Вы находитесь на странице: 1из 10

## Author(s): Colin Tripp

Source: The Mathematical Gazette, Vol. 59, No. 408 (Jun., 1975), pp. 98-106
Stable URL: http://www.jstor.org/stable/3616644
Accessed: 02-03-2015 11:08 UTC

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
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.

The Mathematical Association is collaborating with JSTOR to digitize, preserve and extend access to The Mathematical
Gazette.

http://www.jstor.org

This content downloaded from 61.175.193.50 on Mon, 02 Mar 2015 11:08:32 UTC
All use subject to JSTOR Terms and Conditions

98

GAZETTE
THE MATHEMATICAL

COLIN TRIPP

1. Introduction
Manyreaderswill be familiarwiththe puzzleillustratedin Fig. 1, in which
the triangleABC is isosceles and anglesa, b, c are given.t If a, b, c are
integerswhenexpressedin degreesthenit mightappear,at firstsight,that
so mustangle0 be. Butthisis not the case.Thereaderwill quicklyfindthat
0 is not accessibleby simple'anglechasing'.By this I meanmarkingin the
A

B
FIGURE1

## anglesobtainedby makingthe anglesin any trianglesum to 180?and by

makingthe exteriorangle of any triangleequal the sum of the interior
opposites.Thefollowingquestionarises:Givena, b, c canangle0 beobtained
by puregeometry?My conjectureis that the answeris yes in those cases
whenanglesa, b, c and0 areall multiplesof the sameangle.However,I am
farfromprovingthis,andhaveobtainedgeometricalderivationsonlyfor a
limitednumberof cases,as will be seen.
2. Trigonometricformula
for 0
to
A formularelating0 a, b, c can be obtainedby repeatedapplication
of the sine rulefor triangles.This formulamay be written
sin(b + c) sinc(cosa + cos2b)
tn 0 =
sinb(cosa + cos 2c) + cos (b + c) sinc(cosa + cos2b)'
t For a recent instance, see the Bulletin of the Institute of Mathematicsand its
Applicationsfor July/August1974.

This content downloaded from 61.175.193.50 on Mon, 02 Mar 2015 11:08:32 UTC
All use subject to JSTOR Terms and Conditions

99

## Insofaraswe areinterestedin solutionsof thisequationin whicha, b, c and0

are all integers(or all multiplesof the same angle), this trigonometric
equationmay be regardedas a kind of diophantineequation.
angles
DEFINITION1: With referenceto

## Fig. 1, given that angles a, b, c are multiples

if thecorresponding
of 1 , withb > c, the triplet(a, b, c) is adventitious
angle
0 is also a multipleof 1?.We call 0 the derivedangle.
We take b > c to avoid mirrorimages, since obviously,if (a, b, c) is
adventitious,then so is (a, c, b). Also, if c = b we have (a, b, b) whichis
4. An example
Thetriplet(20, 60, 50)t is adventitious.Howevertheproofthatthisis so,
althoughelementary,is not obvious.Thisis one of the casesI haveseenset

FIGURE2
t Strictly (20?, 60?, 50?), but the 'degrees' sign is conveniently omitted.

This content downloaded from 61.175.193.50 on Mon, 02 Mar 2015 11:08:32 UTC
All use subject to JSTOR Terms and Conditions

100

THE MATHEMATICAL
GAZETTE

## as a puzzle, and many hundreds of qualified-man-hoursmust have been

spent in trying to derive the value of 0! Substitution in the trigonometric
formula for 0, after replacing differencesof cosines by products of sines and
cancelling sin 50? top and bottom, gives
tan 0 = sin 20? sin 40? sin 70?/(sin60?sin 30? - sin220? sin 40?).
It is not enough to look up the various cosines and sines in tables (or use an
electronic calculator) because this will only obtain 0 to so many decimal
place accuracy, and will not prove that 0 is an integer. One must reduce the
formula still further, making use of trigonometric identities. Considerable
ingenuity is requiredto solve this equation exactly for 0 (especially without
knowing the solution first!). Rather than follow this through, I give a
derivation using only elementary geometry, although a simple construction
is needed.
This is to mark a point E' on AC such that angle E'BC = 20? (Fig. 2).
It then appears that triangles EBC, BE'Cand DE'Bare isosceles. Therefore
triangle BEE' is equilateral, so triangle EE'D is isosceles. But angle
DE'E= 40, so 0 +40?= 70? giving 0= 30? exactly. This proves that
5. Questionswhicharise
The above result leads to a number of interesting questions:
(a) Is there any procedure for discovering the construction used in the
proof in Section 4?
(b) How many other adventitious triplets are there (among the total of
113 564 possible triplets that can be formed with the restrictions on a, b, c
given in the definition in Section 3), and what are they ?
(c) Can all the other adventitious triplets be proved to be adventitious by
using only pure geometry?
(d) If geometrical proofs exist, and a construction (such as in the example
of Section 4) is needed in at least some of the proofs, is there any procedure
by which the appropriateconstruction might be deduced?
I have an answer for (a), a conjecture for (b), a statement of belief for (c)
and a partial answer for (d). Questions (a) and (d) will be dealt with in the
next section. I have already stated that I believe the answer to (c) to be yes.
The conjecture for (b) is based on computer results. A computer search for
integer solutions to the trigonometric equation for 0 among the 113 564
possible triplets yielded the information that there are 53 triplets which
appear to be adventitious to five decimal place accuracy, and these are
listed in Table 1. Thus there can be at most 53 adventitious triplets.
If the geometrical figures corresponding to each of these triplets are
drawn, it will be noticed that in nearly all cases there appear special geometrical features. Isosceles triangles occur, or lines cross at right angles.
In some cases, however, if there are special geometrical featuresthey are not

This content downloaded from 61.175.193.50 on Mon, 02 Mar 2015 11:08:32 UTC
All use subject to JSTOR Terms and Conditions

101

ANGLES

## immediatelyevident.I haveprovedthat 33 of these53 tripletsare actually

adventitious,in particularthe 8 tripletswith a = 20?,whichare dealtwith
in the next section.
a

4
8
12
12
12
12
12
12
16
20
20
20
20
24
28
32
36
40
44
48
52
56
72
72
72
72
120

46
47
42
48
57
66
69
72
49
50
60
65
70
51
52
53
54
55
56
57
58
59
39
42
48
51
24

4
8
18
12
33
42
21
42
16
20
30
25
50
24
28
32
36
35
34
33
32
31
21
24
24
39
12

2
4
12
6
15
12
3
6
8
10
10
5
10
12
14
16
18
15
12
9
6
3
12
12
6
9
6

4
8
12
12
12
12
12
12
16
20
20
20
20
24
28
32

46
47
42
48
57
66
69
72
49
50
60
65
70
51
52
53

44
43
30
42
42
54
66
66
41
40
50
60
60
39
38
37

42
39
24
36
24
24
48
30
33
30
30
40
20
27
24
21

40
44
48
52
56
72
72
72
72
120

55
56
57
58
59
39
42
48
51
24

40
44
48
52
56
27
30
42
42
18

20
22
24
26
28
18
18
24
12
12

TABLE
1 (anglesin degrees)
6. The eight triplets with a = 20?
Two of the eight of these triplets listed in Table 1 can be proved to be
adventitious immediately. The triplet (20, 50, 40) produces a 'kite'-shaped
quadrilateraland it follows that 0 = 30? by symmetry. The triplet (20, 50, 20)
produces a 'fan' of equal lengths BC = EC = DC. An equilateral triangle
appears, and it immediately follows that 0 = 10?. The other six triplets all
requirea construction. The fascinating thing is that in each case the required
construction can be obtained by superimposing one of the other cases on
the figure! For example, for (20, 60, 50) which was proved to be adventitious
in Section 4, the construction was obtained by simply superimposing the
mirror image of the (20, 50, 20) case on the figure. A network can be established, in which triplets are connected by arrows if the adventitiousness of

This content downloaded from 61.175.193.50 on Mon, 02 Mar 2015 11:08:32 UTC
All use subject to JSTOR Terms and Conditions

102

THE MATHEMATICAL
GAZETTE

## one tripletcan be used to provethe adventitiousnessof the other. Fig. 3

illustratesthisnetwork.Theangle0 is foundfor (20, 50,40) and(20, 50, 20)
independentlyas alreadydescribed,without the use of a construction.
Theseresultsareusedto find0 for (20, 70, 50) and(20, 60, 50),respectively.
constructionsprovidedby the (20, 60, 50) figurealong with the value of 0
for thattriplet.The0 valuefor (20, 60, 30) facilitatesthe calculationof 0 for
(20, 65, 60), whichin turnenables0 to be calculatedfor (20, 65, 25). The
numbersby the arrowsin Fig. 3 indicatethe threetypesof proofthatarise:
(1) propertiesof isosceles(andequilateral)trianglesare used;
arises,and its propertiesare used;
arise,and theirpropertiesused.
'kite'
(20, 50, 40)
(3)
v

'fan'
(20, 50,20)
(1)
v

(3)

(>

(1)

(2)
V

FIGURE3

## It can be seenthat(20, 65, 25) seemsto be themostinaccessiblecase,in that

at least four othertripletsmustbe provedto be adventitiousfirst.It may,
however,be possibleto arriveat the adventitiousnessof (20, 65, 25) by a
shorterroute, since the paths shown in the figureare not the only ones
of (20, 70, 60) by a proofof type
of (20, 70, 50) impliesthe adventitiousness
(2).I havenot establishedanyotherpaths.It willbe notedthatall thearrows
may be reversed.
It wasthecases(20,60, 50)and(20,70, 60),set as puzzles,whichledto the
the existenceof the other cases, the problemof finding0 is tantalisingly
difficult.Thereadermightliketo tryone or two of theseproofsfor himself,
in orderto appreciatemorethe eleganceand simplicityof each.
7. Cyclic complements

## In this section the idea that the existenceof one adventitioustriplet

impliesthe existenceof othersis developed.It will be noticedfromTable1
thatall the tripletsfallinto pairs(setsideby sideon the sameline),withone

This content downloaded from 61.175.193.50 on Mon, 02 Mar 2015 11:08:32 UTC
All use subject to JSTOR Terms and Conditions

103

exception: (36, 54, 36). In each pair a and b are equal. Theorem 1 explains
this, and also why the difference in the values of c in each pair equals the
difference in the values of 0.
THEOREM1.

## Given that the set (a, b, c) is adventitious,with derivedangle 0,

then (a, b, b - 0) is also adventitious,with derivedangle b - c.

PROOF. Construct the circle through EDC; it will intersect AB again at E',
between E and B (Fig. 4). In the cyclic quadrilateralEE'DC we have angles
AED = DCE' and ECE' = EDE'. Now AED = EBD + EDB, so these two
equations give
and c-c'=O-O',
900-ja-b+0=900?-a-c'

FIGURE4

where c' = E'CB, 0' = E'DB. Thus c' = b - 0 and 0' = b - c. The configuration AE'BCD therefore reveals a second triplet (a, b, c') with derived angle
0', where c' = b - 0, 0' = b - c. It follows that c' and 0' are both integers if
b, c and 0 are.
This completes the proof. Note that it is the general version of the type
(2) proof referred to in Section 6. The other two types of proof may be
generalised in the same way, but in these two cases restrictions must be
specified on the angles a and b.
The following definition now seems a natural one to make:
DEFINITION2. Given that the triplet (a, b, c) has derived angle 0, then the
triplet (a, b, b - 0) is its cyclic complement.

This content downloaded from 61.175.193.50 on Mon, 02 Mar 2015 11:08:32 UTC
All use subject to JSTOR Terms and Conditions

104

THE MATHEMATICAL
GAZETTE

## InTable1cycliccomplementshavebeenwrittenon thesameline.A triplet

is its own cycliccomplementif 0 = b - c. This is foundto be the case for
(36, 54, 36) (with0 = 18?).I havenot provedthatthereis only one adventitious tripletwhichis its own cycliccomplement.
8. 'Kites'and'fans'
It is possibleto provethe existenceof a set of adventitioustripletsindependently.These are the ones givingrise to 'fans'.Notice that the cyclic
complementof a 'fan'is a 'kite'.
2. Thereexist 27 adventitioustripletsgiven by the 14 triplets
THEOREM
(a, 45? + la, a), having derivedangles 0 = la, with a = 4?, 8?, ..., 56? along
the tripletwitha = 36?beingits own cyclic
withtheircyclic complements,
complement.

C
FIGURE5

## ThustriangleEDC is isosceles,so that 0 = la. Now 45?+ ia is aninteger

if a is a multipleof 4, andthisin turnensuresthat0 is an integer.All angles
on the figure must be positive, and in particular,angle DCE> 0, i.e.
a > 0, so that a < 60?. This restricts a to the values 4, 8,..., 56. By

90? -

## Theorem1 each of these 14 tripletshas a cycliccomplement(a, 45?+ la,

45? - ia) with derived angle 45? - ia. Only one of these triplets is its own
cyclic complement since 45? - la = a only for a = 36?. This completes the

proof.

This content downloaded from 61.175.193.50 on Mon, 02 Mar 2015 11:08:32 UTC
All use subject to JSTOR Terms and Conditions

105

## Two of the 27 tripletsproved to be adventitious in Theorem 2 have a = 20?,

so altogether I have proved that 25 + 8 = 33 of the 53 triplets suggested by
the computer calculations are definitely adventitious. There remain 10
triplets with a = 12?, 8 triplets with a = 72? and 2 triplets with a = 120?that
still have to be proved adventitious!
9. Generalisations
The idea of adventitious angles may be generalised in two ways.Firstly,
there is no reason why a basic unit other than the degree (=7r/180radians)
should not be used.
DEFINITION3. Given that a, b, c, are each multiples of n/Nradians,

where Nis
a positive integer and b > c, the triplet (a, b, c) is N-adventitious if the
corresponding angle 0 is a multiple of r/N.
Table 2 gives the numberof N-adventitious tripletsfor various values of N,
suggested by computer calculations.
N

## Total numberof triplets

J(JN- 1)(N- 2)(IN- 3)

Conjecturednumberof

14
16
24
42
70
120
168
180
210

20
35
165
1 140
5984
32 509
91 881
113 564
182 104

2
2
4
10
11
39
34
53
49

TABLE2

## Secondly, the condition that triangle ABC is isosceles may be dropped.

This introduces another degree of freedom. The problem really only concerns the quadrilateralDCBE (see Fig. 6).
^,D

FIGURE6

This content downloaded from 61.175.193.50 on Mon, 02 Mar 2015 11:08:32 UTC
All use subject to JSTOR Terms and Conditions

106

THE MATHEMATICAL
GAZETTE

DEFINITION
4. With reference to Fig. 6, given that the angles B, b, C, c are

angle0 is also a multipleof 1?.
It is possiblethatthesequadrupletsprovidea keyto provingsomeof the
COLIN TRIPP

## MathematicsDepartment,Brunel University,KingstonLane, Uxbridge,

Middlesex

Primefactorsandrecurringduodecimals
CEDRIC A. B. SMITH
not obvious; whetherthey were novel I do not know. It occurs to me
(whetherstrictlynovel or not) they mightwell be worthpublishing,as of
generalinterestand as a tributeto a greatandkindlyman.However,since
Prof. Davenport'sletter was not intendedfor publicationas written,I
paraphraseit herewith explanationof the background.
Numbersof the form 12"+ 1 can be dividedinto 4 classes:
(B) 122k+ 1, (C) 122k+ - 1, (D) 122k+ + 1.
(A) 122k-1,
Now 122k- 1 = (12k+ 1)(12k- 1); by using this identity (repeatedly if

## necessary)we can factorisenumbersof class (A) into those of forms (B),

(C),or (D). GeorgeS. Terry,of Hingham,Massachusetts,
found
that
he
other
empiricallythatall factorsof
things)
saying(among
5
or
to
1
were
numbers
(mod 12);similarlyall factorsof (C)
congruent
(B)
numbers 1 or -1, andfactorsof (D) numbers 1 or -5. I had writtento
Davenportaskingwhetherthis was true in general,and he replied:"The
resultsyou mentionare all simpleexamplesof the rule for the quadratic
characterof 3 to a primemodulus."
Davenportobservedthat it is sufficientto prove the resultsfor prime
factorsof (12"+ 1), sinceif theseare truethe resultsfor compositefactors
follow immediately.For, if we can showin case (B) that all primefactors
... are congruentto 1 or 5 (mod 12), the same holds for any product
PiP2P3.... Similarly for cases (C), (D). He then appealed to the theory of

P1,P2,P3,

andprovedin Chapter3 of Davenport'sown book [1].
We recallthat an integerx is saidto be quadratic(modulop) (or a quadratic residue)when it is a perfectsquarein arithmeticmodulop, that is

This content downloaded from 61.175.193.50 on Mon, 02 Mar 2015 11:08:32 UTC
All use subject to JSTOR Terms and Conditions