The function tangent and polynomials
Andrzej Nowicki Version atg-01. May 15, 2015
It is known (see for example [1], [4]) that the numbers tanπ8, tan3π8 , tan5π8 , tan7π8 are distinct roots of the equation x4− 6x2+ 1 = 0, and the numbers tanπ7, tan2π7 , . . . , tan6π7 are distinct roots of the equation x6− 21x4+ 35x2− 7 = 0.
This note contains some proofs of the above facts and their generalizations. Our note is based on [3], [2] and [5].
1 The polynomials f
n(x) and g
n(x)
If n is a nonnegative integer, then we denote by fn(x) and gn(x) the polynomials in one variable x defined by
(1 + ix)
n= f
n(x) + ig
n(x)
.The polynomials fn(x) and gn(x) belong to the polynomial ring Z[x], that is, they are polynomials with integer coefficients.
f0(x) = 1, f1(x) = 1,
f2(x) = −x2+ 1 = −(x − 1)(x + 1), f3(x) = −3x2 + 1,
f4(x) = x4− 6x2+ 1 = (x2 + 2x − 1)(x2− 2x − 1), f5(x) = 5x4− 10x2+ 1,
f6(x) = −x6+ 15x4− 15x2+ 1 = (x2 + 2x − 1)(x2− 2x − 1), f7(x) = −7x6 + 35x4− 21x2+ 1,
f8(x) = x8− 28x6+ 70x4− 28x2+ 1
= (x4− 4x3− 6x2+ 4x + 1)(x4+ 4x3− 6x2− 4x + 1), f9(x) = 9x8− 84x6+ 126x4− 36x2+ 1
= (3x2− 1)(3x6− 27x4+ 33x2− 1),
f10(x) = −x10(x) + 45x8− 210x6+ 210x4− 45x2+ 1
= −(x − 1)(x + 1)(x3− 4x3− 14x2− 4x + 1)(x4+ 4x3− 14x2+ 4x + 1), f11(x) = −11x10(x) + 165x8− 462x6+ 330x4− 55x2+ 1,
g1(x) = x, g2(x) = 2x,
g3(x) = −x3 + 3x = −x(x2− 3),
g4(x) = −4x3+ 4x = −4x(x − 1)(x + 1), g5(x) = x5 − 10x3+ 5x = x(x4− 10x2 + 5), g6(x) = 6x5− 20x3+ 6x = 2x(3x2− 1)(x2− 3),
g7(x) = −x7 + 21x5− 35x3+ 7x = −x(x6− 21x4+ 35x2− 7),
g8(x) = −8x7+ 56x5 − 56x3+ 8x = −8x(x − 1)(x + 1)(x2+ 2x − 1)(x2− 2x − 1), g9(x) = x9 − 36x7+ 126x5− 84x3+ 9x = x(x2− 3)(x6− 33x4+ 27x2− 3),
g10(x) = 10x9− 120x7+ 252x5 − 120x3+ 10x = 2x(5x4− 10x2+ 1)(x4− 10x2+ 5).
Each fn is an even polynomial, and each gn is an odd polynomial divisible by x.
Thus we have the equalities
fn(−x) = fn(x), gn(−x) = −gn(x).
Moreover,
deg f2s = 2s, deg f2s+1 = 2s, deg g2s = 2s − 1, deg g2s+1 = 2s + 1.
It is clear that we have the following equalities
(1 − ix)n = fn(x) − ign(x), fn(x) = (1+ix)n+(1−ix)2 n, gn(x) = (1+ix)n2i−(1−ix)n.
If h is a nonzero polynomial in a variable x, then let us denote by w(h) the initial coefficient of h. For example, if h = 5x3 − 4x + 7, then w(h) = 5. It is easy to prove:
Proposition 1.1. w (f2s) = (−1)s, w (f2s+1) = (−1)s(2s + 1), w (g2s) = (−1)s+12s, w (g2s+1) = (−1)s.
Since fn(x) + ign(x) = (1 + ix)n, we have
fn(x) + ign(x) = 1 + in1x1−n2x2− in3x3+n4x4+ in5x5− · · · . In the case n = 4k we have
f4k(x) = x4k−4k−24k x4k−2+4k−44k x4k−4− · · · −4k2x2+ 1, g4k(x) = −4k−14k x4k−1+4k−34k x4k−3− · · · −4k3x3 +4k1x.
Hence:
f4k(x) =
2k
X
j=0
(−1)j4k2jx2j, g4k(x) =
2k−1
X
j=0
(−1)j2j+14k x2j+1. By the same way we obtain the following 8 equalities:
f4k(x) = P2k
j=0
(−1)j4k2jx2j, g4k(x) = 2k−1P
j=0
(−1)j2j+14k x2j+1, f4k+1(x) = P2k
j=0
(−1)j4k+12j x2j, g4k+1(x) = P2k
j=0
(−1)j4k+12j+1x2j+1, f4k+2(x) = 2k+1P
j=0
(−1)j4k+22j x2j, g4k+2(x) = P2k
j=0
(−1)j4k+22j+1x2j+1, f4k+3(x) = 2k+1P
j=0
(−1)j4k+32j x2j, g4k+3(x) = 2k+1P
j=0
(−1)j4k+32j+1x2j+1. Thus, we have the following 4 equalities:
f
2s(x) =
Psj=0
(−1)
j2s2jx
2j, g
2s(x) =
s−1Pj=0
(−1)
j2j+12sx
2j+1, f
2s+1(x) =
Psj=0
(−1)
j2s+12jx
2j, g
2s+1(x) =
Psj=0
(−1)
j2s+12j+1x
2j+1,
.
The polynomials gn(x) are divisible by x. Let us introduce the notation:
hn(x) = gn(x) x ,
for all n. Observe that if n is odd, then the polynomials fn(x) and hn(x) are of the same degree n − 1.
Proposition 1.2. For every integer k > 0 we have the following two equalities:
x4kf4k+11x= h4k+1(x), x4k+2f4k+31x= −h4k+3(x).
Hence
x
2sf
2s+1 1x= (−1)
sh
2s+1(x).
. Proof. Assume that n = 4k + 1. Then we have:x4kf4k+11x = x4k P2k
j=0
(−1)j4k+12j x12j
= P2k
j=0
(−1)j4k+12j x4k−2j = x1 P2k
j=0
(−1)j4k+1−2j4k+1 x4k+1−2j
= 1x P2k
j=0
(−1)j4k+12j+1x2j+1 = 1xg4k+1(x)
= h4k+1(x).
Now let n = 4k + 3:
x4k+2f4k+3
1 x
= x4k+22k+1P
j=0
(−1)j4k+32j 1x2j
= 2k+1P
j=0
(−1)j4k+32j x4k+2−2j = x12k+1P
j=0
(−1)j4k+3−2j4k+3 x4k+3−2j
= −1x2k+1P
j=0
(−1)j4k+32j+1x2j+1 = −1xg4k+3(x)
= −h4k+3(x).
This completes the proof.
2 Divisibility properties of f
n(x) and g
n(x)
Proposition 2.1. For all nonnegative integers m, n we have the following two equali- ties:
f
n+m(x) = f
n(x)f
m(x) − g
n(x)g
m(x) g
n+m(x) = f
n(x)g
m(x) + f
m(x)g
n(x)
.Proof.
fn+m(x) + ign+m(x) =
1 + ix
n+m
=
1 + ix
n
1 + ix
m
=
fn(x) + ign(x)
fm(x) + igm(x)
=
fn(x)fm(x) − gn(x)gm(x)
+ i
fn(x)gm(x) + fm(x)gn(x)
. This completes the proof.
In particular, f2n(x) = fn(x)2− gn(x)2 and
g
2n(x) = 2f
n(x)g
n(x)
.Proposition 2.2 ([5], [2]). The polynomials fn(x), gn(x) are relatively prime in Q[x].
Proof. Since Z[i][x] is a unique factorization domain, it is enough to prove that the polynomials fn(x) and gn(x) are relatively prime in the domain Q(i)[x].
Let n be a fixed positive integer and suppose that these polynomials are not relatively prime. Then fn(x) and gn(x) are divisible by an irreducible polynomial h(x) ∈ Q(i)[x]. Obviously deg h(x) > 1. Then h(x) divides the polynomials (1 + ix)n and (1 − ix)n. As h(x) is irreducible, we deduce that the polynomials (1 + ix) and (1 − ix) are divisible by h(x). This implies that h(x) divides the number 2. But it is a contradiction, because deg h(x) > 0.
Note next propositions concerning the divisibility of fn(x) and gn(x).
Proposition 2.3. If m | n, then gm(x) | gn(x) (divisibility in Z[x]).
Proof. We will prove by an induction that for every positive integer k, the poly- nomial gm divides the polynomial gkm. It is obvious in the case k = 1. Assume that k > 1 and gm divides gkm in Z[x]. Let gkm = u · gm with u ∈ Z[x]. Then by Proposition 2.1, we have
g(k+1)m = gkm+m = fkmgm+ fmgkm
= fkmgm+ fmugm =
fkm+ ufm
gm and so, gm divides g(k+1)m in Z[x].
We may prove also:
Proposition 2.4 ([2]). If m | n, then gn(x) = vm(x)gm(x), where vm(x) ∈ Z[x], and the polynomials vm(x) and gm(x) are relatively prime in Q[x].
Proof. Let n = dm, where d is a positive integer. If m = n, then vm(x) = 1 and we are done. Assume that m < n. Then d> 2 and we have
fn+ ign = (1 + ix)n = (1 + ix)dm = (fm+ igm)d
= fmd + idfmd−1gm−d2fmd−2gm2 − id3fmd−3g3m+ · · · ,
and hence gn = dfmd−1gm + u · gm2, where u = u(x) is a polynomial belonging to Z[x].
Put
vm(x) = dfmd−1+ ugm.
Then gn(x) = vm(x)gm(x). Since gcd(fm, gm) = 1 (see Proposition 2.2), we have gcd(gm, vm) = gcdgm, dfmd−1+ ugm= gcdgm, dfmd−1= 1.
Therefore, the polynomials gm(x) and vm(x) are relatively prime in Q[x].
Observe that the polynomials f3 = −3x2 + 1, f5 = 5x4 − 10x2 + 1 and f7 =
−7x6+ 35x4− 21x2+ 1 are irreducible in Q[x] (and also in Z[x]). Now we will prove that the same is true for all the polynomials fp, where p is an odd prime number. Let us recall that hn(x) = gnx(x).
Proposition 2.5. If p > 3 is a prime number, then the polynomials fp(x) and hp(x) are irreducible in Z[x].
Proof.
fp(x) + igp(x) = (1 + ix)p
= 1 +p1ix −p2x2− ip3x3+ · · · ±p−1p xp−1∓ ixp.
This implies that fp(x) = 1 −p2x2 +p4x4 + · · · ±p−1p xp−1 and gp(x) = x · hp(x), where
hp(x) =p1−p3x2+ · · · ∓p−2p xp−3± xp−1.
As all the numbers p1, p2, . . . , p−1p are divisible by p, we deduce by Eisenstein’s theorem that the polynomials fp(x) and hp(x) are irreducible in Z[x].
Proposition 2.6 ([5]). If n, k are nonnegative integers, then:
(1) fn | g2kn; (2) fn | f(2k+1)n;
(3) the polynomials fkn and gn are relatively prime;
(4) gcd (gkn+r, gn) = gcd (gr, gn) for integers r > 0.
Proof. (1). Since g2n = 2fngn, we have fn| g2n. But g2n | g2kn(see Proposition 2.3).
Hence, fn| g2kn.
(2). It is obvious for k = 0. Let k > 0 and assume that fn| f(2k+1)n. Since f(2k+3)n = f(2k+1)n+2n= f(2k+1)nf2n− g(2k+1)ng2n
and fn| g2n, we have fn | f(2k+3)n.
(3). It is obvious for k = 0, because f0 = 1. If k = 1, then it follows from Proposi- tion 2.2. Let k > 1 and assume that gcd (fkn, gn) = 1. Then we have: gcdf(k+1)n, gn= gcd (fkn+n, gn) = gcd (fknfn− gkngn, gn) = gcd (fknfn, gn) = gcd (fkn, gn) = 1.
(4). We use the previous properties:
gcd (gkn+r, gn) = gcd (fkngr+ gknfr, gn) = gcd (fkngr, gn) . But gcd (fkn, gn) = 1, so gcd (gkn+r, gn) = gcd (gr, gn).
Now, by Proposition 2.6 and the Euclid algorithm, we have:
Theorem 2.7. If n, m are positive integers, then
gcd (g
m, g
n) = g
gcd(m,n) .3 Zeros of the polynomials f
n(x) and g
n(x)
Let n> 1 be a fixed integer. Consider the numbers of the form
t
j= tan jπ n
!
,
where {0, 1, 2, . . . , n − 1} and n 6= 2j.
If n jest odd, then we have n distinct real numbers t0 = 0, t1, . . . , tn−1. If n = 2m is even, then we have n − 1 distinct real numbers
t0 = 0, t1, . . . , tm−2, tm−1, tm+1, tm+2, . . . , tn−1.
Theorem 3.1. All zeros of each polynomial gn(x) are real numbers.
More precisely, if n is odd, then all the real numbers t0, t1, . . . , tn−1 are distinct zeros of gn. If n = 2m is even then all the real numbers t0, t1, . . . , tm−1, tm+1, tm+2, . . . , tn−1 are distinct zeros of gn.
Proof. Assume that j ∈ {0, 1, . . . , n − 1}, n 6= 2j. Then we have:
fn(tj) + ign(tj) = (1 + itj)n=1 + i tanjπnn =
1 + isin
jπ n
cosjπn
n
= cosjπn−ncosjπn + i sinjπnn
= (−1)jcosjπn−n and this implies that gn(tj) = 0.
Theorem 3.2. All zeros of each polynomial fn(x) are real numbers.
Proof. It follows from Theorem 3.1 and the equality g2n= 2fngn. As a consequence of the above theorems we obtain
Proposition 3.3. For every n > 1, the polynomials fn(x) and gn(x) are square-free.
The next proposition is also a consequence of the above theorems.
Proposition 3.4. If q 6= 12 is a rational number, then the number tan (qπ) is algebraic.
Proof. Put r = tan(qπ). Then ±r = tannjπ, where j, n are nonnegative integers with n > 1 and j < n, and of course n 6= 2j. We know, by Theorem 3.1, that ±r is a zero of the polynomial gn(x). All the coefficients of gn(x) are integers, so ±r is algebraic, and so r is algebraic.
Theorem 3.5. If q is a rational number, then all the numbers sin(qπ), cos(qπ), tan (qπ), cot(qπ) (except when undefined) are algebraic.
Proof. It is a consequence of Proposition 3.4 and the following known equalities:
sin α = 2t
1 + t2, cos α = 1 − t2
1 + t2, cot α = 1 tan α, where t = tanα2.
Theorem 3.6. If q is a rational number, then all the numbers sin(qo), cos(qo), tan (qo), cot(qo) are algebraic.
Proof. Since the angle 180o (180 degrees) equals π radians, we have qo = 180q π.
Therefore, this theorem is a consequence of the previous theorem.
4 Equalities with tangents
Look again at Theorems 3.1, 3.2 and the initial coefficients described in Proposition 1.1.
Observe that as a consequence of the above facts we obtain the following equalities.
Proposition 4.1.
f
2m(x) = (−1)
m 2mQk=1
x − tan
(2k−1)π4m
, f
2m+1(x) = (−1)
m(2m + 1)
Qk∈A
x − tan
(2k+1)π4m+2
,
where m> 1 and A = {0, 1, 2, . . . , 2m} r {m}.
Proposition 4.2.
g
2m+1(x) = (−1)
mx
2mQk=1
x − tan
2m+1kπ, g
2m(x) = (−1)
m+12mx
Qk∈A
x − tan
2mkππ
,
where m> 1 and A = {0, 1, 2, . . . , 2m − 1} r {m}.
Since tan (π − α) = −tan (α), we may rewrite the above two propositions in the following way.
Proposition 4.3.
f
2m(x) = (−1)
m mQk=1
x
2− tan
2(
(2k−1)π4m)
f
2m+1(x) = (−1)
m(2m + 1)
m−1Qk=0
x
2− tan
2(
(2k+1)π4m+2)
g
2m(x) = (−1)
m+12mx
m−1Qk=1
x
2− tan
2(
2mkπ)
g
2m+1(x) = (−1)
mx
mQk=1
x
2− tan
2(
2m+1kπ)
.
Let us recall (see Proposition 1.2) that x2mf2m+1
1 x
= (−1)mh2m+1(x).
where h2m+1(x) = g2m+1x (x). The zeros of the polynomial h2m+1(x) are real numbers of the form tan2m+1kπ , where k = 1, . . . , 2m. Thus, we have:
Proposition 4.4. The zeros of the polynomial f2m+1(x) are real numbers of the form cot kπ
2m + 1, k = 1, 2, . . . , 2m.
Thus, we have the equalities
f
2m+1(x) = u
m 2mQk=1
x − cot
2m+1kπ= u
m mQk=1
x
2− cot
22m+1kπ,
where um = (−1)m(2m + 1).
Recall that hn(x) = gnx(x). Look again at the polynomial hn(x) in the case n = 4k+1:
h4k+1(x) =
2k
X
j=0
(−1)j4k+12j+1x2j =
2k
Y
j=1
x2− tan2(4k+1jπ ).
Comparing the constant terms, we obtain the equality 4k+11 = Q2k
j=1
tan2(4k+1jπ ). Com- paring the coefficients of x4k−2, we obtain the equality 4k+12 = P2k
j=1
tan2(4k+1jπ ). Hence, we have the following two identities:
2k
Y
j=1
tan
jπ 4k + 1
=√
4k + 1,
2k
X
j=1
tan2
jπ 4k + 1
= 2k(4k + 1).
Using the same in the case n = 4k + 3, we obtain:
2k+1
Y
j=1
tan
jπ 4k + 3
=√
4k + 3,
2k+1
X
j=1
tan2
jπ 4k + 3
= (2k + 1)(4k + 3).
Assume now that n is even. For n = 4k we have:
h4k(x) = −4k
2k−1
Y
j=1
x2− tan2(jπ4k)=
2k−1
X
j=0
(−1)j2j+14k x2j,
Comparing the constant terms, we obtain the equality 4k1 = 4k2k−1Q
j=1
tan2(jπ4k). Com- paring the coefficients of x4k−4, we obtain the equality 4k3 = 4k2k−1P
j=1
tan2(jπ4k). Hence, we have the following two identities:
2k−1
Y
j=1
tan
jπ 4k
= 1,
2k−1
X
j=1
tan2
jπ 4k
= 1
3(2k − 1)(4k − 1).
Using the same in the case n = 4k + 2, we obtain:
2k
Y
j=1
tan
jπ 4k + 2
= 1,
2k
X
j=1
tan2
jπ 4k + 2
= 1
32k(4k + 1).
Therefore, we proved the following proposition.
Proposition 4.5.
m−1 Q
j=1
tan
2mjπ= 1,
m−1Pj=1
tan
22mjπ=
13(m − 1)(2m − 1),
m Q
j=1
tan
2m+1jπ= √
2m + 1,
Pmj=1
tan
22m+1jπ= m(2m + 1).
.
By the same method, using Proposition 4.4, we obtain the following equality with cotangents.
Proposition 4.6 ([1] 123).
m X
j=1
cot
2jπ
2m + 1 = m(2m − 1)
3
.5 Examples and applications
5.1. Multiples of π5 = 36o. (1) tanπ5 · tan 2π5 =√
5.
(2) tan2π5+ tan22π5 = 10.
(3) cot2π5+ cot22π5 = 2.
(4) tanπ5 =
q
5 − 2√
5, tan2π5 =
q
5 + 2√ 5.
(5) The numbers tanπ5, tan2π5 , tan3π5 , tan4π5 are zeros of the polynomial h5(x) = x4− 10x2+ 5.
(6) The numbers cotπ5, cot2π5 , cot3π5 , cot4π5 są are zeros of the polynomial f5(x) = x4− 10x2+ 1.
5.2. Multiplies of π7.
(1) tanπ7 · tan 2π7 · tan 3π7 =√ 7.
(2) tan2π7+ tan22π7 + tan23π7 = 21.
(3) cot2π7+ cot22π7 + cot23π7 = 5.
(4) The numbers tanπ7, tan2π7 , . . . , tan6π7 are zeros of the polynomial h7(x) = x6− 21x4+ 35x2− 7.
(5) The numbers cotπ7, cot2π7 , . . . , cot6π7 are zeros of the polynomial f7(x) = −7x6+ 35x4− 21x2+ 1.
5.3. Multiplies of π8 = 22, 5o. (1) tanπ8 · tan 2π8 · tan 3π8 = 1.
(2) tan2π8+ tan22π8 + tan23π8 = 7.
(3) tanπ8 =
q
3 − 2√
2, tan2π8 = 1, tan3π8 =
q
3 + 2√ 2.
(4) The numbers tanπ8, tan3π8 , tan5π8 , tan7π8 are zeros of the polynomial f4(x) = x4− 6x2+ 1.
5.4. Multiplies of π9 = 20o.
(1) tanπ9 · tan 2π9 · tan 3π9 · tan 4π9 = 3.
(2) tan2π9+ tan22π9 + tan23π9 + tan24π9 = 36.
(3) cot2π9+ cot22π9 + cot23π9 + cot24π9 = 283.
(4) The numbers tanπ9, tan2π9 , . . . , tan8π9 are zeros of the polynomial h9(x) = x8− 36x6+ 126x4− 84x2 + 9 = (x2− 3)(x6− 33x4+ 27x2− 3) (5) The numbers cotπ9, cot2π9 , . . . , cot8π9 are zeros of the polynomial
f9(x) = 9x8− 84x6+ 126x4− 36x2+ 1 = (3x2− 1)(3x6− 27x4+ 33x2− 1).
5.5. Multiplies of 10π = 18o.
(1) tan10π · tan 2π10 · tan 3π10 · tan4π10 = 1.
(2) tan210π+ tan22π10+ tan23π10+ tan24π10= 12.
(3) tan10π =q5−2
√ 5
5 , tan3π10 =q5+2
√ 5 5 .
(4) All numbers of the form tanjπ10, where j ∈ {1, 2, 3, 4, 6, 7, 8, 9}, are zeros of the polynomial
h10(x) = 10x8− 120x6+ 252x4− 120x2+ 10 = 2(5x4− 10x2+ 1)(x4− 10x2+ 5).
(5) The numbers tan10π, tan3π10, tan7π10, tan9π10 are zeros of the polynomial f5(x) = 5x4− 10x2+ 1.
Literatura
[1] T. Andreescu, Z. Feng, 103 Trigonometry Problems. From the training of the USA IMO team, Birkh¨auser, Boston - Basel - Berlin, 2005.
[2] X. Lin, Infinitely many primes in the arithmetic progression kn − 1, The American Mathematical Monthly, 122(1)(2015), 48-50.
[3] T. Nagell, Introduction to Number Theory, Chelsea Publishing Company, New York, 1964.
[4] A. Nowicki, Real Numbers and Functions (in Polish), Podróże po Imperium Liczb, part 10, Second Edition, OWSIiZ, Toruń, Olsztyn, 2013.
[5] A. Nowicki, S. Spodzieja, Polynomial imaginary decomposition for finite extensions of fields of characteristic zero, Bull. Pol. Sci. Acad., 51(2)(2003), 157-168.
Nicolaus Copernicus University, Faculty of Mathematics and Computer Science, 87-100 Toruń, Poland, (e-mail: anow@mat.uni.torun.pl).