6 views

Uploaded by Trinh Chien

- Quantum Meditation: The Self-Spirit Projection
- Hilbert
- Spatial Statistics
- 1363
- 5
- 14-20.txt
- A.I.Bobenko, D.Matthes and Yu.B. Suris- Discrete and smooth orthogonal systems: C∞-approximation
- 2008e
- ALAp 2
- QFT.pdf
- Matrices
- [2004!10!18 HeeJoo]ImageCompressionUsingtheHaarWaveletTransform 2
- Java Printing
- Soleymani
- The Subtle Art of Not Giving a F*ck: A Counterintuitive Approach to Living a Good Life
- Sapiens: A Brief History of Humankind
- Hidden Figures: The American Dream and the Untold Story of the Black Women Mathematicians Who Helped Win the Space Race
- Shoe Dog: A Memoir by the Creator of Nike
- The Library Book
- The Unwinding: An Inner History of the New America

You are on page 1of 6

OMAR HIRZALLAH1 , FUAD KITTANEH2 AND MOHAMMAD SAL MOSLEHIAN3 Abstract. Let A1 , An be operators acting on a separable complex Hilbert Pn space such that i=1 Ai = 0. It is shown that if A1 , An belong to a Schatten p-class, for some p > 0, then 2p/2 np1

n X i=1

Ai

p p

n X i,j=1

Ai Aj

p p

for 0 < p 2, and the reverse inequality holds for 2 p < . Moreover,

n X i,j=1

Ai Aj

2 p

2n2/p

n X i=1

Ai

2 p

for 0 < p 2, and the reverse inequality holds for 2 p < . These inequalities are related to a characterization of inner product spaces due to E.R. Lorch.

1. Introduction Let B(H) be the algebra of all bounded linear operators on a separable complex Hilbert space H. Let A B(H) be compact, and let 0 < p < . The Schatten p-norm (quasi-norm) for 1 p < (0 < p < 1) is dened by A p = (tr|A|p )1/p , where tr is the usual trace functional and |A| = (A A)1/2 . Clearly |A|2 p/2 = A 2 for p > 0. For p > 0, the Schatten p-class, denoted by Cp , is dened to be p the set of those compact operators A for which A p is nite. When p = 2, the Schatten p-norm A 2 = (tr|A|2 )1/2 is called the HilbertSchmidt norm of A. For p > 0, Cp is a two-sided ideal in B(H). For 1 p < , Cp is a Banach space; in particular, if A1 , , An Cp , then the triangle inequality for p asserts that

n n

(1.1)

i=1

Ai

p

i=1

Ai

p.

However, for 0 < p < 1, the quasi-norm . p does not satisfy the triangle inequality. It has been shown in [2] (see, also, [3]) that if A1 , , An Cp are positive and 0 < p 1, then

n n

(1.2)

i=1

Ai

i=1

Ai

p

For more information on the theory of the Schatten p-classes, the reader is referred to [1, 6].

2000 Mathematics Subject Classication. Primary: 47A30; Secondary: 46C15, 47B10, 47B15. Key words and phrases. Schatten p-norm, positive operator, inequality, inner product space.

1

It is well-known that a normed space X is an inner product space if and only if for every x, y X, we have (1.3) x+y

2

+ xy

= 2( x

+ y 2) .

The identity (1.3) is known as the parallelogram law. A version of this fundamental law holds for the Hilbert-Schmidt norm. Generalizations of the parallelogram law for the Schatten p-norms have been given in the form of the celebrated Clarkson inequalities (see [3] and references therein). These inequalities have proven to be very useful in analysis, operator theory, and mathematical physics. Another known characterization of inner product spaces is due to E.R. Lorch [5]. He proved that a normed space X is an inner product space if and only if for a xed integer n 3, and x1 , , xn X with n xi = 0, we have i=1

n n

(1.4)

i,j=1

xi xj

= 2n

i=1

xi 2 .

Since C2 is a Hilbert space under the inner product A, B = tr(B A), it follows that if A1 , , An C2 with n Ai = 0, then i=1

n n

(1.5)

i,j=1

Ai Aj

2 2

= 2n

i=1

Ai

2 2

In this paper, we establish operator inequalities for the Schatten p-norms that form natural generalizations of the identity (1.5). Our inequalities presented here seem natural enough and applicable to be widely useful. 2. Main results To achieve our goal, we need the following lemma (see [2, 4]). Lemma 2.1. Let A1 , , An Cp for some p > 0. If A1 , , An are positive, then

n n p n p1 i=1

Ai p p

i=1

Ai

p

i=1

Ai

p p

(a)

n n p n

Ai p p

i=1

i=1

Ai

p

np1

i=1

Ai

p p

(b)

for 1 p < . A commutative version of Lemma 2.1 can be formulated for scalars as follows: If a1 , , an are nonnegative real numbers, then

n n p n

p1 i=1

ap i

i=1

ai

i=1

ap i

(2.2)

i=1

ap i

i=1

ai

np1

i=1

ap i

for 1 p < . These inequalities follow, respectively, from the concavity of the function f (t) = tp , t [0, ) for 0 < p 1, and the convexity of the function f (t) = tp , t [0, ) for 1 p < . Our rst main result can be stated as follows. It furnishes a generalization of (1.5). Theorem 2.2. Let A1 , , An , B1 , , Bn Cp , for some p > 0, such that n i,j=1 Ai Bj = 0. Then 2

p/2

n 1 p1

Ai

i=1

p p

+

i=1

Bi

p p

i,j=1

Ai Bj

p p

n

Ai Bj

i,j=1

p p

p/2

n 1 p1

Ai

i=1

p p

+

i=1

Bi

p p

n n

Ai Bj

i,j=1

p p

=

i,j=1 n

|Ai Bj |2

i,j=1

|Ai Bj |2

p/2

n p/2

=

i,j=1 n

|Ai | + |Bj |

n

A Bj i

p/2

Bj Ai p/2

=

i,j=1

|Ai |2 +

i,j=1 n

|Bj |2

p/2 n p/2

p/2 i=1

|Ai | +

i=1 n

|Bi |

2 p/2 n

(2n)p/21 np/2

i=1 n

Ai

p p

+

i=1 n

Bi

p p

i=1

Ai

p p

+

i=1

Bi

p p

This proves the rst part of the theorem. Based on Lemma 2.1(b), one can employ an argument similar to that used in the proof of the rst part of the theorem to prove the second part. An application of Theorem 2.2 can be seen in the following result, which is a natural generalization of (1.5).

n n

Ai = 0.

2p/2 np1

i=1

Ai

p p

i,j=1

Ai Aj

p p

n n

Ai Aj

i,j=1

p p

2p/2 np1

i=1

Ai

p p

for 2 p < . In particular, (1.5) holds in the case where p = 2. Proof. Since i=1 Ai = 0. Consequently, i=1 Ai = 0, we have i=1 Ai n A Aj = 0. Utilizing Theorem 2.2 with Bj = Aj (1 j n), we obtain i,j=1 i the result. Using the same reasoning as in the proof of Theorem 2.2, one can obtain the following result concerning operators having orthogonal ranges. Recall that ranges of two operators A, B B(H) are orthogonal (written ranAranB) if and only if A B = 0. Theorem 2.4. Let A1 , , An Cp , for some p > 0, such that ranAi ranAj for i = j. Then

n n n n n

i=1

Ai

p p

i,j=1

Ai Aj

p p

n n

Ai Aj

i,j=1

p p

i=1

Ai

p p

for 2 p < . Our second main result, which also leads to a generalization of (1.5), can be stated as follows. Theorem 2.5. Let A1 , , An , B1 , , Bn Cp , for some p > 0, such that n i,j=1 Ai Bj = 0. Then

n n

Ai Bj

i,j=1

2 p

n2/p

i=1

|Ai |2 + |Bi |2

1/2 2 p

n

n for 2 p < .

2/p i=1

|Ai | + |Bi |

2 2 1/2 p

i,j=1

Ai Bj

2 p

n n

Ai Bj

i,j=1

2 p

=

i,j=1 n

|Ai Bj |2

p/2

i,j=1 n

|Ai Bj |2

p/2

(by(1.2))

=

i,j=1 n

=

i,j=1 n

|Ai |2 +

i,j=1

|Bj |2

p/2

n

i=1 n

|Ai |2 + |Bi |2

p/2 2/p p/2 |Bi |2 p/2

n

i=1

|Ai | +

n

n n

i=1 n 2/p i=1

|Ai | + |Bi |

p 2 1/2 p 1/2 2 p

2/p

|Ai |2 + |Bi |2

(by the second inequality of (2.2) ). This proves the rst part of the theorem. Based on (1.1), the rst inequality of Lemma 2.1(b), and the rst inequality of (2.1), one can employ an argument similar to that used in the proof of the rst part of the theorem to prove the second part. An application of Theorem 2.5, which is another natural generalization of (1.5), can be seen as follows. Corollary 2.6. Let A1 , , An Cp , for some p > 0, such that Then

n n n i=1

Ai = 0.

Ai Aj

i,j=1

2 p

2n2/p

i=1 n

Ai

2 p

Ai

i=1

2 p

i,j=1

Ai Aj

2 p

for 2 p < . In particular, (1.5) holds in the case where p = 2. Using the same reasoning as in the proof of Theorem 2.5, one can obtain the following result concerning operators having orthogonal ranges.

Theorem 2.7. Let A1 , , An Cp , for some p > 0, such that ranAi ranAj for i = j. Then

n n

Ai Aj

i,j=1

2 p

2n2/p1 (n 1)

i=1 n n

Ai

2 p

i=1

Ai

2 p

i,j=1

Ai Aj

2 p

[1] R. Bhatia, Matrix Analysis, Springer-Verlag, New York, 1997. [2] R. Bhatia and F. Kittaneh, Cartesian decompositions and Schatten norms, Linear Algebra Appl. 311 (2000), 109116. [3] R. Bhatia and F. Kittaneh, Clarkson inequalities with several operators, Bull. London Math. Soc. 36 (2004), 820832. [4] F. Kittaneh, Norm inequalities for certain operator sums, J. Funct. Anal. 143 (1997), 337348. [5] E.R. Lorch, On certain implications which characterize Hilbert space, Ann. of Math. (2) 49 (1948), 523532. [6] B. Simon, Trace Ideals and their Applications, Cambridge University Press, Cambridge, 1979.

1 Department

2 Department

3 Department of Pure Mathematics, Ferdowsi University of Mashhad, P.O. Box 1159, Mashhad 91775, Iran; Center of Excellence in Analysis on Algebraic Structures (CEAAS), Ferdowsi University of Mashhad, Iran. E-mail address: moslehian@ferdowsi.um.ac.ir and moslehian@ams.org

- Quantum Meditation: The Self-Spirit ProjectionUploaded bySEP-Publisher
- HilbertUploaded byCepa Ugol
- Spatial StatisticsUploaded byRezzy Eko Caraka
- 1363Uploaded bySilviu
- 5Uploaded byJonel Pagalilauan
- 14-20.txtUploaded bytamilselvi
- A.I.Bobenko, D.Matthes and Yu.B. Suris- Discrete and smooth orthogonal systems: C∞-approximationUploaded byGremnDL
- 2008eUploaded byP6E7P7
- ALAp 2Uploaded byInacio Sergio
- QFT.pdfUploaded byJuan Sebas Tian
- MatricesUploaded byPranivoid
- [2004!10!18 HeeJoo]ImageCompressionUsingtheHaarWaveletTransform 2Uploaded bySai Adarsh
- Java PrintingUploaded byCemal Abin
- SoleymaniUploaded byAnonymous wWOWz9UnW

- Qm Lecture NotesUploaded by3334333
- homeworksUploaded by7747252
- la09.pdfUploaded byNATUREISNATURE
- Textbook - Week 04Uploaded byaditya sharma
- 215Chw2solUploaded byi'm not telling
- KMtalkUploaded bychaltihawa
- Introduction to Matrix Methods and ApplicationsUploaded byatomoneutrino
- Linear Algebra Werner GreubUploaded byDani Greentown
- commsys-2-2012-6Uploaded byJoseph John
- Linear Algebra Done Wrong - Sergei TreilUploaded byNiflheim
- Paulsen_An Introduction to the Theory of Reproducing Kernel Hilbert SpacesUploaded bygrzerysz
- SyllabusMSc MathematicsUploaded bysrajubasava
- Chapter6(6.1~6.3)Uploaded byAnisa
- CSIR / NET Complete Information Brochure for Session 2014 15Uploaded byIMS Institute of Mathematical Sciences, New Delhi
- [Non-QU] Linear Algebra by Dr.Gabriel Nagy.pdfUploaded byamr
- math6338_hw1Uploaded byRicardo E.
- Linear Algebra by NayakUploaded byTanmoyDey
- Orthogonal GroupsUploaded byCieva Grace Dabalos Cabatchete
- Cauchy-Schwarz Inequality and Heisenberg's Uncertainty PrincipleUploaded byMark Eichenlaub
- 9.17 Reading Quiz, Orthonormal FunctionsUploaded byNathan Saunders
- Financial Econometrics lecture notes 2Uploaded bycdertert
- 45 Griewank Andreas BroydenUploaded byLuiz Henrique Melo
- Linear_Algebra_and_Geometry__Algebra__Logic_and_Applications_.pdfUploaded byAcis Assis
- Bernard Zygelman-A First Introduction to Quantum Computing and Information-Springer (2018)Uploaded byAakash Rabbii
- Calculus on Manifolds. Spivak. SolutionsUploaded byEcdee Fts
- William C. Brown-A Second Course in Linear Algebra-Wiley-Interscience (1988).pdfUploaded byJacob Sloat
- MATH 120-Linear Algebra With Differential Equations-Imran Naeem-Shaheen NazeerUploaded byHaris Ali
- A First Course in Optimization Theory. Rangarajan K. SundaramCambridge University Press (1996)Uploaded bySilvio Paula
- ECE CSP MTechSyllabusUploaded byStanly Kurian
- Menaldi 2015-12-12 Measure Integration (1)Uploaded byagaasfa