Hadamard Product Versions Jagjit Singh Matharu and
Jaspal Singh Aujla vol. 10, iss. 2, art. 51, 2009
Title Page
Contents
JJ II
J I
Page1of 13 Go Back Full Screen
Close
HADAMARD PRODUCT VERSIONS OF THE CHEBYSHEV AND KANTOROVICH INEQUALITIES
JAGJIT SINGH MATHARU AND JASPAL SINGH AUJLA
Department of Mathematics National Institute of Technology Jalandhar 144011, Punjab, INDIA
EMail:matharujs@yahoo.com aujlajs@nitj.ac.in
Received: 10 February, 2009
Accepted: 15 April, 2009
Communicated by: S. Puntanen
2000 AMS Sub. Class.: Primary 15A48; Secondary 15A18, 15A45.
Key words: Chebyshev inequality, Kantorovich inequality, Hadamard product
Abstract: The purpose of this note is to prove Hadamard product versions of the Chebyshev and the Kantorovich inequalities for positive real numbers. We also prove a generalization of Fiedler’s inequality.
Acknowledgements: The authors thank a referee for useful suggestions.
Hadamard Product Versions Jagjit Singh Matharu and
Jaspal Singh Aujla vol. 10, iss. 2, art. 51, 2009
Title Page Contents
JJ II
J I
Page2of 13 Go Back Full Screen
Close
Contents
1 Introduction 3
2 The Chebyshev and Kantorovich Inequalities: Matrix Versions 5
Hadamard Product Versions Jagjit Singh Matharu and
Jaspal Singh Aujla vol. 10, iss. 2, art. 51, 2009
Title Page Contents
JJ II
J I
Page3of 13 Go Back Full Screen
Close
1. Introduction
In what follows, the capital letters A, B, C, . . . denote m × m complex matrices, whereas the small letters a, b, c, . . . denote real numbers, unless mentioned other- wise. By X ≥ Y we mean that X − Y is positive semidefinite (X > Y mean X −Y is positive definite). For A = (aij)and B = (bij), A◦B = (aijbij) de- notes the Hadamard product of A and B. According to Schur’s theorem [4, Page 23] the Hadamard product is monotone in the sense thatA ≥ B, C ≥ D implies A◦C ≥B◦D. The tensor productA⊗B is them2×m2 matrix
(1.1)
a11B · · · a1mB ... ... am1B · · · ammB
.
Marcus and Khan in [10] made the simple but important observation that the Hadamard product is a principal submatrix of the tensor product. The inequality
(1.2)
n
X
i=1
wiai
! n X
i=1
wibi
!
≤
n
X
i=1
wiaibi
holds for all a1 ≥ a2 ≥ · · · ≥ an ≥ 0, b1 ≥ b2 ≥ · · · ≥ bn ≥ 0 and weights wi ≥ 0, i = 1, . . . , n. Hardy, Littlewood and Polya [6, page 43] attribute this inequality to Chebyshev. For0< a≤ai ≤b, wi ≥0, i= 1,2, . . . , n, Kantorovich’s inequality states that
(1.3)
n
X
i=1
wiai
! n X
i=1
wi
ai
!
≤ (a+b)2 4ab
n
X
i=1
wi
!2
.
Hadamard Product Versions Jagjit Singh Matharu and
Jaspal Singh Aujla vol. 10, iss. 2, art. 51, 2009
Title Page Contents
JJ II
J I
Page4of 13 Go Back Full Screen
Close
In Section2, we state and prove matrix versions of inequalities (1.2) and (1.3) involv- ing the Hadamard product. A generalization of Fiedler’s inequality is also proved in this section. There are several generalizations of Kantorovich and Fiedler’s inequal- ity; see [2,3,8,9].
Hadamard Product Versions Jagjit Singh Matharu and
Jaspal Singh Aujla vol. 10, iss. 2, art. 51, 2009
Title Page Contents
JJ II
J I
Page5of 13 Go Back Full Screen
Close
2. The Chebyshev and Kantorovich Inequalities:
Matrix Versions
We begin with a Hadamard product version of inequality (1.2).
Theorem 2.1. LetA1 ≥ · · · ≥An≥0andB1 ≥ · · · ≥Bn ≥0. Then
(2.1)
n
X
i=1
wiAi
!
◦
n
X
i=1
wiBi
!
≤
n
X
i=1
wi
! n X
i=1
wi(Ai◦Bi)
! ,
wherewi ≥0, i= 1, . . . , n,are weights.
Proof. We have
n
X
i=1
wi
! n X
i=1
wi(Ai◦Bi)
!
−
n
X
i=1
wiAi
!
◦
n
X
i=1
wiBi
! (2.2)
=
n
X
i,j=1
(wiwj(Aj ◦Bj)−wiwj(Ai◦Bj))
= 1 2
n
X
i,j=1
wiwj(Aj ◦Bj)−wiwj(Ai◦Bj)
+wjwi(Ai◦Bi)−wjwi(Aj◦Bi)
= 1 2
n
X
i,j=1
wiwj(Ai−Aj)◦(Bi−Bj).
Since the Hadamard product of two positive semidefinite matrices is positive semidef- inite, therefore the summand in2.2is positive semidefinite.
Hadamard Product Versions Jagjit Singh Matharu and
Jaspal Singh Aujla vol. 10, iss. 2, art. 51, 2009
Title Page Contents
JJ II
J I
Page6of 13 Go Back Full Screen
Close
Our next result is a Hadamard product version of inequality (1.3) .
Theorem 2.2. LetA1, . . . , An be such that 0 < aIm ≤ Ai ≤ bIm, i = 1, . . . , n (hereIm denotes them×midentity matrix). Then
(2.3)
n
X
i=1
Wi1/2AiWi1/2
!
◦
n
X
i=1
Wi1/2A−1i Wi1/2
!
≤ a2+b2 2ab
n
X
i=1
Wi
!
◦
n
X
i=1
Wi
!
for allWi ≥0, i= 1, . . . , n.
Proof. We first prove the inequality
(2.4) P1/2AP1/2◦Q1/2B−1Q1/2+P1/2A−1P1/2◦Q1/2BQ1/2 ≤ a2 +b2
ab (P ◦Q), when0< aIm ≤ A, B ≤bIm andP, Q≥ 0. LetA =U DU∗ andB =VΓV∗ with unitaryU andV, and diagonal matricesDandΓ. Then
A⊗B−1+A−1⊗B = (U ⊗V)(D⊗Γ + Γ−1⊗D)(U ⊗V)∗
≤(U ⊗V)
a2+b2
ab (Im⊗Im)
(U ⊗V)∗
= a2+b2
ab (Im⊗Im), where the inequality follows from (1.3). Thus we have
P1/2AP1/2⊗Q1/2B−1Q1/2 +P1/2A−1P1/2⊗Q1/2BQ1/2 (2.5)
Hadamard Product Versions Jagjit Singh Matharu and
Jaspal Singh Aujla vol. 10, iss. 2, art. 51, 2009
Title Page Contents
JJ II
J I
Page7of 13 Go Back Full Screen
Close
= (P1/2⊗Q1/2)(A⊗B−1 +A−1⊗B)(P1/2⊗Q1/2)
≤ a2+b2
ab (P ⊗Q).
Since the Hadamard product is a principal submatrix of the tensor product, the in- equality (2.4) follows from (2.5). On takingB =AandQ=P in (2.4) we see that (2.3) holds forn = 1. Further, by (2.4) we have
Wi1/2AiWi1/2◦Wj1/2A−1j Wj1/2+Wi1/2A−1i Wi1/2◦Wj1/2AjWj1/2 ≤ a2+b2
ab (Wi◦Wj) fori, j = 1, . . . , n. Summing overi, j, we have
(2.6) 2
n
X
i,j=1
h
Wi1/2AiWi1/2
◦
Wj1/2A−1j Wj1/2i
≤
a2+b2 ab
n X
i,j=1
(Wi◦Wj),
which implies that
n
X
i=1
Wi1/2AiWi1/2
!
◦
n
X
i=1
Wi1/2A−1i Wi1/2
!
≤
a2+b2 2ab
n
X
i=1
Wi
!
◦
n
X
i=1
Wi
! .
The next corollary follows on takingWi =wiIm,i= 1, . . . , n.
Corollary 2.3. LetA1, . . . , Anbe such that0< aIm ≤Ai ≤bIm,andwi ≥0, i= 1, . . . , nbe weights. Then
n
X
i=1
wiAi
!
◦
n
X
i=1
wiA−1i
!
≤
a2 +b2 2ab
n X
i=1
wi
!2
Im.
Hadamard Product Versions Jagjit Singh Matharu and
Jaspal Singh Aujla vol. 10, iss. 2, art. 51, 2009
Title Page Contents
JJ II
J I
Page8of 13 Go Back Full Screen
Close
Remark 1. The casen= 1of Corollary2.3is proved in [7]. The example A=
2 1 1 1
, a= 3−√ 5
2 , b= 3 +√ 5 2 shows that the inequality
A◦A−1 ≤ (a+b)2 4ab I2 need not be true.
For our next result we need the following lemma.
Lemma 2.4. Let0≤r≤1. ThenAr+A−r≤A+A−1 for allA >0.
Proof. Suppose thatA=UΓU∗with unitaryU and diagonal matrixΓ. Then Ar+A−r =U(Γr+ Γ−r)U∗
≤U(Γ + Γ−1)U∗ =A+A−1
sincexr+x−r ≤x+x−1for any positive real numberxand0≤r ≤1.
Theorem 2.5. Let0≤α < β. Then
n
X
i=1
Wi1/2AαiWi1/2
!
◦
n
X
i=1
Wi1/2A−αi Wi1/2
!
≤
n
X
i=1
Wi1/2AβiWi1/2
!
◦
n
X
i=1
Wi1/2A−βi Wi1/2
!
for allAi >0andWi ≥0,i= 1, . . . , n.
Hadamard Product Versions Jagjit Singh Matharu and
Jaspal Singh Aujla vol. 10, iss. 2, art. 51, 2009
Title Page Contents
JJ II
J I
Page9of 13 Go Back Full Screen
Close
Proof. We first prove the inequality
(2.7)
Wi1/2AαiWi1/2
◦
Wj1/2A−αj Wj1/2 +
Wi1/2A−αi Wi1/2
◦
Wj1/2AαjWj1/2
≤
Wi1/2AβiWi1/2
◦
Wj1/2A−βj Wj1/2 +
Wi1/2A−βi Wi1/2
◦
Wj1/2AβjWj1/2
for0≤α < β. Let0≤r≤1. Then
Wi1/2AriWi1/2
⊗
Wj1/2A−rj Wj1/2 +
Wi1/2A−ri Wi1/2
⊗
Wj1/2ArjWj1/2
=
Wi1/2⊗Wj1/2
Ari ⊗A−rj +A−ri ⊗Arj
Wi1/2⊗Wj1/2
=
Wi1/2⊗Wj1/2 Ai⊗A−1j r
+ Ai⊗A−1j −r
Wi1/2⊗Wj1/2
≤
Wi1/2⊗Wj1/2 Ai⊗A−1j
+ Ai⊗A−1j −1
Wi1/2⊗Wj1/2
where the inequality follows from Lemma2.4. Takingr=α/βand replacingAi by Aβi andAj byAβj, we have
Wi1/2AαiWi1/2
⊗
Wj1/2A−αj Wj1/2 +
Wi1/2A−αi Wi1/2
⊗
Wj1/2AαjWj1/2
≤
Wi1/2AβiWi1/2
⊗
Wj1/2A−βj Wj1/2
+
Wi1/2A−βi Wi1/2
⊗
Wj1/2AβjWj1/2
. Again using the fact that the Hadamard product is a principal submatrix of the tensor
Hadamard Product Versions Jagjit Singh Matharu and
Jaspal Singh Aujla vol. 10, iss. 2, art. 51, 2009
Title Page Contents
JJ II
J I
Page10of 13 Go Back Full Screen
Close
product, the preceding inequality implies (2.7). Summing overi, j in (2.7), we have
n
X
i=1
Wi1/2AαiWi1/2
!
◦
n
X
i=1
Wi1/2A−αi Wi1/2
!
≤
n
X
i=1
Wi1/2AβiWi1/2
!
◦
n
X
i=1
Wi1/2A−βi Wi1/2
! .
Corollary 2.6. Let0≤α < β. Then
n
X
i=1
Aαi
!
◦
n
X
j=1
A−αj
!
≤
n
X
i=1
Aβi
!
◦
n
X
j=1
A−βj
!
for allAi >0,i= 1, . . . , n.
Proof. TakingWi =Im in Theorem2.5we get the desired result.
Corollary 2.7. Let0≤β. Then Im ≤
n
X
i=1
Wi1/2AβiWi1/2
!
◦
n
X
i=1
Wi1/2A−βi Wi1/2
!
for allAi >0andWi ≥0,i= 1, . . . , n, wherePn
i=1Wi =Im. Proof. Takingα = 0in Theorem2.5gives the desired inequality.
Remark 2. Corollary2.7is another generalization of Fiedler’s inequality [5]
A◦A−1 ≥Im.
Hadamard Product Versions Jagjit Singh Matharu and
Jaspal Singh Aujla vol. 10, iss. 2, art. 51, 2009
Title Page Contents
JJ II
J I
Page11of 13 Go Back Full Screen
Close
Next we prove a convexity theorem involving the Hadamard product.
Theorem 2.8. The function
f(t) =A1+t◦B1−t+A1−t◦B1+t
is convex on the interval[−1,1]and attains its minimum att= 0for allA, B >0.
Proof. Sincef is continuous we need to prove only thatf is mid-point convex. Note that forA, B >0ands, tin[−1,1]the matrices
A1+s+t A1+s A1+s A1+(s−t)
,
A1−(s+t) A1−s A1−s A1−(s−t)
, B1+s+t B1+s
B1+s B1+(s−t)
,
B1−(s+t) B1−s B1−s B1−(s−t)
are positive semidefinite. Hence the matrix X=
A1+s+t◦B1−(s+t)+A1−(s+t)◦B1+s+t A1+s◦B1−s+A1−s◦B1+s A1+s◦B1−s+A1−s◦B1+s A1+(s−t)◦B1−(s−t)+A1−(s−t)◦B1+(s−t)
is positive semidefinite. Similarly, the matrix Y=
A1+(s−t)◦B1−(s−t)+A1−(s−t)◦B1+(s−t) A1+s◦B1−s+A1−s◦B1+s A1+s◦B1−s+A1−s◦B1+s A1+(s+t)◦B1−(s+t)+A1−(s+t)◦B1+s+t
is positive semidefinite. Hence
(2.8) X+Y =
f(s+t) +f(s−t) 2f(s) 2f(s) f(s+t) +f(s−t)
is positive semidefinite, which implies that f(s)≤ 1
2[f(s+t) +f(s−t)].
Hadamard Product Versions Jagjit Singh Matharu and
Jaspal Singh Aujla vol. 10, iss. 2, art. 51, 2009
Title Page Contents
JJ II
J I
Page12of 13 Go Back Full Screen
Close
This proves the convexity off. Further, note that f(t) = f(−t). This together with the convexity off implies thatf attains its minimum at 0.
Corollary 2.9. The function
g(t) = At◦B1−t+A1−t◦Bt
is decreasing on [0,1/2], increasing on [1/2,1], and attains its minimum at t = 12 for allA, B >0.
Proof. The proof follows on replacingA, B byA1/2, B1/2 andtby 1+t2 in Theorem 2.8.
A norm|||·|||onm×mcomplex matrices is called unitarily invariant if|||U XV|||=
|||X|||for all unitary matricesU, V. IfAis positive semidefinite andXis any matrix, then
|||A◦X||| ≤max aii|||X|||
for all unitarily invariant norms||| · |||[1]. Thus the proof of the following corollary follows from Corollary2.9using the fact thatg(1/2)≤g(t)≤g(1) =g(0).
Corollary 2.10. Let0≤t≤1. Then,
2|||A1/2◦B1/2||| ≤ |||At◦B1−t+A1−t◦Bt||| ≤ |||A+B|||
for all unitarily invariant norms||| · |||and allA, B >0.
Hadamard Product Versions Jagjit Singh Matharu and
Jaspal Singh Aujla vol. 10, iss. 2, art. 51, 2009
Title Page Contents
JJ II
J I
Page13of 13 Go Back Full Screen
Close
References
[1] T. ANDO, R.A. HORN AND C.R. JOHNSON, The singular values of the Hadamard product: A basic inequality, Linear Multilinear Algebra, 21 (1987), 345–365.
[2] J.K. BAKSALARYAND S. PUNTANEN, Generalized matrix versions of the Cauchy-Schwarz and Kantorovich inequalities, Aequationes Math., 41 (1991), 103–110.
[3] R.B. BAPAT AND M.K. KWONG, A generalisation of A◦A−1 ≥ I, Linear Algebra Appl., 93 (1987), 107–112.
[4] R. BHATIA, Matrix Analysis, Springer Verlag, New York, 1997.
[5] M. FIEDLER, Über eine Ungleichung für positiv definite Matrizen, Math.
Nachrichten, 23 (1961), 197–199.
[6] G.H. HARDY, J.E. LITTLEWOOD ANDG. POLYA, Inequalities, Cambridge University Press, Cambridge, 1959.
[7] J. MI ´CI ´C, J. PECARIC ANDY. SEO, Complementary inequalities to inequal- ities of Jensen and Ando based on the Mond-Peˇcari´c method, Linear Algebra Appl., 318 (2000), 87–107.
[8] A.W. MARSHALL AND I. OLKIN, Matrix versions of the Cauchy and Kan- torovich inequalities, Aequationes Math., 40 (1990), 89–93.
[9] M. SINGH, J.S. AUJLA AND H.L. VASUDEVA, Inequalities for Hadamard product and unitarily invariant norms of matrices, Linear Multilinear Algebra, 48 (2000), 247–262.
[10] M. MARCUSANDN.A. KHAN, A note on Hadamard product, Canad. Math.
Bull., 2 (1950), 81–83.