SEMIGROUP THEORY

A LECTURE COURSE
VICTORIA GOULD

1. The Basic Concept

Definition 1.1. A semigroup is a pair (S, ∗) where S is a non-empty set and ∗ is an
associative binary operation on S. [i.e. ∗ is a function S × S → S with (a, b) 7→ a ∗ b and
for all a, b, c ∈ S we have a ∗ (b ∗ c) = (a ∗ b) ∗ c].

n Semigroups Groups
1 1 1
2 4 1
3 18 1
4 126 2
5 1160 1
6 15973 2
7 836021 1
8 1843120128 5
9 52989400714478 2
The number (whatever it means) of semigroups and groups of order n

We abbreviate “(S, ∗)” by “S” and often omit ∗ in “a ∗ b” and write “ab”. By induction
a1 a2 . . . an is unambiguous. Thus we write an for

aa . . . a} .
| {z
n times
Index Laws For all n, m ∈ N = {1, 2, . . . }:

an am = an+m

m
an = anm .
Definition 1.2. A monoid M is a semigroup with an identity, i.e. there exists 1 ∈ M
such that 1a = a = a1 for all a ∈ M.

Putting a0 = 1 then the index laws hold for all n, m ∈ N0 = {0, 1, 2, . . .}.
1
2 VICTORIA GOULD

Note. The identity of a monoid is unique.

Definition 1.3. A group G is a monoid such that for all a ∈ G there exists a b ∈ G with
ab = 1 = ba.
Example 1.4. Groups are monoids and monoids are semigroups. Thus we have

Groups ⊂ Monoids ⊂ Semigroups.

The one element trivial group {e} with multiplication table

e
e e

is also called the trivial semigroup or trivial monoid.

Example 1.5. A ring is a semigroup under ×. If the ring has an identity then this
semigroup is a monoid.
Example 1.6. (1) (N, ×) is a monoid.
(2) (N, +) is a semigroup.
(3) (N0 , ×) and (N0 , +) are monoids.
Example 1.7. Let I, J be non-empty sets and set T = I × J with the binary operation

(i, j)(k, ℓ) = (i, ℓ).

Note


(i, j)(k, ℓ) (m, n) = (i, ℓ)(m, n) = (i, n),


(i, j) (k, ℓ)(m, n) = (i, j)(k, n) = (i, n),
for all (i, j), (k, ℓ), (m, n) ∈ T and hence multiplication is associative.
Then T is a semigroup called the rectangular band on I × J.
Notice: (i, j)2 = (i, j)(i, j) = (i, j), i.e. every element is an idempotent.
This shows that not every semigroup is the multiplicative semigroup of a ring, since any
ring where every element is an idempotent is commutative. However, a rectangular band
does not have to be commutative.
Adjoining an Identity Let S be a semigroup. Find a symbol not in S, call it “1”. On
S ∪ {1} we define ∗ by

a ∗ b = ab for all a, b ∈ S,
a∗1 =a = 1∗a for all a ∈ S,
1 ∗ 1 = 1.
 SEMIGROUP THEORY A LECTURE COURSE 3

j ℓ

i (i, j) (i, ℓ)

k (k, ℓ)

Figure 1. The rectangular band.

Then ∗ is associative (check this) so S ∪ {1} is a monoid with identity 1. Multiplication in

S ∪ {1} extends that in S.
The monoid S 1 is defined by
(
S if S is a monoid,
S1 =
S ∪ {1} if S is not a monoid.
Definition 1.8. S 1 is “S with a 1 adjoined if necessary”.
Example 1.9. Let T be the rectangular band on {a} × {b, c}. Then T 1 = {1, (a, b), (a, c)},
which has multiplication table

1 (a, b) (a, c)
1 1 (a, b) (a, c)
(a, b) (a, b) (a, b) (a, c)
(a, c) (a, c) (a, b) (a, c)

The Bicyclic Semigroup/Monoid B

If A ⊆ Z, such that |A| < ∞, then max A is the greatest element in A. i.e.
(
a if a > b,
max{a, b} =
b if b > a.
We note some further things about max:
4 VICTORIA GOULD

• max{a, 0} = a if a ∈ N0 ,
• max{a, b} = max{b, a},
• max{a,
 a} = a, 
• max a, max{b, c} = max{a, b, c} = max max{a, b}, c .
Thus we have that (Z, max) where max(a, b) = max{a, b} is a semigroup and (N0 , max) is
a monoid.
Note. The following identities hold for all a, b, c ∈ Z

a + max{b, c} = max{a + b, a + c},
(⋆)
max{b, c} = a + max{b − a, c − a}.

Put B = N0 × N0 . On B we define a ‘binary operation’ by

(a, b)(c, d) = (a − b + t, d − c + t),

where t = max{b, c}.
Proposition 1.10. B is a monoid with identity (0, 0).
Proof. With (a, b), (c, d) ∈ B and t = max{b, c} we have t − b > 0 and t − c > 0. Thus we
have a − b + t > a and d − c + t > d. Therefore, in particular (a − b + t, d − c + t) ∈ B so
multiplication is closed. We have that (0, 0) ∈ B and for any (a, b) ∈ B we have

(0, 0)(a, b) = (0 − 0 + max{0, a}, b − a + max{0, a}),

= (0 − 0 + a, b − a + a),
= (a, b),
= (a, b)(0, 0).
Therefore (0, 0) is the identity of B.
We need to verify associativity.
Let (a, b), (c, d), (e, f ) ∈ B. Then

(a, b)(c, d) (e, f ) = a − b + max{b, c}, d − c + max{b, c} (e, f ),
= a − b − d + c + max{d − c + max{b, c}, e},


f − e + max{d − c + max{b, c}, e} .



(a, b) (c, d)(e, f ) = (a, b) c − d + max{d, e}, f − e + max{d, e} ,
= (a − b + max{b, c − d + max{d, e}}


f − e − c + d + max{b, c − d + max{d, e}} .
Now we have to show that
 SEMIGROUP THEORY A LECTURE COURSE 5

 
a−
✘✘✘b − d + c + max d − c + max{b, c}, e = ✘
a−
✘✘b + max b, c − d + max{d, e} ,
 
f✘
✘ −✘
e + max d − c + max{b, c}, e = ✘
f✘−✘
e − c + d + max b, c − d + max{d, e} .
We can see that these equations are the same and so we only need to show

 
c − d + max d − c + max{b, c}, e = max b, c − d + max{d, e} .

Now, we have from (⋆) that this is equivalent to

 
max max{b, c}, c − d + e = max b, c − d + max{d, e} .
The RHS of this equation is
 
max b, c − d + max{d, e} = max b, max{c − d + d, c − d + e} ,

= max b, max{c, c − d + e} ,
= max{b, c, c − d + e},

= max max{b, c}, c − d + e .
Therefore multiplication is associative and hence B is a monoid. 
Definition 1.11. With the above multiplication, B is called the Bicyclic Semigroup/Monoid.
Example 1.12. For any set X, the set TX of all maps X → X is a monoid. (See Lecture
3).
Definition 1.13. A semigroup S is commutative if ab = ba for all a, b ∈ S.
For example N with + is commutative. B is not because

(0, 1)(1, 0) = (0 − 1 + 1, 0 − 1 + 1) = (0, 0),

(1, 0)(0, 1) = (1 − 0 + 0, 1 − 0 + 0) = (1, 1).
Thus we have (0, 1)(1, 0) 6= (1, 0)(0, 1). Notice that in B; (a, b)(b, c) = (a, c).
Definition 1.14. A semigroup is cancellative if

ac = bc ⇒ a = b, and
ca = cb ⇒ a = b.
NOT ALL SEMIGROUPS ARE CANCELLATIVE

For example in the rectangular band on {1, 2} × {1, 2} we have

(1, 1)(1, 2) = (1, 2) = (1, 2)(1, 2)

6 VICTORIA GOULD

B is not cancellative as e.g.

(1, 1)(2, 2) = (2, 2)(2, 2).
Groups are cancellative (indeed, any subsemigroup of a group is cancellative). N0 is a
cancellative monoid, which is not a group.
Definition 1.15. A zero “0” of a semigroup S is an element such that, for all a ∈ S,

0a = a = a0.

Adjoining a Zero Let S be a semigroup, then pick a new symbol “0”. Let S 0 = S ∪ {0};
define a binary operation · on S 0 by

a · b = ab for all a ∈ S,
0·a=0=a·0 for all a ∈ S,
0 · 0 = 0.
Then · is associative, so S 0 is a semigroup with zero 0.
Definition 1.16. S 0 is S with a zero adjoined.

2. Standard algebraic tools

Definition 2.1. Let S be a semigroup and ∅ = 6 T ⊆ S. Then T is a subsemigroup of S if
a, b ∈ T ⇒ ab ∈ T . If S is a monoid then T is a submonoid of S if T is a subsemigroup
and 1 ∈ T .
Note T is then itself a semigroup/monoid.

Example 2.2. (1) (N, +) is a subsemigroup of (Z, +).

(2) R = {cx | x ∈ X} is a subsemigroup of TX , since
cx cy = cy
for all x, y ∈ X.
R is a right zero semigroup (See Ex.1).
(3) Put E(B) = {(a, a) | a ∈ N0 }.
From Ex. 1, E(S) = {α ∈ B : α2 = α}
Claim E(B) is a commutative submonoid of B.
Clearly we have (0, 0) ∈ E(B) and for (a, a), (b, b) ∈ E(B) we have

(a, a)(b, b) = (a − a + t, b − b + t) where t = max{a, b},

= (t, t),
= (b, b)(a, a).
 SEMIGROUP THEORY A LECTURE COURSE 7

Definition 2.3. Let S, T be semigroups then θ : S → T is a semigroup (homo)morphism

if, for all a, b ∈ S,

(ab)θ = aθbθ.
If S, T are monoids then θ is a monoid (homo)morphism if θ is a semigroup morphism and
1S θ = 1T .
Example 2.4. (1) θ : B → Z given by (a, b)θ = a − b is a monoid morphism because

(a, b)(c, d) θ = (a − b + t, d − c + t)θ t = max{b, c}
= (a − b + t) − (d − c + t)
= (a − b) + (c − d)
= (a, b)θ + (c, d)θ.
Furthermore (0, 0)θ = 0 − 0 = 0.
(2) Let T = I × J be the rectangular band then define α : T → TJ by (i, j)α = cj . Then
we have

(i, j)(k, ℓ) α = (i, ℓ)α,
= cℓ ,
= cj cℓ ,
= (i, j)α(k, ℓ)α.
So, α is a morphism.
Definition 2.5. A bijective morphism is an isomorphism.

Isomorphisms preserve algebraic properties (e.g. commutativity).

See handout for further information.

Embeddings Suppose α : S → T is a morphism. Then Im α is a subsemigroup (sub-

monoid) of T . If α is 1:1, then α : S → Im α is an isomorphism, so that S ∼
= Im α. We
say that S is embedded in T .
8 VICTORIA GOULD

Theorem 2.6 (The “Cayley Theorem” – for Semigroups). Let S be a semigroup. Then S
is embedded in TS 1 .
Proof. Let S be a semigroup and set X = S 1 . We need a 1:1 morphism S → TX .
For s ∈ S, we define ρs ∈ TX by xρs = xs.
Now define α : S → TX by sα = ρs .
We show α is 1:1: If sα = tα then ρs = ρt and so xρs = xρt for all x ∈ S 1 ; in particular
1ρs = 1ρt and so 1s = 1t hence s = t and α is 1:1.
We show α is a morphism: Let u, v ∈ S. For any x ∈ X we have

x(ρu ρv ) = (xρu )ρv = (xu)ρv = (xu)v = x(uv) = xρuv .

Hence ρu ρv = ρuv and so uαvα = ρu ρv = ρuv = (uv)α. Therefore α is a morphism.
Hence α : S → TX is an embedding. 
Theorem 2.7 (The “Cayley Theorem” - for Monoids). Let S be a monoid. Then there
exists an embedding S ֒→ TS .
Proof. S 1 = S so TS = TS 1 . We know α is a semigroup embedding. We need only check
1α = IX .
Now 1α = ρ1 and for all x ∈ X = S we have

xρ1 = x1 = x = xIX
and so 1α = ρ1 = IX . 
Theorem 2.8 (The Cayley Theorem - for Groups). Let S be a group. Then there exists
an embedding S ֒→ SS .
Proof. Exercise. 

2.1. Idempotents
S will always denote a semigroup.
Definition 2.9. e ∈ S is an idempotent if e2 = e. We put
E(S) = {e ∈ S | e2 = e}.
Now, E(S) may be empty, e.g. E(S) = ∅ (N under +).
E(S) may also be S. If S = I × J is a rectangular band then for any (i, j) ∈ S we have
(i, j)2 = (i, j)(i, j) = (i, j) and so E(S) = S.

For the bicyclic semigroup B we have from Ex. 1


E(B) = (a, a) | a ∈ N0 .
If S is a monoid then 1 ∈ E(S).
If S is a cancellative monoid, then 1 is the only idempotent: for if e2 = e then ee = e1 and
so e = 1 by cancellation. In particular for S a group we have E(S) = {1}.
 SEMIGROUP THEORY A LECTURE COURSE 9

Definition 2.10. If E(S) = S, then S is a band.

Definition 2.11. If E(S) = S and S is commutative, then S is a semilattice.
Lemma 2.12. Let E(S) 6= ∅ and suppose ef = f e for all e, f ∈ E(S). Then E(S) is a
subsemigroup of S.
Proof. Let e, f ∈ E(S). Then

(ef )2 = (ef )(ef ) = e(f e)f = e(ef )f = (ee)(f f ) = ef

and hence ef ∈ E(S). 

From Lemma 2.12 if E(S) 6= ∅ and idempotents in S commute then E(S) is a semilattice.

Example 2.13. (1) E(B) = (a, a) | a ∈ N0 is a semilattice.
(2) A rectangular band I × J is not a semilattice (unless |I| = |J| = 1) since (i, j)(k, ℓ) =
(k, ℓ)(i, j) ⇔ i = k and j = ℓ.
Definition 2.14. Let a ∈ S. Then we define hai = {an | n ∈ N}, which is a commutative
subsemigroup of S. We call hai the monogenic subsemigroup of S generated by a.
Proposition 2.15. Let a ∈ S. Then either
(i) |hai| = ∞ and hai ∼ = (N, +) or
(ii) hai is finite. In this case ∃n, r ∈ N such that

hai = {a, a2 , . . . , an+r−1 }, |hai| = n + r − 1

{an , an+1 , . . . , an+r−1 } is a subsemigroup of hai and for all s, t ∈ N0 ,

an+s = an+t ⇔ s ≡ t (mod r).

Proof. If ai 6= aj for all i, j ∈ N with i 6= j then θ : hai → N defined by ai θ = i is an

isomorphism. This is case (i).

Suppose that in the list of elements a, a2 , a3 , . . . there is a repetition, i.e. ai = aj for some
i < j. Let k be least such that ak = an for some n < k. Then k = n + r for some r ∈ N –
where n is the index of a, r is the period of a. Then the elements a, a2 , a3 , . . . , an+r−1 are
all distinct and an = an+r .
DO NOT CANCEL
Let s, t ∈ N0 with
s = s′ + ur, t = t′ + vr
with
0 ≤ s′ , t′ ≤ r − 1, u, v ∈ N0 .
10 VICTORIA GOULD

Then
′
an+s = an+s +ur
′
= as an+ur in S 1
′
= as an+r a(u−1)r
′
= as an a(u−1)r
′
= as an+(u−1)r
..
.
′
= as an
′
= an+s .
′
Similarly, an+t = an+t . Therefore
′ ′
an+s = an+t ⇔ an+s = an+t ⇔ s′ = t′ ⇔ s ≡ t( mod r).
Notice that

an+ur = an
for all u.
We have shown
{a, a2 , . . . , an , an+1 , . . . , an+r−1} = hai
and
|hai| = n + r − 1.
Clearly
{an , an+1 , . . . , an+r−1}
is a subsemigroup. In fact
an+s an+t = an+u
where u ≡ s + n + t (mod r) and 0 ≤ u ≤ r − 1. This is case (ii).
We can express this pictorially:

an+1

a a2 a3 a4 an an+2

an+r−1

Lemma 2.16 (The Idempotent Power Lemma). If hai is finite, then it contains an idem-
potent.
 SEMIGROUP THEORY A LECTURE COURSE 11

Proof. Let n, r be the index and period of a. Choose s ∈ N0 with s ≡ −n (mod r). Then
s + n ≡ 0 (mod r) and so s + n = kr for k ∈ N. Then

(an+s )2 = an+n+s+s = an+kr+s = an+s

and so an+s ∈ E(S). 
In fact, {an , an+1 , . . . , an+r−1 } is a cyclic group with identity an+s .
Corollary 2.17. Any finite semigroup contains an idempotent.

2.2. Idempotents in TX
We know cx cy = cy for all x, y ∈ X and hence cx cx = cx for all x ∈ X. Therefore
cx ∈ E(TX ) for all x ∈ X. But if |X| > 1 then there are other idempotents in TX as well.
Example 2.18. Let us define an element
 
1 2 3
α= ∈ E(TX ).
2 2 3
Then
     
2 1 2 3 1 2 3 1 2 3
α = · = ,
2 2 3 2 2 3 2 2 3
thus α is an idempotent.
Definition 2.19. Let α : X → Y be a map and let Z ⊆ X. Then the restriction of α to
the set Z is the map
α|Z : Z → Y, z 7→ zα for every z ∈ Z.
NOTE: Sometimes we treat the restriction α|Z as a map with domain Z and codomain
Zα.
Example 2.20. Let us define a map with domain {a, b, c, d} and codomain {1, 2, 3}:
 
a b c d
α= .
1 3 1 2
Then α|{a,d} is the following map:
 
a d
α|{a,d} = .
1 2
We can see that α is not one-to-one but α|{a,d} is.
Let α ∈ TX (i.e. α : X → X). Recall that

Im α = {xα : x ∈ X} ⊆ X = Xα.
12 VICTORIA GOULD

Example 2.21. In T3 we have Im c1 = {1}, Im I3 = {1, 2, 3} and

 
1 2 3
Im = {2, 3}.
3 2 3
The following lemma gives a rather useful characterization of the idempotents of a trans-
formation monoid.
Lemma 2.22 (The E(TX ) Lemma). An element ε ∈ TX is idempotent ⇔ ε|Im ε = IIm ε .
Proof. ε|Im ε = IIm ε means that for all y ∈ Im ε we have yε = y.
Note that Im ε = {xε : x ∈ X}.
Then

ε ∈ E(TX ) ⇔ ε2 = ε,
⇔ xε2 = xε for all x ∈ X,
⇔ (xε)ε = xε for all x ∈ X,
⇔ yε = y for all y ∈ Im ε,
⇔ ε|Im ε = IIm ε . 
Example 2.23. Let
 
1 2 3
α= ∈ T3 ,
2 2 3
this has image Im α = {2, 3}. Now we can see that 2α = 2 and 3α = 3. Hence α ∈ E(T3 ).
Example 2.24. We can similarly create another idempotent in T7 , first we determine its
image: let it be the subset {1, 2, 5, 7}. Our map must fix these elements, but can map the
other elements to any of these:
   
1 2 3 4 5 6 7 1 2 3 4 5 6 7
→ ∈ E(T7 ).
1 2 5 7 1 2 5 7 5 2 7
Using Lemma 2.22 we can now list all the idempotents in T3 . We start with the constant
maps, i.e. ε ∈ E(T3 ) such that | Im ε| = 1. These are
     
1 2 3 1 2 3 1 2 3
, , .
1 1 1 2 2 2 3 3 3
Now consider all elements ε ∈ E(T3 ) such that | Im ε| = 2. These are
     
1 2 3 1 2 3 1 2 3
, , ,
2 2 3 1 1 3 3 2 3
     
1 2 3 1 2 3 1 2 3
, , .
1 3 3 1 2 1 1 2 2
 SEMIGROUP THEORY A LECTURE COURSE 13

Now there is only one idempotent such that | Im ε| = 3, that is the identity map
 
1 2 3
.
1 2 3

3. Relations
Please see the handout ‘Functions and Relations’.

In group theory, homomorphic images of groups are determined by normal subgroups.

The situation is more complicated in semigroup theory, namely the homomorphic images
of semigroups are determined by special equivalence relations. Furthermore, elements of
semigroups can be quite often ‘ordered’. For example there is a natural notion of a map
being ‘bigger’ than another one: namely if its image has a bigger cardinality. These
examples show that relations play a central role in semigroup theory.
Definition 3.1. A (binary) relation ρ on A is a subset of A × A.

Convention: we may write “a ρ b” for “(a, b) ∈ ρ”.

3.1. Some special relations

Properties of the relation 6 on R:

a6a for all a ∈ R,

a 6 b and b 6 c ⇒ a 6 c for all a, b, c ∈ R,
a 6 b and b 6 a ⇒ a = b for all a, b ∈ R,
a 6 b or b 6 a for all a, b ∈ R.

Thus, the relation 6 is a total order on R (sometimes we say that R is linearly ordered by
6).

Recall that if X is any set, we denote by P(X) the set of all subsets of X (and call it the
power set of X. Properties of the relation ⊆ on a power set P(X) of an arbitrary set X:

A⊆A for all A ∈ P(X)

A ⊆ B and B ⊆ C ⇒ A ⊆ C for all A, B, C ∈ P(X)
A ⊆ B and B ⊆ A ⇒ A = B for all A, B ∈ P(X)

Notice that if |X| > 2 and x, y ∈ X with x 6= y then {x} 6⊆ {y} and {y} 6⊆ {x}, thus ⊆ is
a partial order but not a total order on P(X).
14 VICTORIA GOULD

Recall that

[a] = {b ∈ A | a ρ b}.
If ρ is an equivalence relation then [a] is the equivalence-class, or the ρ-class, of a.

We denote by ω the UNIVERSAL relation on A: ω = A × A. So x ω y for all x, y ∈ A,

and [x] = A for all x ∈ A.

We denote by ι be the EQUALITY relation on A:


ι = (a, a) | a ∈ A .
Thus x ι y ⇔ x = y and so [x] = {x} for all x ∈ A.

3.2. Algebra of Relations

If ρ, λ are relations on A, then so is ρ ∩ λ. For all a, b ∈ A we have

a (ρ ∩ λ) b ⇔ (a, b) ∈ ρ ∩ λ
⇔ (a, b) ∈ ρ and (a, b) ∈ λ
⇔ a ρ b and a λ b.
We note that ρ ⊆ λ means a ρ b ⇒ a λ b.

Note that ι ⊆ ρ ⇔ ρ is reflexive and so ι ⊆ ρ for any equivalence relation ρ.

We see that ι is the smallest equivalence relation on A and ω is the largest equivalence
relation on A.
Lemma 3.2. If ρ, λ are equivalence relations on A then so is ρ ∩ λ.
Proof. We have ι ⊆ ρ and ι ⊆ λ, then ι ⊆ ρ ⊆ λ, so ρ∩λ is reflexive. Suppose (a, b) ∈ ρ∩λ.
Then (a, b) ∈ ρ and (a, b) ∈ λ. So as ρ, λ are symmetric, we have (b, a) ∈ ρ and (b, a) ∈ λ
and hence (b, a) ∈ ρ ∩ λ. Therefore ρ ∩ λ is symmetric. By a similar argument we have
ρ ∩ λ is transitive. Therefore ρ ∩ λ is an equivalence relation. 

Denoting by [a]ρ the ρ-class of a and [a]λ the λ-class of a we have that,

[a]ρ∩λ = {b ∈ A | b ρ ∩ λ a},
= {b ∈ A | b ρ a and b λ a},
= {b ∈ A | b ρ a} ∩ {b ∈ A | b λ a},
= [a]ρ ∩ [a]λ .
We note that ρ ∪ λ need not be an equivalence relation. On Z we have
 SEMIGROUP THEORY A LECTURE COURSE 15

3≡1 (mod 2),

1≡4 (mod 3).



If ≡ (mod 2) ∪ ≡ (mod 3) were ) to be transitive then we would have



(3, 1) ∈ ≡ (mod 2) ∪ ≡ (mod 3)



(1, 4) ∈ ≡ (mod 2) ∪ ≡ (mod 3)



⇒ (3, 4) ∈ ≡ (mod 2) ∪ ≡ (mod 3)
⇒ 3 ≡ 4 (mod 2) or 3 ≡ 4 (mod 3)
but this is a contradiction!

3.3. Kernels
Definition 3.3. Let α : X → Y be a map. Define a relation ker α on X by the rule

a ker α b ⇔ aα = bα.
We call ker α the kernel of α.
We may sometimes write a ≡α b. It is clear that ker α is an equivalence relation on X.
The ker α classes partition X into disjoint subsets; a, b lie in the same class iff aα = bα.
Example 3.4. Let α : 6 → 4 where
 
1 2 3 4 5 6
α= .
3 2 3 2 2 1
In this case the different ker α-classes are {1, 3}, {2, 4, 5}, {6}.
Note that if α : A → B is a map then α is one-one if and only if ker α = ιA and α is
constant if and only if ker α = ωA .
Definition 3.5. An equivalence relation ρ on a semigroup S is a congruence if

(a ρ b and c ρ d) ⇒ ac ρ bd.
Lemma 3.6 (The Kernel Lemma). Let θ : S → T be a semigroup morphism. Then ker θ
is a congruence on S.
Proof. We know ker θ is an equivalence relation on S. Suppose a, b, c, d ∈ S with

(a ker θ b) and (c ker θ d).

Then aθ = bθ and cθ = dθ, so

(ac)θ = aθcθ = bθdθ = (bd)θ.

Therefore ac ker θ bd, so that ker θ is a congruence. 
16 VICTORIA GOULD

Note. Some remarks on the notion well-defined : usually we define a map on a set by
simply stating what the image of the individual elements should be, e.g:
α : N → Z, nα = the number of 9’s less the number of 2’s in the decimal form of n.
But very often in mathematics, the set on which we would like to define the map is a set
of classes of an equivalence relation (that is, the factor set of the relation). In such cases,
we usually define the map by using the elements of the equivalence classes (for usually we
can use some operations on them). For example let
ρ = {(n, m)|n ≡ m (mod 4)} ⊆ N × N.
Then ρ is an equivalence relation having the following 4 classes:
A = {1, 5, 9, 13, . . .}, B = {2, 6, 10, 14, . . .},
C = {3, 7, 11, 15, . . .}, D = {4, 8, 12, 16, . . .}.
Thus, the factor set of ρ is X = {A, B, C, D}. We try do define a map from X to N by
α : X → N, [n]ρ α = 2n .
What is the image of A under α? We choose an element n of A (that is, we represent A
as [n]ρ ): 1 ∈ A, thus A = [1]ρ . So Aα = [1]ρ α = 2. However, 5 ∈ A, too! So we have
Aα = [5]ρ α = 25 = 32. Thus, Aα has more than one values. We refer to this situation as
‘α being not well-defined’.
Keep in mind that whenever we try to define something (a map, or an operation) on a
factor set of an equivalence relation by referring to ELEMENTS of the equivalence classes,
it MUST be checked, that the choice of the elements of the equivalence classes does not
influence the result.
For example in the above-mentioned example let
β : X → N0 , [n]ρ β = n,
where n denotes the remainder of n on division by 4 (that is, 0, 1, 2 or 3). In this case β is
well-defined, because all elements in the same class have the same remainder, for example
Aβ = [1]ρ β = 1 = [5]ρ β = [9]ρ β = . . .

The following construction and lemmas might be familiar...

Let ρ be a congruence on S. Then we define


S/ρ = [a] | a ∈ S .
Define a binary operation on S/ρ by

[a][b] = [ab].
We need to make sure that this is a well-defined operation, that is, that the product [a][b]
does not depend on the choice of a and b. If [a] = [a′ ] and [b] = [b′ ] then a ρ a′ and b ρ b′ ;
 SEMIGROUP THEORY A LECTURE COURSE 17

as ρ is a congruence we have ab ρ a′ b′ and hence [ab] = [a′ b′ ]. Hence our operation is

well-defined. Let [a], [b], [c] ∈ S/ρ then we have

[a] [b][c] = [a][bc],
 
= a(bc) ,
 
= (ab)c ,
= [ab][c],


= [a][b] [c].
If S is a monoid, then so is S/ρ because we have

[1][a] = [1a] = [a] = [a1] = [a][1]

for any a ∈ S. Hence we conclude that S/ρ is a semigroup and if S is a monoid, then so
is S/ρ.
Definition 3.7. We call S/ρ the factor semigroup (or monoid) of S by ρ.
Now, define νρ : S → S/ρ by

sνρ = [s].
Then we have

sνρ tνρ = [s][t] definition of νρ ,

= [st] definition of multiplication in S/ρ,
= (st)νρ definition of νρ .

Hence νρ is a semigroup morphism. Moreover if S is a monoid then 1νρ = [1], so that νρ

is a monoid morphism. We now want to examine the kernel of νρ :

s ker νρ t ⇔ sνρ = tνρ definition of ker νρ ,

⇔ [s] = [t] definition of νρ ,
⇔sρt definition of ρ.
Therefore ρ = ker νρ and so every congruence is the kernel of a morphism.
Theorem 3.8. [The Fundamental Theorem of Morphisms for Semigroups] Let θ : S → T
be a semigroup morphism. Then ker θ is a congruence on S, Im θ is a subsemigroup of T
and S/ ker θ ∼
= Im θ.
Proof. Define θ̄ : S/ ker θ → Im θ by [a]θ̄ = aθ. We have
18 VICTORIA GOULD

[a] = [b] ⇔ a ker θ b

⇔ aθ = bθ
⇔ [a]θ̄ = [b]θ̄.
Hence θ̄ is well-defined and one-one. For any x ∈ Im θ we have x = aθ = [a]θ̄ and so θ̄ is
onto. Finally,


[a][b] θ̄ = [ab]θ̄ = (ab)θ = aθbθ = [a]θ̄[b]θ̄.
Therefore θ̄ is an isomorphism and S/ ker θ ∼ = Im θ. 
Note that the analogue of Theorem 3.8 holds for monoid to give us the The Fundamental
Theorem of Morphisms for Monoids.
Example 3.9. θ : B → (Z, +) given by (a, b)θ = a − b is a monoid morphism. Check that
θ is onto, so by FTH we have
B/ ker θ ∼
= Z.
Moreover, ker θ is the congruence given by
(a, b) ker θ (c, d) ⇔ a − b = c − d.

4. Ideals
Ideals play an important role in Semigroup Theory, but rather different to that they hold
in Ring Theory. The reason is that in case of rings, ALL homomorphisms are determined
by ideals, but in case of semigroups, only some are.

4.1. Notation
If A, B ⊆ S then we write

AB = {ab | a ∈ A, b ∈ B},
A2 = AA = {ab | a, b ∈ A}.
Note. A is a subsemigroup if and only if A 6= ∅ and A2 ⊆ A.
We write aB for {a}B = {ab | b ∈ B}.
For example

AaB = {xay | x ∈ A, y ∈ B}.

Facts:
(1) A(BC) = (AB)C therefore P(S) = {S | A ⊆ S}, equipped by the above-defined
operation, is a semigroup – the power semigroup of S.
(2) A ⊆ B ⇒ AC ⊆ BC and CA ⊆ CB for all A, B, C ∈ P(S).
 SEMIGROUP THEORY A LECTURE COURSE 19

(3) AC = BC 6⇒ A = B and CA = CB 6⇒ A = B, i.e. the power semigroup is

not cancellative - think of a right zero semigroup, there AC = BC = C for all
A, B, C ⊆ S.
(4) A is isomorphic to the subsemigroup {{a} | a ∈ A} of P(A).
(5) S 1 S = S = SS 1 .

Definition 4.1. Let ∅ = 6 I ⊆ S then I is

(1) a left ideal if SI ⊆ I (i.e. a ∈ I, s ∈ S ⇒ sa ∈ I);
(2) a right ideal if IS ⊆ I;
(3) an (two-sided) ideal if IS ∪ SI ⊆ I, that is, I is both a left and a right ideal.
Note that if S is commutative, (1),(2) and (3) above coincide.
If ∅ =
6 I ⊆ S then we have:
I is a left ideal ⇔ S 1 I ⊆ I;
I is a right ideal ⇔ IS 1 ⊆ I;
I is an ideal ⇔ S 1 IS 1 ⊆ I.
Note that any (left/right) ideal is a subsemigroup.
Example 4.2. (1) Let i ∈ I then {i} × J is a right ideal in a rectangular band I × J.
(2) Let m ∈ N be fixed. Then Im = {(x, y) | x > m, y ∈ N0 } is a right ideal in the
0

bicyclic semigroup B.
Indeed, let (x, y) ∈ Im and let (a, b) ∈ B. Then
(x, y)(a, b) = (x − y + t, b − a + t),
where t = max{y, a}. Now, we know that x ≥ m and that t ≥ y, so t − y ≥ 0.
Adding up these two inequalities, we get that x − y + t ≥ m, thus the product is
indeed in Im .
(3) If Y ⊆ X then we have {α ∈ TX | Im α ⊆ Y } is a left ideal of TX .
(4) For any n ∈ N we define

S n = {a1 a2 . . . an | ai ∈ S}.
This is an ideal of S. If S is a monoid then S n = S for all n, since for any s ∈ S
we can write

. . . 1} ∈ S n .
s = s |11 {z
n−1
(5) If S has a zero 0, then {0} (usually written 0), is an ideal.
Definition 4.3. Let S be a semigroup.
(1) We say that S is simple if S is the only ideal.
(2) If S has a zero 0, then S is 0-simple if S and {0} are the only ideals and S 2 6= 0.
Note that S 2 is always an ideal, so the condition S 2 6= 0 is only required to exclude the 2-
element null semigroup. A null semigroup is a semigroup with zero such that every product
equals 0 - notice that every subset containing 0 is an ideal.
20 VICTORIA GOULD

Example 4.4. Let G be a group and I a left ideal. Let g ∈ G, a ∈ I then we have

g = (ga−1 )a ∈ I
and so G = I. Therefore G has no proper left/right ideals. Hence G is simple.
Exercise: G0 is 0-simple
Example 4.5. We have (N, +) is a semigroup. Let n ∈ N. Now define In ⊆ (N, +) to be

In = {n, n + 1, n + 2, . . . },
which is an ideal. Hence N is not simple.
Note. {2, 4, 6, . . . } is a subsemigroup but not an ideal.
Example 4.6. The bicyclic semigroup B is simple.
Proof. Let I ⊆ B be an ideal, say (m, n) ∈ I. Then (0, n) = (0, m)(m, n) ∈ I. Thus
(0, 0) = (0, n)(n, 0) ∈ I. Let (a, b) ∈ B. Then

(a, b) = (a, b)(0, 0) ∈ I

and hence B = I ⇒ B is simple. 

4.2. Principal Ideals

We make note of how the S 1 notation can be used. For example

S 1 A = {sa | s ∈ S 1 , a ∈ A},
= {sa | s ∈ S ∪ {1}, a ∈ A},
= {sa | s ∈ S, a ∈ A} ∪ {1a | a ∈ A},
= SA ∪ A.
In particular, if A = {a} then S 1 a = Sa ∪ {a}. So,

S 1 a = Sa ⇔ a ∈ Sa,
⇔ a = ta
for some t ∈ S. We have S 1 a = Sa for a ∈ S if:
• S is a monoid (then a = 1a).
• a ∈ E(S) (then a = aa).
• a is regular, i.e. there exists x ∈ S with a = axa (then a = (ax)a).
But in (N, +) we have 1 6∈ 1 + N. Dually,

aS 1 = aS ∪ {a}
and similarly
 SEMIGROUP THEORY A LECTURE COURSE 21

S 1 aS 1 = SaS ∪ aS ∪ Sa ∪ {a}.
Claim. aS 1 (S 1 a, S 1 aS 1 ) is the “smallest” right (left, two-sided ideal) containing a.
Proof. (for aS 1 ).
We have a = a1 ∈ aS 1 and (aS 1 )S = a(S 1 S) ⊆ aS 1 . So, aS 1 is a right ideal containing a.
If a ∈ I and I is a right ideal, then aS 1 ⊆ IS 1 = I ∪ IS ⊆ I. 
Definition 4.7. We call aS 1 (S 1 a, S 1 aS 1 ) the principal right (left, two-sided) ideal gen-
erated by a.
If S is commutative then aS 1 = S 1 a = S 1 aS 1 .
Example 4.8. In a group G we have

aG1 = G = G1 a = G1 aG1
for all a ∈ G.
Example 4.9. In N under addition we have

n + “N1 ” = In = {n, n + 1, n + 2, . . . }
Example 4.10. B is simple, so

B(m, n)B = B 1 (m, n)B 1 = B

for all (m, n) ∈ B. However:

Claim. (m, n)B = (m, n)B 1 = (x, y) | x > m, y ∈ N0
Proof. We have

(m, n)B = (m, n)(u, v) | (u, v) ∈ B

⊆ (x, y) | x > m, y ∈ N0 .
Let x > m then



(m, n) n + (x − m), y = m − n + n + (x − m), y ,
= (x, y).

Therefore (x, y) ∈ (m, n)B ⇒ (x, y) | x > m, y ∈ N0 ⊆ (m, n)B. Hence we have proved
our claim. 

Dually we have B(m, n) = (x, y) | x ∈ N0 , y > n .
Lemma 4.11 (Principal Left Ideal Lemma). The following statements are equivalent;
i) S 1 a ⊆ S 1 b,
ii) a ∈ S 1 b,
22 VICTORIA GOULD

iii) a = tb for some t ∈ S 1 ,

iv) a = b or a = tb for some t ∈ S.
Note. If S 1 a = Sa and S 1 b = Sb, then the Lemma can be adjusted accordingly.
Proof. It is clear that (ii), (iii) and (iv) are equivalent.
(i) ⇒ (ii): If S 1 a ⊆ S 1 b then a = 1a ∈ S 1 a ⊆ S 1 b ⇒ a ∈ S 1 b.
(ii) ⇒ (i): If a ∈ S 1 b, then as S 1 a is the smallest left ideal containing a, and as S 1 b is a
left ideal we have S 1 a ⊆ S 1 b. 
Lemma 4.12 (Principal Right Ideal Lemma). The following statements are equivalent:
i) aS 1 ⊆ bS 1 ,
ii) a ∈ bS 1 ,
iii) a = bt for some t ∈ S 1 ,
iv) a = b or a = bt for some t ∈ S.
Note. If aS = aS 1 and bS = bS 1 then aS ⊆ bS ⇔ a ∈ bS ⇔ a = bt for some t ∈ S.
The following relation is crucial in semigroup theory.

Definition 4.13. The relation L on a semigroup S is defined by the rule

a L b ⇔ S 1a = S 1b
for any a, b ∈ S.
Note.
(1) L is an equivalence.
(2) If a L b and c ∈ S then S 1 a = S 1 b, so S 1 ac = S 1 bc and hence ac L bc, i.e. L is
right compatible. We call a right (left) compatible equivalence relation a right (left)
congruence. Thus L is a right congruence.
Corollary 4.14. We have that

a L b ⇔ ∃ s, t ∈ S 1 with a = sb and b = ta.

Proof.
a L b ⇔ S 1a = S 1b
⇔ S 1 a ⊆ S 1 b and S 1 b ⊆ S 1 a
⇔ ∃ s, t ∈ S 1 with a = sb, b = ta
by the Principal Left Ideal Lemma. 
We note that this statement about L can be used as a definition of L.
Remark.
(1) a L b ⇔ a = b or there exist s, t ∈ S with a = sb, b = ta.
 SEMIGROUP THEORY A LECTURE COURSE 23

(2) If Sa = S 1 a and Sb = S 1 b, then a L b ⇔ ∃ s, t ∈ S with a = sb, b = ta.

Dually, the relation R is defined on S by

a R b ⇔ aS 1 = bS 1
and

a R b ⇔ ∃ s, t ∈ S 1 with a = bs and b = at,

⇔ a = b or ∃ s, t ∈ S with a = bs and b = at.
We can adjust this if aS 1 = aS as before. Now R is an equivalence; it is left compatible
and hence a left congruence.

Definition 4.15. We define the relation H = L ∩ R and note that H is an equivalence.

The relations L, R, H are in fact three of the so-called Greens’ relations.

Example 4.16. (1) If S is commutative, L = R = H.
(2) In a group G,
G1 a = G = G1 b and aG1 = G = bG1 for all a, b ∈ G.
So a L b and a R b for all a, b ∈ G. Therefore L = R = ω = G × G and hence
we have H = ω.
Example 4.17. In N under + we have

a + N1 = {a, a + 1, . . . }
and so a + N1 = b + N1 ⇔ a = b. Hence L = R = H = ι.
Example 4.18. In B we know

(m, n)B 1 = (x, y) | x > m, y ∈ N0
and so we have

(m, n)B 1 = (p, q)B 1 ⇔ m = p.

Hence (m, n) R (p, q) ⇔ m = p. Dually,

(m, n) L (p, q) ⇔ n = q.
Thus (m, n) H (p, q) ⇔ (m, n) = (p, q), which gives us H = ι.
24 VICTORIA GOULD

4.3. L and R in TX
Claim. αTX ⊆ βTX ⇔ ker β ⊆ ker α.
(Recall ker α = (x, y) ∈ X × X | xα = yα ).
Proof. (⇒) Suppose αTX ⊆ βTX . Then α = βγ for some γ ∈ TX . Let (x, y) ∈ ker β. Then

xα = x(βγ) = (xβ)γ = (yβ)γ = y(βγ) = yα.

Hence (x, y) ∈ ker α and so ker β ⊆ ker α.
(⇐) Suppose ker β ⊆ ker α. Define γ : X → X by

z z∈/ Im β
zγ =
xα z = xβ

xβ
β
x γ

α
xα

If z = xβ = yβ, then (x, y) ∈ ker β ⊆ ker α so xα = yα. Hence γ is well-defined. So γ ∈ TX

and βγ = α. Therefore α ∈ βTX so that by the Principal Ideal Lemma, αTX ⊆ βTX . 
Corollary 4.19 (R–TX -Lemma). α R β ⇔ ker α = ker β.
Proof. We have

α R β ⇔ αTX = βTX
⇔ αTX ⊆ βTX and βTX ⊆ αTX
⇔ ker β ⊆ ker α and ker α ⊆ ker β
⇔ ker α = ker β.

Fact: TX α ⊆ TX β ⇔ Im α ⊆ Im β (See Exercises).
Corollary 4.20 (L − TX -Lemma). α L β ⇔ Im α = Im β.
Consequently α H β ⇔ ker α = ker β and Im α = Im β.
Example 4.21. Let us define
 
1 2 3
ε= ∈ E(T3 )
2 2 3
 SEMIGROUP THEORY A LECTURE COURSE 25

Now we have Im ε = {2, 3}. We can see that ker ε has classes {1, 2}, {3}. So

α H ε ⇔ Im α = Im ε and ker α = ker ǫ

⇔ Im α = {2, 3} and ker α has classes {1, 2}, {3}.
So we have
   
1 2 3 1 2 3
α= or α=ε=
3 3 2 2 2 3

ε α
ε ε α
α α ε

which is the table of a 2-element group. Thus the H-class of ε is a group.

5. Subgroups of Semigroups
Let S be a semigroup and let H ⊆ S. Then H is a subgroup of S if it is a group under the
restriction of the binary operation on S to H; i.e.
• a, b ∈ H ⇒ ab ∈ H
• ∃ e ∈ H with ea = a = ae for all a ∈ H
• ∀ a ∈ H ∃ b ∈ H with ab = e = ba
Remark.
(1) S does not have to be a monoid. Even if S is a monoid, e does not have to be 1.
However, e must be an idempotent, i.e. e ∈ E(S).
(2) If H is a subgroup with identity e, then e is the only idempotent in H.

S
H
e

Figure 2. e is the only idempotent in H.

(3) If e ∈ E(S),
 then {e}
 is a trivial
 subgroup.

1 2 3 1 2 3
(4) With α = and ǫ = we have the H-class {ǫ, α} is a subgroup
3 3 2 2 2 3
of T3 .
26 VICTORIA GOULD

(5) SX is a subgroup of TX . Notice

α H IX ⇔ Im α = Im IX and ker α = ker IX ,

⇔ Im α = X and ker α = ι,
⇔ α is onto and α is one-one,
⇔ α ∈ SX .
Therefore SX is the H-class of IX .
Definition 5.1. In the sequel, we are going to denote by La the L-class of a; by Ra the
R-class of a and by Ha the H-class of a.

Now La = Lb ⇔ a L b and Ha = La ∩ Ra . For example, in B, we have L(2,3) = (x, 3) |
x ∈ N0 .
We are going to show that the maximal subgroups of semigroups are just the H-classes of
idempotents. As a consequence, we will see that whenever two subgroups are not disjoint,
then they are both contained within a subgroup, as the following figure shows.

S
He
e

Figure 3. Existence of a Maximal Subgroup.

Lemma 5.2 (Principal Ideal for Idempotents). Let a ∈ S, e ∈ E(S). Then

(i) S 1 a ⊆ S 1 e ⇔ ae = a
(ii) aS 1 ⊆ eS 1 ⇔ ea = a.
Proof. (We prove part (i) only because (ii) is dual). If ae = a, then a ∈ S 1 e so S 1 a ⊆ S 1 e
by the Principal Ideal Lemma. Conversely, if S 1 a ⊆ S 1 e then by the Principal Ideal Lemma
we have a = te for some t ∈ S 1 . Then

ae = (te)e = t(ee) = te = a.

Corollary 5.3. Let e ∈ E(S). Then we have

a R e ⇒ ea = a,
a L e ⇒ ae = a,
a H e ⇒ a = ae = ea.
 SEMIGROUP THEORY A LECTURE COURSE 27

Thus, idempotents are left/right/two-sided identities for their R/L/H-classes.

Lemma 5.4. Let G be a subgroup with idempotent e. Then G ⊆ He , thus, the elements of
G are all H-related.
Proof. Let G be a subgroup with idempotent e. Then for any a ∈ G we have ea = a = ae
and there exists a−1 ∈ G with aa−1 = e = a−1 a. Then
)
ea = a
⇒aRe
aa−1 = e
)
ae = a
⇒aLe
a−1 a = e
⇒ a H e.
Therefore a H e for all a ∈ G, so G ⊆ He . 
Theorem 5.5 (Maximal Subgroup Theorem). Let e ∈ E(S). Then He is the maximal
subgroup of S with identity e.
Proof. We have shown that if G is a subgroup with identity e, then G ⊆ He .
We show now that He itself is a subgroup with identity e.
We know that e is an identity for He . Suppose a, b ∈ He . Then b H e, so b R e hence
ab R ae (R is left compatible) so
ab R ae = a R e.
Also, a L e ⇒ ab L eb = b L e hence ab H e so ab ∈ He . It remains to show that for all
a ∈ He there exists b ∈ He with ab = e = ba.
Let a ∈ He . Then, by definition of H = R ∩ L, there exist s, t ∈ S 1 with

at =e=
|{z} | {zsa} .
aRe aLe
We have

a(ete) = (ae)te = ate = ee = e = · · · = (ese)a.

Let x = ete, y = ese so x, y ∈ S and ex = xe = x, ey = ye = y. Also e = ax = ya. Now

x = ex = (ya)x = y(ax) = ye = y.
So let b = x = y. Then

eb
| = b {zba = e} be
| = b {zab = e}
bRe bLe
so b H e, thus b ∈ He . Hence He is indeed a subgroup. 
28 VICTORIA GOULD

Let e, f ∈ E(S) with e 6= f . Since He and Hf are subgroups containing the idempotents e
and f , respectively, He 6= Hf . This implies that He ∩ Hf = ∅.
Theorem 5.6. [Green’s Theorem] If a ∈ S, then a lies in a subgroup iff a H a2 .
Proof. See later. 
Corollary 5.7. Let a ∈ S. Then the following are equivalent:
(i) a lies in a subgroup,
(ii) a H e, for some e ∈ E(S),
(iii) Ha is a subgroup,
(iv) a H a2 .
Proof. (i) ⇒ (ii): If a ∈ G, then G ⊆ He where e2 = e is the identity for G. Therefore
a ∈ He so a H e.
(ii) ⇒ (iii): If a H e, then Ha = He and by the MST, He is a subgroup.
(iii) ⇒ (i): Straightforward, for a ∈ Ha .
(iii) ⇒ (iv) If Ha is a subgroup, then certainly Ha is closed. Hence a, a2 ∈ Ha therefore
a H a2 .
(iv) ⇒ (i) This follows from Greeen’s Theorem (Theorem 5.6).


Subgroups of Tn
We use Green’s Theorem to show the following.
Lemma 5.8. Let α ∈ Tn . Then α lies in a subgroup of Tn ⇔ the map diagram has no tails
of length > 2.
Proof. We have that
α lies in a subgroup ⇔ α H α2
⇔ α L α2 , α R α2
⇔ Im α = Im α2 , ker α = ker α2 .
We know Im α2 ⊆ Im α (as Tn α2 ⊆ Tn α). Let ρ be an equivalence on a set X. Recall

X/ρ = [x] | x ∈ X
We have seen that

|n/ ker α| = | Im α|.

We know that ker α ⊆ ker α (α Tn ⊆ αTn ), which means that the ker α2 -classes are just
2 2

unions of ker α-classes:

Claim. For α ∈ Tn , Im α = Im α2 ⇔ ker α = ker α2 .
 SEMIGROUP THEORY A LECTURE COURSE 29

ker α classes
ker α2 classes

Figure 4. The classes of ker α and ker α2 .

Proof.
n/ ker α2 = | Im α2 | ≤ | Im α| = |n/ ker α| .
Thus ker α and ker α2 have the same number of classes if and only if | Im α| = | Im α2 |. It
follows that ker α = ker α2 if and only if Im α = Im α2 . 

We now continue with the proof of Lemma 5.8:

We have that α lies in a subgroup ⇔ Im α = Im α2 . Note that elements of Im α \ Im α2

are exactly those second vertices of tails in the map diagram of α which are not members
of a cycle. Thus, Im α2 = Im α if and only if no such vertices exist, thus if and only if all
tails have length smaller than or equal to 1. 

An arbitrary element of Tn looks like:

α α
.
∈ Im α \ Im α2

Example 5.9.
(1) We take an element of T5 to be
 
1 2 3 4 5
α= ∈ T5 .
3 1 4 3 1
This has map diagram
30 VICTORIA GOULD

1 3 4

5
Now α has a tail with length > 2 and therefore α doesn’t lie in any subgroup.
(2) Let us take the constant element c1 ∈ T5
 
1 2 3 4 5
c1 = .
1 1 1 1 1
This has the following map diagram

2 3 4 5

Now c1 has no tails of length > 2, therefore c1 lies in a subgroup and hence c1 lies
in a subgroup. Note that actually c21 = c1 .
Now for any β,

β ∈ H c 1 ⇔ β H c1 ,
⇔ β R c1 and β L c1 ,
⇔ ker β = ker c1 and Im β = Im c1 ,
⇔ ker β has classes {1, 2, 3, 4, 5} and Im β = {1},
⇔ β = c1 .

Therefore the maximal subgroup containing c1 is Hc1 = {c1 }.

(3) Take the element
 
1 2 3 4 5
α= .
2 3 2 3 5
This has map diagram
 SEMIGROUP THEORY A LECTURE COURSE 31

3 4

1 2
5
No tails of length > 2. Therefore α lies in a subgroup. Hence α lies in a maximal
subgroup. Hence the maximal subgroup containing α is Hα . For any β

β ∈ Hα ⇔ β H α,
⇔ β R α and β L α,
⇔ ker β = ker α and Im β = Im α,
⇔ Im β = {2, 3, 5} and ker β has classes {1, 3}, {2, 4}, {5}.
We now figure out what the elements of Hα are. We start with the idempotent.
We know that the image of the idempotent is {2, 3, 5} and that idempotents are
identities on their images. Thus we must have
 
1 2 3 4 5
ε= .
2 3 5
We also know that 1 and 3 go to the same place and 2 and 4 go to the same place.
Thus we must have
 
1 2 3 4 5
ε= .
3 2 3 2 5
We now have what the idempotent is and then the other elements of Hα are (note
that 1 and 3 must have the same images, just as 2 and 4):

 
1 2 3 4 5
2 3 2 3 5
  
1 2 3 4 5 1 2 3 4 5
5 2 5 2 3 3 5 3 5 2
  
1 2 3 4 5 1 2 3 4 5
.
5 3 5 3 2 2 5 2 5 3

These are all 6 elements.

Check Hα ≃ S3 .
32 VICTORIA GOULD

6. D, J and Green’s Lemmas

Recall S 1 aS 1 = {xay | x, y ∈ S 1 }.

Definition 6.1. We say that a J b if and only if

a J b ⇔ S 1 aS 1 = S 1 bS 1
Check:
a J b ⇔ ∃ s, t, u, v ∈ S 1 with a = sbt b = uav.
Note. If a L b, then S 1 a = S 1 b so S 1 aS 1 = S 1 bS 1 so a J b, i.e. L ⊆ J , dually R ⊆ J .
Recall: S is simple if S is the only ideal of S. If S is simple and a, b ∈ S then

S 1 aS 1 = S = S 1 bS 1 so a J b
and J = ω (the universal relation). Conversely if J = ω and I is an ideal of S, then pick
any a ∈ I and any s ∈ S. We have

s ∈ S 1 sS 1 = S 1 aS 1 ⊆ I.
Therefore I = S and S is simple.

We have shown that that

S is simple ⇔ J = ω.

Similarly if S has a zero, then {0} and S \ {0} are the only J -classes iff {0} and S are the
only ideals.

6.1. Composition of Relations

Definition 6.2. If ρ and λ are relations on A we define

ρ ◦ λ = (x, y) ∈ A × A | ∃ z ∈ A with (x, z) ∈ ρ and (z, y) ∈ λ .
Lemma 6.3. If ρ, λ are equivalence relations and if ρ◦λ = λ◦ρ then ρ◦λ is an equivalence
relation. Also, it is the smallest equivalence relation containing ρ ∪ λ.
Proof. Put ν = ρ ◦ λ = λ ◦ ρ
• for any a ∈ A, a ρ a λ a so a ν a and ν is reflexive.
• Symmetric - an exercise.
 SEMIGROUP THEORY A LECTURE COURSE 33

• Suppose that a ν b ν c then there exists x, y ∈ A with

a ρ x λ b λ y ρ c.
(Note that first we use that ν = ρ ◦ λ, and next we use that ν = λ ◦ ρ.)
From x λ b λ y we have x λ y, so

a ρ x λ y ρ c.
Therefore x ν c hence there exists z ∈ A such that x ρ z λ c, therefore a ρ z λ c
and hence a ν c. Therefore ν is transitive.
We have shown that ν is an equivalence relation. If (a, b) ∈ ρ then a ρ b λ b so (a, b) ∈ ν.
Similarly if (a, b) ∈ λ then a ρ a λ b so (a, b) ∈ ν. Hence ρ ∪ λ ⊆ ν.
Now, suppose ρ ∪ λ ⊆ τ where τ is an equivalence relation. Let (a, b) ∈ ν. Then we have
a ρ c λ b for some c. Hence a τ c τ b so a τ b as τ is transitive. Therefore ν ⊆ τ . 
The smallest equivalence relation containing any ρ and λ is denoted by ρ ∨ λ; we have
shown that if ρ and λ commute, then ρ ∨ λ = ρ ◦ λ.
Definition 6.4. D = R ◦ L, i.e. a D b ⇔ ∃ c ∈ S with a R c L b.
Lemma 6.5 (The D Lemma). R ◦ L = L ◦ R
Proof. We prove that R ◦ L ⊆ L ◦ R, the proof of the other direction being dual. Suppose
that a R ◦ L b. Then there exists c ∈ S with

aRcLb
1
There exists u, v, s, t ∈ S with

a = cu c = av c = sb b = tc.
(1) (2) (3) (4)

Put d = bu then we have

a = cu = sbu = sd,
(1) (3)

d = bu = tcu = ta.
(4) (1)

Therefore a L d. Also

b = tc = tav = tcuv = buv = dv.

(4) (2) (1) (4)

Therefore b R d and hence a L ◦ R b. 

Hence D is an equivalence relation and D = L ∨ R.
By definition
34 VICTORIA GOULD

H = L ∩ R ⊆ L ⊆ D,
H = L ∩ R ⊆ R ⊆ D.
As J is an equivalence relation and L ∪ R ⊆ J we must have D ⊆ J . This has Hasse
Diagram
J

L R

Notation: Da is the D class of a ∈ S and Ja is the J -class of a ∈ S.

Note. Ha ⊆ La ⊆ Da ⊆ Ja and also Ha ⊆ Ra ⊆ Da ⊆ Ja .

Egg-Box Pictures
Let D be a D-class. Then for any a ∈ D we have Ra ⊆ D = Da , and La ⊆ D. We
denote the R-classes as rows and the L-classes as columns. The cells (if non-empty) will
be H-classes - we show they are all non-empty!
Let u, v ∈ D then u D v. This implies that there exists h ∈ S with u R h L v, so
Ru ∩ Lv 6= ∅, that is, no cell is empty. Moreover

Ru ∩ Lv = Rh ∩ Lh = Hh .
As D is an equivalence, S is the union of such “egg-boxes”: the rows represent the R-
classes, and the columns represent the L-classes.

u h
v

6.2. Structure of D-classes

Let S be a semigroup, s ∈ S 1 . We define ρs : S → S by aρs = as for all a ∈ S
Lemma 6.6 (Green’s Lemma). Let a, b ∈ S be such that a R b and let s, s′ ∈ S be such
that

as = b and bs′ = a.
 SEMIGROUP THEORY A LECTURE COURSE 35

Then ρs : La → Lb and ρs′ : Lb → La are mutually inverse, R-class preserving bijections

(i.e. if c ∈ La , then c R cρs and if d ∈ Lb then d R dρs′ ).
Proof. If c ∈ La then

cρs = cs L as = b,
because L is a right congruence. So cρs L b therefore ρs : La → Lb . Dually ρs′ : Lb → La .
Let c ∈ La . Then c = ta for some t ∈ S. Now

cρs ρs′ = tasρs′ = tass′ = tbs′ = ta = c.

So ρs ρs′ = ILa , dually, ρs′ ρs = ILb .
Again, let c ∈ La . Then

cs = c · s,
c = cs · s′ .
Therefore c R cs = cρs . 

Continuing Lemma 6.6. For any c ∈ La we have ρs : Hc → Hcs is a bijection with

inverse ρs′ : Hcs → Hc . In particular – put c = a then

ρs : Ha → Hb and ρs : Hb → Ha
are mutually inverse bijections.

Let s ∈ S 1 . Then we define λs : S → S by aλs = sa.

Lemma 6.7 (Dual of Green’s Lemma). Let a, b ∈ S be such that a L b and let t, t′ ∈ S
be such that ta = b and t′ b = a. Then λt : Ra → Rb and λt′ : Rb → Ra are mutually
inverse L-class preserving bijections. In particular, for any c ∈ Ra we have λt : Hc → Htc ,
λt′ : Htc → Hc are mutually inverse bijections. So, if c = a we have λt : Ha → Hb ,
λt′ : Hb → Ha are mutually inverse bijections.
Corollary 6.8. If a D b then there exists a bijection Ha → Hb .
Proof. If a D b then there exists h ∈ S with a R h L b. There exists a bijection Ha → Hh
by Green’s Lemma and we also have that there exists a bijection Hh → Hb by the Dual of
Green’s Lemma. Therefore there exists a bijection Ha → Hb . 
Thus any two H-classes in the same D-class have the same cardinality (just like any two
R- and L-classes).
Theorem 6.9 (Green’s Theorem – Strong Version). Let H be an H-class of a semigroup
S. Then either H 2 ∩ H = ∅ or H is a subgroup of S.
36 VICTORIA GOULD

Proof. We prove that if H 2 ∩ H 6= ∅, then H is a subgroup. This is exactly the statement

of the theorem.
So suppose H 2 ∩ H 6= ∅. Then there exists a, b, c ∈ H such that ab = c. Since a R c,
ρb : Ha → Hc is a bijection. But Ha = Hc = H so ρb : H → H is a bijection. Hence
Hb = H. Dually, aH = H.
Let u, v ∈ H. Then av ∈ H so that as above, Hv = H. But then uv ∈ H and H is a
subsemigroup. Further, vH = H so that by a standard argument (see Exercises 1), H is a
subgroup of S.
Alternatively Since b ∈ H, b = db for some d ∈ H. As b R d, d = bs for some s ∈ S 1 and
then d = bs = dbs = d2 . Hence H contains an idempotent, so (by the Maximal Subgroup
Theorem) it is a subgroup. 
Corollary 6.10. a H a2 ⇔ Ha is a subgroup.
Proof. We know Ha is a subgroup ⇒ a, a2 ∈ Ha so a H a2 .
Conversely, if aHa2 , then a2 ∈ Ha ∩ (Ha )2 . Hence Ha ∩ (Ha )2 6= ∅. So, by Green’s Lemma,
Ha is a subgroup. 

7. Rees Matrix Semigroups

Just as the main building blocks of groups are simple groups, the main building blocks of
semigroups are 0-simple semigroups.
In general, the structure of 0-simple semigroups is very complicated. In the finite case and,
more generally, in case certain chain conditions hold, their structure is transparent - they
can be described by a group and a matrix.

Construction: Let G be a group, let I, Λ be non-empty sets and let P be a Λ × I matrix

over G ∪ {0} such that every row and every column of P contains at least one non-zero
entry.
M0 = M0 (G; I, Λ; P ) is the set

I × G × Λ ∪ {0}
with binary operation given by 0n = 0 = n0 for all n ∈ M0 and
(
0 if pλk = 0,
(i, a, λ)(k, b, µ) =
(i, apλk b, µ) if pλk 6= 0.
Check that M0 (G; I, Λ; P ) is a semigroup with zero 0.

Definition 7.1. M0 = M0 (G; I, Λ; P ) is called a Rees Matrix Semigroup over G.

 SEMIGROUP THEORY A LECTURE COURSE 37

Definition 7.2. a ∈ S is regular if there exists x ∈ S with

a = axa.
S is regular if every a ∈ S is regular.

If S is regular then a R b ⇔ aS = bS ⇔ there exists s, t ∈ S with a = bs and b = at, etc.

Proposition 7.3. Rees matrix facts Let M0 = M0 (G; I, Λ; P ) be a Rees Matrix Semi-
group over a group G.
(1) (i, a, λ) is idempotent ⇔ pλi 6= 0 and a = p−1
λi .
0
(2) M is regular.
(3) (i, a, λ) R (j, b, µ) ⇔ i = j.
(4) (i, a, λ) L (j, b, µ) ⇔ λ = µ.
(5) (i, a, λ) H (j, b, µ) ⇔ i = j and λ = µ.
(6) The D = J -classes are {0} and M0 \ {0} (so 0 and M0 are the only ideals).
(7) M0 is 0-simple.
(8) The so-called rectangular property:

xy D x ⇔ xy R x
∀ x, y ∈ M0
xy D y ⇔ xy L y
Proof. (1) We have that

(i, a, λ) ∈ E(M0 ) ⇔ (i, a, λ) = (i, a, λ)(i, a, λ),

⇔ pλi 6= 0, (i, a, λ) = (i, apλi a, λ),
⇔ pλi 6= 0, a = apλi a,
⇔ pλi 6= 0 and pλi = a−1 .
(2) 0 = 000 so 0 is regular. Let (i, a, λ) ∈ M0 \ {0} then there exists j ∈ I with pλj 6= 0
and there exists µ ∈ Λ with pµi 6= 0. Now,

(i, a, λ)(j, p−1 −1 −1

λj a pµi , µ)(i, a, λ) = (i, a, λ)
and hence M0 is regular.
(3) {0} is an R-class. If (i, a, λ) R (j, b, µ) then there exists (k, c, ν) ∈ M0 with

(i, a, λ) = (j, b, µ)(k, c, ν) = (j, bpµk c, ν)

and so i = j. Conversely, if i = j, pick k with pµk 6= 0. Then

(i, a, λ) = (j, b, µ)(k, p−1 −1

µk b a, λ)
and together with the dual we have (i, a, λ) R (j, b, µ)
(4) Dual.
(5) This comes from (3) and (4) above.
38 VICTORIA GOULD

(6) {0} is a D-class and a J -class. If (i, a, λ), (j, b, µ) ∈ M0 then

(i, a, λ) R (i, a, µ) L (j, b, µ)

so (i, a, λ) D (j, b, µ) and so (i, a, λ) J (j, b, µ). Therefore D = J and {0} and
M0 \ {0} are the only classes.
(7) We have already shown that the only J -classes are {0} and M0 \ {0}. Let i ∈ I,
then there exists λ ∈ Λ with pλi 6= 0 so (i, 1, λ)2 6= 0. Therefore (M0 )2 6= 0 and so
M0 is 0-simple.
(8) If xy R x, then clearly xy D x, because R ⊆ D. For the other direction, suppose
that xy D x. Notice that the two D-classes are zero and everything else. If xy = 0,
then necessarily x = 0, because D0 = {0}. If xy 6= 0, then necessarily x, y 6= 0, so
we have that

x = (i, a, λ) y = (j, b, µ).

Then xy = (i, apλj b, µ), so xy R x. The result for L is dual. 

Some more facts!

(9) Put Hiλ = (i, a, λ) | a ∈ G . By (5) we have Hiλ is an H-class (Hiλ = H(i,e,λ) ). If
pλi 6= 0 we know (i, p−1
λi , λ) is an idempotent and so Hiλ is a group, by the Maximal
Subgroup Theorem. The identity is (i, p−1 λi , λ) and (i, a, λ)
−1
= (i, p−1 −1 −1
λi a , pλi , λ).
(10) If pλi 6= 0 and pµj 6= 0 then Hiλ ≃ Hjµ . It is clear that (i, a, λ) 7→ (j, a, µ) is a
bijection, but this is not in general a morphism. Exercise: find a morphism!

Chain conditions
A finitary property is a property held by all finite semigroups: chain conditions are one
kind of finitary property.

Definition 7.4. A semigroup S has ML if there are no infinite chains

S 1 a1 ⊃ S 1 a2 ⊃ S 1 a3 ⊃ . . .
of principal left ideals. ML is the descending chain condition (d.c.c.) on principal left
ideals.
The left/right dual is MR .
Lemma 7.5 (The Chain Lemma). The semigroup S has ML if and only if any chain

S 1 a1 ⊇ S 1 a2 ⊇ . . .
terminates (stabilizes) i.e. there exists n ∈ N with

S 1 an = S 1 an+1 = . . .
 SEMIGROUP THEORY A LECTURE COURSE 39

Proof. If every chain with ⊇ terminates, then clearly we cannot have an infinite strict chain

S 1 a1 ⊃ S 1 a2 ⊃ . . .
So S has ML .
Conversely, suppose S has ML and we have a chain

S 1 a1 ⊇ S 1 a2 ⊇ . . .
Let the strict inclusions be at the ji th steps:

S 1 a1 = S 1 a2 = · · · = S 1 aj1 ⊃ S 1 aj1 +1 = S 1 aj1 +2

= · · · = S 1 aj2 ⊃ S 1 aj2 +1 = . . .
Then
S 1 aj1 ⊃ S 1 aj2 ⊃ . . .
As S has ML , this chain is finite with length n say. Then

S 1 ajn +1 = S 1 ajn +2 = . . .
and our sequence has stabilised. 

Definition 7.6. The ascending chain condition (a.c.c.) on principal ideals on left/right
ideals M L (M R ) is defined as above but with the inclusions reversed.
The analogue of the Chain Lemma holds for M L and (M R ).
Example 7.7. Every finite semigroup has ML , MR , M L , M R . For example, if

S 1 a1 ⊃ S 1 a2 ⊃ S 1 a3 ⊃ . . . ,
then in every step, the cardinality of the sets must decrease at least by one, so the length
of a strict sequence cannot be greater than |S|.
Example 7.8. The Bicyclic semigroup B has M L and M R . We know

B(x, y) = (p, q) | q > y
and so
B(x, y) ⊆ B(u, v) ⇔ y > v,
and inclusion is strict if and only if y > v. If we had an infinite chain

B(x1 , y1 ) ⊂ B(x2 , y2 ) ⊂ B(x3 , y3 ) ⊂ . . .

then we would have

y1 > y2 > y3 > . . . ,

which is impossible in N.
Hence M L holds, dually M R holds.
40 VICTORIA GOULD

However, since 0 < 1 < 2 < . . . we have

B(0, 0) ⊃ B(1, 1) ⊃ B(2, 2) ⊃ . . .

so there exists infinite descending chains. Hence B does not have ML or MR .
Example 7.9. Let M0 = M0 (G; I; Λ; P ) be a Rees Matrix Semigroup over a group G.
Then M0 has ML , MR , M L and M R .
Proof. We show that the length of the strict chains is at most 2. Suppose αM0 ⊆ βM0 .
We could have α = 0. If α 6= 0 then αM0 6= {0} so β 6= 0 and we have α = (i, g, λ),
β = (j, h, µ) and α = βγ for some γ = (ℓ, k, ν). Then

(i, g, λ) = (j, h, µ)(ℓ, k, ν) = (j, hρµℓ k, ν).

This gives us that i = j and so α R β and αM0 = βM0 .
Summarising, 0M0 ⊂ αM0 for all non-zero α. But if α 6= 0 and αM0 ⊆ βM0, then
αM0 = βM0 . Hence M0 has MR and M R ; dually M0 has ML and M L . 
Definition 7.10. A 0-simple semigroup is completely 0-simple if it has MR and ML .

By above, any Rees Matrix Semigroup over a group is completely 0-simple. Our aim is to
show that every completely 0-simple semigroup is isomorphic to a Rees Matrix Semigroup
over a group.
Theorem 7.11 (The D = J Theorem). Suppose
(
∀ a ∈ S, ∃ n ∈ N with an L an+1 ,
(⋆)
∀ a ∈ S, ∃ m ∈ N with am R am+1 .
Then D = J .
Example 7.12.
(1) If S is a band, a = a2 for all a ∈ S and so (⋆) holds.
(2) Let S be a semigroup having ML and let a ∈ S. Then
S 1 a ⊇ S 1 a2 ⊇ S 1 a3 ⊇ . . . .
Since S has ML , we have that this sequence stabilizes, so there exists n ∈ N such
that S 1 an = S 1 an+1 which means that an L an+1 . Similarly, if S has MR , then for
every a ∈ S there exists m ∈ N such that am R am+1 .
Proof. of D = J Theorem
We know D ⊆ J . Let a, b ∈ S with a J b. Then there exists x, y, u, v ∈ S 1 with

b = xay, a = ubv.
Then

b = xay = x(ubv)y = (xu)b(vy) = (xu)2 b(vy)2 = · · · = (xu)n b(vy)n

 SEMIGROUP THEORY A LECTURE COURSE 41

for all n ∈ N. By (⋆), there exists n with (xu)n L (xu)n+1 . Therefore

b = (xu)n b(vy)n L (xu)n+1 b(vy)n = xu (xu)n b(vy)n = xub.
Therefore b L xub, so

S 1 b = S 1 xub ⊆ S 1 ub ⊆ S 1 b.
So S 1 b = S 1 ub, which means that b L ub. Dually, b R bv. Therefore a = ubv R ub L b. So
a D b and J ⊆ D. Consequently, D = J . 
As a consequence we have the following:
Corollary 7.13. If a semigroup S has ML and MR , then it satisfies (⋆) and thus D = J .
In the same vein we have:
Lemma 7.14. The Rectangular Property:
Let S satisfy (⋆). Then for all a, b ∈ S we have
(i) a J ab ⇔ a D ab ⇔ a R ab,
(ii) b J ab ⇔ b D ab ⇔ b L ab.
Proof. We prove (i), (ii) being dual. Now,

a J ab ⇔ a D ab
as D = J . Clearly if a R ab then a D ab; as R ⊆ D.
Conversely, If a J ab then there exists x, y ∈ S 1 with

a = xaby = xa(by) = xn a(by)n

for all n. Pick n with (by)n R (by)n+1 . Then

a = xn a(by)n R xn a(by)n+1 = xn a(by)n by = aby.

Now
aS 1 = abyS 1 ⊆ abS 1 ⊆ aS 1 .
Hence aS 1 = abS 1 and a R ab. 

7.1. Completely 0-simple semigroups

Let S have a 0. Recall that S is 0-simple if and only if 0 (properly, {0}) and S are the
only ideals and S 2 6= 0. If in addition S has MR and ML , then S is completely 0-simple.
Lemma 7.15. [0-Simple Lemma] Let S have a 0 and S 2 6= 0. Then the following are
equivalent:
(i) S is 0-simple,
(ii) SaS = S for all a ∈ S \ {0},
(iii) S 1 aS 1 = S for all a ∈ S \ {0},
42 VICTORIA GOULD

(iv) the J -classes are {0} and S \ {0}.

Proof. (i) ⇔ (iii) ⇔ (iv) is a standard exercise.

(ii) ⇒ (iii): Let a ∈ S \ {0}. Then

S = SaS ⊆ S 1 aS 1 ⊆ S
and therefore S = S 1 aS 1 .

(i) ⇒ (ii): Since S 2 6= 0 and S 2 is an ideal, then S 2 = S. Therefore

S 3 = SS 2 = S 2 = S 6= 0.
Let I = {x ∈ S | SxS = 0}. Clearly 0 ∈ I and hence I 6= ∅. If x ∈ I and s ∈ S, then

0 ⊆ SxsS ⊆ SxS = 0.
Therefore SxsS = 0 and so xs ∈ I. Dually sx ∈ I; therefore I is an ideal. If I = S, then

S 3 = SIS,
[
= SxS,
x∈I
= 0.
This is a contradiction, therefore I 6= S. Hence I = 0. Let a ∈ S \ {0}. Then SaS is an
ideal and as a 6∈ I we have SaS 6= 0. Hence SaS = S. 
Corollary 7.16. Let S be completely 0-simple. Then S contains a non-zero idempotent.
Proof. Let a ∈ S \ {0}. Then SaS = S, therefore there exists a u, v ∈ S with a = uav. So,

a = uav = u2 av 2 = · · · = un av n
for all n. Hence un 6= 0 for all n ∈ N. Pick n, m with un R un+1 , um L um+1 . Notice

un+1 R un+2
as R is a left congruence. Similarly,

un+2 R un+3
we deduce that un R un+t for all t > 0. Similarly um L um+t for all t > 0. Let s =
max{m, n}. Then us R u2s , us L u2s so us H u2s = (us )2 . Hence by Corollary 5.7, us lies
in a subgroup. Therefore us H e for some idempotent e. As us 6= 0 and H0 = {0}, we have
e 6= 0. 
Theorem 7.17 (Rees’ Theorem - 1941). Let S be a semigroup with zero. Then S is
completely 0-simple ⇔ S is isomorphic to a Rees Matrix Semigroup over a group.
 SEMIGROUP THEORY A LECTURE COURSE 43

Proof. If S ∼
= M0 (G; I; Λ; P ) where G is a group, we know M0 is completely 0-simple (by
Proposition 7.3, Rees Matrix facts and Example 7.9), hence S is completely 0-simple.
Conversely, suppose that S is completely 0-simple. By the D = J Theorem, D = J (as
S has MR and ML , it must have (⋆)). As S is 0-simple, the D = J -classes are {0} and
S \ {0}. Let D = S \ {0}. By Corollary 7.16, D contains an idempotent e = e2 .
Let {Ri | i ∈ I} be the set of R-classes in D (so I indexes the non-zero R-classes). Let
{Lλ | λ ∈ Λ} be the set of L-classes in D (so Λ indexes the non-zero L-classes).
Denote the H-class Ri ∩ Lλ by Hiλ . Since D contains an idempotent e, D contains the sub-
group He (Maximum Subgroup Theorem or Green’s Theorem). Without loss of generality
we can assume that both I and Λ contain a special symbol 1, and we can also assume that
e ∈ H11 . Put G = H11 , which is a group.
For each λ ∈ Λ let us choose and fix an arbitrary qλ ∈ H1λ (take q1 = e).
Similarly, for each i ∈ I let ri ∈ Hi1 (take r1 = e).
Notice that

e = e2 , e R qλ ⇒ eqλ = qλ
Thus, by Green’s Lemma,

ρqλ : He = G → H1λ
is a bijection. Now,
e = e2 , e L ri ⇒ ri e = ri .
By the dual of Green’s Lemma

λri : H1λ → Hiλ

is a bijection. Therefore for any i ∈ I, λ ∈ Λ we have

ρqλ λri : G → Hiλ

is a bijection.
Note. By the definition of ρqλ and λri , we have that
aρqλ λri = ri aqλ
for every a ∈ G, i ∈ I and λ ∈ Λ.
So, each element of Hiλ has a unique expression as ri aqλ where a ∈ G. Hence the mapping

θ : (I × G × Λ) ∪ {0} → S
given by 0θ = 0, (i, a, λ)θ = ri aqλ is a bijection.
Put pλi = qλ ri . If pλi 6= 0 then qλ ri D qλ D ri . By the rectangular property
e R qλ R qλ ri L ri L e
so that qλ ri ∈ G.
44 VICTORIA GOULD

L1 Lλ
R1 a qλ

Ri ri ri aqλ

So, P = (pλi ) = (qλ ri ) is a Λ × I matrix over G ∪ {0}. For any i ∈ I, by the 0-simple
Lemma (Lemma 7.15) we have Sri S = S. So, uri v 6= 0 for some u, v ∈ S. Say, u = rk bqλ
for some k, λ and b. Then

pλi = qλ ri 6= 0
as rk bqλ ri v 6= 0. Therefore every column of P has a non-zero entry. Dually for rows.
Therefore

M0 = M0 (G; I; Λ; P )
is a Rees Matrix Semigroup over a group G. For any x ∈ M0 (x = 0 or x is a triple) then

(0x)θ = 0θ = 0 = 0(xθ) = 0θxθ.

Also, (x0)θ = xθ0θ. For (i, a, λ), (k, b, µ) ∈ M0 we have
(

0θ if pλk = 0,
(i, a, λ)(k, b, µ) θ =
(i, apλk b, µ)θ if pλk 6= 0,
(
0 if pλk = 0,
=
ri apλk bqµ if pλk 6= 0,
= ri apλk bqµ ,
= ri aqλ rk bqµ ,
= (i, a, λ)θ(k, b, µ)θ.
Therefore θ is a morphism, and since it is bijective, it is an isomorphism. 

8. Regular Semigroups
Definition 8.1. We say that a ∈ S is regular if a = axa for some x ∈ S. The semigroup
S is regular if every a ∈ S is regular.
 SEMIGROUP THEORY A LECTURE COURSE 45

Examples of regular semigroups: any band, Rees matrix semigroups, groups.

Examples of non-regular semigroups: (N, +), (Z, ∗)
Nontrivial null (or zero) semigroups i.e. S = X ∪ {0} with X 6= ∅ and all products are 0.
Note that (N, +) has no regular element.

Definition 8.2. An element a′ ∈ S is an inverse of a if

a = aa′ a and a′ = a′ aa′ .
We denote by V (a) the set of inverses of a.
If G is a group then V (a) = {a−1 } for all a ∈ G.
Caution: Inverses need not be unique. For example, in a rectangular band T = I × Λ,

(i, j)(k, ℓ)(i, j) = (i, j)

(k, ℓ)(i, j)(k, ℓ) = (k, ℓ)
for any (i, j) and (k, ℓ). So every element is an inverse of every other element.
Lemma 8.3. If a ∈ S, then a is regular ⇔ V (a) 6= ∅.
Proof. If V (a) 6= ∅, clearly a is regular. Conversely suppose that a is regular. Then there
exists x ∈ S with a = axa. Put a′ = xax. Then

aa′ a = a(xax)a = (axa)xa = axa = a,

and
a′ aa′ = (xax)a(xax) = x(axa)(xax)
= xa(xax) = x(axa)x = xax = a′ .
So a′ ∈ V (a). 
Note. If a = axa then

(ax)2 = (ax)(ax) = (axa)x = ax

so ax ∈ E(S) and dually, xa ∈ E(S). Moreover

a = axa ax = ax ⇒ a R ax,
a = axa xa = xa ⇒ a L xa.

Definition 8.4. S is inverse if |V (a)| = 1 for all a ∈ S, i.e. every element has a unique
inverse.
Example 8.5.
(1) Groups are inverse; V (a) = {a−1 }.
(2) A rectangular band T is regular; but (as every element of T is an inverse of every
other element) T is not inverse (unless T is trivial).
46 VICTORIA GOULD

a ax

xa

Figure 5. The egg box diagram of Da .

(3) If S is a band then S is regular as e = e3 for all e ∈ S; S need not be inverse.

(4) B is regular because (a, b) = (a, b)(b, a)(a, b) for all (a, b) ∈ B. Furthermore, B is
inverse - see later.
(5) M0 is regular (see “Proposition 7.3, Rees Matrix Facts”).
(6) TX is regular (see Exercises).
(7) (N, +) is not regular as, for example 1 6= 1 + a + 1 for any a ∈ N.
Theorem 8.6. [Inverse Semigroup Theorem] A semigroup S is inverse iff S is regular and
E(S) is a semilattice (i.e. ef = f e for all e, f ∈ E(S)).
Proof. (⇐) Let a ∈ S. As S is regular, a has an inverse by Lemma 8.3. Suppose x, y ∈ V (a).
Then

a = axa x = xax a = aya y = yay,

(1) (2) (3) (4)

so ax, xa, ay, ya ∈ E(S). This gives us that

x = xax = x(aya)x = (xa)(ya)x = (ya)(xa)x = y(axa)x

(2) (3)

= yax = y(aya)x = y(ay)(ax) = y(ax)(ay) = y(axa)y = yay = y.

(1) (3) (1) (4)

So |V (a)| = 1 and S is inverse.

Conversely, suppose S is inverse. Let a′ denote the unique inverse of a ∈ S.

Certainly S is regular. Let e ∈ E(S). Then e is an inverse of e, because e = eee and
e = eee, so the inverse of any idempotent e is just itself: e′ = e.

Let e, f ∈ E(S). Let x = (ef )′ . Consider the element f xe. Then

(f xe)2 = (f xe)(f xe) = f (xef x)e = f xe

as x = (ef )′ . So f xe ∈ E(S) and therefore f xe = (f xe)′ .
 SEMIGROUP THEORY A LECTURE COURSE 47

We want to show that f xe and ef are mutually inverse:

ef (f xe)ef = ef 2 xe2 f = ef xef = ef,
(f xe)ef (f xe) = f xe2 f 2 xe = f (xef x)e = f xe.

Therefore we have ef = (f xe)′ = f xe ∈ E(S), so the product of any two idempotents is

an idempotent. Therefore E(S) is a band. Let e, f ∈ E(S). Then

ef (f e)ef = ef 2 e2 f = ef ef = ef

and f e(ef )f e = f e similarly. Therefore we have ef = (f e)′ = f e. 

Example 8.7.
(1) Let B be the Bicyclic Semigroup. Then

E(B) = (a, a) | a ∈ N0 ,
and

(a, a)(b, b) = (t, t) = (b, b)(a, a)

where t = max{a, b}. So E(B) is commutative, and since B is regular, we have
that it is inverse. Note that (a, b)′ = (b, a).
(2) TX – we know TX is regular. For |X| > 2 let x, y ∈ X with x 6= y we have
cx , cy ∈ E(TX ). Then cx cy 6= cy cx so TX is not inverse.
(3) If S is a band, then S is regular. Furthermore we have

S is inverse ⇔ ef = f e for all e, f ∈ E(S),

⇔ ef = f e for all e, f ∈ S,
⇔ S is a semilattice.

(4) Let M0 = M0 (G; I, Λ; P ). If pλi , pµi are both non-zero, then

(i, p−1 −1 0
λi , λ), (i, pµi , µ) ∈ E(M )

and
(i, p−1 −1 −1 −1
λi , λ)(i, pµi , µ) = (i, pµi , µ)(i, pλi , λ)

if and only if λ = µ. So for M0 to be inverse, for every i ∈ I there must be exactly

one λ ∈ Λ with pλi 6= 0; dually for each κ ∈ Λ there exists exactly one j ∈ I with
pκj 6= 0.
It is an Exercise to check that, conversely, if the above condition holds then M0
is inverse and isomorphic to a Brandt semigroup.
48 VICTORIA GOULD

8.1. Green’s Theory for Regular D-classes

If e ∈ E(S) then He is a subgroup of S (by the Maximal Subgroup Theorem or Green’s
Theorem). If e D f then |He | = |Hf | (by the Corollary to Green’s Lemmas). We will show
that He ∼
= Hf .
Lemma 8.8. We have that
(i) If a = axa then ax, xa ∈ E(S) and ax R a L xa,
(ii) If b R f ∈ E(S), then b is regular;
(iii) If b L f ∈ E(S), then b is regular.
Proof.
(i) We have already proven this.
(ii) If b R f then f b = b. Also, f = bs for some s ∈ S 1 . Therefore b = f b = bsb and it
follows that b is regular.
(iii) Dual to (ii).

From Lemma 8.8 an element a ∈ S is regular if and only if it is R-related to an idempotent.
Dually, a ∈ S is regular if and only if it is L-related to an idempotent.
Lemma 8.9 (Regular D-class Lemma). If a D b then if a is regular, so is b.
Proof. Let a be regular with a D b . Then a R c L b for some c ∈ S.

a e c

Figure 6. The egg box diagram of D.

There exists e = e2 with e R a R c by (i) above. By (ii), c is regular. By (i), c L f = f 2 .

By (iii), b is regular. 
Corollary 8.10. [Corollary to Green’s Lemmas] Let e, f ∈ E(S) with e D f . Then
He ∼
= Hf .
Proof. Suppose e, f ∈ E(S) and e D f . There exists a ∈ S with e R a L f .
As e R a there exists s ∈ S 1 with e = as and ea = a. So a = asa. Put x = f se. Then
 SEMIGROUP THEORY A LECTURE COURSE 49

a e

ax = af se = ase = e2 = e
and so a = ea = axa. Since a L f there exists t ∈ S 1 with ta = f . Then

xa = f sea = f sa = tasa = ta = f.
Also

xax = f x = f f se = f se = x.
So we have
e = ax a = axa x = xax f = xa.
We have e R a and ea = a therefore ρa : He → Ha is a bijection. From a L f and xa = f
we have λx : Ha → Hf is a bijection. Hence ρa λx : He → Hf is a bijection.
So we have the diagram
ρa

e a

λx

x f

Let h, k ∈ He . Then

h(ρa λx )k(ρa λx ) = (xha)(xka) = xh(ax)ka =

xheka = xhka = hk(ρa λx ).
So, ρa λx is an isomorphism and He ∼
= Hf . 
50 VICTORIA GOULD

It is worth noting that the previous proof also allows us to locate the inverses of a regular
element.
Lemma 8.11. If a ∈ S is regular, and x ∈ V (a), then there exist idempotents e = ax and
f = xa such that
a R e L x, a L f R x.
Conversely, if a ∈ S and e, f are idempotents such that
a R e, a L f,
then there exists x ∈ V (a) such that ax = e and xa = f (and then
e L x, f R x.)

a e = ax

f = xa x
Proof. For the first part, one just has to define e = ax and f = xa. As we have seen, e
and f are idempotents satisfying the required properties.
The converse follows directly from the proof of Corollary 8.10 (Corollary to Green’s Lem-
mas). 
Example 8.12.
0 0 0
 M = M (G; I; Λ; P ) we know that M \ {0} is a D-class. We have Hiλ =
(1) For
(i, g, λ) | g ∈ G . If pλi 6= 0, Hiλ is a group H-class. If pλi , pµj 6= 0 then Hiλ ∼
= Hjµ
(already seen directly). 
The Bicyclic Monoid B is bisimple with E(B) = (a, a) | a ∈ N0 and H(a,a) =
(2) 
(a, a) . Clearly H(a,a) ∼= H(b,b) .
(3) In Tn , then α D β ⇔ ρ(α) = ρ(β) where ρ(α) = | Im(α)|. By Corollary 8.10 , if
ε, µ ∈ E(Tn ) and ρ(ε) = ρ(µ) = m say, then Hε ∼ = Hµ . In fact Hε ∼ = Hµ ∼= Sm .