You are on page 1of 10

MASTERS EXAMINATION IN MATHEMATICS

PURE MATH OPTION, Fall 2013


Algebra
A1. Prove that every group automorphism of S
3
is inner.
Suggestion: given an automorphism of S
3
follow -images of the 3 transpositions.
Solution. Since (ab)(abc) = (bc) and (abc)(ab) = (ac), it is easy to see that the center of S
3
is the trivial subgroup. Therefore, the group of inner automorphisms of S
3
is isomorphic to
S
3
and has size 6.
On the other hand, S
3
has 3 transpositions. These must be permuted by an automor-
phism of S
3
, and a nontrivial automorphism induces a nontrivial permutation. This gives an
injective homomorphism from Aut(S
3
) to S
3
, which shows that the group of automorphisms
of S
3
has size at most 6.
It follows that Aut(S
3
)

= Inn(S
3
)

= S
3
.
A2. Dene a homomorphism Z[X] R by sending X to 1 +

2.
Is the kernel of this homomorphism a principal ideal?
Solution. Denote the homomorphism by . A polynomial f(x) is in the kernel of if and
only if f(1 +

2) = 0.
The polynomial g(x) = x
2
2x 3 satises f(1 +

2) = 0 and is irreducible over Q, and


hence over Z (since it is a quadratic with an irrational root).
Any polynomial f(x) such that f(1 +

2) = 0 therefore is divisible by g(x), so ker() =


(g(x)) is principal.
A3. Prove that a group G of order 72 cannot be simple.
Solution. Suppose that G is a group of order 72. Let n
3
(G) be the number of Sylow
3-subgroups of G. Since n
3
(G) 1(mod3) and n
3
(G)|8, we must have n
3
(G) = 1 or 4.
If n
3
(G) = 1, then there is a unique Sylow 3-subgroup, P say, and since all conjugates of
P are Sylow 3-subgroups, P must be normal, so G is not simple.
On the other hand, if n
3
(G) = 4, then G acts by conjugation on the set of Sylow 3-
subgroups. This action is nontrivial, so there is a homomorphism : G S
4
with nontrivial
image. Since 72 ||S
4
| = 24, the kernel of is a nontrivial proper subgroup of G. Therefore
in this case G is not simple either.
1
Complex Analysis
C1. Compute

2
0
cos
5 + 4 cos
d .
Solution. To compute the integral, we recall that
cos =
e
i
+ e
i
2
,
and so, by plugging this into the integral and remembering that z() = e
i
, 0 2 is a
parametrization of the unit circle C, positively oriented, we get:

2
0
cos
5 + 4 cos
d =

2
0
e
i
+ e
i
10 + 4(e
i
+ e
i
)
d =

2
0
1 + e
2i
4e
i
+ 10 + 4e
i
e
i
d
=
1
i

C
1 +
1
z
2
4z + 10 +
4
z
dz =
1
i

C
z
2
+ 1
4z
3
+ 10z
2
+ 4z
dz
The last integral can easily be computed using the Residue Theorem. Indeed, notice that
the denominator can be factored as
4z
3
+ 10z
2
+ 4z = 2z(2z
2
+ 5z + 2) = 2z(z + 2)(2z + 1)
and so the function we are integrating has two simple poles, at z = 0 and z =
1
2
, inside the
unit circle, with residues
Res
(
z
2
+ 1
4z
3
+ 10z
2
+ 4z
; z = 0
)
=
1
4
and
Res
(
z
2
+ 1
4z
3
+ 10z
2
+ 4z
; z =
1
2
)
=
5
12
.
So, by the Residue Theorem,

C
z
2
+ 1
4z
3
+ 10z
2
+ 4z
dz = 2i
(
1
4

5
12
)
=
i
3
.
Putting it all together we get

2
0
cos
5 + 4 cos
d =

3
.
C2. Let
f(z) =
z
z
2
+ 4
.
Find all the Laurent series expansions of f about z
0
= 2i and specify the regions in which
they are valid.
2
Solution. One can immediately see that
f(z) =
z
(z 2i)(z + 2i)
=
(
1 +
2i
z 2i
)

1
z + 2i
.
Since the rst factor on the right-hand side already has the form of a Laurent series about
z
0
= 2i, we concentrate on the second factor,
1
z+2i
. The rst important observation is that
1
z+2i
has only one singularity, at z = 2i. Therefore, if we wish to expand
1
z+2i
about z
0
= 2i
we will do it in two separate regions: |z 2i| < 4 and |z 2i| > 4, respectively.
Indeed, for |z 2i| < 4, we obtain that:
1
z + 2i
=
1
(z 2i) + 4i
=
1
4i

k=0
(1)
k
(4i)
k
(z 2i)
k
,
and hence
f(z) =
(
2i
z 2i
+ 1
)

1
4i

k=0
(1)
k
(4i)
k
(z 2i)
k
=
1
4i
[
2i
z 2i
+

k=0
(
2i
(1)
k+1
(4i)
k+1
+
(1)
k
(4i)
k
)
(z 2i)
k
]
=
1
4i
[
2i
z 2i
+
1
2

k=0
(1)
k
(4i)
k
(z 2i)
k
]
On the other hand, for |z 2i| > 4, we obtain that
1
z + 2i
=
1
(z 2i) + 4i
=
1
z 2i

k=0
(4i)
k
(z 2i)
k
,
and hence
f(z) =
(
1 +
2i
z 2i
)

k=1
(4i)
k1
(z 2i)
k
=
1
z 2i
+
1
2

k=2
(4i)
k1
(z 2i)
k
.
C3. Find the number of zeroes of f(z) = z
5
z
4
+ z
3
6z
2
+ 2z +
1
2
in the annulus
A = {z C| 1 |z| 3}.
Solution. We proceed by using Rouches Theorem twice. Indeed, note that on the circle
|z| = 1, we have that
|6z
2
| = 6 > 5 +
1
2

z
5
z
4
+ z
3
+ 2z +
1
2

We conclude then that in the disk |z| < 1 the number of zeroes of f is the same as the
number of zeroes of 6z
2
, namely 2.
On the other hand, on the circle |z| = 3, we have that
|z
5
| = 243 > 168 +
1
2

z
4
+ z
3
6z
2
+ 2z +
1
2

,
3
and hence in the disk |z| < 3 the number of zeroes of f is equal to the number of zeroes of
z
5
, namely 5.
In conclusion, we can deduce from above that the number of zeroes of f in the annulus
A = {z C| 1 |z| 3} is 5-2=3.
PURE MATH OPTION, Fall 2013
Number Theory
N1. Let (n) denote the number of divisors of n. Show that, as x ,

nx
(n) = xlog x + O(x).
Solution. We start by writing

nx
(n) =

nx

|n
1
=

|n
nx
1
=

m
x

1 ( after writing n = m)
=

x
[
x

]
=

x
(
x

+ O(1)
)
= x

x
1

+ O(x).
Next, examining the area under the graph of y = 1/t shows that for each k 1
1
+ 1
<

+1

dt
t
<
1

.
Summing up
[x]

=2
1

<

[x]
1
dt
t
<
[x]1

=1
1

.
As a result

x
1

[x]
1
dt
t
+ O(1) = log x + O(1).
4
Finally,

nx
(n) = x

x
1

+ O(x) = x(log x + O(1)) + O(x) = xlog x + O(x).


N2. Let (s) =

n1
1
n
s
, s C with Re(s) > 1, denote the Riemann zeta function. Let (n)
denote the Mobius function. Show that, for any s C with Re(s) > 1, we have

n1
(n)
n
s
=
1
(s)
.
Solution. We note that for each s = + it with > 1 we have

n1

(n)
n
s

n1
1
n

< +.
Consequently the series is absolutely convergent. Next for Re(s) > 1
(s)

n1
(n)
n
s
=
(

m1
1
m
s
)(

n1
(n)
n
s
)
=

k1
1
k
s

mn=k
(n)
=

k1
1
k
s

n|k
(n).
Now we use the well-known property of the function:

n|k
(n) =
{
1 if n = 1;
0 otherwise.
This implies
(s)

n1
(n)
n
s
= 1
from which the desired identity follows.
N3. Show that, for every two natural numbers m and n, the expression
gcd(m, n)
n
(
n
m
)
is an integer.
Solution. If m > n the assertion is obvious. Otherwise, there are r and s such that
rm + sn = gcd(m, n).
5
Then
gcd(m, n)
n
(
n
m
)
=
rm + sn
n
(
n
m
)
= r
m
n
(
n
m
)
+ s
(
n
m
)
.
Next
m
n
(
n
m
)
=
m
n
n!
m!(n m)!
=
(n 1)!
(m1)!(n m)!
=
(
n 1
m1
)
.
Consequently
gcd(m, n)
n
(
n
m
)
= r
(
n 1
m1
)
+ s
(
n
m
)
.
This last expression is clearly integral.
Real Analysis
R1. Show that for any M > 0,

n=1
n
n
3
+x
3
converges uniformly on (0, M) to a dierentiable
function f(x) such that on (0, M) the function f

(x) is equal to the sum of the derivatives


of the terms of this series.
Solution. Let f
n
(x) =
n
n
3
+x
3
. Then f

n
(x) =
3nx
2
(n
3
+x
3
)
2
, so that on [0, M] one has |f

n
(x)|
3nM
2
(n
3
+0
3
)
2
=
3M
2
n
5
. Since

n=1
1
n
5
converges, by the Weierstrass M test

n=1
f

n
(x) converges
uniformly on [0, M] to a continuous function. Furthermore,

n=1
f
n
(0) =

n=1
1
n
2
con-
verges. Thus by the theorem on termwise dierentiation,

n
f
n
(x) converges uniformly on
[0, M] to a function f(x) such that f

(x) =

n=1
f

n
(x) on (0, M).
R2. Let f be a function which is continuous on the interval (1, 1), and is dierentiable at
every point of (1, 1) except possibly the point 0. Assume that the limit L = lim
x0
f

(x)
exists. Prove that f is dierentiable at 0 and that f

(0) = L.
Solution. The fastest way is Lhopitals rule applied to f(x) f(0):
lim
x0
f(x) f(0)
x 0
= lim
x0
(f(x) f(0))

(x 0)

= lim
x0
f

(x)
= L
One may also use the denition of derivative in conjunction with the mean-value theorem
(imitating the proof of lHopital). Namely, by the mean value theorem f(x) f(0) = xf

(c
x
)
for some c
x
between 0 and x. Thus
f(x)f(0)
x0
= f

(c
x
). For any sequence x
n
0, c
xn
also
approaches zero, so
lim
n
f(x
n
) f(0)
x
n
0
= lim
n
f

(c
xn
)
= L
6
R3. Let f be a polynomial of degree n. Suppose that for k = 0, ..., n, both f
(k)
(0) and
f
(k)
() are integers. Prove that

0
f(x) sin xdx is an integer.
Solution. One integrates by parts n + 1 times. After the rst one we get


0
f(x) sin xdx = f() cos() + f(0) cos(0) +


0
f

(x) cos xdx


After the second one, we have


0
f(x) sin xdx = f() cos() + f(0) cos(0)
+f

() sin() f

(0) sin(0)


0
f

(x) sin xdx


After n + 1 integrations by parts one has


0
f(x) sin xdx =
n

k=0
a
k
f
(k)
() +
n

k=0
b
k
f
(k)
(0)
Here each a
n
is cos() or sin() and each b
n
is cos(0) or sin(0), all integers. There
is no integral term now since f
n+1
(x) = 0 for all x, f(x) being a polynomial of degree n.
Since each f
(k)
() and each f
(k)
(0) is an integer by assumption, and each a
k
and b
k
is an
integer, we conclude

0
f(x) sin x dx is an integer too.
Logic
L1.
(a) State the Compactness Theorem for rst-order logic.
Solution. The Compactness Theorem states that for any set of sentences, if every nite
subset of is satisable, then itself is satisable.
(b) The Soundness Theorem asserts that for any formula and any set of formulas , if
, then . The Completeness Theorem asserts that for any formula and any set
of formulas , if , then .
Use the Soundness Theorem and Completeness Theorem to prove the Compactness The-
orem.
Solution. Suppose that is a set of sentences such that every nite subset of has a model.
Suppose, towards a contradiction, that does not have a model. Then |= x(x = x)
(vacuously). By the Completeness Theorem, x(x = x). Since proofs are nite, there is
a nite
0
such that
0
x(x = x). By the Soundness Theorem,
0
|= x(x = x). By
assumption, there is |=
0
, whence |= x(x = x), a contradiction.
L2.
7
Recall that a class K of L-structures is L-axiomatizable if there exists an L-theory T such
that, for any L-structure M, we have M K if and only if M |= T.
Let L
rv
be the language that consists of:
a constant symbol 0;
a 2-place function symbol +;
for each real number r, a 1-place function symbol f
r
.
It is then straightforward to verify that the class of vector spaces over R is L
rv
-axiomatizable;
here the function symbols f
r
are being interpreted as scalar multiplication by r. For exam-
ple, to state the distribute law, we need to include, for each pair of real numbers r, s, the
sentence x(f
r
(x) + f
s
(x) = f
r+s
(x)).
Prove that the class of nite-dimensional real vector spaces is not L
rv
-axiomatizable. (Hint:
Use the Compactness Theorem.)
Solution. Suppose, towards a contradiction, that the class of nite-dimensional vector
spaces was
rv
-axiomatizable by the set of sentences . Add innitely many new constants
{c
1
, c
2
, c
3
, . . .} to the language and consider the set

of sentences (in the new language)


consisting of together with all of the sentences of the form
f
r
1
c
1
+ f
r
2
c
2
+ + f
rn
c
n
= 0,
where n is any positive natural number and r
1
, . . . , r
n
is an n-tuple of real numbers not
all of which are 0. Any nite subset of

has a model (which can just be taken to be R


n
for n suciently large and where the c
i
s are interpreted as basis vectors). Thus, by the
compactness theorem, there is a model V of

. Note that {c
V
1
, c
V
2
, c
V
3
, . . .} is an innite
linearly independent set in V , contradicting that V |= .
L3.
Recall that for L-structures M and N, M is a substructure of N if |M| |N| and the
interpretations of the symbols in M are the restrictions of the interpretations of the symbols
in |N|.
Suppose that L is a language and M and N are L-structures such that M is a substructure
of N. We say that M is an elementary substructure of N, written M N, if for any formula
( x) and any tuple a from M, we have M |= ( a) if and only if N |= ( a).
(a) Suppose that M
0
is a substructure of M
1
which is in turn a substructure of M
2
; in
symbols: M
0
M
1
M
2
. Further suppose that M
0
M
2
and M
1
M
2
. Prove that
M
0
M
1
.
Solution. Suppose that ( x) is an -formula and a is a tuple from M
0
such that M
0
|= ( a).
Since M
0
M
2
, we have M
2
|= ( a). Since a also lives in M
1
and M
1
M
2
, we have
M
1
|= ( a).
(b) Give an example of a substructure M of a structure N such that M is not an elementary
substructure of N.
8
Solution. Let M = (N, <) and let N = (Z, <). Then M is a substructure of N but M N
(so in particular M N) as M |= xy(x = y x < y) while N |= xy(x = y x <
y)).
Topology
T1. Let X be a compact Hausdor space, p a point of X and U an open neighborhood of
p. Prove that there is an open neighborhood V of p such that the closure of V is contained
in U.
Solution. For each x / U let A
x
and B
x
be disjoint open sets with p A
x
and x B
x
.
These exists as X is Hausdor. The collection {B
x
}
x/ U
cover U
c
. Now, since U
c
is a closed
subset of a compact space it is compact, so there is a nite subcover. Let B
x
1
, B
x
2
, . . . , B
xn
be such a subcover. Let V = A
x
1
A
x
2
. . . A
xn
and let W = B
x
1
. . . B
xn
. We claim V
has the properties requested.
First, V is a neighborhood of p since p A
x
for all x, and V is a nite intersection of open
sets, hence is open. Second, since V is disjoint from each B
x
i
, V is a subset of W
c
. Since W
is a union of open sets it is open, and W
c
is closed. Thus cl(V ) W
c
. Since the B
x
i
cover
U
c
, W
c
U. Thus cl(V ) U as claimed.
T2. Prove that if X has a dense subset which is connected in the subspace topology then
X is connected.
Solution. Let A X be dense and connected in the subspace topology.
Let X = U V with U and V disjoint and open. We need to see that one of U or V is
empty. By dention, AU and AV are open in the subspace topology on A. As they are
disjoint and cover A, and A is connected, one of them is empty. Without loss of generality,
assume A U is empty. This means A U
c
. Since A is dense, and U
c
is closed, we have
X = cl(A) U
c
. Thus U is empty as claimed.
T3. Let X be the square [0, 1] [0, 1], and let T be the collection of subsets of X whose
intersections with each vertical segment {x} [0, 1] and each horizontal segment [0, 1] {y}
are open sets in the subspace topology on the segment (viewed as a subspace of the usual
R
2
).
(1) Show that (X, T ) satises the axioms for a topological space.
(a) The empty set has trivial intersection with each segment, and the empty set
is open in the subspace topology on the segment, so the empty set is in T .
Similarly, X intersects any segment in the entire segment, and this is also open
in the subspace topology. Thus X is in T .
(b) Suppose U and V are in T . For any segment L, L U and L V are open in
L. Since L (U V ) = (L U) (L V ), L (U V ) is an intersection of
two open sets in L and hence is open in L. Since this holds for all vertical and
horizontal segments L, U V is in T .
9
(c) Let {U
a
}
aA
be any collection of sets in T . For any segment L and any a A,
L U
a
is open in L. As
aA
(U
a
L) = (
aA
U
a
) L, (
aA
U
a
) L is a union
of open sets in L, and so it is open in L. Since this holds for all vertical and
horizontal segments L,
aA
U
a
is in T .
(2) For t [0, 1] show that the set U
t
= {(x, y) | x = y} {(t, t)} is open in (X, T )
For any segment L of the form s [0, 1] or [0, 1] s for s = t, L U
t
is the
complement of a point in L and hence open in L ( since points are closed in the
standard topology on the segments). For L equal to either t [0, 1] or [0, 1] t,
L U
t
= L which is open in L. Thus U
t
is open in (X, T ).
(3) Is (X, T ) Hausdor?
Yes. Note that by the denition of subspace topology, any open set U in the
standard topology on R
2
has U X open in (X, T ). Thus, since R
2
is Hausdor in
the standard topology, (X, T ) is also Hausdor.
(4) Is (X, T ) compact?
No. The collection of sets U
t
dened in the second part form a cover of X as
any point (x, y) with x = y is in all the U
t
and any point (x, x) is in U
x
. For any
subcollection U
t
1
, U
t
2
, , U
tn
, let x be any element of [0, 1] not in {t
1
, t
2
, , t
n
}.
The point (x, x) is then not in any U
t
i
and so the subcollection does not cover. Thus
there is no nite subcover, proving (X, T ) is not compact.
(5) Is (X, T ) connected?
Yes. Let U and V be disjoint, non-empty open sets in (X, T ) with U V = X.
For any t [0, 1] let L
t
= t [0, 1]. By denition U L
t
and V L
t
are open in L
in the standard topology. We have (U L
t
) (V L
t
) = (U V ) L
t
= L
t
and
(U L
t
) (V L
t
) = (U V ) L
t
= . Since L
t
is connected in the standard
topology, one of U L
t
or V L
t
is empty and the other is all of L
t
. As U is non-
empty, there is therefor some a such that a[0, 1] U. Applying the same argument
to the segments [0, 1] t and the fact that V is non-empty, we have some b so that
[0, 1] b V . Thus the point (a, b) is in U V , which is a contradiction.
10

You might also like