C -Algebra B H (I) Consisting of Bessel Sequences in a Hilbert Space

Size: px

Start display at page:

Download "C -Algebra B H (I) Consisting of Bessel Sequences in a Hilbert Space"

Sabrina McDonald
6 years ago
Views:

1 Journal of Mathematical Research with Applications Mar., 2015, Vol. 35, No. 2, pp DOI: /j.issn: C -Algebra B H (I) Consisting of Bessel Sequences in a Hilbert Space Zhihua GO 1,, Maoren YIN 2, Huaixin CAO 1 1. College of Mathematics and Information Science, Shaanxi Normal niversity, Shaanxi , P. R. China; 2. Department of Mathematics, Junior College of Xinzhou Teachers niversity, Shanxi , P. R. China Abstract Let H be a separable Hilbert space, B H (I), B(H) and K(H) the sets of all Bessel sequences {f i} i I in H, bounded linear operators on H and compact operators on H, respectively. Two kinds of multiplications and involutions are introduced in light of two isometric linear isomorphisms α H : B H(I) B(l 2 ), β : B H(I) B(H), respectively, so that B H (I) becomes a unital C -algebra under each kind of multiplication and involution. It is proved that the two C -algebras (B H (I),, ) and (B H (I),, ) are -isomorphic. It is also proved that the set F H(I) of all frames for H is a unital multiplicative semi-group and the set R H (I) of all Riesz bases for H is a self-adjoint multiplicative group, as well as the set K H (I) := β 1 (K(H)) is the unique proper closed self-adjoint ideal of the C -algebra B H (I). Keywords C -algebra; Bessel sequence; Hilbert space; frame; Riesz basis MR(2010) Subject Classification 39B52 1. Introduction Frames in Hilbert spaces were firstly introduced by Duffin and Schaeffer [1] in the study of nonharmonic Fourier series in Recently, frame theory plays an important role in mathematics, science, and engineering [2 4], and various generalizations of frames have been obtained. For example, Sun in [5,6] introduced and discussed the concept of a G-frame for a Hilbert space, which generalizes the concepts of frames [7], pseudoframes [8], oblique frames [9,10], outer frames [11], bounded quasi-projectors [12,13], frames of subspaces [14,15], operator frames for B(H) (see [16]). In [17], (p, Y )-operator frames for a Banach space X were introduced, which makes a G- frame {T j } j Λ for a Hilbert space H with respect to a sequence {K j } j Λ of closed subspaces of a Hilbert space K be a (2, K)-operator frame for H. Hence, the concept of a (p, Y )-operator frame for a Banach space generalizes all of the concepts of frames. In Banach space setting, X d frames, X d Bessel sequences, tight X d frames, independent X d frames, and X d Riesz basis for a Banach Received December 12, 2013; Accepted November 22, 2014 Supported by the National Natural Science Foundation of China (Grant Nos ; ; ), China Postdoctoral Science Foundation (Grant No. 2014M552405) and the Natural Science Research Program of Shaanxi Province (Grant No. 2014JQ1010). * Corresponding author address: guozhihua@snnu.edu.cn (Zhihua GO); yinmaoren@126.com (Maoren YIN); caohx@snnu.edu.cn (Huaixin CAO)

2 192 Zhihua GO, Maoren YIN and Huaixin CAO space were introduced and discussed, a necessary and sufficient condition for a Banach space X to have an X d frame or to have an X d Riesz basis as well as a necessary and sufficient condition for an X d frame to have a dual frame were obtained in paper [18], some relations among basis, X d frame and X d Riesz basis in a Banach space were also established there. In paper [19], the properties of frames and atomic decompositions for a Banach space were explored in terms of the theory of frames for a Hilbert space, a sufficient condition for a complete Bessel sequence in a Banach space X to be a Banach frame was given and a sufficient condition for a complete Bessel sequence {y n } to have a sequence {x n } such that ({y n }, {x n }) becomes an atomic decomposition for X was also established and a relation between Banach frames and atomic decompositions was discussed there. Multivariate Riesz multiwavelet bases with short support in (L 2 (R s )) r 1 have applications in many areas, such as image processing, computer graphics and numerical algorithms. Pan in [20] characterized an algorithm to derive Riesz bases from refinable function vectors and several other important results about Riesz wavelet bases in (L 2 (R s )) r 1 were also given there. Traditionally, Gabor and wavelet analysis were studied by using classical Fourier analysis methods. But in recent years, more and more abstract tools have been introduced such as operator theory, operator algebra, abstract harmonic analysis and group-representations, etc. Especially, the author in [7] proved that the set B H (I) of all Bessel sequences {f i } i I in a Hilbert space H is a Banach space and established an isometric isomorphism α from B H (I) onto the operator space B(H, l 2 ) defined by α({f i } i I )(x) = { x, f i } i I, which suggests a relationship between wavelet analysis and operator theory. The aim of this paper is to introduce multiplication and involution on the Banach space B H (I) so that it becomes a unital C -algebra. This enables us to establish a corresponding connection between wavelet analysis and operator algebra. Now, let us recall some definitions. Throughout this paper, H denotes a complex Hilbert space, I is an index set so that dim(h) = I (the cardinality of I), and B(H) is the C -algebra of all bounded linear operators on H, F denotes either R or C, l 2 l 2 (I) is the Hilbert space of all square-summable complex sequences with the inner product defined by {c n }, {d n } = c n d n, which has the canonical orthonormal basis e = {e n }, where e n = {δ n,k } k I. Moreover, S H (I) = { {f n } : f n H( n I) }, and denote by O H (I) the set of all orthonormal bases for H. Definition 1.1 ([7]) Let f = {f n } S H (I). If there exists a positive constant B such that x, f n 2 B x 2, x H, (1.1) then we call f is a Bessel sequence in H. Denote by B H (I) the set of all Bessel sequences in H.

3 C -algebra B H(I) consisting of Bessel sequences in a Hilbert space 193 For every f, g B H (I), define αf + βg = {αf n + βg n }, α, β F; ( f 2 = sup x, f n 2). x 1 Then (B H (I), ) becomes a normed linear space over F and f = inf{b : B satisfies (1.1)}, f B H (I). For every f B H (I), put T f : H l 2, T f x = { x, f n }, x H. It is evident that T f B(H, l 2 ) and T f = f for all f in B H (I). Thus, the map f T f induces an isometric homomorphism α from B H (I) into B(H, l 2 ). In the sequel, we shall need the following. Theorem 1.2 (1) T f {c n} = c nf n, f B H (I), {c n } l 2 (I); T f A. (2) α : f T f is an isometrically linear isomorphism; (3) (B H (I), ) is a Banach space; (4) If A B(H) and f = {f n } B H (I), then Af := {Af n } B H (I) and T Af = Proof Similar to the proof of Corollary 2 in [7]. Definition 1.3 ([7]) A Bessel sequence f = {f n } in H is called a frame if there exists a positive constant A such that for every x H, A x 2 x, f n 2. (1.2) Denote by F H (I) the set of all frames for H. If conditions (1.1) and (1.2) hold for every x H and the bounds A, B coincide, then the frame is called tight, f is called a Parseval frame if A = B = 1. Definition 1.4 ([7]) Let f = {f n } S H (I). If there exist two positive constants C, D such that C {c n } 2 c n f n 2 D {c n } 2, {c n } l 2, (1.3) and span{f n n I} = H, then f is said to be a Riesz basis for H. Denote by R H (I) the set of all Riesz bases for H. It is well-known that a frame has the unique dual if and only if it is a Riesz basis. Theorem 1.5 Let f = {f n } S H (I). Then (1) f B H (I) T f B(H, l 2 ); (2) f F H (I) T f is below-bounded T f T f is invertible; (3) f R H (I) T f is invertible;

4 194 Zhihua GO, Maoren YIN and Huaixin CAO (4) f O H (I) T f is unitary. Proof Similar to the proof of Theorem 2 in [7]. Definition 1.6 If f = {f n }, g = {g n } are two frames for H and x = x, f n g n = g n f n, x H, x, then we say that g is a dual frame of f. In this case, f is also a dual frame of g, so {f, g} is called a dual pair of frames. For every frame f = {f i } i I for H, Theorem 1.2(2) implies that the operator Tf T f is invertible. Thus, f := {(Tf T f ) 1 f i } i I is a dual frame of f, called the canonical dual of f. 2. Main results In this section, we will define two kinds of multiplications and involutions of Bessel sequences in a Hilbert space H by using two maps α H : B H (I) B(l 2 ) and β : B H (I) B(H), respectively, and then obtain C -algebra structures on the Banach space B H (I). To define a multiplication on the Banach space B H (I), we first assume that H = l 2 (I). Let e = {e n } be the canonical orthonormal basis for l 2 (I). Then every element c of l 2 (I) can be written as c = { c, e n }. Definition 2.1 For every f = {f n }, g = {g n } B l 2(I), we define f g = {(f g) n } = Tg f = {Tg f n }, (2.1) f = {(f ) n } = T f e = {T f e n }. (2.2) Clearly, f g and f are both elements of B l 2(I). By Theorem 1.2(4), we get that T f g = T T g f = T f T g and T f = T Tf e = T e Tf = T f. Theorem 2.2 B l 2(I) is a unital C -algebra with identity e = {e n } and the mapping α l 2 : B l 2(I) B(l 2 ) is an isometrically -algebraic isomorphism. Proof (i) For all f, g, h B l 2(I), we compute that f (g h) = Tg hf = (T g T h ) f = Th Tg f = (f g) h, (λf) g = Tg (λf) = Tλgf = f (λg) for all λ F and e f = T f e = T f e = f = T e f = f e; f (g + h) = T g+hf = (T g + T h ) f = (T g + T h )f = f g + f h. Similarly, (f + g) h = f h + g h. For all f, g B l 2(I), we have f g = T f g = T f T g T f T g = f g.

5 C -algebra B H(I) consisting of Bessel sequences in a Hilbert space 195 So, B l 2(I) is a Banach algebra. (ii) For all f, g B l 2(I) and a, b F, we have (af + bg) = {T af+bg e n } = {at f e n } + {bt g e n } = af + bg, (f g) = {T f g e n } = {T f T g e n } = T f g = g f, (f ) = {T f e n } = {T f e n} = f. (iii) f B H (I), f f = T f T f = T f 2 = f 2. Hence, B l 2(I) is a unital C -algebra with identity e = {e n }. Since α l 2(f ) = T f = Tf = (α l 2(f)), α l 2(f g) = T f g = T f T g = α l 2(f)α l 2(g), and α l 2 : B l 2(I) B(l 2 ) is an isometrically linear isomorphism, α l 2 : B l 2(I) B(l 2 ) is an isometrically -algebraic isomorphism. It is well-known that every separable infinite-dimensional Hilbert space H with a basis {ε i } i I is isomorphic to l 2 (I) in light of the unitary : H l 2 defined by ( i I c iε i ) = {c i } i I. Thus, f = {f n }, g = {g n } B H (I), we have f := {f n } B l 2(I), g := {g n } B l 2(I). By Definition 2.1, we know that f g B l 2(I). Put π (f) = f = {f n } for every f B l 2(I). Then π (f) = T π (f) = T f = T f T f 1 2 = Tf T f 1 2 = Tf = f, and so we obtain an isometrically linear isomorphism π : B H (I) B l 2(I) since is a unitary. Definition 2.3 For all f, g B H (I), we define f g = π 1 ((f) (g)), f = π 1 ((f) ). (2.3) It is clear that f g B H (I) and f B H (I). Theorem 2.4 B H (I) is a unital C -algebra with identity e H = π 1 (e) and the mapping α H α l 2π : B H (I) B(l 2 ) is an isometrically -algebraic isomorphism. Proof Clearly, the Banach space B H (I) becomes a Banach algebra with the multiplication defined by (2.3). For all f, g B H (I) and a, b F, we compute that (i) (af + bg) = π 1 (((af + bg)) ) = π 1 (ii) (f g) = π 1 (((f g)) ) = π 1 (a(f) + b(g) ) = af + bg ; ((g) (f) )) = g f ; (iii) (f ) = π 1 ((f ) ) = π 1 ({T f e n}) = π 1 ({T f e n }) = f. Thus, B H (I) becomes a Banach -algebra with identity e H. Moreover, f B H (I), we have f f = T f T f = T f 2 = f 2. This shows that B H (I) is a unital C -algebra with identity e H. Since π (f g) = (f) (g) = π (f) π (g), π (f ) = (f) = (π (f)),

6 196 Zhihua GO, Maoren YIN and Huaixin CAO we get that π is an isometrically -algebraic isomorphism. It follows from Theorem 2.2 that α H is an isometrically -algebraic isomorphism. Remark 2.5 For f, g B H (I), the equation f g = g f does not necessarily hold, so C -algebra B H (I) is not abelian in general. For every f B H (I), define S f = α H (f). (2.4) By the definition of α H, it is evident that S f = T f = T f B(l 2 ) and S f = f. For f, g B H (I), we have S f g = T (f g) = T (f) (g) = T f T g = T f T g = S f S g ; S f = T f = T (f) = Tf = (T f ) = Sf. From Theorem 1.5, we can get following results: Corollary 2.6 Let f B H (I). Then (1) f F H (I) if and only if S f is below-bounded; (2) f R H (I) if and only if S f is invertible; (3) f O H (I) if and only if S f is unitary. Corollary 2.7 Let f, g B H (I). Then A(f g) = f (Ag) whenever A B(H). Proof se the fact that A(f g) = ATg f = (T g A ) f = TAg f = f (Ag). The map α H defined in (2.4) depends on the isomorphism from H onto l 2 and builds an isomorphism between B H (I) and B(l 2 ). There is another way to obtain a new kind of multiplication of two Bessel sequences and a new involution of a Bessel sequence in light of a fixed orthonormal basis for H. Take an orthonormal basis δ = {δ n } for H, then for arbitrary f B H (I) and x H, put R f x = f n δ n. (2.5) x, Clearly, x, f n δ n = Tδ T f x H and so R f = Tδ T f B(H) and R f = T f = f. Now, we define a map β : B H (I) B(H), f R f. It is clear that β is injective. For every A B(H), take f = α 1 (T δ A), then β(f) = R f = Tδ T f = A, so β is surjective. Thus, β is an isometrically linear bijection. Definition 2.8 B H (I) and f, g B H (I), define f g = β 1 (R f R g ), f = β 1 (R f ). Clearly, f g, f β(f g) = R f R g = β(f)β(g), β(f ) = R f = β(f). Theorem 2.9 B H (I) is a unital C -algebra with identity δ and the mapping β : B H (I) B(H) is an isometrically -algebraic isomorphism. Proof Similar to the proof of Theorem 2.4.

7 C -algebra B H(I) consisting of Bessel sequences in a Hilbert space 197 From R f = T δ T f and Theorem 1.5, we can get following result: Corollary 2.10 Let f B H (I). Then (1) f F H (I) if and only if R f is below-bounded; (2) f R H (I) if and only if R f is invertible; (3) f O H (I) if and only if R f is unitary. Even though the maps from B H (I) into B(l 2 ) and B(H), respectively, are different, we obtain the same conclusion, which says that (B H (I),, ) and (B H (I),, ) are unital C -algebras with two different kinds of multiplications and involutions. The following theorem shows that these two C -algebras are isomorphic. Theorem 2.11 Set ψ (A) = A, then the mapping ϕ = β 1 ψ 1 α H is an isometric - isomorphism from (B H (I),, ) onto (B H (I),, ), and C -algebras (B H (I),, ) and (B H (I),, ) are isomorphic. Moreover, ϕ(f H (I)) = F H (I), ϕ(r H (I)) = R H (I). Proof Clearly, the mapping ψ : B(H) B(l 2 ) is an isometric -isomorphism and then ϕ = β 1 ψ 1 α H is an isometric -isomorphism from (B H (I),, ) onto (B H (I),, ). By the fact that ϕ(f) = β 1 ψ 1 α H(f) = β 1 ( S f ) = β 1 ( T f ) = β 1 ( T δ R f ), and Corollary 2.10(1), we get that ϕ(f) F H (I) β(ϕ(f)) is below bounded f F H (I). So, ϕ(f H (I)) = F H (I). Similarly, ϕ(r H (I)) = R H (I). 3. Some applications In this section, we will study some properties of some subsets of B H (I). Theorem 3.1 (1) F H (I) is a semi-group with identity δ with respect to multiplication ; (2) R H (I) is a self-adjoint multiplicative group. Proof (1) Clearly δ F H (I). If f, g F H (I), then R f R g is still a bounded-below operator, by Corollary 2.10, f g F H (I). This implies F H (I) is a multiplicative semi-group with identity δ. (2) It is clear that R H (I) is a semi-group with identity δ. The following discussion will prove that every f R H (I) has a unique inverse in R H (I). Let f R H (I). Then R f is invertible. Thus, there is a unique g B H (I) such that β(g) = R 1 f, that is, R g = R 1 f. Since R f g = R f R g = I = R δ and R g f = R g R f = R δ, f g = g f = δ. Hence, g is the inverse of f. Clearly, g R H (I). For arbitrary frame f for H, define f = {g g is a dual frame of f}. Clearly, f f for every frame f for H. Theorem 3.2 (1) If f F H (I), then g f = δ for all g f. (2) Let f F H (I), g B H (I). If g f = δ, then g F H (I) and g f.

8 198 Zhihua GO, Maoren YIN and Huaixin CAO (3) If f F H (I), then f F H (I) if and only if f R H (I); (4) If f F H (I), then f is idempotent if and only if f is the identity δ of B H (I). Proof (1) If f F H (I) and g f, then x = x, f n g n = g n f n, x H, x, by the definitions of T f and T g, we have x = T g T f x = T f T gx. So, T g T f = T f T g = I and thus we have T g T f = T g f = I, so g f = δ. (2) If g f = δ, then T f T g = R f R g = R gr f = β(g)β(f) = β(g f) = I. So we can see that T g is left-invertible. This implies that g F H (I) and g f. (3) Let f F H (I). Suppose that f F H (I). By Corollary 2.10, we have R f, R f are both bounded-below. So R f is invertible and thus f R H (I). Conversely, we suppose that f R H (I). Then R f is invertible, so Rf is also invertible, therefore f R H (I) F H (I). (4) f is idempotent if and only if R f is idempotent. Since f F H (I) implies that R f : l 2 l 2 is below-bounded, now we only need prove that an idempotent injective operator R f is the identity I. Suppose that ran(r f ) {0}, where ran(r f ) denotes the range of R f, then there exists x ran(r f ) \{0} such that R f x x. However, R f (R f x x) = 0. This contradicts the fact that R f is injective. So, R f = I. Hence, f is idempotent if and only if f is the identity δ of B H (I). Remark 3.3 In F H (I), the dual frame of f is not unique, so its left-inverse element is also not unique. But the convex combination of left-inverse elements of f is still a left-inverse element of f. In the following, we consider ideals of B H (I) as a C -algebra and obtain an important result. Denote by K(H) the set of all compact operators on H, and denote K H (I) := β 1 (K(H)). c 0 (H) = { {f n } : f n 0 (n ) } Theorem 3.4 c 0 (H) B H (I). K H (I) is the unique proper self-adjoint closed ideal of B H (I) and K H (I) Proof Since K(H) is the unique non-trivial closed ideal of C -algebra B(H) and β is an isometrically -isomorphism, K H (I) is the unique proper closed ideal of B H (I). Since T K(H) if and only if T K(H), we see that K H (I) is self-adjoint. Also, f K H (I), we have R f K(H). Take the identity δ of B H (I), then f n = Rf δ n, so we compute that f n = Rf δ n 0(n ). So, K H (I) c 0 (H). Clearly, K H (I) B H (I). This shows that K H (I) c 0 (H) B H (I). Remark 3.5 The of Theorem 3.4 may be a proper inclusion in general. For instance, let

9 C -algebra B H(I) consisting of Bessel sequences in a Hilbert space 199 {e n } be an orthonormal basis for H and define the sequence {f n } by { e1, e 2, e 2, e 3, e 3, e 3,... } It is easily seen that {f n } is a Parseval frame for H and {f n } c 0 (H). But R f is a co-isometry, i.e., R f R f = I. Hence, R f is not a compact operator. Applying properties of an ideal, we also derive following results. Corollary 3.6 (1) K H (I) F H (I) = Ø; (2) If f B H (I) and A K(H), then Af K H (I). Proof (1) Suppose that K H (I) F H (I) Ø, that is, at least there exists f K H (I) F H (I), then we have R f K(H) and furthermore R f K(H). On the other hand, for f F H(I), take g f, then by Definition 1.6 and Theorem 1.5, we have R f R g = T f T δt δ T g = I. Since K(H) is the unique ideal of B(H), I K(H). This is a contradiction. Therefore, K H (I) F H (I) = Ø. (2) Since R Af = T Af T δ = AT f T δ K(H), Af K H (I). References [1] J. DFFIN, A. C. SCHAEFFER. A class of nonharmonic Fourier series. Trans. Amer. Math. Soc., 1952, 72: [2] Deguang HAN, D. R. LARSON. Frames, bases and group representations. Mem. Amer. Math. Soc., 2000, 147(697): [3] E. J. CANDÈS. Harmonic analysis of neural networks. Appl. Comput. Harmon. Anal., 1999, 6(2): [4] R. H. CHAN, S. D. RIEMENSCHNEIDER, Lixin SHEN, et al. frame: An efficient way for high-resolution image reconstruction. Appl. Comput. Harmon. Anal., 2004, 17(1): [5] Wenchang SN. g-frames and g-riesz bases. J. Math. Anal. Appl., 2006, 322(1): [6] Wenchang SN. Stability of g-frames. J. Math. Anal. Appl., 2007, 326(2): [7] Huaixin CAO. Bessel sequences in a Hilbert space. Gongcheng Shuxue Xuebao, 2000, 17(2): [8] Shidong LI, H. OGAWA. Pseudoframes for subspaces with applications. J. Fourier Anal. Appl., 2004, 10(4): [9] O. CHRISTENSEN, Y. C. ELDAR. Oblique dual frames and shift-invariant spaces. Appl. Comput. Harmon. Anal., 2004, 17(1): [10] Y. C. ELDAR. Sampling with arbitrary sampling and reconstruction spaces and oblique dual frame vectors. J. Fourier Anal. Appl., 2003, 9(1): [11] A. ALDROBI, C. CABRELLI,. MOLTER. Wavelets on irregular grids with arbitrary dilation matrices and frame atomics for L 2 (R d ). Appl. Comput. Harmon. Anal., 2004, 17: [12] M. FORNASIER. Quasi-orthogonal decompositions of structured frames. J. Math. Anal. Appl., 2004, 289(1): [13] M. FORNASIER. Decompositions of Hilbert spaces: Local construction of global frames. Constructive theory of functions, , DARBA, Sofia, [14] M. S. ASGARI, A. KHOSRAVI. Frames and bases of subspaces in Hilbert spaces. Math. Anal. Appl., 2005, 308(2): [15] P. G. CASAZZA, G. KTYNIOK. Frame of Subspaces. Amer. Math. Soc., Providence, RI, [16] Chunyan LI, Huaixin CAO. Operator Frames for B(H). Birkhäuser, Basel, [17] Huaixin CAO, Lan LI, Qingjiang CHEN, et al. (p, Y )-Operator frames for a Banach space. J. Math. Anal. Appl., 2008, 347(2): [18] Chunyan LI, Huaixin CAO. X d Frames and Riesz bases for a Banach space frames and Riesz bases for a Banach space. Acta Math. Sinica (Chin. Ser.), 2006, 49(6): [19] Jiayun ZHO, Yu LI. The property of frame and atomic decomposition for Banach space. Acta Math. Sinica (Chin. Ser.), 2004, 47(3): [20] Yali PAN. Multivariate Riesz multiwavelet bases. Acta Math. Sinica (Chin. Ser.), 2010, 53(6):

arxiv:math/ v1 [math.fa] 5 Aug 2005

arxiv:math/ v1 [math.fa] 5 Aug 2005 arxiv:math/0508104v1 [math.fa] 5 Aug 2005 G-frames and G-Riesz Bases Wenchang Sun Department of Mathematics and LPMC, Nankai University, Tianjin 300071, China Email: sunwch@nankai.edu.cn June 28, 2005