1. Given that tan θi ’s, where i = 1, 2, . . . , n, are n zeros of xn − a1 xn−1 + a2 xn−2 + . . . + (−1)n an = 0.

Prove that 
a1 −a3 +a5 −...+(−1)k a(2k+1)
 1−a2 +a4 −...−(−1)k−1 a(2k) , if n = 2k + 1
n
!  
X
tan θi =
k−1
 a1 −a3 +a5 −...+(−1) a(2k−1) , if n = 2k

i=1 
1−a2 +a4 −...−(−1)k−1 a2k

Pn
Solution We need to find tan ( i=1 θi ). Start with a polynomial x2 − a1 x + a2 = 0 with roots tan θ1
and tan θ2 :
tan θ1 + tan θ2 a1
tan(θ1 + θ2 ) = = ,
1 − tan θ1 tan θ2 1 − a2
and a polynomial x3 − a1 x2 + a2 x − a3 = 0 with roots tan θ1 , tan θ2 and tan θ3 :
tan θ1 +tan θ2
1−tan θ1 tan θ2 + tan θ3
tan(θ1 + θ2 + θ3 ) = tan θ1 +tan θ2
1 − 1−tan θ1 tan θ2 tan θ3
tan θ1 + tan θ2 + tan θ3 (1 − tan θ1 tan θ2 )
=
1 − tan θ1 tan θ2 − (tan θ1 + tan θ2 ) tan θ3
P3 Q3
i=1 tan θi − i=1 tan θi
= P
1 − 1≤i<j≤3 tan θi tan θj
a1 − a3
=
1 − a2

Without loss of generality, we denote the coefficients of the polynomial with degree v as:

(uv )
 
X u
Y
Au (v) =  tan θt  , (1)
1≤t1 <...<tu ≤v 1≤t≤v

where u ≤ v.

For those not familiar with combinatorics, we provide a simple example for the explanation of how to use
the notation as follows:

If we need to get the sum of three pairs of products among T1 , T2 and T3 , then the summation
would be
(32) (32)
 
X X Y2
T1 T2 + T1 T3 + T2 T3 = Ti1 Ti2 =  Tit  .
1≤i1 <i2 ≤3 1≤i1 <i2 ≤3 1≤it ≤3
Q P
The criteria of choosing terms are listed under the notations and , and the number of
items is listed above the notations. Some books may omit the number of items on top of the
notation.

Email: tedwong314@[Link]
Notice that for u + 1 ≤ v − 1, we have
Au+1 (v − 1) + tan θv Au (v − 1)

= (tan θ1 tan θ2 . . . tan θu tan θu+1 ) + . . . + (tan θ1 tan θ2 . . . tan θu tan θv−1 ) + . . .
| {z } | {z }
u+1 u+1
+(tan θ1 tan θv−u . . . tan θv−2 tan θv−1 ) + . . . + (tan θv−u−1 tan θv−u . . . tan θv−2 tan θv−1 )
| {z } | {z }
h u+1 u+1
+ tan θv (tan θ1 tan θ2 . . . tan θu−1 tan θu ) + . . . + (tan θ1 tan θ2 . . . tan θu−1 tan θv−1 ) + . . .
| {z } | {z } (2)
u u i
+(tan θ1 tan θv−u+1 . . . tan θv−2 tan θv−1 ) + . . . + (tan θv−u tan θv−u+1 . . . tan θv−2 tan θv−1 )
| {z } | {z }
u u

v
P(u+1)
Q 
u+1
= 1≤t1 <...<tu+1 ≤v 1≤t≤v tan θt

= Au+1 (v),
and also for u + 1 = v:
tan θu+1 Au (v − 1) = Au+1 (v) = Av (v) (3)
We have seen the cases when n = 2, 3. By using mathematical induction, we assume that it is true when
n = m. that is


A1 (2k+1)−A3 (2k+1)+A5 (2k+1)−...+(−1)k−1 A(2k−1) (2k+1)+(−1)k A(2k+1) (2k+1)
m
! 

 1−A2 (2k+1)+A4 (2k+1)−...−(−1)k−2 A(2k−2) (2k+1)−(−1)k−1 A(2k) (2k+1)
, if m = 2k + 1
X
tan θi = .
i=1
 A1 (2k)−A3 (2k)+A5 (2k)−...+(−1)k−2 A(2k−3) (2k)+(−1)k−1 A(2k−1) (2k)
, if m = 2k


1−A2 (2k)+A4 (2k)−...−(−1)k−2 A(2k−2) (2k)−(−1)k−1 A(2k) (2k)

There are two cases that m is even or odd.

1. m = 2k + 1
For n = m + 1, i.e. n = 2k + 2, we have
Pm
tan (θm+1 + i=1 θi )

A1 (2k+1)−A3 (2k+1)+A5 (2k+1)−...+(−1)k−1 A(2k−1) (2k+1)+(−1)k A(2k+1) (2k+1)

tan θ2k+2 +
1−A2 (2k+1)+A4 (2k+1)−...−(−1)k−2 A(2k−2) (2k+1)−(−1)k−1 A(2k) (2k+1)
= A1 (2k+1)−A3 (2k+1)+A5 (2k+1)−...+(−1)k−1 A(2k−1) (2k+1)+(−1)k A(2k+1) (2k+1)
1−tan θ2k+2
1−A2 (2k+1)+A4 (2k+1)−...−(−1)k−2 A(2k−2) (2k+1)−(−1)k−1 A(2k) (2k+1)

tan θ2k+2 [1−A2 (2k+1)+A4 (2k+1)−...−(−1)k−2 A(2k−2) (2k+1)−(−1)k−1 A(2k) (2k+1)]

= 1−A2 (2k+1)+A4 (2k+1)−...−(−1)k−2 A(2k−2) (2k+1)−(−1)k−1 A(2k) (2k+1)
to be continued
+A1 (2k+1)−A3 (2k+1)+A5 (2k+1)−...+(−1)k−1 A(2k−1) (2k+1)+(−1)k A(2k+1) (2k+1)
− tan θ2k+2 [A1 (2k+1)−A3 (2k+1)+A5 (2k+1)−...+(−1)k−1 A(2k−1) (2k+1)+(−1)k A(2k+1) (2k+1)]

A1 (2k+2)−A3 (2k+2)+A5 (2k+2)−...+(−1)k A(2k+1) (2k+2)

= 1−A2 (2k+2)+A4 (2k+2)−...−(−1)k−1 A(2k) (2k+2)−(−1)k A(2k+2) (2k+2)

Email: tedwong314@[Link]
 This indicates that the identity is true when m is odd. Notice that we use the result 2 except the
last term in the denominator in which we use 3.
2. m = 2k
For n = m + 1, i.e. n = 2k + 1. then
Pm
tan (θm+1 + i=1 θi )

A1 (2k)−A3 (2k)+A5 (2k)−...+(−1)k−2 A(2k−3) (2k)+(−1)k−1 A(2k−1) (2k)

tan θ2k+1 +
1−A2 (2k)+A4 (2k)−...−(−1)k−2 A(2k−2) (2k)−(−1)k−1 A(2k) (2k)
= A1 (2k)−A3 (2k)+A5 (2k)−...+(−1)k−2 A(2k−3) (2k)+(−1)k−1 A(2k−1) (2k)
1−tan θ2k+1
1−A2 (2k)+A4 (2k)−...−(−1)k−2 A(2k−2) (2k)−(−1)k−1 A(2k) (2k)

tan θ2k+1 [1−A2 (2k)+A4 (2k)−...−(−1)k−2 A(2k−2) (2k)−(−1)k−1 A(2k) (2k)]

= 1−A2 (2k)+A4 (2k)−...−(−1)k−2 A(2k−2) (2k)−(−1)k−1 A(2k) (2k)
to be continued
+A1 (2k)−A3 (2k)+A5 (2k)−...+(−1)k−2 A(2k−3) (2k)+(−1)k−1 A(2k−1) (2k)
− tan θ2k+1 [A1 (2k)−A3 (2k)+A5 (2k)−...+(−1)k−2 A(2k−3) (2k)+(−1)k−1 A(2k−1) (2k)]

A1 (2k+1)−A3 (2k+1)+A5 (2k+1)−...+(−1)k−2 A(2k−1) (2k+1)+(−1)k−1 A(2k+1) (2k+1)

= 1−A2 (2k+1)+A4 (2k+1)−...−(−1)k−2 A(2k−2) (2k+1)−(−1)k−1 A(2k) (2k+1)

One may compare this to the identity and found it matching exactly. It is indeed this proves the
identity as the degree increases by one and becomes the case of odd m.

Email: tedwong314@[Link]