Eigenvaluesandeigenvectors

FromWikipedia,thefreeencyclopedia

Inlinearalgebra,aneigenvectororcharacteristicvectorofalineartransformationisanonzerovectorthatdoesnotchangeitsdirectionwhenthatlinear
transformationisappliedtoit.Moreformally,ifTisalineartransformationfromavectorspaceVoverafieldFintoitselfandvisavectorinVthatisnotthezero
vector,thenvisaneigenvectorofTifT(v)isascalarmultipleofv.Thisconditioncanbewrittenastheequation

whereisascalarinthefieldF,knownastheeigenvalue,characteristicvalue,orcharacteristicrootassociatedwiththeeigenvectorv.
IfthevectorspaceVisfinitedimensional,thenthelineartransformationTcanberepresentedasasquarematrixA,andthevectorvbyacolumnvector,rendering
theabovemappingasamatrixmultiplicationonthelefthandsideandascalingofthecolumnvectorontherighthandsideintheequation

Thereisacorrespondencebetweennbynsquarematricesandlineartransformationsfromanndimensionalvectorspacetoitself.Forthisreason,itisequivalent
todefineeigenvaluesandeigenvectorsusingeitherthelanguageofmatricesorthelanguageoflineartransformations.[1][2]
Geometrically,aneigenvectorcorrespondingtoareal,nonzeroeigenvaluepointsinadirectionthatisstretchedbythetransformationandtheeigenvalueisthe
factorbywhichitisstretched.Iftheeigenvalueisnegative,thedirectionisreversed.[3]

Contents
1 Overview
2 History
3 Eigenvaluesandeigenvectorsofmatrices
3.1 Eigenvaluesandthecharacteristicpolynomial
3.2 Algebraicmultiplicity
3.3 Eigenspaces,geometricmultiplicity,andtheeigenbasisformatrices
3.4 Additionalpropertiesofeigenvalues
3.5 Leftandrighteigenvectors
3.6 Diagonalizationandtheeigendecomposition
3.7 Variationalcharacterization
3.8 Matrixexamples
3.8.1 Twodimensionalmatrixexample
3.8.2 Threedimensionalmatrixexample

3.8.2 Threedimensionalmatrixexample
3.8.3 Threedimensionalmatrixexamplewithcomplexeigenvalues
3.8.4 Diagonalmatrixexample
3.8.5 Triangularmatrixexample
3.8.6 Matrixwithrepeatedeigenvaluesexample
4 Eigenvaluesandeigenfunctionsofdifferentialoperators
4.1 Derivativeoperatorexample
5 Generaldefinition
5.1 Eigenspaces,geometricmultiplicity,andtheeigenbasis
5.2 Zerovectorasaneigenvector
5.3 Spectraltheory
5.4 Associativealgebrasandrepresentationtheory
6 Dynamicequations
7 Calculation
7.1 Eigenvalues
7.2 Eigenvectors
8 Applications
8.1 Eigenvaluesofgeometrictransformations
8.2 Schrdingerequation
8.3 Molecularorbitals
8.4 Geologyandglaciology
8.5 Principalcomponentanalysis
8.6 Vibrationanalysis
8.7 Eigenfaces
8.8 Tensorofmomentofinertia
8.9 Stresstensor
8.10 Graphs
8.11 Basicreproductionnumber
9 Seealso
10 Notes
11 References
12 Externallinks
12.1 Theory
12.2 Demonstrationapplets

Overview

Eigenvaluesandeigenvectorsfeatureprominentlyintheanalysisoflineartransformations.TheprefixeigenisadoptedfromtheGermanwordeigenfor"proper",
"inherent""own","individual","special""specific","peculiar",or"characteristic".[4]Originallyutilizedtostudyprincipalaxesoftherotationalmotionofrigid
bodies,eigenvaluesandeigenvectorshaveawiderangeofapplications,forexampleinstabilityanalysis,vibrationanalysis,atomicorbitals,facialrecognition,and
matrixdiagonalization.
Inessence,aneigenvectorvofalineartransformationTisanonzerovectorthat,whenTisappliedtoit,doesnotchangedirection.ApplyingTtotheeigenvector
onlyscalestheeigenvectorbythescalarvalue,calledaneigenvalue.Thisconditioncanbewrittenastheequation

referredtoastheeigenvalueequationoreigenequation.Ingeneral,maybeanyscalar.Forexample,maybenegative,inwhichcasetheeigenvectorreverses
directionaspartofthescaling,oritmaybezeroorcomplex.
TheMonaLisaexamplepicturedatrightprovidesasimpleillustration.Eachpointonthepaintingcanbe
representedasavectorpointingfromthecenterofthepaintingtothatpoint.Thelineartransformationin
thisexampleiscalledashearmapping.Pointsinthetophalfaremovedtotherightandpointsinthebottom
halfaremovedtotheleftproportionaltohowfartheyarefromthehorizontalaxisthatgoesthroughthe
middleofthepainting.Thevectorspointingtoeachpointintheoriginalimagearethereforetiltedrightor
leftandmadelongerorshorterbythetransformation.Noticethatpointsalongthehorizontalaxisdonot
moveatallwhenthistransformationisapplied.Therefore,anyvectorthatpointsdirectlytotherightorleft
withnoverticalcomponentisaneigenvectorofthistransformationbecausethemappingdoesnotchange
itsdirection.Moreover,theseeigenvectorsallhaveaneigenvalueequaltoonebecausethemappingdoes
notchangetheirlength,either.
Lineartransformationscantakemanydifferentforms,mappingvectorsinavarietyofvectorspaces,sothe
eigenvectorscanalsotakemanyforms.Forexample,thelineartransformationcouldbeadifferential
operatorlike ,inwhichcasetheeigenvectorsarefunctionscalledeigenfunctionsthatarescaledbythat
differentialoperator,suchas

Inthisshearmappingtheredarrowchangesdirection
butthebluearrowdoesnot.Thebluearrowisan
eigenvectorofthisshearmappingbecauseitdoesn't
changedirection,andsinceitslengthisunchanged,its
eigenvalueis1.

Alternatively,thelineartransformationcouldtaketheformofannbynmatrix,inwhichcasetheeigenvectorsarenby1matricesthatarealsoreferredtoas
eigenvectors.IfthelineartransformationisexpressedintheformofannbynmatrixA,thentheeigenvalueequationaboveforalineartransformationcanbe
rewrittenasthematrixmultiplication

wheretheeigenvectorvisannby1matrix.Foramatrix,eigenvaluesandeigenvectorscanbeusedtodecomposethematrix,forexamplebydiagonalizingit.
Eigenvaluesandeigenvectorsgiverisetomanycloselyrelatedmathematicalconcepts,andtheprefixeigenisappliedliberallywhennamingthem:
Thesetofalleigenvectorsofalineartransformation,eachpairedwithitscorrespondingeigenvalue,iscalledtheeigensystemofthattransformation.[5][6]
ThesetofalleigenvectorsofTcorrespondingtothesameeigenvalue,togetherwiththezerovector,iscalledaneigenspaceorcharacteristicspaceof
T.[7][8]
IfthesetofeigenvectorsofTformabasisofthedomainofT,thenthisbasisiscalledaneigenbasis.

History
Eigenvaluesareoftenintroducedinthecontextoflinearalgebraormatrixtheory.Historically,however,theyaroseinthestudyofquadraticformsanddifferential
equations.
Inthe18thcenturyEulerstudiedtherotationalmotionofarigidbodyanddiscoveredtheimportanceoftheprincipalaxes.[9]Lagrangerealizedthattheprincipal
axesaretheeigenvectorsoftheinertiamatrix.[10]Intheearly19thcentury,Cauchysawhowtheirworkcouldbeusedtoclassifythequadricsurfaces,and
generalizedittoarbitrarydimensions.[11]Cauchyalsocoinedthetermracinecaractristique(characteristicroot)forwhatisnowcalledeigenvaluehisterm
survivesincharacteristicequation.[12][13]
FourierusedtheworkofLaplaceandLagrangetosolvetheheatequationbyseparationofvariablesinhisfamous1822bookThorieanalytiquedelachaleur.[14]
SturmdevelopedFourier'sideasfurtherandbroughtthemtotheattentionofCauchy,whocombinedthemwithhisownideasandarrivedatthefactthatreal
symmetricmatriceshaverealeigenvalues.[11]ThiswasextendedbyHermitein1855towhatarenowcalledHermitianmatrices.[12]Aroundthesametime,
Brioschiprovedthattheeigenvaluesoforthogonalmatriceslieontheunitcircle,[11]andClebschfoundthecorrespondingresultforskewsymmetricmatrices.[12]
Finally,WeierstrassclarifiedanimportantaspectinthestabilitytheorystartedbyLaplacebyrealizingthatdefectivematricescancauseinstability.[11]
Inthemeantime,LiouvillestudiedeigenvalueproblemssimilartothoseofSturmthedisciplinethatgrewoutoftheirworkisnowcalledSturmLiouville
theory.[15]SchwarzstudiedthefirsteigenvalueofLaplace'sequationongeneraldomainstowardstheendofthe19thcentury,whilePoincarstudiedPoisson's
equationafewyearslater.[16]
Atthestartofthe20thcentury,Hilbertstudiedtheeigenvaluesofintegraloperatorsbyviewingtheoperatorsasinfinitematrices.[17]Hewasthefirsttousethe
Germanwordeigen,whichmeans"own",todenoteeigenvaluesandeigenvectorsin1904,[18]thoughhemayhavebeenfollowingarelatedusagebyHelmholtz.
Forsometime,thestandardterminEnglishwas"propervalue",butthemoredistinctiveterm"eigenvalue"isstandardtoday.[19]
Thefirstnumericalalgorithmforcomputingeigenvaluesandeigenvectorsappearedin1929,whenVonMisespublishedthepowermethod.Oneofthemost
popularmethodstoday,theQRalgorithm,wasproposedindependentlybyJohnG.F.Francis[20]andVeraKublanovskaya[21]in1961.[22]

Eigenvaluesandeigenvectorsofmatrices
Eigenvaluesandeigenvectorsareoftenintroducedtostudentsinthecontextoflinearalgebracoursesfocusedonmatrices.[23][24]Furthermore,linear
transformationscanberepresentedusingmatrices,[1][2]whichisespeciallycommoninnumericalandcomputationalapplications.[25]
Considerndimensionalvectorsthatareformedasalistofnscalars,suchasthethreedimensionalvectors

Thesevectorsaresaidtobescalarmultiplesofeachother,orparallelorcollinear,ifthereisascalarsuchthat

Inthiscase=1/20.
NowconsiderthelineartransformationofndimensionalvectorsdefinedbyannbynmatrixA,

MatrixAactsbystretchingthevectorx,
notchangingitsdirection,soxisan
eigenvectorofA.

or

where,foreachrow,
.
Ifitoccursthatvandwarescalarmultiples,thatisif

(1)

thenvisaneigenvectorofthelineartransformationAandthescalefactoristheeigenvaluecorrespondingtothateigenvector.Equation(1)istheeigenvalue
equationforthematrixA.
Equation(1)canbestatedequivalentlyas

(2)

whereIisthenbynidentitymatrix.

Eigenvaluesandthecharacteristicpolynomial
Equation(2)hasanonzerosolutionvifandonlyifthedeterminantofthematrix(AI)iszero.Therefore,theeigenvaluesofAarevaluesofthatsatisfythe
equation

(3)

UsingLeibniz'ruleforthedeterminant,thelefthandsideofEquation(3)isapolynomialfunctionofthevariableandthedegreeofthispolynomialisn,the
orderofthematrixA.ItscoefficientsdependontheentriesofA,exceptthatitstermofdegreenisalways(1)nn.Thispolynomialiscalledthecharacteristic
polynomialofA.Equation(3)iscalledthecharacteristicequationorthesecularequationofA.
ThefundamentaltheoremofalgebraimpliesthatthecharacteristicpolynomialofannbynmatrixA,beingapolynomialofdegreen,canbefactoredintothe
productofnlinearterms,

whereeachimayberealbutingeneralisacomplexnumber.Thenumbers1,2,...n,whichmaynotallhavedistinctvalues,arerootsofthepolynomialand
aretheeigenvaluesofA.
Asabriefexample,whichisdescribedinmoredetailintheexamplessectionlater,considerthematrix

Takingthedeterminantof(MI),thecharacteristicpolynomialofMis

(4)

Settingthecharacteristicpolynomialequaltozero,ithasrootsat=1and=3,whicharethetwoeigenvaluesofM.Theeigenvectorscorrespondingtoeach
eigenvaluecanbefoundbysolvingforthecomponentsofvintheequationMv=v.Inthisexample,theeigenvectorsareanynonzeroscalarmultiplesof

IftheentriesofthematrixAareallrealnumbers,thenthecoefficientsofthecharacteristicpolynomialwillalsoberealnumbers,buttheeigenvaluesmaystill
havenonzeroimaginaryparts.Theentriesofthecorrespondingeigenvectorsthereforemayalsohavenonzeroimaginaryparts.Similarly,theeigenvaluesmaybe
irrationalnumbersevenifalltheentriesofAarerationalnumbersoreveniftheyareallintegers.However,iftheentriesofAareallalgebraicnumbers,which
includetherationals,theeigenvaluesarecomplexalgebraicnumbers.
Thenonrealrootsofarealpolynomialwithrealcoefficientscanbegroupedintopairsofcomplexconjugates,namelywiththetwomembersofeachpairhaving
imaginarypartsthatdifferonlyinsignandthesamerealpart.Ifthedegreeisodd,thenbytheintermediatevaluetheorematleastoneoftherootsisreal.
Therefore,anyrealmatrixwithoddorderhasatleastonerealeigenvalue,whereasarealmatrixwithevenordermaynothaveanyrealeigenvalues.The
eigenvectorsassociatedwiththesecomplexeigenvaluesarealsocomplexandalsoappearincomplexconjugatepairs.

Algebraicmultiplicity
LetibeaneigenvalueofannbynmatrixA.ThealgebraicmultiplicityA(i)oftheeigenvalueisitsmultiplicityasarootofthecharacteristicpolynomial,that
is,thelargestintegerksuchthat(i)kdividesevenlythatpolynomial.[8][26][27]
SupposeamatrixAhasdimensionnanddndistincteigenvalues.WhereasEquation(4)factorsthecharacteristicpolynomialofAintotheproductofnlinear
termswithsometermspotentiallyrepeating,thecharacteristicpolynomialcaninsteadbewrittenastheproductdtermseachcorrespondingtoadistincteigenvalue
andraisedtothepowerofthealgebraicmultiplicity,

Ifd=nthentherighthandsideistheproductofnlineartermsandthisisthesameasEquation(4).Thesizeofeacheigenvalue'salgebraicmultiplicityisrelated
tothedimensionnas

IfA(i)=1,theniissaidtobeasimpleeigenvalue.[27]IfA(i)equalsthegeometricmultiplicityofi,A(i),definedinthenextsection,theniissaidtobea
semisimpleeigenvalue.

Eigenspaces,geometricmultiplicity,andtheeigenbasisformatrices
GivenaparticulareigenvalueofthenbynmatrixA,definethesetEtobeallvectorsvthatsatisfyEquation(2),

Ononehand,thissetispreciselythekernelornullspaceofthematrix(AI).Ontheotherhand,bydefinition,anynonzerovectorthatsatisfiesthisconditionis
aneigenvectorofAassociatedwith.So,thesetEistheunionofthezerovectorwiththesetofalleigenvectorsofAassociatedwith,andEequalsthenullspace
of(AI).EiscalledtheeigenspaceorcharacteristicspaceofAassociatedwith.[7][8]Ingeneralisacomplexnumberandtheeigenvectorsarecomplexnby
1matrices.Apropertyofthenullspaceisthatitisalinearsubspace,soEisalinearsubspaceofn.
BecausetheeigenspaceEisalinearsubspace,itisclosedunderaddition.Thatis,iftwovectorsuandvbelongtothesetE,written(u,v)E,then(u+v)Eor
equivalentlyA(u+v)=(u+v).Thiscanbecheckedusingthedistributivepropertyofmatrixmultiplication.Similarly,becauseEisalinearsubspace,itisclosed
underscalarmultiplication.Thatis,ifvEandisacomplexnumber,(v)EorequivalentlyA(v)=(v).Thiscanbecheckedbynotingthatmultiplication
ofcomplexmatricesbycomplexnumbersiscommutative.Aslongasu+vandvarenotzero,theyarealsoeigenvectorsofAassociatedwith.
ThedimensionoftheeigenspaceEassociatedwith,orequivalentlythemaximumnumberoflinearlyindependenteigenvectorsassociatedwith,isreferredto
astheeigenvalue'sgeometricmultiplicityA().BecauseEisalsothenullspaceof(AI),thegeometricmultiplicityofisthedimensionofthenullspaceof(A
I),alsocalledthenullityof(AI),whichrelatestothedimensionandrankof(AI)as

Becauseofthedefinitionofeigenvaluesandeigenvectors,aneigenvalue'sgeometricmultiplicitymustbeatleastone,thatis,eacheigenvaluehasatleastone
associatedeigenvector.Furthermore,aneigenvalue'sgeometricmultiplicitycannotexceeditsalgebraicmultiplicity.Additionally,recallthataneigenvalue's
algebraicmultiplicitycannotexceedn.

TheconditionthatA()A()canbeprovenbyconsideringaparticulareigenvalueofAanddiagonalizingthefirstA()columnsofAwithrespectto's
eigenvectors,describedinalatersection.TheresultingsimilarmatrixBisblockuppertriangular,withitstopleftblockbeingthediagonalmatrixIA().Asa
result,thecharacteristicpolynomialofBwillhaveafactorof()A().TheotherfactorsofthecharacteristicpolynomialofBarenotknown,sothealgebraic
multiplicityofasaneigenvalueofBisnolessthanthegeometricmultiplicityofasaneigenvalueofA.Thelastelementoftheproofisthepropertythatsimilar
matriceshavethesamecharacteristicpolynomial.
SupposeAhasdndistincteigenvalues1,2,...,d,wherethegeometricmultiplicityofiisA(i).ThetotalgeometricmultiplicityofA,

isthedimensionoftheunionofalltheeigenspacesofA'seigenvalues,orequivalentlythemaximumnumberoflinearlyindependenteigenvectorsofA.IfA=n,
then
TheunionoftheeigenspacesofallofA'seigenvaluesistheentirevectorspacen
AbasisofncanbeformedfromnlinearlyindependenteigenvectorsofAsuchabasisiscalledaneigenbasis
AnyvectorinncanbewrittenasalinearcombinationofeigenvectorsofA

Additionalpropertiesofeigenvalues
LetAbeanarbitrarynbynmatrixofcomplexnumberswitheigenvalues1,2,...,n.EacheigenvalueappearsA(i)timesinthislist,whereA(i)isthe
eigenvalue'salgebraicmultiplicity.Thefollowingarepropertiesofthismatrixanditseigenvalues:
ThetraceofA,definedasthesumofitsdiagonalelements,isalsothesumofalleigenvalues,
[28][29][30]

ThedeterminantofAistheproductofallitseigenvalues,
[28][31][32]

TheeigenvaluesofthekthpowerofA,i.e.theeigenvaluesofAk,foranypositiveintegerk,are1k,2k,...,nk.
ThematrixAisinvertibleifandonlyifeveryeigenvalueisnonzero.
IfAisinvertible,thentheeigenvaluesofA1are1/1,1/2,...,1/nandeacheigenvalue'sgeometricmultiplicitycoincides.Moreover,sincethe
characteristicpolynomialoftheinverseisthereciprocalpolynomialoftheoriginal,theeigenvaluessharethesamealgebraicmultiplicity.
IfAisequaltoitsconjugatetransposeA*,orequivalentlyifAisHermitian,theneveryeigenvalueisreal.Thesameistrueofanysymmetricrealmatrix.
IfAisnotonlyHermitianbutalsopositivedefinite,positivesemidefinite,negativedefinite,ornegativesemidefinite,theneveryeigenvalueispositive,
nonnegative,negative,ornonpositive,respectively.
IfAisunitary,everyeigenvaluehasabsolutevalue|i|=1.

Leftandrighteigenvectors
Manydisciplinestraditionallyrepresentvectorsasmatriceswithasinglecolumnratherthanasmatriceswithasinglerow.Forthatreason,theword"eigenvector"
inthecontextofmatricesalmostalwaysreferstoarighteigenvector,namelyacolumnvectorthatrightmultiplesthenbynmatrixAinthedefiningequation,
Equation(1),

TheeigenvalueandeigenvectorproblemcanalsobedefinedforrowvectorsthatleftmultiplymatrixA.Inthisformulation,thedefiningequationis

whereisascalaranduisa1bynmatrix.AnyrowvectorusatisfyingthisequationiscalledalefteigenvectorofAandisitsassociatedeigenvalue.Takingthe
conjugatetransposeofthisequation,

ComparingthisequationtoEquation(1),thelefteigenvectorsofAaretheconjugatetransposeoftherighteigenvectorsofA*.Theeigenvaluesoftheleft
eigenvectorsarethesolutionofthecharacteristicpolynomial|A**I|=0.BecausetheidentitymatrixisHermitianand|M*|=|M|*forasquarematrixM,the
eigenvaluesofthelefteigenvectorsofAarethecomplexconjugatesoftheeigenvaluesoftherighteigenvectorsofA.RecallthatifAisarealmatrix,allofits
complexeigenvaluesappearincomplexconjugatepairs.Therefore,theeigenvaluesoftheleftandrighteigenvectorsofarealmatrixarethesame.Similarly,ifA
isarealmatrix,allofitscomplexeigenvectorsalsoappearincomplexconjugatepairs.Therefore,thelefteigenvectorssimplifytothetransposeoftheright
eigenvectorsofATifAisreal.

Diagonalizationandtheeigendecomposition
SupposetheeigenvectorsofAformabasis,orequivalentlyAhasnlinearlyindependenteigenvectorsv1,v2,...,vnwithassociatedeigenvalues1,2,...,n.The
eigenvaluesneednotbedistinct.DefineasquarematrixQwhosecolumnsarethenlinearlyindependenteigenvectorsofA,

SinceeachcolumnofQisaneigenvectorofA,rightmultiplyingAbyQscaleseachcolumnofQbyitsassociatedeigenvalue,

Withthisinmind,defineadiagonalmatrixwhereeachdiagonalelementiiistheeigenvalueassociatedwiththeithcolumnofQ.Then

BecausethecolumnsofQarelinearlyindependent,Qisinvertible.RightmultiplyingbothsidesoftheequationbyQ1,

orbyinsteadleftmultiplyingbothsidesbyQ1,

Acanthereforebedecomposedintoamatrixcomposedofitseigenvectors,adiagonalmatrixwithitseigenvaluesalongthediagonal,andtheinverseofthematrix
ofeigenvectors.Thisiscalledtheeigendecompositionanditisasimilaritytransformation.SuchamatrixAissaidtobesimilartothediagonalmatrixor
diagonalizable.ThematrixQisthechangeofbasismatrixofthesimilaritytransformation.Essentially,thematricesAandrepresentthesamelinear
transformationexpressedintwodifferentbases.Theeigenvectorsareusedasthebasiswhenrepresentingthelineartransformationas.
Conversely,supposeamatrixAisdiagonalizable.LetPbeanonsingularsquarematrixsuchthatP1APissomediagonalmatrixD.LeftmultiplyingbothbyP,
AP=PD.EachcolumnofPmustthereforebeaneigenvectorofAwhoseeigenvalueisthecorrespondingdiagonalelementofD.SincethecolumnsofPmustbe
linearlyindependentforPtobeinvertible,thereexistnlinearlyindependenteigenvectorsofA.ItthenfollowsthattheeigenvectorsofAformabasisifandonlyif
Aisdiagonalizable.
Amatrixthatisnotdiagonalizableissaidtobedefective.Fordefectivematrices,thenotionofeigenvectorsgeneralizestogeneralizedeigenvectorsandthe
diagonalmatrixofeigenvaluesgeneralizestotheJordannormalform.Overanalgebraicallyclosedfield,anymatrixAhasaJordannormalformandtherefore
admitsabasisofgeneralizedeigenvectorsandadecompositionintogeneralizedeigenspaces.

Variationalcharacterization
IntheHermitiancase,eigenvaluescanbegivenavariationalcharacterization.Thelargesteigenvalueof isthemaximumvalueofthequadraticform
.Avalueof thatrealizesthatmaximum,isaneigenvector.

Matrixexamples
Twodimensionalmatrixexample
Considerthematrix

Thefigureontherightshowstheeffectofthistransformationonpointcoordinatesintheplane.The
eigenvectorsvofthistransformationsatisfyEquation(1),andthevaluesofforwhichthedeterminantofthe
matrix(AI)equalszeroaretheeigenvalues.
TakingthedeterminanttofindcharacteristicpolynomialofA,

Settingthecharacteristicpolynomialequaltozero,ithasrootsat=1and=3,whicharethetwo
eigenvaluesofA.
For=1,Equation(2)becomes,

ThetransformationmatrixA=

Anynonzerovectorwithv1=v2solvesthisequation.Therefore,

preservesthe

directionofvectorsparalleltov=1=[11]T(in
purple)andv=3=[11]T(inblue).Thevectorsin

isaneigenvectorofAcorrespondingto=1,asisanyscalarmultipleofthisvector.
For=3,Equation(2)becomes

Anynonzerovectorwithv1=v2solvesthisequation.Therefore,

isaneigenvectorofAcorrespondingto=3,asisanyscalarmultipleofthisvector.

redarenotparalleltoeithereigenvector,so,their
directionsarechangedbythetransformation.See
also:Anextendedversion,showingallfour
quadrants.

Thus,thevectorsv=1andv=3areeigenvectorsofAassociatedwiththeeigenvalues=1and=3,respectively.
Threedimensionalmatrixexample
Considerthematrix

ThecharacteristicpolynomialofAis

Therootsofthecharacteristicpolynomialare2,1,and11,whicharetheonlythreeeigenvaluesofA.Theseeigenvaluescorrespondtotheeigenvectors
and
,oranynonzeromultiplethereof.

Threedimensionalmatrixexamplewithcomplexeigenvalues
Considerthecyclicpermutationmatrix

Thismatrixshiftsthecoordinatesofthevectorupbyonepositionandmovesthefirstcoordinatetothebottom.Itscharacteristicpolynomialis13,whoseroots
are

wherei=

istheimaginaryunit.

wherei=

istheimaginaryunit.

Fortherealeigenvalue1=1,anyvectorwiththreeequalnonzeroentriesisaneigenvector.Forexample,

Forthecomplexconjugatepairofimaginaryeigenvalues,notethat

Then

and

Therefore,theothertwoeigenvectorsofAarecomplexandare

and

witheigenvalues2and3,respectively.Notethatthe

twocomplexeigenvectorsalsoappearinacomplexconjugatepair,

Diagonalmatrixexample
Matriceswithentriesonlyalongthemaindiagonalarecalleddiagonalmatrices.Theeigenvaluesofadiagonalmatrixarethediagonalelementsthemselves.
Considerthematrix

ThecharacteristicpolynomialofAis

whichhastheroots1=1,2=2,and3=3.TheserootsarethediagonalelementsaswellastheeigenvaluesofA.
Eachdiagonalelementcorrespondstoaneigenvectorwhoseonlynonzerocomponentisinthesamerowasthatdiagonalelement.Intheexample,theeigenvalues
correspondtotheeigenvectors,

respectively,aswellasscalarmultiplesofthesevectors.
Triangularmatrixexample
Amatrixwhoseelementsabovethemaindiagonalareallzeroiscalledalowertriangularmatrix,whileamatrixwhoseelementsbelowthemaindiagonalareall
zeroiscalledanuppertriangularmatrix.Aswithdiagonalmatrices,theeigenvaluesoftriangularmatricesaretheelementsofthemaindiagonal.
Considerthelowertriangularmatrix,

ThecharacteristicpolynomialofAis

whichhastheroots1=1,2=2,and3=3.TheserootsarethediagonalelementsaswellastheeigenvaluesofA.
Theseeigenvaluescorrespondtotheeigenvectors,

respectively,aswellasscalarmultiplesofthesevectors.
Matrixwithrepeatedeigenvaluesexample
Asinthepreviousexample,thelowertriangularmatrix

hasacharacteristicpolynomialthatistheproductofitsdiagonalelements,

Therootsofthispolynomial,andhencetheeigenvalues,are2and3.Thealgebraicmultiplicityofeacheigenvalueis2inotherwordstheyarebothdoubleroots.
ThesumofthealgebraicmultiplicitiesofeachdistincteigenvalueisA=4=n,theorderofthecharacteristicpolynomialandthedimensionofA.
Ontheotherhand,thegeometricmultiplicityoftheeigenvalue2isonly1,becauseitseigenspaceisspannedbyjustonevector[0111]Tandistherefore1
dimensional.Similarly,thegeometricmultiplicityoftheeigenvalue3is1becauseitseigenspaceisspannedbyjustonevector[0001]T.Thetotalgeometric
multiplicityAis2,whichisthesmallestitcouldbeforamatrixwithtwodistincteigenvalues.Geometricmultiplicitiesaredefinedinalatersection.

Eigenvaluesandeigenfunctionsofdifferentialoperators
ThedefinitionsofeigenvalueandeigenvectorsofalineartransformationTremainsvalideveniftheunderlyingvectorspaceisaninfinitedimensionalHilbertor
Banachspace.Awidelyusedclassoflineartransformationsactingoninfinitedimensionalspacesarethedifferentialoperatorsonfunctionspaces.LetDbea
lineardifferentialoperatoronthespaceCofinfinitelydifferentiablerealfunctionsofarealargumentt.TheeigenvalueequationforDisthedifferentialequation

ThefunctionsthatsatisfythisequationareeigenvectorsofDandarecommonlycalledeigenfunctions.

Derivativeoperatorexample
Considerthederivativeoperator

witheigenvalueequation

Thisdifferentialequationcanbesolvedbymultiplyingbothsidesbydt/f(t)andintegrating.Itssolution,theexponentialfunction

istheeigenfunctionofthederivativeoperator.Notethatinthiscasetheeigenfunctionisitselfafunctionofitsassociatedeigenvalue.Inparticular,notethatfor=
0theeigenfunctionf(t)isaconstant.
Themaineigenfunctionarticlegivesotherexamples.

Generaldefinition
Theconceptofeigenvaluesandeigenvectorsextendsnaturallytoarbitrarylineartransformationsonarbitraryvectorspaces.LetVbeanyvectorspaceoversome
fieldKofscalars,andletTbealineartransformationmappingVintoV,

WesaythatanonzerovectorvVisaneigenvectorofTifandonlyifthereexistsascalarKsuchthat

ThisequationiscalledtheeigenvalueequationforT,andthescalaristheeigenvalueofTcorrespondingtotheeigenvectorv.NotethatT(v)istheresultof
applyingthetransformationTtothevectorv,whilevistheproductofthescalarwithv.[33]

Eigenspaces,geometricmultiplicity,andtheeigenbasis
Givenaneigenvalue,considertheset

(5)

whichistheunionofthezerovectorwiththesetofalleigenvectorsassociatedwith.EiscalledtheeigenspaceorcharacteristicspaceofTassociatedwith.
Bydefinitionofalineartransformation,

for(x,y)VandK.Therefore,ifuandvareeigenvectorsofTassociatedwitheigenvalue,namely(u,v)E,then

So,bothu+vandvareeitherzerooreigenvectorsofTassociatedwith,namely(u+v,v)E,andEisclosedunderadditionandscalarmultiplication.The
eigenspaceEassociatedwithisthereforealinearsubspaceofV.[8][34][35]Ifthatsubspacehasdimension1,itissometimescalledaneigenline.[36]
ThegeometricmultiplicityT()ofaneigenvalueisthedimensionoftheeigenspaceassociatedwith,i.e.,themaximumnumberoflinearlyindependent
eigenvectorsassociatedwiththateigenvalue.[8][27]Bythedefinitionofeigenvaluesandeigenvectors,T()1becauseeveryeigenvaluehasatleastone
eigenvector.
TheeigenspacesofTalwaysformadirectsum.Asaconsequence,eigenvectorsofdifferenteigenvaluesarealwayslinearlyindependent.Therefore,thesumofthe
dimensionsoftheeigenspacescannotexceedthedimensionnofthevectorspaceonwhichToperates,andtherecannotbemorethanndistincteigenvalues.[37]
AnysubspacespannedbyeigenvectorsofTisaninvariantsubspaceofT,andtherestrictionofTtosuchasubspaceisdiagonalizable.Moreover,iftheentire
vectorspaceVcanbespannedbytheeigenvectorsofT,orequivalentlyifthedirectsumoftheeigenspacesassociatedwithalltheeigenvaluesofTistheentire
vectorspaceV,thenabasisofVcalledaneigenbasiscanbeformedfromlinearlyindependenteigenvectorsofT.WhenTadmitsaneigenbasis,Tis
diagonalizable.

Zerovectorasaneigenvector
Whilethedefinitionofaneigenvectorusedinthisarticleexcludesthezerovector,itispossibletodefineeigenvaluesandeigenvectorssuchthatthezerovectoris
aneigenvector.[38]
Consideragaintheeigenvalueequation,Equation(5).DefineaneigenvaluetobeanyscalarKsuchthatthereexistsanonzerovectorvVsatisfying
Equation(5).Itisimportantthatthisversionofthedefinitionofaneigenvaluespecifythatthevectorbenonzero,otherwisebythisdefinitionthezerovector
wouldallowanyscalarinKtobeaneigenvalue.Defineaneigenvectorvassociatedwiththeeigenvaluetobeanyvectorthat,given,satisfiesEquation(5).

Giventheeigenvalue,thezerovectorisamongthevectorsthatsatisfyEquation(5),sothezerovectorisincludedamongtheeigenvectorsbythisalternate
definition.

Spectraltheory
IfisaneigenvalueofT,thentheoperator(TI)isnotonetoone,andthereforeitsinverse(TI)1doesnotexist.Theconverseistrueforfinitedimensional
vectorspaces,butnotforinfinitedimensionalvectorspaces.Ingeneral,theoperator(TI)maynothaveaninverseevenifisnotaneigenvalue.
Forthisreason,infunctionalanalysiseigenvaluescanbegeneralizedtothespectrumofalinearoperatorTasthesetofallscalarsforwhichtheoperator(TI)
hasnoboundedinverse.Thespectrumofanoperatoralwayscontainsallitseigenvaluesbutisnotlimitedtothem.

Associativealgebrasandrepresentationtheory
Onecangeneralizethealgebraicobjectthatisactingonthevectorspace,replacingasingleoperatoractingonavectorspacewithanalgebrarepresentationan
associativealgebraactingonamodule.Thestudyofsuchactionsisthefieldofrepresentationtheory.
Therepresentationtheoreticalconceptofweightisananalogofeigenvalues,whileweightvectorsandweightspacesaretheanalogsofeigenvectorsand
eigenspaces,respectively.

Dynamicequations
Thesimplestdifferenceequationshavetheform

Thesolutionofthisequationforxintermsoftisfoundbyusingitscharacteristicequation

whichcanbefoundbystackingintomatrixformasetofequationsconsistingoftheabovedifferenceequationandthek1equations
givingakdimensionalsystemofthefirstorderinthestackedvariablevector
intermsofitsoncelagged
value,andtakingthecharacteristicequationofthissystem'smatrix.Thisequationgiveskcharacteristicroots
foruseinthesolutionequation

Asimilarprocedureisusedforsolvingadifferentialequationoftheform

Calculation
Eigenvalues
Theeigenvaluesofamatrix canbedeterminedbyfindingtherootsofthecharacteristicpolynomial.Explicitalgebraicformulasfortherootsofapolynomial
existonlyifthedegree is4orless.AccordingtotheAbelRuffinitheoremthereisnogeneral,explicitandexactalgebraicformulafortherootsofapolynomial
withdegree5ormore.
Itturnsoutthatanypolynomialwithdegree isthecharacteristicpolynomialofsomecompanionmatrixoforder .Therefore,formatricesoforder5ormore,
theeigenvaluesandeigenvectorscannotbeobtainedbyanexplicitalgebraicformula,andmustthereforebecomputedbyapproximatenumericalmethods.
Intheory,thecoefficientsofthecharacteristicpolynomialcanbecomputedexactly,sincetheyaresumsofproductsofmatrixelementsandtherearealgorithms
thatcanfindalltherootsofapolynomialofarbitrarydegreetoanyrequiredaccuracy.[39]However,thisapproachisnotviableinpracticebecausethecoefficients
wouldbecontaminatedbyunavoidableroundofferrors,andtherootsofapolynomialcanbeanextremelysensitivefunctionofthecoefficients(asexemplifiedby
Wilkinson'spolynomial).[39]
Efficient,accuratemethodstocomputeeigenvaluesandeigenvectorsofarbitrarymatriceswerenotknownuntiltheadventoftheQRalgorithmin1961.[39]
CombiningtheHouseholdertransformationwiththeLUdecompositionresultsinanalgorithmwithbetterconvergencethantheQRalgorithm.Forlarge
Hermitiansparsematrices,theLanczosalgorithmisoneexampleofanefficientiterativemethodtocomputeeigenvaluesandeigenvectors,amongseveralother
possibilities.[39]

Eigenvectors
Oncethe(exact)valueofaneigenvalueisknown,thecorrespondingeigenvectorscanbefoundbyfindingnonzerosolutionsoftheeigenvalueequation,that
becomesasystemoflinearequationswithknowncoefficients.Forexample,onceitisknownthat6isaneigenvalueofthematrix

wecanfinditseigenvectorsbysolvingtheequation

,thatis

Thismatrixequationisequivalenttotwolinearequations
thatis
Bothequationsreducetothesinglelinearequation
eigenvalue
.

.Therefore,anyvectoroftheform

,foranynonzerorealnumber ,isaneigenvectorof with

Thematrix abovehasanothereigenvalue
.Asimilarcalculationshowsthatthecorrespondingeigenvectorsarethenonzerosolutionsof
is,anyvectoroftheform
,foranynonzerorealnumber .
Somenumericmethodsthatcomputetheeigenvaluesofamatrixalsodetermineasetofcorrespondingeigenvectorsasabyproductofthecomputation.

Applications
Eigenvaluesofgeometrictransformations
Thefollowingtablepresentssomeexampletransformationsintheplanealongwiththeir22matrices,eigenvalues,andeigenvectors.

,that

scaling

unequalscaling

rotation

horizontalshear hyperbolicrotation

illustration

matrix

characteristic
polynomial
eigenvalues

algebraicmultipl.
geometricmultipl.

eigenvectors

Allnonzerovectors

Notethatthecharacteristicequationforarotationisaquadraticequationwithdiscriminant
,whichisanegativenumberwhenever isnotan
integermultipleof180.Therefore,exceptforthesespecialcases,thetwoeigenvaluesarecomplexnumbers,
andalleigenvectorshavenonreal
entries.Indeed,exceptforthosespecialcases,arotationchangesthedirectionofeverynonzerovectorintheplane.
Alineartransformationthattakesasquaretoarectangleofthesamearea(asqueezemapping)hasreciprocaleigenvalues.

Schrdingerequation
Anexampleofaneigenvalueequationwherethetransformation isrepresentedintermsofadifferentialoperatoristhetimeindependentSchrdingerequation
inquantummechanics:

where ,theHamiltonian,isasecondorderdifferentialoperatorand ,thewavefunction,isoneofits

eigenfunctionscorrespondingtotheeigenvalue ,interpretedasitsenergy.
However,inthecasewhereoneisinterestedonlyintheboundstatesolutionsoftheSchrdingerequation,one
looksfor withinthespaceofsquareintegrablefunctions.SincethisspaceisaHilbertspacewithawelldefined
scalarproduct,onecanintroduceabasissetinwhich and canberepresentedasaonedimensionalarray
(i.e.,avector)andamatrixrespectively.ThisallowsonetorepresenttheSchrdingerequationinamatrixform.
Thebraketnotationisoftenusedinthiscontext.Avector,whichrepresentsastateofthesystem,intheHilbert
spaceofsquareintegrablefunctionsisrepresentedby
.Inthisnotation,theSchrdingerequationis:

where
isaneigenstateof and representstheeigenvalue. isanobservableselfadjointoperator,the
infinitedimensionalanalogofHermitianmatrices.Asinthematrixcase,intheequationabove
is
understoodtobethevectorobtainedbyapplicationofthetransformation to
.

Molecularorbitals
Inquantummechanics,andinparticularinatomicandmolecularphysics,withintheHartreeFocktheory,the
atomicandmolecularorbitalscanbedefinedbytheeigenvectorsoftheFockoperator.Thecorresponding
eigenvaluesareinterpretedasionizationpotentialsviaKoopmans'theorem.Inthiscase,thetermeigenvectoris
usedinasomewhatmoregeneralmeaning,sincetheFockoperatorisexplicitlydependentontheorbitalsandtheir
eigenvalues.Thus,ifonewantstounderlinethisaspect,onespeaksofnonlineareigenvalueproblems.Such
equationsareusuallysolvedbyaniterationprocedure,calledinthiscaseselfconsistentfieldmethod.Inquantum
chemistry,oneoftenrepresentstheHartreeFockequationinanonorthogonalbasisset.Thisparticular
representationisageneralizedeigenvalueproblemcalledRoothaanequations.

Thewavefunctionsassociatedwiththebound
statesofanelectroninahydrogenatomcanbe
seenastheeigenvectorsofthehydrogenatom
Hamiltonianaswellasoftheangular
momentumoperator.Theyareassociatedwith
eigenvaluesinterpretedastheirenergies
(increasingdownward:
)and
angularmomentum(increasingacross:s,p,d,
...).Theillustrationshowsthesquareofthe
absolutevalueofthewavefunctions.Brighter
areascorrespondtohigherprobabilitydensity
forapositionmeasurement.Thecenterofeach
figureistheatomicnucleus,aproton.

Geologyandglaciology
Ingeology,especiallyinthestudyofglacialtill,eigenvectorsandeigenvaluesareusedasamethodbywhichamassofinformationofaclastfabric'sconstituents'
orientationanddipcanbesummarizedina3Dspacebysixnumbers.Inthefield,ageologistmaycollectsuchdataforhundredsorthousandsofclastsinasoil
sample,whichcanonlybecomparedgraphicallysuchasinaTriPlot(SneedandFolk)diagram,[40][41]orasaStereonetonaWulffNet.[42]

Theoutputfortheorientationtensorisinthethreeorthogonal(perpendicular)axesofspace.Thethreeeigenvectorsareordered

bytheireigenvalues

[43]

thenistheprimaryorientation/dipofclast, isthesecondaryand isthetertiary,intermsofstrength.Theclastorientationisdefinedas

thedirectionoftheeigenvector,onacompassroseof360.Dipismeasuredastheeigenvalue,themodulusofthetensor:thisisvaluedfrom0(nodip)to90
(vertical).Therelativevaluesof , ,and aredictatedbythenatureofthesediment'sfabric.If
,thefabricissaidtobeisotropic.If
,thefabricissaidtobeplanar.If

,thefabricissaidtobelinear.[44]

Principalcomponentanalysis
Theeigendecompositionofasymmetricpositivesemidefinite(PSD)matrixyieldsanorthogonalbasisofeigenvectors,
eachofwhichhasanonnegativeeigenvalue.TheorthogonaldecompositionofaPSDmatrixisusedinmultivariate
analysis,wherethesamplecovariancematricesarePSD.Thisorthogonaldecompositioniscalledprincipalcomponents
analysis(PCA)instatistics.PCAstudieslinearrelationsamongvariables.PCAisperformedonthecovariancematrixor
thecorrelationmatrix(inwhicheachvariableisscaledtohaveitssamplevarianceequaltoone).Forthecovarianceor
correlationmatrix,theeigenvectorscorrespondtoprincipalcomponentsandtheeigenvaluestothevarianceexplainedby
theprincipalcomponents.Principalcomponentanalysisofthecorrelationmatrixprovidesanorthonormaleigenbasisfor
thespaceoftheobserveddata:Inthisbasis,thelargesteigenvaluescorrespondtotheprincipalcomponentsthatare
associatedwithmostofthecovariabilityamonganumberofobserveddata.
Principalcomponentanalysisisusedtostudylargedatasets,suchasthoseencounteredinbioinformatics,datamining,
chemicalresearch,psychology,andinmarketing.PCAispopularespeciallyinpsychology,inthefieldofpsychometrics.In
Qmethodology,theeigenvaluesofthecorrelationmatrixdeterminetheQmethodologist'sjudgmentofpractical
significance(whichdiffersfromthestatisticalsignificanceofhypothesistestingcf.criteriafordeterminingthenumberof
factors).Moregenerally,principalcomponentanalysiscanbeusedasamethodoffactoranalysisinstructuralequation
modeling.

Vibrationanalysis
Eigenvalueproblemsoccurnaturallyinthevibrationanalysisofmechanicalstructureswithmanydegreesoffreedom.The
eigenvaluesarethenaturalfrequencies(oreigenfrequencies)ofvibration,andtheeigenvectorsaretheshapesofthese
vibrationalmodes.Inparticular,undampedvibrationisgovernedby

or

PCAofthemultivariateGaussian
distributioncenteredat
witha
standarddeviationof3inroughlythe
directionandof1in
theorthogonaldirection.Thevectors
shownareuniteigenvectorsofthe
(symmetric,positivesemidefinite)
covariancematrixscaledbythe
squarerootofthecorresponding
eigenvalue.(Justasintheone
dimensionalcase,thesquarerootis
takenbecausethestandarddeviation
ismorereadilyvisualizedthanthe
variance.

thatis,accelerationisproportionaltoposition(i.e.,weexpect tobesinusoidalintime).
In dimensions, becomesamassmatrixand astiffnessmatrix.Admissiblesolutionsarethenalinearcombinationof
solutionstothegeneralizedeigenvalueproblem

where istheeigenvalueand isthe(imaginary)angularfrequency.Notethattheprincipalvibrationmodesaredifferent

fromtheprincipalcompliancemodes,whicharetheeigenvectorsof alone.Furthermore,dampedvibration,governedby

leadstoasocalledquadraticeigenvalueproblem,

ModeShapeofaTuningForkat
Eigenfrequency440.09Hz

Thiscanbereducedtoageneralizedeigenvalueproblembycleveruseofalgebraatthecostofsolvingalargersystem.
Theorthogonalitypropertiesoftheeigenvectorsallowsdecouplingofthedifferentialequationssothatthesystemcanberepresentedaslinearsummationofthe
eigenvectors.Theeigenvalueproblemofcomplexstructuresisoftensolvedusingfiniteelementanalysis,butneatlygeneralizethesolutiontoscalarvalued
vibrationproblems.

Eigenfaces
Inimageprocessing,processedimagesoffacescanbeseenasvectorswhosecomponentsarethebrightnessesofeachpixel.[45]Thedimensionofthisvectorspace
isthenumberofpixels.Theeigenvectorsofthecovariancematrixassociatedwithalargesetofnormalizedpicturesoffacesarecalledeigenfacesthisisan
exampleofprincipalcomponentsanalysis.Theyareveryusefulforexpressinganyfaceimageasalinearcombinationofsomeofthem.Inthefacialrecognition
branchofbiometrics,eigenfacesprovideameansofapplyingdatacompressiontofacesforidentificationpurposes.Researchrelatedtoeigenvisionsystems
determininghandgestureshasalsobeenmade.
Similartothisconcept,eigenvoicesrepresentthegeneraldirectionofvariabilityinhumanpronunciationsofaparticularutterance,suchasawordinalanguage.
Basedonalinearcombinationofsucheigenvoices,anewvoicepronunciationofthewordcanbeconstructed.Theseconceptshavebeenfoundusefulinautomatic
speechrecognitionsystemsforspeakeradaptation.

Tensorofmomentofinertia
Inmechanics,theeigenvectorsofthemomentofinertiatensordefinetheprincipalaxesofarigidbody.Thetensorofmomentofinertiaisakeyquantityrequired
todeterminetherotationofarigidbodyarounditscenterofmass.

Stresstensor
Insolidmechanics,thestresstensorissymmetricandsocanbedecomposedintoadiagonaltensorwiththeeigenvaluesonthe
diagonalandeigenvectorsasabasis.Becauseitisdiagonal,inthisorientation,thestresstensorhasnoshearcomponentsthe
componentsitdoeshavearetheprincipalcomponents.

Graphs
Inspectralgraphtheory,aneigenvalueofagraphisdefinedasaneigenvalueofthegraph'sadjacencymatrix ,or
(increasingly)ofthegraph'sLaplacianmatrixduetoitsDiscreteLaplaceoperator,whichiseither
(sometimescalled
thecombinatorialLaplacian)or
(sometimescalledthenormalizedLaplacian),where isadiagonal
matrixwith

equaltothedegreeofvertex ,andin

,the thdiagonalentryis

.The thprincipal

eigenvectorofagraphisdefinedaseithertheeigenvectorcorrespondingtothe thlargestor thsmallesteigenvalueofthe

Laplacian.Thefirstprincipaleigenvectorofthegraphisalsoreferredtomerelyastheprincipaleigenvector.

Eigenfacesasexamplesof
eigenvectors

Theprincipaleigenvectorisusedtomeasurethecentralityofitsvertices.AnexampleisGoogle'sPageRankalgorithm.The
principaleigenvectorofamodifiedadjacencymatrixoftheWorldWideWebgraphgivesthepageranksasitscomponents.Thisvectorcorrespondstothe
stationarydistributionoftheMarkovchainrepresentedbytherownormalizedadjacencymatrixhowever,theadjacencymatrixmustfirstbemodifiedtoensurea
stationarydistributionexists.Thesecondsmallesteigenvectorcanbeusedtopartitionthegraphintoclusters,viaspectralclustering.Othermethodsarealso
availableforclustering.

Basicreproductionnumber
Thebasicreproductionnumber( )isafundamentalnumberinthestudyofhowinfectiousdiseasesspread.Ifoneinfectiouspersonisputintoapopulationof
completelysusceptiblepeople,then istheaveragenumberofpeoplethatonetypicalinfectiouspersonwillinfect.Thegenerationtimeofaninfectionisthe
time, ,fromonepersonbecominginfectedtothenextpersonbecominginfected.Inaheterogeneouspopulation,thenextgenerationmatrixdefineshowmany
peopleinthepopulationwillbecomeinfectedaftertime

Seealso
Antieigenvaluetheory
Eigenplane
Eigenvaluealgorithm
Introductiontoeigenstates
Jordannormalform

haspassed.

isthenthelargesteigenvalueofthenextgenerationmatrix.[46][47]

Listofnumericalanalysissoftware
Nonlineareigenproblem
Quadraticeigenvalueproblem
Singularvalue

Notes
1.Herstein(1964,pp.228,229)
2.Nering(1970,p.38)
3.Burden&Faires(1993,p.401)
4.Betteridge(1965)
5.Press(2007,p.536)
6.WolframResearch,Inc.(2010)Eigenvector(http://mathworld.wolfram.com/Eigen
vector.html).Accessedon20160401.
7.Anton(1987,pp.305,307)
8.Nering(1970,p.107)

9.Note:
In1751,LeonhardEulerprovedthatanybodyhasaprincipalaxisof
rotation:LeonhardEuler(presented:October1751published:1760)"Du
mouvementd'uncorpssolidequelconquelorsqu'iltourneautourd'unaxe
mobile"(https://archive.org/stream/histoiredelacad07unkngoog#page/n196/
mode/2up)(Onthemovementofanysolidbodywhileitrotatesarounda
movingaxis),Histoiredel'Acadmieroyaledessciencesetdesbelles
lettresdeBerlin,pp.176227.Onp.212(https://archive.org/stream/histoire
delacad07unkngoog#page/n232/mode/2up),Eulerprovesthatanybody
containsaprincipalaxisofrotation:"Thorem.44.Dequelquefigureque
soitlecorps,onypeuttoujoursassigneruntelaxe,quipasseparson
centredegravit,autourduquellecorpspeuttournerlibrement&d'un
mouvementuniforme."(Theorem.44.Whateverbetheshapeofthebody,
onecanalwaysassigntoitsuchanaxis,whichpassesthroughitscenterof
gravity,aroundwhichitcanrotatefreelyandwithauniformmotion.)
In1755,JohannAndreasSegnerprovedthatanybodyhasthreeprincipal
axesofrotation:JohannAndreasSegner,Specimentheoriaeturbinum
[Essayonthetheoryoftops(i.e.,rotatingbodies)](Halle("Halae"),
(Germany):Gebauer,1755).Onp.XXVIIII(i.e.,29)(https://books.googl
e.com/books?id=89NMAAAAcAAJ&pg=PR29#v=onepage&q&f=false),
Segnerderivesathirddegreeequationint,whichprovesthatabodyhas
threeprincipalaxesofrotation.Hethenstates(onthesamepage):"Non
autemrepugnattresesseeiusmodipositionesplaniHM,quiainaequatione
cubicaradicestresessepossunt,ettrestangentistvalores."(However,it
isnotinconsistent[thatthere]bethreesuchpositionsoftheplaneHM,
becauseincubicequations,[there]canbethreeroots,andthreevaluesof
thetangentt.)
TherelevantpassageofSegner'sworkwasdiscussedbrieflybyArthur
Cayley.See:A.Cayley(1862)"Reportontheprogressofthesolutionof
certainspecialproblemsofdynamics,"ReportoftheThirtysecondmeeting
oftheBritishAssociationfortheAdvancementofScienceheldat
CambridgeinOctober1862,32:184252seeespeciallypages225226.
(https://books.google.com/books?id=S_RJAAAAcAAJ&pg=PA225#v=onep
age&q&f=false)

10.SeeHawkins1975,2
11.SeeHawkins1975,3
12.SeeKline1972,pp.807808
13.AugustinCauchy(1839)"Mmoiresurl'intgrationdesquationslinaires"
(Memoirontheintegrationoflinearequations),Comptesrendus,8:827830,
845865,889907,931937.Fromp.827:(http://gallica.bnf.fr/ark:/12148/bpt6k29
67c/f833.item.r=.zoom)"Onsaitd'ailleursqu'ensuivantlamthodedeLagrange,
onobtientpourvaleurgnraledelavariableprinicipaleunefonctiondans
laquelleentrentaveclavariableprincipalelesracinesd'unecertainequationque
j'appellerail'quationcaractristique,ledegrdecettequationtantprcisment
l'orderdel'quationdiffrentiellequ'ils'agitd'intgrer."(Onknows,moreover,
thatbyfollowingLagrange'smethod,oneobtainsforthegeneralvalueofthe
principalvariableafunctioninwhichthereappear,togetherwiththeprincipal
variable,therootsofacertainequationthatIwillcallthe"characteristicequation",
thedegreeofthisequationbeingpreciselytheorderofthedifferentialequation
thatmustbeintegrated.)
14.SeeKline1972,p.673
15.SeeKline1972,pp.715716
16.SeeKline1972,pp.706707
17.SeeKline1972,p.1063

18.See:
DavidHilbert(1904)"GrundzgeeinerallgemeinenTheoriederlinearen
Integralgleichungen.(ErsteMitteilung)"(http://www.digizeitschriften.de/d
ms/img/?PPN=PPN252457811_1904&DMDID=dmdlog11&LOGID=log11
&PHYSID=phys57#navi)(Fundamentalsofageneraltheoryoflinear
integralequations.(Firstreport)),NachrichtenvonderGesellschaftder
WissenschaftenzuGttingen,MathematischPhysikalischeKlasse(Newsof
thePhilosophicalSocietyatGttingen,mathematicalphysicalsection),pp.
4991.Frompage51:(http://www.digizeitschriften.de/dms/img/?PPN=PPN
252457811_1904&DMDID=dmdlog11&LOGID=log11&PHYSID=phys57#
navi)"InsbesondereindiesererstenMitteilunggelangeichzuFormeln,die
dieEntwickelungeinerwillkrlichenFunktionnachgewissen
ausgezeichnetenFunktionen,dieichEigenfunktionennenne,liefern:(In
particular,inthisfirstreportIarriveatformulasthatprovidethe[series]
developmentofanarbitraryfunctionintermsofsomedistinctivefunctions,
whichIcalleigenfunctions:)Lateronthesamepage:"DieserErfolgist
wesentlichdurchdenUmstandbedingt,daichnicht,wieesbisher
geschah,inersterLinieaufdenBeweisfrdieExistenzderEigenwerte
ausgehe,"(ThissuccessismainlyattributabletothefactthatIdonot,
asithashappeneduntilnow,firstofallaimataproofoftheexistenceof
eigenvalues,)
Fortheoriginandevolutionofthetermseigenvalue,characteristicvalue,
etc.,see:EarliestKnownUsesofSomeoftheWordsofMathematics(E)
(http://jeff560.tripod.com/e.html)
19.SeeAldrich2006
20.Francis,J.G.F.(1961),"TheQRTransformation,I(part1)",TheComputer
Journal,4(3):265271,doi:10.1093/comjnl/4.3.265andFrancis,J.G.F.(1962),
"TheQRTransformation,II(part2)",TheComputerJournal,4(4):332345,
doi:10.1093/comjnl/4.4.332
21.Kublanovskaya,VeraN.(1961),"Onsomealgorithmsforthesolutionofthe
completeeigenvalueproblem",USSRComputationalMathematicsand
MathematicalPhysics,3:637657.Alsopublishedin:"
"[Oncertainalgorithms
forthesolutionofthecompleteeigenvalueproblem],
(JournalofComputational
MathematicsandMathematicalPhysics),1(4):555570,1961
22.SeeGolub&vanLoan1996,7.3Meyer2000,7.3
23.CornellUniversityDepartmentofMathematics(2016)LowerLevelCoursesfor
FreshmenandSophomores(http://www.math.cornell.edu/m/Courses/Catalog/lower
level).Accessedon20160327.

24.UniversityofMichiganMathematics(2016)MathCourseCatalogue(https://ww
w.lsa.umich.edu/UMICH/math/Home/Undergrad/Ugrad_Courses.pdf).Accessed
on20160327.
25.Press(2007,pp.38)
26.Fraleigh(1976,p.358)
27.Golub&VanLoan(1996,p.316)
28.Beauregard&Fraleigh(1973,p.307)
29.Herstein(1964,p.272)
30.Nering(1970,pp.115116)
31.Herstein(1964,p.290)
32.Nering(1970,p.116)
33.SeeKorn&Korn2000,Section14.3.5aFriedberg,Insel&Spence1989,p.217
34.Shilov1977,p.109
35.Lemmafortheeigenspace
36.Schaum'sEasyOutlineofLinearAlgebra(https://books.google.com/books?id=pk
ESXAcIiCQC&pg=PA111),p.111
37.Foraproofofthislemma,seeRoman2008,Theorem8.2onp.186Shilov1977,
p.109Hefferon2001,p.364Beezer2006,TheoremEDELIonp.469and
Lemmaforlinearindependenceofeigenvectors
38.Axler,Sheldon,"Ch.5",LinearAlgebraDoneRight(2nded.),p.77
39.Trefethen,LloydN.Bau,David(1997),NumericalLinearAlgebra,SIAM

40.Graham,D.Midgley,N.(2000),"Graphicalrepresentationofparticleshapeusing
triangulardiagrams:anExcelspreadsheetmethod",EarthSurfaceProcessesand
Landforms,25(13):14731477,Bibcode:2000ESPL...25.1473G,
doi:10.1002/10969837(200012)25:13<1473::AIDESP158>3.0.CO2C
41.Sneed,E.D.Folk,R.L.(1958),"PebblesinthelowerColoradoRiver,Texas,a
studyofparticlemorphogenesis",JournalofGeology,66(2):114150,
Bibcode:1958JG.....66..114S,doi:10.1086/626490
42.KnoxRobinson,C.Gardoll,StephenJ.(1998),"GISstereoplot:aninteractive
stereonetplottingmoduleforArcView3.0geographicinformationsystem",
Computers&Geosciences,24(3):243,Bibcode:1998CG.....24..243K,
doi:10.1016/S00983004(97)001222
43.Stereo32software(http://www.ruhrunibochum.de/hardrock/downloads.htm)
44.Benn,D.Evans,D.(2004),APracticalGuidetothestudyofGlacialSediments,
London:Arnold,pp.103107
45.Xirouhakis,A.Votsis,G.Delopoulus,A.(2004),Estimationof3Dmotionand
structureofhumanfaces(PDF),NationalTechnicalUniversityofAthens
46.DiekmannO,HeesterbeekJA,MetzJA(1990),"Onthedefinitionandthe
computationofthebasicreproductionratioR0inmodelsforinfectiousdiseasesin
heterogeneouspopulations",JournalofMathematicalBiology,28(4):365382,
doi:10.1007/BF00178324,PMID2117040
47.OdoDiekmannJ.A.P.Heesterbeek(2000),Mathematicalepidemiologyof
infectiousdiseases,Wileyseriesinmathematicalandcomputationalbiology,West
Sussex,England:JohnWiley&Sons

References
Akivis,MaxA.Goldberg,VladislavV.(1969),Tensorcalculus,Russian,Science
Publishers,Moscow
Aldrich,John(2006),"Eigenvalue,eigenfunction,eigenvector,andrelatedterms",
inJeffMiller(Editor),EarliestKnownUsesofSomeoftheWordsof
Mathematics,retrieved20060822
Alexandrov,PavelS.(1968),Lecturenotesinanalyticalgeometry,Russian,
SciencePublishers,Moscow
Anton,Howard(1987),ElementaryLinearAlgebra(5thed.),NewYork:Wiley,
ISBN0471848190
Beauregard,RaymondA.Fraleigh,JohnB.(1973),AFirstCourseInLinear
Algebra:withOptionalIntroductiontoGroups,Rings,andFields,Boston:
HoughtonMifflinCo.,ISBN039514017X
Beezer,RobertA.(2006),Afirstcourseinlinearalgebra,Freeonlinebookunder
GNUlicence,UniversityofPugetSound

Betteridge,HaroldT.(1965),TheNewCassell'sGermanDictionary,NewYork:
Funk&Wagnall,LCCN587924
Bowen,RayM.Wang,ChaoCheng(1980),Linearandmultilinearalgebra,
PlenumPress,NewYork,ISBN0306375087
Brown,Maureen(October2004),IlluminatingPatternsofPerception:An
OverviewofQMethodology
Burden,RichardL.Faires,J.Douglas(1993),NumericalAnalysis(5thed.),
Boston:Prindle,WeberandSchmidt,ISBN0534932193
Carter,TamaraA.Tapia,RichardA.Papaconstantinou,Anne,LinearAlgebra:
AnIntroductiontoLinearAlgebraforPreCalculusStudents,RiceUniversity,
OnlineEdition,retrieved20080219
CohenTannoudji,Claude(1977),"ChapterII.Themathematicaltoolsofquantum
mechanics",Quantummechanics,JohnWiley&Sons,ISBN0471164321
Curtis,CharlesW.(1999),LinearAlgebra:AnIntroductoryApproach(4thed.),
Springer,ISBN0387909923

Demmel,JamesW.(1997),Appliednumericallinearalgebra,SIAM,ISBN0
898713897
Fraleigh,JohnB.(1976),AFirstCourseInAbstractAlgebra(2nded.),Reading:
AddisonWesley,ISBN0201019841
Fraleigh,JohnB.Beauregard,RaymondA.(1995),Linearalgebra(3rded.),
AddisonWesleyPublishingCompany,ISBN0201839997
Friedberg,StephenH.Insel,ArnoldJ.Spence,LawrenceE.(1989),Linear
algebra(2nded.),EnglewoodCliffs,NewJersey07632:PrenticeHall,ISBN0
135371023
Gelfand,I.M.(1971),Lecturenotesinlinearalgebra,Russian,Science
Publishers,Moscow
Gohberg,IsraelLancaster,PeterRodman,Leiba(2005),Indefinitelinearalgebra
andapplications,BaselBostonBerlin:BirkhuserVerlag,ISBN3764373490
Golub,GeneF.vanderVorst,HenkA.(2000),"Eigenvaluecomputationinthe
20thcentury",JournalofComputationalandAppliedMathematics,123:3565,
Bibcode:2000JCoAM.123...35G,doi:10.1016/S03770427(00)004131
Golub,GeneH.VanLoan,CharlesF.(1996),Matrixcomputations(3rded.),
JohnsHopkinsUniversityPress,Baltimore,Maryland,ISBN9780801854149
Greub,WernerH.(1975),LinearAlgebra(4thed.),SpringerVerlag,NewYork,
ISBN0387901108
Halmos,PaulR.(1987),Finitedimensionalvectorspaces(8thed.),NewYork:
SpringerVerlag,ISBN0387900934
Hawkins,T.(1975),"Cauchyandthespectraltheoryofmatrices",Historia
Mathematica,2:129,doi:10.1016/03150860(75)900324
Hefferon,Jim(2001),LinearAlgebra,Onlinebook,StMichael'sCollege,
Colchester,Vermont,USA
Herstein,I.N.(1964),TopicsInAlgebra,Waltham:BlaisdellPublishing
Company,ISBN9781114541016
Horn,RogerA.Johnson,CharlesF.(1985),Matrixanalysis,Cambridge
UniversityPress,ISBN0521305861
Kline,Morris(1972),Mathematicalthoughtfromancienttomoderntimes,Oxford
UniversityPress,ISBN0195014960
Korn,GraninoA.Korn,TheresaM.(2000),"MathematicalHandbookfor
ScientistsandEngineers:Definitions,Theorems,andFormulasforReferenceand
Review",NewYork:McGrawHill(2ndReviseded.),DoverPublications,
Bibcode:1968mhse.book.....K,ISBN0486411478

Kuttler,Kenneth(2007),Anintroductiontolinearalgebra(PDF),Onlineebookin
PDFformat,BrighamYoungUniversity
Lancaster,P.(1973),Matrixtheory,Russian,Moscow,Russia:SciencePublishers
Larson,RonEdwards,BruceH.(2003),Elementarylinearalgebra(5thed.),
HoughtonMifflinCompany,ISBN0618335676
Lipschutz,Seymour(1991),Schaum'soutlineoftheoryandproblemsoflinear
algebra,Schaum'soutlineseries(2nded.),NewYork:McGrawHillCompanies,
ISBN0070380074
Meyer,CarlD.(2000),Matrixanalysisandappliedlinearalgebra,Societyfor
IndustrialandAppliedMathematics(SIAM),Philadelphia,ISBN978089871
4548
Nering,EvarD.(1970),LinearAlgebraandMatrixTheory(2nded.),NewYork:
Wiley,LCCN76091646
(Russian) Pigolkina,T.S.Shulman,V.S.(1977)."Eigenvalue".InVinogradov,I.
M.MathematicalEncyclopedia.5.Moscow:SovietEncyclopedia.
Press,WilliamH.Teukolsky,SaulA.Vetterling,WilliamT.Flannery,BrianP.
(2007),NumericalRecipes:TheArtofScientificComputing(3rded.),
ISBN9780521880688
Roman,Steven(2008),Advancedlinearalgebra(3rded.),NewYork:Springer
Science+BusinessMedia,LLC,ISBN9780387728285
Sharipov,RuslanA.(1996),CourseofLinearAlgebraandMultidimensional
Geometry:thetextbook,arXiv:math/0405323 ,Bibcode:2004math......5323S,
ISBN5747700995
Shilov,GeorgiE.(1977),Linearalgebra,TranslatedandeditedbyRichardA.
Silverman,NewYork:DoverPublications,ISBN048663518X
Shores,ThomasS.(2007),Appliedlinearalgebraandmatrixanalysis,Springer
Science+BusinessMedia,LLC,ISBN0387331948
Strang,Gilbert(1993),Introductiontolinearalgebra,WellesleyCambridge
Press,Wellesley,Massachusetts,ISBN0961408855
Strang,Gilbert(2006),Linearalgebraanditsapplications,Thomson,
Brooks/Cole,Belmont,California,ISBN0030105676

Externallinks
WhatareEigenValues?(http://www.physlink.com/education/AskExperts/ae520.cfm)nontechnicalintroduction
fromPhysLink.com's"AsktheExperts"

TheWikibookLinear

EigenValuesandEigenVectorsNumericalExamples(http://people.revoledu.com/kardi/tutorial/LinearAlgebra/Eigen
ValueEigenVector.html)TutorialandInteractiveProgramfromRevoledu.
IntroductiontoEigenVectorsandEigenValues(http://khanexercises.appspot.com/video?v=PhfbEr2btGQ)lecture
fromKhanAcademy
Hill,Roger(2009)."Eigenvalues".SixtySymbols.BradyHaranfortheUniversityofNottingham.
"ABeginner'sGuidetoEigenvectors".Deeplearning4j.2015.

Theory

Algebrahasapageonthe
topicof:Eigenvaluesand
Eigenvectors
TheWikibookTheBook
ofMathematicalProofs
hasapageonthetopicof:
Algebra/Linear
Transformations

Hazewinkel,Michiel,ed.(2001),"Eigenvalue",EncyclopediaofMathematics,Springer,ISBN9781556080104
Hazewinkel,Michiel,ed.(2001),"Eigenvector",EncyclopediaofMathematics,Springer,ISBN9781556080104
Eigenvalue(ofamatrix)(http://planetmath.org/?op=getobj&from=objects&id=4397)atPlanetMath.org.
Eigenvector(http://mathworld.wolfram.com/Eigenvector.html)WolframMathWorld
EigenVectorExaminationworkingapplet(http://ocw.mit.edu/ans7870/18/18.06/javademo/Eigen/)
SameEigenVectorExaminationasaboveinaFlashdemowithsound(http://web.mit.edu/18.06/www/Demos/eigenappletall/eigen_sound_all.html)
ComputationofEigenvalues(http://www.sosmath.com/matrix/eigen1/eigen1.html)
Numericalsolutionofeigenvalueproblems(http://www.cs.utk.edu/~dongarra/etemplates/index.html)EditedbyZhaojunBai,JamesDemmel,Jack
Dongarra,AxelRuhe,andHenkvanderVorst
EigenvaluesandEigenvectorsontheAskDr.Mathforums:[1](http://mathforum.org/library/drmath/view/55483.html),[2](http://mathforum.org/library/dr
math/view/51989.html)

Demonstrationapplets
Javaappletabouteigenvectorsintherealplane(http://scienceapplets.blogspot.com/2012/03/eigenvaluesandeigenvectors.html)
WolframLanguagefunctionalityforEigenvalues,EigenvectorsandEigensystems(http://reference.wolfram.com/language/guide/DifferentialEquations.html)
Retrievedfrom"https://en.wikipedia.org/w/index.php?title=Eigenvalues_and_eigenvectors&oldid=737070968"
Categories: Mathematicalphysics Abstractalgebra Linearalgebra Matrixtheory Singularvaluedecomposition
Thispagewaslastmodifiedon31August2016,at15:48.
TextisavailableundertheCreativeCommonsAttributionShareAlikeLicenseadditionaltermsmayapply.Byusingthissite,youagreetotheTermsofUse
andPrivacyPolicy.WikipediaisaregisteredtrademarkoftheWikimediaFoundation,Inc.,anonprofitorganization.