, xi, ∀x∈ H.
So
X
j∈I
vi2hΛiPHix,ΛiPHixi ≤Chx, xi, ∀x∈H.
Therefore Λ is a g−fusion Bessel sequence for H. Now we cant define the g−fusion frame operatorSΛ onH. So
X
j∈J
v2jhΛjPWjx,ΛjPWjxi=hSΛx, xi, ∀x∈H.
SinceSΛ is positive, self-adjoint, then hSΛ12x, S
1 2
Λxi=hSΛx, xi, ∀x∈H.
That implies
A||hx, xi|| ≤ ||hSΛ12x, SΛ12xi|| ≤B||hx, xi||, ∀x∈H.
Frome lemma1.7there exist two canstantsA0, B0 >0 such that A0hx, xi ≤ hSΛ12x, S
1 2
Λxi ≤B0hx, xi, ∀f ∈H.
So
A0hx, xi ≤X
i∈I
vi2hΛiPHix,ΛiPHixi ≤B0hx, xi, ∀x∈H.
Hence Λ is ag−fusion frame forH.
140141
142 Fakhr-dine Nhari, Mohamed Rossafi and Youssef Aribou
3. Tensor product ofg−fusion frame in HilbertC∗−modules
Suppose thatA,Bare unitalsC∗−algebras and we takeA ⊗ Bas the completion ofA ⊗algB with the spatial norm. A ⊗ B is the spatial tensor product of Aand B, also suppose that H is a HilbertA−module andK is a HilbertB−module. We want to defineH⊗K as a Hilbert (A ⊗ B)−module. Start by forming the algebraic tensor productH⊗algK of the vector spaces H, K (over C). This is a left module overh (A ⊗alg B) (the module action being given by (a⊗b)(x⊗y) =ax⊗by(a∈ A, b∈ B, x∈H, y∈K)). For (x1, x2 ∈H, y1, y2∈K) we define hx1⊗y1, x2⊗y2iA⊗B=hx1, x2iA⊗hy1, y2iB. We also know that forz=Pn
i=1xi⊗yiinH⊗algK we havehz, ziA⊗B=P
i,jhxi, xjiA⊗ hyi, yjiB≥0 and hz, ziA⊗B= 0 iffz= 0. This extends by linearity to an (A ⊗algB)−valued sesquilinear from onH⊗algK, which makesH⊗algK into a semi-inner-product module over the pre-C∗−algebra (A ⊗algB). The semi-inner-product on H⊗algKis actually an inner product. see [10]. ThenH⊗algKis an inner-product module over the pre-C∗−algebra (A ⊗algB), and we can perform the double completion discussed in chapter 1 of [10] to conclude that the completionH⊗K ofA ⊗algBis a Hilbert (A ⊗ B)−module. We callH⊗Kthe exterior tensor product ofH andK. WithH, Kas above, we wish to investigate the adjointable operators onH⊗K. Suppose thatS ∈End∗A(H) andT ∈End∗B(K). We define a linear operatorS⊗T onH⊗K byS⊗T(x⊗y) =Sx⊗T y (x∈H, y∈K). It is a routine verification that is S∗⊗T∗ is the adjoint ofS⊗T, so in factS⊗T ∈End∗A⊗B(H⊗K). We note that ifa∈ A+ andb∈ B+, thena⊗b∈(A ⊗ B)+. Plainly ifa, bare Hermitian elements ofAanda≤b, then for every positive elementxofB, we havea⊗x≥b⊗x.
Definition 3.1. Let{vi}i∈I,{wj}j∈J be two families of positive weights, i.e., each vi andwj are positive invertible elements of A, and Λi⊗Γj ∈ End∗A(H ⊗K, Wi⊗Vj) for each i ∈ I and j ∈J. Then the family Λ⊗Γ = {Hi⊗Kj,Λi⊗Γj, viwj}i,j is saide to be a generalized fusion frame org−fusion frame forH⊗Kwith respect to{Hi⊗Kj}i,j if there exist constants 0< A≤B <∞such that for allx⊗y∈H⊗K
Ahx⊗y, x⊗yi ≤X
i,j
vi2wj2h(Λi⊗Γj)PHi⊗Kj(x⊗y),(Λi⊗Γj)PHi⊗Kj(x⊗y)i ≤Bhx⊗y, x⊗yi
where PHi⊗Kj is the orthogonal projection of H⊗K ontoHi⊗Kj. The constants A andB are called the frame bounds of Λ⊗Γ. IfA=B then it is called a tightg−fusion frame. If the family Λ⊗Γ satisfies the inequality, for eachx⊗y∈H⊗K
X
i,j
vi2wj2h(Λi⊗Γj)PHi⊗Kj(x⊗y),(Λi⊗Γj)PHi⊗Kj(x⊗y)i ≤Bhx⊗y, x⊗yi
then it is called ag−fusion Bessel sequence inH⊗Kwith boundB.
141142
Tensor product for g-fusion frame in HilbertC∗−modules 143
Definition 3.2. Fori∈I andj∈J, define the HilbertA ⊗ B−module l2({Wi⊗Vj}) ={{xi⊗yj}:xi⊗yj∈Wi⊗Vj,kX
i,j
hxi⊗yj, xi⊗yjik<∞}
with theA ⊗ B−valued inner product h{xi⊗yj},{x0i⊗yj0}i=X
i,j
hxi⊗yj, x0i⊗yj0i
=X
i,j
hxi, x0iiWi⊗ hyj, y0jiVj
=
X
i∈I
hxi, x0iiWi
⊗
X
j∈J
hyj, yj0iVj
=h{xi}i∈I,{x0i}i∈Iil2({Wi}i∈I)⊗ h{yj}j∈J,{yj}j∈Jil2({Vj}j∈J). Theorem 3.3. The familiesΛ ={Hi,Λi, vi}i∈I andΓ ={Kj,Γj, wj}j∈J areg−fusion frames forH andK with respect to{Wi}i∈I and{Vj}j∈J, respectively if and only if the familyΛ⊗Γ = {Hi⊗Kj,Λi⊗Γj, viwj}i,j is ag−fusion frame for H⊗K with respect{Wi⊗Vj}i,j.
Proof. Suppose that Λ and Γ are g−fusion frame for H and K. Then there exist positive constants (A, B) and (C, D) such that
Akxk2≤ kX
i∈I
v2ihΛiPHix,ΛiPHixiAk ≤Bkxk2,∀x∈H,
Ckyk2≤ kX
j∈J
w2jhΓjPKjy,ΓjPKjyiBk ≤Dkyk2,∀y∈K, then,
ACkxk2kyk2≤ kX
i∈I
v2ihΛiPHix,ΛiPHixiAkkX
j∈J
wj2hΓjPKjy,ΓjPKjyiBk ≤BDkxk2kyk2,
hence,
ACkx⊗yk2≤ kX
i∈I
vi2hΛiPHix,ΛiPHixiA⊗X
j∈J
w2jhΓjPKjy,ΓjPKjyiBk ≤BDkx⊗yk2,
so,
ACkx⊗yk2≤ kX
i,j
v2iw2jhΛiPHix⊗ΓjPKjy,ΛiPHix⊗ΓjPKjyik ≤BDkx⊗yk2.
Therefore, for eachx⊗y∈H⊗K ACkx⊗yk2≤ kX
i,j
vi2wj2h(Λi⊗Γj)PHi⊗Kj(x⊗y),(Λi⊗Γj)PHi⊗Kj(x⊗y)ik ≤BDkx⊗yk2, 142143
144 Fakhr-dine Nhari, Mohamed Rossafi and Youssef Aribou
we conclude that, Λ⊗Γ is a g−fusion frame forH ⊗K. Conversely, suppose that Λ⊗Γ is a g−fusion frame for H ⊗K, then there exist constants A, B > 0 such that for all x⊗y ∈ H⊗K− {0⊗0},
Akx⊗yk2≤ kX
i,j
v2iw2jh(Λi⊗Γj)PHi⊗Kj(x⊗y),(Λi⊗Γj)PHi⊗Kj(x⊗y)ik ≤Bkx⊗yk2,
then,
Akx⊗yk2≤ kX
i∈I
v2ihΛiPHix,ΛiPHixiA⊗X
j∈J
w2jhΓjPKjy,ΓjPKjyiBk ≤Bkx⊗yk2,
hence,
Akxk2kyk2≤ kX
i∈I
vi2hΛiPHix,ΛiPHixiAkkX
j∈J
w2jhΓjPKjy,ΓjPKjyiBk ≤Bkxk2kyk2,
So,
Akyk2 kP
j∈Jw2jhΓjPKjy,ΓjPKjyiBkkxk2≤ kX
i∈I
v2ihΛiPHix,ΛiPHixiAk,
and
kX
i∈I
v2ihΛiPHix,ΛiPHixiAk ≤ Bkyk2 kP
j∈Jw2jhΓjPKjy,ΓjPKjyiBkkxk2.
Therefore, Λ is ag−fusion frame forH. Similarly, it can be schown that Γ is ag−fusion frame
forK.
Definition 3.4. Let Λ⊗Γ ={Hi⊗Kj,Λi⊗Γj, viwj}i,j be ag−fusion frame forH⊗K. The synthesis operatorTΛΓ:l2({Wi⊗Vj})→H⊗K is given by
TΛΓ({xi⊗yj}) =X
i,j
viwjPHi⊗Kj(Λi⊗Γj)∗(xi⊗yj), ∀{xi⊗yj} ∈l2({Wi⊗Vj}).
And the frame operatorSΛ⊗Γ :H⊗K→H⊗Kis described by
SΛ⊗Γ(x⊗y) =X
i,j
(viwj)2PHi⊗Kj(Λi⊗Γj)∗(Λi⊗Γj)PHi⊗Kj(x⊗y), ∀x⊗y∈H⊗K.
Theorem 3.5. If SΛ, SΓ and SΛ⊗Γ are the associated g−fusion frame operators and TΛ, TΓ
andTΛ⊗Γ are the synthesis operators ofg−fusion framesΛ,ΓandΛ⊗ΓforH,KandH⊗K, respectively, thenSΛ⊗Γ=SΛ⊗SΓ,SΛ⊗Γ−1 =S−1Λ ⊗SΓ−1, and ,TΛ⊗Γ=TΛ⊗TΓ,TΛ⊗Γ∗ =TΛ∗⊗TΓ∗. 143144
Tensor product for g-fusion frame in HilbertC∗−modules 145
Proof. Let eachx⊗y∈H⊗K, we have
SΛ⊗Γ(x⊗y) =X
i,j
vi2wj2PHi⊗Kj(Λi⊗Γj)∗(Λi⊗Γj)PHi⊗Kj(x⊗y)
=X
i,j
vi2wj2(PHi⊗Kj)(Λi⊗Γj)∗(Λi⊗Γj)(PHi⊗PKj)(x⊗y)
=X
i,j
vi2wj2(PHiΛ∗iΛiPHix)⊗(PKjΓ∗jΓjPKjy)
= X
i∈I
v2iPHiΛ∗iΛiPHix
⊗ X
j∈J
wj2PKjΓ∗jΓjPKjy
=SΛx⊗SΓy
= (SΛ⊗SΓ)(x⊗y).
Then,SΛ⊗Γ=SΛ⊗SΓ, so SΛ⊗Γ−1 =SΛ−1⊗SΓ−1.
On the other hand, for each{xi⊗yj} ∈l2({Wi⊗Vj}), we have
TΛ⊗Γ({xi⊗yj}) =X
i,j
viwjPHi⊗Kj(Λi⊗Γj)∗(xi⊗yj)
=
X
i∈I
viPHiΛ∗ixi
⊗
X
j∈J
wjPKjΓ∗jyj
=TΛ({xi})⊗TΓ({yj})
= (TΛ⊗TΓ)({xi⊗yj}).
this shows thatTΛ⊗Γ =TΛ⊗TΓ, henceTΛ⊗Γ∗ =TΛ∗⊗TΓ∗.
Theorem 3.6. LetΛ ={Hi,Λi, vi}i∈I andΓ ={Kj,Γj, wj}j∈J beg−fusion frames forH and K with g−fusion frame operators SΛ and SΓ, respectively. Then θ ={SΛ⊗Γ−1 (Hi⊗Kj),(Λi⊗ Γj)PHi⊗KjSΛ⊗Γ−1 , viwj}i,j is ag−fusion frame for H⊗K.
145144
146 Fakhr-dine Nhari, Mohamed Rossafi and Youssef Aribou
Proof. Let (A, B) and (C, D) be theg−fusion frame bounds of Λ and Γ, respectively. Now, we have for eachx⊗y∈H⊗K,
kX
i,j
vi2w2jh(Λi⊗Γj)PHi⊗KjSΛ⊗Γ−1 PS−1
Λ⊗Γ(Hi⊗Kj)(x⊗y),(Λi⊗Γj)PHi⊗KjS−1Λ⊗ΓPS−1
Λ⊗Γ(Hi⊗Kj)(x⊗y)ik
=kX
i,j
vi2w2jhΛi⊗ΓjPHi⊗KjSΛ−1⊗SΓ−1PS−1
Λ Hi⊗PS−1
Γ Kj(x⊗y), Λi⊗ΓjPHi⊗KjSΛ−1⊗SΓ−1PS−1
Λ Hi⊗PS−1
Γ Kj(x⊗y)ik
=kX
i,j
vi2w2jhΛi⊗ΓjPHi⊗PKjSΛ−1PS−1
Λ Hi⊗SΓ−1PS−1
Γ Kj(x⊗y), Λi⊗ΓjPHi⊗PKjSΛ−1PS−1
Λ Hi⊗SΓ−1PS−1
Γ Kj(x⊗y)ik
=kX
i,j
vi2w2jhΛiPHiSΛ−1PS−1
Λ Hix⊗ΓjPKjSΓ−1PS−1
Γ Kjy,ΛiPHiSΛ−1PS−1
Λ Hix⊗ΓjPKjSΓ−1PS−1 Γ Kjyik
=kX
i,j
vi2w2jhΛiPHiSΛ−1x⊗ΓjPKjSΓ−1y,ΛiPHiS−1Λ x⊗ΓjPKjSΓ−1yik
=kX
i,j
vi2w2jhΛiPHiSΛ−1x,ΛiPHiSΛ−1xiA⊗ hΓjPKjSΓ−1y,ΓjPKjSΓ−1yiBk
=kX
i∈I
vi2hΛiPHiSΛ−1x,ΛiPHiSΛ−1xiAkkX
i∈J
w2jhΓjPKjSΓ−1y,ΓjPKjS−1Γ yiBk
≤BkSΛ−1xk2DkSΓ−1yk2
≤ BD
(AC)2kx⊗yk2.
On the other hand for eachx⊗y∈H⊗K,
kx⊗yk4=khx, xiAk2khy, yiBk2
=khX
i∈I
vi2PHiΛ∗iΛiPHiSΛ−1x, xiAk2khX
i∈J
wj2PKjΓ∗jΓjPKjSΓ−1y, yiBk2
=kX
i∈I
vi2hΛiPHiS−1Λ x,ΛiPHixiAk2kX
j∈J
wj2hΓjPKjS−1y,ΓjPKjyiBk2
≤ kX
i∈I
vi2hΛiPHiS−1Λ x,ΛiPHiSΛ−1xiAkkX
i∈I
vi2hΛiPHix,ΛiPHixiAk kX
j∈J
wj2hΓjPKjSΓ−1y,ΓjPKjSΓ−1yiBkkX
j∈J
w2jhΓjPKjy,ΓjPKjyiBk
≤BDkxk2kyk2kX
i∈I
v2ihΛiPHiSΛ−1PS−1
Λ Hix,ΛiPHiSΛ−1PS−1
Λ HixiAk kX
j∈J
wj2hΓjPKjSΓ−1PS−1
Γ Kjy,ΓjPKjSΓ−1PS−1
Γ KjyiBk
=BDkx⊗yk2kX
i,j
(viwj)2h(Λi⊗Γj)PHi⊗KjSΛ⊗Γ−1 PS−1
Λ⊗Γ(Hi⊗Kj)(x⊗y)k,
145146
Tensor product for g-fusion frame in HilbertC∗−modules 147
hence, 1
BDkx⊗yk2≤ kX
i,j
(viwj)2h(Λi⊗Γj)PHi⊗KjSΛ⊗Γ−1 PS−1
Λ⊗Γ(Hi⊗Kj)(x⊗y)k.
Therefore,θis a g−fusion frame forH⊗K.
Proposition 3.7. For the g−fusion frameθ, frame operator isSΛ⊗Γ−1 . Proof. We putG= (Λi⊗Γj)PHi⊗KjSΛ⊗Γ−1 , then
G∗G= ((Λi⊗Γj)PHi⊗KjSΛ⊗Γ−1 )∗(Λi⊗Γj)PHi⊗KjSΛ⊗Γ−1
= (SΛ−1⊗SΓ−1)(PHi⊗PKj)(Λ∗i ⊗Γ∗j)(Λi⊗Γj)(PHi⊗PKj)(SΛ−1⊗SΓ−1)
=S−1Λ PHiΛ∗iΛiPHiSΛ−1⊗SΓ−1PKjΓ∗jΓjPKjSΓ−1 hence,
PS−1
Λ⊗Γ(Hi⊗Kj)G∗GPS−1
Λ⊗Γ(Hi⊗Kj)
= PS−1
Λ Hi⊗PS−1 Γ Kj
G∗G PS−1
Λ Hi⊗PS−1 Γ Kj
=PS−1
Λ HiSΛ−1PHiΛ∗iΛiPHiSΛ−1PS−1
Λ Hi⊗PS−1
Γ KjSΓ−1PKjΓ∗jΓjPKjSΓ−1PS−1
Γ Kj
= (PHiSΛ−1)∗Λ∗iΛiPHiSΛ−1⊗(PKjSΓ−1)∗Γ∗jΓjPKjSΓ−1
=SΛ−1PHiΛ∗iΛiPHiSΛ−1⊗SΓ−1PKjΓ∗jΓjPKjSΓ−1. So, for eachx⊗y∈H⊗K, we have
X
i,j
v2iw2jPS−1
Λ⊗Γ(Hi⊗Kj)G∗GPS−1
Λ⊗Γ(Hi⊗Kj)(x⊗y)
=X
i,j
vi2w2j SΛ−1PHiΛ∗iΛiPHiS−1Λ ⊗SΓ−1PKjΓ∗jΓjPKjSΓ−1 (x⊗y)
=
X
i∈I
v2iSΛ−1PHiΛ∗iΛiPHiSΛ−1x
⊗
X
j∈J
wj2SΓ−1PKjΓ∗jΓjPKjSΓ−1y
=SΛ−1SΛ(SΛ−1x)⊗S−1Γ SΓ(SΓ−1y)
=SΛ−1x⊗SΓ−1y
= (SΛ−1⊗SΓ−1)(x⊗y)
=SΛ⊗Γ−1 (x⊗y),
we conclude that the correspondingg−fusion frame forθ isSΛ⊗Γ−1 . Theorem 3.8. LetΛ ={Hi,Λi, vi}i∈I,Λ0 ={Hi0,Λ0i, v0i}i∈I beg−fusion Bessel sequences with bounds B, D, respectively in H and Γ = {Kj,Γj, wj}j∈J, Γ0 = {Kj0,Γ0j, w0j}j∈J be g−fusion Bessel sequence with boundsE, F, respectively inK. Suppose(TΛ, TΛ0)and(TΓ, TΓ0)are their
146147
148 Fakhr-dine Nhari, Mohamed Rossafi and Youssef Aribou
synthesis operators such that TΛ0TΛ∗ = IH and TΓT∗
Γ0 = IK. Then Λ⊗Γ = {Hi⊗Kj,Λi⊗ Γj, viwj}i,j andΛ0⊗Γ0 ={Hi0⊗Kj0,Λ0i⊗Γ0j, v0iw0j}i,j areg−fusion frames forH⊗K.
Proof. By theorem3.3, Λ⊗Γ and Λ0⊗Γ0 areg−fusion Bessel sequences, respectively inH⊗K.
Now, for eachx⊗y∈H⊗K,
kx⊗yk4=khx, xiAk2khy, yiBk2
=khTΛ∗x, TΛ∗0xiAk2khTΓ∗y, TΓ∗0yiBk2
≤ kTΛ∗xk2kTΛ∗0xk2kTΓ∗yk2kTΓ∗0yk2
≤DFkxk2kyk2kX
i∈I
vi2hΛiPHix,ΛiPHixiAkkX
j∈J
w2jhΓjPKjy,ΓjPKjyiBk
=DFkx⊗yk2kX
i,j
v2iw2jhΛiPHix,ΛiPHixiA⊗ hΓjPKjy,ΓjPKjyiBk
=DFkx⊗yk2kX
i,j
v2iw2jh(Λi⊗Γj)PHi⊗Kj(x⊗y),(Λi⊗Γj)PHi⊗Kj(x⊗y)ik,
then, 1
DFkx⊗yk2≤ kX
i,j
vi2wj2h(Λi⊗Γj)PHi⊗Kj(x⊗y),(Λi⊗Γj)PHi⊗Kj(x⊗y)ik.
Therefore, Λ⊗Γ is ag−fusion frame forH⊗K. Similarly, it can be shown that Λ0⊗Γ0 is also
ag−fusion frame forH⊗K.
4. Frame operator for a pair ofg−fusion Bessel sequences in tensor product of HilbertC∗−modules
Definition 4.1. Let Λ⊗Γ ={Hi⊗Kj,Λi⊗Γj, viwj}i,jand Λ0⊗Γ0 ={Hi0⊗Kj0,Λ0i⊗Γ0j, vi0wj0}i,j
be twog−fusion Bessel sequences in H⊗K. Then the operatorS:H⊗K→H⊗K defined by for allx⊗y∈H⊗K,
S(x⊗y) =X
i,j
viwjv0iw0jPHi⊗Kj(Λi⊗Γj)∗(Λ0i⊗Γ0j)PH0
i⊗K0j(x⊗y)
is called the frame operator for the pair ofg−fusion Bessel sequences Λ⊗Γ and Λ0⊗Γ0.
Theorem 4.2. Let SΛΛ0 and SΓΓ0 be the frame operators for the pair of g−fusion Bessel sequences
Λ ={Hi,Λi, vi}i∈I,Λ0 ={Hi0,Λ0i, v0i}j∈J
and
Γ ={Kj,Γj, wj}j∈J, Γ0 ={Kj0,Γ0j, w0j}j∈J
inH andK, respectively. ThenS=SΛΛ0 ⊗SΓΓ0.
147 148