Sei sulla pagina 1di 5
Measure and Cardinality Author(s): James M. Briggs and Thomas Schaffter Reviewed work(s): Source: The American

Measure and Cardinality Author(s): James M. Briggs and Thomas Schaffter Reviewed work(s):

Source: The American Mathematical Monthly, Vol. 86, No. 10 (Dec., 1979), pp. 852-855 Published by: Mathematical Association of America

Accessed: 18/11/2011 04:55

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

information about JSTOR, please contact support@jstor.org. Mathematical Association of America is collaborating with

Mathematical Association of America is collaborating with JSTOR to digitize, preserve and extend access to The American Mathematical Monthly.

http://www.jstor.org

852

JAMES M. BRIGGS AND THOMAS SCHAFFTER

[December

K. W. SchmidtandEdwardT. H. Wang,TheweightsofHadamardmatrices,J.Combin.TheorySer.A, 23

(1977)257-263.

JiriSedlacek,On magicgraphs,Math.Slov.,26 (1976)329-335;MR 55 #7853;Zbl.348.05114.

- ,Somepropertiesof magicgraphs,GraphsHypergraphsand BlockSystems,ZielonaGora, 1976,

247-253.

Y. Shiloach,A minimumlineararrangementalgorithmforundirectedtrees(tbp).

DavidSingmasterandD. J.Souppouris,A constrainedisoperimetricproblem,Math.Proc.CambridgePhilos.

Soc.,83(1978)73-82.

NeilJ.A. Sloane,A HandbookofIntegerSequences,AcademicPress,1973. W. F. Smyth,A labelingalgorithmforminimumedgeweightsumofcompletebinarytrees(tbp). HerbertTaylor,Oddpathsumsinan edge-labelledtree,Math.Mag.,50 (1977)258-259. ZalmanUsiskin,On thenumberof3-termarithmeticprogressionsin setswithno 4-termones,Math.Mag. (tbp).

CLASSROOM NOTES

EDITEDBYDEBORAHTEPPERHAIMOANDFRANKILNTEPPERHAIMo

Materialforthisdepartmentshouldbe sentto ProfessorDeborah TepperHaimo,DepartmentofMathematical Sciences,UniversityofMissouri,St. Louis,MO 63121.

Adviceto prospectiveauthors:The editorshaverecentlybeenreceivingabouttentimesas many ClassroomNotesas can be used.It willsimplifyourworkifauthorswillsubmitonlyespecially interestingmanuscripts.

R.P.B.

MEASUR1EAND CARDINAL1TY

JAMESM. BRIGGSANDTHOMASSCHAFFrER

The purposeofthispaperis to describehowthesettheoryaxiomswe acceptaffectLebesgue measureon therealline.The resultsare generallyknownamongsettheoristsand easyto state butratherhardto getat forthenon-specialist.A coursein real analysiswouldseemto be an appropriateplace to introducestudentsto theseresults. One ofthefirstpropertieslearnedaboutLebesguemeasureon thereallineis thatcountable sets of real numbersare measurableand theyhave measurezero. This is, of course,an immediateconsequenceof thefactsthatsingletonsetshave measurezero and thatLebesgue measureis countablyadditive.To thestudentwhoknowstheremaybe setsofrealnumbersof cardinalitystrictlybetween so (thecardinalityofthenaturalnumbers)and c (thecardinalityof thereals),thequestionnaturallyarisesas to whatcan be said abouttheirmeasurability.This questioncan be answeredfromtwoperspectives:(1) whatmustholdin thereals,and (2) what mighthold in thereals.More precisely,thedistinctionis between(1) whatis impliedby our basic axioms,i.e.,whatcan be provedfromthem,and (2) whatis consistentwithand whatis independentoftheaxioms.Ofcourse,itis notsufficienttoconsiderjusttheusualaxiomsforthe reals(i.e.,theaxiomsfora complete,orderedfield);we mustalso consideraxiomsofsettheory. Clearly,theanswersgivento theabovequestionswilldependon whatsettheoryaxiomsweuse. Thereis consensusamongmathematiciansthattheaxiomsofZermelo-Fraenkelsettheory(ZF) are acceptablebut thereare controversiesconcerningthe Axiom of Choice (AC) and the ContinuumHypothesis(CH).

1979]

CLASSROOM NOTES

FromtheworkofGodel [31and Cohen [2] weknowthatbothofthestatements,AC and

853

CH,

areindependentofZF, i.e.,neithercan be provedorrefutedinZF. Anotherwaytoputitis that thereare modelsof ZF in whichAC (respectively,CH) holdsand modelsin whichAC (CH) fails.For thosewhobelievein an objectiverealityoftheobjectsofmathematicsthereis strong intuitiveevidenceforAC. (Givena setofnonemptysets,"surely"thereis a setcontainingone elementfromeach. Whywouldn'ttherebe? For a discussionof thisand contraryviewsof mathematicssee Bernays[1]. For a similardiscussionrelatedto CH see Godel [4]. Thereis no such evidenceforCH (whyshouldc be thefirstuncountablecardinal?).Accordingly,most mathematiciansunhesitatinglyuse AC butdo nottakea standon CH. Thus,ouroriginalquery (whatcan be said in ZF + AC (ZFC) aboutthemeasurabilityof setsof realswithcardinality between 40and c) shouldbe meaningfulto mostmathematicians. It is wellknownamongpersonsspecializingin thefoundationsofmathematicsthatifthere areanysetsofrealnumbersofcardinalitybetween mo and c, all suchsetswhicharemeasurable havemeasurezero.We presentan elementaryproofofthisfact.Nextwe discusswhatis known aboutthepossibilityof non-measurablesetsin this"in between"land. Finally,forthosewho considerAC up fordebate,we commenton whatcan be said in theabsenceofthisaxiom.

THEOREM. If E is a measurableset ofrealsofcardinalityless thanc, thenE has measure0.

Proof.We showthatifE has positivemeasurethenthecardinalityof E is c. Accordingly,

supposemE >0 and let F be a closed subsetof E of positivemeasure.Then thereare two disjoint,closed intervals I,JI of lengthless thanmF whichintersectF in sets of positive

measure.Now within Io thereexisttwo disjoint,closed intervals IoJIo1 of

m(F n IO)whichalso intersectF in setsofpositivemeasure.Similarly,we can findintervalsIlo

and III within I, withthecorrespondingproperties. Continuingin thismanner,we produceat thenthstep2" disjoint,closed intervals Ij.,. ,,,

ji E f0,1), each of whichintersectsF in a set of positivemeasure.In the limiteach infinite

lengthless than

sequence jlIJ2

from{0,1} correspondsto an infinitesequence Ij,,Ij,j2,

of nested,closed

intervals,all of whichintersectF. By thecompletenessof therealstheintersectionof sucha sequenceis non-empty,and hencedeterminesat leastonepointofF. The pointsofF produced bytheseintersectionsareclearlydistinct,and thereforethisprocessidentifies2'o = c elementsof

F, givingus thedesiredresult. At this point we are assumingAC. However,we note thatit is not needed for the constructionsin thisproof:theintervalselectionclearlycouldbe madewithoutitsuse; itsuse can also be avoided when choosinga pointfromeach intersectionof nestedintervalsby requiringthelengthsoftheintervalsto go to zero,therebyproducingsingletons. Thus,in all modelsof thereals(withinZFC) whereCH fails(i.e.,wheresetsofrealsof "in between"cardinalityexist),all measurablesetsof "in between"cardinalityhave

But can therebe non-measurablesetsof "in between"cardinality?The classicalproofof the existenceof a non-measurableset,due to Vitali(1905),is of no helpsinceitproducesa setof

cardinalityc (see [5,pp. 69-701). However,RobertSolovayhas providedtheanswer.On theone sidehe has shownthattherearemodelsofZFC inwhichCH failswhereeverysetofcardinality lessthanc is measurable.Hence,assumingthatsetsof"inbetween"cardinalityexist,we cannot prove,usingZFC, thatanysuchsetsare non-measurable.In theterminologyof Foundations, ZFC + - CH cannotprovethereare non-measurablesetsof"in between"cardinality. On theotherside,Solovayhas shownthattherearemodelsofZFC in whichCH failswhere

measurezero.

therearesetsofcardinality81(thefirstuncountablecardinal)whicharenotmeasurable.Hence,

ZFC cannotprovethereare no non-measurablesetsofthekindin question(see [9]). Thus,justas we cannotproveorrefute(in ZFC) thattherearesetsofrealsof"in between" cardinality,we cannot prove or refutethatamong such sets (if thereare any) some are

non-measurable.Stateddifferently,just as ZFC does

ZFC does notdecidewhetherall setsofrealsofcardinalityless thanc aremeasurable.

not decide the ContinuumHypothesis,

854

JAMES M. BRIGGS AND THOMAS SCHAFFTER

[December

Finally,we commenton whatcan happenin theabsenceof AC. Some formof theaxiom mustbe retainedin ordertopreservea reasonablemeasure.In theabsenceofAC itcan happen thatthesetofrealsis a countableunionofcountablesets(Levy[71), forcingus intoa measure whichis eithertrivialor not countablyadditive.A weak formof AC, whichis sufficientto preservethemainpropertiesof Lebesguemeasureis theprincipleofDependentChoices (DC).

DC statesthat,givena binaryrelationR on a setS suchthatforall x E S thereis a y E S for

whichxRy,thenstartingwitha particularxo e S, a countablenumberof choicescan be made, relativeto R, each one dependingon theprevious;moreexactly,given xo e S thereexistsa

countablesequencex I,

Axiomof Choice (everycountablecollectionof nonemptysetshas a choicefunction)and is

clearlyimpliedbyAC. (See Jech[6,p. 791fora proofofthefirstimplication.)

Solovayhas shownin [81thatif thereis a model of ZF whichcontainsan inaccessible cardinal(i.e., iftheexistenceof an inaccessiblecardinalis consistentwithZF), thenthereis a modelofZF +DC in whicheverysetofrealsis measurable.(Roughly,an inaccessiblecardinal is an uncountablecardinalwhichcannotbe reachedfrombelowbytakingunionsorpowersets; i.e.,ifK is inaccessible,thentheunionoffewerthanX sets,each ofcardinalitylessthans, has cardinalitylessthanK, and thepowersetof a setof cardinalitylessthanK has cardinalityless than". Manyset theoristsbelieveit is quiteunlikelythattheexistenceof sucha cardinalis inconsistentwithZF; however,thequestionis openat thistime.)Thus,ifone is willingto give up AC, whichmostmathematiciansare not,it is consistentto assume thatthereare no non-measurablesetsof realsat all (assumingtheconsistencyof theexistenceofan inaccessible cardinal).Solovay'smotivationherewas nottopresentthepossibilityofconsideringa theoryof realsin whichall setsweremeasurable,butratherto showthenecessityofAC in provingthe existenceof a non-measurableset: "Of course,theaxiomof choiceis true,and so thereare non-measurablesets[8,p. 3]." However,ifone is notquiteso certainaboutthetruthofAC, the aboveinterpretationis possible. In thesamepaperSolovayshows,undertheinaccessiblecardinalassumption,thatthereare modelsofZFC, somein whichCH holdsand somein whichitfails,whereall definablesetsof realsaremeasurable.The adequacyofan interpretationofa conceptsuchas definabilitywithin

the formalismof set theoryis always open to question;however,Solovay's translationas

"definablefroma sequenceofordinals"seemstocomeverycloseto capturingthenotion.This

is verystrongevidencethatwe willneverbe able to exhibitexplicitlya non-measurableset. In summary,wehaveseenthatiftheAxiomofChoiceis accepteditcan be provedthatthere

are non-measurablesetsof realsof cardinalityc; and ifthereare uncountablesetsof realsof

cardinalityless thanc whichare measurable,theymusthave measurezero.Further,just as it

cannotbe provedor refuted(fromtheordinaryaxiomsof set theory,includingtheAxiomof Choice)thatsuchuncountablesetsofrealsexist,italso cannotbe provedorrefutedthatamong thesesets(assumingthatthereare some)thereare non-measurableones. We havealso seenthatifone is willingto giveup theaxiomofchoice,it cannotbe proved

thatthereareanynon-measurablesetsofrealsat all (assumingtheconsistencyoftheexistence of an inaccessiblecardinal).Furthermore,thereare modelsof ZFC, somein whichCH holds

and somein whichit fails,in whichall definablesetsof realsare measurable.Thus it is very

x2, such that xoRxRX2

This statementimpliesthe Countable

unlikelywe willeverbe able to produceexplicitlya non-measurableset.

Acknowledgment:We thankLooy Simonoffforhis contributions.

References

1. P. Bemays,On platonismin mathematics,in Philosophy of Mathematics, ed. by P. Benacerrafand H.

Putnam,Prentice-Hall,EnglewoodCliffs,N.J.1964.

2. P. J.Cohen,Theindependenceofthecontinuumhypothesis,Part I, II, Proc.Nat.Acad.Sci. U.S.A., 50

(1963)1143-1148;51(1964)105-110.

1979]

CLASSROOM NOTES

855

3. K. Godel,The consistencyof theaxiomof choiceand thegeneralizedcontinuumhypothesiswiththe

axiomsofsettheory,AnnalsofMathematics,StudyNo. 3,PrincetonUniv.Press,Princeton,N.J.,1940.

4. , Whatis Cantor'scontinuumproblem?,inPhilosophyofMathematics,ed.byP. BenacerrafandH.

Putnam,Prentice-hall,EnglewoodCliffs,N.J.,1964.

5. P. R. Halmos,MeasureTheory,VanNostrand,Princeton,N.J.,1965.

6. T. J.Jech,LecturesinSetTheory,LectureNotesinMathematics,Springer-Verlag,1971.

7. A. Levy,The independenceof certainconsequencesof theaxiomof choice,Fund.Math.,54 (1964)

135-137.

8. R. Solovay,A modelofsettheoryinwhicheverysetofrealsis Lebesguemeasurable,Ann.ofMath.,92,

No. 1(July1970)1-56.

9. R. SolovayandD. Martin,InternalCohenextensions,Ann.Math.Logic,2 (1970)143-178.

DEPARTMENTOF MATHEMATICS,UNIVERSITY OF NEVADA, LAS VEGAS, NV 89154.

THE MATRIC EQUATION X2 =A

W. R. UTZ

The purposeofthisnoteis to givea usefulprocedureforsolvingthematricequationX2 =A. In responseto thecuriosityof sophomoresin linearalgebracoursesI have accumulateda numberof exercisesand theoremsinvolvingpolynomialequationsin matricessuitableforan elementarycourse.Chapter7 of a book [5] of C. C. MacDuffeeprovidesa summaryof basic resultsand is a good sourceof referencesformatricequationsup to 1933.It is adequatefor enrichingan elementarycourse.Matricequationsis stillan activeresearchfield(cf.[2],[4],for example)butI am unawareofa recentsurvey.Accordingto [6,Chapter2],thesolutionofthe matricequation2!OAIX' =0 is equivalentto thesearchforrightdivisorsIX-X of 27 0AiV. In thisconnectionthepaper[3] is ofinterest. Returningto thesimplequadraticX2 = A, a classicaltheorem[1,p. 299]ensuresa (nonsingu- lar) solutionoftheequationifA is nonsingular.I nowdescribea method,oncesuggestedto me byW. E. Roth,thatreducestheproblemofsolvingX2 =2A tolinearequationsregardlessofthe rankofA. IfX is a solutionoftheequation(examplesforwhichthereis no solutionare easilydevised), then

hence

X2-X2I=A -21;

(X-XI)(X+XI)=A

-2I.

If +(X) denotesthecharacteristicfunctionofa solution,then

4(X)4( -X) = det(A-A2I)

(*)

and so thecharacteristicfunctionofa solution,ifone exists,mustbe a divisorofdet(A-2I). For eachofthepossiblesolutions+(X)of(*) onesecurestheequation+(X) = 0 inwhichall even powersof X maybe replacedby powersof theknownmatrixA to reducetheequationto a linearmatrixequation. As an example,considertheequation

For thisA,

X2

9

5

0

4

-5 -81

0(X)0(-X) = det(A -X2I) =- X2(X2 _ 9)(X2 - 4)