Bucharest, March 4

th
, 2011
SOUTH EASTERN EUROPEAN MATHEMATICAL
OLYMPIAD FOR UNIVERSITY STUDENTS
Bucharest, March 4
th
, 2011
Solutions and grading schemes
Problem 1. For a given integer n 1, let f : [0, 1] R be a non-
decreasing function. Prove that
1
_
0
f(x) dx (n + 1)
1
_
0
x
n
f(x)dx.
Find all non-decreasing continuous functions for which equality holds.
Solution. For given n and any x, y [0, 1] we integrate on the interval
[0, 1] in terms of x and in terms of y the obvious inequality
(x
n
y
n
)(f(x) f(y)) 0, (1)
obtaining
_
1
0
__
1
0
(x
n
y
n
)(f(x) f(y)) dx
_
dy 0,
or
_
1
0
x
n
f(x) dx
_
1
0
y
n
dy
_
1
0
f(x) dx
_
1
0
x
n
dx
_
1
0
f(y) dy+
_
1
0
y
n
f(y) dy 0
which gives the desired inequality. As f is continuous, when equality holds
it forces equality in (1) also, that is f must be constant.
Grading scheme.
(n + 1)
1
_
0
x
n
f (x) dx =
1
_
0
f
_
n+1

t
_
dt . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3p
f non-decreasing implies f (x) f (
n+1

x) for x [0, 1] . . . . . . . . . . . . . 3p
1
1
_
0
f (x) dx
1
_
0
f (
n+1

x) dx = (n + 1)
_
1
0
x
n
f (x) dx . . . . . . . . . . . . . . . . . . 1p
Equality holds only for constant functions. . . . . . . . . . . . . . . . . . . . . . . . . . . 3p
Alternative solution.
(x
n
y
n
) (f (x) f (y)) 0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3p
Integrate the above inequality on [0, 1] [0, 1] to obtain
0
1
_
0
1
_
0
((x
n
y
n
) (f (x) f (y))) dxdy =
1
_
0
x
n
f (x) dx
1
_
0
x
n
dx
1
_
0
f (y) dy
1
_
0
y
n
dy
1
_
0
f (x) dx +
1
_
0
y
n
f (y) dy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3p
Using
1
_
0
x
n
dx =
1
n + 1
it obtains the inequality. . . . . . . . . . . . . . . . . . . . . . 1p
Equality holds only for constants function. . . . . . . . . . . . . . . . . . . . . . . . . . . 3p
Remark 1 The use of the Chebyshev inequality (without proof ) . . . . . . . . 7p
Problem 2. Let A = (a
ij
) be a real n n matrix such that A
n
,= 0 and
a
ij
a
ji
0, for all i, j. Prove that there exist two nonreal numbers among
eigenvalues of A.
Solution. Let
1
, . . . ,
n
be all eigenvalues of this matrix. From condition
it follows that a
ii
= 0, therefore
n

k=1

k
= 0. The sum

i<j

j
equals the sum
of principal minors with size 2 for the matrix A, i. e. it equals

i<j
a
ij
a
ji
.
Thus the sum
n

k=1

2
k
= 2

i<j
a
ij
a
ji
0. If a number z is an eigenvalue of A
then so does z.
Grading scheme. a
ii
= 0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1p

i
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1p

i<j

j
0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3p
n

i=1

2
i
0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2p
A
n
,= 0 implies
i
,= 0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1p

2
i
< 0 implies
i
CR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1p

i
CR implies
j
=
i
,=
i
CR . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1p
Alternative solution. Tr(A
2
) 0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4p
Tr(A
2
) =

2
i
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1p

2
i
0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2p
2
A
n
,= 0 implies
i
,= 0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1p

2
i
< 0 implies
i
CR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1p

i
CR implies
j
=
i
,=
i
CR . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1p
Problem 3. Given vectors a, b, c R
n
, show that
_
[[a[[

b, c
__
2
+
_
[[b[[ a, c
_
2
[[a[[[[b[[
_
[[a[[[[b[[ +[

a, b
_
[
_
[[c[[
2
,
where x, y denotes the scalar (inner) product of the vectors x and y and
[[x[[
2
= x, x.
Solution (Trigonometric). We shall rst prove that, given vectors
u, v R
n
, with [[u[[ = [[v[[ = 1 (thus u, v S
n1
), then
sup
||x||=1
_
u, x
2
+v, x
2
_
= 1 +[ u, v [.
Consider the unique vectorial representation x = x

+x

with x

span(u, v)
and x

span(u, v). Then 1 = [[x[[

+ x

, x

+ x

= [[x

[[
2
+
2 x

, x

+[[x

[[
2
= [[x

[[
2
+[[x

[[
2
(Pythagoras relation), so [[x

[[ 1. Now,
[ u, x [ = [ u, x

[ [ u, y [ where y = 0 if x

= 0 and y =
x

[[x

[[
if x

,= 0
(thus [[y[[ = 1). Similarly [ v, x [ = [ v, x

[ [ v, y [. The maximum
asked is therefore obtained when x span(u, v).
Thus our vectorial problem has been transferred in the 2-dimensional
space, with unit vectors u, v and x. The endpoints of the vectors u and v
partition the circle S
1
into four arcs, each of measure at most (and possibly
0, when u = v); the endpoint of x will fall into one of them. Let be the
measure of that arc, and , , with + = , the measures of the arcs
between the endpoint of x and the ends of that arc. Then u, x
2
+v, x
2
=
cos
2
+cos
2
, by the well-known geometric interpretation of the dot-product
(independently of the dimension of the space). But then u, x
2
+ v, x
2
=
cos
2
+ cos
2
= 1 +
1
2
(cos 2 + cos 2) = 1 + cos( + ) cos( )
1 +[ cos [ = 1 +[ u, v [, with equality in the obvious cases
when = = /2, therefore when [ u, x [ = [ v, x [, therefore x =
(u v)/[[ u v[[, where the signs are such that 0 < /2;
when = /2, therefore when u, v = 0, i.e. u v, for all x
span(u, v).
Then, for any not-null a, b, c, let us take u =
a
[[a[[
, v =
b
[[b[[
, x =
c
[[c[[
,
therefore u, x
2
=
1
[[a[[ [[c[[
2
a, c
2
, v, x
2
=
1
[[b[[ [[c[[
2
b, c
2
, u, v =
3
a, b
[[a[[ [[b[[
, so the proven relation becomes
1
[[a[[
a, c
2
+
1
[[b[[
b, c
2

_
1 +
[ a, b [
[[a[[ [[b[[
_
[[c[[
2
,
equivalent with the required inequality, also true for a, b or c equal to zero.
Remarks. We can continue by AM-GM to obtain
[ u, x v, x [
1
2
_
u, x
2
+v, x
2
_

1
2
(1 +[ u, v [) .
A simple computation now yields
[ a, c b, c [
1
2
([[a[[ [[b[[ +[ a, b [)[[c[[
2
,
also true for a, b or c equal to zero, a generalization of the Cauchy-Schwarz
inequality, since for a = b it precisely yields it.
Solution (Quadratic Forms). For [[x[[ = [[u[[ = [[v[[ = 1 we have
0 [[x + u + v[[
2
= x + u + v, x + u + v =
=
2
+
2
+
2
+ 2x, u + 2 x, v + 2 u, v ,
a quadratic form which takes non-negative values for any real parameters
, , , thus corresponding to a positive semidened matrix
_
_
1 x, u x, v
x, u 1 u, v
x, v u, v 1
_
_
The principal minors of order 1 are thus non-negative, which yields the semi-
positivity of the norm; the principal minors of order 2 are non-negative, which
yields the C-B-S inequality, e.g. 1 u, v
2
; nally, also the determinant of
the matrix is non-negative
= 1 (u, v
2
+x, u
2
+x, v
2
) + 2 u, v x, u x, v 0,
which can be arranged as x, u
2
+x, v
2
1+[ u, v [[ u, v [(1+[ u, v [
2[ x, u x, v [). But x, u
2
+ x, v
2
2[ x, u x, v [, which plugged into
the relation yields that (1 [ u, v [)(1 + [ u, v [ 2[ x, u x, v [) 0.
Then either 1 = [ u, v [, when 1+[ u, v [ = 2 2[ x, u x, v [ (by C-B-S),
or 1 > [ u, v [, which implies 1 + [ u, v [ 2[ x, u x, v [. Thus always
x, u
2
+x, v
2
1 +[ u, v [.
4
Solution (Lagrange Multipliers). Dene
L(x, ) = u, x
2
+v, x
2
([[x[[
2
1)
and consider the system
L
x
i
= 2u
i
u, x + 2v
i
v, x 2x
i
= 0,
for 1 i n, and
L

= [[x[[
2
1 = 0. Now
0 =
1
2
n

i=1
x
i
L
x
i
= u, x
n

i=1
u
i
x
i
+ v, x
n

i=1
v
i
x
i

n

i=1
x
2
i
= u, x
2
+
v, x
2
[[x[[
2
= u, x
2
+v, x
2
, so needs be, computed at the critical
point(s), the very expression we are considering (of no relevance, in fact). On
the other hand, 0 =
1
2
n

i=1
u
i
L
x
i
= u, x
n

i=1
u
2
i
+v, x
n

i=1
v
i
u
i

i=1
x
i
u
i
=
u, x [[u[[
2
+v, x u, v u, x = u, x +u, v v, x u, x, and simi-
larly for v, thus reaching the system of two equations (in the variables u, x
and v, x)
_
(1 ) u, x +u, v v, x = 0
u, v u, x + (1 ) v, x = 0
The determinant of the matrix of this particular system is = (1)
2

u, v
2
, and, if not null, the only solution is the trivial u, x = v, x = 0,
when our expression reaches an evident minimum, equal to zero (therefore
since then = 0, this means u, v ,= 1, which translates in u ,= v, and
x span(u, v)). We are therefore interested in = 0 (for all other critical
points), leading to = 1 u, v, thus = 1 +[ u, v [ at a maximum point,
and = 1 [ u, v [ at a critical point. There are some particular cases.
When u = v, thus u, v = 1, the situation is simple. We have a
maximum of 2 when x = u, and a minimum of 0 when x u.
When u v, thus u, v = 0, then (thus the expression) is 1 at the
maximum points, for all x span(u, v), and (thus the expression) is 0 at
the minimum points, for all x span(u, v).
Grading scheme. Any solution. |a| = |b| = |c| = 1 to get a, c
2
+
b, c
2
1 +[a, b[ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2p
Any solution. Case of some norm is 0 forgotten . . . . . . . . . . . . . . . . . 0p
Any solution. only case of norm 0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1p
Solution 1. (trigonometry)
5
Replacing , with cos to get cos
2
+ cos
2
1 +[ cos [ . . . . . . . . . . 1p
Replacing [ cos [ with cos
2
. . . . . . . . . . . . . . . . . . . . +0p, wrong inequality
Some similar cases are lost . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . upto 2p 10p
Solution 2. (coordinates)
Consider c = (1, 0, . . . , 0), a = (a
1
, a
2
, 0, . . . , 0), b = (b
1
, b
2
, . . . , b
n
) to get
a
2
1
+ b
2
1
1 +[a
1
b
1
+ a
2
b
2
[ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3p
Moving and adjusting variables 0 y = b
2
x = a
1
1 to get x
2

y
2
+ x
_
1 y
2
y

1 x
2
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5p
Solution 3. (Sylvester criteria)
Consider inequality |a + b + c|
2
0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3p
Witing down Sylvester criteria for form (, , ) . . . . . . . . . . . . . . . . . . . 5p
Problem 4. Let f : [0, 1] R be a twice continuously dierentiable
increasing function. Dene the sequences given by L
n
=
1
n
n1

k=0
f
_
k
n
_
and
U
n
=
1
n
n

k=1
f
_
k
n
_
, n 1. The interval [L
n
, U
n
] is divided into three equal
segments. Prove that, for large enough n, the number I =
1
_
0
f(x) dx belongs
to the middle one of these three segments.
Solution. Let us nd a suitable representation for L
n
.
Lemma 2 For f C
2
[0, 1]:
L
n
= I
f(1) f(0)
2n
+ O
_
1
n
2
_
, n .
Proof. Denote C = f(1) f(0). Consider
I L
n
=
n1

k=0
_ k+1
n
k
n
_
f(x) f
_
k
n
_
_
dx =
=
n1

k=0
_ k+1
n
k
n
_
f

_
k
n
__
x
k
n
_
+
1
2
f

(
x
kn
)
_
x
k
n
_
2
_
dx =
=
1
2n
2
n1

k=0
f

_
k
n
_
+ r
n
, (1)
where the remainder
[r
n
[
1
2
max[f

[
n1

k=0
1
3
_
x
k
n
_
3

k+1
n
k
n

const
n
2
.
6
Here
x
kn

_
k
n
,
k+1
n

are intermediate points from Taylors theorem in the

dx
1
n
n1

k=0
f

_
k
n
_
=
n1

k=0
_ k+1
n
k
n
f

x
kn
)
_
x
k
n
_
dx = O
_
1
n
_
.
So in the right hand side of (1) we have
1
n
n1

k=0
f

_
k
n
_

_
1
0
f

dx = C, n ,
and the error in the right hand side of (1) is of order O
_
1
n
2
_
.
Thus from (1) we have
I L
n
=
C
2n
+ O
_
1
n
2
_
, n .
Now we return to solution of our problem. We have U
n
= L
n
+
C
n
. Put
x
n
= L
n
+
kC
3n
, k = 1, 2. Then
x
n
= I +
C
n
_
k
3

1
2
_
+ O
_
1
n
2
_
.
For k = 1 we have
k
3

1
2
< 0, so x
n
< I for large enough n; for k = 2 we
have
k
3

1
2
> 0, so x
n
> I for large enough n. Thus for large enough n
L
n
+
C
3n
< I < L
n
+
2C
3N
.
Grading scheme. For stating the problem as a double inequality . . . . . 1p
For stating the idea of approaching the problem that might lead to a
solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1p
For establing the estimate I = L
n
+
f(1) f(0)
2n
+ O
_
1
n
_
. . . . . . . . . . 7p
Complete answer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1p
7