You are on page 1of 4

A Truly Elementary Approach to the Bounded Convergence Theorem

Author(s): Jonathan W. Lewin


Source: The American Mathematical Monthly, Vol. 93, No. 5 (May, 1986), pp. 395-397
Published by: Mathematical Association of America
Stable URL: http://www.jstor.org/stable/2323608 .
Accessed: 09/01/2015 11:54

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.

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

http://www.jstor.org

This content downloaded from 69.91.134.221 on Fri, 9 Jan 2015 11:54:56 AM


All use subject to JSTOR Terms and Conditions
1986] THE TEACHING OF MATHEMATICS 395

These matriceshavethesameeasilycomputablecharacteristicpolynomialsas N-matrices and


theirentriesmaybe selectedso as to yieldresultsanalogousto the N-matrixcases. For 3 x 3
matricesone obtainsthesamesixcases(1)-(6) discussedabove,butin eachof thecasesthereare
in theformoftheeigenvectors.
slightdifferences

A TRULY ELEMENTARY APPROACH TO THE BOUNDED CONVERGENCE THEOREM

JONATHANW. LEWIN
KennesawCollege,Marietta,GA 30061
ofMathematics,
Department

The boundedconvergence theorem followstrivially fromtheLebesguedominatedconvergence


theorem, but at thelevelofan introductory coursein analysis,whentheRiemannintegral is being
studied,how hardis theboundedconvergence theorem? For an answer,we mightlook at Bartle
and Sherbert [2],page203: Theproofofthisresultis quitedelicateandwillbe omitted. Or we might
look at Apostol[1],page 228: TheproofofArzela's theorem is considerablymoredifficult than ...
and willnotbe givenhere.WalterRudinin [4] ignoresthetheoremaltogether in his chapteron
Riemannintegration, presenting it onlyas a corollaryto theLebesguedominatedconvergence
theoremseveralchapters problemin ChapterTwo,Rudinrefers
later,and in [5],in an interesting
his readersto [3]. In [3], Eberleindoes presenta proofwhichfromsome pointsof view is
elementary. Certainly,his proofdoes not requireany notionsof measurability, but it is hardly
elementary fromthepointofviewof a studentwhois firstlearningtheRiemannintegral. So the
answerto theabovequestionseemsto be: veryhard!But thisis notso. In thispaper,we present
theproofof theboundedconvergence theoremin a trulyelementary setting,and in sucha way
thatit couldbe includedforthefirsttimein an introductory course.
We beginby defining an elementary set.A boundedsubsetE of R is said to be elementary if
E is a finiteunionofboundedintervals, or equivalently,if XE is a stepfunction. One can define
the Lebesgue measurem(E) of an elementary set E to be fb'XEI where[a, b] is an interval
includingE, and one can showsimply thaton the familyofelementary sets(whichis closedunder
union,intersection Lebesguemeasureis finitely
and differences), additiveand finitelysubadditive.
Given a Riemannintegrable functionf on an interval[a, b], and an elementary subsetE of
[a, b], we definefEf = JbfXE.If E and F are mutually disjointelementary sets,thenone may
show easilythat fEuFf = fEf + fFf, and if lf(x)l < K foreverypoint x in E, thenIfEf I <
Km(E). One mayalso provesimplythatif E is an elementary setand e > 0, thenone can finda
closed elementary subsetH of E suchthatm(H) > m(E) - ?.
sequenceof boundedsubsetsof R, withan empty
LEMMA. Suppose (A,?) is a contracting
For eachn, define
intersection.
an= subsetofA,n}.
sup{ m( E) {E is an elementary
Thenan -O 0 as n -- oo.
Proof.The sequence(an) is clearlydecreasing.Now, to obtaina contradiction, assumethat
thissequencedoes notconvergeto 0, and choose 8 > 0, suchthata,, > 8 forall n. For each n,
subsetEn of An suchthat
choose a closedelementary
m(En) > an -82n

and define
n
Hn = n E,.
i=1

sequence of closed bounded sets, we can obtain the desired


Since (Hn) is a contracting
forthentheintersection
by showingthateach set Hn is non-empty;
contradiction of all thesets
H,, would be non-emptyeven thoughthelargersets A,nhave an emptyintersection. For this

This content downloaded from 69.91.134.221 on Fri, 9 Jan 2015 11:54:56 AM


All use subject to JSTOR Terms and Conditions
396 JONATHANW. LEWIN [May

Firstly,
twoobservations:
purpose,we makethefollowing subset
forany n, if E is an elementary
of A,n\ E,n,thensince
m(E) + m(En) = m(E U Ej) an and m(E,n) > an-8/2n,
subsetof A,n\ H,1,then
it followsthatm(E) < 8/2n.Secondly,forany n, if E is an elementary
since
E = (E\ E1) U (E\ E2) U (E \ E3) U *...U(E\En)

and since E \E, is an elementary subsetof Ai \E, for everyi = 1,2,. .. n, it followsthat
m(E) < 8.
But foreveryn, because an > 8, the set An musthave an elementary subsetE such that
m(E) > 8, and so it followsthateach set Hnis non-empty.
The Main Result.Suppose(f,) is a sequenceofRiemannintegrable functionson [a, b], suppose
f is a Riemannintegrable
function on [a, b], thatfn-* f pointwiseon [a, b] and thatforsome
constantK > 0, we haveIfnl< Kfor everyn. Thenwehave
|_
fn f-

in assumingthatfn> 0 foreach n and thatf = 0. Let


Proof.Thereis no loss of generality
e > 0, and foreach n, define

A,,=(x [a, b]It(x) > 2(b ) forat leastone naturali > n)

We now applythelemmato (An) to choosea naturalN suchthatwhenever n > N, and E is an


elementarysubsetof An,we have m(E) < e/2K, and theproofwillbe completewhenwe have
shown that whenevern > N, we have fabfn < E. Let n > N. Since the integralof a Riemann
function
integrable is thesameas itslowerintegral,in orderto showthatfabfn< e, it is sufficient
to showthatwhenever and 0 < s < fn,we havefbs < E. Let s be sucha step
s is a stepfunction
functionand define

E= x E [a, b] is(x) > b and F=[a,b]\E.


2(bb-a))
sets,and sinceE
Then E and F are elementary C An_we have m(E) < e/2K. Therefore

fb IE
f IF IE IFf2(6b a) If lf2(b a)

-Km(E)+ (b-a)<e.
2(b -a)
And thatis all thereis to it.Noticethatwhiletheaboveproofemployssomeof thenotationand
conveyssome of theatmosphere of moreadvancedtreatments it keepswellaway
of integration,
fromanything hardtLebesguemeasureis neededonlyforelementary sets;and all themeasureis
in thiscase is thesumof thelengthsof thefinitely manycomponent intervalsthatmakeup an
elementary set. The proofis accessibleto studentswho have neverseen countability and never
seen infiniteseries.Theydon'tevenneed theHeine Boreltheorem if theyknowthata bounded
sequence of real numbersmusthave a partiallimit(clusterpoint) and that,consequently, a
contracting sequenceof nonemptyclosedboundedsetsmusthavea non emptyintersection.
Incidentally,it is easyto adapttheaboveproofto showthatevenifit is notassumedthatthe
limitfunctionf is Riemannintegrable, because (fn(x)) is a Cauchysequenceforeach x, the
sequenceof integralsfaf?mustbe a Cauchysequenceand musttherefore converge. Thismaybe
used to givea revealing explanationof theinadequacyof theRiemannintegral.

This content downloaded from 69.91.134.221 on Fri, 9 Jan 2015 11:54:56 AM


All use subject to JSTOR Terms and Conditions
1986] THE TEACHING OF MATHEMATICS 397

References

1. Tom M. Apostol,MathematicalAnalysis,2nd ed., Addison-Wesley,Reading,MA, 1974.


to Real Analysis,Wiley,New York, 1982.
2. RobertG. Bartleand Donald R. Sherbert,Introduction
3. W. F. Eberlein,Notes on integration convergencetheorem,Comm. Pure Appl. Math.,Vol.
I: theunderlying
X (1957) 357-360.
4. WalterRudin, Principlesof MathematicalAnalysis,McGraw-Hill,New York, 1964.
5. , Real and ComplexAnalysis,McGraw-Hill,New York, 1966.
6. W. A. J. Luxemburg,Arzela's dominatedconvergencetheoremfor the Riemann integral,this MONTHLY,
78 (1971) 970-979.

EXPLAINING SIMPLE COMBINATORIAL ANSWERS

ROGER B. EGGLETON
ofMathematics,
Department Science,
Statisticsand Computer
ofNewcastle,N.S. W.2308,Australia
University

This note illustratestheprinciple:If theanswerto a problemturnsout to be simple,thereis


probablya goodexplanation forit!A simpleanswershouldmotivate us to tryto derivethatanswer
in a way whichmakesit obvious,or at least clarifiesthe underlying reasonforits simplicity.
Simplicityand clarityare of coursesubjectivemeasures,but ones whichare stilluseful.The
practiceof mathematics is an artas wellas a science.
Consider combinatorics. Here it is recognizedthat simpleanswersare oftensatisfyingly
explicablein termsof correspondences. This themewas takenup in [3], forexample,fromthe
viewpointthatcountingtheelements unfamiliar
of a relatively achieved
set X can be satisfyingly
ifwe establisha correspondence betweentheelementsof X and thoseof somerelatively familiar
set A. The correspondence constitutesthedesiredexplanation. In thisnotewe takeup thetheme
fromtheviewpointthatexplanations in termsof correspondences can also be achievedbetween
two sets X and A of equallyfamiliar structure. thiswithseveralexamples,mostof
We illustrate
which"explain"a well-known identity, suitableforclassroomuse.
and are therefore
We shall use lowercase symbolsto denotenaturalnumbers,includingzero,and I(n) will
denotetheset comprising thefirstn naturalnumbers(thatis, thenaturalnumbersless thann).
The familyof k-subsets of I(n) willbe denotedbyI(n, k). We regardthebinomialcoefficients as
thecardinalities of suchsets,by definition:
nk
JlI(n, k) 1.

EXAMPLE 1 (SymmetryofPascal's Triangle).Let A = I(n, k) and X:= I(n, n - k). Pairing


each k-subsetof I(n) withits complementgivesa one-to-onecorrespondence X * A. Hence
=
IXI IAI, so

(n-k) k ()k
EXAMPLE sometimes
2 (Pascal's Identity, called Vandermonde'sIdentity).Let A = I(n + 1,
k + 1) and X:= XO X1,U where =
XO I(n, k) and =
X1 I(n, + 1). Any (k + l)-subsetof
k
I(n + 1) eithercontains theelementn or it does not.In theformercase,pairit withthek-subset
of I(n) obtainedby deletingthen, whilein thelattercase simplypairit withitself,nowregarded
as a (k + 1)-subsetof I(n). Thisgivesa one-to-one correspondenceX * A, sinceXOand X1 are
disjoint.Hence IXI = IXO+ IXii= JAl, so

k Jk + 1 k + 1J
EXAMPLE 3 (Arithmetic SeriesIdentity).The sum of naturalnumbersup to n, inclusive,is
in (n + 1), whichis a barely-disguised How can we explainthebinomial
binomialcoefficient.

This content downloaded from 69.91.134.221 on Fri, 9 Jan 2015 11:54:56 AM


All use subject to JSTOR Terms and Conditions

You might also like