Binary Operations, Monoids, and Groups

W. E. Aitken

September 2022 Edition∗

This is one in a planned series of documents that survey the foundations of

algebra. The documents of this series aim to cover the same sort of topics one would
see in an introductory abstract algebra class, but to explore these topics perhaps a
bit more abstractly and in more depth than one would necessarily want in a first
course. So this series should be thought of as foundational but not introductory. As
such the target audience is a reader who has already had some exposure to abstract
algebra, but wishes to explore or review the foundations of the subject.
This document covers ideas related to the concept of binary operations. This
includes examples, various properties (commutative, associative) that binary op-
erations can have, the ideas of identity and inverse, and so on. This leads to the
definitions of monoids and groups. It also covers basic laws of exponents, the general
associativity laws, and the general commutative law. The ideas related to binary
operations are first step to algebra, and so this document can be thought of as the
“ground floor” for modern algebra built on a solid (but elementary) set-theoretical
base. As such it is not intended to cover group theory in any degree beyond an
introduction to a few basic notions, and it does not cover ring theory at all.

1 Background and Notation

I assume familiarity with some basic set theory including the idea of a Cartesian
product A × B of two sets A and B, basic ideas related to functions, the notion of a
family, and the notion of a total ordering of a set. I will assume as known the basic
number systems (see, for example, in my Number Systems text). These constitute
the logical prerequisites, but, as mentioned above, some previous exposure to groups
or rings, say, is also useful. Otherwise, some of this document might seem a bit
abstract and unmotivated. I use matrices in some examples as well, and the notion
of continuous and differentiable functions come up in a few exercises, but these
concepts are not central to the development of the theory.
In particular, as part of the set-theoretic background and notion of function, I
use the concept of domain and codomain of a function, and use the notation A → B
to indicate that the function f has domain A and codomain B (we sometimes say
∗ Version of September 16, 2022. Copyright c 2017–2022 by Wayne Edward Aitken. This

work is made available under a Creative Commons Attribution 4.0 License. Readers may copy
and redistribute this work under the terms of this license. This document is an modified and
expanded version of a set of handouts written for my Math 470 class, introduction to abstract
algebra, during the period 2017–2020.

1
that A → B is the type of f ). When it comes time to define exponentiation, I will
take as established the concept and basic properties of iteration of functions of the
form f : S → S. Such properties are established in my Number Systems text.
I will use N, Z, Q, R, C for the set of natural numbers, the set of integers, the
set of rational numbers, the set of real numbers, and the set of complex numbers
respectively. We have the inclusion N ( Z ( Q ( R ( C. There are two definitions
of N that are fairly common. I use the version where N includes 0 as an element:

N = {0, 1, 2, . . .}.

The other popular version of the natural numbers excludes 0 and starts at 1; I will
write this set as Z+ . The set Z differs from N by including negative integers.

2 Binary Operations
Definition 1. Let S be a set. A binary operation on S is just a function S ×S → S.

Example 1. Let S = R. Multiplication × : R × R → R is a binary operation since it

takes as input two real numbers (thought of as an ordered pair) and outputs a real
number. Addition and subtraction also give binary operations on R, but division
does not.
Let S = N. Multiplication and addition are binary operations on N. (Subtrac-
tion and division are not). There are many, many other examples. For example,
the function f : N × N → N defined by the rule f (x, y) = 17 is a binary operation.
This operation ignores the inputs x and y and always outputs the constant value
of 17. Such constant operations are not as interesting as the usual operations in
arithmetic, but they do satisfy the official definition.
Example 2. Let Xn = {1, . . . , n}. Let S = F(Xn ) be the set of functions Xn → Xn .
Thus S has nn elements. Then composition is a binary operation. To see this,
suppose g, f ∈ S. In other words, f, g are both functions Xn → Xn . Then the
composition g ◦ f is also a function Xn → Xn . In other words, if g, f ∈ S then
also g ◦ f ∈ S. So composition ◦ is a binary operation on S = F(Xn ).
Let Sn be the set of bijective functions Xn → Xn . We also call such functions
permutations. If g, f ∈ Sn then g◦f is also a bijection, so is in Sn . Thus composition
is a binary operation
Sn × Sn → Sn .
Note that Sn has n! elements.
Example 3. Let Mm,n (R) be the set of real matrices with m rows and n columns.
For example, two elements of M2,3 (R) are
 √   
1 0 2 1 e π
and .
−1 21 11 2 2e 2π

Then matrix addition + is a binary operation on Mm,n (R). When m = n we

write Mn (R) instead of Mm,n (R). Matrix multiplication is a binary operation
on Mn (R).

2
Remark. There are two possible notations we can use with binary operations: (i) or-
dinary functional notation or (ii) infix notation. Infix notation is the standard
notation for binary operations, but it is sometimes illuminating to use functional
notation from time to time.
First we describe ordinary functional notation. If f : S × S → S is a binary
operation, and if a, b ∈ S, then in ordinary functional notation we write

f (a, b)

for the result or output of the function. The element f (a, b) ∈ S is often called
the value. (Ordinary functional notion is sometimes called “prefix” notation, and
sometimes we write f a b as a variant for f (a, b) when using this notation.)
For infix notation we put the function name in between the two inputs. Typi-
cally the function name is not a letter, but a symbol such as ∗, +, or ×. Suppose,
for example, that ∗ : S × S → S is a binary operation, and that a, b ∈ S. Then
instead of writing the value as ∗(a, b) or ∗ a b, in infix notation we write

a∗b

where we put ∗ between the two inputs. In infix notation we require parenthesis
in what is otherwise an ambiguous expression. For example, a ∗ b ∗ c could be
interpreted as (a ∗ b) ∗ c or a ∗ (b ∗ c). If we have established grouping conventions we
can leave the parentheses off and let the reader mentally supply them. For instance
the infix expression x + y · z in various numbers systems (or rings in general) is
interpreted as x + (y · z), not as (x + y) · z.
We use infix notation in the next definition:
Definition 2. Let ∗ : S × S → S be a binary operation. We say that this operation
is commutative if
a∗b=b∗a
for all a, b ∈ S.
Remark. To show that ∗ : S × S → S is not commutative, you just need to find a
specific counterexample. For example, you probably suspect that subtraction

−: Z × Z → Z

is not commutative. To prove it you need to give a specific counterexample. We

choose 1 and 0 as our inputs in Z. This is a counterexample since 1 − 0 6= 0 − 1.
(Note that choosing 1 for both does not give a counterexample, since 1 − 1 = 1 − 1.
So not every choice of a, b ∈ S will give you a counterexample).
Remark. If we use ordinary functional notation instead of infix notation for a binary
operation f : S × S → S, the definition of commutative can be written as requiring

f (a, b) = f (b, a)

for all a, b ∈ S. An example of such a commutative binary operation (with ordinary

functional notation) would be the function defined by the equation

f (x, y) = sin(x2 y 2 )

3
where S = R. On the other hand the function g(x, y) = sin(xy 2 ) does not look
commutative because of the asymmetry between x and y in the formula. But to
prove it is not commutative you need a specific counterexample. (In this case it is
not difficult to find such a counterexample).
Remark. Suppose Sn is the set of permutations Xn → Xn . Then the binary opera-
tion of composition ψ : Sn × Sn → Sn applied to (g, f ) can be written in two ways:
as ψ(g, f ) or as g ◦ f . The first is ordinary functional notation The second is infix
notation using the symbol ◦ to denote the (infix version) of the operation.

2.1 Closure
Sometimes people describe a binary operation as “closed” to indicate that if a, b ∈ S
then a ∗ b ∈ S. For us this term is redundant for a binary operation since ∗ is
required to be a function S × S → S, so if a, b ∈ S then a ∗ b must always be in the
codomain S.
But the concept is useful if we are in a slightly different situation. Suppose you
have a function mapping ordered pairs in S to values in some set T . In other words,
suppose that we have a function ∗ : S × S → T . Suppose also that S is a subset
of T . We say this function is closed on S if a ∗ b ∈ S for all a, b ∈ S. In other words,
for any two inputs in S the value “lands in” S and never at a value which is in T
but outside S. When we are in this situation, we can restricted the codomain to S
and then we would have a true binary operation.1
Informally it is common for people to says that “binary operations must be
closed”, but as mentioned above closure build into the definition of binary operation,
and not an extra condition that needs to be imposed on binary operations. However,
when you define a binary operation S × S → S, you do need to check that the
values are always in S, and you may call this “checking for closure” if you would
like. This is part of making sure the definition is well-defined, and often this check
is left (tacitely) to the reader.
The concept of closed becomes more important when we work with submonoids
and subgroups. Suppose ∗ : S × S → S is a binary operation on S and that A is a
subset of S. We say that ∗ is closed on A if a ∗ b ∈ A for all a, b ∈ A. If that holds
we get a binary operation A × A → A by restricting the domain and codomain of
the original operation.
Example 4. Subtraction N × N → Z is not closed since the image of the function
contains negative integers which are outside N. So subtraction cannot be thought
of as a binary operation on N. Subtraction is a binary operation on Z, though.
1 In mathematics, restriction of codomain is often done tacitly, or without much fanfare, since

the restricted operation is in some sense “the same” and is represented by the same collection of
ordered pairs. But the “type” of the operation is changed since the codomain is changed. More
often authors will be more explicit about restriction of domain, but even this more standard type
of restriction is sometimes done tacitly, even though the “type” of the function has changed.

4
3 Associativity
Suppose that ∗ : S × S → S is a binary operation. An expression such as

a∗b∗c

that involves more than two elements is ambiguous since a binary operation com-
bines only two elements at a time. To avoid ambiguity, you need parentheses to
show what two elements are being combined. For example a∗b∗c can be interpreted
as (a ∗ b) ∗ c, which really means d ∗ c where d = a ∗ b. On the other hand, a ∗ b ∗ c
can be interpreted as a ∗ (b ∗ c), which really means a ∗ d where d = b ∗ c. It turns
out that many important operations are associative, and the two interpretations
of a ∗ b ∗ c give equal values.
Remark. For general operations, that may or may not be associative, a common
convention is to group from left to right. So a ∗ b ∗ c ∗ d would be interpreted as

((a ∗ b) ∗ c) ∗ d.

For example, subtraction is not associative (for Z), so what does

3 − 4 − (−10) − 5

mean? According to the left-to-right convention, it is equal to 4. The grouping

convention is very important for non-associative operations such as subtraction.
For associative operations it doesn’t really make a difference since the value is the
same.
Definition 3. Suppose ∗ : S × S → S is a binary operation. Then ∗ is said to be
associative if (a ∗ b) ∗ c = a ∗ (b ∗ c) for all a, b, c ∈ S.
Remark. If we use ordinary functional notation, and if f : S × S → S is the binary
operation, then associativity means

f (f (a, b), c) = f (a, f (b, c))

for all a, b, c ∈ S.

4 Operation Tables
Suppose ∗ : S ×S → T is given where S is a finite set. We can describe the operation
fully by giving a square table giving all the values of ∗. This involves (1) ordering
the set S, (2) making a row and column for each element of S in the order from
step 1, (3) for each a, b ∈ S writing the value a ∗ b in the a-row and b-column.
If all elements of the table are in S then we conclude that ∗ is closed and so (by
choosing T = S), the table can be regarded as describing a binary operation

S × S → S.

Exercise 1. Given an operation table for addition modulo 4 for the set

S = {0, 1, 2, 3}.

5
Exercise 2. Given an operation table for multiplication modulo 7 for the set
S = {1, 2, 3}.
is this a binary operation (is it closed)?
Exercise 3. Given an operation table for multiplication modulo 7 for the set
S = {1, 2, 4}.
is this a binary operation (is it closed)?
Exercise 4. What has to be true about the operation tables for a binary operation
to be commutative?
Remark. There is no quick criteria that I know of to decide if an operation is
associative by just looking at the operation table.
Exercise 5. Suppose S = {T, F } be a set with two distinct elements called “true”
and “false”. This set is important in logic. Define the “and” the “or” and the
“exclusive or” binary operations. Find the tables for these operators.
Exercise 6. Give a table for the composition operation on S3 . (Use cycle notation
if you are familiar with this).

5 Identity Elements
Definition 4. Let ∗ : S × S → S be a binary operation. An identity element for ∗
is an element e ∈ S such that
e∗a=a and a∗e=a
for all a ∈ S.
Exercise 7. Give examples of identity elements. Not all binary operations have
identity elements. Can you give examples of this as well?
Exercise 8. What are the identity elements (if any) for the logical operations of
Exercise 5.
Theorem 1. Let ∗ : S × S → S be a binary operation. If there is an identity for ∗,
then it is unique.
Proof. Suppose that e1 , e2 ∈ S are both identities and consider e1 ∗ e2 . Since e1 is
an identity, we have
e1 ∗ e2 = e2 .
Since e2 is an identity, we have
e1 ∗ e2 = e1 .
Thus
e1 = e2 .

6
Remark. From now on if we know that an element is an identity for a given binary
operation, then we can call this element the identity because of the above uniqueness
theorem.
Identities are unique, but left or right identities are not necessarily unique. We
begin with the definition of left and right identities.
Definition 5. Let ∗ : S × S → S be a binary operation. A left identity element
for ∗ is an element e ∈ S such that

e∗a=a

for all a ∈ S.

Definition 6. Let ∗ : S × S → S be a binary operation. A right identity element

for ∗ is an element e ∈ S such that

a∗e=a

for all a ∈ S.
Remark. Observe that, by definition, every identity element is a left identity and a
right identity. In fact our definitions make it clear that an element is an identity if
and only if it both a left-identity and a right identity.
Exercise 9. Suppose that ∗ : S × S → S is given by a operation table. Suppose
that e1 ∈ S. What has to be true about the e1 row in order for e1 to be a left
identity? Suppose e2 ∈ S. What has to be true about the e2 column for e2 to be a
right identity?
What has to be true about rows and columns for e to be a (two-sided) identity?

Exercise 10. Define a binary operation with an operation table so that the op-
eration has exactly three distinct left-identities. This shows that left identities are
not unique. Does your operation have any right identities? Is it commutative?

Exercise 11. Suppose that ∗ : S × S → S is commutative. Show that every left

identity is automatically an identity (so left identities are unique in this case since
identities are unique).

As seen above, it is possible in a noncommutative binary operation to have

more than one left identity. In this case there will be no right identities. In fact,
the following theorem gives uniqueness whenever there exists both left and right
identities (even if at first we do not know if the right and left identities are the
same element).
Theorem 2. Let ∗ : S × S → S be a binary operation. Suppose e1 ∈ S is a left
identity and e2 ∈ S is a right identity. Then e1 = e2 . In particular, e1 and e2 are
the unique identity for ∗.

Proof. Use the same type of argument as used for Theorem 1. (In fact, we could
have started with the above theorem, and then made Theorem 1 a corollary).

7
6 Monoids
Definition 7. A monoid M is a set with a given choice of operation ∗ : M ×M → M
such that (1) ∗ is associative, and (2) ∗ has an identity element in M .
Definition 8. A commutative monoid M is a monoid whose binary operation is
commutative.
Example 5. We have many examples of monoids:

• The natural numbers N is a commutative monoid under addition. In this case

the unique identity is 0. (In this document, we accept 0 as a natural number.)
• The natural numbers N is a commutative monoid under multiplication. In
this case the unique identity is 1.
• The set of functions F(Xn ) is a monoid under composition. What is the
identity?
• The set of permutations Sn is a monoid under composition. What is the
identity?
• The integers Z form a commutative monoid under addition. What is the
identity?
• The integers Z form a commutative monoid under multiplication. What is
the identity?
• The set Q is a commutative monoid under addition. What is the identity?
• The set Q is a commutative monoid under multiplication. What is the iden-
tity?
• The set R is a commutative monoid under addition. What is the identity?
• The set R is a commutative monoid under multiplication. What is the iden-
tity?
• The set Mn,m (R) of commutative matrices is a commutative monoid under
addition. What is the identity?
• The set Mn (R) of matrices is a monoid under multiplication. What is the
identity?
• The set C is a commutative monoid under addition. What is the identity?
• The set C is a commutative monoid under multiplication. What is the iden-
tity?
• The set {T, F } is a commutative monoid under the “and” operation. What
is the identity?
• The set {T, F } is a commutative monoid under the “or” operation. What is
the identity?

8
 • The set {T, F } is a commutative monoid under the “exclusive or” operation.
What is the identity?
• The set Zn of integers modulo n is a commutative monoid under addition.
What is the identity?
• The set Zn of integers modulo n is a commutative monoid under multiplica-
tion. What is the identity?
Remark. Two common (infix) notations for monoids are multiplicative notation
(using juxtaposition or some sort of product sign for the binary operation) and
additive notation (using + for the binary operation). In multiplicative notation 1
is a common notation for the unique identity. In additive notation 0 is the common
notation for the identity element. Sometimes, for example in logic, we might want
to avoid either of these standard conventions.
Remark. You may also see the term “semigroup”. A semigroup S is a set with a
given choice of operation ∗ : S × S → S such that ∗ is associative. Note that every
monoid is a semigroup, but it is easy to find examples of semigroups that are not
monoids.
Remark. A monoid has two pieces of data: (i) the underlying set M and (ii) the
binary operation ∗ on M . We sometimes write this package (or “structure”) as an
ordered pair hM, ∗i. If you change either component, you have a different monoid.
For example hZ, +i is distinct from hN, +i. Likewise, hZ, +i is distinct from hZ, ·i.
If the operation on M is clear from context, we sometimes denote the
monoid hM, ∗i simply as M . In other words, it is common to use the name of
the underlying set as also the name for the monoid as well.

7 Inverses
Warning: we cannot talk about inverses until we first have an identity element. So
in this section we will always assume we have an identity element; often we will
work in a monoid where we also have associativity.
Definition 9. Let ∗ : M × M → M be a binary operation with an identity ele-
ment e ∈ M . Let a ∈ M . We say that a is invertible if there is an element b ∈ M
such that
a ∗ b = b ∗ a = e.
Theorem 3. Let M be a monoid with identity element e ∈ M . If a ∈ M is
invertible then there is a unique element b ∈ M such that
a ∗ b = b ∗ a = e.
Proof. Suppose there are two such elements b1 , b2 ∈ M . Consider the resulting
elements b1 ∗ (a ∗ b2 ) and (b1 ∗ a) ∗ b2 in M . We have
b1 ∗ (a ∗ b2 ) = b1 ∗ e = b1 .
and
(b1 ∗ a) ∗ b2 = e ∗ b2 = b2 .
The associativity law now implies that b1 = b2 .

9
Definition 10. Let M be a monoid with identity element e ∈ M . Suppose a ∈ M
is invertible. Then the inverse of a is defined to be the unique element b ∈ M such
that
a ∗ b = b ∗ a = e.
In additive notation (where the operation is written +), we write −a for the inverse
of a. If we are not using additive notation, we usually write a−1 for the inverse
of a.

Exercise 12. Suppose that ∗ : M × M → M is given by a operation table. What

do you have to check about the table to verify that b ∈ M is the inverse of a ∈ M ?

Exercise 13. When talking about inverses, we usually want S to be a monoid (so
that inverses are unique). But we can define the notion of an inverse, even if S
is not a monoid: all we need is that S has an identity. However, inverses are not
necessarily unique when S is not a monoid. Give an example of non-uniqueness
of inverses. Hint: write an operation table for a set S = {e, a, b, c} where e is an
identity, but where b and c are are both inverses of a.
Exercise 14. There is a notion of left inverse and right inverse. In this case you
do not always have uniqueness even in a monoid. Give the definition of left inverse.
Give the definition of right inverse.
Exercise 15. Assume that a ∈ M where M is a monoid. Show that if a has a left
inverse, and that a has a right inverse, then the left inverse is equal to the right
inverse and a is invertible.

Here are some basic laws about inverses. (The proofs are straightforward):
Theorem 4. Suppose e ∈ M is the identity element of a monoid. Then e is
invertible, and its inverse is itself:

e−1 = e.

Theorem 5. Suppose c ∈ M is an invertible element of a monoid. Then c−1 is

invertible, and
(c−1 )−1 = c.

Corollary 6. Suppose c ∈ M is an invertible element of a monoid with c−1 = d.

Then d is invertible with d−1 = c.

Theorem 7. Suppose c, d ∈ M be invertible elements of a monoid. Then c ∗ d is

invertible, and
(c ∗ d)−1 = d−1 ∗ c−1 .

Definition 11. We sometimes call invertible element in a monoid units.

We can rephrase Theorems 4, 5, and 7 in terms of units:

Theorem 8. Suppose M is a monoid. Then the identity element is a unit, the
inverse of a unit is a unit, and the product of units is a unit (where here we are
using multiplicative notion for convenience).

10
8 Cancellation laws
Theorem 9 (Left-cancellation Law). Suppose M is a monoid and c ∈ M is in-
vertible. Suppose that
c∗x=c∗y
with x, y ∈ M . Then
x = y.

Proof. Since c ∗ x = c ∗ y we get the equation

c−1 ∗ (c ∗ x) = c−1 ∗ (c ∗ y).

By the associative law, we can rewrite this equation as

(c−1 ∗ c) ∗ x = (c−1 ∗ c) ∗ y.

By the definition of inverses, we can write this equation as

e ∗ x = e ∗ y.

where e ∈ M is the identity. By the definition of identity, we get

x = y.

Similarly we have the following:

Theorem 10 (Right-cancellation Law). Suppose M is a monoid, and c ∈ M is
invertible. Suppose that
x∗c=y∗c
with x, y ∈ M . Then
x = y.
Remark. For the left-cancellation law, it is enough that c is left-invertible. For the
right-cancellation law, it is enough that c is right-invertible.
Example 6. In the monoid hR, +i, every element is invertible (it turns out to be a
“group”). So the cancellation law for addition holds for all c ∈ R:

c + x = c + y =⇒ x = y.

However, in the monoid hR, ×i, only the nonzero elements are invertible. So the
cancellation law
cx = cy =⇒ x = y
only applies for c 6= 0.

9 Groups
Definition 12. A group is a monoid such that every element is invertible.

11
Definition 13. An Abelian group is a group whose operation is commutative. In
other words, it is a commutative monoid such that every element has an inverse.
We use the term “Abelian” in honor of the mathematician Abel (1802–1829).2

Example 7. Some examples:

• The set of natural numbers N is not a group under addition nor is it a group
under multiplication.
• The monoid of integers Z under addition is an Abelian group, but the corre-
sponding monoid Z under multiplication is not a group (since 2, for example,
has no inverse).
• The set of permutations Sn is a group under composition. The inverse of a
permutation α ∈ Sn is the inverse function α−1 . This group is non-Abelian if
n > 2. (In contrast, the monoid of functions F(Xn ) is not a group if n > 1).
• The set Mn,m (R) of matrices is a group under addition. Describe inverses. Is
it Abelian?
• The set Mn (R) of matrices is not a group under multiplication. However,
the subset GLn (R) of matrices with determinate not equal to zero is a group.
The inverse of A is the matrix A−1 . If n > 1 then GLn (R) is a non-Abelian
group.
• The set R is a group under addition, but not under multiplication. However,
if we remove 0 the set that remains R× = R − {0} is a group.
• The set Q is a group under addition, but not under multiplication. What
element (or elements) of Q under multiplication do not have inverses? What
if you remove just 0 from Q?
• The set C is a group under addition, but not under multiplication. If you
remove 0 from C you do get a group under multiplication.
• The set {T, F } is a group under the “exclusive or” operation. Observe that
the inverse of any element is itself. (Are there any other logical operators
on {T, F } for which we get a group?)
• The set Zn of integers modulo n is a group under addition, but not under
multiplication. Is it Abelian? Is there a connection between Z2 under addition
and {T, F } under exclusive or?
• Let p be a prime. Then the set Z×
p of integers modulo p not congruent to zero
turns out to be an Abelian group under multiplication with p − 1 elements.

For groups we have the cancellation law:

Theorem 11. Let c ∈ G where G is a group. If c ∗ x = c ∗ y with x, y ∈ G,
then x = y. Similarly, if x ∗ c = y ∗ c with x, y ∈ G, then x = y.
2I capitalize “Abelian” to honor Abel, but many authors write “abelian” with a lower-case ‘a’.

12
Proof. This follows from cancellation laws for monoids.

Theorem 12. If G is a group with identity element e then

e−1 = e

For all c ∈ G
(c−1 )−1 = c.
For all c, d ∈ G
(c ∗ d)−1 = d−1 ∗ c−1 .
If G is an Abelian group then for all c, d ∈ G

(c ∗ d)−1 = c−1 ∗ d−1 .

Proof. This follows from the inverse laws for monoids.

Recall that if the operation of a monoid M is written additively using +, then

it is customary to write the identity element as 0 and the inverse of a ∈ M as −a,
if it exists. We can rephrase various results using additive notation. For example,
the above theorem can be stated as follows:
Corollary 13. If G is a group with additive notation then

−0 = 0

For all c ∈ G
−(−c) = c.
For all c, d ∈ G
−(c + d) = (−d) + (−c)
If G is an Abelian group then for all c, d ∈ G

−(c + d) = (−c) + (−d).

Remark. When we use additive notation we are usually working in an Abelian

group or at least a commutative monoid.
Definition 14. Let G be a group with additive notation. Given a, b ∈ G we define
the subtraction operation as follows:
def
a − b = a + (−b).

Corollary 14. If G is an Abelian group with additive notation then for all c, d ∈ G

−(c + d) = (−c) − d.

Exercise 16. Suppose M is monoid such that every element has a left-inverse.
Show that M is a group.

13
9.1 The One-Side Inverse Shortcut
According to the definition of inverse we must show both a ∗ b = e and b ∗ a = e in
order to conclude that b is the inverse of a. If we are working in a group then this
is overkill, and there is a short-cut: we only need to show one of the two equations.
(Of course if ∗ is commutative one equation is enough; but the point here is that
in a group G one equation is enough even if G is non-Abelian.)
Theorem 15. Let G be a group with identity element e. If
a∗b=e
where a, b ∈ G then
b∗a=e
and so
b = a−1 and a = b−1 .
Proof. Note that
a ∗ b = e = a ∗ a−1 .
By the cancellation law, we get
b = a−1 .
In particular,
b ∗ a = a−1 ∗ a = e
and so
a = b−1 .

10 Translation Functions
Think of R2 as the collection of vectors in the plane under the operation of vector
addition. If you fix a vector (c1 , c2 ) then the function
(x, y) 7→ (c1 , c2 ) + (x, y) = (c1 + x, c2 + y)
corresponds to the geometric idea of translation: each point in R2 maps to its
translation by the vector (c1 , c2 ). This operation is used to move points, and so
subsets of R2 , in a way that preserves lengths and angles and so on. In particular,
this map is an example of what we call an “isometry”.
We can generalize this translation map to any monoid hM, ∗i. To do so fix an
element c ∈ M and define a function Tc : M → M by the rule
def
Tc (x) = c ∗ x.
We call this left-translation by c. We define right-translation in a similar manner.
Of course if the monoid is commutative, as in the case of R2 under addition, left and
right translations by c ∈ M agree. If unspecified, a translation will be understood
as left-translation.
The following two results are straightforward:

14
Lemma 16. Let e be the identity in the monoid M . Then the translation func-
tion Te is the identity function M → M .

Lemma 17. Let Ta and Tb be translation functions M → M associated to ele-

ments a, b ∈ M where hM, ∗i is a monoid. Then the composition Ta ◦ Tb : M → M
is also a translation function. In fact

Ta ◦ Tb = Ta∗b .

Exercise 17. Show that had we defined Ta and Tb as right-translations, we would

get Ta ◦ Tb = Tb∗a instead.

We can use the above two Lemmas to prove the following:

Lemma 18. Suppose that c ∈ M is an invertible element of a monoid M . Then
the translation functions Tc and Tc−1 are inverse functions to each other. In other
words the compositions Tc ◦ Tc−1 and Tc−1 ◦ Tc are both the identity function. We
can express this fact nicely with the identity

Tc−1 = Tc−1 .

Corollary 19. Suppose that c ∈ M is an invertible element of a monoid M . Then

the translation functions Tc defined by Tc (x) = c ∗ x is a bijection.
We can apply this result to operation tables of finite monoids M :

Theorem 20. Suppose hM, ∗i is a finite monoid with invertible element c. Sup-
pose M is ordered once and for all, and consider the associated operation table
for ∗. Then the row associated to c has every element of M appearing exactly once.
Similarly the column associated to c has every element of M appearing exactly once.

Proof. Suppose a1 , a2 , . . . , an are the elements of M listed in order. Then the row
associated to c consists of the following finite sequence of elements:

c ∗ a1 , c ∗ a2 . . . , c ∗ an .

Think of this sequence as

Tc (a1 ), Tc (a2 ), . . . , Tc (an ).

Every element appears at most once because Tc is injective. Every element appears
at least once because Tc is surjective. The argument for the c-column is similar.

11 Exponentiation in a Monoid
Throughout this section, let M be a monoid with operation ∗ and identity element e.
For most of what follows we will employ multiplicative notation, allowing us to
write a ∗ b simply as ab. Our goal is to introduce powers au where n ∈ N and,
in fact, when a is invertible for all n ∈ Z. (We translate all the main results to
additive notation later on where we write na).

15
 Let a ∈ M be an element of the monoid. Informally, if n ∈ N then we define an
to be
a ∗ a ∗ ··· ∗ a
where a is repeated n times. If a ∈ M is invertible, we define a−n for negative
integers −n < 0 by defining a−n either as (a−1 )n or (an )−1 . If one follows this path,
then one can derive the usual laws of exponentiation, even for negative powers (if
a is invertible). However, proving them often requires a bunch of different cases
which depends on the sign (positive or negative) of each of the exponents involved
in the given law.
We take another approach to exponentiation using iteration of translation func-
tions. This approach treats positive and negative powers in a more unified manner.
Our plan is to iterate the translation function x 7→ ax starting with x = e yielding
the sequence
e a ∗ e a ∗ a ∗ x a ∗ a ∗ a ∗ e ....
The nice thing about this approach is that we can use negative iteration when the
translation function is invertible. In what follows I will take the concept of iteration,
and some of the basic results for iterated functions, as part of the set-theoretical
background for abstract algebra.3 Here are three propositions summarizing what
we need:

Proposition 21. Suppose f : S → S is a function from a set S to itself. Then

for any nonnegative integer n ∈ N we have the nth iteration f n of f . Properties of
iteration for such a function f : S → S include the following:

• For all n ∈ N, the iteration f n is a function S → S.

• The zeroth iteration f 0 is the identity map S → S.
• The first iteration f 1 is just f : S → S itself.
• For m, n ∈ N, we have f m ◦ f n = f m+n . In particular, for n ∈ N we
have f n+1 = f n ◦ f = f ◦ f n .
• For m, n ∈ N, we have (f m )n = f mn .
Proposition 22. Suppose f : S → S is a bijection from a set S to itself. Then
we can extend the definition of interation f u to all integers u ∈ Z. Properties of
iteration for such a bijection f : S → S include the following:
• For all u ∈ Z, the iteration f u is also a bijection S → S.
• The iteration f −1 is, as the notation suggests, the inverse of f .
• For u, v ∈ Z, we have f u ◦ f v = f u+v .
• For u, v ∈ Z, we have (f u )v = f uv .
• If f is the identity function, then f u is the identity function for all u ∈ Z.
3 See my number systems textbook for a formal account of iteration and a rigorous development
of the properties of iteration.

16
Proposition 23. Suppose f, g : S → S are commuting functions. In other words,
suppose f ◦ g = g ◦ f . Then

(f ◦ g)n = f n ◦ g n

for all n ∈ N. If in addition f and g are both bijections S → S, then

(f ◦ g)u = f u ◦ g u

for all u ∈ Z.
With this we are ready for the formal definition of exponentiation. Recall that
from Corollary 19 that the translation function Ta is a bijection if a is invertible.
Definition 15 (Exponentiation in a multiplicative monoid). Let M be a monoid
that employs multiplicative notation. If n ∈ N then
def
an = Tan (e).

where e ∈ M is the identity element, and where Ta is the translation map x 7→ a∗x.
If a is invertible in M , then we extend the above definition for all u ∈ Z:
def
au = Tau (e)

(This is well-defined since Ta is a bijection in this case).

Theorem 24. Suppose that a ∈ M where M is a multiplicative monoid. Then

a0 = e

where e ∈ M is the identity element, and

a1 = a.

If a is invertible in M then a−1 is the inverse of a in M .

Proof. Observe that a0 = Ta0 (e) = e since Ta0 is the identity map (Proposition 21).
Next, by Proposition 21 and the above definition,

a1 = Ta1 (e) = Ta (e) = ae = a.

Finally, suppose a is invertible, and let b be the inverse of a in M . By Lemma 18,

the translation Tb is the inverse of Ta , so Ta−1 = Tb by Proposition 22 (where Ta−1
here mean the −1 ∈ Z iteration of Ta ). Thus

a−1 = Ta−1 (e) = Tb (e) = be = b

where a−1 means the −1 power of a (not, a priori, the inverse of a which we denote
simply as b).

Remark. The last part of the above shows that both meanings of a−1 agree: as an
inverse and as a negative power.

17
 To prove additional properties of exponentiation, we will need a few lemmas
about translations functions. The first states that the iteration of translation is a
translation.

Lemma 25. Consider the iteration Tau of the translation Ta where a ∈ M and
where u ∈ N or more generally, if a is invertible, where u ∈ Z. Then there is an
element c ∈ M such that Tan = Tc .

Proof. Fix a ∈ M . First we show that the claim holds when u ∈ N by induction.
The base case holds since Ta0 = Te where e is the idenitity. Suppose that Tan = Tc
for a particular n ∈ N and c ∈ M . Then

Tan+1 = Tan ◦ Ta = Tc ◦ Ta = Tca . (Prop 21 and Lemma 17)

Thus the claim holds for n+1 with ca as the associated element of M . By induction
we accept the claim for u ≥ 0.
We have established the claim for for all a ∈ M and all u ≥ 0, and wish to
establish the claim for negative u. So let a ∈ M be invertible with inverse b ∈ M ,
and suppose u = −n for some n ∈ N. We have Tb = Ta−1 by Lemma 18. So, using
the identity (f u )v = f uv from Proposition 22,

n
Tau = Ta−n = Ta−1 = Tbn .

We have established that Tbn = Tc for some c ∈ M . Thus Tau = Tc as desired.

We can actually be more specific about what c is in the previous lemma.

Lemma 26. Suppose that a ∈ M . Then

Tau = Tau

for all u ∈ N. If a is invertible, then this holds more generally for all u ∈ Z.

Proof. By Lemma 25, Tau = Tc for some c ∈ M . Apply Tau and Tc to the identity
element e ∈ M :
def
au = Tau (e) = Tc (e) = ce = c.

Now we are ready to prove some important identities.

Theorem 27. Let a ∈ M where M is a multiplicative monoid. If m, n ∈ N then

am+n = am an .

More generally, if a is invertible in M and if u, v ∈ Z then

au+v = au av .

18
Proof. Observe that

Tam+n = Tam+n (Lemma 26)

= Tam ◦ Tan (Proposition 21)
= Tam ◦ Tan (Lemma 26)
= T am an (Lemma 17)

From Tam+n = Tam an we get

am+n = Tam+n (e) = Tam an (e) = am an .

The above proof is for m, n ∈ N. To generalize to u, v ∈ Z, replace m with u,

replace n with v, and replace the reference to Proposition 21 with Proposition 22
in the above argument.

We are not assuming that M is a commutative monoid. However, we have the

following:
Corollary 28. Let a ∈ M . Then a power of a commutes with any power of a.

Proof. This follows since addition is commutative in Z.

Theorem 29. Let a ∈ M where M is a multiplicative monoid. If m, n ∈ N then

(am )n = amn .

If a ∈ M is invertible and if u, v ∈ Z then

(au )v = auv .

Proof. We have Tam = Tam (Lemma 26), so

(Tam )n = (Tam )n = Tamn

where the last equality is based on a property of iterations (Proposition 21). This,
together with Definition 15, yields

(am )n = (Tam )n (e) = Tamn (e) = amn .

To generalize to u, v ∈ Z, when a is invertible, replace m with u, n with v, and

the reference to Proposition 21 with Proposition 22 in the above argument.

Theorem 30. Let a ∈ M where M is a monoid. If a is invertible, then so is au

for all u ∈ Z.

Proof. Observe that a−u is the inverse of au by previous exponentiation laws.

In a commutative monoid we have that (ab)n = an bn . We will show this, but

first we need a simple lemma.
Lemma 31. Let a, b ∈ M . Suppose ab = ba (which holds, for example, when M is
a commutative monoid). Then Ta and Tb commute as functions: Ta ◦ Tb = Tb ◦ Ta .

19
Proof. This follows from Lemma 17 since Tab = Tba in this case.

Theorem 32. Let a, b ∈ M in a multiplicative monoid. If ab = ba (which is true,

for example, if M is a commutative monoid or an Abelian group) then for all n ∈ N

(ab)n = an bn .

If in addition a and b are invertible, then so is ab, and for all u ∈ Z

(ab)u = au bu .

Proof. By Lemma 31, Ta and Tb commute. So

n
Tab = Ta ◦ Tb )n (Lemma 17)
= Tan ◦ Tbn (Prop. 23)
= Tan ◦ Tbn (Lemma 26)
= Tan bn (Lemma 17)

This, together with Definition 15, yields

def
n
(ab)n = Tab (e) = Tan bn (e) = an bn e = an bn .

If a and b are invertible, then ab is invertible since it has inverse b−1 a−1 . The
above argument for n ∈ N generalizes to u ∈ Z in this case.

Theorem 33. Let e ∈ M be the identity element in a multiplicative monoid.

Then eu = e for all u ∈ Z.

Proof. We know that Te is the identity map by Lemma 16. So, by Proposi-
tion 22, Teu is also the identity map. Thus
def
eu = Teu (e) = e.

11.1 Exponentiation in Additive Notation

When M is a monoid or group written in additive notation, we use a different
notation for exponentiation. The results are the same, but are just expressed in a
different notation.
When M is a monoid with its binary operation written as + then the identity
element is generally written as 0. For n ∈ N and a ∈ M we write na for the additive
version of exponentiation. Similarly if a is invertible and u ∈ Z, we write ua for
exponentiation. All the results of the previous section hold for M , but one just has
to translate to additive notation. For the convenience of the reader, we summarize
the main results when the monoid is a group.
Theorem 34 (Exponentiation in an additive group). Suppose A is a group written
in additive notation. Suppose a ∈ A. Then we have the following

20
 • 0 a = 0. (Here the first 0 is in Z and the second is in G).
• 1a = a
• (−1)a = −a
• (u + v)a = ua + vb for all u, v ∈ Z
• (uv)a = u(va) for all u, v ∈ Z
• ua commutes with va for all u, v ∈ Z.

If a, b ∈ G commute, a + b = b + a, then
• u(a + b) = ua + ub for all u ∈ Z

Finally,

• u 0 = 0 for all u ∈ Z (here 0 is in G)

Remark. We use the normal elementary algebra notational conventions for supress-
ing parenthesis. For example, the expression ua + vb really means (ua) + (vb) (and
not u(a + (vb)) say).
Remark. Most groups that use additive notation are Abelian, so the law

u(a + b) = ua + ub

holds for all u ∈ Z and a, b ∈ G.

12 The Induced Operation on Closed Subsets

Definition 16. Let ∗ : S × S → S be a binary operation. If A is a subset of S then
we say that A is closed under ∗ if the following holds:

∀a, b ∈ A, a ∗ b ∈ A.

Example 8. Consider N under addition, and let A be the subset of even numbers
greater than 11. Then A is closed under +. Let B be the set of odd natural
numbers. Then B is not closed under + since 7 + 3 is not in B.

Definition 17. Let ∗ : S × S → S be a binary operation. Let A ⊆ S be a subset

that is close under ∗. Then we define a binary operation on ∗A : A × A → A using
the following rule: if a, b ∈ A then a ∗A b is defined to be a ∗ b. This operation is
called the induced operation on A.

We warn the reader that the term induced operation can refer to different things
depending on context. The above is what is meant when A is a subset of S. There
are also induced operation corresponding to sets of functions, or sets of cosets. The
common idea is that an induced binary operation closely matches, in some sense, a
previously given binary operation, but often has a different, albeit related, domain
or codomain.

21
 The induced operation of Definition 17 is often thought of as the operation
obtained by restricting the operation ∗ from S to the subset A. The assumption
that A is closed plays a key role here. It is always legal to restrict the domain
of a function. So we can always restrict ∗ : S × S → S to A × A to obtain a
function A × A → S. It is not always legal, however, to restrict a codomain. You
can only restrict to a smaller codomain if the this proposed codomain contains the
image. So we can only restrict from A × A → S to A × A → A if the image of the
restriction A × A → S is a subset of A. This is exactly the requirement that A
is closed under ∗: when we say “A is closed” we just mean that the image of ∗
restricted to A × A is contained in A.
By definition we have the law
a ∗A b = a ∗ b
for all a, b ∈ A. So ∗ and ∗A look like the same operation; the only real difference
is the domain and codomain. The bottom line is that, as long as a, b ∈ A, the
expressions a ∗ b and a ∗A b are completely interchangeable.
The operation ∗A : A × A → A is sometime called the restriction of ∗ to A. We
also say that the operation S × S → S an extension of the operation A × A → A
to S. It is common practice to use the same symbol for both operations S × S → S
and A × A → A. When we do so, we rely on context to tell us which is meant
by the symbol in any given usage. (When applied to elements in A, both notions
give the same result, so one can be be safely ambiguous. For example, if A = N
and S = Z then 7 + 3 is 10 whether or not you are thinking of + : Z × Z → Z or
the induced operation + : N × N → N.)
Lemma 35. Let ∗ : S × S → S be a binary operation. Let A ⊆ S be a subset that
is close under ∗.
If ∗ : S × S → S is associative, then the induced operation A × A → A is also
associative.
If ∗ : S × S → S is commutative, then the induced operation A × A → A is also
commutative.
Proof. We will prove the claim related to the associative law. The proof of the
commutative claim is similar. So let a, b, c ∈ A. We have
def
(a ∗A b) ∗A c = (a ∗ b) ∗ c
by definition of ∗A . Similarly, we have
def
a ∗A (b ∗A c) = a ∗ (b ∗ c).
Since a, b, c ∈ A, and since A is a subset of S, we have a, b, c ∈ S. Since ∗ is
associative on S by assumption, we have
(a ∗ b) ∗ c = a ∗ (b ∗ c).
Putting these equations together we get the desired equation:
(a ∗A b) ∗A c = a ∗A (b ∗A c).

22
 From this we get the following:
Theorem 36. Suppose M is a monoid with operation ∗ and identity element e.
If A is a subset closed under ∗ and if A contains e then A is itself a monoid under
the induced operation. The identity of A with the induced operation is the identity
of M .

13 Submonoids
Informally a submonoid is a subset A of a monoid M that is itself a monoid under
the induced operation. This will only make sense if the subset A is closed under
the operation of M , so this will always be required. One question is whether we
want the identity of A to be the same element as the identity of M . Based on the
answer to this question we have two notions of submonoid. Here we will pursue
the notion where M and A have the same identity element: this will be a second
requirement.4 As we saw in Theorem 36, these two requirements suffice. So we
define submonoid as follows:
Definition 18. Suppose hM, ∗i is a monoid with identity element e. A submonoid
is a subset A ⊆ M such that (1) A is closed under ∗ and (2) e ∈ A. As we saw above,
every submonoid is itself a monoid under the induced operation with identity e.
Example 9. The set of even integers forms a submonoid of hZ, +i. The set of odd
integers fails to be a submonoid. If A is the set of even numbers greater than 11
then A is a closed subset of Z under addition, but is not a submonoid since it does
not contain 0. (And in fact, this A has no identity element).
We restate Theorem 36:

Theorem 37. Suppose A is a submonoid of a monoid M . Then A is itself a

monoid under the induced operation.

The following is clear:

Theorem 38. Every submonoid of a commutative monoid is itself a commutative

monoid (under the induced operation).
Exercise 18. Consider the monoid hZ10 , +i. Let A be the submonoid {0, 2, 4, 6, 8}.
Make an operation table for A using the induced operation +A .
Theorem 39. Suppose M is a monoid and that B is a submonoid of M . Let A be
a subset of B. Then A is a submonoid of M if and only if A is a submonoid of B.

Proof. Let e be the identity element of M . Since B is a submonoid, e is the identity

of B. Let ∗ be the binary operation on M . Then, as we have seen, the induced
operation ∗B is the binary operation on the submonoid B.
4 The set M = N × N under multiplication is a monoid under componentwise products with
identity (1, 1). The subset A = N × {0} is a monoid under the induced operation with identity
element (1, 0). However, by our official definition, A is not a submonoid of M since the identity
of A is different from that of M .

23
 Suppose that A is a submonoid of M . Then e ∈ A. But e is the the identity
of B. Suppose a, b ∈ A. Then a ∗B b = a ∗ b since ∗B is the induced operation.
Since A is a submonoid of M , the element a ∗ b must be in A. Thus a∗B b is in A.
We conclude that A is closed in B. We have established both conditions needed
for A to be a submonoid of B.
Suppose that A is a submonoid of B. Then the identity e of B is in A. So
the identity of M is in A since e is also the identity of M . Suppose a, b ∈ A.
Then a ∗B b = a ∗ b since ∗B is the induced operation. Since A is a submonoid
of B, the element a ∗B b must be in A. Thus a ∗ b is in A. We conclude that A is
closed in M . We have established both conditions needed for A to be a submonoid
of M .

The following lemmas give compatibility between notions in a monoid and in a

submonoid:

Lemma 40. Suppose A is a submonoid of M . If a ∈ A has an inverse b in the

monoid A, then b is the inverse of a in M as well.
Similarly, if a has inverse b in M , and if a, b ∈ A, then a has inverse b in A.

Proof. Let e be the identity of M , which is also the identity of A since A is a

submonoid.
Suppose a ∈ A has inverse b in A. By definition of inverse a ∗A b = b ∗A a = e.
But a ∗A b = a ∗ b and b ∗A a = b ∗ a since a, b ∈ A. Thus a ∗ b = b ∗ a = e in M .
Now suppose a has inverse b in M , so a ∗ b = b ∗ a = e. Suppose a, b ∈ A.
Then a ∗A b = a ∗ b and b ∗A a = b ∗ a. Thus a ∗A b = b ∗A a = e.

Lemma 41 (Multiplicative Version). Suppose A is a submonoid of M . Let a ∈ A.

Then an as defined in M is the same as an as defined in A where here n ∈ N. In
particular A is closed under powers: an ∈ A for all a ∈ A and n ∈ N where an is
as defined in M .
If in addition a is invertible in A then the above extends to all n ∈ Z.

Proof. This follows by induction for n ∈ N. The more general statement for a
invertible follows from the properties of exponents.

Exercise 19. Suppose that M is a multiplicative monoid. Let a ∈ M . Show

that A = {an | n ∈ N} is a submonoid of M . Show that this A is the smallest
submonoid containing a in the sense that any submonoid of M containing a must
contain A as a submonoid.
Show that if a ∈ M is invertible, then A0 = {au | u ∈ Z} is also a submonoid, and
that in A0 is an Abelian group. Show that it is the smallest submonoid containing a
that is a group.

14 The Unit Group

In this section hM, ∗i will be a monoid with identity e. There is a submonoid of M
that is of particular interest:

24
Definition 19. Let hM, ∗i be a monoid. Then M inv is defined to be the set of
elements of M that are invertible. Elements of M inv are called units. As we will
see, M inv forms a group under the induced operation, and we call M inv the unit
group of M .

The following is just a restatement of the definition of group:

Theorem 42. A monoid M is a group if and only if M inv = M .

Example 10. (Availing ourselves of some linear algebra). Let M = Mn (R) be the
set of n-by-n matrices under matrix multiplication. This is a monoid. The unit
group M inv consists of invertible matrices is a group. It is an important group in
mathematics called the general linear group, and is often denoted as GLn (R). As
we learn in linear algebra, GLn (R) is the set of n-by-n matrices with determinant
not equal to zero.
Example 11. (Assuming familarity with rings) Let M = R be a ring. If we ignore
addition, and use multiplication as the operation, we get a monoid. In this case Rinv
consists of all elements in R with multiplicative inverses. We sometimes write this
subset as R× , and call this the group of units of R. Note that if F is a field, then F ×
is F − {0}.

Exercise 20. Consider M = N, M = Z, and M = Q under multiplication. What

is M inv in each of these cases? Note that we get a group under multiplication in
each case.
Now do the same exercise for addition: identify M inv in each of these cases.
As mentioned above we will determine that M inv is a group. First we convince
ourselves that M inv is a submonoid, and hence M inv is itself a monoid.

Lemma 43. Let hM, ∗i be a monoid. Then M inv is a submonoid.

Proof. The follows from Theorem 8.

Theorem 44. Let hM, ∗i be a monoid. Then M inv is a submonoid of M that is

in fact a group under the induced operation. Furthermore, M inv is the largest such
group in the following sense: if G is a submonoid of M that is a group under the
induced operation, then G ⊆ M inv .

Proof. Since M inv is a submonoid, it is in fact a monoid under the induced oper-
ation. Does every element of M inv have an inverse in M inv ? By definition, every
element of M inv has an inverse in M , but Theorem 8 tells us more: the inverse is
guaranteed to be in M inv itself.
Suppose G is a submonoid of M that is a group under the induced operation.
Then every element of G is a unit in M (see Lemma 40). Thus G ⊆ M inv .

Exercise 21. If hM, ∗i be a commutative monoid, then can you conclude that M inv
is an Abelian group under the induced operation?
Exercise 22. Find the group of invertible elements in hZ8 , ×i. Make an operation
table for this group.

25
15 Subgroups
Suppose hG, ∗i is a group with identity element e. Some subsets A ⊆ G will be
groups under the induced operation, but others will not. If we want A to be a
group under the induced operation, we will certainly need A to be closed under ∗.
If we want A and G to share the same identity element (as we did with monoids),
we would want A to contain e, and then we would need A to be “closed under
inverses” as well in the sense that a−1 ∈ A for all a ∈ A where here a−1 is the
inverse of a in G. The following shows that these properties are indeed necessary
and sufficient:
Lemma 45. Suppose hG, ∗i is a group with identity element e, and that A ⊆ G
is a subset closed under ∗. Then A is a group under the induced operation if and
only if (1) e ∈ A, in other words A is a submonoid, and (2) if a ∈ A then a−1 ∈ A
where a−1 is the inverse of a in G.

Proof. Suppose A is a group under the induced operation. Then A must contain
an identity element e0 . Note that

e0 ∗ e0 = e0 = e0 ∗ e.

So e0 = e by the cancellation law in G. So A is a submonoid of G. By assumption

each a ∈ A has an inverse in A. This inverse is a−1 (the inverse of a in G) by
Lemma 40. Thus (1) and (2) both hold.
Conversely suppose (1) and (2) hold. By (1) A is a submonoid of G, so is a
monoid under the induced operation. By (2) and Lemma 40 every element of A is
invertible. Thus A is a group.

The above lemma motivates the following definitions:

Definition 20. Let hG, ∗i be a group. We say that a subset A ⊆ G is closed under
inverses if the inverse of a is in A for all a ∈ A.
Definition 21. Suppose hG, ∗i is a group. A subgroup is a subset A ⊆ G such that
(1) A is closed under ∗, (2) the identity element of G is in A and (3) A is closed
under inverses.
By Lemma 45 together with Lemma 40 and Lemma 41 we have the following:
Theorem 46. If A is a subgroup of G, then A is itself a group under the induced
(restricted) operation. The identity of A is the identity in G. The inverse of an
element a ∈ A in A is equal to its inverse in G. For each a ∈ A and n ∈ Z the nth
power of a in A is equal to the nth power of a in G.
Exercise 23. Show that if A is a subgroup of an Abelian group, then A is itself
an Abelian group under the induced operation. (Hint: Lemma 35).
Remark. The idea of a subgroup makes it easy to construct examples of groups
without proving all the equations that must hold in a group: the associative law
equation, the identity law equations, and the inverse law equations. Now that
we have the above theorem, we get these equations for free when dealing with a
subgroup of a known group.

26
 If A is not a subgroup of a known group, and you want to show it is a group,
you have to do it the hard way and prove all the required equations (including
associativity, identity equations, and inverse equations).
There are some short-cuts to checking that A is a subgroup. The first allows
you to replace checking that the identity is in A with checking that A nonempty.
Proposition 47. Suppose hG, ∗i is a group. Then a subset A ⊆ G is a subgroup
if and only if (1) A is nonempty, (2) A is closed under ∗, (3) A is closed under
inverses.

Proof. First suppose that A is a subgroup. Then since the identity e is in A we

know that it is (1) nonempty. Conditions (2) and (3) also hold by the definition of
subgroup.
Conversely suppose (1), (2), and (3) hold. Since A is nonempty, it contains an
element a ∈ A. By (3) we have a−1 ∈ A. So by (2) we have a ∗ a−1 ∈ A. In other
words e ∈ A. This, together with (2) and (3), means that A is a subgroup by the
definition of subgroup.

The following gives you a way to check two conditions instead of three (we
use multiplicative notation for convenience, but it can be translated into additive
notation):

Proposition 48. Suppose hG, ∗i is a group. Then a subset A ⊆ G is a subgroup if

and only if (1) A is nonempty and (2) a ∗ b−1 ∈ A for all a, b ∈ A.

Exercise 24. Prove the above proposition. Hint: assuming (1) and (2), first show
that e ∈ A, then show A is closed under inverses. Finally show A is closed under ∗.

For finite groups the situation is easier and there is a nice shortcut. Actually
the enclosing group G does not even need to be finite, just the subset A.

Proposition 49. Suppose hG, ∗i is a group, and that A is a finite subset of G.

Then A is a subgroup of G if and only if (1) A contains the identity, and (2) A is
closed under ∗.

Proof. One direction is clear, so we focus on the other direction. Suppose a ∈ A.

Consider the translation function Ta : A → A given by the rule

Ta (x) = a ∗ x.

This function is well-defined since A is closed under ∗. This function is injective

by the cancellation law. Since A is finite, this function must be surjective as well
by properties of finite sets. Since e ∈ A, there must be an element b mapping to e.
Since Ta (b) = e we conclude that a ∗ b = e. Thus b is the inverse of a (Theorem 15).
Thus A is closed under inverses. This together with (1) and (2) gives the
result.

Corollary 50. Every finite submonoid of a group is a subgroup.

27
Exercise 25. Suppose hG, ∗i is a group, and that A is a finite subset of G. Suppose
that (1) A is nonempty and (2) A is closed under ∗. Show that the identity must
be in A. Conclude that A must be a subgroup.
In other words, (1) in the above Proposition can be replaced by “(1) A is
nonempty”. Hint: if a ∈ A, then all its powers are in A. Two powers with distinct
exponents are equal so the identity e is a power of a.
Exercise 26. Suppose M is a monoid with identity e, then show that M and {e}
are submonoids of M . Suppose G is a group with identity e. Then show that G
and {e} are subgroups of G.

16 Products and Sums of Finite Sequences

We will use the term finite ordered sequence in a set S for any family (ai )i∈I of
elements of S such that I is a finite, totally ordered set. We define an equivalence
relation for such finite ordered sequences: suppose (ai )i∈I and (bJ )i∈J are two finite
ordered sequences in S then we say that these finite ordered sequences are order
equivalent if (i) the cardinality of I and J are equal, and (ii) the unique order-
preserving bijection γ : I → J has the property that bγ(i) = ai for all i ∈ I. The
reader can verify the following:
Lemma 51. Order equivalence is an equivalence relation among finite ordered se-
quences in a set S.
The definition of order equivalence makes use of the following fact from set
theory (that can be proved by induction):
Lemma 52. If I and J are finite totally ordered sets of the same size, then there
is a unique order-preserving bijection I → J.
We say that a finite ordered sequence is nonempty if I is nonempty. We now
define the product of finite ordered sequences:
Definition 22. Fix Q a set S and a binary operation ∗ : S × S → S. Then we
define the product (ai )i∈I of any nonempty finite ordered Q sequence (ai )i∈I in S
recursively as follows. If I has one element, call it α, then (ai )i∈I is defined to
be just aα . If I has more than one element then let ω be the last element of I, and
let I 0 be I − {ω}. Then we define the product by the recursive equation
Y 
def
Y
(ai )i∈I = (ai )i∈I 0 ∗ aω

where (ai )i∈I 0 denotes the restriction of the given family to I 0 . Q

We will sometimes follow the common convention and write (ai )i∈I as
Y
ai .
i∈I

If I = [m, n]Z is the interval of integers from m to n (inclusive) then we can employ
the notation
Y n
ai
i=m

28
 Q
for the product (ai )i∈I . If there is no danger of ambiguity, we can use abbrevia-
tions for products such as
Y Y
ai or ai .
i

If the Q
operation
P is written + then we usually use additive notation where we re-
place with and use the term sum instead of product.
Definition 23. If M is a monoid or a group, then we extend the above definition
to include the empty sequence: we define the product of the empty finite ordered
sequence to be the identity element of M .
Theorem 53. Let S be a set with binary operation ∗ : S × S → S. If (ai )i∈I
and (bJ )i∈J are order equivalent nonempty finite ordered sequences, then
Y Y
(ai )i∈I = (bj )i∈J .

(If S = M is a monoid, then this extends to the empty sequence.)

Proof. A straightforward induction argument sufficies.

17 The General Associativity Laws

Now we consider various generalizations of the associativity law. These are all
based on the following result:
Theorem 54 (General Associative Law: First Form). Let S be a set with an asso-
ciative binary operation ∗ : S ×S → S. Suppose (ai )i∈I is a nonempty finite ordered
sequence in S. Suppose also that I is the disjoint union of two nonempty subsets I1
and I2 and that every element of I2 is an upper bound of I1 . Then
Y Y Y
ai = ai ∗ ai .
i∈I i∈I1 i∈I2

Proof. If I2 = {ω} is a singleton then ω must be the maximum in I, and the result
now follows from the recursive definition of the product. We proceed by induction
and assume I2 has more than one element. Let ω be the maximum element of I2
and let I20 = I2 − {ω}. Note also that ω is the maximum of all of I. Observe that,
by the associative law and the induction hypothesis (on the size of I2 )
      
Y Y Y Y Y
 ai  ∗ aω =  ai ∗ ai  ∗ aω = ai ∗  ai  ∗ aω 
i∈I1 ∪I20 i∈I1 i∈I20 i∈I1 i∈I20
Q
By definition of product, the left-hand side is just (ai )i∈I and the right-hand side
simplifies as desired:
  
Y Y Y Y
ai ∗  ai  ∗ aω  = ai ∗ ai .
i∈I1 i∈I20 i∈I1 i∈I2

29
 For monoids we can extend this law to possibly empty sequences:
Corollary 55 (General Associative Law: First Form for Monoids). Let hM, ∗i be
a monoid. Suppose (ai )i∈I is a finite ordered sequence in M . Suppose also that I
is the disjoint union of two subsets I1 and I2 where every element of I2 is an upper
bound of I1 . Then Y Y Y
ai = ai ∗ ai .
i∈I i∈I1 i∈I2

Proof. If I1 and I2 are nonempty then use the theorem. If I1 or I2 are empty, replace
the corresponding product with the identity element, and the result is clear.

We can freely remove or add terms that are the identity element:

Theorem 56. Let hM, ∗i be a monoid with identity element. Suppose (ai )i∈I is a
finite ordered sequence in S. Suppose I 0 is a subset of I such that ai = e if i ∈ I −I 0 ,
in other words, the ordered subsequence (ai )i∈I 0 contains all the nontrivial terms
of (ai )i∈I . Then Y Y
ai = ai .
i∈I i∈I 0
Q
In particular, if ai = e for all i ∈ I then i∈I ai is equal to e, the product of the
empty subsequence.

Proof. This can be proved by induction on the size of I. We distinguish two cases:
where I 0 contains the maximum element of I, and where I 0 does not contain this
maximum.

We can extend this law to more than two terms:

Theorem 57 (General Associative Law: Second Form). Let S be a set with an
associative binary operation ∗ : S × S → S. Suppose (ai )i∈I is a nonempty finite
ordered family in S. Suppose also that I is the disjoint union of k ≥ 1 nonempty
subsets I1 , . . . , Ik such that every element of a given It is an upper-bound for all Is
with s < t. Then
Y k Y
Y
ai = ai .
i∈I j=1 i∈Ij

Proof. If k = 1 then the result is clear by the definition of singleton products. We

proceed by induction to k ≥ 2. Suppose I 0 = I1 ∪ · · · ∪ Ik−1 . Then by Theorem 54,
and the induction hypothesis,
 
Y Y Y k−1
Y Y Y
ai = ai ∗ ai =  ai  ∗ ai .
i∈I i∈I 0 i∈Ik j=1 i∈Ij i∈Ik

The right-hand side simplifies to the desired express (using the definition of prod-
uct).

For monoids we can drop the requirement that each Ij is nonempty. We can
even allow k = 0:

30
Corollary 58 (General Associative Law: Second Form for Monoids). Let hM, ∗i
be a monoid with identity element e. Suppose (ai )i∈I is a nonempty finite ordered
family in M . Suppose also that I as the disjoint union of k ≥ 0 subsets I1 , . . . , Ik
such that every element of a given It is an upper-bound for all Is with s < t. Let J
be the set of nonnegative integers up to k. Then
Y Y Y
ai = ai .
i∈I j∈J i∈Ij
Q
Proof. If I is empty, then we get a product j∈J e of identity elements which is
the identity, and the result follows. If I is nonempty then we can remove from J
all j such that Ij is empty (Theorem 56), so we reduce to the case where each Ij is
nonempty. We then can use the above theorem to establish the result.

Next we consider the idea that we can rearrange the parenthesis for associative
operators without affecting the result. Suppose ∗ is associative, then we regard
expressions such as ((a1 ∗ a2 ) ∗ a3 ) ∗ a4 and (a1 ∗ a2 ) ∗ (a3 ∗ a4 ) as representing
different calculations of the same result (by associativity). We formalize the notion
of “calculation” as follows:5
Definition 24. Let S be a set with a binary operation ∗ : S ×S → S. We define the
concept of a “calculation” of a nonempty finite sequence with terms in S recursively
as follows: If I = {i0 } is a singleton set, the only calculation of the sequence (ai )i∈I
is ai0 . If I has size n ≥ 2 then the calculations of (ai )i∈I are the elements of S
of the form b1 ∗ b2 where I1 and I2 are nonempty subsets partitioning I such that
every element of I2 is an upper bound of I1 , where b1 is a calculation of (ai )i∈I1 ,
and where b2 is a calculation of (ai )i∈I2 .

Example 12. Consider the sequence 4, −3, 2, 1, 3 with terms in Z. Consider sub-
traction as our binary operation. Then


4 − (−3) − (2 − 1) − 3 = 5

and
(4 − (−3)) − ((2 − 1) − 3)) = 9
are two calculations. The results are not equal, which illustrates that subtraction
is not associative.
Theorem 59 (General Associative Law: Third Form). Let S be a set with an as-
sociative binary operation ∗ : S × S → S. Suppose (ai )i∈I is a nonempty finite
Q family in S. Then all calculations of (ai )i∈I using ∗ are equal to the prod-
ordered
uct (ai )i∈I . In particular, all calculations of (ai )i∈I are equal.

Q If I = {i0 } then the only calculation of (ai )i∈I is ai0 , which is equal
Proof.
to (ai )i∈I by definition. So we can assume the size of I is greater than one,
and proceed by induction. Suppose b is a calculation of (ai )i∈I , so b = b1 ∗ b2
5 With a bit more work we could formalize “calculation” in terms of the binary treelike structure

used to describe how to calculate the result. Then the theorem would say that any two such trees
associated to a given sequence a1 , a2 , . . . , an .

31
 Q Q
where b1 = (ai )i∈I1 , where b2 = (ai )i∈I2 , where I1 and I2 partition I, and
where every element of I2 is an upper bound of I1 . By the induction hypothesis
and by Theorem 54
Y Y Y
b = b1 ∗ b2 = ai ∗ ai = ai .
i∈I1 i∈I2 i∈I

Finally we mention that our definition of an is equivalent to the other popular

definition:6

Theorem 60. Let a ∈ M where M is a multiplicative monoid. If n is a positive

integer then
Yn
n
a = a.
i=1

Proof. We can prove this by induction on n. The base case reduces to the equa-
tion a1 = a established above. The induction step can be proved as follows:
n
! n+1
Y Y
n+1 n 1
a =a a = a a= a.
i=1 i=1

Exercise 27. Translate the results of this section to additive notation.

18 General Commutativity Laws

In this section we will develop general commutativity laws for multiplicative
monoids. We use monoids since the statements and proofs are a bit more ele-
gant than in a more general setting. But we can certainly generalize these results
to “commutative semigroups” where we do not require an identity element (you just
need to be careful about empty sequences). Also it is straightforward to convert
the results to additive notation.
First we show we can move any term to the end:
Lemma 61. Let hM, ∗i be a commutative monoid. If (ai )i∈I is a finite ordered
sequence of elements in M and if i0 ∈ I then
!
Y Y
ai = ai ∗ ai0
i∈I i∈I 0

where I 0 = I − {i0 } with the restricted ordering.

6 We
used iteration to define powers, but other authors such as Bourbaki in Algebra use the
equation in Theorem 60 instead, at least for positive n.

32
Proof. Let I1 be the set of elements of I strictly smaller than i0 and let I2 be the
set of elements of I strictly larger than i0 . From Corollary 55 and the recursive
definition of products we have
! !
Y Y Y Y Y
ai = ai ∗ ai = ai ∗ ai0 ∗ ai .
i∈I i∈I1 ∪{i0 } i∈I2 i∈I1 i∈I2

Using associativity and commutivity we have

! ! ! ! !
Y Y Y Y Y Y
ai ∗ ai0 ∗ ai = ai ∗ ai0 ∗ ai = ai ∗ ai ∗ ai0 .
i∈I1 i∈I2 i∈I1 i∈I2 i∈I1 i∈I2

Finally, by associativity and Corollary 55 we have

! ! ! !
Y Y Y Y Y
ai ∗ ai ∗ ai0 = ai ∗ ai ∗ ai0 = ai ∗ ai0 .
i∈I1 i∈I2 i∈I1 i∈I2 i∈I 0

From this lemma we can prove the first commutative law which states that the
order of the index set does not affect the product:

Theorem 62 (General Commutative Law: First Form). Let hM, ∗i be a commu-

tative monoid. Let I and J be two ordered finite sets with the same underlying set
which we call K. Let (ak )k∈K be a family of elements of M indexed by K, which
we we can think of as a finite ordered sequence either as (ai )i∈I or as (aj )j∈J . Note
these sequences have the same terms, but perhaps given in a different order. Then
Y Y
ai = aj .
i∈I j∈J

Proof. The proof is by induction on the size of K. If K is empty, then both sides
of the equation in question are the identity element by definition of the empty
product. So assume K is not empty, and fix an element k ∈ K.
Let I 0 = I − {k} and J 0 = J − {k} where these sets have the restricted order
from I and J respectively. Note that I 0 and J 0 have the same underlying set. By
the above lemma and the induction hypothesis we have
!  
Y Y Y Y
ai = ai ∗ ak =  aj  ∗ ak = aj .
i∈I i∈I 0 j∈J 0 j∈J

Definition 25. Let hM, ∗i be a commutative monoid and let Q (ak )k∈K be a family
of elements of M indexed by a finite set K. Then we define (ak )k∈K to be the
value resulting from any chosen total ordering of K. By the above theorem, all
orderings give the same value, so this new type of product is well-defined, and
agrees with the old definition (Definition 22) for any ordering of K.

33
 Recall that in Section 16 we considered an equivalence relation for finite ordered
sequences called order equivalence, and showed that order equivalent sequences
have the same product. Recall that two finite ordered sequence (ai )i∈I and (bJ )i∈J
with values in a given set are order-equivalent if (i) the cardinality of I and J are
equal, and (ii) the unique order-preserving bijection γ : I → J has the property
that bγ(i) = ai for all i ∈ I. We now define another equivalence relation where
we drop the requirement that γ be order-preserving: we declare two finite fami-
lies (ai )i∈I and (bJ )i∈J with values in a given set to be multiset-equivalent if there
is a bijection γ : I → J with the property that bγ(i) = ai for all i ∈ I. (Note: we
can define a multiset as an equivalence class under this relation. However, we will
not need multisets, per se, in this document.)
Theorem 63 (General Commutative Law: Second Form). Let (ai )i∈I and (bj )j∈J
be two finite families of elements of M where hM, ∗i is a commutative monoid.
If (ai )i∈I and (bj )j∈J are multiset-equivalent then
Y Y
ai = aj .
i∈I j∈J

Proof. Let γ be a bijection γ : I → J with the property that bγ(i) = ai for all i ∈ I.
Fix any ordering on I. If J comes with an order, then ignore it. Instead, impose
the unique order on J such that γ is an order preserving map. Under this choice of
order we have that (ai )i∈I and (bj )j∈J are order-equivalent sequences and so, using
these orderings, Y Y
ai = bj .
i∈I j∈J

Note that this equality continuous to hold when we drop the orderings on I and J
(Theorem 62 and Definition 25).

19 Induced Operation on Functions

Above in Section 12 we used the term “induced operation” for the restriction of
a binary operation to a closed subset. We will switch gears a bit here and define
another kind of “induced operations”, namely induced operations on functions. The
motivation is observation that if we have a binary operation on a set M then we
get a natural binary operation on the set of functions S → M where S is any given
domain:
Definition 26. Suppose S and M are sets. Then let F(S, M ) be the set of func-
tions S → M .
Definition 27. Let M be a set with a binary operation ∗ : M × M → M and let S
be be a fixed set to serve as a domain. Then we define a binary operation ∗F on
the set of functions F(S, M ) as follows: given a pair of functions f, g ∈ F(S, M )
define f ∗F g by the rule
def
(f ∗F g)(x) = f (x) ∗ g(x).

34
We call ∗F the induced operation on functions. We will often write ∗F as just ∗
and use context to distinguish this operation from that on M .
Remark. Observe that ∗F (or just ∗) gives a binary operation
F(S, M ) × F(S, M ) → F(S, M ).
The following lemmas are proved in a straightforward manner.
Lemma 64. Suppose that ∗ : M × M → M is an associative operation. Then the
induced operation on F(S, M ) is also associative.
Lemma 65. Suppose that ∗ : M × M → M is a commutative operation. Then the
induced operation on F(S, M ) is also commutative.
Lemma 66. Suppose that ∗ : M × M → M is a binary operation with identity
element e ∈ M . Let e : S → M be the constant function defined by the equa-
tion e(x) = e for all x ∈ S. Then e is an identify element for the induced operation
on F(S, M ).
Theorem 67. Suppose M is a monoid and that S is a set. Then F(S, M ) is a
monoid under the induced operation. The identity element is the constant func-
tion S → M whose values are given by the identity element of M . If M is a
commutative monoid then so is F(S, M ).
Theorem 68. Suppose hG, ∗i is a group and that S is a set. Then F(S, G) is
a group under the induced operation. The identity element is the constant func-
tion S → G whose values are given by the identity element of G.
Proof. By Theorem 67, F(S, G) is a monoid. To show it is a group we need to
show every element f ∈ F(S, G) has an inverse. Let g : S → G be defined by the
rule x 7→ f (x)−1 (for convenience we adopt multiplicative notation for G, and use e
for the identity in G). Then for any x ∈ S we have
(f ∗ g)(x) = f (x) ∗ g(x) = f (x) ∗ f (x)−1 = e.
In other words, f ∗ g is the constant function e whose values are e. Similarly, we
can show g ∗ f = e. Thus g is the inverse of f in F(S, G).
Exercise 28. Suppose M monoid and that S is a set. Describe the elements in
the group F(S, M )inv of invertible elements.
Exercise 29. (Taking as given the notions of continuity and differentiability) Con-
sider hR, +i. Let G = F(R, R). As we have seen, G is a group under the induced
addition +. Let H1 be the subset of continuous functions. Show that H1 is a
subgroup of G. Let H2 be the subset of differentiable functions. Show that H2 is a
subgroup of H1 and G.
Exercise 30. (Taking as given the notions of continuity and differentiability) Con-
sider hR, ·i. Let S be the open interval (0, 1), and let M = F(S, R). As we have
seen, M is a monoid under the induced multiplication. A typical element of M is
the function defined by f (x) = 1/x. Let H1 be the subset of continuous functions.
Show that H1 is a submonoid of M . Let H2 be the subset of differentiable functions.
Show that H2 is a submonoid of H1 and G.

35
19.1 Cartesian Powers of a Monoid
We now consider the special case of sets of functions F(S, M ) where S = {1, . . . , n}
where n is a positive integer. Then an element f of F(S, M ) can be completely
described simply by specifying n values:
f (1), f (2), . . . , f (n).
In other words, elements of F(S, M ) can be thought of as finite se-
quences (a1 , . . . , an ) with ai ∈ M . Here we identify (a1 , . . . , an ) with the func-
tion i 7→ ai which is an element of F(S, M ) where S = {1, . . . , n}.
Definition 28. Let M be a set, and let n be a positive integer. Then M n is
defined to be F({1, . . . , n}, M ). As discussed above, we think of M n as the set of
finite sequences of the form (a1 , . . . , an ) where each ai ∈ M . We call M n the nth
Cartesian power of M .
If M has a binary operation ∗ : M ×M → M , then we get an induced operation ∗
on M n using Definition 27. The induced operation can be written as
(a1 , . . . , an ) ∗ (b1 , . . . , bn ) = (a1 ∗ b1 , . . . , an ∗ bn )
where the operation ∗ in the left-hand side is the induced operation, and the oper-
ation in the right hand side is the operation in M .
The following is just a special case of previous results for F(G, S):
Theorem 69. Let n be a positive integer. If M is a monoid, then so is M n (under
the induced operation). If M is a commutative monoid, then so is M n . If G is a
group then so is Gn . If G is an Abelian group, then so is Gn .
Exercise 31. (Taken as given the vector spaces R2 and R3 ). Consider R as a group
under addition. Show that the induced operation on R2 and R3 are the usual vector
addition. Conclude that R2 and R3 are Abelian groups under vector addition.
Exercise 32. Suppose H is a subgroup of G. By extending codomains, you can
identify every element of F(S, H) with an element of F(S, G), and so can iden-
tify F(S, H) with a subset of F(S, G). Show that when you do this, the operation
on F(S, H) is the restriction of the operation on F(S, G), and that F(S, H) can
be regarded as a subgroup of F(S, G). Note: as a special cases we see H n can be
regarded as a subgroup of Gn for all positive n.
(Note: the analogous property holds for monoids and submonoids, and this can
be shown along the way.)
Remark. For any set G it is common to define G0 to be a set with one element.
When G is a group or monoid, it is common to regard G0 as a group using the
unique binary operation on this one-point set G0 . (This is consistent with thinking
of G0 as the set F(∅, G) ).

20 Composition of Functions
Above we considered one sort of operation on functions which gives a monoid,
assuming the codomain is a monoid. Now we describe another way to form monoids

36
whose elements are functions. Let X be a given set and consider F(X, X), the set
of functions from X to itself. In this case we write F(X, X) simply as F(X). Note
that composition ◦ gives a binary operation

◦ : F(X) × F(X) → F(X)

and that if X is finite of size n then F(X) has nn elements. The following theorems
are straightforward consequences of basic set-theoretical facts:

Theorem 70. Let X be a set. Then F(X) is a monoid under composition. The
identity element is the identity function X → X.
Theorem 71. Let X be a set. Then a function f ∈ F(X) is invertible in the
monoid F(X) if and only if it is invertible as a function X → X. The inverse of
an invertible f in the monoid F(X) is the inverse function f −1 : X → X.
Definition 29. Let X be a set. The group of units of F(X) is called the sym-
metric group on X, which we can write as S(X) or SX , and its elements are called
permutations of X.

Remark. Note that if X is finite with n elements then the symmetric group on X
has n! elements. If X = {1, . . . , n} then the symmetric group is sometimes just
written as Sn (we sometimes write it as Sn )
Remark. Observe that if A is a group or monoid then F(A, A) has two standard
operations: composition and the induced operation. We have to be clear of what
is meant in any given situation, especially since we sometimes use multiplicative
notation f g for both operations. Similarly, f −1 can have two possible meanings
depending on which monoid, and hence which binary operation, is under consider-
ation.

Exercise 33. Consider hR, ·i, and let M = F(R, R). Let f, g ∈ M be defined by
the equations f (x) = x + 1 and g(x) = x2 . What is f g, gf, f ◦ g, and g ◦ f ?

37