International Journal of Pure and Applied Mathematics

Volume 82 No. 5 2013, 695-707

ISSN: 1311-8080 (printed version); ISSN: 1314-3395 (on-line version)
url: http://www.ijpam.eu
doi: http://dx.doi.org/10.12732/ijpam.v82i5.3
AP
ijpam.eu

GAUSS’ METHOD OF CONSTRUCTING

A REGULAR HEPTADECAGON

Yutaka Nishiyama
Department of Business Information
Faculty of Information Management
Osaka University of Economics
2, Osumi Higashiyodogawa Osaka, 533-8533, JAPAN

Abstract: This article explains how to construct a regular heptadecagon

according to the theory of cyclotonic equations which was discovered by Gauss
in 1796. The author also shows how to construct any root or fraction.

AMS Subject Classification: 20A02, 00A09, 97A20

Key Words: constructing a regular heptadecagon, theory of cyclotonic equa-
tions, modulo, prime number and primitive root, Constructing roots and frac-
tions

1. Introduction

We know about historically famous theories in mathematics, but there are many
cases where we don’t know the proofs. Fermat’s last theorem, Galois’ theorem,
Gödel’s incompleteness theorem..., once you start, there’s nowhere to stop. Un-
til recently I did not know the proof supporting Gauss’ method for constructing
a regular heptadecagon - a polygon with 17 sides. The construction method
for an arbitrary regular n-sided polygon may be explained as follows, according
to the ‘Dictionary of Mathematics’ (Iwanami Shoten Publishing). The neces-

c 2013 Academic Publications, Ltd.

Received: October 13, 2012 url: www.acadpubl.eu
696 Y. Nishiyama

sary and sufficient conditions for a regular n-sided polygon to be possible to

construct were established by Gauss. Given the prime factorization of n ,

n = 2λ P1 · · · Pk (λ ≥ 0) (1)

all of P1 , · · · , Pk must be distinct primes taking the form 2h + 1 (Fermat num-

bers). Inserting values for λ and h into this formula, the following values may
be obtained for n.
n = 3, 4, 5, 6, 8, 10, 12, 15, 17, · · ·
When I reproduced the following quote from a certain document, “It is well-
known that Gauss obtained a geometrical method for constructing a regular
heptadecagon”, I received the query, “By the way, how do you construct the
regular heptadecagon?” I had copied verbatim from the dictionary, and didn’t
know the answer. The construction methods for regular triangles and regular
pentagons were established by Euclid, the ancient Greek. However, considering
whether we in our generation have completely mastered even the construction
of regular pentagons, it is doubtful. For the regular pentagon, there are two
construction methods. One begins by establishing a single edge, while the
other contacts a circle internally. Neither requires a protractor, and both may
be drawn using only a pair of compasses and a ruler. Since it’s not the main
topic of this chapter the explanation is abbreviated, but it utilizes the facts
that √
π 1+ 5
cos = (2)
5 4
and p √
π 10 − 2 5
sin = . (3)
5 4
I am not an algebra specialist, so in the explanation that follows I’d like for
the reader to understand the construction method for the regular heptadecagon
attempted by Gauss as far as I have investigated it.

2. From Gauss’ Diary

Chapter 1 of Takagi Teiji’s ‘History of Modern Mathematics’ reports that when

on the 30th March 1796, the 19-year-old Gauss opened his eyes and arose from
his bed, a method for the outstanding problem of constructing a regular hep-
tadecagon occurred to him and was thus recorded in his diary. An outline of
the method goes as follows.[3]
GAUSS’ METHOD OF CONSTRUCTING... 697

If it is only required to prove the possibility of constructing a regular hep-

tadecagon, the solution is clear and simple. Taking

360◦ = 17φ,

if the value of cos φ may be expressed as a square root then it is possible

to construct the figure. cos φ represents the x coordinate of a point on the
circumference of a unit circle. Gauss demonstrated the computational process.
Let’s build on the explanation by looking at the method.
Firstly, Gauss made the following definitions.

cos φ + cos 4φ = a,
cos 2φ + cos 8φ = b,
cos 3φ + cos 5φ = c,
cos 6φ + cos 7φ = d.

The thing to pay attention to here is the substitution of the parameters

a, b, c, d for the values of cos φ to cos 8φ . Perhaps this kind of substitution is
only applied to this problem. If these are combined at random, then in total
there are 8 C2 × 6 C2 × 4 C2 = 2520 possibilities. Do you think Gauss might have
investigated every case? He did not. We will look at this issue in detail later,
and it is closely related to the theory of cyclotomic equations.
Next, setting
a + b = e,
c + d = f,
then as is widely known,
1
e+f =− . (1)
2
In order to understand Equation (1) it’s sufficient to remember a problem
which often comes up in university entrance examinations.
Theory. For natural number n, setting

Sn = cos φ + cos 2φ + · · · + cos nφ,

implies that
φ 2n + 1 φ
2Sn sin = sin φ − sin .
2 2 2
698 Y. Nishiyama

The proof of this theory involves the application of the product → sum
φ
formula for 2 cos kφ sin (k = 1, · · · , n), so that the intermediate term vanishes
2
2π
and the formula is simplified. Then, substituting n = 8, φ = into
17
2n + 1 φ φ
Sn = (sin φ − sin )/2 sin
2 2 2
2n + 1 1
yields sin φ = 0 . Therefore, S8 = − .
2 2
The products formed by each pair among a, b, c, d may now be obtained.
By means of a simple calculation, and taking note of the fact that cos nφ =
cos(17 − n)φ the results are
1
2ab = e + f = − ,
2
2ac = 2a + b + d,
2ad = b + c + 2d,
2bc = a + 2c + d,
2bd = a + 2b + c,
1
2cd = e + f = − .
2
Looking at one of the equations above, for example, 2ab, it is obtained as
follows

2ab = 2(cos φ + cos 4φ)(cos 2φ + cos 8φ)

= 2 cos φ cos 2φ + 2 cos φ cos 8φ + 2 cos 4φ cos 2φ + 2 cos 4φ cos 8φ
= (cos 3φ + cos φ) + (cos 9φ + cos 7φ) + (cos 6φ + cos 2φ) + (cos 12φ + cos 4φ).

Here the substitution cos 9φ = cos 8φ, cos 12φ = cos 5φ is made, and the
result rearranged, yielding
= (cos φ + cos 4φ) + (cos 2φ + cos 8φ) + (cos 3φ + cos 5φ) + (cos 6φ + cos 7φ)
=a+b+c+d
=e+f
1
=−
2
ergo,
2ac + 2ad + 2bc + 2bd = 4a + 4b + 4c + 4d,
 GAUSS’ METHOD OF CONSTRUCTING... 699

i.e.,
2ef = −2,
or alternatively,
ef = −1. (2)
The following is a solution method utilizing the relationship between the
solutions and factors of 2nd order equations. In particular, from (1) and (2),
the equations for e and f are the roots of
1
x2 + x − 1 = 0.
2
r r
1 17 1 17
One is thus − + , while the other is − − . A glance is sufficient
4 16 4 16
to reveal from their values that the first is e, and the second is f .
Now, the following equation has roots a and b.
1
x2 − ex − = 0.
4
The values of the roots are

1 1√ √
r q
1 1 1 2 1
e± + e =− + 17 ± 34 − 2 17.
2 4 4 8 8 8
It is clear here that a is the upper (positive) sign, and b is the lower (nega-
tive) sign. The reason is that, trivially,

a − b = (cos φ − cos 2φ) + (cos 4φ − cos 8φ)

so in exactly the same way,

1 1√ 1 √
q
c=− − 17 + 34 + 2 17,
8 8 8
and
1 1√ √
q
1
d=− − 17 − 34 + 2 17
8 8 8
Now finally, cos φ and cos 4φ are clearly the roots of the following 2nd order
1
equation (because the product cos φ · cos 4φ = c).
2
1
x2 − ax + c = 0
2
700 Y. Nishiyama

Consequently, r
1 1 2 1
cos φ = + a + a − c,
2 4 2
and r
1 1 2 1
cos 4φ = + a − a − c.
2 4 2
Rearranging however yields,

2a2 = 2 + b + 2c

so,
1√ √
r q
1 1 1 1 1 1
cos φ = a + + b− c=− + 17 + 34 − 2 17
2 r 4 8 4 16 16 16
1 √ √ √
q q
+ 17 + 3 17 − 34 − 2 17 − 2 34 + 2 17
8
which is the value Gauss obtained for cos φ.

3. Gauss’ Theory of Cyclotomic Equations

We have seen how the value of cos φ needed for the construction of a regular
heptadecagon can be obtained, but this calculation was just a confirmation. So
far we have not even touched upon the essential question of why it turns out
as it does, i.e., the reason for the substitution of the four parameters a, b, c, d
for the terms from cos φ to cos 8φ must be stated.
The reason is not written in Gauss’ diary. In order to find out, one must
enlist the help of Kurata Reijirou’s ‘Gauss’ Theory of Cyclotonic Equations’,[2]
or ‘Gauss’ Theory of Numbers’ translated by Takase Masahito.[1] Think about
the following equation.
xn − 1 = 0 (4)
It goes without saying that the roots of this equation are the nth roots of 1,
and as is widely known,
2πki
e n (k = 0, 1, 2, · · · , n − 1).

This has a relationship with Euler’s equation.

eiθ = cos θ + i sin θ (5)

Among the n roots, those which first equal 1 when raised to the power n
2πi
are known as the primitive nth roots of unity. e n is a primitive nth root of
GAUSS’ METHOD OF CONSTRUCTING... 701

2π
unity. This is a point with angle on a unit circle in the complex plain, i.e.,
n 2πki
it expresses an nth equal part of a complete circumference. e n expresses the
point on the circumference with k-times the angle.
Through the intermediate agency of the complex plain, Equation (4) is tied
to th n-sided regular polygon figures originating in ancient Greece. This was
Gauss’ underlying perspective, and the first mathematician to introduce the
complex plain was Gauss.
Now, Equation (4) is

xn − 1 = (x − 1)(xn−1 + xn−2 + · · · + x + 1),

so excluding 1, all the roots of Equation (4) are roots of the following function.

F (x) = xn−1 + xn−2 + · · · + x + 1 (6)

This is known as a cyclotomic equation, or alternatively, as a circle-partitio-

ning equation.
In xn − 1 = 0 , if a primitive nth root of unity, denoted ω , is already known,
then the solutions of this equation are

1, ω, ω 2 , · · · , ω n−1 .

For example, thinking about x3 − 1 = 0 ,

x3 − 1 = (x − 1)(x2 + x + 1) = 0,
√ √ √
−1 ± 3i −1 + 3i −1 − 3i
so the 3rd root is 1, . Taking = ω, we have = ω2,
2 2 2
and it can be seen that the roots are 1, ω and ω 2 . In this case, 3θ = 2π so
ω = cos θ + i sin θ = eiθ
ω 2 = cos 2θ + i sin 2θ = ei2θ
ω 3 = cos 3θ + i sin 3θ = ei3θ = 1
and it can be seen that the 3 roots cycle.
At this point I’d like to introduce the following theorem which utilizes the
concept of a remainder.

Theorem. When p is a prime number, the set of roots of the cyclotomic

equation
F (x) = xn−1 + xn−2 + · · · + x + 1.
702 Y. Nishiyama

may be written as Ω, and apart from 1, all the elements r ∈ Ω (which satisfy
xn − 1 = 0) are complex numbers. Furthermore, for a positive or negative
integer e which is not divisible by p, the following are satisfied.
(1) r p = 1, r 2p = r 3p = · · · = 1, r ep = 1;
(2) for integers λ and µ

λ ≡ µ (mod p) ⇔ r λ = r µ ;

(3) for r ∈ Ω

Ω = {r e , r 2e , · · · , r e(p−1) }, r e + r 2e + · · · + r e(p−1) = −1.

Gauss defined a parameter with a value representing an f -step cycle.

Definition. For an odd prime p, and a primitive pth root of unity, denoted
r, and taking p − 1 = f e, g as a primitive root of p, and λ as an arbitrary
integer, the value of the f -step cycle, denoted (f, λ), is defined as follows

(f, λ) = [λ] + [λh] + [λh2 ] + · · · + [λhf −1 ]

(note that h = g e ).
At this point let’s perform the calculation for a regular heptadecagon. Tak-
ing 3 as a primitive root of the prime number 17,
p = 17, p − 1 = 16 = 16 × 1 = f × e, g = 3, h = g e = 31 = 3,
and the 16-step cycle is as follows.
(16, 1) = [1] + [3] + [9] + [10] + [13] + [5] + [15] + [11] + [16] + [14]
+[8] + [7] + [4] + [12] + [2] + [6]
That is to say, when 3 is taken as the primitive root for {[1], · · · , [16]}, the result
is a cyclic group modulo 17. For example, the 4th term on the right-hand side
of the equation above, [10], can be obtained as

λh3 = 1 × 33 = 27 = 10 (mod17).

An explanation of primitive roots, modulo, and cyclic groups is deferred to

books specialized in algebra, number theory, groups, and so on.
At any rate, primes and primitive roots have a truly ingenious relationship.
I first heard of the term ‘primitive root’ when I studied subroutines for gen-
erating pseudorandom numbers on computers. As a consequence, computers
GAUSS’ METHOD OF CONSTRUCTING... 703

can be made to generate randomly ordered sequences of all the integers they
are capable of expressing. For the problem in question, the significant point
is that the 16 roots can be reordered according to the concept of remainders.
This was how the 16-step cycle could be obtained. Gauss presented a theory
decomposing an f -step cycle.

Theorem. When p − 1 = abc, the bc-step cycle (bc, λ) is the sum of b c-step
cycles
(bc, λ) = (c, λ) + c, λga + c, λg2a + · · · + c, λg a(b−1)

Let’s attempt a decomposition of the 16-step cycle into 2 8-step cycles as

described on p27-28 of reference (Kurata, 1988).[2] From p = 17, p − 1 = 16 =
1 × 2 × 8 = a × b × c,
(16, 1) = (8, 1) + (8, 3).
Defining (8, 1) and (8, 3) we have the following equations.

(8, 1) = [1] + [9] + [92 ] + [93 ] + [94 ] + [95 ] + [96 ] + [97 ]

= [1] + [9] + [81] + [729] + [6561] + [59049] + [531441] + [4782969]
= [1] + [9] + [13] + [15] + [16] + [8] + [4] + [2] (mod17)
= [1] + [2] + [4] + [8] + [9] + [13] + [15] + [16]
(8, 3) = [3] + [3 · 9] + [3 · 92 ] + [3 · 93 ] + [3 · 94 ] + [3 · 95 ] + [3 · 96 ] + [3 · 97 ]
= [3] + [27] + [243] + [2187] + [19683] + [177147] + [1594323]
+[14348907]
= [3] + [10] + [5] + [11] + [14] + [7] + [12] + [6] (mod17)
= [3] + [5] + [6] + [7] + [10] + [11] + [12] + [14]

Here, the 16-step cycle (16, 1) is decomposed into 2 8-step cycles, (8, 1)
and(8, 3). If the decomposition theory is applied repeatedly, the decomposition
can be continued down to a final 1-step cycle. Figure 1 shows a breakdown of
the whole decomposition process. The parameters a, b, c, d, e, f and values of
cos φ to cos 8φ recorded in “Gauss’ diary” are compiled in this figure. Can you
understand the reason why Gauss chose the values
cos φ + cos 4φ = a, cos 3φ + cos 5φ = c,
704 Y. Nishiyama

cos 2φ + cos 8φ = b, cos 6φ + cos 7φ = d

a + b = e, c + d = f

by means of this result?

For p = 17, the 16 roots besides 1 may be calculated as shown in Table 1, in
a similar way. Figure 2 shows a diagram of a regular heptadecagon. By means
of such a decomposition, it becomes possible to factorize the 2nd term on the
right-hand side of the equation

Figure 1: x17 − 1 = (x − 1)(x16 + x15 + · · · + x + 1).

Why it turns out like this, at present, I do not know. In order to find out it will
be necessary to make further detailed study, and this is surely a place where
the profound beauty of algebra lies. I have a feeling that I have experienced a
glimpse of this beauty.
GAUSS’ METHOD OF CONSTRUCTING... 705

Figure 2: The decomposition process of the f -step cycle

[1], [16] 0.9324722294 ± 0.3612416662i

[2], [15] 0.7390089172 ± 0.6736956436i
[3], [14] 0.4457383558 ± 0.8951632914i
[4], [13] 0.0922683595 ± 0.9957341763i
[5], [12] −0.2736629901 ± 0.9618256432i
[6], [11] −0.6026346364 ± 0.7980172273i
[7], [10] −0.8502171357 ± 0.5264321629i
[8], [9] −0.9829730997 ± 0.1837495178i

Table 1. The 16 roots

4. Constructing roots, and more...

The value of cos φ was expressed using roots, so let’s look at their construction
process specifically. Construction problems in mathematics are not solved by
calculating numerical values using a computer and drawing diagrammatic illus-
trations. The figure must be drawn using only a ruler and compasses, and given
two points on a plane (points 0 and 1, i.e., a line segment) as a unit length.
The ruler is only used for drawing straight lines, and is considered incapable of
measurement.
The problem is then how to draw the figure using root lengths. It is possible
to construct the length of any root, in a manner similar to those shown in Figure
3. First, a √
unit square is drawn. By the triple angle
√ theory, the diagonal edge
has length 2. Secondly, a circular arc with radius 2 is drawn, and this length
is transferred to the line extended from the√ base edge of the square. In this
way, a rectangle with height 1, and width 2 can be constructed.
706 Y. Nishiyama

Once again √applying the triple angle theory to this rectangle,

√ the
√ length
√ √of
the diagonal is 3. By proceeding in this manner, the lengths√ 4, 5, 6, 7
and so on can be obtained. The construction of the length 17 can also be
thus achieved.
Next, rather than the roots of integers, let’s think about the construction
of roots of arbitrary numerical values. Pages 50-51 of Kurata (1988) discuss
this in more detail and contain an explanation of how to construct fractions
a √
(of the form ) and roots (of the form a). See Figures 4 and 5. Even in the
b
present age, when computers have advanced so far, the construction method
for the regular heptadecagon discovered by Gauss around 200 years ago in 1796
still impresses anew. For me it simply reaffirmed Gauss’ magnificent mental
capacity and creative abilities.[?]

Figure 3: The length of a root

a
Figure 4: The construction of a fraction ( )
b
GAUSS’ METHOD OF CONSTRUCTING... 707

√
Figure 5: The construction of a root ( a)

References

[1] C.F. Gauss, (trans. by M. Takase), Gauss Seisuron [Gauss’ Number The-
ory, Disquisitiones Arithmeticae], Tokyo: Asakura, (1995).

[2] R. Kurata, Gauss Enbun Hoteishikiron [Gauss’ Theory of Cyclotonic Equa-

tions], Nagoya: Kawai cultural education research laboratory, (1988).

[3] T. Takagi, Kinsei Sugakushidan [History of Modern Mathematics], 3rd

edition, Tokyo: Kyoritsu, (1987).
708