A Gardner's Workout (Gnv64)

A Gqlcnelis WOlkout
1irrgtlc^rJrJ 'rtcrrgtlc'|it
A K Peters
Natick, Massachusetts
Editorial, Sales, and Customer Service Office
A K Peters, Ltd.
63 South Avenue
Natick, MA 01760
www.akpeters.com
Copyright 2001 by A K Peters, Ltd.
All rights reserved. No part of the material protected by this copyright notice
may be reproduced or utilized in any form, electronic or mechanical, including
photocopying, recording, or by any information storage and retrieval system,
without written permission from the copyright owner.
Library of Congress Cataloging-in-Publication Data
Gardner, Martin
A gardner's workout: training the mind and entertaining the spirit /
Martin Gardner.
p. cm.
ISBN 1-56881-120-9
1. Mathematical recreations. 1. Title.
QA95 .G 158 2001
793.7' 4--dc21
Printed in Canada
05 04 03 02 01 10 9 8 7 6 5 4 3 2 1
2001021427
To all the underpaid teachers of mathematics,
everywhere, who love their subject
and are able to communicate that love
to their students.
Tqbl e of Contents
Pefce xi
Pqr I 1
1 The Op'que Cube 3
2 The Squ'e Root of 2 1.414 213 562 373 095 ... 9
3 Flip, the Psychic Robot 21
4 The Poposition'l c'ieulus with Diede G'phs 25
5 M'them'tics 'n Worpl'y 35
6 Steiner Trees on ' Checkebo'r 39
7 Tiling the Bent Tromino 61
8 Covering ' Cube with Conguent Polygons 73
9 M'gic Tricks on ' Computer 77
10 V'i'tions on the 12345679 Trick 83
11 Moe c'ieul'tor Whimsies 85
12 !(sP'rov's DeFe't by Deep Blue 91
13 Computes Ne' the TheshoW 97
14 Comehng the King 109
15 T oroi'l Cunency 117
VII
VIII A Gardner's Workout
16 Six ChCllenging Dissection TCSks 121
17 Lewis CCrroll's Pillow-Problems 129
18 Lewis CCrroll's Wor LCers 133
19 The Ant on 1 x 1 x 2 139
20 Three-Point Tiling 143
21 Lucky Numbers Cn 2187 149
22 3 x 3 MCgic SCuCres 157
23 Some New Discoveries Cbout 3 x 3 MCgic SCuCres 167
24 Primes in Arithmetic Progression 175
25 Prime MCgic SquCres 181
26 The Dominono GCme 187
27 The Growth or RecreCtionCI MCthemCtics 191
28 MCximum Inscribe SquCres, RectCngles, Cn TriCngles 203
29
30
31
32
33
34
SeriCI Isogons or 90 Degrees
Aroun the SolCr System
Ten AmCzing MCthemCticCI Tricks
Moeling MCthemCtics with PICying CCrs
The Asymmetric Propeller Theorem
Chess Queens Cn MCximum VnCttCcke Cells
Pq r I I
35 Lion Hunting
36 Two Books on InAnity
215
231
235
241
249
257
269
271
275
Table or Contents IX
37 The Universe cmd the Te'cup 281
38 A New Result on PerFect M'gic SCu'res 285
39 The Number Devil 291
40 Prob'bility 1 295
41 Fuzzy New New M'th 301
Prefqce
.rzyearsuactueu.n.rancpleasure.rritin,tue:atuemat
i.al Games ..lumn in Scientifc Amercan. All tu.se..lumns uave
n.beenreprintec, ituupcatin,, inrineenv.lumes, startin, itu
The Scientifc American Book of Mathematical Puzzles and Diversions
ancencin,ituLast Recreations.
sin.e st.ppecritin,tue..lumnuaven.mtimet.time..n
tributec arti.les ancb..krevies ab.ut matuemati.s t.b.tua.ac
emi.].urnals ancp.pularma,azines r.rty.ne.rtuesepie.es are
,atuerecuere 1uem.st..ntr.versial is tuerinalrevieinui.u
.riti.izea.urrenttea.uin,rackn.nastuenene:atu
Hytuetimetuisb..kispublisuec.ulc,uessancu.petuatne
nematuisbein,abanc.necalm.stasrapiclyastue.lcnematu
racec ..ulcber.n, nany.ase, itmaybece.acesber.re.urpub
li.ecu.ati.n isablet. attra.t..mpetenttea.uers u.uavelearnec
u. t. tea.u matu t. pre..lle,e stucents itu.ut puttin, tuem t.
sleep 1uereare, .r..urse, many tea.uers u. ceserven.tuin,but
praise tist.tuemuavececi.atectuisb..k.
XI
Martin Gardner
Hendersonville, NC
Ch'pter 1
The 0
p'que Cu be
The Opqque Cube
antt.pr.p.setuer.ll.in,uns.lvecpr.blem Asraraskn.,
amtuerirstt. askit

wuatistueminimalarea.rsurra.esinsiceatransparent.ube
tuatillrencerit.paque:

Hy.paquemeanif te suraces are opaque, no ray of light, enter
ing te cube from any direction, wilpass trough it
1ueansermay .rmayn.tbetueminimal surace spannin,tue
telveec,es.rtue.ube. 1uis questi.n als. is unanserec seetue
cis.ussi.n.ritbyc.urantancr.bbins, l } , cu )
1uattueminimal spannin, surra.e mayn.tbetue ansert.tue
.paque.ubepr.blemissu,,estecbytuera.ttuattueminimalsteiner
tree spannin, tue r.ur ..rners .ra square isn.ttue anser t. tue
.paquesquarepr.blem 1uebestkn.ns.luti.nr.rtuesquare,als.
n.t pr.vec minimal) is su.n in i,ure l 1ue square pr.blem is
cis.ussecbyr.ssn.nsber,er, z} ,p zz)
mybests.luti.nr.rtue.ubeist.].intue.entert.alltue..rners
1ueselines.utline lztrian,lesitua..mbinecarea.r+,
believetue.paque.ubepr.blemt. beextremely cirri.ult t is
keepin,meaakeatni,ut '
IThe note appeared i n the Dutch periodical Cubism for Fun (No. 23, March 1990,
p. 15). I followed this with a second note, "The Opaque Cube Again, " i n the same
periodical (No. 25, December 1990, Part 1, p. 14) .
3
4 A G<rd nds Workout
Figue 1. Total length 2.639+.
Refrences
[ I ] R. Courant and H. Robbins. Wat is Mathematics? , 4th edition, Oxford Univer
sity Press, 1 947.
[2] R. Honsburger. "Mathematical Morsels, " The Mathematical Association of Amer
ica, 1 978.
The Opqque Cube Agq i n
s.raraskn., n..neuasyetpr.vectuattues.luti.nr.rtue.paque
squareisminimal,n.ruasany.nepr.vecaminimalsurra.espannin,
tueec,es.ra.ube.
rerert.tuepr.blemp.secincrz-,mar.u l vvo) ,p l inc
tue leastarea surra.e tuat ill bl..k any li,ut ray tryin, t. pass
tur.u,uaunit.ube
r.m stepuen narvey, Dunecin, Ne zealanc, re.eivec a .al
.ulati.n .rtue area .r a surra.e su.u as pi.turec in ri,ure z+o .r
c.urantancr.bbins Wat Is Matematics ?'. 1uisarea(A ! z!z)
is sli,utlylesstuan+
r.mn s m c.xeter,1.r.nt.,canaca,re.eivecaletterinui.u
ue su.stuat tue spannin, surra.e .rc.urant anc r.bbins.ann.t
uavestrai,utec,es, ui.umakestuisareaveryciui.ultt...mpute.
Lventuissurra.e.anbeimpr.vecsli,utly, u.ever,asissu.nby
kennetuHrakke,matuemati.sDepartment,susqueuannatniversity,
selin,s,r.ve, rennsylvania ne uas amarvel.us ..mputerpr.,ram
tuatsear.uesr.rminimalsurra.es. ner.uncaminimalsurra.espan
nin,tue.ube sec,estuatuasan area.rA ! z+z1 1uisistuebest
s.luti.nyetr.rtueOpaquecube
1. The 0P.'ue (ure
Opqque Cubes by Kenneth Btqkke
Figure 2. Twelve triangles from the
edges to the centre. Area 3v ~
4.2426.
Figure 4. Generalization of opaque
square sol ution to three di mensions.
Topologically, this is the opaque square
solution stretched vertically with the
top and bottom faces of the cube added.
Area, 4.2343.
Figur 3. Soap flm formed by dipping a
cubical frame in a soap sol ution. Area
' 4.2398.
Figure 5. My best solution. Thi s i s
l i ke Figure 4, except i t has three-fold
symmetry i n place of two-fold. Area
' 4.2324.
5
kennetu Hrakke ill ,ive a ruller treatment .ruis results else
uere meanuileue kincly all.ec tue ecit.rs .rcrt. publisu
s.me.ruis..mputerprint.uts, seeri,uresz) rtueyinspireany
reacert.rincabetters.luti.n,ue.ulcliket.uearab.utit
Attuem.mentitseemsunlikelytuattueminimums.luti.nr.rtue
.paque.ubeill..nsist.r,cis..nne.tec)surra.estuatc.n.tspantue
.ube sec,esitua sin,le surra.e n.ever, tuisisrarn.mpr.vec
Postso- ipt
1uepi.ture.rHrakke sbests.luti.nt.tue.paque.ube, repr.cu.ec
uereasi,ure,macetue..ver.rThe Amercan Mathematical Monthly
v.l vv, ,N.vember l vvz) 1ue..verillustratecHrakke spaper,1ue
Opaque cube rr.blem, tuatran in ri.uarc Guy s tns.lvecr.b
lems..lumn Hrakkeals.cis.ussestue.paquespuerepr.blem
1ue.paquesquare, ituitssin,lesteinerp.int,.bvi.usly,ener
alizest..paquere,ularp.ly,.ns. 1uelimitin,.ase, astuenumber.r
p.ly,.n sicesin.reases, is, .r..urse, tue.ir.le. 1ue.paque.ir.leis
pr.bablys.lvecbyaren.e.rlen,tuz-tuatistuelimit.raninrinite
series 1ues.lverisHerncka.ul,apr.ress.r.rmatuemati.sattue
tniversity.rc.l.,ne smatuemati.alnstitute. nis s.luti.n is ,iven
in1ueOpaquesquareanctueOpaquecir.le, intuepr..eecin,s.r
a..nreren.ein Ober.lra.u,reprintecintue International Seres of
Numercal Matematics, v.l lz-, l vv), pp --v-+s
i,uresanc su.tuebestkn.ns.luti.nsr.rtue.paquepen
ta,.n anc uexa,.n N.te tuatea.u ist.tallyla.kin, in symmetry
ka.ul s..n]e.turecs.luti.nr.rtue.paque.ir.leisbasec.nextrap
.latin,rr.mtuep.ly,.nal.ases seeals.ka.ul sarti.lesymmetry
.rN.t:, inThe Mathematical Intelligencer, v.l. zo, l vvs), pp l szz,
inui.ub.tutuesquareanc.ir.le.asesarecis.ussec
1 . The 0p'que Cube
. .
" ..,

Figure 6. The opaque pentagon. Length 3. 528.
| `
|
|
|
|
1-
/:
|
|
|
|
l
.
.
|
|
|
|
|
|
:

: .
:
.
. .
| .
( .
:
Fi gure 7. The opaque hexagon. Length (7 + y)/2 4.366.
7
Chqpter 2
The Squq re Root of 2
1.41 4 21 3 562 373 095
r.sesarerec,
vi.letsareblue.
Onep.int+l +. .
stue squarer..t.rt.
..nresstuat r.te tue ab.ve]in,le .nly t. uave s.me li,utverse
t.p tuis arti.le 1ue c.ts at tue enc .rtue tuircline inci.ate tuat
tuece.imalrra.ti.nisenclessancn.nrepeatin, n.tuer.rcs, v
is irrati.nal. Altu.u,u its ce.imal ci,its, liketu.se.r.tuer ram.us
irrati.nalssu.uas ancL look likeasequen.e.rranc.mci,its,tuey
arerarr.mranc.mbe.auseiry.ukn.uattuenumberisy.u.an
alays.al.ulatetuenextci,itarteranybreakintuesequen.e su.u
irrati.nalsals.su.ulcn.tbe.allecpatternlessbe.ausetueyuavea
patternpr.vicecbyanyr.rmulatuat.al.ulatestuem 1uesquarer..t
.rt.,r.rexample, is tue limit .rtue r.ll.in, ..ntinuec ,encless)
rra.ti.n
v
=
-- ;--
z
1
z
z1
r.mtuis..ntinuecrra.ti.n.ne.anceriverati.nalrra.ti.ns,rra.
ti.ns ituinte,ers ab.ve ancbel.tueline)tuat,iveV t.anyce
sireca..ura.y. 1uesequen.el/l ,-/z, /, l / l z, +l/zv, vv/o, z-v/l sv,
/+os, l -o-/vs . is s.metimes.allecLuc.xus laccer arter an
an.ientGreekastr.n.mer anc ,e.metri.ian 1uerra.ti.nsare alter
natelyui,uerancl.ertuantueirlimit,ui.uisV. La.urra.ti.nis
.l.sert.tuanitsprece.ess.r 1uebestappr.ximati.nitunumer
at.r anc cen.minat.r n.tex.eecin,turee ci,its is/+os t,ives
This article frst appeared in Math Horzons (April 1997) .
9
J t.rivece.imalpla.es rarra.ti.nintuissequen.eisrepresentec
byo,/, tue nextrra.ti.n illbe(o+ z/,}(o+ /, N.tetuan .n ea.u
run,.rtuelaccertuenumerat.ristuesum.ritscen.minat.ranc
tuecen.minat.r.rtuepre.ecin,na.ti.n
Davicwells, inuis Penguin Dictionary of Curous ad Interesting
Numbers ,pa,es-+-),ivess.me stran,epr.perties.rtuemultiples
.rJ. r.rexample,ritein alinetuemultiples, .mittin, tuena.
ti.nalpart r.rexample, l times J, i,n.rin, tuece.imalci,its is
l 1i.e J, i,n.rin, tue ce.imals, is z ntuis ayy.u.btaintue
r.ll.in,sequen.e l , z, +, , , s, . .
Heneatu tuis sequen.e put c.n tue numbers missing n.m tue
rirstsequen.e
l z +
- s l o l -
s v l l l z
l zo z- z -o
1ue ciueren.e beteen tue t.p anc b.tt.mnumbers at ea.u
_
p.siti.nis alaysti.e.
NOl mq l Nu mbels
Any
_
r..t .ra p.sitive inte,er ,in alltuat r.ll.s inte,er ill

meanap.sitiveinte,er)n.titselrannth p.erisirrati.nal Altu.u,u
allsu.uirrati.nalr..tsuavece.imalci,itstuatareneituerranc.m
n.rpatternless, tueyareall, s.rarasany.nekn.s, n.rmal 1uis
meanstuatiry.uspe.iryanypattern.rci,its, su.u asasin,leci,it,
pairs.rac] a.ent ci,its, triplets .rac] a.ent ci,its, anc s..n, intue
l.n, run tue pattern ill appear itu]ust tue rrequen.y y.u .ulc
expe.t .n tue assumpti.n tuattue pr.bability .rrincin, any ,iven
ci,itatany,ivenpla.eisalays l/ lo.
1uepattern neecn.tinv.lveac] a.entci,its 1uey.anbespa.ec
any ay y.ulike r.rexample, y.u mi,ut..nsicertue pattern o/,
uere o anc / are separatec by, say, seven ci,its, anc / anc are
separatec by, say, l oo ci,its. All tests s. rar t. cetermine tue rre
quen.y.rsu.upatterns uavesu.ntuat allirrati.nalr..ts, inany
basen.tati.n,aren.rmal
1ue m.st extensive tests r.r tue n.rmal.y .r.ertain irrati.nals
uavebeenmacer.r be.ause uasn.been.al.ulatect.uuncrecs
.rmilli.ns.rci,its, butsimilartests.r.tuerram.usirrati.nalssu.u
asL anctue,.lcenrati.uavesu.nn.ceviati.nsrr.mn.rmal.y. c.
n.tkn.u.rarJ uasbeen.al.ulatec, tu.u,uuavearereren.e
2. The Squ<re Root or 2 1 . 414 213 562 373 095 ... 11
,1tML, O.t.ber z, l vl ) t.ituavin, been .arriec t.m.re tuan a
milli.nci,itsin l v l by1a.quesDutka,tuen ac.lumbiatniversity
matuemati.ian
Onemi,utima,inetuatallirrati.nalsaren.rmal,butitiseasyt.
seetuattuisisn.ttue.ase. Ap.pularexampleistuebinaryrra.ti.n
l ol ool oool oooo . . 1uenumber.learlyis n.trati.nalanc]ustas
.learlyisrarrr.mn.rmal
v
,
d DIOWh i hg qt Seq
1uecis..very.rirrati.nalr..tsasrirstmacebytuerytua,.reans, a
se.retbr.tueru..ctuatrl.urisuecinan.ientGree.e 1ueircis..very
.rtue rirstirrati.nal number, tue squarer..t.r z, as amilest.ne
intueuist.ry.rmatuemati.s n,e.metri.alr.rmtuissaystuattue
cia,.nal.raunitsquareisin..mmensurableitutue square ssice
N.ruler,n.matteru.rinely,racuatec,.ana..uratelymeasuretue
t.linese,ments. rtuesice.rasquareisrati.nal,tuecia,.nalill
beirrati.nal, ancvi.eversa
1uerearet.le,encsab.uttueexpl.siveerre.t.rtuiscis..very
One is tuat arytua,.reannamecnippasusas s.rn n.tt.reveal
tuecis..verybe.auseitsuatterectuerytua,.reanbeliertuatinte,ers
a..uratelymeasure alltuin,s. nippasusbr.ketue v. Asaresult
ue as cr.nec at sea eituer by sui.ice, murcer, .r tue ratu .r
tue,.cstuele,encuasmanyvariati.ns 1ue.tuerle,encuastue
ytua,.reans .elebratin, tueir ,reat cis..very by sa.riri.in, many
.xen t. tue ,.cs 1ue cis..very .rin..mmensurable line se,ments
uacapr.r.uncinrluen.e.nrlat.,u.r.teab.utitinuisLaws.
I hfh ite Desceht
1ueGreekspr.vectuein..mmensurability.rasquare ssiceancci
a,.nal by a .leverinriniteces.entpr..rusin, tue cia,ram su.n
ini,ure l
Assumetuattuesice.rtuelar,estsquareis..mmensurableitu
itscia,.nal rs., ea.u.rtuet.linese,mentsillbemultiples.r
aunittuate.all k. Draasmallersquare.rsiceb, .u..sin,p.int
J s.tuata c. siceb .rtuis squareillbe..mmensurableituits
Figure 1. An infinite descent proof that y i s irrational.
cia,.nal be.ause ea.u is a multiple .r k. Next e sele.t p.int y s.
tuatd b. A,ain,tuesiceanccia,.nal.rtuissmallersquareillbe
..mmensurableiturespe.tt. k.
1uispr..ess.anbe..ntinuect.inrinityassu,,estecbytuer.urtu
tinysquare 1uesices.ralltuesesquares.ann.tbezer.,butats.me
p.intintue encless..nstru.ti.n erea.ua squareitu a siceless
tuan k. A len,tulesstuan k .ann.tbe a multiple .rk, s. euave
en..untereca..ntraci.ti.npr.vin,tuat.urassumpti.n,tuattuesice
anccia,.nal.rasquareare..mmensurable,isralse. rtuesquare s
siceis l , tuecia,.nalis weuavesu.ntuatisirrati.nal.
we.anexpresstuepr..ran.tueray weseemt.,etaninrinite
series.rinte,ers,multiples.rk), ea.usmallertuantueprevi.us.ne,
butsu.uaseries.bvi.uslymustberinite
nu,. steinuaus, intuerirst.uapter .rMathematical Snapshots,
,ivesaciuerent,e.metri.alpr..rbyinriniteces.ent tis basec.n
tuere.tan,le su.ninri,ure z. ts sices areinarati. su.utuatir
tuere.tan,leis sli.ec inualrassu.n, ea.uualrillbeare.tan,le
similar t. tue .ri,inal .ne. rtue sices are labelec as inci.atec, a
anc b ill beintuesamerati.asa/2 ancb. 1ueequati.nrecu.est.
a
2
2b
2
, s.irb 1, a ill be
Assumetuat a anc b are ..mmensurable, ea.u sice amultiple.r
unitk. Or..urse, k .anbeanyunit,in.ues, .entimeters, .ruatever
ni,ure-euaveatta.uect.tuel.n, sice.rre.tan,leab a..n
,ruentre.tan,letuatuasbeen,iven aquarterturn.l..kise. 1uis
pr.cu.es a lar,er re.tan,le .r sices b anc (a + b) . Hy .uttin, t.
squares .r sice b rr.m tuis lar,e re.tan,le e pr.cu.e tue smaller
2. The Square Root at 2 1 . 414 213 562 373 095 ... 13
b
al2
a
Figure 2. Another infinite descent proof.
suacecre.tanie. tssicesareb anc(a - b). He.ausea ancb areinte
,ers, (a - b) mustais.beaninteer. 1uerer.retue suacecre.tan,le
mustuavesicestuataremuitipies.rk.
we.anrepeattuepr..ecureb,.uttintw.squaresn.mtuesuacec
re.tan,let..reateastiilsmallerre.tan,le,simiiart.tuesuacec.ne,
itusicestuatais.mustbemultipies.rk. Asintueprevi.uspr..r,
i tuis pr..ess is ..ntinuec we s..n pr.cu.e a re.tan,ie witu sices
smaiiertuank. weuaverea.ueca..ntraci.ti.n. 1uepr..ecure.an
be.arriect.inrinity, but.ne.ann.tuave an inrinite sequen.e.rin
teerstuatkeepettinsmaileranc smailer. 1uerer.re a anc b are
in..mmensurable,anc . isirrati.nai. nriniteces.entpr..rs.anbe
,iven al,ebrai.r.rms, many .rwui.u,eneraiizet. pr.vin,tuat an,
nth r..ta.tannth p.wer is irrati.nai .
.
..
.
.
.
.
.
.
O
a+b
Figure 3. Steinhaus' infinite descent proof.
.ranappli.ati.n.rtuel x y re.tan,let.ama,i.tri.kinv.lvin,
tue repeatecr.lcin, .r a playin, .arc, see tue .uapter .n reptiles
inmy Unexpected Hanging and Oter Matematical Diversions. 1ue
re.tan,leis .allecan .rcerz reptile be.auseit.anbe.utint.t.
parts ea.u similar t. itselr Hritisu anc Lur.pean sueets .rpaper
usuallyuave sicesina l t. y rati.s.tuatuenualvec,quarterec,
ancs..n,tuesueetsremainsimilar.
Odds ,d Evehs
1uean.ientGreeksals.uacanele,antay.rusin,tuelas.r.cc
anc even numbers t.pr.ve y isirrati.nal t.an beexpressecin
numer.usays, buttuer.ll.in, seemstuesimplest
ieta stancr.r tueuyp.tenuse.rari,utis.s.elestrian,leancb
r.rits sice wekn.rr.mtuerytua,.reantue.remtuata
2
2b
2
, .r
a
2
1b
2
2. 1uerra.ti.nalb .bvi.uslyisbeteenl anc z. Assumeitis
recu.ect.l.esttermstuatis, itst.pancb.tt.mnumbersuaven.
..mm.ncivis.r.tuertuan l wekn.b is,reatertuat l, .tuerise
alb .ulcbeaninte,er
1ueri,utsice.ra2 2b2 iseven,tuerer.retuelertsicea2 isals.
even, anc a is even be.ausetue square r..t .rany even number is
even .ra e.ansubstitute2x uerex isanyinte,er squarin,2x
,i ves4x
2
, s.e.anrite4x
2
2b
2
. 1uisrecu.est.2X
2
b
2
1uelert
siceiseven,tuerer.reb2 isevenancb iseven. He.auseb.tua ancb are
even, ea.u.anbecivicecby z 1uis ..ntraci.tstueassumpti.ntuat
alb uasbeenrecu.ect.l.estterms. weuavepr.vectuatalb .ann.t
bearati.nalrra.ti.nbeteen l ancz,tuerer.re y isirrati.nal.
Lu.lic ,ave tuis pr..r in H..k lO, anc Arist.tle alluces t. it in
many pla.es A...rcin, t. rlat. inuis cial.,ue Theaetetus ,se.ti.n
l +), 1ue.c.rus.rcyrene, abrilliantpuil.s.pueranc,e.metri.ian,
als.pr.vectue irrati.nality.rtue square r..ts .rall n.nsquares.r
- tur.u,u l Alas, n.ne.ruis ritin,s survive, s. e c.n tkn.
u.uecicit,.ruyue st.ppecat l n.icentally, 1ue.c.rus as
banisuecrr.mcyrenebe.auseuec.ubtectueexisten.e.rtueGreek
,.cs witusuitablem.ciri.ati.ns, parity,.cceven)pr..rs.ry .an
be,eneralizect.allnth r..ts.rinte,erstuataren.tnth p.ers
La.u.rtuer.re,.in,pr..rsisareductio ad absurdum .rincire.t
pr..rin ui.u assumpti.n is mace tuen later pr.vec ralse by a
..ntraci.ti.n Auimsi.alincire.tpr..r.rtueirrati.nality.ry is
2. The Squ<re Root or 2 1 . 414 213 562 373 095 ... 15
basec.ntuerinalci,it.rsquarenumbers tiseasyt.seetuattuis
ci,itmustbeo, l , +, , s, .rv c.nsicera,aintueequati.na2 2b2 ,
uere alb isrecu.ect.l.estterms, b ,reater tuan l
1ueterminalci,it.rb.tua2 ancb2 mustbe.ne.rtue sixlistec
ab.ve Ontueri,ut sice.ra2 2b2 , b2 ismultipliecbyz, tuerer.re
tue rinalci,it.r2b
2
mustbe o, z, .r s. t.ann.tbez .r s be.ause
tuere isn. z .r s astuelastci,it.ra
2
. 1ue.nlymat.u isO. s.
a
2
anc 2b
2
mustea.u enc in zer.. t r.ll.stuata must enc in o,
ancb mustencino.r neituer.aseb.tua ancb arecivisibleby
, ..ntraci.tin,tueassumpti.ntuatalb isrecu.ect. l.estterms
nen.ealb isirrati.nalanc V isirrati.nal
similarterminalci,itpr..rs .rtueirrati.nality.rV .anber.r
mulatecin.tuerbasen.tati.ns nbinaryn.tati.n,r.rexample, tue
pr..r is unusually simple 1ue lert sice.ra2 2b2 terminatesinan
even number.rzer.sanctueri,utsiceterminatesinan.ccnumber
.rzer.s.
many ele,ant pr..rs .rtue irrati.nality .rV are basec.ntue
runcamental tue.rem .raritumeti., ui.u states tuat every inte,er
istuepr.cu.t.rauniqueset.rprimes nereis.ne.rtueeasiestt.
r.ll.
As ber.re, e use tue equati.n a
2
2b
2
uere alb is a rati.nal
rra.ti.nrecu.ect.l.estterms, b ,reatertuan l 1ueterma
2
must
uave an even number .rprime ra.t.rs. wuy: He.ause ira is tue
pr.cu.t.reitueran.cc.ranevennumber.rprimes, itssquareill
uaveti.easmanyprimera.t.rs
c.nsicern.tueri,utsice.ra
2
2b
2
tilluavean.ccnumber
.rprimera.t.rsbe.auset.tueevennumber.rprimera.t.rs.rb2 e
acctueprimera.t.rz weuavepr.cu.eca..ntraci.ti.nbe.ausetue
number.rprimera.t.rsr.rtuet. sices .rtue equati.n .ann.tbe
even.n.ne siceanc .cc.ntue.tuer. tisn.tciui.ultt. seetuat
tuepr..rappliest.tuesquarer..t.ranyprime,.rt.aninte,eritu
an.ccnumber.rprimera.t.rs.
rimecivis.rspr.viceasimplepr..rtuatanysquarer..tn.tan
inte,er is irrati.nal we apply itrirstt. V. rr.m a
2
2b
2
e .an
cerivetue equati.nb
2
a
2
/2 ui.uis tue same asa times a/2. ra
prime civicestuepr.cu.t.rt.inte,ers J anc y, it.bvi.uslymust
civiceeituerJ .r y. ieta2 anca betuet.inte,ersu.sepr.cu.t
isa2 /2. 1uere mustbeaprime tuatcivicesb2 be.auseb is ,reater
tuan l . 1uis sameprime mustcivice tueri,utsice.rtueequati.n,
tuerer.reitmustcivice a/2 .ra. n eituer .aseitcivicesa be.ause
iritcivicesualr.ra, it ill als.civicea. c.ntraci.ti.n weuave
su.ntuat a prime civicesb.tua anc b, tuerer.re a/b .ann.tbe a
rati.nalrra.ti.nrecu.ect.l.estterms
substitute r.r z anyinte,er u.se square r..tisn.taninte,er
anctuer.re,.in,pr..ru.lcs witururtuer,eneralizati.nstuepr..r
ill applyt.allnth r..ts.rinte,erstuataren.tnth p.ers.
An.tuersimplepr..r.rtueirrati.nality.rv isbasec.ninequal
ities. ra/b is v recu.ect.l.estterms, tuenb islesstuan a, anc
a islesstuan2b, tuerer.re (a - b) islesstuanb. startitua
2
2b
2
,
ancmaketuer.ll.in,.uan,es
a
2
- ab
a( a - b)
a/b
2b
2
- ab
b( 2b - a)
( 2b - a) /( a - b)
Aseuave seen, (a - b) issmallertuanb. weuave..ntraci.tec
tue assumpti.n tuat a/b is recu.ec t. l.est terms 1uis pr..rals.
,eneralizest.any
|
r..t.ranynumbern.tan
|
p.er
1uereare c.zens .r.tuer ays t. pr.vetue irrati.nality .rtue
squarer..ts.rinte,erstuataren.t squares, many .rui.uextenc
easilyt. nth r..ts 1uey all ..mec.n t. tuer.ll.in,tue.rem r
a/b isarati.nalrra.ti.nin l.estterms, b ,reatertuan l , tuen any
p.er.ra/b illals.bearati.nalrra.ti.ntuat.ann.tberecu.ect.
l.erterms
1uis.anbepr.vecbytuer.ll.in,ar,umentinv.lvin,primera.
t.rs. Assume tuat a/b, itu b ,reater tuan l , is recu.ec t. l.est
terms 1ueprimera.t.rs.ra illuaven.ra.t.rsin..mm.nitub,
.tuerisetue..mm.nra.t.rs.an.el.utanca/b isrecu.ec. c.nsicer
n.tuesquare.ra/b. 1uera.t.rsab.vetuelineillbetuesameas
ber.re,ea.urepeatecti.e,anctuesamer.rtueprimera.t.rsbel.
tueline 1uerearestilln...mm.nra.t.rst..an.el 1uismeanstuat
tue square.rarati.nalrra.ti.nrecu.ect. l.est terms isan.tuer
rra.ti.nrecu.ect.l.estterms, s.it.ann.tbeaninte,er nbrier,
n.inte,ern.tasquare.anuaveasquarer..ttuatisrati.nal
1uear,ument.bvi.uslyappliest..ubesancallui,uerr..ts .r
example, a3 /b3 is (a x a x a) / (b x b x b) . 1uis t.. is a n.nrecu.ible
rra.ti.nbe.ausetuerearen...mm.nprimera.t.rsab.veancbel.
tuelinet. be.an.elec stuere any simpler, easier t. ..mpreuenc,
ayt.su.tuatnth r..ts.rinte,ersn.tnth p.ersareirrati.nal:
wuen as in ui,u s.u..l anc rirst learnec tuat v ..ulc n.t
beexpressecasarati.nalrra.ti.n,..ulcn tbelieveit squancerec
2. The Squ<re Root or 2 1 . 414 213 562 373 095 ... 17
manyu.ursinstucyperi.cstryin,t.rincsu.uana.ti.n Lventually
..nvin.ec myselrit..ulcn tbe c.ne, butt.cay uave n. mem.ry
.ru. pr.vec it, irinceec cic like t. tuinkit as .ne .rtue
pr..rs,ivenintuisarti.le t.ulcbeinterestin,t.kn.u.many
matuemati.ians, rar ,reater tuan , uac a similar experien.e uen
tueyereveryy.un,
N.te tuat all tue pr..rs intuis arti.le are reductio ad absurdem
pr..rs 1uey illustrateu. p.errul tuis type .r pr..ris As G n
narcyputitinuisram.usMathematician's Apology
t is a rar riner ,ambit tuan any .uess ,ambit a .uess
player may .rrer tue sa.riri.e .r a pan .reven a pie.e,
butamatuemati.ian.uers the game.
Refrences
nuncrecs.rb..ks ..ntainpr..rs.rtueirrati.nality.rJ ancm.re
,eneralpr..rs.rtueirrati.nality.ranynth r..t.raninte,ern.tan
nth p.er. wuatr.ll.sarerereren.esineasilya..essibleperi.ci.als.
[ 1 ] Beckenbach, Edwin. "On the Positive Square Root of Two. " Math. Teacher 62,
April 1 969, 261-267.
[2] Bloom, David. "A One-Sentence Proof that J is Irrational . " Math. Mag. 68,
Oct. 1 995, 286.
[3] Bumcrot, Robert. "Irrationality Made Easy. " Colege Math. J. 17, May 1 986,
243-233.
[4] Estermann, T. "The Irrationality of J." Math. Gazette 59, June 1 975.
[5] Fine, Nathan. "Look Ma, No Primes. " Math. Mag. 49, Nov. 1 976, 249. See also
letters in April 1 977, 1 75.
[6] Goodstein, R. L. "The Irrationality of a Root of a Non-Square Integer." Math.
Gazette 53, Feb. 1 969, 50.
[7] Harris, V. C. "Terminal Digit Proof that J Is Irrational . " Math. Gazette 53,
Feb. 1 969, 65.
[ 8] Harris, V. C. "On Proofs of the Irrationality of J." Math Teacher 64, January
1 971 , 1 9-2 1 . See also his "J Sequel," 64, Dec. 1 97 1 , 760.
[9] Lange, L. J. "A Simple Irrationality Proof for nth Roots of Positive Integers. "
Math. Mag. 42, November 1 969, 242-243.
[ 1 0] Lindstrom, Peter. "Another Look at J." Math. Teacher 72, May 1 979, 346 -347.
[ 1 1 ] Maier, E. A. , and Ivan Niven. "A Method Of Establishing Certain Irrationalities. "
Math. Mag. 37, Sept. /Oct. 1 96 4, 208-210.
[ 1 2] Randall, T. J. "J Revisited. " Math. Gmette 6 7, December 1 983, 442.
[ 1 3] Rothbart, Andrea. "Back to J." Math. Teacher 6 5, November 1 972, 66 7-66 8.
[ 1 4] Shibata, Toshio. "On a Proof of the Irrationality of J." Math. Teacher 6 7, Feb.
1 974, 1 1 9.
[ 1 5] Strickland, Warren. "A More General Proof for J." Math. Teacher6 5, February
1 972, 1 09.
[ 16 ] Subbarao, M. V. "A Simple Irrationality Proof for Quadratic Surds. " Amer. Math.
Monthly 75, Aug. /Sept. 1 96 8, 772-773.
[ 1 7] Waterhouse, William. "Why Square Roots Are Irrational . " Amer. Math. Monthly
93, March 1 986 , 21 3-214.
[ 1 8] Zoll , Edward. "A Fourteenth Proof for J." Math. Teacher 6 5, Jan. 1 972, 30.
Postso- ipt
nerearet.su.rtincire.tpr..rs,n.tinmy arti.le,tuattuesquare
r..t.rzisirrati.nal.
l . r J is a rati.nal rra.ti.n tuere must be a smallest p.sitive
inte,er k tuat .ulc make kJ an inte,er. Hut kJ - k is a
smaller su.uinte,er c.ntraci.ti.n
z Asmace.learintuis.uapter,irtuesquarer..t.rzisaninte,ral
rra.ti.n beteen l anc z, tue equati.n a2 2b2 .ulc uave a
s.luti.nininte,ers nbase-n.tati.n,tuelastn.nzer.ci,it.r
asquaremustbe l 1uisappliest.tuelertsice.rtueequati.n
Hut tue last n.nzer. ci,it .r2b
2
is z, a ..ntraci.ti.n pr.vin,
tuata
2
2b
2
uasn.s.luti.nininte,ers
m.ntezer,er.allecmyattenti.nt.anamazin,ayt.,enerate
alltue..nver,ents.rtue squarer..t.rz. ner.unctuis explainec
inapaperbyD v Ancers.ninThe Mathematical Gazette ,N.vember
l vvs, pp +) startitutuesequen.e.rp.ers.rz,ea.up.er
2. The Squ<re Root or 2 1 . 414 213 562 373 095 ... 1 9
appearin,ti.e, tuenaccac] a.entci,itsintuemanner.ras.al stri
an,le 1uerirstt.entries.rea.ur.aretuesu..essive..nver,ents
tuatlistecearlierastuerun,s.rLuc.xus laccer
l l z z + + s s
z - + s s l z l s
l o l + zo zs
lz l z+ -+ +s
zv +l s sz
o vv l+o
l sv z-v
assurprisec anc u.n.rec uen tue r.re,.in, arti.le .n tue
l vvs 1rev.r Lvans Aarc .r :zo, ,iven ea.u year t. arti.les pub
lisuecin Math Horzons. 1rev.rLvans asa cistin,uisuec matue
mati.ianatLm.ryuniversity
Ch'pter 3
Fl i
p
, the Ps
y
ch i c Robot
You Outwit q Mi hdl ess Automqtoh?
nere sa.uan.et.take.nan.pp.nentu.uasn...ntr.l.ver tue
m.vesuemakes still,innin,maybe t.u,uer tuany.utuink
1ue ,ame is mat.uin, pennies Y.u rlip a ..in anc tuen rlip,
tuer.b.t, ill c.tue same rlip srlipmat.uesy.urs,ueacs arter
y.u uave tur.n ueacs, .r tails arter y.ur tails), rlip ins r n.t
,tailsarter y.urueacs, .r ueacs arter y.urtails), y.u in. lip uas
.uallen,ecy.ut.az,amemat.u
nere su.it.rks .rea.u,ame, l z), tur.a..inancn.te
uetuery.uture ueacs.rtails Hein,amere sueet.rpaper,lip
.an tt.ssr.ruimselr,s.y.un.,ett.telluimuatt.tur. 1uus,
iry.uture ueacs, y.u may ant t.tell rlip t.tur. tails. Hut
tuere sa.at.u s.metimeslipill.beyy.ur..mmancs, s.metimes
ue.n t Y.uuaven.ay.rkn.in,inacvan.euenuell.beyanc
uen uellcis.bey
On.ey.u ve.u.senueacs.rtailsr.rlip,ancn.tber.re) , r.ll.
alinetuatleacsrr.mtuatresp.nsealltueayt.itsenct.cis..ver
lip struerlip nr.ll.in,tuelines, y.umayn.t.uan,ecire.ti.ns
atinterse.ti.ns cue.ktueresult.rea.u,ameber.repr..eecin,t.
tuenext
keeparunnin,t.tal.rtuenumber.r,ames.nbyy.uanctue
number.nbylip
Y.umaybe surprisect.rinctuateverytimey.uplaytuis ,ame
lipillin nere suy
1ue,ameiscesi,nect...untertueaype.plemake.u.i.esuen
tryin,t.beatama.uine resear.uuassu.ntuatm.stpe.pleinsu.u
asituati.nrallint.apreci.tablepsy.u.l.,i.alpattern,tue,ameuses
tuispatternincesi,nin,tuer.b.t sresp.nses,uetuerue ll.bey.r
This article frst appeared in Gaes Magazine (October 1984) .
21
cis.bey)in a ay tuat .rks t. lip s acvanta,e n eue.t, lipis
psy.uin,.utuumanstryin,t.psy.u.utrlip
1ue,ameisbasec.na.tualexperimentsinartiri.ialintelli,en.e
n l vsv, s.vietresear.uerspr.,rammec a..mputert.play a,ame
equivalent t. tuis ..inmat.uin, test anc r.unc tuat nearly turee
quarters.rtuesl uumanstueytestecl.st. rauumanplayercicn.t
tryt.psy.u.uttuer.b.tbutsimply.u.serlip sresp.nsesatranc.m,
ue.rsue.ulcuaveaoo.uan.e.rinnin,.
cesi,necrlipanerreacin,ab.uttuerussian..mputerpr.,ram
inrusswalter s The Secret Guide to Computers ,HirkuauserH.st.n,
l vs-)
3. Flip, the Psychic Robot 23
Figure 1. Flip, the psychic robot.
Ch'pter 4
The Propos iti on' l c' l cu l us with
Di rected Gr'
p
hs
many r.rmall.,i.s.anberepresentecby,e.metri.alcia,ramstuat
areuserulint.ays 1uey arevisualaicstuatuelp stucents,rasp
tue stru.ture .rstatements in tue l.,i., anc in s.me .ases tue cia
,rams.anbemanipulatecinsu.uaaytuattue.rems.anbepr.vec
ancpr.blemss.lvecaserri.ientlyasitual,ebrai.te.uniques. ,.r
auist.ry.rl.,i.cia,ramsseeGarcner l } )
1ue rirst ,..c ay t. cia,ramtue pr.p.siti.nal .al.ulus as by
usin,venn.ir.les, pr.p.secinl ssobytueHritisul.,i.ian1.unvenn
ne cis.ussec uis system at mu.u ,reater len,tu in Symbolic Logic
z} , uereueappliecitalm.stex.lusivelyt.a.lassinterpretati.n.r
tueneH..leanal,ebra,espe.iallyt.tuecia,rammin,.rtraciti.nal
syll.,isms He.ausevenn.ir.lesapply]ustasa..uratelyt.tuebinary
relati.ns.rtue pr.p.siti.nalinterpretati.n.rH..lean al,ebra, it is
.uri.us tuat even t.cay intr.cu.t.ry textb..ks .rr.rmal l.,i. limit
venncia,ramsex.lusivelyt..lassin.lusi.nl.,i..
i,ure l su.s u. a pair .r venn .ir.les are suacec t. repre
sentea.u.rtue..nne.tives.rtuepr.p.siti.nal.al.ulus. ,1raciti.nal
symb.lsr.rtue..nne.tivesareals.su.n ) vennall.ectueentire
area.utsiceuis.ir.lest.representtue .lass..nsistin,.rtuene,a
ti.ns .rallterms, butis bestt. ..nrine tuis .lass t. a small .ir.le
asinci.atec 1uis makestuere,i.nmu.ueasiert. suace Anycia
,rammecstatement.anbe.uan,ect.itsne,ati.nbyex.uan,in,tue
uiteanccarkareas, tuene,ateccia,ramissimplytuepu.t.,rapui.
ne,ative 1uree.ir.lesilluancleallbinaryrelati.nstuat..n.ern
turee.rreerterms, ea.u.ir.lestancin,r.rapr.p.siti.ntuatmust
beeituertrue.rralse. 1uesystemisreacilyextencect.cia,ramr.r
This article, coauthored with the noted graph theorist Frank Harary, frst appeared
in Eueka (March 1988) .
25
A:B Implication (" If_ then_" )

0
B:A
0
A vB Disjunction, alternation

(" Either_ or_orboth")
A" B Exclusive disjunction, non-equivalence

@
(" Either or _ but not both")
AlB Non-conjunction
0
(" Not both _ and _" )
<
A=B Eq Uiva Ie nce
0
(" If and only if _ then _" )
A B Conjunction
@
C_and_" )
Figure 1 . Venn diagrams for binary connectives in the propositional calculus.
r.ur .r m.re termsbyusin, .l.sec.urves.tuer tuan.ir.les, anct.
matrixr.rmspr.p.secby1.unmarquanc, ieiscarr.ll, anc.tuers
A metu.c .rcia,rammin, tue pr.p.siti.nal .al.ulus itu unci
re.tec net.rks aspr.p.sec by Garcner l } in tue late l vos n
tueearly l vsosDr Garritmmes, aDut.ub.rnsur,e.nattuemec
i.al center, kru,ersc.rp, Ani.a, impr.vec tue net.rk metu.c by
accin,arr.st.itslines. nissystemasneverpublisuec,tu.u,uit
isbrierlymenti.necbyGarcner l } . n l vrranknararyincepen
centlytu.u,ut.rtuesamete.unique 1uepurp.se.rtuispaper ist.
explaintuemetu.c 1ueautu.rsbelieveitisauserulecu.ati.nalce
vi.e N.t.nlyc.esitcia,ramstatementsintuepr.p.siti.nal.al.ulus
inaaytuatpermitstueeui.ients.lvin,.relementarypr.blems,but
4. The Propositi on'l ('Ieu Ius wi th Di rectec Gr'phs
A 8

w
A -8
Figure 2. A is true, B fal se.
27
its.lvestueminaaytuatcuarlesreir.elikect..alli..ni.tuat
is, in a manner tuat str.n,ly resembles tue r.rmal stru.ture bein,
analyzec m.re.ver, itmakesuse.rcia,rams.l.selyrelatect. tue
net.rks.rl.,i..ir.uitsint.cay s..mputer.uips
1ue ci,rapu systemuasan.tuer ,reat acvanta,e.ver venn cia
,rams Artercia,rammin,tuepremises.rapr.blemintuepr.p.si
ti.nal.al.ulus, usin,venn.ir.les, itisimp.ssiblet.cistin,uisutue
suacin,.r.nepremisen.mtuesuacin,.ran.tuer. 1uismakesitex
tremelycirri.ultt.experimentitutuepr.blemstru.turebyalterin,
itspremisest. seeuattue .uan,eentails. n tueci,rapu system,
asesuallsee,ea.upremiseuasacia,ramis.latecrr.mtue.tuers
1uismakesr.r,reateaseinexpl.rin,tuet.talstru.ture.rtuepr.b
lem, seein, u.it.uan,esuenanypremiseisrem.vec.ralterec,
.rnepremisesareaccec
1ue runcamental cia,ramis simple 1ue t.p.ssible values.r
any term X are inci.atec by t. p.ints, .ne ab.ve tue .tuer Hy
..nventi.n, tue t.p p.intis X anc tue b.tt.mp.intisits ne,ati.n,
' X. 1ue trutu value .ra term, uen it is kn.n, is inci.atec by
crain,atiny.ir.lear.unctueappr.priatep.int,assu.nini,ure
z 1ueupperlert.ir.leinci.atestuattuepr.p.siti.nlabelecA istrue
1uel.erri,ut.ir.le su.stuatpr.p.siti.n B isralsetuatis, B s
ne,ati.nis auirmec
1uebinary..nne.tivesareinci.atecbycire.teclinesassu.nin
tuerirst..lumn.rri,ure- 1uese..nc..lumnsu.stuesamecia
,rams, butsimpliriecbyrepla.in,ea.uc.ubleline,arr.s,.in,b.tu
ays) itua sin,le uncire.tecline tisassumectuatanuncire.tec
line.anbetraversecinb.tucire.ti.ns. rtuetrutuvalue.ranyterm
iskn.n,e.antravelrr.mtue.ir.lectermal.n,anylineatta.uec
t.it,pr.vicectuelineisuncire.tec.ranarr.permitsit,t.an.tuer
termtuat.antuenbe.ir.lec.
1.examplesillmake tuis .lear c.nsicertue..n] un.ti.n "A
anc B t asserts tuat statements A anc B are b.tu true As tue
A B A B
Q
-
A.B

A. -B
..
AB

BA

. .
AvB
. .
AlB
<
-
A=B
-
-
.
A*B
Figure 3.
cia,ram r.rtuis ..nne.tiveinci.ates, ireituer A .r B is .ir.lec, e
.antravelal.n,tuelinet.tue.tuerp.intanc.ir.leit 1uecia,ram
r.rimpli.ati.n, rA, tuen B, su.s at.n.etuatirA is.ir.lece
mayr.ll.tuecire.teclinet. B anc.ir.leitals., bute.ann.t,.
a,ainsttuearr.rr.mBt.A. similarly,ir -Bis.ir.lecemay,.
t.A anc.ir.leit,bute.ann.t,.rr.m-A t.-B n.tuer.rcs,
kn.in, B istruetellsusn.tuin, ab.utA, anckn.in, A isralse
tellsusn.tuin,ab.utB 1uemeanin,s.rtue.tuer..nne.tivesare
immeciately.lear rr.m tueir ci,rapulines 1ue ,rapu is, .r..urse,
merelyaay.rcisplayin,tuestru.ture.rtue..nne.tive strututable.
ietusseeu.tuesystemappliest.ana.tualpr.blem.rtues.rt
.rtenr.uncinelementary l.,i.textb..ks.
4. The Propo5iti on'l ('l cul u5 wi th Di rectec Gr'ph5 29
Problem 1. 1uere areturee .men, Amy( A), Hertua ,B) , anc
car.l(e). wetake b.rnin1exas t.betue equivalent.rtrue, anc
n.tb.rn in 1exas t.be tue equivalent .rralse we are ,iven tue
r.ll.in,premises
l rAmyasb.rnin1exas, tuenHertuaasb.rnin1exas
z Lituer Hertua asb.rn in1exas, .rcar.l asb.rn in1exas,
butn.tb.tu.
- Lituer Amy asb.rnin1exas, .r car.l asb.rnin 1exas, .r
b.tu
+ Hertuaasb.rnin1exas
ntuen.tati.n.rtuepr.p.siti.nal.al.ulus, tuepremisesare
l A=B
z B= e
- . AVe
+. B
we isut. learn, irp.ssible, uetuer Amy anc car.l ere b.rn in
1exas t is p.ssible tuat tue premises may uarb.r a ..ntraci.ti.n
inui.u.asen...n.lusi.ns.anberea.uec 1ue..mbinecpremises
als.mayleave.pentuequesti.n.ruereAmyasb.rn,.rcar.l,.r
b.tu
i,ure+ ,t.p) su.su.tuepremisesareci,rapuec. wesimply
take tuem in .rcer Artertue turee ,rapus are cranr.rtue rirst
tureepremises, ancB is.ir.lec,.ntuebasis.rtuer.urtupremise),
all icenti.al terms are ].inec by uncire.tec lines t. make a sin,le
..nne.tec,rapu
westartexpl.rin,itatBbe.auseekn.Bistrue we.ann.t
travelestal.n,tuelinerr.mtuerirst.ir.lecB be.ausetue arr.
pr.uibits it, but e.an ,.rr.mtue se..nc .ir.lec B t. a ' e. we
.ir.le e assu.nini,ure+. rr.m e e.an travel t.an.tuer
' e anc .ir.le it. rr.m tuere, e .an ,. al.n, tue cire.tec line
t. A anc .ir.le it rr.m A e ,. t. tue lenm.st A anc .ir.le it
1uepatuleacsba.kt. B. 1ueci,rapun.l..ksassu.n. wesee
tuatall tue As, Bs anc Cs areauirmec we..n.lucetuatAmy
anc Hertua ereb.rn in 1exas, car.l as n.t 1uere is n. ay t.
30 A G<rd nd5 Workout
A B B
c A c
A : B B 7 C A X C
Figure 4. Solving Problem I with digraphs.
..ntinueexpl.rin,tue,rapu N...ntraci.ti.n,arrirmin,b.tuaterm
ancitsne,ati.n) asen..unterec,tuerer.retuepr.blemis s.lvec r
premise + uac assertec tuat Hertua as n.t b.rn in 1exas (" B),
a similar expl.rati.n .rtue ci,rapu .ulc uave auirmec A, " B
ancc.
t is easy t. see u.tuete.unique .an be extencec in a .uain
t. take .are .r any number .r binary premises, anc inv.lvin, any
number .rterms 1uere are te.uniques r.rusin, su.u ci,rapus t.
uancle..mp.uncstatements, statementsituparentueses)but,asis
tue.aseals. ituvenn .ir.les, tue systembe..mes t....mpli.atec
t.be.rmu.uinterest.
rtuepremisesc.n.tauirmtuevalue.ranysin,leterm,tueymay
stillpermitceterminin,tuevalue.r.ne.rm.reterms nsu.u.ases
.nemustexpl.retueci,rapubymakin,assumpti.nsab.utvaluest.
rinc.utiranyassumpti.ns.anbeeliminatecbya..ntraci.ti.n 1.
see u.tuis .rks ..nsicertuer.ll.in,pr.blem
Problem 2. A.manmay .rmayn.tbeAnn( A .r" A), may .r
mayn.tbebeautirul(B .r B), ancmay .rmayn.tbe.lever,c.r
c) wearet.lc
l . 1ue.maniseituerAnn.rsueis.leverbutn.tb.tu.
z sue isn.tb.tubeautirulanc.lever
- sueiseituerAnn.rbeautirul, .rb.tu
A c B c A B
A 7 C B C A B
Figure 5. Solving Problem 2 with dgraphs.
nn.tati.n
l A-c
z. B I c
- . AVB
wuat.anbececu.ecab.uttue.man
1ueci,rapuis su.ninri,ure 1.expl.reit, e startituany
assumpti.neplease iet sassumetue.man isAnn we.ir.letue
t.As 1raversin,tuelinesall.sust..ir.le t.cs sue is n.t
.lever N...ntraci.ti.nis en..unterec,buteuavelearnecn.tuin,
ab.ut uetuer Ann is beautirul. we try a,ain, erasin, tue .ir.les
anc.ir.lin, At. seeuatr.ll.sireassumetue.manisn.t
Ann. r.ll.in,tuelinesqui.klyleacst...ntraci.ti.nsr.r.in,ust.
auirmalltueterms anctueirne,ati.ns. weneec,.n.rurtuer 1ue
premisesare..nsistent.nly itu tue assumpti.ntuattue .man is
Ann,u.isn.t.lever wuetuersueisbeautirulisunce.icable
reassumetuattue.manisbeautirul, erinctuatsuemust
beAnn,ancn.t.lever Hute.anals.assume sueisn.tbeautirul,
ancuen.erinc tuat sueisAnnancn.t.lever reassume sueis
.lever,tueci,rapuleacsint...ntraci.ti.ns rsueisn.t.lever,tue
ci,rapu su.ssuemustbeAnn, ituuerbeautyunce.icec
t,.esitu.utsayin,tuat,asituall,..cl.,i.cia,rams,tueci
,rapupr.vicesasimpleayt.pr.vetaut.l.,y rt.statementsare
icenti.al, tueirci,rapusillbeicenti.al. r.rexample, De m.r,an s
ellkn.n pair .rlasasserts tuat (A. B) ,ui.u is tue same
as AlB) is equivalent t. Av " B; anc (A VB) is equivalent t.
-A-" B. 1ueequivalen.esare.bvi.us uen eci,rapuea.usice
.reituerlaancseetuattueci,rapusareicenti.al
A B A B
`
A B A . -B (or A I B)
Figure 6. Negating a term.
1..uan,etue value.ra sin,leterm X .r ' X in abinaryrela
ti.n, ima,inetuattuep.intsatp.siti.ns X ancX areex.uan,ec,
.arryin, itutuemanylinestuat are atta.uec 1uink .ra lineas
an elasti. strin,, .neenc.rui.um.vesituam.vin,p.intuile
its.tuerencremainsrixec i,ure s su.s, r.r example, A = B is
.uan,ect.A =' B.
1. .uan,ea ci,rapu.ran entirebinary relati.n t. its ne,ati.n,
rirst.uan,e ea.u uncire.tec line t. a c.uble line itu arr.s,.in,
b.tuays 1aketuerr.ntenc.rea.ulineancm.veitt.tueterm s
.tuerp.inttuat is, m.ve iteituer up.rc.n rinally, irtueresult
is a c.uble line, repla.eit itu an uncire.tec .ne i,ure su.s
an example we ant t. ne,ate A = B. wuen tue rr.nt encs .r
tue t. cire.teclinesare sit.uec, e ,et tuemiccleci,rapu 1ue
c.ublelineituarr.s,.in,b.tuays.ann. berepla.ecituan
uncire.tec line as su.n .n tueri,ut we see tuat tue ne,ati.n .r
A = B is A. B. n .rcs, irit is ralset. asserttuat A impliesB,
tuen A mustbetrueancB must beralse
A B A B A B
A B
A -B A -B
Figure 7. Negating a connective.
1uel.,i. ci,rapus .an be m.ciriec ancextencec in many ays
tuatsu,,estrurtuerstucy wuatistuebestayt.ci,rapu..mp.unc
statements: can ci,rapusuancle .lassin.lusi.n l.,i. itu existen
tial quantiriers su.uastue syll.,ismspremise s.me A is B: can
ci,rapusbeappliect..tuerl.,i.s su.u astuel.,i..rstri.timpli.a
ti.n anc.tuerm.cal l.,i.s: can tuey uanclemultivaluecl.,i.s by
accin,m.rep.ints t. ea.utermancperuaps usin, ciuerent ..l.rs
r.rtue lines t. inci.atetue p.ssible values.rtue..nne.tives: can
ci,rapusbeappliect.ruzzyl.,i.:
1ue autu.rs .ulc be please t. uear rr.m any.ne u. investi
,atestuesequesti.ns, .ru.intr.cu.esci,rapusinelementaryl.,i.
..ursesanc.arest.rep.rt.nu.tuestucentsresp.ncec
Refrences
[ 1 ] Gardner, M. , Logic Machines and Diagrams, revised edition, University of Chicago
Press, Chicago, 1 982.
[ 2] Venn, J. , Symbolic Logic, Macmillan, London, New York, 1 894.
[3] Harary, F. , Norman, R. , and Cantright, D. , Structural Models: an Introduction
to the Theory of Directed Graphs, Wiley, New York, 1 96 5.
Chqpter 5
Mqthemqti cs q nd Word
p
l q
y
many matuemati.ians en] .y.rcplay, ancr.r.bvi.usreas.ns tis
alm.stabran.u.r..mbinat.rialmatuemati.s. 1uepleasurecerivec
rr.ms.lvin,a..mbinat.rialpr.blemisverymu.uliketuepleasure
.rs.lvin,a.rypt.,ram.r a.r.ss.rcpuzzle, .r..nstru.tin,a,..c
palincr.me Given tue r.rmal system.raritumeti., an.ient matue
mati.ians askectuemselvesuetuertue ci,its l tur.u,uv ..ulc be
pla.ecinatureebytureematrix s.tuatr.s, ..lumns, anctuet.
cia,.nalsuactuesame sum 1uisisn.tmu.uciuerentrr.maskin,
ir, ,iven tue r.rmal rules .r Ln,lisu, .ne .an ..nstru.t a tureeby
turee.rcsquareinui.uea.ur., ..lumn,ancmaincia,.nalisa
ciuerent.rc.
1uereis, .r..urse, acirreren.ebeteen..mbinat.rialmatuemat
i.s anc.rcplay. matuemati.s is emb.ciecintue stru.ture .rtue
universe Altu.u,umatuemati.al systems are nee inventi.ns.ruu
man mincs, tuey uave ast.nisuin, appli.ati.ns t. nature N. .ne
expe.tecn.nLu.lician,e.metryt.beuserul, butitpr.vect.bees
sential t.relativity tue.ry H..lean al,ebra seemec uselessuntil
surpriseitturnec.utt.m.celtueele.tri.alnet.rks.r..mputers
1uereareuuncrecs.r.tuer.utstancin,instan.es.ruatpuysi.ist
Lu,enewi,ner uas.allectueunreas.nable eue.tiveness.rmatue
mati.s.
1uink .rtue letters anc .rcs .ra lan,ua,e, t.,etuer itu its
rules, asar.rmal system Altu.u,utue .rcsuavearbitrarymean
in,sassi,nect.tuembymincs, anctueremaybeaceepstru.ture
.rsyntaxtuat..nr.rmst.l.,i.,tue.rcstuemselvesuaven.reality
apartn.ma.ulture Hutterrliesare all.vertue.rlc, buty.u ill
n.trinctue.rcbutterrlybyl..kin,tur.u,uateles..pe.rmi.r.
s..pe. On.e tue .rc be..mes atta.uect.butterrlies, u.ever, itis
amusin,t.n.tetuatbutterrliesrlutterby He.auselan,ua,e, unlike
This article frst appeared in Word Ways (February 1993) .
35
matuemati.s, isartiri.ial, .rcplayuasm.rein..mm.nitu, say,
inventin,.arctri.ks .rplayin,.uess
1ue..mbinat.rialnature.r.rcplayisuncers..recbytuere.ent
use.r..mputersr.r s.lvin,.rcpr.blems Disks..ntainin,alltue
.rcs.ralan,ua,earen.available. witusuitablepr.,ramstuey
.anbeusect...nstru.t.rcsquares, rincana,rams, su.rtest.rc
laccers, ancs..n .ulcn tbesurprisecirs.mecay..mputerss.lve
..mpli.atec.r.ss.rcsaseasilyastueyn.s.lve.uesspr.blems
tis .rtun.tin,tuat,inb.tumatuemati.sanc.rlcplay, s.lv
in,apr.blemis.uri.uslylike..nrirmin,atue.ry. ns.lvin,a.ryp
t.,ram, r.rexample, .ne rirstmakes ..n] e.tures. s a sin,leletter
.rcA.r, .rmaybeO:sAHcAtue.rctuat: su.u..n]e.tures
aretuentestect. seeirtueyleac t. ..ntraci.ti.ns rtueyleact.
.tuer .rcs, tuey,ain intueirpr.bability .rbein,..rre.t Lventu
allyap.intisrea.uecatui.u.neis.ertaintuata.rypt.,ramuas
been.ra.keceventu.u,un.tallitslettersarekn.n
Oneistemptect.saytuatuenall.rcsarekn.n, .ne.anbe
abs.lutely.ertaina.rypt.,ramuasbeens.lvec. 1uisisn.ttue.ase
be.ause, espe.iallyirtue.rypt.,ramissu.rt,tuere..ulcbean.tuer
s.luti.n tuat tue ..mp.ser .rtue puzzle intencea r, u.ever, tue
.rypt.,ram is l.n,, su.u an un.ertainty be..mes vanisuin,ly small
1uis istrue als. ins.ien.e wuen tuere is alar,eabuncan.e.rra.ts
explainecbyatue.ry, su.uasbytuec.perni.antue.ry.rtuetue.ry
.rev.luti.n, .ertaintyrea.uesapr.bability.ro +++++++++ .
N. r.r a ceep metapuysi.alquesti.n r.uess uac n.tbeenin
ventec, istuerea sense in ui.u tue.rems ab.ut.uess.anbe saic
t. exist: Assumin, tuer.rmal system.r.uess, anc ,iven a .ertain
p.siti.n .n tue b.arc, is it permissible t. say tuat tuere is a mate
intureem.veseven irn..ne uasp.sectuepr.blem: Assumin,tue
stru.ture.race.k.r.arcs, istuereasenseinui.ua,..c.arctri.k
is s.meu. .uttuere, in a rlat.ni. realm.runiversals, even irn.
.arcsexistec:
supp.setuereeren.Ln,lisulan,ua,e w.ulcitbemeanin,rul
t.saytuat,ivensu.ualan,ua,e, tuereisa senseinui.ua.ertain
ana,ramexistsevenirn..ne sp.keLn,lisu: t is s.metuin, like
askin,ira.ertainnumberituamilli.nci,itsisprime.r..mp.site
ber.reany.netestectuenumbert.rinc.ut. well,n.tquite,be.ause
aritumeti..ertainlyexistsasar.rmalsystem Anyay, m.stmatu
emati.ians are rlat.nists u. believe tuat, n. matter u. bizarre,
.ru.rarrem.vecrr.mreality a system.anbe, tuey cis..verits
5. M'them'tics 'nd Wordpby 37
tue.remsratuertuaninventtuem Lventu.u,uLn,lisuis auuman
..nstru.ti.n, n.ueret.ber.uncinnature, istuereasenseinui.u
its.rcplaysarerealber.reany.nerincstuem: leaveanserin,
tuisquesti.nt.myreacers
PostsO' ipt
menti.nec tuetask .r ..nstru.tin, a tureebyturee .rc square
.nui.ueveryr.,..lumn,ancmaincia,.nalisa..mm.nLn,lisu
.rc nl vvvD.nalcknutu,tuen.tec..mputers.ientistn.retirec
rr.mstanr.rctniversity, macea..mputersear.ur.rsu.usquares
1uepr.,ramusecalist.rso+tureeletter.rcstuatarein..mm.n
use, ex.lucin,pr.per names, a.r.nyms, ancabbreviati.ns tr.unc
l+ss.luti.ns nerearet.examples
wnO
ON
1LL
rLA
trN
HAY
rcupli.ate lettersareex.lucec, tuen su.u a squarepr.vicesan
amusin,,ameis.m.rpui.ituti.ta.t.e 1uesquare snineletters
are.nnine.arcs. rlayerstaketurnscrain,a.arcanctuerirstt.
r.rm .ne .rei,ut .rcs is tue inner He.ause ti.ta.t.e is als. a
tie n.ever,irtue .rc squareis n.tsu.n, apers.nkn.in,tue
square,anckn.in,u.t.play aperre.t,ame.rti.ta.t.e,.anin
easilya,ainst.pp.nentsunaare.rtueis.m.rpuism
wuenknutu.ue.kecuis l+s .rcsquaresr.r s.luti.nsitun.
cupli.ateletters, uer.uncn.ne
nasu.rtarti.lein Word Ways ,Au,ustl vvv)askectuequesti.n
.ruetuersu.uasquareexists 1uearti.leisreprintecuere
Ti c-Tqc-Toe Pl qyed qS q Wotd Gq me
npr.blem+sin Your Move ,m.Granill, l v l ) , Davic silverman
ces.ribecalin,uisti.versi.n.rti.ta.t.e..nsistin,.ra st..kpile.r
tue.rcsArmY, cnA1,sn, Gri,nOrN,kN1, sOtr, swAN,
ancvO1L. rlayers alternately sele.t.rcs, anctuerirstt. ..lle.t
turee.rcssuarin,a..mm.nletteristueinner.
n tue sin,leletter anal.,ue .r silverman s ,ame, players cra
alternatelyn.mast..kpile.rninecirrerentletters, tuerirstt.sele.t
tuelettersr.rmin,.ne.raspe.irieclist.rnine.rcsistueinner
1. make it easy t. remember tuese .rcs, tuey .an be ritten in
tuer.rm.ratureebyturee.rcsquareinui.ub.tucia,.nalsare
als..rcs
squares are easy t. rinc ir y.u all. abbreviati.ns, a.r.nyms,
pr.per names .r r.rei,n .rcs n.ever, m ualr..nvin.ec tuat
tuereisn.s.luti.nitu..mm.n everycay .rcs. t salm.stsp..ky
u.y.u.anrincseven .rcsbutn.ttueei,utu. mynearmisses
NOs ArL nOr
LA1 sN LAr
wrY rOD skY
HLD
OAr
wnY
iu
Ar
1LN
DOs
LA1
wrY
ntuerirst square, tuerirst.rc.anbeusecin a senten.esu.u
as 1ueNOsuaveit summarizin, a v.te cuan,etueN t. i, anc
tuewt.D, anctuerirst.rc, iOs, istuerirstualr.ri.sAn,eles.
r,ivennamesareall.ec, ..nsicertuese..ncsquareiturO.1ue
tuircsquareusestue..ntra.ti.nnL s, tuesixtuusesDOs, asintue
purase c.s ancc.n ts. 1uere must be c.zens .r even uuncrecs .r
squares usin, .rcs rr.m, say, tue Orri.ial s.rabble rlayers Di.ti.
nary, buttueseinvariablyempl.y lessramiliar .rcstuan tue .nes
intuesquaresab.ve
r.ssL.kler,eneratecaninterestin,set.rsquarestuatallusetue
ren.u.rcLAt, r.uncinLn,lisupuraseslike eau ce..l.,ne.r
eaucevie
OG
LAt
DrY
nOG
LAt
rrY
iOG
LAt
1rY
iOH
LAu
1rY
rOH
LAu
DrY
GOH
LAt
1rY
I hhod udi oh
Chqpter 6
Stei ner Trees on q
Checkerboq rd
supp.sea rinite set.r p.intsareranc.mly s.atterecab.utin tue
plane n..antueybe].inecbyanet.rk.rstrai,utlinesitutue
su.rtestp.ssiblet.tallen,tu: 1ue s.luti.n t.tuispr.blemuas pra.
ti.al appli.ati.nsintue..nstru.ti.n.ravariety.rnet.rksystems,
su.u asr.acs, p.erlines, pipelines, ancele.tri.al.ir.uits
t is easy t. see tuat tue su.rtest net.rk must be a tree, tuat
is, a..nne.tecnet.rk..ntainin,n..y.le ,A cycle is a.l.secpatu
tuatall.s.net.travelal.n,a..nne.tecpaturr.ma,ivenp.intt.
itselritu.utretra.in,anyline. ) rn.nep.ints.anbeaccect.tue
.ri,inalset.rp.ints, tuesu.rtestnet.rk..nne.tin,tuemis .allec
aminimum spanning tee.
Aminimumspannin,tree is n.t ne.essarilytue su.rtestnet.rk
spannin, tue.ri,inal set.rp.ints nm.st.asesasu.rternet.rk
.anber.uncir.ne is all.ect.accm.rep.ints .rexample, sup
p.sey.uantt.].intureep.intstuatr.rmtueverti.es.ranequilat
eraltrian,le 1.sices.rtuetrian,lemakeupaminimumspannin,
tree 1uisspannin,tree.anbesu.rtenecbym.retuanl - per.entby
accin,anextrap.intattue.enteranctuenmakin,..nne.ti.ns.nly
beteentue.enterp.intancea.u..rner, seeri,ure l ) La.uan,le
attue.enteris l zo
Aless.bvi.usexampleistueminimumnet.rkspannin,tuer.ur
verti.es.ra square Onemi,ut supp.se.ne extrap.intin tue .en
ter .ulc ,ivetue minimumnet.rk, butitc.esn.t 1ue su.rtest
net.rkrequires, inra.t, two extrap.ints , seei,urez). A,ainall
This article, coauthored with Fan Chung and Ron Graham, frst appeared in Math
ematics Magazine (April 1989) .
39
length 2
length 1
Figure 1 .
an,lesar.unctueextrap.intsintuenet.rkare l zo . 1uenet.rk
itu.neextrap.intintue.enteruaslen,tu2y, .rab.utz szs 1ue
net.rkitut.extrap.intsrecu.estuet.tallen,tut. 1 + /, .r
ab.utz -z
One.rtuerirstmatuemati.ianst.investi,atesu.unet.rksas
1a..b steiner, an eminent siss ,e.meter u.ciecin l ss-. 1ue
extrane.us p.ints tuatminimize tue len,tu .rtue net.rk are n.
.allec steiner p.ints tuas been pr.vec tuatall steiner p.ints are
]un.ti.ns.rturee lines r.rmin,turee l zo an,les. 1ue su.rtestnet
.rk,all.in,steinerp.ints, is.allecaminimum Steiner tree. mini
mumsteinertreesarealm.stalayssu.rtertuanminimumspannin,
trees, buttuerecu.ti.ninlen,tuusuallycepencs.ntuesuape.rtue
.ri,inal spannin,tree tuasbeen..n] e.turec v} tuatr.r any,iven
set .rp.ints in tue plane, tue len,tu .rtue minimum steiner tree
.ann.tbelesstuanara.t.r.r//2, .rab.uto sss, timestuelen,tu
.rtueminimumspannin,tree, tueresultuas been pr.vec, u.ever,
.nlyr.rturee,r.ur, ancrivep.ints l o} , l z}
.
.
length 3
length 2/ length 1 1
(a) (b) ( c)
Figure 2.
6. Stei ner Trees on < Checkerbo<rd 41
many pr.perties .r minimum steinertrees .an be r.unc in tue
ex.ellent ,but s.meuat .ut.rcate) survey paper .r L N Gilbert
anc n O .llak v} 1ue best .urrent l.er b.uncr.rtue rati. .r
tueminimumsteinertreet.tueminimumspannin,treeiso sz+ l
, see+} )
1uerearemanyayst...nstru.taminimumspannin,tree. One
.rtuesimplestmetu.csiskn.nasa,reecyal,.ritum, be.auseat
ea.ustepitbites .utuem.stcesirable pie.e irstrinct.p.ints
tuatareas.l.set.,etuerasany.tuert.anc].intuem rm.retuan
.nepair.rp.intsareequally.l.se,.u..seanysu.upair repeattuis
pr..ecureitutueremainin,p.intsinsu.uaaytuat].inin,apair
never..mpletesa.ir.uit 1uerinalresultisa spannin, tree.rmini
mumlen,tu 1uisal,.ritumiscuet.kruskalina l vspaper l l } .
Given tue simpli.ity .r kruskal s ,reecy al,.ritum r.r tue ..n
stru.ti.n.rminimumspannin,trees, .nemi,utsupp.setuere.ulc
be ..rresp.ncin,ly simple al,.ritums r.r rincin, minimum steiner
trees tnr.rtunately, u.ever, tuis isalm.st.ertainlyn.ttue.ase
1uistaskbel.n,st.aspe.ial.lass.ruarcpr.blemskn.nin..m
puter s.ien.e asN..mpletepr.blems wuentuenumber.rp.ints
in anet.rkissmall, say lo t. zo, tuerearekn.nal,.ritums } ,
l -} r.r rincin, minimum steiner treesi na reas.nably su.rt time
As tue number .rp.ints,r.s, u.ever,tue..mputin,timeneecec
in.reasesat arapiclya..eleratin,pa.e Lvenr.r arelatively small
number.rp.ints tue best al,.ritums .urrently available ..ulc take
tu.usancs .r even milli.ns .ryeast. terminate. m.st matuemati
.ians believen. eui.iental,.ritums exist r.r ..nstru.tin, minimum
steinertrees.narbitrary sets.rp.intsintueplane } , s}
ma,ine,u.ever,tuattuep.intsarearran,ecinare,ularlatti.e
.runitsquares, liketuep.intsattue..rners.rtue.ells.ra.ue.ker
b.arc stuerea,..cal,.ritumr.rrincin,aminimumsteinertree
spannin, tuep.ints.rsu.ure,ularpatterns: n parti.ular, uatis
tue len,tu .rtue minimum steiner tree tuat].ins tue s l p.ints at
tue ..rners .ra stancarc.ue.kerb.arc: stuetree in i,ure - tue
s.luti.n:
manypr.blemsinv.lvin,patustur.u,up.intsintueplane, ui.u
areciui.ultuen tuep.ints are arbitrary, be..metrivialuen tue
p.intsn.mre,ularlatti.es. Onemi,utexpe.ttuattuetask.rspan
nin,p.intsinsu.uarraysbyminimumsteinertrees.ulcbeequally
trivial Ontue..ntrary,tuispr.blemseemst.besurprisin,lyelusive
tpt.n., .nlyminimumsteinertreesr.rzby re.tan,ulararrays
length 73. 033
Figure 3.
.rp.intsuavebeen..nstru.tec -} Asicerr.mtuisspe.ial.ase, very
littleseemst.bekn.nab.utu.t.rincminimumsteinertreesr.r
re.tan,ulararraysuentuenumber.rp.ints.nea.usiceis ,reater
tuanz
ntuis paper, eill summarize vari.uspr.blems, ..n]e.tures,
ancs.mepartialresults.ntueminimumsteinertreesr.rre.tan,u
lar arrays se.ti.n z ..ntains tue su.rtest kn.n trees r.r square
latti.es.rsmallsize. 1uesetrees..nsist.r..pies.rtuesymmetri.al
tree.nr.urp.ints,seeri,urez,.)), ui.ue.all\rr.mn..n,t.
,etuerituasmallnumber.rex.epti.nalpie.es .rexample,tue
..n]e.turecs.luti.nr.r s+p.intsisauni.n.rzl \ s, seei,ure).
nse.ti.n-, e,iveapr..rtuatare.tan,ulararray.anbespannec
byasteinerneemaceupentirely.r\ sirana.nlyirtuearrayisa
squareanctue.rcer.rtuesquareis ap.er.rz urtuerquesti.ns
arepr.p.secinse.ti.n+
Tree Sybol Lngth

E e = 1

T
1 + 1
t 1 1 . 93185 . . .
`
X J 1 + 1 2. 73205 . . .

L I ]5 + 201 8. 34512 . . .
Figure 4.
shott Stei net Ttees on Squq te Lqtti ces
n tuis se.ti.n, e illrirst su. tue su.rtest steinertreese.ur
rentlykn.r.rsquarelatti.es.rsizeup t. l+ by l +. we illtuen
cis.ussa s.uemer.r..nstru.tin, steiner treesr.r lar,e square lat
ti.esr.rmtue small.nes Am.n,alltue..nstru.ti.ns, .nlytuepat
ternsr.r tue z x z, - x - anc+ x + squaresuavebeenpr.vect.be
minimumsteinertrees,unpublisuecresults.rL 1 c..kayne) 1ue
..nstru.ti.ns r.r square latti.es .r.rcers z t. v ere ..ntainec in
tue1une l vssissue.rScientifc American s} 1ue treesr.rsquare
latti.es.rsizes l oby l oanczzbyzzintue samearti.leere s..n
impr.vec by many reacers 1ue .urrent best tree r.r tue l oby l o
squarelatti.ei s cuet..ne.rtueautu.rs,riG)anctuebesttreer.r
tue zzby zz squarelatti.eiscuet.Lri.carls.n l } . nis..nstru.ti.n
uastue same t.tal len,tu as.ur ,eneral ..nstru.ti.n Overall, tue
..nstru.ti.ns rall naturally int. six .lasses, cepencin, .n uat n is
m.cul.s, itutuerare,ancremarkable)ex.epti.nstuat...uruen
isap.er.rz tseemstuattue steinertreesr.r squarelatti.es
arealaysr.rmec by atta.uin, smallminimum steinertrees asan
ec,eE ,r.rtue l byz array) , \ ,r.r tue z byz array), atrian,leT
,r.rmecrr.mtureeverti.es.rasquare) ancL ,r.rtuezby array) .
weillusetuen.tati.n,iveni nri,ure+.
1uetreeL isaminimumsteinertree.ntuere,ularzby array
-} . 1ue minimum steiner tree r.r tue z by array itu even is
]ustmaceup.r\ s].inect.,etuerbyec,es Ontue.tueruanctue
minimum steiner tree r.r tue z by n array itu n .cc uas len,tu
( (n( 2 + V) 2) 2 + 1 ) 1 /2 . 1ue ..nstru.ti.ns r.r n x n, n 14, are
illustratecinri,ure
N Conjectured Minimum Steiner Tree
Length
2
x " 2.73205 . .
3
2x J 2 " 7.46410 . . .
"known lO be optima
Figure 5.
6. Stei ner Trees on ' Checkerbo'rd
45
N Conjectured Minimum Stemer Tree Length
4 5x 1.ZO. . .
5 7x O
22.12436 . . .
6 llx I
O.UO+
7 15x *
43.98076 .
Figure 5 continued.
46 A Gard ner' s Workout
n
Conjectured Mi ni mum Steiner Tre Lngth
8
2lx " 57.37307 . . .
9 26x 2 = 73.03332 . .
10 3x + I 9.36. . .
Figue 5 continued.
6, Stei ner Trees on ' Checkerbo'rd
47
0 Conjectured Minimum Steiner Tree tugID
11 39x 3
109,54998
12 47x I 130,33824 " ,
13 5x 3 " 15,26279
Figure 5 continued,
n Conjectured Minimum Steiner Tree
Length
14
62x 1 = 177.73227
Figure 5 continued.
1...nstru.tsteinertreesr.rlar,esquarelatti.esitu.rcersn.t
equalt.ap.er.rz, eillalaysusea..resquareituar.lcec
banc.rictu - rappec ar.uncitin vari.usays ,1ue.nly..re
squareseneec ares x s, l ox l o, l+ x l+ r.r even, anco x o, +
x + anc s x sr.r .cc ) 1ue,eneralpatternis su.nini,ures
m
|
|
| |
...
. . .
I
I
I
I
I
I
N OVCD
H odd
Figure 6.
'
Figure 7.
La.u acciti.nalr.lc.rtuestrip accs st.tuesize.rtue,ric
.rexample,r.rn = 22, eseetuat22 :o,m.cs)s.eusea l o
x l o..reitu- ,c.ublec)r.lcs.rtuebancassu.ni nri,ure
Or ..urse, tuestripmustbebr.kenanc..nne.tect.tue..re. ,see
tuecetailecpi.tureini,ure 8. wen.te tuat in .rcert.make tue
..nne.ti.n t. tue strip, tue ..rresp.ncin, pla.e in tue..re must be
striplike ) wuen n is 6k, tue..re is 0 x o, i e , empty, s. ec.n t
uavet...nne.titt.tuebanc ntuis.ase, e.nlyuavet.breaktue
bancancre..nne.ttuet.is.latecp.intsituaT.
.r odd n, tuebanc c.esn tr.rma.y.lebut is .pen at ea.u enc,
leavin,t.is.latecp.intsattueenc. Y.u.anseetuisuappenin,in
tue..n]e.turecminimumsteinertreesr.rv x vanc l - x l - . ntuis
.ase,uenn = +,m.cs)e.nlyneecanEt...nne.ttue..ret.tue
banc summarizin,tueseresults, euaver.rn 15, n = 2t :
n
6k
6k + 1
6k + 2
6k + 3
6k + 4
6k + 5
ien,tu.r..n]e.turecminimumsteinertreer.r G
n
( 12k2 - J + t)
( 12k2 + 4k - l) x + 3
( 12k2 + 8k - 2) x + I
( 12k2 + 12k + 2) x + 2
( 12k2 + 16k + 2) x + I
( 12k2 + 20k + 7) x + 3
uerex = 1 + V, 1 = )35 + 20V, anct = ( 1 + V) /v. Or..urse, r.r
n = 2t , anyri,uttuinkin,pers.n .ulc,uesstuattuelen,tu.rtue
minimumsteinertreer.rG
2
t is]ust , ( 4t - l) x butunr.rtunatelye
.an tevenpr.vetuisr.rt
=
3 !
Figure 8.
SqUq te Lqtti ces (at Powets o( 2
nereeill,ivetuepr..r.rtuemainresultintuispaper.
Theorem
If a rectangular array can be spanned by a Steiner tree made up entirely
of \'s, then the array is a square of size 2t by 21 for some t ;: 1 .
Prof we startby ,ivin,ea.uz x z.ellapair.r...rcinates ( i, j)
i ntue .bvi.us ay, uere tue l.er lertuanc .ell uas ...rcinates
(0
,
0) .
iet s .all a .ell (i, j) even iri + j is even Als., let s .all a .ell
occupied irituasan \init Suppose .ura x b arrayuasan \tree,
tuatis, asteinertreer.rmecby\ s
Fact 1. Onlyeven .ells.anbe...upiec
|

1

+-array
0
o 2
Figure 9.
Proof 1ue ...upiec .ells mustbe ..nne.tec rurtuerm.re, tuere
.an t be t. adjacent ...upiec .ells 1uus, ...upiec .ells .an .nly
t.u.uea.u.tuercia,.nally ,as su.n) , inui.u .ase tuey areb.tu
even.rb.tu.cc n.ever, tue..rner.ell ( 0
,
0) is...upiecanceven
1uus, all ...upiec.ellsareeven.
Asa..nsequen.e, eseetuata ancb mustb.tubeeven. callan
even.ell ( i , j) doubly even iri ancj areb.tueven Otuerise, .allit
doubly odd.
Fact 2. Lvery c.ublyeven.ellmustbe ...upiec
(forbidden)
(allowed)
Figure 1 0.
52
(2u, 2 v )

Figure 11 .
A G<rd nd5 Workout
-- -(2i , 2j )-unocupie
Proof irst n.te tuat al tue even .ells .n tue b.uncary mustbe
...upiecancarec.ublyeven supp.seeuac s.me interi.rc.ubly
even .ell ( 2i
,
2j) tuat as n.t ...upiec 1uen, e must be able t.
cra a patu P rr.m tue .enter.r ( 2i
,
2j) tuat ,.es t. tue .utsice.r
tuearrayancc.esn tpasstur.u,uany...upiec.ell. ,1uisisbe.ause
tue..mplement.r.ur\treemustbea..nne.tecset. )Hutiranyeven
.ellisunoccupied tuenal four .ritsevennei,ub.rsmustbeoccupied
, sin.e,.tuerise, .ne.rits..rnerp.ints.ulcbecis..nne.tec )1uis
n. impliestuat tue .nly even .ells P .anpasstur.u,u are c.ubly
even.nes, ancneverac.ubly.cc.ne
N., r..us .nac.ubly even .ell ( 2u, 2v) ]ustinsicetueb.uncary
tuatP triest. pass tur.u,u.nitsayt.tue.utsice sin.e ( 2u, 2v)
mustbeun...upiec ,be.auseP is,.in,tur.u,uit) , all.ritsnei,u
b.rs mustbe...upiec n parti.ular, tuis r.rms a barer itu tue
ac] a.ent...upiecb.uncary.ellstuatpreventsP n.m,.in,tur.u,u
t. tue .utsice uere Hut tuisuappensuerever P triest.rea.utue
.utsice sin.e all tue even .ells.ntueb.uncary are ...upiec 1uus,
P .an neverrea.utue .utsice, ui.uisa ..ntraci.ti.n. nen.e, tue
uyp.tuesistuattuereis anun...upiecc.ubly even.ellisuntenable,
anctueasserti.nispr.vec.
1uerer.re, in .rcert.kn.uat.ur\tree is, e.nlyuavet.
kn. ui.u ,acciti.nal) doubly odd .ells are ...upiec i..k at tue
pi.turer.ran 8 x 8 arrayinri,ure l z
Next t. tue 8 x 8 array, a + x + array is cran N.ti.e tuat
tuere is natural ..rresp.ncen.e beteen tue array .r v c.ubly .cc
8 4 x 4 array
(shaded) doubly odd cels in 8 X 8 array
Figure 1 2.
.ells, suacec)intue 8 x 8 anctue array.rv.ellsintue+ x +. 1ue
key .bservati.n n., ui.u is n.t uarc t. .ue.k, is tuat the set of
occupied doubly odd cells in the 8 x 8 must corespond exactly to an
X-tree in tue+ x +array Namely, ..nse.utivec.ubly.cc.ellsin tue
samer..r ..lumn .ann.tb.tube...upiec,.re ,eta.y.le), anc
allp.intsintuearraymustbe].inect.,etuer
nerre.t,tuec.ubly.cc.ells.ntue8 x 8 arrayr.rmastret.uec
.utversi.n.ralltue.ells.ra+ x +array
m.re,enerally, tuisar,umentsu.stuatirana x b array uas an
\treetuena 2A, b z, anc, rurtuerm.re,tuesmallerA x array
must als.uave an\tree
we n. .an apply tuisrepeatecly , similart.ermat smetu.c .r
inriniteces.ent, ex.epttuat e st.p at z x z) t. ,ettue ..n.lusi.n
tuat.nly2t x 21 arrays.anuave\trees 1uis..mpletestuepr..r.r
tuetue.rem
Concl ud i ng Remq tks
Or..urse, tuemain.penpr.blemist.ceterminetueminimumsteiner
treesr.rall,.reveninrinitelymany) squarelatti.es tisembarrass
in,tuatevenr.r2t by2t arrays, estill.an tpr.ve.ptimalityr.rtue
.bvi.usly..rre.t\tree
reassumetuecistan.ebeteenac] a.entverti.esintuelatti.e
isz,tuenitisn.tuarct.su.tuatanysteinertree,minimum.rn.t)
uaslen,tu.rtuer.rm, a + bV uere a ancb areinte,ers,p.ssibly
ne,ative) c.n.eivably,tuisra.t..ulcbe.ruelpinpr.vin,.ptimality
in s.me.ases
Aninterestin,relatecquesti.nist.use.nlytue 2 x 2 minimum
steinertreesanctuesmallestp.ssiblenumber.rsin,leec,es,.allec
E s) t.r.rm a spannin,tree.rG( m, n) . wekn.tuatr.r m anc n
b.tu lar,e en.u,u, .nly a b.uncec number .rE sare ever neecec
wuen m is small u.ever ,uere e .an assume m n) , e may
neec arbitrarily many E s. s.me examples.rtuis are illustratecin
i,ure l -
Acknowl edgements
1ue autu.rs .ulcliket.a.kn.lec,etuemeti.ul.usuelp.rNan.y
Davics.ninpreparin,tueri,uresr.rtuispaper
m = 1
m = 3

E E E E

Figure 13.
al E' s
0 .
2- E's needed
Only
2
E' s needed
6. Stei ner Trees on a Chec kerboard
m = 4
m = 5
m = 6
m = 7
Figure 1 3 continued.
n
1
|
E' s needed
A bo
n
ded number
of E' S needed
(at most 4)
At most
2 E' s needed
At most
4 E 's needed
55
Refrences
[ 1 ] E. Carlson, private communication.
[2] F. R. K. Chung and R. L. Graham, "Steiner trees for the regular simplex, " Bull.
Inst. Math. Acad. Sinica 4 ( 1 976 ) , 3 1 3-325.
[3] F. R. K. Chung and R. L. Graham, "Steiner trees for ladders, " Annals of Discrete
Math. 2 ( 1 9780, 1 73-200.
[4] F. R. K. Chung and R. L. Graham, "A new bound for Euclidean Steiner minimal
trees, " Annals of the N Y Acad. Sci. 440 ( 1 985), 328-346 .
[5] E. J. Cockayne and D. E. Hewgill , "Exact computation of Steiner minimal trees
in the plane," Infor. Prc. Letters 22 ( 1 986 ), 1 5 1-1 56 .
[6 ] M. Gardner, "Mathematical games," Scientifc Amercan (June 1 986 ) , 16 -22.
[ 7] M. R. Garey, R. L. Graham and D. S. Johnson, "The complexity of computing
Steiner minimal trees, " SIAM J Appl. Math 32( 1 977), 835-859.
[8] M. R. Garey and D. S. Johnson, Computers and Intractability, a Guide to the
Theory of NP-Completeness, W. H. Freeman and Co. , San Francisco, CA, 1 979.
[ 9] E. N. Gilbert and H. O. Pollak, "Steiner minimal trees, " SIAM J. Appl. Math.
16 ( 1 96 8), 1-29.
[ 1 0] R. Kallman, "On a conjecture of Gilbert and Pollak on minimal trees," Stud.
Appl. Math. 52 ( 1 973), 1 41 -1 5 1 .
[ 1 1 ] J . B. Kruskal, "On the shortest spanning subtree of a graph and the traveling
salesman problem," Prc. Amer. Math, Soc. 7( 1 956 ) , 48-50.
[ 1 2] H. O. Pollak, "Some remarks on the Steiner problem," J. Combin. Theory Ser. A
24(3) ( 1 978), 278-295.
[ 1 3] P. Winter, "An algorithm for the Steiner problem in the Euclidean plane, " Net
works 15 ( 1 985), 323-345.
Postso-i pt
Her.retuer.re,.in,paperaspublisuec,uaccev.tectuelast.rmy
matuemati.alGames..lumnsinScientifc Amercan ,1une l vss)t.
steiner trees wuen an upcatecreprint.rtuis ..lumn appearec in
Last Recreations ,sprin,erverla, l vv)accecan accencumanca
l.n,erbibli.,rapuy wuatr.ll.suereistuataccencumanctuelist
.rrereren.es
1.ma].rbreaktur.u,usrelatin,t.ms1s,minimalsteinertrees)
uave...urrecsin.etuis.uapterasritteninl vss nl vsst.Hell
iabsmatuemati.ians, n Or.ll.kancL NGilbert,..n]e.turectuat
tuerati..rtuelen,tu.ranms1t.tuelen,tu.rtueminimalspannin,
treer.r tue same set.rp.intsisatleast V/2 " o sssv v v , a savin,s
inlen,tu.rab.ut l - +/. 1uisistuerati.r.rtuet.kincs.rtrees
tuat].intue ..rners .ranequilateral trian,le. , see tue l vsspaper
by Gilbert anc r.ll.k ) n l vs r.nalc Grauam anc uis ire ran
cuun,raisectuel.erb.unc.rtuerati.t.o sz+l 1uepr..rass.
u.rrible, Grauam saic, tuatueur,ectu.seinterestecnot t.l..kup
tueirpaper
1ue pr.blem as imp.rtant en.u,u t. Hell iabs, uere rincin,
su.rternet.rksis an.bvi.us ..st savin,, r.rGrauamt..uer:oo
t. any.ne u...ulc pr.ve tue V/2 ..n]e.ture 1ueprizeas.n
in l vvobyt.cuinesematuemati.ians, Din,zuuDu, tuen ap.st
,racuatestucentatrrin.et.nuniversity, ancrranknan,, .rHell
iabs ,seetueir l vvzpaper )
A simplexisare,ularp.lyuecr.n, inanycimensi.n, ituamin
imumnumber .rsices, su.u astue - spa.e tetrauecr.n ms1s are
kn.n .nlyr.rsimplexestur.u,urivecimensi.ns , seetue l vspa
perbycuun, ancGilbert. ) cal.ulatin,tuemr.r ui,uercimensi.ns
israrrr.ms.lvec 1uems1r.rtue..rners.raunit.ubeissu.nin
i,ure l + tslen,tuiss l vs
Figure 14.
nan,ancDu, intueirl vvl paper,stucyms1s.nis.metri.,equi
lateraltrian,le)latti.ep.ints
1ue.tuerbreaktur.u,u, byriveAustralianmatuemati.ians, as
a..mpletes.luti.nt.rincin,ms1sr.rb.tusquareancre.tan,ular
latti.es.rp.intsinamatrix.runit squares , seetue l vvresear.u
rep.rtbym Hrazilancuisr.urass..iates ) nal vvspaperbyHrazil
ancriveass..iatestuey..nrirmectueunpublisuecpr..rbyGrauam
anc cuun, ab.ut tuer.rm.rms1sr.r sets .rp.ints attue verti.es
.ra2
k
x 2
k
squarelatti.e
1uereisa,r.in,literature.nminimalrectilinear steinertrees
trees itu .nly u.riz.ntal anc verti.al lines. 1uey uave imp.rtant
appli.ati.ns t. ele.tri.al .ir.uit cesi,n see Dana ri.uarc s ast
neuristi.Al,.ritumsr.rre.tilinearsteiner1rees, inAlgorthmica,
v.l +, l vl zo, l vsv.
nale.ture.nsteinertreesbyGrauam,ui.uuactuepleasure
.rattencin,, uein.lucectuer.ll.in,p.ints
1a..bsteinermacen...ntributi.nst.tuetue.ry.rsteinertrees
ex.eptt.,etuisnameatta.uect.tuem 1uep.intsereearlier.allec
ermat p.ints, buttueir existen.e askn.nevenber.re ermat s
time
1uerirstpr..rtuatrincin,steinertreesr.rn p.intsisNr..mplete
as in tue l v paper by Grauam anc t. Hell iabs ..llea,ues,
mi.uaelGareyancDavic1.uns.n. Als.Nr..mpleteistuepr.blem
.r.al.ulatin,tueexa.tlen,tu.raminimalspannin,tree. ntuitively
it seems as irtue ,reecy al,.ritum .ulcmaketuis easy. tisn.t
easybe.ausetuep.intsmayn.tbeatinte,er...rcinates.ntueplane.
Assu.up.intsin.reaseinnumber,.al.ulatin,tueexa.tlen,tu.rtue
treerapiclybe..mesm.recim.ult
Refrences
[ 1 ] E. N. Gilbert and H. O. Pollok, "Steiner minimal trees," SIAM Joural of Applied
Mathematics, Vol. 1 6( 1 ) , pp. 1-29, January 1 968.
[ 2] Fan Chung and E. N. Gilbert, "Steiner trees for the regular simplexes," Bulletin
of the Institute of Mathematics Academy Sinica, Vol . 4, pp. 3 1 3-325, 1 976.
[3] F. H. Hwang, "On Steiner minimal trees with rectilinear distance, " SIAM Jourl
of Applied Mathematics, Vol. 30, pp. 1 04-1 14, 1 976.
[4] M. R. Garey, R. L. Graham, and D. S. Johnson, "The complexity of computing
Steiner minimal trees, " SIAM Journal of Applied Mathematics, Vol. 32, pp. 835-
859, 1 977.
[5] F. R. K. Chung and R. L. Graham, "Steiner trees for ladders," Annals of Discrete
Mathematics, Vol . 2, pp. 1 73-200, 1 978.
[6 ] Dale T. Hoffman, "Smart soap bubbles can do calculus, " The Mathematics Teacher,
Vol. 72(5), pp. 377-385, 389, May 1 979.
[7] F. Chung and R. Graham, "A new bound for Euclidean Steiner minimal trees, "
Annals of the New York Academy of Sciences, Vol. 440, pp. 328-346 , 1 985.
[8] M. Bern and R. Graham, "The shortest-network problem," Scientifc Amercan,
Vol. 26 0, pp. 84-89, 1989.
[ 9] F. Chung, M. Gardner, and R. Graham, "Steiner trees on a checkerboard," Math
ematics Magazine, Vol. 6 2, pp. 83-96 , April 1 989.
[ 1 0] F. K. Hwang and D. Z. Du, "Steiner minimal trees on chinese checkerboards, "
Mathematics Magazine, Vol. 6 4, pp. 332-339, Decemer 1 99 1 .
[ 1 1 ] F. K. Hwang, D. S . Richards, and P. Winter, "The Steiner tree problem," Annals
of Discrete Mathematics, Vol. 53, Amsterdam, 1 992.
[ 1 2] R. B. Cohen, "Optimal Steiner points, " Mathematics Magazine, Vol. 6 5, pp. 323-
329, December 1 992.
[ 1 3] D. Z. Du and F. K. Hwang, "A proof of the Gilbert-Pollok conjecture on the Steiner
ration," Algorithmica, Vol. 7, pp. 1 21 -1 35, 1 992.
[ 1 4] R. Bridges, "Minimal Steiner trees for the three-dimensional networks, " The
Mathematical Gazette, Vol. 78, pp. 1 57-16 2, July 1 994.
[ 1 5] M. Brazil, J. H. Rubinstein, J. F. Weng, N. C. Wormald, and D. A. Thomas,
"Full minimal Steiner trees on lattice sets, " Research Report 14, Department of
Electrical Engineering, University of Melbourne, Australia, pp. 1-0, 1 995.
[ 16 ] M. Brazil, J. H. Rubinstein, D. A. Thomas, J. R. Weng, and N. C. Wormald,
"Minimal Steiner trees for rectangular arrays of lattice points, " Researh Report
24, University of Melbourne, Australia, pp. 1-28, 1 995.
[ 1 7] M. Brazil, T. Cole, J. H. Rubinstein, D. A. Thomas, J. F. Weng, and N. C.
Wormald, "Minimal Steiner trees for 2
k
X 2
k
square laticces, " Journal ofCom
binatoral Theory, Series A. , Vol. 73, pp. 91 -1 09, January 1 996 .
chqpter 7
Ti l i ng the Bent Tromi no
1uem.n.min.,unitsquare) , c.min., anc strai,uttr.min. .an ea.u
be.utint.n ..n,ruentparts, r.r anyinte,er n. 1uis is.bvi.us be
.auseea.uisare.tan,le, ancare.tan,le.anbe.utint.n ..n,ruent,
parallelstrips 1uebenttr.min.,.ritr.min.)is,tuerer.re, tuesim
plestp.ly.min.tuatraisesunanserecquesti.nsab.uttuevalues.r
n tuatall.r.rperre.ttilin,
. Gebelaskesr.ruatvalues.rn tuebenttr.min..anbe.utint.
n ..n,ruentparts l } . nerepr.cu.estureecisse.ti.ns, in z, - , anc+
..n,ruentparts, su.nuereini,ure l , ancasksirtuetr.min..an
be.utint.rive..n,ruentparts
He.ausea square.anbe.utint. n ..n,ruentstrips, itr.ll.sat
.n.en.mtuen " - .asetuatperre.t s.luti.nsexistr.ranyn tuatis
amultiple.r- 1uen " +.ases.lvesanan.ientpuzzleusually,iven
interms.ra rarmtuatist.becivicecint.r.ur..n,ruentre,i.ns
1ueunique s.luti.n usestiles.rtue same suape astue rarm. ns
w. G.l.mb stermin.l.,y, su.usuapesare.allecreptiles.
n 2
n = 3 n P 4
Figure i .
This article frst appeared i n the Joual of Recreational Mathematics, Vol . 22(3) ,
1990, pp. 1 85-191 .
61
Figure 2.
He.auseea.u.rtuesereptiles.ansimilarlybe.utint.r.ursmaller
repli.as, itr.ll.stuatperre.ts.luti.nsexistr.ranyotuatisap.er
.r+. Alltureetilin,sinri,ure l aresurelyunique,tu.u,ukn..r
n.pr..rs
1uetiler.rtueo" z .aseisan.tuerreptile i,urez su.su.
it.anbe.utint.r.ursmallerrepli.ast.pr.viceas.luti.nr.ro" s,
ancr.rs.luti.nsuenotakesvaluesi ntueseriess, -z, l zs, l z,
Gebel als.statestuatit is n.tuarct. rinc s.luti.nsuen ouas
tuer.rm +.r
.
.r z
.
, ancasksirtuereareperre.t s.luti.nsr.r
any on.t.rtuesetureer.rms l } s.luti.ns,ivenuerer.ro" l oanc
l+ ansertuisquesti.nauirmatively N.s.luti.nsarekn.nuen
oisprime,reatertuan- , n.rareimp.ssibilitypr..rskn.nr.rsu.u
.ases
raperre.ttilin,isimp.ssible,e.anseekr.ratilin,by..n,ru
ent tiles, ea.u l/otu tue tr.min. s area, tuat ill ..ver a maximum
area.rtuetr.min. 1uetilesmustn.t, .r..urse, .verlap, buttuey
ne.essarilyill.verlap,ats.mesp.t.rsp.ts, tuetr.min. sb.rcer
i,ures-t.s su.tuebestresultsuaver.uncr.ro" , , l o,
l l , l - , ancl + La.uiscran.n,rapupaperitutuetr.min. ssu.rt
sicesequal t. o1uetile, tuerer.re, ill uave an area.r+o, anctue
tr.min. sarea ill be+o
.
n ea.u .asetue tile uas tuer.rmsu.n
inri,urevuencran.ntue,rapu.
iet s.allri,urevtue.an.ni.altiler.r a,iveno. tismyrirst
..n]e.turetuatr.reveryoabest s.luti.n.anbea.uievecby.an.n
i.altilin,. wuenoisamultiple.r- , tuesimplesttilin,is, .r..urse,
byre.tan,les.rsize+ x o, butinevery.asetuetr.min..anals.be
tilec.an.ni.ally
7. Ti l i ng the Bent Tromi no
63
Figure 3.
t
~
,
-.
-_.
I
I
I
-
-
I
I
f
-

-. 1- f.-.
Figure 4.
64 A Gardner' s Workout
M - 1 0
Figue 5.
M .. 1 1
Figue 6.
7. Ti l i ng the Bent Tromi no
r --' -TL
T
T g -=p~
--- . - ~ -
1 --f , t--I- f--
I ---
j == - , . _. .. -- --
1-' 1 --' -- f--
j-- 1--.-
:
: "-
I --i- - -
i
-
-
=.
r--;--
r
-
|
-
1
-
-
r
-
-
1----
!
-
-
i--
i-- $~~
-
-
1-
1-- - --
i
-
-
, -
-
!
-
-
-
-
- --
=
!
i
-
-
--
--
!
-
-
1
-
-
f--r-
I
1
-
-
" "
= a =
, I
i
--
1
-
-
1- -
i
-
-
i
-
-
1
-
-
t
-
-
i-
!---
1
-
-
(.,
r
-
r-
!-
!
-
-
I--t-
!
-
-
i
-
-
1- t-
' -_ - ~ *! ~ ~r~ r ~~ _~ = _ ...
[
~
_
...
- 1
-
1
wm .=f- 1-----
|"
. .= ~+~|
m : f-
| |
1gm-~- +_1
|
|
| |
f - --. -- ~~!

--
:-- ! --.
l
-
-
!--- _.- - .
-+- ! --'---
,
-- f-- - - i-- i - - = . - -- - - r- ] =
i

-- ----- ! --
t
----
- !---
,-
r -
-
- - - i - +- : _ j=
|
-+-.,--
t
-- ' '
-1 -- 1- 1 ----- ,--j~-
Lma
|--
-
-
i
1-
-
I i
I
|
--
r
-
-
1
--
i -- !--
!
-
-,....
- - ! - -
t
-
-
1--t-
1--
I
-
-
|
--
!--l-
f--
.. ..
1-
|
- .Ja..
n - 1 3
Fjgue 7,
65
-,--,-- g~~p-~p-
-
--," - T--r--r- "--'- -T-- r--r-..-,--,--r--r--,-_~- , *~@ ~ @ ~ -,-_--q~-
7
~=_
f t t | | t t t t | t t | t | | t | t t t | t 1 | | f t
1-
-++++++- -- - - f - - - +- +++-l
| | t t t | I | t | | t t t | | t | t t t | t t t , f |
- -t--t---+-'--1--t---+-+-1--t--t- "+"q""|""}"++"'""[""J"++"+"[""|""|
| | 1 t
I I
| I | | t l t t t t | t t | t t | | t
t-- -t-- , -----,--i--
t t t
" j"{""||"" -" |]+"""[""|" "
t t t _ t | t t t | t t | t t t t t t | f |
~ -4----_
.
_---- ~~~~~-+-~+ ^ - ~~~~~~+~~+~~~~~^^^^+^^^=^=<m+=~
I t t t $ I t | I t t t | | t | t | t t t | t t |
t t t t l t t t | t t | t | | t t | t t t t | t t l
t~~
--r--r-""-;--1--- -"-;--1----t-...-"--"--"--"--t-..
-
,--,-~,
t t I t t t | | t I t t | | | t | | | f | t |
L. _! .. L.. ..a.+ . . J.. mm |+JmJm=4mm L_ =JJmJ. . L.. ..... m J..I.. L.. L..J.. ! __ J
t t t t t t I t t t t I t t t | | t | t t t t t
t | t t t t | | t t t t t t t | | t t | t t , t , t | t
* ~~- <^~~*9-^*m*^+*9^*-^%**-**
t t t t t t t t t I | | | t | | t t I | t t t t t t t
| t t t t t t t t I t t | t | | t t t t t t t |
@- ~q~~~~~_-q~~q~~~~p~~p~~q~-=~p= -~q~~- ~g~~~~-~~~~ ,--y~~p-@*\~-,--,
t t t t t t t | t t t t I t | , t t | | t t t | t
1-- ++++++- -+++++-++++++-+++-/
- -l--L--L-.-J--l---L-.-.L---!--L- -.-J--J--L--L-.-----!--L-.--.---!--l
t t t , t t t t t t t t t | | t t f t t t t l I t
| t I t t I | t t t t
t t | | t t t f t t t t t t
,~ ~~p~=p=q===~~.p~ ~_~~~~,~,_~~^*~,*_~_~+,~~~~,
t t t t t t I t t , t t t t t | t | t f t t t t t
\- +++-++-.--f
-
+-+-++-+-+ ++-!++++++++++
t I t I t t t t t t | t
,
: t | t | I t l t t t t $ t t t
r-
-t-t--t--t--r-t-i--t--t--t-t-i--t--t--I
, I I I I I I I I I I I I I t I t t t t t t t , ,
-l--r--r-T-'--"--T--r-T-'--'--T--r- -,-,--,--r--r-T-,--"--T--r--"-"-'--T--'
t t t t t t | t t t t | t t t | | t t t t t t t t t t
1-
+++
-+-1--
1--
--I
"+"""|"*["""+^""{""+"""""_""1"* --t
m
t
, `
t t
I =~ ---i--+- -f- -+---t-- t--- +- '
t
|
--i
| t f t t | , f t | t t t t t t t | | t | t t t t
j-- ---- --(--|--|-- __- --|
| t t t t t l | t | | | t t t | t 1 t | | | t t t
cm . ..L. . .. ...
.. .. .. t
| | I t t I I t I t | | : I : : I | I | I | ; I t I
. ~..qm;... (.. ..__ .. _+_ =~a..... -4---"--"--"--"--"----"-+--"-- --.
t t t t t t t t I t t | t
. t
1
_
t t t t t t f | $ 9
- ++++++-
-
-
I
t t f t t , _ t l t t | t t t | t I t t | t $ t
.. ma..&...... .mam mm..
...... &mmm mu_ a.. ..tmm....... ... ..mm ... m m:

t t I t l t t t t | t t t t l t f t t t , | t t t t
t t | l t t t t I : t t t | t t t t | t r t
~~ -~~~~<""-~ -+-~+-~~~~-~ !
t t t t t t t t : t | | | t | t t l | t t t t t
-------- -|--j------- --
~ ,.a.. .. . L== ...
: I : L : | J J. I | I | | I
t
: I : l : I l I : :
g
-
---
,
g - ---]--- - , --r--r- , -i-1-- --I
l- ... .. . . .. ,..
t t t t t | t t
-4- - ' ---+-+-4-- --t

.
|-|
(|
,
_-` -(]|_
{--|-]|-|q-
| | t I t , 1 t : l l t I t t t t t t t t
..L..1I.!L.1JLl:1:J__ t mml_LLL_l_LLLJ_J.
n P 1 4
Figure 8.
n/2
`
|
2
n/2
|
.
n
Figure 9.
7. Ti l i ng the Bent Tromi no 67
1uesepatterns, anc similarpatternsr.r ui,uer otuatare easily
r.unc, su,,esttuer.ll.in,acciti.nal..n]e.tures
l . rerre.t.an.ni.altilin,sexistr.reveryo,reatertuan z.
z. 1uerearen.perre.t.an.ni.altilin,sr.rany.cco,reatertuan
-tuatisn.tamultiple.r- 1uisset.r..ursein.lucesallprimes
ab.ve- .
- r.reveryotuatuasn.perre.ttilin,,a.an.ni.altilin,ill..ver
allbuto.rtuet.talarea,ancn.bettertilin,isp.ssible
1. .btain a .an.ni.al tilin,, tue pr..ecure is simple. 1ile c.n
tuetr.min. slertsiceitu6 x ore.tan,les, ea.u.rui.u.anbe.ut
int. t. .an.ni.al tiles wuen tuis ..lumn .rre.tan,les .ann.t be
..ntinuec, tiles,.attueb.tt.minapatterntuatvariesituo
Artertueb.tt.mtilesaresuitablypla.ec, ..ntinueuparc.ntue
ri,ut rois.ccancn.tamultiple.r- , tuet.pm.sttile.ntueri,ut
illpr.]e.tab.vetueb.rceritua 1 x re.tan,le 1uisne.essarily
leaves un..verec a re.tan,le .rtue same size anc suape As o in
.reases, tueun..verec spa.ebe..mes a smalleranc smallerrra.ti.n
.rtuet.tal aea, appr.a.uin,butneverrea.uin, zer.astuelimit
can.ni.altilin,.learlyilln.ttileperre.tlyuenois.ccancn.t
amultiple.r- 1uel.est..mp.site.aseis z. n.ever,asGc bel
re..,nizec, squareos.anbeperre.tlytilecitubenttr.min.es ri,
urelosu.stilin,r.r squares.r- tur.u,u. He.auseea.utr.min.
.anbe.utint...n,ruentualves, seeri,ure l ) , .bvi.uslyanysquare
otilin,illpr.viceatilin,r.rzo.
N.tuin,intuisarti.lerules.utperre.ttilin,suenois.cc, n.t
square, ancn.t civisible by - .ulcel..me anypr..rs.r ..un
terexamplest.my..n]e.tures
Refrences
[ I ] F. G(bel, Problem 1 77 1-The L-Shape Dissection Problem, Joural of Recre
ational Mathematics, Vol. 22( 1), pp. 6 4- 6 5, 1 990.
Postso-i pt
mi.uaelHeeler,intueJournal of Recreational Mathematics ,v.l. z+,
N.. l , l vvz, pp. ssv) ,invertecmyappr.a.ut.tilin,itr.min.es
itu small stret.uec itr.min.es. nsteac, ue tiles u.riz.ntally
stret.uecitr.min.esitun.rmalbenttr.min.es, tuenstret.uestue
Figue 10.
stru.tureverti.allyt.makean.rmalbenttr.min.tuatistilecbytue
stret.uec.nes tsin,tuiste.unique,uepr.vestuatr.rany..mp.s
iten tuebenttr.min..anbetilecitustret.uec.nes iert.penis
tuequesti.n.ruetuerabenttr.min..anbecisse.tecint.aprime
number.r..n,ruentsuapes,reatertuan-
Heelerals.su.stuatirn i s asquare,asymmetri.altilin,bybent
tr.min.esisalaysp.ssible. ri,ure l l su.su.tuisisc.neuen
n 9
2

iet k betuenumber.r.ells.n asu.rt sice.ra bent tr.min., as
su.ninri,urel z rk isn.tamultiple.r- , tuebenttr.min..ann.t
betilecitu strai,uttr.min.es. 1ue cia,.nal..l.rin, insurestuat
every strai,uttr.min. pla.ec .n tue pattern must ..verturee.ells
tuat..ntainallturee..l.rs. Huttuenumber.r.ells.rea.u..l.rare
not equal 1uey areinaritumeti.alpr.,ressi.n. ntue.asepi.turec
(k " +),tuereare l ,ray.ells, l s bla.k,anc l uite.
7. Ti l i ng the Bent Tromi no 69
Figure 11 . Beeler's symmetrical tiling when T 92.
1ue cia,.nai -..i.rin,pr.vicesapieasant..mpani.nt.tue.ic
brain teaserab.utwuetuer a .uess b.arc, witucia.naily.pp.site
..rner .elis rem.vec, .an be ..verec itu - i n.n.veriappin c.mi
n.es. nem.ve]ust.ne..rner.eii. cantues-remainin,unitsquares
be..verecwituz i strai,uttr.min.es: 1uecia,.nal..i.rin, r.ves
imp.ssibiiity.
Figure 12. A three-color proof that the bent tromino with a short side of 4 cannot be
tiled with straight trominoes .
1 1 1 1
2 2 2 2
3 3 3 3
1 1 1 1
2 2 2 2 2 2 2 2
3 3 3 3 3 3 3 3
1 1 1 1 1 1 1 1
2 2 2 2 2 2 2 2
Figure 13. Another impossibility proof.
wuensentmy..l.rin,pr..rt.D.nalcknutuuerealizecat.n.e
tuatit..ulcpr.vemu.um.re. ne su.ectuatirtue squares.r
a n.nstrai,ut .n.m. arecivicecint.n x n unit squares, tue
.n.m..ann.tbetilecitun2 strai,ut.nim.esuenn isrelatively
primet. . 1uus an.n strai,utpent.min.,itsrive squares civicec
int.n x n unit squares, .anbetilecitun
2
strai,utpent.min.esir
anc.nlyirn isamultiple.r A6 x 6 square.an tbepa.kecituv
strai,uttetr.min.es , 4, ui.uisrelativelyprimet.n +) Anc
tuereareencless.tuerexamples
1ueseresults.an als.be.btainecbylabelin,tuer.s.rap.ly
.min. l , z, -, l , z, - , v o v i,ure l - su.su. tuis applies t.tue
itr.min.t.pr.veit.ann.tbetilecitu strai,uttr.min.esunless
itssu.rt siceisamultiple.r- Anystrai,uttr.min.pla.ec.ntue
patternill..verturee.ellsituasumtuatisamultiple.r- nen.e
irtuepattern .anbe..verecitu.ut.verlappin,, tuet.tal.rallits
.ellsmustbeamultiple.r- ntuis.asetuet.tal,vz,isn.tamutliple
.r- .
1uemainuns.lvecpr.blemab.utitr.min.tilin,, ui.umaybe
very cim.ult, is t. su. tuat it .an t be tilec iturive ..n,ruent
suapes reruapstuisillrall.ut.ram.re,eneralpr..rtuattilin,
isimp.ssibleirtuetr.min. ssu.rtsiceisaprime,reatertuan-
cuttin, tue bent tr.min. int. r.ur ..n,ruent parts i s a puzzle
r.unc in manyearlypuzzle b..ks 1uis pa,e is rr.m Ge.r,e carl
s.n sPeter Puzzlemaker, a l vzzb..kr.r.uilcren
7. Ti l i ng the Bent Twmi no
.
<
; `` v
. . .-
e -
+ ,

9
-* "
B
BAKER b,ked , d,""oo, ,,1,ln k, ood wId , q"""" of "
;
: The cake was a square, and the piece cut from it was also a
::
; square. Four hungr kids enter the shop to buy the rest of the
; cake. The baker wants three dollars for it, but the kids have
only two dollars. As it is the youngest boy' s birthday, the
baker will let them have the cake for two dollars if they can tell
him how to cut it into four pieces of exactly the same size and shape. How
will they get the cake?
Figure 14.
71
Ch'pter 8
Coveri ng ' Cube with
Cong ruent Pol
ygons

cantue surra.e .ra.ubebe..verec, itu.ut.verlap, itu
..n,ruentp.ly,.ns, ueren is anyinte,er,reatertuan l :
1uepr.blemi s uns.lvec

1uink.rtuep.ly,.nsaspie.es.rpapertuat.anbepastec.ntue
surra.e .ra .ube, r.lcin, tuem uen ne.essary. Altuernatively, e
.an tuink.r..l.rin, tue .ube itu ..l.rs s. tuattuere,i.ns .an
beunr.lcect.r.rm ..n,ruentp.ly,.ns. 1ue.ase.r 2 iseasily
s.lvecbyt.re.tan,les.rsize+x 1 r.rtueunit.ube. He.auseea.u
re.tan,le .an becivicecint. n ..n,ruentre.tan,les, itr.ll.stuat
tue.ri,inalpr.blem.anbes.lvecr.r any even n.
1ue .ase .rn + isals. easily s.lvecbyturee 2 x 1 re.tan,les
A,ain, ea.ure.tan,le.anbe.utint.n ..n,ruentre.tan,les, tuerer.re
tue.ri,inaltaskis s.lvecr.rany tat is a multiple of 3.
wen.inquire uetuertuereis as.luti.nr.rn 5. 1ueanser
isyes. 1ue .ube .an be ..verecbyrive..n,ruent Latin crosses, as
su.nini,urel
1uispr.blemasp.secinMatematics Magazine, 1anuary l vs,
byveitLlser.rsan1.se,calir.rnia A s.luti.nbytuepr.p.seranc
bymi.uaelG.lcber,.rwasuin,t.n, D c. , appearecinMathematics
Magazine, v.l. o , l v), pp l ssl v.
1uenextui,uer tuatremainsuns.lvecin 7. Ca the cube be
covered with seven congruent polygons ? 1uisquesti.nisunanserec
The above noted appeared in Cubism for Fun (No. 25, December 1990, Part 1 ,
pp. 1 3) .
73
Figure 1. A cube covered by five congruent Latin Crosses. The folds are indicated by the
dotted lines.
I here ofer U. S.
$
50 to the fst person who either fnds a solution or
proves it impossible.
Postsct" pt
1.myvastsurpriseAnneke1reepn.t.nly.nmyrirtyc.llars, sue
r.uncr.ur cirrerent strips tuat .ulc rap ar.unc a .ube, ea.u .r
ui.u ..ulc be civicec int. n ..n,ruent parts 1uey are su.n in
i,ure z i,ure- su.su.tuerirststrip is easily.utint.n..n
,ruentparts. 1reepexplainectuisinuerarti.le c.verin, a cube,
in Cubism for Fun ,N.. z, De.ember l vvl , pp l sl ) .
8. Cover i ng a Cube with Congruent Pol ygons 75
h.
C
Figure 2.
Figure 3.
ChC
pter 9
MCg i c Tri cks on C Computer
several yeas a,. as inv.lvec in a s.rtare pr.]e.t inv.lvin, a
ciskette tuat .ulc play tue r.le .r a ma,i.ian, perr.rmin, ma,i.
tri.ks, s.me .rui.u .ulcmystiry evenapers.nrunnin, tuepr.
,rams N.tuin,.ame.rtuispr.]e.t,butinplannin,materialr.rtue
ciskette ,..n]urin, is my prin.ipal u.bby) a..umulatec an exten
siverile.rtri.kstuat..ulc easilybepr.,rammec s.mema,i.ians
uavetu.u,utal.n,similarlines, ancpr.,ramsuavebeenpublisuec
in ma,i. peri.ci.als, buttuerielcremains ice .pen t. expl.rati.n
byma,i.iansin..llab.rati.nitu.leverpr.,rammers. nerearer.ur
typi.aliceas.
A C tq Pteq i di on
1uetri.kbe,ins itutue..mputertur.in,.ntue s.reenapi.ture
.rtueba.k.raplayin,.arc tremainstuereuntiltuetri.k s.limax
Y.uaret.lc ,in ab..kleta...mpanyin,tueciskette)t.suurrlea
ce.k .r.arcs tu.r.u,uly, tuen type int. tue ..mputertuenames .r
tueten.arcs, insequen.e, tuatareattueb.tt.m.rtuesuuulecce.k
ieavetueten.arcsintuesamep.siti.nintuece.k
n.lcin,tuece.kra.ec.n, startcealin,.arcsn.mitra.eupt.
r.rmapile. As y.uceal,..untba.karcn.ml ot. l Allra.e.arcs
areassi,necavalue.rl o ra.arc svaluemat.uesa.allecnumber,
st.pcealin, ancstartase..ncpile r.rexample, assumey.u..unt
l o, v, s, ,s, ancasixiscealt.ntue..unt.rsix startase..ncpile,
leavin,tuerirstpileitutuesixra.eup.nt.p 1uisvalue.bvi.usly
asceterminecbypure.uan.e
1ureem.repilesare r.rmeci nexa.tlytue sameay. rn...in
.icen.e.r.arcancnumber...urs, killtuepilebyputtin,tuenext
This article frst appeared in Algorthm (Jan/Feb 1990) .
77
.arc ra.e c.n .nt.p.rit rall r.ur piles arekillec ,tuis selc.m
uappens) , tuen suuule tue r.ur piles t.,etuer, repla.e tue .arcs .n
t.p.rtueuncealt.arcs, ancstart.ver
Artertuer.urpilesare..mpletec, acctuevalues.rtue.arcs.n
t.p.rtueunkillecpiles .rexample, .nepilemayuavebeenkillec,
ancaturee,queen,ancninearera.eup.ntue.tuertureepiles. 1ue
queen ..unts as lo, s.tuesumis- + l o+ v" zz c.untt.tuezznc
.arcintueremaincer.rtuece.kancn.teitsra.e
A ..mputerkey, spe.irieci nacvan.e,itisalaystue samekey),
isn.pressec sl.lytue.arc.ntues.reenr.tatest.cis.l.seits
.tuersice tillbetuevery.arcy.useemin,ly.u.seatranc.m
n. c.estue ..mputerkn.: wuen y.u type tue names .rtue
ten .arcs at tue b.tt.m .rtue suumec ce.k, int. tue ..mputer, it
i,n.res all tue names ex.epttue nintu 1uis istue.arc u.se ra.e
tue..mputerputs.ntues.reenastuera.e.rtuer.tatin,.arc
Abit.relementary al,ebraill..nvin.ey.utuattuenintu.arc
rr.m tue ce.k s b.tt.mis alays tue .ne sele.tec. 1ue tri.k s m.st
amusin,aspe.tistuatspe.tat.rsuaveastr.n,,but..mpletelyralse)
impressi.n tuat s.meu. tue ..mputer as able t. put tue .u.sen
.arc.ntues.reen,itsra.euiccen,ber.retuece.kassuuulec
Guess i ng A Wotd
n tuistri.ky.u assist tue ..mputerin a sneaky ay 1ue tri.k is
intencecr.r a,r.up.r.nl..kersu.c.n.tsuspe.ttuaty.uplay a
r.lein its .rkin,
s.me.neneelynamesar.urletter.rc 1ue..mputertuenasks
tuat pers.n a series .r r.ur questi.ns Onl..kers at.u .ver y.ur
su.ulcerasy.utypeuis.ruerreplies.nt.tues.reen
1ue pr.,ram .an u.lc alist.r, say, auuncrecquesti.ns s. tuat
tue r.ur it asks are ciuerent ea.u time tue tri.k is repeatec. 1ue
questi.nsaret.tallyirrelevant .rexample, tuerirstquesti.n..ulc
be nuat.ityancstateerey.ub.rn:
1uenextquesti.n..ulcbe wuatistuerirstline.r.ne.ry.ur
rav.ritep.ems: 1uenextt.mi,utbe wu.playectuemaleleac
in casablan.a: anc wuat s tue sum .r ten tu.usanc anc rirteen
tu.usanc: nea.u.asey.utypeexa.tlyuattuepers.ntellsy.u
wuen a .ertain key is pressec, tue .u.sen .rc rlasues .n tue
s.reen.
9. Magi c Tri ck5 on a Computer 79
Y.umust,.r..urse, se.retlysi,naltue.rc 1uerearemanyays
t.c.tuis nereis.ne.rtuesimplest 1ue..mputerispr.,rammec
n.tt. su..ntue s.reen tuelettery.utypearteruittin, tue spa.e
barr.r tuerirsttime supp.setue.u.sen.rcispink wueny.u
type tue anser t.tue rirstquesti.n, starttue se..nc .rc itu p,
tuen itu.utuesitati.n ,..nt. tueletters .rtue se..nc.rc. 1ue
s.reenilln.tsu.p.rm.vetue.urs.r spe.tat.rsillbeat.uin,
tues.reen, n.ty.urrin,ers. ntue sameayy.u,ivetue..mputer
i, n,kasy.utypeanserst.tuelasttureequesti.ns
1ue rirst publisuec tri.k.rtuis type, i nui.u y.u se.retly .ue
tue..mputeruiley.utypeansers, asinventecbycurism.r,an,
r.rmer ecit.r.rByte. nm.r,an sversi.n, tue icentity.ra .u.sen
.arcis..nveyecbyuittin,.rn.tuittin,tuespa.ebarartery.urinisu
typin,tueanserst. yes.r n.questi.ns Y.u llrinctuis ex.ellent
tri.kces.ribecbykeeDecneyinuisScientifc Amercan ..lumnr.r
Au,ust l vss, anc by m.r,anin Byte, all l vv m.r,anpr.vices a
..mpletepr.,raminHAsc
A Hqt Tli ck
1ues.reensu.sar..r,say,tureeuats, brimsc.n,ancallalike
1uey arelabelecAHc rressin,a..rresp.ncin, letterraisesauata
re in.ues A se..ncpressin, l.ers it. Hesice tue uats is a small
bunny.
stanc.ntuerasice.rtuer..m,ba.kt.tueaucien.e, uiles.me
.neraisesauat,usestuem.uset.puttuerabbituncerit,tuenl.ers
tueuat.vertuerabbit Y.uturnar.unc, returnt.tue..mputer,anc
at.n.epresstuekeytuatlirtstueuattuatuicestuebunny n.c.
y.ukn.ui.uuatt.raise:
ntueprevi.ustri.ky.use.retly.uetue..mputer ntuis.ne,it
se.retly.uesy.u. wuiletueuatisbein,raisec, tue..mputerimper
.eptibly alters tue pi.ture.rtue uat.ne stept. tueri,ut ,Assume
AHct.be.y.li. )1uis.anbec.nebypla.in,atinyc.ts.meuere.n
tueuat,.rbyrem.vin,atinyc.t,.rany.tuersmall.uan,et..small
t.ben.ti.eableunlessy.ukn.ueret.l..k Ontue.u.uan.etuat
apers.n is.l.selyat.uin,tueuatbein,raisec, tue.uan,eismace
.nan.tueruat
1uetri.k.bvi.uslyuas manyvariants ar..ralnutsuellsanc
apea,ar..rinvertec.upsancasmall.b]e.tsu.uasarin,, marble,
Your victim hides the rabbit under a hat.
Figure 1 .
.rmat.ur.lcer, ar..renvel.pesancatu.usancc.llarbill, ar.
.rc..rsbeuincui.ualepre.uaunuices, ancs..n.
Guess i ng A Di g it
Askany.ner.rac.llarbill, anc.allattenti.nt.itsei,utci,itnum
ber reactueseci,its al.uc sl.ly Asy.u c. speak, asktue bill s
.nert.tuinksilently.ranyn.nzer.ci,itnamec
nancba.ktuebillanctelltuepers.nt.,ivetue..mputer,inany
.rcer, alltue ci,its ex.epttue .nementally sele.tec 1ue..mputer
s.reen remains blank wuile tue .ner c.es tuis, y.u stanc a.r.ss
tuer..mituy.urba.kturnec 1uusy.uuaven.inklin, .ruat
ci,itue.r suetu.u,ut.r
Y.u n. return t. tue ..mputer anc type wuat ci,it is .
tuinkin, .r: ,in.lucin,tuepers.n sname in tueblankspa.e) 1ue
.u.senci,itinstantlyappears.ntues.reen
1uese.ret Asy.u.all.utueci,itsattuestart.rtuetri.k, men
tallyacctueminy.urueac,.astin,.utninesasy.u,.al.n, 1uis
is easilyc.nebyaccin,tue ci,itsin anyt.ci,it sum ancremem
9. Magi c Tri ck5 on a Computer 81
berin,.nlytuesin,leci,itresult .rexample,irtuerirstnumberis
vanctuese..nciss,y.usayt.y.urselrvplussisl , anclplusis
s Y.uremembers 1uisiskn.nastueci,italr..t.rtuesequen.e
Y.umustse.retlytelltue..mputertuisnumber Oneayt.c.it
ist.presstueletterkeycire.tlybel.tueci,it.ntuekeyb.arc st.p
r.. Asber.re, y.u.anc.tuis]ustanertue spa.ebeteentuerirst
t. .rcs.rtuequesti.n. 1ue..mputer subtra.tsrr.m tuis ci,ital
r..ttueci,italr..t.rtueset.rseven numbers earlier ,ivent.itby
tuebill s.ner ,1ue..mputeruas.r..urseceterminectueci,ital
r..t.rtu.se seven numbers ber.rey.u senc ittue ci,italr..t.rall
ei,ut ) rtueresultisp.sitive, tuatistuenumber.u.sen rzer..r
ne,ative, tue.u.sennumberistueresultplusv
u.petueser.urtri.kssu,,esttuerlav.rancvariety.ramusin,
ci,italma,i.tuat..mputers.anperr.rm 1uereareenclessp.ssibil
itiesinv.lvin,.arcs, ..ins, ci.e, mat.ues, anc.tuer..mm.n.b]e.ts,
asellas.rcsancnumbers nmany.asesitis.nlyaquesti.n.r
acaptin,aellkn.nma,i.tri.kt.a..mputerpresentati.n.
Ch(
pter 1 0
V( ri (ti ons on the 12345679
Tri ck
An.lcnumberstunttuatstillneverrailst.entertain.uilcrenstarts
itu l z-+sv,n.tetuemissin, s)intuereac.ut .rtue .al.ulat.r
Ask a.uilct. tell y.uuisrav.riteci,it.rl tur.u,u v, .all tuisci,it
A 1elluimt.multiply l z-+svby A, anctuentuatpr.cu.tby v
1ueresultisAAAAAAAAA
re.ently amusecmyselrby.rkin,.ut s.me variati.ns.ntuis
.uestnut. .r example, tue tri.k,eneralizes easily t. .u.sen num
bers.rm.retuan.neci,it, su.uasstartin,itul l zz--++sssv
sele.tanyt.ai,itnumber,prererablyituciuerentai,its, .alltuis
numberAH,p.siti.naln.tati.n) multiplytuebi,numberbyAH,anc
tuenbyv, ancy.u,etanAHAHAH strin,
1ue .ri,inal stunt .rks, .r ..urse, be.ause v times l z-+sv
equals l l l l l l l l l 1ue ,eneralizati.n .rks be.ause v times tue
startin, number is l ol ol ol N. r.r s.me variati.ns basec .n
numbers.btainecbycivicin,repunits, l l l l l l l )byaninte,erun
tiltuereisn.remaincer
rut-o-intuereac.ut multiplybyA, tuenby-, t.,etAAAAAA
1ue number y.u start itu .an be a l.n, number tuat repeats
-o-o-
uts+intuereac.ut multiplybyA,tuenbyl - , t.,etAAAAAA
Oruses+oos+oos+
ut-oo-oo-intuereac.ut,.ranumberbasec.n-oo-oo-oo-
multiplybyA, tuenby-
.rzs+v,.rzs+voozs+v ) , multiplybyA, tuen-v
.rz l ,.rz l ooz l ) , multiplybyA, tuen+l
.r l zzl ,.r l zzl ool zz l ) , multiplybyA, tuenvl
He.ausetue.veruelmin,number.rrepunitsare..mp.site,tuere
isaninrinitenumber.rstartin,numberstuat.perateintuisrasui.n
This article frst appeared in the REC Newsletter (Dec. 1991 ) .
83
nereisacirrerentvariati.n, tuis.nebasec.n tueprimera.t.rs
.rl z-+sv, - anc---ss rut---ssintuereac.ut multiply by
A anctuenby---t.,etAAAAAAAAA 1uis.rksbe.ause---" v
x - itu - ara.t.r.r l z-+sv
Or..urse, y.u.anc.tuisstuntinreverse ut-intuereac.ut
multiplybyA anc tuen by-oo-oo- ,tuepr.cu.t.r---ss anc v)t.
,etAAAAAAAAA
1uere is als. an inrinity .rstartin,numberstuat.an beusecr.r
,eneratin, anAHAHAH strin, 1uey are.btainecbycivicin,tue
numbers l ol ol byinte,ersuntiltuereisn.remaincer. Lxamples
startitu l ++- ,.r l ++-ool++- ) , multiplybyAH, anctuenby
Orstartitu--s ,.r --soo--s), multiplybyAH,anctuenby
- He.auseitiseasyen.u,ut.multiplyAHby - mentally, y.u.an
requesttuat--sbemultipliecbytueresult,- x AH) 1uis,enerates
tueAHAHAH strin,
startitu,.rooo), multiplybyAH, anctuenbyl -
A m.nstr.us ma,i. number lencs itselr t. a cem.nstrati.n .r
li,utnin, .al.ulati.n write c.n vl sz-s++s-zs l v l ,1. aic
mem.ry, n.tetue stru.ture .r..mplementary ci,its summin,t. l o
inpairs, r.ll.ecbyarinal l . )
Ask s.meb.cy t. .all.ut a t.ci,itnumber, AH, tue t. ci,its
prererably cirrerent ny.urueac, multiplyAH by l l ,easyen.u,u)
ancrequesttuattue,iantnumberbemultipliecby tuis pr.cu.t, l l
x AH) Or..urse, y.u ll neec s.meb.cy t. uave ..mputer a..esst.
verirytueresult.ranAHAHAH strin,
Ch(
pter 11
More C( l cu l (tor Wh i ms i es
..ketanccesk.al.ulat.rsaren.t.nlyuserul, tuey als..anbeusec
t. surpriseancentertainy.urselrancy.urrriencs nereis a.u.i.e
sele.ti.n.r.al.ulat.ramusements
l. sele.tany numberkey ,.tuertuano) ancpressi ttureetimes
Divicetuenumber.ncisplayby-, tuencivicetueresultbytue
number.ntuekeyy.urirstpun.uec. 1ueresult: -
z rut.ncisplay any ci,it n.m l tur.u,u s Divicebyv 1ue
anser ill stutter y.ur .ri,inal number clear tue cisplay,
ana enter a number beteen l o anc vs. 1uen aivice by vv
1ue.al.ulat.rilla,ainrepeattuesele.tecnumber N..lear
ancenteratureeci,itnumberlesstuatvvv, anccivicebyvvv
Guessuat:
- run.u in vss+- l z N.te tuat tue l anc z are in tue r.n,
.rcer. Diviceby s 1ue anser illsurprisey.u ,ry.ur.al
.ulat.ru.lcreertuannineci,its, y.u lluavet.c.tuis.neby
uanc )
+ rlu, in at.ci,itnumber, like -o reversetue ci,its ,o-, .r
]ust -) anc acc tue ne number t. tue .ne y.u pi.kec Y.u
,et a palincr.mea number tuat reacs tue same r.rarc anc
ba.karc witus.menumbers, y.uneect.reversetuesumanc
acc a,ainber.re y.u ,et results .r example, -v + v- " l -z,
anctuen l -z + z- l " -s-. Y.umayneect.repeattuesesteps
,reversin,ancaccin,)manytimesber.rey.u,etapalincr.me.
warnin, Don' t even ty 89 with a calculator.
1uepalincr.me.rsviss, s l - , zoo, oz- , l ss,ue ) Anc ]ust
in.asey.ureinterestec, l vsistuesmallestnumbertuat.an tbe
This article frst appeared in Zigzag, (Sept . 1995) .
85
turnecint.apalincr.metuis ay matuemati.ians uavetriec
reversin, anc accin, its ci,its m.retuan +, ooo times itu.ut
su..ess
Y.u ll neec a re,ular, six sicec cie ,like y.u .an rincin many
,ames)r.rtuistri.k. Givearrienctuecieanca.al.ulat.r. Ask
y.urrrienct.r.lltuecie, uiley.uuavey.urba.kturnec s.
y.u.antseeit. N.telly.urrriencrirstt.multiplytuenumber
.nt.p.rtueciebyvvvvvv, anctuent.civicetueresultby.
Lventu.u,uy.uc.n tkn.tuenumber.ntuecie,y.u.an
.all .ut,inranc.m.rcer, ea.uci,itintuepr.cu.t 1ue se.ret:
call.ut,inanysequen.ey.uliketueci,its l , z, +, , , ancs
s write-vvo.napie.e.r paperancpla.eitra.ec.n.natable
Ask .ne pers.n t. enter tueyear ue .r sue as b.rn .n tue
.al.ulat.r Ask ase..nc pers.n t.accuis .rueryear.rbirtu
Ask tuerirstpers.nt. acctue a,e ue.r sue ill be attueenc
.rtuis year. 1uenuavetue se..ncpers.nc.tue same 1urn
.ver tue papert. su. y.urpreci.ti.n. tmat.uestuet.talin
tue.al.ulat.r
. sele.tanyr..rtureenumberkeysancpresstueminany.rcer.
1.tuisnumberaccatureeci,itnumber.btainecbypun.uin,,
i nany .rcer, tuekeysi nan.tuerr. Acc atuircnumberby
pun.uin, in tuekeys ineituer cia,.nal inranc.m.rcer Acc
an.tuernumberusin,tuekeysintue.tuercia,.nalinranc.m
.rcer Y.usu.ulcn.uavear.urci,itnumber
writetuisnumberc.n, tuenaccall.ritsci,its. rtuesum
ism.retuan.neci,it, acctue t. ci,itst.,eta sin,leci,it.
Amazin,ly, y.uillrinctuattueci,it.ncisplayiss.
s. wuen.ertainnumbersintuereac.utarevieecupsicec.n,
tueymake.rcs nereareturee su.utri.ks nea.u.ase, arter
y.u c. tue matu, turn y.ur .al.ulat.r ar.unc t. reac uat it
says
sayuit.tuema.uine 1uencivices l szbys
wuat stue.apital.rcau.: multiplysby+
wuatcicsantaclaussayuenruc.lpusu.ecuim.ne.r
tuesestunts: multiplyo os-+bys
wuat c. y.u .all pe.ple u. .an t multiply s by -: 1ry
l v, --stimes-
11 . More ('i cu i 'tor Whi m5i e5 87
Onm.st.al.ulat.rs, tueupsicec.nreac.utsilll..kal.tlike
uLii O, HOsL, ancuOuOu O s.me.al.ulat.rs, su.uastu.se
builtint.au.me..mputer,mayn.tcisplaytuezer.t.tuelert
.rtuece.imalp.intinananser,preventin,tuerirstanctuirc
.rtuesetri.ksrr.m.rkin,
v Lntervss+-z Divicebys 1ueresultillbel z-+sv, itu
tue ei,ut stran,ely missin,. multiply tuis numberby y.urra
v.rite ci,itbeteen l anc v 1uen multiply tuatresult by v
Y.u ll,ety.urrav.riteci,itrepeatec.veranc.vera,ain.
Postso-i pt
nereares.mem.re..mputertri.ks.
l o rutl++-.ncisplay Asks.me.net.telly.uuera,e ,suemust
be.lcertuan v.) multiply l ++-byuer a,e,tuenmultiplyby
1ue..mputerillstutteruera,e
l l . Y.u llneecr.urpapermat.uesr.rtuistri.k 1uink.ranumber
rr.m l tur.u,uv,tuenputit.ncisplay rla.eamat.u.nt.p.r
tuerow .rkeystuat..ntaintuenumber,ancputan.tuermat.u
.ntuecolumn tuat..ntainstuenumber
1. tuenumberin tuereac.ut, acc any ci,it representec
byan uncovered key A,ain, ..veritsr.anc..lumnitut.
m.remat.ues Onekeyillremainun..verec Acctuisnumber
t.tueprevi.us sum
Lventu.u,utueturee ci,itsere sele.tec atranc.m,tue
rinal sumillbe l
l z nere sasimple,amer.rt.players 1uerirstplayerpusuesany
keyt.enteraci,it 1uese..ncplayeraccsanyci,itrepresentec
byakeyadjacent t.tuekeyprevi.uslypusuec ,Ac] a.entmeans
next t. tue key eituer t. tue lert.rri,ut, .r ab.ve .r bel.,
.rcia,.nally ) rlayerstake turns accin, ac] a.entci,its 1ue
inneristuerirstt.,.ab.veo.
Y.u.anvary tue ,ame by .u..sin, .tuer numbers r.r tue
,.al
l - 1uisisaayt.pr.vey.uarepsy.ui. Askarrienct.tuink.r
anytureeci,itnumber callitAHc 1elluimt.puttuenumber
.n cisplay ti.e t. make anumber.rtuer.rm AHcAHc ,.r
example,irue.u.se-s, ue.ulcenter-s-s ) wuileuec.es
tuis, stanc ituy.urba.kturnecs.y.u .an tseetuenumber
uesele.tec.
mstartin,t.,ets.mevibesrr.my.urnumber, y.u say
1uey tell me it s exa.tly civisible by tue unlu.ky number l -
rleaseciviceby l - anctellmeir mri,ut Y.uare 1uereis
n.remaincer
wituy.urba.kstillturnec, say,uaveastr.n,impressi.n
tuattuenumbern..ncisplayisexa.tlycivisiblebytuelu.ky
number l l wuenuecivicesbyl l , sureen.u,u, a,aintuereis
n.remaincer
rub y.ur r.reueac as ir..n.entratin, rreci.t tuat irue
civicesby tue lu.ky number , .n.e m.re tuere ill be n. re
maincer ri,uta,ain
1ell uimt.take a,..cl..kattuenumberintuereac.ut.
till beAHc, tuenumberuerirsttu.u,ut.r
l+. Asks.me.net. put.n cisplay anyr.urci,itnumber. n y.ur
minc, subtra.t z rr.m tuis number, tuenput z in rr.nt .rtue
result .rexample,assumeuis numberis -szl subtra.tin,z
,ives -sl v, anc puttin, z in n.nt .ritpr.cu.es z-sl v write
tuisnumber.napie.e.rpaper ancturnitra.ec.n
Ask y.ur rrienc t. put a se..nc r.urci,it number uncer
tue previ.us .ne Y.u tuen supply a tuircr.urci,it number,
apparently aranc.m.ne. A.tually, r.r ea.u ci,ity.u se.retly
sele.t a ci,ittuat uen accect. tue .necire.tly ab.veitill
t.talv
Y.ur rrienc n. rites a r.urtu r.urci,it number. Y.u
ritearirtu, a,ain.u..sin, ci,itstueayy.ucicber.re 1ue
rinalacciti.npr.blemilll..ks.metuin,liketuis
-szl
v+-
os+
zvl
zos
11 . More ('i cu i 'tor Whi m5i e5 89
1.taltuerivenumbers, usin,tue.al.ulat.r 1urn.vertue
papert.reveal y.ur..rre.tpreci.ati.n
l write c.n any tureeci,it number pr.vicec its rirst anc last
ci,itsareciuerent reversetueci,itst.r.rman.tuernumber
1aketuesmallerrr.mtuelar,er rtueresultuast.ci,its, put
azer.inrr.nt reversetuisnumberancaccitt.tueprevi.us
.ne surprisin,ly,tueanserillbel osv
1uenumber losvuasmany.uri.uspr.perties r.rexam
ple,multiplyitbyanyci,itex.ept uttuereverse.rtuisne
number .n cisplay, civicebytue number cire.tly .pp.sitetue
lastkeyy.upusuec r.rexample,iry.umultipliecby, civice
by- Y.uareba.kt. l osv
l s wuiley.urba.kisturnec,askarrienct.ritec.ntuenumber
.nanyc.llarbill, tuens.rambletueci,itsanyaysuelikest.
make a se..nc number tsin, a .al.ulat.r, sue subtra.ts tue
smaller numberrr.mtuelar,er
wituy.urba.kstillturnec, askuert.rixinuermincany
ci,it n.t zer. in tue anser, tuen sl.ly .all .ut every ci,it
ex.epttue.ne sue sele.tec Ass..nas suerinisues, y.uname
tue.mittecci,it
se.ret As sue .alls .ut ci,its, acc tuem in y.ur ueac
wuenevertue sum,.esab.ve.ne ci,it, acctuet.ci,itsanc
remember .nly tue sum r.r example,irsue .alls .utv anc s,
acc tuemt. ,et l , tuen acc l anc ancremember s Arter
sue uas .allecalltueci,itsex.epttue.u.sen.ne, subtra.ttue
sin,leci,ity.uuaverr.mv. 1ueresultillbetueci,itsuelert
.ut
l Lnterzl +-iny.ur.al.ulat.r ,N.te tuattuis number is aper
mutati.n .r l z-+) Divice by zz, tuen uit tue squarer..tkey
ti.e 1ue reac.ut ill ,ive ..rre.t t. ei,ut ce.imal ci,its
1ueappr.ximati.n is basec.n anicentity cis..verecby srini
vanaramanu] anin l vl +.
chq pter 12
J(qsp
q rov/ s Defeqt b
y
Dee
p
B l ue
nerbert sim.n .ulc.alluat Deepbluec.estuinkin,
1.unsearle Hal.ney
---quoted by Bruce Weber in "A Mean Chess-Playing Program Tears at the
Meaning of Thought, " ' New York Times (February 1 9, 1 996).
ma,ineapers.n..nversin,itut.uiccenentities, .neauuman
bein,, tue.tuer a..mputer rtuepers.nisunable t. ce.ice ui.u
isui.u, esuallber.r.ect. acmittuattuema.uineuasa.uievec
uumanintelli,en.e 1uatastue s..allec1urin,test, pr.p.secby
Hritisumatuemati.ianAlan1urin,asl.n,a,.asl vo.
nepreci.tectuatbytueyearzooo,..mputers.ulcspeaknuently
en.u,ut.ce.eivean avera,einterr.,at.ratleast -oper.ent.rtue
timearterab.utriveminutes.rcial.,ue 1uis.auti.uspr.pue.ymay
ell uavebe..metrue Hutill..mputerseveracvan.et.astateat
ui.utueir..nversati.n,.veral.n,peri.c.rtime,illce.eiveeven
intelli,entinterr.,at.rs:
1.caytuatquesti.nsuarplycivicesmatuemati.iansancartiri.ial
intelli,en.eresear.uers. many Aersbelievetuatas ..mputers,r.
in..mplexityancp.eritis.nlyamatter.rtimeuntiltueybe..me
aare.rtueirexisten.e, ituanintelli,en.etuatmayeven surpass
.urs. matuemati.ians ..nsicer su.upreci.ti.ns u.,asu. 1.tuem
tue..mputeris n.m.re tuan at..lr.r]u,,lin, numbers s.rapicly
tuatitisn.l.n,erne.essaryt.asteu.ursmakin,lar,e.al.ulati.ns
byuanc, itis.nly speec, a..ura.y ancrlexibility tuatcistin,uisua
brainless..mputerrr.mabrainlessaba.us
Onsuncay,may l l , Hm s.uessplayin,..mputerDeepHlue.n
asix,amemat.uinmanuattana,ainstrussia sGarykaspar.v, tue
.rlc.uampi.nbelievect.betue best.uessplayerever. kaspar.v
.n.ne,ame,tiecturee,ancl.stt. 1uerinal,ameasa.rusuin,
This essay frst appeared in The Washington Post (May 25, 1997) .
91
cereat tlastecab.ut u.urancencecarternineteenHluem.ves
kaspar.v sbluncer as a..eptin, Hlue s sa.riri.e .r a kni,ut r.r a
pan ne railec t. see uat Deep Hlueuac seen, tuat by l.sin, a
kni,ut,Hlue.btainecan.veruelmin,p.siti.nalacvanta,e
kaspar.vasasuakenancan,ryl.ser. Arterap.l.,izin,r.ruis
p..rplay, ueimmeciatelylaun.uecint.abitteratta.k.nDeepHlue,
.uallen,in,itt.entert.urnamentsancplayreal.uess ne,uar
anteectuatirtuisuappenecue.ulctearDeepHluet.pie.es
kaspar.v scereatasuailecasamilest.ne, n.t.nlyin.uessuis
t.rybutals.intuepr.,ress.rAnmy.pini.nitasaverymin.r
milest.ne
Lxa.tly uatc...mputers c.: 1uey are minclessma.uinesce
si,nect.manipulatebinaryci,its.nesanczer.sm.celecbyele.
tri.al impulses sit.uec uere anc tuere al.n, ires 1ue simplest
example.rsu.ua cevi.eistueaba.us Onesarem.celecbybeacs,
zer.s by empty spa.es al.n,r.cs sit.ues are pr.vicec byrin,ers
tuatslicetuebeacsa...rcin,t.al,.ritumspr..ecurestuat,ivein
stru.ti.ns t. tue rin,ers mus.les .rtue uanc anc armrurnisutue
ener,y. Or..urse, tuep.er.ranaba.us isseverelylimitecbytue
smallnumber.rr.csancbeacsancbytuel.n,timeittakest..p
eratetuecevi.e.
me.uani.al..mputers arem.re eui.ient,1uey.anbemaceitu
..,ueels, itu]ets.raterrl.in,tur.u,uanet.rk.rtubes, itu
levers ancpulleys, itu balls r.llin, c.n in.linesinceecitu al
m.stanytuin,tuat.anbemanipulatecbyener,y uavea.arcb.arc
cevi.e, printecyearsa,.as an acvertisin,premium, tuat plays un
beatableti.ta.t.ebyusin, a slicin, strip anc ar.tatin, cisk N.t
l.n,a,.a,r.up.r.lever..mputerua.kersbuiltati.ta.t.ema.uine
itu tinker t.ys n prin.iple, a tinkert.y ma.uine .an c. every
tuin,asuper..mputer.anc.pr.vicecitislar,een.u,uanc,iven
en.u,utime
super..mputerscirrerrr.mme.uani.al.al.ulat.rsin.nly.nerun
camental ay Hyusin,ele.tri.ityanctiny sili..nsit.uest.m.ve
.nes anc zer.s tur.u,u ires net.rks it ,ainsin.recible speec r
y.u.alluatitc.estuinkin,, y.umi,ut]ustasellsaytuebeacs
.ranaba.usaretuinkin,uiletueyaccnumbers
Asuper..mputer saes.mespeecenablesitt.ansermatuemat
i.al questi.ns n..ne ..ulc anserbyuanc. N. uuman minc ..ulc
uave.al.ulatec, as..mputersuaveeasilyc.ne, t.milli.ns.rce.
imalci,its 1uera.ttuat..mputerpr.,rams.an play ,rancmaster
12. Ka5parov
'
5 DeFeat b
y
Deep Bl ue 93
.uessisn.m.resurprisin,tuantueirabilityt.multiply,i,anti.num
bersrastertuananyuumanli,utnin,.al.ulat.r DeepHluecereatec
kaspar.vin at.tallyminclessay tn.m.rekneitas playin,
.uesstuanava.uum.leanerkn.sitis.leanin,aru, t.aresn.t
auituetueritins.rl.ses
numan .uess players examine a re ruture m.ves at a rate .r
severalperse..nc, usin,experien.eancintuiti.nt.av.ic..nsicerin,
irrelevantm.ves. DeepHlueexaminesal p.ssibleruturep.siti.nsr.r
l o.r m.re m.vesaueac at arate.rzoomilli.n p.siti.ns a se..nc
tistuis rantasti. speec, ..mbinecitusele.tivityrulesr.rratin,
p.siti.ns, tuat ,ives DeepHlueitsen.rm.usbruter.r.ep.er Anc
itins,ames Asis.rtenp.intec.ut,airplanesrlyrastertuanbircs
butitu.utnappin,tueirin,s
.ryears, .uesspr.,ramsuavecereatec,rancmastersuenm.ves
mustbemacerapicly Anctuepresent.ue.ker.uampi.n.rtue.rlc
isa..mputerpr.,ram cue.kersiss.mu.usimplertuan.uesstuat
in a ce.ace .r t. tue ,ame may be s.lvecapr.,ram ill play a
perre.t,ame
cuess is rar rr.m bein, s.lvec Hut ..mputers uave passec a
1urin, .uesstest. A ,rancmaster .ann.t kn. uetuer uis uiccen
.pp.nent isan.tuer,rancmaster .r a ..mputerpr.,ram Hut tuat
a.uievementisarar.ryrr.mtue..mplexity.ruumanintelli,en.e
c.mplexity uas be..me a buzz.rc, pre.isely cerinec i n ..m
puters.ien.e uil.s.puersuavebr.acenectuetermt. applyt.tue
ev.luti.n .rtue universe arter it expl.cec int. existen.e Altu.u,u
tueuniverseasau.leisin.reasin,inentr.py,cis.rcer) ,tuereare
re,i.nsuereanctuereuerecis.rcer,ivesayt.beautirul.rcerin
tueemer,en.e.reverm.re..mplexsystems 1uer.rmati.n.r,alax
ies, stars, ancplanetsarestrikin,examples Onatleast.neplanet,
lireuasemer,ecancev.lvecintuecire.ti.n.reverin.reasin,..m
plexity, .ulminatin, in tue brains .rsu.u stran,e.reatures as y.u
ancme.
N. tue n.ti.n tuat, as ..mplexity in.reases, ast.nisuin, ne
pr.pertiesemer,eisas.lcastuean.ientGreektuinkers w.ncr.us
pr.perties appearec uen at.msr.rmecn.mquarks anc ele.tr.ns
Lven m.re amazin, pr.perties emer,ecuen at.ms].inect.make
m.le.ules nycr.,en anc .xy,en are simple elements itu simple
pr.perties. ruttuem t.,etuer anc y.u ,etater, a substan.e itu
remarkableattributesunliketu.se.reituerelement,pr.pertiestuat
maybeabs.lutelyessentialr.rtueemer,en.e.rlire.
s.ill..mputers, .rtues.rtekn.u.t.builc, everrivaltue
..mplexity.ruumanintelli,en.e: nl vss, nansm.rave.,u.ueacs
ar.b.ti.slab.rat.ryatcarne,iemell.nuniversity,r.teab..kti
tlecMind Children: The Future of Robot and Human Intelligence in
ui.uuepreci.tectuat..mputers.ulcbesurpassin,uumanmincs
inlesstuanualra.entury similarvies, tu.u,uitumu.ul.n,er
timerrames, uavebeenacvan.ecbyAresear.uersmarvinminsky,
nerbert sim.n, ancD.u,lasn.rstacter ruil.s.puerDanielDennett
is n.tintueleastmystiriec by..ns.i.usness, believin,ueexplainec
itanctuat..mputersill.necayuaveitinuisbest sellin, l vvl
b..kConsciousness Explained.
m.rave.,puysi.istrrank1ipler ancare .tuers a.tuallybelieve
tuat..mputersilleventuallyrencertueuumanra.e.bs.lete 1uey
illbe..me.urminc.uicren,cestinect.take.vertuetak.r..l.
nizin,tue..sm.sasuumanity,.estueay.rtuecin.saurs
m.rave. is tue str.n,est .r uat are .allec str.n, Aers ne
rites, 1.cay.urma.uinesarestillsimple.reatures Hutituin
tue .entury tuey ill mature int. entities as ..mplex as .urselves,
anceventuallyint.s.metuin,trans.encin,everytuin,ekn..
1ue.rcsuavearamiliarrin, nereissamuelHutlerritin,in
uis l szrantasy Lreu.n 1uere isn. se.urity a,ainst tue ulti
matecevel.pment.rme.uani.al..ns.i.usness rerle.tup.ntue ex
tra.rcinary acvan.e ui.uma.uinesuavemacecurin,tue lastre
uuncrecyears, ancn.teu. sl.ly tue animal ancve,etablekin,
c.msareacvan.in, . 1uepresentma.uinesaret.tuerutureastue
earlysaurianst.man
1ue cirreren.e beteen Hutler s remarks anc similar sentiments
byuarcAersintuatHutlerasritin, satire
Opp.sin, tuese ile rantasies is a ,r.up .r tuinkers s.metimes
.allec mysterians, anc ri,utly s., be.ause tuey believe tuat .ur
mincsremainapr.r.uncmystery neretuemysterians, am.n,u.m
..unt myselr, ].in tue matuemati.ians u. .rkitu..mputers
Am.n, tue m.st .utsp.ken mysterians are turee Ameri.an puil.s.
puers, 1.un searle, 1u.mas Na,el, anc c.lin m.Ginn, tue Hritisu
matuemati.al puysi.ist sir r.,er renr.se, anc psy.u.l.,ist steven
rinker renr.se s t. b..ks, The Emperor's New Mind anc its se
quelShadows of the Mind, aretuestr.n,estatta.ksyet.ntuebelier
tuat..mputersills..n.r.ssaturesu.lc.r..mplexitymakin,tuem
aare.ru.tuey are, ablet. reelpleasure ancpain, t..reate anc
lau,u at].kes, t. l.ve anc uate, t. make m.ral ce.isi.ns, t. rite
12. Ka5parov
'
5 DeFeat b
y
Deep Bl ue 95
,reat p.etry, musi. anc n.vels, t.make ne s.ientiri. cis..veries,
anct.mecitate.npuil.s.pui.alanctue.l.,i.alquesti.ns
m.stmysteriansc.n.tbelieveas.ulexistsapartrr.mtuebrain
1uey a..epttuevietuat.urselr, ituits..ns.i.usnessancrree
ill,t.namesr.rtuesametuin,), isarun.ti.n.ramaterialbrain,
a..mputermace.rmeat,asminskylikest.say. 1uey..ntenctuat
.urbrainsares.mu.um.re..mpli.atectuant.cay s..mputerstuat
euave.nlytuecimmest..mpreuensi.n.ru.tuey.perate.
Neur.s.ientistsaremakin,pr.,ress, butasyettueyc.n.teven
kn.u.mem.riesarest.recancretrievec renr.sesimplyinsists
tuattue abilities.rtueuuman mincaren.t ,.in, t. emer,e rr.m
..mputer ..mplexity as l.n, as ..mputers ..nsist .r n.tuin, m.re
tuan ele.tri. .urrents m.vin, tur.u,uiresinamanner ci.tatecby
s.rtare tntil ekn.m.re ab.utu.brainsc.uattuey c., e
illn.tbeablet...nstru.t..mputerstuatill..me.l.set.rivalin,
uumanmincs enr.seeven..ntencstuatn.su.u..mputersillbe
builtuntilekn.m.reab.utlas.rpuysi.sceepertuanquantum
me.uani.s.
DeepHlue s cereat.rkaspar.vinn.ay si,nalstue emer,en.e
am.n,..mputers.ranytuin,raintlyresemblin,uumanintelli,en.e
wuatDeepHluec.esisn.tuin,qualitativelyciuerentn.muatan
.lcrasui.necaccin,ma.uinec.es tmerelyticclesnumbersraster
tuaname.uani.alma.uine reruapss.mecay,irquantum..mputers
areevermace.perati.nal,tueyillbe.ntueirayt.arcs.metuin,
resemblin,uumantu.u,ut
.n.er.tetuatm.rave.surrersrr.muavin,react..mu.us.i
en.eri.ti.n A pr.minent s.ien.e ri.ti.n autu.rt..kme t. taskr.r
tuisremark, insistin, tuat n. .ne .anreact.. mu.us.ien.eri.ti.n
wuat meant, .r..urse, astuatm.rave.t..kt..un.riti.ally tue
st.riesuereacab.uttue..min,.rintelli,entr.b.ts. tisal.n,, l.n,
cistan.en.mtue.ir.uitry.rDeepHluet.tueminc.ram.use
Chqp
ter 1 3
Computers Neq r the Th res hol d ?
1uen.ti.ntuatitis p.ssiblet...nstru.tintelli,entma.uines.ut.r
n.n.r,ani.materialisas.lcasGreekmytu.l.,y vul.an, tuelame
,.c .rrire, rabri.atec y.un, .men .ut.r,.lc t. assistuimin uis
lab.rs. neals.macetuebr.nze,iant1alus, u.,uarcectueislanc
.rcrete by runnin, ar.uncittureetimes a cay anc ueavin, uu,e
r..ksatenemysuips Asin,levein.ri.u.r,tuebl..c.rtue,.cs)ran
rr.m1alus sne.kt.uisueels. neblect.ceatuuenueas.uncec
in tueankle.r, a...rcin,t. an.tuermytu, uen abrasspininuis
ueelasrem.vec
Artertuenaustrialrev.luti.n,ituits.ncerrulma.uinery,rit
ers be,an t. spe.ulate ab.uttue p.ssibility tuatuumans as ell as
,.cs..ulcs.mecaybuilcintelli,entma.uines ncuapters z-, z+,
anc z.ruisn.velErewhon , l sz) , samuel Hutlerr.teab.uttue
..min,.rsu.ur.b.ts 1ik1.k,.ne.rtueearliestme.uani.almenin
ri.ti.n, asaincup..pperpers.nu.maceuisrirstappearan.ein
i rankHaum s Ozma of Oz , l vo) ne as manura.turec inLv, a
lancac] a.entt.Oz, bytuerirm.rsmitu anc1inker A plate .nuis
ba.ksaictuattuer.b.t1uinks, speaks, A.ts, ancD.esLverytuin,
Hutiive.
Artertue..mputerrev.luti.npr.cu.ecele.tr.ni..al.ulatin,ma
.uines, itu tueir .uri.usresemblan.e t. tue ele.tri.al net.rks .r
uumanbrain, tuep.ssibility.r..nstru.tin, intelli,entr.b.tsbe,an
t.betakenseri.usly,espe.iallybyleacers.rA,artiri.ialintelli,en.e)
resear.u, ancby arerell.travelin,puil.s.puers. nansm.rave.
cire.tsar.b.tlab.rat.ryatcarni,iemell.ntniversity, inuisb..k
Mind Children , l vss), uepreci.teatueappearan.e.rr.b.tsituuu
manintelli,en.eber.re tue enc.rtue next rirty years H.tuue anc
rank1ipler, a1ulanetniversitypuysi.ist,are ..nvin.ectuat..m
This essay frst appeared i n Mysteres of Life and The Univere. ( 1 992) , edited by
William H. Shore.
97
putersills..n exceed uumanintelli,en.e, makin, tueuumanra.e
superrlu.us c.mputersilltuentake.vertueburcenancacventure
.r..l.nizin,tueuniverse
nereisapassa,err.mErewhon tuat..ulcuavebeenrittense
ri.uslyby1ipler.rm.rave.
1uereisn.se.uritya,ainsttueultimatecevel.pment.r
me.uani.al..ns.i.usness, intuera.t.rma.uinesp.ssess
in,little..ns.i.usnessn.. Am.llus.uasn.tmu.u..n
s.i.usness. rerle.tup.ntueextra.rcinaryacvan.eui.u
ma.uines uave mace curin, tue last re uuncrec years,
ancn.teu. sl.ly tue animal ancve,etable kin,c.ms
are acvan.in,. 1ue m.re ui,uly .r,anizec ma.uines are
.reatures n.t s. mu.u .r yestercay, as .r tue last rive
minutes, s. t. speak, in ..mparis.n itu past time As
sumer.rtuesake.rar,umenttuat..ns.i.usbein,suave
existecr.rs.metentymilli.nyears seeuatstricesma
.uinesuavemaceintuelasttu.usanc mayn.ttue.rlc
lasttentymilli.nyearsl.n,er: rs., uatilltueyn.t
intueencbe..me:
An.tuerpassa,e, remarkablypr.pueti.,rr.mtuesameb..k
D.n.tletmebemisuncerst..caslivin,inrear.rany
a.tuallyexistin,ma.uine, tuereispr.bablyn.kn.nma
.uineui.uism.retuanapr.t.type.rrutureme.uani.al
lire 1uepresentma.uinesare t. tue ruture astue early
saurianst.man. 1uelar,est.rtuemillpr.bably,reatly
ciminisuinsize
A.tually, tuese are n.t tue narrat.r s .rcs but senten.es tuat
Hutlerattributes, t.n,uerirmlyin.ueek,t.anLreu.nianpr.ress.r
1uepr.ress.r s.pini.nspr.mpttueLreu.nianst.cestr.yalltueir
ma.uinesber.re tuey surpassuuman intelli,en.e anc take .vertue
.rlc.
nMind Children m.rave.putsittuisay
1.cay.urma.uinesarestillsimple.reati.ns, requirin,
tueparental.are ancu.verin, attenti.n .rany neb.rn,
uarcly .rtuy .rtue.rc intelligent. Hutituin tuenext
13. Computers Ne'r the Thres hol d!
.entury tuey ill mature int. entities as..mplex as.ur
selves, anceventuallyint. s.metuin, trans.encin,every
tuin,ekn.inu.me.antakepriceuentueyrerer
t.tuemselvesas.urces.encants
99
1uem.stp.errulatta.k.nsu.u.pini.ns, ui.uuave..met.be
.allecsn.n,A, isr.,errenr.se sbest seller The Emperor's Mind
, l vsv) Naturally, tueb..kasvi,.r.uslylambastecbystr.n,Aers
He.auser.tetueb..k sr.re.rc,t..uavebeencen.un.ecr.rmy
.btuseness 1uisessayisaneu.rtt.setc.ninm.recetailpre.isely
uatbelieveab.uttuep.ssibilitytuat..mputers ill s..nbeable
t...nverseituusinaysincistin,uisuablerr.mtue..nversati.ns
.ruumanbein,s
irst,su.ulcmake.leartuatamn.tavitalistu.tuinkstuere
is a,u.stintuema.uinea s.ul cistin.tn.mtuebrain believe
tuattueuumanminc, liketueminc.ranyl.eranimal,isarun.ti.n
.r a material lump .r .r,ani. matter Altu.u,u remain .pen t.
tue rlat.ni. p.ssibility .ra cisemb.ciec s.ul, as am .pen t. any
metapuysi.al n.ti.n n.t l.,i.ally..ntraci.t.ry, tueevicen.e a,ainst
itseems.veruelmin,. str.n,ar,umentsr.rarun.ti.nalvie.rtue
mincaret..ramiliart.neec summarizin,.
rauumanuasan.nmaterials.ul,itisuarct.seeuytuesame
su.ulc n.tbe saic .ran am.eba, a plant, .r even a pebble A re
panpsy.ui.m.nistssu.uascuarlesnartsu.rnea.tuallyc.saytuis,
but..nsiceritanabsurcmisuse.r.rcs, a.ate,.rymistake, t.
talk.rap.tat.inacark.ellarasuavin,uatHutler.alleca.ertain
ce,ree.r.unnin, nerearet.panpsy.ui.qu.tati.nsrr.mHutler s
ima,inaryLreu.nianpr.ress.r
Hutu..ansaytuattuevap.ren,ineuasn.takinc.r
..ns.i.usness: wuerec.es..ns.i.usnessbe,in,ancuere
enc: wu..an cra tue line: wu..an cra any line: s
n.teverytuin,inter.venitueverytuin,:sn.tma.uin
erylinkecituanimallireinaninrinitevariety.rays:
suallesaytuattueplantc.esn.tkn.uatitisc.
in,merelybe.auseituasn.eyes, .rears, .rbrains: re
saytuatita.tsme.uani.ally, ancme.uani.ally.nly, suall
en.tber.r.ect.acmittuatsuncry.tuerancapparently
very celiberate a.ti.nsare als.me.uani.al: rit seemt.
us tuattueplantkillsanceatsanyme.uani.ally, mayit
n.tseemst.tueplanttuatamanmustkillanceatsueep
me.uani.ally:
100 A G<rdnd5 Workout
a,reeituArist.tletuattueselristuer.rm.rtueb.cy, .rin
m.cerntermin.l.,y, apattern .rtuem.le.ular stru.ture.r.r,ani.
matter insice .ur skull Or..urse tue pattern is rarm.re ..mplex
tuantuepattern.ravase.rtueLmpirestateHuilcin,
iire cicinceecev.lve al.n, ..ntinua,buttuere are sp.ts ,albeit
ituruzzy ec,es) uere ice .uasms ere.r.ssecancnepr.per
ties.rmatteremer,ec 1uerirst,reatturesu.lcastueemer,en.e.r
lirerr.mlireless..mp.uncs Anctuelast.rtue,reatturesu.lcs, tue
,reatest.rtuemall, as tueev.luti.n.rtuebrain itu su.upr.per
tiesas..ns.i.usness, selraareness), expancecrreeill,ituallits
m.ralimpli.ati.ns), asense.rri,utancr.n,, asense.ruum.r,tue
p.ert...mmuni.ate..mpli.ateciceasbyspee.uancritin,, anca
rart.r.reativeskills su.u astueabilitiest...mp.sep.etry ancmu
si., paint pi.tures, cis..ver si,niri.ant matuemati.al tue.rems, anc
invent s.ientiri. tue.ries.apable.rbein,testec. witutuelastskill
.ameaes.mep.ert...ntr.ltuepr..ess.rev.luti.n anc steerit
innecire.ti.ns, asellastuep.ert.terminatetuepr..ess.
t,.esitu.ut sayin, tuat many.rtueseuuman traits are p.s
sessec t. a eak ce,reeby animals cuimpanzees seem t. uave a
l.level aareness.rtuemselves anc an ability t. make ce.isi.ns.
(Free will anc selfawareness, bytue ay, are r.r met.names r.r
tue samepuen.men.n, likeLinstein sprin.iple.rtue equivalen.e.r
,ravity anc inertia, .r, uat am.unts t. tue same tuin,, tue equiv
alen.e .r,ravitati.nal ancinertial mass .ann.t ima,ine myselr
uavin,rreeillitu.utbein, selraare,n.r.an..n.eive.rbein,
selraareitu.uts.mece,ree.rrreeill) m.nkeysuave areeble
sense .ruum.r, y.u .an at.u tuem play pranks .n ea.u .tuer in
a .a,e H.er bircs may uave a cim sense.rvisual beauty Apes
.an ..mmuni.ateituuumans by si,ns, anc s..an a c., .r.at A
.uimp.anmakeanctest..n] e.turesab.utu.t.,etabananarr.m
tue.eilin,irtuere areb.xeslyin,ar.unc. Anc s..n 1uatanimals
reelem.ti.ns.rl.ve ancpain isunceniable. tis equallyunceniable
tuatama].r,ulr.rs.mes.rtas.r.ssecuenuumansev.lvecrr.m
bestialan.est.rs.
1uequesti.nisn.tuetuer.uruumantraitsemer,ecasarun.
ti.n .ran ev.lvin, brain, as assume tuey cic, but uetuerit ill
be easy .r ciui.ult, peruaps even imp.ssible, t. builc a .al.ulatin,
ma.uine ..mplex en.u,u t. leap tue same turesu.lc. At tuis p.int
et.u.utue.entraltueme.rr.,errenr.se sbrilliant,..ntr.versial
b..k The Emperor's New Mind
13. Computers Near the Thres hol dl 1 01
1ueciui.ultyin.r.ssin,tueturesu.lc, renr.sear,ues, istuate
c.n tyetkn.en.u,uab.utmattert.kn.u.t.c.it clearlye
kn. rar m.re ab.utparti.lestuatcicDem..ritus, but e are still
al.n,ayrr.muncerstancin,tu.separti.les n stancactue.ries,
matterismace.rlept.nsancquarks, anctueseparti.lesaretakent.
be,e.metri.alp.ints, .ratleastpointlike, aspuysi.istsprerert.say
nre.ent superstrin, tue.riestuey are n.tp.ints but in..n.eivably
tiny l..ps n eituer .ase, Net.n suarclittlepebblestuatb.un.e
a,ainst.nean.tuertuekinc.rmatterHisu.pHerkeleyrici.ulecas
astupic, tu.u,utlesss.meuatuaven.t.tally ciss.lvec. wuat
islertismatuemati.s iept.nsancquarks, uetuerp.ints.rl..ps,
aren.tmace .ranytuin, 1ueir rielcs are]ust as ,u.stly On tue
quantumlevel,t.putitbluntly, tuereisn.tuin,ex.eptmatuemati.al
patterns
myreacerskn.u.impatientamitus.mepra,matists, pue
n.men.l.,ists, anc sub]e.tive icealists .rvari.us s.u..ls u. ueap
s..rn.ntuen.ti.ntuatmatuemati.alstru.turesare.uttuereitu
arealitytuatisn.tminccepencent r.rtuesetuinkers, matuemati
.alrealityis l..atecituinuumanexperien.e iikerenr.seanctue
.veruelmin,ma].rity.reminentmatuemati.ianspastancpresent,
am a rlat.nist in tue sensetuat believe matuemati.al patterns
are cis..verec, n.tinventec Or..urse, tuey are stillinventec, ina
sense. Lverytuin, uumans c. anc say is uat uumans c. anc say
matuemati.s .bvi.usly ispart.ruuman .ulture, butt.say s.ist.
say s.metuin,utterlytrivial. 1uera.ttuat .nly uumans.antalk.r
rite ab.ut matuemati.s anc las .rpuysi.s c.es n.t mean tuat it
isuserult.cenytuatmatuemati.sanclas.rpuysi.sareembeccec
inanen.rm.usuniversen.tmacebyus, but.rui.ueareapart,
ancanin..n.eivablytinypartattuat
Asuavesaicber.re, irt.cin.saursmett..tuersin ar.rest
.learin,, tuere .ulcuave been r.ur cin.saurs tuereeven tu.u,u
tuebeastseret..stupict...untanctuereeren.uumansar.unc
t.at.u believetuatalar,einte,erisprimeber.rematuemati.ians
pr.ve it prime believe tuat tue Ancr.meca ,alaxy uac a spiral
stru.tureber.reuumansar.se.nLartut..allitspiral Astuen.tec
Helliabsmatuemati.ianr.nalcGrauamre.entlyputit,matuemati.s
isn.t.nlyreal, itistueonly reality
s.me e..entri. puil.s.puers prerer t. tuink tuat uuman mincs
al.nearereallyreal. 1uereareevenpuysi.ists, .veruelmecbytue
s.lipsisti. tin,es .rQm ,quantum me.uani.s), u. like t. talk tue
102 A Gardner's Workout
sameay Huttueuumanmincismace.rm.le.ules, ui.uarein
turnmace.rat.ms, ui.uareinturnmace.rele.tr.ns, pr.t.ns, anc
neutr.ns 1uepr.t.nsanc neutr.ns are mace.rquarks wuatare
quarksancele.tr.nsmace.r: N.tuin,ex.eptequati.ns iet sra.eit
Y.uanc, attuel.estkn.nlevel.r.urmaterialb.cies, aremace
.rmatuemati.s, purematuemati.s, matuemati.s un..ntaminatecby
anytuin,else
1uem.stele,anttue.ry.rmattert.cay is.r..urse Qm tnr.r
tunately, it is ricclec itu mysteri.us parac.xes n re.ent years
Linstein sLrrparac.x,namecitutueinitials.rLinsteinanct.
.ruis..llea,ues)uasbeentuem.stcebatec n..antuemeasure
ment.r.neparti.le.ausetueemer,en.e.rapr.perty.na..elatec
parti.letuat.anbemilli.ns.rli,utyearsaay:
t seemst. uappen eituer instantane.usly.r itu a speecraster
tuanli,ut.antravelbeteentueparti.les. ntuerirst.ase, tuepara
c.xvi.latestuec.,matuatpr.uibitsinstanta.ti.natacistan.e n
tue se..nc.ase, it seemst.vi.laterelativity, ui.upr.uibits inr.r
mati.nn.mtravelin,rastertuanli,ut
N.ne.rtuemanypr.p.secres.luti.nsissatisra.t.ry. 1uemany
.rlcsinterpretati.n.rQmseemst.,etric.rtueparac.x,buttuere
isanen.rm.uspri.et.bepaic Onemustp.sitbilli.nsup.nbilli.ns
.reverpr.lireratin, paralleluniversesin ui.ueverytuin, tuat.an
uappen,c.es Otuererr.rtst.s.lvetueLrrparac.xc.n.m.retuan
restateitin a ciuerent lan,ua,e t is n. ,..c, r.r example, t. say
tuattuet...elatecparti.lesarepart.rasin,lequantumsystem
u.se ave run.ti.n ,.r state ve.t.r in an.tuerlan,ua,e) ..llapses
allat.n.e, s.naturallyueny.u..llapseitbymeasurin,.neparti
.le y.u.btaininr.rmati.nab.uttue.tuer 1uis simplyrestatestue
r.rmalism Y.u n. uave t. explain u. t. parti.les, li,utyears
apart,.anremain..rrelatec.
.ryearsrenr.seuasmaintainec, al.n, ituDavicH.um,raul
Dira.,Lrin s.ur.cin,er, anc.tuer,reatpuysi.ists, tuatQmisn.t
tueultimatetue.ry.rrielcs anc parti.les 1uis as Linstein s.n
vie nceec, it asLinstein u. rirstpr.p.sectuen.t.ri.us Lrr
parac.xinaneu.rtt.su.tuatQmasin..mplete w.rkin,puysi
.ists,r.rtuem.stpart,never.rryab.utsu.utuin,s Asl.n,asQm
.rks, anc .r..urseit.rks ma,niri.ently, tuey simply a..epttue
ra.ttuat ,asri.uarceynmanlikect.say)Qmis.razy. D.n ttry
t. uncerstancu.it.rks, reynman arnecuis stucents, be.ause
n.b.cykn.su.it.rks. su.ulcpuysi.istsleaveit attuat: N.,
13. Computers Near the Thres hol dl 1 03
insistH.um ancrenr.se, be.ausetuat tencst.cis..ura,eresear.u
tuatmays.mecayrinctuatQm,likeNet.n s,ravity,is.nlya,..c
appr.ximati.n .r a ceeper tue.ry enr.se uimselr is tryin, t. ,.
ceeper, itua,e.metri.altue.ry.rparti.lesancrielcsab.utui.u
amn.t..mpetentt.uavean.pini.n
renr.se ..ntencs, anc a,ree, tuat until e kn. m.re ab.ut
matter.nalevelbey.nc Qm, eilln.tuncerstancu..urmincs
.an be a run.ti.n .r.ur,raymatter until ekn. tu.se ceeper
las, eilln.teven..me.l.set...nstru.tin, ama.uinetuat.an
c.everytuin,.urmincs.anc.
nrenr.se s.pini.n,tue,reatmistakebeuinctue.ptimisti.pre
ci.ti.ns.rstr.n,Aersistueassumpti.ntuatma.uinesmace.rires
ancsit.ues, .peratin,itual,.ritumi.s.rtare,.an.r.sstue,reat
turesu.lc. iet sl..kattuisassumpti.nm.re.l.sely. wekn.rr.m
tue .rk.rAlan1urin, anc .tuers tuat it is p.ssibleinprin.iplet.
builc..mputers.ut.rany kinc .requipmenttuattransmitsener,y
al.n, .uannels, itu sit.ues t. tell tue ener,y uere t. ,. Y.u
.anbuilc..mputersitunet.rks.rpipestuatu.lcarl.in,liquic
Y.u.anbuilctuemitur.tatin,,ears, itustrin,ancpulleys, itu
littleballstuatr.llc.nin.lines.rsliceal.n,iresas.nanaba.us
me.uani.alcevi.es.rtueses.rts uavebeen..nstru.teci ntuepast
ry.uareinterestec,y.u.anreacab.uttueminmyLogic Machines
and Diagrams.
Lveryma.uine, tuepuil.s.puermatuemati.iancuarlesreir.e.n.e
.bservec, isal.,i.ma.uineinuavin,aspe.tstuatm.cell.,i.run.
ti.ns 1ueblaces.rane,,beaterr.tatein.necire.ti.niranc.nly
iry.uturntueuancle.l..kise 1uey,.tue.tuerayiranc.nlyir
y.uturntue uancle..unter.l..kise An.lcme.uani.altyperiter
isa]un,le.rbinaryl.,i.relati.ns ,reir.e, in.icentally, astuerirst
t. su.u. asimplel.,i.ma.uineuanclin,binaryrun.ti.ns..ulc
be built itu ele.tri.al .urrents anc sit.ues ) A re years a,. a
,r.up.r..mputerua.kers..nstru.tecama.uine.ut.r1inker 1.ys
tuatplayecperre.tti.ta.t.e 1uereisn.reas.nuy,intue.ry, .ne
..ulcn.tbuilca1inker1.y..mputertuat..ulcc.everytuin,acray
..mputer.anc..r,inceec,uatanysuper.al.ulatin,ma.uine.rtue
ruture ..ulc c. Or..urse, it .ulc uave t. be m.nstr.usly lar,e
ancint.lerablysl. w.ulcitsslu,,isunessciluteits..ns.i.usness:
c.ulcitstillritea,reatn.vel,pr.vicecituacaretu.usancyears
t.c.it:
N.,n..neinuisri,utminc.ulcsaytuata1inker1.yti.ta.
t.ema.uinekn.sitisplayin,ti.ta.t.eanym.retuanava.uum
104 A G<rd ner'5 Workout
.leanerkn.sitis.leanin,aru,.r a lanm.erkn.sitis.ut
tin, ,rass s.puisti.atec ..mputer pr.,ramstuatn. play :aster
.uessciuern.mti.ta.t.epr.,rams.nlyintue..mplexity.rtueir
al,.ritums A..mputer itu su.u apr.,ramis n.m.re aareitis
playin,.uesstuanane,,beaterisaareitisbeatin,e,,s
str.n, Aers believe tuat as..mputers .rtue s.rt epresently
kn. u. t. ..nstru.t keep ,r.in, in tue ..mplexity .rtueir.ir
.uitry anc s.rtare, tuey ill eventually .r.ss a turesu.lc anc be
..me..ns.i.us.ruattueyarec.in, r.nebelievestuis, isn.t.n.e
r.r.ect.saytuata1inker1.yma.uine.r..mparable..mplexity,.r
even .ne mace itur.llin,marbles, ill.r.sstuesameturesu.lc:
acmittuatalltuismays.mecaybep.ssible,buta,reeituren
r.seanc su.u.pp.nents.rstr.n,Aastuepuil.s.puer1.unsearle
tuatitseems extremelyunlikely wekn.verylittleab.utu.tue
brain.rarisu.r abirc.rks wec.n.teven kn.u.mem.ries
are st.recintueminc.ran ant tistruetuatele.tri.alpulsesare
silently suirtec ab.ut insice tue skulls .ranimals, but tuis is c.ne
in amannerrar rr.muncerst..c wuatrenr.seistellin,usistuat
ev.luti.n, .rkin,.n..mputersmace.rmeat,.r.ssecaturesu.lcin
a ay tuat inv.lveslas .rpuysi.s n.tyetkn.n. t..ulcbetuat
ir anc uen tu.se ceeper las are cis..verec anc ekn. exa.tly
u..urbrainc.esuatitc.es, eillbeablet...nstru.tarepli.a
,peruapsmace.rn.n.r,ani.matter,peruapsrequirin,.r,ani.m.le
.ules) tuatillsimulateauumanminc Hutt.expe.ta.al.ulatin,
ma.uinemaceitu..mp.nents.rtues.rtn.inuse.rima,inect.
.r.sstueturesu.lcseemsu.pelesslyunarrantec
wuatc.esa..mputerc.: tticcles symb.lssymb.lstuatare
meanin,lessuntil e atta.u meanin,st. tuem tticcles tuem in
blinc.becien.et. synta.ti.alrulespr.vicecbytues.rtare Hut.ur
mincsc.m.retuanticclesymb.ls 1uey als. ticclemeanin,s.r
symb.ls. .an easilyima,ine am.nstr.usma.uine mace .r1inker
1.ys tuat.anplay Granc:aster.uess, but.ann.t ima,ine itill
kn.itis playin, .uess Hy tue enc.rtuis.entury expe.t a.uess
pr.,ramt.beablet.cereat any ,rancmaster uileplayin, uncer
tueusualtimerestraints Lvenn..uesspr.,rams.an.rusu,ranc
mastersuenm.vesmustbemaceituinarese..ncs expe.ttuat
p.errul..mputersillsteacilyimpr.veintueirabilityt.c.alls.rts
.rextra.rcinary tuin,s, buttuesetuin,sill all bec.neby symb.l
ticclin, c.n.tbelievetuattue..mplexity.rtueir.ir.uitry ill
pusutuema.r.sstuema,i.turesu.lc.
13. Computers Ne'r the Thres hol dl 105
Lncless n.vels, st.ries, plays, anc even .peras uave been rit
ten ab.ut intelli,entr.b.ts, anc itisn. a..icent, suspe.t, tuat s.
many str.n, Aers ere s.ien.eri.ti.nburrsin tueir y.utu my ra
v.rite n.vel ab.ut r.b.ti.s is a littlekn.n .ne by i.rc Dunsany,
inexpli.ably never publisuec in tueunitec states 1itlec The Last
Revolution , l vl ) , it..n.erns arebelli.na,ainstuumanity.rsuper
intelli,ent,selrrepr.cu.in,ma.uines, tues.rt.rrebelli.ntuat1ipler
ancm.rave.believep.ssible. 1ueb..k srunniests.ene...ursuen
tuenarrat.rplays.uessitutuerirstpr.t.type tsinvent.r,Ablarc
rencer,livesituuisauntmary nepretencst.incupuis,ac,et
ituakeys.asn.tt.rri,utenuerbylettin,uerkn. itis alive
trequires .nly arem.vesr.r tuenarrat.r, u.seruyiepez
.penin,qui.klytakesabizarreturnn.tinany.uessmanual,t.real
izeueisplayin,n.t.nlya,ainstanintelli,en.esuperi.rt.uis.n
buta,ainstamincaare.ruatitisc.in,.
wuenrencer s,irlrrienc, Ali.ia,rirstseestue.rablike,r.urle,,ec
m.nsterancitseyeslikea...kr.a.u s, tuereisanintuitiverlasu.n
uerra.e liker.rkecli,utin, sue sensesimmeciatelytuattuetuin,
isalive Ac.,,rri,utenecbys.metuin,itt..kn.sislivin,buttuat
uas n. smell, u.ls anc bites tue ir.n 1ue tuin, tears tue c., t.
pie.es
1uebrain .rrencer s r.b.t ..nsists .rrine ire tuat transmits
ele.tri.alpulses. Dicy.umakeitentirelyy.urselr: Ali.iaasks.
Yes, .r..urse, rencerreplies D.n ty.ulikeit:
1ime, saysAli.ia,illuavet.su.tuat.
Postsci pt
As tuis essay makes .lear, am am.n, tue ranks .rtu.se u. are
.allec mysterians wec. n.t ceny tuat tue minc emer,esrr.m a
..mplex pattern .rm.le.ules, but e believe tuat tue brain s ..m
plexityiss.vasttuatatpresentesimplyc.n.tuncerstancu.it
pr.cu.esselraarenessancrreeill weals.believetuat..mputers,
.peratin,ituiresancsit.uestuatsuirtele.ni.al.urrentsar.unc
a net.rk in .becien.e t. s.nare al,.ritums, ill never .r.ss tue
turesu.lcatui.uanytuin,resemblin,tueuumanbrainillemer,e
sin.e ritin, tuis essay .ame a.r.ssaninterestin, l szletter
rr.msamuelHutlert.cuarlesDarin,reprintecinDain saut.bi
.,rapuy nitHutlerexplainstuattue se.ti.ninErewhon, uereue
106 A G<rdner'5 Workout
su,,eststuatma.uinesills..nmakeuumans.bs.lete,asritten
in]est nereisuatHutleruast.say
ventureup.ntueliberty.rritin,t.y.uab.utap.r
ti.n.rtuelittleb..kErewhon ui.uuavelatelypublisuec
ancui.uamanaicuasbeena,..ccealmisuncerst..c
rerert.tue.uapter.nma.uinesin ui.uuavecevel
.pecanc.rkec.uttue.bvi.uslyabsurctue.rytuattuey
areab.utt.supplanttueuumanra.eancbecevel.pecint.
aui,uerkinc.rlire.
wuenrirst,.tu.lc.rtueiceacevel.pecitr.rmere
run, anc be.ause it amusec me anc tu.u,ut it .ulc
amuse .tuers, butitu.ut a parti.le .rseri.us meanin,,
but cevel.pec i tanc intr.cu.ec ,it) int. Erewhon itu
tue intenti.n .rimplyin, seeu. easy it is t. be plau
sible, anc uat absurcpr.p.siti.ns .an becerencecby a
littlein,enuity anccist.rti.n anc ceparturerr.m stri.tly
s.ientiri.metu.cs, ancuacHutler sAnalogy inmyueac
astueb..kat ui.uit su.ulc beaimec, butprererrect.
..n.eal my aim r.r many reas.ns. irstly tue b..k as
alreacy as ueavily ei,utec ituueter.c.xy as it .ulc
bear, anccaren.t,ivean.tuerualr.un.elestitsu.ulc
breaktue.amel sba.k, se..nclyit.ulcuaveinterrerec
itutueplausibility.rtuear,ument, ancl..kect.tuis
plausibility asavaluableaict.tue,enerala..eptati.n.r
tueb..k,tuirclyitism.reamusin,itu.utanys.rt.rex
planati.n,anctu.u,uttuecrierparttuatuac,.neber.re
antecalittlerelievin,, als. tue m.re eni,mati. a tuin,
.rtuis s.rtis, tuem.repe.pletuinkr.r tuemselvesab.ut
it, .n tueprin.iple tuat acvertisers askwuereisLliza:
ancwu. sGriuitus: tuerer.re tu.u,utitunne.essary
t.,iveanycis.laimer.ranintenti.n.rbein,cisrespe.trul
t. tue Orgin of Species a b..kr.r ui.u .an never be
suui.iently ,raterul, tu.u,u am ell aare u. utterly
in.apableam.rr.rmin,any.pini.n.nas.ientiri.sub]e.t
ui.uis.rtuam.ment s..nsicerati.n
ni rankHaum s The Scarecrow of Oz, tuereisas.eneinui.u
a,rassu.pperrests.ntues.are.r. sseverecueac Arey.ualive:
tue,rassu.pperasks.
13. Computers Ne'r the Thres hol dl
1uat is a questi.n uave never been ablet.ce.ice,
saic tue s.are.r. s ueac wuen my b.cy is pr.perly
stuuec uave animati.n anc.anm.ve ar.unc as ell as
any live pers.n 1ue brains in tue ueac y.u are n. ..
.upyin, asatur.ne,are.rvery superi.rquality ancc.a
l.t.rvery.levertuinkin, Hutuetuertuatisbein,alive,
.rn.t,.ann.tpr.vet.y.u,r.r .neu.livesisliablet.
ceatu, uileam.nly liablet.cestru.ti.n
107
Chqpter 1 4
Corneri ng the J(i ng
n a l vvz issue.rmi.uael L.ker s REC Newsletter p.sec tue r.l
l.in, .uess pr.blem. 1ue queen is at tue l.er ri,ut ..rner .r a
re.tan,ularrielc, anctuekin,is attueupperlert..rner Onsquare
re.tan,lestuekin,isin.ue.k,s.uem.vesrirst. Otuerisetuequeen
,.esrirst. rlayersalternatem.ves. Onuatrielcs.antuequeenr.r.e
tuekin,int.tueupperlert..rner: 1uepr.blem,eneralizesapuzzle
by marek enszk., .rr.lanc, ui.uappearec in Games ma,azine,
tuecate.rui.uuavel.st
1ue kin, .learly .an t be ..rnerec .n 2 x n rielcs .r .n a 3 x 3
square ne can be ..rnerec.n.ertain 3 x n rielcs uen n > 3. 1ue
smallestb.arc .rinterest,tuerer.re, istue3 x 1b.arc
Assumin,tuekin,alaysmakesuisbestm.ves, uat stuemini
mumnumber.rqueenm.vesrequirect...rnertuekin,.ntue3 x 1:
t s n.taseasyt. s.lve asit seems Ancuat s tue st.ry.n lar,er
b.arcs:
..n]e.turecinREC Newsletter tuattuekin,..ulcn.tbe..rnerec
.nanysquareb.arc,tu.u,uuacn.pr..r als...n]e.turectuattue
queen..ulcin.nallre.tan,ularn x m b.arcs, n > 2 ancm > + l .
Ancyiiu,tuea.epr.blem.reat.r anc s.lverattueuniversity.r
Alberta, canaca, as intri,uecbytue pr.blem ne anc suarec a
byline.ntuer.ll.in,arti.le, ritten byiiu
A Royq l Ptobl em: And Al i ce I s (q ug ht i n the Mi dd l e
1uerecQueenasruri.us, asusual ner.urrentireasbr.u,ut.n
bytue absen.e.rtue reckin, rr.m uis rala.e Onuerrarevisits,
sueexpe.tect. see u.m sueuac..met.see
This article frst appeared in Quatum (July/August 1993) .
109
Hrintue.lcr..lba.kuere,.relse r.arectuenecQueen,wu.
asrelatect.tueQueen.rnearts.
Oreiseuat:askecAli.e, but.niyarterner:a]estyuacswept
raciantiy.ut.rearsu.tba.kt. aersice.rtuerala.e.
Ouwituy.urueac 1weeciecumsaic.
wuatelse:accec1weecleceeruet.ri.aily.
Ou,cear, saicAii.e,tuisputs aaewmeanint.tenper.ent.u
tuet.p. wuat suall c.: c.nteven kn.w wuere tnerec kin, .s.
1uetinsbr.u,ut.utamap.rtueianc. twastueramiliar8 x 8
.uessb.arciniure l .
betkn.wwuerenis:a]estyis, saic1weeciecum.
Onu ex.iaimec1weeciecee.
n.w c.y.uknewtuat:Ali.easkec.
weil, saic1weeciecum, tue rec kin piays it sare. ne never
ventures.ut.ruiskin,o.mint.tueH.rcerianc.
ne ais. reruses t..r.ss .ver t. tue Queen sice, accec 1wee
clecee.
8
7
6
5
4
3
2
Queen Side
King Side
!I
r
-
, .
..
''
a b c d e f g
Fi gure 1 .
Red Kingdom
Bmdedand
Wbite Kingdom
b
Palace
14. Comer i n9 the Ki n9 111
s.ueis..nrinec t.telve squares 1uat is uelprul, but1 still
c.n tseeu.y.u.anbes.suretuatueis.nus
nisma]esty likes t.be as rar aay rr.m tue recQueen as ue
p.ssibly.an, 1eeclecumsaic
A.tually, as rar rr.m tue rec Queen s rala.e as p.ssible, ..r
re.tec1eeclecee. ne uas n...ntr.l .ver tue uereab.uts.rner
ma]esty.
1uereisan.tuerpr.blem, saicAli.e. rtue reckin, c.esn.t
antt...meba.kt.es, u..an1 persuaceuima,ainstuisisues:
1uetinstu.u,utr.rauile,ancr.u,utr.r auile]ustt.pass
tuetime 1uentueyb.tu.ameupituabrillianticea N.tsurpris
in,ly,itastuesameicea
Arey.u in m.rtal rea.rtue rec Queen: 1eeclecum askec
Ali.e.
Or..urse wu.isn t:
Orallpe.ple,u.rearsuertuem.st:askec1eeclecee
narct. say, Ali.e repliec 1uen it ...urrec t.uer 1ue rec
kin,, .r..urse
ri,ut saic1eeclecum ne..ulcn.trisk,ettin,.au,utina
matin, situati.n itutuewuite Queen
s.iry.u cis,uisey.urselras tuat ,..c lacy, y.u .an crive nis
ma]estyba.kuere, ce.larec1eecleceetriumpuantly.
tis.rtuatry, saicAli.e, s.meuaten..ura,ec su.ulcn.t
asteanytimebyventurin,.utsice.rtu.setelvesquareseituer
makesurey.uc.n t..rnernisma]esty.nus, 1eeclecumac
visecAli.e.
Als., c.n.t crive uimint. tue H.rcerlanc, saic 1eeclecee
nisma]estymayrinc.uttuat it isn.tas can,er.us as ue makes
it.utt.be
well, cbetteruurryancbrin,nisma] estyba.kass..nas1 .an
1uerecQueen spatien.eis su.rtertuanuertemper
Probl em 1.
On te miniature chessboard in Figure 2, White has a lone Queen on e8,
and Red has a lone King on h6. lite moves frst, and wins if the Red
King is drven back to e8 within 1 0 moves. If this is not accomplished,
then Red wins. Other than what is noted above, normal chess rules
apply. With perfect play, which royalty wins ?
8
V
7
6
M
e
f g
h
Figure 2.
Ali.e as able t. a...mplisu uer missi.n, .nly t. uave tue rec
kin, slip .ut a,ain numpty Dumpty, in uis l.rty p.siti.n .n tue
all, sp.ttecnisma]esty.nu+tuistime
Ali.e..rre.tlycecu.ectuattuereckin,stilluarb.recn.tu.u,ut
.r.r.ssin, .ver t. tue Queen sice. wuile ue uactemp.rarily ..n
querecuisrear .rtueH.rcerlanc, ue asn.tyetillin,t. venture
int.tuewuitekin,c.m.
navin,l.stmu.utimeina...mplisuin,uerrirstmissi.n,Ali.eset
.utimmeciatelyt.reena.ttuecrama,but.nanenlar,ecsta,e
Problem 2.
On te miniature chessboard in Figure 3, lite has a lone Queen on e8
and Red has a lone King on h4. White moves frst, and wins if te Red
King is driven back to e8 witin 14 moves. If this is not accomplished,
then Red wins. Oter tan what is noted above, noral chess rules
apply. Wit perect play, which royalty wins ?
8
7
6
5
4
V
&
e f g
h
Figure 3.
Ali.ecr.vetuereckin,ba.kt.uisala.e]ustintime
c.meal.n,, r.arectuerecQueen weuavet.attencasummit
..nreren.eitutuewuiteQueen ancuer..ns.rt
wuatistuemattertuistime,cear:askectuereckin,timicly
we uave beencis.ussin, tue patiti.n .rtueH.rcerlanc 1uere
ist..mu.u,.in,s.nuere, espe.ially.nu+,.r s.1 uear.
.an tima,ineuat, murmurectuereckin,
Anyay,tuewuiteQueenanc1 uavea,reect.establisu.urb.r
cersbeteenranks+anc we]ustmeett.r.rmalizetueceal
ry.usays., cear.
Ass..nastuenetreatyassi,nec,tuereckin,ueacecr.ru,
tuerurtuestuavenituinuisc.main Ali.eascispat.uecarteruim
atuirctime
Poblem 3.
On the miniature chessboard in Figure 4, White has a lone Queen on
e8, and Red has a lone King on h5. Red moves frst because the King
is already in check. White wins if the Red King is drven back to e8
within 12 moves. If this is not accomplished, then Red wins. Other
than what is noted above, normal chess rules apply. With perfect play,
which royalty wins ?
8
V
7
6
5
M
e
f g
h
Figure 4.
Sol uti ohS
l wuite Queeninsasr.ll.s
, l ) rs u,a)
,z) rs ,s
, -) us f
,+) u rs ,b)
,) ,s e
,s) , es
,) u rs
,s) ,s e
,v) ,s rs
, l o) u es
N.tes ,a)r, l ) . ,s, tuen,z)usf, anc..ntinuerr.m,+) .
,b) r,+) . . es, ..ntinueasber.re r,+) . . rs,..ntinuerr.m, s)
z wuiteinsasr.ll.s
, l ) e ,+
,z) rs u
, -) r+ ,s
,+) e us,a)
,) r ,
,s) es us,b)
,) us ,s
,s) ,s us,.)
,v) , u
, l o) rs ,s
, l l ) us f
, l z) , rs ,c)
, l -) ,s e
, l +) r es
N.tes ,a)r,+) f, ..ntinuerr.m, l z) r,+) u,..ntinuerr.ml o
,b) r,s) rs,..ntinuerr.m, l -) r,s) u,..ntinuerr.m, l o)
, .) r,s) rs,tuen, v) ues
,c)r, l z) . . es, tuen, l -)r+ e, anc..ntinuerr.m, l +) .
- reckin,ins Anarti.lei ntuenextissueillpresentanar,u
mentsupp.rtin,tuis..n.lusi.nbys.lvin,tuer.yalrr.blemr.r all
! ? .uessb.arcs, ! ;: ;: 3.
Postsci pt
1ue ,eneral pr.blem as ..mpletely s.lvec by Ancy iiu s amazin,
stucents 1uree.rtuem, 1esse cuan, reteriauin, ancDaii , allin
tentu,race,,avetueirs.luti.ninmartinGarcner s r.yalrr.blem,
inQuatum ,september/O.t.berl vv-) 1uey..nrirmecmy..n]e.ture
tuat tue kin, .an tbe ..rnerec .n square rielcs, but su.t c.n my
.tuer..n]e.ture,tuattuequeenins.nly.nre.tan,ularb.arcsuen
> 2 anc! > + 1. tu.u,utuacsu.ntuattuekin,ins.ntue
4 ? 5 b.arc,butAncy sstucentssurprisecmebyrincin,aqueenin
.ntuisb.arc.
eruaps su.ulc acctuat all.uessrulesapply rtuequeenis
ac]a.entt.tuekin,, tuekin, .an take tuequeen, tuereby innin,
tue,ame Anctuekin,insirueisstalematec.nasquare.utsice
tueupperlen..rner, .riruea.uievesperpetual.ue.k.utsice tuat
..rner
chqpter 1 5
T oroi cq Cu rrenc
y
c.ubt ir many matuemati.ians .utsice Australia are aare tuat
tuereserveHank.rtuatnati.n uas issuect. bankn.testuatare
t.r.ical 1uey aretue : bill ,i,ure l ) printecin l vvzanc tue : l o
bill,i,urez)printecin l vv-.
1ue rr.nt anc ba.k.rea.u .r tuese uancs.me n.tes are su.n
Observetuat.nb.tusices.rea.ubilltuepatternsatt.pancb.tt.m
rapar.unc, asellastuepatterns.ntuelertancri,utec,es As
allt.p.l.,istskn., irtueri,utanclertsices.rare.tan,le, ancits
t.pancb.tt.m,are].inec,tueresultistueramiliart.rus, .rc.u,unut
suape. r]ust .nepair.rsicesis..nne.tecbyreversin,.ne.rtue
ec,es, tue stru.tureisakleinb.ttle rtuereversal appliest.b.tu
pairs.rec,es, tue stru.tureisapr.]e.tiveplane. reruapss.mecay
anati.nillprintpr.]e.tiveplane.rkleinb.ttlen.tes
1ue purp.se.rtue t.r.ical rapar.uncsist.makeituarcert.
..unterreit tue bills. s.puisti.atec ..l.r ..piers uave mace ..unter
reitin,mu.ueasierall.vertue,l.be.
An acciti.nal anti..unterreitin, reature .rea.un.teis a trans
parentinc.atal.er..rner,itappearsbla.kasprintecuere)
still an.tuer su.u cevi.e istuelittle .ir.le su.in, r.ur p.ints .ra
star.n.nesiceanctureep.ints.ntue.tuer. n.lceituerbillupt.a
str.n, li,utanctuep.intsruset.r.rm aperre.t sevenp.intecstar,
symb.lizin,Australia sseven.ri,inalstates 1uesli,utestvariati.n
inre,ister.ulccist.rttuestar
1uera.e.ntue :n.te.bvi.uslyistuat.rQueenLlizabetu 1ue
.man .n tue : l on.te is Dame mary Gilm.re, an Australian p.et
u..rkectirelesslyt.battlein]usti.esintuenati.n, espe.iallyin
tuetreatment.rtuenativeab.ri,ines 1ueman.ntue.pp.site sice
isA H Han]. aters.n, aballac sin,er anc].urnalistbestkn.n
r.r uavin, ritten tue .rcs .rwaltzin, matilca, Australia s un
This article frst appeared in Quatum (September/October 1994) .
117
118
Figure 1 .
official national anthem. (Matilda. by the way. is not a woman but
a knapsack. ) "The Man from Snowy Ri ver" i s another of Paterson s
popular songs.
The $ 1 0 note has been made mm longer than the $5 bill to help
sight-i mpaired persons di sti ngui sh between the two. Both notes are
made of locally produced polymer rather than imported paper. The
pol ymer l asts longer. stays cleaner. and can be recycled for plastic
products.
Melboure artist Max Robinson designed the new $ 1 0 note. Behind
Paterson s profi l e. in mi cropri nti ng. are l i nes from Paterson s verse.
making the bill even more difficult to counterfeit.
15. Toroki Currenc
y
119
Figure 2.
Chqpter 1 6
S i x Chq l l eng i ng Di ssecti on Tqs ks
karls.uerer,a..mputers.ientistinAu.klanc, Nezealanc, re.ently
p.sectuer.ll.in,sixtasks
l cutasquareint.turee..n,ruentparts
z. cutasquareint.tureesimilarparts, ]ustt..rui.uare
..n,ruent
-. cutasquareint.tureesimilarparts, n.t...n,ruent.
+ cutanequilateraltrian,leint.turee..n,ruentparts
5. cutanequilateraltrian,leint.turee similarparts, ]ustt..r
ui.uare..n,ruent
s. cut anequilateraltrian,leint.tureesimilarparts, n.t...n
,ruent
1ues.luti.nt.tuerirsttask is.bvi.us,seei,ure l ) . tis surely
unique, tu.u,u kn. .rn. pr..r an steart anc A. w.mstein
uave su.n tuat n. re.tan,le .an be civicec int. turee ..n,ruent
p.ly.min.esunlesstuepie.esarere.tan,les
Figure 1 .
This article frst appeared i n Quantum (May / June 1 994) .
Joural of Combinatorial Theory, Series A, Vol. 61 , September 1992, pp. 130-36.
121

a

b c
Figure 2
i,urez su.sturee s.luti.nst.tuese..nccisse.ti.ntask
1ask- ism.reciui.ult s.uererr.unctuepattern su.nini,
ure-. 1ues.luti.nisn.tunique, be.ausetueslantin,line.anassume
aninrinity.rp.siti.ns 1ue.nesu.nmaybetue.neinui.uline
se,mentsuavetuesmallestp.ssibleinte,erlen,tus
Asmatuemati.ian r.bert wainri,ut .rrlainvie, Ne1ersey,
uas.bservec,i,urezbresultsuentueslantin,lineis.rtu.,.nal.
1 8 3
" - " ' .
v
z:

^
I
II
9 1 2
Figure 3.
weturnn.t.tuetureeequilateraltrian,letasks
16. Six Ch'i i en9 i n9 Di sseci on T's ks 123
Fi9ue 4.
1uer.urtutask.bvi.uslyuasaninrinity.rs.luti.ns, .btainecby
r.tatin,tuetureetrise.tin,linesab.uttue.entralp.int,i,ure+)
1uetrise.tin,linesneecn.tbestrai,ut 1uey.anbeasi,,lyasy.u
like,pr.vicectuattueyareicenti.alancc.n.tinterse.t,i,ure+b)
s.uererr.uncanele,ants.luti.nt.tuerinutask,i,ure5) . t s
believect.beunique N.te its similarityt.i,urez..
Fi9ue 5.
1ue sixtu task is easily s.lvec ,i,ure s) . t s pr.bably unique,
tu.u,un.pr..riskn.n
Fi9ue 6.
1
?
Figure 7.
my.nly..ntributi.nt.tuesixtasks astuerecis..very.rase.
.ncs.luti.nt.tuetuirctask,i,ure) laterlearnecrr.ms.uerer
tuatueuacr.uncityearsearlier wuatistue value.rx, assumin,
tuesmallersice.rtuesmallestre.tan,leisI ? tu.u,uttuis.ulcbe
asimplequesti.nt.anser rx isn trati.nal,surelyit sare..,niz
ableirrati.nal, su.uas l -z . . ,tuesquarer..t.r-) , .r l . sl s . . ,tue
,.lcenrati.,.rten.allecpui) ,.rs.me.tuerellkn.nirrati.nal.
1.myamazement,x turnec.utt.beanirrati.nalnumberuac
neveren..unterecber.re.
1ue.ubi. equati.nrelatin, tue rati..rtue sices.rtue smallest
re.tan,let.tuerati..rtuesices.rtuesimilarlar,estre.tan,leis
1
J
x3 - 2x2 + x - I = 0,
( x2 - x) ( x - i , = 1 .
1uece.imalexpansi.n.rx is l +ssssz+ssvzs Aswain
ri,utp.intec .ut, tuenumberis .l.sely relatec t. pui , tue ,.lcen
rati. 1uere.ipr..al.rpuiequalspuiminus.ne 1uere.ipr..al.rx
equals (x - 1 )
2
. Otuerequalitiesare
i
2
= y- l
;
x
i
Y
=
.
x - I
pr.p.se.allin,tuisnumberui,upui D.nalcknutu, stanr.rc
tniversity sn.tec..mputers.ientist, su,,estec,ivin, ittue symb.l
<, inui.utuelittle.ir.le.rpuiisraisec nep.intec.utinaletter
u. .l.se a m.ciriec rra.ti.n r.r ui,upui resembles tue ..ntinuec
rra.ti.nr.rpui uiistuelimit.r
i
i + -
1 + 1
1 + 1
1 + 1
1 + 1
Accsquarer..tsi,nsancy.u,ettuem.ciriec..ntinuecrra.ti.n
r.rui,upui
Asknutur.te, tueseries..nver,esm.rerapiclytuantueseries
r.r pui, ,ivin, values tuat are alternately .veranc uncer tuetrue
value l , z, l . l , l s, l -, l . +, . .
knutuals..allecattenti.nt.tuer.ll.in,equalityr.rui,upui
1 i
1 +-< + -
< - 1
<
karls.uererp.ints.uttuattuetureere.tan,lesinmyri,ureuave
areas.r.. ancX4 . Ancirtue.ri,inal squareuasasicelen,tu.r
l , tuere.tan,lesuaveareas.ri;. i;.
anc 1 /X4 . 1uis su.stuat

1 1 / .+i ; .+1 / X4 , anctuerati..rtuelar,estre.tan,let.tuerest
.rtuesquareis0.
s.uerer su,,eststue terms puit., puituree, anc s. .n, r.rtue
rirstterms.rtueseries.rs.luti.nsr.rtueequati.n
1
- .- i ,
.
ne..n]e.turestuat l is tue sum.rtue inrinite series.rtuere.ip
r..als.rpuit.,puituree, puir.ur,ancs..n nbrier,
D
'"

cananyreacerpr.ve.rrerutetuis..n]e.ture: sitn.tsurprisin,
tuat su.u a simple ,e.metri.al ..nstru.ti.n .ulc ,enerate su.u a
.uri.usnumber: N.tetuatsss, tuenumber .rtuebeastintueH..k
.rrevelati.ns, r.ll.sits rirst six ce.imalci,its
PostSGi pt
r.bertwainri,utr.uncar.urtuayt..utasquareint.tureesim
ilaparts, ]ustt...n,ruent, assu.nini,ure s. tresemblesmy
tuircpattern ,i,urez.) , buttuepr.p.rti.nsareciuerent
Figure 8.
r.c.lr.mar.el.kur.uan,inHuen.sAires, anctuecanacianmatu
emati.ian Ancy iiu incepencently ..nsicerec tue task .r .uttin, a
squareint.r.ursimilarparts, meetin,tuer.ll.in,pr.vis.s
l . Allr.ur..n,ruent.
z 1ustturee..n,ruent
-. 1...n,ruent,ancacirrerentpairals...n,ruent.
+ 1...n,ruent,t.n.t
5. N.t...n,ruent.
1ueyr.uncs.luti.nst.allbuttuerirtutask D.essu.uapattern
exist:
samuel1 maltby, intueJournal of Combinatoral Theory, series
A ,v.l ss, April l vv+, pp +oz) pr.vectuatiranyre.tan,le ,in
.lucin,, .r..urse, tuesquare)iscisse.tecint.turee..n,ruentparts,
tuepie.esmustbere.tan,les. still.penisuetuertuis isals.true
.rall..nvexquacrilaterals.
severalreacersn.ti.ectuatmyi,ure scisse.ti.n.anbeeasily
extencect.pr.cu.e smallerancsmallersimilartrian,les
s.uerer s square pr.blem as ,eneralizec by Hyun,kyu cuun,
Ancy iiu, anc Daniel van vlietin Disse.tin, squares int. similar
re.tan,les, in Crux Mathematicorum ,v.l zz, l vvs, pp z+lz+s)
1ueyces.ribectueir,eneralizati.nasr.ll.s
we,eneralizetuepr.blem.rs.uerer ancGarcnerr.r
tue squareas r.ll.s Givenany inte,er! > i anc any .r
its2(m-l ) ..mp.siti.ns, .r.rcerecpartiti.ns, ! al + a
2
+
. . . + an , cisse.tasquareint.! similarpie.ess.tuattuere
are al ..n,ruent pie.es .r tue lar,est size, a2 ..n,ruent
pie.es.rtue nextlar,est size, anc s..n n tue .ri,inal
pr.blem,! 3 anctue..mp.siti.nsare , l ) -, ,za) l z,
,zb)z+ l, , -) l + l + l .
Ourmainresulti s tuattue cisse.ti.n pr.blemalays
acmits a s.luti.n usin, re.tan,ular pie.es ir anc .nly ir
tue ..mp.siti.n is n.t .rtue r.rm k + 1 , uere k is any
p.sitive inte,er. 1uese s.lvable.asesare..verecby t.
..nstru.ti.nsui.uare.nlysli,utlycirrerent
s.uerer s ..n]e.ture, ui.u ,ave at tueenc .rmy arti.le, as
su.tc.nbys.uereruimselr.
an steart s Scientifc Amercan ..lumn r.r 1une l vvs, anc a
brier n.te attue enc .ruis N.vember l vvs..lumn, ..n.ernec uat
ue.allstueplasti.number 1uisastuename,ivent.itre.ently
byri.uarcrac.van, an talianar.uite.tu..recitectue number s
cis..veryt.aren.uar.uite.tinl vz+ 1uenumbercerivesrr.mtue
sequen.el , l , l , z, z, -, +, , , v, l z, l s, zl ,inui.uea.unumber
istuesum.rtue se..nc anctuircnumberspre.ecin,it
1uesequen.eis,e.metri.allyci,rammecbytuespiralin,equilat
eraltrian,lessu.nini,ure v iiketueramiliarspiral.rsquares
tuat ,enerate tue ib.na..i numbers, a l.,aritumi. spiral .an be
cran.ntuepattern ntue.ase.rtueuirlin, squares, tuespiral
isins.ribec ntue.ase.rtueuirlin,trian,les, itis.ir.ums.ribec
1.ac]a.entnumbers.rtueib.na..isequen.euaverati.stuat
..nver,e.npui ,tue,.lcenrati.l sl s o-- 1uisirrati.nalnumber
istuevalue.r.intueequati.n .
- . - i O. 1ue.tuer sequen.e
is.l.selyrelatec rati.s.rac]a.entterms..nver,e.ntueirrati.nal
numberl -z+ l v ,tuevalue.r.intueequati.n.
- . - i O.
micuat1 Gazale, anL,yptian ele.tri.al en,ineer u.retirecin
l vv-aspresicent.rA1&1ran.e, sentmearas.inatin,unpublisuec
paper.ntueplasti.numberancitspr.perties, in.lucin,tueuirlin,
trian,lesanctueir.ir.ums.ribecl.,aritumi. spiral Gazaletu.u,ut
tue term plasti. number t.. u,ly , a,ree) nere.uristenecittue
silvernumbert.u.n.rits.l.seauinityt.tue,.lcenrati. su,,est
.allin,tuesequen.etuat,eneratesittuesilversequen.e
Ancn.r.rabi,surprise nmyi,ure, tuerati..r.
.
.+ 1
t..istuesilvernumber n.tuer.rcs, ui,upui, l + ) squarec,
minus ui,u pui, plus l , civicec by ui,u pui, is tue silver number
l -z+ 1uereareevensimplerr.rmulasrelatin,tuet.irrati.nals
i
2
a
nd < - <o
< - l
Figure 9
Chqpter 1 7
Lewi s Cq rrol 1 / s pi l l ow- Probl ems
ieis carr.ll as tue pen name .r cuarles D.c,s.n, u. tau,ut
matuemati.satcuristcuur.u,.ne.rtueOxr.rctniversity..lle,es
inLn,lanc neisbestkn.n, .r..urse, astueautu.r.rt.imm.rtal
rantasiesab.ut Ali.e anca l.n, n.nsenseballac.allec The Hunting
of the Snark.
n l sv-carr.llpublisuecalittleb..k.rseventyt..ri,inalmatu
emati.alpuzzles, many.rtuemn.teasilys.lvec 1ueb..k stitleas
Pillow-Prblems Thought Out Durng Sleepless Nights. .rtueb..k s
se..nc eciti.n ue .uan,ectue last t. .rcs t. akerul u.urs s.
reacers .ulcn t tuink ue suuerec rr.m .ur.ni. ins.mnia A ne
prera.easaccect. tue r.urtu eciti.n , l sv). carr.llintencectue
b..kt.bepart.ruatue .allec Curosa Matematica. rart, A
New Theory of Parallels, ast..seri.ust.be.allecre.reati.naleven
tu.u,uitasrittenitutueusualcar.llianuum.r
1uem.stinterestin,puzzlesinPillow-Problems ..n.ernpr.babil
ity 1uerirst.ne,pr.blem, issimplet.statebutextremely..nrusin,
t.analyze..rre.tly

Aba,..ntains.ne..unter,kn.nt.beeitueruite.rbla.k
Auite..unterisputin,tueba,suaken, anca..untercran
.ut, ui.u pr.ves t. be uite wuat is n. tue .uan.e .r
crain,auite..unter:

As carr.ll rites, .ne is temptect.anser l/z Her.retueuite
..unterisitucran, tue ba, is assumect. u.lc ituequal pr.ba
bilityeituer.nebla.kanc.neuite..unter, .rt.uite..unters
rtue..untersintueba,are bla.k anc uite, abla.k..unter ill
remainartertueuite.neistaken. rtue..untersareb.tuuite,
This article frst appeared in Quatum (Marchi April 1995) .
129
auite..unterillremainarter auite.neiscran He.ausetue
t.states.rtueba, areequallypr.bable, itseemstuatarterauite
..unter is taken, tue remainin, ..unter ill be bla.k .r uite itu
equalpr.bability
carr.ll .laims ..rre.tly tuat tue ab.ve ar,ument, tu.u,u intu
itivelyplausible, isceac r.n, iet A stancr.r a uite..unterin
tue ba, at tue .utset, B r.r a bla.k ..unter, anc C r.r tue accec
uite..unter Arter a uite..unteris taken, tuereareturee, n.t
t., equallyp.ssiblestates
l C uasbeentaken,leavin,A.
z A uasbeentaken, leavin,C.
- C uasbeentaken,leavin,B.
ntuerirstt..asesauite..unterremainsintueba, ntue
tuirc.ase, tueremainin,..unterisbla.k 1ue s.meuat surprisin,
anser,tuerer.re,isz/- .
1ue pr.bability .rrirst crain, a uite ..unter is -/+, anc tue
pr.babilitytuattueremainin,..unterisuiteisals. -/+ Or..urse,
as s..n asy.useetuattue..untertaken isuite, tue pr.babilities
alter rbla.k, tue.tuer ..unter is uiteitu.ertainty ruite,
tue.tuer..unteris uite ituapr.bability.rz/- , ancbla.kitu a
pr.bability .r l/- All tuis .an be mace .lear itu an invertec tree
cia,ram, seei,ure l )
A is taken.
B remains
B is taken.
A remains
Figure 1 .
A is taken.
C remains
C is taken.
A remains
17. Lewis Carrol l ' s Pi l l ow-Probl ems 131
1ue rra.ti.ns represent pr.babilities 1ue pr.bability .r ea.u .r
tuer.ur.ut..mes,b.tt.mr.)is l/ztimes l /z, .r l/+ 1uecia,ram
su.stuattureetimes.ut.rr.urauite..unterillbecran, anc
tureetimes.ut.rr.urauite..unterremainsintueba, rcrain,
a bla.k ..unter is n.t ..nsicerecassume tuat irtuis uappens tue
bla.k..unter ill berepla.ec anc crain, ..ntinuecuntil a uite
..unter is takentue remainin, ..unter is uite t. times .ut .r
turee
1uepr.blemiseasilym.celecituplayin,.arcs. suuuleace.k,
spreacitra.ec.n, ancrem.ve a .arcitu.utl..kin, at its ra.e
Hesicei tpla.era.ec.na.arcy.ukn.t.berec 1urny.urba.k
uile aniencmixestuep.siti.ns.rtuet..arcs 1urnar.uncanc
putarin,er.n.ne.arc 1ue.uan.etuatit srecis-/+, anctue.uan.e
tue.tuer.arcisrecisals. -/+ 1urn.vertue.arcy.u ret.u.uin,
rit sbla.k,tue.tuer.arcmustberec rit srec,tuepr.babilitytue
.tuer.arcisrec,.esc.nt.z/-
1ueb..k slastpr.blem,N.. z, uasbeentuesub]e.t.rmu.u..n
tr.versy.

Aba,..ntainsz..unters, ast.ui.un.tuin,iskn.nex.ept
tuat ea.u is eituer bla.k .r uite. As.ertain tueir ..l.urs
itu.uttakin,tuem.ut.rtueba,

nereis carr.ll ssurprisin,anser
wekn.tuat,iraba,..ntainec-..unters, zbein,bla.kanc.ne
uite,tue.uan.e.rcrain, abla.k.ne.ulcbez/- , anctuatany
oter state.rtuin,s.ulcnot ,ivetuis.uan.e
N.tue.uan.es, tuattue,ivenba,..ntains( a) BB, ( ,) BW, ( r) WW,
arerespe.tively l/+, l/z, l /+
Accabla.k..unter
1uen tue .uan.es, tuat it ..ntains ( a) BBB, ( ,) BW B, ( r) WW B,
are, asber.re, l/+, l/z, l/+
nen.etue.uan.e, .rn.crain,abla.k.ne,
1 /4 1 + 1/2 . 2/3 + 1 /4 1/3 2/3.
nen.etueba,n...ntainsBBW ,sin.eanyother state.rtuin,s
.ulcnot ,ivetuis.uan.e)
nen.e, ber.re tue bla.k..unter asaccec, it..ntainec BW, i e
.nebla.k..unteranc.neuite.
1ue pr..ris s..bvi.uslyralse tuat it s uarc t. ..mpreuenc u.
several t.p matuemati.ians ..ulc uave taken it seri.usly anc .itec
it as an example .ru. little carr.lluncerst..cpr.bability tue.ry
1uere is, u.ever, n.ttue sli,utestc.ubttuatcarr.ll intencecitas
a].ke neanserecalltuirteen .rtue.tuerpr.babilityquesti.nsin
uisb..k..rre.tly. ntueb..k sintr.cu.ti.nue,ivestueu.axaay
rany.rmyreacerssu.ulcreelin.linect.repr.a.ume
itu uavin, .rkec t.. unir.rmly in tue re,i.n .r c.m
m.npla.e, ancitunever uavin, venturect. ancer.ut
.rtue beaten tra.ks, .an pr.ucly p.int t. .ne rr.blem
in 1rans.encentalrr.babilities asub]e.tinui.u,be
lieve, very little uas yet been c.neby eventuem.sten
terprisin,.rmatuemati.alexpl.rers 1.tue.asualreacer
itmay seemabn.rmal, anceven parac.xi.al, but .ulc
uave su.uareaceraskuimselr, .ancicly, tuequesti.n s
n.tiireitselrararac.x:
tas.uara.teristi..rcarr.lltuatueencecuisb..kitua.u.i.e
spe.imen.rcarr.lliann.nsense
Chqpter 1 8
Lewi s Cq rrol 1 / s Word Lqdders
D.ublet tasks ..nsist .r .uan,in, .ne .rc t. an.tuer by alterin,
sin,le letters at ea.u stept.makea cirrerent.rc 1uet. .rcs
attuebe,innin,ancenc.rsu.ua.uainmust,.r..urse, betuesame
len,tu, anctueysu.ulcberelatect.ea.u.tuerins.me.bvi.usay
1ueymustn.tuaveicenti.allettersi ntuesamep.siti.ns All.rcs
intue.uainsu.ulcbe..mm.nLn,lisu.rcs, pr.pernamesex.lucec
A perre.t s.luti.n uas a number .r stepsequal t. tue number .r
pla.estue,iven.rcsciuer .rexample c.lc,c.rc, carc, warc,
warm r a perre.t .uan,e is n.t p.ssible, tuebest s.luti.n is tue
su.rtest .uain .r playin, c.ublets as a ,ame itu t. .r m.re
players, ieis carr.ll inventec a set .r s..rin, rules t. cetermine
u.ins
1ue rirstmenti.n.rtue ,ameincarr.ll s Diois.nmar.u lz,
l ss, uencarr.llrep.rtstea.uin,w.rciinks,uis.ri,inalname
r.rtue,ame)t.,uestsatacinnerparty neuacinventectue,ame,
uetellsusinapampulet,.ncuristmasDayin l sr.rt.little,irls
u.r.uncn.tuin,t.c..
carr.ll suancletterecw.rciinks AGamer.r1.rlayers, .ra
r.unc GameasritteninApril l ss. iatertuatyearueprinteca
revisecversi.nasar.urpa,epampulet. startin,itutuemar.uzv,
l sv, issue.r Vanity Fair, carr.ll..ntributeca series.rarti.les.n
c.ublets 1ue rirstasr.ll.ecbyanarti.le ann.un.in, a c.ublet
..mpetiti.nancatuircarti.le,ivin,anemetu.c.rs..rin,. nl sv
ma.millan,atuerectue Vanity Fair arti.lesint.a-vpa,eb..k,itu
recc.t..vers, titlecDoublets: A Word Puzzle. A l ssose..nceciti.n
as enlar,ec t. - pa,es The Lewis Carl Picture Book reprints
part .r tuis eciti.n iater tuat same year ma.millan publisuec a
tuirceciti.n,revisecancenlar,ect. s pa,es
carr.llt..ktuenamec.ubletsrr.maline.rtueit.uesin.anta
This article frst appeared in Math Horzons (November 1 994) .
133
ti.nin suakespeare sMacbet: D.uble,c.uble,t.ilanctr.ublea
linecarr.llpla.ec.ntuetitlepa,e.ruisb..k
Onmay l l , l ss, carr.ll menti.ns in uis Diary tuatueuas ex
tencec uis list.rsevenletter .rcpairstuat.anbelinkect.,etuer
t.m.retuanoo.
D.ubletsbe.ame aparl.r .razeini.nc.n,ancuasbeenamu.u
pra.ti.ecr.rm.r.rcplayeversin.e 1ueyuavebeen.allecby.tuer
names, su.uas.rclaccers, anc,invlacimirNab.k.v sn.velPale
Fire) .rc,.lr Ln.rm.us ener,iesuavebeenexpencec.nrincin,
su.rtest laccers r.r a ,iven pair .r.rcs. c.mputer s.rtare ..n
tainin,allLn,lisu.rcsisn..btainable,ancpr.,ramsuavebeen
rittenr.rrincin,minimum.uains in]ust arese..ncs 1uetask
isequivalentt.rincin,tue su.rtestr.utes..nne.tin,t.p.ints.n
a,rapu
D.nalcknutu,stanr.rcuniversity sn.tec..mputers.ientist,uas
..nstru.tec a ,rapu.n ui.u , .rtuem.st ..mm.n riveletter
Ln,lisu.rcs,pr.pern.unsex.lucec)arerepresentecbyp.ints, ea.u
].inecbyalinet.every.rct.ui.uit.anbe.uan,ecbyalterin,
]ust.ne letter 1ue ,rapu uas l+, l - lines On.e in a ..mputer s
mem.ry, pr.,rams.anberittentuatillcetermineina splitse.
.nc tue su.rtest .rc laccer ].inin, any t. .rcs .n tue ,rapu
knutur.unctureeletter .rcs t.. simple, anc six letter .rcs less
interestin,be.ausen.tt..many.anbe..nne.tec
m.stpairs.rriveletter.rcs.nknutu slist.anbe].inecbylac
cers s.meknutu.allstuemal..r .rcsbe.ause.ne.rtuemis
tue .rc al..ruave n. nei,ub.rs 1ue,rapuuas s l al..r.rcs,
su.u as eartu, ..ean, bel., su,ar, lau,u, rirst, tuirc, nintu 1.
.rcs, bares anc ..res, are ..nne.tec t. z .tuer .rcs, n.ne t.
a ui,uernumber 1uereare l o- .rc pairs itu n. nei,ub.rs ex
.ept ea.u .tuer, su.u as .cium.pium, anc m.nac,.nac. knutu s
l vvzcuristmas.arcreaturectuesmallestlaccer,eleven steps)tuat
.uan,ess.rct.pea.ebyusin,.nly.rcsr.uncintueHible sre
visecstancarcversi.n.
uavetakentueab.veinr.rmati.nrr.mtueei,utpa,escev.tect.
c.ubletsintuerirst.uapter.rknutu sb..k The Stanford GraphBase
,Accis.nwesley, l vv-) knutuill..vertuet.pi.m.rerullyinuis
r.rtu..min,tureev.lume.rk.n..mbinat.ri.sinuis.lassi.Art.r
c.mputerrr.,rammin, series .r uints .n u. t. s.lve c.ublets
itu.uta..mputerseeuisarti.leieiscarr.ll sw.rc, warc, ware,
Dare,Dame,Game, in Games ma,azine,1ulyAu,ust, l vs).
18. Lewis Carrol l ' s Word Ladders 135
tuasbeenp.intec.uttuatc.ubletsresembletueay in ui.u
ev.luti.n.reates ane spe.iesbymakin, smallranc.m.uan,esin
tue,enestuatareintervalsal.n,tueueli.alDNAm.le.ule carr.ll
uimselr,altu.u,uaskepti..rDarin stue.ry, ev.lvecmann.mApe
insixsteps
ALArLLrLLrrLArmArmAN
wuen,avetuiss.luti.ninaScientifc Amercan ..lumn.nmatu
emati.al Games ,tue ..lumn is reprintec as cuapter + in my New
Mathematical Diversions), t.reacerspr.cu.eca su.rter s.luti.n
ArLAr1Or1OA1mA1mAN
n a letter .rmar.u l z, l svz, , seem.rt.n c.uen s The Letters of
Lewis Caroll, v.lume z, pa,esvs) , carr.ll accecarule tuatall.s
.net.rearran,etueletters.rany.rc,..untin,tuisasastep witu
su.u in.reasec rreec.m, ue p.intec .ut, many imp.ssible c.ublets,
su.uas.uan,in,r.nt.ieac, .anbea.uievec
rONcONcONcOrNcOrDiOrDiOADiLAD
tiscirri.ultbutn.timp.ssibler.ra.rc.uaint.r.rmasenten.e
n Vanity Fair ,1ulyzs, l sv), .ne.rcarr.ll sc.ubletsaskecwnY
is it better NO1 t. marry: t. .uan,e wnY t. NO1 ue accec tuis
pr.vis. tue .uain mace wnY t. NO1} su.ulc emb.cy tue r.l
l.in, .bservati.n tuat l.vers, curin, tue temp.rary insanity .r
..urtsuip, t.. .rten rail t. re..,nize tue ,rave prucential reas.ns
ui.u su.ulc ceter tuem rr.m takin, tuis ratal step nere is car
r.ll s.levers.luti.n
wnYwnOwOOwO1NO1
1uematuemati.ianancriter.rs.ien.eri.ti.n,rucyru.ker,uas
likenecc.ubletst.ar.rmal system 1uerirst.rcistue,ivenax
i.m. 1uesteps.beytransr.rmati.nrulesanctuerinal.rcistue
tue.rem One seeks t. pr.ve tue tue.rem by tue su.rtest set .r
transr.rmati.ns
manypapers.nc.ubletsuaveappearecintue].urnal Word Ways,
aquarterlycev.tect.lin,uisti.amusements An arti.leintue eb
ruary l vvissueexpl.rec.uainstuatreversea.rc, su.uas1rAm
t.mAr1,iOGt.GOi,iOOrs t.srOOi,ancs..n 1ueautu.r
asksiranexample.anber.uncusin,asixletter.rc.
stuerea.l.sec.uain, .ncer,tuat.uan,essrrNGt. stm
mLrt.Au1tmNt.wN1Lr,tuenba.kt. srrNG:rs., uatis
tuesu.rtests.luti.n:
A k Deceny, in a c.mputer re.reati.ns ..lumn in Scientifc
Amercan ,Au,ust l vs), .allsa,rapu..nne.tin, all .rcs.r let
ters, a.rceb. ne su.stuatallzletter.rcsareeasily].inec
bysu.uaeb,ancasksirany.ne.an..nstru.ta..mplete.rceb
r.r -letter.rcs
ROGUE QUELL KETTLE COSTS SHOES
vogue quill settle posts sloes
vague quilt settee pests foes
value guilt setter tests foss
valve guile better tents gloss
halve guide betted tenth glass
helve glide belted tench class
heave glade bolted teach crass
leave grade bolter peach cress
lease grave bolder peace crest
least brave HOLDER PENCE CRUST
BEAST BRAVO
BLACK BEANS GRASS STEAL WHEAT
clack beams crass steel cheat
crack seams cress steer cheap
track shams tress sheer cheep
trick shame trees shier creep
trice shale frees shies creed
trite shall freed shins breed
write shell greed chins BREAD
WHITE SHELF GREEN COINS
FURIES TEARS PITCH FLOUR RAVEN
buries sears pinch foor riven
buried stars winch food river
burked stare wench blood riser
barked stale tench brood MISER
barred stile tenth broad
BARREL SMILE TENTS BREAD
Tb[ e 1 , These are Carrell' s best solution to fifteen doublets, Can any be improved?
18. Lewis Carrol l ' s Word Ladders 137
Postsci pt
my askin, r.r a .l.sec .uain .rtue seas.ns surely uas n. s.luti.n
D.nalcknutup.intec.uttuatAt1umN uas n.nei,ub.r ne su,
,estec ac.ptin, carr.ll s br.acer rules tuatall. ana,rammin, a
.rc aner a letter .uain mt1uAi, r.r instan.e, tuen be..mes a
nei,ub.r.rAt1umN Lven itutuesenerules stillc.n tkn.ir
a.l.sec.uainexists.
nere is u. knutu impr.vec s.me .rcarr.ll s s.luti.ns r.r tue
c.ubletstuat,ave
SHOES BLACK COSTS COSTS GRASS GRASS
shops brack coots coats gross crass
chops brace clots boats grows cress
crops trace plots blats grown tress
cross trice plats plats groat trees
cress trite plate plate great treed
crest write place place greet greed
CRUST WHITE peace peace GREEN GREEN
PENCE PENCE
Ch'ptel 1 9
The Ant on 1 ` i Z 2
An ant is at corer A of a i? 1 ? 2 box. [t crawls along the surface
along a geodesic, the shortest possible path, to a point B. Where is B
located to make the path as long as possible?
Figure 1. A 1 X 1 X ' bicube. U sol id formed by joining two cubes.
Intuitively one would guess B to be at the corer marked X because
thi s i s the poi nt the farthest from corner A. Yoshiyuki Kotani , a pro
fessor of mathematics in Saitama, Japan, recently made a surprising
discovery. Point B i s not at X, but one-fourth of the way down the
di agonal of the square face as shown!
The Geodesic from A t o X is easily traced by unfolding the solid
along a hinged edge as shown in Figure 2. The Pythagorean theorem
gives the path as the square root of 8, or 2. 828. . . . If you trace the
geodesic from A to B, the ant can take either of the two routes shown i n
Figures +and +. By symmetry there are two si mi l ar routes along the
hidden sides of the solid. Two of the four paths go over two sides, and
This [t ick Irst appeared in Math HLons (February 1996) .
1 39
1 40
the other two go over three sides. Applying the Pythagorean theorem
to these four paths. they all have the same length of 2. 850 about
0. 022 longer that the path from A to X!
x
2
A
2
Figure 2.
I do not know whether Kotani generalized the problem to 1 ? 1 ?
l sol i ds. In any case. physicist Richard Hess. who fi rst called my
attention to the problem. and four mathematicians to whom I sent
the problem (Ken Knowlton. Robert Wainwright. Dana Richards. and
Brian Kennedy) each i ndependently sol ved thi s more general case. I
expected that calculus would be required. but it turns out that by
unfol di ng the sol i d along hi nged edges, and appl yi ng basic algebra.
the formula for the location of B is not too dificult to find.
A -
2. 75
Figure 3.
19. The Ant C 1 7 1 7 2 141
iabeltuecia,.nal.rtuesquarera.eituoatX ancl attueci
a,.nal sb.tt.m..rner 1uecistan.e.rHrr.mX, al.n,tuiscia,.nal,
asarun.ti.n.rn, is (n - 1) /2n. rn 2, tuer.rmulaputsHaquarter
.rtueayc.ntuecia,.nal rn 3, H is l/- .rtueayc.n As
n appr.a.uesinrinity,H appr.a.ues l/zattuelimit,pla.in, itattue
.enter .rtue cia,.nal. Or..urse n .an take anyreal value ,reater
tuan l ancn.tne.essarilyaninte,er.
1 . 75
A
2. 25
Fi9ue 4.
r.blemsab.utaspiceratsp.tA.ntueall.rar..mtuat.rals
al.n,a ,e.cesi.t. .at.u a rly at sp.tH arein many.lassi. puzzle
b..ks nenryDuceney,tueHritisupuzzlemaker, ,ivessu.uapr.b
lem in The Canterbury Puzzles, anc uis Ameri.an ..unterpart sam
i.yc uas tue samepr.blemin uis Cyclopedia of Puzzles ,pa,ezl v)
1ueren.umatuemati.ianmauri.ekrait.uikp.sesasimilarpuzzle
inMathematical Recreations, ituanillustrati.nsu.in,allays.r
unr.lcin,tuer..m
su.upr.blems .an be rurtuer ,eneralizect. s.lics,.rr..ms) .r
cimensi.ns ? b ? C. Am.reciui.ultquesti.n,su,,estecby..mputer
s.ientistD.nalcknutu,ist.rincmaximumlen,tu,e.cesi.s.nsu.u
s.lics .rexample, tuemaximum,e.cesi..ntue 1 ? 1 ? 2 s.licis
not n.mtue.enter .r.ne square ra.e t.tue.enter .rtue.tuer, a
cistan.e.r- nessuas macetuesurprisin,cis..verytuattuemaxi
mumpatuuasalen,tu.rsli,utlym.retuan- ol Hutness sresults
.nmaximumlen,tu,e.cesi.s, asyetunpublisuec,areal.n,st.ry
chqpter 20
Th ree- Poi nt Ti l i ng
s.metilin,pr.blemsareinruriatin,be.ausetueyl..kasirtueyuave
easys.luti.nsbuttuenturn.utt.been.rm.uslyciui.ult 1.,..c
examplesaret..l.selyrelatecpr.blemsinv.lvin,tuetilin,.rp.ints
arran,ecintuer.rm.requilateraltrian,lesitu p.ints.ntuesice
1ue Greeks usec su.u latti.est. cia,ram uat tuey .allectrian
,ularnumbersnumberstuatarepartial sums.rl , z, - , . . rtue
number .r p.ints is even, it is easy t. tile any trian,le, itu.ut
.verlap,bylinese,mentstuat].int.ac] a.entc.ts.
1.uarcquesti.nsn.arise
rtuenumber.rp.intsisamultiple.r-, tuen
a).antuetrian,lebetilecitu-p.intstrai,utlines:
b) .an it be tilec itu little trian,les ..nsistin, .rturee
mutual ac] a.entp.ints:
1.puttuesamet.questi.nsinacirrerentr.rm,.anequilateral
trian,lesr.rmecbyac] a.entuexa,.nsuentuenumber.ruexa,.ns
isamultiple.r-betilecitueituer.rtuet.triuexes,as.allec
tuemin .ne .rmy Scientifc American ..lumns), tue strai,uttriuex
anc tue trian,ular triuex: we ill ..nsicer b.tu tasks intue p.int
r.rmratuertuantuep.lyuexr.rm
Any trian,le, u.sep.ints are a multiple .r-, .an betilecitu
amixture.rstrai,utlinesanctrian,lesbyac.ptin, tue .an.ni.al
patternsu.nini,ure l Hut.ansu.utrian,lesbetilecitueituer
tileal.ne:
c. n.t kn. tue .ri,in .r eituer .rtuese t. seemin,ly easy
pr.blems,butasrarasamaare,tueyererirstanserecinapi
.neerin,paperby1. n. c.nayanc1 C. ia,arias, 1ilin,itur.ly
.min.esancc.mbinat.rialGr.up1ue.ry, intueJournal of Combi
natoral Theory series A, v.l - , l vvo) pp ls-zos 1ue autu.rs
This article frst appeared in Cubism for Fun (No. 28, April 1992) .
143
.
I I I
I I
. I I I
I I I I I
I I I I
. I I I
I I I
I I I I
/ I I I I I
I I I I I I I I
Figure 1 . Canonical tiling for U = 1 2.
ereablet.pr.vetuatno trian,le.anbetilecitu-p.intlines, anc
tuattuey.anbetilecitu -p.inttrian,lesiranc.nlyir o : +
.r l l ,m.cul. l z) H.tupr..rs are l.n, anccirri.ult, makin,use.r
uattueautu.rs .allb.uncaryinvariants 1ueseare..mbinat.rial
,r.upinvariants ass..iatec itutueb.uncaries.rtuetilesanctue
b.uncaries .rtue re,i.ns t.betilec 1ue autu.rs acmit tuat tueir
pr..rsares.meuat..mpli.atec,ancitisreas.nablet.askirsim
plers.luti.nsexist.
Neituerpr.blem, tuey als.maintain, .an be s.lvecby a..l.rin,
pr..ecure 1uismaybetrue, buts.l.m.nG.l.mb, attueuniversity
.r s.utuern calir.rnia, r.unc a .lever ..l.rin, pr..rr.r tue linear
tiles tuat pr.ves imp.ssibility r.r all values .r ex.ept r.r o
.r -i ,m.cul. v) nis pr..rrirstappearecin uis ..lumn G.l.mb s
Gambits, intue John Hopkins Magazine, O.t.ber l vvo. t,.es as
r.ll.s iabeltuer.s.rtuetrian,leituturee..l.rs, a, b, ., tuat
repeat.y.li.allyassu.nini,urez.ntue i i trian,le
La.u verti.al tile ,slantin, eituer ay) ..vers .ne p.int .rea.u
..l.r La.u u.riz.ntal line ..vers turee p.ints .r tue same ..l.r
clearly, ire acctuenumber.r..verecsp.ts.rea.u..l.r,anypair
.rsumsmustciuerbyamultiple.r- n.ever, uen e..unttue
number .r sp.ts .rea.u ..l.rtueyc.n.tciuern.m.nean.tuerby
multiples.r- 1uisis a.lear..ntraci.ti.ntuatpr.vesimp.ssibility
tnr.rtunately, tuepr..rrailsuen o.r -1 ,m.cul.v).
A similar pr..r.rmy.nisbasec.n labelin, tuer.s l , z, -.
He.auseea.uverti.altileuasasum.rs, ancea.uu.riz.ntaltileuas
20. Three-Poi nt Ti l i n9 145
Figure 2. Coloring patter for Golomb' s proof.
a sum .r-, s, .r v, tue sum .ralltuetiles must bea multiple .r-,
ex.ept uen o 0 .r -1 ,m.cul. v) .r a uile tu.u,ut uac a
simpleimp.ssibility pr..rr.r all lineartiles, but itasqui.kly anc
incepencentlysu.tc.n byancuun,ancnerbert1ayl.r
A quite ciuerent pr..rtuat applies t. b.tu tilin, tasks uas been
cis..verecbyD.nalc c west, at suNY, lattsbur,u, NY, ancmay
appear in a r.rtu..min, paper t, t.., is neituer su.rtn.r simple,
restin,.naay.rtransr.rmin,patternst.tue.an.ni.alr.rm
i,ure- su.su.anotrian,le.anbetilecitutuelittletrian
,lesiro 9 .rl l H.tupatternsuaveapleasanttureer.lcr.tati.nal
symmetry 1ue pattern r.r o 9 is unique 1ue o 1 1 trian,le
als.uasanunsymmetri.tilin,,u.ever,ui.uis.btainecbya l so
Figure 3. Triangle tiling for U = 9 and U = 1 1 .
Figure 4. Triangle tiling for U - 12.
degree rotation of a hexagonal region of 8 small triangles (indicated
by dotted lines) . This tiling can also be obtained by adding a bottom
row of triangles (alternating in orientation) to the 9 tiling.
In Figure + we show a tiling for the 1 2 triangle, obtained by
adding two rows of triangles to the 9 tiling. Another tiling may be
obtained from this one by a l 80-degree rotation of a hexagonal region
equal to the one in the 11 triangle. The number of patterns for
the 12 triangle, and any larger triangle, is unknown.
Michael Beeler, from Newton, MA, found a general way to tile
any large N-triangle, N 0, 2 , 9, 1 1 (modulo l z)with small triangular
tiles. For both problems, the crucial lemma proves the existence of
tileable strips of width s and of arbitrary length, at least z For the
present triangle problem these strips are constructed as follows: First
two types and b of building blocks, each consisting of two small
triangles, are defined ( see Figure 5) .
- /
. .
Figure 5. Beeler' s method of tiling strips of width 6 .
20. Three-Poi nt Ti l i n9 147
Two blocks form a strip A of width 6 and length 2. Three blocks
b form a strip B of width 6 and length - Any strip of width 6, and of
even length, can be filled by strips A. If a 6-wide strip has odd length,
a strip B is taken off and a strip of even length remains, which can
be filled by strips A. Now, all ingredients needed to tile a large N
triangle are collected. Figure 6 shows the tiling plan, which makes
use of a combination of 6-wide strips (according to Figure 5), side- 1 2
triangles (according to Figure +) and one triangle of side 2, v, or 1 1
(according to Figure -).
2 . 9 Df 1 1
k-- 6
6
Figure 6. Tiling plan for large N = 0, 2, 9, 1 1 [mod 1 2] triangles.
Questions left over:
Is there a simpler way of proving the Conway/Lagarias results?
What is the number of different triangle tilings for large N?
What is the number of rotational- symmetric triangle tilings?
Chqpter 21
Luck
y
Nu mbers q nd 21 87
The house where I grew up as a child in Tulsa, Oklahoma, has an
address of 21 87 S. Owasso. Of course I never forgot this number.
Many years ago, when I was visiting my imaginary friend Dr. Irving
Joshua Matrix, the world' s most famous numerologist, I asked him if
there was anything remarkable about 21 87.
He immediately replied: "It is - raised to the power of 7. If you
write it in base -notation it is 1 0, 000,000. "
"I' m amazed you would know that ! " I exclaimed. "Anything else
unusual about 21 87?"
"My dear chap, " Dr. Matrix responded with a heavy sigh,
"
every
number has endless unusual properties. Exchange the last two digits
of 21 87 to make 21 78, multiply by +, and you get 87 1 2, the second
number backward. Take 21 87 from vvvv and the result is 781 2, its
reversal. Multiplying 21 by 87 produces 1 827, the same digits in a
different order. They are in correct order in the product of 27 and 8 l .
And have you noticed that the first four digits of the constant t , 2 , 7, 1 ,
8 , and the number of cubic inches in a cubic foot, 12
3
1728, are each
permutations of 21 87? You might ask your readers how quickly they
can insert plus or minus signs inside 21 87 to make the expression add
to zero. "
I was struggling to jot all this down on my notepad when Dr. Ma
trix added: "And 21 87 is, of course, one of the lucky numbers. "
I had never heard of lucky numbers. What follows is a summary of
what I learned about them from Dr. Matrix, and from the references
listed at the end of this article.
The notion of lucky numbers originated about 1 955 with Stanislaw
Ulam, the great Polish mathematician who co-invented the H-bomb
and was the father of cellular automata theory. It is one of the most
studied types of what are called "sieve numbers. " The oldest, most
This article frst appeared in The Mathematical Intelligencer (Vol . 19, No. 2, 1 997) .
149
important sieve numbers are the primes. They are called sieve num
bers because they can be generated by what is known as the Sieve of
Eratosthenes.

Imagine all the positive integers written in counting order.
Cross out all multiples of 2, except 2. The next uncrossed
out number is 3. Cross out all multiples of 3, except 3 .
Continue in this way, sieving out multiples of 5, 7, 1 1 , and
so on. The numbers that remain (except for the special case
of 1) are the primes.

The sieving process is slow and tedious, but if continued to infinity it
will identify every prime.
Using a sieve for generating lucky numbers is similar. Curiously,
it produces numbers closely related to primes even though they are
mixtures of primes and composites (non-primes). Here is how the
procedure works.
Step 1: Cross out every second number: 2, +, 6, 8, . . . , leaving only
the odd integers.
Step 2: Note that the second uncrossed-out integer is 3. Cross out
every third number not yet eliminated: 5, 1 1 , 1 7, 23, . . . .
Step 3: The third surviving number from the left is 7. Cross out
every seventh integer not yet crossed out: 1 v, 3v, . . . .
Step 4: The fourth number fom the left is v Cross out every
ninth number not yet eliminated, starting with 27.
As you continue in this fashion you will see that certain integers
permanently escape getting killed. Ulam called them "lucky num
bers. " Figure 1 lists all the luckies less than 1 , 000.
Eratosthenes' s sieve abolished all numbers except the primes. The
procedure is based on division. Ulam' s sieve, on the contrary, is based
entirely on a number' s position in the counting series. Using Eratos
thenes' s sieve you have to count every integer as you go along. Using
Ulam' s sieve you count only the integers not previously eliminated.
Although the luckies are identified by a sieving process completely
diferent from Eratosthenes' s sieve, the amazing thing is that luckies
share many properties with primes. The density of luckies in a given
interval among the counting numbers is extremely close to the density
of primes in the same interval. For example, there are 25 primes less
21 . Luck
y
Nu mbers and 2187 151
1 3 7 9 1 3 1 5 2 1 2 5 3 1 33 3 7 4 3 4 9 5 1 6 3
6 7 6 9 7 3 75 7 9 87 93 9 9 1 0 5 1 1 1 1 1 5 1 2 7 1 2 9 1 3 3 1 3 5
1 4 1 1 5 1 1 5 9 1 6 3 1 6 9 1 7 1 1 8 9 1 93 1 9 5 201 2 0 5 2 1 1 2 1 9 22 3 2 3 1
2 3 5 2 37 24 1 2 5 9 2 6 1 2 67 2 7 3 2 8 3 2 8 5 2 8 9 2 9 7 303 307 3 1 9 3 2 1
327 3 3 1 3 3 9 34 9 357 3 6 1 3 67 38 5 3 9 1 3 93 3 9 9 4 09 4 1 5 4 2 1 4 2 7
4 2 9 4 33 4 5 1 4 63 4 7 5 4 1 7 4 8 3 4 8 7 4 8 9 4 95 5 1 1 5 1 7 5 1 9 5 2 9 535
537 5 4 1 553 5 5 9 5 7 7 5 7 9 58 3 5 9 1 601 6 1 3 6 1 5 6 1 9 6 2 1 6 3 1 6 3 9
64 3 64 5 6 51 6 55 673 6 7 9 6 8 5 6 9 3 6 9 9 7 1 7 7 2 3 7 2 7 7 2 9 735 739
7 4 1 7 4 5 7 6 9 7 7 7 7 8 1 7 8 7 80 1 8 0 5 8 1 9 8 2 3 8 3 1 8 4 1 8 5 5 8 6 7 8 7 3
883 8 8 5 8 9 5 8 9 7 9 03 9 2 5 9 2 7 9 3 1 9 33 9 3 7 957 9 6 1 9 7 5 97 9 9 8 1
9 91 993 997
Figure 1 . A computer printout of lucky numbers less than 1 ,000, supplied by Charles
Ashbacher. Note that '99 will be a lucky year.
than 1 00, and 23 luckies less than 1 00. The overall asymptotic density
for each type of number is the same!
The distances between successive primes and the distances be
tween successive luckies keep growing longer as the numbers grow in
size. These distances also are almost the same for both number types.
The number of twin primes-primes that difer by 2-is close to the
number of twin luckies. There are eight twin primes less than 1 00,
and seven twin luckies in the same interval . Although primes play a
much more significant role in number theory than luckies, the simi
larities suggest that many of the properties of primes are less unique
than previously assumed. Their properties may be more a product of
sieving than anything else'
The most notorious unsolved problem involving primes, now that
Fermat' s Last Theorem has been proved, is the Goldbach conjecture.
It states that every even number greater than 2 is the sum of two
primes. There is a similar unsolved conjecture about luckies: that
every even number is the sum of two luckies. This has been computer
tested for integers up to 1 00, 000, and perhaps further than that, in
recent years, without finding an exception.
In a 1 996 booklet about number problems, Charles Ashbacher, of
Cedar Rapids, Iowa, conjectures that every lucky number appears at
the tail of a larger lucky. For example, 7 is at the end of 37; 9 at the
end of 49; 1 5 at the end of 61 5; and so so on. Lucky 87 is at the end of
my oId house number 21 87. Lucky 579 is at the end of lucky 96579.
Ashbacher wrote a computer program that verified his conjecture for
22 of the first 1 00 luckies. This suggests, he writes, that his conjecture
is a good bet.
It is easy to determine if certain large numbers are not lucky. Con
sider 98765. We can quickly tell it is not lucky because it has a digital
root of 8. The digital root of a number is its equivalence modulo 9-
that is, the remainder when divided by 9. If there is no remainder, the
digital root is 9. A digital root is quickly obtained by adding the digits
of a number, then adding again if the sum has more than one digit,
and continuing this way until just one digit remains. Lucky numbers
display all digital roots except 2, 5, and 8. Why? Because all numbers
with those three digital roots have the form 3k +2 and the first two
sieving steps eliminate them all .
Dr. Matrix called my attention to the curious fact that 1 3 , con
sidered the most unlucky of all numbers, is the fifth lucky, the sixth
prime, and the seventh Fibonacci number.
A few weeks after my meeting with Dr. Matrix I received from him
a fax message listing the following identities:
21 87 + 1 234 " 3421
21 87 + 1 2345 " 14532
21 87 + 1 23456 " 1 25643
21 87 + 1 234567 " 1 236754
21 87 + 1 2345678 " 1 2347865
21 87 + 1 23456789 " 1 23458976
Note how the sums on the right are permutations of the numbers
added to 2 1 87.
It has been proved that no polynomial formula will generate only
primes, and I would guess that the same is true for the luckies. How
ever, simple quadratic formulas will generate sequences of primes and
luckies. One way to search for such formulas was invented by Ulam.
On a square grid write the integers in a spiral fashion as shown in
Figure 2, and indicate the luckies by color. Note that nine luckies
clump along a diagonal . Applying the calculus of finite differences to
these luckies we discover that they are generated by 4x2 +:.+1 , as
x takes the values -3
,
-2, -: . 0
,
1 , 2, 3
,
4
,
5.
The spiral can start with any higher number to reveal clumps along
different diagonals. Leonhard Euler found that x
2 +x +41 generates
forty primes by letting x take values 0 through 39. If you write the
integers in a spiral starting with 41 , these primes will fill the entire
diagonal of a 40 ? 40 grid! Is there a quadratic formula equally rich, or
perhaps even richer, in finding a clump of luckies? I will be interested
in hearing from any reader who finds such a formula.
21 . Luck
y
Numbers and 2187 153
I !2 ! ! Jl4
! I! 7 I ! B I l2 l 2l
f
l
I
! ! 74 7 77 7B 7 B0 B ! B2
l
I
1 0 72 f 4 4 4
f 47 f 4B BJ

l
I
! B 7l 4Z 22 ZJ Z4 2 2 5! B4
I I I I
r T I
! 7 4! Z B 1 27 Z B
I
I I
I I
! 40 I Z I ! 2B J B
l
! B J 2B 4 I Z 2 4
L
I
I
I I I
I4 JB !
f I 4 J BB

l
I I
I0J f J J J4 JZ B
I I
T I
l2
~
4 Z I f
~
B 7 0

I
!0I ! 98 - 7 f 4 92
-
91
Figure 2. Ulam' s spiral technique for finding quadratic lucky-rich formulas.
There is a classic proof by Euclid that there is an infinity of primes.
Although it is easy to show there is also an infinity of lucky numbers,
the question of whether an infinite number of luckies are primes re
mains, as far as I know, unproved. Also unsolved is whether there is
an infinity of twin luckies.
Dr. Matrix enjoys practical jokes. When we talked about 21 87
he pointed out that if this number is divided by 9999 the quotient is
0. 21 8721 8721 87 = . . = I was momentarily surprised until I realized that
any integer of digits, not made entirely of nines, when divided by
a number consisting of nines, produces a decimal fraction in which
the original number is repeated endlessly as the quotient' s period.
"Ulam discovered lucky numbers with his lucky imagination, " Dr.
Matrix added. "Note the letters at positions 2, 1 , 8, and 7 in LUCKY
IMAGINATION. What do they spell?"
The first three lucky numbers are 1 , 3 , and 7. Now 1 37 not only is
a prime but it is one of the most interesting of all three-digit numbers.
It is, of course, the notorious fine-structure constant, the most myste-
rious of all constants in physics. I mentioned this to Dr. Matrix. This
prompted him to talk for twenty minutes about 1 37. Here are some
highlights of what he said:
Check the King James Bible' s first chapter, third verse, and sev
enth word. The word is "light. " Dr. Matrix reminded me that the
fine-structure constant is intimately connected with light.
The reciprocal of 1 37, or 1 divided by 1 37, produces the decimal
fraction 007299270072992700 . . . . The period is a palindrome!
Partition 1 37 into 1 3 and 7. The thirteenth letter of the alphabet
is M and the seventh is G-my two initial s!
Chlorophyll, which takes light from the sun to give energy to plants,
is made of exactly 1 37 atoms.
Dr. Matrix asked me to write my oid house number twice, 21 8721 87,
and put the number into my hand calculator. This number, he in
formed me, is exactly divisible by 1 37. I performed the division and
sure enough, the read-out displayed the integer 1 5965 1 . I got an even
greater surprise when Dr. Matrix asked me to turn the calculator
upside down. The number was the same inverted!
Dr. Matrix next asked me to divide 1 5965 1 by 73. The result was
21 87 ' I later discovered that this was another of Dr. Matrix' s hoaxes.
Any number of the form of ABCDABCD is evenly divisible by 1 37 and
73. The reason? ABCDABCD is the product of ABCD and 1 0001 . The
two prime factors of 1 0001 are 1 37 and 73, so dividing ABCABCD by
those two numbers will naturally restore ABCD. Of course the quotient
after the first division is not likely to be invertible.
"Is there any connection, " I asked, "between the lucky numbers and
666, the famous number of the Beast in New Testament prophecy?"
Dr. Matrix put his fingertips together and closed his emerald eyes
for a full minute before he spoke.
"Consider your old house number 21 87, and the first four luckies
1 , 3 , 7, 9. Omit the 1 in each number to leave 287 and 379. Add the
two numbers and you get 666. By the way, I forgot to mention earlier
that if you divide 1 8, the middle digits of 21 87, by 27, the first and
last digits, the quotient is 0. 66666666 . . . . "
I conclude with a mind-reading trick of my own that involves 21 87.
Ask someone to put this number into a calculator' s display. With your
back turned, tell him to multiply it by any number he likes without
revealing this number to you. He next calls out, in any order, each
digit in the product except one nonzero digit. You at once name the
missing digit.
21 . Luck
y
Nu mbers and 2187 155
How do you do it? As he calls out digits, keep adding them in your
head until you know the digital root of their sum. This is easily done
by casting out nines as explained earlier. If the digital root is 9, he
omitted 9. If less than 9, he left out a digit equal to the difference
between 9 and the digital root. For example, if the digital root is 2,
he omitted 7.
I leave it to you to figure out why this always works with 21 87.
Hint: 2 1 87 has a digital root of 9.
Refrences
[ 1 ] "On Certain Sequences of Integers Defined by Sieves. " Verna Gardiner, R. Lazarus,
Nicholas Metropolis, and Stani slaw Ulam. Mathematics Magazine, Vol. 3 1 , 1 956,
pp. 1 1 7-122.
[ 2] "The Lucky Number Theorem." David Hawkins and W. W. Briggs. Mathematics
Magazine, Vol. 3 1 , 1 957-58, pp. 8 1 -84, 277-280.
[3] A Collection of Mathematical Prblems. Stanislaw Ulam. (Interscience, 1 960), p.
1 20.
[4] "Prime-like Sequences Generated by a Sieve Process. " W. E. Briggs. Duke Math
ematical Joural, Vol. 30, 1963, pp. 297-3 1 2.
[ 5] "Sieve-generated Sequences. " Marvin Wunderlich. The Canadi"1 Joural of
Mathematics, Vol . 1 8. 1 966, pp. 291-299.
[6] "A General Class of Sieve-generated Sequences. " Marvin Wunderlich, ACf Arith
metica, Vol. 1 6, 1 969, pp. 41 -56.
[7] Excursions in Number Theory. C. Stanley Ogilvy and John T. Anderson. (Dover,
1 988), pp. l 00-1 02.
[8] The Penguin Dictionary of Curous and Intersting Numbers. David Wells. (Pen
guin, 1 986), pp. 1 3, 32-33.
[9] Unsolved Prblems in Number Theory. Richard K. Guy. (Second edition, Springer
Verlag, 1 994) , pp. 1 08-109.
[ 1 0] Zer to Lazy Eight. Alexander Humez, et al. (Simon and Schuster, 1 993), pp.
1 98-200.
[ 1 1 ] Collection of Prblems on Smarandache Notions. Charles Ashbacher. (Erhus
University Press, 1 996), pp. 5 1 -52.
Postsct" pt
My numerology involving 21 87 inspired several readers to find other
curiosities involving the number.
Monte Zerger discovered that 12
+ 22
+ a o a 218722 3481976250, a
number containing each of the ten digits ! He also pointed out that in
Which Way Did the Bicycle Go? ( 1 996), the authors Joseph Konhauser,
Dan Velleman, and Stan Wagon prove on page 1 34 (Problem 82) that
21 78 is the only four-digit number that reverses when multiplied by 4.
Charles Ashbacher provided a 3 x 3 magic square, made entirely
with lucky numbers; that has the lowest mathematical sum.
1 95 9 1 1 1
21 1 05 1 89
99 201 1 5
The square' s magic constant i s 3 1 5.
Owen O' Shea, i n Ireland, noticed that 872 - 2 12 71 28. He also
found the following result: 21 + 87 108, and 87 - 21 66. Multiply
1 08 by 66 and you get 7 1 28.
These from Dr. Matrix:
Write down all two-digit numbers that can be made with 2, 1 , 8,
and 7, without duplicating any digit. They add to 594. Multiply by 3
and you get 1 782.
Add the fourth powers of 2, 1, 8, and 7. The sum is 65 1 4. Now add
the fourth powers of 6, 5, 1 , and 4. The result is 21 78.
Ch(
pter 22
3 ` 3 M(
g
i c Squ( res
Perhaps the oldest of all combinatorial problems about numbers i s the
task of placing the first nine counting numbers in a + x + matrix so
that each row, column, and main diagonal has the same sum. It turns
out that, not counting rotations and reflections as diferent, there is
only one solution:
2 9 4
7 5 3
6 1 8
Legend has it that in the 23rd century B. C. the mythical King Yu
saw the pattern as spots on the back of sacred turtle in the River
Lo. Many modern scholars doubt that 10 shu, the most common name
in China for the magic square, is that ancient. They believe that the
pattern is not older than the tenth century A. D. At any rate, the name
means Lo River writing. The Chinese identify it with their familiar
yin-yang circle of light and dark regions. The four even digits (shown
shaded above) are identified with the dark yin. The Greek cross made
of the five odd digits is identified with the light yang. For centuries the
10 shu has been used as a charm on jewelry and other objects. Today
large passenger ships fequently arrange their shufleboard numbers
in the 10 sh u pattern.
How can we prove that the pattern is unique? The simplest way
known to me is first to note that the nine digits add to 45. If each of
the three rows (or columns) has the same sum, then the sum must be
one third of 45, or 1 5. We next list the eight possible triads of digits
that add to 1 5:
This article frst appeared in Quantum (January/February 1996) .
1 57
158
9 + 5 + 1
9 + 4 + 2
8 + 6 + 1
8 + 5 + 2
8 + 4 + 3
7 + 6 + 2
7 + 5 + 3
6 + 5 + 4
A G<rd nds Workout
The square' s center digit belongs to four lines of three. Five is the
only digit to appear in four of the triads; therefore, 5 must go in the
center. Digit 9 is in only two triads, so it must go in a side cell, and
the digit at the other end of the same line must of course be 1 . Digits
3 and 7 are also in just two triads. For symmetry reasons it does not
matter how they are placed in the other two side cells. This gives us
The remaining empty cells can now be filled with even digits in only
one way to complete the magic.
Various mnemonic methods allow one to form the 10 shu without
having to memorize the entire pattern. For me the easiest is to place
the even digits in the corners, in sequence from left to right, top to
bottom. The odd digits, also in sequence starting with 9 and going
backward, form the zigzag pattern show below:
Note that 1 , 2, 3 and 7, 8, 9 mark the corners of two isosceles triangles,
with 4, 5, 6 along a main diagonal.
In The Mathematical Gazette (December 1 970, p. 376) R. Holmes
pointed out a surprising property of the 10 shu. Take each row, column,
22. 3 Z 3 Mag i c Squares 159
and diagonal (including the four "broken" diagonals) as a three-digit
number to be read both forward and backward. The following identi
ties hold:
orogona1s
( 294 + 753 + 618) 2 (492 + 357 + 816) 2
( 276 + 951 + 438) 2 (672 + 159 + 834)
2
diagonals
(654 + 132 + 879) 2 (456 + 231 + 978) 2
( 852 + 174 + 639) 2 ( 258 + 471 + 936) 2
The identities are unafected if the middle digits of the numbers are
deleted, or any two corresponding two digits.
Imagine the 10 shu to be toroidally connected-that is, imagine the
pattern wrapped around both vertically and diagonally, as it would be
if drawn on a torus divided into nine cells. I was amazed to discover
years ago that if you add the four digits in each 2 x 2 square, the sums
are the nine consecutive numbers from 1 6 through 24.
Of course, an infinity of 3 x 3 magic squares can be constructed
with other numbers, not necessarily in counting sequence, and includ
ing all real numbers. To exclude trivial examples, we assume that no
two numbers are alike. For the matrix to be magic, however, the nine
numbers must fall into three triads, in each of which the numbers
are in the same aithmetic progression (that is, with the same difer
ences) . Moreover, the smallest numbers of the three triads must also
be in arithmetic progression, though not necessarily with the same
diferences as the triads. We can express these rules algebraically as
follows:
a + x a + 2y + 2x a + y
a + 2y a + y + x a + 2x
a + y + 2x a a + 2y + x
Note that each line of three cells has a sum of 3a+3y+3x, or 3( a+y+x) ,
proving that the magic constant must be a multiple of 3 , and that the
center number is one third of the constant.
With the square' s algebraic structure in mind, it is possible to con
struct a fascinating variety of magic squares based on given restraints.
160 A G<rdnds Workout
The box below shows a sampling of such squares. Each is the simplest
square meeting the restraints, defining "simplest" as the square with
the lowest magic constant. Do you see why the number 2 cannot be
used in a magic square of primes?
c=27
15 1 1 1
5 9 13
7 17 3
odd integers
c=1 1 1
67 1 43
13 37 61
31 73 7
pnmes
including 1
c=354
1 21 1 14 1 1 9
1 1 6 1 1 8 120
1 1 7 122 115
consecutive
composites
c=30
1 6 2 12
6 10 13
8 18 4
even integers
without zero
c=1 77
71 89 17
5 59 1 13
101 29 47
pnmes
excluding 1
c=54
27 6 21
1 2 18 24
15 30 9
composites in
athmetic prgression
c=24
14 0 1 0
4 8 1 2
6 16 2
even integers
with zero
c=31 77
1669 199 1249
619 1 039 1459
829 1879 409
pnmes In
arthmetic prgression
c=636
222 101 313
303 21 2 121
111 323 202
al numbers and c
ar paindrmes
Three-by-three magic squares of special types with lowest constant (c). The palindrome
square was constructed by Rudolph Ondrejka of Linwood, New Jersey.
In 1 987 I ofered $ 1 00 to anyone who found a 3 ? 3 magic square
made with consecutive primes. The prize was won by Harry Nelson
of Lawrence Livermore Laboratories. He used a Cray computer to
produce the following simplest such square:
1 , 480, 028, 201 1 , 480, 028, 1 29 1 , 480, 028, 1 83
1 , 480, 028, 153 1 , 480, 028, 1 71 1 , 480, 028, 1 89
1 , 480, 028, 159 1 , 480, 028, 213 1 , 480, 028, 141
Martin LaBar, i n The College Mathematics Joural (January 1 984,
p. 69) , asked if a 3 x 3 magic square exists with nine distinct square
numbers. ( Such squares exist with eight distinct squares plus a zero. )
Neither such a square nor a proof of impossibility has been found.
Nelson believes it exists, but beyond the reach of any of today' s super
computers running a reasonable amount of time.
I here ofer : l ooto the first person to construct such a square. If it
exists, its numbers are sure to be monstrously large. John Robertson
has shown that the task is equivalent to finding an elliptic curve of
the form y
2
x
3
- n
2
x with three rational points, each the double of
another rational point on the curve, having x-coordinates in arithmetic
progression.
Henry Ernest Dudeney, in Amusements in Matematics (pp. l z+
z) and in his article on magic squares in the fourteenth edition of
the Encyclopaedia Brtannica, defines magic squares based on sub
traction, multiplication, and division, and gives 3 x 3 examples of each
kind. There are also 3 x 3 antimagic squares with the property that
no two sums of three lines are alike. If certain provisos are met, a
variety of interesting anti magic combinatorial problems result. These
variants, however, will have to be the topic of another article.
Surely the most fantastic 3 ? 3 magic square ever discovered is one
constructed by Lee Sallows, a British electronics engineer who works
for the University of Nijmegen in Holland:
5 22 18
28 15 2
1 2 8 25
It would be hard to guess its amazing property. For each cell,
count the number of letters in the English word for its number, then
place these counting numbers in the corresponding cell of another 3 x 3
matrix. For example, "five" has four letters, so 4 goes into the top left
corner of the new matrix. Here is the result:
4 9 8
1 1 7 3
6 5 10
Not only is i t another magic square, but i ts integers are i n consecutive
order! Sallows calls the first square the Ii shu (Ii for his first name
Lee) , and the second square its alphamagic partner. His computer
investigations of alphamagic squares in more than twenty languages
are reported in his two-part article "Alphamagic Squares" in Abacus
(Vol. 4, 1 986, pp. 28-5, and 1 987, pp. 20-29, 43).
Now for three easy puzzles to work on before the answers are given
in the next issue.
1. Construct a 3 ? 3 magic square with nine consecutive positive
integers that has a magic constant of 666-the notorious number
of the Beast in the Bible' s Apocalypse.
2. Place 1 in a corner cell of the 3 ? 3 matrix, then fill the remaining
cells with consecutive nonnegative digits to form a magic square.
3. Arrange nine playing cards like this:
1 2 3
3 1 2
2 3 1
All rows and columns have a sum of 6, as well as one main diagonal,
but the other diagonal has a sum of 3 . Change the positions of three
cards to make the square completely magic.
Answets
l
219 224 223
226 222 218
221 220 225
2. This is the 10 shu with I taken from each number.
1 8 3
6 4 2
5 0 7
3 . There are two solutions. Shift the entire bottom row of cards to the
top of the square, or move the entire leftmost column of cards to the
square' s right side.
Addend u m
The following addendum to "3 x 3 Magic Squares" appeared in Quan
tum (March/April l vvs).
This will update my ofer of : l ooin "The Magic of 3 ? 3" for an
order-3 magic square made with nine distinct square numbers. Lee
Sallows, mentioned in my article, wrote a program that found many
almost magic squares in which only one diagonal failed to give the
magic sum. His square with the lowest constant is shown in Figure l
Such semimagic squares exist, as John Robertson of Berwyn, Penn
sylvania, has shown, if and only if they consist of three triplets of num
bers in arithmetic progression, all with the same diferences between
adjacent terms. Corresponding terms in the triplets need not be in
arithmetic progression, as required for the square to be fully magic.
Robertson has also shown that finding all such squares is equivalent
to finding all the rational points on certain elliptic curves.
127
2
46
2
58
2
2
2
113
2
94
2
742 822 972
Figure 1 . constant= 1472
In most cases found by Sallows, the constant is also a square, as in
the example given (Figure 1 ) . However, this is not true of all partial
magic squares, as shown by the counterexample in Figure 2, discov
ered by Michael Schweitzer, a Gottingen mathematician.
35
2
3495
2
2958
2
36422 2 1252 1 7852
2775
2
2058
2
3005
2
Figure 2. consWnf= 20966014
The constant for rows, columns, and one diagonal is the nonsquare
20966014. In my article I said that order-3 squares made of squares
are possible with zero in one cell. I should have added that squares of
this type are magic only in rows and columns.
Robertson sent a variety of 4 ? 4 magic squares made with distinct
squares, and called my attention to R. D. Carmichael' s Diophantine
Analysis. Order-3 magic squares for powers of are impossible unless
three powers can be in arithmetic progression. For this to be true, the
equation a
n
+ b
n
2c
n
must have solutions with distinct integers
for a, b, and c. Leonhard Euler proved there are no solutions for
3. This rules out order-3 squares made with cubes or multiples
of cubes. Carmichael also shows impossibility for 4 and multiples
of fourth powers. I have been informed by Noam Elkies that with
Andrew Wiles' s proof of Fermat' s Last Theorem it can be shown that
a
n
+ b
n
2c
n
has no solution for greater than 2.
Even though three squares can be in arithmetic progression, there
may be no way to construct a 3 x 3 fully magic square with nine
distinct squares. Schweitzer has shown that if such a square exists,
the central term must have at least nine digits, and if the entries have
no common divisor greater than one, all entries must be odd.
Postsci pt
By 1 900 it was known that no order- 3 magic square could be made with
powers of 3 , 4, and 5. It is now known that the equation x
n
+ y
n
2z
n
has no solution if exceeds 2, hence no 3 x 3 magic square, with
distinct digits, can be made with powers greater than 2. The final
proof was given by Henri Darmon and LoYe Morol in their paper (it' s
in English) "Winding Quotients and Some Variants of Fermat' s Last
Theorem," in Journal fr die reine und angewandte Mathematik Vol.
490 ( 1 997) , pp. 8 1-1 00.
My ofer of $ 1 00 for a magic square made with squares remains.
Here I extend the same ofer for a proof of impossibility.
John Robertson discusses the 2 case, and gives the historical
background, in "Magic Squares of Squares, " Mathematics Magazine,
Vol . 69 (October 1 996), pp. 289-283.
Magic squares made with squares are possible for squares of or
der 4 and have been constructed for many higher orders. Robertson
conjectures that such squares exist for all orders higher than 3 .
ChCpter 23
Some New Di scoveri es C bout
3 ` 3 MC
g
i c SquC res
One afternoon, while sitting i n a car waiting for my wife to finish
supermarket shopping, I located a pencil and paper and drew on the
sheet the unique 10 shu, or magic square made with distinct digits 1
through 9:
2 7 6
9 5 1
4 3 8
Having nothing better to do, I decided to see what happens if in
stead of adding digits in each row and column I multiplied them. The
top row' s product (2 ? 7 ? 6) is 84. The second row' s product is 45, and
the third row' s product is 96. I added the three products to see if the
sum was of any interest. It was 225, the square of the 10 shu' s magic
constant of 1 5 . Would the sum of the column products be equally in
teresting? To my amazement, it also was 225. This equality obviously
holds regardless of how the square is rotated or reflected. Could this
be just a coincidence? What about the two main diagonals? Might
their products also add to 225? No, the sum turned out to be 200.
Back home I began experimenting with other 3 x 3 magic squares
having numbers not necessarily in sequence. In every case the sum
of the products of rows equaled the sum of the products of columns.
Like so many discoveries in mathematics, far more significant than
this, I had found a result by experimenting with numbers in a manner
exactly like the way physical laws are discovered. The big diference,
of course, is that mathematical conjectures, obtained experimentally,
can be confirmed or falsified by rigorous proofs.
This article frst appeared in Math Horzons (February 1998) .
1 67
I recalled the following algebraic structure for all order- 3 magic
squares:
a + b a - b - c a + c
a - b + c a a + b - c
a - c a + b + c a - b
When distinct real values (including fractions, negative numbers,
irrational roots, J, L and so on) , even complex numbers, are substi
tuted for a, b, and c, the result is always magic. The matrix provides
a simple though tedious way to prove the conj ecture that for all 3 ? 3
magic squares the sum of low products must equal the sum of column
products.
I thought I had stumbled on a new property of order- 3 magic squares
not noticed before, but Lee Sallows, a computer scientist in Holland,
quickly disabused me. Hwa Suk Hahn, of West Georgia College, in
Carrollton, Georgia, reported the result in "Another Property of Magic
Squares" ( The Fibonacci Quarterly, Vol. 73, 1 975, pages 205-208). He
called magic squares with the sum of row products equal to the sum
of column products a "balanced square, " and proved that all order- 3
magic squares are balanced. The property was rediscovered by D. B.
Eperson, who mentioned i t i n a brief note to The Mathematical Gazette
(Vol. 79, July 1 995, pages 1 82-83) .
Hahn found similar squares of orders 4 and 5, but not of any higher
order. Shown below are balanced squares of orders 4 and 5.
1 7 13 19 25
1 14 7 12
14 20 21 2 8
15 4 9 6
22 3 9 15 16
10 5 16 3
10 1 1 1 7 23 4
8 1 1 2 13
18 24 5 6 12
After reading a first draft of this article, John Robertson suggested
adding here the following paragraph:
We have seen above that all order-3 magic squares are
balanced while for any higher order, some squares are bal-
23. Some New Di scoveri es about 3 7 3 Magic Squares
anced and some are not. Hahn also found a more sub
tle difference between magic squares of order 4 and magic
squares of higher order. He proved that if a constant is
added to every element of any balanced order-4 magic square,
the result is always another balanced magic square. If
a constant is added to a balanced magic square of order
greater than 4 the result will be a magic square, but it
might or might not be balanced. If a constant is added to
every element of the order-5 magic square in the diagram
above, the result is always a balanced magic square. (For
proof, note that you need only test five nonzero constants
as the sum of the row or column products is a fifth-degree
polynomial in the constant. ) There is a known order-5 bal
anced magic square, with non-integral entries, that loses
the property of being balanced when a certain constant is
added. It would be of interest to find an order-5 balanced
magic square with integral entries that produces a non
balanced magic square when some constant is added, or
prove there are none.
169
A question arises. Are there order- 3 magic squares for which the
sum of the products of the two diagonals is also the same as the other
two sums? I was unable to construct such a square.
In correspondence with Robertson, of Berwyn, PA, who is far more
skilled than I in solving problems in number theory, I mentioned my
discovery about the rows and columns, and wondered if diagonals
could have the same magic sum. A few days later he sent a proof
that there is an infinity of such squares.
Here' s how to construct them. Start with any sequence of three
square numbers in arithmetic progression, x
2
, y
2
, and z
2
. For in the
algebraic matrix substitute the value 2y. For b substitute the value of
x, and for C substitute the value of z.
Let' s take the simplest example: 1
2
, 5
2
,
7
2
. The three squares, 1 ,
25, 49, are in arithmetic progression with a difference of 24. Within
the matrix, then, has a value of 2 ? 5 10, b has a value of 1 , and
c has a value of 7. Lo and behold, like sorcery, the result is a magic
square of the type we are seeking (shown at the top of the following
page) . The sum of the row products is 1 , 500. This is also the sum of
the column products, and the sum of the diagonal products.
Figure I shows examples, provided by Robertson, of three higher
squares with the same strange property. Above each square I show the
1 1 2 1 7
1 6 1 0 4
3 18 9
arithmetic progression of square numbers that generates the square,
and the values taken by a, b, c in the matrix. Below each square is its
magic product-addition sum.
These magic squares-let' s call them Robertson squares-are re
markable patterns. Not only are they magic in the traditional sense,
they are also magic in an entirely diferent sense. As known to the
ancients, the additive constant of any order-3 magic square is 3a, or
three times the square' s central number. Robertson squares have, in
addition, what we can call the multiplicative-additive constant. It is
a( 2a2 - b2 - c2 ) , or more simply, 3a2 /2.
Here is a simplified account of how Robertson discovered the pro
cedure for constructing such squares. From the algebraic matrix it
is easy to determine that the sum of the low (or column) products is
a( 3a2 - 3b2 - 3c2 ) . As mentioned above, the sum of the diagonal prod
ucts is a( 2a2 - b2 - c2 ) . Setting these equal and simplifying yields
a2 = 2 (b2 + c2 ) . Because a must be even, we can rewrite the equation
as 2 (a/2) 2 = b2 + c2 .
The last expression tells us that b2 , ( a/2) 2 , and c2 are squares in
arithmetical progression. There is an infinity of such triples, and well
known formulas for producing them. This makes it easy to construct
an infinity of Robertson squares.
72 , 132 , 1 72 72 , 1 72 , 232 1 72 , 252 , 312
a = 26 a = 34 a = 50
b = 7 b = 7 b = 1 7
c = 1 7 c = 23 c = 31
33 2 43 41 4 57 67 2 81
36 26 1 6 50 34 18 64 50 36
9 50 19 1 1 64 27 19 98 33
26364 58956 187500
Figure 1. Three Robertson Squares.
23. Some New Di scoveri es about 3 7 3 Magic Squares 171
At the most, three squares can be in arithmetic progression. Such
triples are closely related to Pythagorean triangles-right triangles
with integral sides. The smallest square root of a number in the
progression is the difference between the triangle' s legs, the largest
square root is the sum of the legs, and the middle square root is the
Pythagorean triangle' s hypotenuse.
Consider, for example, the familiar 3, 4, 5 Pythagorean triangle.
The difference between its legs is 1 , the hypotenuse is 5, and the sum
of its legs is 7. The squares of these numbers, 1 , 25, and 49, form
the progression that generates the smallest Robertson square. The
three squares shown in Figure 1 correspond to Pythagorean triangles
with sides 5, 12, 1 3 ; 8, 1 5, 1 7; and 7, 24, 25. Thus from any given
Pythagorean triangle you can construct a Robertson magic square.
A question remains. Are there order-3 squares other than the 10
shu such that the sum of the row products equals the sum of the
column products, and such that this sum also equals the square of the
magic constant? (Diagonals are not considered. ) Robertson has shown
that the 10 shu is the only 3 ? 3 square in distinct positive integers
that has this property. If zero is allowed, the square below meets the
provisos.
14 0 10
4 8 12
6 16 2
The magic constant is 24. The sum of the rows (or columns) is 576,
and this equals 242 Similar squares exist if negative numbers are
allowed. For example, here is such a square found by Sallows.
1 -4 3
2 0 -2
-3 4 -1
Its sum of row products, sum of column products, and sum of diag
onal products in each case is zero. Zero also is the square of the magic
constant. As Sallows pointed out, adding 5 to each number produces
the 10 shu.
Robertson has found a simple way to construct order-3 magic squares
with distinct positive integers such that the sum of the row (or column)
products is an integral multiple of the square of the magic constant.
He is writing a note explaining the procedure.
An interesting property of order-3 magic squares, long known, is
that the sum of the squares of the first row equals the sum of the
squares of the third row. This is also true of course of the sums of the
squares of the first and third columns, though the two sums are never
the same.
Another curious and little-known property of all order-3 magic
squares, called to my attention by Monte Zerger, involves treating
such squares as matrices. If such a square matrix is multiplied by
itself the result is never another magic square. However, if the ma
trix is cubed, the result is always another magic square! For example,
when the 10 shu is multiplied twice by itself, the result is the following
magic square with a constant of 153 3375.
1 053 1 1 73 1 149
1 221 1 125 1029
1 1 01 1077 1 197
For a proof of this theorem as well as other results based on magic
matrices see N. Gauthier' s paper "Singular Matrices Applied to 3 ? 3
Magic Squares, " in The Matematical Gazette (Vol 8 1 , July 1 997, pages
225-220) .
Robertson adds:
For order-3 magic squares that are nonsingular when con
sidered as matrices, it is known that all odd matrix pow
ers are magic squares, including negative powers (Problem
E3440, Amercan Mathematical Montly, proposed Vol. 98,
No. 5, May 1 991 , page 437, solved Vol. 99, No. 1 0, De
cember 1 992, pages 966-967). Robertson has shown that
an arbitrary matrix product of an odd number of order- 3
magic squares is a magic square. It is straightforward to
show this for a product of three squares, and then proceed
by induction. He conjectures that both of the above re
sults (inverse is magic, arbitrary products of odd numbers
of squares are magic) are true of all magic squares of odd
order. A proof or disproof would be of interest.
23. Some New Di scoveri es about 3 7 3 Magic Squares 173
Postsci pt
Owen O' Shea, a correspondent in Ireland, sent the following curiosi
ties involving the 10 shu.
For each digit in the 10 shu, substitute its corresponding Fibonacci
number-that is, is position number in the familiar sequence 1 , 1 , 2,
3 , 5, 8, 1 3 , 2 1 , 35.
8 1 21
13 5 2
1 34 3
The sum of the sums of each row is 88, and this is also the sum of
the sums of each column. The sum of the products of each row is 600.
This is also the sum of the products of each column. The sum of the
products of each diagonal is 225, the square of 1 5.
Cross out the middle column of the 10 shu, in any of its orientations,
to form three two-digit numbers 68, 73, 24.
6
\
8
7

3
2

4
The sum of 68, 73, and 24 is 1 65. Reverse each number. The sum
of 86, 37, and 42 is also 1 65.
Consider the following chart of products and sums of the same set
of numbers and their reversals:
4624
5329
576
10529
7396
1369
1 764
10529
All these results hold regardless of how the 10 shu is rotated and
reflected.
Ch'pter 24
Pri mes i n Arith meti c
Pro
g
ress i on
Euclid proved that the number of primes is infinite, and for more than
200 years it has been known that there are arithmetic progressions
that contain an infinite number of primes. A much deeper question,
far fom resolved, is whether there are arbitrarily long sequences of
primes alone that are in arithmetic progression.
In the early 1 940s Paul Erd6s conjectured that any infinite se
quence of increasing integers, such that the sum of their reciprocals
diverges, contains arbitrarily long APs (arithmetic progressions). Be
cause the sum of the reciprocals of the primes is infinite (this was
proved by Leonhard Euler in 1 737) it follows that if Erd6s' s conjec
ture is true, the sequence of primes will contain arbitrarily long APs.
Although Erd6s ofered $3000 in 1977 for a proof of his conjecture, the
prize has not been collected.
In 1 976 L. J. Lander and T. R. Parkin made a stronger conj ecture:
For every k there are k consecutive primes in AP. "This conjecture is
undoubtedly true, " Erd6s has written, "but is completely unattackable
by the methods at our disposal. " As far as I know, the longest known
sequence of consecutive primes in AP are the six terms that start with
1 21 1 748 1 1 , and have a common difference of 30. This was reported
by Lander and Parkin in the same paper in which they made their
conjecture.
Let k be the number of primes in AP and d be the common dif
ference. If d 1 there is only one sequence: the doublet 2, 3 . If
d 2 there may or may not be an infinity of what are called "twin
primes", or primes that difer by 2. This is the notorious unsolved
twin-prime problem. Curiously, the sum of the reciprocals of all twin
This article frst appeared in The Rome Press, Raleigh, NC, Mathematical Sciences
Calendar 1998.
175
primes converges, which suggests (but certainly doesn' t prove) that
the set of twin primes could be finite.
When d 2 and k 3, the only triplet in AP is 3 , 5, 7. With d
greater than 2, Sarvadaman Chowla showed in 1 944 that the number
of prime triplets in AP is infinite. It has long been known that any
prime AP starting with 3 cannot have more than three terms, but
is there an infinity of such triplets starting with 3? This question
remains undecided.
In a note on "Prime Arithmetic Progressions" ( Crx Mathematico
1L1, Vol. 7 [ 1 98 1 ] , p. 68-69) Charles Trigg listed the first 25 triplets
that start with 3 in increasing value of the second term.
3 5 7 3 23 43 3 53 1 03 3 101 199 3 157 311
3 7 1 1 3 31 59 3 67 131 3 107 2 1 1 3 167 331
3 1 1 19 3 37 71 3 71 139 3 1 13 223 3 181 359
3 13 23 3 41 79 3 83 163 3 127 251 3 191 379
3 17 31 3 43 83 3 97 191 3 137 271 3 193 383
Prime triples in arithmetic progression starting with 3.
Trigg cited the triplet 3, 5003261 , 1 00065 1 9 as an example of one
with a large second term. Surely triplets with larger second terms
have since been calculated, but I do not know the current record.
Except for the doublet 2, 3, the difference in any prime AP ob
viously must be even because otherwise it would produce an even
number following a prime. But we can say much more about the dif
ference. It has long been known that if k, the number of terms in a
prime AP, is greater than 2, and the difference does not start with k
as the first term, d will be divisible by all primes less than or equal to
k. If the sequence starts with k, d will be divisible by all primes less
than k. This of course sets lower bounds for the difference in any AP
of k primes. If the difference is not the minimum set by the product of
primes, it will be a multiple of the minimum. For example, suppose
k is 25 . The diference cannot be less than the product of all primes
from 2 through 23, or not less than 223 , 092, 870. If d is more than
that, it will be a multiple of 223 , 092, 870. Knowing these divisibility
properties greatly simplifies computer searching for long sequences of
primes in AP.
Given a sequence of length k, we can ask what sequence of that
length maximizes or minimizes certain values such as the common
24. Pr i mes in Ari thmeti c Progressi on 17
difference, or the first or last term, or the sum of the terms. Here
we will consider prime APs with the smallest known first term, given
length k. If there is more than one sequence of length k that starts
with the same term, we will choose the progression with the lowest
diference.
When k 3, the first sequence is 3 , 5, 7 (d 2). When k 4 or 5,
the first is 5, 1 1 , 1 7, 23, 29 (d 6). When k 6, the first sequence is
7, 37, 67, 97, 1 27, 1 57 (d 30). When k 7, the first is 7, 1 57, 307,
457, 607, 757, 907 (d 150) .
When k 8, 9, or 1 0, the first sequence starts with 199 (d 210).
This progression was first reported in 1 91 0 by E. B. Escott. Magic
square bufs realized at once that it made possible a third-order (3 ? 3)
square of nine primes in AP. No third-order square of this type has
a lesser constant (magic sum of rows, columns, and main diagonals).
We shall indicate the constant with c.
1669 199 1 249
619 1039 1459
c 31 1 7
829 1879 409
For k 1 1 , the first known prime AP starts with 1 0, 859 (d 21 0) .
It has been credited to Malcolm H. Tallman, but I do not know the
date of its discovery.
For k 12, the first known starts with 23, 143 (d 30, 030) found
by V. A. Golubev in 1 967. A 1 2-term sequence starting with 1 66, 601
has both a lower difference, 1 1 , 550, and a lower last term.
When k 13, the first known starts with 4943 (d 60, 060) . I do
not know the discoverer.
When k 14, the first known starts with 1 2, 1 07, 737 (d 35, 735, 700) .
It was reported by Sol Weintraub in 1 976.
When k 15 or 1 6, the first known (also discovered by Weintraub
and reported in 1 976) starts with 1 3 , 81 6, 843 (d 35, 735, 700) . This
makes possible the construction of a fourth-order prime square that
is also pandiagonal (see below) . A pandiagonal magic square ( some
times called a diabolical or Nasik square) is one on which all broken
diagonals, as well as the two main diagonals, also add to the magic
sum. Put another way, if such a square is drawn on a torus surface,
all orthogonal and all diagonal rows add to the same number.
371 , 1 73, 843 49, 552, 543 228, 231 , 043 478, 380, 943
156, 759, 643 549, 852, 343 299, 702, 443 1 21 , 023, 943
335, 438, 143 85, 288, 243 192, 495, 343 514, 1 16, 643
263, 966, 743 442, 645, 243 406, 909, 543 13, 816, 843
A pandiagonal magic square of primes in arithmetic progression C - 1 , 127, 338, 372.
It is not hard to prove, fom the algebraic structure of third-order
magic squares, that no pandiagonal square of this size is possible.
The order-4 square, therefore, is the smallest that can be pandiagonal.
Whether such a square can be made with a lower constant than one
based on Weintraub' s sequence of 1 6 primes in AP remains an open
question.
When k 1 7, the first known prime AP (reported in 1 977 by Wein
traub starts with 3 ,430, 75 1 , 869 (d 87, 297, 2 1 0) .
When k 18 or 1 9, the first known sequence (Weintraub, 1 977)
starts with 8, 297, 644, 387 (d 4, 1 80, 566, 390) .
When k 20, the first known sequence (found in 1987 by James
Fry) starts with 803 , 467, 381 , 001 (d 2, 007, 835, 830) .
If a sequence with k greater than 20 has been found, it has not
come to my attention. Of course, if and when a sequence of length
25 is discovered, it will make possible a pandiagonal fifth-order magic
square of primes in AP.
Postsci pt
When I wrote this piece I gave six consecutive primes in arithmetic
progression. In 1 995 Harvey Dubner and Harry L. Nelson raised the
record to seven consecutive primes. The first prime has 97 digits and
the diference is 21 0. In 1 997 Dubner and four associates found a
24. Pr i mes in Ari thmeti c Progress i on 179
sequence of eight consecutive primes, and in 1 998 extended the record
to nine and ten primes ! Along the way their computer found 27 new
sequences of eight, and hundreds of sequences of seven.
The jump to eleven consecutive primes is a huge one. Nelson es
timates that if such a sequence exists the numbers will be at least a
thousand digits long. As Richard Guy wrote in "What' s Left?" (Math
Horzons, April 1 998), "It will be a while before anyone finds eleven. "
As far as I know the record for non-consecutive primes in arithmetic
progression is 22, found in 1 993 . Guy, in the article cited above, gives
two examples. It is still not known if there is a limit on the number
of consecutive primes in arithmetic progression.
Ch(pter 25
Pri me M(
g
i c Squ( res
Constructing magic squares with primes not necessarily i n AP (arith
metic progression) is one of the more esoteric aspects of magic square
theory, though one in which research with the aid of computers can
surely make advances that were not within the reach of earlier re
searchers. Literature on the topic is scattered in obscure places and
many questions remain unanswered. Here I will give only a few main
results. The famous British puzzle maker Henry Ernest Dudeney, in
his Amusements in Mathematics ( 1 91 7) , claims that he was the first
to consider prime squares (in The Weekly Dispatch, July 22 and Au
gust 5, 1 900) . He summarizes some findings, but at that time 1 was
considered a prime so many of these early results concern squares in
with 1 appears.
Accepting today' s convention that 1 is not a prime, the unique
third-order prime square with the lowest constant is not given by Du
deney. It is
17 1 13 47
89 59 29 c=l 77
71 5 101
There are four essentially different forms of the fourth-order prime
square with lowest constant. All four are given by Allan W. Johnson,
Jr. , in Crux Mathematicorum, Vol. 5 ( 1 979), page 241 . One of them is
at the top of the following page.
This article frst appeared i n The Mathematical Sciences Calendar in 1988, issued
by the Rome Press ( Raleigh, NC) .
181
3 61 19 37
43 31 5 41
c=120
7 1 1 73 29
67 17 23 13
The constant must be doubled to obtain a pandiagonal prime square
of order 4. The one shown below was published by Johnson in the
Joural of Recreational Mathematics, Vol. 1 2 ( 1 979-80), page 207,
and later proved by him to have the lowest possible magic sum for
such a square.
41 109 31 59
37 53 47 103
c=240
89 61 79 1 1
73 17 83 67
A second pandiagonal prime square of order 4, with diferent num
bers but the same constant, appears in what I believe to be the first
paper ever published on the topic, "Pan diagonal Prime Number Magic
Squares" by Charles D. Shuldham, in The Monist, Vol. 24 ( 1 91 4) , pages
608-61 3 . (Shuldham' s address is given as Wyoming, N. J. Either there
was such a town in New Jersey, or it was a misprint for N. Y. ) The
square is
73 41 13 1 13
23 103 83 31
c=240
107 7 47 79
37 89 97 17
25. Pr i me Magi c Squares 183
Johnson has proved that the fifth-order prime square shown below
has the lowest constant. (It was first published in Crux Mathemati
corum, Vol . 8 [ 1 982] , page 97, in Johnson' s article "Magic Squares in
Minimal Primes. ") Note that its constant also is prime.
3 83 41 101 5
89 67 1 1 29 37
19 7 103 31 73 c=233
79 53 1 7 13 71
43 23 61 59 47
In the same article Johnson gives what he has proved to be a sixth
order prime square of lowest constant (c = 432). A fifh-order pandi
agonal square of known lowest constant (c = 395) was published by
Johnson in the same journal, Vol . 6 ( 1 980) , p. 1 75. His best result for a
pandiagonal prime square of order 6 (c = 486) has not been published,
nor has it been shown minimal.
Little is known about pandiagonal prime squares higher than order
6. Shuldham' s examples for orders 7 and 8 use 1 as a prime. He was
unable to construct an order 9 with or without 1 , but he did construct
an order 1 0, with 7 its smallest prime and a constant of 990.
Even less is known about magic squares made with consecutive
primes. Johnson has constructed five such squares of order 4, of which
the one with lowest constant (published in the Joural of Recreational
Matematics, Vol. 1 4 ( 1 98 1-82), page 1 52) is shown below:
37 83 97 41
53 61 71 73
c=258
89 67 59 43
79 47 31 101
In his Crux Mathematicorum article of 1 982, Johnson gave consec
utive prime squares for orders 5 and 6, both of which he has shown to
have the lowest constants (3 1 3 and 484 respectively) . Note that the
fifth-order square' s constant is a prime. Researches in Magic Squares,
a Japanese book by Akira Hirayama and Gakuho Abe (published in
Osaka in 1 983) , contains diferent consecutive prime squares for or
ders 5 and 6 (with the same constants as Johnson' s), and consecutive
prime squares for orders 7, 8, and 9, with constants of 797, 201 6, and
221 1 respectively. I do not know whether the last three have been
proven minimal .
Is it possible to construct a magic square with consecutive odd
primes that start with 3? The surprising answer is yes, but not until
a square reaches the size of order 35. Such a square was first con
structed in 1957 by Akio Suzuki, of Japan. Another order- 35 square
of the same sort, but with diferent numbers, was constructed in 1980
by Captain William H. Benson, a retired U. S. Navy Oficer who coau
thored, with Oswald Jacoby, two original Dover paperbacks: New
Recreations with Magic Squares ( 1 976) and Magic Cubes ( 1 981 ) . Ben
son, who has not yet published his square, did not know of Suzuki' s
earlier discovery.
Does a consecutive-prime magic square exist for order 3? It proba
bly does, but the constraints on the order 3 squares, are so tight that
it could be that no such square is possible. It would require three
sets of triplets, each triplet in arithmetic progression with the same
common difference for each progression, and with the lowest terms
of the triplets in an arithmetic progression. If such a square exists,
the primes required may or may not be beyond the reach of computer
programs that run in a reasonable time. It is an outstanding unsolved
problem in magic square theory.
I will give $ 1 00 to the first person who can produce an order- 3
magic square formed with nine distinct primes that are consecutive.
Postsci pt
Harry Nelson, in 1 988, using a Cray computer at the University of
California, won my hundred-dollar prize. He found 22 solutions. What
is almost certainly the one with the lowest constant possible for such
a square is shown on the following page.
25. Pr i me Magi c Squares 185
1 , 480, 028, 201 1 , 480, 028, 1 29 1 , 480, 028, 1 83
1 , 480, 028, 153 1 , 480, 028, 1 71 1 , 480, 028, 1 89
1 , 480, 028, 159 1 , 480, 028, 213 1 , 480, 028, 1 41
Ch'pter 26
The Domi nono G' me
Dominono is a recently discovered relative of tic-tac-toe. Although it
appears easy to analyze, it is actually quite complex.
Dominono is played on a square grid such as the two boards shown
below. It can be played as a pencil-paper game like tic-tac-toe. Players
take turns putting their mark on any vacant square, with one player
using X' s, the other O' s. The person who forms a domino-that is, who
marks two squares that share an edge-loses. Note that it' s all right
to play in squares that only touch diagonally.
It' s more fun to play Dominono with checkers, counters, or coins of
two diferent colors. Players take turns placing one of their pieces on
a vacant square. The winner is the first to force his opponent to make
a domino.
On a 2 ? 2 grid, the second player obviously wins. Indeed, if the
board' s side has an even number of squares, the second player can
always win by using a symmetry strategy. After each move by his
opponent, he simply plays on the square symmetrically opposite the
opponent' s move on a line drawn through the board' s center. Because
of this strategy, only odd-sided boards offer playable games.
On a 3 ? 3 board, the second player wins when both sides play their
best, but it takes considerable analysis to prove it.
If the opening move is in the center, the second player wins by
taking, on his or her first two turns, two corner squares that are not
diagonally opposite one another. Playing a third move on any safe
square then forces the opponent to make a domino.
If the opening move is on a side square, the first player clearly can
not later take the center without losing. The second player, therefore,
wins by playing symmetrically.
If the opening move is in a corner, I thought there was no simple
winning strategy, until Fred Galvin, a mathematician at the Univer-
This article frst appeared in Games Magazine ( April 1999, p. 53) .
1 87
sity of Kansas, surprised me by finding a succinct strategy that takes
care of all opening moves. Here are the rules, based on the square
coloring shown below.
1 . Play on a white square as long as possible.
2. Never let your opponent take two opposite corners.
3. Don' t make a domino.
After you run out of white squares, if your opponent hasn't already
lost, he or she will lose on his next move.
The 5 ? 5 board shown, as well as all larger boards, remain un
solved. Perhaps a reader can write a computer program that will
decide whether the first or second player can always win, or whether
the game is a draw if both sides play perfectly. It has been conjectured
that the second player can always win on all square boards, but this
is far from established. Try playing on the 5 ? 5. You' ll quickly see
how enjoyable the game is even on so small a board, and also how
complex!
7 7
Now for a puzzle. Suppose the first player takes the square
marked X on the 5 ? 5 board. Can you prove that this is a
losing opening?

'cn. cc6c) q
'cn. cc
26. The Domi nono Game 189
Answer: like the side opening on the 3 ? 3, this starting move
prevents the first player fom later taking the central square. The
second player wins easily by symmetry play.
Postsci pt
On all even/even rectangular boards there is a central intersection
point that allows the second player to force a win by always playing
symmetrically opposite that point after each of his opponent' s moves.
If the rectangle is odd/odd, or odd/even, no general strategy is known,
and only small fields have been solved.
I explained how the second player wins on the 3 ? 3. A computer
program by A. E. Brouwer, in Holland, proved a second player win on
the 5 ? 5. Fred Galvin, in Lawrence, Kansas, showed by hand that the
second player also wins on the 3 ? 5 and 3 ? 7, but the 5 ? 7 and 7 ? 7
remain unsolved. He conjectures that the second player wins on all
odd/odd boards.
On even/odd boards the second player has an easy win on the 2 x 3.
Galvin found second player wins on the 2 ? 5, 3 x 4, 3 ? 6, and 4 x 5.
The 4 ? 7 case is still open. Galvin has proved that on all even/odd
boards the second player can either win or draw, but never lose.
On 1 ? fields, Galvin confirmed an earlier result by Richard Guy
that the game is a draw if 1 , 2, 4, 6. Otherwise the second player
wins.
It is surprising that a game with such simple rules is so complicated
and intractable on odd/odd and odd/even boards.
Ch'pter 27
The Growth of
Recre'ti on' l M'them'ti cs
Amusement is one of the fields of
applied mathematics,
-William F. White
My "Mathematical Games" column began in Scientifc Amercan's De
cember 1 956 issue with an article on hexaflexagons, curious hexagonal
paper structures that keep changing their faces when properly flexed.
Physicist Richard Feynman was one of their inventors. At that time
only a few books on recreational mathematics were in print. W. R.
Rouse Ball' s classic Matematical Recreations and Essays ( 1 892) had
been updated by the Canadian geometer H. S. M. Coxeter. A transla
tion from the French of Maurice Kraitchik' s Recreational Matemat
ics was available as a Dover paperback. Dover had not yet reprinted
Henry Dudeney' s Amusements in Mathematics, or my two-volume se
lection of mathematical puzzles by Sam Loyd. Aside fom a few other
puzzle collections, that was about it.
Since 1 956 there has been a remarkable explosion of books about
recreational math. Many are by such distinguished mathematicians
as Ian Stewart, who for some time wrote Scientifc American's column
on the topic; Clifford Pickover, of IBM' s Laboratory; John Conway, of
Princeton University; Richard Guy, of Calgary University; and Elwyn
Berlekamp, of the University of California at Berkeley. Matemat
ics: Problem Solving Through Recreational Mathematics, a pioneering
textbook by Bonnie Averbach and Orin Chein, of Temple University,
was published by W. H. Freeman in 1 980. New books in the field are
being written every year by mathematicians too numerous to mention.
A heavily revised and cut version of this paper appeared in Scientifc American (Vol .
279, August 1998) .
191
Articles on recreational topics run with increasing frequency in
mathematical periodicals. The quarterly Journal of Recreational Math
ematics began publishing in 1 96 1 . The National Council of Teachers
of Mathematics (NCTM) issued the first volume of William Schaffs
extensive Bibliography of Recreational Mathematics in 1 955. New ref
erences have required three sequels.
Obviously no sharp line separates entertaining math from serious
math. One reason is that creative mathematicians regard their work
as much a form of play as golf is to a professional golfer. In general,
math is considered recreational if it has a strong aspect of play that
can be appreciated by any layman interested in mathematics. This
can take many forms. They include elementary problems with ele
gant, at times surprising, solutions; mind-bending paradoxes, bewil
dering magic tricks, games that are easy and fun to play, ingeniously
concealed fallacies; topological curiosities such as Mobius bands and
KJein bottles; puzzles involving placing or moving counters, sliding
blocks, mazes, logic, dissections, tiling, and many other categories.
Almost every branch of mathematics below calculus has areas that
can be considered recreational . Let me give four amusing examples.

Jones puts three cards face down in a row. Only one is an ace, but
you don' t know which. You place a finger on a card. The probability
it is on the ace clearly is 113 . Jones now secretly peeks at each card,
then turns face up a card that is not the ace. What is the probability
your finger is on the ace? Most everyone thinks it has risen from 113 to
112. After all, only two cards are face down, and one must be the ace.
Actually, the probability remains 113 . If you now switch your finger to
the other face-down card, the probability it is on the ace jumps to 2/3 .
I introduced this problem in my Scientifc Amercan column (Oc
tober 1959) in the form of a warden and three prisoners. In 1 990
Marilyn vos Savant, in her popular Parade column, gave it with three
doors and a car behind one of them. Her answer was correct, but
she received thousands of angry letters, many from Ph. D. mathemati
cians, berating her for her ignorance of elementary probability! The
fracas generated a front-page story in the New York Times.

The statement "All crows are black" is logically equivalent to the
statement "All non black objects are not crows. " Finding a black crow
obviously confirms the first statement. You pick up a red tomato and
27. The Growh of Recreati onal Mathemati cs 193
see that it is not a crow. This just as obviously confirms the second
statement. Is a red tomato, therefore, a confirming instance of "All
crows are black?" If not, why not?
My April 1 957 column introduced this question, known as Hempel' s
paradox after philosopher Carl Hempel, who discovered it. I am still
getting letters about it, and a raf of papers claiming to resolve the
paradox have appeared in philosophical journals.

Arrange a deck so the card colors alternate. Cut the deck about
in half, making sure that the bottom cards of each half are not the
same color. Rife shuffle the two halves together as thoroughly or
as carelessly as you please. Take cards from the top of the deck in
pairs. Each pair will consist of a black and a red card! This is known
to magicians as the Gilbreath principle afer its discoverer, Norman
Gilbreath. I first explained it in my column of August 1 960, discussed
it again in July 1972, and added a proof when the last column was
reprinted in New Mathematial Diversions.
Magicians have invented more than a hundred card tricks based
on this principle. It generalizes. Arrange the deck so the suits are in
cycles of, say, spades, hearts, clubs, and diamonds. Deal any multiple
of four cards to the table, thus reversing the suit ordering, until you
have a pile of 24 or 28 cards. Riffle shufle the two piles together.
Every quadruplet of cards, taken fom the top, will contain the four
suits. Try this with two decks. Arrange one deck in any random order,
and the other deck with cards in the reverse order. Shufle the decks
together, then divide the cards exactly in half. Each half will be a
complete deck!
Although Gilbreath had no purpose in mind other than to furnish
a basis for card tricks, last year Donald Knuth, Stanford University' s
noted computer scientist, discovered a wonderful application of the
principle to the eficient input and output of computers.

Consider the square patterns shown in Figure 1 . Each is made
with the same sixteen pieces, yet the square on the right has a large
hole in its center! What happened to the missing area? It would spoil
the fun to explain it here. This strange fallacy, which I devised for
a May 1 961 column, was recently made in China as a puzzle with
colorful plastic pieces, for sale in a dollar store chain.

194 A G4rdner' s Wor:out
Figure 1. The authOI's vanishing area paradox. The 16 polygons on the left fOI'm a sol i d
square. The same 1 6 pieces. in a diffrent patter on the right. form a square wi th a
central hole. Where did the missing area go?
Al though Mathematics Teacher, a monthl y publi shed by the NCTM,
has increasingly carried articles on recreational topics, teachers them
selves have, by and l arge, i gnored such materi al . For forty years I
have done my best to convince them that entertaining math is more
than what they call "enrichment" material . It should be i ntroduced
regularl y as a way to arouse the i nterest of young students in the
wonders of mathematics, So far, movement in this direction has been
at a snail ' s pace. I have often told the story of how one day, i n a
high school math study period, I had finished my assignment and was
trying to decide whether the first player can always win a game of
tic-tac-toe. The teacher saw my scribbl i ng, snatched away the sheet,
and said, "Mr. Gardner, when you' re in my class I expect you to work
on mathematics and nothi ng else. "
The situation is much the same today. I know of no recommenda
tion by the NCTM that asking students to solve the game of tic-tac-toe
is a superb way to i ntroduce combi natori al mathematics, symmetry,
graph theory, game theory, and even probability. Moreover, it ties
i n strongly wi th every student' s experience. Who has not, as a chi ld,
played tic-tac-toe?
I recently purchased the 1 997 yearbook of the NCTM, titled Mul
ticulturalism and Gender Equity in the Mathematical Classrom: The
Gift of Diversity. The book has 28 chapters written by teachers who
defend a current trend that goes by the name of "new new math" to
distinguish it from the failures of the old "new math" that had been so
disastrously backed by the NCTM in the late 1 960s. The new fad tries
to do away with lecturing by forming the class into small groups of stu
dents who are given problems and asked to solve them by cooperative
reasoning.
For example, instead of explaining the Pythagorean theorem on
a blackboard, students are asked to cut right triangles and squares
from graph paper and discover the theorem all by themselves. This
may take days, and even end in failure. "Interactive learning, " as
it is called, is substituted for passive listening to a teacher. Rote
memory of addition and multiplication tables, and learning how to do
long division, is replaced by the use of calculators. The result: When
students get to college they are unable to multiply or add even small
numbers without reaching for their calculator. As for long division,
forget it!
The new new math also places a heavy emphasis, admirably so, on
multiculturalism and gender equity. Translated from the yearbook' s
mind-numbing jargon, this means that all ethnic groups should be
treated equally, and girls treated the same as boys. It is the new new
math' s positive side.
What struck me most about the yearbook was that not a single
teacher had anything good to say about the value of recreational math
in hooking student interest. Let me propose to teachers the following
experiment. Ask each group of students to think of any three-digit
number ABC, then put into their calculator the number ABCABC.
Thus i f they thought of 237, they would give their computer the num
ber 237237. Tell them you have the psychic power to predict that if
they divide their ABCABC number by 1 3 there will he no remainder.
This proves to be true. Now ask them to divide the result by 1 1 . Again,
as you predicted, no remainder. Finally, ask them to divide by 7. Lo
and behold, their original ABC number is now in the readout. I' ll wa
ger that this task will create much more excitement than asking them
to invent the Pythagorean theorem. I know of no better introduction
to number theory and the properties of primes than asking students
to explain why this trick always works.
One of the greatest joys of writing the Scientifc Amercan column
over a period of thirty years was getting to know so many authentic
mathematicians. I myself am little more than a journalist who loves
mathematics and can write about it glibly. I took no math in college. If
you go through my fifteen book collections of columns you will see that
the math grew increasingly sophisticated as I learned more. But the
key to the column' s popularity springs not so much fom my writing
as from the material I was able to coax from genuine mathematicians.
Solomon Golomb, of the University of California, was one of the
first to supply grist for the column. I had the pleasure of introducing
his polyominoes, shapes formed by joining unit squares along their
edges. Polyomino problems are now a flourishing branch of recre
ational math. Two books are devoted entirely to these fascinating
little shapes, one by Golomb, one by George Martin, at SUNY Albany.
Golomb also provided material on what he named "rep-tiles, " identical
polygons that fit together to form larger replicas of themselves. Five
examples are shown in Figure 2. Any rep-tile obviously will tile the
plane by making larger and larger replicas.
The late Piet Hein, Denmark' s illustrious poet and inventor, be
came a good friend. I devoted a column to his topological game of Hex
(July 1957 and June 1975), to his superellipse and superegg ( Sept.
1 965), to his Soma Cube (Sept. 1958, July 1 969, and Sept. 1 972), and
other of his inventions. There is a beautiful proof that the first player
can always win Hex on a board of any size, though it tells you nothing
about how to go about it. The Soma Cube consists of seven polycubes
unit cubes joined at their faces. They fit together like the seven "tans"
of tangrams to form a cube, as well as endless other forms. Conway
was among the first to prove that the cube has exactly 240 solutions.
John Conway is one of the world' s undisputed geniuses. He came
to see me one day and asked if I had a Go board. I did. On it he
demonstrated his now-famous cellular automata game called Life. By
applying three ridiculously simple rules that tell when a cell on the
board changes to full or empty, an astonishing variety of Life forms
are created, some that move across the field like tiny insects. That
column became an instant hit with computer bufs. It is said that
for many weeks business firms and research labs were almost shut
down while Life enthusiasts experimented with Life forms on their
computer screens.
Conway later invented an entirely new way to define numbers that
generates infinite classes of weird numbers never met before. I ex
plained this in a September 1 976 column. Donald Knuth wrote a
novel entitled Sureal Numbers about how a boy and girl found some
ancient stone tablets that enabled them to develop Conway' s number
system.
27. The Gw\h of Recreti or1 Mthemtics 197
b
C d
figure 2. Five low-o.de. .ep-tiles. ( a) A .ep-2 t.iangle. ( b) A .ep-3 t.iangle. ( c) A .epA
quad.il ateml . (d) A .epA hexagon. (e) The Sphinx: the only known .epA pentagon.
Conway collaborated with Guy and Berlekamp on a two-volume
work called Winning Ways ( 1 982)1 which I consi der to be the greatest
contribution to recreational mathematics this century. It i s a won
drous source of completely new materi al . One of its hundreds of gems
i s a two-person board game called Phutball. Phutball was invented by
`Reissued b A K Peters, Ltd. in D four volume edition.
Conway as an example of a game with no distinction between moves
made by each player; only the goal, to get a piece called the phut
ball across a player' s goal line, difers. I'm surprised the game has
never been marketed. Winning Ways also contains Berlekamp' s path
breaking analysis of the children' s pencil game of dots and squares.
Ron Graham, Bell Labs' s distinguished mathematician and juggler,
supplied material on Ramsey theory (Nov. 1 977) and Steiner trees
(May 1 986) . Frank Harary, now at New Mexico State University,
generalized tic-tac-toe to the goal of forming a specified polyomino
instead of a straight row (Apr. 1 979). Ronald Rivest, then at M. LT. ,
allowed me to be the first to reveal the "public-key" or "trapdoor cipher"
system of which he was coinventor (Aug. 1 977) . It was the first of such
ciphers to revolutionize cryptology.
Other new developments in recreational mathematics that I had
the pleasure of introducing include Robert Abbott' s card game Eleusis
(June 1 959 and Oct. 1970) which so remarkably models the role of
induction in science; the bewildering paradox of the unexpected hang
ing (March 1963); Scott Kim' s magic lettering of words and phrases
that are the same upside down or mirror reflected (June 1 981 ) ; the
mathematical art of Maurits Escher (Apr. 1 966) ; and the non periodic
tiling discovered by Roger Penrose, England' s famous mathematical
physicist.
Penrose tiles are a marvelous example of how a discovery made
solely for the fun of it can turn out to have an unexpected practical
application. His two tiles, kites and darts, tile the plane in such a
way that there is no fundamental region that repeats itself. I had the
honor of explaining the tiles in a January 1 977 column. The issue' s
cover picture was drawn by Conway, who made many amazing dis
coveries about properties of Penrose tiling. A few years later, a three
dimensional form of Penrose tiles became the basis for constructing
a hitherto unknown type of crystal. Hundreds of papers on what are
now called quasicrystals have been published. The outstanding un
solved question in tiling theory is whether a single tile exists that can
tile the plane only nonperiodically.
The two columns that generated the greatest number of letters
were my hoax column and one on Newcomb' s paradox. The April Fool
column (Apr. 1975) purported to cover great breakthroughs in science
and math. They included a refutation of relativity theory, a psychic
motor, the discovery that Leonardo da Vinci had invented the flush
toilet, that the opening move of pawn to king rook four was a certain
win in chess, and that e raised to the power of 7V163 was exactly equal
to the integer 262, 537,41 2, 640, 768,744. To my amazement, thousands
of readers failed to recognize the column as ajoke. I gave a picture of a
complicated map that I said required five colors. Hundreds of readers
sent me copies of the map colored with four colors, many saying the
task had taken days.
Newcomb' s paradox, named after its discoverer physicist William
Newcomb, was first described in a technical paper by Harvard philoso
pher Robert Nozick. I received so many letters after writing about the
paradox (July 1 973) that I packed them into a large carton and person
ally delivered them to Nozick, who analyzed them in a guest column
(Mar. 1974) .
The paradox concerns two boxes, A and B. Box A contains a thou
sand dollars. Box B may either contain a million dollars or be empty.
You are allowed to choose either B, in the hope of getting a million,
or you can take both boxes.
A superbeing, God if you like, has the power of knowing in advance
how you will choose. If he predicts that out of greed you will take both
boxes, he leaves B empty and you get only the thousand in A. If he
predicts you will take only Box B, he has put a million dollars in it.
You have watched this game played many times with others. In every
case if a player took both boxes, he or she found the box empty. Every
time a player took only B, he or she became a millionaire.
How should you choose? The pragmatic argument is that because
of the previous games you have witnessed, you can assume that the
superbeing does indeed have the power to make accurate predictions.
You should, therefore, take only B and know you will get a million.
But wait ' The superbeing made his prediction before you play the
game, and has no power to alter it. There are two possibilities. Either
B is empty or it contains a million. If empty, you lose nothing by
taking both boxes. If it contains a million, you gain a million plus a
thousand by taking both boxes.
Each argument seems impeccable. Yet each cannot be the best
strategy. Nozick concluded that the paradox, which belongs to a
branch of mathematics called decision theory, remains unresolved.
My personal opinion is that the paradox proves, by leading to a logical
contradiction, the impossibility of a superbeing' s ability to predict a
person' s decision. In any case, the paradox remains controversial, and
a raft of papers have grappled with it since Scientifc Amercan gave
it wide currency.
Figure 3. The only magic hexagon possible with consecutive number starting with l .
Every straight row of 3 , 4, or 5 cells has a sum of 38.
Magic squares have long been a part of recreational math. The
order- 3 (3 x 3) square, using digits 1 through 9, is unique apart
from rotations and refections. Squares of order 4 have 880 patterns,
and the number of arrangements increases rapidly for higher orders.
Suprisingly, this is not the case with magic hexagons. When I pub
lished the order-3 hexagon shown in Figure 3 (Aug. 1963) , Charles
Trigg, a retired Californa math teacher, proved that not only is this
elegant pattern the only order- 3 magic hexagon, but that no magic
hexagons of any other size are possible.
Magic squares need not be restricted to numbers in consecutive
order starting with 1 . If the only requirement is that the numbers
be distinct, a wide variety of order- 3 magic squares, in which rows,
columns, and the two main diagonals have the same sum, can be
constructed. For example, there is an infinity of such squares with
distinct prime numbers. Can an order- 3 square be made with nine
distinct square numbers? A few years ago I ofered $ 1 00 for such a
pattern. It has been proved that no 3 ? 3 magic square can consist
of distinct cubes or any higher powers, but no one has yet proved
impossibility for a "square of squares. " If it exists, its numbers will
be huge, perhaps beyond the reach of today' s supercomputers. Such
a square would be as useless as the magic hexagon. Why, then, are
mathematicians trying to find it? Because it might be there.
Every year or so I would devote a column to an imaginary inter
view with a numerologist I called Dr. Irving Joshua Matrix. (Note
the 666 provided by the letters of his names. ) The doctor would ex-
pound on unusual properties of numbers as well as bizarre forms of
word play. Many readers thought Dr. Matrix and his beautiful half
Japanese daughter Iva Toshiyori were real. I recall a letter from a
puzzled Japanese reader who told me that Toshiyori was a most pe
culiar surname in Japan. I had taken it from a map of Tokyo. My
informant said that in Japanese the word means "street of old men. "
I regret I have not asked the doctor hi s opinion on the recent pre
posterous bestseller The Bible Code. It concerns a cipher system Dr.
Matrix could have invented and applied, not just to the Old Testa
ment, but also to the New Testament, the Koran, to great epics of
poetry, or even to the absurd book in which it is detailed.
The last time I heard from Dr. Matrix he was in Hong Kong inves
tigating the accidental appearance of in well-known works of fiction.
He cited, for example, the following sentence in Part 2, Chapter 9, of
H. G. Wells' s War of the Worlds: "For a time I stood regarding . . . . " The
letters in the words give to six digits !
Postso-i pt
It would of course require a book to report on high points in the de
velopment of recreational mathematics during recent decades. I can
only refer readers to the books I mentioned, books I didn' t mention,
and the sixteen volumes that are collections of my Scientifc American
columns over a period of a quarter-century.
I cannot, however, resist adding a note about my final paragraph.
It is hard to believe, but Michael Keith, writing in Word Ways (August
1 999, pp. 1 89-90) bettered the H. G. Wells quote by finding two lit
erary sentences in which the word order gives to seven digits ! They
are:
"For I have a great sacrifice to . . . . " (King James Bible, I Kings
1 0: 19).
"And a pair 0' Hells wherewith to . . . . " (Rudyard Kipling' s poem
"Ballad of the Clamperdown, " line 4).
How Keith located these sentences boggles my mind.
Cc|pe 28
M|x. ~. ~ ' . . |e( 'q.| e,
<e.| q ` e, | ( T. | q ` e
It is easy to construct the largest rectangle that will ft inside a given
triangle. Simply draw lines from the midpoints of any two sides to
meet the third side perpendicularly as shown in Figure 1 . Rectangle
ABCD wi l l be the largest rectangle that wi l l go i nsi de the triangle.
There are three such rectangles (not necessarily the same shape) i n
any acute triangle. Right triangles have two, and obtuse triangles
have only one, which rests on the side opposite the obtuse angle.
It is not hard to show by calculus or algebra that the area of the
maxi mum rectangle is exactl y half the tri angle' s area, and the rectan
gle' s base i s half the triangle' s base. Thi s result goes back to Euclid' s
Book 6, Proposition 27. A neat way to demonstrate it is by paper
fol di ng. Cut the tri angle from paper, then fol d over the three corners
as shown in Figure 2, folding along the rectangle' s three interior sides.
Figure 1 . Constructing a maxi mum rectangle i n any triangl e.
This paper frst appeared in MCth Horizons (September 1997) .
204
/
/
/"
/
"
/
"
A G4rdner' s WorKout
Figure 2. Paper fold demonstration that the maxi mum rectangle i s half the tri an
gle's area.
The faps will ft snugly into the rectangle_ Thi s provides a simple
way to construct a maxi mum rectangle by paper fol di ng_
Paper foldi ng al so wi l l demonstrate that no other i nscribed rectan
gle can be larger. Try drawing one that is shorter and fatter, or taller
and thi nner. Fol d over the corners as before by fol di ng along the new
rectangle' s three sides. You' lI find that the faps either overlap each
other, or overlap the rectangle, showing that its area is less than half
that of the tri angle.
A question now ari ses. Could there be a l arger triangle entirely
inside-that is, one that docs not have a side resting on a side of the
triangle? ( See Figure +i . The answer is no, although proving it is not
so easy. You' lI fi nd proofs i n the publ i shed references l i sted at the end
Figure 3. An i nterior rectangle not inscribed.
28. M<xi mum I nscri bed Squ<res, Rect<ngi es, <nd Tri<ngi es 205
of this article. An informal demonstration can be made by paper fold
ing. Cut along the dotted line, then fold over the flaps by folding along
the rectangle' s sides. You' ll find the flaps overlapping the rectangle,
proving that their combined area exceeds the rectangle' s area.
It is not possible, by the way, for any rectangle to be inscribed in
any triangle without resting on one of the triangle' s sides. Do you see
why? A polygon is said to be "inscribed" in a larger polygon only if all
its vertices are on the perimeter of the circumscribed polygon. Apply
the pigeonhole principle. A rectangle has four corners but a triangle
has only three sides, therefore two corners must touch the same side.
When we seek the largest square that will go inside any given tri
angle, the task becomes more interesting. Like maximum rectangles,
such squares must rest on a side of the triangle. Call this the triangle' s
base. Let a represent the base and b the triangle' s altitude. The for
mula giving the side of the maximum inscribed square is wonderfully
simple:
ab
a + b
The late geometer Leon Bankof, in a letter cited here as a refer
ence, provided two algebraic proofs of this formula. The simpler one
makes use of the triangle shown in Figure 4 with its largest inscribed
square. Triangle Ade is similar to triangle ABC, so we can write
b - x b
x a
xb ab - ax
ax + xb ab
x( a + b) ab
ab
x
a
a + b
There are different ways to construct such maximum squares. One
of the simplest is shown in Figure 5 with respect to acute, right, and
obtuse triangles. Erect a square on the outside of the triangle' s base,
then draw lines from A and B to C. They will intersect the triangle' s
base at points that mark the base of the side of the largest square
resting on that base. (Another way to construct such squares can be
found in George Polya' s How To Solve It, 1 945, Part 1 , Problem 1 8) .
Unlike maximum rectangles i n triangles, maximum squares that
are on diferent sides of a triangle need not have the same area. On
206 A G4rdner' s Wor:out
B -
U ~
Figure 4, Algebraic proof that J - ,
( b+c)
the right i sosceles triangle in Figure 6, for example, the square shown
is slightly larger than the l argest square resting on the triangle' s hy
potenuse,
Figure 7 displays the largest square that goes inside an equilat
eral triangle of side 1 . It is slightly smaller than the largest inscribed
rectangle. On thi s tri angle the three maxi mum squares on the tri
angle' s three sides are, of course, equal . Is thi s the case only for the
equilateral triangle?
C'
I`
A "
Figure 5, Constructing the maximum interior square on a rriangle' s side,
207
Figure 6. Largest square in a right i soscel es trangl e.
Amazi ngly, the answer i s no. There i s j ust one other tri angle for
which thi s also is true. It is the obtuse i sosceles triangle shown in
Figure 8. I found i t i n The Book of Number, by John Conway and
Richard Guy ( 1 996), page 206, where it i s credited to Eugenio Calabi,
a geometer at the University of Pennsylvania. The ratio of base to
si de is 1 . 55 1 38 . . . , a number that is the positive value of in the
equation 2x3 - 2x2 - 3x + 2 o.
The formula ab/a +b, applied to the largest square within the equi
lateral tri angl e of si de I , gi ves the square' s side as J -3 2J -3 "
0. 46410161513 . . . Note this number carefully. We will meet it again in
a completely unexpected place.
How can one construct a triangle of largest area that will ft inside
a given rectangle? Simply draw lines from the corners of the rec-
Figure 7. Dimensions of the largest square in an equil ateral triangle of side I .
208 A G4rdner' s WorKout
Figure 8. The only triangle, aside fmm the equilateral. with three equal maximum
squares.
tangle' s base to any spot on the opposite side. There obviously i s an
uncountable i nfinity of such triangles. Like rectangles of the largest
area in any triangle, the largest triangles in any rectangle are each
hal f the rectangl e s area. We can' t demonstrate thi s by foldi ng, but if
you snip off the two small triangles at each side of a maximum trian
gle, they will fit neatly into the large triangle. Figure 9 shows how
thi s can be proved geometrically. Triangles and
f
and triangles b
and b' , obviously are congruent.
Consi der now the largest equil ateral tri angle that wi l l fi t wi thi n a
unit square. Slightly larger than such a triangle resting on a square' s
base, it i s shown i n Figure 1 0. To be of largest area its corers must
touch all four sides of the square, but since a square has four sides, one
of the triangle' s corers must be at a corer of the square. An informal
pr oof that thi s tri angle cannot be enlarged is to i magi ne it rotated
in either direction, keeping its lower left corner fixed and sliding the
Figure 9. Prof that the maxi mum-area tri angl e i nscribed in any rectangl e has an area
equal to half the ,ectangle s area.
28, M'xi mum l fscri bed 5'u'res. Rect'f:l es. 'fd Tri 'f:l es
\ F
' ,
------- C
= 2v3 - 3
" .464 1 .
2 -
.
U
Figure 10, The l argest equi l ateral triangle i nsi de a uni t square.
209
other two corners along the square' s other two sides. Obvi ously thi s
will make one side of the triangle shorter than the other two so the
triangle will no longer be equilateral. I do not know if a formal proof
by calcul us would be easy or hard.
The si de of the equil ateral triangle is the secant of 15 . which is
. - J = 1. 03527 . . . . We can now determine the triangle' s area by
finding its altitude ( ,6 - 3V " 0. 8965 . . . . ) then halving the prod
uct of alti tude and base, or by usi ng Heron' s formula for the area:
, s(s - a) (s - b) (s - l) where s is the semi perimeter (half the perime
ter of the triangle). and a, b, l are the triangle' s sides. The area of the
maxi mi zi ng tri angle turns out to be 2 V - 3, the same as the side of
the largest square that fits into an equilateral triangle of side I ! Has
thi s been noticed before? I found it extremely mystifyi ng until Pri nce
ton' s John Conway came up with a beautiful proof that I wiII explain
in a moment.
Abul Wefa ( A. D. 940998), a Persian geometer, gave five different
ways to construct thi s largest equil ateral triangle wi th compass and
straightedge. Three are reproduced in David Well' s Penguin Book
of Curious and Interesting Puzzles ( 1 992, Puzzle 38) . Wells refers
readers to Episodes in the Mathematics of Medieval Islam ( 1 986), by
J. L. Berggren, a book I have not seen.
21 0 A G4rdner' s WorKout
Figure 11. A pattern l i nking the largest square i n an equi l ateral triangle of side I with
the maxi mum equi l ateral in a unit square.
Henry Ernest Dudeney, in Puzzles and Curious Problems ( 1 93 1 ,
Problem 201 ) shows how to construct the triangle by folding a square
sheet of paper.
An easy way to draw the triangle (I do not know i f i t is one of We fa' s
methods) i s shown in Figure 1 1 . As science writer Barry Cipra dis
covered, the same lines that construct the largest square in the equi
lateral triangle also mark the points on the top of the larger square
28. M'xi mum l fscri bed 5'u'res. Rect'f:l es. 'fd Tri 'f:l es 211
. 3
Figure 12. John Conway' s dissection pmof of a curious equal ity.
that are the top corers of the two maximum equilateral triangles that
fit within the unit square. By drawing these two triangles we get a
lovely bilaterally symmetrical patter that shows how closely related
the two constructions are.
212 A Gard ner's Workout
I mentioned to Cipra the surprising fact that the area of the largest
equilateral triangle inside a unit square exactly equals the side of
the largest square inside a unit equilateral triangle. Cipra in turn
mentioned this to Conway. Almost at once Conway saw how to prove
the equality by a dissection. On the diagram of Figure 1 2 draw the
shaded parallelogram. Its area clearly is 1 times 2V 1, the side of
the square at the top. By slicing the parallelogram into three pieces
as shown, they will fit precisely into the equilateral triangle inscribed
in the bottom square!
Because the narrow-right triangles colored in light gray in Figure
1 1 have a combined area exactly equal to the area of the dark gray
right isosceles triangle, a pretty geometrical dissection problem arises.
What is the smallest number of pieces into which the two light gray
triangles can be cut so they will fit without gaps or overlaps into the
dark gray triangle? Clearly the number of pieces cannot be less than
four because each light gray triangle must be sliced at least once to
fit inside the dark gray triangle. A solution is possible in four pieces
( see Figure 1 3) .
Are there other interesting properties or puzzles based on the pat
tern of Figure I I ?
I wish to thank Don Albers and Peter Renz for many good sugges
tions that I have followed in writing this article.
Refrences
[ I ] Bankoff, Leon. Letter in Mathematics Teacher 79, May 1986, 322.
[2] Bird, M. T. "Maximum Rectangle Inscribed in a Triangle. " Mathematics Teacher
24, December 1 97 1 , 759-760.
[3] Embry-Wardrop, Mary. "An Old Maxi-Min Problem Revisited." Amercan Math
ematical Monthly 97, May 1 990, 421-23.
[4] Lange, Lester H. "What Is the Biggest Rectangle You Can Put Inside a Given
Triangle?" Mathematics Teacher 24, May 1 993, 237-240.
[5] Niven, Ivan. Maxima and Minima Without Calculus. Mathematical Association
of America, 1 98 1 .
28. Maxi mu m I nscri bed Squares, Rectangl es, and Triangl es 213
Figure 13. Solution to dissection problem.
Postsni pt
Solomon Golomb pointed out that it is not strictly accurate to say that
the area of the largest equilateral triangle inside a unit square exactly
equals the side of the largest square inside a unit equilateral triangle.
Reason: One is a linear measurement, the other an area. A more
precise way to put it is to think of 2. 3 as two equal ratios.
Tom M. Apostol and Mamikon Mnatsakanian, writing on "Trian
gles in Rectangles" (Math Horzons, February 1 998) explained a vari
ety of unusual theorems related to this chapter. For example, my Fig
ure 10 shows a special case of the following theorem: "An equilateral
triangle inscribed in, and having a common vertex with a rectangle,
cuts of three right triangles inside the rectangle. " Call the areas A,
B, and C, as shown in Figure 1 4. Areas A + B C. The article goes
on to state other related theorems.
Fi gure 14. A + B - C.
chqpter 29
Seri q l I sogons of 90 Deg rees
In 1988 the first author devised a computer program to search a unit
square grid for closed paths with the following properties. The path
starts along a lattice line with a segment of unit length, turns 90 de
grees in either direction, continues for 2 units, turns again in either
direction, continues for 3 units, and so on. In other words, the seg
ments of the path are in serial order 1, 2, 3, . . . N, with a right angle
turn at the end of each segment. A path of N segments-the number
is of course the same as the number of turns or corners-is said to be
a path of order N.
If the path returns to its starting point, making a right-angle with
its first segment, we call it a serial isogon of 90 degrees. The isogon
is allowed to self-intersect, to touch at corners, and to overlap along
segments. Think of it as a serial walk through a city of square blocks
and returning to the starting corner, or as the moves of a rook on a
large enough chessboard.
It is not obvious that such paths exist. However, with a little
doodling you will discover the unique isogon of order 8, the lowest
order a 90-degree isogon can have. It outlines a polyomino of 52 unit
squares that, as Figure 1 shows, tiles the plane. Indeed, it satisfies
the "Conway criterion" [ 1 ] for identifying tiling shapes. The boundary
of the polyomino can be partitioned into six parts, the first and fourth
of which (AB and ED in Figure 2) are equal and parallel, while the
other four ( BC " 3 ; CD " 4, 5, 4; EF "7; FA " 1 , 8, 1) each have
rotational symmetry through 1 800 about their midpoints (black dots
in Figure 2). It may be the only plane-tiling polyomino with a serial
boundary, but we are unable to prove this.
Puzzle: Can you tile this polyomino with thirteen L- shaped tetro
minoes?
This article, coauthored with Lee Sallows, Richard Guy, and Donald Knuth, frst
appeared in Mathematics Magazine (Vol . 64, December 1 991 ) .
21 5
Fi gure 1. The only known serial-sided polyomino that tiles the plane.
It is easy to prove that N must be a multiple of 4. One way is to
consider N rook moves on a bicolored chessboard. Assume without loss
of generality that the rook begins on a black cell and makes its first
move horizontally. To close the path, its final move must be vertical,
and end on a black square. Because moves of odd and even length
6
D
4
5
c
4
2
EL. -----Fr
8
Fi gure 2. Applying Conway' s criterion to prove that the polyomino tiles the plane.
29. Seri <! i sogons of 90 Degrees 217
alternate, the sequence of colors at the end of each segment forms the
repeating sequence: WW BBWW BB . s . . The rook returns to a black
cell , after a vertical move, if and only if the number of moves 0
mod 4.
Experimenting on graph paper will quickly convince you that there
is no serial isogon of order 4. (You might call this "a 4-gon conclusion". )
We have exhibited one of order 8. After the computer program exhaus
tively plotted all serial isogons of orders through 24, a surprising fact
emerged. No serial isogons were found except when N is a multiple
of 8. This led to several proofs of the following theorem:
For any 90-degree seral isogon, N must be a multiple of 8.
Assume that a closed path begins with a unit move to the east,
and that moving east or north is positive, and moving south or west is
negative. A path can be uniquely described by placing a plus or minus
sign in front of each number in the sequence of moves to indicate the
direction of the move. For example, the order-8 polyomino has the
following formula:
+1 + 2 3 4 5 6 + 7 + 8.
It is obvious that if the path closes, the sum of all horizontal
moves-the odd numbers-must be zero, otherwise the path will not
return to the vertical lattice line that goes through the starting point.
Similarly, the sum of all vertical moves-even numbers-must be zero
or the path will not return to the horizontal lattice line going through
the origin point. The sum of all the numbers will, of course, also
be zero.
We know that N is a multiple of 4, say 4k. Then the north-south
moves are the even-length ones, 2, 4, .
.
. , 4k. The total north-south
distance is therefore 2 ( 1 + 2 +
.
.
.
+ 2k) 2k( 2k + 1 ) . Half of this,
k( 2k + 1) , must be north and half of it south. But if k is odd, this
distance is odd, and cannot be the sum of even-length moves.
We can make this clearer by taking N 1 2 as an example. Even
numbers in this path' s formula ( 2, 4, 6, 8, 10, 12) add to 42. If the for
mula describes a closed path, the sum of the positive numbers in this
sequence must equal 42/2 2 1 . But no set of even numbers can add
to the odd number 21 . Consequently, no formula can be constructed
that will describe a closed path of order 1 2.
With reference to the grid, this tells us that i f N i s a multiple of
4 but not of 8, the last segment of the path, which is vertical, cannot
return to the horizontal lattice line that goes through the path' s origin

I

7 W
|
Figure 3. A bicoloring of the grid for proving that N is a multiple of 8 for all orders of
closed serial path. The order-8 polyomino is shown in outline.
point because the sum of the positive segments going north cannot
equal the absolute sum of the negative segments going south. The
path' s end will always be an even number of units above or below the
zero horizontal line.
Is there a coloring pattern that proves the N 0 mod 8 theorem?
Yes, the simple coloring shown in Figure 3 (found by the second au
thor) will do the trick.
Start a path on any black cell, then make your first move hor
izontally in either direction. Regardless of your choices of how to
turn at the end of each segment, the colors at the ends of segments
will endlessly repeat the sequence: WWWWWWBB, WWWWWWB,
WWWWWWBB, . . . . The path will enter a black cell (which it must
do if the path closes) in a direction perpendicular to the first segment
if and only if the number of moves is a multiple of 8. Shown in the
illustration is the path that outlines the order-8 tiling polyomino.
Although all closed paths have formulas in which the total sum of
the signed terms is zero, this is not suficient to produce a formula
for a path. A serial path closes if and only if the even numbers in its
formula add to zero, and likewise the odd numbers. (Nonserial closed
paths may meet this proviso, and be of any order of 4 or greater that
is a multiple of 2. The formula +1 + 2 1 2, for example, describes
a closed path that outlines a domino. ) If this proviso is not met, the
formula gives the location of the path' s final point with respect to the
origin. If the sum of the signed even numbers is positive, it gives the
number of units where the path ends above the origin; if negative, it
gives the number of units below the origin. Similarly for the sum of
the signed odd numbers. If positive, it gives the end point' s distance
east of the origin; if negative, the distance west.
If the signs of all even numbers are changed, or if the signs of all
odd numbers are changed, it reflects the isogon along an orthogonal
axis. If all signs are changed, it rotates the isogon 1 80 degrees.
We have shown that N 0 mod 8 is a necessary condition for
a closed serial path. Is it also suficient? Yes. Here is one way to
arrange plus and minus signs in a formula that will always describe
a serial isogon: Put plus signs in front of the first and last pairs of
numbers. Put minus signs in front of the next to last pairs of numbers
at each end, and continue in this way until all pairs of numbers are
signed. This ensures that all even numbers add to zero, and likewise
all odd numbers, therefore the formula must describe a closed path.
It produces, for instance, the unique formula for N 8. Applied to
N 1 6 it gives the formula +1 + 2 3 4 + 5 + 6 7 8 - 9 - 1 0 + 1 1 +
1 2 - 13 - 14 + 15 + 16, which describes the isogon at position 0
2
E
5
in
Figure 4.
Is there a procedure guaranteed to construct a serial isogon for
any order N 0 mod 8 that outlines a polyomino? The answer is
again yes. Each formula has 8n numbers. If we make positive all
numbers in the first fourth of the formula, and also in the last fourth,
and make negative all numbers in the half in between, we produce a
serial isogon. Applied to N 16, it gives +1 + 2 + 3 + 4 - 5 - 6 - 7 - 8 -
9 - 1 0 - 1 1 - 1 2 + 13 + 14 + 15 + 16, which describes the polyomino at
Ol E
4
in Figure 4. The polyomino generated by this procedure always
takes the form of a snake that grows longer as N increases. Figure
5 shows the polyomino snake for order 32. A diagonal line, running
from the extreme corners of the snake' s head and tail, going between
all the interior corners, is almost, but not quite, straight.
We now turn to the more dificult task of enumerating all possible
serial isogons (not counting rotations and reflections as diferent) for a
given isogon. As mentioned earlier, there is only one isogon of order 8,
the tiling polyomino. For N 16, the computer program found the 28
solutions shown in Figure 4. Note that only three ( OI EI , OI E
4
, OI E7)
are polyominoes. For N 24, the program produced 2, 1 08 distinct
Figure 4. The 28 distinct order- 1 6 serial i sogons of 90
.
isogons, of which 67 bound polyominoes. For N 32 the program' s
running time became too long to be feasible.
No formula is known for enumerating all distinct serial isogons of
order N, or for counting the polyominoes of a given order. However,
there are procedures by which the number of isogons can be counted
by hand to a value of N that goes well beyond N 24.
Here is how the fourth author describes one such procedure:
Suppose N 16; it' s easy to see that this quantity is the constant
term if you expand the algebraic product
into powers of J and y. To get the number of ways for the odd sum to
cancel, we want the constant term of
which is the coeficient of xH
3
+
5
+
, , +1
5
x
6
4
in
which is the coeficient of x3
2
in
29. Seri 'i [ so
g
ons of 90 De
g
rees 221
Figure 5. The snake polyomino of order 32. The diagonal l i ne separating the snake' s
two sides i s al most. but not quite. straight.
which is 8. To get the number of ways fOl- the even sum to cancel, we
want the constant term of
which is the constant term of
which is the coefficient of
y1
+2+
. .
~C :l:6
in
which is the coefficient of yl 8 in
which i s 1 4. So the total number of closed paths is 8 x 14; divide by
4 to get 28 closed paths that are distinct under refectional symme
try. Suppose we start with +1 + 2; then the four ways to do the odd
numbers are
222
01 +1 + 3 - 5 - 7 - 9 - 1 1 + 13 + 15
O2 " + 1 - 3 + 5 - 7 - 9 + 1 1 - 13 + 15
0
3
+ 1 - 3 - 5 + 7 + 9 - 1 1 - 13 + 1 5
04 +1 - 3 - 5 + 7 - 9 + 1 1 + 13 - 15
A G<rd nds Workout
and the seven ways to do the even numbers are
El +2 + 4 + 6 + 8 - 1 0 - 12 - 14 + 16
E
2
+2 + 4 + 6 - 8 + 10 - 12 + 14 - 16
E
3
+2 + 4 - 6 + 8 + 10 + 12 - 14 - 16
E4 +2 + 4 - 6 - 8 - 1 0 - 12 + 14 + 16
E
5
+2 - 4 + 6 - 8 - 10 + 12 - 14 + 16
E
6
+2 - 4 - 6 + 8 + 10 - 12 - 14 + 16
E7 +2 - 4 - 6 + 8 - 1 0 + 12 + 14 - 16
The three serial polyominoes are the snake 01 E4 and two other
solutions Ol El , 01 E7. (It' s curious that only 01 can be completed. The
case 02 El almost works, but that path gives a degenerate polyomino
whose width is zero at one point. Paths 0
3
El , 0
3
E
4
, and 04E
4
fail in
the same way. )
In general when N Sn, the number of closed paths is the product
of the coefficient of x
8
n2 in
and the coeficient of
y
4n2
+
2
in
These numbers, for small n (divided by 2 to remove symmetry) , are
N n odds/2 evens/2 product
8 1 1 1 1
1 6 2 4 7 28
24 3 34 62 21 08
32 4 346 657 227322
40 5 3965 7636 30276740
48 6 48396 93846 4541 77 1 01 6
56 7 61 5966 1 1 99892 739092675672
64 8 8082457 1 5796439 1 27674038970623
It seems certain that the vast majority of these isogons will not
bound polyominoes. The paper of Bhattacharya and Rosenfeld [ 3] is
concerned with the problem of avoiding self-intersections in isogons:
They treat the general problem in which the sides are of arbitrary
length, not just our particular case of consective integers.
Here is how the third author has made an asymptotic estimate of
the number of serial isogons:
We have seen that the number of isogons of a given order is the
product of half the number of possible choices of sign in 2 4
6 . . . (8n - 2) 8n 0 with half the number of choices of sign in
1 3 5 . . . (8n - 3) 8n O.
The first of these is the number of partitions of half of the sum
2 + 4 + 6 + . . . + 8n
into distinct even parts, of size at most 8n, i . e. , the number of parti
tions of n( 4n + 1 ) into d distinct parts of size : 4n. Subtract 1 , 2, .
.
. , d
from these parts, now no longer necessarily distinct, of size : 4n - d.
We require the number of partitions of 4n
2
+ n - ( 1 /2) d( d + 1 ) into at
most d parts, no longer necessarily distinct, of size : 4n-d. Here d, the
number of parts, lies in the approximate range (4 - 2V) n < d < 2Vn.
In the same way the second number is the number of partitions
of 4n
2
- ( 1 /2) d
2
into at most d parts, not necessarily distinct, of size
: 4n - d, with d in approximately the same range as before, but with
d necessarily even!
The main contribution comes from d 2n and the distribution, as
we shall see, is essentially the binomial distribution, so that a good
estimate of the whole is obtained by multiplying this central term by
. 47n, except that we halve the "odd" estimate since only alternate
terms (d even) are taken.
From formula (75) in [2] we lea that the number of partitions of
j into at most a parts, with each part :: b, is asymptotically equal to
_1 ( a + b )
(j - ab/2 )
,
a, b
a
a, b
where
b
ab ( a + b + 1) /12 and

(x) e-x
2
/2 /y.
In bo
t
h the even and the odd cases, a d, b 4n - d, a + b + 1
4n + 1 , and the central term is given by d 2n a b for which
j 4n
2
+ n - n( 2n + 1 ) and j 4n
2
- 2n
2
, i. e. , 2n
2
in either case, and
j - ab/2 0, so that the central term is asymptotically equal to
1 v 4n
y 2nJ4n + 1 2n
i. e. , asymptotically equal to
v
24n
2nyJ 4n + 1 J 47n
by Stirling' s formula. Multiply by J 47n to estimate the total number
of partitions in the even case, and by half of that in the odd case. The
product of the halves of these numbers (i. e. , not counting the E. -W.
or N. -S. reflections of the isogons as different) is thus
Compare this estimate with the actual values obtained above.
n
1
2
3
4
5
6
7
8
3
X 2
8
n
-6
/7n2 (4n + 1 )
0. 76
27. 2
2140
235604
3 1 248698
466647228 1
75661 8728785
1 30321 844073 1 00
Figure 6. The unique, smallest examples of serial isometric isogons. At top is the only
order-9 isogon with 60-degree angles. In the middle is the only order- 1 2 isogon with
l 20-degree angles. At bottom, shown tiling the plane in two ways, is the only order-5
isogon that mixes the two angles.
The first author has since explored serial isogons on isometric
grids. They are of two types: those with 60-degree angles, and those
with l 20-degree angles. In the 60-degree case, serial isogons exist if
and only if N 9 and N 1 mod3. In the l 20-degree case, they exist
if and only if N is a multiple of 6. If each angle of a serial polygon
can be either 60 or 1 20 degrees, such polygons exist for any N that
is 5 or greater. In all three cases, the smallest order is unique and is
a poJyiamond-a polygon formed by joining unit equilateral triangles.
The three are shown in Figure 6. Note that the order-5 polyiamond
tiles the plane in two different ways, see [4] .
226
T FI OI U I RI T
E
TIW E L viE E
L
E
N TI E N
E
E JG H
E N
F
N Z I E R
o TI W
N
0
F Dl u l R
I
V
s I I X E
S
E
T N
A G<rd nds Workout
0
T
E
Figure 7. A 1 6-sided polyomino, its sides determined by the words for zero through 1 5.
For a time the first author believed that closed serial paths existed
only when angles are 60, 90, or 1 20 degrees, but then he suddenly
discovered such a path with 1 08-degree angles, the angles of a regular
pentagon. More recently Hans Cornet, a retired mathematician in
The Hague, has put forward a proof that at least one serial isogon
can be constructed for every angle a that is a rational multiple of 360
degrees, that is, when C 27n/m radians, with m and n positive
integers; see [4] .
The first author has also investigated closed paths on square and
isometric grids that have segment lengths in sequences other than the
counting numbers, such as Fibonacci sequences, consecutive primes,
and so on. He has produced a whimsical class of piominoes-polyomin
oes whose sides in cyclic order are the first n digits of 7 with zeros
omitted. Because the digits of 7 are pseudo-random, the task of enu
merating 7-isogons of 90 degrees is related to determining the prob
ability of self-intersecting random walks on a square lattice. He has
even experimented with closed paths based on the letters of number
words. An example is shown in Figure 7. It is unquestionably one of
the most useless polyomino outlines ever constructed, yet does it not
have a curious charm?
Ltetmin
Figure 8. A coloring patter on the tilling poliomino for proving it cannot be dissected
into 13 L-tetrominoes.
Solution to Puzzle: To prove that 1 3 L-tetrominoes will not tile
the polyomino, divide the polyomino into alternatively colored vertical
stripes as shown in Figure 8. No matter how an L-tetromino is placed
within this pattern, it will cover an odd number of cells of each color.
Thirteen odd numbers add to an odd number, but the polyomino has an
even number of cells of each color. The tiling is therefore impossible.
Postsci pt
The appearance of a prepublication version of this paper [4] resulted
in a good deal of correspondence and misunderstanding. John Leech
observed that the argument there, showing that N is a multiple of
8, is faulty and exacerbated by the statement that "even numbers
hardly entered into the discussion," whereas we know that it is the
even lengths that clinch the matter. Leech also gives an argument in
volving those gridlines that "quarter" a tetromino, which is equivalent
to that of Figure 8. The difculty of printing a complicated formula in
a popular article resulted in an oversimplification that was no longer
an asymptotic formula in the sense described. Finally, it was implied
that calculation of the exact numbers of serial isogons was more difi
cult than is actually the case, which prompted several people to reach
for their computers. Ilan Vardi [6] has explained how to compute the
numbers rapidly via the Chinese Remainder Theorem, and he lists
the values for N 400 and N 1 000. Calculations up to N 200
were carried out by Sivy Farhi , and by Pierre Barnouin, who gave the
following values:
16 28
24 2108
32 227322
40 30276740
48
4541 771016
56 739092675672
64
127674038970623
72 23085759901 610016
80 4327973308197103600
88 835531 767841 066680300
96 165266721954751 746697155
104 333641 81 61 6540879268092840
1 1 2 685401 741 60982278361 06023048
120 14293682585863432461 84813682344
128 302023498629081603279538134332922
136 64557914743374337032608546756101824
144 139412471 2589399758445771 127308712231 0
1 52 3038225349257507092516361163813831321438
160 6675754757919568321 916769534550748349821 00
168 1477737884739369237247153822487269905821 50405
1 76 32931659242107964657022264548538525142956914056
184 73839877297802965850639446296210651 23001927478725
192 1 664961555710273709724126262313969341483976633058164
200 377359709872056562198423857053288570232577607987443492
ReFetences
[ 1 ] Doris Schattschneider, "Will It Tile? Try the Conway Criterion! " Mathematics
Magazine 53 ( 1 980), 224-232.
[ 2] Lajos Takacs, "Some Asymptotic Fonnulas for Lattice Paths, " Journal of Statis
tical Planning and Inference 14 ( 1 986), 1 23-1 42.
[ 3] Prabir Bhattacharya and Azriel Rosenfeld, "Contour Codes of Isothetic Polygons,"
Computer Vision, Graphics, and Image Prcessing 50 ( 1 990) , 353-363.
[4] Lee Sallows, "New Pathways in Serial Isogons, " The Mathematical Intelligencer
1 4. ( 1 992), 56-67.
[5] A. K. Dewdney, "Mathematical Recreations: An Odd Journey Along Even Roads
Leads to Home in Colygon City," Scientific Amercan 263 #1 (July 1 990), 1 1 8-1 21 .
[6] Ilan Vardi, Computational Recreations in Mathematics, Addison-Wesley, Read
ing, MA, 1 991 , Section 5. 3.
ch'pter 30
Arou nd the Sol ' l S
y
stem
In this article we examine a fantastic mathematical magic trick. I will
present it as a puzzle-why does it always work?-but of course you
can demonstrate it to friends as an amazing feat of ESP.
To the audience, this is how the trick appears. While your back
is turned, someone is asked to put a dime on any of the nine squares
shown on the facing page. Without turning around, you give instruc
tions for moving the dime about at random over the matrix, as if it
were a spaceship touring the solar system. As these random moves
are made, you keep blocking of certain cells by directing that pen
nies be placed on them. Finally, eight cells are occupied by pennies.
With your back still turned, you can name the planet on which the
"spaceship" came to rest.
Pause at this point and see that you have on hand a dime and
eight pennies. Instead of pennies you can use buttons, checkers, or
anything else that will serve as counters. I will now assume the role
of magician while you assume the role of spectator.
Select any one of the nine cells and put the dime on it. This is a
completely free choice on your part, and obviously I have no way of
knowing what choice you made. When you move the dime according to
my instructions, you must move it one cell at a time in any horizontal
or vertical direction. No diagonal moves are allowed. At each move
you spell a letter in the name on the cell where you first put the
dime. For example, if you stat on Mars you spell M-A-R- S, moving
the dime one square east, west, north, or south at random, one move
for each letter.
When you finish spelling the name on the starting square, put
a penny on Venus. I am, of course, betting that no matter where
you began, or how you moved the dime, it will not have come to rest
on Venus.
This article i s reprinted from my Riddles of the Sphinx ( 1 988) .
231
232
Neplune ,
Figure 1 .
A Gard ner's Workout
From now on, at each step of your "tour" of the solar system, you
move the "spaceship" just seven times, regardless of the name on the
cell. These moves are made randomly, as before, but are confined to
unoccupied cells. The number of these vacant squares will become
fewer and fewer as more and more pennies go on the matrix.
After making seven moves, put a penny on Mars.
Move seven times. Put penny on Mercury. As New York City' s
former mayor Ed Koch likes to say, "How'm I doin' ?" Are all the
pennies landing on vacant cells?
Move seven times. Put a penny on Uranus.
Move seven times. Put a penny on Neptune.
Move seven times. Put a penny on Saturn.
Move seven times. Put a penny on Jupiter.
Move seven times. Put a penny on the Moon.
If you followed instructions correctly, the dime should now be on
Pluto!
When you show the trick to someone, turn your back while you
give the above instructions. If you like, you can add to the mystery
by allowing the spectator, at any time, to move by spelling S-E-V-E-N
instead of counting seven. After he puts a penny on the moon, you can
tell him, without turning around to look, that his dime is on Pluto.
30. Around the Sol ar System 233
Why does the trick always work? The answer will introduce you
to the concept of "parity. " It is a concept of enormous importance both
in combinatorial mathematics and in modern particle physics.
Answer: Note that the names on the shaded cells have an odd num
ber of letters and those on the unshaded cells have an even number.
Mathematicians say that the two sets of squares, along with their
names, are of opposite parity. One has even parity, the other odd.
Each time the dime moves to an adj acent cell it changes parity.
If you start the dime on any cell and move to spell the letters of
the name on that cell, the coin is sure to end on an un shaded cell.
Because the dime has now acquired even parity, all shaded cells must
be unoccupied, so it is safe to request that a penny be placed on shaded
Venus.
From now on, at each step, the dime moves an odd number of times.
It makes no diference whether it is moved seven times (or any other
odd number), or moved to spell S-E-V-E-N, or any other word with an
odd number of letters. If some spectator' s first or last name has an
odd number of letters, you can use his or her name for spelling. Every
time the dime is moved, its parity alters. This allows you to direct
that a penny be placed on a vacant cell of a parity opposite that of
the dime. After eight steps, the only unoccupied cell will be the moon,
with the dime resting on Pluto.
If you want to repeat the trick with a different final result, you' ll
have to work out a different set of instructions. With suitable instruc
tions you can, of course, cause the dime to end on any of the shaded
squares. Be careful, though, to eliminate the cells in such an order
that the moving "spaceship" always has access to all the remaining
unoccupied cells.
chqpter 31
Ten Amqzi n
g
Mqthemqti cq l Tri cks
Magicians have invented a fantastic variety of self-working mathe
matical tricks in which the outcome of seemingly random choices can
be predicted in advance. Some of these tricks are described in my
1 956 Dover paperback Mathematics, Magic, ad Mystery, but hun
dreds of new mathematical tricks have been invented since that book
was published. What follows is a small selection.
Readers are urged to follow the instructions of each trick carefully,
then check the predictions at the end of this chapter. I will not spoil
the fun by explaining why each magic trick works. If you can figure
this out for yourself, you'll find the tests pleasant exercises in mathe
matical problem solving.
1 . Twi h kl e Twi h kl e
Twinkle, twinkle, little star,
How I wonder what you are
Up above the world so high,
Like a diamond in the sky.
Twinkle, twinkle, little star,
How I wonder what you are.
Select any word in the first two lines of the above familiar poem.
Count the number of its letters. Call this number n. Now count ahead
n words, starting with 1 for the word following the word you selected.
Count the number of letters in this second word. Call the number k.
Count ahead k words to arrive at a third word.
This article frst appeared in Math Horzons (September 1998) .
235
Continue in this manner until you can' t go any farther.
On what word does your count end?
2. 24691 3578
Enter the above strange number in your calculator. You may feely
choose to do any of the following:
Multiply the number by 2, 4, 5, 7, 8, 10, 1 1 , 1 3, 1 6, 20, 22, 25, 26,
3 1 , 35, 40, 55, 65, 125, 1 75, or 875 .
Or, if you prefer, divide the number by 2, 4, 5, or 8.
After you have done one of the above, rearrange the digits of the
result in serial order from the smallest digit to the largest. Ignore any
zeros among the digits.
The result will be a number of nine digits.
What is this number?
3. Tty Th i s on q Dol l q t Bi l l
Write down the serial number of any dollar bill. Scramble the digits
any way you like-that is, mix up their order. Jot down this second
number.
Using your calculator, subtract the smaller number from the larger.
From the difference, subtract 7.
Copy the digits now on display, then add them all together. If the
sum is more than one digit, add the digits once more.
Keep adding digits in the sum until just one digit is obtained.
What is it?
4. Cou nt the Mqtches
From an unused folder of 20 paper matches, tear out any number of
matches less than ten and put them in your pocket.
Count the number of matches remaining in the folder. Add the
two digits of the count, then remove that number of matches fom the
folder. Put them in your pocket.
Tear out three more matches.
How many matches are left in the folder?
31 . Ten Amazi ng Mathemati cal Tri cks 237
5. A Test with Two Di ce
Roll a pair of dice on the table. Call them A and .
Write down the following four diferent products:
1 . The product of the top numbers on the dice.
2. The product of their bottom numbers.
3. The product of the top of A and the bottom of
+ The product of the top of and the bottom of A.
Add the four products. What' s the sum?
6. Fol d q hd Tti m
Fold a sheet of paper in half four times, and then unfold it. The creases
will form a +? +matrix of cells as shown in Figure 1 . Number the cells
from 1 through 16 as illustrated. Fold each crease forward and back
a few times so the paper will fold easily either way along each crease.
Now fold the sheet into a packet the size of one cell. You can make
the folds as tricky as you please, folding any way you like. You may
even tuck folds between folds. In other words, make the folding as
random as possible until you have a packet the size of a single cell .
With scissors trim away all four edges of the packet so that it
consists of sixteen separate pieces. Spread these pieces on the table.
1 2 3 4
5
6
7 8
9 1 0 1 1 1 2
1 3
1 4 1 5 1 6
Figure 1 .
Some will have their number side up, others will have their number
side down.
Add all the numbers on the face-up pieces.
What is the sum?
7. At the Apex
Copy the triangle of circles in Figure 2.
Put any four digits you like in the four vacant circles of the bottom
row. They needn' t be all diferent, and you may include one or more
zeros if you like.
The remaining circles are filled with digits as follows: Add two
adj acent pairs of numbers, divide the sum by 5, and put the remainder
in the circle just above the adj acent pair of numbers.
For example, suppose two adj acent numbers in the bottom row are
6 and 8. They add to 14. Dividing 14 by 5 gives a remainder of 4, so
you put 4 directly above the 6 and 8. If there is no remainder ( such
as 6 + 4 * 1 0) then put a zero above the 6 and 4.
Continue i n this way, going up the triangle, until all the circles
have digits.
What digit is at the apex of the triangle?
Figure 2.
31 . Ten Amazi ng Mathemati cal Tri cks 239
8. The Red a hd the Bl ack
Shuffle a deck of cards, then deal 30 cards to the table to form a pile.
Count the number of black cards in the pile. From this number
subtract the number of red cards in the rest of the deck.
What' s the diference?
9. Fow Fi l e Ca tds
Write 39 on one side of a file card, and 51 on the other side. On a
second file card write 26 and 34 on the two sides. A third card gets
65 and 85 . A fourth card gets 52 and 68.
Place the four cards on the table so the numbers showing on top
are 26, 39, 52, and 65.
Slide any card out of the row, then turn over the three remaining
cards. Slide out another card, and turn over the remaining two cards.
Slide out a third card. Turn over the card that remains.
You now have a choice of leaving the cards as they are, or turning
all of them over.
With your calculator, multiply all the numbers showing. What is
the product?
1 0. A 3 ? 4 Test
Copy the 3 ? 4 matrix in Figure 3.
+
Figure 3.
Put digits 1 through 9 in the cells in any way you like. Three
empty cells will remain. In those cells put either three ones, or three
fours, or three sevens.
Treat each row of the matrix as a 3-digit number. Add these four
numbers by writing the sum over the four lines below the matrix.
Add the sum' s four digits. If the result is more than one digit, add
those two numbers. Keep adding until only one digit remains.
What is this digit?
Pted i cti ans
1 . The count ends on the word "you. " The trick i s based on what ma
gicians call the Kruskal Count, a principle discovered by Prince
ton mathematician Martin Kruskal .
2. The number is 1 23456789.
3. The final digit is 2.
4. Six matches remain in the folder.
5. The sum of the four products is 49.
6. The face-up numbers add to 68.
7. The triangle' s top number is 4.
8. The difference between the black and red cards is 4.
9. The product of the four numbers is 5860920.
10. The final digit is 3.
Ch'pter 32
Model i ng M'them'ti cs with
pl '
y
i ng C' rds
Because playing cards have values 1 through 1 3 (acks 1 1 , queens 1 2,
kings 1 3) , come i n two colors, four suits, and have fronts and backs,
they provide wonderfully convenient models for hundreds of unusual
mathematical problems involving number theory and combinatorics.
What follows is a choice selection of little-known examples.
One of the most surprising of card theorems is called the Gilbreath
principle, after magician Norman Gilbreath who first discovered it.
Arrange a deck so the colors alternate. Cut it so the bottom cards of
each half are different colors, then rife shufe the halves together.
Take cards from the top in pairs. Amazingly, every pair will consist
of a red and black card!
Here is a simple proof by induction. Assume that the first card to
fall on the table during the shufe is black. If the next card to fall
is the card directly above it in the same half, that card will be red.
This places on the table a redfblack pair. If the next card after the
first one comes fom the other half, it too will be red to put a redlblack
pair on the table. In either case, after two cards have dropped, the
bottom cards of each half will be of diferent colors, so the situation
is exactly the same as before and the same argument applies for the
rest of the cards. No matter how careful or careless the shufe, it will
pile redlblack pairs on the table.
Gilbreath' s principle generalizes. Arrange the deck so the suits
are in an order, say spades, hearts, clubs, diamonds, that repeats
throughout the pack. Deal as many cards as you like to form a pile.
This of course reverses the order of the suits. When the pile is about
the same size as the remaining portion of the deck, rife shufe the
This article frst appeared in The College Mathematics Joural (Vol. 31 , May 2000) .
241
two portions together. If you now take cards in quadruplets from the
top of the shufed pack you will find that each set of four contains all
four suits.
The ultimate generalization is to shufe together two decks, one
with its cards in the reverse order of the other deck. After the shufe,
divide the 1 04-card pack exactly in half. Each half will be a complete
deck of 52 diferent cards !
What mathematician David Gale has called the "non-messing-up
theorem" is another whimsical result. From a shuffled deck deal the
cards face up to form a rectangle of any proportion. In each row re
arrange the cards so their values do not decrease from lef to right.
In other words, each card has a value higher than the one on its lef,
or two cards of the same value are side by side.
After ordering the rows, do the same thing with the columns. This
of course drastically alters the order of cards in the rows. After re
arranging the columns, you may be amazed to find that the rows are
still ordered!
The theorem is at least a hundred years old. You will find it proved
as the answer to a problem in American Mathematical Monthly, Vol.
70, February 1 963, pp. 21 2-1 3, and in a monograph by Gale and
Richard Karp, published in 1 97 1 by the operations research center
of the engineering school of the University of California, Berkeley.
Donald Knuth discusses the theorem in the third volume of The Art of
Computer Programming in connection with a method of sorting called
"shellsort. " In my The Last Recreations, Chapter 1 1 , I describe a clever
card trick based on the theorem.
Is it possible to arrange a deck so that if you spell the name of
each card by moving a card from top to bottom for each letter, then
turning over the card at the end of the spell and discarding it, it will
always be the card you spelled? For example, can you so arrange the
cards that you can first spell all the spades, taking them in order from
ace through king, then do the same thing with the hearts, clubs, and
diamonds?
You might imagine it would take a long time to find out how to
arrange the deck, assuming it is possible to do so, in a way that permits
the spelling of all 52 cards. Actually, finding the order is absurdly
easy. First arrange the deck from top down in the order that is the
reverse of your spelling sequence. Take the King of Diamonds from
the top of the deck, then take the queen, place it on top of the king, and
spell "Queen of Diamonds" by moving a card at each letter from bottom
32. Model i ng Mathematics with Pl ayi ng Cards 243
to top. In brief, you are reversing the spelling procedure. Continue
in this way until the new deck is formed. You are now all set to spell
every card in the predetermined order. Of course you can do the same
thing with smaller packets, such as the thirteen spades, or with cards
bearing pictures, say of animals whose names you spell.
Remember the old brain teaser about two glasses, one filled with
water, the other with wine? You take a drop of water, put it into the
wine, stir, then take a drop of the mixture, move it back to the water,
and stir. Is there now more or less wine in the water than water in
the wine?
The answer is that the two quantities are exactly equal . The sim
plest proof is to realize that, after the transfers, the amount of liquid
in each glass remain the same. So the quantity missing from the wa
ter is replaced by wine, and amount of wine missing from the other
glass is replaced by the same amount of water.
This is easily modeled with cards. Divide the deck into two halves,
one of all red cards, the other of all black. Randomly remove n red
cards, insert them anywhere in the black half, and shufe. Now ran
domly remove n cards fom the half you just shuffled, put them back
among the reds, and shuffle. Inspection will show that the number
of black cards in the red half exactly equals the number of red cards
in the black half. It doesn' t matter in the least if the red and black
portions are not equal at the start.
Closely related to this demonstation is the following trick. Cut a
deck exactly in half, turn over either half and shufe the two parts
together. Cut the mixed-up deck in half again, and turn over either
half. You' ll find that the number of face-down cards in either half
exactly equals the number of face-down cards in the other half. The
same is true, of course, for the face-up cards. Do you see why this is
the case? The trick is bafing to spectators if they don' t know that
the deck is initially divided exactly in half, and if you secretly turn
over one half as you spread its cards on the table.
Playing cards provide a wealth of counter-intuitive probability ques
tions. A classic instance involves three cards that are face down on
the table. You are told that one card only is an ace. Put a finger on
a card. Clearly the chances you have selected the ace is 1/3. A friend
now secretly peeks at all three cards and turns face up a card that is
not the ace. Two cards remain face down, one of which you know is
the ace. What now is the probability your finger is on the ace? Many
persons, including mathematicians who should know better, think the
probability has risen from 113 to 112. A little refection should convince
you that it remains 113 because your fiend can always turn a non-ace.
Now switch your finger from the card it is on to the other face-down
card. It may be hard at first to believe, but the probability you have
now chosen the ace jumps from 1/3 to 2/3 ! This is obvious from the
fact that the card you first selected has the probability of 1/3 being
the ace. Because the ace must be one of the two face-down cards, the
two probabilities must add to 1 , or certainty.
A similiar-seeming paradox also involves three face-down cards
dealt from a shuffled deck. A friend looks at their faces and turns
over two that are the same color. What' s the probability that the re
maining face-down card is the same color as the two face-up cards?
You might think it is 112. Actually it 114. Here' s the proof. The proba
bility that three randomly selected cards are the same color is two out
eight equal possibilities, or 114. Subtract 114 from 1 (the card must
be red or black) and you get 3/4 for the probability that the face-down
card difers in color from the two face-up cards. This is the basis for
an ancient sucker bet. If you are the operator, you can ofer even odds
that the card is of opposite color from the two face-up cards, and win
the bet three out of four times.
Here' s a neat problem involving a parity check. Take three red
cards from the deck. Push one of them back into the pack and take
out three black cards. Push one of them back into the deck and remove
three reds. Continue in this manner. At each step you randomly select
a card of either color, return it to the deck and remove three cards of
opposite color. Continue as long as you like. When you decide to stop
you will be holding a mixture of reds and blacks. Is it possible that the
number of black cards you hold will equal the number of red? Unless
you think of a parity check it might take a while to prove that the
answer is no. After each step you will always have in your hand an
odd number of cards, therefore the two colors can never be equal .
Magicians have discovered the following curiosity. Place cards with
values ace through nine face down in a row in counting order, ace at
the left. Remove a card from either end of the row. Take another
card fom either end. Finally, take a third card from either end. Add
the values of the three cards, then divide by six to obtain a random
number n. Count the cards in the row fom left to right, and turn over
the nth card. It will always be the four!
I leave it to readers to figure out why this works and perhaps to
generalize it to longer rows of numbers. For example, use twelve cards
with values 2, 3, 4, 5, 6, 7, 8, 9, 10, J, Q, K to make the row. Take
a card three times from either end, divide the sum of the values by
9, and call the result n. The nth card fom the left will always be
the five.
A classic card task, going back more than two centuries, is to
arrange all the aces, kings, queens, and j acks-sixteen cards in all-in
a square array so that no two cards of the same value, as well as no
two cards of the same suit, are in the same row, column, or diagonal.
Counting the number of different solutions is not trivial. W. W.
Rouse Ball, in his classic Mathematical Recreations and Essays, said
there are 72 fundamental solutions, not counting rotations and reflec
tions. This was a mistake that persisted through the book' s eleventh
edition, but was dropped from later editions revised by H. S. M. Cox
eter. Dame Kathleen Ollerenshaw, a noted British mathematician
who was once Lord Mayor of Manchester, found there are twice as
many fundamental solutions, 1 44, making the number of solutions in
cluding rotations and reflections 8 ? 144 1 1 52. She recently described
a simple procedure for generating all 1 1 52 patterns in an article writ
ten for the blind. (Dame Ollerenshaw, now 87, is slowly losing her
vision, and energetically learning how to read Braille. )
This is her procedure: Number the sixteen positions in the array
from 1 through 1 6, left to right, top down. Place an arbitrary card,
say the Ace of Spades, in position 1 , the top left corner. A second ace,
say the Ace of Hearts, goes in the second row. It can' t go in the same
column or diagonal as the Ace of Spades, so it must go in either space
7 or 8. Place it arbitrarily in space 7.
Two aces remain to go i n rows 3 and 4. Put the Ace of Diamonds
in the third row. It can go only in space 1 2. The Ace of Clubs is
now forced into space 14 of the bottom row. Had the second ace gone
in space 8, the last two aces would have been forced into spaces 1 0
and 1 5.
Consider the other three spades. They can' t go in the top row or
leftmost column, or in a main diagonal. This forces them into spaces
8, 1 0, and 1 5. Arbitrarily place the King of Spades in 8, the Queen
of Spades in 10, the Jack of Spades in 1 5. The pattern now looks
like this:
A.
A.
K.
Q. A+
A'
J.
The remaining nine cards are forced into spaces that complete the
following pattern:
A. K.
Q
+ J'
Q
'
J
+
A. K.
J . Q.
K' A
+
K
+
A'
J.
Q
.
Multiply the number of choices at each step, 16 ? 3 ? 2 ? 2 ? 3 ? 2,
and you get the total of 1 1 52 patterns.
For more examples of mathematical theorems, problems, and tricks
with playing cards, see my Dover paperback Mathematics, Magic, and
Mystery, and Karl Fulves' s Self- Working Card Trcks, also a Dover
paperback, and the following chapters in my collections of Scientifc
Amercan columns: The Scientifc Amercan Book of Mathematical
Puzzles and Diversions, Chapter 10; Mathematical Carnival, Chap
ters 1 0 and 1 5; Mathematical Magic Show, Chapter 7; Weels, Life,
and Other Mathematical Amusements, Chapter 1 9; Penrose Tiles to
Trapdoor Ciphers, Chapter 1 9, and The Last Recreations, Chapter 2.
Now for two puzzles that can be modeled with cards. Solutions
appear at the end of the chapter.
1 . Arrange nine cards as shown in Figure 1 . Assume the aces have
a value of 1 . Each row, each column, and one diagonal has a sum
of 6. The task is the alter the positions of three cards so that the
matrix is fully magic for all rows, columns, and diagonals.
A 2 3
3 A 2
2 3 A
Figure 1 .
2. Nine cards arranged as shown in Figure 2 has the property of
minimizing the sum of all absolute diferences between each pair
of cells that are adj acent vertically and horizontally. Assume
that the matrix is toroidal-that is, it wraps around in both di
rections. The sum of the diferences is 48. This was proved
minimal by Friend Kirstead, Jr. , in the Joural of Recreational
Matematics, Vol. 1 8 ( 1 985-86) , p. 301 .
A 2 3
4 5
6
7 8 9
Figure 2.
The challenge is to take nine cards of distinct values (court cards
may be used) and form a toroidal square that will maximize the sum
of all absolute diferences.
Sol uti ons to the Puzzl es :
1 . Move the bottom row of three cards to the top, or move the left
most column to the right, Either change produces the desired
magic square:
2 :
A
A
2
:
:
A
2
2. The figure below shows how nine cards of distinct values can be
placed in a toroidal square so as to maximize the sum of absolute
differences of adj acent values in rows and columns. Any value 4
through 1 0 can go in cells J, y, and < to make a total of 1 20. This
was proved maximal by Brian Maxwell, of Middlesex, England,
in the Journal of Recreational Mathematics, Vol. 1 8 ( 1 985-86),
p. 300.
A K
Z
Q Y
2
X
:
J
chqpter 33
The As
y
mmetri c Pro
p
el l er
Theorem
The late Leon Bankof (he died in 1 997) was a Beverly Hills, Califor
nia, dentist who was also a world expert on plane geometry. (For G.
L. Alexanderson' s interview with Bankof, see [ 1 ] . ) We became good
friends. In 1 979 he told me about a series of fascinating discoveries
he had made about what he called the asymmetric propeller theorem.
He intended to discuss them in an article, but never got around to it.
This is a summary of what he told me.
The original propeller theorem goes back at least to the early 1 930s
and is of unknown origin. It concerns three congruent equilateral
triangles with corners meeting at a point as shown shaded in Figure
1 . The triangles, which resemble the blades of a propeller, need not
form a symmetrical pattern, but may be in any position. They may
touch one another or even overlap. Chords BG, DE, and FA are
drawn to form a hexagon inscribed in a circle. The midpoints of the
three chords mark the vertices of an equilateral triangle.
A proof of the theorem, using complex numbers, appeaed in [2] as
the answer to Problem B- 1 in the annual William Lowell Putnam Com
petition. H. S. M. Coxeter sent the proof to Bankof on a Christmas
card, asking him if he could provide a Euclidian proof of the theorem.
Bankof had no dificulty finding such a proof. In a paper titled
"The Asymmetric Propeller" [ 3] , Bankof, Paul Erd6s, and Murray
Klamkin made the first generalization of the theorem. They showed
that the three equilateral triangles need not be congruent. They can
be of any size, as shown in Figure 2, and the theorem still holds. Two
proofs are given, one a simple Euclidian proof, the other with complex
This art icle frst appeared in The College Mathematics Joual (Vol. 30, January
1999, pp. 1 8-22) .
249
250 A G.rdner's Wor:out
F
Figure 1 .
numbers. As before, and i n all subsequent extensions, the triangles
may touch one another or even overl ap.
Later, Bankof made three further general i zati ons. As far as I
know they have not been publi shed.
Figure 2.
33. The Asymmetri c Propel l er Theorem 251
Figure 3.
Second generalization: The propel ler triangles need not meet at a
point. They may meet at the corers of any equil ateral triangle, as
shown in Figure +.
Third generalization: The propeller triangles need not be equi l at
eral ! They need only be similar triangles of any sizes that meet at a
point. The midpoints of the three added l ines wi l l then form a triangle
simil ar to each of the propellers, as shown i n Figure 4.
Figure 4.
E
G
F
Figure 5.
Fourth generalization: The similar triangles need not meet at a
point! If the propellers meet at the corners of a fourth triangle of any
size, provided it is similar to each propeller, the midpoints of the added
lines will form a triangle similar to each propeller. Vertices of the
interior triangle must touch corresponding corners of the propellers.
Here is how Bankoff proved his final generalization on a sheet that
he typed in 1 973. It makes use of Figure 5.
The propellers shown are right triangles, although they can be any
type of triangle. Perhaps the proof that follows can be simplified.
If ABC, AH J, DBE, and FCC are similar triangles, all labeled in
the same sense and situated so that corresponding angles meet at the
vertices of triangle ABC, then X, Y, Z, the midpoints of DF, CH and
JE, are vertices of a triangle similar to the other four.
Prof Let LBCA LCCF a; LDBE LABC /; LJAH
LCAB " "; and let P, Q, R, S denote the midpoints of the segments
DC, AC, AE and CH respectively. We proceed stepwise to show that
triangles PQR, PSZ and finally XYZ are similar to triangle ABC.
33. The Asymmetric Propel l er Theorem 253
If triangle ABD is pivoted about B so that AB falls along BG and
DB along EB, it is seen by the relation AB/BG " DB/BE and by
the equality of LABD and LGBE that triangles ABD and GBE are
similar and that EG / AD BG / AB, with LEG, AD LBG, BA " /.
Since RQ is parallel to and equal to half EG while QP is par
allel to and equal to half AD, we extend the previous relation to
read RQ/QP EG/AD BG/BA, with LRQ, QP " LEG, AD
LBG, BA /. It follows that triangles PQR and ABG are similar.
In like manner, because of the relationship of AJ and AH to RZ
and QS as well as to RP and QP in both relative length and in di
rection, we find triangles ZRP and SQP similar. Then ZP/SP
ZR/QS " AJ/AH AG/AB, with the angles between the segments
in the numerator and in the denominator all equal to ' As a result,
triangles PSZ and ABG are similar.
Continuing as before, we find triangles Z P X and Z SY similar
since PX/SY " GF/GB " GA/Ge and LPX, SY LGF, Ge
LGA, GB
_
.
Noting that in the similar triangles Z P X and Z SY we have
ZX/ZY ZP/ZS " GA/GB and LZX, ZY LGA, GB "
_
we con
clude that triangles XYZ and ABG are similar.
And now a question for interested readers to explore. Do the pro
pellers have to be triangles? It occurred to me that if squares are
substituted for triangles, as in Figure 6, that equilateral triangle still
shows up.
Figure 6.
I have written this piece as a tribute to one of the most remarkable
mathematicians I have been privileged to know.
Refrences
[ 1 ] G. L. Alexanderson, A conversation with Leon Bankoff, College Mathematics
Joural 23: 2 ( 1 992) 98-1 1 7.
[2] Amercan Mathematical Monthly 75: 7 ( 1 968) 732-739.
[ 3] Leon Bankoff, Paul Erdis, and Murray Klamkin, The asymmetric propeller,
Mathematics Magazine, 46: 5 ( 1 973) 270-272.
Postso-i pt
When I discussed the propeller theorem with Leon Bankof, he showed
me the following somewhat similar theorem. If two squares of arbi
trary size touch at a corner, as shown in Figure 7, and their centers
are joined to midpoints on lines connecting pairs of corners, the re
sult is another square. He said he was working on a proof. I cannot
now recall if the theorem was original with Leon or whether he had
encountered it in a periodical.
33. The Asymmetri c Propel l er Theorem 255
Figure 7.
Ch(
p
ter 34
Chess Queens ( nd
M(xi mu m Un(t(cked Cel l s
There is now an enormous literature on the old classic task of placing
eight queens on a chessboard so that no queen can attack another.
There are twelve solutions, not counting trivial rotations and reflec
tions. The task naturally generalizes to enumerating the number of
solutions for n non-attacking queens on an n x n board. (See Chapter 1 6
of my Unexpected Hanging and Oter Mathematical Diverions, 1 969. )
Less well known is an intriguing similar task. Let me introduce it
with a dificult little puzzle that I thought of and first explained in my
May 1 972 Scientifc American column, and gave again in a February
1978 column. Can you place five white queens and three black queens
on a 5 x 5 board so that no queen of one color can attack a queen of
another color? The only solution is shown in Figure 1 . To prove it
unique, the best way is to explore how three queens can be placed to
leave five cells unattacked.
This suggests the following generalization: How can n queens be
placed on an order-n board so as to maximize the number of un attacked
vacant cells? And how many different ways can this be done for any n?
Figure 1.
This article frst appeared i n Mat Horzons (November 1999) .
257
Let' s generalize further. Given a board of order n, what is the
largest number of un attacked vacant cells that k queens will allow,
and in each case, how many distinct patterns solve the problem? The
same question can be more colorfully framed along the lines of the
order-5 problem. Given an order-n board and k queens of one color,
how can they be placed so as to maximize the number of queens of a
diferent color that can be put on the board so no queen of one color
can attack a queen of the other color?
The earliest publication known to me of a problem of this sort is in
W. W. Rouse Ball' s classic Mathematical Recreations and Essays (third
edition, L 896). In his chapter on chess recreations Ball attributes the
problem to his fiend Captain W. H. Turton. The task is to place
eight queens on the chessboard so as to Leave un attacked the largest
number of vacant cells. One solution is given. Ball comments: "Is it
possible to place the eight queens so as to Leave more than eleven cells
out of check? I have never succeeded in doing so, nor in showing that
is impossible to do it. "
Henry Ernest Dudeney' s Problem 3 1 6, in his Amusements in Math
ematics ( 1 9 17), involves eight queens placed on the chessboard as
shown in Figure 2. The puzzle, which Dudeney says is based on Cap
tain Turton' s problem, is to alter the positions of three queens so that
eleven vacant cells are unattacked. The unique solution is shown in
Figure L 4.
"I will hazard the statement," Dudeney writes, "that eight queens
cannot be placed on the chessboard so as to leave more than eleven
squares unattacked. It is true that we have no rigid proof of this yet,
but I have entirely convinced myself of the truth of the statement. "
Figure 2. Dudeney' s eight queens puzzle.
34. Chess Queens :md M'xi rur Vn't'cked Cel l s
259

A B

c D

E F
Figue 3. The six other solutions to the eight queen' s task. The black dots are unattacked
rel l
n 1 2 3 4 5 6 7 8 U 10 1 1 12 13 14 15 16
unattacked cell 0 0 0 1 3 5 7 1 1 18 22 30 36 47 56 72 82
DumUCIof solutions 0 0 0 25 1 3 38 7 1 1 2 7 1 4 3
Figure 4. Maximum unattacked cells and number of solutions for n queens on order-n
boards, n - 1 through 1 6.
Dudeney' s solution is unique only when given his initial conditions.
Without such conditions, how many different ways can eight queens
be placed so as to leave eleven vacant cells unattacked? Dudeney says
he knows of "at least five. " In 1 995 Robert Trent, of Hardinsburg,
Kentucky, wrote a computer program that verified that eleven is in
dead maximum, and found seven distinct solutions. I later learned
that Bernd Schwarzkopf had published them in Die Schwalbe, a Ger
man periodical, in 1 982. The six difering fom Dudeney' s are shown
in Figure 3. Solution A is the one given in Ball' s book.
In the past few years a raft of mathematicians around the world
have studied the general problem, with or without computer aid. They
are notably Ronald Graham and Fan Chung, of the University of Cali
fornia at San Diego; Hiroshi Okuno, of Tokyo; Thur Row, of St. Louis;
Robert Wainwright, of New Rochelle, New York; Mario Velucchi, a
computer scientist at the University of Pisa, Italy; and mathemati
cians unknown to me, from at least eight diferent nations, who corre
sponded with Velucchi . Early results are discussed in Stephen Ains
ley' s Matematical Puzzles (G. Bell & Sons, 1 977), and more recent
results in Les Jeux Matematiques (University of France Press, Paris,
1 997), by Michel Criton.
If the task is limited to cases where k n, the table shown in
Figure 4, supplied by Velucchi , shows the known maximum number
of unattacked cells for each n 1 through 1 6, and the number of
diferent solutions for each n. The unique patterns for n 9, 1 0, 13,
and 1 6 are shown i n Figure 5 . Patterns for 9 and 1 6 are symmetric
with respect to a main diagonal. Note that for n 4 through 8 the
number of unattacked cells is a sequence of five consecutive primes
(taking 1 to be prime) . Unfortunately, this sequence does not continue.
34. Chess Queens ;md Mqxi mum Vnqtqcked Cel l s 261
i

I.

I.

I.

.
r
.

Ie
|
.

I
e ie
H : k =
t .: .1
o = k |t|
I

.
!.

i.

I
i
O
i

.'
I
.
I .
i . [. I.
I.

i.,
i. i.
I
..

I i e l

L
e
0

I .

I
I
.
tt = k = | .l
I
.
1
.i
l

I
.

I
e
I-

.
1
.
1
I.

, . , .

,e
,. l
e
I
e

1

I
e
I
e

I .

e
I
e

I
.
Q k
Figure 5. Unique patters for 7 - k - 9, 1 0, 1 3, and 1 6.
Queens (k)
2 3 4 5 6 7 8 9 10 1 1 1 2
3 2 1 0 0 0 0 0 0 0 0 0 0
4 6 3 2 1 1 1 0 0 0 0 0 0
5 12 7 5 4 3 2 2 1 1 1 1 1

6 20 1 3 9 8 6 5 4 4 3 2 2 2
"
7 30 21 1 6 1 6 1 1 10 7 6 6 6 5 4
1
8 42 31 25 24 1 7 15 1 3 1 1 1 0 9 8 7
C
.
9 56 43 36 36
27 24 20 1 9 1 8 1 4 1 3 12
10 71 57 49 48 37 33 30 26 24 22 19 18
1 1 90 73 6 6 51 46 42 39 36 32 30 29
1 2
1 10 91 81 80 65 59 56 50 46 42 38 36
Figure 6, Maximum number of unattacked cells, U 3 through 1 2, k 1 through 1 2.
The chart i n Figure 6, based on computer results from Okuno and
Trent, covers the more general case of k queens on order-n boards.
Very little is known about the number of distinct solutions for given
values of k and n except for very low orders.
Now for some amazing discoveries by Graham and Chung. I pub-
lish them here, with their permission, for the first time.
Instead of listing the maximum number of unattacked cells for k
queens, we list the minimum number of attacked cells for k queens,
assuming that each queen attacks the cell it is on. This yields the com-
plementary chart shown in Figure 7. Each number on this chart ob-
Queens (k)
2 3 4
5 6 7 8 9 10 1 1 1 2
3 7 8 9 9 9 9 9 9 9 9 9 9
4 1 0 1 3 14 1 5 15 15 1 6 16 16 1 6 16 16
5 1 3 1 8 20 21 22 23 23 24 24 24 24 24

6 16 23 27 28 30 31 32 32 33 34 34 34
7 19 28 33 33 38 39 42 43 43 43 4 45
1
8 22 33 39 40 47 49 51 53 54 55 56
57
.
9 25 38 45 45 54 57 61 62 63 67 68 69
10 28 43 51 52 63 67 70 74 76 78 81 82
1 1 31 48 57 57 70 75 79 82 85 89 91 92
12 34 53 63 6 79 85 88 9 98 1 02 1 06 108
Figure 7, Minimum number of attacked cells, U 3 through 1 2, k 1 through 1 2.
34. Chess Queens and Maxi mum Vnatacked Cel l s
k = 1
3 lines
k = 4
7 lines
k ^ 7
9 lines
'
k = 2
5 lines
k = 5
8 lines
k = 3
6 lines
k " 6
9 lines
k = 8
1 0 lines
263
Figure 8. The number of lines generated by k spots on a hexagonal lattice correlates
with the differences between consecutive values for the minimum number of attacked
cell s by k queens as T (a board' s order) increases.
viously is obtained by subtracting each number on the previous chart
from n
2
, the number of squares on an n ? n board.
Look down the column for k 1 , a single queen on an order-n
board. The best spot for it is at a corner. As n increases, the values
in the table have a constant difference of 3 . The values in the column
for two queens have a constant difference of 5 . The values for three
queens start with a diference of 5 , followed by 6, 7, 6, 6, 6, . . . . The
sixes continue to infinity. Indeed, Graham and Chung have proved
that for every value of k, the consecutive differences, as n increases,
eventually become constant !
What is more astonishing, the value of this constant, with the sole
exception of k 4, is exactly equal to the number of lines generated by
k points that spiral around a fixed point on a hexagonal lattice? Figure
8 shows how this works for points 1 through 8. The chart shown in
Figure 9 gives the value of the constant diference for each k, from
1 through 1 9. Note: The starred numbers are hexagonal numbers of
the form 3x
2
- 3x + 1 .
consecutive consecutive
k diferences k diferences
* 1 3 1 1 1 2
2 5 12 1 2
3 6 13 1 3
4 5/7 14 1 3
5 8 15 14
6 9 16 14
*7 9 1 7 15
8 10 1 8 15
9 1 1 * 1 9 15
10 1 1
Figure 9 . For every k , beyond a certain value, the consecutive differences between
minimum attacked cells, as n increases, become constant.
Graham and Ghung were unable to obtain a precise expression
for the value of n beyond which the consecutive diferences become
constant. They estimate it to be within the range of 4k or 5k.
In the exceptional case of k 4, the consecutive constant diference,
beyond n 5, oscillates between 5 and 7. The minimum number of
attacked cells is 6n - 8 when n is even, and 6n - 9 when n is odd. This
arises fom the fact that a best pattern for four queens, when n exceeds
4, is obtained by placing the queens in the board' s four corners. When
n is odd and greater than 3 , the board' s two main diagonals, along
which the queens attack, share the board' s central cell, so the formula
subtracts 9 from 6n rather than 8 in the even case where the two main
diagonals do not intersect.
Formulas for the maximum number of un attacked empty cells are
known only for small k. For example, if there is only one queen, a
maximum number is obtained by putting the queen in a corner. For
every n the number of unattacked cells is n
2
- 3n + 2. If there are two
queens, solutions are obtained by putting one queen in a corner, and
the other in the same row or column, on the third cell from the corner
occupied by the other queen. In this case the number of unattacked
cells is n
2
- 5n +
7
.
Wainwright found numerous "pretty" solutions. Figure 1 0, for ex
ample, is one of many symmetrical solutions he found for n k
12. Trent suggested the narrower task of finding symmetric pat
terns on which the number of queens equals the maximum number
of un attacked cells. In such cases, of course, the queens and the cells
not in check can be exchanged. Two bilaterally symmetric examples
34. Chess Queens 4nd M4xi mum Vn4t4cked Cel l s 265

o
A A
A
0
O
o
G
Figure 10. Wai nwright' s bi l ateral l y symmetric sol ution for k - H - 12.
on order-9 boards are shown i n Figure 1 I . Figure 1 2 shows how seven
queens fai l to attack the maximum of seven cel l s on an order-7 board.
Two questions remain open:
1 . Find a formul a that, given the number of queens, wi l l determine
the constant difference between the numbers of attacked cel l s as
i ncreases beyond a certai n poi nt.
Figure 11. Two symmetric patterns frm Bob Trent on the 9 X 9 fiel d wi th k. the number
of queens. equal to the maxi mum number of unattacked cel l s.

Figure 12. Seven queens on a 7-board have a maximum of 7 cells unattacked. This is
one of 38 solutions.
2. Find a formula that, given k and n, will enumerate the num
ber of diferent patterns that produce a maximum number of
un attacked cells. This could be an enormously dificult problem
in combinatorics. Moreover, there is no reason at present to as
sume that such a formula even exists.
I have not given any of the neat proofs by which Graham and Chung
arrived at their results because they are a bit too advanced for this
magazine. Perhaps the two will provide them some day in a technical
paper.
Similar tasks involving rooks, bishops, and knights have been in
vestigated by Velucchi. I know of no work done on queens that move
on boards with triangular or hexagonal cells. I give here in Figure 1 3
two pleasant little puzzles based on fields drawn on isometric paper.
On each field a queen moves like a rook in six different directions,
Figure 13. Two un attacked cells problems.
34. Chess Queens nd Mxi mum Vntcked Cel l s 267
horizontal l y left or right, and up and down in two diagonal directions.
In other words, a queen moves along rows of cel l s that are between
adjacent paral l el l i nes. Pl ace four queens on each board maxi mi zi ng
the number of unattacked cell s. The tri angul ar patter has onl y one
solution. The hexagonal patter has three.
Perhaps some readers will be i nterested in generalizing these re
sults to larger isometric fields. Anyone wishing to correspond with
Vel ucchi can reach hi m at Vi a Emi li a, 1 06, 1 -56 1 2 1 , Pi sa, Italy .

Figure 14. Dudeney' s solution to his eight queen puzzle.

|3" ''
chqpter 35
Li on Hu nti ng
Lion Hunting and Other Mathematical Pursuits: A Collec
tion of Mathematics, Verse and Stores by Ralph P. Boas,
Jr. Gerald L. Alexanderson and Dale H. Mugler, eds. xii +
308 pp. The Mathematical Association of America, 1 995. $35.
Ralph Philip Boas, Jr. , died i n 1 992 without seeing the publication of
this splendid tribute to his distinguished career. He was not only a
creati ve mathematician of top rank, but also an inspired teacher, long
time chairman of Northwestern University' s mathematics department
and one-time editor of Matematical Reviews and The Amercan Math
ematical Monthly.
Mathematicians Alexanderson and Mugler have done an admirable
job of assembling this collection of the lighter side of Boas' writings,
although some of the papers may not seem "light" to readers unfa
miliar with advanced mathematics. Interspersed among such serious
articles are short stories, humorous verse, amusing anecdotes and col
orful reminiscences.
Among Boas' popularly written papers the two best known are
"Contributions to the Mathematical Theory of Big Game Hunting, "
and "Snowfalls and Elephants, Pop Bottles, and Pi . " The first of these
outlines 16 ways to capture a lion. For example, you enter a spher
ical cage in the desert, lock it, then perform a space inversion with
respect to the sphere. This puts you outside the cage and the lion in
side. Boas published the paper under the pseudonym of H. Petard. He
later explained that Petard' s full initials are H. W. O. , which stand for
"Hoist With Own. " Several mathematicians were so amused by this
paper that they later published articles, all included in this anthology,
giving additional methods for lion trapping.
This article frst appeared i n Amercan Scientist (Marchi April 1996) .
271
The paper about snowfalls and elephants is an introduction to how
certain statistical problems can be solved by careful sampling. Boas
first takes up the way to calculate the expected number of heavy snow
falls in a particular place or during a person' s lifetime, then shows how
similar techniques apply to finding an extremely heavy elephant in a
herd without having to weigh them all and, finally, how to determine
the number of soft-drink bottles that will not explode when the tem
perature soars.
Two other nontechnical papers merit special mention. "Distribu
tion of Digits in Integers" reveals some startling consequences of the
fact that almost all real numbers are "normal" in the sense that every
digit or finite run of digits occurs equally often in the digital expansion
of that number in any base notation. Imagine that the Encyclopedia
Brtannica is coded by numbers that form one monstrous integer. If
such a real number as J, e or v is indeed normal, as suspected though
not proved, then the Brtannica, in its numerical form, will occur an
infinite number of times in the decimal expansion of the number.
"Traveling Surprises" applies the mean value theorem to a car that
goes from here to there with an average speed of 50 m. p. h. Boas proves
there must be an instant at which the car' s speed is precisely 50 m. p. h. ,
and that there is a short interval during which the average speed is 50
m. p. h. Even more surprising, i f the car travels longer than an hour,
there must be one hour during which the car goes exactly 50 miles.
The paper includes a graph with an arrow pointing to a spot on a
space-time curve. The arrow is labeled "Aha! "
Many of the book' s anecdotes are so technical that only mathe
maticians will find them funny. Some even require a knowledge of
German. I particularly enjoyed those anecdotes that convey Boas' s
love of word play. That "ghoti" spells "fish" is so well known that
James Joyce mentions it in Finnegans Wake. (You pronounce "gh" as
in laugh,
as in women, and "ti" as in nation. ) Not well known

is Boas' discovery that "ghghgh" spells "puff." Pronounce "gh" as in
hiccough, Edinburgh and laugh.
Norbert Wiener once said he would like to write a story about a
girl from Walla Walla who went to Pago Pago to dance the hula hula.
This inspired Boas to write
Lulu from Walla Walla was a devotee of dance:
She did a wicked can-can in a tutu sent fom France.
She said, "I' m going gaga; in toto, life' s a bore, "
So she went to Bora Bora and to Pago Pago' s shore,
35. Li on Hu nti ng
Where she studied hula hula and tried for an A. A. ,
But her work was only so-so, and they wouldn' t let her stay.
Now she' s gone to Baden Baden, to a go-go cabaret;
Billed as Lulu in the muu muu, she' s performing every day.
When she lets the muu muu drop, the old folks drop their teeth
When they see a lava-lava is all she has beneath.
273
In Princeton' s Fine Hall, Boas recalls, someone once posted a "Scale
of Obviousness":
If Wedderburn says it' s obvious, everybody in the room has
seen it ten minutes ago.
If Bohnenblust says it' s obvious, it' s obvious.
If Bochner says it' s obvious, you can figure i t out i n half
an hour.
If von Neumann says it' s obvious, you can prove it in three
months if you' re a genius.
If Lefschetz says it' s obvious, it' s wrong.
Boas was puzzled by a Japanese paper that kept referring to "strick
en mass distributions. " Unable to figure out what this meant, he wrote
to the journal' s editor. It turned out that a referee' s report had told
the author, "The term ' generalized mass distribution' is no longer used.
The word ' generalized' should be stricken. "
There are simple ways to determine whether a number is divisible
by each digit except 7. Boas recalls a letter proposing a way to test
for 7. "Because an integer is divisible by 10 if and only if its last digit
in base 10 is 0, then to test its divisibility by 7 you need only write
the number in base seven and look at the final digit. "
Boas remembers Grunwald' s theorem in algebraic number theory.
"It had a very long proof," he writes, "but after a while Whiples found a
shorter proof. Later, somebody found a counterexample. " Papers often
have uninformative titles. Boas cites the worst example he ever came
across. It was titled "On a Certain Theorem. " He recalls a review of
another paper that said, "This paper contains two theorems. The first
is due to the referee and the second is wrong. "
The short story by Boas that I enjoyed most is about Valentina, a
mathematical whiz who creates a computer program that plays perfect
chess, winning every game. Rather than destroy the game, she locks
it in a safety deposit vault for 1 00 years, hoping that by then the
chess world will be able to accept it. I have, of course, touched only a
few aspects of this delightful anthology that conveys so well the depth
and wit of a remarkable man. The book closes with a bibliography
of more than 280 references to books and articles by Boas, including
his translations of Russian books, but excluding hundreds of his book
reviews.
chqpter 36
Two Books on I nfn ity
Since ancient times, mathematicians, scientists, philosophers and the
ologians have struggled with the dark mysteries of infinity. Our uni
verse seems to be of finite size, expanding at an unknown rate ever
since the Big Bang, but is there an infinite space into which it is ex
panding? Time also seems to have started with the bang, but was
there some sort of super time before then? And if our universe has
enough matter to halt the expansion and send the cosmos backward
toward a Big Crunch, will time continue to march on?
For Immanuel Kant, the infinities of time and space were contra
dictory "antinomies. " It is dificult to imagine space extending forever,
but it is just as hard to suppose it stops, because that will lead you
to ask: "What' s outside the boundary?" It is hard to imagine time
without beginning and end but equally hard to suppose otherwise.
Cosmologists today are speculating, as did the 1 6th century Italian
philosopher Giordano Bruno and others before him, that there may be
an infinity of isolated universes, each with its own set of laws. Only a
tiny subset of these universes allows life to arise. As someone has said,
a universe is something that happens now and then. Time, someone
else said, is what keeps everything from happening at once.
Infinity also extends inward. Is there an ultimate particle, or does
matter have infinite levels of structure like a set of Russian dolls?
Science fiction writers have long played with the concept of infinite
universes. H. G. Wells wrote a story in which our cosmos is an atom
in a ring worn by a gigantic hand. Other writers have turned our
atoms into other universes. Cyberpunk writer Rudy Rucker, in his
novel Spacetime Donuts, introduced a wild circularity. He imagines
the paradox of a person getting smaller and smaller, visiting ever
tinier universes, until he finds himself back in the universe where he
started without having to change his size.
This review appeared in The Los Angeles Times Book Review ( May 18, 1997) .
275
Mathematicians speak of parallel lines meeting at infinity and
of infinite series of factions whose partial sums converge at infin
ity on a well-defined limit. Calculus rests on the summation of infi
nitely small quantities that German philosopher and mathematician
Gotfried Wilhelm von Leibniz called infinitesimals. They were much
derided by Bishop George Berkeley in the early 1 8th century and con
sidered meaningless by most mathematicians until they became re
spectable again in what is known as nonstandard analysis.
A regular polygon turns into a circle when it has an infinity of sides.
The digits of irrational numbers like go on endlessly without repeat
ing a particular sequence of numbers. A revolution in mathematics
occurred when German mathematician George Cantor, in papers pub
lished between 1 874 and 1 884, found a way to define a hierarchy of
infinities.
The lowest infinity represents the number of integers, including
integral fractions. The next higher one counts the real numbers. The
next counts all the curves that can be drawn on a postage stamp.
Cantor' s transfinite numbers, which he called alephs, are themselves
infinite in number.
Theology, too, has its infinities. Does God have infinite attributes?
Is God subject to time and change or does God exist in eternity, outside
space and time altogether? After we die, do we sufer eternally in hell
or live forever in paradise?
A number of interesting books have been written about infinity, in
cluding a nonfiction book by Rucker, Infnity and te Mind. The most
recent are Richard Morns' Achiles in the Quantum Universe (Henry
Holt, 1 997), J. V. Field' s The Invention of Infnity (Oxford University
Press, 1 997), and Eli Maor' s To Infnity and Beyond (Springer Ver
lag, 1 991 ) .
Morris, author of several books about science, has done his usual
excellent job of ranging over his topic in lucid, entertaining prose. He
covers all the important questions and more. The Achilles of his book' s
title is the Greek warrior in one of Zeno' s famous paradoxes of motion.
Achilles is unable to catch a tortoise because when he reaches the
spot where the turtle started, the reptile has crawled ahead a finite
distance. When he goes that distance, the turtle has moved ahead
by a smaller distance. The distances the turtle moves ahead keep
shrinking but never vanish, so how does Achilles manage to overtake
the tortoise? The two runners model dimensionless points that move
along a straight line. If Achilles were to pause a minute after running
36. Two Books on I n Fi nit
y
27
each segment, he would never reach the turtle. As it is, if both go at
a steady rate, the time it takes Achilles to traverse each segment also
gets smaller, converging on zero, allowing him to reach the tortoise in
a finite time.
Morris covers all the infinities that plague cosmology, particle phys
ics, relativity theory and quantum mechanics. The infinities that arise
in black holes are especially troublesome. In a black hole, the volume
of matter becomes zero and its density goes to infinity. No one knows
just what happens after that. Does the matter explode, as some cos
mologists suggest, from a white hole in another universe?
Morris does his best to explain a bizarre model of the universe
recently proposed by Stephen Hawking. In ordinary time, our universe
has a beginning and end, but in "imaginary time" (based on imaginary
numbers), the universe is infinite in both directions. As far as I can
tell, no one except Hawking has taken his model seriously.
It is a credit to Morris that in writing about infinity, he draws upon
understandable and entertaining analogies that have historically been
used to describe infinity. For example, in explaining Cantor' s alephs,
Morris introduces the notorious aleph hotel, which has an infinity of
rooms numbered 1 , 2, 3 , . . . Every room is occupied. Can the manager
accommodate the arrival of an infinity of guests? Easy. He simply
moves each occupant to a room with a number twice his or her previous
number. This opens up an infinity of rooms, namely all those with
odd numbers. As Morris informs us, this paradox was recognized by
Galileo, but it took Cantor to clarify the mathematics involved.
Morris concludes his book with the concept that "the dificulties
encountered by modern physicists show us that the infinite is still as
much a mystery as it was in the time of Zeno. " Both atheists and
theists, he rightly maintains, can accept the now fashionable notion
of a plurality of universes, perhaps even an infinity of them, exploding
here and there, now and then, in some kind of super time and space.
Meditating on such possibilities can arouse in one a sense of awe so
intense that if it persisted more than a minute or two, one could go
mad. Morris' final sentence is a memorable quote from Pascal, as
pious a theist as he was a great mathematician: "The eternal silence
of these infinite spaces frightens me. "
Field' s The Invention of Infnity, in spite of its title, has very lit
tle in common with Morris' book. A research fellow in art history
at Birkbeck College, University of London, Field has written books
about Kepler' s cosmology, the geometry of Girard Desargues and Sci-
ence in Ar, beautiful work featuring paintings in London' s National
Gallery that portray aspects of science and technology. The Invention
of Infnity is an equally handsome volume reproducing hundreds of
striking pictures, alas none in color, that relate to mathematical ideas
of their time.
Although Field makes side excursions into the history of geometry
and algebra, the primary focus of his book is the development of projec
tive geometry and its application to perspective in painting. It is here
that infinity comes into play as the "distance point" on the horizon
at which parallel lines meet at an infinite distance from the viewer.
Field' s book is essentially a detailed, erudite treatise on the collabo
ration of eminent Renaissance mathematicians with artists who had
mastered the art of perspecti ve.
Field reproduces and discusses numerous paintings by Italian art
ists, many of whom also wrote treatises on perspective. Mathemati
cians who contributed to perspecti ve and the concept of infinity include
Kepler, Pascal, Descartes and, above all, Desargues, the French math
ematician who wrote the first great work on projective geometry in the
1 7th century.
Field does not neglect the conic section curves that result when a
right circular cone is sliced by a plane. If the plane is parallel to the
cone' s base, the section is a circle. Tilt the cone ever so slightly and the
circle' s center splits into two foci to create an ellipse, the shape a circle
has when seen in perspective. As the plane assumes steeper angles,
one focus moves father from the other until the plane is parallel with
the cone' s side. At that point, the focus traveled to infinity and the
ellipses have become a parabola with arms that are parallel when they
are infinitely far fom the other focus. As the cutting plane tips even
more-until it is perpendicular to the cone' s base-the cross-sections
tur into parabolas. Their arms meet their asymptotes at infinity.
In his last chapter, Field also refers to Pascal' s remark about his
fear of infinite space. Field seems to think that Pascal is expressing
not his own fears but the fears of an atheist. I believe this is a mis
reading of Pascal. It is Pascal' s own terror, the terror of a devout
believer in God, that Pascal is describing. Here is a longer quota
tion from Pascal' s "Pen sees, " a work saturated with wonder about the
awesome mysteries of space and time:
When I consider the short duration of my life, swallowed
up in the eternity before and after, the small space which
I fill, or even can see, engulfed in the infinite immensity of
36. Two Books on I n Fi nit
y
spaces whereof I know nothing, and which know nothing of
me, I am terrified, and wonder that I am here rather than
there, for there is no reason why here rather than there,
or now rather than then. Who has set me here? By whose
order and design have this place and time been destined
for me?
279
chqpter 37
The Un i verse q nd the Teqcu
p
The Universe ad the Teacup: The Mathematics of Truth
ad Beauty by K. C. Cole. Harcourt Brace, 1 998.
On an acknowledgments page, science writer K. C. Cole says she was
surprised when her editor told her she had written a book about math
ematics. She had intended to write a book that surveyed trends in
contemporary science, and she has, delightfully, but math is the com
mon thread that binds it all together. What do the largest galaxy and
the smallest teacup have in common? Answer: mathematical struc
ture. Cole' s book is a loving paean to the awesome power and beauty
of mathematics.
Cole' s early chapter on giant numbers stresses how hard i t is to
grasp their applications. Our galaxy contains some 200 billion stars,
and there are more than 200 billion galaxies. "Number numbness" sets
in when you contemplate the vastness of space and time. A geologist,
after chalking on the blackboard a line that runs from zero to a trillion,
asks where to put the point that indicates a billion. Amazingly, it is
extremely close to zero! "Compared to a trillion, " Cole comments, "a
billion is peanuts. "
Few realize how rapidly numbers grow by a simple process of dou
bling. Put two grains of wheat on the first square of a chessboard,
four on the next, then eight and so on. You can' t get very far. After
only 1 0 steps, a square requires 1 024 grains. The last square needs
more grains than have ever been produced on Earth. Big numbers.
Cole tells us, "creep up on us unawares. " Disaster looms when pop
ulation grows exponentially on a planet with a surface as finite as a
chessboard.
Widespread failure to understand simple probabilities receives care
ful treatment. We worry more about alar in apples, Cole writes, than
This review appeared in The Los Angeles Times Book Review (February 1, 1998) .
281
the greater threat of cigarette smoke. A woman may avoid fatty meats,
yet not mind sleeping with strangers. Everywhere there is what Cole
calls "skewed perception of risks. "
In a chapter on measurements, Cole introduces the central mys
tery of quantum mechanics. How can an electron have no definite
properties, such as position or spin, but acquire precise attributes as
soon as they are measured? She likens this to a spinning coin, which
is neither heads nor tails until it falls flat. Dificulties in measuring
intelligence are also considered.
There are also severe limits on how large certain things can be.
No teacup can be as big as Jupiter because gravity would pull it into
a sphere. Gravity would fracture the thighs of a person 60 feet tall.
A flea can jump high in relation to its size, but the jump' s height,
approximately one meter, is about the same as what an average person
can achieve. Cole calls it an "interesting invariant" that most animals
can' t leap higher than that.
A thousand other fascinating facts and shrewd observations crowd
into Cole' s lyrical praise of science, from the wonders of galaxies to the
invisible creatures that live on our eyelashes. She is good at emphasiz
ing how unpredictable properties continually emerge as the universe
evolves from primeval simplicity to the enormous complexity of a hu
man brain. What could be more unlike a gas than water? Yet when
two gases, hydrogen and oxygen, combine, the more complex proper
ties of water emerge. Stars and rocks result fom even more com
plex combinations of molecules. At some time in the distant past, life
sprang into existence fom the complexity of a self-replicating mole
cule. As it evolved from simple one-cell forms, the magic of increasing
complexity produced trees, butterflies, dinosaurs, frogs and elephants.
You and I somehow emerged, after billions of years, from the quan
tum fields that predated the big bang. Biology is not applied physics,
Cole reminds us, and psychology is not applied biology. Knowledge
of sounds reveals nothing about a Mozat symphony. "You can learn
everything there is to know about the atoms that make up a cat, "
Cole writes, "and that still will not tell you whether it will scratch the
furniture or sleep on your head. "
Slow increments of change lead to "tipping points, " where a quali
tative change occurs suddenly. Water cools gradually, then, presto, it
turns into ice. Tipping points are all over the human scene: an unan
ticipated drop in crime, a plunge of the Dow, an angry man' s murder of
his wife. Chaos theory studies systems that start out orderly, then, by
37. The Vn i verse ,md the Teacup 283
altering in a completely deterministic way, suddenly cross a threshold
to become unpredictably random.
"The Signal in a Haystack" is Cole' s title for a chapter on how
science does its best to filter significant facts from vast amounts of
irrelevant information that often contaminate research results. As
tronomical data are still not free enough from cosmic noise to deter
mine how fast the universe is expanding or exactly how old it is. Are
there really black holes at the centers of galaxies? Does star wobbling
indicate planets? Are wiggly forms on a meteorite fossils from Mars?
"The sky is like a cocktail party," Cole writes, "with too many conver
sations going on at once. " Scientists listen to the babble in the hope
of hearing what the universe is trying to tell them. As Cole writes:
"That' s exactly how the search for truth is supposed to work. You see
something, and then you try everything you can think of to make it
go away; you turn it upside down and inside out, and push on it from
every possible angle. If it' s still there, maybe you' ve got something. "
The Universe ad the Teacup contains informative chapters on the
inevitable flaws in every voting system and on how things can be
divided fairly. Fair division is a piece of cake when only two people
want half of a piece of cake. One person cuts; the other chooses. The
trick has endless applications to social and political conflicts, such as
who gets what after a divorce, assuming each party wants to be "envy
free. " If more than three players are involved, the task of fair division
becomes more difficult. Ingenious solutions by various experts are
skillfully outlined.
The conflict between altruism and egoism also gets Cole' s attention.
Does self- sacrifice have survival value for an evolving animal species?
Can bitter conflicts between business firms or nations be settled am
icably by applying game theory? The famous tit-for-tat strategy is
explained. This involves competition between two "players, " such as
the efort of one nation to establish fee trade with another nation
or to obtain mutual disarmament. The tit-for-tat strategy is to make
the first move in the desired direction. After that, do whatever your
adversary does. Eventually it dawns on both players that in the long
run, cooperation maximizes self-interest.
Near her book' s close, Cole tackles the question of whether scien
tific truth is always fallible in contrast to the certainty of math and
logic. Kurt Godel' s famous undecidability theorem (it states that every
formal system of mathematics that includes arithmetic contains true
statements that cannot be proved true within the system) is consid-
ered and so is the current fad of fuzzy logic. The book' s final chapter is
a tribute to Emmy Noether, an eminent German mathematician. She
proved that behind every invariant in physics lurks a symmetry that
can be defined by an algebraic structure called a group, an insight
essential to relativity theory and particle physics. Immediately after
the big bang, the universe was inconceivably hot and perfectly sym
metrical. As it cooled, various symmetries were shattered to fashion
the cool and broken universe we know and love.
I have touched on only a tiny fraction of the topics covered in this
dense and passionate book. You will put it down sharing Cole' s awe
and wonder at the vastness and intricacy of what G. K. Chesterton
once said an atheist must view as the most exquisite mechanism ever
constructed by nobody.
chqpter 38
A New Res u l t C|
Perfect Mq
g
i c Squq res
Beauty i s the first test:
There is no permanent place in the world
for ugly mathematics.
-G. H. Hady, A Mathematician's Apology.
Dame Kathleen Ollerenshaw, one of England' s national treasures, has
solved a long- standing, extremely dificult problem involving the con
struction and enumeration of a certain type of magic square. But first,
some background.
For many centuries mathematicians, especially those concerned
with combinatorics, have been challenged by magic squares. These
are arrangements of n
2
distinct integers in an n ? n array such that
each row, column, and main diagonal has the same sum. The sum
is called the magic constant, and n is the square' s order. Traditional
magic squares, known as "normal," are made with consecutive inte
gers starting with 0 or 1 . If a magic square starts with 0 it can, of
course, be changed to start with 1 simply by adding 1 to each cell.
No order-2 square is possible. The order- 3 , shown in Figure 1 ,
barely exists. Why? Because there are just eight diferent triplets of
distinct digits from 1 through 9 that add to 1 5, the square' s constant.
(If a square starts with 1 , the constant is half the sum of n and n
3
. )
Each triplet appears as one of the square' s eight straight lines of three
numbers. The pattern is unique except for trivial rotations and mirror
reflections, which are never considered diferent.
This, little gem of combinatorial number theory was called the 10
shu in ancient China. Legend has it that in the 23rd century B. C.
A cut version of this review appeared in Nature ( Vol. 395, September 17, 1998) .
285
286 A G.rdner's Wor;out
2 7 6
9 5 1
4 3 8
Figure 1. The 10 shu. the only magic square of order 3.
a mythical Ki ng Yu saw the pattern on the back of a sacred turtle
in the River Lo. Modern hi storians, however, find no evidence that
the patter was known before the fourth or fifth century B. C. At any
rate, the name means Lo Ri ver Wri ti ng. The Chi nese identify it with
their familiar yin-yang circle. The even digits, here shown shaded,
are l i nked to the dark yi n. The Greek cross of odd di gi ts i s l i nked to
the light yang. For centuries the 10 shu has been used as a charm on
j ewelry and other objects. Today, large passenger ships ofen use the
10 shu as a shuffleboard pattern.
The number of order-4 magic squares j umps to 880. Among them
is a famous subset of 48 squares called pandiagonal . (In the past they
have been called diabolic and nasik squares. ) A pandiagonal square
with a constant of 30 i s shown in Figure 2.
Al l pandiagonal squares of order 4 have three amazing properties:
I . Each "broken di agonal" al so adds to 30. (Cel l s 0, +, 4, 1 5, 1 2
and 7, 1 3, 8, 2, and so on. are examples of broken diagonals. )
0 13 6 1 1
7 10 1 12
9 4 15 2
14 3 8 5
Figure 2. An order-4 most-perfct square. The magic constant is 30.
38. A New Res ul t C PerFec Magi c Squares 287
Imagine an endless array of this square, or any other order-4
pandiagonal, placed side by side in all directions to make a wall
paper pattern. Every 4 ? 4 square drawn on this pattern will be
a pandiagonal magic square. Every straight line of numbers, or
thogonally or diagonally, will add to 30. Similar properties hold
for pandiagonal squares of all higher orders. A 5 ? 5, for exam
ple, will form a wallpaper pattern on which every 5 ? 5 square is
a pan diagonal.
2. Every 2 ? 2 square on the wallpaper also adds to 30. Imagine
that the square wraps around horizontally and vertically to form
the surface of a torus. Any 2 ? 2 square on this torus will have
a sum of 30.
3. Along every diagonal of the wallpaper, any two cells separated
by one cell add to 1 5 .
If rotations and reflections are included, there are 8 ? 48 384 order-4
pandiagonal squares. If a square' s order is even, it must be a multiple
of 4. If its order is odd, it can be any number.
A pandiagonal square is "most-perfect" if, like the order-4 pandi
agonals, all its rows, columns, and diagonals (main and broken) add
to the magic constant, and all the 2 ? 2 squares on its wallpaper pat
tern have the same sum. If the sguare starts with 0, its constant is
1/2 (n3 + n) - n, and each 2 ? 2 square adds to 2 (n
2
- 1 ) . moreover,
along every diagonal (main and broken) , numbers that are n/2 cells
apart add to n2 - 1 .
Although all order-4 pandiagonal squares have been known to be
perfect for three centuries, very little has been known about most
perfect squares of higher orders. There was no known method for con
structing all of them or determining the number of such squares for
any given order. These were the two unanswered questions finally
settled by Kathleen Ollerenshaw. Her solution is the main topic of a
wonderful little book titled Most-Perfect PandiagonaJ Magic Squares:
Their Construction and Enumeration. Written with David Bne, and
with a foreword by the cosmologist Sir Hermann Bondi, the book is
scheduled for publication by the Institute of Mathematics and Its Ap
plication, at South-end-on Sea, Essex.
Because all most-perfect squares are pandiagonal, their even or
ders must be a multiple of 4. However, unlike the pandiagonals, they
do not exist with odd orders. Figure 3 shows a most-perfect square of
order 8. As a wallpaper pattern, every diagonal line of eight cells will
0 62 2 60 1 1 53 9 55
15 49 13 51 4 58 6 56
16 46 18 44 27 37 25 39
31 33 29 35 20 42 22 40
52 10 54 8 63 1 61 3
59 5 57 7 48 14 50 12
36 26 38 24 47 17 45 19
43 21 41 2 3 32 30 34 28
Figure 3 . A most-perfect square o f order 8. The magic constant i s 252.
add to the square' s magic constant of 252. This is the same as saying
that all its broken diagonals add to 252. Every 2 ? 2 square on the
pattern adds to 2 (8
2
- 1) 126, and every pair of numbers that are
8/2 4 cells apart along a diagonal add to 8
2
- 1 63.
Because the orders of most-perfect magic squares increase by leaps
of 4, the number of essentially different most-perfect squares increases
very rapidly as increases. The authors have determined for the first
time a method for constructing all most-perfect squares of any order,
and a complicated formula that counts the number of such squares
for any order. The number of essentially different order- 8 squares
is 368, 640. The number of order- 1 2 squares is 2. 22953 ? 10
10
. When
you reach order 36 the number is 2
.
7
6
754 ? 10
44
.
Assuming that the
Big Bang occurred 20 million years ago, the authors estimate that
the number of order- 36 most-perfect squares is more than a thousand
times the number of pico-pico- seconds since the Bang. (A pico is one
trillionth of a number. )
The author' s elegant solution of one of the most frustrating of un
solved problems in magic square theory was an achievement that
would have been remarkable for a mathematician of any age. In Dame
38. A New Res ul t C PerFec Magi c Squares 289
Kathleen' s case this was even more remarkable because she was 85
when she and Bree finally proved the conjectures she had earlier made
intuitively. Most mathematicians do their most creative work when
quite young. Dame Kathleen is a striking exception.
Kathleen Ollerenshaw is surely one of the most amazing women
in England. Born in Manchester in 1 91 2, she obtained her doctorate
in mathematics at Somerville College, Oxford. It would take many
inches j ust to list her many awards and the positions she has held
in her long and colorful life. They include president of the Institute
of Mathematics, deputy president of the University of Manchester In
stitute of Technology, and the Lord Mayor ( 1 975-76) of Manchester.
Dave Bree (bor 1 939), coauthor of the book, is a professor of artificial
intelligence at the University of Manchester.
The author' s marvelous solution to the problem of constructing and
counting all most-perfect magic squares is too technical to go into here,
but her book explains it all as clearly as possible. At the back of the
book, in a section headed "A Personal Perspective, " Dame Kathleen
concludes:
The manner in which each successive application of the
properties of binomial coefficients that characterize the Pas
cal triangle led to the solution will always remain one of the
most magical mathematical revelations that I have been
fortunate enough to experience. That this should be af
forded to someone who had, with a few exceptions, been out
of active mathematics research for over forty years will, I
hope, encourage others. The delight of discovery is not a
privilege reserved solely for the young.
ReFetences
[ 1 ] Andrews, W. S. Magic Squares and Cubes (Dover, New York, 1 960) .
[2] Benson, William A. , and Jacoby, Oswald. New Recrations with Magic Squares
(Dover, New York, 1 976) .
[3] Hirayama, Akira, and Abe, Gakuho. Researches in Magic Squares (Osaka Ky
oikutusho Co. , Osaka, Japan, 1 983. )
[4] Ollerenshaw, Dame Kathleen, and Bondi , Si r Herman. Magic squares of Order
Four. Philosophical Transactions of the Royal Society, 306, 443-552 ( 1 982).
[5] Rosser, B. , and Walker, R. J. The algebraic theory of diabolic magic squares. Duke
Mathematical J0Ual, 5, 705-728 ( 1 939).
[6] Stewart, Ian. Most-Perfect Magic Squares. Scientifc Amercan, November 1 999.
A good summary of Dame Ollerenshaw' s book.
Ch'pter 39
The Nu mber Devi l
No book about mathematics, written for young children, could less re
semble a textbook than The Number Devil (Metropolitan Books, 1 998).
The author, Hans Magnus Enzensberger, who lives in Munich, is a
writer and scholar, but not a mathematician. This may explain how
he manages to introduce number theory in such an entertaining way
that his book became a best seller in Germany. Translated by Michael
Heim, and amusingly illustrated by Rotraut Susanne Berger, this is
just the book to give to an intelligent child who falls asleep in mathe
matics classes.
Enzensberger imagines a twelve-year-old Robert who hates math
because his teacher Mr. Bockel-the name means an obstinate goat in
German-is such a stupid and dull teacher. One night, after some un
pleasant nightmares, Robert dreams about a friendly imp with shining
eyes who calls himself a number devil. In this and eleven later dreams,
the devil explains elementary number theory in such a refreshing way
that Robert, instead of being bored, is instantly intrigued.
The devil' s instructions during the first dream couldn' t be simpler.
Every integer, he explains, is reached by adding ones. Because this
process can go on forever, there must be an infinity of integers and
no such thing as a largest number. Similarly, _|_ , a generates
a infinity of smaller and smaller factions, never reaching a smallest
fraction.
When 1 1 is multiplied by 1 1 you get 1 21 , a number palindrome
that is the same backward, like such words as MADAM and ROTA
TOR. and such sentences as "Straw? No, too stupid a fad. I put soot on
warts. " Number palindromes also result from 1 1 1 ? 1 1 1 " 1232 1 , and
1 1 1 1 ? 1 1 1 1 " 1 234321 . Does this continue to produce palindromic prod
ucts as the number of ones increases? No. Robert correctly guesses
A cut version of thi s review appeared in The Los Angeles Times Book Review (No
vember 8, 1998) .
291
that the pattern fails beyond ten ones. Moral: You can' t trust a gen
eralization until it is proved.
Things are slightly more complicated in Robert' s second dream.
The number devil convinces him that Roman numerals were such a
clumsy notation that it held mathematics back for centuries. Robert
learns the value of the decimal system, with digits ordered from right
to left, the last digit indicating a multiple of one, the preceding digit
a multiple of ten, the next digit a multiple of one hundred, and so on,
with zero serving as what mathematicians call a "place holder. "
In his third dream Robert learns the importance of prime num
bers, numbers evenly divisible only by themselves and one, and how
to find them by a sieving method. He is told about a famous theorem
(known as Goldbach' s conjecture) , still not proved, that every even
numbers greater than two is the sum of at least one pair of primes.
(For example, 1 998 is the sum of primes 1 993 and 5. )
The devil reveals a curious fraction i n Robert' s fourth dream.
One divided by seven produces an endless decimal fraction
0. 1 428571 42857 1 42857 . . . . Its repeating pattern 1 42857 has the sur
prising property that when it is multiplied by any digit from one
through six the quotient has the same digits in the same cyclic or
der. Irrational fractions, which have no repeating pattern in their
decimal expansion, come next. The devil proves that the diagonal of
a square of side one has a length equal to the square root of two, an
irrational number that begins 1 . 41 42 1 3 . . . .
The devil has his own whimsical terminology. Irrational num
bers are called "unreasonable numbers. " Roots are called "rootaba
gas. " Primes are "prima donnas. " At the back of the book there is
a handy list of the devil' s terms translated into the terminology of
modern mathematics.
Triangular and square numbers appear in Robert' s fifth dream.
Triangular numbers ( 1 , 3, 6, 10, 1 5 , . . . ) are integers that can be
modeled by dots in triangular arrays, like the fifteen pool balls or the
ten bowling pins at the start of a game. Square numbers ( 1 , 4, 9, 1 6,
. . . ) are modeled by dots i n square arrays. Figure 1 shows the devil' s
elegant "look-see" proof that every partial sum of the series of odd
numbers is a square number.
The fascinating properties of Fibonacci numbers ( l, 1 , 2, 3 , 5, 8, 1 3 ,
. . . ) are the topic of the sixth dream. Each number i s the sum of its
two predecessors. They are followed in the next dream by the wonders
of Pascal' s famous number triangle. Dream 8 concerns permutations
39. T:e Numrer Devi l 293

.

1 3 5 7 9
Figure 1 . A look-see proof.
and factorial numbers such as 6! where the exclamation mark tells
you to mul ti pl y 1 ? 2 ? 3 ? 4 ? 5 ? 6. It numbers the diferent ways si x
students can sit i n a row.
Here are four of the topics covered i n the last four dreams:
I . The proof ( frst noticed, by the way, by Galileo) that there are as
many odd numbers as there are counting numbers. You simply
pair them like so:
I 2
.
+
+ 4
. . .
v
The number of numbers i n both sets i s a "transfi ni te" number
that mathematicians know as aleph null, the smallest of an inf
nite set of transfinite numbers. Although the odd numbers form
an i nfi nite set, and the counti ng numbers form another i nfi ni te
set, the two infinities have a sum that is the same transfinite
number.
2. The beautiful properti es of <, the golden rati o i . o i s. . . It i s a
famous irrational number that is the limit approached by the
ratios of adj acent Fibonacci numberso
+. The five Platonic solids, the only convex polyhedrons with faces
that are regular polygons. The devil teaches Robert how to make
them by cutting and folding paper.
4. The difficult and as yet unsolved task of finding the shortest
route that visits each of n cities when n is large. Mathematicians
call it the "traveling salesman problem. "
In the final dream the devil takes Robert to a party where he meets
some eminent mathematicians of the past. They include Lord Rus
tle (Bertrand Russell) , and Happy Little (Felix Klein) . Little shows
Robert a strange glass bottle he has invented. Its edgeless surface is
closed like the surface of a sphere, yet it has no inside or outside. It
is a one- sided surface like the surface of a Mobius band.
The book ends with Robert back in Mr. Bockel' s class, where he
surprises the teacher by finding a rapid way to solve a problem. If
Mr. Bockel gives a pretzel to one student, two to another, three to a
third student, and so on for - s students, how many pretzels does he
hand out?
Adults who know little about math will find this book as enlight
ening as will younger readers. It closes with a valuable index that
lists in standard terminology all of the number devil' s topics.
chqpter 40
Probq bi l it
y
1
If one is on a hunt, it is better
not to assume at the start that there is no game, or,
you won' t get what little there is.
-William James ( i n a 1 902 letter)
The debate over whether there is intelligent life elsewhere in the uni
verse is an ancient one. Greek and Roman atomists (Leucippus, Dem
ocritus, Epicurus, Lucretius) all defended a plurality of worlds. Plato
and Aristotle thought otherwise. Medieval Christian thinkers, from
Augustine to Aquinas, followed Aristotle and the Bible by confining
mortal life to Earth. Pluralism revived during the Renaissance, no
tably in the writings of the German cardinal Nicholas of Cusa and
the Italian ex-monk Giordano Bruno. (It is often said that Bruno was
burned at the stake for his belief in many worlds, but it was mainly
for other heresies. )
The Copernican revolution, which removed Earth from the center
of the cosmos, surely played a role in the intense upwelling of plu
ralism in post-Renaissance centuries. Kant, Newton, Pope, Voltaire,
Paine, Emerson, together with hundreds of writers and scientists, ea
gerly embraced the notion that intelligent life was everywhere, most
likely on Mas, perhaps even on the moon. The leaders of America' s
two greatest Adventist movements-Joseph Smith, who started Mor
monism, and Ellen White, founder and prophetess of the Seventh-day
Adventists-each defended a plurality of inhabited worlds. White, like
the Swedish mystic Emanuel Sweden borg, even had visions of human
like beings on other planets. (Two eminent English exceptions to this
belief were the naturalist Alfed Russel Wallace and philosopher of
science William Whewell . Each argued strenuously against the possi
bility of extraterrestrial life. )
This review appeared i n The Los Angeles Times Book Review ( November 29, 1 998) .
295
A widespread view among today' s scientists is that the universe
probably teems with life. There are billions of galaxies, each with
billions of stars, and in recent years evidence has been increasing for
a plenitude of other solar systems. So far only planets huge enough
to cause detectable wobblings of their mother suns have been found,
but there is every reason to believe that most suns, perhaps all suns,
have planets of all sizes orbiting them.
In recent years, Carl Sagan was the most vocal scientist to trum
pet a belief in the plurality of inhabited planets and to urge continual
funding of searches for radio messages from extraterrestrial intelli
gence. Frank Tipler, a physicist at Tulane University, is Sagan' s
chief detractor. Tipler thinks the probability of extraterrestrials is
zero and that listening for their signals is a big waste of time and
money. Less aggressive skeptics include physicist John Wheeler, bi
ologists Ernst Mayr and Jacques Monod, mathematician John Casti,
astronomer Martin Rees and a raft of other scientists.
Amir D. Aczel, a statistician at Bentley College in Waltham, Mass. ,
whose previous book, Fermat's Last Theorem, won a 1 996 Los Angeles
Times Book Prize, is even more certain than Sagan that sentient life is
not limited to Earth. In his new book, Probabilty 1 (Harcourt Brace,
1 998) as its title suggests, he maintains that the probability of beings
on at least one other planet is so close to 1 as to be indistinguishable
from certainty. Although his book covers the same ground as earlier
books by others, there are two reasons for recommending it to any
person interested in the debate: It is clearly and gracefully written,
and i t is up to date in its astronomical data.
In 1 950, Aczel reminds us, the Italian physicist Enrico Fermi asked
a famous question now called Fermi' s paradox. The universe is so
vast and so old, Fermi said, that if intelligent life is out there, we can
expect many civilizations to have a technology far in advance of our
own. Long ago they should have visited Earth. "Where are they?"
Fermi asked. Because we have not seen them or heard from them,
they probably don' t exist. Like almost all scientists, Fermi dismissed
UFO mania as popular superstition.
Many scientists refused to buy Fermi' s argument. There are too
many reasons, they said, why extraterrestrials would find it dificult
to cover the vast distances between solar systems. Moreover, why
would they select Earth from billions of other planets in our galaxy?
Aczel retells the story of how, in the late 1 950s, physicist Philip
Morrison, astronomer Frank Drake and others concluded that ex-
40. Probabi l ity 1 297
traterrestrials might be sending messages using a certain frequency
of radio waves. Drake' s Project Ozma, named after the Princess of
Oz in L. Frank Baum' s popular Oz series, was the first attempt to
listen for such messages using a radio telescope at Green Bank, W.
Va. Others soon joined in the search. In 1 993, a skeptical Congress
stopped funding these searches. Drake, Aczel tells us, has continued
his work with private funding from corporations. The most eficient
search, SERENDIP III, is being conducted by Stuart Bowyer, using the
world' s largest radio telescope in Arecibo, Puerto Rico. Since 1 992, it
has examined 500 trillion signals without detecting a message.
In subsequent chapters, Aczel skillfully covers the recent discover
ies of some dozen large planets circling nearby suns, theories about
how life arose on Earth and the remote possibility that life could be
based on elements other than carbon, as found on Earth. Aczel con
siders silicon and sulfur, concluding that neither has anywhere near
the flexibility of carbon in forming large protein molecules so essential
for the specialized tasks they perform in Earth' s life forms. If there
is extraterrestrial life, Aczel concludes, it is probably carbon-based.
Another chapter deals with "panspermia," the conjecture that life on
Earth began with a seeding of organic material carried here by comets
or meteors. This old conjecture has been revived by astronomer Fred
Hoyle and biologist Francis Crick.
The famous Mars meteorite discovered in Antarctica in 1 996, con
taining what could be fossil bacteria, is evaluated by Aczel and found
to be inconclusive. A chapter is devoted to the influence on Earth' s
evolving life by the impacts of large meteors, asteroids and comets.
Aczel defends the view that intelligence is not just an accidental prod
uct of evolution, as Stephen Jay Gould and others believe, but the
inexorable outcome of a tendency for life, once it starts, to evolve crea
tures with ever superior brains.
The chapter "Does God Play Dice?" introduces the random aspects
of quantum mechanics, which Einstein found so distasteful, along with
the determinism of chaos theory and Benoit Mandelbrofs fractals.
These topics lead into deep questions about why evolution on Earth
produced creatures capable of creating art, music, poetry, mathemat
ics, science and philosophy, none of which seems to have much Dar
winian survival value. Are they mere byproducts of a brain designed
by nature only to increase eficiency in hunting, fishing, fighting and
growing food? Or do they have survival value in some deeper sense
not yet understood?
Another chapter concers what statistICIanS call the "inspection
paradox. " Assume, writes Aczel, that a bus arrives at a bus stop on the
average of every 1 0 minutes. You go there at random moments. One
might suppose that the average waiting time would be five minutes,
but this is not the case. It' s a trifle longer than five minutes. Why? If
the bus arrived at the stop at exactly 1 0-minute intervals, the average
wait would indeed be five minutes. The bus, however, arrives at 1 0-
minute intervals only on the average. If you go to the stop at a random
time, the probability is higher that you will arrive at the stop within a
long time interval between bus arrivals than during a short interval.
Your average wait, therefore, will be longer than five minutes.
To further clarify the paradox, Aczel gives another example. As
sume that a fashlight battery has an average life of - hours. Each
time you pick up the flashlight you are selecting a random point of
time between the random durations between battery use. Like the
bus rider, you are more likely to pick up the flashlight between long
time intervals than short ones. Result: Your battery will last a trifle
longer than its average life.
Aczel applies these examples to Fermi' s question. "If your arrival
on Earth is viewed as a random event, you are more likely to land on
a longer-lived planet than a shorter-lived one in the same way that a
dart is more likely to land on a wider section of the dart-board than
on a narrower one. " Our sun is known to have a longer-than-average
lifetime. "It is very likely that, as galactic civilizations go, we are
on the above-average development level, and possibly way up there
among the most advanced . . . . [W] e may be one of the most advanced
civilizations anywhere. "
The book' s final chapter is an effort to prove that the probability
of life on at least one other planet, though not absolutely certain, is
so close to I that we can assume certainty. This argument requires
a long chain of probability questions that ae not, as mathematicians
like to say, well-formed. It seems likely there are billions of planets
in every galaxy, but astronomers do not yet know how probable it
is that a planet meets all the conditions necessary for life. A planet
must be neither too hot nor too cold (Aczel calls this the Goldilocks
criterion) . Does it have water? A suitable atmosphere? How likely is
it that carbon is present? Because we do not know the probability that
a planet will have all these features, it is impossible, in my opinion,
to make a reliable estimate of the chances that conditions on other
planets will be conducive to life.
40. Probabi l ity 1 299
Even if all necessary environmental conditions are present, no one
has the slightest notion of how to estimate the probability that car
bon molecules, jiggling in a primordial soup, will form self-replicating
molecules. It may turn out, as more is learned, that under ideal condi
tions, it is almost certain that DNA molecules will appear. Or it may
turn out, as antipluralists insist, that the chance of this happening
is so close to zero that it is extremely unlikely that life has arisen
anywhere else in our galaxy, perhaps not even in other galaxies.
The view that we are alone in the universe will, of course, be dealt
a near-fatal blow if our space probes find some form of life on Saturn' s
moon Titan, so rich in atmosphere, or in the seas below the surface
ice of Jupiter' s moon Europa. But even if there are low forms of life
on Titan or Europa, they will tell us nothing about the probability
that given enough time, these life forms will evolve into creatures as
intelligent as ourselves. "A sad spectacle! " wrote Thomas Carlyle as
he considered the possibility of extraterrestrial life on other planets.
"If they be inhabited, what a scope for pain and folly; and if they be
not inhabited, what a waste of space! "
I for one would be delighted, albeit a bit terrified, i f tomorrow we
received a radio signal from outer space that repeated the counting
numbers or the prime numbers. As several scientists have warned:
"If we hear such a message, don' t answer' " I am equally undismayed
by the prospect that Earth is the only planet with life (although we
could never be sure of this) and that we are destined to colonize our
galaxy. It is said that William Randolph Hearst once cabled a famous
British astronomer to request 1 000 words on the question "Is there
intelligent life elsewhere in the universe? Hearst got back an answer
in which "Don' t know" was repeated five hundred times.
I, too, have often felt the emotions of that anonymous cowboy who
wrote "Home on the Range":
How often at night,
when te heavens are bright
wit te light from te glittering stars,
have I stood there amazed
and asked as I gazed,
does their glory exceed that of ours ?
chqpter 41
Fuzzy New New Mqth
Multicultural and Gender Equity in the Mathematics Class
rom: The Gift of Diversity ( 1 997 Yearbook) ; edited by Janet
Trentacosta and Margaret J. Kenney. National Council of
Teachers of Mathematics, 248 pp. , $22. 00
Focus on Algebra: An Integrated Approach by Randall I.
Charles, Alba Gonzalez Thompson, et al. Addison-Wesley,
843 pp. , $56. 00
Life by the Numbers: Mat As You' ve Never Seen It Before
narrated by Danny Glover. Seven boxed videotapes pro
duced by WQED, Pittsburg, $ 1 29. 00
Surveys have shown for many decades that the mathematical skills
of American high school students lag far behind those of their coun
terparts in Japan, Korea, Singapore, and many European countries.
In the United States whites do better than blacks, Hispanics, and
Native Americans. Males outscore females. Students from high so
cioeconomic backgrounds do better than those from lower strata.
These are troubling statistics because, in an advanced technologi
cal society such as ours, a firm grasp of basic mathematics is increas
ingly essential for better-paying jobs. Something clearly is wrong with
how math is being taught in precollege grades, but what?
In the late 1 960s the National Council of Teachers of Mathematics
(NCTM) began to promote a reform movement called the New Math.
In an efort to give students insight into why arithmetic works, it
placed a heavy emphasis on set theory, congruence arithmetic, and
the use of number bases other than ten. Children were forbidden to
call, say, 7 a "number. " It was a "numeral" that symbolized a number.
This review appeared in The New York Review of Books ( September 24, 1998) .
301
The result was enormous confusion on the part of pupils, teachers, and
parents. The New Math fad faded after strong attacks by the physicist
Richard Feynman and others. The final blow was administered by the
mathematician Morris Kline' s 1 973 best seller lJy Johnny Can' t Add:
The Failure of the New Math.
Recently, the NCTM, having learned little fom its New Math fi
asco, has once more been backing another reform movement that goes
by such names as the new new math, whole math, fuzzy math, stan
dards math, and rain forest math. Like the old New Math, it is cre
ating a ferment among teachers and parents, especially in California,
where it first caught on. It is estimated that about half of all precol
lege mathematics in the United States is now being taught by teachers
trained in fuzzy math. The new fad is heavily influenced by multicul
turalism, environmentalism, and feminism. These trends get much
attention in the twenty-eight papers contributed to the NCTM' s 1 997
yearbook, Multicultural and Gender Equity in the Mathematics Class
room: The Gift of Diversity.
It is hard to fault most of this book' s advice, even though most
of the teachers who wrote its chapters express themselves in mind
numbing j argon. "Multiculturalism" and "equity" are the book' s most
used buzzwords. The word "equity, " which simply means treating all
ethnic groups equally, and not favoring one gender over another, must
appear in the book a thousand times. A typical sentence opens Chap
ter Eleven: "Feminist pedagogy can be an important part of building
a gender-equitable multicultural classroom environment. " Over and
over again teachers are reminded that if they suspect blacks and fe
males are less capable of understanding math than Caucasian males,
their behavior is sure subtly to reinforce such beliefs among the stu
dents themselves, or what one teacher calls, in the prescribed j argon,
a student' s "internalized self-image. "
"Ethnomathematics" is another popular word. It refers to math as
practiced by cultures other than Western, especially among primitive
African tribes. A book much admired by fuzzy-math teachers is Mar
cia Ascher' s Ethno-mathematics: A Multicultural View of Mathemat
ical Ideas ( 1 991 ) .
1
"Critical-mathematical literacy" is an even longer
j awbreaker. It appears in the NCTM yearbook as a term for the ability
to interpret statistics correctly.
Knowing how pre-industrial cultures, both ancient and modern,
handled mathematical concepts may be of historical interest, but one
must keep in mind that mathematics, like science, is a cumulative
41 . Fuzz
y
New New Math 303
process that advances steadily by uncovering truths that are every
where the same. Native tribes may symbolize numbers by using dif
ferent base systems, but the numbers behind the symbols are identi
cal. Two elephants plus two elephants makes four elephants in every
African tribe, and the arithmetic of these cultures is a miniscule por
tion of the vast jungle of modern mathematics. A Chinese mathe
matician is no more concerned with ancient Chinese mathematics, re
markable though it was, than a Western physicist is concerned with
the physics of Aristotle.
Fuzzy-math teachers are urged by contributors to the yearbook to
cut down on lecturing to passive listeners. No longer are they to play
the role of "sage on stage. " They are the "guide on the side. " Classes
are divided into small groups of students who cooperate in finding
solutions to "open-ended" problems by trial and error. This is called
"interactive learning. " The use of calculators is encouraged, along
with such visual aids as counters, geometrical models, geoboards, wax
paper (for folding conic section curves) , tiles of diferent colors and
shapes, and other devices. Getting a correct answer is considered
less important than shrewd guesses based on insights, hence the term
"fuzzy math. " Formal proofs are downgraded.
No one can deny the usefulness of visual aids. Teachers have
known for centuries that the best way to teach arithmetic to small
children is by letting them "interact" with counters. Each counter
models anything that retains its identity-an apple, cow, person, star.
What' s the sum of 5 and 2? A girl who knows how to count moves into
a pile five counters, then two more, and counts the heap as seven.
Suppose she first moves two, then five. Does it make a difference?
Similar procedures teach subtraction, multiplication, and division.
After a few days of counter playing it has been traditional for chil
dren to memorize the addition table to at least 9. Later they learn the
multiplication table to at least 1 0. "Hands-on" learning first, then rote
learning. Unfortunately, some far-out enthusiasts of new new math
reject anything resembling what they call "drill and kill" memorizing.
The results, of course, are adults who can' t multiply 1 2 by 1 2 without
reaching for a calculator.
Aside from its j argon, another objectionable feature of the yearbook
is that its contributors seem wholly unaware that the best way to keep
students awake is to introduce recreational material that they perceive
as fun. Such material includes games, puzzles, magic tricks, fallacies,
and paradoxes. For example, determining whether the first or second
player can always win at tic-tac-toe, or whether the game is a draw
if each player makes the best moves, is an excellent way to introduce
symmetry, combinatorics, graph theory, and game theory. Because all
children know the game, it ties strongly into their experience.
For what the yearbook likes to call a "cognitively challenging" task,
give each child a sheet with a checkerboard on it. Have each of them
cut of two opposite corner squares. Can the remaining sixty-two
squares be covered by thirty-one dominoes? After a group finds it
impossible, see how long it takes for someone to come up with the
beautiful parity (odd-even) proof of impossibility.
If new new math teachers are aware of such elegant puzzles, and
there are thousands, there is no hint of it in the yearbook. This is
hard to understand in view of such best- selling textbooks as Harold
Jacobs' s Mathematics: A Human Endeavor ( 1 970; third edition, 1 994) ,
which has a great deal of recreational material; Mathematics: Prob
lem Solving Through Recreational Mathematics, a textbook by Bonnie
A verbach and Orin Chein ( 1 980); and scores of recent books on enter
taining math by eminent mathematicians.
I seldom agree with the conservative political views of Lynne Ch
eney, but when she criticized extreme aspects of the new new math
on the Op-Ed page of The New York Times on August 1 1 , 1 997,
2
I
found myself cheering. As Cheney points out, at the heart of fuzzy
math teaching is the practice of dividing students into small groups,
then letting them discover answers to problems without being taught
how to find them. For example, teachers traditionally introduced the
Pythagorean theorem by drawing a right triangle on the blackboard,
adding squares on its sides, and then explaining, perhaps even prov
ing, that the area of the largest square exactly equals the combined
areas of the two smaller squares.
According to fuzzy math, this is a terrible way to teach the the
orem. Students must be allowed to discover it for themselves. As
Cheney describes it, they cut from graph paper squares with sides
ranging from two to fifteen units. ( Such pieces are known as "manip
ulatives. ") Then they play the following "game. " Using the edges of
the squares, they form triangles of various shapes. The "winner" is the
first to discover that if the area of one square exactly equals the com
bined areas of the other two squares, the triangle must have a right
angle with the largest square on its hypotenuse. For example, a trian
gle of sides - , 4, 5 . Students who never discover the theorem are said to
41 . Fuzz
y
New New Math 305
have "lost" the game. In this manner, with no help from teacher, the
children are supposed to discover that with right triangles a2 + b2 c2
"Constructivism" is the term for this kind of learning. It may take
a group several days to "construct" the Pythagorean theorem. Even
worse, the paper game may bore a group of students more than hearing
a good teacher explain the theorem on the blackboard.
One of the harshest critics of fuzzy math is the writer John Leo,
whose article on the subject, "That So-Called Pythagoras," was pub
lished last year in US News and World Report (May 26, 1 997). (His
title springs from a reference he found in a book on ethno-mathematics
to "the so-called Pythagorean theorem. ") Leo tells of Marianne J en
nings, a professor at Arizona State University, whose daughter was
getting an A in algebra but had no notion of how to solve an equa
tion. After obtaining a copy of her daughter' s textbook, Jennings soon
understood why. Here is how Leo describes this book:
It includes Maya Angelou' s poetry, pictures of President
Clinton and Mali wood carvings, lectures on what envi
ronmental sinners we all are and photos of students with
names such as Taktuk and Esteban "who ofer my daugh
ter thoughts on life. " It also contains praise for the wife of
Pythagoras, father of the Pythagorean theorem, and asks
students such mathematical brain teasers as "What role
should zoos play in our society? However, equations don' t
show up until Page 1 65, and the first solution of a linear
equation, which comes on Page 21 8, is reached by guessing
and checking.
Romesh Ratnesar' s article "This is Math? ( Time, August 25, 1 997)
also criticizes the new new math. It describes fifth-graders who were
asked how many handshakes would occur if everyone in the class
shook hands with everyone else. At the end of an hour, no group
had the answer. Unfazed, the teacher said they would be trying again
after lunch. Professor Jennings makes another appearance. She told
Ratnesar that she became angry and worried when she saw her daugh
ter use her calculator to determine 10 percent of 470.
Curious about her daughter' s textbook, which is now widely used,
I finally obtained a copy by paying a bookstore $59. 1 2. Titled Focus on
Algebra: An Integrated Approach, this huge text contains 843 pages
and weighs close to four pounds. (In Japan, the average math textbook
is two hundred pages. ) It is impossible to imagine a sharper contrast
with an algebra textbook of fifty years ago.
"Integrated" in the subtitle has two meanings:
( 1 ) Instead of being limited to algebra, the book ranges all over
the math scene with material on geometry, combinatorics, probability,
statistics, number theory, functions, matrices, and scatter graphs, and
of course the constant use of calculators and graphers. Fifty years ago
high school math was given in two classes, one on algebra, one on
geometry. Today' s classes are "integrated" mixtures.
( 2) The book is carefully integrated with respect to gender and
to ethnicity, with photographs of girls and women equal in number to
photographs of boys and men. Faces of blacks and whites are similarly
equal, though I noticed few faces of Asians.
On the positive side is the book' s lavish use of color. Only a few
pages lack full-color photos and drawings, all with eye-catching lay
outs. When it comes to actual mathematics the text is for the most
part clear and accurate, with a strong emphasis on understanding why
procedures work, and on inducements to think creatively. "Afer all,"
the text says on its first page, "what good is it to solve an equation if
it is the wrong equation?" The trouble is that the book' s mathematical
content is often hard to find in the midst of material that has no clear
connection to mathematics.
Not having taught mathematics myself, I have no opinion about
the value of students working in small groups as opposed to sitting
and listening to a teacher talk. Nor have I found research studies
that make a decisive case in favor of either method. Clearly a great
deal depends on the qualities of particular teachers, and these would
be hard to appraise in any survey. The authors justify the group
approach by saying it anticipates the workplaces in which students
will find themselves as adults. John Donne' s remark about how no
man is an island is quoted. The book' s first "exercise" is a question:
"In general, do you prefer to work alone or in groups?"
An emphasis on ethnic and gender equity is, of course, admirable,
though in this textbook it seems overdone. For example, twelve faces
of boys and girls of mixed ethnicity reappear in pairs throughout the
pages. Each has something to say. "Taktuk thinks . . . " is followed
with "Esteban thinks . . . , " "Kirti thinks . . . " is followed by what "Keisha
thinks . . . ," and so on. These pairings become mechanical and pre
dictable.
The book jumps all over the place, with transitions as abrupt as the
dream episodes of Alice in Wonderland. I think most students would
find this confusing. Eight full pages are devoted to statements by adult
41 . Fuzz
y
New New Math 307
professionals, with their photographs. Each statement opens with a
sentence about whether they liked or disliked math in high school,
followed by generally banal remarks. For example, Diana Garcia
Prichard, a chemist, writes: "I liked math in high school because all
the problems had answers. Math is part of literacy and the framework
of science. For instance, film speed depends on chemical reactions. I
use math to model problems and design experiments. I like getting
results that I can publish and share. " Presumably such statements
are intended to convince students that math will be useful later on
in life.
Many of the book' s exercises are trivial. For example, on page 20
students are asked to play forward and backward a VCR tape of a
skier, then answer the question: "How will this afect the way the
skier appears to move?" On page 1 1 : "A circle graph represents 1 80
kittens. What does 1 14 of the circle represent?" (Answer, to be found
in the back of the book: 45. ) A chapter on "the language of algebra"
opens with a page on the origin of such phrases as "the lion' s share, "
"the boondocks, " and "not worth his salt. " It is not clear what this has
to do with algebra.
Many pictures have only a slim relation to the text. Magritte' s
painting of a green apple foating in front of a man' s face accompanies
some problems about apples. Van Gogh' s self-portrait is alongside
a problem about the heights and widths of canvases. A picture of
the Beatles accompanies a problem about taxes only because of the
Beatles' song "Taxman. " My favorite irrelevant picture shows Maya
Angelou talking to President Clinton. Beside it is the following extract
from one of her prose poems:
Lift up your eyes upon
This day breaking for you
Give birth again
To the dream.
Women, Children, Men,
Take it into the palm of your hands.
Mold it into the shape of your most
Private need. Sculpt it into
The image of your most public self.
Why is this quoted? Because the "parallel" phrases shown under
lined are similar to parallel lines in geometry! Is this intended to
"integrate" geometry and poetry?
The book is much concerned with how the environment is being
polluted. Protecting the environment is obviously a good cause, but
here its connection with learning math is often oblique, if not arbi
trary. A chapter on functions opens with a page headed "Unstable
Domain. " Its first question is "What other kinds of pollution besides
air pollution might threaten our planet?" Page -ohas a picture of
crude oil being poured over a model of the eath. It accompanies a set
of questions relating to the way improper disposals of oil are contam
inating ground water.
A page headlined "Hot Stuff' shows three kinds of peppers to illus
trate how they are used in cooking. Two of the "exercises" are: "The
chili cook-off raises money for charity. Describe some ways the orga
nizers could raise money in the cook-of," and "How would you set up
a hotness scale for peppers?" This page introduces a chapter on how
to solve linear equations.
Another section on equations opens with pictures of zoo animals. It
discusses what can be done to prevent species from becoming extinct.
The first question is "What role should zoos play in today' s society?"
The book' s index, under the entry "Animal study and care, " lists thirty
two page references.
A section on mathematical inequalities is preceded by a page on
how Mary Rodas became vice-president of a toy company, and how
Linda Johnson Rice found a creative way to market Ebone cosmetics
for black women. Under a photo of a smiling Mary, the first questions
are: "Would you like to own your own business someday? Why or
why not?"
On page 67, a picture of Toni Morrison is used to illustrate a prob
lem about how many ways four obj ects can be placed in a row. The text
then introduces four students who each read an excerpt from some
thing Morrison has written. In how many different orders, the text
asks, can the four excerpts be read? A man from Mysore, India, who
creates shadow pictures on the wall with his fingers is featured on
page 421 . What this has to with the following section on solving sys
tems of inequalities is not evident. A photo of Alice Walker on page
469 illustrates the question: "Is the time it takes to read an Alice
Walker novel always a function of the number of pages?" This and
other such references give the impression that well-known writers are
being dragged into the text.
The most outrageous page-it opens a section on linear functions
concerns the Dogon culture of West Afica. Students are told that this
41 . Fuzz
y
New New Math 309
primitive tribe, without the aid of telescopes, discovered that Jupiter
has satellites, that Saturn has rings, and that an invisible star of
great density orbits Sirius once every fifty years. Presumably the
Dogon had supernormal powers. It has long been known, however,
that the Dogon made no such discoveries. They merely leaned these
astronomical facts fom missionaries and other Western visitors.
3
Like the authors of the NCTM yearbook, those who fashioned this
huge textbook seem wholly uninterested in recreational material. The
book' s only magic trick (page 246) is a stale, utterly trivial way to
guess a number. Although strongly favoring the use of calculators,
the authors don' t seem aware that the hundreds of amazing number
tricks that can be done with them provide excellent exercises. A child
can learn a lot of significant number theory in discovering why they
work. None is in the book.
An old brain teaser involves a glass of wine and a glass of water.
A drop is taken from the wine and added to the water. The water is
stirred, then a drop of the mixture goes back to the wine. Is there now
more water in the wine than wine in the water, or vice versa? The
surprising answer is that the two amounts are precisely equal.
Students will be fascinated by the way this principle can be mod
eled with a deck of cards. Divide the deck in half, one half consisting of
all the red cards, the other half consisting of all the black cards. Take
as many cards as you like from the red (wine) half and insert them
anywhere among the blacks (water) . Shufe the black half. From
it remove from anywhere the same number of cards you took from
the reds and put them back among the reds. You' ll find the number
of blacks among the reds is exactly the same as the number of reds
among the blacks. Students will enj oy proving that this is always the
case. But will it work if the two starting portions of the deck are un
equal? (Yes. It doesn' t matter if the two glasses in the brain teaser
are not the same size; nor does it matter how many cards are in the
black and red piles. )
This secondary math textbook has an index that is not very helpful.
What value are more than l sopage references for the entry "Science"?
What use is a similar quantity of page numbers for the entry "Indus
try"?
WQED' s boxed set of seven video-tapes, Life by the Numbers, was
funded mainly by the National Science Foundation and the Alfred
P. Sloan Foundation. The photography is superb. There are scenes of
men and women mathematicians seated at computer consoles, driving
cars, or walking down a street or through the woods. There are many
close-ups of their faces, dazzling glimpses of mountains and skyscrap
ers, baseball games, martial arts contests, blossoming flowers, wild
animals, and everything imaginable that has little to do with math.
The tapes rate high on special efects, low on mathematical content.
The seventh tape covers a typical new new math class. To discover
that the longer a pendulum, the slower its swing, students tie weights
to the ends of string and swing the weights back and forth while other
students keep charts of string lengths and pendulum periods. After
several days they learn that the period of a pendulum is a function of
its length. This discovery enables them to calculate whether the victim
in Poe' s horror story "The Pit and the Pendulum" has enough time to
escape from the huge pendulum which threatens to cut him in half as
it swings lower and lower over his reclining body. It is assumed that
because students have fun swinging weights they will remember the
function better than if a teacher takes a few minutes to demonstrate
it by swinging a weight and slowly lengthening the string.
One of the most telling attacks on new new math is Bernadette
Kelly' s article "Dej a Vu? The New 'New Math, ' " in the professional
journal Efective School Practices (Spring 1 994) . Kelly summarizes
four case studies by four supporters of fuzzy math in which they report
on four fifth-grade teachers.
4
Two of the teachers, called Sandra and
Valerie, are enthusiastic users of new new math techniques. The other
two teachers use traditional methods.
Sandra was very good at getting students to cooperate in groups.
However, in one exercise she told students that one could obtain the
perimeter of a rectangular field by mUltiplying its length by its width'
In another proj ect she calculated the volume of a sandbox by multi
plying together its length and width in yards, then mUltiplying the
product by the box' s height in feet!
In an interview Sandra said that while working on the sandbox
problem her pupils asked what a cubic foot was. "You know, the thing
is that I couldn' t really answer that question. Then I thought and
thought, then I remembered how to measure a cube. " Neither Sandra
nor her students were ever aware of her two huge mistakes. In spite
of these errors, the author of the article about her said she was an
"exemplary teacher. " Sandra is praised for getting her students to
enjoy their cooperative eforts to solve problems "in the context of real
world situations. " Finding a correct answer was less important than
having fun in working on the problem.
41 . Fuzz
y
New New Math 311
Valerie made an equally astonishing blunder. The task was to
determine the average number of times her thirty students had eaten
ice cream over a period of eight days. This was "solved," by dividing
-oby s,to get - , which Valerie rounded up to 4!
As with Sandra, neither Valerie nor her students ever became
aware that they obtained a totally wrong answer. Nevertheless, the
author of the paper about her forgives her mistake on the grounds
that she had succeeded so well in getting her students to work on a
problem in the context of their experience. Moreover, the work had
impressed on the students the "usefulness and relevance of averages. "
No matter that they completely failed to find an average.
As for Jim and Karen, the two teachers who used more traditional
methods, the authors of the case studies are unimpressed by their stu
dents having scored high on tests. Both are castigated for failing to
appreciate the methods of the new new math. What is deplorable, as
Bernadette Kelly' s article points out, is not so much that the case stud
ies revealed the incompetence of two teachers, who come through as
ignoramuses, as the authors' praise of Sandra and Valerie for finding
ways to get their pupils working joyfully on problems. Little wonder
that new new math is called fuzzy. Insights are deemed significant
even when they are wrong.
The mathematician Sherman Stein, i n his 1 996 book Stengt in
Numbers, devotes a chapter to a history of math reform movements.
His hopes for the new new math are dim. "I am disturbed, " he writes,
that the authors of the [new new math books] do not cite
any pilot project or any school district as a model to show
that their goals can be achieved in the real world. That
means that they are proposing to change the way an en
tire generation learns mathematics without checking the
feasibility of their recommendations. A manufacturer in
troduces a new soap with more care, first testing its re
ception in a few stores or towns before committing to mass
production.
But evaluating the efficacy of fuzzy math will not be easy. Too
many variables are involved, including the skill of teachers and the
educational background of parents, to mention only two. A glaring ex
ample of how research can be biased is provided by a recent testing of
precollege math students around the world by the Third International
Mathematics and Science Study. Results announced last. February
revealed that American students did better than students in just two
other countries, Cyprus and South Afica. A cartoon in The New York
Times (March s) showed a car' s bumper sticker that said "My kid' s
math scores beat kids in Cyprus and South Africa. " Inside the car the
son is giving a thumbs-up sign.
These statistics are worthless. In many cases the students in a
foreign country were much older than students here at the same grade
level. More significantly, in most foreign nations students in early
grades who show no aptitude for math are sent of to trade schools or
to jobs, if they can find them. In the US such students are required
to continue attending high school. Obviously our high school students
will do less well on math tests than students in countries where poor
students are quickly moved out of the system.
Although we lack clear, systematic evidence that methods of fuzzy
math are inferior to older methods, education oficials in California,
the nation' s largest customer for math textbooks, have suddenly turned
against the new new math. The change in state policy was mainly in
response to the outrage of parents who complained that their children
were unable to do the simplest arithmetical calculations. Their out
rage was backed by many top mathematicians and scientists. Michael
McKeown, for example, a distinguished molecular biologist at the
Salk Institute, heads a parental group called Mathematically Correct.
"We' re not opposed to teaching concepts, " he told Newsweek ("Sub
tracting the New Math, " December 1 5, 1 997). "I am opposed to failing
to give a kid tools to solve a problem. "
In a vote of ten to zero (one person abstained) the eleven members
of California' s Board of Education recommended this spring a broad
return to basics in math teaching. The decision is sure to have an
efect in other states. The board said students should learn the multi
plication table by the end of the third grade, and that fourth-graders
should know how to do long division without consulting a calculator.
It banned the use of calculators on state tests. Teachers were urged
not to introduce calculators before grade six.
Defenders of fuzzy math are, of course, dismayed. They branded
the board' s decisions a product of nostalgia, and a contribution to our
country' s dumbing down. The National Science Foundation, which
has given more than :omillion to California districts for research on
new math teaching, is furious. It has threatened to withdraw further
funding to any California district that adopts the board' s recommen
dations.
41 . Fuzz
y
New New Math 313
The conflict is bitter and far from over. It may be many years be
fore it becomes clear how to sift out from the new new math what is
valuable while retaining worthy aspects of older teaching methods.
5
My own opinion is that the most important question concerning the
teaching of math is not how big and colorful textbooks are, how many
visual aids are used, how the classroom is physically arranged, or even
what methods are used in it. The greatest threat to good math teach
ing is surely the low pay that keeps so many excellent teachers and
potential teachers out of our schools. What matters more than any
thing else is having trained teachers who understand and love math
ematics, and are capable of communicating its mystery and beauty tp
their pupils.
Notes
1 . Textbooks emphasizing multiculturalism are prliferating rapidly. Here are a
few: Afca Counts: Number and Patter in Afican Culture, by Claudia Za
slavsky (Lawrence Hill, 1 997); Multiculturalism in Mathematics. Science and
Technology, by Miriam Barios-Chacon and others (Addison-Wesley 1 993); Multi
cultural Mathematics: Teaching Mathematics fom a Global Perspective, by David
Nelson, George Gheverghese Joseph, and Julian Williams (Oxford University
Press, 1 993); Teaching wit a Multicultural Perspective: A Prctical Guide, by
Leonard Davidman and Patricia T. Davidman (Perseus, 1 996). Striking multi
cultural math posters are available from teaching supply houses.
2. See also the letters in The New York Times of August 17, 1 997, and an earlier
article by Cheney in the Weekly Standard (August 4, 1 997).
3. On the myth of Dogon astronomy, see Carl Sagan, Brca's Brin (Random House,
1 979), pp. 63-64 and Chapter Six; Ian Ridpath, "Investigating the Sirius ' Mys
tery,'" in The Skeptical Inquirer, Vol. 3 (Fall 1 978), pp. 56-62; and Terence
Hines, Pseudoscience and the Paranoral (Prometheus Books, 1 988), pp. 21 6-
21 9.
4. RT. Putnam, RM. Heaton, RS. Prewat, and J. Remillard, "Teaching Mathe
matics for Understanding, " in Elementary School Joural, Vol. 93 ( 1 992), pp.
21 3-228.
5. That the new new math has positive aspects goes without saying. It is important
that students understand the basic concepts of math and not just memorize pro
cedures that work; and to give students such conceptual understanding teachers
themselves must have such understanding. This is the theme of a recent mono
graph, Middle Grade Teachers ' Mathematical Knowledge and Its Relationship
to Instrction, by Judith Sowder, Randolph Philipp, Barbara Anstrong, and
Bonnie Schappelle ( State University of New York, 1 998).
The monograph reports on a two-year investigation of five teachers, with a
primary emphasis on how they taught fractions . Why, for example, in dividing
one fraction by another do you flip upside down the divi sor fraction, then multi
ply the numerators and denominators? Should teachers be content with letting
students accept this as a trick that works like magic, or try to answer a student
who asks, "Why is thi s division?" The monograph defends the admirable aspects
of math reform without going to fuzzy-math extremes.
An anonymous piece of humor circulating on the internet purports to give capsule
summaries of how a simple problem is handled at four stages in the recent history
of mathematics teaching in the United States.
The problem: A logger sells a truckload of lumber for $1 00. His production
cost is 4/5 of the price. What is his proft?
1960, the old math wanes - - -
A logger sells a truckload of lumber for $100. His cost of production is four-ffths
of the price, or $80. What is his proft?
1970, the new math . . .
A logger exhanges a set "L" of lumber for a set "M" of money. The cardinality
of set "11" is 1 00, and each element is worth $1 . Make 100 dots representing the
elements of the set "M" . The set "C" of the costs of production contains 20 fewer
points than set "11" . Represent the set "C" as a subset of "11" , and answer the
following question: What is the cardinality of the set "P" of profts?
1980, leveling the playing feld . . .
A logger sells a truckload of lumber for $100. His cost of production is $80, and
his proft is $20. Your assignment: underline the number 20.
1990, The New New Math . . .
By cutting down beautiful forest trees, a logger makes $20. What do you think
of this way of making a living? Topic for class participation: How did the forest
birds and squirrels feel?
41 . Fuzz
y
New New Math 315
Postsci pt
The following two letters appeared in The New York Review of Books
(December -, 1 998) , with my reply:

There are many aspects of Martin Gardner' s critique of the current re
form movement in mathematics education that invite rebuttal . In this
communication I would like to take up his treatment of the Pythagor
ean theorem.
Gardner objects to having students discover this important math
ematical principle through the use of "manipulatives. " He describes
a game in which students "cut from graph paper squares with sides
ranging from two to fifeen units. . . . Using the edges of the squares,
they form triangles of various shapes. " The "winner," he says, "is the
first to discover that if the area of one square exactly equals the com
bined areas of the other two squares, the triangle must have a right
angle with the largest square on its hypotenuse . . . . ' Constructivism'
is the term for this kind of learning. It may take a group several
days to 'construct' the Pythagorean theorem. Even worse, the paper
game may bore a group of students more than hearing a good teacher
explain the theorem on the blackboard. "
I teach a college-level geometry course for future elementary teach
ers in which we spend several days studying the Pythagorean theorem
and analyzing a few of the four hundred known proofs of this impor
tant proposition. At the end of this experience, I ask students to write
in their journals about any new insights they have obtained from this
reexamination of the Pythagorean theorem.
More often than not, students write something along these lines.
"I always knew that the Pythagorean theorem was expressed by the
formula a
2
+ b
2
c
2
, but I was surprised to learn that the theorem
actually describes a relationship among areas of squares. "
Why is it that these college students find the connection between
the Pythagorean theorem and the areas of squares such a revelation?
From my experience most high school geometry textbooks prove the
theorem using properties of similar right triangles. While this is one of
the shorter proofs, it obscures the fact that the terms in the formula
a
2
+ b
2
c
2
represent areas of actual squares. Too often students
simply memorize the formula without any real understanding of what
a, b, and c stand for.
I wish our students would come to the university having learned
the Pythagorean theorem through the method that Gardner ridicules.
The activity he describes gives students the opportunity to work to
gether and look for patterns. In the process they may discover more
than just the Pythagorean theorem. They should also observe that in
obtuse triangles the square of the longest side is greater than, and
in acute triangles it is less than, the sum of the squares of the other
two sides. They should also observe or rediscover that the sum of
the lengths of any two sides must be greater than the length of the
third side.
These discoveries, however, cannot be left to chance, but are more
likely to take place under the guidance of a teacher. Contrary to Gard
ner' s assertion, cooperative learning does not imply that children re
ceive "no help" from the teacher, nor that there are "winners" who
discover what they are supposed to and "losers" who don' t. A skilled
teacher guides the entire class in a discussion of what they have ob
served and in formulating the appropriate generalizations. Gardner
is correct to observe that some students will become bored if the "ex
ploration" phase is prolonged just as many students have been bored
by brilliant lectures delivered by "good teachers. " The art of teach
ing involves determining an appropriate mixture of teacher-led and
student-directed investigation and in determining when to intervene
in cooperative group discussions.
Contrary to i ts most vocal critics, the mathematics education re
formers do not advocate that teachers abdicate their roles as leaders
in the classroom, nor that we abandon precision and rigor. It is ironic
that Gardner accuses the reform movement of promoting "fuzzy" math,
when traditional methods of teaching seem to result in so much fuzzy
understanding among our students.
Timothy V. Craine
Associate Professor
Department of Mathematical Sciences
Central Connecticut State University
New Britain, Connecticut
41 . Fuzz
y
New New Math

317
I am the author of the case of "Sandra, " one of the cases of fifth
grade teachers Martin Gardner discusses and quotes from. Gardner
draws on his knowledge of Sandra and the other three teachers from
two sources. One source is Beradette Kelly' s article (Efective School
Practices, Spring 1 994) and the other is a summary article based on
the four cases (Elementary School Joural, Volume 93, 1 992) . The
original four cases of Sandra, Valerie, Jim, and Karen, which were
written and published as four individual articles (Elementary School
Journal, Volume 93, 1 992) , are not referenced by Gardner yet he criti
cizes the authors and texts of these cases. For example, I am criticized
by Gardner for saying Sandra was an "exemplary teacher. " In the text
of my case, I accurately reported from my sources that "Sandra likes
to teach mathematics and is identified by both district personnel and
a university-based staf developer as someone who teaches in ways
consistent with the Framework. She is, from their perspective, an
exemplary teacher within her school district" (p. 1 55). The label, "ex
emplary," was one given to Sandra by sources I interviewed, not one
I attributed to her. Gardner' s other criticisms of my work are based
on statements from my case pulled out of context and similar to those
used by Kelly (Effective School Practices, Spring 1 994) .
As a qualitative educational researcher, it is my responsibility
to report what I see and to try to understand what I see from the
participant' s point of view. Sandra' s practice seemed outrageous and
perplexing to me as an observer. However, in writing the case, my
aim was to try to explain what I observed. Gardner states, "What
is deplorable, as Bernadette Kelly' s article points out, is not so much
that the case studies revealed the incompetence of two teachers, . . . as
the authors' praise of Sandra and Valerie for finding ways to get their
pupils working joyfully on problems. " What Gardner and Kelly inter
pret as praise is my effort as a researcher not to bash a teacher and to
try to understand what I observed fom Sandra' s perspective. In the
text of my case, I state, "My observations in Sandra' s classroom illus
trate what can happen when a teacher who has the best of intentions
tries to teach a concept that he or she does not understand. Although
subj ect matter knowledge alone is insuficient for teachers to teach for
understanding, it is nonetheless crucial . Sandra' s case dramatizes the
role of subject matter knowledge-its need becomes evident through
the consequences of its absence" (p. 1 6 1 ) . The title of my case, "Who
is Minding the Mathematics Content? A Case Study of a Fifth Grade
Teacher," also represents my concer over attention to subject matter
in elementary mathematics instruction. Had Gardner read the orig
inal text of my case of Sandra, I think he would have seen that he
and I agree on the need for elementary teachers to understand the
mathematics they are aiming to teach.
/ rep/lec q5 flow5:

Ruth M. Heaton, Ph. D.
Assistant Professor
University of Nebraska-Lincoln
Lincoln, Nebraska
I am even more astounded than Mr. Craine to learn that college stu
dents in a math class would not know that the Pythagorean theorem
is about geometrical squares. Surely this testifies to the miserable
quality of pre-college math teaching.
The question is how best to teach this theorem to children in lower
grades. Class time is precious. I cannot believe that a group of young
sters, struggling for days to invent the Pythagorean theorem by ma
nipulating paper cutouts, will remember the theorem any better than
hearing a good teacher explain it on a blackboard.
The University of Indiana' s mathematician/logician Raymond
Smullyan once taught high school geometry. He would chalk a right
triangle on the blackboard and its three squares. Assuming the squar
es are made fom sheets of gold, he would then ask his students which
would they rather have: the large gold square or the two smaller
squares. His students would be flabbergasted when told it made no dif
ference because regardless of the right triangle' s shape the big square' s
area always equals the total area of the two smaller ones.
I think this demonstration of the theorem would become more
firmly fixed in a child' s mind than spending days in a group trying
(perhaps unsuccessfully) to discover the theorem by themselves. More
over, it allows the teacher time to explain how the theorem applies to
semicircles and other similar shapes erected on the triangle' s sides,
and other fascinating related facts. As a homework assignment, stu
dents could be asked to search for one of the hundreds of proofs men
tioned by Craine. Unfortunately, fuzzy math educators are not much
interested in formal proofs.
41 . Fuzz
y
New New Math 319
It is true that my comments on Ruth Heaton' s article about Sandra
were based entirely on Bernadette Kelly' s paper. I am pleased that
Ms. Heaton is given this opportunity to clarify her opinion of Sandra' s
failure to understand the math she was supposed to teach.

A Gardner's Workout (Gnv64)

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

A Gardner's Workout (Gnv64)

Uploaded by

Copyright:

Available Formats

A Gqlcnelis WOlkout

r..t .ra p.sitive inte,er ,in alltuat r.ll.s inte,er ill

A:B Implication (" If_ then_" )

A" B Exclusive disjunction, non-equivalence

...... &mmm mu_ a.. ..tmm....... ... ..mm ... m m:

anc 1 /X4 . 1uis su.stuat

126 A G<rdner'5 Workout

Figure 14. Dudeney' s solution to his eight queen puzzle.

as in women, and "ti" as in nation. ) Not well known

You might also like

A Gardner&#39;s Workout (Gnv64)

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

A Gardner&#39;s Workout (Gnv64)

Uploaded by

Copyright:

Available Formats

A Gqlcnelis WOlkout

r..t .ra p.sitive inte,er ,in alltuat r.ll.s inte,er ill

A:B Implication (" If_ then_" )

A" B Exclusive disjunction, non-equivalence

...... &mmm mu_ a.. ..tmm....... ... ..mm ... m m:

anc 1 /X4 . 1uis su.stuat

126 A G<rdner'5 Workout

Figure 14. Dudeney' s solution to his eight queen puzzle.

as in women, and "ti" as in nation. ) Not well known

You might also like

A Gardner's Workout (Gnv64)

A Gardner's Workout (Gnv64)