Matrices: Theory and Applications

291 Pages·2010·3.906 MB·English

by  Denis Serre

Graduate Texts in Mathematics 216
Editorial Board
S. Axler
K.A. Ribet
For other titles in this series, go to
http://www.springer.com/series/136
Denis Serre
Matrices
Theory and Applications
Second Edition
1 C
Denis Serre
Unité de Mathématiques Pures et Appliquées
École Normale Supérieure de Lyon
69364 Lyon Cedex 07
France
[e-mail]
Editorial Board:
S. Axler K. A. Ribet
Mathematics Department Mathematics Department
San Francisco State University University of California at Berkeley
San Francisco, CA 94132 Berkeley, CA 94720
USA USA
[e-mail] [e-mail]

ISSN 0072-5285
ISBN 978-1-4419-7682-6 e-ISBN 978-1-4419-7683-3
DOI 10.1007/978-1-4419-7683-3
Springer New York Dordrecht Heidelberg London
Library of Congress Control Number: 2010938321
Mathematics Subject Classification (2010): 15-XX, 22-XX, 47-XX, 65-XX
© Springer Science+Business Media, LLC 2010
All rights reserved. This work may not be translated or copied in whole or in part without the written
permission of the publisher (Springer Science+Business Media, LLC, 233 Spring Street, New York, NY
10013, USA), except for brief excerpts in connection with reviews or scholarly analysis. Use in connec-
tion with any form of information storage and retrieval, electronic adaptation, computer software, or by
similar or dissimilar methodology now known or hereafter developed is forbidden.
The use in this publication of trade names, trademarks, service marks, and similar terms, even if they are
not identified as such, is not to be taken as an expression of opinion as to whether or not they are subject
to proprietary rights.
Printed on acid-free paper
Springer is part of Springer Science+Business Media (www.springer.com)
ToPascale,Fanny,Paul,andJoachim
Contents
PrefacefortheSecondEdition ....................................... xi
PrefacefortheFirstEdition ......................................... xiii
1 ElementaryLinearandMultilinearAlgebra ...................... 1
1.1 VectorsandScalars ......................................... 1
1.2 LinearMaps............................................... 5
1.3 BilinearMaps ............................................. 9
2 WhatAreMatrices............................................. 15
2.1 Introduction ............................................... 15
2.2 MatricesasLinearMaps..................................... 19
2.3 MatricesandBilinearForms ................................. 28
3 SquareMatrices ............................................... 31
3.1 Determinant ............................................... 31
3.2 Minors ................................................... 34
3.3 Invertibility ............................................... 38
3.4 EigenvaluesandEigenvectors ................................ 42
3.5 TheCharacteristicPolynomial................................ 43
3.6 Diagonalization ............................................ 48
3.7 Trigonalization............................................. 49
3.8 Rank-OnePerturbations ..................................... 52
3.9 AlternateMatricesandthePfaffian............................ 54
3.10 CalculatingtheCharacteristicPolynomial ...................... 56
3.11 IrreducibleMatrices ........................................ 59
4 TensorandExteriorProducts ................................... 69
4.1 TensorProductofVectorSpaces.............................. 69
4.2 ExteriorCalculus........................................... 72
4.3 TensorizationofLinearMaps ................................ 77
4.4 APolynomialIdentityinM (K).............................. 78
n
vii
viii Contents
5 MatriceswithRealorComplexEntries........................... 83
5.1 SpecialMatrices ........................................... 83
5.2 EigenvaluesofReal-andComplex-ValuedMatrices.............. 86
5.3 SpectralDecompositionofNormalMatrices.................... 91
5.4 NormalandSymmetricReal-ValuedMatrices................... 92
5.5 FunctionalCalculus ........................................ 94
5.6 NumericalRange........................................... 98
5.7 TheGershgorinDomain.....................................102
6 HermitianMatrices ............................................109
6.1 TheSquareRootoverHPD .................................109
n
6.2 RayleighQuotients .........................................110
6.3 FurtherPropertiesoftheSquareRoot..........................114
6.4 SpectrumofRestrictions ....................................115
6.5 SpectrumversusDiagonal ...................................117
6.6 TheDeterminantofNonnegativeHermitianMatrices.............119
7 Norms ........................................................127
7.1 ABriefReview ............................................127
7.2 Householder’sTheorem .....................................133
7.3 AnInterpolationInequality ..................................135
7.4 VonNeumann’sInequality...................................137
8 NonnegativeMatrices ..........................................149
8.1 NonnegativeVectorsandMatrices ............................149
8.2 ThePerron–FrobeniusTheorem:WeakForm ...................150
8.3 ThePerron–FrobeniusTheorem:StrongForm ..................151
8.4 CyclicMatrices ............................................154
8.5 StochasticMatrices.........................................156
9 MatriceswithEntriesinaPrincipalIdealDomain;JordanReduction163
9.1 Rings,PrincipalIdealDomains ...............................163
9.2 InvariantFactorsofaMatrix .................................167
9.3 SimilarityInvariantsandJordanReduction .....................170
10 ExponentialofaMatrix,PolarDecomposition,andClassicalGroups 183
10.1 ThePolarDecomposition....................................183
10.2 ExponentialofaMatrix .....................................184
10.3 StructureofClassicalGroups.................................188
10.4 TheGroupsU(p,q).........................................191
10.5 TheOrthogonalGroupsO(p,q) ..............................192
10.6 TheSymplecticGroupSp ..................................195
n
Contents ix
11 MatrixFactorizationsandTheirApplications .....................207
11.1 TheLU Factorization .......................................208
11.2 CholeskiFactorization ......................................213
11.3 TheQRFactorization .......................................214
11.4 SingularValueDecomposition................................216
11.5 TheMoore–PenroseGeneralizedInverse.......................218
12 IterativeMethodsforLinearSystems ............................225
12.1 AConvergenceCriterion ....................................226
12.2 BasicMethods.............................................227
12.3 TwoCasesofConvergence ..................................229
12.4 TheTridiagonalCase .......................................231
12.5 TheMethodoftheConjugateGradient.........................235
13 ApproximationofEigenvalues ...................................247
13.1 GeneralConsiderations......................................247
13.2 HessenbergMatrices........................................249
13.3 TheQRMethod............................................253
13.4 TheJacobiMethod .........................................259
13.5 ThePowerMethods ........................................266
References.........................................................277
IndexofNotations..................................................279
GeneralIndex .....................................................283
CitedNames .......................................................289
Preface for the Second Edition
Itisonlyafterlongusethatanauthorrealizestheflawsandthegapsinhisorher
book. Having taught from a significant part of it, having gathered more than three
hundredexercisesonapublicwebsitehttp://www.umpa.ens-lyon.fr/
serre/DPF/exobis.pdf,andhavinglearnedalotwithineightyearsofread-
ing after the first edition was published, I arrived at the conclusion that a second
editionofMatrices:TheoryandApplicationsshouldbesignificantlydifferentfrom
thefirstone.
First of all, I felt ashamed of the very light presentation of the backgrounds in
linear algebra and elementary matrix theory in Chapter 1. In French, I should say
cen’e´taitnifait,nia` faire(neitherdone,nortobedone).Ithusbeganbyrewriting
this part completely, taking this opportunity to split pure linear algebra from the
introductiontomatrices.IhopethatthereaderissatisfiedwiththenewChapters1
and2below.
When teaching, it was easy to recognize the lack of logical structure here and
there.Forthesakeofamoreelegantpresentation,Ithereforemovedseveralstate-
ments from one chapter to another. It even happened that entire sections was dis-
placed,suchasthoseaboutsingularvaluedecomposition,theLeverrieralgorithm,
theSchurcomplement,orthesquarerootofpositiveHermitianmatrices.
Next, I realized that some important material was missing in the first edition.
Thishasledmetoincreasethesizeofthisbookbyaboutfortypercent.Thenewly
addedtopicsare
• Dunforddecomposition
• Calculuswithrank-oneperturbations
• ImprovementbyPreparataandSarwateofLeverrier’salgorithmforcalculating
thecharacteristicpolynomial
• Tensorcalculus
• PolynomialidentityofAmitsurandLevitzki
• Regularityofsimpleeigenvaluesforcomplexmatrices
• FunctionalcalculusandtheDunford–Taylorformula
• Stableandunstablesubspaces
xi
xii PrefacefortheSecondEdition
• Numericalrange
• Weylinequalities
• Concavityof(detH)1/noverHPD
n
• vonNeumanninequality
• ConvergenceoftheJacobimethodwithrandomchoice(perhapsanewresult).
Withsomanyadditions,thechapteronrealandcomplexmatricesextendedbeyond
areasonablesize.Thisinturnledmetosplitit,bydedicatingaspecificchapterto
thestudyofHermitianmatrices.Becausetensorcalculus,togetherwithpolynomial
identities,alsoformsanewchapter,thenumberofchaptershasincreasedfromten
tothirteen.
The reader might wonder why I included the new fourth chapter, because it is
essentially not used in the sequel. Several reasons led me to this choice. First of
all,tensorandexteriorcalculusarefundamentalinspectralanalysis,indifferential
geometry, in theoretical physics and many other domains. Next, I think that the
theoremofAmitsurandLevitzkiisoneofthemostbeautifulinmatrixtheory,which
remained mysterious until a quite direct proof using exterior algebra was found,
allowing it to be presented in a graduate textbook. The presence of this result
isconsistentwithmyphilosophythateverychapter,beyondtheintroductoryones,
shouldcontainatleastoneadvancedstatement.Mylastmotivationwastoinclude
morealgebraicissues,becauseitisapermanenttendancyofmatrixanalysistobe
tooanalytic.
Followingthispointofview,IshouldhavelovedtoincludeaproofofA.Horn’s
conjecture, after W. Fulton, A. Klyachko, A. Knutson and T. Tao: the statement,
whichhasananalyticflavorinasmuchasitconsistsininequalitiesbetweeneigen-
values of Hermitian matrices, is undoubtedly a great achievement of algebraists,
involving,asitdoes,representationtheoryandSchubertcalculus.However,thema-
terialtobeincludedwouldhavebeenfartooadvanced.Thistheoryisstilltoofresh
and will not be part of textbooks before it has been digested, and this will take
perhapsafewdecades.ThusIdecidedtobecontentwithdiscussingtheWeyland
Lidski˘ıinequalities,whicharethelowerstepsinHorn’slist.
Eventhoughthisnewversionhadexpandedalot,Ithoughtlongandhardabout
including a fundamental and beautiful (to my opinion) issue, namely Loewner’s
theory of operator monotone functions. I eventually abandoned this idea because
the topic would have needed too much space, thus upsetting the balance between
elementaryandadvancedmaterial.
Finally,Ihaveincludedmanynewexercises,manyofthemofferingawaytogo
furtherintothetheoryandpractice.Thewebsitementionedaboveremainsavailable,
andis expectedtogrow overtime,but thereferencesin itwillremain unchanged,
in order that it may be usable by the owners of the first edition, until I feel that it
absolutelyneedstoberefreshed.
Lyon,France DenisSerre
November2009
Preface for the First Edition
The study of matrices occupies a singular place within mathematics. It is still an
areaofactiveresearch,anditisusedbyeverymathematicianandbymanyscientists
workinginvariousspecialities.Severalexamplesillustrateitsversatility:
• Scientificcomputinglibrariesbegangrowingaroundmatrixcalculus.Asamatter
offact,thediscretizationofpartialdifferentialoperatorsisanendlesssourceof
linearfinite-dimensionalproblems.
• Atadiscretelevel,themaximumprincipleisrelatedtononnegativematrices.
• Controltheoryandstabilizationofsystemswithfinitelymanydegreesoffreedom
involvespectralanalysisofmatrices.
• ThediscreteFouriertransform,includingthefastFouriertransform,makesuse
ofToeplitzmatrices.
• Statisticsiswidelybasedoncorrelationmatrices.
• Thegeneralizedinverseisinvolvedinleast-squaresapproximation.
• Symmetric matrices are inertia, deformation, or viscous tensors in continuum
mechanics.
• Markovprocessesinvolvestochasticorbistochasticmatrices.
• Graphscanbedescribedinausefulwaybysquarematrices.
• Quantum chemistry is intimately related to matrix groups and their representa-
tions.
• The case of quantum mechanics is especially interesting. Observables are Her-
mitianoperators;theireigenvaluesareenergylevels.Intheearlyyears,quantum
mechanicswascalled“mechanicsofmatrices,”andithasnowgivenrisetothe
development of the theory of large random matrices. See [25] for a thorough
accountofthisfashionabletopic.
Thistextwasconceivedduringtheyears1998–2001,ontheoccasionofacourse
that I taught at the E´cole Normale Supe´rieure de Lyon. As such, every result is
accompanied by a detailed proof. During this course I tried to investigate all the
principalmathematicalaspectsofmatrices:algebraic,geometric,andanalytic.
In some sense, this is not a specialized book. For instance, it is not as detailed
as [19] concerning numerics, or as [40] on eigenvalue problems, or as [21] about
xiii