1.2. Semigroup Ring Theory 15
If S is a semigroup, then we denote by S
1
the smallest monoid containing
S. So
S
1
=
S , if S already has an identity element;
S ∪ {1} , if S does not have an identity element,
with s1 = 1s = s, for all s ∈ S
1
. Similarly, we denote by S
0
the smallest
semigroup with a zero containing S. So
S
0
=
S , if S already has a zero element;
S ∪ {θ} , if S does not have a zero element,
with sθ = θs = θ, for all s ∈ S
0
. In particular, for a group G, we say that
G
0
is a group with zero.
Adjoining a zero normally simplifies arguments, while adjoining an iden-
tity is often useless, as the structure theory of semigroups relies on s ubsemi-
groups and ideals, which do not have an identity in general.
The following example describes a kind of matrix semigroup that is of
utmost importance in the algebraic theory of semigroups.
Example 1.2.2 (Rees Matrix Semigroup). Let G be a group, G
0
=
G ∪ {θ} be the group with zero obtained from G by the adjunction of a zero
element θ (as in Definition 1.2.1), and I and M be arbitrary nonempty sets.
By an I × M matrix over G
0
we mean a mapping A : I × M → G
0
, for which
we use the notation a
i,m
:= A(i, m), for (i, m) ∈ I × M, and (a
i,m
) := A.
By a Rees I × M matrix over G
0
we mean an I × M matrix over G
0
having at most one nonzero element. For g ∈ G, write (g)
i,m
, with i ∈ I
and m ∈ M, for the I × M matrix having g in the (i, m)-entry, its remaining
entries being θ. For any i ∈ I and m ∈ M, the expression (θ)
i,m
denotes the
I × M zero matrix, which will be also be denoted by θ.
Now let P be a fixed arbitrary M × I matrix over G
0
. We define a
multiplication operation in the set of all Rees I × M matrices over G
0
as
AB := A ◦ P ◦ B, where ◦ denotes the usual multiplication of matrices and,
in p erforming this, we agree that, for g ∈ G
0
, θ + g = g = g + θ. We call P
the sandwich matrix with respect to this multiplication. Clearly, the set of
all Rees I × M matrices over G
0
is closed under this operation, which is also
associative. So, we can consider the Rees I × M matrix semigroup over the
group with zero G
0
with sandwich matrix P and denote it by M
0
(G, I, M, P ).
When the sets I and M are finite, say |I| = n and |M| = m, we will write
M
0
(G, I, M, P ) simply as M
0
(G, n, m, P ).
In fact, this type of semigroups is very natural, for instance:
Preliminaries