RUTGERS UNIVERSITY

GRADUATE PROGRAM IN MATHEMATICS

Written Qualifying Examination
January 2011
Questions and Solutions
First DayPart I: Answer each of the following three questions
1. Let f be a complex valued measurable function on R. Let be the
Lebesgue measure and suppose that for each a < b,

Z b



f d b a.

a

Prove that |f (x)| 1 for almost every x.

First Solution. Invoke the Lebesgue Differentiation Theorem: If x is in the
Lebesgue set for f (which has complement of measure zero), then
Z x+h
1
f d.
f (x) = lim
h0 2h xh
By assumption, the integral is absolute value at most 2h, hence |f (x)| 1.
Second Solution. Suppose not. The set E = {x : |f (x)| > 1} is the union
of the countable family of sets


E(, ) = {x : Re ei f (x) > 1 + }
with ,  rational and  > 0. Thus if (E) > 0 then there exists and an
 > 0 such that the set F = E(, ) has positive Lebesgue measure. Let U
be any open containing F such that (U ) (1 + /2)(F ) (such sets exits
by the outer regularity of Lebesgue measure). Then U is a countable union
of disjoint intervals (aj , bj ). Since the measure is countably additive, the
assumption on f gives

Z
  Z bj

 X

 X

 f d = 
f d
(bj aj ) = (U ) (1 + /2)(F ) .

 

U
aj
j=1

j=1

On the other hand, since Re(ei f ) (1 + ) on F ,

 Z
 Z

Z
 
 


i
i
 f d =  e f d  Re(e f ) d (1 + )(F ) .
 
 


F

Since

Z
 Z
 Z
 
 

 f d  f d 

 
 
U

U F



f d ,
c

the last two inequalities give

Z

(1 + /2)(F ) (1 + )(F ) 

U F



f d .
c

Z




Hence 
f d (/2)(F ) for every such open set U containing F .
U F c
Since (F ) > 0 this last inequality is aR contradiction. Indeed, since f is
integrable, we can choose such a U with U F c |f | d < (/2)(F ).
2. Use contour integration to evaluate
Z
1
dx.
(1 + x2 )2
0
Be clear about any computation of residues and about any computations of
limits of integrals.
1
. Then the poles of f are at z = i, i and
(1 + z 2 )2
are of order 2. Let R > 1. and consider the counter clockwise closed path
consisting of
Solution. Let f (z) =

1 = {y = 0, x [R, R]} and 2 = {Re z 0, |z| = R} .

Then by the Residue Theorem,
Z
o
dn

f (z)dz = 2i Resf (i) = 2i
(z i)2 f (z) 
dz
z=i

n 2 o


= 2i
 = .
3
(z + i) z=i
2
2

Written Qualifying Examination, January 2011

On the other hand, if R > 2 there is a constant C > 0 such that |f (z)|
4
C/R
for z on 2 . Since the length of 2 is R, this gives the estimate
R
|f
(z)dz|
C/R3 . Therefore
2
Z
0

Z
Z
1
1
1 R
1
1
dx =
dx = lim
dx
2
2
2
2
R 2 R (1 + x2 )2
(1 + x )
2 (1 + x )
Z

Z

1
f (z) dz
f (z) dz = .
= lim
2 R
4
2

3. Let S9 denote the symmetric group on {1, 2, ..., 9} and let S9 be given
(in table form) by


1 2 3 4 5 6 7 8 9
=
.
9 5 8 1 7 2 6 3 4
As usual C(), the centralizer of in S9 , is defined to be C() = {
S9 | = }. Find |C()| and justify your answer.
First Solution. In cycle form = (194)(2576)(38). S9 acts on itself
by conjugation and C() is the stabilizer of . The orbit of consists of
9!(2!3!1!)/(3!4!2!) = 9!/(24) elements. This is because there are 9!/(3!4!2!)
partitions of {1, . . . , 9} into subsets of cardinalities 3, 4, and 2 and there are
(k 1)! distinct k-cycles permuting a set of k elements. Hence |C()| =
|S9 |/|S9 | = (9!)/(9!/24) = 24.
Second Solution. In cycle form = (194)(2576)(38). S9 acts on itself
by conjugation and C() is the stabilizer of . If S9 then 1 =
( (1) (9) (4))( (2) (5) (7) (6))( (3) (8)). Thus if 1 = then is
determined by the choices of (1) (3 possibilities), (2) (4 possibilities), and
(3) (2 possibilities). This gives a total of 3 4 2 = 24 elements in C().
First DayPart II: Answer three of the following questions. If you
work on more than three questions, indicate clearly which three
should be graded.
4. Recall that a set X in a topological space is called a G set when it is a
countable intersection of open sets, and it is called an F set when it is a

Rutgers Graduate Program in Mathematics

countable union of closed sets. Let denote Lebesgue measure on R. Show

that for every Borel set A R there is a G set G and an F set F such that
F A G and (G F c ) = 0. Here F c = R \ F .
Solution. Use the inner and outer regularity of to get for each k > 0 an
open set Uk and a closed set Ck such that Ck A Uk and
(A Ckc ) 1/(2k),

(Uk Ac ) 1/(2k) .

Without loss of generality

we may assume that Uk+1 S
Uk and Ck Ck+1
T
for all k. Let G = k Uk , which is a G set. Let F = k Fk , which is an F
set. Then
(G F c ) = lim (Uk Fkc ) .
k

Since
(Uk Fkc ) = (Uk Ac ) + (A Fkc ) < 1/k ,
the assertion is proved.
5. Let f be analytic on the unit disc D, and assume that |f (z)| < 1 for all
z D. Prove that if there exist two distinct points a and b in the disc which
are fixed points, that is, f (a) = a and f (b) = b, then f (z) = z for all z D.
az
for z D be the Mobius transform. Then a
1a
z
is an automorphism of D with inverse a . Consider F (z) = a f a (z).
Then F (0) = 0 and


ab
ab
F
=
6= 0 .
(?)
1a
b
1a
b
Solution. Let a (z) =

Furthermore, |F (z)| 1 for z D. By the Schwarz lemma, F (z) = wz for

some w D. Clearly w = 1 by (?). Hence f (z) = z for all z D.
6. Prove that there exists no simple group of order 80.
Solution. Let G be a group of order 80 = 24 5. Let n5 be the number
of 5-Sylow subgroups in G. If n5 = 1 then G has a normal subgroup since
all 5-Sylow subgroups are conjugate. Hence G is not simple in this case. If
n5 > 1 then, by the theorem on the number of Sylow subgroups, the group

Written Qualifying Examination, January 2011

has 16 distinct 5-Sylow subgroups. Since these subgroups are cyclic of prime
order, their pairwise intersections are trivial, so in this case G has at most
80 16 4 = 16 elements whose orders are powers of 2. These elements form
only one Sylow 2-subgroup, and therefore that subgroup is normal. Hence G
is not simple.
7. Let X and Y be topological spaces and f : X Y and g : X Y be
continuous functions. Prove that if Y is a Hausdorff space then {x X :
f (x) = g(x)} is closed.
Solution. let E = {x X : f (x) = g(x)}. Suppose z 6 E; we show that
there is a neighborhood of z disjoint from E. Since f (z) 6= g(z) there are
neighborhoods Nf of f (z) and Ng of g(z) that are disjoint. By continuity of
f and g, f 1 (Nf ) and g 1 (Ng ) are neighborhoods of z. Let N = f 1 (Nf )
g 1 (Ng ). Then N is a neighborhood of z. Now we claim that N E = .
Let w N . Then f (w) Nf and g(w) Ng and the disjointness of Nf and
Ng implies f (w) 6= g(w) so w 6 E.
8. Let (fn ) be a sequence of nonnegative integrable functions on [0, 1] converging almost everywhere to a function f (x). Prove that if
Z
Z
f d
fn d =
lim
n

[0,1]

[0,1]

then
lim fn = f

in L1 [0, 1].
First Solution. Since fn 0, we have 0 min{fn , f } f . Since f
L1 [0, 1] and min{fn , f } f pointwise as n , the Lebesgue Dominated
Convergence Theorem implies that
Z
Z
lim
min{fn , f } d =
f d .
n

[0,1]

[0,1]

Hence the relation |fn f | = fn + f 2 min{fn , f } gives

Z
Z
Z
lim
|fn f | d = lim
fn d
f d = 0 .
n

[0,1]

[0,1]

[0,1]

Rutgers Graduate Program in Mathematics

Second Solution. Set X = [0, 1] and let  > 0R be given. Since f is integrable
and nonnegative, there exists > 0 such that E f d <  for any measurable
set E X with (E) < . Since (X) < , Egorovs Theorem implies
that there exists a measurable set E with (E) < and {fn } converging to
f uniformly on X \ E. Since (X \ E) < , the uniform convergence on
X \ E and the assumed convergence of the integrals implies that there exits
an integer N such that
Z

Z


|fn f | d <  and  (fn f ) d <  for all n N .
(?)
X\E

Assume n N . Since fn 0 we can use (?) to estimate

Z
Z
Z
f d
fn d = (fn f ) d +
0
E
ZE
ZE
Z
=
(fn f ) d
(fn f ) d +
f d
X

X\E

3 .
From this estimate we obtain
Z
Z
|fn | + |f | d 4 .
|fn f | d
E

Thus

|fn f | d
X

|fn f | d +
X\E

|fn f | d 5
E

for all n N . Since  > 0 was arbitrary, this proves the convergence in L1 .
9. Let A and B be commuting 8 by 8 diagonalizable matrices over the real
numbers with characteristic polynomials
det(A I) = ( 1)3 ( 3)5
and
det(B I) = 2 ( 4)6 .
Suppose the minimum polynomial of A B is
(2 1)(2 9) .

Written Qualifying Examination, January 2011

Find the dimension of the vector space of all 8 by 8 real matrices that commute with both A and B.
Solution. Let V1 , V3 be the eigenspaces with eigenvalues = 1 and = 3
for A. Then dim V1 = 3 and dim V3 = 5 since A is diagonalizable. Likewise,
let W0 , W4 be the eigenspaces with eigenvalues = 0 and = 4 for B. Then
dim W0 = 2 and dim W4 = 6 since B is diagonalizable. Let Vi,j = Vi Wj for
i = 1, 3 and j = 0, 4. Since B commutes with A,
R8 = V1,0 V1,4 V3,0 V3,4
and A and B act by the scalars i and j, respectively, on Vi,j . We have
dim V1,0 + dim V3,0 = 2 ,
dim V1,0 + dim V1,4 = 3 ,

dim V1,4 + dim V3,4 = 6 ,

dim V3,0 + dim V3,4 = 5 .

(?)

Since A B has minimum polynomial (2 1)(2 9), the eigenvalues of

A B are 1 and 3. Hence V1,0 and V3,4 are nonzero eigenspaces for A B
with eigenvalues 1, 1, respectively. Likewise, V3,0 and V1,4 are nonzero
eigenspaces for A B with eigenvalues 3, 3 respectively. It follows from
(?) that dim V1,0 = dim V3,0 = 1, and hence dim V1,4 = 2 and dim V3,4 = 4.
A matrix commutes with both A and B if and only if it maps each joint
eigenspace Vi,j to itself. The action of the matrix on Vi,j can be any linear
transformation. So, the dimension of the space of all 8 by 8 real matrices
that commute with both A and B is 12 + 22 + 12 + 42 = 22.

Day 1 Exam End

RUTGERS UNIVERSITY
GRADUATE PROGRAM IN MATHEMATICS
Written Qualifying Examination
January 2011, Day 2
Second DayPart I: Answer each of the following three questions

1. Let f (x) be a function on [0, 1] and suppose that f 0 (x) is defined for all
0 x 1. Prove that f 0 (x) is a measurable function.
Solution. Take a differentiable extension of f (x) to the right of x = 1. For
example, set f (1 + a) = f (1) + af 0 (1) for a > 0. Set


1

n (x) = n f x +
f (x) .
n
Each n (x) is continuous and therefore measurable. Since f 0 (x) = limn n (x)
it is measurable.
2. Prove that all the roots of z 7 5z 3 + 12 = 0 lie in {z C : 1 |z| 2}.
Solution. Let f (z) = z 7 5z 3 + 12, g(z) = z 7 , and h(z) = 12. On |z| = 1,
|f (z) h(z)| = |z 7 5z 3 | 6 < 12 = |h(z)| .
Therefore f (z) has no zeros in |z| < 1 by Rouches theorem, since h(z) has
no zeros there. On |z| = 2,
|f (z) g(z)| = |5z 3 12| 5 23 + 12 = 26 < 27 = |g(z)| .
Therefore f (z) has 7 zeros in |z| 2 by Rouches theorem, since g(z) has 7
zeros there (counting multiplicities). By the first part, the zeros of f (z) all
have modulus greater than 1, and since f (z) has degree 7, these are all of its
zeros.

Written Qualifying Examination, January 2011

3. Are the quotient rings Z[x]/(x3 +1) and Z[x]/(x3 +2x2 +x+1) isomorphic?
Provide full justification for your answer.
Solution. The rings are not isomorphic. In fact, the first ring contains zero
divisors since x3 + 1 = (x + 1)(x2 x + 1) is not irreducible in Z[x]. For the
second ring, note that 1 is not a root of p(x) = x3 + 2x2 + x + 1. Since
p(x) is monic and the product of its (complex) roots is the constant term 1,
it follows that p(x) has no integer roots. Hence p(x) is irreducible in Z[x],
since any factorization of it would include at least one linear factor (because
p(x) has degree 3). Since Z[x] is a unique factorization domain, it follows
that p(x) is prime. Thus Z[x]/(p(x)) is an integral domain. This proves that
the rings are not isomorphic.
Second DayPart II: Answer three of the following questions. If
you work on more than three questions, indicate clearly which
three should be graded.
4. Find the Laurent expansion of f (z) = (1 z 2 )e1/z around z = 0. Compute
the residue at 0.
Solution. The Laurent series for e1/z is given by
1/z

X
1
.
=
n!z n
n=0

Hence



X
X
1
1
n2 + 3n + 1 n
2
f (z) =

z .
=
z

z
+
n
n2
n!z
n!z
(n
+
2)!
n=0
n=0
The coefficient of z 1 in the series is (12 + 3 1 + 1)/3!. Thus Resf (0) = 5/6.
5. Let f be a complex-valued measurable function on R. Let be Lebesgue
measure and suppose that for each g L2 (), the function f g L1 (). Show
that f L2 ().
Solution. For each positive integer N define
Z
2
FN = {g L () :
|f g| d N } .

10

Rutgers Graduate Program in Mathematics

The assumption on f is that

L () =

FN .

(?)

N =1
2

We first show that FN is closed in L (). Indeed, if {gj } is a sequence in

FN that converges to a function g in the L2 () norm, then by passing to a
subsequence we may assume that {gj } that converges to g almost everywhere.
Since f gj converges to f g almost everywhere, Fatous Lemma implies that
Z
Z
|f g| d lim inf |f gj | d N .
j

Since L2 () is a complete metric space, Baires Theorem asserts that one of

the sets on the right side of (?) must have an interior point. Hence there
exists an N , a function g0 FN , and a real number r > 0 such that the ball
of radius r around g0 is contained in FN . Thus for all unit vectors h L2 (),
Z
Z
Z
Z
r |hf |d |(rh + g0 )f |d + |g0 f |d N + |g0 f |d .
This proves that
Z
M = sup

|hf |d < .

(??)

khk2 =1

R
Hence the linear functional g 7 F (g) = f g d, for g L2 (), is bounded
with bound M . By the Riesz Representation
Theorem, there exists a function
R
2
L () such that F (g) = g d for all g L2 (). Hence f = almost
everywhere, so f L2 ().
(??) can also be proved without the help of the Baire Category Theorem as
follows. First one may assume that f 0, as the assumption on f is also
valid for the real and imaginary parts of f , and then for their positive and
negative parts. If (??) is Rnot valid, then for any k N, there exists gk with
||gk ||L2 () = 1 such that f gk d = ak k. One may even take gk 0.
Define
l
X
hl =
(kak )1 gk ,
k=1

Pl

P
then ||hl ||L2 () k=1 k , and f hl d = lk=1 k 1 . It now follows by
Monotone Convergence Theorem that

X
h=
(kak )1 gk L2 (),
2

k=1

Written Qualifying Examination, January 2011

but

f hd

f hl d for all l, which leads to

11
f gd = , a contradiction.

6. Let x = (x1 , x2 ) and y = (y1 , y2 ) be vectors over the field F = Z/3. Show
that the bilinear forms B(x, y) = x1 y1 x2 y2 and D(x, y) = x1 y1 + x2 y2
are equivalent.
Solution. Interpreting x and y as column vectors, then D(x, y) = xt y and
B(x, y) = xt (I)y. Thus we need a matrix


a b
P =
c d
with coefficients in F such that P t (I)P = I, or P t P = I. Thus entries of
P must satisfy a2 + c2 = 1, b2 + d2 = 1, and ab + cd = 0. One solution is
a = b = d = 1 and c = 1, since 2 = 1 in F .
7. Consider the curve S = {(x, sin(1/x)) : x (0, 1]} R2 . Let T =
S ({0} [1, 1]). Show that T is a connected subset of R2 .
Solution. Suppose A, B is a pair of disjoint non-empty open sets of R2
whose union contains S. S is the image of the connected set (0, 1] under the
continuous map f : (0, 1] R2 given by f (x) = (x, sin(1/x)) and is hence
connected. Thus S is a subset of A or B; assume S A. Every point (0, y)
of {0} [1, 1] = T S is a limit point of S. Indeed, let b = arcsin(y) and
for k a positive integer let xk = 1/(b + 2k). Then f (xk ) = (xk , y), which
converges to (0, y) as k . So if (0, y) B then B contains points of S
since B is open. This contradicts the assumption A B = , so we conclude
that T A, and hence T is connected.
8. Let G be a finite group. Prove that G is cyclic if and only if G has exactly
one subgroup of order n for each positive integer n dividing |G|.
Solution. Let N = |G|. If G is a cyclic group then G
= Z/N Z. The
subgroups of G are in one-to-one correspondence with subgroups kZ of Z
containing N Z, i.e., with k | N , by the isomorphism theorems. Since
kZ/N Z
= Z/(N/k)Z there is exactly one subgroup of order n = N/k for
each n dividing N .

12

Rutgers Graduate Program in Mathematics

Conversely, suppose that G has a unique subgroup Gn of order n for each n

dividing N . We proceed by induction on N . Let p be a prime dividing N .
By Cauchys Theorem every Gn with p dividing n has a subgroup of order
p, so Gp Gn for all such n. Therefore by the isomorphism theorems, G/Gp
has a unique subgroup of each order dividing N/p. By induction G/Gp is
cyclic. Choose x G such that Gp x generates G/Gp . Then N/p divides the
order of x.
If x has order N , then G is cyclic. So assume that x has order N/p. Also
hxiGp = G so N = |G| = |hxi||Gp |/|hxi Gp | = (N/p)p/|hxi Gp |. Therefore
hxiGp = 1. Every subgroup H of G is normal, since H is the only subgroup
of order |H|. Therefore, G = hxi Gp . If p divides N/p, then hxi has a
subgroup of order p, contradicting the uniqueness of Gp . Therefore p does
not divide N/p, so G
= ZN/p Zp
= ZN by the Chinese Remainder Theorem.

9. Exhibit a conformal map f : U D, (that is, a bijective map f from U

to D, such that both f and its inverse are holomorphic), where D is the unit
disc {z C : |z| < 1} and U is the set {z D : Re z > 0}.
Solution. First we let

iz
.
i+z
Since fractional linear transformations carry circles (a line being a circle
through ) to circles, and g has the values g(i) = 0, g(0) = 1, g(i) = ,
and g(1) = i, it follows that g maps the imaginary axis to the real axis, and
maps the unit circle to the imaginary axis. Also
g(z) =

g(iy) =

1y
,
1+y

so g maps {iy : 1 < y < 1} to the positive real axis. Since 1/2 U
and g(1/2) = (3 + 4i)/5 is in the first quadrant, it follows that g maps U
conformally to the first quadrant
Q = {x + iy C : x > 0 and y > 0} .
Next, the map z 7 z 2 sends Q conformally to the upper half-plane
H = {x + iy C : y > 0} .

Written Qualifying Examination, January 2011

13

Finally, if we let

zi
,
z+i
then h maps H to D, because |h(z)| = 1 if z is real (since in that case
h(z) = w/w if w = z i), and h(i) = 0. So, the composite map f given by

 

i z 2
i z 2
2
f (z) = h(g(z) ) =
i
+i
i+z
i+z
h(z) =

sends U conformally onto D.

Remark. It is recommended that maps constructed in this problem be
illustrated with appropriate sketches. Relevant points and boundaries in the
domains and ranges should be labeled.

Exam Day 2 End