Problems

Author(s): Allen J. Schwenk, Mihaly Bencze, R. S. Luthar, Huseyin Demir, Nick Lord, Neil
Eklund, Chris Caldwell, Murray S. Klamkin, Andy Liu, Ronald L. Graham, Con Amore
Problem Group, Keith Loseke, David Callan, Fullerton Problem Group, University of
Wyoming Problem Circle, Jerry M. Metzger, Ravi Vakil, James O. Chilaka and Reiner Martin
Source: Mathematics Magazine, Vol. 64, No. 2 (Apr., 1991), pp. 131-138
Published by: Mathematical Association of America
Stable URL: http://www.jstor.org/stable/2690766 .
Accessed: 23/01/2015 12:20

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
Mathematics Magazine.

http://www.jstor.org

This content downloaded from 128.235.251.160 on Fri, 23 Jan 2015 12:20:45 PM

All use subject to JSTOR Terms and Conditions
 P RO BL EM S
LORENC. LARSON,editor
St. OlafCollege

GEORGEGILBERT,associateeditor
Texas Christian
University

Proposals
To be consideredforpublication,solutions
shouldbe receivedby September1, 1991.

1368. Proposed by Allen J. Schwenk,WesternMichigan University,

Kalamazoo,
Michigan.
In mykitchenI have a set of threemeasuringcups withcapacitiesc, 1 - c, and 1
where 0 < c < 1/2. Startingwiththe largestcup full,I can measureseveral addi-
tionalfractions of a cup by sequentiallypouringfromone cup to another.I always
pouruntileitherthereceivingcup is fullor untilthedispensingcup is empty.I never
spill,waste,or consumeanyoftheliquid.Definetheaccuracyofthisset ofcups to be
the lengthE ofthelongestinterval(a, a + e) in [0,1] forwhichno x E (a, a + 8) can
be measured.For example,when c = 1/n I can measure2/n by fillingthe smallest
cup, transferringthis 1/n to the middlecup, refillingthe smallest,and transferring
again. Clearlyeach fractioni/n can be measuredin this way, so for c = 1/n the
accuracyis also 1/n.
For each value of c, findthe accuracyof the set.
1369. Proposedby MihalyBencze,Bra,ov, Romania.
a. Find all naturalnumbersx, y,z such that3x + 4Y = 5z
b.* GivennaturalnumbersA > B and C, findall naturalnumbersx, y,z such that

((A2 - B2)C) + (2ABC)Y = ((A2 + B2)C)

ASSISTANT EDITORS: CLIFTON CORZATT, BRUCE HANSON, RIcIIARo) KLEBER, KAY SMITII and TIIEO13ORE
VESSEY, St. Olaf Collegeand MARK KRUSEMEYER, CarletonCollege.We invitereadersto submitproblems
believed to be new and appealing to studentsand teachersof advanced undergraduatemathematics.
Proposalsshouldbe accompaniedbysolutions, ifat all possible,and byanyotherinformationthatwill assist
theeditorsand referees.A problemsubmitted as a Quickieshouldhave an unexpected, succinctsolution.An
asterisk() nextto a problemnumberindicatesthatneithertheproposernortheeditorssupplieda solution.
Solutionsshouldbe written in a styleappropriateforMathematics Magazine.Each solutionshouldbegin
on a separatesheetcontaining thesolver'snameandfull address.
Solutionsand new proposals should be mailed in duplicate to Loren C. Larson, Departmentof
Mathematics, St. Olaf College,Northfield,MN 55057.

131

This content downloaded from 128.235.251.160 on Fri, 23 Jan 2015 12:20:45 PM

All use subject to JSTOR Terms and Conditions
132 MATHEMATICS MAGAZINE

1370. Proposed by R. S. Luthar, Universityof Wisconsin Center, Janesville,

Wisconsin.
a. If x, y, and z are positivereal numberswithx + y + z = 1, provethat

2 ln(x5 + y5 +z5) 5
3 ln(x7+y7+z7) <

b. Generalizethe resultofparta.
1371. Proposed by HiiseyinDemir, Middle East Technical University,
Ankara,
Turkey.
Let A, B, and C be verticesof a triangleand let D, E, and F be pointson the
sides BC, AC, and AB, respectively.Let U,X,V,Y,W,Z be the midpointsof,
BD, DC, CE, EA, AF, FB. Provethat
respectively,

Area( A UVW) + Area( A XYZ) - Area( A DEF)

is a constantindependentof D, E, and F.
1372. Proposedby NickLord,TonbridgeSchool,Tonbridge,Kent,England.
For whichangles0, a rationalnumberofdegrees,is it thecase thattan20 + tan220
is rational?

Quickies

Answersto the Quickiesare on page 137.

Q775. Proposedby Neil Eklund,CentreCollege,Danville,Kentucky.
Evaluate the integral
ixk_ 1d.
lnx dx.

Q776. Proposedby Chris Caldwell, University of Tennesseeat Martin,Martin,

Tennessee.
Show thatif P(x) is a real polynomialwithreal zeros,then P'(X)2 > P(x)P"(x) for
all real x.
Q777. ProposedbyMurrayS. Klamkinand AndyLiu, University ofAlberta,Edmon-
ton,Alberta, Canada.
T1 and T2 are twoacute trianglesinscribedin the same circle.If the perimeterof
T, is greaterthanthe perimeterof T2, mustthe area of T, also be greaterthanthe
area of T2?

This content downloaded from 128.235.251.160 on Fri, 23 Jan 2015 12:20:45 PM

All use subject to JSTOR Terms and Conditions
VOL. 64, NO. 2, APRIL 1991 133

Solutions

A Curious Recursion April 1990

1343. Proposedby Ronald L. Graham,AT&T Bell Labs, MurrayHill, New Jersey.
a. Whatis thebehavior
oftherecursive
sequencedefined
byxn+2 = (1 + X,1+1)/Xn,
withxo,x1 arbitrary positivenumbers?
b. Answerthe same questionforx +3 =(1 + xn+ + xn2)/xn.
c.* Generalize(a) and (b).
Solutionby Con AmoreProblemGroup,Royal Danish Schoolof EducationalStudies,
Copenhagen,Denmark.
a. The sequence maybe tracedbackwardsusing x,,= (1 + x + 1)/X n+2 and there
is no harmin extendingthisto negativeindices.We thenhave

l+X1 1+ (1 +x )/x0 1 +x0+X1

X2 x =
xo XI xoxl

as well as
1+X0 1+ (1+x0)/x _ 1+x0+x
x XI x_2 = xo x oxI

So x-2 = X3. Since it is obviouslyimmaterial

whichtwo successivetermsare called
xo and xl we concludethatforall integersn, Xn = Xn+5. That is, the sequence is
periodicwithperiod5.
b. We proceedas in parta.

1 + xi + X2 1 + X2 + ((1 + X1 + X2)/Xo) 1 + X0 + XI + X2 + XOX2

xo '1 xl xo

1 + X0 + xi + x0 + + XO + X1)/X2) 1 + X0 + X1 + X2 + XoX2
X-1 X2
'
-2 xl x1x2

So X_2X2 = X0X4. As abovewe concludethatXnXn+4 = Xn +2Xn+6 forall integers

n.
But we thenalso have xn +2xn+6 n+8' so XnXn+4 = xn +4xn+8, whence xn
= Xn+4x

xn+8 (since xn+4 is positive).That is, the sequence is periodicwithperiod8.

Also solved, in some cases using Mathematica or MAPLE, by Charles Ashbacher, Duane Broline,
Francis M. Henderson, R1. High, Kee-Wai Lau (Hong Kong), Stanley Wu-Wei Liu, Reiner Martin
(Germany), Mike Pinter, Robert L. Raymond et al., Volkhard Schindler (Germany), Heinz-JiirgenSeiffert
(Germany), Stephen L. Snover, John S. Sumner, University of Wyoming Problem Circle, Ravi Vakil
(student, Canada), Michael Vowe (Switzerland), XAMPLE (Xavier Mathematics Problem League), and the
proposer. There was one partially correct solution.
Con Amore Problem Group, Henderson, and Sumner investigated the possibility of the sequence having
a shorterperiod(1 in parta and 1, 2, or 4 in partb) forspecificinitialvalues.For example,in partb, period
2 occurswhen x2 = xO > 1, xI = (xO + 1)/(Xo - 1).
Severalsolversinvestigated possiblegeneralizations, rangingfrom

xn+k = (1 + Xn++ '. . +Xn+k-l)/Xnf

throughsimilarexpressionsin whichthe xi are movedto the denominator. Con AmoreProblemGroup,

High, Liu, Raymondet al., and Snover reported"pseudoperiodicity," usuallyon the basis of computer
experiments.Specifically,for k = 4 and startingvalues such as x= =x2 = x3= 1, the sequence
"almost" has period 11. Liu presented an especially detailed discussion with various observations,

This content downloaded from 128.235.251.160 on Fri, 23 Jan 2015 12:20:45 PM

All use subject to JSTOR Terms and Conditions
 134 MATHEMATICS MAGAZINE

conjectures,and numericalevidence about the behaviorof recurrencessuch as x,, 4 = (1 + +,3

x +
xn + )/(X n+2 + x11)(and manyothers).However,no solverfounda generalization
of the resultsofpartsa
and b, and severalspeculatedthatno such generalization
mightexist.

Generalizationof ged-lemFormula April1990

1344. Proposedby KeithLoseke,student,BuhlerHigh School,Buhler,Kansas.
Provethe following
generalization
of the formulagcd(a, b) = ab/lcm(a, b):

pO
gcd(aI,a2,...Iak)-
PE

where aI, a2,. ak are positiveintegers,"gcd" and "lcm" are abbreviationsfor

greatestcommondivisorand leastcommonmultiple,PO is theproductofthelcm'sof
all subsetsof {al, a2,..., ak) withan odd numberof elements,and PE is the similar
productof the non-empty subsetsof even numbersof elements.
Solutionby David Callan, University of Wisconsin,Madison,Wisconsin.
For 1 < i ? k we can let a =H11> pJi',where the pj's are distinctprimesand
Cij > 0. Then forI a subsetof {1,2,..., k),
n i

lcm{ai}ie1= 7JpmaxieI{cij} and gcd{ai})ic- 1pniniE={cij)

j=1 j=1

Fix j and denote cij by ci. Withoutloss ofgenerality,

assume cl < c2 S* < cn.
Then we mustshowthat

E max{cs}- E max{ci}=cl.
I:l I odd iI I: Ill even i I

Clearlycl occursoncein thefirst sumon theleftsideandnotat all in thesecond.

For i> 1, ci+1 occurs i + ( + ( + * 2- 1 timesin thefirstsumand ()
+ (3) + = 2i-1 timesin thesecond,
andhencecancels.The resultfollows.
Also solved by Duane M. Broline,Con AmoreProblemGroup (Denmark), David Doster, Newcomnb
Greenleaf,RussellJay Hendel, R. High, Kee-Wai Lau (Hong Kong), SandorLehoczky(student),N. J.
Lord (England), David E. Manes, Jean-MarieMonier (France), Allan Pedersen (Denmark), Vivek
Prabhakaran(student),Brian Ruder (student),Heinz-Jiirgen (Germany),JohnS. Sumner,Ravi
Seiffert
Vakil(student,Canada), Ellen Veed, University ofWyoming ProblemCircle,XAMPLE (XavierMathemat-
ics ProblemLeague), Paul J. Zwier,and theproposer.
Manes notes thatthis problemis Theorem7.4, p. 11, in L. K. Hua's book Introduction to Number
Theory,Springer-Verlag, New York,1982. Noticealso the relatedresult,Theorem7.3: lem (al, a2,..., ak)
= Go/GE, whereGO is the productof the gcd's of all subsetsof {a,, a2,.. ., ak) withan odd numberof
elementsand GE is thesimilarproductofthenonempty subsetsofeven numbersofelements.This related
resultalso appearsas Problem29, p. 117,and p. 308, in G. P6lyaand G. Szegb, Problemsand Theoremsin
AnalysisII, Springer-Verlag, New York, 1976, where the proof is based on the inclusion-exclusion
principle.

Limitofan InverseImage April1990

1345. Proposed bytheFullertonProblemGroup,California
StateUniversity,
Fuller-
ton,California.
Let A(?) be theunionofall maximalboundedclosedintervals
of(0,1) suchthat
-E < sin(l/x) < E foreveryx E A(s). Denote by S(W)the sum of the lengthsof all
theseintervals.
Provethat
1
lim S( ?)

This content downloaded from 128.235.251.160 on Fri, 23 Jan 2015 12:20:45 PM

All use subject to JSTOR Terms and Conditions
VOL. 64, NO. 2, APRIL 1991 135

Solutionby the Universityof WyomingProblemCircle, University

of Wyoming,
Laramie,Wyoming.
If 0 <e < 1, then the set of y's for which -s < sin y < s is the union of the
intervals [k r - a(s), k-7r+ a(s)] as k ranges over the set of integers. Here, a(s)
sin1 E. It follows that

A() kU [ k + a(s) ' kw-a(s)I

k=1-\

so that

S(F) 1 f ( 1 - 1 8

E 8 k 1
- a(s) krT+a()
4 )
1 2a(?)
k=E k21rT2-a2(s)
a(s)
_ 2 1
8 72 k= 1 k2 _a )

72

Now, by L'Hopital's Rule,

lim =lim s = lim 1

- ,0 E
? ?0 ? -----0 1

Also, the series defines a continuous functionof E on the interval [0, 1] since it is a
uniformlyconvergent series of continuous functions.In fact,the Weierstrass M-Test
can be applied since

1 1

k_a 2(\2
(s?) k2 1

and the series E(k2 - (1/4))0-1 converges. Consequently,

0 + ? 7 kE
6 ~2 2 1 = 2 rT2 I
si __

Also solvedby Seung-Jin

Bang (Korea), S. F. Barger,LarryBlaineand T. A. Giebutowski,
Paul Br-ackeni
(Canada), David Callan, Con AmoreProblemGroup (Denmnark), Jesse Deutsch, David Doster, Robert
Doucette,MordechaiFalkowitz(Israel), H. Guggenheimer, Kee-Wai Lau (Hong Konag),SandorLehoczky
(student), Peter W. Lindstrom,N. J. Lord (Englanad),Reiner Martin,Jean-MarieMonaier(France),
WilliamA. Newcomb,StephenNoltie,StephenG. Pentrice, C. Ray Rosentrater,Heinz-Jiirgen Seiffert
(Germany),JohnS. Sumner,Gan Wee Teck (student,Singapore),WesternMarylandCollege Problenm
Group,Michael Vowe, and theproposer.

This content downloaded from 128.235.251.160 on Fri, 23 Jan 2015 12:20:45 PM

All use subject to JSTOR Terms and Conditions
136 MATHEMATICS MAGAZINE

SolitaireEquivalence April1990
1346. Proposedby JerryM. Metzger,University of NorthDakota, Grand Forks,
NorthDakota.
Considera solitairegameplayedas follows.Froman ordinary deck of52 cards,lay
out the top 8 cards face up in a row. If two or morecards have the same rank(for
example,two eightsor two jacks), pick two of themand cover themby two cards
dealt faceup fromthe top ofthedeck.Continuein thiswayas longas at least twoof
the cards showthe same rank.The game ends ifone cannotplay,and is a win ifthe
deck is exhausted.Provethatthegameis a win workingfromthetopto thebottomif
and onlyifit is a win withthe deck turnedover(thatis, workingfromthe bottomto
the top).
Solutionby Ravi Vakil,student,Universityof Toronto,Toronto,Canada.
The game is a win workingfromthe top of the deck ifand onlyif,forall even n,
thereare at most7 unmatchedcards in the top n cards of the deck. But thenthis
means thatthereare at most7 unmatchedcards in the bottom52 - n cards forall
even n. (Each unmatchedcard in the top n cardsgivesrise to an unmatchedcardof
the same rankin the bottom52 - n cards.)A game is a win if and onlyifthe game
withthe deck turnedoveris a win.
Also solved by Larry Blaine, Con Amore ProblemGroup (Denmark), Roger B. Eggleton(Brunei
Darussalam), R. High, Johnand LibbyKrussel,MaryAnn Martin(student),WilliamJ. Metzger,Adam
Riese, HarrySedinger,Paul J. Zwier,and theproposer.

Centerofthe GeneralLinear Group April1990

1347. ProposedbyJames0. Chilaka,LongIsland University, Greenvale,New York.
Let G = GL(n, R) be the groupundermultiplication of n X n matriceswithreal
entriesand non-zerodeterminant. Find theleastnumberk ofelementsg1,g2, . gk
in G suchthatif a is in G and a gi = gi a, i =1,2..., k, thena -g = g -a forall g
* *

n G.
Solutionby ReinerMartin,student,University of California,Los Angeles,California.
case, n > 2, the answeris k = 2.
In the non-trivial
Let g1 E G be the diagonalmatrixwithentries1,2,. n and let g2 = (Oio(j)) e G
be thepermutation matrixwithu(k) = k + 1 fork = 1 .., n - 1 and o(n) = 1. Now
let a-(aij) E G be arbitrary. Then a *g1 = g1 a implies aij = O for i ' j and
a g2 = g2 a impliesall = = ann,so bothconditionstogetherimplythata is a
scalarmultipleof the identitymatrix, hence a *g = g *a forall g E G.
It remainsto show that k cannotbe improvedto 1. Therefore,we assume that
thereexistsa g, E G with the propertythatif a E G commuteswith g, then a
commuteswith every element of G. Choosing a = g1, we see that an arbitrary
elementof G commuteswith gl. So every elementof G commuteswith every
elementof G, a contradiction to the factthatG is notcommutative.
Also solved by RicardoAlfaro,S. F. Barger, Duane M. Broline,David Callan, Con AmoreProblem
Group (Denmark), David W. Koster,MarvinMarcus and Susan Franklin,Daniel B. Shapiro,JohnS.
Sumner,WesternMarylandCollegeProblemsGroup,and theproposer.

This content downloaded from 128.235.251.160 on Fri, 23 Jan 2015 12:20:45 PM

All use subject to JSTOR Terms and Conditions
VOL. 64, NO. 2, APRIL 1991 137

Answers
to theQuickieson p. 132.
Solutions
A775. Solution1. Note thatforx > 0,

fetlnxdt= lnet = 14 (eklnxxx1 1) =

0 ln x t-O In x In x

Therefore,

xI l dx= f fetlnxdtdx.

we get
the orderof integration
If we interchange

1= k| xt dxdt= kt 1 dt=ln(k + 1).

Solution2. Let
ixk - x
g(k) = n x dx.

Then

g'(k) fxkdx 1

back,we findthatg(k)=
Integrating ln(k + 1).
Comment.This integralcan be foundin variousintegralhandbooks,and the case
k = 2 appearsin Mathematical
Spectrum,Vol. 22, no. 1, problem22.4, 1989-90.
A776. If x is one of the roots rl, r2,...,rn of P(x), then the resultis obvious.
Otherwise,we have
n
P ( X) I

P(x) x-ri

each side,
and differentiating

P ( X) P ,(X)- P,(
2
X))2
-E1~=1(x-
En - < O,
<0
P(X) ri)

whichcompletesthe proof.(Notice thatequalityholdsif and onlyif x is a multiple

rootof P.)
A777. By considering two triangles with angles (800,500,500) and (700, 700,400) the
answeris in the negative.
The resultwould be valid fortwo generaltrianglesif the angles of T2 majorized
thoseof Tl, thatis, if Al > B1 > C1, A2 >B2? C2, then A2 >A1 and A2+B2>A
+ B1. Then by the majorization inequality,

F(A1) + F(B1) + F(C1) > F(A2) + F(B2) + F(C2) (1)

This content downloaded from 128.235.251.160 on Fri, 23 Jan 2015 12:20:45 PM

All use subject to JSTOR Terms and Conditions
138 MATHEMATICS MAGAZINE

forconcavefunctionsF. The restfollowssince the perimeterand area of a triangle

ABC is given by 2R(sin A + sinB + sinC) and 2R2 sinA sin B sinC, respectively,
and sinx and lnsinx are concaveon (0, 7).
Comments.For the special case when T1 and T2 have a commonangle (or
equivalentlya commonside), thentheanglesof T2 majorizethoseof T1. It wouldbe
of interestto give an elementarygeometricproofof (1) forF(x) = sinx or lnsinx.

ProofwithoutWords:
AlgebraicAreas

(a+b +c) ++(a+b -c) +(a-b +c) ++(a-b -c) = 22(a2 + b2+c2)

a + b+ c

2c a +b-c

a-b+c~~~

2a-- * 2b 2c

-SAM POOLEY
K. ANN DRUDE
MILLSAPS COLLEGE
JACKSON,MS 39210

This content downloaded from 128.235.251.160 on Fri, 23 Jan 2015 12:20:45 PM

All use subject to JSTOR Terms and Conditions