Abstract Algebra I

(Math 5550)

25
◦ .................
...... ....................
....... ... ..........
.......
.........
.... .....
.
◦
.
...
......
..... .............
........
........
........ 13
41◦ .... . .
. ..... .... ..... .
.
........
. . .
.....
.
34 .......
.......
........
...
....... ... ............
◦
.. .
. . .. . .......... ......
...... . ..... ..
.. ......
. . .. ... ............. ...........
.... . ......... . ....... ..
.. ... ... ........ ...........
51 ◦
.. . .... .
....
.
12 . .
. ......
... ...
.
. .... .
.. ............................
◦ 42
.
....
.
.
..
. ..
◦
. ..
.
.
.
... .........
... .. ...... ....
.
.
.
..
..... ...
..
... ................... ... . ...
53 ◦
..........
..
... ...
... .. ... .. .. ... .. ...◦ 54
...
. ...
.
...
. ...
...
32 ◦
... .... ....... ......... ....... . .. ...
... ... ..
. ...
.. .. ... . .
.
... .. ... . .. ...
.. ... .... . ..
.. ... ... ... ...
... .. ... ... .
.. ...
.. ... .. .. ...
.. . .
... .. ...
... .. ... .... .
. ...
..
... ... ... .. .... ...... . .
.
..................... ◦ 23 ....
45 ◦ .............................................. ... ... ...
..
...
..
.
.... .......
. ..
...
...
...
....... .. .
...
. .
...... ....
.
. ◦ 35
.
.. . ...
◦
........ ... . .
.. .

24 ◦
....
......... ....... ... .. ..... ... ..
...
◦ 15
......
...... .........
......
.
...
..... .....
..
21 ...
...
... ... .
.
...
...
...... ....... .....
..
...... ....... ..... .
... ...
......
......
......
...... .. ...........
.
.....
....... .. ◦
◦ ........
........
........
.......
43
.......
.......
.....
.
.. .
. .
.......
....... ... .
14
31 ........
◦
........ ... .
........ ...
........ . .......
.......
.......
........ .. ............
◦
...............

52

c 1993, 2004, 2005


G. Eric Moorhouse
ii
 Preface
These lecture notes were used in teaching Abstract Algebra I (Math 709, later Math 5550) at
the University of Wyoming in the Spring 1990, 1993 and 2004 semesters. Many of the exercises
included were also used in homework assignments and tests during the semester. I expect there
remain several gaps, misprints and errors in these notes. These will be posted on the website
[Link]
I would appreciate receiving any suggestions and corrections which I can include at this site, and
which will be useful in future revisions of these notes.
The main topics presented are groups, rings, fields and Galois theory. Several standard topics
are omitted (such as group presentations, factorization theorems for rings, linear representations
and characters of finite groups) or stated without proof (e.g. the Jordan-Hölder Theorem) in order
to reach the goal of rudimentary Galois Theory by the end of the course, a goal we proved to
be practical and worthwhile. In such a course as this, which is required for all our graduate
students, not just those specializing in pure mathematics, I believe that abstract group theory
is less significant than understanding the concept of group actions. Accordingly I have stressed
permutation actions of groups, which shows how groups arise ‘in nature’ and stays fairly close to
the historical origins of finite group theory (as groups of permutations, including Galois actions on
zeroes of polynomials).
In these notes, I assume no prior expertise in groups or rings. I do expect the reader has
reasonable experience with sets and functions, elementary number theory, and an appreciation for
the notion of a proof. Some knowledge of linear algebra is also assumed, in particular matrix
arithmetic, and the concepts of basis and dimension. I do not expect that these notes would be
suitable, however, for any of our undergraduate courses, except perhaps for individual study in
special cases.
With so many good abstract algebra textbooks on the market, why these lecture notes? Math-
ematical textbooks tend to include far more material in every chapter than a typical first algebra
course can reasonably cover. This is the natural outcome of the motivation of authors and pub-
lishers to offer a small number of current titles of a very comprehensive nature, in order to accom-
modate all the whims of every instructor, whose job it is then to pick and choose what portion
of the textbook appeals to her/him. Unfortunately the student must then work harder to extract
from the textbook what the instructor requires. By contrast, these contain no more and no less
than students in my class are expected to know.
In order to get through as much of the most basic material as possible, many of the standard
results are found only in the homework exercises. Accordingly, students are expected to solve all of
the exercises; any exercises which cannot complete themselves as assigned homework, they should
later seek help in learning to solve. In order to help keep these expectations realistic, ample detailed
hints are provided for many of the exercises. Exception: Those problems designated by asterisks
(∗ ) are considered more challenging and may be considered supplementary.
My desire in these notes has been to be as informal and visual as possible. For example, many
results are outlined in discussion format without being dignified as Theorem so-and-so.
Eric Moorhouse
1993, 2004, 2005

iii
iv
 Prologue
The goal of this course is to cover certain topics in groups (especially finite groups), rings
and fields. Many standard topics in group theory and ring theory will be bypassed so that
we will have time to reach our goal by the end of the course, namely some basic Galois
theory.
By the time we have done a little Galois Theory, you will have attained an appreciation
for the interrelationships between different areas of mathematics. It is these interrelation-
ships that so often make mathematics, and especially Galois theory, so beautiful.
When I was in secondary school in Saint John, the Head of our school’s Math/Science
Department showed us a problem that he said was too difficult for him. Of course this
made the problem irresistible to me! The problem was as follows:

A 10 ft. × 10 ft. building has a 100 ft. flagpole erected on one of its upper corners.
During a storm, the flagpole is cracked at a point of height x above the roof, such that
the tip of the flagpole just touches the ground, while the flagpole also just touches
another corner of the building (see diagram).

Find x.

This led me to a cubic polynomial equation for x, namely

11x3 − 490x2 + 100x + 500 = 0.

Numerical approximation shows that this has three real roots: 44.317, 1.133, and the
extraneous root −0.905. However, I was not interested in approximations. I wanted the

v
real thing! After all, the quadratic equation ax2 + bx + c = 0 could be solved exactly using
the Almighty Formula √
b ± b2 − 4ac
x= .
2a
Every day I would rush home from school to my pile of rough work, in a futile attempt to
factor my cubic equation, or transform it into a quadratic equation, or even find another
equation for x which would simplify the problem. It was weeks later that I discovered that
every cubic equation could be transformed (by an affine change of variable) to one of the
form x3 + px = q, for which Fontana’s solutions are



3 q p3 q2 3 q p3 q2
x1 = + + + − + ,
2 27 4 2 27 4
x2 = . . . , x3 = . . . .

I also discovered that similar solutions existed for the general quartic (degree 4) polynomial
equations, and that these general solutions required the extraction of fourth roots, as
expected. It was only much later that I learned why the general quintic (degree 5) equation
has no such solution in terms of fifth roots (or in fact using any radicals. Indeed Galois
theory shows that the question of whether a given polynomial equation is solvable by
radicals, reduces to the question of whether a certain group (now known as the Galois
group of the polynomial) is ‘solvable’.
We mention another couple of classical problems whose answers depend (somewhat
surprisingly) on field theory. One is the impossibility of trisecting an arbitrary angle using
straightedge and compass. You will see once and for all why this is so. Another is the
impossibility of finding an antiderivative

2
ex dx

in elementary terms. This topic we may possibly present in an extra seminar, once we
have covered enough field theory.
Let’s forge ahead, then, with an introduction to group theory.

vi
0. Integers

We review some preliminary facts concerning the ring Z of integers. (Although the defini-
tion of a ring does not appear until Section 14, nevertheless all these facts about Z should
be familiar to you.)
Let a, b ∈ Z. We say that a divides b (or a is a divisor of b, or b is a multiple

of a)
if b = da for some d ∈ Z. We
write the statement ‘a divides b’ symbolically as a
b; if a
does not divide b, we write a
b. The following results are well known.

0.1 Proposition. Let a, b, c ∈ Z.

(i) If a
b and b
c, then a
c.





(ii) If a
b and a
c, then a
(b + c) and a
(b − c).

0.2 Theorem (Division Algorithm). Let a, d ∈ Z and suppose that d > 0. Then
there exist unique integers q, r such that a = qd + r and 0 ≤ r < d.

The algorithm (i.e. procedure) for determining the quotient q and remainder r is familiar;
for example when we divide 103 by 7, we obtain 14 as the quotient and 5 as the remainder:

14
.................................
...
7103 ..
..
.
7
....................
33
28
....................
5

Thus 103 = 14 · 7 + 5. Calling Theorem 0.2 an ‘algorithm’ is rather a misnomer, but a

popular one.
The divisors of 12 are ±1, ±2, ±3, ±4, ±6, ±12; the divisors of 18 are ±1, ±2, ±3, ±6,
±9, ±18. Thus the common divisors of 12 and 18 are ±1, ±2, ±3, ±6. The greatest (i.e.
largest) of these common divisors is therefore 6. We write gcd(12, 18) = 6 (the greatest
common divisor of 12 and 18 is 6). Given any two integers a, b ∈ Z, not both zero,
we similarly denote their greatest common divisor by gcd(a, b). (Note that gcd(0, 0) is
undefined since the common divisors of 0 and 0 are all the integers.) We say m and n are
relatively prime if gcd(m, n) = 1; for example, 12 and 35 are relatively prime.
Similarly, the multiples of 12 are 0, ±12, ±24, ±36, ±48, . . .; the multiples of 18 are
0, ±18, ±36, ±54, . . .; and the common multiples are 0, ±36, ±72, . . .. Of all the positive
common multiples of 12 and 18, the least is 36, and we write lcm(12, 18) = 36 (the least
common multiple is 36).
2 0. INTEGERS

By the Fundamental Theorem of Arithmetic, every positive integer has a unique prime
factorization. We see from the prime factorizations 12 = 22 31 and 18 = 21 32 that
gcd(12, 18) = 21 31 = 6. Similarly, lcm(12, 18) = 22 32 = 36. More generally, if m, n ∈ Z
are nonzero, let p1 , p2 , . . . , pk be the primes dividing at least one of m and n. Write the
prime factorizations of m and n as
 
m= pri i , n= psi i
i∈I i∈I

where ri , si ≥ 0 for all i ∈ I = {1, 2, . . . , k}. Then

0.3 Proposition.
 min{ri ,si }  max{ri ,si }
(i) gcd(m, n) = pi ; and (ii) lcm(m, n) = pi .
i∈I i∈I

From these formulae it follows easily that mn = gcd(m, n) lcm(m, n), so determining
the greatest common divisor of two given integers is just as hard (or just as easy) as
finding their least common multiple. For large numbers, prime factorizations are not easily
computed, so Proposition 0.3 will be of little value in computing gcd(m, n). However,
Euclid’s Algorithm gives us the answer quickly, by repeated application of the Division
Algorithm. For example, we determine gcd(108, 74):

108 = 1 · 74 + 34
74 = 2 · 34 + 6
34 = 5 · 6 + 4
6 = 1·4 + 2
4 = 2·2 + 0

and the last nonzero remainder gives gcd(108, 74) = 2. Moreover reversing these steps, we
are able to express 2 as an integer-linear combination of 108 and 74, thus:

2=6−4
= 6 − (34 − 5 · 6)
= 6 · 6 − 34
= 6(74 − 2 · 34) − 34
= 6 · 74 − 13 · 34
= 6 · 74 − 13(108 − 74)
= 19 · 74 − 13 · 108.

More generally
 0. INTEGERS 3

0.4 Theorem (Euclid’s Algorithm). Let m and n be integers, not both zero, and
let d = gcd(m, n). Then there exist integers x, y such that d = mx + ny.

The algorithm described above (by example) succeeds very quickly even for very large inte-
gers m and n because, with repeated application of the Division Algorithm, the remainders
decrease in size rather quickly. Again, it is a popular misnomer (to which we accede) to
refer to Theorem 0.4 as Euclid’s Algorithm, which is more precisely the algorithm described
above for computing d, x and y.
It is perhaps already apparent that the properties listed above hold for rings other
than Z; in particular the ring R[X] of polynomials in X with real coefficients, admits a
Division Algorithm, and so Euclid’s Algorithm applies also in R[X]. The rings Z and R[X]
are examples of the class of rings known as Euclidean domains, considered in Section 15,
for which these results apply. But because the elementary properties of groups make use
of these elementary properties of the integers, we have reviewed these properties at the
outset.
Exercises 0.
1. Let m and n be positive integers. Prove that there exist relatively prime integers m
and n
such that
m
| m, n
| n and lcm(m, n) = m
n
.
 
Hint: Write m = i∈I pri i and n = i∈I psi i in the 
notation of Proposition 0.3. Let I1 = {i ∈ I :

r s
ri ≥ si } and I2 = {i ∈ I : ri < si }. Consider m
= i∈I1 pi i and n
= i∈I2 pi i .
 CHAPTER I

Groups
1. Definitions
A binary operation on a set G is a function G × G → G. Examples of binary operations
include:
addition of real numbers, i.e. + : R × R → R, (x, y) → x + y;
multiplication of real numbers, i.e. × : R × R → R, (x, y) → x × y;
punctiliation of real numbers, i.e. ∗ : R × R → R, (x, y) → (x4 − y 3 )/(x2 + y 2 + 1).
(I’m sorry, I don’t know what punctiliation is good for. I made it up just now.) Some books
use a symbol like ∗ or ◦ to denote an arbitrary binary operation. However in the general
case we shall use juxtaposition of elements to denote the binary operation, so that the image
of the pair (x, y) ∈ G × G is simply denoted xy ∈ G. It is important to remember, though,
that this operation need not correspond to any usual notion of ‘multiplication’; indeed, the
operation (x, y) → xy may represent actual addition, or something very unfamiliar, like
punctiliation.
Certain choices of a binary operation on a set G will make G into a group, while
others will not. In the above three examples, R is a group under addition, but not under
multiplication or punctiliation. For this reason, we might say that the pair (R, +) is a
group, while (R, ×) and (R, ∗) are not. However, if G is a set on which we have one clearly
defined binary operation, it is unambiguous and acceptable to say simply that G itself is
a group.
A group is a set G, together with a binary operation on G (here denoted simply by
juxtaposition of elements of G), such that
(i) there exists e ∈ G such that xe = ex = x for all x ∈ G;
(ii) for all x ∈ G, there exists y ∈ G such that xy = yx = e; and
(iii) the binary operation is associative, i.e. (xy)z = x(yz) for all x, y, z ∈ G.
If G is a group, then by (i), G contains an identity element for the binary operation.
In particular, every group is nonempty. It is easy to see that the identity of G is unique,
for if e1 and e2 are identities for G, then e1 = e1 e2 = e2 . We will denote the (unique)
identity of G by e (or sometimes 1 or 0).
Furthermore, ‘inverses’ (as in (ii)) are unique in G. For if x ∈ G has two inverses
y1 , y2 ∈ G such that xy1 = y1 x = e = xy2 = y2 x then

y1 = y1 e = y1 (xy2 ) = (y1 x)y2 = ey2 = y2 ,

6 I. GROUPS

as claimed. We will usually denote the (unique) inverse of x in G by x−1 . Furthermore we

denote
m times

 
x = ((· · · ((xx)x)x) · · · x)x = xx · · · x
m

for any positive integer m. (The parentheses above are irrelevant, of course, by associativ-
ity.) It is then straightforward to check that

xm xn = xm+n , (xm )n = xmn

for all integers m, n, with the additional conventions that x0 = e and x−m = (xm )−1 =
(x−1 )m for m < 0.
Note that the binary operation for a group is not required to be commutative. If
xy = yx for all x, y ∈ G, then G is said to be abelian; otherwise G is nonabelian.
The order of G is by definition |G|, i.e. the cardinality of the set of group elements.
There is no reason why |G| must be finite, but if it is, we say that G is a finite group.
The order of an element x ∈ G is the smallest positive integer m such that xm = e. (If
no such integer exists, we say that x has infinite order.) For example, e is the unique
element of order 1.

Exercises 1.
1. For elements x and y in a group G, prove that (xy)−1 = y −1 x−1 . (This is the so-called Shoe-Sock
Theorem: The opposite of putting on socks and shoes, is to first remove the shoes, then remove the
socks.)
2. Prove that if x2 = e for every element x in a group G, then G is abelian.
3. Let G be a group, and suppose x ∈ G has order m. Show that xi = xj iff i ≡ j mod m. In particular,
xk = e iff k is divisible by m.
Hint: Suppose xk = e. By the Division Algorithm 0.2, write k = qm + r where 0 ≤ r < m. Deduce
that xr = e and so r = 0. More generally if xi = xj then xi−j = e and the previous reasoning
applies.
4. Let x, y be elements of a group G such that xy = yx. Prove that if m = |x| and n = |y| are relatively
prime integers, then |xy| = mn.
Hint: Suppose (xy)k = e. Since x and y commute, this means that xk y k = e. Consider the element
z = xk = y −k . Then z m = (xk )m = (xm )k = e and z n = (y −k )n = (y n )−k = e. By Exercise 1.3, |z|
divides both m and n. . .
5. (a) Show that every element of a finite group has finite order.
(b) Find an example of an infinite group in which every element has finite order.
6. Let x and y be elements of a group G such that xy = yx. Suppose that x has finite order m, and y
has finite order n. Show that there exist r, s ∈ Z such that |xr y s | = lcm(m, n).
Hint: There exist integers m
, n
as in Exercise 0.1. Let r = m/m
and s = n/n
and apply
Exercise 1.4.
7. Let G be a finite group of even order. Prove that G has an element of order 2.
Hint: If G has no element of order 2, then the nonidentity elements of G are partitioned into pairs
{g, g −1 }.
 I. GROUPS 7

2. Examples

Example: Multiplicative Rationals. The set of nonzero rational numbers under

×
multiplication, i.e. Q = Q {0}, forms an infinite abelian group. Of course e is really 1
........

in this case. Likewise both R× and C× are infinite abelian groups under multiplication.
(However, the nonzero integers do not form a multiplicative group, owing to the lack of
multiplicative inverses.)

Example: Additive Integers. The set of integers under addition forms an infinite
abelian group. (In this case note that xy is really x + y, e is really 0, xm means x + x +
· · · + x = mx, and x−1 is really −x.) Similarly each of the sets R, Q and C forms an
infinite abelian group under addition.

Example: General Linear Groups of Degree n. The set of all invertible n × n

matrices with real entries, denoted GLn (R), forms an infinite group, under the usual matrix
multiplication. This group is nonabelian for n > 1. Similarly we have the groups GLn (C)
and GLn (Q).

Example: Symmetric Groups. A permutation of degree n is a bijection from

the set {1, 2, 3, . . . , n} to itself. For a given n, the set of all such bijections forms a group
under composition, called the symmetric group of degree n, and denoted Sn . This
group has order n!, and is nonabelian for n > 2.
We introduce notation for permutations as follows. Every permutation is the product
(i.e. composite) of disjoint ‘cycles’. For example in S6 , the cycle of length 4 (or 4-cycle)
denoted (2634) is the permutation defined by

1 2 3 4 5 6
(2634) ↓ ↓ ↓ ↓ ↓ ↓ .
1 6 4 2 5 3

In this example, (2634) takes 1 to 1, 2 to 6, etc. We could also have denoted this per-
mutation by (1)(2634)(5), but it is customary here to suppress writing the cycles (1) and
(5) of length 1. Note also that (2634) = (3426) = (4236) = (6342). We will also express
the action of (2634) by writing 1(2634) = 1, 2(2634) = 6, etc. (This is preferable to saying
(2634)(1) = 1, (2634)(2) = 6, etc., since we don’t want our use of parentheses to start
getting ambiguous.) An example of a permutation which is the product of two disjoint
3-cycles is (124)(365), which has the effect

1 2 3 4 5 6
(124)(365) ↓ ↓ ↓ ↓ ↓ ↓ .
2 4 6 1 3 5
8 I. GROUPS

Composing the above permutations, we have

1 2 3 4 5 6
(2634) ↓ ↓ ↓ ↓ ↓ ↓
1 6 4 2 5 3 ,
(124)(365) ↓ ↓ ↓ ↓ ↓ ↓
2 5 1 4 3 6

which shows that (2634)(124)(365) = (1253). We multiply (or compose) permutations

from left to right like this, since we want to be able to say
 (124)(365)
2(2634) = 6(124)(365) = 5 = 2(1253) = 2(2634)(124)(365).

(Warning: Some books define multiplication of permutations using right-to-left compo-

sition!) It is customary to write the identity of Sn as (1), which is an abbreviation for
(1)(2)(3) · · · (n), the permutation fixing every point.

Example: Cyclic Groups.  of order n may be written as Cn =

The cyclic group
2
{e, x, x , . . . , xn−1
} where x = e, and x = e iff n  m; see Exercise 1.3. In this example,
n m

the order of x is n, which is the same as the order of Cn .

Example: Isometry Groups. An isometry of the Euclidean plane is a transfor-

mation R2 → R2 which preserves distance. Isometries of the plane include rotations,
reflections, translations and glide reflections. Examples of these are the maps

(x, y) → (y, −x); (x, y) → (x, −y);

(x, y) → (x + 2, y − 3); (x, y) → (x + 1, −y)

respectively. The set of all isometries of the plane forms an infinite nonabelian group under
composition. We may also consider the isometry group of Euclidean 3-space, an even larger
nonabelian group.

Example: Dihedral Groups. The dihedral group of degree n and order 2n is the
group Dn consisting of all symmetries of a regular n-gon in the plane. (Warning: Some
books use the notation D2n in place of Dn .) By a symmetry of an object in 2-space or
in 3-space, we mean an isometry of the corresponding 2-space or 3-space which preserves
the object.
Consider a regular n-gon in the plane with vertices labeled 1, 2, 3, . . . , n in a counter-
clockwise fashion as shown. This n-gon has n rotational symmetries I, R, R2, . . . , Rn−1
where R represents a counter-clockwise rotation through an angle of 2π n about the center
of the n-gon. It also has n reflective symmetries. The axes of these reflections consist of
lines through the center of the n-gon, passing through the vertices of the n-gon and the
 I. GROUPS 9

midpoints of its sides. Let T be the reflective symmetry whose axis is the line joining
vertex 1 with the center of the n-gon.

4 3 4 3
......
•
...............................................................
...... •
......
......
......
......
• •
...............................................................
......
......
..
..
....... ......
..
..
....... ......
.... ...... .... ......
...... .. ......
. ...... . . ......
...... 2 ........ 2
......
. ...... ..... ........... . ......
.
.
.
. ..
...
...... .....
• ...
...
... .
.
.
• ...
...
...
.
. ...... ...
...
...
... . ...
...
.
. ....
.
. ...
...
R
...
... .
. ...
...
...
.
. ... ... ... . ...
. ..
..
..
. ... 2π
...
..
...
.. . .
......
.
.
......
.
...
... T....
. . ..
...... ..
...
...
... . ... ... ... .......
.. .. n .. .
. .
. ... ...
... ...
.
. .
... ◦
.......... ....... ....... ....... ....... ....... ....... ....... ......... ......... ..
.. .
.
•. .
..
. .
..
.
◦ .
....... ........ ....... ....... ........... ....... ....... ....... ....... ....... ....... ........... ....... ....... ....... .......... ....... .......... .
..
.
• .... ...
. 1
... . ........
....
........
....
1 ... .........
. ... . ... ...
.
..... . .
...
.
. .. . ...
.. ...
. ... . .
. .... . .
...
. . . ..
... ...
. .. .
.
. .............. • .
.
•
......
.
......
..
.
. .....
...... n .
. .....
......n
.
. . . ... .
. . . ...
. .
. . . . . . . .
. . . . . . . . . . . . . . . .

Every element of Dn has the form Ri T j where i ∈ {0, 1, 2, . . . , n−1}, j ∈ {0, 1}, and it is
not hard to check that elements of Dn multiply according to the law

i j k  Ri+k T  , if j is even;
(R T )(R T ) = i−k +1
R T , if j is odd.

Exercises 2.
1. List all orders of elements of S5 , and the number of elements of each order. Check that the total
number of elements equals 5! = 120.
2. How many elements of S6 commute with (12)? (Do not list them all!)
3. A transposition is a 2-cycle in Sn . For example, S3 has exactly three transpositions: (12), (13)
and (23). Observe that every 3-cycle in S3 is a product of two transpositions: (123) = (12)(13),
(132) = (13)(12). Show that every permutation in Sn is a product of at most n − 1 transpositions.
(These transpositions, however, are not necessarily disjoint!) This says that Sn is generated by its
transpositions.
4. Let n ≥ 2 be an integer. Consider the polynomial
f (X1 , X2 , . . . , Xn ) = (Xj − Xi ),
1≤i<j≤n

which is homogeneous of degree n(n − 1)/2. For any σ ∈ Sn , define a new polynomial by
fσ (X1 , X2 , . . . , Xn ) = f (X1σ , X2σ , . . . , Xnσ ).
For example if n = 3 then f (X1 , X2 , X3 ) = (X2 − X1 )(X3 − X1 )(X3 − X2 ) and f(12) (X1 , X2 , X3 ) =
f (X2 , X1 , X3 ) = (X1 − X2 )(X3 − X2 )(X3 − X1 ) so that f(12) = −f .
(a) For any σ ∈ Sn , show that fσ = ±f . Thus we may define the sign of σ, denoted sgn σ, by
sgn : Sn → {1, −1}, fσ = (sgn σ)f.

(b) If σ ∈ Sn and τ ∈ Sn is a transposition, show that sgn(στ ) = − sgn σ. In particular for σ = (1),
this says that the sign of every transposition is −1.
10 I. GROUPS

(c) Using (b) and Exercise 3, prove that sgn(σπ) = sgn(σ) sgn(π) for all σ, π ∈ Sn .
(d) Prove that the set An = {σ ∈ Sn : sgn(σ) = 1} is a group of order n!/2. This is called the
alternating group of degree n.
(e) Show that any permutation σ ∈ Sn is expressible either as a product of an even number of
transpositions (in which case σ ∈ An ), or as a product of an odd number of transpositions (in
which case σ ∈ Sn ...... An ); but that no permutation is expressible both as a product of an even
number of transpositions, and as a product of an odd number of transpositions. So we call a
permutation even if its sign is +1, or odd if its sign is −1.
(f) List all the permutations in A4 .
5. Show that every isometry of the Euclidean plane is the product of at most three reflections.
6. Show that every isometry of the Euclidean plane is (a) continuous, and (b) bijective.
7. Let G be the symmetry group of the solid in R3 consisting of all points (x, y, z) such that |x| ≤ 2,
|y| ≤ 2, and |z| ≤ 0.1. Say as much as you can about the structure of G, including its order, the
number of elements of each order, whether or not G is abelian, and if possible relating G to any groups
we have studied so far.

3. Isomorphism
If G and H are groups, then an isomorphism from G to H is a bijection φ : G → H such
that φ(xy) = φ(x)φ(y) for all x, y ∈ G. If there exists an isomorphism from G to H, then
G and H are said to be isomorphic, and we write G ∼ = H. Clearly, group isomorphism
is an equivalence relation. It is also clear that isomorphic groups have the same order, as
well as the same number of elements of each order. In fact isomorphic groups have all the
same abstract group-theoretical properties, so that two groups which are isomorphic are
usually considered to be the same group.
For a finite group G, a Cayley table (or group table or multiplication table) is
a square table specifying the products of all pairs of elements of the group. For example,
if G1 = {1, 3, 5, 7} is the group whose operation is multiplication modulo 8, then a Cayley
table for G1 is given by
1 3 5 7
1 1 3 5 7
3 3 1 7 5
5 5 7 1 3
7 7 5 3 1
Similarly, if G2 is the group consisting of all diagonal 2 × 2 matrices with entries ±1, under
the usual matrix multiplication, then a Cayley table for G2 is given by
1 0  1 0  −1 0  −1 0 
0 1 0 −1 0 1 0 −1
1 0 1 0 1 0
 −1 0  −1 0

0 1 0 1 0 −1 0 1 0 −1
1 0
 1 0
 1 0 −1 0
 −1 0 
0 −1 0 −1 0 1 0 −1 0 1
−1 0  −1 0  −1 0  1 0 1 0 
0 1 0 1 0 −1 0 1 0 −1
−1 0  −1 0  −1 0  1 0  1 0
0 −1 0 −1 0 1 0 −1 0 1
 I. GROUPS 11

Observe that
 the Cayley  table for G1 becomes the
−1Cayley
 table for G2 if we replace
1 → 10 01 , 3 → 10 −1 0
, 5 → −1 0
0 1 , and 7 →
 0
0 −1 throughout. This defines an
isomorphism from G1 to G2 . In general, two finite groups G and H are isomorphic iff a
group table for H may be obtained from a group table for G by renaming the elements
appropriately, and possibly permuting rows and columns.
It is clear that given n, there are only finitely many groups of order n up to isomor-
phism; for if G = {x1 =e, x2 , . . . , xn }, then there are only finitely many ways to fill the
n2 entries of the table for G, and only a few of these possible tables can be expected to
form groups. Clearly any row (or column) of a Cayley table for G must consist of all the
elements x1 =e, x2 , . . . , xn in some order. (Why?) We may therefore ask for a list of all
groups of order n up to isomorphism.
If G is a group of order 2, say G = {e, x}, then there is only one way to complete the
table
e x
e e x
x x
to a Cayley table, namely
e x
e e x .
x x e
This shows that any group of order 2 is cyclic, i.e. isomorphic to C2 .
If G is a group of order 3, say G = {e, x, y}, then we start with a Cayley table for G
as follows:
e x y
e e x y
x x
y y
The next row of the table fills in as either
e x y e x y
e e x y e e x y
or .
x x e y x x y e
y y y y

But we cannot have two y’s in the same column, so the table must complete uniquely as

e x y
e e x y
x x y e
y y e x
12 I. GROUPS

Thus G = {e, x, x2 =y} ∼ = C3 , i.e. every group of order 3 is cyclic.

How many isomorphism classes of groups are there of groups of order 4? Let G =
{e, x, y, z}. Starting with
e x y z
e e x y z
x x ,
y y
z z
the next row completes as either

e x y z e x y z e x y z
e e x y z e e x y z e e x y z
(i) x x e z y , (ii) x x y z e , or (iii) x x z e y .
y y y y y y
z z z z z z

However, cases (ii) and (iii) are equivalent, since interchanging the last two columns of (iii),
then interchanging the last two rows, then renaming y ↔ z, yields case (ii). So disregard
case (iii). It is easy to see that (ii) leads uniquely to

e x y z
e e x y z
x x y z e .
y y z e x
z z e x y

In this case G = {e, x, x2 =y, x3 =z} ∼

= C4 , i.e. G is cyclic. Case (i) may be completed to a
group table in just two ways:

e x y z e x y z
e e x y z e e x y z
(i.a) x x e z y , or (i.b) x x e z y .
y y z e x y y z x e
z z y x e z z y e x

However, case (i.b) is {e, y, y 2=x, y 3 =z}, which is cyclic, isomorphic to case (ii). Case (i.a)
is not cyclic, since it has g 2 = e for all g ∈ G. Case (i.a) is called the Klein 4-group.
(Note that examples G1 and G2 above belong to this isomorphism class.) Thus there are
exactly two groups of order 4 up to isomorphism: the cyclic group C4 , and the Klein
4-group.
 I. GROUPS 13

Of course for infinite groups, or even large finite groups, tables are not helpful for spec-
ifying the group or checking isomorphism. As an example of an isomorphism of infinite
groups, consider R under addition, and the positive real numbers (0, ∞) under multiplica-
tion. Clearly log : (0, ∞) → R is an isomorphism; here log(xy) = log(x) + log(y) for all
x, y ∈ (0, ∞). (Here we have been careful to write ‘+’ for the operation in R, but multi-
plication for the operation in (0, ∞).) However, R× = (−∞, 0) ∪ (0, ∞) is not isomorphic
to the additive group of R. One way to see this is to note that R× has an element −1
of order two, whereas the additive group R has no element of finite order other than the
identity.
Exercises 3.
1. Let G be a group of order 5. By considering possibilities for the Cayley table of G, prove that G is
cyclic.
2. Let H and K be groups, with respective group operations. On the usual set-theoretic Cartesian
product
H × K = {(h, k) : h ∈ H, k ∈ K},
we define componentwise multiplication:
(h, k)(h
, k
) = (hh
, kk
) ∈ H × K for (h, k), (h
, k
) ∈ H × K.
Show that this makes (H, K) into a group. This is called the direct product of H and K. Observe
that the Klein 4-group is isomorphic to C2 × C2 , so that the only two groups of order 4 (up to
isomorphism) are C4 and C2 × C2 .
3. If n = n1 n2 · · · nr where the positive integers n1 , n2 , . . . , nr are mutually relatively prime (i.e.
gcd(ni , nj ) = 1 whenever i = j), show that
Cn ∼
= Cn1 × Cn2 × · · · × Cnr .
4. Show that the multiplicative group R× of nonzero real numbers, is isomorphic to the direct product
of (0, ∞) with a group of order 2.
5. Show that two cycles in Sn commute iff they are disjoint, or one is a power of the other.
6. Let σ ∈ Sn . Show that |σ| is the least common multiple of the lengths of the disjoint cycles in σ.
7. Let Fp = {0, 1, 2, . . . , p−1} be the field of integers modulo p, with addition and multiplication modulo p.
Define the general linear group over Fp to be the multiplicative group of all invertible n×n matrices
over Fp .
(a) Determine the order of GLn (Fp ).
Hint: For a typical element A ∈ GLn (Fp ), consider how many possible choices there are for the
first column of A, then how many choices for the second column of A, then how many choices for
the third column of A, etc.
(b) Show that GL2 (F2 ) ∼
= S3 .
8. Let G = C3 × C3 × C3 , and let H be the group of all 3 × 3 matrices of the form
⎛ ⎞
1 a b
⎝0 1 c⎠
0 0 1
with a, b, c ∈ F3 . Then G and H are groups of order 27.
(a) How many elements of each order does G have? How many elements of each order does H have?
(b) Is H isomorphic to G? Explain.
Hint: Is G abelian? Is H abelian?
14 I. GROUPS

4. Subgroups and Cosets

A subset H ⊆ G is called a subgroup of G (denoted H ≤ G) if H forms a group with

respect to the binary operation on G, restricted to H. It is easy to see that H is a subgroup
of G, iff H ⊆ G is a nonempty subset which is closed under multiplication and inversion;
or equivalently, Ø = H ⊆ G and xy −1 ∈ H whenever x, y ∈ H.
Examples of subgroups include the trivial subgroup {e}, often denoted simply by 1,
and G itself. Also each x ∈ G generates a cyclic subgroup, the set of all powers of x,
denoted by
x = {xk : k ∈ Z}.
If x is of finite order m, we simply have x = {e, x, x2 , . . . , xm−1 } ∼ = Cm . In particular,
note that |x| = |x|.
More generally, if x1 , x2 , . . . , xr ∈ G, then x1 , x2 , . . . , xr  denotes the subgroup
generated by x1 , x2 , . . . , xr ; this is the set of all finite products formed by the generators
x1 , x2 , . . . , xr and their inverses x−1 −1 −1
1 , x2 , . . . , xr . This may also be described as the
unique smallest subgroup of G containing x1 , x2 , . . . , xr ; see Exercise 4.2.
The set of all subgroups of a finite group is often represented in a diagram showing
their inclusions. For example, the subgroups of the cyclic group C12 = {e, x, x2 , . . . , x11 }
(where x12 = e) are all of the form x12/d  ∼ = Cd for d  12, and their inclusions are all
represented by the diagram:

x=C12
... .....
.... .....
..... .....
..
......
. .....
.....
..... .....
.... .....
..... .....
..
......
. .....
.. .....
..... .....
..... .....
.... .

x2 ∼
=C6 x3 ∼
=C4
... ..... ...
..... ..... .....
..... ..... .....
..
...... .....
..... ..
......
..... ..... .....
..... ..... .....
..... ..... .....
.
......
. ..... .
......
.
. ..... .
..... ..... .....
..... ..... .....
..... . .....

x4 ∼
=C3 x6 ∼
=C2
..... .
..... .....
..... ....
..... ..
......
.
..... .....
..... .....
..... .....
..... ....
..... ..
..
..... ..
..... .....
..... .....
. .....

e

For an arbitrary group G, we define the center of G as

Z(G) = {z ∈ G : zx = xz for all x ∈ G}.

We check that Z(G) is a subgroup of G, as follows. First, ex = x = xe for all x ∈ G, so

e ∈ Z(G). If z1 , z2 ∈ Z(G), then

(z1 z2 )x = z1 (z2 x) = z1 (xz2 ) = (z1 x)z2 = (xz1 )z2 = x(z1 z2 )

 I. GROUPS 15

for all x ∈ G, so that z1 z2 ∈ Z(G). Also z1 x = xz1 implies

z1−1 x = z1−1 (xz1 )z1−1 = z1−1 (z1 x)z1−1 = xz1−1

for all x ∈ G, so z1−1 ∈ Z(G). This shows that Z(G) ≤ G as claimed. By definition, we
have Z(G) = G iff G is abelian.
If A and B are subsets of a group G, we define

AB = {ab : a ∈ A, b ∈ B}.

In particular for H ≤ G and x ∈ G, we have Hx = {hx : h ∈ H}, called the right coset
of H containing x. It is easy to see that, for x, y ∈ G, H ≤ G,

y ∈ Hx ⇐⇒ yx−1 ∈ H ⇐⇒ H = Hyx−1 ⇐⇒ Hx = Hy ⇐⇒ Hx ∩ Hy = Ø.

(For example if Hx ∩ Hy = Ø, then h1 x = h2 y for some h1 , h2 ∈ H, so that yx−1 =

h1 h−1
2 ∈ H, and the other conclusions follow similarly.) This means that the right cosets
of H partition the elements of G. Furthermore for x ∈ G, the map

H → Hx, h → hx

is clearly bijective. Especially this shows that if H is finite, then all right cosets of H
have the same size, namely |H|. We define the index of H in G, denoted [G : H], as the
number of distinct right cosets of H in G. Since the right cosets of H have equal size and
they partition the elements of G, we have

|G| = [G : H] · |H|.

This proves

4.1 Lagrange’s Theorem. If G is a finite group with subgroup H, then |H| divides
|G|.

The truth of Lagrange’s Theorem is clear in the example

 of C12 detailed above. Warning:
the ‘converse’ of Lagrange’s Theorem is false! If d  |G|, it does not necessarily follow that
G has a subgroup of order d; see Exercise 4.7.
In the proof of Lagrange’s Theorem, we partitioned G into right cosets Hx, x ∈ G. It
is also true that G is partitioned into left cosets xH, x ∈ G. There are [G : H] left cosets,
each of size |H|, which gives another proof of Lagrange’s Theorem. However, left cosets
do not necessarily coincide with right cosets! For example if G = S3 and H = (12) < G,
then the right cosets of H are

H = {(1), (12)}, H(13) = {(13), (123)}, H(23) = {(23), (132)}

16 I. GROUPS

whereas the left cosets are

H = {(1), (12)}, (13)H = {(13), (132)}, (23)H = {(23), (123)}.

The number of cosets is the same in each case (namely, [G : H] = |G|/|H| = 62 = 3) and
the size of each coset is the same (namely |H| = 2), but the two partitions are not the
same:
H H(13) H(23) H

 
 
 
 
.......................................................................................................................................................................................................... ..........................................................................................................................................................................................................
... ........................................................ ........................................................ ........................................................ ..
.... ... .... .... .... .... .... ....
... ........................................................ ....................................................................................................................... ..
.... ... .... .... .... ....
⎫
. ..
.... ....
.. ..
.. ..
... ...
.. ..
.. ..
... ...
.. ..
.. ..
... ...
.. ..
. ..
.... ....
.. ..
.. ..
... ...
.. ..
.. ..
... ...
.. ..
⎪
⎬
... ... .. ... .. ... .. .. ... ... .. ... .. ..
. . . . . . ... ... . . . . ... ...
.... .... (12) .... .... (123) .... .... (23) .. .. .... .... (12) .... .... (123) (23) .. .. (23)H
.. ..
... ...
.. ..
.. ..
... ...
.. ..
.. ..
... ...
.. ..
.. ..
... ...
.. ..
.. ..
... ...
.. ..
.. ..
... ...
.. ..
.. ..
... ...
.. ..
⎪
⎭
... ... .. ... .. ... .. ... ... ... .. ... .. ...
.. .. ... .. ... .. ... .. .. .. ... .. . .
... ... .. ... .. ... .. .. ... ... .. .......................................................................................................................... ...
.
.... ....
.. ..
. .
.... ....
.. ..
. .
.... ....
.. ..
. .. ...
... ..
.. ..
.
.... ....
.. ..
. .
.... ......................................................................................................................... ....
.. .. .. ..
.
⎫
... ...
.. ..
... ...
.. ...
... ..
.. ...
.. ...
... ..
.. ...
.. ...
... ..
.. ...
... ...
.. ..
... ...
.. ...
... ..
.. ...
.. ...
... ..
.. ...
⎪
⎬
.. .. .. .. .. .. .. .. .. .. .. .. .. ..
... ... ... ... ... ... ... .. ... ... ... ... ... ..
(1) .
.... ....
. . (13)
... ....
. .
... ....
.(132) .. ... (1) .
.... ....
. .
... .... (13)
. (132) .. ... (13)H
.. ..
... ...
... ..
.. ...
... ..
.. ...
.. ..
... ..
.. ...
.. ..
... ...
... ..
.. ...
.. ..
... ..
.. ...
⎪
⎭
.. .. .. .. .. .. .. .. .. .. .. .. .. ..
... ... ... ... ... ... . . ... ... ... ... . .
.. .......................................................... .......................................................... ............................................................ .... .. .......................................................... ........................................................................................................................... ....
... . ... .
....................................................................................................................................................................................................... .......................................................................................................................................................................................................
 
  

G G
This motivates the definition of normal subgroup, in the next section.
Our formula |G| = [G : H]|H| makes sense in the case of infinite groups, with the
usual conventions (such as m∞ = ∞ whenever 1 ≤ m ≤ ∞). An example of an infinite
subgroup of finite index is given by the subgroup (0, ∞) in the multiplicative group R× .
In this case the index [R× : (0, ∞)] = 2 since there are just two cosets, namely (0, ∞) and
(−∞, 0). Note that the subgroup −1 = {1, −1} of order 2 has infinite index in R× , while
the infinite subgroup Q× < R× has infinite index (see Exercise 4.8).

4.2 Corollary. Let G be a finite group. Then every element of G has order dividing
|G|.

Proof. Let g ∈ G. Then |g| = |g|, which divides |G| by Lagrange’s Theorem.

Exercises 4.
1. If H, K ≤ G, show that H ∩ K ≤ G. More generally, if S is any nonempty collection of subgroups of
G, show that
S = H ≤ G.
H∈S

2. Let S be any subset of a group G, and let

S be the set of all products formed by elements of S and
inverses of elements of S. (If S = Ø, then
S = 1.)
(a) Show that
S is a subgroup of G. (We call
S the subgroup generated by S.)
(b) Show that
S is the intersection (see Exercise 4.1) of all subgroups of G containing S.
 I. GROUPS 17

3. Let H, K ≤ G. Recall that HK = {hk : h ∈ H, k ∈ K}. Is the product HK necessarily a subgroup?

Prove this or give a counterexample.
|H||K|
4. If H, K ≤ G where G is a finite group, show that |HK| = |H∩K|
.

5. Consider a cyclic group Cn =

g = {e, g, g 2 , . . . , g n−1 } here g n = e. Prove that for every divisor d | n,
Cn has exactly one subgroup of order d, namely the cyclic subgroup
g n/d ; and that these are the only
subgroups of Cn .
6. Prove that every group of prime order is cyclic.
7. Prove that A4 has no subgroup of order 6.
8. Let G = R× , the multiplicative group of nonzero real numbers.
(a) Show that the subgroup (0, ∞) has index 2.
(b) Show that the subgroup
−1 has infinite index.
(c) Show that the subgroup Q× has infinite index.
Hint: Consider cardinalities.
9. Let G be a group. Define the centralizer of an arbitrary element a ∈ G as CG (a) = {g ∈ G : ga =
ag}. Show that CG (a) is a subgroup of G.
10. Show that the center of GL2 (C) consists of nonzero scalar multiples of the identity matrix.
11. Let G be a group of finite order n. By Corollary 4.2, we have xn = e for all x ∈ G. The smallest
integer k ≥ 1 such that xk = e for all x ∈ G, is called the exponent of G.
(a) Prove that the exponent k divides n.
Hint: Using the Division Algorithm, write n = qk + r for some integers k, r where 0 ≤ r < k.
Then use Corollary 4.2.
(b) Suppose moreover that G is abelian. Prove that the exponent k is the maximum of |x| for all
x ∈ G. [This is not generally true for nonabelian groups, since S3 has exponent 6, but its elements
have orders 1, 2, 3 only.]
Hint: Use Exercise 1.6.
12. Let G be a group. A maximal subgroup of G is a proper subgroup M < G such that there is no
subgroup strictly between M and G, i.e. if a subgroup H satisfies M ≤ H ≤ G then either H = M or
H = G. The Frattini subgroup Φ(G) ≤ G is defined as the intersection of all maximal subgroups of G
(see Exercise 4.1).
(a) Prove that Φ(G) is a normal subgroup of G.
(b) Consider a subset S ⊆ G such that S ∪Φ(G) generates G (see Exercise 4.2). Prove that S generates
G. (Thus P hi(G)

5. Normal Subgroups and Quotient Groups

A subgroup H ≤ G is said to be a normal subgroup if xH = Hx for all x ∈ G, i.e. if

the partition of G into left cosets of H, coincides with the partition of G into right cosets
of H. (Warning: This does not mean that xh = hx for all h ∈ H.) We write H ≤ G if H
is a normal subgroup of G; or H < G if H is a normal proper subgroup of G. Note that
xH = Hx ⇐⇒ x−1 Hx = x−1 xH = H, by associativity. Furthermore if x−1 Hx ⊆ H
for all x ∈ G, then replacing x by x−1 gives (x−1 )−1 Hx−1 ⊆ H, i.e. H ⊆ x−1 Hx. This
proves
18 I. GROUPS

5.1 Proposition. A subgroup H ≤ G is normal in G iff x−1 Hx ⊆ H for all x ∈ G.

One nice thing about a normal subgroup H ≤ G is that the product of two cosets of H is
again a coset of H, since

(Hx)(Hy) = H(xH)y = H(Hx)y = (HH)(xy) = H(xy)

for all x, y ∈ G. (Note that HH ⊆ H since H is closed under the group operation;
conversely, HH ⊇ H since e ∈ H.) Multiplication of cosets (or of any subset of G) is
associative:    
(Hx)(Hy) (Hz) = (Hx) (Hy)(Hz) ,
simply because multiplication of elements of G is associative. Also H acts as an ‘identity’
among cosets of H, since

H(Hx) = (HH)x = Hx, (Hx)H = H(xH) = H(Hx) = (HH)x = Hx

for all x ∈ G. Furthermore for any coset Hx we have

(Hx)(Hx−1 ) = H(xx−1 ) = He = H, (Hx−1 )(Hx) = H(x−1 x) = He = H,

i.e. H(x−1 ) acts as an inverse for Hx among the cosets of H. So whenever H ≤ G, we

are led to define the quotient group (or factor group) of G modulo H, as the set of all
cosets of H in G, under the operation of subset multiplication:

G/H = {Hx : x ∈ G}.

We have already shown that this is in fact a group of order [G : H] with identity H and
inverses (Hx)−1 = H(x−1 ).
For example, consider G = S3 . We have already seen that the subgroup (12) is not
normal in G. However, consider H = (123) of order 3. We have a partition of G into
[G : H] = 63 = 2 left cosets of H, and likewise into 2 right cosets of H. The left cosets are
in fact
H = {(1), (123), (132)}, (12)H = {(12), (13), (23)},
while the right cosets are

H = {(1), (123), (132)}, H(12) = {(12), (13), (23)}.

Since these two partitions are the same, we have H ≤ G. As an example of multiplication
of two cosets of H, we have
  
H(12) H(12) = {(12), (13), (23)} · {(12), (13), (23)} = {(1), (123), (132)} = H.
 I. GROUPS 19

Thus G/H = S3 /(123) is a cyclic group of order 2, with Cayley table

H H(12)
H H H(12) .
H(12) H(12) H

The same argument shows that if G is any group having a subgroup H of index 2,
then H is normal in G. This is because if x ∈ G .........H then the left cosets form a partition
{H, xH} of G, while the right cosets form a partition {H, Hx} of G, so that the two
partitions must coincide.
As further examples, note that G ≤ G always, and G/G is the trivial group. Also the
trivial subgroup e is normal in G, and the left (and right) cosets of e are the singleton
subsets {x} for x ∈ G so that G/e ∼ = G, where x → x is an obvious isomorphism. We
say that a nontrivial group G is simple if its only normal subgroups are 1 and G itself.
Cyclic groups of prime order are simple (see Exercises 4.6, 5.3), and there are also many
nonabelian simple groups, of which the smallest is the alternating group A5 of order 60
(see Exercise 9.7).
If G is an abelian group, then every subgroup of G is clearly normal. More generally,
if every element of G commutes with every element of a subgroup H ≤ G (equivalently,
H ≤ Z(G)), then H ≤ G.
Exercises 5.
1. If H, K ≤ G, show that H ∩ K ≤ G. More generally, if S is any nonempty collection of normal
subgroups of G, show that
S = H ≤ G.
H∈S

Compare with Exercise 4.1.

2. If H ≤ G and K ≤ G, show that HK ≤ G. If H, K ≤ G, show that HK ≤ G.
3. Prove that an abelian group is simple iff it is cyclic of prime order.
4. If H ≤ K ≤ G, does it necessarily follow that H ≤ G? Explain.
Hint: Consider the group A4 .

6. Homomorphisms

We have already defined group isomorphisms. If we drop the requirement of bijectivity,

then we have simply a homomorphism of groups. If G and H are groups, then a map
φ : G → H is called a homomorphism if φ preserves the respective group operations,
i.e. φ(xy) = φ(x)φ(y). We remind the reader that the two binary operations arising
in the latter equation are interpreted quite differently: the operation between x and y
takes place inside G, whereas the operation between φ(x) and φ(y) takes place inside
H. To interpret the requirement of being a homomorphism in terms of group tables,
is trickier than the interpretation for isomorphisms. [See however Gorenstein, Scientific
20 I. GROUPS

American (Dec. 1985), pp.104–115 for such an interpretation.] A homomorphism is called

an epimorphism if it is surjective (that is, ‘onto’); or a monomorphism if it is injective
(that is, ‘one-to-one’); and an isomorphism if it is bijective (that is, ‘one-to-one’ and
‘onto’).
As examples, note that every isomorphism is a homomorphism. Moreover if G is
any group, then the unique map from G to 1 is a homomorphism, the so-called trivial
homomorphism. A more interesting example is the map θ : G → G, x → x2 which is
a homomorphism iff G is abelian. To see this, note that the equation xyxy = θ(xy) =
θ(x)θ(y) = x2 y 2 simplifies to xy = yx for all x, y ∈ G. Moreover θ need not be either
bijective or trivial in this case; for example if G = C4 = {e, x, x2 , x3 } then θ(e) = θ(x2 ) = e
and θ(x) = θ(x3 ) = x2 .
Suppose that φ : G → H is a homomorphism. Then φ(e) = φ(ee) = φ(e)φ(e), so

φ(e) = e.

Of course this means that φ(eG ) = eH where eG and eH are the respective identities of G
and H. Also
φ(xk ) = φ(x)k
for all x ∈ G and k ∈ Z; for example φ(x−1 )φ(x) = φ(x−1 x) = φ(e) = e implies that
φ(x−1 ) = φ(x)−1 , and by induction on k we verify the above formula more generally.
Define the kernel and image of a homomorphism φ : G → H by

ker φ = {g ∈ G : φ(g) = e}, φ(G) = {φ(g) : g ∈ G}.

It is easy to show that ker φ ≤ G and φ(G) ≤ H. More significant is the fact that kernels
of homomorphisms are always normal subgroups. (This is not true, however, for images.)
To see that ker φ ≤ G, note that

φ(x−1 gx) = φ(x)−1 φ(g)φ(x) = φ(x)−1 eφ(x) = e

if x ∈ G and g ∈ ker φ, so that x−1 (ker φ)x ⊆ ker φ for all x ∈ G. Next we prove a
converse of this fact: Every normal subgroup is the kernel of some homomorphism.
Given a group G and a normal subgroup K ≤ G, we have the canonical homomor-
phism
π : G → G/K, g → Kg.
This map π is onto by definition, and the fact that π is a homomorphism follows from
(Kx)(Ky) = Kxy. Since K is the identity of G/K, we have

ker π = {g ∈ G : π(g) = K} = {g ∈ G : Kg = K} = K.

Thus the given normal subgroup K is the kernel of a homomorphism (namely, the canonical
homomorphism π : G → G/K).
 I. GROUPS 21

Now let φ : G → H be any group homomorphism. Since ker φ ≤ G, we may form the
quotient group G/ ker φ. We show

6.1 First Isomorphism Theorem. G/ ker φ ∼

= φ(G).

Proof. Let K = ker φ ≤ G, and define φ : G/K → φ(G) by φ(Kg) = φ(g) for g ∈ G.
We first observe that φ is well-defined, for if Kx = Ky for some x, y ∈ G, then y = kx for
some k ∈ K and φ(Ky) = φ(y) = φ(kx) = φ(k)φ(x) = eφ(x) = φ(x) = φ(Kx).
Now if φ(Kx1 ) = φ(Kx2 ) for some x1 , x2 ∈ G, then φ(x1 ) = φ(x2 ), i.e. φ(x1 x−1
2 ) =
−1 −1
φ(x1 )φ(x2 ) = e so that x1 x2 ∈ ker φ = K and Kx1 = Kx2 . This proves that φ is
one-to-one.
Clearly φ is onto: every element of φ(G) is of the form φ(g) = Kg for some g ∈ G,
and this is the same element as φ(Kg). Thus φ is bijective.
The fact that φ is an isomorphism now follows from the fact that
 
φ (Kx)(Ky) = φ(Kxy) = φ(xy) = φ(x)φ(y) = φ(Kx)φ(Ky)
for all x, y ∈ G.

Exercises 6.
1. Let φ : G → H be a group homomorphism. Show that φ is one-to-one iff ker φ = 1.
2. Let φ : G → H be a group homomorphism. Prove that for every g ∈ G of finite order, the image φ(g)
has finite order dividing |g|.
3. Let G be a group. Define φ : G → G, x → x2 and define ψ : G → G, x → x−1 . Prove that G is
abelian iff φ is a homomorphism iff ψ is a homomorphism.
4. Let G = GL2 (C), the multiplicative group of invertible 2 × 2 matrices with comples entries. Let
Z = Z(G). Recall (Exercise 4.10) that Z = {αI : 0 = α ∈ C} where I ∈ G is the identity matrix.
Since Z ≤ G, we may define the projective general linear group
P GL2 (C) = G/Z.
Let C∗ = C ∪ {∞} be the one-point compactification of the complex plane. Let H be the set of all
fractional linear transformations on C∗ , i.e. mappings of the form
az + b
C∗ → C∗ , z →
cz + d
where ad−bc = 0. (Here az+b
cz+d
means the same as lim aw+b
, in order to make these transformations
w→z cw+d
defined, and in fact continuous, on C∗ .) Prove that H is a group under composition, and that
H∼ = P GL2 (C).
Hint: Use the First Isomorphism Theorem.
∗ 5. Let G be a finite group, and suppose that φ : G → G is a homomorphism such that
(i) φ(φ(x)) = x for all x ∈ G; and
(ii) φ(x) = x only for x = e.
22 I. GROUPS

Show that G is abelian. Can you find a counterexample if G is infinite?

6. Let K ≤ G. If H is a subgroup of G containing K, then clearly K ≤ H and we have H/K ≤ G/K.
(a) Show that H → H/K defines a bijection between S1 (the collection of all subgroups of G which
contain K) and S2 (the collection of all subgroups of G/K).
(b) For H ∈ S1 , show that
H≤ G ⇐⇒ H/K ≤ G/K.

7. Let H be a proper normal subgroup of G. We say that H is maximal normal in G if H is maximal

among all proper normal subgroups of G, i.e. if H ≤ K < G implies H = K. Show that H is maximal
normal in G iff G/H is simple.
Hint: Use Exercise 6.6.
8. Let H ≤ G, K ≤ G. Recall that HK ≤ G; see Exercise 5.2.
(a) Show that H ∩ K ≤ H and K ≤ HK. Thus we may construct the quotient groups H/(H ∩ K)
and HK/K.
(b) Prove the Second Isomorphism Theorem for groups:
H/(H ∩ K) ∼
= HK/K.
Hint: Consider the map φ : H → HK/K, h → hK and apply the First Isomorphism Theorem.
9. Suppose that K ≤ H ≤ G and K ≤ G. By Exercise 6.6, we have H/K ≤ G/K.
(a) Show that the map φ : G/K → G/H, Kg → Hg is well-defined, and an epimorphism.
(b) Prove the Third Isomorphism Theorem for groups:
(G/K) ∼
(H/K) = G/H
by applying the First Isomorphism Theorem.
10. Let K ≤ G be groups such that K is contained in the center of G. If G/K is cyclic, show that G is
abelian.
11. (a) Give an example of an infinite group G and a homomorphism φ : G → G which is one-to-one but
not onto.
(b) Give an example of an infinite group G and a homomorphism φ : G → G which is onto but not
one-to-one.
12. Let G be a group. Define the commutator of two elements x, y ∈ G by
[x, y] = x−1 y −1 xy.
(a) Show that [x, y] = e iff x and y commute.
(b) Define the derived subgroup G
≤ G to be the subgroup generated by all commutators in G:
G
=
[x, y] : x, y ∈ G.
Show that G
≤ G and that G/G
is abelian.
(c) Let K ≤ G. Show that K ⊇ G
iff G/K is abelian. Thus G
is the smallest normal subgroup of
G whose quotient group is abelian. (It makes sense to speak of the smallest such normal subgroup,
because this is the same as the intersection of all normal subgroups K ≤ G such that G/K is
abelian, as in Exercise 4.1.)
13. Let G be a group. Define the commutator of two subgroups H, K ≤ G to be the subgroup [H, K] ≤ G
generated by all commutators [h, k] such that h ∈ H, k ∈ K. This generalizes the derived subgroup
G
= [G, G] considered in Exercise 6.12. Now suppose that H, K ≤ G satisfy H ∩ K = 1.
(a) Show that every element of H commutes with every element of K.
Hint: Show that [H, K] ⊆ H ∩ K.
 I. GROUPS 23

(b) Recall that HK ≤ G; see Exercise 5.2. Show that HK ∼

= H × K.
Remarks: A subgroup of the form HK is known as an internal direct product of the subgroups H
and K.
14. Let n ≥ 1 and let p be prime. Consider the group G = GLn (Fp ) of invertible n × n matrices over
the field Fp of order p; see Exercise 3.7. Recall that the determinant map det : G → F× p is a
homomorphism, where F× p = {1, 2, . . . , p−1} is the multiplicative group of nonzero scalars.
(a) Show that det is surjective. Using the First Isomorphism Theorem, determine the order of the
special linear group SLn (Fp ) = {A ∈ G : det(A) = 1}.
(b) Show that GL3 (F5 ) ∼
= SL3 (F5 ) × C4 .
Hint: Use Exercise 6.13.
15. Let G be a finite abelian group and let p be a prime divisor of |G|. Show that G has an element of
order p. (This is a special case of Cauchy’s Theorem, Exercise 8.7.)
Hint: We may assume that G has a nontrivial proper subgroup H (why?) and by induction on |G|, we
may assume that p | |H|. Show that G/H has an element of order p by induction, and use Exercise 6.2
to obtain an element of order p in G.
16. Let G be a finite abelian group, and let g ∈ G of order equal to the exponent m of G. (See Exercise 4.11
where it was shown that such an element g ∈ G necessarily exists.)
(a) If G is not cyclic, prove that there exists an element k ∈ G of prime order such that k ∈
/
g.
Hint: The quotient group G/
g has an element k
g of prime order p; thus there exists k ∈ G
such that kp ∈
g but k ∈ /
g. Since (kp )m/p = hm = e and the subgroup
g has exactly m/p
solutions of xp = e, namely the elements of
g p  (see Exercise 4.5) we have kp ∈
g p , i.e. kp = g jp
for some integer j. Verify that k
= kg −j has the required properties.
(b) Prove that there exists a subgroup H ≤ G complementary to
g, i.e. G =
gH with
g ∩ H = 1,
and thus G ∼=
g × H by Exercise 6.13.
Hint: If G is not cyclic, there exists a nontrivial subgroup K ≤ G such that K ∩
g = 1 (why?).
Denote the quotient group G = G/K and a typical element x = xK ∈ G. By induction G =
gH
with
g ∩ H = 1. Let H ≤ G be the preimage of H in G, as in Exercise 6.6.
17. Prove the Fundamental Theorem of Finite Abelian Groups: Every finite abelian group G is isomorphic
to a direct product of cyclic groups.
Hint: Use Exercise 6.16.

7. Automorphisms

For any group G, we may consider the possibilities for an isomorphism from G to itself.
Such a map is called an automorphism of G.
We have already seen that if G is abelian, then the map

θ : G → G, x → x2

is a homomorphism. In order for θ to be an automorphism of G, θ must be bijective.

Assuming that G is finite, θ is bijective iff θ is one-to-one iff θ is onto. And this is
equivalent to ker θ = 1 (see Exercise 6.1). All this says that for a finite group G, the map
x → x2 is an automorphism iff G is abelian with no elements of order two.
24 I. GROUPS

Given a group G, let Aut G denote the set of all automorphisms of G. Clearly Aut G
contains the identity map G → G, x → x. Now Aut G is a group under composition,
called the automorphism group of G, or the full automorphism group of G (to
distinguish it from other groups of automorphisms of G, i.e. subgroups of Aut G).
A special set of automorphisms of G is the set of inner automorphisms Inn G =
{ψg : g ∈ G}. These automorphisms are defined by

ψg : G → G, x → xψg = g −1 xg.

(We will write xψg instead of ψg (x), since ψg simply permutes the elements of G, and
we wish to be consistent with our usual notation in placing permutations as superscripts.)
If g −1 xg = g −1 yg, then cancelation yields x = y, so that ψg is one-to-one. Also, given
y ∈ G, we have (gyg −1)ψg = y, which shows that ψg is onto. For x, y ∈ G, we have

xψg y ψg = (g −1 xg)(g −1yg) = g −1 xyg = (xy)ψg ,

so that ψg is a homomorphism, and so ψg ∈ Aut G. Now

 ψg ψh
x = h−1 (g −1 xg)h = (gh)−1 x(gh) = xψgh

for all x ∈ G, and so ψg ψh = ψgh for all g, h ∈ G. This means that Inn G is closed under
composition. It is also easy to chow that (ψg )−1 = ψg −1 , so in particular Inn G is ‘closed
under inverses’. Together this shows that Inn G ≤ Aut G. (Of course, if G is abelian,
then ψg is the identity map G → G, x → x. In this case Inn G is the trivial subgroup of
Aut G.)
More is true: we in fact have Inn G ≤ Aut G. To prove this, let g ∈ G, σ ∈ Aut G.
Then
−1  −1  ψg  σ  −1 σ  σ  −1 σ σ  σ −1 σ
xσ ψg σ
= xσ = g −1 xσ g = g −1 xσ g = g xg

for all x ∈ G, i.e. σ −1 ψg σ = ψg σ ∈ Inn G. This shows that Inn G ≤ Aut G as claimed.
Exercises 7.
1. Let G be a cyclic group of order n. Show that Aut G is abelian of order ϕ(n), using the notation of
Euler’s function ϕ(n) = |{k ∈ Z : 1 ≤ k ≤ n, gcd(k, n) = 1}|.
Hint: Let G = {e, x, x2 , . . . , xn−1 }. Show that every automorphism of G has the form τk : G → G,
xi → xik for some integer k relatively prime to n.
2. Determine the automorphism group of the Klein 4-group.
3. Let G = Cp × Cp × · · · × Cp (n times). Determine Aut G. (This is a group you have encountered
previously in this course.)
Remark: Compare with Exercises 3.7(b) and 7.2.
4. Let G = SLn (F ) and consider the inverse-transpose map φ : g → g −T .
(a) Prove that φ is an automorphism of G.
 I. GROUPS 25

(b) Prove that the automorphism φ is inner, iff n ≤ 2.

5. A subgroup H ≤ G is characteristic (denoted H char G) if φ(H) = H for every φ ∈ Aut(G). Prove

that
(a) Every characteristic subgroup of G is normal.
(b) The derived subgroup of G is characteristic (see Exercise 6.12).

6. A nontrivial group G is characteristically simple if its only characteristic subgroups are 1 and G itself.
Prove that G is characteristically simple iff G is a direct product of isomorphic simple groups, i.e.
G∼= K n = K × K × · · · × K (n times) for some simple group K and some positive integer n.
7. Let H ≤ G be a minimal normal subgroup, i.e. H is a nontrivial normal subgroup for which the only
normal subgroups of G contained in H, are 1 and H itself. Prove that every minimal normal subgroup
of G is characteristic.

8. Permutation Groups and Actions

If S is any set, the symmetric group on S is by definition the set of all permutations
of S (i.e. bijections S → S), denoted Sym S. This forms a group under composition. We
will generally assume that |S| = n < ∞, in which case Sym S is clearly isomorphic to Sn .
A permutation group on S is by definition a subgroup G ≤ Sym S. In this case we
say that G is a permutation group of degree n. The stabilizer of an element a ∈ S is by
definition
Ga = {g ∈ G : ag = a}.

It is not hard to verify that Ga is in fact a subgroup of G. The orbit of a ∈ S is by

definition the set of all possible images of a under G:

aG = {ag : g ∈ G}.

If S consists of a single orbit under G, we say that G is transitive on S, or that G

permutes S transitively; otherwise G is intransitive.
As a first example, consider the group G = (1234), (24), a dihedral group of order 8.
This is a permutation group of degree 4 which may be considered as the symmetry group
of the square
3.......................................................2
.. ...
.. ...
.. ...
.. ...
... ...
.. ...
... ...
... ..
..
...............................................
4 1
In this case the stabilizer of the vertex 1 is G1 = (24) = {(1), (24)}, and the orbit of the
vertex 1 is 1G = {1, 2, 3, 4}. We observe that |G1 ||1G | = 2 · 4 = 8 = |G|, as evidence of a
property that holds more generally; see Proposition 8.1 below. The subgroup H = (24)
has three orbits on vertices:

1H = {1}; 2H = {2, 4} = 4H ; 3H = {3}.

26 I. GROUPS

Observe that the orbits of H partition the vertex set {1, 2, 3, 4} into three orbits, each of
size dividing |H| = 2. We have |H1 ||1H | = |H||{1}| = 2 · 1 = 2 = |H|; also |H2 ||2H | =
|(1)||{2, 4}| = 1 · 2 = 2 = |H|. Note that G permutes {1, 2, 3, 4} transitively; the subgroup
H is intransitive.
In our example above, the vertices played a special role. Our original abstract def-
inition of a dihedral group, in Section 2, gave no special heed to vertices of the n-gon.
The dihedral group just as well permutes edges of the n-gon, or diagonals, or points of the
entire Euclidean plane. In our next example we consider some of these other actions of
the dihedral group.
A permutation action (or permutation representation) of a group G on a set
S is a homomorphism φ : G → Sym S. This associates, to each g ∈ G, a permutation
φ(g) ∈ Sym S. If φ is one-to-one, we say that the action is faithful; in this case φ(G) ∼ = G,
so we may identify G with the permutation group φ(G) ≤ Sym S. However, if φ is not
one-to-one, then distinct elements g, h ∈ G may give rise to the same permutation φ(g) =
φ(h) ∈ Sym S; in this case we say φ is unfaithful.
As an example, we again consider the dihedral group G = D4 of order 8, but this time
using our original notation of Section 2, thus: G = {I, R, R2, R3 , T, RT, R2T, R3 T }. We
consider five actions of this group:
µ, the action of G on the vertex set V = {1, 2, 3, 4};
σ, the action of G on the set of edges E = {a, b, c, d};
δ, the action of G on the set of diagonals D = {D, D
};
τ , the action of G on the set of coordinate axes A = {X, Y }; and
γ, the action of G on the set consisting of the center O = {O}.
One verifies that these five actions are as listed in the following table, in which () denotes
the identity permutation:
.. ....
...............................
g µ(g) σ(g) δ(g) τ (g) γ(g) .
R
.........
........
.....
..
2•
. .............
.. . ......
......
......
..... ......
.
..... .
..
.. .. ...... ......
.
I (1) () () () () .. .
.....
. . ..... ............

...
.....
. ......
D ....
...
D ..... ......... Y .....
..... . ...
R (1234) (abcd) (D D
) (X Y ) () .. .....
..
.. ..
.....
.
.
.
.
.
..... ....
..
..
.. ...
.....
...
...
...
.. .
. .. . .. ..... ...
2 .. ..... . .. ..... ...
R (13)(24) (ac)(bd) () () () b . .
.......... ..
.
.
.
.
....
.. a......
. ..... ..
...
.. .
. ..... .
..... .. ... ......
. ... ....
T
.
3 .. . .. ..... ........
(D D
) ..
.. ....... . .
..
. . ..
R (1432) (adcb) (X Y ) () ... ..... .
3• ◦ X •
.. .
.......... ....... ....... ....... ....... .............. ....... ....... ....... ....... ......... ........
. ......... .....
..... .
. .... ....
..... ... .........
T (24) (ad)(bc) (D D
) () () .....
..... .. O ....
....... 1 ....
.
..
.. ....
..... .....
..... .....
. ..... .....
c ..... ..
d .....
..... ..... .... ..... ........
RT (14)(23) (ac) () (X Y ) () ......
. .....
.......
.
..... ...
..... ........ .... ....... ....
2 . .
(D D
) ..
. ...... .. .. .....
. .
R T (13) (ab)(cd) () () ..
.. .
.....
..... .... .........
.. ..
.....
..... .....
..... . .........
. ..
3 . ............ .....
4•
.. .
R T (12)(34) (bd) () (X Y ) () .. .. ..

We readily verify that each of these actions is a homomorphism. It is natural to require

this property; for example the fact that µ(gh) = µ(g)µ(h) says that the action of gh on
the vertex set V, is the same as first applying g to the vertices, then applying h to the
vertices.
 I. GROUPS 27

Note that the actions µ and σ are faithful: distinct elements of G give distinct permu-
tations of the vertices, and of the edges. The other actions listed are unfaithful: distinct
elements of G may give the same permutations on D, A and O. It is possible to identify
G with its image µ(G) ≤ Sym V; certainly this permutation group is isomorphic to G, via
the isomorphism µ. There is nothing special however about vertices; one may equally well
identify G with the permutation group σ(G) ≤ Sym E. However, the groups δ(G), τ (G),
γ(G) have orders 2, 2, 1 respectively; these groups cannot be identified with G. Indeed, the
action γ is trivial: every element of G fixes the center.
We speak of orbits and stabilizers not only for permutation groups, but more generally
for permutation actions. In the example above, G acts transitively on D, giving an orbit
DG = {D, D
}. The stabilizer of D is the subgroup GD = {I, R2 , RT, R3 T } and we check
that |GD ||DG | = 4 · 2 = 8 = |G|.
In our previous example, µ, σ, δ, τ, γ were five actions of the group G on distinct sets
V, E, D, A, O. Next we consider a situation where a single group G can act on one set in
more than one way.
Let G = C4 = {e, x, x2 , x3 }, a cyclic group of order 4 with x4 = e. We can let G act
on a cube as a group of rotational symmetries. One natural way to do this would be to let
x act as a 90◦ counter-clockwise rotation about a vertical axis, as shown. We have labeled
the vertices of the cube as V = {1, 2, . . . , 8}.
...
..
..
..
..
φ(x) = (1234)(5876) ∈ Sym V
............
......
. ... .
.. ... ....
. . .. . . .....
....
..
.................
..•4
...........................
. .. ..
...
...
...
................. .... .... ........
...
1• .............. ... .....
.................... .
..
. .....
... ...... ... ....
... ......
... ...... ◦ .
...
.
.....
....
...
...
....
.....
....
.
.
.
. ..........................
.
.
.
. ..
... . .
... .. .. .
.
....
................. •3
... ..... ....
... ..... ................................. ... ..
...
.....
2• .........
.
.
....
.
.
.
.
.
..
..
...
..
.. .. . .
. ..
... ... . .
. ...
.. .. . .
. ..
... .. . .
. ..
. . . .
. ...
.... .... .. ..
..
...
..
... .
. ... .......
. ...... ..... ..........
. ...
... •5 .....
..
..
...
.. . .. .
.... . . ..
... ...... ..... .............. . ..... ..
• .........
.....
...
.
.
.
..
....
. ..... ...
...
8 .....
....
.....
.
.
..
. ◦ .
.
. .. ..
.......
....
..... ...
.. .
.
. ................................
................
.... •
.... ..
..... ..
....... ............................. . .
...
... 6
• ....... ...
..
7 ..
...
... ............
......
. .... ..
.. ... ....
. . ... . . .....
..
.

Clearly G permutes the vertices intransitively in this case; there are two orbits, {1, 2, 3, 4}
and {5, 6, 7, 8}, each of length 4. Each vertex has trivial stabilizer in G, and we check that
|1G ||1G | = 1 · 4 = 4 = |G|. However, our choice of action of G on the cube was rather
arbitrary; we could instead let x act as a 90◦ rotation about a different axis, giving rise to
28 I. GROUPS

the action η : G → Sym V indicated:

................
...•4
................ ...........
...
...
..................
. .. .......
... . .. . .....
1• ...........
.................... ... .....
....
... ..... . .
... ......
... ......
..
.
.....
.....
.... η(x) = (1872)(3456) ∈ Sym V
.....
............ .......
...
...
.....
.....
.... . .
............
.
.
.
. ....... ..
... ... •3
...
................... ......
.
.
.
.. ..... ........ . .
... .... ............................... . ..
. .
2• ............ .. . ........
..
... ... ...
...
. ...
...
. .. . . ...
................
.....
.
..
...
............ ....... ..
.
...
.
.
.
. .... ....
.
.
. . .. .. ◦
.... .
. .. . . . .
.
.
.
.
.
.
.
.... . ... .... ...... ... .
. ... . .... ...... .. ... ... . .
. ... ........... . ....... .. ..
............... . . ..◦.
. . .
.
.
.
.
.
.
..
..
...

................
..
..
......... . .....
.
. ...
.
....
.. .... ...... ..
•5
....
.... ...... .. .....
.
....
..
. ... .... ...... .. .....
. ... ... ...... ................ .. ..... ...
.
...
. • ........ ...
.....
.
.
..
.
..
....
.
..
..
..
8 ....
.....
....
.
.
.
.
...
... ...
........
.....
..... .
....
................
.............• .
.... ..
..... .. ...........
. ................... 6
....... ......................
• .............
7

As an example of an unfaithful representation, consider the action ψ : G → Sym V in

which ψ(x) is a reflection in a horizontal plane through the cube’s center; in this case

ψ(x) = (18)(27)(36)(45).

In such cases where more than one action of G on a given set S is considered, it is ambiguous
to write ag for a ∈ S and g ∈ G; we must explicitly denote the action as, for example, aφ(g) .
In our example of C4 acting on V, we have 1φ(x) = 2, whereas 1η(x) = 1ψ(x) = 8. The orbit
depends on the choice of action; in our example, 1φ(G) = {1, 2, 3, 4}, 1η(G) = {1, 2, 7, 8},
1ψ(x) = {1, 8}. Moreover the stabilizer of a point a ∈ S may in general depend on the
choice of action. While these examples may seem somewhat contrived, there are important
examples in which a group G can act in two or more ways on the same set. In particular,
we will see how every group G can act on itself by right-multiplication or by conjugation.
(Yet another action, by left-multiplication, can be considered.)
Finally, we prove

8.1 Proposition. |aG | = [G : Ga ] = |G|/|Ga |.

Proof. Write Ga \G = {Ga g : g ∈ G}, which is just the set of right cosets of Ga in G.
(Warning: This is not a group unless Ga is normal in G.) Of course |Ga \G| = [G : Ga ] =
|G|/|Ga |, so it is sufficient to find a bijection between Ga \G and aG . So we will show that,
given g, h ∈ G,
Ga g = Ga h ⇐⇒ ag = ah .

If Ga g = Ga h, then g = xh for some x ∈ Ga , i.e. ax = a, in which case ag = axh = (ax )h =

−1
ah . Conversely, if ag = ah , then agh = a, so gh−1 ∈ Ga , which implies that Ga g = Ga h.
It follows that Ga g → ag is a bijection Ga \G → aG .
 I. GROUPS 29

We next show the remarkable fact that every finite group may be faithfully represented
as a permutation group!

8.2 Cayley Representation Theorem. If G is a finite group, then G is isomorphic

to a permutation group (i.e. a subgroup of Sn for some n).

Proof. We let G act on itself by right-multiplication, as follows. For g ∈ G, we define

ρg : G → G, x → xg.

Clearly this is a bijection on G, so we have ρg ∈ Sym G. If g, h ∈ G, then

(xρg )ρh = (xg)h = x(gh) = xρgh

for all x ∈ G, i.e. ρg ρh = ρgh , and so we have a homomorphism

ρ : G → Sym G, g → ρg .

Thus ρ is a representation (called the right regular representation) of G. If ρg fixes

every element of G, then e = eρg = eg = g, so ρ is one-to-one (i.e. the representation is
faithful), and so ρ gives an isomorphism from G to a subgroup of Sym G. If |G| = n, this
is just as good as an isomorphism from G to a subgroup of Sn .

One nice thing about this result is that permutations are somewhat familiar objects
which multiply in a predictable way. They are also quite recognizable to a computer. If
we have a computer program designed to accept a permutation group as input data, then
such a program can be used to handle any finite group G, if G is first represented as a
permutation group. However, in order for this permutation representation to be useful in
a program, the degree of the representation should not be too large. In other words, we
want to write G as a group of permutations of {1, 2, . . . , n} where n is not too large. The
problem with our proof of the Cayley Representation Theorem is that it realizes G as a
subgroup of Sn for a typically large value of n, namely n = |G|. It is often possible to
improve on this, as we shall see later. But just as an example of how significant this point
really is, note that Cayley’s Theorem manages to express S5 as a set of 120 permutations
of {1, 2, 3, . . . , 120}; clearly the natural representation of S5 as a set of 120 permutations
of {1, 2, 3, 4, 5} is much more concise!
There is a way to generalize the basic idea in the proof above, as follows. Let H be any
subgroup of G. Then G permutes the right cosets of H by right multiplication. Writing
the set of right cosets of H as H\G = {Hx : x ∈ G} (again, this is merely a set, not a
group in general), then |H\G| = [G : H] = |G|/|H|. For g ∈ G we define

ρg : (H\G) → (H\G), Hx → Hxg.

30 I. GROUPS

Clearly this gives a permutation of the right cosets of H, i.e. ρg ∈ Sym(H\G). (If we had
the trivial subgroup H = {e}, then this would be the same action as in the proof of the
Cayley Representation Theorem.) Moreover G permutes these cosets transitively! And
we have a permutation representation of G of degree [G : H]. If we can choose a large
subgroup H < G, and thereby make [G : H] correspondingly small, then we will have
succeeded in finding a smaller degree permutation representation of G than that realized
by our proof of Theorem 8.2 above. This is good news. The bad news, however, is that
this action might not be faithful. Shortly, however, we will find an explicit expression for
ker ρ.

Exercises 8.
1. Let H be the subgroup of S9 generated by (123)(789) and (345). How many orbits does H have on
{1, 2, 3, . . . , 9}, and what are their sizes?

2. Suppose that H ≤ G ≤ Sn where H acts transitively on {1, 2, 3, . . . , n}. Show that G = G1 H where
G1 = {g ∈ G : 1g = 1}.

3. (a) Let G be the rotational symmetry group of a cube. What is the order of G? Is G abelian? What
else can you say about G? Can you identify G up to isomorphism, as a group we have studied?
(b) Do the same for a regular octahedron in place of the cube. A regular octahedron has eight triangular
faces, all equilateral of the same size.

4. Suppose that G acts on S. Let H be any normal subgroup of G, and let SH = {a ∈ S : ah = a for all
h ∈ H}, the set of fixed points of H. Prove that G preserves SH (i.e. ag ∈ SH whenever a ∈ SH ) and
so G induces a permutation group on the subset SH ⊆ S.
This latter exercise is so important that we must give an example: Label the vertices of a regular
octahedron as 1,2,3,4,5,6 in such a way that vertices 1,2,3,4 form a square. The group of all rotational
symmetries of the regular octahedron (Exercise 8.3) has a subgroup G ∼ = D4 preserving a square
formed by the vertices 1,2,3,4. The cyclic subgroup H ≤ G of order 4 cycles 1,2,3,4 and fixes the
remaining two vertices 5,6. Therefore G must preserve SH = {5, 6}. You can see directly that in fact
G does permute the pair {5, 6} of antipodal vertices.

5. Let p be prime, and recall the construction of the group GLn (Fp ) in Exercise 3.7.
(a) Show that for every n ≥ 1, G has a subgroup isomorphic to Sn consisting of permutation matrices
(matrices in which every row and column has a single nonzero entry equal to 1).
(b) Show that every finite group is isomorphic to a subgroup of GLn (Fp ) for some n ≥ 1.
Hint: Apply Theorem 8.2.

6. Consider the puzzle depicted in the illustrations below, in which nineteen circular disks (numbered
1,2,. . . ,19) are free to slide around the two loops of a track shaped in a figure ‘8’. After sliding disks
around one of the loops, one disk must come to rest at the center position (where the two loops meet)
before sliding disks around the other loop. The first illustration shows the original configuration,
according to the manufacturer’s instructions. The second illustration shows the positions of the disks
after the puzzle has been played with extensively. I have been trying, with no success, to restore the
puzzle to its original configuration, and I am beginning to suspect that my child has dropped the puzzle
on the floor, causing some disks to come loose from their tracks, then pushed the loose disks back into
their tracks in a configuration which cannot be solved (i.e. restored to the original configuration using
only legal moves).
 I. GROUPS 31

6 7 18
17
16 5 2 7 19 9
8 19 18 6
5 15 17 1
9 12
4 10 14 11 10
1 4

8
3 2 11 12 13 15 14 16 13
3
Puzzle: Original Position Puzzle: Altered Position

(a) Do you think it is possible to restore the puzzle from the altered position shown, to its original
position? Explain.
(b) If some disks pop out and are pushed back into their tracks in a random configuration, what do
you think is the probability that the resulting configuration is solvable (i.e. can be restored to the
original configuration using only legal moves)? Explain.
Hint: Recall Exercise 2.4.

7. Let G be a finite group of order divisible by a prime p. Prove that G has an element of order p. (This
is Cauchy’s Theorem; compare Exercise 6.15 where we settled the case G is abelian.)
Hint: Let S be the set of all p-tuples (g1 , g2 , . . . , gp ) of elements of G satisfying g1 g2 · · · gp = 1, and
define σ : (g1 , g2 , . . . , gp ) → (g2 , g3 , . . . , gp , g1 ). Show that S is invariant under σ and that σ permutes
S in orbits of size 1 and p. Determine |S| and observe that σ ∈ Sym S fixes (1, 1, . . . , 1) ∈ S.

9. Conjugation

Let G be any finite group. Given g, h ∈ G, we say that g is conjugate to h in G (written

g ∼ h) if g = x−1 hx for some x ∈ G. Note that g = e−1 ge, so that every group element
is conjugate to itself. Also if g ∼ h, say g = x−1 hx, then h = (x−1 )−1 g(x−1 ), so h ∼ g.
Furthermore if g ∼ h ∼ k, say g = x−1 hx and h = y −1 ky, then g = x−1 y −1 kyx =
(xy)−1 k(xy), so that g ∼ k. This shows that conjugacy is an equivalence relation on the
elements of G. The equivalence classes for this relation are called the conjugacy classes
of G. In particular, the conjugacy classes partition the elements of G.
For example in S3 , we have (23) = (123)−1 (12)(123) and so (23) ∼ (12). Indeed the
conjugacy classes of S3 are

{(1)}, {(12), (13), (23)}, {(123), (132)}.

Note that the size of each of these conjugacy classes divides |S3 | = 6. This is no accident,
as we shall soon see.
Recall that the inner automorphism ψx ∈ Inn G is defined by ψx : g → g ψx = x−1 gx.
Recall also that ψx ψy = ψxy for all x, y ∈ G. Thus we have a homomorphism

ψ : G → Inn G, x → ψx ,
32 I. GROUPS

which is in fact a permutation representation of G, called the conjugation action of G

on itself. Clearly, the orbits under this action are precisely the conjugacy classes of G.
The stabilizer of an element g ∈ G is just the centralizer of g in G:

{x ∈ G : g ψx = g} = {x ∈ G : x−1 gx = g} = {x ∈ G : gx = xg} = CG (g);

see Exercise 4.9. So the conjugacy class of an element g ∈ G is the orbit g G = g ψ(G) , which
has length [G : CG (g)] = |G|/|CG (g)|. In particular, the size of each conjugacy class divides
 
|G|. Returning to our S3 example, we have CS3 (123) = (123) = {(1), (123), (132)} of
order 3, and the corresponding conjugacy class {(123), (132)} has size [S3 : CS3 ((123))] =
6
3 = 2.
So far we have used two actions of G on the elements of G: the right regular repre-
sentation ρ, and the conjugation action ψ. If we were to subsequently write xg , you might
well wonder whether we meant xρg = xg or xψg = g −1 xg. In fact we will typically mean
the latter:
xg = g −1 xg.

Just as we talk about conjugate elements of G, we can talk about conjugate subgroups.
If H is any subgroup of G, and x ∈ G, then

H x = x−1 Hx = {x−1 hx : h ∈ H}

is also a subgroup of G (since it is the image of H under ψx , and the image of any group
homomorphism is a subgroup.) Again, G acts on the set of subgroups of G, by conjugation.
The set of conjugates of H will be one orbit under this action. What is the length of this
orbit? That is, how many distinct conjugates does H have inside G? The stabilizer of H
in this action is otherwise known as the normalizer of H in G, written

NG (H) = {x ∈ G : H x = H} = {x ∈ G : x−1 Hx = H} = {x ∈ G : Hx = xH}.

Observe that H ≤ NG (H) ≤ G, and that NG (H) = G iff H ≤ G. Now the number of

subgroups of G which are conjugate to H is [G : NG (H)] = |G|/|NG (H)|, and in particular
this number divides |G|.
For example consider (123) < S4 . Then S4 has exactly four subgroups conjugate to
(123), namely
(123), (124), (134), (234).
 
This agrees with our formula, since NS4 (123) = (123), (12) = S3 of order 6, which
has index 4 in S4 .
Next, we show how conjugate subgroups arise naturally in the study of permutation
groups: points in the same orbit have conjugate stabilizers. For suppose that G acts
transitively on a set S, i.e. φ : G → Sym S is a transitive permutation action. We have
 I. GROUPS 33

seen that |S| = |aG | = [G : Ga ] for a ∈ S. If b ∈ S also, then by transitivity we have

b = ag for some g ∈ G. We claim that Gb = (Ga )g = g −1 Ga g . Firstly, if x ∈ Ga , then
−1 −1
bg xg
= ((bg )x )g = (ax )g = ag = b,

which shows that g −1 xg ∈ Gb . The reverse inclusion Gb ⊆ g −1 Ga g is equivalent to

−1
gGb g −1 ⊆ Ga , which follows by the same reasoning since a = bg . Thus Gb = g −1 Ga g
as claimed. Our superscript notation makes everything easy to remember: the stabilizer
of a is Ga , so the stabilizer of ag is (Ga )g .
Exercises 9.
1. Let H ≤ G, and let x, y ∈ H. If x and y are conjugate in H, must they be conjugate in G? If
they are conjugate in G, must they be conjugate in H? In each case, prove the statement or supply a
counterexample.
2. (a) List the conjugacy classes of S4 .
(b) Show that a subgroup H ≤ G is normal iff H is a union of conjugacy classes of G (i.e. H = S
where S is a collection of conjugacy classes of G).
(c) Using (b), determine all normal subgroups of S4 .
3. Consider the permutations σ = (1942)(35)(68) and τ = (137)(28) in S9 .
(a) Determine σ −1 τ σ and (1σ 3σ 7σ )(2σ 8σ ). What do you observe?
(b) Generalize your observation to show that two permutations in Sn are conjugate iff they have the
same cycle structure, i.e. the same number of disjoint cycles of each length.
4. Let G be a group of order 2n where n ≥ 1 is odd. Prove that G has a subgroup of order n. (Do not
use the Sylow Theorems!)
Hint: Let ρ : G → Sym G be the right regular representation of G. By Exercise 1.7, there exists τ ∈ G
of order 2. Show that the permutation ρτ is odd, by considering its cycle structure; then consider the
kernel of the homomorphism G → {±1}, g → sgn(ρg ).
5. Let G be any group, and let a ∈ G. Show that a ∈ Z(G) iff {a} is a conjugacy class of G. Conclude
that Z(G) is the union of all singleton conjugacy classes of G.
6. Two elements A, B ∈ GLn (C) are conjugate iff they are similar as matrices.
(a) List all properties that you can think of, shared by two similar matrices. (For example, similar
matrices have the same trace.)
(b) How does one check whether or not two given matrices are similar? Give necessary and sufficient
conditions similar to the criterion for conjugacy of permutations, given in Exercise 9.3.
7. Complete the following steps (a)–(f) to show that An is a simple group for n ≥ 5. Assume that n ≥ 5
and suppose 1 = K ≤ An . We must show that K = An .
(a) Show that An is generated by its 3-cycles.
Hint: First express an arbitrary product of two 2-cycles as a product of at most two 3-cycles, not
necessarily disjoint.
(b) Show that if K contains a 3-cycle, then K = An .
Hint: Recall Exercise 9.3.
(c) If σ = (i1 i2 i3 · · · ik ) × (cycles disjoint from i1 , i2 , . . . , ik ) ∈ K where k ≥ 4, show that K contains
a 3-cycle.
Hint: Compute the commutator (i1 i2 i3 )−1 σ −1 (i1 i2 i3 )σ.
34 I. GROUPS

(d) If σ = (i1 i2 )(i3 i4 ) × (cycles disjoint from i1 , i2 , i3 , i4 ) ∈ K, show that K contains a 3-cycle.
Hint: Similar to (c).
(e) If σ = (i1 i2 i3 )(i4 i5 i6 ) × (cycles disjoint from i1 , i2 , i3 , . . . , i6 ) ∈ K, show that K contains a 3-cycle.
(f) Complete your proof using the steps above. What goes wrong with your proof if n ≤ 4?
8. Consider the dihedral group Dn =
R, T  of order 2n, using the notation of Section 2. Find all
conjugacy classes of Dn .
9. Let G be a group. Define the left and right regular representations of G by λ, ρ : G → Sym G
where
λa : G → G, x → a−1 x;
ρa : G → G, x → xa.
(a) Prove that λa λb = λab and ρa ρb = ρab with left-to-right composition as usual. Thus λ and ρ are
permutation representations of G. Why did we require the a−1 in the definition of λa ?
(b) Show that G ∼=
λa : a ∈ G ∼
=
ρa : a ∈ G ≤ Sym G.
Hint: Use the First Isomorphism Theorem.
(c) Let M =
λa , ρa : a ∈ G. Show that M ∼ = (G × G)/Z where Z = {(z, z) : z ∈ Z(G)}.
Hint: Use the First Isomorphism Theorem.

10. More on Transitive Permutation Representations

We begin by giving two examples of permutation actions φ, ψ of the group G = S4 . In the
action φ, we let G permute the set S = {A, B, C} consisting of the following three squares:
2.......................................................3 3.......................................................2 4.......................................................2
... ... ... ... ... ...
... ... ... ... ... ...
.. ... .. ... .. ...
... ... ... ... ... ...
..
..
A ...
...
..
..
B ...
...
..
..
C ...
...
... ... ... ... ... ...
.. . .. . .. .
................................................. ................................................. .................................................
4 1 4 1 3 1
The action of G is defined naturally by its action on the vertex labels {1, 2, 3, 4}. For
example, consider (12) ∈ G, which is a symmetry of square A, so  A(12)
 = A; however, it
(12) (12)
interchanges the other two squares: B = C, C = B. Thus φ (12) = (BC) ∈ Sym S
in the usual cycle notation. One of the conjugacy classes of G is C = {α=(12)(34),
β=(13)(24), γ=(14)(23)} and we denote by ψ the action of G on C by conjugation. For
example, consider the action of (12) ∈ G:
α(12) = α, β (12) = γ, γ (12) = β
 
so we may write ψ (12) = (βγ) ∈ Sym C. Writing () for the identity permutation on both
S and C, a table of values of the two actions φ, ψ is as follows:
g φ(g) ψ(g) g φ(g) ψ(g) g φ(g) ψ(g)
(1) () () (132) (ABC) (αβγ) (13)(24) () ()
(12) (BC) (βγ) (124) (ABC) (αβγ) (14)(23) () ()
(13) (AC) (αγ) (142) (ACB) (αγβ) (1234) (AC) (αγ)
(14) (AB) (αβ) (134) (ACB) (αγβ) (1243) (AB) (αβ)
(23) (AB) (αβ) (143) (ABC) (αβγ) (1324) (BC) (βγ)
(24) (AC) (αγ) (234) (ABC) (αβγ) (1342) (AB) (αβ)
(34) (BC) (βγ) (243) (ACB) (αγβ) (1423) (BC) (βγ)
(123) (ACB) (αγβ) (12)(34) () () (1432) (AC) (αγ)
 I. GROUPS 35

It should be clear from this table that the two permutation representations φ, ψ are equiva-
lent: every element g ∈ G permutes the three squares A, B, C in exactly the same way as it
permutes the three involutions α, β, γ. In other words, if we simply rename the three points
being permuted, each of the permutations φ(g) becomes ψ(g). This choice of renaming is
formally expressed by the bijection

.................................................... ..................................................
...
.. ... .. ...
...
..
A ...
θ
.......................................................................
... α
...
..
.. ..
... ... ... ...
.. ... ... ..
.. ... ... ..
... . . ...
.. .... .... . ..
.. ................................................................... ..
...
..
B ...
..
...
..
β ...
..
.. ... ... ..
... . . ...
.. .... .... ..
.. .. .. ..
... ................................................................... γ ...
..
..
C ..
...
.. ..
...
..
..
.. .
................................................ .
................................................

S C

The equivalence of the two permutation representations of G is expressed by the rule that
φ(g)θ = θψ(g) for all g ∈ G, using left-to-right composition. For example the fact that
φ((12)) ψ((12))
B −−−−→ C gives, after using θ to rename points, the fact that B θ −−−−→ C θ . In other
words, the effect of the composite map

φ((12)) θ
B −−−−→ C −→ C θ

is the same as the effect of the composite map

θ ψ((12))
B −→ B θ −−−−→ C θ .

In other words, φ((12))θ = θψ((12)). We are ready to formally define the notion of
equivalence of permutation actions. This is the appropriate equivalence relation for actions,
just as isomorphism is the appropriate equivalence relation for groups.
Let φ : G → Sym S and φ
: G → Sym S
be two permutation representations of
the same group G. Then we say that φ is equivalent to φ
if there exists a bijection
θ : S → S
such that φ(g)θ = θφ
(g) for all g ∈ G, i.e. the following diagram commutes:
θ .
S.. ............................................ S..

... ...
... ...
... ...
.. φ(g) .. φ
(g)
... ...
... ...
............ ............
.. ..
θ .
S ............................................ S

The map θ which achieves this equivalence, is called an intertwining map. To say that
this diagram commutes means that if we compose maps indicated by arrows, in this case
from upper left to lower right, it does not matter which of the two possible routes we take
(i.e. ‘right, then down’ gives the same result as ‘down, then right’).
36 I. GROUPS

There is a third permutation representation of G, equivalent to both of the two actions

above, that we need to mention: the action of G by right-multiplication on the three right
cosets of the subgroup H = GA = CG (α). This action is the map ρ : G → Sym(H\G)
defined by g → ρg where

ρg : (H\G) → (H\G), Hx → Hxg.

Explicitly, we list the right cosets of H as

H = {(1), (1324), (12)(34), (1423), (13)(24), (14)(23), (12), (34)},

H(14) = {(14), (132), (1243), (234), (1342), (23), (124), (143)},
H(13) = {(13), (243), (1234), (142), (24), (1432), (123), (134)}

We determine for example the action of (12) ∈ G:

H ρ(12) = H(12) = H,
 ρ
H(14) (12) = H(14)(12) = H(142) = H(13),
 ρ
H(13) (12) = H(13)(12) = H(132) = H(14).
 
We may write ρ(12) = H(14) , H(13) , if we are willing to stretch the customary cycle
notation a little. In fact, for every g ∈ G, ρg permutes right cosets of H in exactly the
same way as φ(g) permutes squares A, B, C; also the same way as ψ(g) permutes α, β, γ.
An intertwining map from φ to ρ is the map

................................................................... ....................................................................
.. ... ...
.. .. .. . ...
... . ...
.. A ....................................................................................... H ..
.. ... ... ..
... .. .. ...
.. ... ... ..
.. .. .. ..
... ... ... ...
.. . . ..
.. B ....................................................................................... H(14) ..
... .. .. ...
.. ... ... ..
.. .. .. ..
... ... ... ...
.. .. .. ..
.. . . ..
... ...................................................................................... H(13) ...
..
..
C ...
..
...
..
..
.
................................................................ ..................................................................

S H\G

This is typical of the more general result

10.1 Theorem. Let G be a group. Every transitive permutation action of G is

equivalent to the action of G, by right-multiplication, on the right cosets of a subgroup
H ≤ G. This subgroup is actually the stabilizer of one of the points.

Proof. Let φ : G → Sym S be a transitive permutation representation, and let a ∈ S.

We write simply ag in place of aφ(a) . Let H = Ga = {g ∈ G : ag = a}. In the course
 I. GROUPS 37

of proving Proposition 8.1, we showed that for all x, y ∈ G, ax = ay iff Hx = Hy. In

particular we have a well-defined bijection

θ : S → H\G, ax → Hx.

Let g ∈ G. Then for every x ∈ G we have

(ax )θρg = ((ax )θ )g = (Hx)g = Hxg,
whereas
(ax )φ(g)θ = ((ax )g )θ = (axg )θ = Hxg.
Thus θρg = φ(g)θ for all g ∈ G as required.

Return now to our example G = S4 with action φ : G → Sym S where S = {A, B, C}.
We see that ker φ = {(1), (12)(34), (13)(24), (14)(23)}, so φ is not faithful. Evidently it
is impossible for S4 to have a faithful permutation representation of degree 3, since the
order |S4 | = 24 exceeds 3! = 6. Moreover by the First Isomorphism Theorem, G/ ker φ ∼ =
24
φ(G) ≤ Sym S. We have equality: φ(G) = S, either by comparing orders ( 4 = 6 = 3!), or
by observing, from our table of values of φ(g), that φ is surjective.
What is the kernel of an arbitrary transitive permutation representation? Let φ : G →
Sym S be an arbitrary transitive permutation representation, and let a ∈ S. As before,
let H = Ga . We saw in Section 9 that the stabilizer of an arbitrary point ag ∈ S is the
conjugate subgroup H g = g −1 Hg. The set of all elements g ∈ G fixing every point of S is
the intersection of all these point stabilizers. This intersection is known as the core of H
in G. We have just proved

10.2 Proposition. Let φ : G → Sym S be a transitive permutation representation,

and let H be the stabilizer of some point of S. Then ker φ = CoreG (H).

Here we define, for an arbitrary group G and arbitrary subgroup H ≤ G,

CoreG (H) = Hg = g −1 Hg.
g∈G g∈G

This is a normal subgroup of G, since for all x ∈ G, we have

 x
g
H = H gx = Hy = Hg.
g∈G g∈G y∈G g∈G

Alternatively, CoreG (H) ≤ G follows from the fact that CoreG (H) = ker φ. In fact, this is
the largest normal subgroup of G contained in H, in the sense that if K ≤ G and K ⊆ H,

then K ⊆ CoreG (H), since K = K g ⊆ H g for all g ∈ G, and K ⊆ g∈G H g .
38 I. GROUPS

Finally, we are in a position to answer our earlier query about finding faithful permu-
tation representations of small degree for a given group G. (This is equivalent to expressing
G itself, not a quotient thereof, as a subgroup of Sn for small n.) Let H be a (fairly large)
subgroup of G, so that G acts transitively on the right cosets of H by right-multiplication.
Denote by ρ the corresponding permutation representation. In addition to choosing H
large (so that the degree [G : H] of the representation will be small), we require H to be

corefree in G, i.e. CoreG (H) = g∈G H g = 1. This will ensure that the representation is
faithful.
Exercises 10.
1. Prove that equivalence of permutation actions is an equivalence relation.

2. Up to equivalence, there are just two transitive subgroups of S3 , namely

(123) and S3 itself. Up to
equivalence, how many transitive permutation groups of degree 4 are there? Of degree 5?

3. Show that every finite group G is isomorphic to a subgroup of An for some n ≥ 1. What is the
minimum possible value of n if G ∼
= S3 × S3 ?

4. Let G be the group of rotational symmetries of a cube, so that |G| = 24. Follow steps (a)–(d) to show
that G ∼
= S4 .
(a) Define a diagonal of the cube to be a line joining two antipodal vertices. There are four such
diagonals, each of which passes through the center of the cube. Label the four diagonals by
D = {1, 2, 3, 4}. Clearly G acts on D; denote the corresponding action by φ. Show that this action
is transitive.
(b) Let H, the subgroup of S4 = Sym D generated by the eight 3-cycles. Show that H = S4 .
Hint: Lagrange’s Theorem limits the possibilities for a subgroup of S4 .
(c) By considering 120◦ rotations about the diagonals, show that the image φ(G) contains the eight
3-cycles.
(d) By considering the orders of G and S4 , show that | ker φ| = 1 and that φ is an isomorphism
G → S4 .

5. Show that the full isometry group G of a cube is isomorphic to C2 × S4 .

Hint: The rotational symmetry group H < G satisfies H ∼ = S4 by Exercise 10.4. Argue that |G| = 48
and so H < G. Observe that there is an element τ ∈ G mapping each vertex of the cube to its
antipode (the vertex furthest away). We may consider every element of G as a linear transformation
R3 → R3 , with origin at the center of the cube. By considering the matrix representing τ , show that
τ ∈ Z(G). Use Exercise 6.13.

6. Consider the dodecahedron pictured on the title page of these notes.

(a) Show that the rotational symmetry group of the dodecahedron is isomorphic to A5 .
(b) Show that the full symmetry group of the dodecahedron is isomorphic to C2 × A5 .
Hint: Recall Exercises 10.4 and 10.5.

7. Recall (see Exercise 3.7) that the group G = GL2 (F3 ) has order 48.
(a) Let H be the set of all upper-triangular matrices in G. Show that H is a subgroup of order 12.
(b) Determine K = CoreG (H).
(c) Using the action of G on right cosets of H, conclude that G/K ∼
= S4 . Thus, in notation similar to
Exercise 6.11, we have P GL2 (F3 ) ∼
= S4 .
 I. GROUPS 39

11. p-Groups and Sylow p-Subgroups

For a prime p, a p-group is a group whose order is a power of p. Thus, for example, a 2-
group has order ∈ {1, 2, 4, 8, . . .}. One property which makes p-groups special is that they
have nontrivial centers, as the following proposition shows. It is easy to find non-p-groups
with trivial centers, such as S3 .

11.1 Proposition. If P is a nontrivial p-group, then P has a nontrivial center, i.e.

|Z(P )| = 1.

Proof. Express P as a disjoint union of its conjugacy classes, thus:

P = {e} ∪ C2 ∪ C2 ∪ · · · ∪ Cr .
Each |Ci | divides |P | and so is a power of p. Then for some i ∈ {2, 3, . . . , r}, we must
have that |Ci | has size 1; otherwise |P | ≡ 1 mod p, a contradiction. This shows that {e}
is not the only singleton conjugacy class in P . By Exercise 9.5, it follows that Z(P ) is
nontrivial.

As simple as the latter proof is, it uses an important trick which will be encountered again
in the next section, in proving the Sylow Theorems.
Now let G be any finite group. A subgroup P ≤ G is called a p-subgroup if P is a
p-group. In case |P | is the highest power of p dividing |G|, we call P a Sylow p-subgroup

of G. Clearly if p  |G| then {e} is the only Sylow p-subgroup of G. But if p  |G|, it is
not obvious that G has any nontrivial p-subgroups. The existence of Sylow p-subgroups is
assured by the famous Sylow theorems, which we shall prove in Section 12. Their existence
may be regarded as a partial converse to Lagrange’s Theorem, in the case of prime-power
divisors of |G|. Recalling that G does not necessarily have subgroups of every order dividing
|G|, this shows another respect in which prime power orders are special.
Observe also that if P is a Sylow p-subgroup of G, then |P g | = |P | for every g ∈ G,
so clearly every conjugate of P is also a Sylow p-subgroup. It is less obvious that every
Sylow p-subgroup of G is conjugate to P , but this is another of Sylow’s theorems treated
in the next section.
In the remainder of this section we will be content to show how Sylow p-subgroups
may be constructed for one very important family of groups:

11.2 Proposition. For every positive integer n and every prime p, the group Sn has
a Sylow p-subgroup.

We proceed to construct only a Sylow 3-subgroup of Sn ; however you should have no

trouble in modifying our construction for a general prime p. Please observe that Sn may
have many Sylow 3-subgroups, and at present we are interested only in finding one. Note
40 I. GROUPS

that, for example, |S4 | = 24 = 23 · 3 has four Sylow 3-subgroups, namely (123), (124),
(134) and (234), each of which is cyclic of order 3.
Since |S1 | = 1 and |S2 | = 1, these groups have only the trivial Sylow 3-subgroup
{(1)}. So we may suppose that n ≥ 3. Now |S3 | = 6 = 2 · 3, |S4 | = 24 = 23 · 3 and
|S5 | = 120 = 23 · 3 · 5, so for 3 ≤ n ≤ 5, a Sylow 3-subgroup of Sn is given by (123). We
illustrate the action of our chosen Sylow 3-subgroup with a picture in each case, intended
to emphasize how it cyclically permutes 1, 2, 3:

123 123 4 123 45

−−−−→ −−−−→ −−−−→
n=3 n=4 n=5
Since |S6 | = 720 = 24 · 32 · 5, a Sylow 3-subgroup of S6 has order 32 ; we may choose
(123), (456), which permutes 1, 2, 3 cyclically, and independently permutes 4, 5, 6 cycli-
cally. This group is isomorphic to C3 × C3 , and we represent its action by the picture:

123 456
−−−−→ −−−−→
n=6
Since |S7 | = 24 · 33 · 5 · 7 and |S8 | = 27 · 32 · 5 · 7, we may also choose (123), (456) as a Sylow
3-subgroup of S7 and of S8 :

123 456 7 123 456 78

−−−−→ −−−−→ −−−−→ −−−−→
n=7 n=8
Now |S9 | = 27 · 34 · 5 · 7, so (123), (456), (789) does not suffice. However, a Sylow 3-
subgroup of S9 is given by the nonabelian subgroup
(123), (456), (789), (147)(258)(369) = (123), (147)(258)(369)
of order 34 , where (123), (456) and (789) each cyclically permutes the points within a block
of size three, and (147)(258)(369) cyclically permutes these three blocks themselves:

123 345 789

−−−−→ −−−−→ −−−−→
−−−−−−−−−−−−−−−−−→
n=9
Let’s call the latter pattern of three blocks a superblock. Now let’s jump ahead and see
what happens when n reaches the next power of 3, namely 27. In this case |S27 | =
223 · 313 · 56 · 73 · 112 · 132 · 17 · 19 · 23, so we may construct a Sylow 3-subgroup of S27 which
cyclically permutes three superblocks of size nine inside one superduperblock, and acts
individually on each superblock as in the case n = 9 above:
 I. GROUPS 41

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27
−−−−→ −−−−→ −−−−→ −−−−→ −−−−→ −−−−→ −−−−→ −−−−→ −−−−→
−−−−−−−−−−−−−−−−−→ −−−−−−−−−−−−−−−−−→ −−−−−−−−−−−−−−−−−→
−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−→
n = 27

Such a Sylow 3-subgroup has order 313 . Now for general n, consider the ternary (i.e.
base 3) expression for n, say
n = n0 + 3n1 + 32 n2 + 33 n3 + · · ·
where each ni ∈ {0, 1, 2}. Then a Sylow 3-subgroup of Sn has n0 fixed points, plus n1
blocks, plus n2 superblocks, plus n3 superduperblocks, etc. For example, 17 is written as
122 (base 3), so we need 2 fixed points, plus 2 blocks, plus 1 superblock, thus:

123 456 789 10 11 12 13 14 15 16 17

−−−−→ −−−−→ −−−−→ −−−−−→ −−−−−→
−−−−−−−−−−−−−−−−−→
n = 17

Explicitly, our Sylow 3-subgroup of S17 is given by

(1 2 3), (4 5 6), (7 8 9), (1 4 7)(2 5 8)(3 6 9), (10 11 12), (13 14 15)
= (1 2 3), (1 4 7)(2 5 8)(3 6 9), (10 11 12), (13 14 15)
of order 36 . It may be shown that the highest power of p dividing |Sn | = n! is pe where
     
n n n
e= + 2 + 3 +···
p p p

and as usual, x denotes the greatest integer ≤ x. Note that the latter sum has only
finitely many nonzero terms; for example
       
17 17 17 17
+ + + + · · · = 5 + 1 + 0 + 0 + · · · = 6,
3 9 27 81

so a Sylow 3-subgroup of S27 has order 36 as above.

Exercises 11.
1. List all Sylow 2-subgroups, all Sylow 3-subgroups, and all Sylow 5-subgroups of S5 .

2. Let H be a proper subgroup of a p-group P , and let N = NP (H). Prove that H < N.
Hint: The key conclusion here, namely the fact that H is properly contained in N, follows by an
argument similar to the proof of Proposition 11.1.

3. Let G = GLn (Fp ), the multiplicative group of invertible n × n matrices over Fp = {0, 1, 2, . . . , p−1};
see Exercises 3.7 and 8.5. By considering upper-triangular matrices, find an explicit Sylow p-subgroup
of G.
42 I. GROUPS

12. The Sylow Theorems

The Theorems of Sylow are among the most celebrated in group theory. Their proofs will
require a little more work than previous results.

12.1 Sylow Theorems. Let G be a finite group, and let p be a prime. Then the
following statements hold.
(i) G has a Sylow p-subgroup.
(ii) Any two Sylow p-subgroups of G are conjugate in G.
(iii) Every p-subgroup of G is contained in a Sylow p-subgroup of G.
(iv) The number of Sylow p-subgroups of G is a divisor of |G| of the form kp + 1 for
some integer k ≥ 0.

Note that (iv) implies (i); however, we shall prove (i) before the remaining conclusions.
The first Sylow Theorem (i) is an easy consequence of the following Proposition 12.2. To
see why, let H be any finite group. If n is large enough, then by the Cayley Representation
Theorem 8.2, we may consider H to be a subgroup of Sn . Now Sn has a Sylow p-subgroup
P by Section 11. Then Proposition 12.2 says that we can obtain a Sylow p-subgroup of H
by intersecting H with a suitable conjugate of P in Sn . [Alternatively, H is isomorphic
to a subgroup of GLn (Fp ) by Exercise 8.5, and GLn (Fp ) has a Sylow p-subgroup by
Exercise 11.3.] Later we will also see how conclusions (ii) and (iii) also follow from this
same Proposition; then we will need another argument to obtain (iv). But let’s first verify
the following

G .....
... .. ..........
.
.... ....
. .....
.....
. ...
. .....
12.2 Proposition. Let G be a finite group .
.
.
... ....
.
.....

..
.
.
. .
.
.
. ....
H
with subgroup H. If P is a Sylow p-subgroup of .
...
.
... ..
... ... ...
...
G, then there exists g ∈ G such that P g ∩ H is P P
.
g.
.....
....
...
..... ...
..... ..
a Sylow p-subgroup of H. .....
.....
..... ...
..
. .
P g ∩H

Proof. As usual we denote the set of right cosets of P in G by

P \G = {P g : g ∈ G}.

The number of cosets, m = [G : P ] = |G|/|P |, is not divisible by p since |P | is the highest

power of p dividing |G|. Recall from Section 8 that G acts transitively on P \G by right
multiplication; and that the stabilizer of P in this action is P . Using observations from
Section 9, the stabilizer of any right coset P g ∈ P \G is the conjugate P g .
 I. GROUPS 43

P P g2 P g3 · · · P gm

 

P \G
Consider now the action of H on P \G by right-multiplication. In general we cannot
expect H (like G) to permute the right cosets of P in G transitively; rather, P \G may
split into several H-orbits, as shown:

one H-orbit another H-orbit last H-orbit

 
 
 

P Pa ··· Pb Pc ··· Pd ··· Px Py ··· Pz

 

P \G
However, one thing is certain: the H-orbits do not all have length divisible by p, since the
total of the lengths of H-orbits, namely m, is not divisible by p. So we may choose a coset
P g belonging to an H-orbit of length not divisible by p. Since the stabilizer of P g in G is
P g (see Section 9), the stabilizer of P g in H is

{x ∈ P g : x ∈ H} = P g ∩ H.

So by Proposition 8.1, the length of the H-orbit containing P g is [H : P g ∩H] = |H|/|P g

∩H|, which is not divisible by p. However, P g ∩H ≤ P g is a p-group. This means that
|P g ∩H| is the highest power of p dividing H, as required.

We now complete the proof of the Sylow Theorems. Conclusion (i) has been verified,
as noted earlier.
To prove (ii), let P and Q be two Sylow p-subgroups of G. Applying Proposition 12.2
with Q in place of H, there exists g ∈ G such that P g ∩Q is a Sylow p-subgroup of Q. Since
|P g ∩ Q| is the highest power of p dividing |Q|, we must have P g ∩ Q = Q, i.e. Q ⊆ P g .
But |Q| = |P g | is the highest power of p dividing |G|, so Q = P g , which proves (ii).
To prove (iii), let P be a Sylow p-subgroup of G, and let H be any p-subgroup of G.
By Proposition 12.2, there exists g ∈ G such that P g ∩ H is a Sylow p-subgroup of H. As
before, this means that P g ∩ H = H, so H is contained in P g , a Sylow p-subgroup of G.
Finally, to prove (iv), we make use of the following:

12.3 Proposition. If P = Q are two Sylow p-subgroups of G, then Q does not

normalize P , i.e. Q ⊆ NG (P ).
44 I. GROUPS

Proof of Proposition 12.3. Suppose that Q ⊆ NG (P ). Since P ⊆ NG (P ), we observe the

inclusions indicated by the diagram
G..
....
...
..
.
NG. (P
.
)
...
... ...
..
....
.
...
...
.. ...
..
P Q
Since |P | = |Q| is the highest power of p dividing |G|, it must also be the highest power
of p dividing |NG (P )|. So P and Q are Sylow p-subgroups of NG (P ). By (ii), there exists
g ∈ NG (P ) such that P g = Q. But since g ∈ NG (P ), by definition we have P g = P . Thus
P = P g = Q, contradicting the assumption that P and Q are distinct.

We now return to prove (iv). Let P be a Sylow p-subgroup of G, whose existence is

guaranteed by (ii). Let S = {P1 =P, P2 , P3 , . . . , Ps } be the class of all Sylow p-subgroups
of G. If s = 1 then we are done; hence we assume that s ≥ 2. By (ii), every member of S
is a conjugate of P , or in words, G acts transitively on S by conjugation:

P P2 P3 ··· Ps

 

S
Recall from Section 9 that NG (P ) is the stabilizer of P in this action, and so s = [G :
NG (P )] = |G|/|NG (P )| is a divisor of |G|. We must show that s ≡ 1 mod p.
Consider the action of P on S by conjugation. This will no longer be transitive! In
particular P normalizes P , so P fixes P in this action. The remaining members of S are
partitioned into P -orbits of undetermined size, thus:

fixed
by P another P -orbit yet another P -orbit last P -orbit

 
 
 
 

P Pi1 Pi2 ··· Pit P j1 P j2 ··· P ju ··· P1 P2 ··· Pv

 

S
By Proposition 8.1, the length of every P -orbit divides |P |, and so is one of 1, p, p2 , . . . , |P |.
But if i = 1 then by Proposition 12.3, P does not normalize Pi . This means that Pi is
not fixed in the conjugation action of P , so the length of the orbit containing Pi is one of
p, p2 , . . . , |P |. In particular S is a disjoint union of P -orbits, all having length divisible by
p, except for the singleton orbit {P }, which has length 1. Thus s = |S| ≡ 1 mod p as
required.
 I. GROUPS 45

Exercises 12.
1. If G is a group of order 42, how many subgroups of order 7 does G have? Why?
2. Show that every group of order 15 is cyclic.
3. If H is any subgroup of S11 of order 110, must H act transitively on {1, 2, 3, . . . , 11}? Why or why
not?
4. Let G be a finite group, P a Sylow p-subgroup of G for some prime p, and N = NG (P ). If H ≤ G
has index [G : H] not divisible by p, show that HN = G.
Hint: Use Exercise 8.2.
∗ 5. Show that every simple group of order 60 is isomorphic to A5 .
Hint: Consider the action of G on its Sylow 2-subgroups by conjugation.

13. Composition Series

A normal series for a group G is a chain of subgroups of the form

1 = G0 < G1 < G2 < · · · < Gk−1 < Gk = G,

this example having length k. Sometimes we may relax the condition Gi−1 < Gi to
Gi−1 ≤ Gi ; but in such cases the repeated subgroups may be deleted until the remaining
inclusions are proper. It does not follow (cf. Exercise 5.4) that every Gi is normal in G;
rather, we say that Gi is subnormal in G. A second normal series for G,

1 = H0 < H1 < H2 < · · · < H−1 < H = G,

is called a refinement of the former series, if every Gi is an Hj for some J, i.e. if the chain
Hj : 0 ≤ j ≤  is obtained from the chain Gi : 0 ≤ i ≤ k by ‘squeezing in’ additional sub-
groups in a way that preserves normality of adjacent terms in the chain, and in particular
 ≥ k. Every group G = 1 has a trivial normal series 1 < G , of which all other normal
series are refinements. Observe that G is simple iff the series 1 < G cannot be properly
refined.
Assume that 1 < |G| < ∞. Starting with any normal series for G (such as the trivial
series), after a finite number of refinements we arrive at a normal series for G which may not
be further refined. Such a series is called a composition series for G. Equivalently, any
normal series 1 = G0 < G1 < G2 < · · · < Gk−1 < Gk = G is a composition series iff Gi−1
is maximal normal in Gi for all i = 1, 2, . . . , k (see Exercise 6.6), iff each factor Gi /Gi−1 is
simple. Accordingly, the k quotient groups Gi /Gi−1 are called the composition factors
of G, and k is the composition length of G. The composition factors and composition
length of G are independent of the choice of composition series:

13.1 Jordan-Hölder Theorem. Any two composition series for a finite group G
have the same length and factor groups, counting multiplicity.
46 I. GROUPS

We will not prove this result here, although it is fundamentally important for us to be able
to use the result. The Jordan-Hölder Theorem generalizes the Fundamental Theorem of
Arithmetic (see Section 0). Many problems in finite group theory are tackled by induction
on the composition length of the group, typically by reducing the problem to the case of
a simple group.
Consider, for example, the cyclic group of order 12, denoted C12 = {e, x, x2 , . . . , x11 },
whose subgroups are described explicitly in Section 4. Since every subgroup is normal, C12
has exactly three composition series, each of length three:

1 < x6  < x3  < C12 1 < x6  < x2  < C12 1 < x4  < x2  < C12
C2 C2 C3 C2 C3 C2 C3 C2 C2
 
  
  

composition factors composition factors composition factors

Whichever composition series we choose for C12 , the composition factors are C2 , C2 and
C3 . As another example, a composition series of length 4 for S4 is shown:

1 < (12)(34) < (12)(34), (13)(24) < A4 < S4

C2 C2 C3 C2
 

composition factors

Only two other composition series for S4 exist, found by replacing (12)(34) by either
(13)(24) or (14)(23). Observe that from the composition series above, A4 has the same
composition factors as C12 . So unlike a positive integer, which is uniquely determined
by its prime factors (counting multiplicity), a group is not uniquely determined by its
composition factors. A different sort of example is S5 , whose unique composition series is

1 < A5 < S5
A5 C2
 

composition factors

We say that G is solvable if all its composition factors are cyclic of prime order. Con-
versely, G is nonsolvable if at least one nonabelian simple group is among the composition
factors of G. The examples above show that C12 and S4 are solvable, whereas S5 is non-
solvable. Solvability is an important property in the study of groups, and especially so
for us since it relates directly to the question of whether a given polynomial equation is
solvable by radicals . . . more about this later in the course!

13.2 Proposition. If G is solvable, then so is every subgroup of G.

Proof. Let H ≤ G, where G has a composition series

1 = G0 < G1 < G2 < · · · < Gk−1 < Gk = G

 I. GROUPS 47

in which each factor Gi /Gi−1 is cyclic of prime order. Intersect every term in the latter
series with H to obtain Hi = H ∩Gi . It is easily checked that Hi−1 ≤ Hi for i = 1, 2, . . . , k.
Furthermore
Hi /Hi−1 = Hi /(Hi ∩ Gi−1 ) ∼
= Hi Gi−1 /Gi−1 ≤ Gi /Gi−1
using the Second Isomorphism Theorem (Exercise 6.8). Since Gi /Gi−1 is cyclic of prime
order, Hi /Hi−1 is either trivial or cyclic of prime order. So after removing duplications
from the sequence
1 = H0 ≤ H1 ≤ H2 ≤ · · · ≤ Hk−1 ≤ Hk = H,
we obtain a composition series for H, in which all composition factors are cyclic of prime
order.

It is also true that homomorphic images of solvable groups are solvable. A stronger
statement is the following

13.3 Proposition. Suppose that H < G. Then G is solvable iff both H and G/H
are solvable.

Proof. We may assume that 1 = H = G. Refine the normal series 1 < H < G to obtain
a composition series for G which includes H:
1 = G0 < G1 < G2 < · · · < Gk−1 < Gk = H < Gk+1 < · · · < G−1 < G = G.
From this we can immediately read off a composition series for H:
1 = G0 < G1 < G2 < · · · < Gk−1 < Gk = H,
and the composition factors for H are the first k of the composition series for G, namely
Gi /Gi−1 for i = 1, 2, . . . , k. Also by Exercise 6.6, we obtain a normal series for G/H:
1 = H/H = Gk /H < Gk+1 /H < · · · < G−1 /H < G /H = G/H.
By the Third Isomorphism Theorem (Exercise 6.9), the factors in this series are

(Gi /H) (G /H) ∼= Gi /Gi−1 for k < i ≤ ,
i−1

all of which are simple, so we have a composition series for G/H. Thus the composition
factors of H, together with those of G/H, make up those of G, counting multiplicity. The
result is then immediate.

Exercises 13.
1. Prove that every abelian group is solvable.
2. Prove that every dihedral group is solvable.
3. Prove that every p-group is solvable.
4. Prove that Sn is solvable iff n ≤ 4.
 CHAPTER II

Rings
14. Definitions and Examples

Unfortunately there are many definitions to absorb in this new topic! But fortunately, the
definitions are reasonable, and are supported by examples, which we shall supply as soon
as possible.
A ring is a set R with two binary operations, usually addition (denoted by ‘+’) and
multiplication (denoted by juxtaposition of elements), such that
(i) R is an abelian group under addition, with additive identity denoted by 0;
(ii) R has associative multiplication, i.e. (ab)c = a(bc) for all a, b, c ∈ R; and
(iii) multiplication is distributive over addition, i.e. a(b+c) = ab+ac and (a+b)c = ac+bc
for all a, b, c ∈ R.
If ab = ba for all a, b ∈ R, we say that R is commutative. If R has a two-sided
multiplicative identity 1 ∈ R with 1 = 0, then R is a ring with unity, or a ring with
identity. (This is often called a ring with unit, which is somewhat misleading, since units,
as we are about to define shortly, include more than just 1.) We require 1 = 0 in this case,
in order to eliminate the trivial ring R = 0. It is easy to see that if a two-sided identity
exists, then it is unique (by the same trick as in Section 1 for groups). Warning: Many
authors use the term ‘ring’ to refer to a commutative ring with identity. When referring
to a book for results on ‘rings’, be sure to check carefully at the beginning of the section,
chapter or book to see what the author means by this terminology.
The additive inverse of a ∈ R is denoted by −a, and b − a means b + (−a).

14.1 Proposition. For all a, b ∈ R, we have

(i) 0a = a0 = 0, and
(ii) (−a)b = a(−b) = −(ab), so we denote this simply by −ab.

Proof. We have 0a = (0 +0)a = 0a +0a, and similarly on the other side, which proves (i).
Also 0 = 0b = (a + (−a))b = ab + (−a)b, which proves that (−a)b = −(ab); similarly,
a(−b) = −(ab).
50 II. RINGS

More generally, we define integer multiples of ring elements as follows: for each positive
integer k and each a ∈ R, we define
k times

 
ka = a + a + · · · + a ∈ R;

and for each negative integer k, define ka = −(|k|a) where |k|a is defined as above. It is
easy to check that
k(a) = (k)a, (ka)b = k(ab),

(k + )a = ka + a, k(a + b) = ka + kb

for all k,  ∈ Z and a, b ∈ R. If it happens that the integer k is in R as well as in Z, there
is no need to worry about whether the product ka should be regarded as a product in R
or as defined above, for in such cases, both answers must agree. For each a ∈ R we write
Za = {ka : k ∈ Z}; clearly this is the additive subgroup of R generated by a.
Let R be a ring. If a, b ∈ R such that a, b = 0 but ab = 0, then a and b are called
zero divisors. If R is a ring with unity, we say an element a ∈ R is invertible (or that
a is a unit) if ab = ba = 1 for some b ∈ R. If such an inverse exists, it is easy to see
that it is unique, and so we write it as b−1 . Note that no unit is a zero divisor; if bc = 0
where b is a unit, then left-multiplication by b−1 gives c = 0. It is easy to show that in
any ring with unity, the set of units forms a multiplicative group. This group is called the
unit group of R, and is denoted R× (or sometimes R∗ ).
An integral domain is a commutative ring with unity and no zero divisors. A
skewfield (or division ring) is a ring in which the nonzero elements form a multiplicative
group. A field is a ring in which the nonzero elements form an abelian group. In other
words, a commutative ring with unity is a field if every nonzero element is a unit; if we
remove the requirement of commutativity, we get a skewfield. (Warning: Some older
books use the term ’field’ for skewfield.) Since units are not zero divisors, every field is an
integral domain. Here is a picture showing all implications between the classes of rings we
have defined:
commutative commutative
integral ......................................... ring with .........................................
............
.. ... domain
. . ring
..........
........
unity
.......
...... ......
...... ......
field . ....
..
......
............ ............. ............
.......... . .
...........
........
skewfield ..
........................................
ring with ..
........................................ ring
unity

None of these implications is reversible.

Two rings R and S are isomorphic (written R ∼ = S) if there exists an isomorphism
φ : R → S, i.e. a bijection φ : R → S such that φ(x + y) = φ(x) + φ(y) and φ(xy) =
φ(x)φ(y) for all x, y ∈ R. Isomorphism is clearly an equivalence relation on the class of all
rings, and two rings in the same isomorphism class will be considered essentially the same.
 II. RINGS 51

Examples. Some well-known fields include R, C and Q. The ring of integers, Z, is

an integral domain, whose group of units is {1, −1}. The set of all continuous real-valued
functions defined on the interval [a, b], under the operations of pointwise addition and
multiplication of functions, is a commutative ring with unity, denoted C([a, b]). Clearly
C([a, b]) has zero divisors, for example f g = 0 where f and g are nonzero elements of
C([a, b]) with graphs as shown:
.. ..
.. f (x)=
..
..
max{ 0,x− a+b } 2 ................
.
........ ..
..
..
......... g(x)=
.........
.........
max{ 0, a+b
2 −x }
... .
.......
.
. ... .........
.
.. ......... .. .......
... .......... ... ..........
.. ...... .. .........
... .........
.
.
. ...
........
.........
....... ........
... ......... ... ..........
.. ......... .. .........
..
... .............
.
.
.
..
...
........
........
. ......... .. ..
................................................................................................................................................................................ ........................................................................................................................................................................................
.... ....
0 a
..
...
a+b b 0 a
..
...
a+b b
...
..
2 ...
..
2
.. ..
. .

The group of units of C([a, b]) is {f ∈ C([a, b]) : f (x) = 0 for all x ∈ [a, b]}. Note
that if pointwise multiplication is replaced by composition, then we don’t get a ring since
f ◦ (g + h) = f ◦ g + f ◦ h.
If R is any ring, then we have the ring of n × n matrices over R,
⎧⎛ ⎞ ⎫
⎪ a11 a12 ··· a1n ⎪
⎪
⎨⎜ a21 ⎪
⎬
a22 ··· a2n ⎟
Rn×n = ⎜
⎝ ... .. .. ⎟
.. ⎠ : aij ∈ R ,
⎪
⎪ . . . ⎪
⎪
⎩ ⎭
an1 an2 · · · ann

under the usual matrix addition and multiplication. (Sometimes Rn×n is denoted Mn (R).)
If R has unity 1, then Rn×n has identity given by the usual n × n identity matrix I.
Typically Rn×n is noncommutative; for example F n×n is noncommutative for every field
F and all n ≥ 1. Also if n ≥ 2 and R = 0, then Rn×n has zero divisors, for example
 a 0  0 0  0 0
0 0 0 a = 0 0

where 0 = a ∈ R. The group of units of Rn×n is GLn (R); see Exercise 3.7.
The ring of integers modulo n is

Z/nZ = {0, 1, 2, . . . , n−1}

with addition and multiplication modulo n. This is a commutative ring with unity 1. For
example in Z/6Z, we have 2 + 3 = 5, 4 + 3 = 1, 5 · 2 = 4, 2 · 3 = 0 = 3 · 4. In particular, we
have zero divisors 2, 3, 4 ∈ Z/6Z. The group of units is (Z/6Z)× = {1, 5}. Note that our
symbols for elements of Z/nZ are ambiguous unless the value of n is clear from context.
(Warning: Some authors write Zn in place of Z/nZ. This unfortunate practice conflicts
with the usage of Zp for the ring of p-adic integers, which we will not define here. The
notation Z/nZ is standard notation for a quotient ring, which we will define in Section 15.)
52 II. RINGS

Every finite field has size equal to a power of some prime p; see Exercise 18.5. More-
over, for every q = pe where p is prime and e ≥ 1, there exists (up to isomorphism) a
unique field with exactly q elements, called the finite field (or Galois field) of order q,
denoted by Fq or GF (q). When q = p is prime, this is nothing other than Z/pZ; however
for q = p2 , p3 , . . ., the ring Z/qZ has zero divisors, and so it is quite different from the field
Fq .
If R is any ring and X is an indeterminate (i.e. a symbol with no numerical value),
then we have the polynomial ring R[X] consisting of all polynomials p(X) = a0 +
a1 X + a2 X 2 + · · · + ak X k where a0 , a1 , . . . , ak ∈ R, k ≥ 0, with the usual addition
and multiplication of polynomials. In this case R is usually commutative with unity 1,
in which case R[X] is commutative with unity 1 ( = the constant polynomial 1). Note
that X commutes with every element of R. The additive identity of R[X] is the zero
polynomial, which is the polynomial all of whose coefficients are zero. The degree of
any nonzero polynomial p(X), denoted deg p(X), is the largest k such that the coefficient of
X k in p(X) is nonzero. We often extend this definition to say that the zero polynomial has
degree −∞; this convenience allows us to state Theorem 14.2 below, for all polynomials,
including the zero polynomial. We also define a constant polynomial to be a polynomial
of degree ≤ 0, i.e. an element of R interpreted as a polynomial with no X k terms for k ≥ 1.
 
14.2 Theorem. If R is an integral domain, then so is R[X], and deg f (X)g(X) =
deg f (X) + deg g(X) for all f (X), g(X) ∈ R[X].

Proof. We may suppose that f (X) and g(X) are not both zero; otherwise the result holds
with the usual convention that −∞ + k = −∞ whenever k ∈ Z ∪ {−∞}. Thus we may
write f (X) = a0 + a1 X + a2 X 2 + · · · + ak X k and g(X) = b0 + b1 X + b2 X 2 + · · · + b
X

where ak , b
= 0. Then

f (X)g(X) = a0 b0 + (a0 b1 + a1 b0 )X + · · · + ak b
X k+
.

 
If R is an integral domain then ak b
= 0, in which case deg f (X)g(X) = k +  and
f (X)g(X) = 0, so that R[X] has no zero divisors.

It is important to realize that each element of R[X] is simply a formal expression of

the form p(X) = a0 +a1 X +a2 X 2 +· · ·+ak X k , not to be confused with a function R → R.
To make this point clear, consider the case R = F2 = {0, 1}, which is the field of integers
modulo 2. The polynomial p(X) = X + X 2 ∈ F2 [X] represents the zero function R → R,
and yet f (X) is by definition not the zero polynomial, since not all its coefficients are
zero! So in the proof above, the conclusion f (X)g(X) = 0 does not mean that f (X)g(X)
has no zeroes in R; nor does it mean that the function R → R is not identically zero.
Rather, it means that f (X)g(X) is not the zero polynomial, i.e. that f (X)g(X) has at
least one nonzero coefficient. Moreover, 1+X +X 2 ∈ F2 [X] is by definition not a constant
 II. RINGS 53

polynomial, although it represents a constant function F2 → F2 . And finally, when we

write f (X) = g(X), we mean simply that they are the same polynomial, i.e. that all their
corresponding coefficients are equal, which is stronger than saying that they represent the
same function R → R. For example, X = X 2 , even though they represent the same
function F2 → F2 .
If you are used to working only with the fields R and C, you might be a little shaken
up, or at least a little confused, by these examples, as I once was. If so, I hope you will
sleep better tonight if I assure you that two polynomials f (X), g(X) ∈ R[X] represent
the same function (i.e. f (a) = g(a) for all a ∈ R), then they are the same polynomial
(i.e. all corresponding coefficients are the same). This is more generally true with any
infinite coefficient field in place of R. Yet this is not obvious, and needs to be proved! (See
Exercise 15.9.) Until it is clear to you how to think of a polynomial as distinct from a
function, or how to think of a0 +a1 X +a2 X 2 +· · ·+ak X k as a formal expression, you would
be better to treat elements of R[X] as finite sequences over R of the form (a0 , a1 , a2 , . . . , ak ),
of arbitrary length. Addition is componentwise (after the shorter sequence has been padded
with 0’s on the right until the sequences have the same length). And multiplication is
defined by
(a0 , a1 , a2 , . . . , ak )(b0 , b1 , b2 , . . . , b
) = (a0 b0 , a0 b1 +a1 b0 , . . . , ak−1 b
+ ak b
−1 , ak b
).

Padding 0’s on the right of any sequence doesn’t change its meaning, so (a0 , a1 , . . . , ak ) =
(a0 , a1 , . . . , ak , 0, 0, . . . , 0). Of course the use of an indeterminate X makes the notation
much more readable and the rule for multiplication much easier to remember; yet the X
serves merely as a placeholder, and the expressions involving X have exactly the same
meaning as these sequences we have just introduced. And that’s what polynomials really
are, as distinct from whatever functions they might represent.
Now let’s consider the ring of polynomials in n indeterminates X1 , X2 , . . . , Xn and co-
efficients in R, denoted R[X1 , X2 , . . . , Xn ]. Of course the indeterminates X1 , X2 , . . . , Xn
commute with each other, as well as with every element of R. Clearly, every f (X1 , X2 , . . . ,
Xn ) ∈ R[X1 , X2 , . . . , Xn ] may be uniquely expressed as a polynomial in Xn with coeffi-
cients in R[X1 , X2 , . . . , Xn−1 ]; this gives an isomorphism

R[X1 , X2 , . . . , Xn ] ∼
= (R[X1 , X2 , . . . , Xn−1 ])[Xn ].

So by induction, using Theorem 14.2 we obtain

14.3 Theorem. If R is an integral domain, then so is R[X1 , X2 , . . . , Xn ].

I still owe you an example of a noncommutative skewfield. It is a fact (Wedderburn’s

Theorem) that every finite skewfield is a field, so our example will necessarily be infinite.
Everyone’s favorite example is the set H of real quaternions, which are expressions of
the form
x = a + bi + cj + dk, a, b, c, d ∈ R,
54 II. RINGS

so that H is a 4-dimensional vector space over R with basis {1, i, j, k}, and the usual
componentwise vector addition. Multiplication of real quaternions is associative, and works
according to the famous rules of Hamilton:
i2 = j2 = k2 = ijk = −1.
Furthermore every real number commutes with i, j and k. Then H is a ring with unity 1,
but H is not commutative, since for example, ij = k = −ji, as you should verify using the
rules above. Define the conjugate of an arbitrary x = a + bi + cj + dk ∈ H by
x = a − bi − cj − dk,
and the norm of x by
√ 
||x|| = xx = a2 + b2 + c2 + d2 .
This is just the usual Euclidean norm when H is identified naturally with the real vector
space R4 . One checks that xy = y x for all x, y ∈ H, and from this it follows that
||xy|| = ||x|| ·||y|| for all x, y ∈ H. From the identity xx = ||x||2 we see that every nonzero
real quaternion is a unit; indeed if x = a + bi + cj + dk = 0, then

x a b c d
x−1 = 2
= 2
− 2
i− 2
j− 2k.
||x|| ||x|| ||x|| ||x|| ||x||

A subset S of a ring R which is itself a ring with respect to the binary operations of R
restricted to S, is called a subring of R. It is easy to check that a nonempty subset S ⊆ R
is a subring iff S is closed under products and differences, i.e. if ab, a − b ∈ S whenever
a, b ∈ S. For example, Z is a subring of Q, which is a subring of R, which is a subring of
C. It is possible to see C as a subring of H in several different ways; for example,

{a + bi : a, b ∈ R} ∼
= {a + cj : a, c ∈ R} ∼
= {a + dk : a, d ∈ R} ∼
= C.

The set of all constant polynomials in R[X] is a subring isomorphic to R, so we naturally

identify R with this subring. The matrix ring Rn×n has many subrings; for example the
set of diagonal matrices; or the set of upper-triangular matrices. The set of all scalar
matrices (matrices of the form aI, a ∈ R) is a subring isomorphic to R.
If a field E has a subring F ⊆ E which is also a field, then we say that F is a subfield
of E, or that E is an extension of F . It follows directly from the axioms in this case that
E is a vector space over F . The dimension of this vector space is known as the degree of
the extension E ⊇ F , and this number (finite or infinite) is denoted [E : F ]. We have for
example that [R : Q] = ∞; see Exercise 14.11. Extensions of degree 2, 3, 4, 5 are called
quadratic, cubic, quartic, quintic respectively. For example the extension C ⊃ R is
quadratic since {1, i} is a basis for C over R. A tower of extension fields is a sequence of
extensions F1 ⊇ F2 ⊇ · · · ⊇ Fk .
If α ∈ E ⊇ F , then we define
F [α] = {f (α) : f (X) ∈ F [X]}.
 II. RINGS 55

Clearly F [α] is the smallest subring of E containing both F and α. We have E ⊇ F [α] ⊇ F ,
but F [α] may not be a field. The smallest subfield of E containing both F and α is the
subfield is  
g(α)
F (α) = : g(X), h(X) ∈ F [X], h(α) = 0 .
h(α)
We now have E ⊇ F (α) ⊇ F [α] ⊇ F in which E ⊇ F (α) ⊇ F is a tower of extension
fields. Notice that the construction of the field F (α) from the ring F [α] closely resembles
the construction of the field Q from the ring Z. We call F (α) the quotient field of F [α]. In
a similar way, every integral domain is extendible to its quotient field; see Exercise 14.13.
We similarly define the ring F [α1 , . . . , αk ] and its quotient field F (α1 , . . . , αk ).
It frequently happens that the ring F [α] is already √ a field,√in which case its quotient
√
field F√(α) = F [α]. For example, we √ have R ⊇ Q( 5) ⊇ Q[ 5] ⊇ Q, where Q[ 5] =
{a + b 5 : a, b ∈ Q}. We note that Q[ 5] is a field. To verify this, we only need to check
that it is closed under division. This follows from the familiar process√of rationalizing the
denominator; for example, here we divide two typical elements of Q[ 5]:
√ √ √ √
3
2 − 1
3 5 18 − 4 5 9 − 6 5 282 − 144 5 √
√ = √ · √ = = − 94
33 +
16
11 5.
3
4 + 1
2 5 9+6 5 9−6 5 81 − 180
√ √ √
Thus
√ Q( 5) = √Q[ 5] = {a + b 5 : a, b ∈ Q}, is a quadratic extension of Q, having basis
{1, 5}, i.e. [Q[ 5] : Q] = 2.
Finally, for an example to show that F [α] need not be a field, see Exercise 14.12. The
precise conditions under which F (α) = F [α] will be presented in Theorem 18.2.
Exercises 14.
1. Let R be a ring with unity. Show that the units of R[X] are just the units of R.

2. Let R and S be rings. Define R ⊕ S = {(r, s) : r ∈ R, s ∈ S} with addition and multiplication defined
componentwise by
(r, s) + (r
, s
) = (r+r
, s+s
); (r, s)(r
, s
) = (rr
, ss
).
(a) Show that R ⊕ S is a ring. This is known as the direct sum of R and S.
(b) If R and S are rings with unity, show that R ⊕ S is also a ring with unity, and that its unit group
is given by (R ⊕ S)× = R× × S × .

3. Let R be a ring. Define the center of R by

Z(R) = {a ∈ R : ra = ar for all r ∈ R},
and the centralizer of an element r ∈ R by
CR (r) = {a ∈ R : ar = ra}.
Show that Z(R) and CR (r) are subrings of R.

4. Show that the unit group of Zn×n (the ring of all n × n matrices with integer entries) is the group
GLn (Z) = {A ∈ Zn×n : det(A) = ±1}.
56 II. RINGS

5. Can a ring equal the union of two of its proper subrings? Give an easy example, or prove that this is
not possible.

6. Let R be any ring, and let S = R2×2 . Show that the ring S 2×2 is naturally isomorphic to R4×4 . Try
to generalize this result.

7. (a) Show that

a b
: a, b ∈ R
−b a
is a subring of R2×2 isomorphic to C.
(b) Use similar ideas to show that H is isomorphic to a subring of R4×4 , and isomorphic to a subring
of C2×2 .

8. Let R be a nontrivial ring. Show that

⎧⎛ ⎞ ⎫
⎨ a b c ⎬
S = ⎝ c a b ⎠ : a, b, c ∈ R
⎩ ⎭
b c a
is a subring of R3×3 , and that S has zero divisors.

9. Prove that no two of the rings Z, Z[X], Z[X, Y ] are isomorphic. Is there a ring R such that R[X] ∼
= R?
Justify your answer.

10. Consider the ring R = {a + bI + cJ + dK : a, b, c, d ∈ C} where I 2 = J 2 = K 2 = IJK = −1. Show

that R is not a skewfield.

11. Prove that [R : Q] = ∞.

Hint: Show that for every integer n ≥ 1, the vector space Qn is countably infinite.
1
12. Show that π
∈
/ Q[π] and conclude that Q(π) ⊇
.. Q[π].

Hint: You man assume the fact that π is not a zero of any nonzero f (X) ∈ Q[X]. This is the statement
that π is transcendental; more about this in Section 18.

13. The following exercise shows that every integral domain R is a subring of some field. In examples
we have seen so far, this was clear since R was chosen as a subring of a known field. But given any
integral domain R, we can nevertheless construct a field containing R, and whose elements are simply
the quotients of the original ring R. The construction follows:
Let R be an integral domain. For a, b, c, d ∈ R, we say that (a, b) ∼ (c, d) iff ad = bc. This defines a
relation on R × R.
(a) Prove that ∼ gives an equivalence relation on the set S = {(x, y) ∈ R × R : y = 0}.
a
(b) Let K denote the set of equivalence classes, and denote by b
the equivalence class of (a, b). Define
addition and multiplication on K by
ad+bc
a
b
+ c
d
= bc
, a
b
· dc = ac
bd
.
Prove that K is a field. We call K the quotient field of R, and this construction naturally
generalizes the construction of Q from Z.
(c) Prove that the set of all a1 with a ∈ R is a subring of K, naturally isomorphic to R. Thus every
integral domain embeds naturally in its quotient field.
Example: For any integral domain R, the polynomial ring R[X] is also an integral domain by Theo-
rem 14.2. The quotient field of R[X], denoted R(X), is the set of all rational functions f (X)/g(X),
where f (X), g(X) ∈ R[X] with g(X) = 0.
 II. RINGS 57

14. Let F be a field, and define a (formal) power series in X with coefficients in F , to be an expression
of the form a0 + a1 X + a2 X 2 + · · · where ai ∈ F for all i ≥ 0. The set of all such power series forms
a ring, denoted F [[X]], having F [X] as a subring. Like polynomials, these are strictly formal objects,
not functions (and so questions of convergence need not arise). More generally, a (formal) Laurent
series in X with coefficients in F , is an expression of the form ak X k + ak+1 X k+1 + ak+2 X k+2 + · · ·
where k ∈ Z and ai ∈ F for all i ≥ k. The set of all such Laurent series forms a ring, denoted
F ((X)).
(a) Show that F ((X)) is a field, and that it is in fact (up to isomorphism) the quotient field of F [[X]];
see Exercise 14.13. Thus we have the following subrings of F ((X)):
F ((X))
.. ..
..... .....
..... .....
..... .....
..
......
.
.....
.....
.
F (X)
..
F [[X]] .
..... .....
..... .....
..... .....
..... ....
..... ..
.
..
F [X]
For example, we have the rational function X/(1 + X) = X − X 2 + X 3 − X 4 + · · · ∈ F (X) .
(b) Let F = F2 and consider the rational function f (X) = (1 + X + X 3 )/(X 2 + X 3 + X 4 ) ∈ F2 (X).
Expand f (X) in its Laurent series, showing all terms up to and including the X 5 term.

15. Ideals and Quotients

Let R be a ring. An ideal of R is a subring which is invariant under left- and right-
multiplication by elements of R. That is, a subset A ⊆ R is an ideal iff
(i) A = Ø;
(ii) a − b ∈ A for all a, b ∈ R, and
(iii) ra, ar ∈ A for all a ∈ A and r ∈ R.
For example, every ring is an ideal of itself, and every ring has the trivial ideal {0}. The
ring Z of integers has ideals of the form mZ = {ma : a ∈ Z} for each m ∈ Z.
An ideal A of a ring R is proper if A ⊆.. R. It is clear (but worthy of explicit mention)

that if R has unity 1, then no proper ideal of R contains 1 (or for that matter, any units).
Given two ideals A, B ⊆ R, we define their sum as

A + B = {a + b : a ∈ A, b ∈ B}

and their product AB as the set of all sums a1 b1 + a2 b2 + · · · + ak bk where all ai ∈ A and
bi ∈ B.

15.1 Proposition. (i) If A and B are ideals of a ring R, then so are AB, A + B and
A ∩ B.

(ii) If {Aα }α is a nonempty collection of ideals of R, then their intersection α Aα is
an ideal of R.

Warning: To define AB, in general it does not suffice to take just products ab with a ∈ A
and b ∈ B, for this does not always give an ideal. For example consider R = Z[X], which
58 II. RINGS

has ideals A = {f (X) ∈ Z[X] : f (0) is even} and B = {f (X) ∈ Z[X] : f (1) is even}. Now
AB contains X(X + 1) + (2 − X)(X − 1) = 4X − 2, which is not of the form f (X)g(X)
with f (X) ∈ A and g(X) ∈ B.

Proof of Proposition 15.1. Suppose that A and B are ideals of R, and let x = i ai bi ∈
    
AB, y = j aj bj ∈ AB. Then x − y = i ai bi + j (−aj )bj ∈ AB by definition. If also

r ∈ R, then rx = (rai )bi ∈ AB since every rai ∈ A; and similarly, xr ∈ AB. This
proves that AB is an ideal of R. The proof that A + B is an ideal, is left as an exercise.
The fact that A ∩ B is an ideal is a special case of (ii), whose proof follows.

Let {Aα }α be a nonempty collection of ideals of R, and let I = α Aα . Since 0
belongs to each Aα , we have 0 ∈ I, and so I = Ø. If x, y ∈ I then x and y belong to

each Aα , so x − y belongs to each Aα , so x − y ∈ α Aα = I. Similarly if r ∈ R and
x ∈ I , then rx, xr ∈ I.

For example, given two ideals mZ and nZ in Z, the sum, product and intersection are
given by mZ + nZ = gcd(m, n)Z, (mZ)(nZ) = (mn)Z, and mZ ∩ nZ = lcm(m, n)Z; see
Exercise 15.2.
You should immediately observe the strong analogy between group theory and ring
theory results: Many statements of ring theory results are obtained from group theory
results simply by replacing the word ‘group’ by ’ring’; replacing ‘subgroup’ by ‘subring’;
and replacing ‘normal subgroup’ by ‘ideal’. This should help considerably in absorbing
these new results! Already you can see this pattern arising in Proposition 15.1, since
the class of normal subgroups of a given group G is similarly closed under product and
intersection (see Exercise 5.1).
Just as we formed quotients of groups modulo normal subgroups, we form quotients
of rings modulo ideals. We have already encountered an example of this: the quotient ring
Z/nZ. More generally, given a ring R and an ideal A ⊆ R, the quotient ring, denoted
R/A, is the set of all cosets r + A = {r + a : a ∈ A}, for r ∈ R. (Because addition is
commutative, it does not matter whether we write left cosets or right cosets.) Addition
and multiplication of cosets is defined by

(r + A) + (s + A) = (r + s) + A, (r + A)(s + A) = rs + A.

Why are these operations well-defined? It is easy to see that addition of cosets is well-
defined. (These are, after all, cosets of an additive subgroup of the abelian additive group
of R, so this statement has already been proved.) But for the multiplication of cosets to
be well-defined, we really need to use the fact that A is an ideal of R, not just an arbitrary
subring. Suppose that r + A = r  + A and s + A = s + A. Then r + 0 = r  + a for
some a ∈ A, which says that r − r  ∈ A. Similarly, s − s ∈ A. Now we have ‘defined’
(r + A)(s + A) as rs + A, which equals r  s + (r − r  )s + r(s − s ) + A = r  s + A since
(r − r  )s ∈ A and r(s − s ) ∈ A.
Are we justified in calling the set of cosets of A a quotient ring? Yes:
 II. RINGS 59

15.2 Theorem. If A is an ideal of R, then R/A is a ring with respect to the addition
and multiplication defined above.

Proof. Let r, s, t ∈ R. Then

(r + A) + (s + A) = (r + s) + A = (s + r) + A = (s + A) + (r + A),
(r + A) + A = (r + A) + (0 + A) = (r + 0) + A = r + A,
(r + A) + ((−r) + A) = (r + (−r)) + A = 0 + A = A,

which shows that R/A is an additive abelian group with identity A = 0 + A. Also
   
(r + A) (s + A) + (t + A) = (r + A) (s + t) + A
= r(s + t) + A
= (rs + rt) + A
= (rs + A) + (rt + A)
= (r + A)(s + A) + (r + A)(t + A),

which proves one of the distributive laws. The other distributive law, and the associativity
of multiplication, are proved similarly.

Let R and S be rings. A homomorphism from R to S is a function φ : R → S such

that
(i) φ(x + y) = φ(x) + φ(y), and
(ii) φ(xy) = φ(x)φ(y)
for all x, y ∈ R. (We say that φ preserves both addition and multiplication.) The kernel
and image of φ are

ker φ = {r ∈ R : φ(r) = 0}, φ(R) = {φ(r) : r ∈ R}.

Given an ideal A ⊆ R, the canonical homomorphism from R to R/A is defined by
π : R → R/A, r → r + A.
It is easy to check that π is in fact a homomorphism, and that ker π = A, which shows
that every ideal is the kernel of some homomorphism. Conversely, the kernel of any ring
homomorphism is an ideal; this is part of Theorem 15.3 below. Note that these statements
are the ring-theoretic analogues of statements from Section 6.
Of course, a bijective homomorphism is the same thing as an isomorphism. We now
translate the three isomorphism theorems for groups (Theorem 6.1 and Exercises 6.7,8)
into the language of ring theory.
60 II. RINGS

15.3 First Isomorphism Theorem. If φ : R → S is a homomorphism of rings,

then ker φ is an ideal of R, and R/ ker φ ∼
= φ(R).

Proof. Let K = ker φ. We have φ(0) = φ(0 + 0) = φ(0) + φ(0), which implies that
φ(0) = 0. This shows that 0 ∈ K and so K = Ø. If x, y ∈ K then φ(x−y) = φ(x)−φ(y) =
0 − 0 = 0, and so x − y ∈ K. If x ∈ K and r ∈ R then φ(rx) = φ(r)φ(x) = φ(r)0 = 0
and so rx ∈ K; similarly, xr ∈ K. This proves that K is an ideal of R.
Define φ : R/K → φ(R) by φ(r + K) = φ(r). First we check that φ is well-
defined: If r + K = r  + K then r − r  ∈ K, which says that φ(r − r  ) = 0, and so
φ(r + K) = φ(r) = φ(r  ) = φ(r  + K). Thus φ is well-defined.
 
Let r, s ∈ R. Then φ (r + K)(s + K) = φ(rs + K) = φ(rs) = φ(r)φ(s) = φ(r +
 
K)φ(s + K). Similarly, φ (r + K) + (s + K) = φ(r + K) + φ(s + K), and so φ is a
homomorphism.
Clearly φ is onto φ(R), since each element of φ(R) has the form φ(r) = φ(r + K) for
some r ∈ R. Also, if φ(r + K) = 0 then φ(r) = 0, whence r ∈ ker φ = K and r + K = K,
so that ker φ contains only the identity K ∈ R/K, i.e. φ is one-to-one. Thus φ is an
isomorphism.

The proofs of the following two theorems are left as exercises.

15.4 Second Isomorphism Theorem. Let R be a ring with ideal A and subring
S. Then A ∩ S is an ideal of S, and S/(A ∩ S) ∼
= (S + A)/A.

15.5 Third Isomorphism Theorem.  Let R be a ring with ideals A ⊆ B. Then

B/A is an ideal of R/A, and (R/A) (B/A) ∼
= R/B.

Given a subset X ⊆ R, we speak of the ideal generated by X, denoted by (X); this is

the unique smallest ideal of R containing X. This ideal exists by Proposition 15.1(ii); it
is the intersection of all ideals of R which contain X. Usually X is a finite subset, and if
X = {x1 , x2 , . . . , xk }, we alternatively write (X) = (x1 , x2 , . . . , xk ). In case X = {x} is a
singleton subset, this gives an ideal (x) generated by the single element x ∈ R, known as
a principal ideal. If R is a commutative ring with unity, then (x) = Rx = {rx : r ∈ R}
by Exercise 15.1. As an example, the ideal of Z generated by integers m1 , m2 , . . . , mk is
the same thing as the principal ideal generated by m = gcd(m1 , m2 , . . . , mk ), so we have
(m1 , m2 , . . . , mk ) = (m) = mZ.
The familiar ‘modulus’ notation for integers extends to arbitrary rings with ideals.
Every ideal A of a ring R determines an equivalence relation on the elements of R: we say
that x ≡ y mod A if x − y ∈ A. Note that the equivalence classes of this relation are just
the elements of R/A. Also observe that for integers, we have x ≡ y mod (m) iff x ≡ y
mod m (in the usual notation).
 II. RINGS 61

Every ideal of Z is principal. For suppose that A = (0) is an ideal of Z. Since A

contains a nonzero integer, and since A is closed under multiplication by −1 ∈ Z, there
must be a positive integer in A. Let m be the smallest positive integer in A. By definition
of an ideal, (m) = mZ ⊆ A. This must be equality, for if n ∈ A, then we may write
n = qm + r for some integers q, r with 0 ≤ r < m; now 0 ≤ r = n − qm ∈ A, and since
m is the smallest positive element of A, we obtain r = 0, i.e. n = qm ∈ (m).
Define a principal ideal ring to be a ring in which every ideal is principal. Also a
principal ideal domain (or P.I.D.) is a principal ideal ring which is an integral domain.
We have just proved:

15.6 Proposition. Z is a principal ideal domain.

We wish to generalize Proposition 15.6 to a larger class of rings which includes polynomial
rings F [X] where F is an arbitrary field. Define a Euclidean ring as a commutative ring
R = 0 together with a function deg : R → {−∞, 0, 1, 2, 3, . . .} such that for all a, b ∈ R,
(i) deg(a) = −∞ iff a = 0;
(ii) deg(a) ≤ deg(ab); and
(iii) if b = 0, then there exist q, r ∈ R such that a = qb + r and deg(r) < deg(b).
Property (iii) (and the usual algorithm for computing the quotient q and remainder r)
are together known as the Division Algorithm for R. A Euclidean ring which is an
integral domain is called a Euclidean domain. For example, Z is a Euclidean domain
with deg(a) = |a| for nonzero integers a; and deg(0) = −∞. Also if F is any field, then
the polynomial ring F [X] is a Euclidean domain with the usual degree function. However,
Z[X] is not a Euclidean domain; the usual degree function for Z[X] does not satisfy the
Division Algorithm (iii). Of course this is not enough to prove that Z[X] is not a Euclidean
ring, since it is still conceivable that a different degree function might work; see however
Exercise 15.4.

15.7 Theorem. Every Euclidean ring is a principal ideal ring with unity.

15.8 Corollary. If F is a field, then F [X] is a principal ideal domain.

Proof of Theorem 15.7. Let A be an ideal of a Euclidean ring R; we must show that A
is principal. If A = (0) then we are done. So assume that A has a nonzero element, and
choose a nonzero element a ∈ A for which deg(a) is minimal. Clearly Ra ⊆ (a) ⊆ A. Now
let x ∈ A. We may write x = qa + r for some q, r ∈ R such that deg(r) < deg(a). Then
r = x − qa ∈ A, so by minimality of the degree of a, we must have r = 0, which says
that x = qa ∈ Ra. Now A ⊆ Ra ⊆ (a) ⊆ A, so equality must hold: A = Ra = (a). In
particular, R is a principal ideal ring.
Since R is an ideal of itself, we have R = Ru = (u) for some nonzero u ∈ R. In
particular, we have u = eu = ue for some e ∈ R. For every x ∈ R, we have x = bu for
some b ∈ R, and so xe = bue = bu = x, so e is a multiplicative identity for R. Finally,
e = 0 since u = 0, so R is a ring with unity.
62 II. RINGS

Exercises 15.
1. (a) Let R be a commutative ring, and let a ∈ R. Show that (a) = Za + Ra, i.e. (a) = {ka + ra : k ∈
Z, r ∈ R}.
(b) Conclude that if R is a commutative ring with unity, then (a) = Ra.
2. Let R be an integral domain. Given a, b ∈ R, we say a divides b (or a is a divisor of b, denoted a | b)
if b = ca for some c ∈ R. Prove that the following three conditions are equivalent:
(i) a | b and b | a;
(ii) a = ub for some unit u ∈ R× ;
(iii) (a) = (b).
Elements a and b having the above relation are called associates. This relation is an equivalence
relation on R, as is clear from (iii). For example, the associates of 10 ∈ Z are 10 and −10.
3. Let R be a Euclidean domain, and let a, b ∈ R. An element g ∈ R is a common divisor of a and
b if g divides both a and b. We say g is a gcd (greatest common divisor) of a and b if (i) g is a
common divisor of a and b, and (ii) deg g ≥ deg h for every common divisor h of a and b. In general,
the gcd is not unique.
(a) Prove that if a, b ∈ R are not both zero, then a and b have a gcd g; that the associates of g are all
the gcd’s of a and b; and that (a) + (b) = (g). (See Exercise 15.2.)
Hint: Since R is a Euclidean domain, (a) + (b) = (g) for some g ∈ R. Prove that g has the
required properties.
(b) In the notation of (a), conclude that g = as + bt for some s, t ∈ R. This fact (and the well-known
algorithm for determining such s, t ∈ R, generalizing the example of Section 0) are together known
as Euclid’s Algorithm or the Euclidean Algorithm.
(c) With a, b, g as above, prove that the common divisors of a and b are precisely the divisors of g.
4. Show that Z[X] is not a P.I.D., and hence by Theorem 15.7, Z[X] is not a Euclidean ring.
5. Prove the Second Isomorphism Theorem 15.4.
Hint: It may help to imitate Exercise 6.7.
6. Prove the Third Isomorphism Theorem 15.5.
Hint: It may help to imitate Exercise 6.8.
7. Let F be a field. We say that a constant a ∈ F is a zero of a polynomial f (X) ∈ F [X] if f (a) = 0.
Given a ∈ F , show that a is a zero of f (X) iff f (X) = (X − a)q(X) for some q(X) ∈ F [X].
Hint: In order to show that f (X) is divisible by X − a, first divide f (X) by X − a and obtain a
quotient and remainder using the Division Algorithm.
Remark: More generally, a zero of f (X) ∈ F [X] may refer to an element a in an extension field
E ⊇ F such that f (a) = 0.
8. Let f (X) ∈ F [X] be a polynomial of degree n where F is a field, and let E ⊇ F be an extension
field. Show that f (X) has at most n zeroes in E.
Hint: Use Exercise 15.7.
9. Let f (X) ∈ F [X] where F is an infinite field. Show that f (X) represents the zero function F → F iff
f (X) = 0. This means that the only polynomial vanishing at every field element is the zero polynomial.
10. Let F b e a field, and let G be a finite subgroup of the multiplicative group of F × = F ...... {0}. Prove
that G is cyclic.
Hint: By Exercise 6.15, G ∼ = Cn1 × Cn2 × · · · × Cnr for some positive integers n1 , n2 , . . . , nr . If
gcd(ni , nj ) = 1 for some i = j, then choose a prime p dividing both ni and nj and count the number
of zeroes of X p − 1 ∈ F [X] to obtain a contradiction. Then use Exercise 3.3.
 II. RINGS 63

16. Maximal and Prime Ideals

An ideal P of a ring R is prime if

(i) P ⊆.. R, and

(ii) for any two ideals A, B ⊆ R, if AB ⊆ P then either A ⊆ P or B ⊆ P .

The prime ideals of Z are (0) and (p) for p prime (in the usual sense of ‘prime’). This
follows easily from Proposition 15.6, and the fact that (m)(n) = (mn) for ideals of Z.
A useful characterization of prime ideals is the following

16.1 Proposition. Let P be an ideal of a ring R.

(i) Suppose that ab ∈ P implies that either a ∈ P or b ∈ P . Then P is prime.
(ii) If R is commutative, then the converse of (i) holds.

Proof. (i) Suppose that A and B are ideals of R such that AB ⊆ P with A ⊆ P . Choose
a ∈ A ........ P . For every b ∈ B we have ab ∈ AB ⊆ P , so that b ∈ P ; hence B ⊆ P . Thus P
is prime.
(ii) Suppose that R is commutative with a prime ideal P ⊂ R, and that ab ∈ P where
a, b ∈ R. By Exercise 15.1, (a) = Za + Ra = {na + ra : n ∈ Z, r ∈ R}. Similarly,
(b) = Zb + Rb. From this it is not hard to see that (a)(b) ⊆ (ab) ⊆ P . Thus either (a) ⊆ P
(in which case a ∈ P ), or (b) ⊆ P (in which case b ∈ P ).

An ideal M of a ring R is maximal if

(i) M ⊆.. R, and
(ii) there is no ideal A such that M ⊆ ⊆
.. A .. R.

16.2 Proposition. Let R be a ring with unity. Then every maximal ideal of R is
prime.

Proof. Let M be a maximal ideal of R, and suppose that AB ⊆ M , A ⊆ M , B ⊆ M for

certain ideals A, B of R. Then M ⊆
.. A + M , so by maximality of M , we have A + M = R,

and similarly B + M = R. Thus 1 = a + m = b + m for some a ∈ A, b ∈ B and

m, m ∈ M . Now 1 = (a + m)(b + m ) = ab + am + mb + mm ∈ M , contradicting
M⊆ .. R.

Every maximal ideal of Z is of the form (p) for p prime. So every nonzero prime ideal
of Z is maximal. It is not generally true, however, that nonzero prime ideals are necessarily
maximal, even for commutative rings with identity. For example, in the ring Z ⊕ Z with
componentwise addition and multiplication, the nonzero ideal Z ⊕ 0 = {(x, 0) : x ∈ Z} is
64 II. RINGS

prime but not maximal, since it lies inside the larger proper ideal Z⊕2Z, which is maximal.
The most important characterization of prime and maximal ideals is

16.3 Theorem. Let R be a commutative ring with unity, and let A ⊆ R be an ideal.
Then
(i) A is prime iff R/A is an integral domain.
(ii) A is maximal iff R/A is a field.

Proof. (i) Suppose that A is a prime ideal of R. If (x + A)(y + A) = 0 + A = A, then

by definition xy + A = A, so either x ∈ A or y ∈ A, i.e. either x + A = A or y + A = A.
Thus R/A has no zero divisors. Since R has an identity 1, an identity for R/A is 1 + A;
and since R is commutative, so is R/A. Thus R/A is an integral domain. The converse is
just as easy.
(ii) Suppose that A is a maximal ideal of R. By Proposition 16.2, A is a prime ideal, so
R/A is an integral domain. To show that R/A is in fact a field, we must show that every
nonzero element x + A ∈ R/A is invertible. Since x + A is a nonzero element of R/A, we
have x ∈ / A, so by maximality of A, we have (x) + A = R. This means that 1 = rx + a
for some r ∈ R, a ∈ A. But then (r + A)(x + A) = 1 + A, so the inverse of x + A ∈ R/A
is the element r + A ∈ R/A.

For example, for each prime p, the ideal (p) = pZ is maximal, and Z/pZ = {0, 1, 2, . . . ,
p−1} is the finite field Fp . In the next section, we will see more examples of fields con-
structed as quotient rings.
Exercises 16.
1. Give an example of a ring R ⊇ Z such that R has a unique maximal ideal M = 0. Justify your answer.

2. Let a < b be real numbers. For each c ∈ [a, b], define Mc to be the set of all f ∈ C([a, b]) such that
f (c) = 0.
(i) Prove that Mc is a maximal ideal of C([a, b]).
Hint: Consider the map C([a, b]) → R, f → f (c).
(ii) Is every maximal ideal of C([a, b]) of the form Mc for some c ∈ [a, b]? Justify your answer.

3. Use Zorn’s Lemma (see the Appendix) to prove that every proper ideal of a ring R is contained in a
maximal ideal.

17. Irreducibility

Let R be a commutative ring with unity. An element f ∈ R is irreducible in R if

(i) f is not a unit of R, and
(ii) whenever f = gh for two elements g, h ∈ R, at least one of g, h is a unit of R.
 II. RINGS 65

Note that an irreducible element is always nonzero. Also observe that the property of
irreducibility is always relative to a particular choice of ring R; for example X 2 + 1 is
irreducible in the polynomial ring Z[X] (also in Q[X] and in R[X]), but is reducible in
C[X] since X 2 + 1 = (X + i)(X − i). Also, 2X + 2 = 2(X + 1) is reducible in Z[X],
but irreducible in Q[X] (also in R[X] and in C[X]). We are interested in irreducibility
primarily in the case of polynomial rings.
Note that the irreducible elements of the ring Z are simply the numbers ±p where
p ∈ Z is an ordinary prime. The fact that an ordinary prime p ∈ Z is irreducible in Z
is actually by definition; to conclude from this that the ideal (p) ⊂ Z is prime requires
Euclid’s Lemma or something more general, such as Theorem 17.2 below.

17.1 Theorem. Let R be an integral domain, and let 0 = f ∈ R. If the ideal (f ) is

prime, then f is irreducible in R.

Proof. Suppose that the ideal (f ) ⊆ R is prime, and that f = gh for some g, h ∈ R. By
Proposition 16.1(ii), at least one of g, h lies in (f ). We may suppose that g = qf for some
q ∈ R. Then (1 − qh)f = 0, and since R is an integral domain, this implies that qh = 1.
This means that h is a unit of R, and so f is irreducible in R.

17.2 Theorem. Let R be a P.I.D., and let 0 = f ∈ R. Then the following three
statements are equivalent.
(i) The element f ∈ R is irreducible.
(ii) The ideal (f ) ⊆ R is maximal.
(iii) The ideal (f ) ⊆ R is prime.

Proof. First suppose (i) holds, so that f ∈ R is irreducible. We will show that (ii) holds.
By definition, f ∈ / R× , so it generates a proper ideal (f ) ⊆
.. R. Suppose that (f ) is properly

contained in an ideal (b) ⊆ R. (There is no loss of generality in calling this ideal (b) for
some b ∈ R since R is a P.I.D.). Now f ∈ (b) implies that f = qb for some q ∈ R. Clearly
q is not a unit, since (f ) ⊆ .. (b). Since f is irreducible in R, we must have that b is a

unit of R, so (b) = R is the unique ideal properly containing (f ). This proves that (f ) is
maximal.
If (ii) holds, then (iii) follows by Proposition 16.2.
Finally, suppose that (iii) holds, i.e. the ideal (f ) ⊂ R is prime. We must verify (i).
Since the ideal (f ) ⊂ R is proper, f is not a unit of R. Suppose that f = gh for some
g, h ∈ R. By Proposition 16.1, at least one of g, h lies in (f ); without loss of generality,
assume g = qf for some q ∈ R. Thus (1 − qh)f = 0 so h is a unit, proving (i).
66 II. RINGS

As an example, consider R = Z[X]. The ideal (X) ⊂ R is prime, with quotient

Z[X]/(X) ∼= Z. This isomorphism follows from the First Isomorphism Theorem using the
homomorphism Z[X] → Z, f (X) → f (0). The ideal (X, 2) ⊂ R is maximal; it is the kernel
of the homomorphism Z[X] → F2 , f (X) → f (0) mod 2. Again, the First Isomorphism
Theorem gives Z[X]/(X, 2) ∼ = F2 .
We are interested in irreducible polynomials largely as a means of producing extension
fields:

17.3 Theorem. Let F , and let f (X) ∈ F [X] be irreducible in F [X]. Let n =
deg f (X). Then
(i) E = F [X]/(f (X)) is a field.
(ii) E contains a subfield {a + (f (X)) : a ∈ F } naturally isomorphic to F (and
therefore we identify this subfield with F ).
(iii) [E : F ] = n.
(iv) f (X) has a zero α = X + (f (X)) ∈ E, and E = F [α] = F (α).

Proof. (i) follows from Theorems 16.3 and 17.2.

(ii) Let π : F [X] → F [X]/(f (X)) denote the canonical homomorphism g(X) → g(X) +
(f (X)). Since n ≥ 1, the restriction of π to F is one-to-one, thereby giving an isomorphism
F → π(F ) = {a + (f (X)) : a ∈ F }.
(iii) Using the Division Algorithm, every element of E looks like a0 + a1 X + a2 X 2 +
· · · + an−1 X n−1 + (f (X)) for some a0 , a1 , . . . , an−1 ∈ F . This shows that the elements
π(X i ) = X i + (f (X)) for i = 0, 1, 2, . . . , n−1 span E over F . Moreover this set is
linearly independent, for if a0 π(1) + a1 π(X) + a2 π(X 2 ) + · · · + an−1 π(X n−1 ) = (f (X)),
then a0 + a1 X + a2 X 2 + · · · + an−1 X n−1 ∈ (f (X)), and comparing degrees, this gives
a0 = a1 = · · · = an−1 = 0. So E is an n-dimensional vector space over π(F ) ∼ = F.
(iv) We have f (α) = f (X) + (f (X)) = (f (X)), which is the zero element of E. Also
every element of E is of the form g(X) + (f (X)) = g(α) for some g(X) ∈ F [X], so that
E ⊆ F [α] ⊆ F (α) ⊆ E.

For example, suppose F is a field and d ∈ F has√ no square root in F . Then X 2 −d ∈ F [X] is
irreducible in F , and E = F [X]/(X 2−d) ∼ =√ F [ d] is a quadratic extension of E. We have
E = {a 2 ∼ 2
√ + bX + (X −d) : a, b ∈ F } = {a + b d : a, b ∈ F }. The map a + bX + (X −d) →
∼
a + b d√ is an isomorphism in this case. Particular cases give C = R[i] = R[X]/(X +1) 2

and Q[ 5] ∼ = Q[X]/(X 2 −5).

In view of the utility of Theorem 17.3 for constructing extension fields, it is helpful to
have available quick methods for determining irreducibility of polynomials. Unfortunately,
in general it is difficult to prove irreducibility. However, some criteria do exist, especially
the following.
 II. RINGS 67

17.4 Theorem. Let f (X) ∈ Z[X]. If f (X) is irreducible in Z[X], then f (X) is
irreducible in Q[X].

Proof. Suppose that f (X) is reducible in Q[X]. Then by factoring out all integer common
factors, we have f (X) = ab g(X)h(X) for some fraction ab ∈ Q in lowest terms, and some
g(X), h(X) ∈ Z[X] such that the coefficients of g(X) have no common integer divisor
greater than 1, and similarly for h(X). We must show that b = ±1. If not, then we may
choose some prime p dividing b. Reducing the polynomial equation bf (X) = ag(X)h(X)
modulo p, we obtain g(X)h(X) = 0, where the polynomials g(X), h(X) ∈ Fp [X] are
obtained by reducing the coefficients of g(X), h(X) ∈ Z[X] modulo p. By construction,
both g(X) and h(X) are nonzero polynomials, and this violates the fact that Fp [X] has
no zero divisors (see Theorem 14.2).

As an example, the polynomial X 3 + X + 2 is irreducible in Q[X]. For otherwise, by

Theorem 17.3, it is irreducible in Z[X], so that X 3 + X + 2 = (aX + b)(cX 2 + dX + e) for
some a, b, c, d, e ∈ Z; but then a ∈ {−1, 1} and b ∈ {−2, −1, 1, 2}, so that either X − 2
or X + 2 divides X 3 + X + 2, a contradiction. So by Theorem 17.3, we have a cubic
extension of Q given by Q(X)/(X 3 + X + 2).
Checking irreducibility of f (X) ∈ F [X] is usually easier when F is a finite field, since
in this case, there are only finitely many polynomials to check as possible factors of f (X).
For example, we easily see that the polynomial f (X) = X 3 + X + 1 ∈ F2 [X] is irreducible
in F2 [X], since the only polynomials of degree 1 are X and X + 1, neither of which divides
f (X). So we have a cubic extension F8 = F2 [X]/(f (X)) ⊆ F2 . The elements a + bX +
cX 2 +(f (X)) ∈ E may be abbreviated ‘abc’, so the eight elements of E are ‘000’ = (f (X)),
‘001’ = X 2 + (f (X)), ‘010’ = X + (f (X)), . . . , ‘111’ = 1+X+X 2 + (f (X)). Addition of
elements of E is addition of the coordinates modulo 2; for example, ‘011’ + ‘110’ = ‘101’.
Multiplication is not too hard, e.g.

‘011’ · ‘110’ = (X+X 2 )(1+X) + (f (X)) = X+X 3 + (f (X)) = 1 + (f (X)) = ‘100’.

Another very useful irreducibility criterion is the following.

17.5 Eisenstein Irreducibility Criterion. Let f (X) = a0 + a1 X + · · · + an X n ∈

Z[X]. Suppose there exists a prime p such that

(i) p  a0 ,
(ii) p  a 1 , a2 , . . . , an , and
(iii) p2   an .
Then f (X) is irreducible in Q[X].
68 II. RINGS

Proof. We may suppose that the gcd of the coefficients in f (X) is 1; otherwise factor
out the gcd, and this does not affect irreducibility in Q[X]. We will show that f (X)
is irreducible in Z[X], and then by Theorem 17.4 the result will follow. Suppose on the
contrary that f (X) is reducible in Z[X]. Then there exist polynomials g(X), h(X) ∈ Z[X],
neither of which is a unit (i.e. g(X), h(X) ∈
/ {1, −1}), such that f (X) = g(X)h(X). And
neither of the factors g(X), h(X) is an integer constant, since the gcd of the coefficients in
f (X) is 1. So g(X) and h(X) are polynomials of degree ≥ 1, say

g(X) = b0 + b1 X + b2 X 2 + · · · + bk X k ,
h(X) = c0 + c1 X + c2 X 2 + · · · + cn−k X n−k

in Z[X], where 1 ≤ k ≤ n − 1.
Let denote the reduction modulo p, so that Z = Z/pZ = Fp . By hypothesis (ii)
we have f (X) = a0 ∈ Fp [X], a constant polynomial, and by (i), this is not the zero
polynomial, so deg(f (X)) = 0. Since f (X) = g(X)h(X), by Theorem 14.2 we conclude
that both the polynomials g(X), h(X) ∈ Fp [X] are constant (degree 0). In particular,
bk = cn−k = 0, i.e. p divides both bk and cn−k . Then p2 divides an = bk cn−k , contrary
to (iii).

Observe that

a0 +a1 X +· · ·+an X n = (b0 +b1 X +b2 X 2 +· · ·+bk X k )(c0 +c1 X +c2 X 2 +· · ·+cn−k X n−k )

if and only if

an + an−1 X + · · · + a0 X n = (bk + bk−1 X + · · · + b0 X k )(cn−k + cn−k−1 X + · · · + c0 X n−k ).

1
(This is easily verified by replacing X → X , then finding a common denominator.) There-
fore the irreducibility of a polynomial is unaffected by reversing its list of coefficients.
For this reason, the Eisenstein Criterion 17.5 is sometimes stated with the alternative
hypotheses

(i ) p  an ,
(ii ) p  a 0 , a1 , . . . , an−1 , and
(iii ) p2   a0 .
At first sight, the hypotheses (i)–(iii) [or (i )–(iii )] might seem to be too restrictive to
be widely applicable. However, we use a trick: although a polynomial f (X) itself might
not satisfy the hypotheses (i)–(iii), it may nevertheless be possible to perform a change of
variable X → X + m for some integer m, to obtain a new polynomial which does satisfy
the criteria. As an example, consider f (X) = X 3 − 3X + 5 ∈ Z[X]. There is no prime
p satisfying conditions (i)–(iii) or (i )–(iii ) for f (X). However, substitute X = Y + 1 to
obtain f (X) = (Y +1)3 −3(Y +1)+5 = Y 3 +3Y 2 +3. The latter polynomial g(Y ) ∈ Z[Y ]
satisfies (i )–(iii ) for the prime 3, and so g(Y ) is irreducible in Q[Y ]. It clearly follows
 II. RINGS 69

that f (X) is irreducible in Q[X], since any factorization of f (X) in Z[X] gives, by change
of variable, a factorization of g(Y ) in Z[Y ]. This same trick is useful in proving

17.6 Theorem. Let p be a prime. Then the polynomial

Φp (X) = 1 + X + X 2 + · · · + X p−1

is irreducible in Q[X].

Proof. Note that (X − 1)Φp (X) = X p − 1. Substitute X = Y + 1 to obtain

p    p
1 p
p−1
(Y + 1)p − 1 j p−1
g(Y ) = Φp (Y + 1) = = Y =Y + Y j−1 .
Y Y j j
j=1 j=1
p
Observe that for 1 ≤ j ≤ p−1, the coefficient j = j!(p−j)! p!
is divisible by p. Also, the
constant term in g(Y ) is p, which is not divisible by p . Thus g(Y ) satisfies (i )–(iii ), so
2

g(Y ) is irreducible in Q[Y ]. This means that Φp (X) ∈ Z[X] is irreducible in Q[X].

The polynomial Φp (X) above is an important example of a cyclotomic polynomial,

which will be useful in Section 23. Exercise 17.9 gives a slight generalization of Theo-
rem 17.6.
Exercises 17.
1. Let f (X) = X 4 + X 2 + 1. Factor f (X) into irreducible factors
(a) in Q[X]; (b) in R[X]; (c) in C[X]; (d) in F2 [X];
(e) in F3 [X]; (f) in F5 [X]; (g) in F7 [X].
2. Let f (X) ∈ Q[X]. Show that if f (X 2 ) is irreducible in Q[X], then so is f (X). Does the converse hold?
Justify your answers.
3. Show that X 3 − 3X − 1 is irreducible in Q[X].
4. Construct a field of order 27.
5. Show that X 5 − X + 1 is irreducible in Q[X].
Hint: If X 5 − X + 1 is reducible in Q[X], then use Theorem 17.4 to obtain a factorization in Z[X].
Now reduce this factorization modulo 5, to obtain a factorization in F5 [X].
6. Show that for every prime p, there exists an irreducible polynomial f (X) ∈ Fp [X] of degree 2. Conclude
that for every prime p, there exists a field of order p2 .
7. Let p be prime. As in Exercise 17.6, prove that there exist irreducible polynomials in Fp [X] of degree
3 and 4. Conclude that there exist fields of order p3 and p4 .
8. Show that every quadratic extension of a given field F may be obtained by Theorem 17.3. In other
words, if [E : F ] = 2, show that E ∼
= F [X]/(f (X)) for some irreducible f (X) ∈ F [X] of degree 2.
9. Let q = pn be a power of a prime p, where n ≥ 1. Define
n
Xp − 1 n−1 n−1 n−1
Φq (X) = n−1
= 1 + Xp + X 2p + · · · + X (p−1)p ∈ Z[X].
Xp −1
Prove that Φq (X) is irreducible in Q[X].
Hint: Imitate the proof of Theorem 17.6.
 CHAPTER III

Fields
18. Algebraic Extensions

We begin with some standard definitions. Let F be a field. Since F contains 1, for every
positive integer it also contains an element 1 + 1 + · · · + 1 (n times), abbreviated simply as
n. However, it may happen that n = 1 + 1 + · · · + 1 = 0 in F , for some positive integer n.
If so, then the smallest positive integer n for which this happens must be prime; otherwise
by the distributive law,
n1 times n2 times n times

 
 
 
n = n1 n2 = (1 + 1 + · · · + 1) (1 + 1 + · · · + 1) = (1 + 1 + · · · + 1) = 0
where n1 , n2 < n, so that n1 = 0 or n2 = 0, contradicting the minimality of n. This
minimum n is called the characteristic of F . If there is no positive integer n such that
n = 0 in F , we say that F has characteristic zero. We denote the characteristic of
an arbitrary field F by char F . So either char F = p, a prime, in which case pa =
a + a + · · · + a = (1 + 1 + · · · + 1)a = 0 for all a ∈ F ; or char F = 0, in which case
na = a + a + · · · + a is never zero for any a ∈ F × and any positive integer n.
The unique smallest subfield of F is called the prime field of F . This is the subfield of
F generated by 1, and it is isomorphic to Fp ∼= Z/pZ if char F = p is prime; the prime field
is isomorphic to Q if char F = 0. In particular every finite field has prime characteristic,
since it cannot have an infinite subfield Q. We will devote more time, however, to studying
fields of characteristic zero.
First, a result showing the multiplicativity of degrees of extensions:

18.1 Theorem. For every tower K ⊇ E ⊇ F of fields, [K : F ] = [K : E][E : F ].

Proof. Let {α1 , α2 , . . . , αm } be a basis for K over E, and let {β1 , β2 , . . . , βn } be a basis
for E over F . Then it is easy to show (Exercise 18.1) that {αi βj : 1 ≤ i ≤ m, 1 ≤ j ≤ n} is
a basis for K over F , so that [K : F ] = mn = [K : E][E : F ].

Let E ⊇ F be any extension of fields. For each α ∈ E, we say that α is algebraic over
F if f (α) = 0 for some nonzero polynomial f (X) ∈ F [X] . If no such √
nonzero polynomial
exists, we say that α is transcendental over F . Thus, for example, 2 ∈ R is algebraic
72 III. FIELDS

over Q. Also π ∈ R is transcendental over Q, but algebraic over R, as it is a zero of the

polynomial X − π ∈ R[X]. If α ∈ E is algebraic over F , then {f (X) ∈ F [X] : f (α) = 0}
is a nonzero ideal in F [X], and since F [X] is a P.I.D., this ideal is principal, say
 
{f (X) ∈ F [X] : f (α) = 0} = f0 (X)

where f0 (X) ∈ F [X]. Moreover we may assume that f0 (X) is monic, i.e. that its
leading coefficient (the coefficient of the highest power of X) is 1, for otherwise we may
adjust f0 (X) by multiplying by the appropriate constant. Clearly f0 (X) is irreducible
in F [X]; otherwise one of the factors of f0 (X) would be a polynomial of degree less than
deg f0 (X) having α as a zero. Since f0 (X) is the unique lowest degree monic polynomial
in F [X] having α as a zero, we rename it as Irrα,F (X). The degree of α over F is by
definition the degree of Irrα,F (X). Thus, for example, Irr√5,Q (X) = X 2 − 5, whereas
√ √ √ √
Irr√5,E (X) = X − 5 if E = Q[ 5]; so 5 is algebraic of degree 2 over Q, but 5 is
√
algebraic of degree 1 over E = Q[ 5].

18.2 Theorem. Let E ⊇ F be an extension of fields, and let α ∈ E. Then the

following conditions are equivalent.
(i) α is algebraic over F .
(ii) [F (α) : F ] < ∞.
(iii) F (α) = F [α].

Proof. (i)⇒(iii) Suppose α is algebraic over F , so that f (α) = 0 where f (X) =

Irrα,F (X). We must show that F [α] is a field. Consider a nonzero element of F [α];
this has the form g(α) where g(X) ∈ F [X] is not divisible by f (X). Then by Euclid’s
Algorithm (see Exercise 15.3) there exist polynomials u(X), v(X) ∈ F [X] such that
u(X)f (X) + v(X)g(X) = 1.
Substituting for α in place of X gives v(α)g(α) = 1, i.e. g(α) is a unit of F [α].
(iii)⇒(i) We may assume that α = 0. If F (α) = F [α] then
α−1 = a0 + a1 α + a2 α2 + · · · + an αn
for some n ≥ 0, ai ∈ F . This means that
1 − a0 α − a1 α2 − · · · − an αn+1 = 0,
where the coefficients of the powers αi are not all zero (in fact the coefficient of α0 is 1),
so α is algebraic over F .
(ii)⇒(i) If [F (α) : F ] = n < ∞, then the set {1, α, α2 , . . . , αn } is linearly dependent
over F , and so there exist coefficients ai ∈ F , not all zero, such that
a0 + a1 α + · · · + an−1 αn−1 + an αn = 0,
i.e. α is algebraic over F .
 III. FIELDS 73

{(i),(iii)}⇒(ii) If α is algebraic over F , then f (α) = 0 where Irrα,F (X) is of degree n,

say. Each element of F (α) = F [α] is of the form g(α) for some g(X) ∈ F [X]. Since
g(X) = u(X)f (X) + r(X) for some u(X), r(X) ∈ F [X] with deg r(X) < n, we have
F (α) = F [α] = {r(α) : r(X) ∈ F [X], deg r(X) < n}.
It follows that {1, α, α2 , . . . , αn−1 } is a basis for F (α) = F [α] over F , so [F (α) : F ] =
n < ∞.

18.3 Corollary. Let E ⊇ F be an extension of fields, and let α ∈ E be algebraic

over F . Then the degree of α over F equals [F [α] : F ].

Proof. The key ideas in proving this result appear in the proof of Theorem 18.2; the
details are left as an exercise.

  E ⊇ F is a field extension such that E = F [α] for

18.4 Corollary. Suppose that
∼
some α ∈ E. Then E = F [X] Irrα,F (X) .

Proof. Clearly themap π : F [X] → E, g(X) → g(α) is a homomorphism. By definition,

ker π = Irrα,F (X) . Also the image is π(E) = F [α] = E, so the result follows from the
First Isomorphism Theorem 15.3.

A field extension E ⊇ F is called algebraic if every element of E is algebraic over F .

If [E : F ] < ∞ then the extension is algebraic, since for each α ∈ E we have E ⊇ F (α) ⊇
F and [E : F ] = [E : F (α)][F (α) : F ] < ∞, so that [F (α) : F ] < ∞, i.e. α is algebraic
over F . We restate this fact formally:

18.5 Corollary. Every finite field extension E ⊇ F is algebraic. In particular if

α ∈ E is algebraic over F , then the extension F (α) ⊇ F is algebraic.

Let E ⊇ F be an extension of fields. If α, β ∈ E are both algebraic over F , then so are

α + β and αβ. To see this, first note that [E : F ] < ∞ where E = F (α) = F [α]. Now β
is a zero of some nonzero polynomial f (X) ∈ F [X] ⊆ E[X], and so β is algebraic over E,
i.e.
[E(β) : F ] = [E(β) : E][E : F ] < ∞.
This means that E(β) = F (α, β) is an algebraic extension of F , and in particular the
elements α + β, αβ ∈ F (α, β) are algebraic over F . So define the algebraic closure of
F in E as
{α ∈ E : α is algebraic over F }.
This is a field by the preceding arguments, and so is an algebraic extension of F . (Of
course this is nothing other than E, if E is already algebraic over F . More generally, the
74 III. FIELDS

algebraic closure of F in E is the largest subfield of E containing F , which is algebraic

over F .)
A field is algebraically closed if, for every extension E ⊇ F , the algebraic closure
of F in E is F itself. For example, C is algebraically closed. Also, there exist algebraically
closed fields of every possible characteristic.
Warning: The terminology now starts to sound slightly ambiguous unless one is careful
to observe the prepositions! An algebraic closure of a field F is an extension F ⊇ F
such that (i) F is algebraically closed, and (ii) the extension F ⊇ F is algebraic. It may be
shown that any two algebraic closures of the same field F are isomorphic. If one accepts
Zorn’s Lemma (or, equivalently, the Axiom of Choice) then every field has an (unique)
algebraic closure. For example, the algebraic closure of R is C. It will be convenient for us
to assume the existence of algebraic closures, although this is not strictly necessary since
all our Galois theory can be accomplished using finite extensions.
Exercises 18.
1. Fill in the missing details in the proof of Theorem 18.1.

2. Let E ⊇ F be an extension of degree 3, and suppose α ∈ E but α ∈

/ F . Show that F [α] = E.

3. Recall that if A is a linear transformation V → V where V is an n-dimensional vector space over a

field F , and if g(X) ∈ F [X] is its characteristic polynomial, then g(A) = 0 (the zero matrix) by the
Cayley-Hamilton Theorem; and that g(X) is divisible by m(X) ∈ F [X], the minimal polynomial of
A over F , i.e. the (unique) minimum degree monic polynomial having A as a zero.
(a) Let F be a field, and let
f (X) = a0 + a1 X + a2 X 2 + · · · + an−1 X n−1 + X n ∈ F [X].

The companion matrix of f (X) is the n × n matrix

⎡ 0 0 · · · 0 −a ⎤
0
⎢ 1 0 · · · 0 −a1 ⎥
⎢ ⎥
⎢ ⎥
Af = ⎢ ⎢ 0 1 · · · 0 −a 2 ⎥ ∈ F n×n .
⎢ .. .. . . . . ⎥
⎣. . . .. .. ⎥⎦
0 0 ··· 1 −an−1
Find the characteristic polynomial of Af .
(b) Let E ⊇ F be an extension of degree n, and let α ∈ E. The map
E → E, x → αx
is a linear transformation Aα of E over F . Show that the minimal polynomial of Aα equals
Irrα,F (X). If E = F (α), show that the characteristic polynomial of Aα equals Irrα,F (X).

4. Prove Corollary 18.3.

5. Let F be a finite field. Show that |F | = pr for some prime p and some integer r ≥ 1.
Hint: Let p = char F . Then F is a vector space of finite dimension r, say, over its prime field Fp .
√ √ √ √
6. Since 2 and 3 are algebraic over Q, so is α = 2 + 3 by the comments following Corollary 18.5.
Determine Irrα,Q (X).

7. Use Zorn’s Lemma to show that every field has an algebraic closure.
 III. FIELDS 75

8. Assume the remarks at the end of Section 18 regarding existence of algebraic closures. Show that there
exist algebraically closed fields F which are proper extensions of C.

9. By Exercise 17.9, the polynomial Φ9 (X) = 1 + X 3 + X 6 is irreducible in Q[X]. Let ζ = e2πi/9 .

(a) Show that ζ is a zero of Φ9 (X). Conclude that E = Q(ζ) ⊃ Q is an extension of degree 6.
(b) Let α = ζ + ζ −1 = 2 cos( 2π
9
) ∈ R and let F = Q(α). Show that α is a proper subfield of E.
(c) Show that ζ is a zero of X 2 − αX + 1 ∈ F [X]. Conclude that [E : F ] = 2 and [F : Q] = 3.
(d) Determine Irrα,Q (X).

10. (Straightedge-and-Compass Constructions) Consider the points P0 = (0, 0) and P1 = (1, 0)

in the Euclidean plane R2 . Suppose that, for each n ≥ 2, the point Pn = (xn , yn ) ∈ R2 is con-
structible from {P0 , P1 , . . . , Pn−1 } by an elementary straightedge-and-compass construction, i.e. one of
the following holds:
(i) Pn is the intersection of two distinct straight lines, each joining two points of {P0 , P1 , . . . , Pn−1 };
(ii) Pn is a point of intersection of a straight line joining two points of {P0 , P1 , . . . , Pn−1 }, and a circle
with center Pi and radius Pj Pk , 0 ≤ i, j, k < n, j = k; or
(iii) Pn is a point of intersection of two distinct circles, each having as center a point in {P0 , P1 , . . . ,
Pn−1 }, and radius equal in length to a segment joining two of {P0 , P1 , . . . , Pn−1 }.
A point P ∈ R2 is constructible if it belongs to some sequence P0 , P1 , P2 , . . . of points, each point
obtainable by an elementary construction from the previous points. Let Fn = Q(x0 , y0 , x1 , y1 , . . . ,
xn , yn ), which is the smallest subfield of R containing the coordinates of P0 , P1 , . . . , Pn .
(a) Show that [Fn : Fn−1 ] = 1 or 2.
(b) Show that [Fn : Q] = 2t for some t ≤ n − 1.
√
(c) Show that the point P = ( 12 , 23 ) is constructible, and hence the angle π
3
(formed by P1 , P0 , P )
is “constructible” by straightedge and compass.
(d) Show that the point Q = (cos π9 , sin π9 ) is not constructible, and hence the angle π
3
may not be
trisected using straightedge and compass.
Hint: Use Exercise 18.9.
π π
(e) Is the point (cos 5
, sin 5
) constructible? Explain.

19. Normal Extensions

Let f (X) ∈ F [X]. We wish to find an extension E ⊇ F in which f (X) splits completely
into linear factors. Clearly it is sufficient to be able to do this when f (X) is irreducible in
F [X]. In this case, E = F [X]/(f (X)) has at least one zero α = X + (f (X)). However,
f (X) might not split into linear factors in E[X].
For example, X 3 − 2 is irreducible in Q[X], and F = Q[X]/(X 3 − 2) is a cubic
extension of Q in which f (X) has a zero α = X + (f (X)) ∈ F , and f (X) = (X − α)(X 2 +
αX + α2 ). But the latter quadratic factor is irreducible in F [X]. (This is not hard to
see. For if β ∈ F is a zero of X 2 + αX + α2 , then [Q[β] : Q] divides [F : Q] = 3. But
clearly β ∈/ Q, so the degree of β over Q is 2, whence [Q[β] : Q] = 2 by Corollary 18.3, a
contradiction.) So we take E = F [T ]/(T 2 + αT + α2 ), which is a quadratic extension of
F in which f (X) = (X − α)(X − ωα)(X − ω 2 α), where ω ∈ E is a zero of T 2 + T + 1.
Altogether we have [E : Q] = [E : F ][F : Q] = 2 · 3 = 6. This extension E is a splitting field
76 III. FIELDS

for the polynomial X 3 − 2 over Q. In general, an extension E ⊇ F is called a splitting

field for a polynomial f (X) ∈ F [X], if
(i) f (X) splits into linear factors in E[X], i.e. f (X) = (X − α1 )(X − α2 ) · · · (X − αn ),
where αi ∈ E; and
(ii) E = F (α1 , α2 , . . . , αn ).
The reason for the last restriction is that f (X) splits into linear factors in F (α1 , α2 , . . . , αn )
⊆ E, and we want E to be the smallest possible extension of F in which f (X) splits into
linear factors. For example, while X 3 − 2 splits into linear factors in C, we would not
call C a splitting field for X 3 − 2. However, the field E = Q(α, ωα, ω 2 α) = Q(α, ω) is a
splitting field for X 3 − 2.

19.1 Theorem. Any two splitting fields for the same polynomial f (X) ∈ F [X] are
isomorphic.

We shall not prove this fact, but it says for example that every splitting field for X 3 − 2
over Q, is isomorphic to the extension E = Q(α, ω) constructed above. Therefore we are
justified in calling E the splitting field for X 3 − 2 over Q, rather than merely a splitting
field for X 3 − 2 over Q.
In the example above, the extension F ⊃ Q has the somewhat unfortunate property
that X 3 − 2 has a zero in F , yet it does not split completely into linear factors in F [X].
The extension E ⊃ Q is nicer in this respect: it is a fact that every irreducible polynomial
in Q[X] having a zero in E, splits into linear factors in E[X]. (This fact follows from
Theorem 19.2 below, which we also state without proof.)
We say that an extension L ⊇ K is normal if every polynomial f (X) ∈ K[X] which
is irreducible in K[X] and has a zero in L, splits completely into linear factors in L[X].
The choice of terminology ‘normal’ for this property is directly related to the property of a
subgroup being normal, as we shall see later. The finite normal extensions are characterized
as the splitting fields of polynomials, thus:

19.2 Theorem. A finite extension L ⊇ K of fields is normal iff L is the splitting

field of some polynomial f (X) ∈ K[X] over K.

Note that the polynomial f (X) in Theorem 19.2 is not required to be irreducible; for
example √ √ √ √ √
Q[ 2, 5] = {a + b 2 + c 5 + 10 : a, b, c, d ∈ Q}
√ √ √
is a normal extension of Q of degree 4 (with basis {1, 2, 5, 10}), since it is the splitting
field of (X 2 − 2)(X 2 − 5) over Q.
Exercises 19.
1. Find an extension E ⊃ F of degree 4 which is not normal. Explain.
2. Give an example of normal extensions E ⊃ F and L ⊃ E such that the extension L ⊃ F is not
normal. Justify your answer. (This is just like the situation for groups; see Exercise 5.4).
 III. FIELDS 77

3. Prove that every quadratic extension of fields is normal. This is the analogue of which result in group
theory? (see Section 5).

4. If L ⊇ E ⊇ F are fields such that the extension L ⊇ F is normal, does it follow that the extension
L ⊇ E is normal? or that the extension E ⊇ F is normal? Explain.

5. Prove Theorem 19.2.

20. Separable Extensions

Another ‘unfortunate’ situation that may occur in an algebraic extension E ⊇ F , is that

if α ∈ E and f (X) = Irrα,F (X) is the minimal monic polynomial of α over F , then
f (X) may have repeated roots. This never happens in extensions of Q as we shall see, but
consider the field

f (T )
E = F2 (T ) = : f (T ), g(T ) ∈ F2 [T ], g(T ) = 0 ,
g(T )
i.e. E is the field consisting of all rational functions in a single indeterminate T , with
coefficients in the field F2 = {0, 1}; see Exercise 14.13. Now E has a subfield F = F2 (T 2 )
consisting of all rational functions of T 2 with coefficients in F2 , i.e. F is the quotient
field of F2 [T 2 ]. (Actually F ∼
= E, but that is irrelevant here; F is a proper subfield of
E.) The element T ∈ E is algebraic of degree 2 over F , since T ∈ / F is a zero of the
polynomial f (X) = X 2 − T 2 ∈ F [X]. Since E = F (T ), it follows from Corollary 18.5
that the extension E ⊃ F is algebraic of degree 2. (In fact every quadratic extension is
algebraic; see Exercise 19.2.) Note that f (X) = X 2 − T 2 = (X − T )2 is irreducible in
F [X], and has a double zero T ∈ E.
We wish to avoid such situations, so whenever E ⊇ F is an extension and α ∈ E is
algebraic over F , we say that α is separable over F if α is a simple zero of Irrα,F (X).

20.1 Theorem. Let E ⊇ F be an extension of fields of characteristic zero, i.e.

E ⊇ F ⊇ Q. Then every element α ∈ E which is algebraic over F , is separable
over F .

Proof. Let f (X) = Irrα,F (X) have degree n. Since f (X) is monic, its leading term is X n .
If (X − α)2 divides f (X) in E[X], then (X − α) divides f  (X) in E[X]. But the leading
term of f  (X) is nX n−1 = 0, and so α is a zero of the nonzero polynomial f  (X) ∈ F [X]
of degree n − 1 < deg f (X), contradicting f (X) = Irrα,F (X).

Note why it is that the proof above fails in the situation f (X) = X 2 − T 2 = (X − T )2 ∈
F2 [X]: in this case T is a zero of f  (X) alright, but f  (X) = 0 (the zero polynomial).
An extension E ⊇ F is called separable if every element of E is separable over F .
(This definition implicitly requires that the extension E ⊇ F is algebraic.) We have just
shown that every algebraic extension of a field of characteristic zero, is separable. It may
also be shown that if E ⊇ F are finite fields, then the extension E ⊇ F is separable.
78 III. FIELDS

Exercises 20.
1. Suppose that L ⊇ E ⊇ F is a tower of fields such that the extension L ⊇ F is separable. Prove that
the extensions L ⊇ E and E ⊇ F are separable.

21. Field Isomorphisms

Let F be any field. An automorphism of F is a ring isomorphism σ : F → F , i.e. a

bijection F → F such that (x + y)σ = xσ + y σ and (xy)σ = xσ y σ for all x, y ∈ F . The
set of all automorphisms of F is a group under composition, called the automorphism
group of F , denoted Aut F . For every σ ∈ Aut F , let Fσ = {x ∈ F : xσ = x}. It is clear
that Fσ contains both 0 and 1, and is closed under addition, subtraction, multiplication
and division by nonzero elements of Fσ ; hence Fσ is a subfield of F which contains the
prime field. We call Fσ the fixed field of σ. For example, the usual complex conjugation
τ : z → z is an automorphism of C, with fixed field Cτ = R.
Now suppose that E ⊇ F is an extension of fields. An automorphism of E which fixes
every element of F , is called an F -automorphism of E. The set of all F -automorphisms
of E is a subgroup of Aut E, denoted G(E/F ). For each σ ∈ G(E/F ), the fixed field Eσ
is an intermediate field, i.e. E ⊇ Eσ ⊇ F . More generally, for each subgroup H ≤ G, the
fixed field EH = {x ∈ E : xσ = x for all x ∈ H} is an intermediate field, E ⊇ EH ⊇ F .
Conversely, given any intermediate field L, i.e. E ⊇ L ⊇ F , we have a subgroup
G(E/L) ≤ G(E/F ). Our goal in Sections 21 and 22, is to develop the rudiments of Galois
theory, which is concerned with the special class of extensions E ⊇ F (namely, finite
normal separable extensions) for which this correspondence H ↔ EH between subgroups
of G(E/F ) and intermediate fields, is bijective.
Let E ⊇ F and E  ⊇ F be extensions of the same field F . An F -isomorphism
from E to E  is a field isomorphism E → E  (i.e. ring isomorphism of fields) which fixes
every element of F . An F -monomorphism from E to E  is a field monomorphism (i.e.
one-to-one ring homomorphism of fields) which fixes every element of F .
....
21.1 Theorem. Let E ⊇ F be a separable ex- E.. ..............
...
.......... C
. .
... ...
tension of degree n, and let C be an algebraically ... ...
... ....
.
... .
... ...
closed field containing F . Then there exist exactly n ... ...
... .....
.. ..
distinct F -monomorphisms from E into C. F

Proof. First consider the special case that E = F [α] for some α ∈ E. Let f (X) =
Irrα,X (X). Since C is algebraically closed, f (X) splits into linear factors in C[X], say
f (X) = (X − α1 )(X − α2 ) · · · (X − αn ) for some αi ∈ C. For each i, observe that
Irrαi ,F (X) = f (X) since f (X) is monic irreducible in F [X] and has αi as a zero.
For each i = 1, 2, . . . , n, define σi : F [α] → C by g(α) → g(αi ) where g(X) ∈ F [X].
Then σi is well-defined, since if g(α) = h(α), then g(X) ≡ h(X) mod (f (X)), in which
case g(αi) = h(αi ). Clearly σi : E → C is a ring homomorphism, fixing every element
 III. FIELDS 79

of F . Also σi is one-to-one, for if g(α)σi = g(αi ) = 0, then f (X) divides g(X), so that
g(α) = 0. So each σi : E → C is an F -monomorphism. The image of σi is the subfield
E σi = F [αi ] ⊆ C.
Now F [αi ] ∼ = F [α] = E is separable over F , so α1 , α2 , . . . , αn are distinct. Since
ασi = αi , the monomorphisms σ1 , σ2 , . . . , σn are distinct.
Finally, let σ be any F -monomorphism from E into C. Then f (ασ ) = f (α)σ = 0σ = 0,
so that ασ ∈ {α1 , α2 , . . . , αn }. Let us say that ασ = αi . Since the ring homomorphisms σ
and σi agree on F and on α, they must agree on F [α] = E, i.e. σ = σi . Thus σ1 , σ2 , . . . , σn
are the only F -monomorphisms from E into C.
Consider now the general case E ⊇ .. F , and let α ∈ E
........
F . We may assume that
F [α] ⊆.. E; otherwise we are done by the previous case. We have E .. F [α] .. F and
⊇ ⊇
n = mt where m = [E : F [α]] and t = [F [α] : F ]. By induction on the degree of
extension, there exist t distinct monomorphisms σ1 , σ2 , . . . , σt : F [α] → C. Let αi = ασi .
C..
.... ...
.. ...
....
.
...
...
... ...
.. ...
... ...
....
. ...
... ...
... ...
... θij ..
σi ............. σ .................
E.. .........................
.... E i
. ..
...................
E.. σi θij
... . ...
... ... ... ...
.... .. ... ..
... ... ... ...
... ....
. ... ....
.
... .. ... ..
... ... ... ...
... ... ... ...
... .
.... ...
... .
....
... .. ..
. ... .. ...
iσ .................
F [α] ...
...................... F [αi ]
... ....
.... ..
... .....
... ..
... ...
... ...
...
... .. ...
... ...
... ...
...

F
Since σi : E → E σi is an F -isomorphism, the extension E σi ⊇ F is separable; hence by
Exercise 20.1, the extension E σi ⊇ F [αi ] is separable. By induction on the degree of ex-
tension, for each i there exist m distinct F [αi ]-monomorphisms θi1 , θi2 , . . . , θim : E σi → C.
The composite maps σi θij : E → C (with left-to-right composition) constitute mt = n
distinct F -monomorphisms. To see that these are the only F -monomorphisms E → C,
suppose that σ : E → C is an F -monomorphism. As before, σ must take α to some αi .
Then σi−1 σ : E σi → C is an F [αi ]-monomorphism, so by induction, σi−1 σ = θij for some
j, whence σ = σi θij as required.

It is often useful to have a single generator for an extension field. The following result
guarantees that such a generator exists for all finite separable extensions. We present a
proof, however, only in the special case char F = 0. This case is slightly easier, and it is
the primary case we are interested in. For a proof in the general case, see e.g. Garling [1].
80 III. FIELDS

21.2 Theorem of the Primitive Element. Let E ⊇ F be a finite separable

extension of fields. Then E = F [α] for some α ∈ E.

Proof in the case char F = 0. Let C be an algebraically closed field containing F . By

Theorem 21.1, there exist distinct F -monomorphisms σ1 , σ2 , . . . , σn : E → C where n =
[E : F ].
We claim that there exists α ∈ E such that the images ασ1 , ασ2 , . . . , ασn ∈ C are
distinct. To see this, we apply Exercise 21.1 as follows. Whenever 1 ≤ i < j ≤ n, the set
Vij = {x ∈ E : xσi = xσj } is a proper subspace of the vector space E over F . Also
|F | = ∞ since char F = 0. Since E cannot be covered by finitely many proper subspaces


Vij , there exists α ∈ E........ 1≤i<j≤n Vij , and this α has the required property: ασi = ασj
whenever i = j.
Since [E : F ] < ∞, we have F (α) = F [α] by Corollary 18.5 and Theorem 18.2. So
we have a tower of extensions E ⊇ F [α] ⊇ F and n = [E : F [α]][F [α] : F ]. Since the
restrictions σ1 , . . . , σn : F [α] → C are distinct F -monomorphisms, we have n ≤ [F [α] : F ]
by Theorem 21.1. Therefore [F [α] : F ] = n and E = F [α].

Remark: The use of an algebraically closed extension C in the proof of Theorem 21.2 was
simply a convenient crutch, and was not really necessary. All that is really required is a
splitting field for f (X), which is a finite extension. This releases us from having to assume
the Axiom of Choice (see comments at the end of Section 18).
Exercises 21.
1. Let V be a vector space over an infinite field F , and let V1 , V2 , . . . , Vn be finitely many proper subspaces
of V . Prove that the union V1 ∪ V2 ∪ · · · ∪ Vn is a proper subset of V . (This problem was used in the
proof of Theorem 21.2. The proof is slightly tricky, but requires only elementary linear algebra.)
√ √
2. Show
√ that
√ Q[ 2, 3] ⊃ Q is a finite normal extension. By Theorem 21.2 there exists α such that
Q[ 2, 3] = Q[α]. Find an explicit choice for such an α. Justify your answer.
3. Let E ⊇ F be a separable extension of degree n, and let C be an algebraically closed field containing
F . By Theorem 21.1, there exist exactly n distinct F -monomorphisms σi : E → C. Prove that the
functions σ1 , σ2 , . . . , σn are linearly independent over F .
Hint: If not, consider a subset of {σ1 , . . . , σn } which is linearly dependent, but having no linearly
dependent proper subset. We may assume {σ1 , . . . , σk } is such a subset, where 1 ≤ k ≤ n. Now there
exist constants a1 , . . . , ak ∈ C, none of which are zero, such that a1 xσ1 + · · · + ak xσk = 0 for all
x ∈ E. Then k ≥ 2, and since σ1 = σ2 , there exists c ∈ E such that cσ1 = cσ2 . Replacing x by
cx in the relation above, we obtain another relation a1 cσ1 xσ1 + · · · + ak cσk xσk = 0 for all x ∈ E.
Eliminating xσ1 from these two relations, obtain a relation involving only xσ2 , . . . , xσk .
4. Let F be a field. Prove that every automorphism of F fixes every element of the prime subfield K ⊆ F .
5. Let F be a field of characteristic p. Show that the map σ : F → F defined by x → xp is an
Fp -monomorphism.
Hint: (x + y)p = xp + y p follows from the Binomial Theorem, since each of the binomial coefficients
p!
j!(p−j)!
is divisible by p for j = 1, 2, . . . , p−1. If xp = y p then (x − y)p = xp − y p = 0 by a similar
argument.
 III. FIELDS 81

22. Galois Extensions

The following gives an effective upper bound for the number of F -automorphisms of E for
any finite separable extension E ⊇ F .

22.1 Theorem. Let E ⊇ F be a finite separable extension. Then |G(E/F )| ≤ [E :

F ].

Proof. Let C be an algebraically closed field such that C ⊇ E ⊇ F . Then every element
of G(E/F ) is an F -monomorphism E → C, so the result follows by Theorem 21.1.

Again, the proofs of Theorem 22.1, and of Theorem 22.2 below, appeal to the exis-
tence of an algebraically closed field C only as a convenience; it is possible to prove all
these results using finite extensions. When does equality occur in the upper bound of
Theorem 22.1? This is answered by

22.2 Theorem. Let E ⊇ F be a finite separable extension of fields. Then the

following three conditions are equivalent.
(i) The extension E ⊇ F is normal.
(ii) |G(E/F )| = [E : F ].
(iii) F is the fixed field of G(E/F ).

Proof. Let n = [E : F ], and let C be an algebraically closed field such that C ⊇ E ⊇ F .

Denote G = G(E/F ).
(i)⇒(ii) Suppose that the extension E ⊇ F is normal. So E is the splitting field of some
polynomial f (X) = (X − α1 )(X − α2 ) · · · (X − αr ) ∈ F [X], i.e. E = F (α1 , α2 , . . . , αr ).
Every F -monomorphism σ : E → C permutes the zeroes α1 , α2 , . . . , αr of f (X), and so
E σ = E, i.e. σ ∈ G. Conversely, if σ ∈ G then σ : E → E ⊆ C is an F -monomorphism.
So G is exactly the set of all F -monomorphisms E → C. By Theorem 21.1, we have
|G| = n = [E : F ].
(ii)⇒(iii) Suppose that |G| = n = [E : F ]. Consider any element α ∈ E which is fixed
by every σ ∈ G. We have E ⊇ F [α] ⊇ F , and by Theorem 21.1, there are only [E : F [α]]
many F [α]-monomorphisms E → C. However, G contains n such F -monomorphisms, so
that [E : F ] = n ≤ [E : F [α]], which implies that F [α] = F , i.e. α ∈ F .
(iii)⇒(i) Suppose that F is the fixed field of G. Let α ∈ E, and define f (X) =

σ∈G (X − α ) ∈ E[X]. Then every τ ∈ G permutes the factors of f (X), so that τ
σ

fixes the polynomial f (X). Since every coefficient in f (X) is fixed by every τ ∈ G, by
assumption we have f (X) ∈ F [X]. Thus Irrα,F (X) divides f (X), which has all of its
zeroes in E, so that the extension E ⊇ F is normal.
82 III. FIELDS

A Galois extension is a finite normal separable extension. Such an extension achieves

equality in the upper bound of Theorem 22.1; or equivalently, satisfies any (and all) of the
conditions of Theorem 22.2. For any Galois extension E ⊇ F , the group G(E/F ) is called
the Galois group of the extension, and is usually denoted Gal(E/F ).
Our next theorem is the central result in this course. It shows that for a Galois
extension E ⊇ F , there is a one-to-one correspondence between intermediate fields of the
extension, and subgroups of the Galois group Gal(E/F ). This is known as the Galois
correspondence. Before proving this result, we give a small example which illustrates
the highlights of this correspondence. Consider the splitting field E of the polynomial
f (X) = X 3√− 2 over Q. We may factor f (X) = (X − α)(X − ωα)(X − ω 2 α) in C[X],
where α = 3 2, the unique real cube root of 2, and ω = e2πi/3 , a primitive cube root of 1,
satisfying ω 2 + ω + 1 = 0. We saw in Section 19 that the extension E = Q[α, ω] ⊃ Q
has degree 6. This is a Galois extension, so its Galois group G = Gal(E/Q) has order 6.
What is G? First observe that since G fixes the polynomial f (X) ∈ Q[X], it must permute
the three zeroes of f (X). Also, any automorphism of E is determined by its action on the
three zeroes α, ωα, ω 2α, since these generate E over Q. Therefore G may be identified as
the set of all six permutations of {α, ωα, ω 2α}, i.e. G ∼
= S3 . Indexing the three zeroes of
f (X) by
α ↔ ‘1’, ωα ↔ ‘2’, ω 2 α ↔ ‘3’,

and identifying G as S3 , it is clear how G acts on E. For example, (23) ∈ G fixes the first
zero α and interchanges ωα ↔ ω 2 α. Therefore (23) interchanges ω ↔ ω = ω 2 and fixes
all real elements of E. In other words, (23) acts on E just the same as complex conjugation!
The remaining automorphisms of E are a little more subtle. For example, (123) ∈ G cycles
2
α → ωα → ω 2 α → α. What does (123) do to ω? We have ω = ωα α
→ ωωαα = ω. So (123)
fixes everything in the quadratic subfield Q[ω]. Since (123) does not fix everything in E,
the fixed subfield of (123) must be just Q[ω] and nothing more.
We illustrate all subgroups of G = S3 and all intermediate fields of the extension
E ⊃ Q in the diagrams:

Q[α, ω] . . .
S..3...
............................................... ............ ............
............ .... ........................... .............. ... ..... ........
.......... ...... .................................................... ................ ... ..... .........
............... .......... .. .....................
.
..
........
. ...... .......... ................................
2 .
.. ..
.
................
.. ... .... ......
.... ......
..........
................ 2 ....
....
..........
.......... .................
................. 2 ................
................
..... .... ......
......
.. .
........
..............
. .
.
......
2
.
. .... .
.............
..........
.................
................ ..
....................... ... ....
.... ......
......
3 ....
.......... . . ... ...
.......
.................
............... ..............
. .
..............
. ...
....
....
.... 3......
......
. ....
........
. . 3 3 ..... ......
Q[ω 2 α]
. ...
.
...
..........
..........
.
...
. .
......
Q[α] Q[ωα] (123) ...
...
....
....
....
......
......
......
......
.......... .... .... ......
.... .. ............ ... .... ......
............ ... ............ .. ......
..........
.. .... ...... .......... ... .... .....
...
.. . ....
. ... .. ...... ..........
............
.
Q[ω] ..
... ...
....
....
.
......
......
...... ..........
..........
............ (23) (13) (12)
................
................ 3 ..
3 ..... ..
. ...... ..........
..........
......
..........
. .
..................
................ .
....
.
.... ......
..
3 ............ ...... .......... ................
................ ... ...... 3 .......... ...... ..
............... .
...
......................
................ .. .... ...... .......... ...... .. .
..........
...
.. ..
................
.
..
................
................
...
. ..
..... ........... ............
.......... 2 ....
......
2 ..........
..........
................
..................
2 ................
................
................
... ....... ...........
. ... ....
. ..........
............ ......
.
..
............... ..............................
.
.
.......... .... .......... .................. . . 2
................ .... .................. .......... ...... ........................
...............................
.. .. ....
Q (1)

Here a double line represents a normal inclusion, and a single line represents an inclusion
which is not normal. Each integer label on a line represents the corresponding index or
 III. FIELDS 83

degree. Notice that the two pictures are almost the same, except each is an upside-down
image of the other. To each intermediate subfield L ⊆ E on the left, there corresponds
the subgroup GL ≤ G on the right, the set of all σ ∈ G fixing every element of L.
And to each subgroup H ≤ G on the right, there corresponds the subfield EH ⊆ E on
the left, the fixed field of H. If L ⊇ L on the left, then on the right this inclusion is
reversed as GL ≤ GL
, and the degree [L : L ] equals the index [GL
: GL ]. Moreover
the extension L ⊇ L is normal iff GL ≤ GL
. There is a nontrivial example of this
in our picture: the extension Q[ω] ⊃ Q is normal of degree 2, and the corresponding
subgroups are (123) < G, of index 2. Finally, since the extension Q[ω] ⊃ Q is normal,
it is Galois, and its Galois group has order 2. This may be identified with the quotient
group G/(123) ∼ = C2 . Why? The group G acts on Q[ω], giving rise to an action
φ : G → Gal(Q[ω]/Q) which takes each σ ∈ G to its restriction to Q[ω]. This map φ is a
homomorphism, and by definition, its kernel is the set of all σ ∈ G which act trivially on
Q[ω], i.e. the set of all σ ∈ GQ[ω] = (123). The fact that φ is onto Gal(Q[ω]/Q) follows
by comparing orders. So G/(123) ∼ = Gal(Q[ω]/Q) follows from the First Isomorphism
Theorem for groups. All these observations we have made in this special example are
consequences of the following theorem.

22.3 Fundamental Theorem of Galois Theory. Let E ⊇ F be a Galois exten-

sion, with Galois group G = Gal(E/F ). Let L be the class of all intermediate fields
L (i.e. E ⊇ L ⊇ F ), and let H be the class of all subgroups of G.
(i) The map which associates to each intermediate field L the subgroup GL ≤ G
fixing L, is a bijection L → H.
(ii) The map which associates to each subgroup H ≤ G its fixed field EH ⊆ E,
is a bijection H → L. This is the inverse of the bijection in (i).
(iii) We have containment L ⊇ L for intermediate fields, iff we have containment
GL
≥ GL of the corresponding subgroups. Moreover, in this case, [L : L ] =
[GL
: GL ].
(iv) Suppose that L ⊇ L for intermediate fields. Then the extension L ⊇ L is
normal (and hence Galois) iff GL ≤ GL
. If these conditions are satisfied,
then Gal(L/L ) ∼
= GL
/GL = Gal(E/L )/ Gal(E/L).

Proof. Let φ : L → H, L → GL , the map in (i); and let ψ : H → L, H → EH ,

the map in (ii). For each L ∈ L, the extension E ⊇ L is Galois with Galois group
GL = Gal(E/L), and by Theorem 22.2, the fixed field of GL is L. This proves that the
φ ψ
composite map L −→ H −→ L is the identity.
Consider an arbitrary subgroup H ≤ G, and let L = EH , the fixed field of H. We
must show that GL = H. By definition, we have GL ≥ H. By Theorem 21.2, E = L[α]

for some α ∈ L. Define f (X) = σ∈H (X − ασ ) ∈ E[X]. Clearly every τ ∈ H permutes
the factors of f (X), and so fixes f (X). Since the coefficients in f (X) are all fixed by
84 III. FIELDS

H, they lie in the fixed field L, i.e. f (X) ∈ L[X]. So the degree of α over L satisfies
[L[α] : L] ≤ deg f (X) = |H|. Using Theorem 22.2, we have |GL | = [E : L] ≤ |H| ≤ |GL |.
ψ φ
Therefore GL = H; in other words, the composite map H −→ L −→ H is the identity.
Therefore the maps φ and ψ are bijections, each the inverse of the other.
If L ⊇ L are intermediate fields, then by definition every L-automorphism of E fixes
the smaller field L , i.e. GL ≤ GL
. Conversely, suppose GL ≤ GL
. Since Gal(E/L) =
GL ≤ GL
fixes every element of L , Theorem 22.2 forces L ⊆ L. In this case [GL
:
GL ] = |GL
|/|GL | = [E : L ]/[E : L] = [L : L ], which proves (iii).
Suppose that L ⊇ L is a normal extension of intermediate fields. We will show that
every σ ∈ GL
preserves L. Given α ∈ L, let f (X) = Irrα,L
(X). Then ασ is also a zero
of f (X), and since the extension L ⊇ L is normal, we have ασ ∈ L, Thus Lσ = L as
required. The restriction GL
→ Gal(L/L ), σ → σ L is clearly a homomorphism. By
definition its kernel is GL , so we have GL ≤ GL
and by the First Isomorphism Theorem
for groups, GL
/GL is isomorphic to a subgroup of Gal(L/L ). Comparing orders, we get
equality: Gal(L/L ) ∼= GL
/GL = Gal(E/L )/ Gal(E/L).
Conversely, suppose that GL ≤ GL
. Then GL
must preserve L, the fixed field of
GL , by Exercise 8.4. So every σ ∈ GL
induces an L -automorphism of L, and the restric-
tion GL
→ G(L/L ), σ → σ L is a homomorphism with kernel GL . Therefore GL
/GL
is isomorphic to a subgroup of G(L/L ). But comparing orders, using Theorem 22.1 we
have |GL
/GL | ≤ |G(L/L )| ≤ [L : L ] = [GL
: GL ]. Therefore equality holds in the upper
bound of Theorem 22.1; in other words, the extension L ⊇ L is normal.

The Galois group of a polynomial f (X) ∈ F [X] over F , is by definition Gal(E/F )

where E is a splitting field for f (X) over F .

22.4 Theorem. Suppose that f (X) ∈ F [X] is a polynomial of degree n ≥ 1, and

let G be the Galois group of f (X) over F . Then G is isomorphic to a subgroup of Sn .
Moreover if f (X) is irreducible over F , then G is isomorphic to a transitive subgroup
of Sn .

Proof. Let E be a splitting field for f (X). We may assume f (X) = (X − α1 )(X −
α2 ) · · · (X − αn ), E = F (α1 , α2 , . . . , αn ), and G = Gal(E/F ). Every σ ∈ G permutes
the zeroes α1 , α2 , . . . , αn of f (X), and since α1 , α2 , . . . , αn generate E over F , σ is de-
termined by its action on {α1 , α2 , . . . , αn }. Thus G may be identified as a subgroup of
Sym{α1 , α2 , . . . , αn } ∼
= Sn . 
Suppose that f (X) is irreducible in F [X]. Let g(X) = σ∈G (X − ασ1 ). Then every
τ ∈ G permutes the factors of g(X), and so G fixes g(X). By Theorem 22.2, we have
g(X) ∈ F [X]. Since α1 is a zero of g(X), it follows that Irrα1 ,F (X) = f (X) divides
g(X). Therefore every αi equals ασ1 for some σ ∈ G. That is, G acts transitively on
{α1 , α2 , . . . , αn }.
 III. FIELDS 85

We provide two illustrations of Theorem 22.4 using the notation of the previous ex-
ample. The polynomial f (X) = X 3 − 2 is irreducible over Q. Its Galois group over Q is
S3 , permuting the three zeroes α, ωα, ω 2 α in all six possible ways.
Also, f (X) = X 3 − 2 is irreducible over Q[ω]. The Galois group of f (X) over Q[ω] is
G = Gal(Q[α, ω]/Q[ω]) ∼ = (123), which cyclically permutes the three zeroes α, ωα, ω 2 α
of f (X). Once again, G acts transitively on the zeroes of f (X).
Recall Cayley’s Representation Theorem 8.2: Every finite group G is isomorphic to
a subgroup of Sn for some n. Now it is natural to ask: Is every finite group isomorphic
to the Galois group of some polynomial over Q? This is the so-called inverse problem of
Galois theory, which has occupied the minds of many brilliant mathematicians. Despite
great advances in this area, to date there is no complete solution known to this problem.
Exercises 22.
√ √
1. Show that the extension E = Q[ 2, 5] ⊃ Q is Galois. Compute the Galois group G = Gal(E/Q).
Give diagrams illustrating the intermediate fields, and the subgroups of G, similar to those given for
the example above.

2. Let F be a finite field of characteristic p. By Exercise 18.5, |F | = q = pr where p is prime and r ≥ 1;

moreover, F ⊇ Fp is an extension of degree r. Define σ : F → F , x → xp .
(a) Prove that σ ∈ Aut F .
Hint: Use Exercise 21.5.
(b) Prove that the extension F ⊇ Fp is separable.
Hint: Suppose α ∈ F is a multiple zero of f (X) = Irrα,Fp (X). Write n = deg f (X). Since α
is a zero of the polynomial f  (X) ∈ F [X] whose degree is less than n, we must have f  (X) = 0.
Deduce that every nonzero term aj X j appearing in f (X) must have j divisible by p, and that
f (X) = g(X)p where g(X) ∈ F [X].
(c) Prove that Aut F = {1, σ, σ 2 , . . . , σ r−1 }.
j i
Hint: If σ i = σ j where 1 ≤ i < j < r, then the polynomial X p − X p of degree pj < q has q
zeroes a ∈ F , which is impossible. Therefore the automorphisms 1, σ, σ 2 , . . . , σ r−1 ∈ Aut F are
distinct. Apply Theorem 22.1.
Remark: This shows that every finite field is a Galois extension of its prime field, with cyclic Galois
group; and so by the Fundamental Theorem of Galois Theory that every extension E ⊇ F where
E and F are finite fields, is Galois with cyclic Galois group.

23. Cyclotomic Extensions

For every positive integer n, the set of all complex n-th roots of unity, i.e. the set of all
complex solutions of z n = 1, is a cyclic group of order n. A primitive n-th root of
unity is a generator of this group. Let ζn denote your favorite n-th root of unity; mine is
e2πi/n , but the choice is not actually relevant. Then the primitive n-th roots of unity are
just the powers ζnk for those values of k ∈ {1, 2, . . . , n} such that gcd(k, n) = 1. Therefore
the number of such primitive roots is just given by Euler’s function

ϕ(n) = |{k ∈ Z : 1 ≤ k ≤ n, gcd(k, n) = 1}|.

86 III. FIELDS

A cyclotomic extension is an extension of the form F (ζn ) ⊇ F , where F is some

extension of Q. Clearly F (ζn ) = F (1, ζn , ζn2 , . . . , ζnn−1 ) is the splitting field of X n − 1,
so cyclotomic extensions are normal, and hence Galois. Let G = Gal(F (ζn )/F ). Then
|G| = [F (ζn ) : F ] ≤ n, since the minimal polynomial Irrζn ,F (X) divides X n − 1. But we
can do much better. Define the n-th cyclotomic polynomial by

Φn (X) = (X − ζnk ).
1≤k≤n
gcd(k,n)=1

This polynomial has degree ϕ(n). (We have already encountered this polynomial in special
cases; see Theorem 17.6 and Exercise 17.9.) Every automorphism of the Galois extension
Q(ζn ) ⊇ Q permutes the primitive n-th roots of unity, and so fixes the polynomial Φn (X);
therefore by Theorem 22.2, we have Φn (X) ∈ Q[X]. (In fact one can show that Φn (X) ∈
Z[X], although we will not need this.) So in an arbitrary cyclotomic extension F (ζn ) ⊇ F ,
the minimal polynomial Irrζn ,F (X) divides Φn (X), and so [F (ζn ) : F ] ≤ ϕ(n).

23.1 Theorem. Let F (ζn ) ⊇ F be a cyclotomic extension. Then the Galois group
G = Gal(F (ζn )/F ) is abelian. Moreover, |G| = [F (ζn ) : F ] ≤ ϕ(n), and equality
holds iff Φn (X) is irreducible in F [X].

Proof. Let σ ∈ Gal(F (ζn )/F ). Then σ maps ζn to another primitive n-th root of unity,
say ζnσ = ζnk , where gcd(k, n) = 1. If also τ ∈ Gal(F (ζn )/F ) then ζnτ = ζn
for some integer
 with gcd(, n) = 1. Then both στ and τ σ map ζn → ζnk
. Since the F -automorphisms
στ and τ σ have the same effect on the generator ζn , we must have στ = τ σ; that is,
Gal(F (ζn )/F ) is abelian. The remaining assertion follows from the remarks above.

We omit the proof of the following result, since it is not required in Section 24. Note
that in the special case n is a prime power, the result follows from Exercise 17.9 together
with Theorem 23.1. For a proof in the general case, see e.g. Garling [1].

23.2 Theorem. The polynomial Φn (X) is irreducible in Q[X]. Therefore [Q(ζn ) :

Q] = ϕ(n), and the Galois group Gal(Q(ζn )/Q) is isomorphic to the group of units of
the ring Z/nZ.

Exercises 23.
1. Compute Φn (X) explicitly for n = 1, 2, 3, . . . , 10.

2. Let n be a positive integer. Prove that X n − 1 = Φd (X), where the product is over all positive
integers d dividing n. d|n

3. Verify Theorem 23.2 in the special case n is a prime power.

 III. FIELDS 87

24. Radical Extensions

In this final section, we briefly indicate the key ideas in the application of Galois theory to
solvability of polynomials by radicals, as alluded to in the Prologue. An extension E ⊇ F
is an elementary radical extension if E = F [α] for some α ∈ E such that αn ∈ F
for some integer n ≥ 1. This means that α is an n-th root of some element of F . More
generally, E ⊇ F is a radical extension (or extension by radicals) if there exist
intermediate subfields

E = Ek ⊇ Ek−1 ⊇ Ek−2 ⊇ · · · ⊇ E1 ⊇ E0 = F

such that each extension Ei ⊇ Ei−1 is an elementary radical extension. Clearly, to say
that a number x is expressible in terms of elements of F using +, −, × and /, together
with the extraction of roots, is equivalent to saying that x lies in some radical extension
of F .
A polynomial f (X) ∈ F [X] is solvable by radicals if there exists a radical extension
E ⊇ F which contains a splitting field for f (X) over F . It is important to realize that we
do not require E itself to be a splitting field for f (X); very often E will be much larger.
It is clear that what we mean by saying that the zeroes of f (X) are expressible in terms
of the coefficients of f (X) using the standard four operations plus extraction of roots, is
precisely the property of solvability of f (X) by radicals. √
In our example of X 3 − 2 = (X − α)(X − ωα)(X − ω 2 α) ∈ Q[X] where α = 3 2
and ω = e2πi/3 , we see that each of the extensions Q[α] ⊃ Q and Q[α, ω] ⊃ Q[α] is an
elementary radical extension, and so Q[α, ω] ⊃ Q is an extension by radicals (but not an
elementary radical extension). Since Q[α] ⊃ Q is itself a radical extension, we see that a
radical extension is not necessarily normal.

24.1 Theorem. Suppose that F is a field of characteristic zero, and let f (X) ∈
F [X]. Then f (X) is solvable by radicals over F , iff the Galois group of f (X) over F
is solvable.

We will not have time to fully prove this, but will give some indication of its proof.
We observed that a radical extension is not necessarily Galois. However, consider an
elementary radical extension F [α] ⊇ F . This will be a Galois extension if F contains
enough roots√of unity, as we proceed to show. For example while Q[α] ⊃ Q is not Galois
where α = 3 3 , yet by adjoining a primitive cube root of unity ω, we obtain F = Q[ω],
and the extension F [α] ⊃ F is Galois.

24.2 Lemma. Let E = F [α] ⊇ F be an elementary radical extension, say αn =

γ ∈ F . If F contains a primitive n-th root of unity, then the extension E ⊇ F is
Galois, and the Galois group Gal(E/F ) is cyclic of order dividing n.
88 III. FIELDS

Proof. Since there exists a primitive n-th root of unity ζ = ζn ∈ F , we have E = F [α] =
n−1
F [α, ζα, ζ 2 α, . . . , ζ n−1 α], which is the splitting field of X n − γ = k=0 (X − ζ k α) over F .
So the extension E ⊇ F is normal, and hence Galois.
Let σ ∈ G = Gal(E/F ). Then σ must permute the zeroes of X n − γ, so ασ = ζ k α
for some integer k. Since α generates E over F , σ is uniquely determined by k. So it is
reasonable to rename this σ as σk . Now we have a one-to-one map G → ζ given by
σk → ζ k . This map is a group homomorphism since (ασk )σ
= (αζ k )ζ
= αζ k+
= ασk+
.
Thus G is isomorphic to a (possibly proper) subgroup of ζ, and so G is cyclic of order
dividing n.

Consider an arbitrary radical extension

E = Ek ⊇ Ek−1 ⊇ Ek−2 ⊇ · · · ⊇ E1 ⊇ E0 = F

where Ei = Ei−1 [αi ] and αi is an ni -th root of γi ∈ Ei−1 . Let n = lcm(n1 , n2 , . . . , nk ).

We adjoin a primitive n-th root of unity to obtain new fields Ei = Ei (ζn ). This gives a
new tower of extensions

E  = Ek ⊇ Ek−1
 
⊇ Ek−2 ⊇ · · · ⊇ E1 ⊇ E0 = F (ζn ) ⊇ F.

Each of the extensions Ei ⊇ Ei−1 

is Galois with cyclic Galois group, by Lemma 24.2.
Also, the last extension F (ζn ) ⊇ F is Galois, with abelian Galois group, by Theorem 23.1.
We are not quite home free, since we still don’t know that E  (or E) is a Galois extension
of F . Fortunately it is possible to reduce to this case; for details, see Garling [1]. As-
suming however that both E and E  are Galois extensions of F , then by the Fundamental
Theorem 22.3, we have a normal series for Gal(E  /F ) given by

1 = Gal(E  /E  ) ≤ Gal(E  /Ek−1


) ≤ Gal(E  /Ek−2

) ≤ ···
· · · ≤ Gal(E  /E1 ) ≤ Gal(E  /E0 ) ≤ Gal(E  /F )

in which the factors are Gal(E /Ek−1

 
), Gal(Ek−1 
/Ek−2 ), . . . , Gal(E1/E0 ), and Gal(E0 /F ),
all of which are abelian. Therefore Gal(E  /F ) is solvable.
By definition, E contains a splitting field K for f (X). Now E  = E(ζn ) ⊇ E ⊇ K ⊇ F ,
in which each containment is normal, and so Gal(K/F ) ∼ = Gal(E  /F )/ Gal(E  /K) is
solvable. This gives a proof of one direction of Theorem 24.1, except with gaps as mentioned
above. That’s all we’ll say about the proof of Theorem 24.1.
If one assumes Theorem 24.1, then every polynomial f (X) ∈ F [X] of degree at most
four is solvable by radicals. This is because, by Theorem 22.4, the Galois group of f (X)
over F is isomorphic to a subgroup of Sn where n = deg f (X) ≤ 4, and this group is
solvable (see Section 13).
This is not the case for polynomials of degree greater than or equal to 5. Suppose
that f (X) ∈ F [X] is irreducible of degree 5, so that the Galois group of f (X) over F is
 III. FIELDS 89

a transitive subgroup of S5 . Up to equivalence, there are five such permutation groups

(Exercise 10.2), and each of these five groups occurs as the Galois group of some irreducible
polynomial of degree 5 over Q. We list representatives of each of these five types, as given
by Garling [1]:
(i) X 5 + X 4 − 4X 3 − 3X 2 + 3X + 1, with Galois group (12345) ∼ = C5 ;
(ii) X 5 − 5X + 12, with Galois group (12345), (25)(34) ∼
= D5 ;
(iii) X 5 − 2, with Galois group (12345), (1243) of order 20;
(iv) X 5 + 20X + 16, with Galois group A5 , and
(v) X 5 − 4X + 2, with Galois group S5 .
Polynomials (i)–(iii) above are solvable by radicals, since they have solvable Galois groups
over Q; polynomials (iv) and (v) are not solvable by radicals, since their Galois groups are
nonsolvable.
Exercises 24.
1. This exercise shows that the polynomial (i) above has solvable Galois group ∼
= C5 , and that it is
solvable by radicals (in fact, using 11th roots of unity).
Let ζ = e2πi/11 ∈ C and let α = ζ + ζ −1 . Since ζ 11 = 1 ∈ Q, the cyclotomic extension E = Q(ζ) ⊃ Q
is an elementary radical extension. Recall (Theorem 23.2) that the Galois group of this extension is
Gal(E/Q) ∼= C10 .
(a) Prove that α ∈ R. Conclude that the field F = Q(α) is a proper subfield of E.
(b) Show that ζ is a zero of a quadratic polynomial in F [X]. Conclude that [E : F ] = 2 and that
Gal(E/F ) ∼
= C2 .
(c) Using the Fundamental Theorem of Galois Theory, show that the extension F ⊃ Q is Galois and
that Gal(F/Q) ∼
= C5 .
(d) Show that α is a zero of the polynomial f (X) = X 5 + X 4 − 4X 3 − 3X 2 + 3X + 1 ∈ Q[X] given
in (i) above.
(e) Show that f (X) is irreducible and that f (X) = Irrα,Q (X).
(f) Conclude that F is the splitting field of f (X) and so the Galois group of f (X) is isomorphic to
C5 .
 Appendix: Zorn’s Lemma

At a couple points during the course we have benefited from Zorn’s Lemma. Here we
outline the statement of Zorn’s Lemma and give an example of its use.
Let S be a set. A partial order on S is a binary relation ≤ such that for all x, y, z ∈ S,
(i) x ≤ x;
(ii) if x ≤ y and y ≤ x, then x = y; and
(iii) if x ≤ y and y ≤ z, then x ≤ z.
Note that there can be many pairs of elements {x, y} in X which are incomparable, i.e.
x ≤ y and y ≤ x. A chain is a subset C ⊆ x such that for all x, y ∈ S, either x ≤ y
or y ≤ x. We write x < y as an abbreviation for the statement that ‘x ≤ y and x = y’.
If S ⊆ X, an upper bound for S is an element b ∈ X such that s ≤ b for all s ∈ S. We
say that S is bounded above if such an upper bound for S exists. Note that b is not
required to belong to the subset S in this case. A maximal element in X is an element
m ∈ X such that no element of X is larger than m; that is, there does not exist x ∈ X
such that m < x.
Example: Z with Divisibility. An example is the relation

of divisibility

on the set
of integers, in which the pair {4, 15} is incomparable since 4 
15 and 15 
4. In this
setting, {1, 2, 4, 8, 16, . . .} is a chain with no upper bound. The chain {3, 12, 36, 1440} has
many choices of upper bound: 1440 is an upper bound (the least upper bound), and 2880
is also an upper bound. There is no maximal element in Z for the divisibility relation.
Example: X ⊂ Z with Divisibility. Now consider the set X consisting of integers
expressible as a product of at most 5 prime factors. For example, X contains 23 31 = 24
and 23 31 71 = 168 but not 23 31 51 71 = 840. We use divisibility as our relation on X.
Every chain in X has at most six elements. Moreover every chain C ⊂ X has an upper
bound: either C = Ø, in which case 1 (or any element of X) is an upper bound for C,
or the largest element of C is an upper bound for C. The element 32 ∈ X (or, for that
matter, any element with exactly 5 prime factors, not necessarily distinct) is a maximal
element of X. Note, however, that 32 is not an upper bound for X.

Zorn’s Lemma. Let X be a nonempty partially ordered set, and suppose every
chain in X is bounded above. Then X has a maximal element.

Like most authors, we assume this result rather than proving it. The reason for this is
that one cannot prove this result without assuming the Axiom of Choice (or something at
least as strong). This is because Zorn’s Lemma is equivalent to the Axiom of Choice, given
the Zermelo-Fraenkel axioms of set theory. It is typically used as a convenient crutch,
where no maximal element is explicitly constructible. This should not be of great concern,
however, since in practical situations where a maximal element is desired, we can typically
get by without one. We will try to make this point clear in the context of an example.
92 APPENDIX: ZORN’S LEMMA

Corollary. Every vector space has a basis.

Proof. Let V be a vector space over a field F . We assume V = 0; otherwise Ø is a basis

for V .
Let I be the collection of all linearly independent subsets of V . Recall that a subset
S ⊆ V is linearly dependent if there exist distinct vectors v1 , v2 , . . . , vk ∈ S and scalars
a1 , a2 , . . . , ak ∈ F , not all zero, such that a1 v1 + a2 v2 + · · · + ak vk = 0. Thus S ∈ I iff
S ⊂ V and S is not linearly dependent. Clearly I is nonempty, since every nonzero vector
v ∈ V gives rise to a linearly independent subset {v} ∈ I.  
Let C ⊂ I be any chain. We claim that  C is
 bounded above by C. (Recall that
 C is
the union of all members of C; that is, C = S∈C S.) We must first show that C ∈ I.
Consider any distinct vectors v1 , v2 , . . . , vk ∈ C and let a1 , a2 , . . . , ak ∈ F . For every
i = 1, 2, . . . , k, the fact that vi ∈ C means that vi ∈ Si for some linearly independent
subset Si ∈ C. Since C is a chain, the Si ’s are totally ordered by inclusion. This means we
may assume that S1 ⊆ S2 ⊆ · · · ⊆ Sk ; at least this will be the case if v1 , v2 , . . . , vk were
listed in a suitable order. But now v1 , v2 , . . . , vk all belong to the linearly independent
set Sk , and so the  scalars a1 , a2 , . . . , ak must all be zero. This shows that C is linearly
independent, so C ∈ I. We still need to show that C is an upper bound for the chain 
C. But this is obvious since for every linearly independent subset S ∈ C, we have S ⊆ C
by definition.
Let B be a maximal element for I, which exists by Zorn’s Lemma. So B is linearly
independent. It remains to be shown that B spans V . Let v ∈ V . We must show that v is
in the span of B. If v ∈ B then this is clear; so we may assume that v ∈ / B, so that B is
a proper subset of B ∪ {v}. Since B is a maximal element of I, it must be the case that
B ∪ {v} is linearly dependent. Thus there exist distinct vectors v1 , v2 , . . . , vk ∈ B ∪ {v}
and scalars a1 , a2 , . . . , ak ∈ F , not all zero, such that
a1 v1 + a2 v2 + · · · + ak vk = 0.
Clearly v ∈ {v1 , v2 , . . . , vk } since B itself is linearly independent; we may assume that
v1 = v. Moreover a1 = 0, for otherwise we have found a nontrivial linear relation between
v2 , v3 , . . . , vk ∈ B, which cannot occur since B is linearly dependent. Thus
 
v = −a−11 a2 v2 + a3 v3 + · · · + ak vk
lies in the span of B, as required. Thus B spans V . Since B is also linearly independent,
B is a basis for V .

For finite dimensional vector spaces, it is very easy to produce bases explicitly, and so
Zorn’s Lemma is not needed in such cases. For many infinite-dimensional vector spaces,
this is not an option. For example, the vector space C([0, 1]) consisting of continuous
functions [0, 1] → R, has a basis, by Zorn’s Lemma. But you will never see an explicit
basis for this vector space! since none can be written down. But in any practical situation
in which C([0, 1]) arises, this is not an issue since we typically deal with only certain
well-known proper subspaces of C([0, 1]) for which explicit bases are known.
 Bibliography
[1] D. J. H. Garling, A Course in Galois Theory, Cambridge Univ. Press, 1986. Quite
readable. Appropriate for the later course material.
[2] D. Gorenstein, ‘The enormous theorem’, Scientific American 253 (1985), pp.104–115.
A layman’s introduction to the classification of finite simple groups.
[3] I. N. Herstein, Topics in Algebra, Wiley, New York, 1975. A standard general algebra
text.
[4] T. W. Hungerford, Algebra, Springer-Verlag, 1974b. Probably more useful as a general
algebra reference than as a textbook.
[5] L. Infeld, Whom the Gods Love, Whittlesey House, New York, 1948. A well-recom-
mended biography of the life of Evariste Galois.
[6] C. C. Pinter, A Book of Abstract Algebra, 2nd ed., McGraw-Hill, 1990.
[7] P. Samuel, Algebraic Theory of Numbers, Kershaw, London, 1972. Very helpful for
ring theory, field theory and Galois theory.
 Index
abelian group . . . . . . . . . . . . . . . 6 degree
action, permutation . . . . . . . . . . . . 26 of a permutation . . . . . . . . . . . . 7
algebraic . . . . . . . . . . . . . . . 71, 73 of a polynomial . . . . . . . . . . . . . 52
algebraic closure . . . . . . . . . . . 73, 74 of a representation . . . . . . . . . . . 25
algebraically closed . . . . . . . . . . . . 74 of an extension . . . . . . . . . . . . . 54
alternating group . . . . . . . . . . . . . 10 of α . . . . . . . . . . . . . . . . . . 72
associate . . . . . . . . . . . . . . . . . 62 derived subgroup . . . . . . . . . . . . . 25
associative . . . . . . . . . . . . . . . . 5 dihedral group . . . . . . . . . . . . . 8, 25
automorphism . . . . . . . . . . . . 23, 78 direct product . . . . . . . . . . . . 13, 23
Axiom of Choice . . . . . . . . . . 74, 80, 91 direct sum . . . . . . . . . . . . . . . . 55
divides . . . . . . . . . . . . . . . . 1, 62
binary operation . . . . . . . . . . . . 5, 49
division algorithm . . . . . . . . . . . . 1, 61
bounded above . . . . . . . . . . . . . . 91
division ring (skewfield) . . . . . . . . . . 50
canonical homomorphism . . . . . . . 20, 59 divisor . . . . . . . . . . . . . . . . . 1, 62
Cauchy’s Theorem . . . . . . . . . . 23, 31
Eisenstein Criterion . . . . . . . . . . . . 67
Cayley Representation Theorem . . . . . . 29
elementary construction . . . . . . . . . . 75
Cayley table . . . . . . . . . . . . . . . 10
elementary radical extension . . . . . . . . 87
center . . . . . . . . . . . . . . . . 14, 55
epimorphism . . . . . . . . . . . . . . . 20
centralizer . . . . . . . . . . . . . 17, 32, 55
equivalent representations . . . . . . . . . 35
chain . . . . . . . . . . . . . . . . . . 91 Euclidean domain . . . . . . . . . . . . . 61
characteristic of a field . . . . . . . . . . . 71 Euclidean ring . . . . . . . . . . . . . . 61
characteristic subgroup . . . . . . . . . . 25 Euclid’s Algorithm . . . . . . . . . . 2–3, 62
characteristically simple group . . . . . . . 25 even permutation . . . . . . . . . . . . . 10
commutative ring . . . . . . . . . . . . . 49 exponent . . . . . . . . . . . . . . . . . 17
commutator . . . . . . . . . . . . . . . 22 extension of a field . . . . . . . . . . . . 54
companion matrix . . . . . . . . . . . . . 74
composition factor group . . . . . . . . . . . . . . . 18
factors . . . . . . . . . . . . . . . . . 45 faithful action/representation . . . . . 26, 29, 38
length . . . . . . . . . . . . . . . . . 45 field . . . . . . . . . . . . . . . . . . . 50
series . . . . . . . . . . . . . . . . . 45 finite field . . . . . . . . . . . . . . 13, 52
conjugacy class . . . . . . . . . . . . . . 31 finite group . . . . . . . . . . . . . . . 6
conjugate . . . . . . . . . . . . . 31, 32, 54 fixed field . . . . . . . . . . . . . . . . 78
constant polynomial . . . . . . . . . . . . 52 fractional linear transformation . . . . . . . 21
core . . . . . . . . . . . . . . . . . . . 37 Frattini subgroup . . . . . . . . . . . . . 17
corefree . . . . . . . . . . . . . . . . . 38 Fundamental Theorem
coset, right . . . . . . . . . . . . . . . . 15 of Arithmetic . . . . . . . . . . . . . 2, 47
coset, left . . . . . . . . . . . . . . . . 15 of Finite Abelian Groups . . . . . . . . . 23
of Galois Theory . . . . . . . . . . . . 83
cubic extension . . . . . . . . . . . . 54, 67
cycle . . . . . . . . . . . . . . . . . . 7 Galois
cycle structure . . . . . . . . . . . . . . 33 correspondence . . . . . . . . . . . . . 82
cyclic group . . . . . . . . . . . . . . . 8 extension . . . . . . . . . . . . . . . 82
cyclotomic field . . . . . . . . . . . . . . . . . . 52
extension . . . . . . . . . . . . . . . 86 group . . . . . . . . . . . . . . . 82, 84
polynomial . . . . . . . . . . . . . 69, 86 general linear group . . 7, 13, 21, 23, 33, 38, 41, 55
96 INDEX

generators . . . . . . . . . . 9, 14, 16, 60, 79 nonsolvable . . . . . . . . . . . . . . . 46

glide reflection . . . . . . . . . . . . . . 8 norm . . . . . . . . . . . . . . . . . . 54
greatest common divisor (gcd) . . . . . . 1, 62 normal
group . . . . . . . . . . . . . . . . . . 5 extension . . . . . . . . . . . . . . . 76
group table . . . . . . . . . . . . . . . 10 series . . . . . . . . . . . . . . . . . 45
subgroup . . . . . . . . . . . . . 17, 20
homomorphism . . . . . . . . . . . . 19, 59
normalizer . . . . . . . . . . . . . . . . 32
ideal . . . . . . . . . . . . . . . . . . 57
odd permutation . . . . . . . . . . . . . 10
identity . . . . . . . . . . . . . . . . 5, 49
orbit . . . . . . . . . . . . . . . . . . 25
image . . . . . . . . . . . . . . . . 20, 59
order . . . . . . . . . . . . . . . . . 6, 52
index . . . . . . . . . . . . . . . . . . 15
inner automorphism . . . . . . . . . . . . 24 partial order . . . . . . . . . . . . . . . 91
integral domain . . . . . . . . . . . . . . 50 p-group, p-subgroup . . . . . . . . . . . . 39
intermediate field . . . . . . . . . . . . . 78 permutation . . . . . . . . . . . . . . . 7
intransitive action . . . . . . . . . . . . . 25 permutation group . . . . . . . . . . . . 25
inverse . . . . . . . . . . . . . . . . 5, 6, 50 permutation action (representation) . . . . . 26
inverse problem . . . . . . . . . . . . . . 85 polynomial ring . . . . . . . . . . . . 52, 53
irreducible . . . . . . . . . . . . . . . . 64 power series . . . . . . . . . . . . . . . 57
isometry . . . . . . . . . . . . . . . . . 8 punctiliation . . . . . . . . . . . . . . . 5
isomorphism . . . . . . . . . . . . 10, 50, 78 prime field . . . . . . . . . . . . . . . . 71
Isomorphism Theorems prime ideal . . . . . . . . . . . . . . . . 63
for groups . . . . . . . . . . . . . 21, 22 primitive element . . . . . . . . . . . . . 80
Theorems for rings . . . . . . . . . . . 60 primitive n-th root of unity . . . . . . . . 85
principal ideal . . . . . . . . . . . . . . 60
Jordan-Hölder Theorem . . . . . . . . . . 45 principal ideal domain (P.I.D.) . . . . . . . 61
kernel . . . . . . . . . . . . . . . . 20, 59 principal ideal ring . . . . . . . . . . . . 61
Klein 4-group . . . . . . . . . . . 12, 13, 24 product of ideals . . . . . . . . . . . . . 57
projective general linear group . . . . . 21, 38
Lagrange’s Theorem . . . . . . . . . . . . 15
Laurent series . . . . . . . . . . . . . . 57 quadratic extension . . . . . . . 54, 66, 69, 77
least common multiple (lcm) . . . . . . . 1, 58 quaternion . . . . . . . . . . . . . . . . 53
quotient
matrix ring . . . . . . . . . . . . . . . . 51 field . . . . . . . . . . . . . . . . 55, 56
maximal group . . . . . . . . . . . . . . . . . 18
ideal . . . . . . . . . . . . . . . . . 63 ring . . . . . . . . . . . . . . . . . . 58
normal subgroup . . . . . . . . . . . . 22
radical extension . . . . . . . . . . . . . 87
subgroup . . . . . . . . . . . . . . . 17
rational function . . . . . . . . . . . . . 56
minimal normal subgroup . . . . . . . . . 25
rationalizing the denominator . . . . . . . 55
minimal polynomial . . . . . . . . . . . . 74
real quaternions . . . . . . . . . . . . . 53
modulo . . . . . . . . . . . 18, 51, 58, 60, 68
refinement of series . . . . . . . . . . . . 45
monic polynomial . . . . . . . . . . . . . 72
reflection . . . . . . . . . . . . . . . 8, 10
monomorphism . . . . . . . . . . . . 20, 78
regular representation . . . . . . . . . 29, 34
multiple . . . . . . . . . . . . . . . . 1, 50
relatively prime . . . . . . . . . . . . . . 1
multiplication table . . . . . . . . . . . . 10
representation, permutation . . . . . . . . 26
nonabelian group . . . . . . . . . . . . . 6 ring . . . . . . . . . . . . . . . . . . . 49
 INDEX 97

rotation . . . . . . . . . . . . . . 8, 30, 38
scalar matrix . . . . . . . . . . . . . . . 54
separable . . . . . . . . . . . . . . . . 77
Shoe-Sock Theorem . . . . . . . . . . . . 6
sign of a permutation . . . . . . . . . . . 9
simple group . . . . . . . . . . . . 19, 22, 25
simple zero . . . . . . . . . . . . . . . . 77
skewfield . . . . . . . . . . . . . . . . . 50
solvable group . . . . . . . . . . . . . . 46
solvable polynomial (by radicals) . . . . . . 87
special linear group . . . . . . . . . . . . 23
splitting field . . . . . . . . . . . . . . . 76
stabilizer . . . . . . . . . . . . . . . 25, 36
straightedge-and-compass construction . . . . 75
subfield . . . . . . . . . . . . . . . . . 54
subgroup . . . . . . . . . . . . . . . . 14
subnormal subgroup . . . . . . . . . . . . 45
subring . . . . . . . . . . . . . . . . . 54
sum of ideals . . . . . . . . . . . . . . . 57
superblock . . . . . . . . . . . . . . . . 40
superduperblock . . . . . . . . . . . . . 40
Sylow p-subgroup . . . . . . . . . . . . . 39
Sylow Theorems . . . . . . . . . . . . . 42
symmetric group . . . . . . . . . . . . 7, 25
symmetry . . . . . . . . . . . . 8, 25, 30, 38
tower of fields . . . . . . . . . . . . . . 54
transcendental . . . . . . . . . . . . 56, 71
transitive action . . . . . . . . . . . . . 25
transposition . . . . . . . . . . . . . . . 9
tricky . . . . . . . . . . . . . . . . . . 80
trivial ring . . . . . . . . . . . . . . . . 49
trivial subgroup . . . . . . . . . . . . . . 14
unfaithful action/representation . . . . . . . 26
unit . . . . . . . . . . . . . . . . . 49, 50
unity . . . . . . . . . . . . . . . . . . 49
upper bound . . . . . . . . . . . . . . . 91
zero divisor . . . . . . . . . . . . . . . 50
zero of a polynomial . . . . . . . . . . . . 62
zero polynomial . . . . . . . . . . . . . . 52
Zorn’s Lemma . . . . . . . . . . . 64, 74, 91