IMC 2015, Blagoevgrad, Bulgaria

Day 2, July 30, 2015

Problem 6. Prove that

X 1
< 2.
n=1
n (n + 1)
(Proposed by Ivan Krijan, University of Zagreb)
Solution. We prove that
1 2 2
< . (1)
n (n + 1) n n+1

Multiplying by n(n + 1), the inequality (1) is equivalent with
p
1 < 2(n + 1) 2 n(n + 1)
p
2 n(n + 1) < n + (n + 1)
which is true by the AM-GM inequality.
Applying (1) to the terms in the left-hand side,
 
X 1 X 2 2
< = 2.
n=1
n (n + 1) n=1 n n+1

Problem 7. Compute Z A
1 1
lim A x dx .
A+ A 1

(Proposed by Jan ustek, University of Ostrava)

Solution 1. We prove that
1 A 1
Z
lim A x dx = 1 .
A+ A 1

For A > 1 the integrand is greater than 1, so

1 A 1 1 A
Z Z
1 1
A x dx > 1dx = (A 1) = 1 .
A 1 A 1 A A
In order to nd a tight upper bound, x two real numbers, > 0 and K > 0, and split the interval
into three parts at the points 1 + and K log A. Notice that for suciently large A (i.e., for A > A0 (, K)
with some A0 (, K) > 1) we have 1 + < K log A < A.) For A > 1 the integrand is decreasing, so we can
estimate it by its value at the starting points of the intervals:
Z A Z 1+ Z K log A Z A 
1 1 1
A dx =
x + + <
A 1 A 1 1+ K log A

1 1 1

= A + (K log A 1 )A 1+ + (A K log A)A K log A <
A
1 1 1
 1
< A + KA log A + A A
1+ K log A = + KA 1+ log A + e K .
A
Hence, for A > A0 (, K) we have
1 A 1
Z
1 1
1 < A x dx < + KA 1+ log A + e K .
A A 1
Taking the limit A we obtain
Z A Z A
1 1 1 1 1
1 lim inf A dx lim sup
x A x dx + e K .
A A 1 A A 1

Now from +0 and K we get

1 A 1 1 A 1
Z Z
1 lim inf A dx lim sup
x A x dx 1,
A A 1 A A 1

so lim inf A1 1A A x dx = lim sup A1 1A A x dx = 1 and therefore

R 1 R 1

A A
Z A
1 1
lim A x dx = 1 .
A+ A 1

Solution 2. We will1 employ l'Hospital's rule. R

Let f (A, x) = A x , g(A, x) = x1 A x , F (A) = 1A f (A, x)dx and G(A) = 1A g(A, x)dx. Since A f and
1 R

A
g are continuous, the parametric integrals F (A) and G(A) are dierentiable with respect to A, and
Z A Z A
0 1 1 1 1 1 1
F (A) = f (A, A) + f (A, x)dx = A A + A x dx = A A + G(A),
1 A 1 x A
and 1
Z A Z A
0 AA 1 1 1
G (A) = g(A, A) + g(A, x)dx = + 2
A x dx =
1 A A 1 x
 A 1 1
1 1 1 1 AA AA 1
=A +A A x = + .
log A 1 A A log A log A
Since limA A A = 1, we can see that limA G0 (A) = 0. Aplying l'Hospital's rule to lim we
1 G(A)
A A
get
G(A) G0 (A)
lim = lim = 0,
A A A 1
so  
0 1 G(A)
lim F (A) = lim A A + = 1 + 0 = 1.
A A A
Now applying l'Hospital's rule to lim F (A)
A
we get
A

A
F 0 (A)
Z
1 1 F (A)
lim A x dx = lim = lim = 1.
A+ A 1 A A A 1

Problem 8. Consider all 2626 words of length 26 in the Latin alphabet. Dene the weight of a word as
1/(k + 1), where k is the number of letters not used in this word. Prove that the sum of the weights of all
words is 375 .
(Proposed by Fedor Petrov, St. Petersburg State University)
Solution. Let n = 26, then 375 = (n + 1)n1 . We use the following well-known
Lemma. If f (x) is a polynomial of degree at most n, then its (n + 1)-st nite dierence vanishes:
f (x) := n+1 i n+1
f (x + i) 0.
n+1
P

i=0 (1) i
Proof. If is the operator which maps f (x) to f (x + 1) f (x), then is indeed (n + 1)-st power
n+1

of and the claim follows from the observation that decreases the power of a polynomial.
In other words, f (x) = i=1 (1) f (x + i). Applying this for f (x) = (n x)n , substituting
Pn+1 i+1 n+1


i
x = 1 and denoting i = j + 1 we get
n n  
(1)j
 
n
X n+1 j n
X n
(n + 1) = (1) (n j) = (n + 1) (n j)n .
j=0
j+1 j=0
j j+1
The j -th summand nj (1) (n j)n may be interpreted as follows: choose j letters, consider all

j

j+1
(n j)n words without those letters and sum up (1) over all those words. Now we change the order of
j

j+1
summation, counting at rst by words. For any xed word W with k absent letters we get kj=0 kj (1)
P
j

j+1
=
1
kj=0 (1)j k+1 1
, since the alternating sum of binomial coecients kj=1 (1)j k+1 vanishes.
P
P

k+1 j+1
= k+1 j+1
That is, after changing order of summation we get exactly initial sum, and it equals (n + 1) . n1

Problem 9. An n n complex matrix A is called t-normal if AAt = At A where At is the transpose of

A. For each n, determine the maximum dimension of a linear space of complex n n matrices consisting
of t-normal matrices.
(Proposed by Shachar Carmeli, Weizmann Institute of Science)
Solution.
Answer : The maximum dimension of such a space is n(n+1) 2
.
The number 2 can be achieved, for example the symmetric matrices are obviously t-normal and
n(n+1)

they form a linear space with dimension n(n+1)

2
. We shall show that this is the maximal possible dimension.
Let Mn denote the space of n n complex matrices, let Sn Mn be the subspace of all symmetric
matrices and let An Mn be the subspace of all anti-symmetric matrices, i.e. matrices A for which
At = A.
Let V Mn be a linear subspace consisting of t-normal matrices. We have to show that dim(V )
dim(Sn ). Let : V Sn denote the linear map (A) = A + At . We have

dim(V ) = dim(Ker ()) + dim(Im ())

so we have to prove that dim(Ker ()) + dim(Im ()) dim(Sn ). Notice that Ker () An .
We claim that for every A Ker () and B V , A(B) = (B)A. In other words, Ker () and Im ()
commute. Indeed, if A, B V and A = At then

(A + B)(A + B)t = (A + B)t (A + B)

AAt + AB t + BAt + BB t = At A + At B + B t A + B t B
AB t BA = AB + B t A A(B + B t ) = (B + B t )A
A(B) = (B)A.

Our bound on the dimension on V follows from the following lemma:

Lemma. Let X Sn and Y An be linear subspaces such that every element of X commutes with every
element of Y . Then
dim(X) + dim(Y ) dim(Sn )

Proof. Without loss of generality we may assume X = ZSn (Y ) := {x Sn : xy = yx y Y }. Dene the

bilinear map B : Sn An C by B(x, y) = tr(d[x, y]) where [x, y] = xy yx and d = diag(1, ..., n) is the
matrix with diagonal elements 1, ..., n and zeros o the diagonal. Clearly B(X, Y ) = {0}. Furthermore, if
y Y satises that B(x, y) = 0 for all x Sn then tr(d[x, y]) = tr([d, x], y]) = 0 for every x Sn .
We claim that {[d, x] : x Sn } = An . Let Eij denote the matrix with 1 in the entry (i, j) and 0 in
all other entries. Then a direct computation shows that [d, Eij ] = (j i)Eij and therefore [d, Eij + Eji ] =
(j i)(Eij Eji ) and the collection {(j i)(Eij Eji )}1i<jn span An for i 6= j .
It follows that if B(x, y) = 0 for all x Sn then tr(yz) = 0 for every z An . But then, taking z = y,
where y is the entry-wise complex conjugate of y , we get 0 = tr(yy) = tr(yyt ) which is the sum of squares
of all the entries of y . This means that y = 0.
It follows that if y1 , ..., yk Y are linearly independent then the equations
B(x, y1 ) = 0, ..., B(x, yk ) = 0
are linearly independent as linear equations in x, otherwise there are a1 , ..., ak such that B(x, a1 y1 + ... +
ak yk ) = 0 for every x Sn , a contradiction to the observation above. Since the solution of k linearly
independent linear equations is of codimension k,
dim({x Sn : [x, yi ] = 0, for i = 1, .., k})

dim(x Sn : B(x, yi ) = 0 for i = 1, ..., k) = dim(Sn ) k.

The lemma follows by taking y1 , ..., yk to be a basis of Y .
Since Ker () and Im () commute, by the lemma we deduce that
n(n + 1)
dim(V ) = dim(Ker ()) + dim(Im ()) dim(Sn ) = .
2

Problem 10. Let n be a positive integer, and let p(x) be a polynomial of degree n with integer coecients.
Prove that   1
max p(x) > n .
0x1 e
(Proposed by Gza Ks, Etvs University, Budapest)
Solution. Let  
M = max p(x).
0x1

For every positive integer k, let Z 1
2k

Jk = p(x) dx.
0
2kn 2kn
Obviously 0 < Jk < M 2k is a rational number. If (p(x))2k = ak,i xi then Jk =
ak,i
. Taking the least
P P
i+1
i=0 i=0
1
common denominator, we can see that Jk .
lcm(1, 2, . . . , 2kn + 1)
An equivalent form of the prime number theorem is that log lcm(1, 2, . . . , N ) N if N . Therefore,
for every > 0 and suciently large k we have
lcm(1, 2, . . . , 2kn + 1) < e(1+)(2kn+1)

and therefore
1 1
M 2k > Jk > (1+)(2kn+1) ,
lcm(1, 2, . . . , 2kn + 1) e
1
M > (1+)(n+ 1 ) .
e 2k

Taking k and then +0 we get

1
M .
en
Since e is transcendent, equality is impossible.

Remark. The constant 1e 0.3679 is not sharp. It is known that the best constant is between 0.4213 and 0.4232.
(See I. E. Pritsker, The GelfondSchnirelman method in prime number theory, Canad. J. Math. 57 (2005),
10801101.)