You are on page 1of 4

# SEEMOUS 2013 PROBLEMS AND SOLUTIONS

Problem 1
Find all continuous functions ) : [1, 8] 1, such that
_
2
1
)
2
(t
3
)dt + 2
_
2
1
)(t
3
)dt =
2
3
_
8
1
)(t)dt
_
2
1
(t
2
1)
2
dt.
Solution. Using the substitution t = n
3
we get
2
3
_
8
1
)(t)dt = 2
_
2
1
n
2
)(n
3
)dn = 2
_
2
1
t
2
)(t
3
)dn.
Hence, by the assumptions,
_
2
1
_
)
2
(t
3
) + (t
2
1)
2
+ 2)(t
3
) 2t
2
)(t
3
)
_
dt = 0.
Since )
2
(t
3
)+(t
2
1)
2
+2)(t
3
)2t
2
)(t
3
) = ()(t
3
))
2
+(1t
2
)
2
+2(1t
2
))(t
3
) =
_
)(t
3
) + 1 t
2
_
2
_
0, we get
_
2
1
_
)(t
3
) + 1 t
2
_
2
dt = 0.
The continuity of ) implies that )(t
3
) = t
2
1, 1 _ t _ 2, thus, )(r) = r
2=3
1, 1 _ r _ 8.
Remark. If the continuity assumption for ) is replaced by Riemann integrability then
innitely many )s would satisfy the given equality. For example if C is any closed nowhere
dense and of measure zero subset of [1, 8] (for example a nite set or an appropriate Cantor type
set) then any function ) such that )(r) = r
2=3
1 for every r [1, 8]C satises the conditions.
Problem 2
Let ', '
2
(C) be two nonzero matrices such that
'
2
=
2
= 0
2
and ' +' = 1
2
where 0
2
is the 2 2 zero matrix and 1
2
the 2 2 unit matrix. Prove that there is an invertible
matrix '
2
(C) such that
' =
_
0 1
0 0
_

1
and =
_
0 0
1 0
_

1
.
First solution. Consider ), q : C
2
C
2
given by )(r) = 'r and q(r) = r.
We have )
2
= q
2
= 0 and )q +q) = id
C
2; composing the last relation (to the left, for instance)
with )q we nd that ()q)
2
= )q, so )q is a projection of C
2
.
If )q were zero, then q) = id
C
2, so ) and q would be invertible, thus contradicting )
2
= 0.
Therefore, )q is nonzero. Let n Im()q) 0 and n C
2
such that n = )q(n). We obtain
)q(n) = ()q)
2
(n) = )q(n) = n. Let = q(n). The vector is nonzero, because otherwise we
obtain n = )() = 0.
Moreover, n and are not collinear since = `n with ` C implies n = )() = )(`n) =
`)(n) = `)
2
Let us now consider the ordered basis E of C
2
consisting of n and .
We have )(n) = )
2
(q(n)) = 0, )() = )(q(n)) = n, q(n) = and q() = q
2
(n) = 0.
Therefore, the matrices of ) and q with respect to E are
_
0 1
0 0
_
and
_
0 0
1 0
_
, respectively.
We take to be the change of base matrix from the standard basis of C
2
to E and we are
done.
1
2
Second solution. Let us denote
_
0 1
0 0
_
by 1
12
and
_
0 0
1 0
_
by 1
21
. Since '
2
=
2
=
0
2
and ', ,= 0
2
, the minimal polynomials of both ' and are equal to r
2
. Therefore, there
are invertible matrices 1, C /
2
(C) such that ' = 11
12
1
1
and = C1
21
C
1
.
Note that 1 and C are not uniquely determined. If 1
1
1
12
1
1
1
= 1
2
1
12
1
1
2
, then (1
1
1
1
2
)1
12
=
1
12
(1
1
1
1
2
); putting 1
1
1
1
2
=
_
a /
c d
_
, the last relation is equivalent to
_
0 a
0 c
_
=
_
c d
0 0
_
. Consequently, 1
1
1
12
1
1
1
= 1
2
1
12
1
1
2
if and only if there exist a C 0 and
/ C such that
1
2
= 1
1
_
a /
0 a
_
. (+)
Similarly, C
1
1
21
C
1
1
= C
2
1
21
C
1
2
if and only if there exist c C 0 and , C such that
C
2
= C
1
_
c 0
, c
_
. (++)
Now, ' +' = 1
2
, ' = 11
12
1
1
and = C1
21
C
1
give
11
12
1
1
C1
21
C
1
+C1
21
C
1
11
12
1
1
= 1
2
,
or
1
12
1
1
C1
21
C
1
1 +1
1
C1
21
C
1
11
12
= 1
2
.
If 1
1
C =
_
r j
. t
_
, the previous relation means
_
. t
0 0
__
0 0
t j
_
+
_
j 0
t 0
__
0 t
0 .
_
= (rt j.)1
2
,= 0
2
.
After computations we nd this to be equivalent to rt j. = t
2
,= 0. Consequently, there are
j, . C and t C 0 such that
C = 1
_
t +
yz
t
j
. t
_
. (+ + +)
According to (+) and (++), our problem is equivalent to nding a, c C 0 and /, , C
such that C
_
c 0
, c
_
= 1
_
a /
0 a
_
. Taking relation (+ + +) into account, we need to nd
a, c C 0 and /, , C such that
1
_
t +
yz
t
j
. t
__
c 0
, c
_
= 1
_
a /
0 a
_
,
or, 1 being invertible,
_
t +
yz
t
j
. t
__
c 0
, c
_
=
_
a /
0 a
_
.
This means
_

_
ct +c
j.
t
+,j = a
cj = /
c. +,t = 0
ct = a
,
and these conditions are equivalent to
_
_
_
cj = /
c. = ,t
ct = a
.
It is now enough to choose c = 1, a = t, / = j and , =
.
t
.
3
Third Solution. Let ), q be as in the rst solution. Since )
2
= 0 there exists a nonzero

1
Ker) so )(
1
) = 0 and setting
2
= q(
1
) we get
)(
2
) = ()q +q))(
1
) =
1
,= 0
by the assumptions (and so
2
,= 0). Also
q(
2
) = q
2
(
1
) = 0
and so to complete the proof it suces to show that
1
and
2
are linearly independent, because
then the matrices of ), q with respect to the ordered basis (
1
,
2
) would be 1
12
and 1
21
respec-
tively, according to the above relations. But if
2
= `
1
then 0 = q(
2
) = `q(
1
) = `
2
so since

2
,= 0, ` must be 0 which gives
2
= 0
1
= 0 contradiction. This completes the proof.
Remark. A nonelementary solution of this problem can be given by observing that the
conditions on ', imply that the correspondence 1
2
1
2
, ' 1
12
and 1
21
extends
to an isomorphism between the subalgebras of /
2
(C) generated by 1
2
, ', and 1
2
, 1
12
, 1
21
respectively, and then one can apply Noether-Skolem Theorem to show that this isomorphism
is actually conjugation by an G|
2
(C) etc.
Problem 3
Find the maximum value of
_
1
0
[)
0
(r)[
2
[)(r)[
1
_
r
dr
over all continuously dierentiable functions ) : [0, 1] R with )(0) = 0 and
_
1
0
[)
0
(r)[
2
dr _ 1. (+)
Solution. For r [0, 1] let
q(r) =
_
x
0
[)
0
(t)[
2
dt.
Then for r [0, 1] the Cauchy-Schwarz inequality gives
[)(r)[ _
_
x
0
[)
0
(t)[ dt _
__
x
0
[)
0
(t)[
2
dt
_
1=2
_
r =
_
rq(r)
1=2
.
Thus
_
1
0
[)
0
(r)[
2
[)(r)[
1
_
r
dr _
_
1
0
q(r)
1=2
q
0
(r)dr =
2
3
[q(1)
3=2
q(0)
3=2
]
=
2
3
__
1
0
[)
0
(t)[
2
dt
_
3=2
_
2
3
.
by (+). The maximum is achieved by the function )(r) = r.
Remark. If the condition (+) is replaced by
_
1
0
[)
0
(r)[
p
dr _ 1 with 0 < j < 2 xed, then
the given expression would have supremum equal to +, as it can be seen by considering
continuously dierentiable functions that approximate the functions )
M
(r) = 'r for 0 _ r _
1
'
p
and
1
'
p1
for
1
'
p
< r _ 1, where ' can be an arbitrary large positive real number.
Problem 4
Let '
2
(Q) such that there is : , : ,= 0, with
n
= 1
2
. Prove that either
2
= 1
2
or
3
= 1
2
.
4
First Solution. Let )
A
(r) = dct( r1
2
) = r
2
:r + j Q[r] be the characteristic
polynomial of and let `
1
, `
2
be its roots, also known as the eigenvalues of matrix . We have
that `
1
+ `
2
= : Q and `
1
`
2
= j Q. We know, based on Cayley-Hamilton theorem, that
the matrix satises the relation
2
: + j1
2
= 0
2
. For any eigenvalue ` C there is an
eigenvector A ,= 0, such that A = `A. By induction we have that
n
A = `
n
A and it follows
that `
n
= 1. Thus, the eigenvalues of satisfy the equalities
`
n
1
= `
n
2
= 1 (+).
Is `
1
R then we also have that `
2
R and from (+) we get that `
1
= `
2
= 1 (and note
that : must be odd) so satises the equation ( + 1
2
)
2
=
2
+ 2 + 1
2
= 0
2
and it follows
that 1
2
=
n
= (+1
2
1
2
)
n
= :(+1
2
) 1
2
which gives = 1
2
and hence
3
= 1
2
.
If `
1
C R then `
2
= `
1
C R and since `
n
1
= 1 we get that [`
1;2
[ = 1 and this implies
that `
1;2
= cos t i sint. Now we have the equalities `
1
+ `
2
= 2 cos t = : Q and `
n
1
= 1
implies that cos :t + i sin:t = 1 which in turn implies that cos :t = 1. Using the equality
cos(: + 1)t + cos(: 1)t = 2 cos t cos :t we get that there is a polynomial 1
n
= r
n
+ of
degree : with integer coecients such that 2 cos :t = 1
n
(2 cos t). Set r = 2 cos t and observe
that we have 1
n
(r) = 2 so r = 2 cos t is a rational root of an equation of the form r
n
+ = 0.
However, the rational roots of this equation are integers, so r Z and since [r[ _ 2 we get that
2 cos t = 2, 1, 0, 1, 2.
When 2 cos t = 2 then `
1;2
are real numbers (note that in this case `
1
= `
2
= 1 or
`
1
= `
2
= 1) and this case was discussed above.
When 2 cos t = 0 we get that
2
+1
2
= 0
2
so
2
= 1
2
.
When 2 cos t = 1 we get that
2
+1
2
= 0
2
which implies that (+1
2
)(
2
+1
2
) = 0
2
so
3
= 1
2
.
When 2 cos t = 1 we get that
2
++1
2
= 0
2
and this implies that (1
2
)(
2
++1
2
) = 0
2
so
3
= 1
2
. It follows that
n

_
1
2
, ,
2
_
. However,
n
= 1
2
and this implies that either
= 1
2
or
2
= 1
2
both of which contradict the equality
3
= 1
2
. This completes the
proof.
Remark. The polynomials 1
n
used in the above proof are related to the Chebyshev poly-
nomials, T
n
(r) = cos(:arccosr). One could also get the conclusion that 2 cos t is an integer by
considering the sequence r
m
= 2 cos(2
m
t) and noticing that since r
m+1
= r
2
m
2, if r
0
were a
noninteger rational
a
/
(/ 1) in lowest terms then the denominator of r
m
in lowest terms would
be /
2
m
and this contradicts the fact that r
m
must be periodic since t is a rational multiple of .
Second Solution. Let :
A
(r) be the minimal polynomial of . Since
2n
1
2
= (
n
+
1
2
)(
n
1
2
) = 0
2
, :
A
(r) must be a divisor of r
2n
1 which has no multiple roots. It is well
known that the monic irreducible over Q factors of r
2n
1 are exactly the cyclotomic polynomials

d
(r) where d divides 2:. Hence the irreducible over Q factors of :
A
(r) must be cyclotomic
polynomials and since the degree of :
A
(r) is at most 2 we conclude that :
A
(r) itself must be a
cyclotomic polynomial, say
d
(r) for some positive integer d with c(d) = 1 or 2 (where c is the
Euler totient function), c(d) being the degree of
d
(r). But this implies that d 1, 2, 3, 4, 6
and since ,
3
cannot be equal to 1
2
we get that :
A
(r) r + 1, r
2
+ 1, r
2
r + 1 and this
implies that either
2
= 1
2
or
3
= 1
2
.