On an Algebraic Identity

September 19, 2008

Roberto Bosch Cabrera

Abstract
We solve problem U23 from Mathematical Reflections (which remained open
for a while.) We use a simple algebraic identity concerning polynomials and
their derivatives, demonstrating its usefulness in solving problems.

U23. Evaluate the sum

n−1
X 1
k=0
1 + 8 sin2 ( kπ
n )

Dorin Andrica and Mihai Piticari

Solution:
It is well known that
eix − e−ix
sin x =
2i
so  
kπ  2kπ 2kπ

1 + 8 sin 2
= 5 − 2 ei n + e−i n .
n
Let
2kπ
ei n = ξk
then we need to find the following sum:
n−1
X ξk
.
−2ξk2 + 5ξk − 2
k=0

But we know the following

x −x 2 1 1 1
= = · − ·
−2x2 + 5x − 2 (2 − x)(1 − 2x) 3 2 − x 3 1 − 2x

Mathematical Reflections 5 (2008) 1

so it follows that
n−1 n−1 n−1
X ξk 2 X 1 1 X 1
= · − · .
−2ξk2 + 5ξk − 2 3 2 − ξk 6 1
− ξk
k=0 k=0 k=0 2

Observe that ξk , k = 0, ..., n − 1 are the n roots of the polynomial p(x) = xn − 1.

So we have to solve the following subproblem:

Let ak (k = 1, ..., n) be the roots of the polynomial p(x) of degree n. Find the
sum:
n
X 1
x − ak
k=1
This question is well known and it was the motivation to write this article. The
answer is quite simple
n
X 1 p0 (x)
= . (1)
x − ak p(x)
k=1

We continue with the proof to the above statement. Because p(x) = (x − a1 ) · · · (x −

an ) then the product rule of differentiation yields
p0 (x) = (x − a2 ) · · · (x − an ) + · · · + (x − a1 ) · · · (x − an−1 )
and we are done.

Now we turn to the solution of U23:
0 n−1
In this case p(x) = xn − 1 so pp(x)
(x)
= nx
xn −1 . We know the following equalities
are true
n−1
X 1 p0 (2) n2n−1
= = n
2 − ξk p(2) 2 −1
k=0
and
n−1 n
X 1 p0 ( 12 ) 2n−1 2n
1 = = = .
k=0 2
− ξk p( 12 ) 1
2n − 1
1 − 2n
Finally
n−1
X 1 2 n2n−1 1 2n n 2n + 1
= · − · = · .
k=0
1 + 8 sin2 ( kπ
n )
3 2n − 1 6 1 − 2n 3 2n − 1

We now discuss some problems and leave others for the reader to solve.

Mathematical Reflections 5 (2008) 2

1. Let p(x) be a non-constant polynomial with real roots. Prove that
p0 (x)2 ≥ p(x)p00 (x) for all x ∈ R.

(Martin Aigner, Gűnter M. Ziegler, Proofs from The Book, Third edition)

Solution:
If x = ai is a root of p(x), then there is nothing to show. Assume that x is not a
root, then
n
p0 (x) X 1
= .
p(x) x − ak
k=1
Differentiating this expression we obtain
n
p00 (x)p(x) − p0 (x)2 X 1
2
=− <0
p(x) (x − ak )2
k=1

and this concludes the proof.


2. Let P be an algebraic polynomial of degree n having only real zeros and real
coefficients.
a) Prove that for every real x the following inequality holds:

(n − 1)P 0 (x)2 ≥ nP (x)P 00 (x) (2)

b) Examine the cases of equality.

(IMC 1998)

Solution:
Observe that both sides of (2) are equal to zero if n = 1. Suppose that n > 1 and
let x1 , ..., xn be the zeros of P . Clearly (2) is true when x = xi , i ∈ {1, ..., n}, and
equality is possible only if P 0 (xi ) = 0, i.e if xi is a multiple zero of P . Now suppose
that x is not a zero of P . Using the identities
n
P 0 (x) X 1 P 00 (x) X 2
= , = .
P (x) x − xi P (x) (x − xi )(x − xj )
i=1 1≤i<j≤n

we find
2 n
P 0 (x) P 00 (x) X n − 1
 X 2
(n − 1) −n = − .
P (x) P (x) (x − xi )2 (x − xi )(x − xj )
i=1 1≤i<j≤n

Mathematical Reflections 5 (2008) 3

But this last expression is simply
 2
X 1 1
−
x − xi x − xj
1≤i<j≤n

and therefore is positive. The inequality is proved. In order for (2) to hold with
equality sign for every real x it is necessary that x1 = x2 = . . . = xn . A direct
verification shows that indeed, if P (x) = c(x − x1 )n , then (2) becomes an identity.

3. Suppose that every zero of the polynomial f (x) is simple. Find the sum of the
reciprocals of the differences between the roots of the equation f 0 (x) = 0 and the
roots of the equation f (x) = 0.

(E.J.Barbeau. Polynomials.)

Solution:
Let ri be the zeros of f (x) and let sj be the zeros of f 0 (x). Then, since
X
f 0 (x) = f (x) (x − ri )−1 ,

for each j we have: X

0 = f 0 (sj ) = f (sj )
(sj − ri )−1 .

Because f (sj ) 6= 0 then we can conclude that (sj − ri )−1 = 0.

P

4. Determine those values of the real number a and positive integer n exceeding 1
for wich
n
X xk + 2
=n−3
xk − 1
k=1

where x1 , ..., xn are the zeros of xn + axn−1 + an−1 x + 1.

(E.J.Barbeau. Polynomials.)

Solution: Let p(x) = xn + axn−1 + an−1 x + 1 then p0 (x) = nxn−1 + a(n − 1)xn−2 +
an−1 . We have
xk + 2 3
=1+
xk − 1 xk − 1
therefore
n n
X xk + 2 X 1 p0 (1)
=n−3 =n−3
xk − 1 1 − xk p(1)
k=1 k=1

Mathematical Reflections 5 (2008) 4

and we have
n + an − a + an−1
n−3· =n−3
an−1 + a + 2
or equivalently (a + 1)(n − 2) = 0, so a = −1 or n = 2. The case a = −1 yields the
polynomial xn − xn−1 − x + 1 = (x − 1)(xn−1 − 1). But in this case, one of the zeros
is 1 and the left side of the given equation is undefined. Hence a 6= −1. The case
n = 2 yields the polynomial x2 + +2ax + 1, whose zeros can be verified to satisfy
the condition, provided a 6= −1.

5. Let −1 < x < 1. Show that
6
X 1 − x2 1 + x7
= 7 · (3)
k=0
1 − 2x cos( 2kπ
7 )+x
2 1 − x7

(Longlist IMO 1988)

Solution:
We will use the follwing fact
2kπ 2kπ
ei + e−i
 
2kπ 7 7
cos = .
7 2
Let
2kπ
ei 7 = ξk
so
1 − x2 2(1 − x2 )ξk
= .
1 − 2x cos( 2kπ
7 )+x
2 −2xξk2 + (2 + 2x2 )ξk − 2x
The discriminant
√ of the quadratic −2xξk2 + (2 + 2x2 )ξk − 2x is D = 4(1 − x2 )2 .
Therefore D = 2 1 − x2 = 2(1 − x2 ) and the zeros are x and x1 . We have
−2xξk2 + (2 + 2x2 )ξk − 2x = −2x(ξk − x)(ξk − x1 ).
Finally,

1 − x2 (x − x1 )ξk 1
x x
= = −
1 − 2x cos( 2kπ
7 )+x
2 (ξk − x)(ξk − x1 ) 1
x − ξk x − ξk

and
6 6 6
X 1 − x2 1X 1 X 1 1 p0 ( 1 ) p0 (x) 1 + x7
= 1 −x = · 1x −x· = 7·
k=0
1 − 2x cos( 2kπ
7 )+x
2 x
k=0 x − ξk k=0
x − ξk x p( x ) p(x) 1 − x7

where p(x) = x7 − 1.

Mathematical Reflections 5 (2008) 5

 

6. Let x1 , x2 , ..., xn be real numbers that satisfy 0 < x1 < x2 < · · · < xn < 1 and
let x0 = 0, xn+1 = 1. Given:
n+1
X 1
=0 i = 1, 2, ..., n
xi − xj
j=0,j6=i

prove that xn+1−i = 1 − xi for i = 1, 2, ..., n.

(Shortlist IMO 1986)

Solution:
Let P (x) = (x − x0 )(x − x1 ) · · · (x − xn )(x − xn+1 ). Then
n+1
0
X P (x)
P (x) =
x − xj
j=0
n+1
XX P (x)
P 00 (x) = .
(x − xj )(x − xk )
j=0 k6=j

Therefore X 1
P 00 (xi ) = 2P 0 (xi )
xi − xj
j6=i

for i = 0, 1, ..., n + 1 and the given condition implies P 00 (xi ) = 0 for i = 1, 2, ..., n.
Consequently
x(x − 1)P 00 (x) = (n + 2)(n + 1)P (x). (4)
It is easy to observe that there is a unique monic polynomial of degree n+2 satisfying
the differential equation (4). On the other hand, the polynomial Q(x) = (−1)n P (1−
x) also satisfies this equation, is monic, and deg Q = n + 2. Therefore (−1)n P (1 −
x) = P (x), and the result follows.

7. Let p(z) be a polynomial of degree n with complex coefficients. Its roots (in the
complex plane) can be covered by a disk of radius r. Show that for any complex
number k, the roots of the polynomial np(z) − kp0 (z) can be covered by a disk of
radius r + |k|.

(Putnam 1957)

Mathematical Reflections 5 (2008) 6

Solution:
Let the roots of p(z) be a1 , a2 , ..., an . Suppose they all lie in a disk with center c
and radius r, then |c − an | ≤ r. Suppose that |c − w| > r + |k|. We will show that w
is not a root of np(z) − kp0 (z). Indeed, we have that |w − ai | ≥ |w − c| − |c − ai | >
0 (z)
r + |k| − r = |k|, thus pp(z)
P 1
= (z−ai ) (note that this is still true if we have
0 0 0
repeated roots), so | pp(w)
(w) n
| < |k| and hence |k pp(w)
(w)
| < n. Thus |n − k pp(w)
(w)
| > 0,
but |p(w)| > 0 (since w lies outside the disk containing all the roots of p(z)), so
0 (w)
|np(w) − kp0 (w)| = |p(w)||n − k pp(w) | > 0.

8. Let p(z) be a polynomial of degree n, all of whose zeros have absolute value 1
in the complex plane and let g(z) = p(z) 0
n . Show that all zeros of g (z) = 0 have
z2
absolute value 1.

(Putnam 2005)

Solution:
n g 0 (z) p0 (z) n
We have z 2 g(z) = p(z), therefore g(z) = p(z) − 2z . Moreover

n n
g 0 (z) X
 
1 1 1 X z + rj
= − =
g(z) z − rj 2z 2z z − rj
j=1 j=1

where r1 , ..., rn are the zeros of p(z). Now if z 6= rj for all j, then

z + rj (z + rj )(z − rj ) |z|2 − 1 + 2Im(zrj )i

= =
z − rj |z − rj |2 |z − rj |2

and so  
n
zg 0 (z) |z|2
 
−1  X 1
Re = · .
g(z) 2 |z − rj |2
j=1

Since the quantity in parentheses is positive g 0 (z) can be 0 only if |z| = 1. If on the
other hand z = rj for some j, then |z| = 1 anyway.

9. For any polynomial g, denote by d(g) the minimun distance of any two of its real
zeros (d(g) = ∞ if g has at most one real zero). Assume that g and g + g 0 both are
of degree k ≥ 2 and have k distinct real zeros. Then d(g + g 0 ) ≥ d(g).

(IMC 2007)

Mathematical Reflections 5 (2008) 7

Solution:
Let x1 < x2 < · · · < xk be the roots of g. Suppose that a and b are roots of g + g 0
satisfying 0 < b − a < d(g). Then a and b cannot be roots of g and

g 0 (a) g 0 (b)
= = −1. (5)
g(a) g(b)
0
Since gg is strictly decreasing between consecutive zeros of g (see problem 1.), we
must have a < xj < b for some j. For all i = 1, 2, ..., k − 1 we have xi+1 − xi > b − a
hence a − xi > b − xi+1 . If i < j both sides of this inequality are negative, if i ≥ j
1
both sides are positive. In any case a−x i
< b−x1i+1 and hence:

k−1 k−1
g 0 (a) X 1 1 X 1 1 g 0 (b)
= + < + =
g(a) a − xi a − xk b − xi+1 b − x1 g(b)
i=1 | {z } i=1 | {z }
<0 >0

This contradicts (5).

Mathematical Reflections 5 (2008) 8

 Problems for independent study

1. From problem 5. deduce that: csc2 ( π7 ) + csc2 ( 2π 2 3π

7 ) + csc ( 7 ) = 8.

(Longlist IMO 1988)

2. Let the roots of an nth degree polynomial P (z), with complex coefficients, lie on
the unit circle in the complex plane. Prove that the roots of the polynomial

2zP 0 (z) − nP (z)

lie on the same circle.

(IMC 1995)

3. If all zeros of a polynomial P (z) lie in the same half plane, then all zeros of the
derivative P 0 (z) lie in the same half plane.

(In a sharper formulation the problem tells us that the smallest convex polygon
that contains the zeros of P (z) also contains the zeros of P 0 (z)).

(Gauss-Lucas Theorem)

Acknowledgements
I would like to thank Mario Garcia Armas.

Mathematical Reflections 5 (2008) 9