volume 6, issue 3, article 87, 2005.
Received 16 July, 2004;
accepted 29 June, 2005.
Communicated by:F. Hansen
Abstract Contents
JJ II
J I
Home Page Go Back
Close Quit
Journal of Inequalities in Pure and Applied Mathematics
A NUMERICAL METHOD IN TERMS OF THE THIRD DERIVATIVE FOR A DELAY INTEGRAL EQUATION FROM BIOMATHEMATICS
MARKUS SIGG
Department of Mathematics and Statistics University of Konstanz
Germany.
EMail:mail@MarkusSigg.de
c
2000Victoria University ISSN (electronic): 1443-5756 135-04
A Minkowski-Type Inequality for the Schatten Norm
Markus Sigg
Title Page Contents
JJ II
J I
Go Back Close
Quit Page2of19
J. Ineq. Pure and Appl. Math. 6(3) Art. 87, 2005
http://jipam.vu.edu.au
Abstract
LetF be a Schattenp-operator andR, Spositive operators. We show that the inequality|F(R+S)1c|pc≤ |F R1c|pc+|F S1c|pcfor the Schattenp-norm| · |pis true forp≥c= 1and forp≥c= 2, conjecture it to be true forp≥c∈[1,2], give counterexamples for the other cases, and present a numerical study for2×2 matrices. Furthermore, we have a look at a generalisation of the inequality which involves an additional factorσ(c, p).
2000 Mathematics Subject Classification:47A30, 47B10
Key words: Schatten class, Schatten norm, Norm inequality, Minkowski inequality, Triangle inequality, Powers of operators, Schatten-Minkowski constant.
This work was supported by the Dr. Helmut Manfred Riedl Foundation.
Contents
1 Introduction. . . 3
2 The Conjecture . . . 5
3 The Casep < cand the Casec6∈[1,2]. . . 6
4 Some Numerical Evidence. . . 10
5 Generalisation of (MS). . . 15
6 Conclusion. . . 18 References
A Minkowski-Type Inequality for the Schatten Norm
Markus Sigg
Title Page Contents
JJ II
J I
Go Back Close
Quit Page3of19
J. Ineq. Pure and Appl. Math. 6(3) Art. 87, 2005
http://jipam.vu.edu.au
1. Introduction
Let H and K be complex Hilbert spaces and 0 < p ≤ ∞. Following [1], we denote bycp(H, K)the space of Schattenp-operatorsT :H −→K, equipped with the Schattenp-norm or quasi-norm| · |p. Note that [1] deals only with the spacescp(H) :=cp(H, H). The generalisationscp(H, K)can be found in text- books like [2] and [3] (there written asBp(H, K)andSp(H, K)respectively).
By L(H) we denote the space of bounded linear operators on H, and by L(H)+ the subset of positive operators. With |T| := (T∗T)1/2 ∈ L(H)+ for T ∈L(H, K)we have forp <∞
|T|pp = tr|T|p forT ∈cp(H, K), and consequently
|T|pp = tr Tp forT ∈cp(H)+:=cp(H)∩L(H)+. Applying|T|p =|T∗|p forT ∈cp(H, K), this shows in case ofp < ∞
F U12
2 p =
U12F∗
2 p =
tr (F U F∗)p2 p2
=|F U F∗|p
2
forF ∈cp(H, K)andU ∈L(H)+. Because| · |∞is the usual operator norm,
F U12
2 p
=|F U F∗|p
2
is also true forp=∞, with the common convention ∞2 :=∞.
Our question, which arose while studying the integration of Schatten opera- tor valued functions in [4], is: For what values ofp∈ (0,∞]andc∈(0,∞)is
A Minkowski-Type Inequality for the Schatten Norm
Markus Sigg
Title Page Contents
JJ II
J I
Go Back Close
Quit Page4of19
J. Ineq. Pure and Appl. Math. 6(3) Art. 87, 2005
http://jipam.vu.edu.au
the Minkowski-like inequality (MS)
F (R+S)1c
c p
≤ F R1c
c p
+ F S1c
c p
true for allF ∈cp(H, K)andR, S ∈L(H)+?
A Minkowski-Type Inequality for the Schatten Norm
Markus Sigg
Title Page Contents
JJ II
J I
Go Back Close
Quit Page5of19
J. Ineq. Pure and Appl. Math. 6(3) Art. 87, 2005
http://jipam.vu.edu.au
2. The Conjecture
LetH, K, p, c, F, R, Sbe as above.
Theorem 2.1. Inequality (MS) is true forp≥c= 1and forp≥c= 2.
Proof. Forp≥c= 1, the triangle inequality for| · |p shows
|F(R+S)|p =|F R+F S|p ≤ |F R|p+|F S|p. Forp≥c= 2, the triangle inequality for| · |p
2 shows
F(R+S)12
2
p =|F(R+S)F∗|p
2 ≤ |F RF∗|p
2+|F SF∗|p
2 =
F R12
2 p+
F S12
2 p.
Theorem2.1suggests the following conjecture.
Conjecture 1. Inequality (MS) is true forp≥c∈[1,2].
For c ∈ (1,2)we have at the present time no proof of this conjecture for other than trivial situations, not even for the special case of 2× 2 matrices.
However, some justification will be given in Section4.
A Minkowski-Type Inequality for the Schatten Norm
Markus Sigg
Title Page Contents
JJ II
J I
Go Back Close
Quit Page6of19
J. Ineq. Pure and Appl. Math. 6(3) Art. 87, 2005
http://jipam.vu.edu.au
3. The Case p < c and the Case c 6∈ [1, 2]
In this section we will demonstrate, by providing counterexamples, that inequal- ity (MS) is not necessarily true for other values of (c, p) than those stated in Conjecture 1. We will offer one example for 0 < p < c < ∞, and one for arbitrary p when c < 1 or c > 2, both examples using 2×2 matrices. The power Ut for t > 0of a non-negative matrix U can be calculated easily with help of the spectral decomposition ofU.
Example 3.1. Inequality (MS) is violated for0< p < c < ∞by F :=
1 0 0 1
, R :=
1 0 0 0
, S :=
0 0 0 1
. Proof. FromUt =U forU ∈ {R, S, R+S}andt∈(0,∞)we get
F U1c
p =|U|p = (trUp)1p = (trU)1p, yielding
F R1c
p
= 1, F S1c
p
= 1,
F(R+S)1c p
= 2p1, and usingp < c,
F R1c
c p+
F S1c
c
p = 2 <2cp =
F(R+S)1c
c p.
A Minkowski-Type Inequality for the Schatten Norm
Markus Sigg
Title Page Contents
JJ II
J I
Go Back Close
Quit Page7of19
J. Ineq. Pure and Appl. Math. 6(3) Art. 87, 2005
http://jipam.vu.edu.au
The second example makes use of an inequality which is interesting in its own right. Seeming simple, it is surprisingly fiddly to prove:
Lemma 3.1. Forx∈(0,1)∪(2,∞)we have
1 + 1
√5
3 +√ 5 2
!x
+
1− 1
√5
3−√ 5 2
!x
<1 + 3x. Proof. Settingr := √
5,α1 := 1 + 1r,α2 := 1− 1r, andω := 3+r2 , we have to show
α1ωx+α2ω−x <1 + 3x.
The case x ∈ (2,∞): Setf(x) := α1ωx, g(x) := α2ω−x, h(x) := 1 + 3x for x ∈ (0,∞). Because α2 > 0 and ω > 1, g is strictly decreasing, thus f(x) +g(x) < f(x) +g(2)forx > 2. We will showf(x) +g(2) < h(x)for x > 2. Because f(2) +g(2) = h(2), this is done if we prove f0(x) < h0(x) for x > 2, which is equivalent to α1(ω3)x lnω < ln 3. This inequality is true forx = 2. All factors of its left side are positive, andω < 3, so the left side is strictly decreasing forx≥2. Hence the inequality is true forx >2as well.
The casex ∈ (0,1): After substitutings := ωx and settingδ := lnln 3ω, we have to prove the equivalent inequality
s+ 1 s + 1
r
s− 1 s
<1 +sδ
fors ∈(1, ω), which can be done by building a sandwich with a suitable poly-
A Minkowski-Type Inequality for the Schatten Norm
Markus Sigg
Title Page Contents
JJ II
J I
Go Back Close
Quit Page8of19
J. Ineq. Pure and Appl. Math. 6(3) Art. 87, 2005
http://jipam.vu.edu.au
nomial function inside: Set ϕ(s) :=s+ 1
s + 1 r
s−1
s
, p(s) := 2
1 + s−1 ω−1
, q(s) := (s−1)(s−ω)
(3−1)(3−ω)(ϕ(3)−p(3)) fors >0. The claim is
ϕ(s)< p(s) +q(s)<1 +sδ
fors ∈ (1, ω). The left inequality is verified by the fact thats·(p(s) +q(s)− ϕ(s))defines a polynomial of degree 3with three zeros {1, ω,3}, where 1 <
ω < 3, and with positive leading coefficientλ := 12(ϕ(3)−p(3))/(3−ω). To prove the second inequality, we inspect
ψ(s) := 1 +sδ−p(s)−q(s)
for s > 0and getψ00(s) = δ(δ−1)sδ−2 −2λ. Because 1 < δ < 2, ψ00 has a unique zero
s0 :=
δ(δ−1) 2λ
2−δ1
, 1< s0 < ω,
withψ00(s)>0fors ∈(0, s0)andψ00(s)<0fors ∈(s0,∞). Nowψ(1) = 0, ψ0(1) >0, andψ00(s) >0fors ∈(1, s0)showψ(s) >0fors ∈(1, s0], while ψ(s0)>0,ψ(ω) = 0,ψ0(ω)<0, andψ00(s)<0fors∈(s0, ω)showψ(s)>0 fors∈[s0, ω).
A Minkowski-Type Inequality for the Schatten Norm
Markus Sigg
Title Page Contents
JJ II
J I
Go Back Close
Quit Page9of19
J. Ineq. Pure and Appl. Math. 6(3) Art. 87, 2005
http://jipam.vu.edu.au
Example 3.2. Inequality (MS) is violated for0< p≤ ∞andc <1as well as c > 2by
F :=
0 0 1 0
, R:=
0 0 0 1
, S :=
2 −1
−1 1
. Proof. Evaluation of the matrix powers fort ∈(0,∞)gives
Rt=R, St =
1
2(α1ωt+α2ω−t) 1r(ω−t−ωt)
1
r(ω−t−ωt) 12(α2ωt+α1ω−t)
! ,
(R+S)t= 1 2
1 + 3t 1−3t 1−3t 1 + 3t
!
withr:=√
5,α1 := 1 +1r,α2 := 1−1r,ω := 3+r2 . ForU ∈ {R, S, R+S}we get in case ofp <∞
|F Ut|p =
tr (F U2tF∗)p21p
=√ ut
withut being the top left entry ofU2t. Using|F Ut|2∞ = |F U2tF∗|∞, the case p=∞yields the same result, thus for allp:
F R1c
p
= 0, F S1c
p
= r1
2(α1ω2/c+α2ω−2/c),
F(R+S)1c p =
r1
2(1 + 32/c).
Substituting 2c byx, we have to proveα1ωx+α2ω−x <1 + 3x forx∈(2,∞) and forx∈(0,1), which is the statement of Lemma3.1.
A Minkowski-Type Inequality for the Schatten Norm
Markus Sigg
Title Page Contents
JJ II
J I
Go Back Close
Quit Page10of19
J. Ineq. Pure and Appl. Math. 6(3) Art. 87, 2005
http://jipam.vu.edu.au
4. Some Numerical Evidence
To justify Conjecture 1, we present the results of a numerical study performed with2×2matrices.
From functional calculus it is known: For an operatorT ≥ 0on a complex Hilbert space the powers Tα, Tβ for α, β ∈ (0,∞) obey the rule TαTβ = Tα+β. IfT is invertible, thenTαcan be defined forα≤0as well, andTαTβ = Tα+β is true for allα, β ∈R.
Before turning to the matrix case, we note the following general lemma.
Lemma 4.1. LetH, K, F be as above andα∈(0,∞).
(a) LetT ∈L(H)+. ThenF Tα = 0if and only ifF T = 0.
(b) Let R, S ∈ L(H)+. ThenF(R+S)α = 0if and only if F Rα = 0 and F Sα = 0.
Proof. (a) SupposeF Tα = 0. Then|F Tα/2|2 =|F TαF∗|= 0, henceF Tα/2 = 0. Repeated application yieldsβ∈(0,1)withF Tβ = 0, thusF T =F TβT1−β
= 0.
Now suppose F T = 0. There is nothing to prove in the case of α = 1, so assume α 6= 1. IfT is invertible, then F Tα = F T Tα−1 = 0. IfT is not invertible, then we have0∈σ(T), the spectrum ofT. Choose polynomialsfn∈ R[t]forn ∈Nsuch thatfn(x)→xαforn → ∞uniformly forx∈σ(T). Then fn(T)→ Tα andF fn(T)→ F Tα forn → ∞, henceF fn(T) =fn(0)F → 0 forn→ ∞, thusF Tα = 0.
A Minkowski-Type Inequality for the Schatten Norm
Markus Sigg
Title Page Contents
JJ II
J I
Go Back Close
Quit Page11of19
J. Ineq. Pure and Appl. Math. 6(3) Art. 87, 2005
http://jipam.vu.edu.au
(b) Part (a) shows:
F Rα = 0 ∧ F Sα = 0 ⇐⇒ F R= 0 ∧ F S = 0
=⇒ F(R+S) = 0
⇐⇒ F(R+S)α = 0.
To prove the missing implication, suppose F(R + S) = 0. Then F RF∗ + F SF∗ = 0. Because F RF∗ ≥ 0 and F SF∗ ≥ 0, we getF RF∗ = 0, thus
|F R1/2|2 = |F RF∗| = 0andF R1/2 = 0. Applying (a) again givesF R = 0.
Symmetry showsF S = 0.
We will also use the following well-known property of2×2matrices:
Lemma 4.2. A complex 2×2matrixM is positive semidefinite if and only if there exista, b∈[0,∞)andγ ∈Cwith|γ|2 ≤absuch that
M =
a γ γ b
.
Lemma4.1(b) shows that, when checking Conjecture1, one may assume the denominator to be non-zero, or setting 00 := 0, in
qc,p(F, R, S) :=
F (R+S)1c
c p
F R1c
c p
+ F S1c
c p
.
We are searching for the supremum ofqc,p(F, R, S)over all complex2×2 matricesF, R, S withR, S ≥ 0. Forr ∈ [0,∞)andx ∈ Cdefiner∧x :=x
A Minkowski-Type Inequality for the Schatten Norm
Markus Sigg
Title Page Contents
JJ II
J I
Go Back Close
Quit Page12of19
J. Ineq. Pure and Appl. Math. 6(3) Art. 87, 2005
http://jipam.vu.edu.au
if |x| ≤ rand r∧x := (r/|x|)xotherwise. Lemma4.2shows that Rhas the structure
R=
α2 |α β| ∧γ
|α β| ∧γ β2
=:P(α, β, γ)
with α, β ∈ R and γ ∈ C, and a corresponding representation is valid for the matrix S. This means that we have to deal with six complex and four real variables, resulting in a 16-dimensional real optimisation problem: For λ = (λ1, . . . , λ16)∈R16we set
Fλ :=
λ1 +λ2i λ3+λ4i λ5 +λ6i λ7+λ8i
, Rλ :=P(λ9, λ10, λ11+λ12i), Sλ :=P(λ13, λ14, λ15+λ16i) and are asking for
σ(c, p) := sup
λ∈R16
qc,p(Fλ, Rλ, Sλ).
To attack this problem, GNU Octave [5], version 2.1.57, was utilised. It offers a function for determining the singular values of a matrix, which can be employed for calculating the Schatten norms. For the optimisation task the implementation [6], version 2002/05/09, with standard parameters of the Down- hill Simplex Method of Nelder and Mead ([7], 10.4) was used. The results are in perfect agreement with Conjecture 1. For visualisation, approximations for σ(c, p)forc ∈ {1.2,1.4,1.6,1.8,2.0}have been calculated and plotted with a step size of0.01forp, see Figure1.
A Minkowski-Type Inequality for the Schatten Norm
Markus Sigg
Title Page Contents
JJ II
J I
Go Back Close
Quit Page13of19
J. Ineq. Pure and Appl. Math. 6(3) Art. 87, 2005
http://jipam.vu.edu.au
Figure 1: Experimental approximations ofσ(c, p).
0.9 1 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2
1 1.2 1.4 1.6 1.8 2
p
c = 1.2 c = 1.4 c = 1.6 c = 1.8 c = 2.0
The apparently smooth shape ofp 7→ σ(c, p)for p ≤ c, together with the fact that for eachpa new random starting pointλwas used for the Nelder-Mead algorihm, gives some confidence in the validity of the data.
A closer inspection of some of the calculated numerical values suggests σ(2,1) = 2, σ 32,1
=σ 95,65
= 212, σ 85,65
=σ 2,32
= 213, σ 54,1
=σ 32,54
=σ 74,75
=σ 2,85
= 214, σ 65,1
=σ 95,32
= 215, which leads to the idea to look at log2 σ(c, p). It seems there is a linear depen-
A Minkowski-Type Inequality for the Schatten Norm
Markus Sigg
Title Page Contents
JJ II
J I
Go Back Close
Quit Page14of19
J. Ineq. Pure and Appl. Math. 6(3) Art. 87, 2005
http://jipam.vu.edu.au
dency oflog2 σ(c, p)fromcifc≥p. This observation will be made precise in the next section.
A Minkowski-Type Inequality for the Schatten Norm
Markus Sigg
Title Page Contents
JJ II
J I
Go Back Close
Quit Page15of19
J. Ineq. Pure and Appl. Math. 6(3) Art. 87, 2005
http://jipam.vu.edu.au
5. Generalisation of (MS)
It is natural to generalise (MS) and to ask for the smallestσ(c, p) ∈ [0,∞]for c∈(0,∞)andp∈(0,∞]such that
F (R+S)1c
c p
≤σ(c, p)
F R1c
c p
+ F S1c
c p
for all F ∈ cp(H, K)and R, S ∈ L(H)+ (and for all complex Hilbert spaces H and K). It is tempting to call σ(c, p) the Schatten-Minkowski constant for (c, p). By choosingF 6= 0and settingRto be the identity andS := 0it can be seen thatσ(c, p) ≥ 1. Now Conjecture1can be re-phrased usingσ(c, p), and, motivated by the numerical results, we add another conjecture:
Conjecture 2. (a) For1≤c≤2andp≥cwe haveσ(c, p) = 1.
(b) For0≤c≤2andp≤cwe haveσ(c, p) = 2cp−1. Again, the casesc= 1andc= 2are not too difficult to prove:
Theorem 5.1.
(a)σ(1, p) =
(1 forp≥1 21p−1 forp≤1 (b)σ(2, p) =
(1 forp≥2 22p−1 forp≤2.
Proof. σ(1, p) ≤ 1 for p ≥ 1 and σ(2, p) ≤ 1 for p ≥ 2 is the subject of Theorem 2.1, while σ(c, p) ≥ 1 is noted above. Example 3.1 tells us that
A Minkowski-Type Inequality for the Schatten Norm
Markus Sigg
Title Page Contents
JJ II
J I
Go Back Close
Quit Page16of19
J. Ineq. Pure and Appl. Math. 6(3) Art. 87, 2005
http://jipam.vu.edu.au
σ(c, p)≥2c/p−1for0< p ≤c <∞, yielding
σ(1, p)≥21p−1 forp≤1 and σ(2, p)≥2p2−1 forp≤2.
Now for the missing ‘≤’ inequalities. For the casec = 1, recall the inequality between the power means of degrees p ≤ 1 and 1, see e.g. [8], 8.12, which reads
αp+βp 2
1p
≤ α+β
2 or equivalently αp+βp ≤21−p(α+β)p forα, β ∈[0,∞). Together with the quasi-norm inequality of| · |p this gives
|F(R+S)|pp ≤ |F R|pp+|F S|pp ≤21−p(|F R|p+|F S|p)p and thus|F(R+S)|p ≤21p−1(|F R|p+|F S|p).
For the casec = 2, start with the power means inequality for the degrees p≤2and2,
αp+βp 2
1p
≤
α2+β2 2
12
or equivalently αp+βp ≤21−p2 (α2+β2)p2 forα, β ∈[0,∞). Together with the quasi-norm inequality of| · |p
2 this gives
F(R+S)12
p p
=|F(R+S)F∗|
p 2p 2
≤ |F RF∗|
p p2 2
+|F SF∗|
p 2p 2
= F R12
p p+
F S12
p
p ≤21−p2
F R12
2 p+
F S12
2 p
p2
A Minkowski-Type Inequality for the Schatten Norm
Markus Sigg
Title Page Contents
JJ II
J I
Go Back Close
Quit Page17of19
J. Ineq. Pure and Appl. Math. 6(3) Art. 87, 2005
http://jipam.vu.edu.au
and consequently
F(R+S)12
2 p
≤22p−1
F R12
2 p
+ F S12
2 p
.
A Minkowski-Type Inequality for the Schatten Norm
Markus Sigg
Title Page Contents
JJ II
J I
Go Back Close
Quit Page18of19
J. Ineq. Pure and Appl. Math. 6(3) Art. 87, 2005
http://jipam.vu.edu.au
6. Conclusion
Starting with Conjecture 1, which we proved for the cases c = 1 and c = 2 in Theorem 2.1, a numerical study of 2 ×2 matrices led to the generalised Conjecture2, which we also proved forc= 1andc= 2in Theorem5.1.
The given proofs make use of the (quasi-) triangle inequality of the Schatten (quasi-) norm. Another ingredient to Theorem 5.1is the power means inequal- ity. Presumably, a combination of these inequalities shall also be central when dealing with the case c 6= 1,2. However, it is unclear how to apply the tri- angle inequality in this situation, because there is no obvious way to get from F(R+S)1/cto an expression whereRandScan be separated.
A Minkowski-Type Inequality for the Schatten Norm
Markus Sigg
Title Page Contents
JJ II
J I
Go Back Close
Quit Page19of19
J. Ineq. Pure and Appl. Math. 6(3) Art. 87, 2005
http://jipam.vu.edu.au
References
[1] C.A. MCCARTHY,cp, Israel J. Math., 5 (1967), 249–271.
[2] J. WEIDMANN, Linear Operators in Hilbert Spaces, Springer, New York 1980.
[3] R. MEISE ANDD. VOGT, Introduction to Functional Analysis, Clarendon Press, Oxford 1997.
[4] M. SIGG, Zur sesquilinearen Integration Schatten-operatorwertiger Funk- tionen, Konstanzer Schriften in Mathematik und Informatik, 200 (2004), http://www.inf.uni-konstanz.de/Preprints/.
[5] J.W. EATON et al,http://www.octave.org.
[6] E. GROSSMANN, http://octave.sourceforge.net/index/
f/nelder_mead_min.html.
[7] W.H. PRESS, B.P. FLANNERY, S.A. TEUKOLSKYANDW.T. VETTER- LING, Numerical Recipes in C: The Art of Scientific Computing, Cam- bridge University Press 1992,http://www.nr.com.
[8] J. HERMAN, R. KU ˇCERA AND J. ŠIMŠA, Equations and Inequalities, Springer, New York (2000).