A Simple Weierstrass Function
Author(s): Donald P. Minassian and John W. Gaisser
Source: The American Mathematical Monthly, Vol. 91, No. 4 (Apr., 1984), pp. 254-256
Published by: Mathematical Association of America
Stable URL: http://www.jstor.org/stable/2322964
Accessed: 20-01-2016 23:24 UTC
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 [email protected].
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 128.6.218.72 on Wed, 20 Jan 2016 23:24:09 UTC
All use subject to JSTOR Terms and Conditions
254 ANDJOHNW. GAISSER
DONALDP. MINASSIAN [April
The dualitybetweenpotentialsat the nodes of a networkand flowsthroughthe edges is
developedmorefullyin [6].
References
1. GilbertStrang, AcademicPress,NewYork,1980.
LinearAlgebraandItsApplications,
2. R. T. Rockafellar,
ConvexAnalysis,
PrincetonUniversity
Press,Princeton,
1970.
3. DavidLuenberger, byVectorSpaceMethods,
Optimization Wiley,NewYork,1969.
4. L. R. FordandD. R. Fulkerson,
Flowsin Networks,
Princeton
University
Press,Princeton,
1962.
5. GilbertStrang,
Maximalflowthrough a domain,Mathematical
Programming, 26 (1983)123-143.
6. , An Introductionto Applied Mathematics,textbookin preparation.
THE TEACHING OF MATHEMATICS
EDITED BY MARY R. WARDROP AND ROBERT F. WARDROP
Materialforthisdepartment RobertF. Wardrop,
shouldbe sentto Professor Department
ofMathematics,
Central
MichiganUniversity,MountPleasant,MI 48859.
A SIMPLE WEIERSTRASS FUNCTION
DONALD P. MINASSIANAND JOHNW. GAISSER
Department
ofMathematical Indianapolis,IN 46208
Sciences,ButlerUniversity,
We giveperhapsone of themosteasilyunderstoodconstructions of a continuous,nowhere
(even one-sidedly)functiong on [0,1]. Brightstudentsin the uppersectionsof
differentiable
elementarycalculusshouldunderstand thereasoning.Our exampleis a newvariantof one given
by van derWaerdenin 1930 [1].
2
(xu, g2(xu))
2 2 40
92
Letthegraphoff1 g1be theunionofthelinesegments from (0,0) to(2, 2) andfrom(2, 2)
to (1,0). Let f2have fivecongruentteethwithheightof each tooth(?)2. Let g2 -fi + f2.
eachf1
Continuein a similarmanner(see illustration); has fivetimesas manyteethas f,1land is
onlyhalfas high;define
This content downloaded from 128.6.218.72 on Wed, 20 Jan 2016 23:24:09 UTC
All use subject to JSTOR Terms and Conditions
1984] THE TEACHING OF MATHEMATICS 255
g.- fj gi-I +fl.
j=1
Theng limgi is thedesiredfunction.
We firstshow thatg, being the uniformlimitof a sequenceof continuousfunctions, is
Fix x in [0,1], and letE > 0 be given.Choosei largeenoughso thatEJ= + I1) J < e/3
continuous.
Then
00 ~~~~00
ig(y) -gi (Y)l E t(Y)4 E (2 F/3
j=i+ 1 j=i+ 1
forall y in [0,1]. (Hence gi convergesto g uniformly).Sinceg, is continuousat x, we maypick
8> 0 suchthatIgi(x') - gi(x)l < e/3 forall x'c[O,1] n (x - 8, x + 8). Hence
ig(x') - g(x)I < Ig(x') - gi(x')I + Igi(x') - g,(x)I + Igi(x) - g(x)I
< e/3 + e/3 + e/3 = -
whenever x' E [0,1] n (x - 8, x + 8). So g is continuous.
The restof thisnoteshowsg does nothavea one-sidedderivative anywhere.
The absoluteslopeoff,is (5/2)i-1. [Eachtimewe halvethemaximum heightand quintuplethe
frequency. It is thiscombination of halvingtheheightand quintupling thefrequency that,as will
be seen,makestheslopesof thepartialsumsbelowincreasewithoutbound.Thisis whatrenders
thisexampledifferent fromvanderWaerden's.] Thisexceedsthemaximum slopeofg, 1fJ,
whichis 02 (5/2)J= (2/3)[(5/2)'-1 - 1]. Hence theslopeoffidominatesin forming theslope
ofgi g,_1 + fl. So, sincetheslopeoff,is alternately positiveand negative, so is thatofg,-and
over the same (equal) intervalson thex-axis(add f3 to g2 in theillustration to verifythisfor
g3= 92 +f3). Thus all abscissasfor "upcorners"(teethhighpoints) forf, are abscissas for
" upcorners" forgi; similarlyfor" downcorners." Also (see illustration),
upcorners forgi1 remain
upcorners forg, exceptthattheygetsharperand higher;thesameholdsfordowncorners, except
thattheydon'trise.Of course,g, has fivetimesas manyupcorners as g, 1.
Thus(see illustration) whenwe addf3to g2 to formg3,upcorner(x", g2(x ,)) risesby at least
as muchas (x, g2(x)) whichrisesby at leastas muchas downcorner (Xd, g2(xd)), whichrisesnot
at all. So, considering slopesof secants,we see that
9g3(X.) - g3(X) > 9g2(X.) - g2(x)
whichalso holdsif"d " replaces"u ". Thiscontinuesto holdforthesamex", x, Xd as we increase
theindexi in gi, so it holdsin thelimitg; i.e.,secantslopescan onlygetworse:
LEMMA.If (xl, gi(xJ))and (X2, gi(X2)) are adjacentcorners gi, thenforany
foranyparticular
point(x, gi(x)) on thejoininglinesegment:
( *) I g(x1) - g(x) a gi(X) - g,(X)
and thesameholdsifx2 replacesxl.
We arenearlydone.Supposeforsomex in [0,1],g had a right-hand sayL. Thenfor
derivative,
all b in some8-interval
to therightof x:
-
(**) L-1 < secant slope g(b) g(x) < L + 1.
Now as i increases,eventuallythereis g, and a < x < b < x + 8 whereA (a, g,(a)) and
B (b, g,( b)) are adjacentcornersforgi on whosejoininglinesegment
lies(x, gi(x)). Further,
if
m denotesslope:
This content downloaded from 128.6.218.72 on Wed, 20 Jan 2016 23:24:09 UTC
All use subject to JSTOR Terms and Conditions
256 DONALD P. MINASSIANAND JOHNW. GAISSER
_ (
(
I ImAB)IImf1
)I IM ? gi-i) forABI=
Im(f1) + m(gi-1) forABI >
Im(f)I - Im(g, l) forABI >
Im(fi ) I - maxIm(g1 ) I = (see fifth
paragraphof thispresentation)
(5/2)1i- 2/3 [ (5/2) i- _] 1/3 (5/2) ??
So we couldhavechosen(and do nowchoose)i largeenoughso Im(AB)I> ILl + 1, and by the
Lemma(in (*) takexl b):
g(b) - g(x)
(* *). A similarargument
contradicting showsthatg has no left-hand at anyx in [0,1]
derivative
Q.E.D.
Reference
1. B. L. van derWaerden,
Ein einfaches
Beispieleinernichtdifferenzierbaren
stetigen
Funktion,
Math.Z., 32
(1930).
MISCELLANEA
125.
(A) The mostabsurdscience,in myopinion,and the mostcapable of stiflingeverysortof
talent,is geometry.
Thisridiculous
sciencedealswithsurfaces,
linesand points,whichdo notexist
in Nature.One can imagineinserting
a hundredthousandcurvesbetweena circleand itstangent,
wherein realityone couldnotinsertthethinnest
straw.
-Voltaire, Jeannotet Colin(1764)
(B) A geometercan proveto youthatyoucan drawinfinitely manycurvesbetweena circleand
a tangent,butnota singleline:youreyesand yourcommonsensesay theopposite.The geometer
willanswersolemnly thatthat'san infinityof secondorder....
You consulta morehonestgeometer, who explainsthe mystery. "We imagine,"he says,
"thingsthatareimpossiblein Nature,linesthathavelengthbutno width.... No actualcurveor
line can pass betweentwo actuallines thattoucheach other:thesearejust intellectualgames,
imaginary chimeras.Truegeometry is theartof measuring
thingsthatexist."
Voltaire,L'Hommeaux quaranteecus(1768)
ANSWERSTO PHOTOS ON PAGE248
Top, J.E. Littlewood;bottom,G. H. Hardy.The Littlewoodpictureis fromthecollectionof
GeorgeP6lya; theHardyphotograph was takenby R. P. Boas in about1939.
This content downloaded from 128.6.218.72 on Wed, 20 Jan 2016 23:24:09 UTC
All use subject to JSTOR Terms and Conditions
