Chapter 2

Regular Γ-semirings

In the theory of semigroups, the notion of regular element was first introduced by

Thierrin as a generalization of inverse element of group. In this chapter, we introduce

regular Γ-semirings. This is a generalization of the concept of regularity in the theory

of semigroups and semirings. Commuting regular rings and semigroups were studied

by H.Doostie, L.Pourfaraj in [19] and by Amir H.Yamini, Sh.A.Safari Sabet in [1]. In

this chapter we also introduce the notion of commuting regular Γ-semirings. We also

discuss some properties of commuting regular Γ-semirings and obtain a necessary and

sufficient condition for Γ-semirings to possess commuting regularity.

24
2.1 Regular Γ-semirings

Definition 2.1.1. Let S be a Γ-semiring. An element a ∈ S is said to be regular if

a ∈ aΓSΓa, where aΓSΓa = {aαbβa; b ∈ S, α, β ∈ Γ}. A Γ-semiring S is said to be

regular if every element of S is regular.

The following examples serve to bring this idea into focus.

Example 2.1.1. Let S be the set of all 3 × 2 matrices and Γ be the set of all 2 × 3

matrices over the field of rational numbers. Then S is a Γ-semiring. Let A ∈ S, where

 
 a b 
 
 c d . Then we choose B ∈ Γ according to the following cases such that
A= 
 
e f

ABABA = ABA = A.

 
 a b 
Case(i): If the submatrix   is non singular, then ad − bc 6= 0. Then
c d
 
d −b
 ad−bc ad−bc 0 
B=  ∈ Γ and for this B, we find that ABA = A.
−c a
ad−bc ad−bc
0
 
f −b
 af −be 0 af −be 
Case(ii): af − be 6= 0. Then B =   ∈ Γ and ABA = A
−e a
af −be
0 af −be

 
f −d
 0 cf −de cf −de
Case(iii): cf − de 6= 0. Then B =   ∈ Γ and ABA = A

−e c
0 cf −de cf −de

25
    
 ad − bc = 0 
   ad − bc = 0 
 
Case(iv): or
 cf − de = 0 
   af − de = 0 
 

If all the elements of A are 0, then the case is trivial. Next we consider atleast

one of the elements of A is non-zero, say aij 6= 0, i = 1, 2, 3 and j = 1, 2. Then we

take bji th element of B as (aij )−1 and other elements of B are zero and we find that

ABA = A. Thus A is regular. Hence S is a regular Γ-semiring.

Example 2.1.2. Let S be the set of all 3 × 2 matrices over the field of rational num-

bers and Γ be the set of all 2 × 3 matrices over the ring of integers. Let A3,2 , B3,2 ∈ S

and D2,3 ∈ Γ. Then with respect to usual matrix product A  ∈ S, S is a

3,2 D2,3 B3,2

 a b 
 
Γ-semiring. We show that S is a regular Γ-semiring. Let A =  c d  ∈ S.
 
 
e f
Case(i): Let ad − bc 6= 0. Then

       
d −b
 a b   a b   a b 
   
2(ad−bc) 2(ad−bc)
 2 0 0   1 0 0 
 
   
 c d   −c a   c d = c d 
2(ad−bc) 2(ad−bc)
        
  0 2 0   0 1 0    
e f 0 0 e f e f

Case(ii): Let af − be 6= 0. Then

       
f −b
 a b   a b   a b 
   
2(af −be) 2(af −be)
 2 0 0   1 0 0 
 
   
 c d   0 0   c d = c d 
        
  0 0 2   0 0 1    
−e a
e f 2(af −be) 2(af −be)
e f e f

26
Case(iii): Let cf − de 6= 0. Then
       
 a b  0 0  a b   a b 
   
 0 2 0   0 1 0 
 
   
 c d  f −d  c d = c d 
   2(cf −de) 2(cf −de)

     
  0 0 2   0 0 1    
−e c
e f 2(cf −de) 2(cf −de)
e f e f

Case(iv):    
 ad − bc = 0 
   ad − bc = 0 
 
If the sub matrices are singular, then either or
 cf − de = 0 
   af − be = 0 
 

If all the elements of A are 0, then the case is trivial. Next we consider atleast

one of the elements of A is non zero, say a11 6= 0.

       
1
 a b  0   a b   a b 
   
2a
 2 0 0   1 0 0 

   
Then 
 c d 

 0
 0 
  c d
 = c d 
  
  0 0 0   0 0 0    
e f 0 0 e f e f

Thus A is regular. Hence S is a regular Γ-semiring.

If S is a Γ-semiring, then for each α ∈ Γ, Sα is a semigroup. If Sα is a group for

some α ∈ Γ, then Sβ is a group for each β ∈ Γ. Now for the regular Γ-semiring, we

have the following observations.

(i) Let S be a Γ-semiring. If Sα is a regular semigroup for some α ∈ Γ, then there

may exist β ∈ Γ such that Sβ is not a regular semigroup.

Example 2.1.3. Let S = {(a, b); a, b ∈ Q} where Q is the field of rational numbers

and Γ = {(5, 3), (0, 2)}. Now S × Γ × S → S is defined by (a, b)(α, β)(c, d) =

(aαc, bβd) where (a, b), (c, d) ∈ S and (α, β) ∈ Γ. We can show that S is a Γ-semiring.

Here S(0,2) is a semigroup, where (a, b)◦(c, d) ∈ S(0,2) is given by (a, b)(0, 2)(c, d). This

27
S(0,2) can not be a regular semigroup. For example, (3, 5) ∈ S. There does not exist

any (c, d) ∈ S such that (3, 5)◦(c, d)◦(3, 5) = (3, 5) for (3, 5)◦(c, d)◦(3, 5) = (0, 100d).

On the otherhand we can show that S(5,3) is a regular semigroup.

(ii) Let S be a Γ-semiring. Sα may not be a regular semigroup for any α ∈ Γ. But

S may be a regular Γ-semiring.

Example 2.1.4. Let S = {(a, 0); a ∈ Q} ∪ {(0, b); b ∈ Q} where Q denotes the field

of rational numbers. Let Γ = {(0, 5), (0, 1), (3, 0), (1, 0)}. Defining S × Γ × S → S

by (a, b), (α, β)(c, d) = (aαc, bβd) for all (a, b), (c, d) ∈ S and (α, β) ∈ Γ, we can show

that S is a Γ-semiring. Sα is not a regular semigroup for any α ∈ Γ. Let (a, 0) ∈ S.

1
If a = 0, then (a, 0) is regular. If a 6= 0, then (a, 0)(3, 0)( 3a , 0)(1, 0)(a, 0) = (a, 0).

Similarly we can show that (0, b) is also regular. Hence S is a regular Γ-semiring.

From the above two observations, we feel that it will be mathematically important

and also interesting to study regular Γ-semirings.

Let us now characterize regularity condition with the help of ideals of Γ-semiring.

Theorem 2.1.5. Let S be a Γ-semiring. S is regular if and only if for any left ideal

A and for any right ideal B of S, A ∩ B = BΓA.

Proof. BΓA ⊂ SΓA ⊂ A as A is a left ideal of S. BΓA ⊂ BΓS ⊂ B as B is a right

ideal of S. Therefore BΓA ⊂ A ∩ B − − − (i)

Let S be regular. Let a ∈ A ∩ B. Since S is regular, there exist b ∈ S and α, β ∈ Γ

such that a = aαbβa ∈ aαA ⊂ BΓA. Hence A ∩ B ⊂ BΓA − − − (ii).

From (i) and (ii), A ∩ B = BΓA.

To prove the converse, let us assume that the given condition holds in S. Let a ∈ S.

Now A = {a} ∪ SΓa is a left ideal and B = aΓS ∪ {a} is a right ideal in S. Hence

28
A = A∩S = SΓ({a}∪SΓa) = SΓa∪SΓSΓa ⊂ SΓa∪SΓa = SΓa. Similarly B ⊂ aΓS.

Then a ∈ A ∩ B ⊂ (SΓa) ∩ (aΓS) = (aΓS)Γ(SΓa) = aΓ(SΓS)Γa ⊂ aΓSΓa. Hence a

is regular. Thus S is a regular Γ-semiring.

Corollary 2.1.6. Let S be a commutative Γ-semiring. Then S is regular if and only

if A = AΓA for all ideals A of S.

Proof. Let S be regular. Let A be an ideal of S. Then by theorem 2.1.5, AΓA =

A ∩ A = A.

Conversely, let us assume that S is a commutative Γ-semiring satisfying the given

condition. Let A and B be two ideals of S. Then BΓA ⊂ SΓA ⊂ A as A is an ideal of

S and BΓA ⊂ BΓS ⊂ B as B is an ideal of S. Therefore BΓA ⊂ A ∩ B − − − (i). Now

A ∩ B is an ideal of S. Then by the given condition, A ∩ B = (A ∩ B)Γ(A ∩ B) ⊂ BΓA.

Therefore A ∩ B ⊂ BΓA − − − (ii). From (i) and (ii), A ∩ B = BΓA. Hence S is a

regular Γ-semiring.

Theorem 2.1.7. Let I be an ideal of a regular Γ-semiring S. Then I is regular and

any ideal J of I is an ideal of S.

Proof. Since I is an ideal of S, I is a sub Γ-semiring of S. Let a ∈ I ⊂ S. Then

a = aα1 bα2 a for some b ∈ S and α1 , α2 ∈ Γ. Let c = bα2 aα1 b. Then c ∈ I.

aα1 cα2 a = aα1 bα2 aα1 bα2 a = a. Hence I is a regular Γ-semiring. We next show that

if a ∈ J ⊂ I and s ∈ S, then both aαs and sαa lie in J where α ∈ Γ. Now aαs ∈ I.

Hence there exist elements d ∈ I and α3 , α4 ∈ Γ such that (aαs)α3 dα4 (aαs) = aαs.

Since sα3 dα4 aαs ∈ I and J is assumed to be an ideal of I, it follows that the product

(aαs)α3 dα4 (aαs) = aαs ∈ J. Similarly we can show that sαa ∈ J.

29
Definition 2.1.2. The centre B of a Γ-semiring S is a subset of S consisting of all

elements a of S such that aγx = xγa for all x ∈ S and for all γ ∈ Γ.

Lemma 2.1.8. The centre B of a regular Γ-semiring S is regular.

Proof. Let a, b ∈ B. Then aγx = xγa and bγx = xγb for all x ∈ S and for all γ ∈ Γ.

Let γ1 ∈ Γ. Then

(aγ1 b)γx = aγ1 (bγx)

= (bγx)γ1 a

= (xγb)γ1 a

= xγ(bγ1 a)

= xγ(aγ1 b)

So aγ1 b ∈ B. Consequently B is a sub Γ-semiring of S. To show that B is regular, let

a ∈ B. Since S is regular, there exist x ∈ S and γ1 , γ2 ∈ Γ such that aγ1 xγ2 a = a.

As a commutes with all the elements of S, we have

aγ1 (aγ2 a)γ1 (xγ2 xγ1 x)γ2 a = (aγ1 xγ2 a)γ1 (aγ2 xγ1 x)γ2 a

= aγ1 (aγ2 xγ1 x)γ2 a

= (aγ1 xγ2 a)γ1 xγ2 a

= aγ1 xγ2 a

= a

Again aγ2 (aγ1 x) = aγ1 xγ2 a = a and xγ1 (aγ2 a) = aγ1 xγ2 a = a.

So, a = aγ2 (aγ1 x) = xγ1 (aγ2 a) commutes with all the elements of S.

30
Hence for every z ∈ S and for every γ ∈ Γ,

xγ((aγ2 a)γ1 z) = (aγ2 a)γxγ1 z (since aγ2 a ∈ B as B is a sub Γ-semiring of S and a ∈ B)

= xγ1 zγaγ2 a

= xγ1 aγ2 aγz

= zγxγ1 aγ2 a (since xγ1 aγ2 a = a ∈ B)

= zγ((aγ2 a)γ1 x)

= (aγ2 a)γzγ1 x

= zγ1 xγ(aγ2 a)

= zγ1 (aγ2 a)γx

= ((aγ2 a)γ1 z)γx

Therefore, (aγ2 a)γ1 z commutes with x.

We shall now show that (aγ2 a)γ1 z also commutes with xγ2 xγ1 x.

(xγ2 xγ1 x)γ((aγ2 a)γ1 z = xγ2 xγ1 ((aγ2 a)γ1 z)γx

= xγ2 ((aγ2 a)γ1 z)γ1 xγx

= ((aγ2 a)γ1 z)γ2 xγ1 xγx

= aγ2 zγ1 aγ2 xγ1 xγx

= aγ2 zγ1 xγ2 aγ1 (xγx)

= aγ2 zγ1 xγ2 (xγx)γ1 a

= aγ2 zγ1 aγ2 xγxγ1 x (since a ∈ B)

= aγ2 zγ1 (xγx)γ2 aγ1 x

= ((aγ2 a)γ1 z)γ(xγ2 xγ1 x)

31
Now,

(aγ2 a)γ1 (xγ2 xγ1 x)γz = (xγ2 xγ1 x)γ1 (aγ2 a)γz

= (xγ2 xγ1 x)γ1 (aγz)γ2 a

= (xγ2 xγ1 x)γ1 (zγa)γ2 a

= (xγ2 xγ1 a)γ1 xγzγ2 a

= xγ2 xγ1 aγ1 (xγa)γ2 z

= xγ2 xγ1 xγaγ1 aγ2 z

= (xγ2 xγ1 x)γ((aγ2 a)γ1 z)

= ((aγ2 a)γ1 z)γ(xγ2 xγ1 x)

= zγ1 (aγ2 a)γ(xγ2 xγ1 x)

= zγ1 aγxγ2 aγ2 xγ1 x

= zγ1 aγaγ2 xγ2 xγ1 x

= aγaγ1 zγ2 (xγ2 xγ1 x)

= zγ(aγ1 a)γ2 (xγ2 xγ1 x)

= zγaγ2 xγ1 aγ2 xγ1 x

= zγ((aγ2 a)γ1 (xγ2 xγ1 x)).

So, (aγ2 a)γ1 (xγ2 xγ1 x) ∈ B. Now, since aγ1 (aγ2 a)γ1 (xγ2 xγ1 x)γ2 a = a,

where (aγ2 a)γ1 (xγ2 xγ1 x) ∈ B, B is regular.

Definition 2.1.3. Let S be a Γ-semiring and a ∈ S. Let b ∈ S and α, β ∈ Γ. b is

said to be an (α, β) inverse of a if a = aαbβa and b = bβaαb. In this case, we shall

write b ∈ Vαβ (a). We sometimes shall denote it by a−1 −1 β

α,β , that is aα,β ∈ Vα (a).

32
If b ∈ Vαβ (a), then a and b are necessarily regular elements of S. On the otherhand

let a = aαcβa, where c ∈ S and α, β ∈ Γ, be a regular element of S. Let b = cβaαc.

Then

aαbβa = aαcβaαcβa

= aαcβa

= a

and

bβaαb = cβaαcβaαcβaαc

= cβ(aαcβa)αcβaαc

= cβ(aαcβa)αc

= cβaαc

= b

Thus if a = aαcβa is a regular element of S, then b is an (α, β) inverse of a where

b = cβaαc.

Definition 2.1.4. Let S be a Γ-semiring. A pair (a, a0 ) of elements of S is said to be

(α, β) regular for some α, β ∈ Γ if aαa0 βa = a and a0 βaαa0 = a0 .

Lemma 2.1.9. Let S be a Γ-semiring. Let (a, a0 ) be a (α, β) regular pair and (b, b0 )

be a (γ, δ) regular pair. Then a0 βaαbγb0 is a δ-idempotent and bγb0 δa0 βa is an α-

idempotent if and only if (aαb, b0 δa0 ) is a (γ, β) regular pair.

33
Proof. Let a0 βaαbγb0 be a δ-idempotent and bγb0 δa0 βa be an α-idempotent. Then,

(aαb)γ(b0 δa0 )β(aαb) = aαa0 βaαbγb0 δa0 βaαbγb0 δb

= aα(a0 βaαbγb0 )δ(a0 βaαbγb0 )δb

= aαa0 βaαbγb0 δb

= aαb

(b0 δa0 )β(aαb)γ(b0 δa0 ) = b0 δbγb0 δa0 βaαbγb0 δa0 βaαa0

= b0 δ(bγb0 δa0 βa)α(bγb0 δa0 βa)αa0

= b0 δbγb0 δa0 βaαa0

= b0 δa0

Conversely let (aαb, b0 δa0 ) be a (γ, β) regular pair.

(a0 βaαbγb0 )δ(a0 βaαbγb0 ) = a0 β(aαbγb0 δa0 βaαb)γb0

= a0 βaαbγb0

Similarly we can show that bγb0 δa0 βa is an α-idempotent.

Lemma 2.1.10. Let S be a regular Γ-semiring. Let e = eαe and f = f βf be two

idempotents of S. Then there exists an idempotent g such that (eαf, g) is a (β, α)

regular pair.

Proof. Now eαf ∈ S. Since S is regular, there exist γ, δ ∈ Γ and x ∈ S such that

eαf γxδeαf = eαf and xδeαf γx = x. Let g = f γxδe. Then gαg = g. Now,

(eαf )β(f γxδe)α(eαf ) = eαf γxδeαf = eαf

(f γxδe)α(eαf )β(f γxδe) = f γ(xδeαf γx)δe = f γxδe

Hence (eαf, g) is a (β, α) regular pair.

34
Lemma 2.1.11. If S is a regular Γ-semiring, then for each α ∈ Γ, there exists an

idempotent e ∈ S such that e = eαe.

Proof. Let a ∈ S. Now aαa ∈ S. Since S is regular, there exist β, γ ∈ Γ and b ∈ S

such that aαaβbγaαa = aαa and bγaαaβb = b. Let e = aβbγa. Then

eαe = aβbγaαaβbγa

= aβ(bγaαaβb)γa

= aβbγa

= e

Hence e is an α-idempotent.

Theorem 2.1.12. A regular Γ-semiring S is a Γ-ring if and only if eαf = f and

eβf = e for any α-idempotent e and for any β-idempotent f of S.

Proof. Suppose S is a Γ-ring. Let e = eαe be an α-idempotent and f = f βf be

a β-idempotent of S. Then e is the identity element of the group Sα and f is the

identity element of the group Sβ . Now f ∈ Sα . Hence eαf = f . Similarly eβf = e.

Conversely, suppose that given conditions hold in a regular Γ-semiring S. From

lemma 2.1.11, it follows that for each α ∈ Γ, there exists an idempotent e ∈ S

such that e = eαe. Let e be an α-idempotent and f be a β-idempotent of S. We shall

show that Sα is a group. Let c ∈ S. Since S is regular, there exist γ1 , γ2 ∈ Γ and

d ∈ S such that c = cγ1 dγ2 c.

35
Now,

(cγ1 d)γ2 (cγ1 d) = (cγ1 dγ2 c)γ1 d

= cγ1 d

and (dγ2 c)γ1 (dγ2 c) = dγ2 (cγ1 dγ2 c)

= dγ2 c

Thus (cγ1 d) is a γ2 -idempotent and dγ2 c is a γ1 -idempotent. Also e is an α-idempotent.

Then by the given condition, we have (dγ2 c)αe = dγ2 c. Then

cαe = (cγ1 dγ2 c)αe

= cγ1 (dγ2 cαe)

= cγ1 dγ2 c

= c

Hence e is the right identity in Sα . Again cαc ∈ S. Then there exist β1 , β2 ∈ Γ and

s ∈ S such that (cαc)β1 sβ2 (cαc) = cαc. Now, cαcβ1 s is a β2 -idempotent. Hence by

the given condition, (cαcβ1 s)β2 e = e. Then cα(cβ1 sβ2 e) = e. Hence for any c in S,

(cβ1 sβ2 e) is the right inverse of c in Sα . Thus Sα is a group. Hence S is a Γ-ring.

2.2 Green’s Relations in Regular Γ-semirings

In this section, we discuss the concepts of Green’s relations in Γ-semirings.

For each element a of a Γ-semiring S, the smallest left ideal containing a is called

the principal left ideal generated by a and shall be denoted by (a)l . We similarly define

(a)r and (a), the principal right ideal and the principal ideal generated by a. We can

show that (a)l = SΓa ∪ {a}, (a)r = aΓS ∪ {a} and (a) = aΓS ∪ SΓa ∪ SΓaΓS ∪ {a}.

36
 Let S be a Γ-semiring and a, b ∈ S. Define binary relations L, R, J, H and D

on S as follows:

(i) aLb if and only if (a)l = (b)l .

(ii) aRb if and only if (a)r = (b)r .

(iii) aJb if and only if (a) = (b).

(iv) aHb if and only if aLb and aRb.

(v) aDb if and only if there exists c ∈ S such that aLc and cRb.

We note the following facts about these relations in a Γ-semiring S.

(a)L, R, J and H are equivalence relations on S.

(b)aLb implies aαcLbαc for all c ∈ S and α ∈ Γ.

(c)aRb implies cαaRcαb for all c ∈ S and α ∈ Γ.

(d)aLb if and only if either a = b or there exist α, β ∈ Γ and c, d ∈ S such that

a = cαb and b = dβa.

(e)aRb if and only if either a = b or there exist α, β ∈ Γ and c, d ∈ S such that

a = bαc and b = aβd.

(f)aHb if and only if L ∩ R

(g)D = L ◦ R = R ◦ L

(h)D ⊂ J

Lemma 2.2.1. L ◦ R = R ◦ L

Proof. Let (a, b) ∈ L ◦ R. Then there exists c ∈ S such that aLc and cRb.

Case (i): a = c. Then aRb. Since aLa and aRb, (a, b) ∈ L ◦ R.

Case (ii): b = c. Then aLb. Since aLb and bRb, (a, b) ∈ L ◦ R.

37
Case (iii): a 6= c and b 6= c. Since aLc and cRb, there exist x, y, u, v ∈ S and

γ, µ, η, θ ∈ Γ such that xγa = c, yµc = a, cηu = b, bθv = c. Let d = yµcηu. Then

aηu = yµcηu = d and dθv = yµcηuθv = yµbθv = yµc = a from which it follows

aRd. Also yµb = yµcηu = d and xγd = xγyµcηu = xγaηu = cηu = b, so dLb. We

deduce that (a, b) ∈ R ◦ L. Therefore L ◦ R ⊆ R ◦ L. Similarly, we can prove that

R ◦ L ⊆ L ◦ R.

The L-class(respectively R-class, H-class, D-class) containing the element a will

be written as La (respectively Ra , Ha , Da ).

Theorem 2.2.2. Let S be a Γ-semiring, α ∈ Γ and e be an α-idempotent. Then

(i) aαe = a for all a ∈ Le .

(ii) eαa = a for all a ∈ Re .

(iii) aαe = a = eαa for all a ∈ He .

(iv) For all a ∈ S, |Ha ∩ Eα | ≤ 1.

Proof.

(i) Let a ∈ Le . Then aLe. It follows that SΓa ∪ {a} = SΓe ∪ {e}. Then a = e or

there exist x ∈ S and γ ∈ Γ such that a = xγe. If a = e, then aαe = eαe = e = a. If

a = xγe, then aαe = (xγe)αe = xγ(eαe) = xγe = a.

(ii) It is similar to (i).

(iii) It follows from (i) and (ii).

(iv) Let e, f ∈ Ha ∩ Eα . Then eHf . So eLf and eRf . Then f ∈ Le and e ∈ Rf . By

(i) and (ii), respectively, we have f αe = f and f αe = e. Therefore e = f . It follows

that |Ha ∩ Eα | ≤ 1.

38
 Here we want to discuss Green’s relations in the context of regular Γ-semirings.

Lemma 2.2.3. Let S be a Γ-semiring. If a = aαa0 βa, a0 ∈ S and α, β ∈ Γ, be

a regular element in S, then (i) (a)r = aΓS = aαS (ii) (a)l = SΓa = Sβa (iii)

(a) = aαS ∪ SΓa ∪ SΓaΓS = aαS ∪ Sβa ∪ SβaαS.

Proof.

(i) We have (a)r = {a} ∪ aΓS. Since a = aαa0 βa, it follows that a ∈ aΓS. Hence

(a)r = aΓS. Let aα1 c ∈ aΓS. Then aα1 c = aαa0 βaα1 c = aα(a0 βaα1 c) ∈ aαS.

Hence aΓS ⊂ aαS. On the otherhand aαS ⊂ aΓS. Thus it follows that (a)r = aαS.

Similarly we can prove (ii) and (iii).

We note that in a regular Γ-semiring Green’s relations will take the following form:

(i) aRb if and only if aΓS = bΓS

(ii) aLb if and only if SΓa = SΓb.

(iii) aRb if and only if there exist α1 , β1 ∈ Γ and c, d ∈ S such that a = bα1 c and

b = aβ1 d.

(iv) aLb if and only if there exist α2 , β2 ∈ Γ and c1 , d1 ∈ S such that a = c1 α2 b and

b = d1 β2 a.

Similar expression can be written for aJb.

Lemma 2.2.4. Let S be a regular Γ-semiring. Let a = aαbβa, b ∈ S and α, β ∈ Γ,

be an element in S such that b ∈ Vαβ (a). Then

(i) aαb is a β-idempotent and bβa is an α-idempotent.

(ii) aRaαbLb, aLbβaRb

39
Proof.

(i) Since b ∈ Vαβ (a), aαbβa = a and bβaαb = b. Now

(aαb)β(aαb) = aα(bβaαb)

= aαb

and (bβa)α(bβa) = bβ(aαbβa)

= bβa

Hence aαb is a β-idempotent and bβa is an α-idempotent.

(ii) From lemma 2.2.3, we have (a)r = aΓS and (aαb)r = (aαb)ΓS. Now (aαb)r ⊂ (a)r

and (a)r = aΓS = (aαb)βaΓS ⊂ (aαb)ΓS. Hence (a)r = (aαb)r . This implies that

aRaαb. Again,

(aαb)l = SΓ(aαb)

= (SΓa)αb ⊂ Sαb ⊂ SΓb

= (b)l

Also,

(b)l = SΓb

= SΓbβ(aαb)

= (SΓb)β(aαb) ⊂ Sβ(aαb) ⊂ SΓ(aαb)

= (aαb)l

Hence, (aαb)l = (b)l . This shows that (aαb)Lb. Similarly we can show that aLbβaRb.

Lemma 2.2.5. Let S be a Γ-semiring. An element a ∈ S is regular if and only if

(a)r = eβS for some β-idempotent e ∈ S, where β ∈ Γ.

40
Proof. Suppose a is regular. Then there exist α, β ∈ Γ and b ∈ S such that a = aαbβa.

Let e = aαb. Then eβe = e. Also, eβa = aαbβa = a. Then by lemma 2.2.3,

(a)r = aαS = (eβa)αS = eβ(aαS) ⊂ eβS. On the otherhand, let eβd ∈ eβS for

some d ∈ S. Then eβd = aαbβd = aα(bβd) ∈ aαS = (a)r . Thus eβS ⊂ (a)r . Hence

(a)r = eβS.

To prove the converse, assume that (a)r = eβS, where e is a β-idempotent in S.

Since (a)r = {a} ∪ aΓS, it follows that e = a or a = eβt for some t ∈ S and

e = aα1 b1 for some α1 ∈ Γ and b1 ∈ S. Now eβa = eβeβt = eβt = a. Hence

a = eβa = aα1 b1 βa ∈ aΓSΓa. Thus it follows that a is regular element of S.

Lemma 2.2.6. Let S be a regular Γ-semiring and let a ∈ S. Suppose e = eαe and

f = f βf, α, β ∈ Γ, be two idempotents of S with eRaLf . Then there exists a unique

b ∈ Vβα (a) such that aβb = e and bαa = f .

Proof. Since eRa, we have eΓS = aΓS. Then there exist α1 , β1 ∈ Γ and c, d ∈ S such

that e = aα1 c and a = eβ1 d. Now, eαa = eαeβ1 d = eβ1 d = a. Again since aLf , we

have (a)l = (f )l . By lemma 2.2.3, (f )l = Sβf . Hence we have SΓa = Sβf . From this

we can show that a = aβf and f = c1 β2 a for some c1 ∈ S and β2 ∈ Γ. Let b = f α1 c.

Then aβb = aβf α1 c = (aβf )α1 c = aα1 c = e. Hence (aβb)αa = eαa = a. Now,

bαaβb = (f α1 c)αaβb

= (c1 β2 aα1 c)αaβb

= c1 β2 (aα1 c)αaβb

= c1 β2 (eαa)βb

= (c1 β2 a)βb = f βb = f βf α1 c = f α1 c = b

Hence b ∈ Vβα (a).

41
Moreover,

bαa = f α1 cαa

= c1 β2 aα1 cαa

= c1 β2 (aα1 c)αa

= c1 β2 (eαa)

= c1 β 2 a

= f

Hence there is atleast one b satisfying aβb = e and bαa = f . We next show the

uniqueness. Suppose now there exist b and b0 in Vβα (a) such that aβb = aβb0 = e and

bαa = b0 αa = f . Then

b0 = b0 αaβb0

= b0 α(aβb0 )

= b0 α(aβb)

= (b0 αa)βb

= bαaβb

= b

Theorem 2.2.7. Let S be a regular Γ-semiring. If e = eαe, α ∈ Γ be an α-idempotent

and a, b be any two elements in an H-class H of S such that e ∈ H, then eαa = aαe = a

and aαb ∈ H.

Proof. Let e = eαe be an α-idempotent in an H-class H of S and let a, b be any two

elements of H. Then eLa and aRa. Hence SΓe = SΓa and eΓS = aΓS. Then there

42
exist α1 , α2 ∈ Γ and m, n ∈ S such that a = mα1 e and a = eα2 n. Hence

eαa = eαeα2 n

= eα2 n

= a

and

aαe = mα1 eαe

= mα1 e

= a

Therefore eαa = aαe = a. Since a, b ∈ H, by lemma 2.2.6, there exist a unique

c ∈ Vαα (a) and a unique d ∈ Vαα (b) such that aαc = cαa = e = bαd = dαb. Now,

(aαb)ΓS = aα(bΓS)

= aα((bαd)ΓS) since bαdαb = b

= aα(eΓS)

= (aαe)ΓS

= aΓS

= (aαc)ΓS by lemma 2.2.5

= eΓS

Hence (aαb)Re. Similarly we can show that aαbLe. Hence aαb ∈ H for any a, b ∈

H.

In semigroup, D-class of the Green’s relation has some interesting property. If

one element of a D-class is regular, then every element of that D-class is regular. Let

us now extend this result to Γ-semiring.

43
Theorem 2.2.8. Let S be a Γ-semiring and a ∈ S. Let Da denote the D-class of S

containing a. If a is regular, then every element of Da is regular.

Proof. Let b ∈ Da . Since D = L ◦ R, we have aLc and cRb for some c ∈ S. Hence

SΓa ∪ {a} = SΓc ∪ {c} and cΓS ∪ {c} = bΓS ∪ {b}. Let a = aαdβa. Then by

lemma 2.2.3, we have Sβa = SΓc ∪ {c}. Therefore there exist α1 , α2 ∈ S and α1 ∈ Γ

such that a = c or a = a1 α1 c and c = a2 βa. Again from cΓS ∪ {c} = bΓS ∪ {b} we

have c = b or c = bα2 a3 and b = cα3 a4 for some α2 , α3 ∈ Γ and a3 , a4 ∈ S. We have

to consider the following cases:

(i) a = c and c = b

(ii) a = c and c = bα2 a3 , b = cα3 a4

(iii) a = a1 α1 c, c = a2 βa and c = b

(iv) a = a1 α1 c, c = a2 βa and c = bα2 a3 and b = cα3 a4 .

Case(i): In this case b is regular since b = c = a.

Case(ii): From b = cα3 a4 = aα3 a4 , we have

aαdβb = (aαdβa)α3 a4

= aα3 a4

= b

Hence,

b = aαdβb

= cαdβb

= bα2 a3 αdβb

= bα2 (a3 αd)βb

Hence b is regular element.

44
Case(iii): Since c = b, we have b = a2 βa. Hence bαdβa = a2 βaαdβa = a2 βa = b.

Then

b = bαdβa

= bαdβa1 α1 b

= bα(dβa1 )α1 b

which implies that b is regular.

Case(iv):

(bα2 a3 )αdβa1 α1 b = (cαd)βa1 α1 (cα3 a4 )

= (cαd)β(a1 α1 c)α3 a4

= cαdβ(aα3 a4 )

= a2 βaαdβaα3 a4

= a2 β(aαdβa)α3 a4

= (a2 βa)α3 a4

= cα3 a4

= b

Hence b is regular.

Theorem 2.2.9. In a regular D-class, each L-class and each R-class contain at least

one idempotent.

Proof. Let a be an element of a regular D-class D in a Γ-semirings S. Then there exist

x ∈ S and α, β ∈ Γ such that a = aαxβa. Then xβa = xβ(aαxβa) = (xβa)α(xβa).

Thus xβa is an α-idempotent. Since a = aα(xβa), aLxβa. Similarly, aαx is a

β-idempotent and aRaαx.

45
Theorem 2.2.10. Let a be an element of a regular D-class D in a Γ-semiring S.

Then

(i) If a0 is an (α, β)-inverse of a, then a0 ∈ D and the two H-classes Ra ∩ La0 and

La ∩ Ra0 contain a β-idempotent aαa0 and an α-idempotent a0 βa, respectively.

(ii)If b ∈ D is such that Ra ∩ Lb and La ∩ Rb contain a β-idempotent e and an

α-idempotent f, respectively, then Hb contains an (α, β)-inverse a∗ of a such that

aαa∗ = e and a∗ βa = f .

(iii) No H-class contains more than one (α, β)-inverse of a for all ordered pairs

(α, β) ∈ Γ × Γ.

Proof.

(i) Let a0 be an (α, β)-inverse of a. Then a = aαa0 βa and a0 = a0 βaαa0 . Thus

aLa0 βa, aαa0 Ra, a0 Laαa0 , a0 βaRa0 . Thus a0 Da, aαa0 ∈ Ra ∩ La0 and a0 βa ∈ La ∩

Ra0 . Therefore a0 ∈ D. Since a = aαa0 βa, a0 βa = a0 βaαa0 βa and aαa0 = aαa0 βaαa0 .

Therefore a0 βa is an α-idempotent and aαa0 is a β-idempotent.

(ii) Since aRe, by theorem 2.2.2(ii), eβa = a. Similarly, from aLf , we deduce that

aαf = a by theorem 2.2.2(i). Again from aRe, it follows that a = e or there exist

x ∈ S and γ ∈ Γ such that aγx = e.

Case (i): a = e. Let e∗ = f βe. Then aαa∗ βa = aα(f βe)βa = (aαf )β(eβa) = aβa =

eβa = a and a∗ βaαa∗ = (f βe)βaα(f βe) = f β(eβa)αf βe = f β(aαf )βe = f β(aβe) =

f βe = a∗ . Then a∗ is an (α, β)-inverse of a. Moreover aαa∗ = aαf βe = aβe = eβe =

e. Further, since aLf, a = f or f = yθa for some y ∈ S and θ ∈ Γ. If a = f , then

a∗ βa = f βeβa = eβeβe = e = f . If f = yθa, then a∗ βa = f βeβa = yθaβeβe =

yθa = f . It now follows easily that a∗ ∈ Le ∩ Rf = Lb ∩ Rb = Hb .

46
Case(ii): aγx = e. Let a∗ = f γxβe. Then aαa∗ βa = aα(f γxβe)βa = (aαf )γxβ(eβa)

= aγxβa = eβa = a and a∗ βaαa∗ = (f γxβe)βaα(f γxβe) = f γxβ(eβa)αf γxβe =

f γxβ(aαf )γxβe = f γxβ(aγx)βe = f γxβeβe = f γxβe = a∗ . Then a∗ is an (α, β)-

inverse of a. Moreover aαa∗ = aαf γxβe = aγxβe = eβe = e. Since aLf , a = f

or there exist y ∈ S and θ ∈ Γ such that f = yθa. If a = f , then a∗ βa =

f γxβeβa = aγxβeβa = eβeβa = eβa = a = f . If f = yθa, then a∗ βa =

f γxβeβa = yθ(aγx)βeβa = yθ(eβe)βa = yθ(eβa) = yθa = f . It now follows easily

that a∗ ∈ Le ∩ Rf = Lb ∩ Rb = Hb .

(iii) Suppose that a0 and a∗ are both (α, β)-inverses of a inside the single H-class

Hb . Since aαa0 and aαa∗ are β-idempotents in the H-class Ra ∩ Lb , aαa0 = aαa∗

by theorem 2.2.2(iv). Similarly, a0 βa = a∗ βa because both are α-idempotents in the

H-class La ∩ Rb . Then a0 = a0 βaαa0 = a∗ βaαa∗ = a∗ .

We have defined L, R, H in terms of principal ideals. On regular Γ-semiring, these

relations can be characterized alternatively in terms of inverses as can be seen in the

following lemma.

Lemma 2.2.11. Let S be a regular Γ-semiring and a, b ∈ S. Then

(i) aLb if and only if there exist α, β, δ ∈ Γ and a0 ∈ Vαβ (a), b0 ∈ Vαδ (b) such that

a0 βa = b0 δb.

(ii) aRb if and only if there exist α, β, γ ∈ Γ and a0 ∈ Vαβ (a), b0 ∈ Vγβ (b) such that

aαa0 = bγb0 .

(iii) aHb if and only if there exist γ, δ ∈ Γ and a0 ∈ Vγδ (a), b0 ∈ Vγδ (b) such that

aγa0 = bγb0 , a0 δa = b0 δb.

47
Proof.

(i) Let aLb and let a0 ∈ Vαβ (a), b00 ∈ Vγδ (b). Then by lemma 2.2.4, we get a0 βa is an

α-idempotent, bγb00 is a δ-idempotent and aLa0 βa, bRbγb00 . Since aLb, a0 βaLbRbγb00 .

Hence by lemma 2.2.6, there exists a unique b0 ∈ Vαδ (b) such that bαb0 = bγb00 and

b0 δb = a0 βa.

Conversely, suppose there exist α, β, δ ∈ Γ and a0 ∈ Vαβ (a), b0 ∈ Vαδ (b) such that

a0 βa = b0 δb. Then aLa0 βa and bLb0 δb. Thus aLa0 βa = b0 δbLb. Hence aLb.

(ii) Proceeding as above we can prove (ii).

(iii) Let aHb. Then aLb and aRb. Let a0 ∈ Vγδ (a). Then aLa0 δa and aRaγa0 . Then

aγa0 RbLa0 δa. Hence it follows from lemma 2.2.6 that there exists a unique b0 ∈ Vγδ (b)

such that bγb0 = aγa0 and b0 δb = a0 δa. Conversely, let there exist γ, δ ∈ Γ and

a0 ∈ Vγδ (a), b0 ∈ Vγδ (b) such that aγa0 = bγb0 , a0 δa = b0 δb. Then aRaγa0 = bγb0 Rb.

Hence aRb. Again aLa0 δa = b0 δbLb. Then aLb. Hence aHb.

2.3 Locating Inverses in Regular Γ-semirings

Suppose that e = eαe and f = f βf are idempotents in a regular Γ-semiring S.

Then there is no guarantee that eαf or f βe is idempotent. However, suppose that

a ∈ Vβα (eαf ). Then eαf βaαeαf = eαf and aαeαf βa = a. Let g = f βaαe.

Then

gαg = f β(aαeαf βa)αe

= f βaαe

= g

48
 gαe = f βaαeαe

= f βaαe

= g

f βg = f βf βaαe

= f βaαe

= g

eαgαf = eαf βaαeαf

= eαf

Thus

(eαf )βgα(eαf ) = eαf βf βaαeαeαf

= eαf βaαeαf

= eαf

gα(eαf )βg = (gαe)α(f βg)

= gαg

= g

Hence g ∈ Vβα (eαf ). Thus eαf has an idempotent (β, α) inverse. We shall denote by

Sαβ (e, f ), the set Sαβ (e, f ) = {gαg = g ∈ S; gαe = g = f βg, eαgαf = eαf }. It is

called the (α, β) sandwich set of e and f .

Lemma 2.3.1. Let S be a regular Γ-semiring and let a, b ∈ S, a0 ∈ Vαβ (a), b0 ∈ Vγδ (b).

For each g ∈ Sαδ (a0 βa, bγb0 ), b0 δgαa0 ∈ Vγβ (aαb).

49
Proof.

(aαb)γ(b0 δgαa0 )β(aαb) = aα((bγb0 )δg)α(a0 βa)αb

= aαgαb since (bγb0 )δg = g and gα(a0 βa) = g

= aα(a0 βaαgαbγb0 )δb

= aαa0 βaαbγb0 δb since g ∈ Sαδ (a0 βa, bγb0 )

= aαb

(b0 δgαa0 )β(aαb)γ(b0 δgαa0 ) = b0 δ(gαa0 βa)α(bγb0 δg)αa0

= b0 δgαgαa0

= b0 δgαa0

Hence b0 δgαa0 ∈ Vγβ (aαb).

Corollary 2.3.2. Let S be a regular Γ-semiring and let a, b ∈ S. Suppose that e = eαe

and f = f βf are idempotents and that eLa, f Rb. Let g ∈ Sαβ (e, f ). Then

(i) aαb = aαgαb = (aαg)α(gαb)

(ii)eαg is α-idempotent and gαf is β-idempotent and aαgLeαgLg, gαbRgαf Rg.

(iii) gαbLaαbRaαg

(iv) aαbDg.

50
Proof. (i) Since eLa, we have aαe = a and since f Rb, we get f βb = b. Then,

aαb = aαeαf βb

= aα(eαf )βb

= aα(eαgαf )βb

= (aαe)αgα(f βb)

= aαgαb

= aαgαgαb

= (aαg)α(gαb)

(ii) Since g = gαe,

(eαg)α(eαg) = eα(gαe)αg

= eαgαg

= eαg

Hence eαg is α-idempotent. Similarly we can show that f αg is β-idempotent. Since

aLe, SΓa = SΓe. Now

SΓaαg = SΓeαg ⊂ SΓg

= SΓgαg

= SΓ(gαe)αg ⊂ Sαeαg ⊂ SΓ(eαg)

Hence SΓ(eαg) = SΓg. Therefore aαgLeαgLg. Similarly we can show that gαbRgαf Rg.

(iii) Let aαg = a1 and gαb = b1 . Then from (ii), we have a1 LgRb1 . We show that

this implies a1 Ra1 αb1 Lb1 .

51
We have,

a1 αb1 ∈ a1 ΓS = a1 αgΓS

= a1 αb1 ΓS since gRb1 implies gΓS = b1 ΓS

Hence a1 Ra1 αb1 .

Similarly a1 αb1 Lb1 . Finally,

a1 αb1 = aαgαgαb

= aαgαb

= aαb by (i)

Thus b1 La1 αb1 Ra1 . Hence gαbLaαbRaαg.

(iv) From (iii) and (ii), we have aαbLgαbRg. Thus aαbDg.

Lemma 2.3.3. Let S be a regular Γ-semiring. Then the following are equivalent:

(i) If e = eαe and f = f βf are any two idempotents of S, then eαf is a β-idempotent

and f βe is an α-idempotent.

(ii) If e = eαe and f = f βf are idempotents of S, then f βe ∈ Sαβ (e, f )

β2 β1
(iii) If a0 ∈ Vαβ11 (a), b0 ∈ Vα2 (b), then b0 β2 a0 ∈ Vα2 (aα1 b).

Proof.

(i) ⇒ (ii): Let g = f βe. Then by (i), f βe is an α-idempotent.

eα(gαf ) = eαf βeαf

= (eαf )β(eαf )

= eαf by (i)

52
 gαe = f βeαe

= f βe

= g

f βg = f βf βe

= f βe

= g

Hence g ∈ Sαβ (e, f ). Thus f βe ∈ Sαβ (e, f ).

β2 0
(ii) ⇒ (iii): By lemma 2.3.1, b0 β2 gα1 a0 ∈ Vαβ1
2
(aα1 b) for g ∈ Sα1 (a β1 a, bα2 b0 ). Hence

by (ii), (bα2 b0 β2 (a0 β1 a) ∈ Sαβ21 (a0 β1 a, bα2 b0 ). Again by lemma 2.3.1, b0 β2 (bα2 b0 β2 a0 β1 a)

α1 a0 ∈ Vαβ1
2
(aα1 b). Hence b0 β2 a0 ∈ Vαβ1
2
(aα1 b).

(iii) ⇒ (i): Let e = eαe and f = f βf be two idempotents in S. Then e ∈ Vαα (e)

and f ∈ Vββ (f ). Hence by (iii), f βe ∈ Vβα (eαf ). Hence eαf βf βeαeαf = eαf and

(eαf )β(eαf ) = eαf . Thus eαf is β-idempotent. Again f βeαeαf βf βe = f βe. Hence

(f βe)α(f βe) = f βe. Thus f βe is α-idempotent.

2.4 Partial Order on Regular Elements of Γ-semirings

In this section we want to introduce a relation on a Γ-semiring which will be a

partial order on regular elements of a Γ-semiring.

Definition 2.4.1. Let S be a Γ-semiring. Let a, b ∈ S. We define a ≤1 b if a is

regular and there exist α, β ∈ Γ and a0 ∈ Vαβ (a) such that a0 βa = a0 βb, aαa0 = bαa0 .

Let S denote the set of all regular elements in S and let S 6= φ.

53
Lemma 2.4.1. The relation ≤1 defined above is a partial order on S.

Proof. Let a ∈ S. Then

(i) It is immediate that ≤1 is reflexive.

(ii) Let a, b ∈ S and let a ≤1 b.

Then by definition there exist α, β ∈ Γ and a0 ∈ Vαβ (a) such that a0 βa = a0 βb, aαa0 =

bαa0 . Now,

aαa0 βb = aα(a0 βb)

= aαa0 βa

= a

and

bαa0 βa = (bαa0 )βa

= aαa0 βa

= a

Hence bαa0 βb = bα(a0 βb) = bαa0 βa = a − − − (1).

Next, let b ≤1 a. Then there exist γ, δ ∈ Γ and b0 ∈ Vγδ (b) such that b0 δb = b0 δa, bγb0 =

aγb0 . Now

bγb0 δa = bγ(b0 δa)

= bγ(b0 δb)

= b − − − (2)

54
Further

bγb0 δa = bγb0 δbαa0 βb since by (1), a = bαa0 βb

= bαa0 βb

= a − − −(3)

From (2) and (3), a = b. Hence ≤1 is antisymmetric.

(iii) For transitivity, let us suppose a ≤1 b and b ≤1 c. Then by definition, there exist

α, β ∈ Γ and a0 ∈ Vαβ (a) such that aαa0 = bαa0 , a0 βa = a0 βb. From b ≤1 c, there exist

γ, δ ∈ Γ and b0 ∈ Vγδ (b) such that bγb0 = cγb0 , b0 δb = b0 δc. Now, bαa0 βb = a = aαa0 βb.

Also, it can be shown that bγb0 δa = a.

aγb0 δb = bαa0 βbγb0 δb since a = bαa0 βb

= bαa0 βb since b0 ∈ Vγδ (b)

= a − − − (4)

Let a∗ = b0 δbαa0 βbγb0 . Then

aγa∗ = aγb0 δbαa0 βbγb0

= (aγb0 δb)α(a0 βb)γb0

= aα(a0 βb)γb0 since by (4), aγb0 δb = a

= aα(a0 βa)γb0 since a0 βb = a0 βa

= aγb0 − − −(5)

55
 bγa∗ = bγb0 δbαa0 βbγb0

= (bγb0 δb)αa0 βbγb0

= (bαa0 )βbγb0

= (aαa0 )βbγb0 since aαa0 = bαa0

= aα(a0 βb)γb0

= aα(a0 βa)γb0 since a0 βb = a0 βa

= aγb0 − − − (6)

From (5) and (6), aγa∗ = bγa∗ − − − (7)

a∗ δa = b0 δbαa0 βbγb0 δa

= b0 δbαa0 β(bγb0 δa)

= b0 δbαa0 βa since bγb0 δa = a

= b0 δ(bαa0 )βa

= b0 δ(aαa0 )βa

= b0 δa − − − (8)

Similarly a∗ δb = b0 δa − − − (9). From (8) and (9), a∗ δa = a∗ δb − − − (10).

Further

aγa∗ δa = (aγa∗ )δa

= (bγa∗ )δb

= aγb0 δb by (6)

= a since by (4) aγb0 δb = a

56
 a∗ δaγa∗ = a∗ δ(bγa∗ )

= a∗ δaγb0 since by (6) bγa∗ = aγb0

= (a∗ δa)γb0

= b0 δaγb0 since by (8) a∗ δa = b0 δa

= b0 δaαa0 βaγb0 since a0 ∈ Vαβ (a)

= b0 δbαa0 βaγb0 since aαa0 = bαa0

= b0 δbαa0 βbγb0 since a0 βa = a0 βb

= a∗

Hence a∗ ∈ Vγδ (a). Thus when a ≤1 b and a0 ∈ Vαβ (a), b0 ∈ Vγδ (b), we can always find

a∗ ∈ Vγδ (a) such that aγa∗ = bγa∗ = aγb0 , a∗ δa = a∗ δb = b0 δa − − − (11).

Now,

aγa∗ = bγa∗

= bγb0 δbαa0 βbγb0

= cγb0 δbαa0 βbγb0 since bγb0 = cγb0

= cγa∗

a∗ δa = a∗ δb

= b0 δbαa0 βbγb0 δb

= b0 δbαa0 βbγb0 δc since b0 δb = b0 δc

= a∗ δc

Hence a ≤1 c. Thus ≤1 is a partial order on the elements of S.

57
Theorem 2.4.2. Let S be a Γ-semiring and let a, b, c, ... be elements of S such that

a ∈ S and α, β, γ, ... be elements of Γ. Then the following hold:

(i) If b is regular, then a ≤1 b. aΓS = bΓS implies that a = b.

(ii) a ≤1 aγd implies that a = aγd where γ ∈ Γ.

(iii) a ≤1 cγa implies that a = cγa where γ ∈ Γ.

(iv) a = (aαxβa) if and only if a ≤1 (aαxβa) where x ∈ S, α, β ∈ Γ.

(v) a = aαa if and only if there is a0 ∈ Vαα (a) such that aαa0 ≤1 a.

(vi) If b is regular, then a ≤1 b if and only if a∗ ≤1 b∗ for some γ, δ ∈ Γ and

a∗ ∈ Vγδ (a), b∗ ∈ Vγδ (b).

(vii) a ≤1 e where e is a γ-idempotent, implies that aγa = a.

(viii) If b is regular, then a ≤1 b if and only if there exist γ, δ ∈ Γ, a∗ ∈ Vγδ (a), b0 ∈

Vγδ (b) such that aγa∗ ≤1 bγb0 , a∗ δa ≤1 b0 δb and bγa∗ δb = a = aγb0 δa.

Proof.

(i) Let a ≤1 b. Then by definition there exist α, β ∈ Γ and a0 ∈ Vαβ (a) such that

a0 βa = a0 βb, aαa0 = bαa0 . Since b is regular, we can write b = bγcδb for some γ, δ ∈ Γ

and c ∈ S. Then by lemma 2.2.3, we get bγS = bΓS and aΓS = aαS. Hence by

hypothesis aαS = bγS. Therefore b = aαd for some d ∈ S. Hence

a = aαa0 βa

= aαa0 βb since a0 βa = a0 βb

= aαa0 βaαd since b = aαd

= aαd

= b

(ii) Let a ≤1 aγd. By definition there exist α, β ∈ Γ and a0 ∈ Vαβ (a) such that

58
a0 βa = a0 βaγd, aαa0 = aγdαa0 . Then aα(a0 βaγd) = aα(a0 βa) = a. Hence aγd =

(aαa0 βa)γd = aα(a0 βaγd) = a.

(iii) Similar as (ii).

(iv) Let a ≤1 (aαxβa). Then by (ii), a = aαxβa.

Conversely let a = aαxβa. Then

aα(xβaαx)βa = (aαxβa)αxβa

= aαxβa

= a

and

(xβaαx)βaα(xβaαx) = xβ(aαxβa)αxβaαx

= xβ(aαxβa)αx

= xβaαx

Hence (xβaαx) ∈ Vαβ (a). Now

xβaαxβa = xβ(aαxβa)

= xβa

aαxβaαx = (aαxβa)αx

= aαx

xβaαxβaαxβa = xβ(aαxβaαxβa)

= xβa

aαxβaαxβaαx = (aαxβaαxβa)αx

= aαx

59
Hence (xβaαx)βa = (xβaαx)β(aαxβa) and aα(xβaαx) = (aαxβa)α(xβaαx). Hence

a ≤1 (aαxβa).

(v) Let aαa0 ≤1 a, where a0 ∈ Vαα (a). Then aαa0 ≤1 aαa0 αa. Now from (iv),

aαa0 = aαa0 αa. Then aαa0 = a. Hence a = aαa0 αa = aαa. Conversely let a = aαa.

Then a = aαaαa. This shows that a ∈ Vαα (a). Hence there exists a0 = a ∈ Vαα (a)

such that aαa0 ≤1 a.

(vi) Let a ≤1 b and b is also regular. Then there exist α, β ∈ Γ and a0 ∈ Vαβ (a) such

that a0 βa = a0 βb, aαa0 = bαa0 . Since b is regular, there exist γ, δ ∈ Γ and b∗ ∈ Vγδ (b).

Then following the proof of (11) of lemma 2.4.1, there exists a∗ ∈ Vγδ (b) such that

aγa∗ = aγb∗ , a∗ δa = b∗ δa. Now a∗ ∈ Vγδ (a) implies that a ∈ Vδγ (a∗ ). Hence a∗ ≤1 b∗ .

The converse follows by symmetry.

(vii) Let a ≤1 e where eγe = e. Then there exist α, β ∈ Γ and a0 ∈ Vαβ (a) such that

a0 βa = a0 βe, aαa0 = eαa0 . Hence aΓS ⊂ eΓS and SΓa ⊂ SΓe. Since a is regular

and a = aαa0 βa, it follows from lemma 2.2.3 that aΓS = aαS and SΓa = Sβa. Then

aαS ⊂ eγS and Sβa ⊂ Sγe. Then a = eγt for some t ∈ S and a = t1 γe for some

t1 ∈ S. Further, aγe = t1 γeγe = t1 γe = a and eγa = eγeγt = eγt = a. Hence

eγa = aγe = a. Then

a = eγa

= eγaαa0 βa

= (eγa)α(a0 βa)

= aα(a0 βe) since a0 βa = a0 βa − − − (12)

60
Hence

aγa = aα(a0 βa)γaα(a0 βa)

= aα(a0 βe)γaα(a0 βe)

= aαa0 β(eγa)α(a0 βe)

= aαa0 βaαa0 βe since eγa = a

= aα(a0 βaγa0 )βe

= aαa0 βe

= a by (12)

(viii) Let b be regular and let a ≤1 b. Since b is regular, there exist γ, δ ∈ Γ

and b0 ∈ Vγδ (b). From a ≤1 b, there exist α, β ∈ Γ and a0 ∈ Vαβ (a) such that

a0 βa = a0 βb, aαa0 = bαa0 . Following the proof of (11) of lemma 2.4.1, there exists

a∗ ∈ Vγδ (a) such that aγa∗ = bγa∗ = aγb0 , a∗ δa = a∗ δb = b0 δa − − − (13). Proceeding

in the same way as in the proof of lemma 2.4.1, we can show bγa∗ δb = a − − − (14),

a∗ δbγa∗ = a∗ δaγa∗ = a∗ − − − (15), bγb0 δa = a = aγb0 δb − − − (16). Since a∗ ∈ Vγδ (a),

aγa∗ is a δ-idempotent. Hence aγa∗ ∈ Vδδ (aγa∗ ). Further

(aγa∗ )δ(bγb0 ) = aγ(a∗ δb)γb0

= aγ(a∗ δa)γb0 since a∗ δb = a0 δa

= aγb0

= aγa∗ since aγb0 = aγa∗

= (aγa∗ )δ(aγa∗ )

61
 (bγb0 )δ(aγa∗ ) = (bγb0 δa)γa∗

= aγa∗ by (16)

= (aγa∗ )δ(aγa∗ )

Hence aγa∗ ≤1 bγb0 .

Again a∗ δa is a γ-idempotent. Hence a∗ δa ∈ Vγγ (a∗ δa).

(a∗ δa)γ(b0 δb) = a∗ δ(aγb0 )δb

= a∗ δ(aγa∗ )δb since aγb0 = aγa∗

= a∗ δb

= (a∗ δb)γ(a∗ δb)

(b0 δb)γ(a∗ δa) = b0 δ(bγa∗ )δa

= b0 δ(aγa∗ )δa since by (13), bγa∗ = aγa∗

= b0 δa

= a∗ δa since by (13), b0 δa = a∗ δa

= (a∗ δa)γ(a∗ δa)

Hence a∗ δa ≤1 b0 δb.

Also, a = aγa∗ δa = aγb0 δa since by (13), a∗ δa = b0 δa. Conversely, let aγa∗ ≤1

bγb0 , a∗ δa ≤1 b0 δb where a∗ ∈ Vγδ (a), b0 ∈ Vγδ (b).

Also, let bγa∗ δb = a = aγb0 δa. From aγa∗ ≤1 bγb0 , there exist γ1 , δ1 ∈ Γ and

x ∈ S such that xδ1 (aγa∗ ) = xδ1 (bγb0 ) and (aγa∗ )δ1 x = (bγb0 )γ1 x. Now bγb0 is a

δ-idempotent. Then bγb0 ∈ Vδδ (bγb0 ). Hence following (11) of lemma 2.4.1, there exist

62
δ ∈ Γ and aγa∗ ∈ Vδδ (aγa∗ ) such that

(aγa∗ )δ(aγa∗ ) = (aγa∗ )δ(bγb0 ) − − − (17)

(aγa∗ )δ(aγa∗ ) = (bγb0 )δ(aγa∗ ) − − − (18)

Similarly from a∗ δa ≤1 b0 δb, there exists a∗ δa ∈ Vγγ (a∗ δa) such that

(a∗ δa)γ(a∗ δa) = (a∗ δa)γ(b0 δb) − − − (19)

(a∗ δa)γ(a∗ δa) = (b0 δb)γ(a∗ δa) − − − (20)

Now,

a∗ = a∗ δ(aγa∗ )

= a∗ δ(aγa∗ δbγb0 ) by (17)

= a∗ δbγb0 − − − (21)

a = (aγa∗ )δa

= (bγb0 δaγa∗ )δa by (18)

= bγb0 δa − − − (22)

Again

a = aγ(a∗ δa)

= aγ(a∗ δaγb0 δb) by (19)

= aγb0 δb − − − (23)

a∗ = (a∗ δa)γa∗

= (b0 δbγa∗ δa)γa∗ by (20)

= b0 δbγa∗ − − − (24)

63
Then

a∗ δb = b0 δbγa∗ δb since by (24), a∗ = b0 δbγa∗

= b0 δa − − − (25) since by assumption bγa∗ δb = a

Hence

a∗ = a∗ δbγb0 by (21)

= b0 δaγb0

Then

a∗ δa = (b0 δaγb0 )δa

= b0 δ(aγb0 δa)

= b0 δa − − − (26) since by assumption aγb0 δa = a

and

aγa∗ = aγ(b0 δaγb0 )

= (aγb0 δa)γb0

= aγb0 − − − (27) since by assumption aγb0 δa = a

Also,

bγa∗ = bγa∗ δbγb0 since by (21) a∗ = a∗ δbγb0

= (bγa∗ δb)γb0 since by assumption bγa∗ δb = a

= aγb0 − − − (28)

From (25) and (26), a∗ δa = a∗ δb. From (27) and (28), aγa∗ = bγa∗ . Hence a ≤1 b.

64
Definition 2.4.2. Let S be a regular Γ-semiring. Let a, b ∈ S. Then we write a ≤ b

if and only if there exist α ∈ Γ and e ∈ S such that e = eαe, a = eαb, eRa and

a ∈ bΓS. The definition of ≤ appears to be one sided. However, this is illusory since

we have the following lemma.

Lemma 2.4.3. Let S be a regular Γ-semiring and let a, b ∈ S. a ≤ b if and only if

there exist β ∈ Γ and f ∈ S such that f βf = f, a = bβf, f La and a ∈ SΓb.

Proof. Let a ≤ b. Then there exist α ∈ Γ and e ∈ S such that eαe = e, a = eαb, eRa

and a ∈ bΓS. Now, since a = eαb, a ∈ SΓb. Again since a ∈ bΓS, a = bγc for some

γ ∈ Γ and c ∈ S. Since b is regular, we can write b = bβdδb for some β, δ ∈ Γ and

d ∈ S.

Then

a = bγc

= bβdδbγc

= bβ(dδa) since bγc = a

Let f = dδa. Then

f βf = dδaβdδa

= dδeαbβdδa since a = eαb

= dδeα(bβdδa)

= dδeαa since bβdδa = a

= dδa since eRa, eαa = a

= f

65
Then a = bβf where f βf = f and a ∈ SΓb. It remains to show that f La. Now,

SΓa = SΓbβf = (SΓb)βf ⊂ SΓf and SΓf = SΓdδa = (SΓd)δa ⊂ SΓa. So,

SΓa = SΓf . Hence f La. A dual argument shows that a = bβf, f La and a ∈ SΓb,

where f βf = f together imply a ≤ b.

Theorem 2.4.4. Let S be a regular Γ-semiring. Let a, b ∈ S. Then a ≤1 b if and

only if a ≤ b.

Proof. Let a ≤1 b. Then there exist α, β ∈ Γ and a0 ∈ Vαβ (a) such that a0 βa =

a0 βb, aαa0 = bαa0 . Then a = aαa0 βa = aαa0 βb = eβb where e = aαa0 is a β-

idempotent. Again from lemma 2.2.4, it follows that aRaαa0 . So aRe. Further

aαa0 = bαa0 . So, a = bαa0 βa. Hence a ∈ bΓS. Thus a ≤ b.

Conversely let a ≤ b where a, b ∈ S. Then there exist α ∈ Γ and e ∈ S such that

e = eαe, a = eαb, eRa and a ∈ bΓS. From lemma 2.4.3, it follows that there

exist β ∈ Γ and f ∈ S such that f βf = f, a = bβf, f La and a ∈ SΓb. Now,

eRa and f La imply that eRaLf . Then from lemma 2.2.6, it follows that there

exists a unique a0 ∈ Vβα (a) such that aβa0 = e, a0 αa = f . We have to show that

a0 αa = a0 αb, aβa0 = bβa0 . Now a = eαb. Therefore a = aβa0 αb since e = aβa0 .

Hence a0 αa = a0 αaβa0 αb = a0 αb. Further since a = bβf , a = bβa0 αa. Hence

aβa0 = bβa0 αaβa0 = bβa0 . Hence a ≤1 b.

Henceforth in a regular Γ-semiring, the relation ≤1 will be written simply by ≤.

66
2.5 Completely Simple Γ-semirings

In this section we want to study completely simple Γ-semirings. Also in this

section we want to define bisimple Γ-semirings.

Lemma 2.5.1. Let S be a regular Γ-semiring. Let e = eαe and f = f βf be any two

idempotents in S. Then

(i) eDf if and only if there exist a ∈ S and b ∈ Vβα (a) such that aβb = e , bαa = f .

(ii) If J(e), J(f ) are the principal ideals generated by e and f respectively, then

J(e) ⊂ J(f ) if and only if there exist a ∈ S, b ∈ Vβα (a) such that aβb = e, (bαa)βf =

bαa = f β(bαa).

(iii) If any two idempotents in S are D-equivalent, then any two elements of S are

D-equivalent.

Proof.

(i) Let eDf . Then there exists a ∈ S such that eRaLf since D = L ◦ R = R ◦ L.

Then by lemma 2.2.6, there exists a unique b ∈ Vβα (a) such that aβb = e and bαa = f .

Conversely, let aβb = e and bαa = f for some b ∈ Vβα (a). Then from lemma 2.2.4, it

follows that aRaβb and aLbαa. Therefore aRe and aLf . Thus eRa and aLf . Hence

eDf .

(ii) Let J(e) ⊂ J(f ). Then SΓeΓS ⊂ SΓf ΓS. From lemma 2.2.3, it follows that

SαeαS ⊂ Sβf βS. Then e = c1 βf βd for some c1 , d ∈ S − − − (1). Let a = eαc1 βf

and b = f βdαe − − − (2).

67
Then

aβbαa = eαc1 βf βf βdαeαa

= eα(c1 βf βd)αeαa

= eαeαeαa

= eαa

= eαeαc1 βf

= eαc1 βf

= a by (2)

bαaβb = bαeαc1 βf βf βdαe

= bαeα(c1 βf βd)αe

= bαeαeαe

= bαe

= f βdαeαe

= f βdαe

= b by (2)

Hence b ∈ Vβα (a).

Now

aβb = eαc1 βf βf βdαe

= eα(c1 βf βd)αe

= eαeαe

= e

68
and

bαa = f βdαeαeαc1 βf

= f β(dαeαc1 )βf

Since b ∈ Vβα (a), it follows from lemma 2.2.4 that bαa is a β-idempotent. Also,

(bαa)βf = bαa and f β(bαa) = bαa. Hence (bαa)βf = bαa = f β(bαa). Conversely,

let aβb = e and (bαa)βf = bαa = f β(bαa) where b ∈ Vβα (a). Then

e = aβb

= aβbαaβb

= aβ(bαa)βb

= aβ(bαaβf )βb since (bαa)βf = bαa)

= (aβbαa)βf βb

= aβf βb ∈ Sβf βS

Hence J(e) ⊂ J(f ).

(iii) Let a, b ∈ S. Since S is regular, a = aαcβa for some α, β ∈ Γ and c ∈ S and

b = bγdδb for some γ, δ ∈ Γ and d ∈ S. Let h1 = cβa and h2 = bγd. Then h1 is

an α-idempotent and h2 is a δ-idempotent. Also h1 = cβa and a = aαh1 imply that

aLh1 and from h2 = bγd and b = h2 δb, we have h2 Rb. Now from the assumption,

h1 Dh2 . Therefore there exists c1 ∈ S such that h1 Lc1 and c1 Rh2 . Hence aLc1 and

c1 Rb. Consequently aDb.

Lemma 2.5.2. A Γ-semiring S is simple if and only if SΓaΓS = S for every a ∈ S.

69
Corollary 2.5.3. Let S be a simple Γ-semiring. Let e = eαe, α ∈ Γ be an α-

idempotent of S. Then SαeαS = S.

Proof. By lemma 2.5.2, SΓeΓS = S. Now, SαeαS ⊂ SΓeΓS −−−(1). Let s1 βeγs2 ∈

SΓeΓS, where s1 , s2 ∈ S and β, γ ∈ Γ. Then s1 βeγs2 = s1 βeαeαeγs2 = (s1 βe)αeα

(eγs2 ) ∈ SαeαS, since s1 βe ∈ S and eγs2 ∈ S. Thus SΓeΓS ⊂ SαeαS − − − (2).

From (1) and (2), we get SαeαS = SΓeΓS. Hence SαeαS = S.

Definition 2.5.1. Let S be a Γ-semiring. Let E be the set of idempotents in S. Let

e, f ∈ E. We write epf if and only if there exist α, β ∈ Γ such that eαe = e, f βf = f

and e = eαf = f βe.

We now show that p is a partial order on E.

(i) Since e is an idempotent, there exists α ∈ Γ such that eαe = e. Thus p is reflexive,

that is epe

(ii) Let epf and f pe. Then by definition there exist α, β, α1 , β1 ∈ Γ such that

eαe = e, f βf = f and e = eαf = f βe, eα1 e = e, f β1 f = f and f = f β1 e = eα1 f .

Then e = eαf = eαeα1 f = eα1 f = f . Hence p is antisymmetric.

(iii) Let epf and f pg. Then since epf , there exist α, β ∈ Γ such that eαe = e, f βf =

f and e = eαf = f βe. Again from f pg, there exist β1 , γ ∈ Γ such that f β1 f =

f, gγg = g and f = f β1 g = gγf .

Now eβe = eαf βe = eα(f βe) = eαe = e.

f β1 e = f β1 (f βe) = (f β1 f )βe = f βe = e.

eβ1 e = eαf β1 e = eαe = e.

Then eβ1 g = eαf β1 g = eαf = e and gγe = gγf β1 e = (gγf )β1 e = f β1 e = e.

So epg. Hence p is transitive. Thus p is a partial order on E.

70
Definition 2.5.2. The relation p defined on E is called the natural partial order on

E.

Definition 2.5.3. Let S be a Γ-semiring. S is called completely simple if it is simple

and contains a primitive idempotent with respect to the partial order p

Definition 2.5.4. A Γ-semiring S is bisimple if D is the universal relation in S.

Lemma 2.5.4. If S is a completely simple Γ-semiring, then

(i) S is a regular Γ-semiring

(ii) S is a bisimple Γ-semiring

(iii) Every idempotent of S is primitive

(iv) Each principal right(left) ideal of S is both maximal and minimal

Proof.

(i) Let S be a completely simple Γ-semiring. Then S is simple and contains a primitive

idempotent e(say). Let e = eαe, α ∈ Γ. By the corollary 2.5.3, SαeαS = S. Let

a ∈ S. Since S is simple, it follows from lemma 2.5.2 that SΓaΓS = S. Therefore

a = cαeαd where c, d ∈ S and e = c2 β1 aβ2 d2 where c2 , d2 ∈ S and β1 , β2 ∈ Γ. Now

a = cαeαd

= cαeαeαeαd

= (cαe)αeα(eαd)

= c1 αeαd1 where c1 = cαe and d1 = eαd

71
Again

e = eαeαe

= eα(c2 β1 aβ2 d2 )αe

= (eαc2 )β1 aβ2 (d2 αe)

Thus e = c3 β1 aβ2 d3 where c3 = eαc2 and d3 = d2 αe. Let f = d1 β2 d3 αc3 β1 c1 . Then

f αf = d1 β2 d3 αc3 β1 c1 αd1 β2 d3 αc3 β1 c1

= (d1 β2 d3 )αc3 β1 (cαeαeαd)β2 d3 αc3 β1 c1

= (d1 β2 d3 )α(c3 β1 aβ2 d3 )αc3 β1 c1 since cαeαd = a

= d1 β2 d3 αeαc3 β1 c1

= d1 β2 d2 αeαeαc3 β1 c1 since d3 = d2 αe

= d1 β2 (d2 αe)αc3 β1 c1

= d1 β2 d3 αc3 β1 c1 since d2 αe = d3

= f

Thus f is α-idempotent. Now,

f αe = d1 β2 d3 αc3 β1 c1 αe

= d1 β2 d3 αc3 β1 cαeαe since c1 = cαe

= d1 β2 d3 αc3 β1 (cαe)

= d1 β2 d3 αc3 β1 c1

= f

Proceeding as above we can show that eαf = f . Hence there exists α ∈ Γ such that

eαe = e, f αf = f and eαf = f αe = f . Thus f pe. But by hypothesis e is primitive.

72
Therefore f = e. Now,

a = c1 αeαd1

= c1 αf αd1

= c1 αd1 β2 d3 αc3 β1 c1 αd1

= cαeαeαdβ2 d3 αc3 β1 c1 αd1

= (cαeαd)β2 d3 αc3 β1 c1 αd1

= (aβ2 d3 )αc3 β1 (cαeαeαd)

= aβ2 d3 αc3 β1 a

= aβ2 (d3 αc3 )β1 a

Thus a is regular. Hence S is a regular Γ-semiring.

(ii) Let S be a completely simple Γ-semiring. Then S is simple and contains a primitive

idempotent e(say). Let e = eαe, α ∈ Γ. Also, let f = f βf, β ∈ Γ be a β-

idempotent of S. Since S is simple by the corollary 2.5.3, SαeαS = S and Sβf βS =

S. Hence SαeαS = Sβf βS. Then f ∈ SαeαS. There exist c, d ∈ S such that

f = cαeαd − − − (1).

Let a = f βcαe − − − (2) and b = eαdβf − − − (3).

73
Then

aαbβa = f βcαeαeαdβf βa

= f β(cαeαd)βf βa

= f βf βf βa

= f βa

= f βf βcαe

= f βcαe

= a

bβaαb = eαdβf βf βcαeαeαdβf

= eαdβf β(cαeαd)βf

= eαdβf βf βf

= eαdβf

= b

Hence b ∈ Vαβ (a).

Now,

aαb = f βcαeαeαdβf

= f β(cαeαd)βf

= f βf βf

= f

bβa = eαdβf βf βcαe

= eαdβf βcαe

74
Since b ∈ Vαβ (a), bβa is an α-idempotent, also,

(bβa)αe = eαdβf βcαeαe

= eαdβf βcαe

= bβa

eα(bβa) = eαeαdβf βcαe

= eαdβf βcαe

= bβa

Hence (bβa)pe. But by hypothesis e is primitive. Hence bβa = e − − − (4). Now

e = bβa and a = (f βc)αe. Thus we get eLa. Again from f = aαb and a = f β(cαe),

we get aRf . Therefore eDf . This shows that any two idempotents in S are D-related.

Since S is completely simple by (i), S is regular. Then by lemma 2.5.1(iii), any two

elements of S are D-related. Hence S is bisimple. Thus (ii) holds.

(iii) From the assumption it follows that S is simple and contains a primitive idem-

potent e(say). Let e = eαe for some α ∈ Γ. Consider a β-idempotent f ∈ S.

Let us show that f is also primitive with respect to the partial order p. In order

to prove we assume that g is γ-idempotent of S such that gpf . Then g = gγf =

f βg − − − (5). Now gβg = gγf βg = gγg = g. Proceeding as in the first part of

the proof (ii), for e, f we get a, b, c, d ∈ S such that f = cαeαd − − − (6), a =

f βcαe − − − (7), b = eαeβf − − − (8), bβa = e − − − (9) and aαbβa = a.

Now aαe = aαbβa = a − − − (10), eαb = eαeαdβf = eαdβf = b − − − (11) and

aαb = f βcαeαeαdβf = f β(cαeαd)βf = f βf βf = f − − − (12).

75
Now,

(bβgβa)α(bβgβa) = bβgβ(aαb)βgβa

= bβgβf βgβa since by (12), aαb = f

= bβgβ(f βg)βa since by (5), f βg = g

= bβgβgβa

= bβ(gβg)βa

= bβgβa since gβg = g

Also, (bβgβa)αe = bβgβ(aαe) = bβgβa since by (10), aαe = a and eα(bβgβa) =

(eαb)βgβa = bβgβa since by (11), eαb = b. Hence (bβgβa)pe. But by hypothesis e

is primitive. Hence bβgβa = e − − − (13). Again from (5), g = gγf = f βg. Hence

g = gγf = gγf βf = (gγf )βf = (f βg)βf .

Now,

g = f βgβf

= aαbβgβaαb since f = aαb

= aα(bβgβa)αb

= aαeαb since by (13), bβgβa = e

= aαb

= f

Hence f is primitive.

(iv) Since S is a regular Γ-semiring, the principal right ideal generated by a is aΓS

which becomes eαS by lemma 2.2.5, where e = eαe, α ∈ Γ. Similarly the principal

right ideal generated by b is bΓS = f βS where f = f βf, β ∈ Γ. Let eαS ≤ f βS i.e.

76
eαS ⊂ f βS. Then e = f βc1 for some c1 ∈ S. Now,

f βe = f βf βc1

= f βc1

= e

and eα(f βe) = eαe

= e

Again,

(eαf )β(eαf ) = (eαf βe)αf

= eαf

Thus eαf is β-idempotent. Also,

(eαf )βf = eα(f βf )

= eαf

and f β(eαf ) = (f βe)αf

= eαf

Hence (eαf )pf . Since S is completely simple, every idempotent of S is primitive

by result (iii). Hence eαf = f . Now, let us take any element f βd from f βS,

where d ∈ S. Therefore f βd = eα(f βd) ∈ eαS implies that f βS ⊂ eαS. Hence

eαS = f βS. Thus aΓS = bΓS. Hence each principal right ideal of S is both maximal

and minimal. Similarly we can show that the result is true for each principal left

ideal. This completes the proof the lemma.

77
Lemma 2.5.5. If S is a regular Γ-semiring and every idempotent of S is primitive,

then S is completely simple Γ-semiring.

Proof. In order to prove the lemma we need only to show that S is simple. Let

e = eαe and f = f βf be any two idempotents in S and let a ∈ Vβα (eαf ). Then

eαf βaαeαf = eαf and aαeαf βa = a − − − (1). Let g = f βaαe. Then g is α-

idempotent, for gαg = f β(aαeαf βa)αe = f βaαe = g by (1). Also

gαe = f βaαeαe

= f βaαe

= g

and f βg = f βf βaαe

= f βaαe

= g

Therefore g = gαe = f βg. Now eαg is α-idempotent since (eαg)α(eαg) = eα(gαe)αg =

eαgαg = eαg. Also,

(eαg)αe = eα(gαe)

= eαg

and eα(eαg) = (eαe)αg

= eαg

Thus (eαg)pe. Hence by hypothesis we get eαg = e. Then from eαg = e and gαe = g

we get gLe. Similarly we can show that gRf . Hence eDf . Thus all the idempotents

of S are D-related. Then by lemma 2.5.1, it follows that any two elements of S are

D-related. Thus S is bisimple and so S is simple. Hence S is completely simple.

78
 Let us now characterize the completely simple Γ-semiring by the partial order ≤

defined in the definition 2.4.1 of this chapter.

Theorem 2.5.6. Let S be a regular Γ-semiring. Then S is completely simple if and

only if the partial order ≤ on S is trivial.

Proof. Let S be a completely simple Γ-semiring. Let a ≤ b. Then there exist α ∈ Γ

and e ∈ S such that eαe = e, a = eαb, eRa and a ∈ bΓS. Since S is completely

simple, it follows from lemma 2.5.4 that each principal right ideal of S is minimal.

Then aΓS = bΓS. Hence aRb so that eRaRb. Now from eRb we get eαS = bΓS

because by lemma 2.2.3 it follows that eΓS = eαS. Then b = eαd for some d ∈ S.

Hence a = eαb = eαeαd = eαd = b.

Conversely, let ≤ be trivial. Let f be an idempotent of S. We claim that f is primitive.

Let us suppose that e is an idempotent of S such that epf . Then there exist α, β ∈ Γ

such that eαe = e, f βf = f and e = eαf = f βe. Let us show that e ≤ f . Now,

we have α ∈ Γ and e ∈ S such that eαe = e, e = eαf, eRe and e ∈ f ΓS since

e = f βe. Hence e ≤ f . But ≤ is trivial. This shows that e = f , so f is a primitive

idempotent. Hence every idempotent of S is primitive. By lemma 2.5.5, it follows

that S is completely simple Γ-semiring.

2.6 Commuting Regular Γ-semirings

Definition 2.6.1. A Γ-semiring S is called commuting regular if for each x, y ∈ S,

there exists an element s ∈ S and α, β, γ ∈ Γ such that xαy = yαxβsγyαx.

Theorem 2.6.1. Let S be a rectangular Γ-band. Then S is commutative if and only

if S is commuting regular

79
Proof. If S is a commutative Γ-semiring, then for each a, b ∈ S, there exists an

α ∈ Γ such that aαb = bαa. Since S is a rectangular Γ-band, there exists an element

c ∈ S and β, γ ∈ Γ such that aαb = bαaβcγbαa. Hence S is a commuting regular

Γ-semiring.

Conversely if S is a commuting regular Γ-semiring, for each x, y ∈ S, there exists an

element z ∈ S and α, β, γ ∈ S such that xαy = yαxβzγyαx. Since S is a rectangular

Γ-band, xαy = yαxβzγyαx = yαx. Hence S is a commutative Γ-semiring.

Theorem 2.6.2. If S is a commuting regular Γ-semiring with set E of the idempo-

tents, then E is a regular sub Γ-semiring of S. Moreover for every element a of E,

there exists an element α ∈ Γ such that a ∈ Vαα (a).

Proof. If S is a commuting regular Γ-semiring, then for each a ∈ S, there exists an

element s ∈ S and α, β, γ ∈ Γ such that aαa = aαaβsγaαa. If aαaβs = b, then

bγb = (aαaβs)γ(aαaβs) = (aαaβsγaαa)βs = aαaβs = b. Hence E is not empty. For

elements xαx = x and yδy = y in E, there exists an element t ∈ S and α, β, γ ∈ Γ

such that xαy = yαxβtγyαx. Now

(xαy)δ(xαy) = xα(yδy)αxβtγyαx

= xα(yαxβtγyαx)

= (xαx)αy

= xαy

Consequently, E is a sub Γ-semiring of S and xαxαx = x yields that E is a regular

sub Γ-semiring of S. Hence x ∈ Vαα (x).

80
Corollary 2.6.3. Let S be a commuting regular Γ-semiring with set E of the idem-

potents. Let a ∈ S and let α, β, γ ∈ Γ. If b ∈ Vαβ (a), then for any γ-idempotent e of

S, (i) aαeγb is β-idempotent (ii) bβeγa is α-idempotent.

Proof. Let b ∈ Vαβ (a). Then (aαb)β(aαb) = (aαbβa)αb = aαb and (bβa)α(bβa) =

(bβaαb)βa = bβa. Since (eγbβa)α(eγbβa) = eγbβa and (eγaαb)β(eγaαb) = eγaαb,

by theorem 2.6.2, eγbβa, eγaαb ∈ E. Now,

(aαeγb)β(aαeγb) = aα(eγbβaαeγbβa)αb

= aαeγ(bβaαb)

= aαeγb

Moreover,

(bβeγa)α(bβeγa) = bβ(eγaαbβeγaαb)βa

= bβeγ(aαbβa)

= bβeγa

Theorem 2.6.4. Let S be a commuting regular Γ-semiring with set E of α-idempotents.

Let e, f ∈ E and α ∈ Γ. Then the set Sαα (e, f ) = {g ∈ Vαα (eαf ) ∩ E; gαe = f αg = g}

is a regular sub Γ-semiring of S.

Proof. Since S is a commuting regular Γ-semiring, there exists an element s ∈ S and

α, β, γ ∈ Γ such that eαf = f αeβsγf αe.

81
Then

(eαf )α(eαf ) = f αeβsγf α(eαe)αf

= (f αeβsγf αe)αf

= eα(f αf )

= eαf

Now, (eαf )α(eαf )α(eαf ) = eαf , (eαf )αe = f αeβsγf α(eαe) = eαf and f α(eαf ) =

(f αf )αeβsγf αe = eαf . This yields (eαf ) ∈ Sαα (e, f ) and we can also prove that

f αe ∈ Sαα (e, f ) which implies Sαα (e, f ) 6= φ. Let x, y ∈ Sαα (e, f ). Since S is a

commuting regular Γ-semiring, there exists an element t ∈ S and α, β, γ ∈ Γ such

that xαy = yαxβtγyαx. Then

(xαy)α(xαy) = xα(yαy)αxβtγyαx

= xα(yαxβtγyαx)

= (xαx)αy

= xαy

(eαf )α(xαy)α(eαf ) = (eαf αxαeαf )αyαeαf

= eαf αyαeαf

= eαf

(xαy)α(eαf )α(xαy) = xα(yαe)α(f αx)αy

= (xαy)α(xαy)

= xαy

82
Hence xαy ∈ Vαα (eαf ) ∩ E. Moreover,

(xαy)αe = xα(yαe)

= xαy

and f α(xαy) = (f αx)αy

= xαy

Let x ∈ Sαα (e, f ). Then

xα(eαf )αx = (xαe)α(f αx)

= xαx

= x

This shows that Sαα (e, f ) is a regular sub Γ-semiring of S.

Remark 2.6.1. The set Sαα (e, f ) is called the (α, α) sandwich set of e and f . It

has an obvious alternative characterization Sαα (e, f ) = {gαg = g ∈ S; gαe = g =

f αg, eαgαf = eαf }.

Lemma 2.6.5. Let S be a commuting regular Γ-semiring. Let a, b ∈ S and let

α, β, γ ∈ Γ. Suppose a0 ∈ Vαβ (a) and b0 ∈ Vβα (b). Then for each g ∈ Sαα (a0 βa, bβb0 ),

b0 αgαa0 ∈ Vββ (aαb).

83
Proof.

(aαb)β(b0 αgαa0 )β(aαb) = aα(bβb0 αg)αa0 βaαb

= aα(gαa0 βa)αb

= aαgαb

= aα(a0 βaαgαbβb0 )αb

= (aαa0 βa)α(bβb0 αb)

= aαb

Moreover

(b0 αgαa0 )β(aαb)β(b0 αgαa0 ) = b0 α(gαa0 βa)α(bβb0 αg)αa0

= b0 α(gαg)αa0

= b0 αgαa0

Hence b0 αgαa0 ∈ Vββ (aαb).

Theorem 2.6.6. Let S be a commuting regular Γ-semiring. Let a, b ∈ S and α, β ∈ Γ.

Then Vβα (b)ΓVαβ (a) ⊆ Vββ (aαb).

Proof. Let a, b ∈ S and let α, β ∈ Γ. Suppose a0 ∈ Vαβ (a) and b0 ∈ Vβα (b). Then by

lemma 2.6.5, b0 αgαa0 ∈ Vββ (aαb) for all g in Sαα (a0 βa, bβb0 ). Now, (a0 βa)α(a0 βa) =

(a0 βaαa0 )βa = a0 βa and (bβb0 )α(bβb0 ) = (bβb0 αb)βb0 = bβb0 . Then by theorem 2.6.4,

bβb0 αa0 βa ∈ Sαα (a0 βa, bβb0 ). By lemma 2.6.5, b0 α(bβb0 αa0 βa)αa0 ∈ Vββ (aαb). Hence

b0 αa0 ∈ Vββ (aαb).

Theorem 2.6.7. Let S be a commuting regular Γ-semiring with set E of idempotents.

Let α, β ∈ Γ. Then Vαβ (e) ⊆ E for every e in E.

84
Proof. Suppose x ∈ Vαβ (e). Then xβeαx = x and eαxβe = e. Since xβe is α-

idempotent and eαx is β-idempotent, xβe ∈ Vαα (xβe) and eαx ∈ Vββ (eαx). Then by

theorem 2.6.6, (eαx)β(xβe) ∈ Vβα (xβeαeαx) which implies that eαxβxβe ∈ Vβα (x).

Now

x = xβ(eαxβxβe)αx

= (xβeαx)β(xβeαx)

= xβx

which implies x ∈ E.

Theorem 2.6.8. Let S be a Γ-semiring. Then S is a commuting regular if and only

if SΓS is a clifford Γ-semiring

Proof. Suppose S is a commuting regular Γ-semiring. Then for x ∈ S and an α-

idempotent e ∈ S, there exists s ∈ S and α, β, γ ∈ Γ such that xαe = eαxβsγeαx.

Now,

eαxαe = (eαe)αxβsγeαx

= eαxβsγeαx

= xαe

Symmetrically, eαxαe = eαx. Hence eαx = xαe which follows that the idempotents

in S are central. For any x, y ∈ S, there exists z ∈ S and α1 , β1 , γ1 ∈ Γ such that

xα1 y = yα1 xβ1 zγ1 yα1 x. Moreover, there exists u ∈ S and α2 , β2 ∈ Γ such that

zγ1 yα1 x = yα1 xγ1 zα2 uβ2 yα1 xγ1 z. Hence xα1 y = (yα1 xβ1 yα1 x)γ1 zα2 uβ2 yα1 xγ1 z.

Since yα1 xβ1 yα1 x is regular in commuting regular Γ-semiring S, yα1 xβ1 yα1 x =

yα1 xβ1 yα1 xα3 vβ3 yα1 xβ1 yα1 x for some v ∈ S and α3 , β3 ∈ Γ. Let e = yα1 xβ1 yα1 xα3 v.

85
Then e is a β3 -idempotent in S such that yα1 xβ1 yα1 x = eβ3 yα1 xβ1 yα1 x and so

xα1 y = eβ3 (yα1 xβ1 yα1 xγ1 zα2 uβ2 yα1 xγ1 z) = eβ3 xα1 y. Now, there exists w ∈ S and

α4 , β4 ∈ Γ such that eβ3 xα1 y = xα1 yβ3 eα4 wβ4 xα1 yβ3 e. Then xα1 y = xα1 yβ3 eα4 wβ4

(xα1 yβ3 e) = (xα1 y)β3 (eα4 wβ4 eβ3 )(xα1 y) which implies that xα1 y is regular. Hence

SΓS is regular.

Conversely suppose SΓS is a clifford Γ-semiring. Since the idempotents of S are

central, for any x, y ∈ S and α, β ∈ Γ, there exists an idempotent e such that

xαy = eβ(xαy). Let u = eαx and v = eαy. As u, v ∈ SΓS, there exist u0 , v 0 ∈ S and

β, γ, δ ∈ Γ such that uβu0 δu = u and vδv 0 γv = v. Since uβu0 and v 0 γv are central

idempotents,

(yαx)β(u0 δv 0 βeγxαyγu0 δv 0 )γ(yαx) = vδ(uβu0 δv 0 βuγvγu0 δv 0 γv)δu

= (vδv 0 γv)δv 0 βuγvγu0 δ(uβu0 δu)

= ((vδv 0 )βu)γvγu0 δu

= uβ(vδv 0 γv)γ(u0 δu)

= (uβu0 δu)γ(vδv 0 γv)

= uγv

= eγ(xαy)

= xαy

Hence S is a commuting regular.

Definition 2.6.2. If a is an element of a Γ-semiring S, the smallest left ideal of S

containing a is SΓa ∪ {a}. An equivalence L on S is defined by the rule that aLb if

and only if SΓa ∪ {a} = SΓb ∪ {b}. Similarly we define the equivalence R by the rule

that aRb if and only if aΓS ∪ {a} = bΓS ∪ {b}.

86
Lemma 2.6.9. Let a and b be elements of a Γ-semiring S. Then aLb if and only if

there exist x and y in S and α, β ∈ Γ such that xαa = b and yβb = a. Also, aRb if

and only if there exist u and v in S and γ, δ ∈ Γ such that aγu = b and bδv = a

Proof. We omit the proof as it is a matter of routine verification.

Lemma 2.6.10. The relations L and R commute

Proof. Let (a, b) ∈ L ◦ R. Then there exists c ∈ S such that aLc and cRb.

Case (i): a = c. Then aRb. Since aLa and aRb, (a, b) ∈ L ◦ R.

Case (ii): b = c. Then aLb. Since aLb and bRb, (a, b) ∈ L ◦ R.

Case (iii): a 6= c and b 6= c. Since aLc and cRb, there exist x, y, u, v ∈ S and

γ, µ, η, θ ∈ Γ such that xγa = c, yµc = a, cηu = b, bθv = c. Let d = yµcηu. Then

aηu = yµcηu = d and dθv = yµcηuθv = yµbθv = yµc = a from which it follows

aRd. Also yµb = yµcηu = d and xγd = xγyµcηu = xγaηu = cηu = b, so dLb. We

deduce that (a, b) ∈ R ◦ L. Therefore L ◦ R ⊆ R ◦ L. Similarly, we can prove that

R ◦ L ⊆ L ◦ R.

Theorem 2.6.11. Let S be a commuting regular Γ-semiring and a, b ∈ S. Then aLb

if and only if aRb.

Proof. Suppose that aLb. By lemma 2.6.9, there are x and y in S and α, β ∈ Γ such

that xαa = b and yβb = a. So, there are t1 , t2 in S and γ1 , γ2 , δ1 , δ2 in Γ such that

b = xαa = aαxγ1 t1 γ2 aαx and a = yβb = bβyδ1 t2 δ2 yβb where u = xγ1 t1 γ2 aαx, v =

yδ1 t2 δ2 yβb. This implies aRb. Proof of the converse is similar.

87
Remark 2.6.2. The equivalence D is a two sided analogue of L and R. Also, we

define the equivalence J by the rule aJb if and only if SΓa ∪ aΓS ∪ SΓaΓS ∪ {a} =

SΓb ∪ bΓS ∪ SΓbΓS ∪ {b} if and only if there exist x, y, u, v ∈ S and α1 , α2 , β1 , β2 ∈ Γ

such that xα1 aα2 y = b and uβ1 bβ2 v = a. It is immediate that L ⊆ J and R ⊆ J.

Hence since D is the smallest equivalence containing L and R, we get D ⊆ J.

Theorem 2.6.12. If S is a commuting regular Γ-semiring, then D = J.

Proof. By remark 2.6.2, it is enough to show that J ⊆ D. For elements a and

b in S, let aJb. Then there are x, y, u, v ∈ S and α1 , α2 , β1 , β2 ∈ Γ such that

xα1 aα2 y = b; uβ1 bβ2 v = a. So there exists an element t1 in S and γ1 , γ2 ∈ Γ such that

a = uβ1 bβ2 v = (uβ1 xα1 a)α2 (yβ2 v) = (yβ2 vα2 uβ1 xα1 aγ1 t1 γ2 yβ2 vα2 u)β1 (xα1 a) =

w1 β1 c where w1 = yβ2 vα2 uβ1 xα1 aγ1 t1 γ2 yβ2 vα2 u and c = xα1 a and so aLc. Com-

bining the relations xα1 aα2 y = b and c = xα1 a, we get cα2 y = b. Then there

exists an element t2 ∈ S and δ1 , δ2 ∈ Γ such that c = xα1 a = (xα1 u)β1 (bβ2 v) =

bβ2 (vβ1 xα1 uδ1 t2 δ2 bβ2 vβ1 xα1 u) = bβ2 w2 , where w2 = vβ1 xα1 uδ1 t2 δ2 bβ2 vβ1 xα1 u. This

shows that cRb. Hence J ⊆ D.

88