J o u r n a l of

Mathematics
and Applications
JMA No 37, pp 39-50 (2014)

Remarkable identities
Jan Górowski, Jerzy Żabowski
Abstract: In the paper a number of identities involving even powers
of the values of functions tangent, cotangent, secans and cosecans are
proved. Namely, the following relations are shown:
m−1  
X πj
f 2n = wf (m),
j=1
2m
m−1  
X 2j + 1
f 2n π = vf (m),
j=0
4m
m  
X
2n πj
f = w̃f (m),
j=1
2m + 1

where m, n are positive integers, f is one of the functions: tangent, cotan-

gent, secans or cosecans and wf (x), vf (x), w̃f (x) are some polynomials
from Q[x].
One of the remarkable identities is the following:
m−1
X (2j + 1)π
sin−2 = m2 , provided m ≥ 1.
j=0
2m

Some of these identities areP used to find, by elementary means, the

∞ 1
sums of the series of the form j=1 j 2n , where n is a fixed positive in-
teger. One can also notice that Bernoulli numbers appear in the leading
coefficients of the polynomials wf (x), vf (x) and w̃f (x).

AMS Subject Classification: 11M06, 11Y60, 10H10

Keywords and Phrases: trigonometric identity, zeta function

In [7] the following formulas have been proved

m m
X πj m(2m − 1) X πj 2m(m + 1)
cot 2
= , sin−2 = , (1)
j=1
2m + 1 3 j=1
2m + 1 3

COPYRIGHT c by Publishing Department Rzeszów University of Technology

P.O. Box 85, 35-959 Rzeszów, Poland
40 J. Górowski, J. Żabowski

where m ∈ N1 . By Nk for a positive integer k we mean N \ {0, 1, 2 . . . , k − 1}. The

P∞ 2
above identities were then used in an elementary proof of the formula k=1 k12 = π6 .
In this paper we develop the ideas from [7] to prove more P∞ generalized identities
than (1). Next we use some of them to find the sum of k=1 k12n , where n ∈ N1 .
The general identities given in this article yield, in particular, the following identity
of uncommon beauty
m−1
X 2j + 1
sin−2 π = m2 , m ∈ N1 .
j=0
2m
P∞ 1
Some elementary methods of finding the sums of the series of the form j=1 j 2n may
be found for example in [1], [3], [5], [6], [8].
We start by recalling some basic facts on symmetric polynomials in m variables.
Put
X m
σn = xnj for n ∈ N1 ,
j=1
X
τk = xj1 xj2 · · · xjk for k ∈ {1, 2, . . . , m}.
1≤j1 <j2 <···<jk ≤m

Moreover, for the convenience set τk = 0 for k > m.

The following lemma comes from [2].
Lemma 1 (Newton). Let n ∈ N1 , then

σn − τ1 σn−1 + τ2 σn−2 − · · · + (−1)n−1 τn−1 σ1 + (−1)n nτn = 0. (2)

In view of Lemma 1 we have

(−1)n+1 nτn (−1)n−2 τn−2 (−1)n−1 τn−1
 
−τ1 τ2 ...
 (−1)n (n − 1)τn−1 1 −τ1 ... (−1)n−3 τn−3 (−1)n−2 τn−2 
 (−1)n−1 (n − 2)τn−2 (−1)n−4 τn−4 (−1)n−3 τn−3
 
0 1 ... 
σn = det 
 
.. .. .. .. .. .. 

 . . . . . . 

 −2τ2 0 0 ... 1 −τ1 
τ1 0 0 ... 0 1
(3)
for every n ∈ N1 . Indeed, putting in (2) instead of n respectively n − 1, n − 2, . . . , 1
we get, together with (2), the system of n equations in n variables: σ1 , . . . , σn . Such
a system is a Cramer’s system and by the Cramer’s rule we get (3).
From now on by Dtan and Dcot we denote the domains of the trigonometric func-
tions tangent and cotangent, respectively.
Lemma 2. The following identities hold true:
Pm 2m
(A) sin 2mx 2j

j
cos2m x cot x = j=0 2j+1 (−1) tan x, (m, x) ∈ N × (Dtan ∩ Dcot );
Pm 2m

(B) cos 2mx
cos2m x =
j 2j
j=0 2j (−1) tan x, (m, x) ∈ N × Dtan ;
Remarkable identities 41

sin (2m+1)x Pm 2m+1

(−1)j tan2j x,


(C) cos2m+1 x cot x = j=0 2j+1 (m, x) ∈ N × (Dtan ∩ Dcot );

sin (2m+1)x Pm 2m+1

(−1)j cot2m−2j x,


(D) sin2m+1 x
= j=0 2j+1 (m, x) ∈ N × Dcot ;

cos (2m+1)x Pm 2m+1

(−1)j tan2j x,


(E) cos2m+1 x = j=0 2j (m, x) ∈ N × Dtan ;

cos (2m+1)x Pm 2m+1

(−1)j cot2m−2j x,


(F) sin2m+1 x
tan x = j=0 2j (m, x) ∈ N × (Dtan ∩
Dcot ).

Proof. It is a known fact that

k  
X k
cosk−j x(i sin x)j = (cos x + i sin x)k = cos kx + i sin kx
j=0
j

for k ∈ N and x ∈ R. Putting k = 2m in the above equation and comparing real and
imaginary parts of the both sides we obtain (A) and (B). Similarly, with k = 2m + 1
we get (C), (D), (E) and (F).

Now we prove the following result.

Theorem 1. For every m ∈ N2 and any n ∈ N1 ,

m−1 m−1
X πj X πj
σn,m (A) = tan2n = cot2n ,
j=1
2m j=1
2m

2m
(2j+1)
where σn,m (A) denotes the determinant given by (3) in which τj = 2m for j ∈
{1, 2, . . . , n}.

Proof. Replace in the identity (A) of Lemma 2, tan2 x by t and set

m  
X 2m
wA (t) = (−1)j tj , (4)
j=0
2j + 1

then wA (t) is a polynomial of order m − 1 in the real variable t.

πl
On the other hand, substituting 2m , where l ∈ {1, 2, . . . , m − 1}, for x in (A) we
get
m  
X 2m πl
0= (−1)j tan2j , l ∈ {1, 2, . . . , m − 1}.
j=0
2j + 1 2m

Hence and by (4) we obtain

  m−1
Y  m−1
Y 
m−1 2m 2 πj m−1 2 πj
wA (t) = (−1) t − tan = (−1) 2m t − tan .
2m − 1 j=1 2m j=1
2m
42 J. Górowski, J. Żabowski

This and the Vieta’s formulas give

2m


X
2πk1 πk2 πkj 2j+1
tan tan2 · · · tan2 =
2m 2m 2m 2m
1≤k1 <k2 <···<kj ≤m−1

and in view of (3) we have

m−1
X πj
σn,m (A) = tan2n .
j=1
2m

πj
As tan 2m = cot π(m−j)
2m for j ∈ {1, 2, . . . , m − 1} we get
m−1 m−1 m−1
πj 2n π(m − j) πj
X X X
2n
tan = cot = cot2n ,
j=1
2m j=1
2m j=1
2m

which completes the proof.

Theorem 2. For every m, n ∈ N1 the following identity holds true:
m−1 m−1
X 2j + 1 X 2j + 1
σn,m (B) = tan2n π= cot2n π,
j=0
4m j=0
4m

2m


where σn,m (B) denotes the determinant given by (3) in which τj = 2j for j ∈
{1, 2, . . . , n}.
Proof. Similarly as in the proof of Theorem 1, replace in the right hand side of the
identity (B) of Lemma 2, tan2 x by t and set
m  
X 2m
wB (t) = (−1)j tj .
j=0
2j

2l+1
Next, substitute 4m π, where l ∈ {0, 1, . . . , m − 1}, for x in (B). This yields
m  
X 2m 2l + 1
0= (−1)j tan2j π, l ∈ {0, 1, . . . , m − 1}.
j=0
2j 4m

Hence and by the definition of wB (t) we get

m−1
Y 
2j + 1
wB (t) = (−1)m t − tan2 π ,
j=0
4m

which in view of the Vieta’s formulas gives

 
X 2k1 + 1 2k2 + 1 2kj + 1 2m
tan2 tan2 · · · tan2 = .
4m 4m 4m 2j
1≤k1 <k2 <···<kj ≤m−1
Remarkable identities 43

By this and (3),

m−1
X 2j + 1
σn,m (B) = tan2n π.
j=0
4m
Using the same argument as in the proof of Theorem 1 we get
m−1 m−1
X 2j + 1 X 2j + 1
tan2n π= cot2n π
j=0
2m j=0
4m

and the proof is completed.

Using identities (C) and (D) of Lemma 2 and the same method as in proofs of
Theorems 1 and 2 one may obtain
Theorem 3. For every m, n ∈ N1 the following identity holds true:
m
X πj
σn,m (C) = tan2n ,
j=1
2m + 1

2m+1


where σn,m (C) denotes the determinant given by (3) in which τj = 2j for j ∈
{1, 2, . . . , n}.
Theorem 4. For every m, n ∈ N1 the following identity holds true:
m
X πj
σn,m (D) = cot2n ,
j=1
2m + 1

1 2m+1


where σn,m (D) denotes the determinant given by (3) in which τj = 2m+1 2j+1 for
j ∈ {1, 2, . . . , n}.
Finally, applying the same reasoning as in the proof of Theorem 1 from (E) and
(F) of Lemma 2 we have
Theorem 5. For every m, n ∈ N1 the following identity holds true:
m−1
X 2j + 1
σn,m (E) = tan2n π,
j=0
2(2m + 1)

1 2m+1


where σn,m (E) denotes the determinant given by (3) in which τj = 2m+1 2j+1 for
j ∈ {1, 2, . . . , n}.
Theorem 6. For every m, n ∈ N1 the following identity holds true:
m−1
X 2j + 1
σn,m (F ) = cot2n π,
j=0
2(2m + 1)

2m+1


where σn,m (F ) denotes the determinant given by (3) in which τj = 2j for j ∈
{1, 2, . . . , n}.
44 J. Górowski, J. Żabowski

The following formulas

n n
1 − sin2 x 1 − cos2 x
 
2n 2n
cot x = , tan x=
sin2 x cos2 x

yield

Lemma 3. The following identities hold true:

Pn−1 n
2j−2n
(G) j
j=0 j (−1) sin x = (−1)n−1 + cot2n x, (n, x) ∈ N1 × Dcot ;
Pn−1 n
(−1)j cos2j−2n x = (−1)n−1 + tan2n x,


(H) j=0 j (m, x) ∈ N1 × Dtan .

Lemma 4. Assume that n ∈ N1 and x ∈ Dcot , then

n n n n
(−1)n−1 + cot2n x (−1)n−1





− − ...

1 2 3 n−1
 (−1)n−2 + cot2n−2 x n−1 n−1 n−2 n−1




 1 − 1 2
... (−1) n−2


1 n−3
+ cot2n−4 x − n−2 n−2
(−1)n−3 n−3



= det  (−1) 0 1 ... .
 
1
2n
sin x  .. .. .. .. .. .. 
 
 . . . . . . 
1 + cot2 x 0 0 0 ... 1

Proof. Replacing n in (G) (Lemma 3) by n − 1, n − 2, . . . , 1, respectively we get,

together with (G), the system of n equations in n variables:

1 1 1
, ,..., .
sin2n x sin2n−2 x sin2 x
Such a system is a Cramer’s system and the assertion follows by the Cramer’s rule.

Using (H) in the same manner as in Lemma 4 we obtain

Lemma 5. Let n ∈ N1 and x ∈ Dtan , then
n n n n
(−1)n−1 + tan2n x (−1)n−1





− − ...

1 2 3 n−1
 (−1)n−2 + tan2n−2 x n−1 n−1 n−1
(−1)n−2 n−2




 1 − 1 2
... 

1 n−3
+ tan2n−4 x − n−2 n−2
(−1)n−3 n−3



= det  (−1) 0 1 ... .
 
1
2n
cos x 
 .. .. .. .. .. .. 

 . . . . . . 
1 + tan2 x 0 0 0 ... 1

To shorten notation from now on we set

n n n n
(−1)n−1





an − − ...

1 2 3 n−1
n−1 n−1 n−1
(−1)n−2 n−2




 a
 n−1 1 − 1 2
... 

− n−2 n−2
(−1)n−3 n−3



µ(an , an−1 , . . . , a1 ) = det  an−2 0 1 ... ,
 
1
 . .. .. .. .. .. 
 . 
 . . . . . . 
a1 0 0 0 ... 1
Remarkable identities 45

thus the identities of Lemmas 4 and 5 can be written as

1
= µ (−1)n−1 + cot2n x, (−1)n−2 + cot2n−2 x, . . . , 1 + cot2 x


2n (5)
sin x
and
1
= µ (−1)n−1 + tan2n x, (−1)n−2 + tan2n−2 x, . . . , 1 + tan2 x ,


2n
(6)
cos x
respectively.

Theorem 7. For every m ∈ N2 and each n ∈ N1 the following identity holds true:
m−1 m−1
X πj X πj
sin−2n = cos−2n (7)
j=1
2m j=1
2m
= µ (−1)n−1 (m − 1) + σn,m (A), (−1)n−2 (m − 1) + σn−1,m (A) ,
. . . , (m − 1) + σ1,m (A)) ,

where the numbers σk,m (A) for k ∈ {1, 2, . . . , n} are defined in Theorem 1.

Proof. In view of (5) we can write

 
πj πj πj πj
sin−2n = µ (−1)n−1 + cot2n , (−1)n−2 + cot2n−2 , . . . , 1 + cot2
2m 2m 2m 2m

for j ∈ {1, 2, . . . , m − 1}. This by the definition of µ, properties od determinants and

Theorem 1 gives
m−1
X πj
sin−2n
j=1
2m

m−1
X  m−1
X 
πj πj
= µ (−1)n−1 + cot2n , (−1)n−2 + cot2n−2 ,
j=1
2m j=1
2m

m−1
X 
πj
..., 1 + cot2 
j=1
2m

= µ (−1)n−1 (m − 1) + σn,m (A), (−1)n−2 (m − 1) + σn−1,m (A) ,

. . . , (m − 1) + σ1,m (A)) .

The same reasoning applies to the second identity.

Analysis similar to that in the proof of Theorem 7 and the use of Theorems 2 – 6
give
46 J. Górowski, J. Żabowski

Theorem 8. For every n, m ∈ N1 the following identities holds true:

m−1 m−1
X 2j + 1 X 2j + 1
sin−2n π= cos−2n π (8)
j=0
4m j=0
4m
µ (−1)n−1 m + σn,m (B), (−1)n−2 m + σn−1,m (B), . . . , m + σ1,m (B) ,


=

m
X πj
sin−2n (9)
j=1
2m + 1
= µ (−1)n−1 m + σn,m (D), (−1)n−2 m + σn−1,m (D), . . . , m + σ1,m (D) ,

m
X πj
cos−2n (10)
j=1
2m + 1
µ (−1)n−1 m + σn,m (C), (−1)n−2 m + σn−1,m (C), . . . , m + σ1,m (C) ,


=

m−1
X (2j + 1)π
sin−2n (11)
j=0
2(2m + 1)
= µ (−1)n−1 m + σn,m (F ), (−1)n−2 m + σn−1,m (F ), . . . , m + σ1,m (F ) ,

m−1
X (2j + 1)π
cos−2n (12)
j=0
2(2m + 1)
µ (−1)n−1 m + σn,m (E), (−1)n−2 m + σn−1,m (E), . . . , m + σ1,m (E) ,


=

where σk,m (B), σk,m (C), σk,m (D), σk,m (E), σk,m (F ) for k ∈ {1, 2, . . . , n} are defined
in Theorems 2 – 6.
Now we show that the general identities from Theorems 1 – 8 yield some particular
equalities, including the one considered by the authors as remarkable.
Theorem 9. If m ∈ N, then
m−1
X πj m2 − 1
sin−2 = , provided m ≥ 2, (13)
j=1
m 3

m−1
X πj (m − 1)(m − 2)
cot2 = , provided m ≥ 2, (14)
j=1
m 3
m−1
X (2j + 1)π
sin−2 = m2 , provided m ≥ 1, (15)
j=0
2m
Remarkable identities 47

Proof. According to Theorem 1 we have

m−1 m−1  
X πj X πj 1 2m
tan2 = cot2 = . (16)
j=1
2m j=1
2m 2m 3

On the other hand, in view of

4
tan2 x + cot2 x = −2
sin2 2x
we get
m−1 m−1 m−1
X πj X πj X πj
tan2n + cot2n =4 sin−2 − 2(m − 1).
j=1
2m j=1
2m j=1
m

Combining this with (16) gives

m−1
X πj m2 − 1
sin−2 =
j=1
m 3

for m ≥ 2. This proves (13).

To prove (14) notice that the identity
1
cot2 x − = −1
sin2 x
yields
m−1 m−1
X πj X πj
cot2 − sin−2 = −(m − 1), m ≥ 2.
j=1
m j=1
m

Thus by (13) we obtain (14).

Finally we show the remarkable (15). Theorem 8 leads to
m−1 m−1  
X (2j + 1)π X (2j + 1)π 2m
sin−2 = cos−2 =m+ = 2m2 (17)
j=0
4m j=0
4m 2

for m ≥ 1. Since
1 1 4
2 + =
2
sin x cos x sin2 2x
we have
m−1 m−1 m−1
X (2j + 1)π X (2j + 1)π X (2j + 1)π
sin−2 + cos−2 =4 sin−2 , m ≥ 1,
j=0
4m j=0
4m j=0
2m

which by (17) implies (15), and the theorem follows.

Next
P∞we use the the identities proved here to find the sums of the series of the
form k=1 k12n , where n ∈ N1 . We begin with the following lemma.
48 J. Górowski, J. Żabowski

Lemma 6. Let n ∈ N1 , then expression σn,m (A), defined in Theorem 1, is a value of

some polynomial from Q[x], where x = m. The order of such a polynomial does not
exceed 2n.
Proof. The proof is by induction on n. For n = 1 we have
 
1 2m 2 1
σ1,m (A) = = m2 − m − ,
2m 3 3 3
and the assertion follows. Fix n ≥ 2 Assuming Lemma 6 to hold for any k ∈ N1 ,
k ≤ n − 1 we prove it for n. By (2),
n−1
X
σn,m (A) = (−1)j−1 τj σn−j,m (A) − (−1)n nτn ,
j=1

1 2m


where τj = 2m 2j+1 for j ∈ {1, 2, . . . , n}. Hence by the inductive assumption
σn,m (A) is a value of some polynomial from Q[x] of order not greater than 2n with
x = m, as claimed.
Theorem 10. For every n ∈ N1 ,
∞
X 1 π 2n σn,m (A)
2n
= lim ,
j=1
j m→∞ (2m)2n

where σn,m (A) is defined in Theorem 1.

Proof. Observe that
1 1  π
0 < cot x < < x ∈ 0, ,
x sin x 2
thus  2n
2n πj 2m 1
cot < < πj
2m πj sin2n 2m
and in consequence
m−1  2n m−1 m−1
X πj 2m X 1 X 1
cot2n < < πj
j=1
2m π j=1
j 2n j=1
sin2n 2m

for n ∈ N1 , m ∈ N2 and j ∈ {1, 2, . . . , n}. By the definitions of σn,m (A) and the
function µ we have
m−1
π 2n σn,m (A) X 1 π 2n
< < µ((−1)n−1 (m − 1) + σn,m (A),
(2m)2n j=1
j 2n (2m) 2n

(−1)n−2 (m − 1) + σn−1,m (A), (18)

. . . , m − 1 + σ1,m (A)).
Remarkable identities 49

The formula for µ and the properties of determinants give

µ (−1)n−1 (m − 1) + σn,m (A), (−1)n−2 (m − 1) + σn−1,m (A), . . . , m − 1 + σ1,m (A)

= (m − 1)µ((−1)n−1 , (−1)n−2 , . . . , (−1)n−n ) + µ(σn,m (A), σn−1,m (A), . . . , σ1,m (A))

= (m − 1)C1 + σn,m (A) + C2 σn−1,m (A) + . . . + Cn σ1,m (A),
where C1 , . . . , Cn are constants depending on n. Hence by Lemma 6 and inequality
(18) we obtain
∞
π 2n σn,m (A) X 1 π 2n σn,m (A)
lim ≤ ≤ lim ,
m→∞ (2m)2n j=1
j 2n m→∞ (2m)2n

which establishes the formula.

Remark 1. Note that in the proof Theorem 10 (the last step of the proof) we have
actually proved more, namely that the order of the polynomial from Lemma 6 equals
exactly 2n. Indeed, if it was not true, we would have
π 2n σn,m (A)
lim =0
m→∞ (2m)2n
and consequently
∞
X 1
2n
≤ 0,
j=1
j
which is impossible.
Remark 2. Treating σn,m (A) as a polynomial in m of order 2n we have
π 2n σn,m (A) π 2n
lim 2n
= a2n n ,
m→∞ (2m) 4
where a2n denotes the leading coefficient of σn,m (A). On the other hand,
∞
22n−1 π 2n X 1
B2n (−1)n−1 = , n ∈ N1 ,
(2n)! j=1
j 2n

where B2n stands for the 2n-th Bernoulli number (see [4], p.320). Thus we get the
following relation between Bernoulli numbers and the coefficients of σn,m (A)
24n−1
a2n = B2n (−1)n−1 , n ∈ N1 .
(2n)!
Remark 3. Similarly as Theorem 10 one can show that
∞
X 1 π 2n σn,m (B) π 2n σn,m (D) π 2n σn,m (F )
= lim = lim = lim , n ∈ N1 .
j=1
j 2n m→∞ (2m)2n m→∞ (2m)2n m→∞ (2m)2n

where σn,m (B), σn,m (D) and σn,m (F ) are defined in Theorems 2, 4 and 6, respectively.
50 J. Górowski, J. Żabowski

References
[1] T. M. Apostol, Another elementary proof of Euler’s formula for ζ(2n), Amer.
Math. Monthly 80 (1973), No. 4, 425-431.

[2] P.M. Cohn, Algebra. Vol. 1, Second edition, John Wiley & Sons, Ltd., Chichester,
1982.
[3] D. Cvijović, J. Klinowski, Finite cotangent sums and the Riemann zeta function,
Math. Slovaca 50 (2000), no. 2, 149-157.

[4] R.L. Graham, D.E. Knuth, O. Patashnik, Concrete mathematics. A founda-

tion for computer science, Second edition, Addison-Wesley Publishing Company,
Reading, MA, 1994.
P∞ 2 Q∞  2

[5] R.A. Kortram, Simple proofs for k=1 k12 = π6 and sin x = x k=1 1 − kx2 π2 .
Math. Mag. 69 (1996), no. 2, 122-125.
P∞ 2
[6] I. Papadimitriou, A simple proof of the formula k=1 k12 = π6 , Amer. Math.
Monthly 80 (1973), No. 4, 424-425.
P∞ 2
[7] T.J. Ransford, An elementary proof of k=1 k12 = π6 , Eureka No. 42, Summer
1982, 3-5.
[8] H. Tsumura, An elementary proof of Euler’s formula for ζ(2m), Amer. Math.
Monthly 111 (2004), no. 5, 430-431.

DOI: 10.7862/rf.2014.4

Jan Górowski – Corresponding author

email: jangorowski@[Link]
Institute of Mathematics,
Pedagogical University,
ul. Podchora̧żych 2,
PL-30-084 Kraków, Poland
Jerzy Żabowski
email: [Link]@[Link]
Pawel Wlodkowic University College in Plock,
al. Kilińskiego 12,
PL-09-402 Plock, Poland
Received 01.10.2013, Accepted 12.01.2014