Carlyle Circles and the Lemoine Simplicity of Polygon Constructions

Author(s): Duane W. DeTemple

Source: The American Mathematical Monthly, Vol. 98, No. 2 (Feb., 1991), pp. 97-108
Published by: Mathematical Association of America
Stable URL: http://www.jstor.org/stable/2323939
Accessed: 15-11-2015 14:57 UTC

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

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

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

http://www.jstor.org

This content downloaded from 130.237.165.40 on Sun, 15 Nov 2015 14:57:45 UTC
All use subject to JSTOR Terms and Conditions
CarlyleCircles and the LemoineSimplicityofPolygonConstructions
DUANE W. DETEMPLE, State University
Washington

DUANE W. DETEMPLE: I received my Ph.D. fromStanfordUniversityin .;:^;? p"'.

1970,whereI wrotemydissertationon bounded univalentfunctionsunder
the supervision I have been at Washington
of M. M. Schiffer. State l
sincethep,withtheexception
University ofthe1981-82yearwhenI wasa
associateprofessor
visiting at ClaremontGraduateSchool.My current m
researchareas includecombinatorialgeometryand graphtheory.

1. Historical Commentsand Introduction.Until very late in the eighteenth

centurythe onlyregularpolygonsknownto have Euclideanstraightedge and
compassconstructionswere preciselythe ones shownin Book IV of Euclid's
Elements.
This situationchangedabruptlyon March30, 1796,whenone month
beforehis nineteenth
birthdayCarl Friedrich
Gauss made the firstentryin his
notebook[7]:
Principiaquibusinnititur
sectiocirculi,ac divisibilitas
geometrica
inseptemdecim
partesetc.
Mart. 30 Bruns[igae].
Principleof the circle'sdivision,and how one geometrically
dividesthe circle
intoseventeenparts,and so forth.
March 30 [Braunschweig]

Gauss had discoveredthatbesides the regularpolygonsof 2' - 3, 2' *4, 2" *5

and 2" * 15 sides,therewerea numberof otherconstructible polygons,
including
the17-gon.Gausssecuredpriority to hisdiscoverybypublishing an announcement
on June1, 1796,whichappearedin the"Intelligenzblatt der Allgemeinen Liter-
aturzeitung," thefirstand onlytimehe publishedin a journalof advancenotices.
In hisshortnotehe wrotethat"Thisdiscovery is reallyonlya corollary
ofa theory
withgreatercontent, whichis not completeyet,butwhichwillbe publishedas
soonas it is complete"([1],[7]).The fullmeaningofGauss'pronouncement came
in 1801withthepublication of hismonumental Disquisitiones
Arithmeticae[6].Its
finaltwosections(articles365 and 366) discusstheissueofpolygon constructibil-
ity.Recastingtheproblemin termsofconstructing theN verticesofthepolygon
on a givencircle,Gaussstatedhisresultas follows.
In generaltherefore
inordertobe able to dividethecirclegeometrically
into
N parts,N mustbe 2 or a higherpowerof2, or a primenumber oftheform
2' + 1,or theproductofseveralprimenumbers ofthisform,or theproduct
ofone or severalsuchprimesinto2 or a higherpowerof2.
The sufficiencyof thecondition followsreadilyfromGauss' analysis.The neces-
sity,however,is notobviousand Gauss neverpublisheda proofofthisassertion.
The firstproofis creditedto PierreL. Wantzel(1814-48)[15].
97

This content downloaded from 130.237.165.40 on Sun, 15 Nov 2015 14:57:45 UTC
All use subject to JSTOR Terms and Conditions
98 DUANE W. DETEMPLE [February

Primesof the form2' + 1, and thereforenecessarilyof the formFk 2- = + 1,

are the Fermatprimes.Gauss knew that Fo = 3, F1 = 5, F2 = 17, F3 = 257 and
F4 = 65537 were prime,and he knew that Euler had shown F5 to be composite.
Even today no other Fermat primes have been found, although the smallest
unsettledcases are F22, F24, and F28. (The primality statusof Fermatnumbersas
of 1983 can be foundin Keller [8], and a shorterbut more recenttable is included
in Young and Buell [17]).
Beyond the brillianttheoreticalbreakthrough of Gauss thereis stillan intrigu-
ing puzzle: how can one devise a sequence of stepswithstraightedge and compass
which constructsa regular N-gon? Gauss expressed interestin this problem,
althoughhe did not offerexplicitgeometricconstructions.Throughoutthe last
century,and even intopresenttimes,numerousconstructions have been contrived
forthe 17-gon(see [1], [2] forhistoricalcomments).J. F. Richelot'sconstruction of
the 257-gon required a total of 194 pages, published in four parts in Crelle's
journal in 1832. A Professor0. Hermes labored forten yearson the construction
and associated algebra of the 65537-gon.The work filled a trunkwhich was
donated to the MathematicalInstituteat Gottingen.Nearly a centurylater it
remainsstoredin an atticthere,in all likelihoodhavingneverbeen read.
In what followswe show thata remarkably uniformprocedureto constructthe
regularFk-gonsis affordedby consistentuse of an idea attributedto the Scottish
historianThomas Carlyle(1795-1881). Before turninghis attentionto the literary
arts, the young Carlyle taught mathematics,translated Legendre's influential
Elementsde Geometrieinto English,and, importantforour purposes,devised an
elegantgeometricalsolutionto quadraticequations.His idea is based on whatwe
will call the Carlylecircle construction,whichwe describein the next section.In
the remainingsections we will demonstrateconstructionsof the 5-, 17- and
257-gonwhichemploythe Carlylecirclemethod,and, at least in a generalway,we
also discussthe method'sapplicabilityto the 65537-gon.
Our unified approach has two nice features: first,the correctnessof the
constructionsis easy to verify,and second, the constructionsare highlyefficient.
For the pentagon the procedure is similar to the popular constructionfrom
antiquityfoundin Ptolemy'sAlmagest,but witha slighttwistat the end. For the
17-gonwe rediscovera methodof Smith[12] whichappeared in 1920,but again we
can offera small improvement. The 257-gonis fairlydifficultat the concretelevel
of actuallyusingthe Euclidean toolswithsufficient accuracyto produce a convinc-
ing construction, but at the conceptuallevel the procedureis remarkablystraight-
forwardin its use of 24 Carlylecircles.Indeed, one becomes convincedthat the
styleof construction needs no change to handle the 65537-gon.
A quantitativemeasureof the simplicity of a geometricconstruction was devised
in 1907 by Emile Lemoine (1840-1912), and is described in Eves [5] this way.
Considerthe followingfiveoperations.
S1: to make the straightedge pass throughone givenpoint.
S2: to rule a straightline.
C,: to make one compass leg coincidewithanypoint.
C2: to make one compass leg coincidewithanypointof a givenlocus.
C3: to describea circle.
If these operationsare performedm1, M2, n1,n2,n3 timesrespectivelyin a con-
structionthen m1S1+ m2S2 + n1C1+ n2C2 + n3C3 is the symbolof the construc-

This content downloaded from 130.237.165.40 on Sun, 15 Nov 2015 14:57:45 UTC
All use subject to JSTOR Terms and Conditions
1991] CARLYLE CIRCLES AND THE LEMOINE SIMPLICITY 99

tion. The total number of operations ml + m2 + n1 + n2 + n3 is called the

simplicityof the construction,where we note that a low value of simplicity
correspondsto an efficientconstruction.In what followswe will assume thatwe
use a singlenoncollapsingcompass'. Simplicity is improvedbyhavingcircleswitha
commonradius,or a commoncenter,describedin succession since the compass
would not have to be reset,or one of itslegswould not requirerepositioning.
As an example,the constructionof a circle and two lines whichcross orthogo-
nallyat the circle'scenter(ifwe beginwiththe circle,C3, thendrawa line through
the circle's center,S1 + S2, and finallyerect the perpendicularbisector,2C, +
2C3 + 2S1 + S2) has symbol3SI + 2S2 + 2C1 + 0C2 + 3C3 and simplicity10. If
the firsttwostepsin the construction just describedare reversed(beginwitha line,
S2, and then draw a circle centered on the line, C2 + C) then the symbolis
2S1 + 2S2 + 2C1 + C2 + 3C3, but the simplicity is still 10.
The constructionswe will describe for regularpolygonshave lower simplicity
measures than theircompetitorsand are perhaps nearlyoptimal.An interesting,
but apparentlyopen, problem is whetheror not it can be proved that a given
construction has optimalLemoine simplicity.
2. CarlyleCircles. A varietyof methodsto constructthe rootsof a quadratic
equation are known,but the one of Carlylewe now describeis especiallyattractive.
It appeared in Leslie's Elementsof Geometrywiththe remark"The solutionof
this importantproblem now insertedin the text,was suggestedto me by Mr.
Thomas Carlyle,an ingeniousyoungmathematician,and formerly mypupil" (see
[5]).
Beginningwiththe usual Cartesian x, y-axesand a unitdistance,our intention
is to constructthe rootsof the quadraticequation x2 - sx + p = 0, wheres and p
are given signed lengths.To this end, plot the points A(0, 1) and B(s, p). The
circlewhichhas the segmentAB as a diameterwillbe called the CarlylecircleCS P
of the givenquadraticequation. The centerof Csp is at M(s/2, (1 + p)/2), which
can be constructedas the midpointof AB. It will later be foundusefulto notice
that M is also the midpointof S(s, 0) and Y(O,1 + p), wherethe advantagederives
frombeing able to locate S and Y on the x- and y-axes,respectively.
If we suppose Cs P crosses the x-axisat H1(x1,0) and H2(x2, O), with x1 >x2
we have the two cases shown in FIG. 1. For p of either sign we observe that
x1 + x2 = s. For p < 0 the intersectingchords theoremfor FIGURE l(a) shows
OH1 OH2 = OA OC; that is, (xI) (-x2) = (1) (-p). For p > 0, the inter-
.

sectingsecantstheoremforFIG. l(b) shows x1 x2 = p. In all cases, then,

- = SX + p, S = X1 + p =
(X X)(X-X2) X2- X2, XIX2,

and so we conclude the followingtheorem:

If theCarlylecircleCs,p intersects
thex-axisat x1 and X2, theseare therootsof
x2 _ sx +p = O.

An alternateproofcan be based on the Pythagoreantheorem.

tThe modernnoncollapsingcompass permitsone to draw with ease a circle centered at a given

pointwhose radiusis the distancebetweentwo otherpoints.If one onlyhas a collapsingcompass,the
cumbersometo transferlengthsand consequentlythe
typeassumed by Euclid, then it is surprisingly
numbersof mostconstructions
simplicity is greatlyincreased.

This content downloaded from 130.237.165.40 on Sun, 15 Nov 2015 14:57:45 UTC
All use subject to JSTOR Terms and Conditions
100 DUANE W. DETEMPLE [February

Y(O,1 +p)

O,1) (O,) (s,p)

H2 S(,O) H,
0
~ __ Ms 1+P)
~s 1 +P)2

- ~~~~~~A(0,1)
Y(O,1i+ p) 0 H2H S(s,O0)
C(O, p) B(s, p)

(a) Case p < O (b) Case p > O

FIG. 1

The quadraticequationswe will considerlater all have real roots,but it should

be pointed out that complexroots are also related to the Carlylecircle.Suppose
the y-coordinate(1 + p)/2 of the centerof Cs,p exceeds the circle'sradius

r [(S)2 (1 p )2

a conditionwhich reduces to havinga negativediscriminantA = S2 - 4p. The

circle C5 p does not intersectthe x-axis,whichreflectsthe factthatthe quadratic
equation has a pair of complexconjugateroots z1, z2 = s/2 ? liv/- . We will
now verifythatany circlecenteredon the x-axisand orthogonalto CS p will meet
the verticalline throughthe centerof CS p at
(21 2 ) and (2 - A).

To see this,considerthe circle centeredat (4, 0) whichis orthogonalto Cs P. Its

equation is
2
(X_f)~~~ 2y2 (2 -P+ -r

and settingx = s/2 we find

2 (1 +P)2 2

Thereforeanysuch orthogonalcirclecan be used to constructz1, and Z2, withthe

y-axisnow interpretedas the imaginaryaxis of the complexplane.
3. Constructionof the Regular Pentagon. We assume that we have already
constructedthe unitcircleand the x, y-axes.FollowingGauss we viewthe problem
as the construction of the rootsof Z5 - 1 = 0. The root zo = 1, correspondingto
the point PO(1,0), can be factored out to show that the remainingpoints
P1, P2, P3, P4 correspondto the roots of the quartic equation z4 + z3 + z2 +

This content downloaded from 130.237.165.40 on Sun, 15 Nov 2015 14:57:45 UTC
All use subject to JSTOR Terms and Conditions
1991] CARLYLE CIRCLES AND THE LEMOINE SIMPLICITY 101

z + 1 = 0. LettingE = e21T//5
these roots are E1, 82E2S, &4, whichwe observesum
to -1.
Two othersumsof rootsare of special interest,namely,

7T0= E + 4 = 2 cos(27/5)
2
771 = + E = 2cos(4w/5).
It is simpleto check that
717+ 77i= -1, 770 77=1, 7o > 0 > T71 (2)
From (2) it followsthat qo and 2)1 are, respectively,
the largerand smallerroot
of the quadraticequation x2 + x - 1 = 0, and thereforetheycan be constructed
by means of the Carlyle circle C_1 -1 which has center M( - 1/2, O). M is
constructedas the midpointof OQ, where Q(- 1,0). The Carlylecirclecenteredat
M and passingthroughA(O, 1) thenintersectsthe x-axisat H&o(1o,0) and Hl(2)1,0).
In view of (1), circles of unit radius centered at Ho and H1 will intersectthe
original unit circle at P1, P2, P3, P4, thereby formingthe desired pentagon
PoP1P2P3P4 as shownin FIGURE2.

A PI

HIN QX |M o /Hopo

P~~~~P

FIG. 2

Beginningwiththe unit circle and the x, y-axesas given,the construction just

describedhas symbol2S1 + S2 + 8C1 + OC2 + 4C3, for a Lemoine simplicityof
15. The popular constructionof Ptolemy(see [4], [15]) proceeds in the same way
throughthe constructionof Ho, but then takes AH, as the distance between
successiveverticesof the pentagon,resultingin a simplicity
of 16 to locate all five
vertices.
4. Constructionof the Regular Heptadecagon (17-gon). Letting 8 = e2,i/17
our task here is to constructthe roots 8 , ,.. ., 16 of the equation Z16 +
-
z15 + ** +Z + 1 = 0. Gauss' idea was to arrangethese roots in a cyclein which
each root is a gth power of its predecessor:
81, E8g 8g2 g15 8g16 (3)
Since ?m = we can see that correspondingto the prime p = 17 there
?m (mod 17)
mustbe founda g forwhich
gr 1 (mod p), 1 < r ?p - 2 and gP 1 (mod p). (4)

This content downloaded from 130.237.165.40 on Sun, 15 Nov 2015 14:57:45 UTC
All use subject to JSTOR Terms and Conditions
102 DUANE W. DETEMPLE [February

A positiveintegerg whichsatisfiesthe conditionsof (4) forp is said to be a

primitiveroot of the prime p. For any Fermat prime p > 3, g = 3 is always a
primitiveroot. This followsfromthe factthat 3 is a quadraticnonresidueby the
quadraticreciprocity theorem,and any nonresidueis a primitiveroot because the
size of the multiplicativegroup is a power of 2. With this choice the cycle (3)
becomes
1 3 9 10 13 5 15 11 16 14 8 7 4 12 2 6
,E ,?, ,8 ,8E , ,8 ,8E ,8E (5) ,8,E , ,E , ,E 8E .
Next,Gauss consideredthe sumsof the termsof varioussubcyclesof (5), which
are called periods.A typicalexampleis theperiod 7,02 = 81 + 89 + 813 + . * +?2.
For notational conveniencea period will be representedby an l-tuple of the
exponentsof E whichoccur in the sum,and so the twoperiodsof lengtheightare
expressedas
710,2 = (1, 9, 13, 15, 16,8,4, 2)
771,2 = (3, 10,5, 11, 14,7, 12, 6). (6)
There are fourperiodsof lengthfour,namely,
'70,4
=
(1,13,16,4), '72,4 =
(9,15,8,2) (7)
(3,5, 14, 12),
'71,4 =
773,4 = (10, 11,7,6).

There are eightperiods of lengthtwo,but onlytwo are needed forthe construc-

tion,
770, 8 = (1, 16) = 2 cos(27r/17) (8)
774,8
= (13,4) = 2cos(8rr/17).
From (8), and using(7), we easilysee that
770,8 + 774,8 = 770,4, '70,8 ' 774,8 = '71,4' '70,8 > 774,8- (9)
to calculate that
Similarlyit is straightforward
770,4 + 772,4 = '70,21 770,4 *172,4 = -1, 770,4 > 172,4 (10)

711,4 + 773,4 = '71,2' 711,4 '173,4 = - 1, 77i,4 > 71734 (11)

710,2 + 71,2- -1,1 2 *71,2

170, = - 4, 770,2> Y71,2 (12)
The inequalitiesin (10) and (11) are geometrically obvious,and since a calculation
shows that (170,4 - 772,4) * (771,4 - '73,4) = 2(7r0,2 - 771,2) it then follows that
710,2> 71,2 as claimed in (12).
Equations (8)-(12) readilylead to the constructionof the regular 17-gon as
shownin FIGURE 3. It is assumed thatwe are giventhe unitcirclecenteredat 0
and the x, y-axes.From (12) we see that 'ri0,2 and '1,2 are, respectively, the larger
and smallerrootsof the quadraticequation x2 + x - 4 = 0 whose corresponding
Carlylecircle C-1 _4 is centeredat MO(- 1/2, - 3/2). Therefore:
(i) Draw the perpendicularbisectorof QO, usinga circleat Q through0 and
the existingcircle. This constructionhas Lemoine symbol2S1 + S2 +
2C1 + C3 and locates Q'.
(ii) Draw a circleat Q' throughP0 (symbol2C1 + C3) to locate Mo.
(iii) Draw the Carlylecircle at Mo (symbol2C1 + C3) to locate Ho, 2(X10,2' 0)
and H1,2071,2, 0).

This content downloaded from 130.237.165.40 on Sun, 15 Nov 2015 14:57:45 UTC
All use subject to JSTOR Terms and Conditions
1991] CARLYLE CIRCLES AND THE LEMOINE SIMPLICITY 103

Mo i~~~~P

FIG. 3

From equations(10) and (11) we see thatCarlylecirclescenteredat M0 2(2-70 2' 0)

and M1,2(21 2, 0) will locateHo, 4(G704, 0) and H, 4(J1, 4 ?0).Therefore:
(iv) Draw the perpendicularbisectorsof OHO 2 and OH1 2, using the same
circleat 0 (symbol4S1 + 2S2 + 3C1 + 3C3). This locates M02 and M12.
(v) Draw the Carlylecirclesat MO,2 and M1,2 (symbol4C1 + 2C3) to locate
Ho 4 and H14.
nextto Equation(9), we see thatHo 8(0,
Turning 8 0) requiresa Carlylecircleat
MO4 Therefore:
(vi) Set the compassto radius QHt 4 and drawa circleat 0 (symbol3Ct + C3)
to locate Y(O,1 + '71,4)*
(vii) Draw the line YHo 4 (symbol2S1 + S2).
(viii) Draw the perpendicularbisectorto YHo 4 (symbol2S1 + S2 + 2C1 + 2C3)
to locate Mo,4.
(ix) Draw the Carlylecircleat Mo,4 (symbol2C1 + C3) to locate Ho08.
In viewof equation(8), the followingstep locates P1 and P16,whichcorrespond
Except forlayingoffthe distanceP P1, thiscompletes
to E1 and E16, respectively.
the construction.
(x) Set thecompassto unitradiusand drawa circleat Ho 8 (symbol3C1 + C3)
to locate P1 and P16.
The ten steps describedabove have total symbollOS1 + 5S2 + 23Ct + OC2 +
13C3, for a simplicityof 51. However,two simple modificationscan reduce the
of the constructionto 45.* First,the costlydouble bisectionof step (iv)
simplicity
can be avoided by usinghalf-scaledCarlylecirclesto locate MO,2 and M1,2. This
requiresa perpendicularbisectorof Q'O, but since the compass is alreadyset at

*Our thanksto the refereeforsuggestingthese modifications.

This content downloaded from 130.237.165.40 on Sun, 15 Nov 2015 14:57:45 UTC
All use subject to JSTOR Terms and Conditions
104 DUANE W. DETEMPLE [February

unitradiusthe symbolis 2S1 + S2 + C1 + C3. We also need a circleof radius OQ'

at 0 (symbol2C1 + C3). Since thereis no more need forstep (iv) the simplicity is
reduced by four. Step (viii) can also be replaced by a less costly procedure.
Following step (vi) the perpendicularbisector of OY can be constructedwith
symbol2SI + S2 + C1 + C3, since the compass need not be resetforthe circleat
Y. Step (vii) now locates MO,4 and so we have furtherreduced the simplicity by
two. Altogetherthe modifiedCarlyle constructionhas symbol8S1 + 4S2 + 22C,
+ 11C3,fora simplicity of 45.
The Smithconstruction [12] is similarto thatdescribedin (i)-(x), but it requires
a perpendicularbisector at Ho04, which increases the simplicityto 58. The
constructionof Richmond[11] (also shownin [13], [16], and elsewhere)seems to
appear more oftenthan its efficiency warrants.Even arrangingforseveral circles
to do double duty,as in the modifiedCarlyle construction,the simplicityis 53
(which constructsP3; the simplicityincreases to 60 to locate P1.) The compara-
tivelyrecent constructionof Tietze [14], also describedin Hall [7], can be done
with simplicity50. Neither referenceprovides a proof of the construction,and
indeed Hall remarksthat "the proofthat it is correctrequiresextensivecalcula-
tion." His assessmentapplies equallywell to the methodof Richmond.
5. Constructionof the Regular 257-gon. Since g = 3 is also a primitiveroot of
257, the rootsof z256 + z255 + * + 1 = 0 are orderedin the cycle
E1 3 E9 , 27 81 243 E215 ,165 E238 E200 86

where E = e2wi/257.The period of length1 = (p - 1)/2' whichcontainsthe term

9g' is denoted by
1-1
7fj k = gJ+f,k k= 2
n= 0

In particular,Th = - 1 and we also observethat 7bjk = mi k when j i (mod k).

To set up the hierarchyof quadratic equations to be solved in succession,it is
necessaryto know the sum, product,and relativesize of pairs of periods. This
allows the constructionof the periodsby means of the appropriateCarlylecircles.
The additionformulas
Thi,2k + ?i+k,2k = Ti,k (13)

are obvious.Formulasforproductsrequireeffortbut perhaps less than expected.

The followingsample calculationof %h 64 32, 64 illustratessome usefulprinciples
of computation. Since %0 64 = (1, 241, 256, 16) and 'q32,64 = (64, 4, 193, 253) are
representedby quadruples of exponents,the productwe seek correspondsto a
modulo 257 additiontable.

32,64
64 4 193 253
0, 64 1 65 5 194 254
241 48 245 177 237
256 63 3 192 252
16 80 20 209 12

This content downloaded from 130.237.165.40 on Sun, 15 Nov 2015 14:57:45 UTC
All use subject to JSTOR Terms and Conditions
1991] CARLYLE CIRCLES AND THE LEMOINE SIMPLICITY 105

The main diagonal,it can be checked,is T3364 = (65,245,192,12), and the

"diagonal" just below it is Ns5 64 = (48, 3, 209,254). Altogetherwe findthat
070,64 ' 732,64 = (14)
?733,64 + T755,64 + 723,64 + 71,64-
Thus productscan be computed by knowingwhich periods of the same length
containthe termsin the firstcolumn.
The rightside of (14) can be simplifiedby use of the additionformulas(13),
giving
%,64 * 32,64 = '?1,32 + q23,32- (15)
The derivationof (15) is actuallyenoughfromwhichto deduce the generalformula
obtainedby increasingthe firstsubscriptsby j. That is,
7j, 64 * j+32,64 = 'b+?1,32 + Ti+?23,32 (16)
Additionaldiscussionof these computationaldetailscan be foundin Rademacher
[10].
The last algebraic difficulty
is the orderingof mj2k and y+ k, 2k. It is obvious
that 7i064 > T732,64 but showing that 7r4,16 > ' 12,16 is somewhat delicate. The
methods described in Klein [9] and Rademacher [10] for the 17-gon can be
extendedto the presentcase however,thusallowingthe identification of the larger
and smallerroot of a pair.
The resultsof the calculationsare shownin TABLE 1. The same information is
containedin Bishop [3],but in a differentform,sincewe have used formula(13) to
simplifythe expressionsfor the period products. Since it will be sufficientto
constructthe point P1 on the unit circle whose x-coordinateis cos(27/257)=
(1/2)% 128' we observe that we need only constructtwo of the period pairs of
length4 and sixof length8.

TABLE 1

Productsof Required Period Pairs Order of Period Pairs

10,2 * 11,2
-
-64 710,2 > 71,2

Thj,4*?j+2,4 = - 16, j =0 , 1 0,4 > 72,4, 71,4 > 773,4

)j,8 *h ?+4,8 -2 + 3NJ+1,2 + 2NJ+2,4 0,8 > 4,81 2,8 > 6,81
j = 0,1,2,3 7)1,8 < 75,8, 7)3,8 <T 7,8

7)i,16 '] ?+8,16 = TJ,2 + TJ,8 + Nj+2,8 + 2TJ+5,8 r10,16 > r18,16, 7)4,16 > 7712,16, 72,16 > 7710,16
j = 0...7 rq6,16 < rq14,16' 771,16> 779,16' 775,16> '713,16'
773,16 > 'q11, 16x 7q7,16 > 'q15,1lh

7Tj,32 *7T+16,32 = 7j,16 + 7?j+1,16 r10,32 > r116,32, 7)8,32 < 7)24,32,
+7j?+2,16 + 7Th+5,16 r11,32 > '117,32, 7)9,32 < 25,32,
j = 0,1, 7,8,9,15 ri7,32 < 23,32, r15,32 > ri31,32

rLj,64 *7j+32,64 = 71j+1,32 + l?j+23,32 r1064 > ri32,64, 24,64 < ri56,64
i = 0,24
770,128 * 64,128 = 7r56,64 r10,128 > 7r64,128

Translatingthe resultsof the table into an explicitconstructionis straight-for-

ward, althoughwe must be alert to carryingout the required procedure in an
efficientmanner.Beginningwiththe firstrow of information in TABLE 1 we see

This content downloaded from 130.237.165.40 on Sun, 15 Nov 2015 14:57:45 UTC
All use subject to JSTOR Terms and Conditions
106 DUANE W. DETEMPLE [February

that '70,2 and 1, 2 are the largerand smallerroots,respectively,

of x2 + x - 64 =
0, and so the first
step is to construct
the centerM0(- 1/2, - 63/2) of the Carlyle
circle C -164- Successive distancedoublingsalong the positivex-axiswill locate
T(31, 0), and thenswingingan arc fromQ'(- 1/2,0) of radius Q'T, will locate Mo
on the perpendicularbisectorx = -1/2 of 0 and Q(- 1,0). As shownin FIGURE
4, the Carlylecircle C1 -64 thenlocates Ho 2(7q02,O) and H1 2(q1 2, ).

HI,2 HO H02 S T

Q 'PO 3 7 15 31

1 63
M 2' 2

FIG. 4

The construction of the fourperiods of length64 is depictedin FIGURE 5. The

point Y(O, - 15) is located by swingingan arc from0 of radius OS, whereS(15, 0)
was constructedin the previousstep. The midpointMo 2 of YHo,2 is the centerof
the Carlyle circle which determinesHo,4(7q04,0) and H2,4(Gq24,0). Similarlythe
points at x = 711,4 and x = 'q3,4 are foundby the Carlylecircle centeredat the
midpointM1,2 of YH1,2

H3,4 \ <HI, 4 /HO,4

FIG. 5

The remainingpairs of periods are constructedin an analogous way,although

extra distance transfersare required because of the more complicatedproduct
expressionsshownin TABLE 1. A circleof unitradiuscenteredat Ho0128 will cross
the unit circle at P1 and P256,whichcompletesthe constructionup to layingoff
the distancePOP1 to locate all of the verticesof the 257-gon.

This content downloaded from 130.237.165.40 on Sun, 15 Nov 2015 14:57:45 UTC
All use subject to JSTOR Terms and Conditions
1991] CARLYLE CIRCLES AND THE LEMOINE SIMPLICITY 107

To optimizethe simplicityof the construction

it is essentialto transferdistances
For examplesome usefulsavingsin constructing0, 16'7 4, 16' '78,16' T12,16
efficiently.
are possiblewhen advantageis takenof the relation h0,16 78, 16 + '74,16 * 12,16 =
37o 2+ 2X1,4. The construction of the threeotherquadruplesof periodsof length
sixteenare aided by the correspondingalgebraicrelations.
Beginning with the unit circle and the x, y-axes, the constructionof
Pj(cos(2 rr/257),
sin(27/257)) has symbol

94S1 + 47S2 + 275C1 + OC2 + 150C3

of 566. Amongthe 150 circlesare 24 Carlylecircles.

fora Lemoine simplicity
6. Remarks on the Constructionof the Regular 65537-gon.We have already
observed that g = 3 is a primitiveroot of p = 65537. The sum, product,and
relativeorder of period pairs must then be computed (one feels sorryfor the
computerlessHermes!). The firstline of the requiredtable of information
is

7,2 + 71,2 = -1, 10,2 1 ,2 = 214, 0,2 > 11,2- (17)

The productformuladerivesfromthe fact that the exponentsums evenlydivide

into quadraticresiduesand nonresidues.Drawingthe Carlylecircle C_1 -2'4 may
place impossibledemands on any compass we own, but at the conceptual level
thereis no problem.
Successive levels in the hierarchyof quadratic equations require 2, 4, 8,...
Carlylecircles.Workingupwardshowever,and assumingthe worstcase possibility
thatthe productformulasand sumsinvolveall distinctperiods,the largestpossible
numberof Carlylecirclesrequiredis 1,2, 6, 30,270,4590,.... Actuallythe 4590 is
spurious,since there are only 210 periods with26 termsand so these can all be
computedwith 29 = 512 Carlyle circles. Even thoughwe know little about the
detailsof the construction,
we can concludethatthe constructionwe seek requires
1 + 2+ 4 + 8 + +512 + 270 + 30 + 6 + 2+1 = 1332 or fewer Carlyle
circles.

REFERENCES
1. R. C. Archibald,The historyof the constructionof the regularpolygonof seventeensides, Bull.
Amer. Math. Soc., 22 (1916) 239-246.
2. , Gauss and the regular polygonof seventeen sides, Amer. MJath.Monthly,27 (1920)
323-326.
3. Wayne Bishop,How to constructa regularpolygon,Amer. Math Monthly,85 (1978) 186-188.
4. H. S. M. Coxeter,Introductionto Geometry,Wiley,New York, 1961.
5. Howard Eves, An Introductionto the Historyof Mathematics,fourthedition,Holt, Rinehart,and
Winston,New York, 1976.
6. C. F. Gauss, DisquisitionesArithmeticae,
Leipzig 1801,translatedfromthe Latin byA. A. Clarke,
Yale University Press, 1965.
7. Tord Hall, Carl FriedrichGauss: A Biography,MIT Press, 1970.
8. WilfredKeller, Factors of Fermat numbersand large primes of the form k x 2' + 1, Math.
Comp. 41 (1983) 661-673.
9. Felix Klein, Famous Problemsof ElementaryGeometry,Dover, 1956 (originalGermaneditionin
1895).
10. Hans Rademacher,Lectureson ElementaryNumberTheory.Blaisdell,New York, 1964.
11. H. W. Richmond,Quart. J. Math.,26 (1893) 296-297.

This content downloaded from 130.237.165.40 on Sun, 15 Nov 2015 14:57:45 UTC
All use subject to JSTOR Terms and Conditions
108 DUANE W. DETEMPLE

of the regularpolygonof seventeensides, Amer.Math. Monthly,27

12. L. Linn Smith,A construction
(1920) 322-323.
13. Ian Stewart,Gauss, Sci. Amer.,237 (1977) 122-131.
14. Heinrich Tietze, Famous Problems of Mathematics,GraylockPress, New York, 1965 (original
German editionin 1949).
15. P. L. Wantzel,Journalde Mathematiques pureset applique'es,2 (1837) 366-372.
16. R. C. Yates, GeometricalTools, Educational Publishers,St. Louis, 1949.
17. JeffYoung and Duncan A. Buell, The twentiethFermat numberis composite,Math. Comp., 50
(1988) 261-263.

Notice to authors: Recentlythe MONTHLY has not been processingpapers

because of a veryhighbacklog.This situationhas now eased somewhatand
we are once again consideringnew manuscriptsfor possible publication.
Thankyou foryourpatience. -The Editors.

This content downloaded from 130.237.165.40 on Sun, 15 Nov 2015 14:57:45 UTC
All use subject to JSTOR Terms and Conditions