You are on page 1of 15

MASTERS EXAMINATION IN MATHEMATICS

PURE MATH OPTION, Spring 2017

Full points can be obtained for correct answers to 8 questions. Each numbered question
(which may have several parts) is worth 20 points. All answers will be graded, but the score
for the examination will be the sum of the scores of your best 8 solutions.

Use separate answer sheets for each question. DO NOT PUT YOUR NAME ON
YOUR ANSWER SHEETS. When you have finished, insert all your answer sheets into
the envelope provided, then seal it.

Any student whose answers need clarification may be required to submit to an oral exam-
ination.

Algebra

A1. Let R be a commutative ring with identity and let φ : M → M be an R-module


homomorphism. Suppose that φ ◦ φ = φ. Prove that M ∼
= ker(φ) ⊕ im(φ).

Solution. For each m ∈ M , we have φ(m − φ(m)) = φ(m) − φ(φ(m)) = φ(m) − φ(m) = 0.
Hence m = (m − φ(m)) + φ(m) ∈ ker(φ) + im(φ). On the other hand, if m ∈ ker(φ) ∩ im(φ),
then there is m0 ∈ M such that m = φ(m0 ). Thus, m = φ(m0 ) = φ(φ(m0 )) = φ(m) = 0 since
m ∈ ker(φ). We have shown that M = ker(φ) + im(φ) and ker(φ) ∩ im(φ) = 0. It is clear
that both ker(φ) and im(φ) are submodules of M . So M ∼
= ker(φ) ⊕ im(φ).
√ √
A2. Determine [Q( 2 + 3) : Q] with justification.
√ √
Solution. 2 + 3 satisfies x4 − 10x2 + 1 = 0. it remains to show that f (x) = x4 − 10x2 + 1
is irreducible in Q[x]. By Gauss’s Lemma, it suffices to check this in Z[x]. Suppose that
f (x) = g(x)h(x). If deg(g) = 1 or deg(h) = 1, then f would have an integer root r. Then we
would have r2 |1 and hence r = 1 or −1. But neither is a root of f . If deg(g) = deg(h) = 2,
write g = x2 + ax + b and h = x2 + cx + d. Then comparing the coefficients, we would have

a + c = 0, ac + b + d = −10, ad + bc = 0, bd = 1

and hence a = −c&c = 0 or a = −c&b = d. In either case, ac + b + d = −10 would have


no integer solutions. Hence it is not possible to have deg(g) = deg(h) = 2. So we must have
either deg(g) = 0 or deg(h) = 0, i.e. f is irreducible.

A3. Let G be a finite group and N be a normal subgroup.


Let Aut(N ) denote {σ : N → N | σ is a group isomorphism}. Suppose that (|G|, | Aut(N )|) =
1. Prove that N is in the center of G.
1
Solution. For each g ∈ G, let cg : N → N be the conjugation, i.e. cg (n) = gng −1 for each
n ∈ N . Since N is normal, cg is well defined. The map α : G → Aut(N ) defined by α(g) = cg
is a group homomorphism. Hence by the First Isomorphism Theorem G/ ker(α) ∼ = im(α).
Hence | im(α)| divides |G|. On the other hand, im(α) is a subgroup of Aut(N ), by Largange’s
Theorem | im(α)| divides | Aut(N )|. Therefore im(α) = {idN }. So gng −1 = n for all g ∈ G
and all n ∈ N . Hence N is in the center of G.

Complex Analysis

C1. Evaluate
e−ix
Z
dx .
R 1 + x + x2
Justify your calculations!

Solution. We compute the integral evaluating


e−iz
Z
2
dz ,
ΓR 1 + z + z

where R > 0 and the contour ΓR consists of the segment [−R, R] (oriented from left to right)
and of CR , the (negatively-oriented) lower half of the circle centered at 0 and of radius R.
Note that the function
e−iz
f (z) =
1 + z + z2
is meromorphic in the complex plane, with simple poles at the zeroes of the polynomial
1 + z + z 2 . But the discriminant

of this quadratic polynomial

is ∆ = 12 − 4 · 1 · 1 = −3, and
so the roots are z1,2 = −1±i 2
3
. Of these, only z2 = −1−i2
3
is in the lower half plane, and so
this is the only pole of f which will contribute to the integral. Indeed, for R large enough,
the Residue Theorem tells us that:
e−iz
Z
2
dz = 2πi · n(ΓR ; z2 ) · Res(f ; z2 ) = −2πi · Res(f ; z2 ) .
ΓR 1 + z + z

Now, √
e−iz2 e(− 3+i)/2
Res(f ; z2 ) = = √ .
z2 − z1 −i 3
Furthermore, for R large enough and for |z| = R, Im(z) < 0, we see that

f (z) ≤ 1
2
,
R −R−1
and hence, by the ML Theorem
Z

≤ πR

f (z) dz R2 − R − 1 .
CR

We conclude that Z
lim f (z) dz = 0 .
R→∞ CR
2
Thus
∞ R
e−ix e−ix
Z Z
dx = lim dx
−∞ 1 + x + x2 R→∞ −R 1 + x + x
2

e−iz
Z
= lim dz
R→∞ Γ 1 + z + z 2
R

2πe(− 3+i)/2
= √ .
3

C2. Let f be an analytic function in the disk D(0; 2) = {z ∈ C | |z| < 2} with
 f (1) = 0 and
|f (z)| < 10 for all z ∈ D(0; 2). Find the best possible upper bound for f 12 .

Solution. To find this best possible bound, consider the function


( f (2z)
10B 1 (z)
, for z 6= 12
g(z) = 2
3 0
20
f (1) for z = 21 .
We then know that g is analytic in the unit disk D(0; 1), and by applying the Maximum
Modulus Theorem in D(0; r) for r < 1 and letting r → 1, we conclude that
|g(z)| ≤ 1 for all z ∈ D(0; 1) .
In particular, we conclude that
    1 1 1
f 1 ≤ 10 1 = 10 4 − 2 = 10 20

B 1 4
= .
2 2 4 1 − 14 · 12 7
8
7
z

Furthermore, this is indeed the best possible constant, since the function 10B 1 2
satisfies
2
all the conditions of the problem, and
 
10B 1 1 = 20 .

2 2 7

C3. Find all z at which f (z) = |z| is differentiable. Find all z at which f (z) is analytic.
p
Solution. We write f (x + iy) = |x + iy| = u(x, y) + iv(x, y) with u = x2 + y 2 , v = 0.
Computing the partial derivatives at (x, y) 6= (0, 0) we obtain ux = √ x2 2
, uy = √ 2y 2 ,
x +y x +y
vx = vy = 0. Clearly the Cauchy-Riemann equations are not satisfied so the function f (z)
is not differentiable (and hence not analytic) anywhere on C \ {0}. The function f (z) is not
differentiable at z = 0 either since the partial derivatives do not exist (|x| is not differentiable
at the origin).

3
Number Theory

N1.
(1) Define the greatest common divisor and least common multiple of two integers a and
b.
(2) Given a, b ∈ Z\{0, −1, 1}, write their unique prime factorizations as
Y
a= pαi i ,
1≤i≤n
Y
b= pβi i ,
1≤i≤n
where αi ≥ 0, βi ≥ 0 for any 1 ≤ i ≤ n. Prove that the greatest common divisor and
the least common multiple of a, b exist and satisfy
Y min{α ,β }
i i
gcd(a, b) = pi ,
1≤i≤n

max{αi ,βi }
Y
lcm(a, b) = pi ,
1≤i≤n
and
ab = gcd(a, b) lcm(a, b).
(3) Prove that for any non-zero natural numbers a, b such that lcm(a, b) = a2 − b, we
have a2 = 2b.

Solution.
1. Given a, b ∈ Z, assuming that their greatest common divisor and their least common
multiple exist, prove that they are unique.
Let d1 , d2 ∈ N be two greatest common divisors of a and b. Applying the definition of a
greatest common divisor for d1 , we get that d2 | d1 . Applying the definition of a greatest
common divisor for d2 , we get that d1 | d2 . Thus d1 = ±d2 . But d1 , d2 ∈ N, hence d1 = d2 .
Let m1 , m2 ∈ N be two least common multiples of a and b. Applying the definition of
a least common multiple for m1 , we get that m1 | m2 . Applying the definition of a least
common multiple for m2 , we get that m2 | m1 . Thus m1 = ±m2 . But m1 , m2 ∈ N, hence
m1 = m2 .

2. Given a, b ∈ Z\{0, −1, 1}, write their unique prime factorizations as


Y
a= pαi i ,
1≤i≤n
Y
b= pβi i ,
1≤i≤n
where αi ≥ 0, βi ≥ 0 for any 1 ≤ i ≤ n. Prove that the greatest common divisor and the
least common multiple of a, b exist and satisfy
Y min{α ,β }
i i
gcd(a, b) = pi ,
1≤i≤n
4
max{αi ,βi }
Y
lcm(a, b) = pi ,
1≤i≤n
and
ab = gcd(a, b) lcm(a, b).

Let
min{αi ,βi }
Y
d := pi
1≤i≤n
and
max{αi ,βi }
Y
m := pi .
1≤i≤n
Since
min{αi , βi } ≤ αi and min{αi , βi } ≤ βi
for any 1 ≤ i ≤ n, we get d | a and d | b. Similarly, since
max{αi , βi } ≥ αi and max{αi , βi } ≥ βi
for any 1 ≤ i ≤ n, we get a | m and b | m.
Now consider d0 , m0 ∈ N such that
d0 | a, d0 | b
and
a | m0 , b | b0 .
By the Fundamental Theorem of Arithmetic, the above divisibility conditions imply that the
prime factorizations of d0 and m0 are
Y γ
d0 = pi i
1≤i≤n

and Y
m0 = pδi i
1≤i≤n
for some integers γi , δi such that γi ≤ min{αi , βi } and δi ≥ max{αi , βi } for any 1 ≤ i ≤ n.
This, in turn, implies that d0 | d and m | m0 , which proves that d = gcd(a, b) and m =
lcm(a, b).
The relation
ab = gcd(a, b) lcm(a, b).
is derived from the prime factorizations of a, b, gcd(a, b), lcm(a, b), and the observation that
min{αi , βi } + max{αi , βi } = αi + βi
for any 1 ≤ i ≤ n.

3. Prove that for any non-zero natural numbers a, b such that lcm(a, b) = a2 − b, we have
a2 = 2b.
Let a, b ∈ N\{0} be such that m := lcm(a, b) = a2 − b. Let d := gcd(a, b) and write
a = dA, b = dB
for some A, B ∈ N\{0}. Then
gcd(A, B) = 1
5
and, recalling that ab = dm = d(a2 − b), we get
d2 AB = d2 (dA2 − B).
Thus
AB = dA2 − B,
that is,
(1) dA2 = B(A + 1).
Recall that gcd(A, B) = 1, hence gcd(A2 , B) = 1. Also, gcd(A2 , A + 1) = 1. Thus, by
(1) and the Fundamental Theorem of Arithmetic, A = 1, giving d = 2B. In turn, this gives
a2 = (2B)2 = b2 .

N2.
X
(1) Let f : N\{0} −→ C and g : N\{0} −→ C be such that g(n) := f (d). Prove that
d|n
if f is multiplicative, then g is multiplicative. X
(2) Consider the function τ : N\{0} −→ Z defined by τ (n) = 1, the number of all
d|n
positive divisors of n, counted with multiplicity. Prove that n = 180 is the smallest
non-zero natural number such that τ (n) = 18. X
(3) Consider the function σ : N\{0} −→ Z defined by σ(n) = d, the sum of all
d|n
positive divisors of n, counted with multiplicity. Prove that if σ(n) is prime, then
n = pα for some prime p and some natural number α such that α + 1 is prime.
X
Solution. 1. Let f : N\{0} −→ C and let g : N\{0} −→ C, g(n) := f (d). Prove that if
d|n
f is multiplicative, then g is multiplicative.
Let a, b ∈ N\{0} such that gcd(a, b) = 1. Then there is a bijection
{d : d|ab} −→ {d1 : d1 |a} × {d2 : d2 |b}
d 7→ (d1 , d2 ),
and so
X
g(ab) = f (d)
d|ab
XX
= f (d1 d2 )
d1 |a d2 |b
  
X X
=  f (d1 )  f (d2 )
d1 |a d2 |b

(using that gcd(a, b) = 1 ⇒ gcd(d1 , d2 ) = 1, and that f is multiplicative)


= g(a)g(b).
6
2. Recall the function
τ : N\{0} −→ Z,
X
τ (n) := 1 = the number of all positive divisors of n, counted with multiplicity.
d|n

Prove that n = 180 is the smallest non-zero natural number such that τ (n) = 18.
Let n ∈ N\{0} be such that
(2) τ (n) = 18.
Write its unique prime factorization as
Y
n= pαi i ,
1≤i≤k

where αi ≥ 1 for all 1 ≤ i ≤ k. Then (2) gives


Y
(3) (αi + 1) = 18.
1≤i≤k

Note that the only nontrivial factorizations of 18 are


18 = 2 · 9 = 3 · 6 = 2 · 3 · 3.
So, from the above and (3),
k = 1, α1 = 17,
or
k = 2, α1 = 1, α2 = 8,
or
k = 2, α1 = 2, α2 = 5,
or
k = 3, α1 = 1, α2 = 2, α3 = 2.
The smallest n for each case are, respectively
217 , 28 · 3, 25 · 32 , 22 · 32 · 5.
Among them, the smallest is 22 · 32 · 5 = 180.

3. Recall the function


σ : N\{0} −→ Z,
X
σ(n) := d = the sum of all positive divisors of n, counted with multiplicity.
d|n

Prove that if σ(n) is a prime, then n = pα for some prime p and some natural number α
such that α + 1 is prime.
Let n ∈ N\{0} be such that
(4) σ(n) is prime.
Write its unique prime factorization as
Y
n= pαi i ,
1≤i≤k
7
where αi ≥ 1 for all 1 ≤ i ≤ k. By the multiplicativity of σ,
Y
(5) σ(n) = σ (pαi i ) .
1≤i≤k

Then, by (4), k = 1, giving


n = pα1 1 ,
and
(6) σ (pα1 1 ) is prime.
Suppose α1 + 1 = ab for some a, b ∈ N\{0, 1}. Then
pab
1 −1
σ (pα1 1 ) = = (pa1 )b−1 + (pa1 )b−2 + . . . + pa1 + 1 · pa−1 a−2
 
1 + p 1 + . . . + p 1 + 1 ,
p1 − 1
thus not a prime, contradicting (6). Therefore α1 + 1 is a prime.

N3.
(1) Prove that for any prime p ≥ 5 we have
7p − 6p ≡ 1( mod 43).
(Hint: You may want to consider different cases according to the residue class
p( mod 6).)
(2) Find the remainder in the division of 100! by 109.  
(3) Let p ≥ 7 be a prime. Prove that there exists a prime q such that q < p and pq = 1.
(Hint: You may want to consider different cases according to the residue class
p( mod 8).)

Solution.
1. Prove that for any prime p ≥ 5 we have
7p − 6p ≡ 1( mod 43).
(Hint: You may want to consider different cases according to the residue class p( mod 6).)
Let p ≥ 5 be a prime. Modulo 6, the only possibilities for p are p = 6k + 1 or p = 6k + 5
for some k ∈ N\{0}. We will consider each case separately.
If p = 6k + 1, then
k k
(7) 7p − 6p = 7 · 76 − 6 · 66 .
Let us note that
2
(8) 76 = 73 ≡ (−1)2 ( mod 43) ≡ 1( mod 43),
2
(9) 66 = 63 ≡ 12 ( mod 43) ≡ 1( mod 43).
Thus, modulo 43, equation (7) becomes
7p − 6p ≡ 1( mod 43).
If p = 6k + 5, then
k k
(10) 7p − 6p = 75 · 76 − 65 · 66 ≡ 75 − 65 ( mod 43)
8
(where we have used once again (8) and (9)).
Now
75 = 72 · 72 · 7 ≡ 36 · 7( mod 43) ≡ −6( mod 43)
and
65 = 63 · 62 ≡ 62 ( mod 43).
Using this in equation (10), we get
7p − 6p ≡ −6 − 36( mod 43) ≡ 1( mod 43).

2. Find the remainder in the division of 100! by 109.


Since 109 is a prime, by Wilson’s Theorem
108! ≡ −1( mod 109).
Thus
100! · (−8)(−7)(−6)(−5)(−4)(−3)(−2)(−1) ≡ −1( mod 109).
Now
(−8)(−7)(−6)(−5)(−4)(−3)(−2)(−1) = 8·7·6·5·4·3·2
= 8 · 7 · 6 · 120
≡ 8 · 7 · 6 · 11( mod 109)
≡ 9 · 11( mod 109)
≡ 99( mod 109).
Thus
100! · 99 ≡ −1( mod 109),
giving
100! · (−10) ≡ −1( mod 109),
that is,
100! · 10 ≡ 1( mod 109).
But
1 ≡ 110( mod 109),
hence
100! · 10 ≡ 110( mod 109).
In other words,
109 | 10 (100! − 11),
giving
100! ≡ 11( mod 109).
 
3. Let p ≥ 7 be a prime. Prove that there exists a prime q such that q < p and pq = 1.
(Hint: You may want to consider different cases according to the residue class p( mod 8).)
Let p ≥ 7 be a prime. Modulo 8, the only possibilities for p are p = 8k + 1, p = 8k + 3,
p = 8k + 5, or p = 8k + 7 for some k ∈ N\{0}. We will consider each case separately.
If p = 8k + 1 or p = 8k + 7, take q := 2.
9
p+1
If p = 8k + 3, then 4
= 2k + 1 has a unique prime factorization, say
p+1 Y
2k + 1 = = qiαi ,
4 1≤i≤n

where αi ≥ 1 for all 1 ≤ i ≤ n. Note that all the primes qi must be odd. Then, by the
Quadratic Reciprocity Law, the prime q1 < p satisfies
 Y α 
        4 qi i − 1
q1 q1 −1 p−1 p q1 −1 p p q1 −1  1≤i≤n
= (−1) 2 · 2 = (−1) 2 ·(4k+1)

= = (−1) 2  
p q1 q1 q1  q1 

 
q1 −1 −1 q1 −1 q1 −1
= (−1) 2 = (−1) 2 · (−1) 2 = 1.
q1
p−1
If p = 8k + 5, then 4
= 2k + 1 has a unique prime factorization, say
p−1 Y
2k + 1 = = qiαi ,
4 1≤i≤n

where αi ≥ 1 for all 1 ≤ i ≤ n. Note that all the primes qi must be odd. Then, by the
Quadratic Reciprocity Law, the prime q1 < p satisfies
 Y α 
        4 qi i + 1  
q1 q1 −1 p−1
· p q1 −1
·(4k+2) p p  1≤i≤n
 = 1 = 1.

= (−1) 2 2 = (−1) 2 = =
p q1 q1 q1  q1  q1

Real Analysis

R1. Prove using the definition of a limit that


sin(n)
lim √ =0
n→∞ n

lim x + 6 = 3
x→3

Solution.Fix an  > 0. We need


sin(n)
√ < .
n
However, | sin(n)| ≤ 1. So, we will ensure a stronger inequality

sin(n)
√ ≤ √1 < .
n n
1 √
< n.

1
< n.
2
10
So, pick N = 1 + 12 . If n > N , then the needed inequality is achieved.
 
Fix  > 0. We have
√ √
√ |( x + 6 − 3)( x + 6 + 3)| |x + 6 − 9| |x − 3|
| x + 6 − 3| = √ =√ ≤ < .
x+6+3 x+6+3 3
So, if δ = 3, then |x − 3| < δ implies the required inequality.

R2. Let g be a differentiable function on interval [−1, 1] with values g(−1) = −1, g(0) = 4,
g(1) = 3.
• Prove that there exists a point a ∈ (−1, 1) such that g(a) = 0.
• Prove that there exists a point b ∈ (−1, 1) such that g 0 (b) = −1.
• Prove that there exists a point c ∈ (−1, 1) such that g 0 (c) = 2.

Solution. First, since g(−1) < 0 and g(0) > 4, by the IVT there exists a such that g(a) = 0.
Second,
g(1) − g(0)
= −1.
1−0
By MVT, there is b such that g 0 (b) = −1. Third,
g(1) − g(−1)
= 2.
1 − (−1)
By MVT, there is c such that g 0 (c) = 2.

R3. Let (
xy
x2 +y 2
for (x, y) 6= (0, 0),
f (x, y) =
0 for (x, y) = (0, 0).
Show that the first partial derivatives exist for all (x, y) ∈ R2 , but that f is not continuous
at the origin.

Solution. That f is not continuous at the origin follows from the fact that for any h > 0
h2 1
lim f (h, h) = lim = 6= f (0).
h→0 h→0 h2 + h2 2
Concerning the partial derivatives, a calculation shows that for any (x, y) 6= (0, 0):
∂f y(x2 + y 2 ) − xy(2x) y 3 − yx2
(x, y) = = ,
∂x (x2 + y 2 )2 (x2 + y 2 )2
and similarly
∂f xy 2 − x3
(x, y) = 2 .
∂y (x + y 2 )2
To see that the partial derivatives also exist at (0, 0), we first note that f (x, 0) = 0 for all x
and hence
∂f f (x + h, 0) − f (x, 0)
(0, 0) = lim = 0.
∂x h→0 h
The same argument shows that ∂f ∂y
(0, 0) = 0.
11
Logic

L1.
(1) Define what it means for a first-order formula to be in a prenex normal form.
(2) Prove that any first-order formula is logically equivalent to a formula in prenex normal
form. (You may argue in any proof system you want.)

Solution. (1) A first-order formula φ is in a prenex normal form if φ is equal to


Qx1 Qx2 · · · Qxn ψ,
where Q stands for any quantifier and ψ is quantifier-free.

(2) The proof is by induction on the structure of φ. If φ is atomic, then φ is already in


the prenex form, and there is nothing to prove.
Suppose φ is of the form ¬φ1 for some first-order formula φ1 , which is logically equivalent
to a formula of the form
Qx1 Qx2 · · · Qxn ψ.
The formula φ is then logically equivalent to
Q̃x1 Q̃x2 · · · Q̃xn ¬ψ,
where Q̃ denotes the change of a quantifier from ∃ to ∀ and vice versa.
Suppose φ is of the form φ1 ∨ φ2 , where φ1 and φ2 are logically equivalent to formulas
Qx1 Qx2 · · · Qxn ψ1 and Qy1 Qy2 · · · Qym ψ2
respectively. Changing names of the variables, we may assume without loss of generality
that xi 6= yj for all i, j. Under this assumption, φ is logically equivalent to
Qx1 Qx2 · · · Qxn Qy1 Qy2 · · · Qym ψ1 ∨ ψ2 .
Finally, suppose φ is of the form Qy φ1 , where φ1 is equivalent to
Qx1 Qx2 · · · Qxn ψ.
Again, we may assume that y 6= xi for all i, and φ is therefore equivalent to
QyQx1 Qx2 · · · Qxn ψ.

L2. Consider the first-order language L = {E}, where E is a binary relation. Let T be the
theory of graphs, in other words T consists of the following sentences.
(1) ∀x ∼ E(x, x) (no loops).
(2) ∀x∀y E(x, y) =⇒ E(y, x) (edges are undirected).
Any graph can naturally be viewed as a model of T and any model of T is a graph. Recall
that a graph is connected if given any two vertices there is a path between them. Prove
that the class of connected graphs is not axiomatizable, i.e., show that there is no set of
L-sentences T 0 ⊇ T such that G |= T 0 if and only if G is a connected graph.
12
Solution. Suppose towards a contradiction that there exists an L-theory T 0 such that a
graph is a model of T 0 if and only if it is connected. Expand L by adding two constants
L0 = {E, c1 , c2 } and let σn , n ≥ 1, be the sentence
∃x1 ∃x2 · · · ∃xn E(c1 , x1 ) ∧ E(x1 , x2 ) ∧ · · · ∧ E(xn−1 , xn ) ∧ E(xn , c2 ).
Informally, σn asserts existence of a path of length at most n from c1 to c2 .
The L0 -theory T 0 ∪ {¬σn : n ≥ 1} is finitely satsifiable — any finite segment of this
theory has a model, because there are connected graphs of arbitrarily large diameter. By
compactness theorem, there exists a model of T 0 ∪{¬σn : n ≥ 1}. Vertices that correspond to
c1 and c2 are in distinct connected components in this model, contradicting the assumption
on T 0 .

L3. Show that the set


{(ω, x) ∈ N × N : φω (x) is defined}
is not decidable. Here φω is the partial recursive function with Gödel number ω ∈ N.

Solution. Consider the function f : N → N defined by


(
0 if φx (x) is undefined
f (x) =
φx (x) + 1 otherwise.
We claim that f is not recursive. Indeed, if f is recursive, then there is ω ∈ N such that
f = φω , and therefore f (ω) = φω (ω), which is impossible.
Finally, if the set
H = {(ω, x) ∈ N × N : φω (x) is defined}
is decidable, then by Church’s Thesis f is recursive, since it is constructed from the char-
acteristic function of H and the universal Turing machine using the successor function. We
conclude that H is undecidable.

13
Topology

T1. Let (X, dX ) and (Y, dY ) be metric spaces. A function f : X → Y is uniformly


continuous if for every  > 0 there exists δ > 0 such that for all x, x0 ∈ X with dX (x, x0 ) < δ
we have dY (f (x), f (x0 )) < .
Show that if X is compact, then every continuous function f : X → Y is uniformly
continuous.

Solution. Denote by B(x, r) the ball of radius r centered at x (in whatever metric space x
belongs to).
Let  > 0. Since f is continuous, for each x ∈ X there exists δx > 0 such that f (B(x, δx )) ⊂
B(f (x), /2).
Balls of the form B(x, δx /2) give an open cover of X. Select a finite subcover given by
{B(x1 , δx1 /2), . . . , B(xn , δxn /2)}. Let δ = 12 mini δxi .
Suppose x, x0 ∈ X satisfy dX (x, x0 ) < δ. Then there is some k with 1 ≤ k ≤ n such that
dX (xk , x) < δxk /2, and therefore
dX (xk , x0 ) ≤ d(xk , x) + d(x, x0 ) < δxk /2 + δ ≤ δxk
where in the last inequality we used δ ≤ δxk /2. Thus both x and x0 lie in B(xk , δxk ), and so
f (x), f (x0 ) ∈ B(f (xk ), /2). By the triangle inequality this implies d(f (x), f (x0 )) < .
This shows that f is uniformly continuous.

T2. Consider topology on X = [0, 1]Z given by the metric


d(x, y) = sup |x(n) − y(n)|.
n∈Z

Let E = (0, 1)Z .


(1) Is E open in X?
(2) Is E dense in X?
In each case, give a proof of your answer.

T3. Let X be a topological space and {Ai }∞ i=0 a sequence of subsets of X. Suppose that
each Ai is connected, and for all i we have Ai ∩ Ai+1 6= ∅. Prove that the set A = ∪i Ai is
connected.

Solution. Suppose for contradiction that A is disconnected. Then there exists a separation
of A, i.e. open sets B, C with A ⊂ B ∪ C and such that the sets A ∩ B, A ∩ C are disjoint
and nonempty. Exchanging the roles of B and C if necessary, we can assume without loss
of generality that A0 ∩ B 6= ∅.
Since Ai ⊂ A, we have Ai ⊂ B ∪ C. This cannot be a separation of Ai , so one of Ai ∩ B
or Ai ∩ C is empty. We conclude:
(∗) For each i we have either Ai ⊂ B or Ai ⊂ C (and not both).
Since A0 ∩ B 6= ∅, it follows from (∗) that A0 ⊂ B.
Next, if Ai ⊂ B, then we have ∅ = 6 Ai ∩ Ai+1 ⊂ B ∩ Ai+1 and applying (∗) gives Ai+1 ⊂ B.
14
S
Thus by induction we have Ai ⊂ B for all i, and A = i Ai ⊂ B. But then A ∩ C = ∅, a
contradiction.

15

You might also like