Sie sind auf Seite 1von 56

Notes on Abstract Algebra

Daniel C. Bastos
dbastos@math.utoledo.edu
Jan 10, 2006
Abstract
These are my notes on abstract algebra; if you see something
wrong, let me know, please. If you found this document with the
help of search engines and youre studying about algebra, I suggest
you look for a book on the subject.
1 Groups
Denition 1.1 (binary operation). A binary operation * on a set S
is a function of S S S. That is, for each pair (a, b) S S, there
is a single element s S such that a b = s. This implies S closed
under *.
Denition 1.2 (group). A group G is a nonempty set with a binary
operation * satisfying: (i) * is associative, (ii) G has an identity e, (iii)
for each g G, there is a g
1
such that gg
1
= g
1
g = e.
Exercise 1.1. Are the even positive integers closed under addition?
Yes. Proof. Because 2n + 2n = 4n is even and positive. Are they
closed under multiplication? Yes. Proof. Because 2n 2n = 4n
2
is
even and positive.
Exercise 1.2. For rectangles, there corresponds to each pair (length
measurement, width measurement), an area measurement. Still, com-
puting the areas is not, strictly speaking, a binary operation. Why?
Because area is something dierent than length.
1
Exercise 1.3. Are the odd positive integers closed under addition?
Yes. Proof. Because 2n+1+2n+1 = 4n+1 is odd and positive. Under
multiplication? Yes. Proof. Because (2n + 1)(2n + 1) = 4n
2
+ 4n + 1
is odd and positive.
Theorem 1.1. Let r be a nonzero integer. Let D be the set of all
divisors of r. The system of all fractions, in lowest terms, with the
property that their denominators is d
i
D, is closed under addition.
Proof. Let 0 ,= r Z. Let D be the set of divisors of r. Let d
1
=
ar, d
2
= br. Then,
p
d
1
+
q
d
2
=
ap +bq
lcd(d
1
, d
2
)
,
where lcd(d
1
, d
2
) D and p, q Z.
Exercise 1.4. Show that the system of fractions having denominators
of 1, 2, or 4 when written in lowest terms is closed under addition.
Proof. By Theorem 1.1, it is closed. Here, r = 4 and D = 1, 2, 4
which are all divisors of 4.
Theorem 1.2. The identity of a group is unique.
Proof. Let e and e

be identities. So, ee

= e

because e is an identity.
Also, ee

= e because e

is an identity as well. Since the binary


operation of a group is a function, and assigns the operation to a
single element in the group, there can only be one identity and so,
e = e

.
Exercise 1.5. Consider Z with the binary operation of subtraction.
Is Z a group? If so, what is the identity? If not, what axioms do not
hold? At rst I thought that 0 could be the identity, because z0 = z,
but for 0 to be an identity, it must be true that z 0 = 0 z = z.
That is not true because z 0 = z ,= 0 z = z. Also, notice that
subtraction is not associative: (a b) c ,= a (b c).
Theorem 1.3. Let G be nite with order 2n, n N. Then, theres
g G such that g
2
= e.
Proof. Let S = g
1
, ..., g
2n
. Step one: remove from S all pairs
(g
1
i
, g
i
) such that g
i
g
1
i
= g
1
i
g
i
= e with g
i
,= g
1
i
. This removes an
even number of elements. Were left with an even number of elements
in S. After step one, e S because e
1
= e. Step two: remove e
2
from S. Were left with an odd number of elements in S. So at least
one of these left will have to be its own inverse since they come in
pairs.
Theorem 1.4. Let G = 1, 2, ..., p 1, p prime. Let be multipli-
cation mod p on G. Then (G, ) is a group.
Proof. Unfortunately, I havent proved this entirely; but heres what
I can say: because p is prime, we have no g
1
g
2
G such that g
1
g
2
divides p, so it never happens that g
i
g
j
= 0 mod p. Multiplication
is a binary operation, but we would need to show closure with mod
p. The identity is always present by hypothesis. Multiplication is
associative, but we would need an argument for inverses.
Theorem 1.5. The multiplication table of a group is a latin square;
that is, each group element occurs once and only once in each row of
the body of the table and in each column of the body of the table.
Proof. Consider the elements of the ith row. They are g
i
g
1
, ..., g
i
g
j
.
The theorem claims that each element occurs once; so, suppose one
does not show up once; that is, two elements of the row must be
equal. In other words, suppose g
i
g
1
= g
i
g
2
, but notice that g
1
,= g
2
.
Then, since were dealing with a group, g
1
i
is present; so, g
1
i
(g
i
g
1
) =
g
1
i
(g
i
g
2
) = (g
1
i
g
i
)g
1
= (g
1
i
g
i
)g
2
= eg
1
= eg
2
= g
1
= g
2
.
So, each element occurs once.
If the group is nite, were done. If the group is not nite, we must still
show that each element still occurs on a row. To prove for columns,
we can repeat the proof replacing the appropriate subscripts.
Denition 1.3 (permutation). A permutation P of a set S is a one-
to-one mapping of S into itself. In other words, P implies the existence
of a function f : S S such that f is one-to-one and onto.
Exercise 1.6. Write the 4! = 24 possible arrangements of all four
letters A, E, P, T. Check which permutations are English words. What
is the probability of making an English word from these four letters by
chance arrangement? Im not going to write them all here, but I will
tell the algorithm Id use. Fix A and permute EPT; then x E and
permute APT, then x P and permute AET; then x T and permute
APT. This will generate 3! + 3! + 3! + 3! = 24 permutations, and its
easy to do it this way. To permute EPT, the same can be done: x
E, permute PT, x P permute ET, x T permute EP.
3
It turns out, we nd the English words PATE, PETA, PEAT, TAPE,
TEPA. They mean, respectively: the human head, 10
15
, carbonized
vegetable matter, tape, a chemical compound: C
6
H
12
N
2
OP. So, we
have 5 chances out of 24 of making a word by chance arrangement;
thats a probability of 5/24 = 0.208 which is aproximatelly 20% of
chances of making a word out of random permutation; a pretty good
chance, if you will.
Exercise 1.7. Let Mr. 1, Mrs 2, Mr. 3 and Mrs 4 be seated on a table.
How many seating arrangements are possible? 4! = 24. Write down all
seating arrangements in which men and women are alternated. There
are 8 possibilities because once you x 1, there are two possibilities
to follow: 2 and 4. Fix 2, there are two possiblities: 1 and 3. Fix 3,
there are two possibilities; x 4, there are two possibilities. Whats
the probabilitity that a random permutation of the four people will
alternate men and women? Thatd be 8/24 = 1/3 which means 33%.
Theorem 1.6. The permutations of a set S form a group under com-
position.
Proof. Let P and Q be two permutations; then, for any s S, there
is a unique Q(s) S because for any s
1
, s
2
S, if Q(s
1
) = Q(s
2
),
then s
1
= s
2
. The fact that Q(s) S comes from the fact that Q is a
function from S into S. Also, because P is one-to-one and because Q
is onto, for all s S, there is a unique P(Q(s)) S.
Last paragraph implies P(Q(s)) is one-to-one. Now, to show compo-
sition is a binary operation, we need to show that the composition of
P and Q is onto; because that would be showing that it is closed in
S.
For all s S, there is q S such that q = Q(s). For all q S, there
is p S such that p = P(q). Then, for all s S, there is a q S such
that q = P(Q(s)) = (P Q)(s).
To show it forms a group, we state that composition is associative by
some theorem which we havent proved yet. The identity would
be the permutation which maps each s S into itself. The inverses
of each permutation is the same permutation with domain and range
swaped.
Denition 1.4. If S is nite with n elements, then the set of all
permutations on S is called the symmetric group S
n
on n elements.
S
n
has n! members.
4
Exercise 1.8. Let
P =

1 2 3 4 5
2 3 1 4 5

What is the image of 3 under the permutation? Its 1. Which elements


are xed under the permutation? 4 and 5. Which are moved? 1, 2, 3.
Write the inverse permutation with respect to composition.
P
1
=

2 3 1 4 5
1 2 3 4 5

.
Exercise 1.9. Let
Q =

1 2 3 4 5
1 4 3 5 2

.
Find P Q and Q P. Find Q
1
.
P Q =

1 2 3 4 5
2 4 1 5 3

Q P =

1 2 3 4 5
4 3 1 5 3

Q
1
=

1 4 3 5 2
1 2 3 4 5

.
Exercise 1.10. Find Q
1
P
1
. Find (P Q) (Q
1
P
1
).
Q
1
P
1
=

2 3 1 4 5
1 5 3 2 4

(P Q) (Q
1
P
1
) =

2 3 1 4 5
2 3 1 4 5

5
Permutations can be written in cycle notation, which is easier to write.
For the permutation P above, we would write (123)(4)(5) meaning 1
goes to 2, 2 goes to 3, 3 goes to one and we close. We close because it
has cycled. The same happens to 4 and 5 because they didnt move
at all; we may as well omit the ones which dont move.
In general, start the rst cycle with the rst element of the permuta-
tion; go on writing which follows which until the rst element appears;
then we close the cycle. If the rst cycle has not exhausted all ele-
ments, start a new cycle with an element that has been used. Repeat
until all elements have been used.
If we omit the 1-cycles, then someone who only sees the permutation in
cycle notation would not know possible elements which were omited.
But usually this is not a problem.
Cycle multiplication is cycle composition. You apply the right side
rst, then the left. For example,
(1234)(2465) =

2 4 6 5 3 1
1 6 5 3 4 2

,
because 2 goes to 4 which goes to 1, 4 goes to 6 which goes to 6 (since
it does not appear), 6 goes to 5 which goes to 5 (since it does not
appear), 5 goes to 2 which goes to 3, 1 goes to 1 which does to 2, 3
goes to 3 which goes to 4.
Exercise 1.11. Q used previously in cycle notation is (245).
Exercise 1.12. Write the permutations of S
3
as cycles. S
3
is the sym-
metric group of all permutations on n = 3 elements. Then, S
3
has 3! =
6 elements; they are: e, (12), (13), (23), (123), (132). Its a bit tough
to be sure that this is correct; but I built the table and it does work.
For example, (12)
1
= (12), (13)
1
= (13), (23)
1
, (23), (123)
1
=
(132), (132)
1
= (123). Its not commutative, though: (12)(13) =
(132) ,= (123) = (13)(12).
Exercise 1.13. Write the following cycles in permutation form.
(134) =

1 3 4 2
3 4 1 2

,
(12)(34) =

3 4 1 2
4 3 2 1

,
6
(1423) =

1 4 2 3
4 2 3 1

,
(13)(12) =

1 2 3 4
2 3 1 4

,
(14)(13)(12) =

1 2 3 4
2 3 4 1

,
Denition 1.5. A transposition is a 2-cycle.
Theorem 1.7. Let G be a group in which every element g G has
the property that g
2
= e. Then G is abelian.
Proof. Let a, b G. Then: (ab)
2
= e = a
2
b(ab) = a = b(ab) =
a = b
2
(ab) = ba = ab = ba.
Denition 1.6 (group isomorphism). Let (G, !) and (H, ) be groups.
We say G is isomorphic to H if and only if there exists a bijective
function f : G H such that for all u, v G, its true that f(u!v) =
f(u) f(v). We write H

= G.
Theorem 1.8. If m n, there is a subgroup of S
n
isomorphic to S
m
.
Proof. Since m n, the set of all permutations of S
m
is a subset
of S
n
, and S
m
is closed because it is a group, so it is a subgroup of
S
n
; then, one isomorphism exists between S
m
and S
m
: the identity
map.
Theorem 1.9. The identity of a subgroup H of G is necessarily the
identity of the group G.
Proof. Let H be a subgroup G; for all h H, e
H
h = he
H
= h. Since
every element of H is also in G, we have e
G
h = he
G
= h, for all h H.
So, e
G
h = e
H
h = e
G
= e
H
by cancellation law.
Theorem 1.10. Inverses are unique.
Proof. Let h G. Now h
1
1
h = e and h
1
2
h = e = h
1
1
= h
1
2
.
7
Exercise 1.14. Find a collection of eight permutations in S
4
that is
a subgroup isomorphic to the dihedral group D
4
= e, r, r
2
, r
3
, s, sr,
sr
2
, sr
3
. Give an explicit isomorphism from D
4
to the subgroup of
S
4
.
Let P
4
= e, (1234), (13)(24), (1432), (12)(34), (14)(23), (13), (24).
Map e
P
with e, (1234) with r, (13)(24) with r
2
, (1432) with r
3
,
(12)(34) with s, (13) with sr, (14)(23) with sr
2
, (24) with sr
3
.
Theorem 1.11. GL(n, R) denotes the group of invertible nn matri-
ces with real entries. The collection H of matrices with determinants
1 forms a subgroup of GL
n
(R).
Proof. We only need to show that H is closed; that is, det(AB) =
1 = det(A)det(B), for any A, B H. Let E be an elementary
matrix; det(E) = 1 whenever E comes from I by interchanging two
rows; det(E) = 1 whenever E comes from I by adding a row to a
multiple of another row of I; every invertible matrix can be expressed
as a product of elementary matrices.
Let P H; express P as E
k
...E
1
. For any E
i
, with 1 i k,
det(E
i
) = 1, then det(P) = det(E
k
...E
1
) = det(E
k
)...det(E
1
) =
1. So, for two matrices P, Q H, det(PQ) = det(E
k
...E
1
Q) =
det(E
k
)...det(E
1
)det(Q) = det(P)det(Q) = 1. Then H is a subgroup
of GL(n, R).
Okay, this proof is not correct because we must also show that the
group is closed under inverses; the group is innite. Also, det(AB) =
det(A) det(B) for any square matrices A and B, so we need not use
elementary matrices here.
On this next example, theres a new notation there. When we write
(1) = 2, we mean that its a cycle in which 1 always goes to 2. So
the set described includes all permutations of S
6
in which 1 goes to 2.
Examples: (12), (123), (1234), (12345), and so on.
Exercise 1.15. Let C = S
6
: (1) = 2. Is C S
6
? No,
because (1256)(1256) = (15...) ,= (1) = 2.
Denition 1.7. Let H G. Dene as left cosets of H in G the sets
gH = gh
1
, gh
2
, ..., gh
k
, ... where g G and h
i
H.
As you pick dierent elements g G, you create [G[/[H[ cosets. This
is called the index of H in G.
8
Denition 1.8. Let H G. Generate all left cosets of H in G.
Count the number of left cosets; call this number the index of H in
G.
These cosets are mutually exclusive and exhaustive, like equivalence
classes; in fact, each coset is an equivalence class. Cosets, however,
can have many names; but since theyre equivalence classes, one way
to know if two names are actually the same coset is to check if their
gs are left equivalent modulo H. See the next example.
Exercise 1.16. Let V = e, (12)(34),(13)(24), (14)(23). Left cosets:
(12)V = (12), (34), (1324), (1423), (13)V = (13), (1234), (24),
(1432), (14)V = (14), (1243), (1342), (23), (123)V = (123), (134),
(243), (142), (124)V = (124), (143), (132), (234), (12)(34)V =
(12)(34), e, (14)(23), (13)(24). What are the right cosets?.
Let W = e, (12), (34), (12)(34). Left cosets: (12)W = (12), e,
(34)(12), (34), (13)W = (13), (123), (134), (1234), (14)W = (14),
(124), (143), (1243), (23)W = (23), (132), (1342), (234), (24)W =
(24), (142), (1432), (243), (1324)W = (1324), (14)(23), (13)(24),
(1423). What are the right cosets?
Take a look at the rst coset of V (12)V . Now, notice that the
coset (1423)V is the same coset, though with a dierent name, as
(12)V . Why? Because (1423) is left congruent to (12) modulo H.
Why? Recal the denition of the left congruence. Two elements, i
and j, of a group P are left congruent (mod Q P) to each other if
theres q Q such that i = jq. So notice that the multiplication by q
happens on the right. In the example above, the subgroup is V , and
(12) is left congruent to (1423) because (12) = (1423)[(14)(23)] and
(14)(23) V .
Denition 1.9. Let H G. Dene g
i

L
g
j
(mod H) if and only
if there is a h
k
H such that g
i
= g
j
h
k
. Call this relation left
congruence.
Theorem 1.12. Left congruence is an equivalence relation.
Denition 1.10. Let H G. Dene g
i

R
g
j
(mod H) if and only
if there is a h
k
H such that g
i
= h
k
g
j
. Call this relation right
congruence.
Theorem 1.13. Right congruence is an equivalence relation.
9
Proof. Let H be a subgroup of G. Suppose that for some g
1
, g
2
G,
there is h H such that g
1
= hg
2
; that is, g
1
g
2
(mod H). Then
g g because g = e
H
g (reexive). Suppose g
1
= hg
2
; then g
2
=
h
1
g
1
(symmetric). Suppose g
1
= h
1
g
2
and g
2
= h
2
g
3
. Then g
1
=
(h
1
h
2
)g
3
implying g
1
g
3
(mod H) (transitive). It is an equivalence
relation.
Theorem 1.14. In a group G, g, g
1
G have the same order.
Proof. Let g
t
= e. Then (g
1
)
t
g
t
= (g
1
)
t
e which implies e = (g
1
)
t
.
Now we must show that t is the smallest such number in which
(g
1
)
t
= e. Suppose not; then, there is n < t such that (g
1
)
n
= e. Then, write (g
1
)
t
as g
1
multiplied t times: (g
1
)
1
...(g
1
)
t
.
Since n < t, we have that (g
1
)
1
...(g
1
)
n
= e. So, (g
1
)
1
...(g
1
)
t
=
e(g
1
)
n+1
...(g
1
)
t
. Now what? Where is the contradiction?
Denition 1.11 (cyclic). Call a cyclic group the group generated
by some element g, and write g). If g) is nite, then call the order
of g the least positive element f for which g
f
= e, where e is the
identity of the group. Call an innite cyclic group the cyclic group
in which g
s
= g
t
if and only if s = t.
Notice that the distinct elements of g) are g, g
2
, ..., g
f
= e. Also,
notice that the powers are not always positive. For example, 1 is a
generator for the integers under addition because 1
0
= 0, 1
1
= 1, 1
2
= 2, 1
3
= 3, and so on. But where are the inverses? Well, we do 1
1
= 1, 1
2
= 2, and so on.
Theorem 1.15. Let G be nite and cyclic. Then G is abelian.
Proof. Because G is nite, it has a nite order; call it n. Any element
of G is expressed g
k
, where 1 k n. Let g
i
, g
j
g). Now g
j
g
i
=
g
j+i
= g
i+j
= g
i
g
j
.
Exercise 1.17. Write, in cycle form, the members of the cyclic group
of order 8 and compare with the octic group. Are both groups abelian?
No; a cyclic group of order 8 is abelian, by the last theorem, but
D
4
is non-abelian. The cycles of the cyclic group of order 8 are e,
(12345678), (1357)(2468), (14725836), (15)(26)(37)(48), (16385274),
(1753)(2864), (18765432).
Theorem 1.16. Let G be nite with g G. Order of g divides [G[.
10
Proof. The order of g is the natural number n such that g
n
= e;
since G is nite, [G[ and n are natural numbers.
Suppose f is a natural number such that g
f
= e; since n is the smallest
such number, f n. By the division algorithm, f = nq +r, where q,
r Z with r < n. So g
f
= g
nq+r
= (g
n
)
q
g
r
= e
q
g
r
= eg
r
= g
r
= e.
But r < n by the division algorithm, and since n is the smallest natural
such that g
n
= e, r must be zero, so n divides f.
The distinct powers of g form a subgroup of G, because theyre closed
under the binary operation. Lagranges theorem states that, here in
particular, [G[ = [g)[ [G : g)[, and [g)[ = n, so n divides [G[ with
[G : g)[ as quotient.
Exercise 1.18. Write down all elements in D
4
, S
4
and S
5
which have
order 2. In D
4
, with order 2, we have: (13), (24), (13)(24), (14)(23),
(12)(34). In S
4
we have (12), (13), (14), (23), (24), (34), (12)(34),
(13)(24), (14)(23). In S
5
we have (12), (13), (14), (15), (23), (24),
(25), (34), (35), (45), (12)(34), (12)(35), (12)(45), (13)(24), (13)(25),
(13)(45), (14)(23), (14)(25), (14)(35), (15)(23), (15)(24), (15)(34),
(23)(45), (24)(35), (25)(34).
Exercise 1.19. We know that S
n
can be generated by transpositions;
that is, every S
n
can be written as a product of transpositions.
It is actually the case that S
n
can be generated by adjacent trans-
positions; that is, those of the form (i, i + 1). Prove this for n = 5
by expressing each (a,b) S
5
as a product of adjacent transpositions.
Proof. (12) = (12), (13) = (12)(23)(12), (14) = (12)(23)(34)(23)(12),
(15) = (12)(23)(34)(45)(34)(23)(12), (23) = (23), (24) = (23)(34)(23),
(25) = (23)(34)(45)(34)(23), (34) = (34), (35) = (34)(45)(34), (45) =
(45).
Exercise 1.20. Let H be a subgroup of a nite group G. If H is
abelian, then H is normal in G. Show this is not true with an example.
Consider the dihedral group with its subgroup V = 1, (12)(34). Let
W = e, (12), (34), (12)(34) S
4
, which is abelian. Now the right
coset W (1234) = (1234), (234), ... ,= (1234) W = (1234), (134),
(123), (13...).
Exercise 1.21. Let K = 1, (12)(34), (14)(23), (13)(24) D
4
= G.
Show that V = 1, (12)(34) is normal in K and that K is normal in
G, but V is not normal in G. We list the cosets; the left cosets are
eV = V and (14)(23)V = (14)(23), (13)(24) and the right cosets
11
are V e = V and V = (14)(23), (13)(24), so V normal in K. Now
the left cosets of K in G are eK = K and (1234)K = (1234), (13),
(24), (1432); the right cosets of K in G are Ke = K and K(1234)
= (1234)K because there are four elements left to be taken, so K is
normal in G. Now V is not normal in G because (1234)V = (1234),
(13) ,= V (1234) = (1234), (24).
Theorem 1.17. The intersection of two subgroups is a group.
Proof. Let H
1
, H
2
G, then e H
1
H
2
because e
1,2
H
1,2
is the
e of G, so its the same identity. Let h H
1
H
2
; because H
1,2
is
a group, h
1
H
1,2
. Associativity comes from the fact that G is
associative. Closure: let h, g H
1
H
2
. Since h, g H
1
and H
1
is
a group, it is closed, then hg H
1
. Similarly, hg H
2
. Therefore,
hg H
1
H
2
.
Exercise 1.22. Let N be a normal subgroup of a group G. Prove that
if n is an element of N, then g
1
ng is an element of N for each element
g G. We start by letting n N. Since N is normal subgroup of
G, we have G/N as the group of cosets of N in G. So, gN G/H;
since n N, nN G/H and g
1
N G/H. Since G/H is a group,
g
1
NnNgN = g
1
ngN. Since n N, it follows that nN = eN = N.
Then g
1
NnNgN = g
1
NeNgN = g
1
egN = g
1
gN = eN = N.
Therefore, g
1
ngN = N which implies g
1
ng N.
Exercise 1.23. Show that the converse of Lagranges Theorem is
false. For example, show that A
4
is a group with 12 elements that con-
tains no subgroup with 6 elements. We start by listing the twelve per-
mutations of A
4
: e, (12)(34), (13)(24), (14)(23), (123), (132), (124),
(142), (134), (143), (234), (243). There are 3 2-cycles, and 1 identity
which amounts to 4 elements, so to get to 6, we will need 3-cycles.
Theorem 1.18. A
4
has no subgroup with 6 elements.
Proof. Let H A
4
such that [H[ = 6. We want to show a contra-
diction. We know that H A
4
because the index is 2. H contains
a 3-cycle, for sure. Now, ghg
1
H for all g A
4
, for some h H.
Lets say that (123) H. Then (124)(123)(142) H which implies
(243) H. Then (243) = g (123) g
1
. Lets say (234) H, then
(234)(123)(243) = (134). Then (134) = g

(123)g
1
. If (123), (134),
(243) H, then we can square each of them because H is a subgroup.
We square the elements until we actually see that we dont get only 6
elements, which we assumed we had.
12
Theorem 1.19. The intersection of two normal subgroups is normal.
Proof. Let subgroups H
1
, H
2
G. Then H
1
H
2
is a subgroup by
Theorem 1.17; so must only show normality. Let g G; we must show
g(H
1
H
2
) = (H
1
H
2
)g, which is true if ghg
1
(H
1
H
2
) for some
h (H
1
H
2
) see some example above. So, we have that gh = h

g
for some h

(H
1
H
2
). We multiply g
1
on the right to get ghg
1
= h

e = h

for some h

(H
1
H
2
). So, ghg
1
(H
1
H
2
) and were
done.
Denition 1.12. The center Z(G) of a group is the set of all elements
of G which commute with every element of G. That is, Z(G) = z G
such that zg = gz for all g G.
Denition 1.13. A subgroup H of G is a normal subgroup of G if
gH = Hg for all g G. We write H G.
Theorem 1.20. Z(G) G.
Proof. Associativity comes from the fact that G is a group. The iden-
tity is in Z(G) because eg = ge = g for all g G. Let z Z(G). Now,
since e Z(G) and e =zz
1
, we have (zz
1
)g = (z
1
z)g = z(z
1
g)
= (gz)z
1
. Since z Z(G), it commutes, so z(z
1
g) = (zg)z
1
. By
cancellation law and associativity, z
1
g = gz
1
, so z
1
Z(G). Now,
normality comes from the fact that gz = zg for all g G, then gZ(G)
= Z(G)g.
To understand normality, we must understand cosets well. Lets think
of cosets for a while. Consider a subgroup H G. What can we say
about the cosets of H? We can say that [gH[ = [H[; that is, the
group formed by the cosets of H has the same order as H. In fact,
such group has a name. Is it really a group? It is, as we prove on the
next theorem, and it also has a name.
Denition 1.14. The cosets of H in G is the quotient group G/H.
Theorem 1.21. The set of left cosets of H in G form a group under
the operation dened as (g
1
H)(g
2
H) = g
1
g
2
H whenever gH = Hg for
all g G.
Proof. This theorem denes the binary operation in which the group
is formed upon. This operation can only be used whenever some
condition about the left and right cosets holds; that is, that the left
13
cosets match the right cosets exactly. In other words, H must be
normal in G.
The denition of the binary operation gives us the fact that (g
1
g
2
H) is
a coset. Do we have an identity? Yes, because g
1
HeH = g
1
eH = g
1
H.
So, H is the identity. Is it associative? Since (g
1
g
2
)g
3
H = g
1
(g
2
g
3
)H,
then (g
1
Hg
2
H)g
3
H = g
1
H(g
2
Hg
3
H). So yes, it is associative. Do
we have inverses? Yes, because (gH)
1
= g
1
H; the proof of that is
(gH)(gH)
1
= gg
1
H = eH = H. So we have inverses, and we have
a group.
Because cosets have many names, it would be nice to have theorems
that would tell us when theyre equal. For instance, if I show you
g
1
H and g
2
H, such that g
1
,= g
2
, does that mean that the cosets are
dierent? Not necessarily. So, how can we tell?
One way, of course, is to take a look at the contents. Since cosets
are sets, if what they have as elements are all the same, in any order,
theyre equal.
Theorem 1.22. Let H G; now g
1
H =g
2
H if and only if g
1
2
g
1
H.
Proof. Because cosets are equivalence classes, g
1
H = g
2
H if g
1

L
g
2
(mod H) were dealing with left cosets here. If this holds, then its
because there is an h H such that g
1
= g
2
h. Now, because G is
a group, we have the inverse of g
2
available; multiply it on the left:
g
1
2
g
1
= eh = h. Then, if g
1
H = g
2
H, then g
1
2
g
1
H.
Now we must show that once g
1
2
g
1
H, it follows that g
1
H = g
2
H.
Well, because g
1
2
g
1
H, we can write g
1
2
g
1
= h, for some h H.
Then, multiply g
2
on the right, so eg
1
= g
2
h = g
1
= g
2
h. Now, this
means that theres an h H, such that g
1
= g
2
h. Therefore, g
1

L
g
2
(mod H). That is, g
1
H = g
2
H.
Theorem 1.23. Let H G; now g
1
H =g
2
H if and only if g
1
1
g
2
H.
Proof. Repeat last proof changing subscripts appropriately.
So, these two theorems give us a nice way to check if cosets are equal.
Next we prove a theorem about normal subgroups. It says that a nor-
mal subgroup of a group is normal only when an element of the group
times an element of the subgroup times the inverse of the element of
the group is in the subgroup. Verbal mathematics is very hard to
understand, but its a good exercise.
Theorem 1.24. A subgroup H of G is normal in G if and only if
ghg
1
is in H, for all g in G.
14
Proof. Suppose H is normal in G. Let h in H and g in G be arbitrary
elements. Then, gh = h

g for some h

in H. Now multiply by g
1
on
the right to get ghg
1
= h

which is in H.
By hypothesis, ghg
1
is in H, so let ghg
1
= h

be in H. Multiply by
g on the right, so gh = h

g, which means that gh is in Hg, and since


h is an arbitrary element of H, gH is a subset of Hg.
Now since g
1
and g are inverses of each other, we can write the
hypothesis as g
1
hg = h

. Now multiply by g on the left to get hg =


gh

which means that hg is in gH and since h is an arbitrary element


of H, Hg is a subset of gH. So, gH = Hg, for all g in G.
Now, if you read this theorem carefully, you may notice that it gives
you a nice message. The message is, if G is abelian, then any subgroup
H of G is normal. How does it say that? Well, suppose G is abelian
and that H G. Then for any g G, gg
1
= e H, right? Now,
multiply h on the right to get gg
1
h = eh. But, since every element
of G commutes with every element of G, including those of H because
H G, we can alter their order; so gg
1
h = eh = ghg
1
= eh =
h H and whenever this happens, H G.
Now, how do we actually build a quotient group? Lets look at an
example. Let G = (Z, +), H = nZ. Because addition is the binary
operation for this group, mathematicians like to write the cosets in
the form g + H to make explicit that the operation is addition. If it
were any other operation, theyd probably write gH as we have been
so far.
Lets try to build the quotient group Z/nZ, for n = 3, so well build
Z/3Z. The elements of Z/3Z are 3Z, 1+3Z, 2+3Z. If youre asking
why, consider the integers mod 3. By the division algorithm, z 3 =
nq + r, where r < 3. So r is either 0, 1 or 2. That explains why the
members are as they are. Now, we set up the Cayley table.
3Z 1 + 3Z 2 + 3Z
3Z 3Z 1 + 3Z 2 + 3Z
1 + 3Z 1 + 3Z 2 + 3Z 3Z
2 + 3Z 2 + 3Z 3Z 1 + 3Z
You may quickly notice that this group is abelian. How would one
quickly notice such thing? Think of the group as a matrix A. Now,
write A
T
and notice that they are equal. The fact that this group is
abelian may get us a new theorem. Once a group G is abelian, is it
true that G/H is also abelian? Yes; if G is abelian, every subgroup is
normal, and so G/H is also abelian.
15
Theorem 1.25. Let H G, with G abelian; then G/H is abelian.
Proof. Let g
1
H, g
2
H G/H. Now (g
1
H)(g
2
H) = (g
1
g
2
)H, right?
Thats the denition for the multiplication (the binary operation) of
a quotient group. But, because G is abelian, every g commutes with
every element of G, so (g
1
g
2
)H = g
2
g
1
H = g
2
Hg
1
H and this shows
that G/H is abelian.
Once I say H G, does it automatically follow that G is abelian? It
cant be, because let H = e which is normal in G and is a subgroup
of any G non-abelian. But what if H is a proper normal subgroup of
G? Does that make G abelian automatically? If so, can you prove? If
not, can you show an example?
But now, once G is abelian, every subgroup of G is normal, and every
subgroup of G is abelian; and once a subgroup is normal, a quotient
group with respect to the subgroup is also abelian. So, abelianism is
a pretty powerful and nice property that a group may have.
Since were talking about abelianism, notice that when G is cyclic,
then theres no way that a subgroup wont be abelian. Think about
it; theres just no way because all elements are expressed as a factor
of the generator, so it doesnt matter which order you multiply any of
them.
Exercise 1.24. Let H G. What is the order of gH G/H? Recall
that the order of gH is the smallest k such that (gH)
k
= H = g
k
H
which happens when g
k
H. Then, the order of gH is the order of
H; that is, [gH[ = [H[.
Theorem 1.26. Let [G : H] = 2. Then H G.
Proof. What do we have to show here? We have to show that once
the index of H is 2, the right cosets of H in G are the same as the
left cosets of H in G. Since the index of H in G is the number of left
cosets of H in G, there are 2 left cosets. Then, each gh is either in
H or in GH. Then we have two left cosets: g
1
H = H and g
2
H =
GH. Now, consider... I dont clearly see this one right now.
Lets talk about ways to produce groups quickly. Which groups do
we know? We know the number systems: (R, +), (R ,= 0, ), (Z, +),
(Q ,= 0, ), (Q, +).
We know how to generate some groups, like the dihedral group D
4
=
r,s : r
4
= s
2
= e, rs = sr
3
. And we can always write the Cayley
16
table for any group, but its very hard to check associativity for crazy
groups in which we may build a table for.
Associativity, then, makes is a hard enough thing to check so that we
may wish to build groups out of things which we already know will
be associtive. For example, any group under composition is associa-
tive, so we like composition as a binary operation. So, a group can
be formed out of invertible functions from a set S into itself under
composition; what kind of groups are those? These are the symmetric
groups S
n
; theyre built out of the symmetry of objects.
For example, think of a rectangle. What can we do with a rectangle?
We can ip it vertically and horizontally: we create two elements of
order 2. The identity is any of these two elements squared. We can
rotate it 180 degrees; which if we square this element, we also get
the identity. This is the symmetries of a rectangle in which we only
have elements which square to the identity; so each element is its own
inverse. This is the Klein 4-group.
Other ways to get away from checking associativity is to form groups
out of other groups. For example, let H G. Then, by using the same
operation of G, we may form a group (H, op(G)) in which associativity
will be given by G, at no charge a pretty good deal. Not to mention
that to check that H G we only need to check that for all h, h

H,
hh

H and h
1
H.
We can use generators. Pick some g G, and take a look at all
powers of g. This algorithm generates the group g) which is cyclic
and, therefore, abelian, which makes every possible subgroup normal.
For example, let G = SL
2
(R). This is the special linear group of
matrices 2 2, under matrix multiplication as binary operation, with
real entries such that the determinant of each of these matrices is 1.
Let
g =

1 1
0 1

,
and notice that g)

= (Z, +).
Another interesting example is the center of every group. Every group
G has a center as a subgroup. Moreover, every center subgroup is
abelian. Let Z(G) = z G : zg = gz for all g G, then Z(G) is an
abelian subgroup of G. Next, we show an interesting theorem about
centers.
17
Theorem 1.27. Suppose G/Z(G) is cyclic. Then G is abelian and so
Z(G) = G.
Proof. Since G/Z(G) is cyclic, we can let gZ(G) generate G/Z(G).
So G/Z(G) = ... g
1
Z(G), Z(G), gZ(G), g
2
Z(G), .... This means
that every element of G is in a coset g
a
Z(G). Now, pick g
1
, g
2
G;
these elements look like g
a
z and g
b
z

in which each one is in some coset


of Z(G) in G, so z, z

Z(G). Because they are in the center of Z,


they both commute with every other element of G; therefore, g
a
zg
b
z

= g
a
zz

g
b
= g
a
z

zg
b
= g
a
z

g
b
z which shows that every element of
G commutes with every other element of G, therefore G is abelian.
Now Z(G) = G because once G is abelian, every element satisfy the
condition to belong in the center. In other words, abelianism implies
global centralization.
Before we move on to rings and other algebras, lets take a look at
the quaternion group. The quaternion group is a non-abelian group
of order 8. It is usually denoted by Q
8
= 1, 1, i, i, j, j, k, k.
The quaternions are actually a number system, similarly constructed
to the complex numbers, but the quaternions dont commute, which
is unsual for number systems.
An interesting thing about the quaternion group is that once we say
that i
2
= j
2
= k
2
= ijk = 1, you can get the whole group. Theres
another way to name the elements of the group, in which I happen to
prefer right now.
Let Q
8
= 1, a, a
2
, a
3
, b, ab, a
2
b, a
3
b. Let a
4
= e, b
2
= a
2
, and so
b
4
= e. Then a
1
= a
3
and it follows that ba = a
1
b. From this, we
can draw the whole multiplication table.
1 a a
2
a
3
b ab a
2
b a
3
b
1 1 a a
2
a
3
b ab a
2
b a
3
b
a a a
2
a
3
1 ab a
2
b a
3
b b
a
2
a
2
a
3
1 a a
2
b a
3
b b ab
a
3
a
3
1 a a
2
a
3
b b ab a
2
b
b b a
3
b a
2
b ab a
2
a 1 a
3
ab ab a
2
b a
3
b b a
3
a
2
a 1
a
2
b a
2
b a
3
b b ab 1 a
3
a
2
a
a
3
b a
3
b b ab a
2
b a 1 a
3
a
2
They dont commute because, for example, ab ,= ba = a
3
b. However,
notice that all of its proper subgroups are abelian, because theyre all
cyclic. They are: 1, a), a
2
), b). So non-abelianism doesnt imply
non-abelian subgroups.
18
With a quick glance at the table and at the nature of the elements of
the quaternion, we could think that they have something to do with
the dihedral D
4
group, but they are not that similar. The quaternion
group has two elements which square to the identity, while D
4
has at
least ve, so D
4
is not isomorphic to Q
8
.
Exercise 1.25. Let G abelian, let H = g G : g
2
= e. Prove
H G. Give an example to show G abelian is necessary for such
H. Proof. We have associativity from G; e
2
= e, so we have identity.
We have closure on H because: since G is abelian, and H is subset of
G, all elements of H also commute, then e
2
= h
2
h
2
= (hh

)(hh

) =
(hh

)
2
H. We also have inverses, because for each h H, h
1
=
h H. Then H G. For the example, we consider S
4
which is not
abelian. Both (12) and (14) square to the identity. But (12)(14) =
(142), but (142)
2
= (124) ,= e, so we dont have closure. So abelianism
is necessary.
Exercise 1.26. Let G be a group and let a be one xed element
of G. Show that H
a
= x G : xa = ax is a subgroup of G.
Proof. Associativity is given by G. The identity is present because
it commutes with every element of G, so it will commute with a just
ne. For inverses now, let x H, so ax = xa, then x
1
ax = x
1
xa
= x
1
axx
1
= x
1
xax
1
= x
1
ae = eax
1
= x
1
a = ax
1
,
so x
1
H. Now we need to show closure; let x, y H
a
. We must
show xy H
a
; xa = ax = (yx)a = yax = yax = (yx)a =
a(yx) = (yx)a.
Exercise 1.27. Show that S
n
is non-abelian for any n 3. Proof.
Any S
n
, where n 3, has (12) and (23) as elements; but since (12)(23)
= (123) ,= (132) = (23)(12), then no S
n
, with n 3, is abelian.
Exercise 1.28. Give an example of a group G which has no elements
of nite order greater than 1, but has a factor group G/H, all of whose
elements have nite order. Let (Z,+) be the group which none of its
elements have nite order; thats possible because theres no power
t > 1 such that z
t
= 1. However, Z/2Z is such that all of its cosets
have order not greater than 2.
Exercise 1.29. Let H G and x g G. Dene gHg
1
= ghg
1
: h H. Prove (a) that gHg
1
G. It is called a conjugate of H.
Suppose G is nite; then prove (b) that [gHg
1
[ = [H[. Now suppose
19
G is a group and H is the unique subgroup of size [H[ in G; then,
prove (c) that H G.
For (a), we show rst the identity is present. Since e H, then geg
1
= gg
1
= e gHg
1
. Associativity comes from G because gHg
1
is composed only of elements of G. We construct the inverses this
way: e = gg
1
= geg
1
= ghh
1
g
1
= gheh
1
g
1
= ghg
1
gh
1
g
1
= (ghg
1
)(gh
1
g
1
) = e. And e = gg
1
= geg
1
= gh
1
hg
1
=
gh
1
ehg
1
= gh
1
g
1
ghg
1
= (gh
1
g
1
)(ghg
1
). So we have in-
verses; therefore, gHg
1
G.
For (b), let H = h
1
, h
2
, ..., h
k
with H G nite. Then
gHg
1
= gh
1
g
1
, ..., gh
k
g
1
.
Now, if I can show that all elements in gHg
1
are distinct, then Ive
shown that theres one and only one ghg
1
for each h H. Therefore,
they have the same number of elements, and therefore, the same order.
So, suppose gh

g
1
= gh

g
1
. By cancellation law, h

= h

and then
theyre all distinct, so they have the same order.
For (c), we have [gHg
1
[ = [H[ G, by (b) and (a) respectively.
The right coset (gHg
1
)g = gh
1
g
1
g, ..., gh
k
g
1
g = gh
1
, ..., gh
k

= gH. Now, since H is the unique subgroup with [H[ elements, then
gHg
1
= H, and so gH = Hg because Hg = (gHg
1
)g implying
H G. However, (b) assumes G is nite, but I think that gHg
1
will
always have the same elements even when G is not nite; so I think
that G-nite in (b) is not a requirement. Theres a not so hard way to
show the same cardinality here, so that we may drop the requirement
that G-nite.
2 Homomorphisms
Denition 2.1. A group homomorphism from (G, ) to (H, ) is a
mapping that maps every g G to some h H so that the image of
an -product is the corresponding ()-product in H; that is (g
1
g
2
)
= (g
1
) (g
2
).
Denition 2.2. Let : G H be a homomorphism. The set (G)
H of all images h H of elements g G is called the homomorphic
image of G. The various elements g sharing the common image (g)
= h form the pre-image of h. Also, we say is a homomorphism onto
H if the mapping is onto; that is, if (G) = H, so that every element
20
of H appears as the image of at least one g G. If is onto and
one-to-one, then is an isomorphism.
Denition 2.3. Let : G G

be a homomorphism. Let H G.
Then we call the set [H] = (h) : h H the homomorphic image
of H on G

.
Denition 2.4. Let : G G

be a homomorphism. The kernel of


is ker() = g G : f(g) = e = f
1
[e].
Theorem 2.1. [H] G

.
Denition 2.5. Let : G G

be a homomorphism. Let H

.
Then we call the set
1
[H

] = g G : (g) H

the pre-image of
H

.
Denition 2.6. Let be a homomorphism from a group G to a
group H. The group K of pre-images of e
H
is called the kernel of the
homomorphism.
Exercise 2.1. The integers can be mapped onto the integers Z
n
=
0, 1, 2, ..., n 1, and this mapping is a homomorphism for the
operation +. Is it also a homomorphism for ? Let f(i + rn) = i
(mod n), and show that f((i +rn)(j +sn)) = ij (mod n).
Proof. We let (p = i +rn) = i (mod n), (q = i +sn) = j (mod n).
Is (pq) = ij (mod n)? ((i +rn)(j +sn)) = (ij +isn+jrn+rsn
2
)
= (ij + (is + jr + rsn)n) = ij (mod n). So it is a homormorphism
for as well.
Exercise 2.2. Let : G H be a group homomorphism. Prove
that [G] is abelian if and only if for all x, y G we have xyx
1
y
1

ker().
Proof. Suppose xyx
1
y
1
ker() for all x, y G. Let a, b in [G],
so we have x, y G such that a = (x), b = (y). We use the s
properties, such as (x
1
) = (x)
1
to write
(xy)(x
1
y
1
) = e
(x)(y)(x
1
)(y
1
) = e
(x)(y)(x)
1
(y)
1
= e
(x)(y)(x)
1
= (y)
(x)(y) = (y)(x)
ab = ba
21
Therefore, [G] is abelian. Now suppose [G] is abelian. Let x, y G.
Must show (xyx
1
y
1
) = e. So
(x)(y) = (y)(x)
(x)(y)(x)
1
= (y)
(x)(y)(x)
1
(y)
1
= e
(x)(y)(x
1
)(y
1
) = e
(xyx
1
y
1
) = e
Therefore xyx
1
y
1
ker() for all x, y G.
Exercise 2.3. Let f : G H be a group homomorphism. Prove that
f is one-to-one if and only if ker(f) = e.
Proof. Assume f is one-to-one. Let x ker(f). Then f(x) = e =
f(e), and since f is one-to-one, we have that x = e, so ker(f) = e.
Now assume ker(f) = e. Let x, y G. Suppose f(x) = f(y). We
now use the fact that f is a homomorphism, to write
f(x) = f(y)
f(x)f(y)
1
= e
f(x)f(y
1
) = e
f(xy
1
) = e
So xy
1
ker(f). Therefore, x = y.
Exercise 2.4. Let G be a group and x g G. Dene a map :
G G by
(x) = gxg
1
.
Show that is (a) one-to-one and (b) onto. Prove is a (c) homo-
morphism.
For (a), let x, y G. Suppose (x) = (y). So (x) = gxg
1
= (y)
= gyg
1
. Apply cancellation law on the left and on the right; then x
= y. So is one-to-one.
For (b) let y G. We must show that there is an x G such that
y = (x) = gxg
1
. Multiply by g on the right, and by g
1
on the left,
so g
1
yg = x; so, is onto.
For (c), let x, y G. Then (x) = gxg
1
and (y) = gyg
1
. So
(x)(y) = gxg
1
gyg
1
= gxyg
1
= (xy)
22
So is a homomorphism.
Exercise 2.5. Let : G H be a homomorphism and K = ker(),
so K G. Let g G. Prove gK =
1
[(g)]. That is, we are proving
that the elements of the coset gK are exactly the elements of G that
map to (g) (g).
Proof. Let : G H be a (group) homomorphism, K = ker(). Let
K G. Let g G. Let gk gK for some k K. Must show gk

1
[(g)]. Since G is a group, gk G; since is a homomorphism,
(gk) = (g)(k) = (g) [G].
Since (gk) [G], gk
1
[(g)].
Now, since f is a homomorphism,
f(g
1
u) = f(g
1
)f(u)
= f(g)
1
f(u)
But f(u) = f(g) because u f
1
[f(g)]. So
f(g)
1
f(u) = f(g)
1
f(g)
= e
So g
1
u is mapped to e, which means g
1
u K. So g
1
u = k for
some k K. Multiply g on the right, so u = gk; so u gK.
Exercise 2.6. Let H G. Dene : G G/H by (g) = gH.
Show that is onto and the kernel of is H.
Proof. As dened, takes each g G into a coset gH G/H; since
H is normal, G/Hs operation is dened as (g
1
H)(g
2
H) = (g
1
g
2
)H.
By denition, (g) = gH, so
(g
1
g
2
) = g
1
g
2
H
= g
1
Hg
2
H
= (g
1
)(g
2
)
Therefore, is a homomorphism.
The kernel of will be all elements which land in the identity of G/H;
that is, they land in H. Which elements g will satisfy (g) = H? By
denition, (g) = gH which is equal to H if and only if (g
1
e) H.
23
Let h H. So (h) = hH = H because h
1
H. Therefore,
H = ker().
To show is onto, let gH G/H. I must nd g

G such that
(g

) = gH; because g G, we certainly have one; let g = g

, so is
onto.
Denition 2.7. A group G ,= e is simple if and only if for every
N G, we have N = G or N = e.
Exercise 2.7. Prove that a nite simple group G which is abelian
must be a cyclic group of prime order.
Proof. By denition, G ,= e, so pick g ,= e; we know g) G.
Suppose g) ,= G; since G is abelian, g) G, so g) is normal and
proper in G, but G is simple, so thats not possible. So, g) = G.
By Theorem 5-3, if some prime p divides [G[, then G has a subgroup
H of order p. Since G is abelian, H must be normal. Since G is
simple, its not true that H is proper, so [G[ = p.
Theorem 2.2. Suppose H is a subgroup of G and N is normal in G.
Dene HN a subset of G by HN = hn : h H, n N. Then HN
is a subgroup of G.
Proof. We know e is in both H and N, so e
2
= e HN; hence HN
is not empty.
For closure, let a, b be in HN. Then there are h, h

H and n, n

N
such that a = hn and b = h

, so ab = hnh

. Now, since N
is normal, hN = Nh for all h G, so nh

Nh

= h

N; hence,
nh

= h

for some n

N, so ab = hh

. Since hh

H and
n

N, ab HN.
For inverses, let a be in HN, so a = hn for some h H and n
N. We know a
1
= n
1
h
1
, so we multiply h on the left to get
ha
1
= hn
1
h
1
which is in N because N is normal. Since ha
1
N,
h
1
(ha
1
) HN, so a
1
HN; hence, HN is a subgroup of G.
Corollary 2.3. If H G, then HN G.
Proof. Let x in HN be arbitrary; then x = uv for some u in H and v
in N. Let g in G be arbitrary; we want to show that gxg
1
is in HN.
Now gxg
1
= guvg
1
= gug
1
gvg
1
. But H and N are normal in G,
so gug
1
is in H and gvg
1
is in N, so gxg
1
is in HN, as desired.
24
Exercise 2.8. Let G be a group. Dene on G by x y if there
exists g G such that
gxg
1
= y.
Prove (a) is an equivalence relation,(b) describe the conjugacy
classes of G if G is abelian, (c) write down the conjugacy classes of
S
3
.
Proof. Let x G. Then x x because xxx
1
= x. Suppose x y.
So exists g such that
gxg
1
= y
xg
1
= g
1
y
x = g
1
y(g
1
)
1
.
Suppose x y, y z. So there exists g, h such that
gxg
1
= y
hyh
1
= z
So h(gxg
1
)h
1
= z. Let g

= hg, and notice g


1
= g
1
h
1
. So
g

xg
1
= z. So x z.
(b) If G is abelian, then the only y which will satisfy the relation is
y = x, but since is an equivalence relation, breaks G apart into
disjoint and exhaustive classes, so G has [G[ conjugacy classes of the
form x.
(c) S
3
= e, (12), (23), (13), (123), (132). A 3-cycle has a 2-cycle
as a conjugate because 3-cycles are even permutations, 2-cycles are
odd, so even-cycle times odd-cycle is odd-cycle. Therefore, one class
C
1
will contain (123), (132). Now, by inspection I nd (12)(13)(12)
= (12)(123) = (23), so (13) (23). Also, (23)(12)(23) = (23)(123) =
(13), so (12) (13). So, another class C
2
will contain (12), (13), (23).
3 Polynomials
We need denitions, notation explained, the theorems, the proofs; the
exercises are already here.
Exercise 3.1. Find all primes p such that x + 2 is a factor of
x
4
+x
3
+x
2
x + 1 Z
p
[x].
Solution. If x +2 is a factor of x
4
+x
3
+x
2
x +1 Z
p
[x], then its
because 2 is a root. By evaluating the polynomial at 2, we get 15
25
as the result. Now, 15 = 0 (mod 3) and 15 = 0 (mod 5). Hence we
have p in 3, 5.
Exercise 3.2. Find all irreducible polynomials of degree 3 in Z
3
[x].
Solution. We can express all polynomials of degree 3 over Z
3
by
ax
3
+bx
2
+cx +d.
The constant term d cannot be zero, or zero is a root. The coecients
cannot sum to zero, or one is a root. Now, to derive a third rule,
evaluate x = 2 = 1: a+b +c +d = 0 which implies b +d = a+c.
That is, whenever the sum of the even coecients equals the sum of
odd coecients, we have 2 = 1 as a root. Since the degree must be
3, then a cannot be 0, so a is either 1 or 2.
Since factorization is unique except for scalar multiple (and order),
consider all monic polynomials and multiply each by 2 to notice that
factorization doesnt change. So we may consider only the monic poly-
nomials; in case we want to distinguish them, we may multiply each
irreducible by 2 to obtain its match, doubling the total of irreducible
polynomials.
So, with these criteria, we have 2 choices for d, 3 choices for c, 3 choices
for b, with 1 choice for a. This gives us 18 polynomials p(x). Suppose
now that p(x) is reducible; then its factors will be of degree 2 and
degree 1; that is,
p(x) = g(x)h(x) = (x
2
+Ax +B)(x +C).
Since d isnt zero, B and C are not zero. So, C = 1 or C = 2, B = 1 or
B = 2. This gives us 6 possibilities for g(x) and 2 possibilities for h(x);
that is, 12 factorizations, which are reducible. We take advantage of
the list to add the irreducibles in the order that they should appear.
We list only the coecients of each polynomial in the order x
3
, x
2
,
x
1
, x
0
.
1 0 0 1
1 0 0 2
1 0 1 1
1 0 1 2
1 0 2 1 irreducible
1 0 2 2 irreducible
1 1 0 1
1 1 0 2 irreducible
26
1 1 1 1
1 1 1 2 irreducible
1 1 2 1 irreducible
1 1 2 2
1 2 0 1 irreducible
1 2 0 2
1 2 1 1 irreducible
1 2 1 2
1 2 2 1
1 2 2 2 irreducible
We found 8 polynomials. If we multiply each by 2, we would get their
matches, yielding 16 irreducible polynomials as shown below.
2 0 0 1
2 0 0 2
2 0 1 1 irreducible
2 0 1 2 irreducible
2 0 2 1
2 0 2 2
2 1 0 1
2 1 0 2 irreducible
2 1 1 1 irreducible
2 1 1 2
2 1 2 1
2 1 2 2 irreducible
2 2 0 1 irreducible
2 2 0 2
2 2 1 1
2 2 1 2 irreducible
2 2 2 1 irreducible
2 2 2 2
Exercise 3.3. Demonstrate that f(x) = x
4
22x
2
+ 1 is irreducible
over Q. Solution. By the quadratic formula,
x
2
= 11

22
2
4/2
= 11 2

30
27
So x =

11 2

30.
Now we can write f(x) as the product of 4 linear factors in R[x]. The
linear factors are irreducible polynomials in R[x]; so, the factorization
of f(x) is unique except for order and scalar multiples. Therefore,
if f(x) were to factor in Q[x], we would see rational roots in these
linear factors, which we dont.
Because the irrationals are not closed under multiplication, we must
also make sure that no factors of degree 2 will have rational coecients.
We take the linear factors and multiply them in all possible ways.
Let
A = x

11 + 2

30
B = x +

11 + 2

30
C = x

11 2

30
D = x +

11 2

30.
So
AB = x
2
11 2

30
AC = x
2
+ (

11 + 2

30

11 2

30)x + 1
AD = x
2
+ (

11 + 2

30 +

11 2

30)x + 1
BC = x
2
+ (

11 2

30 +

11 + 2

30)x + 1
BD = x
2
+ (

11 2

30 +

11 + 2

30)x + 1
CD = x
2
+ 241 44

30
No factor with rational coecients; so f(x) is irreducible in Q[x].
Exercise 3.4. Let : G H be a group homomorphism which is
onto. Suppose G has 60 elements and H has 4 elements. Find [ker()[.
Let h be in H. How many elements of G map to h?
Solution. By the fundamental theorem of homomorphisms, G/ker()

= Image(). Since is onto, H



= Image(), so G/ker()

= H.
Since [H[ = 4, 60/[ker()[ = 4; hence, [ker()[ = 15. Let h be in H;
so [
1
[h][ = 15.
28
Exercise 3.5. Show that =

2 +

3 is algebraic over Q by nding


f(x) in Q[x] such that f() = 0. Repeat for =

2 i.
Solution.

2
= 5 + 2

2
5 = 2

4
10
2
+ 25 = 24

4
10
2
+ 1 = 0
So,

2 +

3 is algebraic over Q; so its an algebraic number.


For the other, we write

2
=
3

2 i

2
i =
3

2
(
2
i)
3
= 2
(
4
2i
2
1)(
2
i) = 2

6
i
4
2i
4
2
2

2
+i = 2

6
3i
4
3
2
+i = 2

6
3
2
2 = 3i
4
i
(
6
3
2
2)
2
= 9
8
+ 6
4
+ 1
(
6
3
2
2)
2
+ 9
8
6
4
1 = 0
Now the coecients are rational. Since is a root of the polynomial
above, and is in C and C is an extension eld of Q, then is algebraic
over Q, so is algebraic number.
Exercise 3.6. The complex numbers s + ti and s ti with s, t in R
are often called conjugates. Show that they satisfy the denition of
conjugates. What happens if t = 0?
Solution. If the coecients of the polynomial
p(x) = (x s +ti)(x s ti),
are real numbers, then we can pick R = F and C = E to satisfy the
denition, because C is an extension eld of R. Also, p(x) would be
irreducible over R because all of its roots are complex numbers. Since
p(x) = x
2
2sx + (s
2
+t
2
),
29
we have p(x) in R[x] satisfying the denition. If t = 0, then the dis-
criminant found in the application of the quadratic formula is greater
than zero which implies that all roots are real which causes p(x) to be
reducible in R failing the denition for the chosen elds here. If we
want to, however, we can pick other elds to satisfy the denition for
when t = 0.
Exercise 3.7. Let F be the eld of rational numbers and let p(x) =
x
2
+x/31. Show that x
3
+1, 1/3x
2
+x+1 and 10x/9+2/3 are in
the same equivalence class. Show that x
4
+x
2
+1 and 28x/27+28/9
are in the same equivalence class.
Solution. We have x
3
+1 = 10x/9+2/3 and x
2
/3+x+1 = 10x/9+
2/3. This shows theyre all in the same equivalence class. For the
other class, we have x
4
+x
2
+ 1 = 28x/27 + 28/9.
Exercise 3.8. Still using Q/p(x) = x
2
+ x/3 1), nd linear poly-
nomials equivalent to x
2
, x
3
, x
4
. Use the results to nd a linear
polynomial equivalent to 3x
4
+x
3
2/3x
2
+x + 1.
Solution. For the rst part, we do long division on each. So
x
2
= x/3 + 1
x
3
= 10x/9 1/3
x
4
= 19x/27 + 10/9.
For the second part, we use the equivalences above to perform the
appropriate substituitions for 3x
4
+ x
3
2/3x
2
+ x + 1. That is, we
can write it as
3(19x/27) + 10/9 + (10x/9) 1/3 + (2/3) (x/3) + 1 +x + 1,
which is equivalent to (46 + 16x)/9.
Exercise 3.9. What is the multiplicative identity of F[x]/p(x))?
Solution. The multiplicative inverse e(x) is an element of the eld
such that k(x)e(x) = e(x)k(x) = k(x). If we take any member of k(x)
and multiply it by a constant term, we get a member of k(x), so all
constants are members of e(x).
Exercise 3.10. Let q(x) = x + 1/2; check if q(x)
1
= 4/5x + 2/5
is the inverse of q(x) in the eld Q[x]/x
2
+ 1).
30
Solution. To check, we verify that q(x) q(x)
1
= e(x) and we divide
e(x) by x
2
+ 1. If we get a constant polynomial, it worked. It does
because
q(x) q(x)
1
= 1/5 4x
2
/5 = 1.
But it couldve been any constant other than 1, since all constants are
members of the class 1.
Exercise 3.11. Find the inverse of x in Q[x]/x
2
+ 1).
Solution. What polynomial that when multiplied by x will yield a
constant mod x
2
+ 1? The polynomial x itself, because (x)(x) = x
2
= 1 mod x
2
+ 1. So x
1
= x.
Exercise 3.12. Divide
2
+ 1 by t + s, express their gcd linearly,
nd the inverse of s+ti, and compare with the preceding result of the
book.
Solution. The division algorithm yields

2
+ 1 = (t +s)(/t s/t
2
) + (1 +s
2
/t
2
).
Since (1 +s
2
/t
2
) is a constant, it is a gcd(
2
+ 1, t +s). We express
it linearly now with

1 +
s
2
t
2

=
2
+ 1

t +s

t

s
t
2

=
2
+ 1

t

s
t
2

t +s

We want to express 1 as a combination of


2
+ 1 and t + s. So we
multiply the equation by 1/(1 +s
2
/t
2
), which yields
1 =

2
+ 1

t

s
t
2

t +s

1
1 +s
2
/t
2
=

1
1 +s
2
/t
2

2
+ 1

1
1 +s
2
/t
2

t

s
t
2

t +s

t
2
t
2
+s
2

2
+ 1

t s
t
2
+s
2

t +s

So
1 +
2
+ 1) =
t s
t
2
+s
2

t +s

+
2
+ 1).
The inverse of s +ti is
s ti
s
2
+t
2
.
31
The comparison we can make here is that
s ti
s
2
+t
2
seems to be the same as

t s
t
2
+s
2
=
s t
t
2
+s
2
,
where plays the role of i.
Exercise 3.13 (7-12).
Exercise 3.14 (7-14).
Exercise 3.15. Find the minimum polynomials over the small eld
of the following elements in the following extensions: (a) i in C : Q,
(b)
3

7 in R : Q, (c) (1 +
5

)/2 in C : Q, (d)

5 +

3 in C : Q.
Solution. We set x = r, where r is the desired root, then we work
with the equation until we get the coecients over the desired eld.
For (a), we have x
2
+1, for (b) x
3
7, (c) x
2
x1, (d) x
4
16x
2
+4.
Exercise 3.16. Find an irreducible polynomial p(x) in Z
3
[x] of degree
2. Explain why it is irreducible. Contruct the eld E = Z
3
[x]/p(x)).
Factor p(x) into two linear terms in E[x].
Solution. For a prime p, Z
p
is a eld, so it has no zero divisors and
its closed under the operations, so a polynomial is reducible if and
only if it has a root. In Z
3
, x
2
+ 1 has no roots.
Let the roots be caled T and T = 2T. We attach the roots to Z
3
,
and we compute what other elements will be appear in the eld by
computing the addition and multiplication tables.
+ | 0 1 2 T 2T 1+T 1+2T 2+T 2+2T
-----------------------------------------------------------
0 | 0 1 2 T 2T 1+T 1+2T 2+T 2+2T
1 | 1 2 0 1+T 1+2T 2+T 2+2T T 2T
2 | 2 0 1 2+T 2+2T T 2T 1+T 1+2T
T | T T+1 T+2 2T 0 1+2T 1 2+2T T
2T | 2T 1+2T 2+2T 0 T 1 1+T 2 2+T
1+T |1+T 2+T T 1+2T 1 2+2T 2 2T 0
1+2T |1+2T 2+2T 2T 1 1+T 2 2+T 0 T
32
2+T |2+T T 1+T 2+2T 2 2+2T 0 1+2T 1
2+2T |2+2T 2T 1+2T 2 2+T 0 T 1 1+T
* | 0 1 2 T 2T 1+T 1+2T 2+T 2+2T
-----------------------------------------------------------
0 | 0 0 0 0 0 0 0 0 0
1 | 0 1 2 T 2T 1+T 1+2T 2+T 2+2T
2 | 0 2 1 2T T 2+2T 2+T 1+2T 1+T
T | 0 T 2T 2 1 2+T 1+T 2+2T 1+2T
2T | 0 2T T 1 2 1+2T 2+2T 1+T 2+T
1+T | 0 1+T 2+2T 2+T 1+2T 2T 2 1 T
1+2T | 0 1+2T 2+T 1+T 2+2T 2 T 2T 1
2+T | 0 2+T 1+2T 2+2T 1+T 1 2T T 2
2+2T | 0 2+2T 1+T 1+2T ---- T 1 2 2T
We can now factor every polynomial in Z
3
(T, 2T) all the way to linear
factors. The roots of x
2
+ 1 are T and 2T, so
x
2
+ 1 = (x +T)(x + 2T).
4 Algebraic eld extensions
Denition 4.1. If every element of a eld F is an element of a eld
E and if F has the same eld operations as E, then F is a subeld of
E and E is an extension of F. We write F E.
Denition 4.2. Let a
0
+a
1
x+a
2
x
2
+... +a
n
x
n
= 0 be a polynomial
equation of degree n with coecients a
i
in a eld F, and let E be an
extension eld of F. If r in E is a root of such an equation, then r is
said to be algebraic over F. If F is Q, then r is an algebraic number.
Denition 4.3. A real number that is not the root of any polynomial
equation with rational coecients is called transcendental.
Knowing that is transcendental from hearing people say it, makes
one wonder if we can show that it is not transcendental. All we have
to do is nd a polynomial with rational coecients which has as a
root. We cant consider x = 0 because we cant assume that is
rational. A rational number times an irrational number is irrational,
so we will need at least a polynomial of degree 2.
33
Why? Because the idea is: if isnt transcendental, then eventually
we would nd some natural n such that
n
= r, where r is a rational
number. Consider the irrational

2. If we square it, we get 2; so this
way we can nd a polynomial to which

2 is a root. For example,
x
2
2 = 0. So, if we can apply some power to and get it to
become a rational number, then we would be done. If I believe the
mathematicians, then I have to say that this is not possible and if not,
then this is another way to think of transcendental numbers: numbers
which wont square to a rational number, ever.
Denition 4.4. An algebraic number is called an algebraic integer
if and only if it is a root of a monic polynomial equation of degree n
with coecients in Z.
Every algebraic number r is a root of an equation with coecients in
Z. Why? Because r is a root of an equation with coecients in Q, by
denition. So, multiply the equation by the least common multiple of
the denominators of the rational coecients and we get an equation
with integral coecients; the polynomial equation wont necessarily
be monic.
Theorem 4.1. Let F be a eld and let r be in E F be algebraic
with respect to F. Then r is a root of a unique monic irreducible
polynomial p(x) in F[x] and p(x) divides each polynomial in F[x]
that has r as a root.
Denition 4.5. Let F be a eld. Then r
1
, r
2
in E F are conjugates
if they are roots of the same irreducible polynomial in F[x].
Denition 4.6. If every element of an extension E of F is algebraic
with respect to F, then E is an algebraic extension of F.
Denition 4.7. Let F be a eld and let p(x) be in F[x] such that p(x)
is irreducible in F[x]. We let F[x]/p(x)) stand for the polynomials
with coecients in F taken modulo p(x)) with equivalence classes
described as r(x) + p(x)q(x), where q(x) is in F[x], represented by
r(x) mod p(x).
The uniqueness of the remainder of a long division is guaranteed by
the Division Algorithm, so the representation is described in the de-
nition above is also unique. Also, we must check that p(x), from the
denition above, is really irreducible.
34
Denition 4.8. We may write the representative of an equivalence
class of polynomials as a polynomial with a bar on top; for example,
(10/9)x + 2/3.
Theorem 4.2 (Kronecker). Let F be a eld, and let p(x) be irre-
ducible in F[x]. Then F[x]/p(x)) is a eld, and contains a subeld
isomorphic to F. The polynomial p(x) has a root in F[x]/p(x)).
Lets review some concepts. Let F be a eld. Suppose r is algebraic
over F which means r is the root of some polynomial over F. If we
require r to be the root of some irreducible polynomial, then theres
only one polynomial to which r is a root of. We call such polynomial,
the minimum polynomial of r over F.
For example, if F = Q and r =
3

2, then p(x) = x
3
2. But if F =
Q, then
3

2 implies the existence of p(x) = x


3

2. So, it depends on
the eld. Now, consider as the root of some polynomial over Q.
Theorem 4.3 (Kronecker, in new words). We have a polynomial p(x)
which is irreducible over F. We can take all polynomials in F[x] and
quotient it out with p(x); that is, we can create F[x]/p(x)); this is a
eld, with nice properties.
The rst property is that the eld, mentioned in the theorem of Kro-
necker, has F inside it: the constant polynomials. Now x is a root
of p(x). For example, x
2
+ 2x + 2 is irreducible by E.I.C. Now take
Q[x]/p(x)) and notice that since p(x) is congruent to zero by deni-
tion, x is a root. To check, just plug x in the polynomial.
Another way to think of the previous example is to declare r to be a
root of p(x). Now, we form a eld F(r) read this as F adjoin r.
Now use the fact that p(r) = 0 to multiply things around what
would this mean?
Going on; if p(x) has degree n, then every element of F(r) is of the
form c
0
+ c
1
r + c
2
r
2
+ ... + c
n1
ra
n1
with c
i
in F. Lets look at an
example.
Declare r a root of x
2
+2x+2. In Q(r), we know that r
2
+2r+2 = 0 or
r
2
= 2r 2. Now Q(r) = x+yr : x, y Q. With this information
now, we write (6 + 2r)(1 3r) = 6 20r 6r
2
= 6 20r 6r
2
= 6 8r because r
2
= 2r 2.
Remark 4.1. So, we have a eld. But how do we divide things in it?
That is, whats 1/a, for example? We can use the divison algorithm to
35
answer that. The multiplication by conjugates doesnt always work;
that is, in C you can dene division in terms of multiplication by
inverses, but it seems to me that this works in C and not on every
eld. So, I should be able to nd one eld in which this does not work,
right? This is my task now, because I dont know one right now, but
I will say that I believe that not all the time we get something like i
2
= 1 in a eld that will answer all of our questions. So I say that C
is a very nice eld, but there are some mean ones.
Denition 4.9. A simple extension is an extension which includes
only one element. The degree of an extension F E is the number
of elements of E that we need to span E over F.
What does this mean? Take
Q(

2,

3) = a +b

2 +c

3 +d

6,
and notice that it is a degree 4 extension of Q, because we absolutely
need 1,

2,

3, and

6.
Now, we have to notice that Q(

2 +

3) = Q(

2,

3), but they
have dierent bases and dierent degrees. Doesnt this sound not
well-dened? This alters perhaps what we call, in this document, a
splitting eld. I owe an explanation here.
Theorem 4.4. If r has minimum polynomial p(x) of degree n over
F, then F(r) : F is an extension of degree n.
So, for example, take p(x) = x
5
+ 2x
2
+ 2x + 6, which is irreducible
over Q, and declare r as a root, so
Q(r) = a +br +cr
2
+dr
3
+er
4

has degree 5.
Now, consider x
2
+ 2x + 2. Here,
Q(r) = a +br : a, b Q.
The rule for multiplication here is r
2
= 2r 2. This is just like the
complex i such that i
2
= 1. Now, we have a root, so we can factor
it. We do the long division between x r and x
2
+ 2x + 2. We get
x
2
+ 2x + 2 = (x r)(x + 2 +r) + 0.
So r and r 2 are conjugates in Q(r).
36
Consider now x
3
2 which is irreducible over Q. So we extend the eld
by adjoining
3

2. So a division should happen with zero remainder.


We get
x
3
2 = (x
3

2)(x
2
+

2 +
3

4) + 0.
Now (x
2
+

2 +
3

4) is irreducible. We then expand the eld to get


it to be reducible. So Q(
3

2,
3

2(1/2 +i

3/2)). I wondered if these


additions to the eld would make the eld completely factorable, but it
doesnt; these additions make these polynomials factorable, but others
are still irreducible, so what happens to C? C somehow is completely
factorable. This is a miracle; it turns out that R[x] mod x
2
+1 makes
C a beautiful eld.
Denition 4.10. Let p(x) be irreducible over F. Suppose F E D
are elds. The polynomial splits if it factors into linear terms. We
say D is a spliting eld for p(x) if p(x) splits in D[x], and p(x) does
not split in E[x] for any F E D.
By the way, all roots of p(x) lie in the splitting eld. Why? Because
if it factors then
p(x) = (x r
1
)(x r
2
)...(x r
n
).
Theorem 4.5. There exists is a unique splitting eld. Take F(r
1
, r
2
,
..., r
n
) which are the roots of p(x).
Lets now see an example. Find a splitting eld of x
4
3x
2
+1. How
do we do it? Find the roots of the polynomial; use the quadratic
formula. So x
2
= (3

5)/2, so x =

(3

5)/2. Now,
Q

(3 +

5)/2,

(3

5)/2

is a splitting eld. Were done. There may be a better way to write


this eld, and this is one of the frustrating thing: there is so many
ways to write it and its not always easy to gure out whether we have
the simplest.
Theorem 4.6. F(r
1
, r
2
) = F(r
1
)F(r
2
) = F(r
2
)F(r
1
) et cetera.
Theorem 4.7. Suppose F E D with degree E : F = m, degree
D : E = n. Then D : F = mn.
37
An example of this theorem is F = Q and E = Q(

2) = a + b

2
: a, b Q. Degree E : F = 2. Now, look at E = Q(

2) and
D = E(
4

2). The minimum polynomial of


4

2 over Q(

2) is x
2

2
so degree D : E = 2 and degree E : Q = 4. I dont really follow this
example, so if theres anything wrong here I wouldnt be able to tell.
Degree Q() : Q = . Now consider Q() : Q(
2
) which has degree
2, and is root of x
2

2
.
In this section, weve studied algebraic numbers and transcendental
numbers. We have seen irreducible polynomials over a eld F. We
have seen Kroneckers theorem which claims the existence of a eld
F[x]/p(x)) which is an extension eld of F where p(x), in particular,
has a root; we can think of F[x]/p(x)) as a simple degree n extension
F(r) where p(r) = 0 degree p(x) = n.
Weve seen minimum polynomials, conjugates of eld elements. On
the next section, we will be looking at roots in which when we permute
them around, we will nd nice properties; that is, well be building a
group of permutations of roots.
Exercise 4.1. Prove that 1, i =

1 form a linear basis for R(i),
where R is the real eld. Show that g =

2 2i is algebraic relative
to R by nding a polynomial having g as a zero.
Solution. Every element of R(i) can be written as a + bi, so 1 and i
span R(i). To show linear independence, we must show that for all r
in R(i), r = a 1 +b i = 0 if and only if a = b = 0.
If a = b = 0, then r = 0. Now if r = 0, then
r = a +ib = 0
a = ib
a
2
= (ib)
2
= (i)
2
b
2
= b
2
a
2
+b
2
= 0
But since a and b are real numbers, a
2
and b
2
are non-negative, so
a = b = 0.
38
To show that g =

2 2i is algebraic relative to R, we write


g
2
= 2 + 4i
2
4i

2
g
2
2 + 4 = 4i

2
(g
2
+ 2)
2
= (4i

2)
2
= 32
(g
2
+ 2)
2
+ 32 = 0
Exercise 4.2. Construct a splitting eld for p(x) = x
4
8x
2
+ 15
over Q.
Solution. We factor p(x) into linear terms and see which elements
were missing into some eld, so that we can build the smallest possible
eld such that p(x) factors completely.
p(x) = (x
2
5)(x
2
3)
= (x +

5)(x

5)(x +

3)(x

3)
So p(x) splits into linear factors in Q(

5,

3).
Exercise 4.3. Is Q() a nite extension? If so, give a basis of Q()
: Q.
Solution. Suppose it is nite. Then we could try 1 and as a basis,
but we need to account for
2
, so we could try, 1, ,
2
as a basis, but
then we would have to account for
3
as well, so we could expand the
basis again and so on, so it doesnt look like this will ever terminate.
Exercise 4.4. Find minimal polynomial of Q[x] and degree of
Q() : Q.
(a) =

6
a
2
= 3

6
a
2
3 =

6
(a
2
3)
2
= 6
(a
2
3)
2
6 = 0
a
4
6a
2
+ 3 = 0
Degree Q() : Q = 4. Let p = 3, so by E.I.C. it is irreducible.
39
(b) =

1
3
+

2
=
1
3
+

1
3
=

1
3

2
= 7

1
3

2
7 = 0

2
2
3

62
9
= 0
9
4
6
2
62 = 0
Degree Q() : Q = 4. Let p = 2, so by E.I.C. it is irreducible.
(c) =

2 +i

2
= 1 + 2i

2
1 = 2i

4
2
2
+ 9 = 0
Degree Q() : Q = 4. E.I.C. doesnt seem to help here. What would
be another approach? The only thing I can think of is to compute the
roots and show they wont allow reducibility over Q.
The polynomial factors into
( +

2 +i)( +

2 i)(

2 +i)(

2 i).
Let A, B, C, D be rst, second, third, forth factors, respectively. Then
AB has an irrational coecient, AC has a complex one, AD has a
complex one, BC has a complex one, BD has a complex one and CD
has a complex one; brute force shows no factors over Q are possible.
(d) =

2 +

2
= 2 + 5 + 2

10
(
2
7)
2
40 = 0

4
14
2
+ 9 = 0
Degree Q() : Q = 4. Brute force again. The roots are r
1
=

2 +

5,
r
2
=

5, r
3
=

2 +

5, r
4
=

5. So it factors over R
as
(x r
1
)(x r
2
)(x r
3
)(x r
4
).
40
Let A, B, C, D be the factors in the other they appear above. So
AB, AC, AD, BC, BD and CD have an irrational coecient. So
factorization over Q is not possible.
Exercise 4.5. For each C and each given eld F, classify as
either algebraic or transcendental over F. If is algebraic, nd the
degree of F() : F.
(a) = i, F = Q.

2
= i
2

2
+ 1 = 0.
So i is algebraic over Q, and degree of Q(i) : Q = 2.
(b) = , F = Q. Transcendental.
(c) =
2
, F = Q(). We have
2
as a root of p() =
2
, so

2
is algebraic over Q() and the degree of Q(
2
) : Q() = 1.
(d) =
2
, F = Q(
3
). Transcendental.
Exercise 4.6. Find degree and basis for the given eld extensions.
(a) Q(

2) : Q. Basis (1,

2), degree 2.
(b) Q(

2,

3,

18) : Q. Basis (1,

2,

3,

6), degree 4.
(c) Q(
3

2,

3) : Q. Basis (1,

3,
3

2,
3

4,

3
3

2,

3
3

4), degree 6.
(d) Q(

2,

3) : Q(

2 +

3). Since

3 +

2 is in the eld, we also


have

6,

2,

3, so the elds here are the same, and we have 1 as a
basis, so degree is 1.
Exercise 4.7. Find all conjugates for =

2 +

3 over Q. Repeat
for =
3

2.
Solution. We look for an irreducible polynomial over Q which has
as a root. So p(x) = x
4
10a
2
+ 1 is irreducible over Q and its roots
are itself, and all of its conjugates which are

2,

3 +

2,

2.
Exercise 4.8. Let , : F F be two eld automorphisms. Prove
that
1
and ( t) are also eld automorphisms.
Proof. Since is an automorphism, it is one-to-one and onto and takes
each f F to some f

F for each f

F. By denition,
1
is the
41
one-to-one and onto function such that (
1
)(f) = (
1
)(f) =
f for all f F.
So, if we show that
1
is a homomorphism, we would have shown
that it is an isomorphism, and therefore, an automorphism. Let a, b
be in F. Since is an automorphism, there are u, v in F with a =
(u), b = (v). So,

1
(ab) =
1
((u)(v))
=
1
((uv))
= uv
=
1
(a)
1
(b).
Similarly for addition,

1
(a +b) =
1
((u) +(v))
=
1
((u +v))
= u +v
=
1
(a) +
1
(b).
Now . Since and are eld automorphisms, they preserve the
operations. Let a and b be in F. So, for multiplication, we do
( )(ab) = ((ab))
= ((a)(b))
= ((a))((b))
= ( )(a)( )(b).
For addition,
( )(a +b) = ((a +b))
= ((a) +(b))
= ((a)) +((b))
= ( )(a) + ( )(b).
So ( ) is a homomorphism. To show that it is onto, let f

be in F
and then we must show that theres an f in F such that ( )(f) =
f

, for all f

in F.
Since is an automorphism, then for each f

in F, there is an f in F
such that (f) = f

. Since is also an automorphism, then for each


f

in F, theres (f) in F such that ((f)) = f

. So for each f

in
F, theres an f in F such that ((f)) = ( )(f) = f

, so its onto,
and hence, an automorphism.
42
Exercise 4.9. Prove that G(C/R) is cyclic with 2 elements.
Proof. By denition, the Galois group G(C/R) is the group of auto-
morphisms of C that x every r R, where the binary operation is
composition. One automorphism is the identity map, which is one-to-
one and onto, and is the identity of the group.
We will try to nd a second one. Dene a map that maps conjugates;
that is, the rule is map a +bi to a bi and notice that every r R
is of the form a + b 0, so for a real number r, (r) = r, so every
r R is xed. Now, if is an isomorphism, then it is a member of
the group, and if we can show that
2
= ( s) = e, then we would
have shown that there are no more elements in the group, hence that
the group is cyclic.
Let a +bi, c +di C. So,
s((a +bi)(c +di)) = s(ac bd + (ad +bc)i)
= ac bd (ad +bc)i
= (a bi)(c di)
= s(a +bi)s(c +di).
Also,
s((a +bi) + (c +di)) = s((a +c) + (b +d)i)
= a +c (b +d)i
= a bi +c di
= s(a +bi)s(c +di)
So is a homomorphism. Let a+bi be in C. We want to show that
is onto; that is, we want to show that there is a c +di in C such that
(c +di) = a +bi.
We know that (c + di) = c di, where c and d are real numbers.
The task was to nd c and d such that c = a, d = b or d = b. There
are such numbers, pick c = a and d = b. So is an isomorphism.
Now we must show that
2
= ( s) = e. Let a + bi in C. Compute
( )(a +bi) = ((a +bi)) = (a bi) = a +bi = e(a +bi); hence,
the group is cyclic with two elements, e and .
Exercise 4.10. Let F be a eld. Dene the characteristic of F to be
the smallest n 1 such that 1 + 1 + ... + 1 = 0 where we are adding
1 to itself n times. If no such n exists, we say F has characteristic
zero. What is char(Q)? What is char(Z
7
)? Suppose F has nite
43
characteristic n. Prove that n must be prime. Hint: elds have no
zero divisors.
We have char(Q) = 0, and char(Z
7
) = 7.
Now char(F) = n. Proof. Let F have nite characteristic n; n must
be composite or prime. Suppose n is composite. Then there is a prime
p < n such that n = pq, for some natural q.
Then (1 + 1 + ... + 1)
p
(1 + 1 + ... + 1)
q
= 0, where the index p
means 1 added p times, and similarly for q. Since elds have no zero
divisors, F isnt a eld if n is composite; hence, n is prime.
5 Galois theory
Denition 5.1. An automorphism of a mathematical system S
with an operation dened on S is a one-to-one correspondence from
S onto itself that is an isomorphism with respect to the operation.
Denition 5.2. An automorphism of a eld F is an automorphism
of F that preserves both eld operations. An element f in F is xed
by an automorphism if maps f to itself; that is, (f) = f; it is
moved by if (f) ,= f.
Theorem 5.1. Let D be a nite extension of F, let E be an inter-
mediate eld F E D. Let G(D/F) be the automorphisms of D
that x every f in F. Let G(D/E) be the automorphisms of D that
x every e in E. Then G(D/F) is a group under composition and
G(D/E) is a subgroup of G(D/F).
Proof. Suppose that G(D/F) is a group under composition. Then its
easy to believe that G(D/E) is a subgroup of G(D/F) because we
know that we would have a similar set with less automorphisms. So
anyway, we will rst prove that G(D/F) is a group.
We show that we have closure under composition. Let
1
,
2
be in
G(D/F). So any d in D has an image
2
(d) in D because
2
is an
automorphism of D, and
1
(
2
(d)) is in D. So the composition of two
automorphisms produces an element in D. Now... oh well, can you
go on?
Denition 5.3. Let E be the splitting eld of a polynomial f(x) in
F[x]. The group G(E/F) of automorphisms of E that x every f in
44
F is the Galois group or group of the polynomial and the group of the
equation f(x) = 0.
Theorem 5.2. Let F be a eld. Let F(r) : F = n. Any automor-
phism in G(F(r)/F) is completely determined by (r). Its true
that [G(F(r)/F)[ n. Let E be the splitting eld of an irreducible
polynomial p(x) in F[x]. The Galois group G(E/F) is isomorphic to a
group of permutations of the k zeros of p(x) in E, hence to a subgroup
of the symmetric group S
k
.
Notice the importance of this theorem. It links group theory to the
permutations of the roots of an equation. The theory of groups, as we
see in the Abstract Algebra books today wasnt very much developed
by Galois or Abel. They studied particular groups formed by the per-
mutations of the roots of an equation, but in 1870, Kronecker isolated
the group properties in their own right.
Theorem 5.3 (). Suppose r is algebraic over F with a minimal
polymonial of degree n. Then [G(F(r)/F)[ n.
Proof. Let be in G(F(r)/F). By denition, the group leaves every
u in F xed, so there are at most n possible images of r under .
Since every is one-to-one and onto, and since every u in F is already
taken, there are at most n possible mappings available because there
at most n roots for the minimal polynomial.
Consider the extension eld Q(r) : Q, and let the minimum polynomial
of r be p(x) = x
5
+
2
3
x
3
+ 2x + 1. How do we know that (r) must
be a root? We know that p(r) = 0. So now apply to r, so (p(r))
= (0) = 0 because 0 is in Q and so it is xed. Now, since is an
automorphism, we can write
(p(r)) = (r
5
+
2r
3
3
+ 2r + 1)
= (r
5
) +
2(r
3
)
3
+ 2(r) + 1
= (r)
5
+
2(r)
3
3
+ 2(r) + 1 = 0
So (r) is a root.
So, suppose that is in G(E/F) and p(x) is in F[x], then permutes
any roots of p(x) which lie in E. We prove this with the argument
above.
45
Denition 5.4. A eld extension E F is normal if and only if any
irreducible p(x) in F[x] which has a root in E must factor completely
in E.
Heres an example of a non-normal extension. Consider Q(
3

2) : Q,
which is not normal; it isnt because x
3
2 has a root, but doesnt
factor completely, and this is a problem for us. We want a bigger eld
with a Galois group.
E : F is normal if and only if every irreducible p(x) in F[x] with a
root in E splits in E, since to split means to factor all the way down
into linear terms.
This is the same as saying that E is the splitting eld of some poly-
nomial. Or equivalently, a E : (a) = a for all G(E/F) = F.
Normal extensions will correspond to normal subgroups.
Denition 5.5. A eld F is called a eld of characteristic zero if F
contains a subeld isomorphic to the eld Q of rational numbers.
Lets review some stu. Suppose E F is a eld extension. The Ga-
lois group G(E/F) = : E E such that is a eld automorphism
and (r) = r for every r in F.
We have to check that eld automorphism form a group under com-
position. I also have to check that if and both x f, then so do

1
and .
Remark 5.1. Let be in G(E/F), then xes polynomials in F[x],
thus permutes roots of polynomials in F[x].
Galois group elements take roots to roots, but the roots must belong
in the eld. Suppose a cubic root is a a root of some polynomial with
only one real solution. Suppose the complex roots do not belong in
the eld. Then the only element will be the identity.
Proposition 5.4. Suppose r is algebraic over F with minimal poly-
nomial of degree n. Then G(F(r)/F) has at most n elements.
Proof. Let r have minimal polynomial
p(x) = a
0
+a
1
x +... +a
n
x
n
.
46
So p(r) = 0. Let u be in G(F(r)/F). So
p(u(r)) = a
0
+a
1
u(r) +... +a
n
(u(r))
n
= a
0
+a
1
u(r) +... +a
n
(u(r)
n
)
= u(a
0
) +u(a
1
r) +... +u(a
n
r
n
)
= u(a
0
) +a
1
r +... +a
n
r
n
)
= u(p(r))
= u(0)
= 0
The last two equations hold because p(r) = 0 and u(0) = 0 because u
is an automorphism, so it maps 0 to 0. So u(r) is also a root. Now, we
notice that every element of G(F(r)/F) corresponds, in a one-to-one
and onto fashion because theyre automorphisms , with a root
of p(x). Since there are at most n roots for p(x), there are at most n
automorphisms in G(F(r)/F).
A word on degrees. The degree of E:F depends on both E and F.
Sure, compare Q(

2,

3) : Q with Q(

2,

3) : Q(

2).
By the way, in this document, our denition of splitting eld is the
smallest possible for a minimal polynomial. There seems to be other
equivalent denitions.
Proposition 5.5. Suppose E is a splitting eld of a polynomial of
degree k in F[x]. Then G(E/F) is a subgroup of S
k
.
Proof. Let r
1
, r
2
, ..., r
k
be roots of p(x) over E. We know that over
E, p(x) splits. Then E = F(r
1
, r
2
, ..., r
k
).
Let be in G(F(r
1
, ..., r
k
)/F). To nish the proof, we must show that
(1) is completely determined by (r
1
), ..., (r
k
). (2) (r
i
) must be
r
j
for some j. To conclude, notice that (1) and (2) tell us that the
Galois group is isomorphic to some subgroup of S
k
.
Denition 5.6. Galois group of p(x) in F[x] is G(E/F) for E a
splitting eld of p(x).
Exercise 5.1. Find the Galois group of x
4
2 over Q.
Solution. We rst look for the roots of x
4
2. The polynomial factors
as
(x
2
+

2)(x
2

2) = (x +
4

2i)(x
4

2i)(x +
4

2)(x +
4

2)
47
Let r =
4

2. So roots are r, r, ri, ri. Whats the splitting eld?


Well, throw, in the basis, the elements r and ri, but notice that we
only need r and i.
What do we need now? Let f be in G(Q(r, i)/Q). To determine f, we
need to know f(r) and f(i). What are the possibilities for f(r)? All
the roots of x
4
2. What about f(i)? It goes to i or i.
So we have 4 possibilities for r, and 2 for i, so 4 2 = 8. This
group has at most 8 elements. It is not necessary that all of these
possibilities will work. We need more theory to show how this could
work, but they are all technically possible so for.
Lets list them. As we read the table, notice that
2
(r) = ((r)) =
r. Thats (r) = ir. So
4
is the identity. We introduce . Now,
we do .
automorphism domain range
1 r i
ir i

2
r i

3
ir i
r i
ir i

2
r i

3
ir i
Do you recognize this table? This group is isomorphic to D
8
, the
dihedral group with 8 elements; the symmetries of a square. We can
describe it by =
3
.
Theorem 5.6. Suppose F E D. Then (1) G(D/E) G(E/F).
(2) Suppose H G(D/F). Let D
H
= d D : (d) = d for all
H. Then F D
H
D.
Recall that E/F is a normal extension if any of the following 3 equiv-
alent conditions hold: (1) E is a splitting eld of some polynomial
in F[x], (2) any irreducible polynomial p(x) in F[x] with a root in E
must split in E, (3) xed eld of G(E/F) is F.
Theorem 5.7 (Fundamental Theorem of Galois Theory). Suppose
the characteristic of F = 0, and F D is a normal extension. Let F
E D; then (1) E G(D/E) gives a one-to-one correspondence
between intermediate elds E and subgroups of G(D/F), (2) The
48
degree [D : E] = [G(D/E)[, (3) F E
1
E
2
D if and only if
e G(D/E
2
) G(D/E
1
) G(D/F), (4) E is normal over F if
and only if G(E/F) G(D/F).
We now solve a series of exercises which will help us to understand
the claims of the Fundamental Theorem of Galois Theory.
Exercise 5.2. Determine the Galois group of
(x
2
2)(x
2
3)(x
2
5),
over Q.
Solution. The polynomial splits in Q(

2,

3,

5). Now, let be
arbitrary in G(Q(

2,

3,

5)/Q). Then, the splitting eld must also


split rs minimal polynomial which is r
2
2 = 0, and so must take

2 to

2 or

2. The same goes for



3; must take

3 to

3 or

3, and the same for



5; must take

5 to

5 or

5.
So, the automorphisms are
automorphisms

2

3

5
1

2

3

5

2

3

3

5

2

3

5

2

3

3

5

5
So the galois group is determined, and it is completely determined by
:

5, :

3 and :

2.
Exercise 5.3. Now determine all subelds of the splitting eld found
in the previous exercise.
Solution. The subelds of the second level of the eld diagram
would be Q(

2,

3), Q(

2,

5), Q(

3,

5), Q(

2,

15), Q(

6,

5),
Q(

6,

10), Q(

3,

10). In the third level we would get Q(

2),
Q(

3), Q(

5), Q(

6), Q(

10), Q(

15), Q(

30). So, including the


trivial ones, we get a total of 16 in the diagram.
What if I wanted to know the xed eld of this Galois group? I would
have to look at every automorphism and see which elements of the
49
basis never get moved. So I collect these elements and throw them in
Q to get the xed eld.
When we look at each of automorphism, we see that
(a +b

2 +c

3 +d

5 +e

6 +f

10 +g

15 +h

30) =
a +b

2 +c

3 d

5 +e

6 f

10 g

15 h

30),
(a +b

2 +c

3 +d

5 +e

6 +f

10 +g

15 +h

30) =
a +b

2 c

3 +d

5 e

6 +f

10 g

15 h

30),
(a +b

2 +c

3 +d

5 +e

6 +f

10 +g

15 +h

30) =
a +b

2 c

3 d

5 e

6 f

10 +g

15 +h

30,
(a +b

2 +c

3 +d

5 +e

6 +f

10 +g

15 +h

30) =
a b

2 +c

3 +d

5 e

6 f

10 +g

15 h

30,
(a +b

2 +c

3 +d

5 +e

6 +f

10 +g

15 +h

30) =
a b

2 +c

3 d

5 e

6 +f

10 g

15 +h

30,
(a +b

2 +c

3 +d

5 +e

6 +f

10 +g

15 +h

30) =
a b

2 c

3 d

5 +e

6 +f

10 +g

15 h

30,
so it looks like that the only eld that is xed by every automorphism
is Q itself.
If now I wanted to know each corresponding subgroup of G(Q(

2,

3,

5)/Q),
then one way would be to work with Z
2
Z
2
Z
2
. We assign (0,0,1)
to because moves only

5, so we consider

5 as the 1 in the third


component of the 3-tuple. Similarly, we assign (0, 1, 0) with , and
(1, 0 ,0) with , and (0,0,0) with the identity map.
We then work with Z
2
Z
2
Z
2
under addition. To get each sub-
group now, we can consider (0,0,0), (0,0,1) and notice that it would
correspond to Q(

(2),

(3)) because it moves only

5. The others
would be found similarly.
Exercise 5.4 (8-3). Let F = Q, D = Q(

2, i), E
1
= Q(

2), E
2
=
Q(i). Then F has characteristic 0 and D = F(r), where r =

2 +fi,
f ,= 0 is in Q. Find (r) for each automorphism G(D/E
1
) and
(r) for each G(D/E
2
). Find the four conjugates of r = r
1
, hence,
by Theorem 8-2, the four automorphisms of G(D/F).
They say D = F(r), but isnt D = Q(

2, i)? So are they saying that


D can be written as F(r) where r is the root of some polynomial that
when adjoined will give us

2 and i? I think so.
50
Such root is r =

2 +fi, with f ,= 0, f Q. Its minimal polynomial
is
r
2
= 2 f
2
+ 2fi

2
r
2
2 +f
2
= 2fi

2
(r
2
2 +f
2
)
2
= 8f
2
(r
2
2 +f
2
)
2
+ 8f
2
= 0
This polynomial has degree 4 and therefore will have at most 4 roots.
Since one root is complex, I guess the fundamental theorem of algebra
implies that we will have 4 distinct roots.
Now I must nd (r) for each automorphism in G(D/E
1
). Since E
1
= Q(

2), then each automorphism moves only i, so s(r) = i or s(r)


= i. So all of the automorphisms here are
automorphism i
1 i
i
Similarly for (r) G(D/E
1
). Since E
2
= Q(i), then each automor-
phism moves

2 only, so (

2) =

2 or s(r) =

2. So all of the
automorphisms here are
automorphism

2
1

2

2
The four conjugates of r
1
=

2 +fi are r
2
=

2 +fi, r
3
=

2
fi, r
4
=

2 fi. Now I must nd, by the theorem above, the 4
automorphisms of G(D/Q). Im failing to see a clear way to use the
theorem, even though I may have gotten the automorphisms correctly
here. The exercise suggests that by the theorem, we are able to get
all the automorphisms. I guess it must be all permutations of the
elements that get moved in G(D/E
1
) and G(D/E
2
).
Any automorphism of G(D/Q) xes Q and moves

2 and i.
automorphism

2 i
1

2 i

2 i

2 i

2 i
51
The theorem says that this Galois group, in particular, is isomorphic
to a group of permutations of 4 zeros of p(x) in Q(

2, i); hence isomor-


phic to a subgroup of the symmetric group S
4
. Every automorphism
squares to the identity, so it should be isomorphic to the Klein-4 which
is isomorphic to the ips of a square.
Exercise 5.5 (8-4). Still with F = Q and D = Q(

2,i). Notice
that G(D/F) is isomorphic to the Klein-4 group. Find the unique
subgroup of G(D/F) that corresponds to each intermediate eld E
1
= Q(

2) and E
2
= Q(i). Also, nd the subeld corresponding to each
subgroup.
Solution. As proper subgroups of G(D/F), we have 1, , 1, , 1,
, which are isomorphic to subgroups of the Klein-4. The exercise
is asking for the subgroups that correspond to the elds E
1
and E
2
.
Which subgroup would correspond to Q(

2)?
It would be the subgroup that is linked to the maps that do not move
Q(

2), so its 1, . For Q(i), it would be the subgroup that is linked


to the maps that do not move Q(i); so its 1, .
Heres a question. The claim iii of the Fundamental Theorem of
Galois Theory seems to say that we have Q(

2, i) Q(i) Q(

2)
Q if and only if 1 G(Q(

2, i)/Q(i)) G(Q(

2, i)/Q(

2))
G(Q(

2, i)/Q). Q(i) is not an of Q(

2); its not even a subset. Is


this a case in which we dont satisfy the hypothesis of the claim iii?
Or is my interpretation wrong?
The exercise also says nd the subeld corresponding to each sub-
group. Isnt this the same task above? Once we nd the subgroup
correspondent to the subeld, then weve found the subeld corre-
spondent to the subgroup.
Exercise 5.6 (8-5). Verify that the claim ii of the Fundamental The-
orem of Galois Theory holds for the example weve been working with.
Solution. The claim says, for this example in particular, that
[Q(

2, i) : Q(i)] = [G(Q(

2, i)/Q(i))[
which is true because the degree of the extension is 2 since the basis
is a +b

2 for a, b Q(i), and for the Galois group we have only two
possible automorphisms; one that takes

2 to

2 and the identity.


The Theorem also claims that
[Q(

2, i) : Q(

2)] = [G(Q(

2, i)/Q(

2))[
52
which is almost the same case as above. Here, we have the basis as
a + bi, for a, b Q(

2). The degree of the extension is 2 and there


are only two possible automorphisms in the Galois group; they are the
identity and another one that takes i to i.
Also, the theorem claims that
[Q(i) : Q] = [G(Q(

2, i)/Q) : G(Q(

2, i)/Q(i))].
I think that this is a link between a degree extension and the number
of left cosets of a subgroup in its group. The degree extension is 2.
G(Q(

2, i)/Q) = G has 4 automorphisms, and G(Q(

2, i)/Q(i)) =
H has 2. Now, by Lagranges theorem, we have that 4 = 2 [G : H],
so [G : H] = 2 as well and it matches the degree extension Q(i) : Q.
Also, the theorem claims that
[Q(

2) : Q] = [G(Q(

2, i)/Q) : G(Q(

2, i)/Q(

2))].
The degree extension is 2,
[G[ = [G(Q(

2, i)/Q)[ = 4,
[H[ = [G(Q(

2, i)/Q(

2))[ = 2.
So again, by Lagranges theorem, 4 = 2 [G : H], so [G : H] = 2
matching the degree extension of Q(

2) over Q.
Exercise 5.7 (8-6). Still with F = Q and D = Q(

2, i). Show
that 1 is a proper subset of G(D/E) which is a proper subset of
G(D/F) for E = E
1
= Q(

2) and for E = E
2
= Q(i) but that
G(D/E
2
) G(D/E
1
). Why are the inclusions reversed in order in
the second clause of the third claim of the Fundamental Theorem of
Galois Theory?
Solution. Let E = Q(

2). Then G(D/E) has two automorphisms 1,


, where takes i to i, so 1 is a proper subset of G(D/E). We
also know that G(D/F) has 4 automorphisms: 1, , , as dened
in Exercise 5.4. So, evidently, G(D/E) is a proper subset of G(D/F).
Let E = Q(i). Then G(D/E) has two automorphisms 1, , where
takes

2 to

2, so 1 is a proper subset of G(D/E). We also


know that G(D/F) has 4 automorphisms: 1, , , as dened in
Exercise 5.4. So, evidently, G(D/E) is a proper subset of G(D/F).
Is G(D/E
2
) G(D/E
1
)? No, because is in G(D/E
2
), but not in
G(D/E
1
). Also, is in G(D/E
1
), but not in G(D/E
2
). That is, my
observations here do not satisfy claim iii.
53
Why are the inclusions reversed? They are reversed because the less
elements the Galois group has to permute, the smaller it gets. So if
the eld increases in elements to be xed, the Galois group decreases
in order.
Exercise 5.8 (8-7). Still working the same D, F, E
1
, and E
2
elds,
show that G(D/E
1
) is normal in G(D/F) and that G(E
1
/F) is iso-
morphic to G(D/F)/G(D/E
1
). Repeat for E
2
.
Solution. We can show that the left cosets match the right cosets,
since were working with a particular case. G(D/E
1
) is 1, = H.
So the left cosets are H, and , = . But the index is 2; so,
its normal, by some theorem weve seen in the past.
To show that G(E
1
/F) is isomorphic to G(D/F)/G(D/E
1
), we need
to nd a one-to-one and onto map that preserves the operations be-
tween them. Thats easy here because we have only two automor-
phisms in G(E
1
/F), and identity goes with identity, so the other map
goes with the other map.
Repeating for E
2
, we know that G(D/E
2
) is normal because the in-
dex is 2; its isomorphic to G(D/F)/G(D/E
2
) for the same reason
described above.
Exercise 5.9. Recall that we determined that the splitting eld of
x
4
2 over was Q(, i) where =
4

2. Also, weve calculated before


that the Galois group was isomorphic to the dihedral group with eight
elements: D
8
= , :
4
=
2
= e, =
3
). Now, let H be the
subgroup 1,
3
. Show by explicit calculation that the xed eld
Q(, i)
H
is equal to Q((1 i)).
Solution. No idea.
Theorem 5.8 (9-3). Let E
0
be a eld of characteristic zero containing
all the nth roots of unity. Let f E
0
and let E be a splitting eld for
x
n
f over E
0
. If w is any zero of x
n
f in E, then E = E
0
(w) and
G(E/E
0
) is cyclic.
Exercise 5.10. In Theorem 9-3, let E
0
= Q, n = 2, f = 7. Find the
splitting eld for x
2
7 over Q. Show that G(E/E
0
) has order 2.
Solution. The splitting eld for x
2
7 over Q is Q(

7) since x
2
7
= (x

7)(x +

7). So, in the theorem, E = Q(

7), and E
0
= Q.
We must show that G(Q(

7)/Q) is cyclic.
54
Here, the w of the theorem is either

7 or

7, so the theorem
says that we can adjoin

7 or

7 and get a cyclic Galois group


G(Q(w)/Q). We let w =

7. So now we must show that G(Q(w)/Q)
has order 2.
Since its cyclic of order 2, we must show that every element squares
to the identity. The identity itself does. The only non-trivial element
is that takes

7 to

7 and squares to the identity as well because


( ) takes

7 to

7 and then back to



7.
Here, Q(

7) is a normal extension of Q because (1) Q is the xed eld


of G(Q(w)/Q) or instead we could justify by saying that (2) Q(

7) is
the splitting eld over Q for x
2
7 over Q.
Exercise 5.11. In Theorem 9-3, let E
0
= Q, n = 2, f = 4. Show that
x
2
f in this case splits in E
0
so that G(E/E
0
) has only the identity
automorphism.
Solution. Its true that x
2
4 = (x 2)(x +2) which evidently splits
in Q. So, in the Galois group of Q over Q, theres really nothing
to permute because the roots of the polynomial were already in Q
anyway. The degree of the extension is 1, so we have only one element:
the identity.
Exercise 5.12. In Theorem 9-3, let E
0
= Q(r) where r is a primitive
third root of unity. Let n = 3, f = 7. Factor x
3
f into linear factors
in its splitting eld over E
0
. Show that the Galois group G(E/E
0
) is
cyclic of order 3.
Solution. The roots of unity for n = 3 are
1,
1
2
+

3i
2
,
1
2
+

3i
2
.
They form a cyclic group under multiplication. We cant let r = 1,
though, otherwise it wouldnt generate the group. Lets let
r =
1
2
+

3i
2
.
So x
3
1 splits over E
0
.
Were interested in x
3
7. Its roots are
3

7,

7
2
+

3
3

7i
2
,

7
2

3
3

7i
2
.
55
So, we really just need
3

7, because we can express the roots as r


3

7,
r
2
3

7, r
3
3

7 noticing that r
3
= 1, the identity of the cyclic group
of roots of unity. So the factorization can be written as
(x r
3

7)(x r
2
3

7)(x r
3
3

7).
So we adjoin
3

7 to get E = Q(r,
3

7). We now consider G(E/E


0
) and
show that its cyclic of order 3.
The Galois group xes Q(r), and roots must be taken to roots. Let
(
3

7) = r
3

7, so
2
(
3

7) = ((
3

7) = (r
3

7) = r
2
3

7 because is,
by denition, multiplying its argument by r. Finally,
3
= r
3
3

7 =
3

7. That shows that we have only 3 elements and that theyre cyclic
with r
3

7 as generator.
Exercise 5.13. Show that the Galois group for x
3
7 over Q is
isomorphic to S
3
and therefore nonabelian.
Solution. Take a look at the Fundamental Theorem. Notice how many
elements the group has, and that there is a theorem that claims that
the Galois group would be isomorphic to some subgroup of S
n
, but
here we will nd 6 elements and this can be seen by using the degree
of the extension eld and therefore it will have to be isomorphic to S
3
itself, which is nonabelian.
Exercise 5.14. Prove that A
5
is not abelian. Why doesnt the chain
S
5
1) provide a solvability chain for S
5
?
Solution. A
5
is not abelian because
(1324)(1432) = (234) ,= (123) = (1432)(1324)
and (1324), (1432) are even permutations.
The denition of solvability requires a chain of subgroups G = G
0

G
1
... G
t
= 1) to have the property that G
i
/G
i+1
is abelian for
i = 0, 1, ..., t 1.
For the chain S
5
1) satisfy the denition, we need G
0
/G
1
abelian,
but G
0
/G
1
= S
5
/1) = S
5
, and S
5
isnt abelian.
56

Das könnte Ihnen auch gefallen