You are on page 1of 22

Semigroup Forum Vol.

70 (2005) 369–390

c 2005 Springer
DOI: 10.1007/s00233-004-0155-8

RESEARCH ARTICLE

Quasi−C−Ehresmann Semigroups and Their


Subclasses

Gang Li, Y.Q. Guo ∗ and K.P. Shum †


Communicated by Steve Pride

Abstract
We introduce the concept of a quasi −C− Ehresmann semigroup. Some spe-
cial quasi −C− Ehresmann semigroups will be considered. We will show that
the quasi −C− Ehresmann semigroups are special U − semiabundant semigroups
and orthodox U − liberal semigroups, which have been recently studied by sev-
eral authors. Some characterization theorems of this new class of semigroups
are obtained. Many results of quasi −C− semigroups in the literature can be
extended to quasi −C− Ehresmann Semigroups.
Key words and phrases: Quasi −C− Ehresmann Semigroups, Orthodox U −
liberal semigroups, U − semiabundant semigroups, Quasi −C− semigroups.
AMS Mathematics Subject Classification (2002): 20M10.

1. Introduction
In generalizing regular semigroups, Fountain considered in [3] a generalization
of Green’s relations, namely, Green’s star relations. He called a semigroup S
an abundant semigroup if every L∗ −class and every R∗ −class of S contain
some idempotents of S . Abundant semigroups and their subclasses have been
extensively studied by a number of authors and fruitful results have been ob-
tained. In particular, the set E(S) of all idempotents of the semigroup S plays
an essential role because the structure of the abundant semigroup S is more or
less dominated by the properties of E(S) .
In recent years, Lawson and others have observed that one can pay special
attention to a subset U of E(S) instead of the whole set E(S) . In particular,
Lawson noticed that if U is a subset of E(S) of a semiabundant semigroup S
then U is perhaps good enough to provide sufficient information for the whole
semigroup S . In order to emphasize that U is a subset of E(S) of a semigroup
S , we usually denote this semigroup by S(U ) . Then, we can denote−analogues
 − U −relations by
of Green’s relations on the semigroup S(U ) , namely, Green
L , R
U  and H
U  , respectively. If every L and every R
U U  U −class of S(U )
contains at least one element of U , then, following the definition of abundant
∗ This research is supported by a NSF grant, China  10471112.
† This research is partially supported by a UGC(HK) grant  2160210(03/05).
370 Li, Guo, and Shum

semigroup, we call the semigroup S(U ) a U −semiabundant semigroup. Unlike


the usual Green’s relations on a semigroup S , Lawson has observed in [12]
that LU is not necessarily a right congruence on S and R  U is not necessarily
a left congruence on S . Due to the above important observation, Fountain,
Gomes and Gould in [4] said that a semigroup S(U ) satisfied the (CR) [ (CL)
] condition if LU is a right congruence [ R
 U is a left congruence ] on S . Then,
they called a U −semiabundant semigroup S(U ) an Ehresmann semigroup if U
is a subsemilattice of S and S(U ) satisfies both the (CR) and (CL) conditions.
An Ehresmann semigroup S(U ) in which the idempotents of U are central is
called a C −Ehresmann semigroup. The structure of C−Ehresmann semigroups
have been described by Gomes and Gould in [5]. In fact, this kind of semigroups
can be regarded as a kind of generalized Clifford semigroups in the range of
U −semiabundant semigroups.
On the other hand, for any element a of a semigroup S , Heyong has
defined in [9] the orbit of a in U ⊆ E(S) by Ua = {u ∈ U | ua = a = au} and
then he defined Q U = {(a, b) ∈ S × S| Ua = Ub } . Clearly, Q  U is an equivalence
a
class of the semigroup S containing a ∈ S . The semigroup S(U ) is called
by Heyong in [9] a U −liberal semigroup if every Q U −class of S contains an

element of U . It is routine to check that a Q −class contains at most one
U

element of U . Denote the unique element in Q  U ∩ U , if it exists, by a◦ . The


a U
structure of U −liberal semigroups and left C−Ehresmann semigroups have also
been recently investigated by Heyong in [9].
For a U −liberal semigroup S(U ) , we call the following condition the
Ehresmann type condition, in brevity, the ET-condition:

(∀a, b ∈ S) (ab)◦U D(U ) a◦U b◦U ,

where
D(U ) = {(e, f ) ∈ U × U | (∃g ∈ U ) eRgLf }.
Now, a U −liberal semigroup S(U ) is called an orthodox U −liberal semigroup
if U is a subsemigroup of S(U ) and the ET-condition holds on S(U ). We
can observe that U −liberal semigroups are not necessarily orthodox U −liberal
semigroups.. For example, if we let S = {1, e, a, b, f, 0} be a semigroup with
the following Cayley table,

. 1 e a f b 0
1 1 e a f b 0
e e e a 0 0 0
a a 0 0 a e 0
f f 0 0 f b 0
b b b f 0 0 0
0 0 0 0 0 0 0

then we can immediately see that S is a finite inverse monoid and so S(E(S))
is a E(S)−liberal semigroup and E(S) is a subsemigroup of S(E(S)) . Since
Li, Guo, and Shum 371

a◦E(S) = 1 , e◦E(S) = e and (ae)◦E(S) = 0 , the ET-condition does not hold on


S(E(S)) . Thus, the above semigroup is an example of U −liberal semigroups
that are not orthodox U −liberal semigroups.
In fact, we will see that the orthodox U −liberal semigroups are general-
ized orthogroups in the range of U −semiabundant semigroups.
As an analogue of regular orthogroups [17] i.e. quasi−C−semigroups
[8], we call an orthodox U −liberal semigroup S(U ) a quasi−C−Ehresmann
semigroup if U is a regular band.
In this paper, we will describe the structure of quasi−C−Ehresmann
semigroups and provide a construction for such kind of semigroups. Many
important results on quasi−C−semigroups in the range of regular semigroups
will be extended to quasi− C− Ehresmann semigroups in the range of U −
semiabundant semigroups. For notations and terminologies not given in this
paper, the reader is referred to the papers [9], [10] and [12]. Throughout this
paper, E(S) denotes the set of all idempotents of a semigroup S and U is a
non-empty subset of E(S) .

2. Basic definitions and results


The following notations and definitions were given by Lawson [12] and Heyong
[9].

Notation 2.1. Let U = ∅ be a subset of E(S) . Then the following set

RegU (S) = {a ∈ S|(∃e, f ∈ U )e L a R f }

is called the set of U −regular elements of S . Clearly U ⊆ RegU (S) . Moreover,


it is clear that a ∈ RegU (S) if and only if a ∈ Reg(S) and VU (a) = {a ∈
V (a)| aa , a a ∈ U } = ∅ .
We call the elements in VU (a) the U −inverses of a ∈ S .

For any a ∈ S . Define Ual = {u ∈ U | ua = a} and Uar = {u ∈ U | au = a} .


Let Ua = Ual ∩ Uar . Then Ua = {u ∈ U | ua = a = au} . We call the set Ua the
orbit of a in U .

Notation 2.2. [12] [9] Define the following equivalence relations on the semi-
group S with respect to U ⊆ E(S) :

LU = {(a, b) ∈ S × S| Uar = Ubr },


 U = {(a, b) ∈ S × S| U l = U l },
R a b
 U = LU ∩ R
H U ,
QU = {(a, b) ∈ S × S| Ua = Ub }.

As usual, the LU , R


U , H
 U and Q
U -classes containing a ∈ S are denoted
by U
L  , H
, R U  and Q
U  respectively. It is noted by Lawson in [12] that LU
U
a a a a
372 Li, Guo, and Shum

is not necessarily a right congruence on S and R  U is not necessarily a left


congruence on S .
We say that a semigroup S(U ) satisfies the (CR) condition if LU is a
right congruence on S and that S(U ) satisfies the (CL) condition if R  U is a
left congruence on S . If the semigroup S(U ) satisfies both the (CR) and (CL)
conditions then we say that S(U ) satisfies the (C) condition.

Notation 2.3. Let S be a semigroup. By S = [Y ; Sα ] , we mean S is a


semilattice Y of the semigroups Sα (α ∈ Y ). In particular, if S is a band, then
[Y ; Sα ] is the greatest semilattice decomposition of S . By S = [Y ; Sα ; φα,β ] ,
we mean S is a strong semilattice of the semigroups Sα (α ∈ Y ) with structure
homomorphisms φα,β . For any equivalence relation ρ on the semigroup S , we
sometimes denote ρ by ρ(S), if necessary.

Definition 2.4. [9] [12] A semigroup S(U ) is called a U −semiabundant



semigroup if every L − and every R
U  U −class of S contains at least one element
 
from U , that is, La ∩ U = ∅ and RaU ∩ U = ∅ for all a ∈ S .
U

Definition 2.5. A semigroup S(U ) is called a U −semi −superabundant


semigroup if every H  U −class of S contains an element of U , that is, H
 U ∩U = ∅
a
for all a ∈ S . In this case, the unique element in H ∩ U is denoted by a .
U
a U

The following results were given in [9] and [12].

Lemma 2.1. Let F be one of Green’s relations L, R or H and FU the


 relations on the semigroup S . Then, for any a, b ∈ S , we
corresponding Green
have
(i) F ⊆ FU and for a, b ∈ RegU (S), (a, b) ∈ FU if and only if
(a, b) ∈ F .
(ii) HU ⊆ QU and Q  U contains at most one element in U .
a
(iii) If S(U ) is a U −semi−superabundant semigroup then S(U ) is a
U −liberal semigroup with Q U = H U .

The following definitions are well known, for example, see Howie [10].

Definition 2.6. Let B be a band. Then


(i) B is called a left regular band if and only if ef e = ef , for all e, f ∈ B .
(ii) B is called a left normal band if and only if f ge = f eg , for all
e, f, g ∈ B .
(iii) B is called a left quasinormal band if and only if ef eg = ef g , for
all e, f, g ∈ B .
(iv) B is called a regular band if and only if ef ege = ef ge, for all
e, f, g ∈ B .
Li, Guo, and Shum 373

The right regular bands, right normal bands and right quasinormal bands
can be dually defined.

Definition 2.7. Let T = [Y ; Tα ] be a semilattice of monoids Tα whose


identity is denoted by 1Tα . If T = [Y ; Tα , φα,β ] is a strong semilattice with
structure monoid homomorphisms φα,β such that φα,β : Tα −→ Tβ is defined
by x → x1Tβ for any β ≤ α , then T is called an E− semilattice of monoids
Tα , where E = {1Tα |α ∈ Y } .

In the above case, the subsemigroup E = {1Tα |α ∈ Y } is isomorphic to


Y and every 1Tα is central in T .
The following important result was proved by Fountain, Gomes and Gould
in [4].

Lemma 2.2. [4] Let T be a semigroup with E ⊆ E(T ). Then, T (E) is a


C−Ehresmann semigroup if and only if T (E) is a E−semilattice of monoids.

In view of the above Lemma, we call T = [Y ; Tα ] a C −Ehresmann


semigroup if T (E) is a E−semilattice of monoids Tα , where E = {1Tα |α ∈ Y } .
In brevity, we just call I × T × Λ , the direct product of a rectangular
band I × Λ and a monoid T , a rectangular monoid. The direct product I × T
of a left zero semigroup I and a monoid T is called a left monoid.
Now, let I = [Y ; Iα ] and T = [Y ; Tα ] be two semigroups. Form the
set product Sα = Iα × Tα and the set union S = Sα . Define the following
α∈Y
mapping η : S → Tl (I) by (i, a) → (i, a) , where Tl (I) is the left transformation
semigroup acting on I . If the mapping η satisfies the following conditions: for
any (i, a) ∈ Sα and (j, b) ∈ Sβ ,
(i) (i, a) j ∈ Iαβ , in particular, (i, a) j = ij when α ≤ β ;
(ii) (i, a) (j, b) = ((i, a) j, ab) ,
then we can verify that S becomes a semigroup under the multiplication

(i, a)(j, b) = ((i, a) j, ab).

According to Zhu, Guo and Shum [19], the above semigroup S is called a left
semi-spined product of I and T with respect to Y and η , denoted by I ×Y,η T .
We call a U −liberal semigroup S(U ) a left C−Ehresmann semigroup
if U is a subsemigroup of S such that S(U ) satisfies the ET-condition and
uS ⊆ Su for all u ∈ U . Heyong then proved the following result in [9].

Lemma 2.3. The following statements are equivalent for a semigroup S :


(i) S(U ) is a left C−Ehresmann semigroup for some U ⊆ E(S);
(ii) S is a semilattice Y of left monoids Sα = Iα × Tα , α ∈ Y and
U = {(i, 1Tα )|i ∈ Iα , α ∈ Y } is a subsemigroup of S ;
374 Li, Guo, and Shum

(iii) S is a left semi-spined product of a left regular band I = [Y ; Iα ] and


a C− Ehresmann semigroup T = [Y ; Tα ], with respect to Y and a structural
mapping η .

In view of the above results, we hence call the semigroup S = [Y ; Sα =


Iα ×Tα ] a left C−Ehresmann semigroup if S is a semilattice of the left monoids
Sα = Iα × Tα with U = {(i, 1Tα )|i ∈ Iα , α ∈ Y } as a subsemigroup of S .
It is now clear that the left C−Ehresmann semigroups can be regarded as
the generalization of the left C−semigroups (see Zhu, Guo and Shum [19]) from
the range of regular semigroups to the range of U −semiabundant semigroups.
The following is a construction theorem for left C−Ehresmann semi-
groups which will be useful in the sequel.

Lemma
 2.4. [9] Let T = [Y ; Tα ] be a C−Ehresmann semigroup and S =
Sα , where Sα = Iα × Tα , Iα is a non-empty set. Let U = {(i, 1Tα )|i ∈
α∈Y
Iα , α ∈ Y } . For any γ ≤ α in Y , define a mapping

Ψα,γ : Sα → Tl (Iγ ) by (i, a) → ψα,γ


(i,a)

satisfying the following conditions for any α, β ∈ Y :


(i,a)
(i) for any (i, a) ∈ Sα , ψα,α is the constant mapping on Iα , whose value
is i;
(i,a) (j,b)
(ii) for any (i, a) ∈ Sα and (j, b) ∈ Sβ , ψα,αβ ψβ,αβ is the constant
(i,a) (j,b)
mapping on Iαβ , whose value is denoted by ψα,αβ ψβ,αβ ;
(k,ab) (i,a) (j,b) (i,a) (j,b)
(iii) for any δ ≤ αβ ∈ Y , ψαβ,δ = ψα,δ ψβ,δ , where k = ψα,αβ ψβ,αβ .
Then S(U ) forms a left C−Ehresmann semigroup under the multiplication

(i,a) (j,b)
(i, a)(j, b) = (ψα,αβ ψβ,αβ , ab),

for all (i, a) ∈ Sα and (j, b) ∈ Sβ .


Conversely, every left C−Ehresmann semigroup can be constructed in the
above manner.

Recently, Li in [11] called a U −liberal semigroup S(U ) a weakly left


C−Ehresmann semigroup if U is a subsemigroup of S(U ) such that S(U )
satisfies the ET-condition and uxuy = uxy for all u ∈ U and x, y ∈ S .
In fact, weakly left C−Ehresmann semigroups are generalized weakly left
C−semigroups (see Guo [7]) in the range of U −semiabundant semigroups.
If S and T are two semigroups having a common homomorphic image
H and if φ : S → H and ψ : T → H are homomorphisms onto H , the spined
product of S and T is defined as

S ×H T = {(s, t) ∈ S × T |sφ = tψ}


Li, Guo, and Shum 375

Li has also observed in [11] that the weakly left C−Ehresmann semi-
groups are related to right normal bands. He proved the following result.

Lemma 2.5. [11] A semigroup S(U ) is a weakly left C−Ehresmann semi-


group for some U ⊆ E(S) if and only if there exist a left C−Ehresmann
semigroup S1 = [Y ; Iα × Tα ] and a right normal B = [Y ; Bα ; φα,β ] such that
S∼= S1 ×Y B .

Notation 2.8. Let X be the subdirect product of the semigroups {Si |i =


1, 2, . . . , n} . Denote the projection of X onto Si × Sj by PSi ×Sj such that
(a1 , a2 , . . . , an )PSi ×Sj = (ai , aj ) , for any (a1 , a2 , . . . , an ) ∈ X . Similarly, the
projection of X onto Si is denoted by PSi such that (a1 , a2 , . . . , an )PSi = ai ,
for any (a1 , a2 , . . . , an ) ∈ X .

3. Properties of orthodox U −liberal semigroups


Recall that a semigroup S(U ) is an orthodox U −liberal semigroup if U is a
subsemigroup of the U −liberal semigroup S(U ) satisfying the ET-condition:
(ab)◦U D(U ) a◦U b◦U ,for any a, b ∈ S .
In the literature,orthogroups and their generalizations have been exten-
sively studied by a number of authors,for example,see Petrich and Reilly in
[17]. In fact,orthodox U −liberal semigroups can be regarded as analogues of
orthogroups in the range of U −semiabundant semigroups. The followings are
some characterization theorems for such semigroups.

Theorem 3.1. The following statements are equivalent for a semigroup S :


(i) S(U ) is a U −liberal semigroup for some U ⊆ E(S) and U itself is
a rectangular band;
(ii) S(U ) is an orthodox U −liberal semigroup such that U is a rectan-
gular band;
(iii) S is isomorphic to a rectangular monoid.

Proof. (i) =⇒ (ii). This part is trivial.


(ii) =⇒ (iii). Let u ∈ U and form Su = {x ∈ S| x◦U = u} , where
x◦U is the unique idempotent in Q  U ∩ U . Then, for any a, b ∈ Su , we have
x
u ∈ Uab = U(ab)◦U so that u ≥ (ab)◦U , by using the natural ordering on U .
Moreover, since U is a rectangular band, we can easily deduce that (ab)◦U = u .
This implies that ab ∈ Su and consequently Su is a monoid with the identity
u . Now, for any v ∈ U , we define a mapping ξv,u : Sv → Su by x → uxu for
x ∈ Sv . Since u ∈ Uuxu , we have (uxu)◦U = u . This shows that ξv,u is well
defined. By the fact that U is a rectangular band, we can immediately deduce
that
xξv,u ξu,v = vuxuv = vuvxvuv = vxv = x , for all x ∈ Sv ,
yξu,v ξv,u = uvyvu = uvuyuvu = uyu = y , for all y ∈ Su .
−1
Consequently, ξv,u is a bijective mapping and ξv,u = ξu,v . Furthermore, we can
376 Li, Guo, and Shum

easily see that for all x, y ∈ S , we have

(xy)ξ(xy)◦U ,u = uxyu = uxx◦U uyU



yu = uxuuyu = xξx◦U ,u yξyU◦ ,u .

Now, it is routine to check that

ξ : S → Su × U by x → (xξx◦U ,u , x◦U )

is an isomorphism, and so, under the above isomorphism ξ , U is mapped


bijectively from S onto {u}×U . Since U is a rectangular band, S is isomorphic
to the rectangular monoid Su × U .
(iii) =⇒ (i). Assume that S = I × T × Λ is a rectangular monoid and
B = I × Λ is a rectangular band. Then, it is routine to verify that S(U ) is a
U −liberal semigroup with U = I × {1T } × Λ a rectangular band. The proof is
completed.

Notation 3.1. If S = I × T × Λ is a rectangular monoid and B = I × Λ


is a rectangular band, then we denote S by S(U ) in order to emphasis that
U = I × {1T } × Λ .

Theorem 3.2. The following statements are equivalent for a semigroup S :


(i) S(U ) is an orthodox U −liberal semigroup for some U ⊆ E(S);
(ii) S = [Y ; Sα (Uα )], where Sα (Uα ) is a rectangular monoid for every
α ∈ Y and U = Uα is a subsemigroup of S ;
α∈Y
(iii) S(U ) is a U −semi-superabundant semigroup satisfying the (C) con-
dition for some U ⊆ E(S) and U is a subsemigroup of S .

Proof. (i) =⇒ (ii). Let S(U ) be an orthodox U −liberal semigroup for


some U ⊆ E(S) . Then we can write U = [Y ; Uα ] . For any α ∈ Y , we form
the set Sα = {x ∈ S| x◦U ∈ Uα } . Since S(U ) satisfies the ET-condition, for all
x ∈ Sα , y ∈ Sβ , we have (xy)◦U D(U )x◦U yU

. This leads to xy ∈ Sαβ and hence
S = [Y ; Sα ] . Since every semigroup Sα (Uα ) is a Uα −liberal semigroup and Uα
is a rectangular band, by Theorem 3.1, Sα (Uα ) is isomorphic to a rectangular
monoid. This proves (ii).
(ii) =⇒ (iii). By (ii), we can write Sα (Uα ) = Iα × Tα × Λα for every
α ∈ Y . Then, we can easily verify that

(i, x, λ)LU (j, y, µ) ⇔ α = β and λ = µ ∈ Λα , for any (i, x, λ) ∈ Sα , (j, y, µ) ∈ Sβ .

Similarly, if (i, x, λ)LU (j, y, λ) for some λ ∈ Λα , then for all (k, z, ν) ∈
Sγ (γ ∈ Y ) , we have

(i, 1Tα , λ)(k, z, ν) = (k  , z  , ν  ) (∈ Sαγ ),


(i, x, λ)(k  , 1Tαγ , ν  ) = (i , x , λ ),
(j, y, λ)(k  , 1Tαγ , ν  ) = (j  , y  , λ ).
Li, Guo, and Shum 377

Consequently, by using the above relations, we derive that

(i, x, λ)(k, z, ν) = (i, x, λ)(i, 1Tα , λ)(k, z, ν)


= (i, x, λ)(k  , z  , ν  )
= (i, x, λ)(k  , 1Tαγ , ν  )(k  , z  , ν  )
= (i , x , λ )(k  , z  , ν  )
= (i , x z  , ν  ),

and

(j, y, λ)(k, z, ν) = (j, y, λ)(i, 1Tα , λ)(k, z, ν)


= (j, y, λ)(k  , z  , ν  )
= (j, y, λ)(k  , 1Tαγ , ν  )(k  , z  , ν  )
= (j  , y  , λ )(k  , z  , ν  )
= (j  , y  z  , ν  ).

Thereby, we obtain that (i, x, λ)(k, z, ν)LU (j, y, λ)(k, z, ν) so that LU is a right
congruence on S .
Similarly, we can show that (i, x, λ)R  U (j, y, µ) ⇔ (∃α ∈ Y ) i = j ∈ Iα ,

and so R is a left congruence on S .
U

Thus, S(U ) is a U −semi-superabundant semigroup. Hence, (iii) holds.


(iii) =⇒ (i). Assume that (iii) holds. Then, by Lemma 2.1, S(U ) is a
U −liberal semigroup and for all a ∈ S(U ) , a◦U = aU . Since S(U ) satisfies the
(C) condition, we have, for all a, b ∈ S(U ) ,

(ab)U R  U ab R
 U ab R U
 U (abU )U LU abU LU aU bU .

 U (ab ) LU a b . Thus, by Lemma 2.1(i), we have


This leads to (ab)U R U U U U
(ab)U R (abU )U L aU bU . Consequently, (ab)◦U = (ab)U D(U ) aU bU = a◦U b◦U
  

holds. This shows that S(U ) satisfies the ET-condition. Hence, (i) holds.

Hereafter we call S(U ) = [Y ; Iα × Tα × Λα ] an orthodox U −liberal


semigroup if S is a semilattice of rectangular monoids Iα × Tα × Λα such
that U = (Iα × {1Tα } × Λα ) is a subsemigroup of S . From the proof of
α∈Y
Theorem 3.2, we see that if S(U ) = [Y ; Iα × Tα × Λα ] is an orthodox U −liberal
semigroup, then for any (i, x, λ) ∈ Iα × Tα × Λα , (j, y, µ) ∈ Iβ × Tβ × Λβ ,

(i, x, λ)LU (j, y, µ) ⇔ λ = µ,


 U (j, y, µ) ⇔ i = j,
(i, x, λ)R
 U (j, y, µ) ⇔ i = j and λ = µ.
(i, x, λ)H

4. Quasi −C−Ehresmann semigroups


As an analogue of regular orthogroups [17] i.e. quasi−C−semigroups [8], we call
an orthodox U −liberal semigroup S(U ) a quasi−C−Ehresmann semigroup if
378 Li, Guo, and Shum

U ⊆ E(S) is a regular band. From Theorem 3.2, we can see that the relation
H U is the same thing in lots of places in this paper (including here) as Q
U . Of
course, any U −liberal semigroup S(U ) with U a regular band is in general not
necessarily a quasi −C−Ehresmann semigroup since it may not satisfy the ET-
condition. The finite U −liberal semigroup described in the introduction of this
paper is such an example. However, there exist plenty quasi−C−Ehresmann
semigroups, for example, all rectangular monoids are quasi−C−Ehresmann
semigroups. Also, by Theorem 3.1, all U −liberal semigroups with U a left
zero bands [resp. right zero bands ] are left [resp. right] monoids and hence
they are quasi−C−Ehresmann semigroups.
On the other hand, by Theorem 3.2 and Lemma 2.2, we can easily see that
a quasi−C− Ehresmann semigroup S(U ) is a C−Ehresmann semigroup if U
forms a semilattice. Thus, quasi −C−Ehresmann semigroups can be regarded
as generalized C−Ehresmann semigroups. Moreover, by Theorem 3.2 and
Lemma 2.3, a quasi −C− Ehresmann semigroup S(U ) is a left C−Ehresmann
semigroup if U is a left regular band. Dually, a quasi−C−Ehresmann semi-
group S(U ) is a right C−Ehresmann semigroup if U is a right regular band.
Thus, quasi−C−Ehresmann semigroups can be regarded as generalized left
C−Ehresmann semigroups and generalized right C−Ehresmann semigroups.
In this Section, we will provide a construction for the quasi−C−
Ehresmann semigroup S(U ) . In view of Theorem 3.2, we can always write S(U )
as a semilattice of rectangular monoids, say, S(U ) = [Y ; Sα (Uα ) = Iα ×Tα ×Λα ] ,
where each Tα is a monoid  and each Uα = Iα × {1Tα } × Λα is a rectangular
band. Recall that U = Uα is a subsemigroup of S(U ) i.e. U = [Y ; Uα ] .
α∈Y
For the sake of brevity, we simply denote the element (i, 1Tα , j) in U by [i, j].
With the above notations, we now prove the following lemmas.

Lemma 4.1. (i) I = Iα forms a left regular band such that I =
α∈Y
[Y ; Iα ], under the multiplication “ ◦ ” defined by

i ◦ j = k ⇔ (∃λ ∈ Λα and ∃µ ∈ Λβ ) ([i, λ][j, µ])PIαβ = k, for all i ∈ Iα , j ∈ Iβ .



(ii) Λ = Λα is a right regular band such that Λ = [Y ; Λα ], under the
α∈Y
following multiplication “ ◦ ” :

λ ◦ µ = τ ⇔ (∃i ∈ Iα and ∃j ∈ Iβ ) ([i, λ][j, µ])PΛαβ = τ, for all λ ∈ Λα , µ ∈ Λβ .

(iii) U ∼
= I ×Y Λ .

Proof. (i) Since U is a regular band, R(U ) is a congruence on U . This


shows that the multiplication on I is well defined and so I ∼ = U/R(U ) .
(ii) The proof of this part is similar to the proof of (i).
(iii) This part is obvious.
Li, Guo, and Shum 379

Lemma 4.2. If (i, x, λ), (i , x, λ) ∈ Sα and (j, y, µ), (j  , y, µ) ∈ Sβ , then

((i, x, λ)(j, y, µ))PTαβ ×Λαβ = ((i , x, λ)(j  , y, µ))PTαβ ×Λαβ .

Proof. Assume that (i, x, λ), (i , x, λ) ∈ Sα and (j, y, µ), (j  , y, µ) ∈ Sβ . Let

(i, x, λ)(j  , 1Tβ , µ) = (k, w, ν).

Then, we can deduce that

((i , x, λ)(j  , y, µ))PTαβ ×Λαβ


= ((i , 1Tα , λ)(i, x, λ)(j  , 1Tβ , µ)(j, y, µ))PTαβ ×Λαβ
= ((i , 1Tα , λ)(k, 1Tαβ , ν)(i, x, λ)(j  , 1Tβ , µ)(j, y, µ))PTαβ ×Λαβ
= ((i , 1Tα , λ)(k, 1Tαβ , ν)(i, x, λ) · (i, 1Tα , λ)(j  , 1Tβ , µ) · (j, y, µ))PTαβ ×Λαβ
= ((i , 1Tα , λ)(k, 1Tαβ , ν)(i, x, λ)(i ◦ j  , 1Tαβ , λ ◦ µ)(j, y, µ))PTαβ ×Λαβ
= ((i , 1Tα , λ)(k, 1Tαβ , ν)(i, x, λ)(i ◦ j, 1Tαβ , λ ◦ µ)(j, y, µ))PTαβ ×Λαβ
( Since Iαβ is a left zero band )
= ((i , 1Tα , λ)(k, 1Tαβ , ν)(i, x, λ) · (i, 1Tα , λ)(j, 1Tβ , µ) · (j, y, µ))PTαβ ×Λαβ
= ((i , 1Tα , λ)(k, 1Tαβ , ν)(i, x, λ)(j, y, µ))PTαβ ×Λαβ
= ((i, x, λ)(j, y, µ))PTαβ ×Λαβ

as required.

Corollary 4.3. The set Sr (Ur ) = (Tα × Λα ) is a right C−Ehresmann
 α∈Y
semigroup, where Ur = ({1Tα } × Λα ), under the following multiplication:
α∈Y
for any (x, λ) ∈ Tα × Λα and (y, µ) ∈ Tβ × Λβ ,

(x, λ)(y, µ) = (z, ν) ⇔ (∃i ∈ Iα and j ∈ Iβ ) ((i, x, λ)(j, y, µ))PTαβ ×Λαβ = (z, ν).

Proof. By Lemma 4.2, the above multiplication is clearly well defined. It is


routine to verify that, under the given multiplication, Sr (Ur ) is a semilattice
of right monoids Tα × Λα (α ∈ Y ) and the set Ur is a right regular band.
Hence, by the dual result of Lemma 2.3, Sr (Ur ) is a right C−Ehresmann
semigroup.

Lemma 4.4. For any (i, x, λ), (i, x, λ ) ∈ Sα and (j, y, µ), (j, y, µ ) ∈ Sβ , we
have
((i, x, λ)(j, y, µ))PIαβ ×Tαβ = ((i, x, λ )(j, y, µ ))PIαβ ×Tαβ .

Proof. The proof is dual to that of Lemma 4.2.



Corollary 4.5. The set Sl (Ul ) = (Iα × Tα ) is a left C−Ehresmann
 α∈Y
semigroup, where Ul = (Iα × {1Tα }), under the following multiplication: for
α∈Y
380 Li, Guo, and Shum

any (i, x) ∈ Iα × Tα and (j, y) ∈ Iβ × Tβ ,

(i, x)(j, y) = (k, z) ⇔ (∃λ ∈ Λα and µ ∈ Λβ ) ((i, x, λ)(j, y, µ))PIαβ ×Tαβ = (k, z).

Proof. The proof is symmetric to Corollary 4.3, we hence omit the


details.

Lemma 4.6. For any (i, x, λ), (i , x, λ ) ∈ Sα and (j, y, µ), (j  , y, µ ) ∈ Sβ ,
we have
((i, x, λ)(j, y, µ))P Tαβ = ((i , x, λ )(j  , y, µ ))PTαβ .
Proof. Assume that (i, x, λ), (i , x, λ ) ∈ Sα and (j, y, µ), (j  , y, µ ) ∈ Sβ .
Then, by Lemma 4.2 and Lemma 4.4, we have

((i, x, λ)(j, y, µ))PTαβ = ((i , x, λ)(j  , y, µ))PTαβ = ((i , x, λ )(j  , y, µ ))PTαβ ,

as required.

Corollary 4.7. The set T (E) = Tα is a C−Ehresmann semigroup,
α∈Y
where E = {1Tα | α ∈ Y } , under the following multiplication: for any x ∈ Tα
and y ∈ Tβ ,

xy = z ⇔ (∃(i, λ) ∈ Uα and (j, µ) ∈ Uβ ) ((i, x, λ)(j, y, µ))PTαβ = z.

Proof. The result follows from Lemma 4.6.

We usually call Sl (Ul ) the left C−Ehresmann semigroup component of


S , Sr (Ur ) the right C−Ehresmann semigroup component of S and T (E) the
C−Ehresmann semigroup component of S .
We now prove the following characterization Theorem for quasi−C−
Ehresmann semigroups.
Theorem 4.8. Let S be a semigroup. Then S(U ) is a quasi−C−Ehresmann
semigroup for some U ⊆ E(S) if and only if S is isomorphic to a spined prod-
uct S1 ×T S2 of a left C−Ehresmann semigroup S1 = [Y ; Iα × Tα ] and a right
C−Ehresmann semigroup S2 = [Y ; Tα × Λα ] having a common C−Ehresmann
semigroup component T = [Y ; Tα ] with respect to the semigroup homomor-
phisms φ : (i, x) → x for (i, x) ∈ S1 and ψ : (x, λ) → x for (x, λ) ∈ S2 .
Proof. ⇒) Assume that, by Theorem 3.2, S(U ) = [Y ; Sα (Uα ) = Iα × Tα ×
Λα ] is a quasi−C−Ehresmann semigroup, where U = [Y ; Uα ] and Sα (Uα ) =
Iα × Tα × Λα (α ∈ Y ) is a rectangular monoid. Let Sl (Ul ) be the left C−
Ehresmann semigroup component of S , Sr (Ur ) the right C−Ehresmann semi-
group component of S and T the C−Ehresmann semigroup component of S .
Then, we can easily verify that

ξ : S(U ) → Sl (Ul ) ×T Sr (Ur ), (i, x, λ) → ((i, x), (x, λ))

is an isomorphism.
Li, Guo, and Shum 381

⇐) Assume, without loss of generality, that S is a spined product


S1 ×T S2 of a left C−Ehresmann semigroup S1 = [Y ; Iα × Tα ] and a right
C−Ehresmann semigroup S2 = [Y ; Tα × Λα ] having a common C−Ehresmann
semigroup component T = [Y ; Tα ] with respect to the semigroup homomor-
phisms φ : (i, x) → x for (i, x) ∈ S1 and ψ : (x, λ) → x for (x, λ) ∈ S2 .
Let
U = U1 ×Y U2
where U1 = [Y ; Iα × {1Tα }], U2 = [Y ; {1Tα } × Λα ] . Then S(U ) is isomorphic to
a semilattice of rectangular monoids Iα × Tα × Λα (α ∈ Y ) and U is a regular
band. Hence, by our Theorem 3.2, we see that S(U ) is a quasi−C−Ehresmann
semigroup.

Corollary 4.9. (i) If S(U ) is a quasi−C−Ehresmann semigroup, then

(∀u ∈ U )(∀x, y ∈ S(U )) uxyu = uxuyu.

(ii) If S(U ) is a quasi−C−Ehresmann semigroup and U is a left quasi-


normal band, then S(U ) is isomorphic to a spined product S1 ×Y Λ of a
left C−Ehresmann semigroup S1 = [Y ; Iα × Tα ] and a right normal band
Λ = [Y ; Λα ; φαβ ]. Thus, by Lemma 2.5, S(U ) is a weakly left C−Ehresmann
semigroup.

Proof. (i) The proof follows directly from Theorem 4.8, we hence omit the
details.
(ii) Assume that S(U ) = [Y ; Sα (Uα ) = Iα × Tα × Λα ] is a quasi−C−
Ehresmann semigroup and U = [Y ; Uα ] is a left quasinormal band. Since
U is a left quasinormal band, L is a congruence on U . Hence the set Λ =
Λα is a right normal band such that Λ = [Y ; Λα ; φα,β ] , under the following
α∈Y
multiplication “ ◦ ” :

λ ◦ µ = τ ⇔ (∃i ∈ Iα and ∃j ∈ Iβ ) ([i, λ][j, µ])PΛαβ = τ, for all λ ∈ Λα , µ ∈ Λβ ,

and, for any (i, x, λ), (i , x , µ) ∈ Sα and (j, y, µ), (j  , y  , µ) ∈ Sβ , we can deduce
that
((i, x, λ)(j, y, µ))PΛαβ = ((i , x , λ)(j  , y  , µ))PΛαβ = λ ◦ µ.
Hence
ξ : S(U ) → Sl (Ul ) ×Y Λ, (i, x, λ) → ((i, x), λ)
is an isomorphism.

We now give a method of construction for quasi−C−Ehresmann semi-


groups.

Theorem 4.10. Let T = [Y ; Tα ] be a C−Ehresmann semigroup, I =


[Y ; Iα ] a left regular band and Λ = [Y ; Λα ] a right regular band, let Tr (Λ) be the
382 Li, Guo, and Shum

right transformation semigroup acting on Λ and Tl (I) the left transformation


semigroup acting on I . If the following mappings

ξ: (Iα × Tα ) → Tl (I), (i, x) → (i, x) ,

α∈Y
η: (Tα × Λα ) → Tr (Λ), (x, λ) → (x, λ)∗
α∈Y

satisfy the following pairs of conditions:

(L1 ) if (i, x) ∈ Iα × Tα and j ∈ Iβ , then (i, x) j ∈ Iαβ ;


(R1 ) if (x, λ) ∈ Tα × Λα and µ ∈ Λβ , then µ(x, λ)∗ ∈ Λαβ ;
(L2 ) if α ≤ β in (L1 ) for α, β ∈ Y, then (i, x) j = i;
(R2 ) if α ≤ β in (R1 ) for α, β ∈ Y, then µ(x, λ)∗ = λ;
(L3 ) if (i, x) ∈ Iα × Tα and (j, y) ∈ Iβ × Tβ ,
then (i, x) (j, y) = ((i, x) j, xy) ;
(R3 ) if (x, λ) ∈ Tα × Λα and (y, µ) ∈ Tβ × Λβ ,
then (x, λ)∗ (y, µ)∗ = (xy, λ(y, µ)∗ )∗ .

Then S(U ) = (Iα × Tα × Λα ) forms a quasi−C−Ehresmann semigroup,
 α∈Y
where U = (Iα × {1Tα } × Λα ) , under the following multiplication:
α∈Y

(i, x, λ)(j, y, µ) = ((i, x) j, xy, λ(y, µ)∗ ).

Conversely, every quasi−C−Ehresmann semigroup can be constructed in


the above manner.

Proof. The proof follow directly from Theorem 4.8, Lemma 2.3 and its dual.
We omit the details.

Following the notation of Guo , Shum and Zhu in [8], the semigroup
constructed in the first part of Theorem 4.10 is called the semi-spined product
of T, I and Λ with respect to Y, ξ and η .
By summarizing the above results, we now provide a handy construction
theorem for quasi−C−Ehresmann semigroups.

Theorem 4.11. Let T = [Y ; Tα ] be a C−Ehresmann semigroup and let Iα


and Λα (α ∈ Y ) be two non-empty sets such that Iα ∩ Iβ =Λα ∩ Λβ = ∅(α =
β).Denote Pα = Iα × Tα , Qα = Tα × Λα (α ∈ Y ) and S = (Iα × Tα × Λα ).
 α∈Y
Form the set U = (Iα × {1Tα } × Λα ). Now, for any α, γ ∈ Y with γ ≤ α ,
α∈Y
define the mappings

ψα,γ : Pα → Tl (Iγ ), (i, x) → ψα,γ


(i,x)

and
φα,γ : Qα → Tr (Λγ ), (x, λ) → φ(x,λ)
α,γ
Li, Guo, and Shum 383

satisfying the following pairs of conditions:


(i,x)
(L1 ) if (i, x) ∈ Pα and j ∈ Iα , then ψα,α (j) = i;
(x,λ)
(R1 ) if (x, λ) ∈ Qα and µ ∈ Λα , then µφα,α = λ;
(i,x) (j,y)
(L2 ) if (i, x) ∈ Pα and (j, y) ∈ Pβ , then ψα,αβ ψβ,αβ is the constant
(i,x) (j,y)
mapping on Iαβ , whose value is denoted by ψα,αβ ψβ,αβ ;
(x,λ) (y,µ)
(R2 ) if (x, λ) ∈ Qα and (y, µ) ∈ Qβ , then φα,αβ φβ,αβ is the constant
(x,λ) (y,µ)
mapping on Λαβ , whose value is denoted by φα,αβ φβ,αβ ;
(k,xy) (i,x) (j,y)
(L3 ) if δ ≤ αβ in (L2 ) for α, β, δ ∈ Y, then ψαβ,δ = ψα,δ ψβ,δ
(i,x) (j,y)
for k = ψα,αβ ψβ,αβ ;
(xy,ν) (x,λ) (y,µ)
(R3 ) if δ ≤ αβ in (R2 ) for α, β, δ ∈ Y, then φαβ,δ = φα,δ φβ,δ
(x,λ) (y,µ)
for ν = φα,αβ φβ,αβ .

Then S(U ) forms a quasi−C−Ehresmann semigroup, under the following mul-


tiplication:
(i,x) (j,y) (x,λ) (y,µ)
(i, x, λ)(j, y, µ) = (ψα,αβ ψβ,αβ , xy, φα,αβ φβ,αβ ).

Conversely, every quasi−C−Ehresmann semigroup can be constructed in


the above manner.

Proof. This theorem is a consequence of Theorem 4.8, Lemma 2.4 and its
dual. We omit the details.

By using the notations of Guo , Shum and Zhu in [8],we call the semigroup
constructed in the first part of Theorem 4.11 the "−product of T, I and Λ with
respect to the structure mappings ψα,γ and φα,γ .

5. Some applications
Let S be a semigroup. The equivalences Q E(S) , LE(S) , R
 E(S) and H E(S) ,
respectively, are simply denoted by Q,  L,
R  and H  , respectively. In fact, the
R
notation of L,  and H  were first mentioned by El-Qallali in [1]. If S(E(S)) is a

E(S)−liberal semigroup, that is, every Q−class of S contains an idempotent of
S , then S is called a liberal semigroup and, in this case, a◦E(S) is simply denoted
by a◦ . Similarly, if S(E(S)) is an orthodox E(S)−liberal semigroup, then we
call S an orthodox liberal semigroup. If S(E(S)) is an Ehresmann semigroup,
then S is called a full Ehresmann semigroup. Full Ehresmann semigroups with
central idempotents are called C−full Ehresmann semigroups.
We call S a semi −superabundant semigroup, a left C−full Ehresmann
semigroup and a quasi−C−full Ehresmann semigroup, respectively, if S(E(S))
is respectively a E(S)−semi −superabundant semigroup, a left C−Ehresmann
semigroup and a quasi−C− Ehresmann semigroup. In the above cases,we
denote, for any a ∈ S , aE(S) by a , if it exists.
384 Li, Guo, and Shum

Recall that a unipotent semigroup S is a monoid containing an unique


idempotent of S . The direct product S = I × T × Λ of a rectangular band I × Λ
and a unipotent semigroup T is called a rectangular unipotent semigroup. In
particular, the direct product S = I × T of a left zero band I and a unipotent
semigroup T is called a left unipotent semigroup. Right unipotent semigroups
can be dually defined.
By using Lemma 2.2 and Lemma 2.3, we see that S is a C−full Ehres-
mann semigroup if and only if it is a semilattice of unipotent semigroups and
that S is a left C−full Ehresmann semigroup if and only if S is a semilattice
Y of left unipotent semigroups.
Hereafter, we call a semigroup T = [Y ; Tα ] a C−full Ehresmann semi-
group if T itself is a C−full Ehresmann semigroup which is expressible as a
semilattice of unipotent semigroups Tα (α ∈ Y ) , and we call the semigroup
S = [Y ; Iα × Tα ] a left C−full Ehresmann semigroup if S is a left C−full
Ehresmann semigroup which is expressible as a semilattice of the left unipotent
semigroups Sα = Iα × Tα (α ∈ Y ) .
By using the arguments in Theorem 3.2 and Theorem 4.8, we obtain
the following characterization theorems for orthodox liberal semigroups and
quasi−C−full Ehresmann semigroups.

Theorem 5.1. The following statements are equivalent for a semigroup S :


(i) S is an orthodox liberal semigroup;
(ii) S is a semilattice of rectangular unipotent semigroups Sα (α ∈ Y )
and E(S) is a subsemigroup of S ;
(iii) S is a semi−superabundant semigroup satisfying the (C) condition
and E(S) is a subsemigroup of S .

Theorem 5.2. Let S be a semigroup. Then S is a quasi−C−full Ehres-


mann semigroup if and only if S is isomorphic to a spined product S1 ×T S2
of a left C−full Ehresmann semigroup S1 = [Y ; Iα × Tα ] and a right C−full
Ehresmann semigroup S2 = [Y ; Tα × Λα ] having a common C−full Ehresmann
semigroup component T = [Y ; Tα ] with respect to the semigroup homomor-
phisms φ : (i, x) → x for (i, x) ∈ S1 and ψ : (x, λ) → x for (x, λ) ∈ S2 .

Fountain adopted Green’s star equivalences L∗ , R∗ , H∗ , D∗ , and J ∗ to


study abundant semigroups in [3]. Now, we call a semigroup S superabundant
if every H∗ −class of S contains at least one idempotent of S .

Lemma 5.3.  = Q,
Let S be a superabundant semigroup. Then H∗ = H 
and, for any a ∈ S , a◦ = a .

Proof. This result follows by observing that H∗  ⊆ Q


⊆ H  and
Lemma 2.1.

Definition 5.1. A superabundant semigroup S is called an orthodox super-


abundant semigroup if E(S) forms a subsemigroup of S . In particular, a su-
Li, Guo, and Shum 385

perabundant semigroup with central idempotents is called a C−superabundant


semigroup. We call an orthodox superabundant semigroup S left C−
superabundant, right C−superabundant and quasi−C− superabundant, re-
spectively, if E(S) forms a left regular band, a right regular band and a regular
band, respectively.

Lemma 5.4. Let S be an orthodox superabundant semigroup. Then S is


an orthodox liberal semigroup which can be expressed as a semilattice of direct
products of cancellative monoids and rectangular bands. In particular, if S is
a left (resp., right) C−superabundant semigroup, then S is a left (resp., right)
C−full Ehresmann semigroup which can be expressed as a semilattice of direct
products of cancellative monoids and left (resp., right) zero bands.

Proof. Assume that S is an orthodox superabundant semigroup. Then, by


definitions, we immediately have

 = Q.
H∗ = H 

This implies that S is a liberal semigroup such that

(∀a ∈ S) aH∗ a◦ .

By observing that R∗ is a left congruence on S and L∗ is right congruence on


S , we have
(ab)◦ R∗ abR∗ ab◦ R∗ (ab◦ )◦ L∗ ab◦ L∗ a◦ b◦ ,

and hence
(ab)◦ R∗ (ab◦ )◦ L∗ a◦ b◦ .

This leads to
(ab)◦ R(ab◦ )◦ La◦ b◦ ,

so that S(E(S)) satisfies the ET-condition. Thus, S is indeed an orthodox


liberal semigroup and so, by Theorem 5.1, S is a semilattice of rectangular
unipotent semigroups Sα = Iα × Tα × Λα (α ∈ Y ) and E(S) is a subsemigroup
of S . Now, by Theorem 3.2, we have

(∀(i, x, λ) ∈ Iα × Tα × Λα ) (i, x, λ) H∗ (i, 1Tα , λ).

This proves that Tα (α ∈ Y ) is a cancellative monoid.

Hereafter, we call a semigroup S = [Y ; Iα × Tα ] (resp., S = [Y ; Tα ×


Λα ] ) a left (resp., right) C− superabundant semigroup if S is a left (resp.,
right) C−superabundant semigroup which can be expressed as a semilattice
of direct products of left zero bands and cancellative monoids Iα × Tα (α ∈
Y ) (resp., cancellative monoids and right zero bands Tα × Λα ). Then, by
Fountain [2], we can easily see that a semigroup S is a C−superabundant
386 Li, Guo, and Shum

semigroup if and only if it is a semilattice of cancellative monoids. Thus, a


C−superabundant semigroup S = [Y ; Tα ] is a C−superabundant semigroup
which can be expressed as a semilattice of cancellative monoids Tα (α ∈ Y ) .
The following theorem is a characterization theorem for quasi−C−
superabundant semigroups. In fact, it also describes the method of construction
for such semigroups.

Theorem 5.5. Let S be a semigroup. Then S is a quasi−C−superabundant


semigroup if and only if S is isomorphic to a spined product S1 ×T S2 of a left
C−superabundant semigroup S1 = [Y ; Iα × Tα ] and a right C−superabundant
semigroup S2 = [Y ; Tα × Λα ] having a common C−superabundant semigroup
component T = [Y ; Tα ] with respect to the semigroup homomorphisms φ :
(i, x) → x for (i, x) ∈ S1 and ψ : (x, λ) → x for (x, λ) ∈ S2 .

Proof. ⇐) Assume that S is a spined product S1 ×T S2 of a left C−


superabundant semigroup S1 = [Y ; Iα × Tα ] and a right C−superabundant
semigroup S2 = [Y ; Tα × Λα ] having a common C−superabundant semigroup
component T = [Y ; Tα ] with respect to the semigroup homomorphisms φ :
(i, x) → x for (i, x) ∈ S1 and ψ : (x, λ) → x for (x, λ) ∈ S2 . Then, by
Theorem 5.2, S is a quasi−C−full Ehresmann semigroup. It can be easily
verified that, for any (i, x) ∈ Iα × Tα , (x, λ) ∈ Tα × Λα ,

(i, x) = (i, 1Tα ) [ resp., (x, λ) = (1Tα , λ)],

holds in S1 [resp., S2 ]. Since H∗ = H  holds in S1 and S2 , for any a =


((j, y), (y, µ)) and b = ((k, z), (z, ν)) ∈ S = S1 ×T S2 ,we have

((i, x), (x, λ))a = ((i, x), (x, λ))b


⇔ (i, x)(j, y) = (i, x)(k, z), (x, λ)(y, µ) = (x, λ)(z, ν)
⇔ (i, 1Tα )(j, y) = (i, 1Tα )(k, z), (1Tα , λ)(y, µ) = (1Tα , λ)(z, ν)
⇔ ((i, 1Tα ), (1Tα , λ))a = ((i, 1Tα ), (1Tα , λ))b

and

((i, x), (x, λ)) = ((i, x), (x, λ))b


⇔ ((i, 1Tα ), (1Tα , λ)) = ((i, 1Tα ), (1Tα , λ))b.

This implies that for any ((i, x), (x, λ)) ∈ S = S1 ×T S2 , we have

((i, x), (x, λ)) L∗ ((i, 1Tα ), (1Tα , λ)),

Similarly, we have

((i, x), (x, λ)) R∗ ((i, 1Tα ), (1Tα , λ)).

Consequently, S is a quasi−C−superabundant semigroup, as required.


Li, Guo, and Shum 387

⇒) Assume that S is a quasi−C−superabundant semigroup. By


Lemma 5.3 and Lemma 5.4, we see that S is, of course, a quasi−C−full Ehres-
mann semigroup satisfying
H∗ = H  = Q.


Let S = [Y ; Sα = Iα × Tα × Λα ] , where Iα × Tα × Λα (α ∈ Y ) is a direct product


of the cancellative monoid Tα and the rectangular band Iα × Λα . Then, by
Theorem 3.2, we have

 = H∗ (j, y, µ) ⇔ i = j, λ = µ.
(∀(i, x, λ), (j, y, µ) ∈ S) (i, x, λ) H

Thus, for any (i, x, λ) ∈ Sα and (j, y, µ) ∈ Sβ , we have

(i, x, λ) L∗ (j, y, µ) ⇔ (i, 1Tα , λ) L (j, 1Tβ , µ) ⇔ α = β, λ = µ.


(i, x, λ) R∗ (j, y, µ) ⇔ (i, 1Tα , λ) R (j, 1Tβ , µ) ⇔ α = β, i = j.
(i, x, λ) H∗ (j, y, µ) ⇔ (i, 1Tα , λ) H (j, 1Tβ , µ) ⇔ α = β, i = j, λ = µ.

Now, let Sr , Sl and T be the semigroups constructed in Corollary 4.3, Corol-


lary 4.5 and Corollary 4.7, respectively. Then, by Theorem 5.2, we can let
S be a spined product Sl ×T Sr of the left C−full Ehresmann semigroup
Sl = [Y ; Iα ×Tα ] and the right C−full Ehresmann semigroup Sr = [Y ; Tα ×Λα ]
having the common C−full Ehresmann semigroup component T = [Y ; Tα ] with
respect to the semigroup homomorphisms φ : (i, x) → x for (i, x) ∈ Sl and
ψ : (x, λ) → x for (x, λ) ∈ Sr . Since every Tα (α ∈ Y ) is a cancellative monoid,
T = [Y ; Tα ] forms a C−superabundant semigroup.
Let (i, x) ∈ Iα × Tα , a = (j, y) ∈ Iβ × Tβ and b = (k, z) ∈ Iγ × Tγ . Then
we consider

(i, x, λ)(j, y, µ) = (i , x , λ ), for λ ∈ Λα , µ ∈ Λβ and ν ∈ Λγ .

Observe that if (i, x)a = (i, x)b then we have

((i, x, λ)(j, y, µ))PIαβ ×Tαβ = (i, x)a = (i, x)b = ((i, x, λ)(k, z, ν))PIαβ ×Tαβ .

Thus, we obtain that

(i, x, λ)(j, y, µ) = (i, x, λ)(k, z, ν)(i , 1Tαβ , λ ),

and consequently, we derive that

(i, 1Tα , λ)(j, y, µ) = (i, 1Tα , λ)(k, z, ν)(i , 1Tαβ , λ ).

Since

((i, 1Tα , λ)(k, z, ν))PIαβ ×Tαβ = ((i, 1Tα , λ)(k, z, ν)(i , 1Tαβ , λ ))PIαβ ×Tαβ
= ((i, 1Tα , λ)(k, z, ν))PIαβ ×Tαβ ,
388 Li, Guo, and Shum

we further infer that


(i, 1Tα )a = (i, 1Tα )b.

Similarly, we obtain that

(i, x) = (i, x)b ⇒ (i, 1Tα ) = (i, 1Tα )b.

By using the fact that (i, x)(i, 1Tα ) = (i, x) , we have

(i, x) L∗ (i, 1Tα ).

If a(i, x) = b(i, x) , then a(i, 1Tα )(i, x) = b(i, 1Tα )(i, x) . If we denote

a(i, 1Tα ) = (i , x ) and b(i, 1Tα ) = (i , x ),

then, for any δ ∈ Λαβ = Λαγ , we finally prove that

a(i, 1Tα )(i, x) = (i , x )(i, x) = ((i , x , δ)(i, x, λ))PIαβ ×Tαβ


= (i , x x) = (i , x (x1Tαβ )),
b(i, 1Tα )(i, x) = (i , x )(i, x) = ((i , x , δ)(i, x, λ))PIαγ ×Tαγ
= (i , x x) = (i , x (x1Tαγ )).

This implies that i = i and x = x , that is,

a(i, 1Tα ) = b(i, 1Tα ).

Similarly, we have

(i, x) = b(i, x) ⇒ (i, 1Tα ) = b(i, 1Tα ).

Hence, we conclude that


(i, x) R∗ (i, 1Tα ).

Summing up the above results, we see that Sl is indeed a left C−superabundant


semigroup. Similarly, Sr is a right C−superabundant semigroup. Thus, the
proof is completed.

Remark 5.6. Quasi−C -superabundant semigroups can be analogously con-


structed by using the arguments in Theorem 4.10 and Theorem 4.11.

It is well known that a regular orthogroup S is a completely regular semi-


group whose set of idempotents E(S) is a regular band. Guo, Shum and Zhu
in [8] called such regular orthogroups quasi−C−semigroups. By observing that
regular orthogroups are in fact quasi−C−superabundant regular semigroups
and that cancellative regular monoids are groups, we can immediately re-obtain
the following known results from our Theorem 5.5.
Li, Guo, and Shum 389

Corollary 5.7. [8] The following statements are equivalent for a semigroup
S:
(i) S is a quasi−C−semigroup;
(ii) S is isomorphic to a spined product S1 ×T S2 of a left C−semigroup
S1 = [Y ; Iα ×Tα ] and a right C−semigroup S2 = [Y ; Tα ×Λα ] having a common
Clifford semigroup component T = [Y ; Tα ] with respect to the semigroup homo-
morphisms φ : (i, x) → x for (i, x) ∈ S1 and ψ : (x, λ) → x for (x, λ) ∈ S2 .

Corollary 5.8. [8] A semigroup S is a quasi−C−semigroup if and only if it


can be expressed as a semi-spined product of a Clifford semigroup, a left regular
band and a right regular band.

Corollary 5.9. [8] A semigroup S is a quasi−C−semigroup if and only if


it can be expressed as a "−product of a Clifford semigroup, a left regular band
and a right regular band.

References
[1] El−Qallali, A., “Structure Theory for Abundant and Related Semigroups”,
Ph.D. Thesis, York University, UK, 1980.

[2] Fountain, J. B., Rpp monoids with central idempotents, Semigroup Forum
13 (1977), 229–237.

[3] Fountain, J. B., Abundant semigroups, Proc. London Math. Soc. 44(3)
(1982), 103–129.

[4] Fountain, J. B., G. M. S. Gomes and V. Gould, A Munn type representation


for a class of E−semiadequate semigroups, J. Algebra 218 (1999), 693–
714.

[5] Gomes, G. M. S. and V. Gould, Fundamental Ehresmann semigroups,


Semigroup Forum 63 (2001), 11–33.

[6] Guo Y.Q., K.P. Shum and P.Y. Zhu, The structure of left C−rpp semi-
groups, Semigroup Forum 50(1) (1995), 9–23.

[7] Guo, Y. Q., On weak left C−semigroups, Chinese Science Bulletin 41(6),
(1996), 462–467.

[8] Guo Y.Q., K.P. Shum and P.Y. Zhu, On quasi−C−semigroups and some
special subclasses, Algebra Colloquium 6 (1999), 105–120.

[9] Heyong, “Some Studies on Regular Semigroups and Generalized Regular


Semigroups”, Ph.D. Thesis, Zhongshan University, China, 2002.

[10] Howie J.M., “Fundamentals of Semigroup Theory”, New York, Oxford


University Press, 1992.
390 Li, Guo, and Shum

[11] Ligang, On weakly left C−Ehresmann semigroups, (submitted for publica-


tion).
[12] Lawson, M.V., Rees matrix semigroups, Proc. Edinburgh Math. Soc. 33
(1990), 23–37.

[13] McAlister, D. B., One to one partial right translations of a right cancellative
semigroup, J. Algebra 43 (1976), 231–251.

[14] Petrich, M., The structure of completely regular semigroups, Trans. Amer.
Math. Soc. 189 (1974), 211–236.

[15] Petrich, M., “Lectures in Semigroups”, Pitman, New York, 1976.

[16] Petrich, M., “Inverse Semigroups”, New York, John Wiley & Sons, 1983.
[17] Petrich, M. and N. R. Reilly, “Completely Regular Semigroups”, Wiley &
Sons, Intersience Publication (1999).

[18] Tang X., On a theorem of C − wrpp semigroups, Comm. in Algebra 25


(1997), 1499–1504.

[19] Zhu P.Y., Y.Q. Guo and K.P. Shum, Struture and chacterizations of left C
semigroups, Science in China, Series A 6 (1992), 791–805.

Department of Mathematics Department of Mathematics


Shandong Normal University Southwest China Normal University
Jinan 250014, P. R. China Chongqing 400715, China
ligng@sohu.com yqguo259@swnu.edu.cn

Faculty of Science
The Chinese University of Hong Kong
Hong Kong, China (SAR)
kpshum@math.cuhk.edu.hk

Received June 8, 2004


Online publication June 2, 2005

You might also like