SEEMOUS 2015 Contest Problems and Solutions

Problem 1. Prove that for every x  (0,1) the following inequality holds:

1
2 2 2
 1  (cos y) dy  x  (sin x) .
0
Solution 1. Clearly
1 x
2 2
 1  (cos y) dy   1  (cos y) dy .
0 0
Define a function F :[0,1]  by setting:
x
F ( x )   1  (cos y )2 dy  x 2  (sin x )2 .
0
Since F (0)  0 , it suffices to prove F '( x )  0 . By the fundamental theorem of Calculus, we have

F '( x )  1  (cos x )2  x sin x cos x .

x 2  (sin x )2
Thus, it is enough to prove that
(1  (cos x)2 )( x 2  (sin x)2 )  ( x  sin x cos x)2 .
But this is a straightforward application of the Cauchy-Schwarz inequality.

1 x

 1   cos y  dy   1   cos y  dy for each fixed x   0,1 . Observe that

2 2
Solution 2. Clearly
0 0
x

 1   cos y  dy is the arc length of the function f  y   sin y on the interval  0, x  which is clearly
2

strictly greater than the length of the straight line between the points  0, 0  and  x,sin x  which in turn

x 2   sin x  .
2
is equal to

Problem 2. For any positive integer n , let the functions f n :  be defined by f n1( x)  f1( f n ( x)) ,
where f1( x)  3x  4 x3 . Solve the equation f n ( x )  0 .

Solution. First, we prove that x  1  f n ( x)  1 holds for every positive integer n. It suffices to

demonstrate the validity of this implication for . But, by assuming x  1 , it readily follows that

f1 ( x)  x 3  4 x 2  3  4 x 2  1 , which completes the demonstration. We conclude that every

solution of the equation f n ( x )  0 lies in the closed interval [1,1] . For an arbitrary such x , set
 
x  sin t where t  arcsin x  [ , ] . We clearly have f1(sin t )  sin3t , which gives
2 2
f n ( x)  sin3n t  sin(3n arcsin x).
Thus, f n ( x )  0 if and only if sin(3n arcsin x)  0 , i.e. only when 3n arcsin x  k for some k Z .
Therefore, the solutions of the equation f n ( x )  0 are given by
x  sin kn ,
3
1 3 n
3n  1
where k acquires every integer value from up to .
2 2
 Problem 3. For an integer n  2 , let A, B, C, D  М п ( ) be matrices satisfying:
AC  BD  I n ,
AD  BC  On ,
where I n is the identity matrix and On is the zero matrix in М п ( ) .
Prove that:
a) CA  DB  I n and DA  CB  On ,
b) det( AC )  0 and (l)n det( BD)  0 .
Solution. a) We have
AC  BD  i( AD  BC )  I n  ( A  iB)(C  iD)  I n ,
which implies that the matrices A  iB and C  iD are inverses to one another. Thus,
(C  iD )( A  iB )  I n  CA  DB  i ( DA  CB )  I n
 CA  DB  I n , DA  CB  On .
b) We have
det((A  iB )C )  det( AC  iBC )
AD  BC On
 det(AC  iAD )
 det( A(C  iD ).
On the other hand,
(C iD )( AiB )  I n
detC  det((C  iD)(A  iB )C )  det((C  iD) A(C  iD))
 det( A) | det(C  iD) |2 .
Thus,
det( AC )  (detA)2 |det(C  iD) |2  0 .
Similarly
det(( A  iB ) D )  det( AD  iBD )
AD  BC On
 det(  BC  iBD )
 ( l)n det( B(C  iD)).
This implies that
(C iD )( AiB )  I n
detD  det((C  iD)( A  iB) D)  ( l)n det((C  iD) B(C  iD))
 ( 1)n det( B) | det(C  iD) |2 .
Thus, (1)n det( BD)  (detB)2 | det(C  iD) |2  0 .
 Problem 4. Let I  be an open interval which contains 0, and f : I  be a function of class
C 2016 ( I ) such that f (0)  0, f (0)  1, f (0)  f (0)  ...  f (2015) (0)  0, f (2016) (0)  0.
i) Prove that there is   0 such that
0  f ( x )  x, x  (0,  ). (1.1)
ii) With  determined as in i), define the sequence ( an ) by

a , a 1  f (a ), n  1.
2 n 1 n (1.2)

Study the convergence of the series  anr , for r  .
n 1

Solution. i) We claim that there exists   0 such that f ( x )  0 for any x  (0,  ). For this, observe
that, since f is of class C1 and f (0)  1  0, there exists   0 such that f ( x )  0 on (0,  ). Since
f (0)  0 and f is strictly increasing on (0,  ), the claim follows.
Next, we prove that there exists   0 such that f ( x )  x for any x  (0,  ). Since
f (2016) (0)  0 and f is of class C 2016 , there is   0 such that f (2016) (t )  0, for any t  (0,  ). By
the Taylor's formula, for any x  (0,  ), there is   0,1 such that
f (0) f (2015) (0) 2015 f (2016) ( x ) 2016
f ( x )  f (0)  1! x  ...  2015!
x  2016!
x , (1.3)
hence
f (2016) ( x ) 2016
g ( x)  2016!
x  0, x  (0,  ).

Taking   min  ,    0, the item i) is completely proven.

ii) We will prove first that the sequence (an ) given by (1.2) converges to 0. Indeed, by relation (1.1)
it follows that
0  an 1  an , n  1,
hence the sequence (an ) is strictly decreasing and lower bounded by 0. It follows that (an ) converges
to some  [0,  ) . Passing to the limit in (1.2) , one gets  f ( ). Taking into account (1.1), we deduce
2
that  0.
In what follows, we calculate
lim nan2015.
n 
From an  0, using the Stolz-Cesàro Theorem, we conclude that
( n 1)  n
lim nan2015  lim n  lim  lim 1
n  2015  2015  1
1 1 1 1
n  n  n 
an2015 an 1 an f ( an ) 2015 an2015

( xf ( x ))2015
 lim 1  lim .
x 0
1  1 x 0 x
2015
 f ( x )2015
f ( x )2015 x 2015
 2015
 2 f ( 2016) (x) 2017 
 x  x 
Observe that, by (1.3)
xf x 2015   2016 !  .
x 2015  f ( x) 2015 f ( 2016) (x) 2016 2014
 x (x  x f ( x)    f ( x) )
2013 2014

2016!
Since f is of class C 2016 , lim f (2016) ( x )  f (2016) (0) and
x 0
( xf ( x ))2015
lim  2016!  0.
x 0 x
2015
 f ( x )2015 2015 f (2016) (0)
 
It means, by the comparison criterion, that the series  anr and  1
r
converge and/or diverge
n 1 n 1 n 2015

simultaneously, hence the series  anr converges for r  2015, and diverges for r  2015.
n 1