Chapter 2
Biordered Sets and Bimorphisms

Definition 2.1. Let E be a nonempty set and let ωr, ωl denote quasi-orders on E such that

ωr (ωl)1 = (ωr)1 ωl = I E. (2.1)

Define

r = ωr (ωr)1,l = ωl (ωl)1, (2.2)

and

ω = ωr ωl.

Let

τ = {τr(e)|e E}{τl(e)|e E} (2.3)

denote a family of partial transformation of E such that for each e E, τr(e) and τl(e) are idempotent mappings of ωr(e) and ωl(e) respectively onto ω(e). Then (E,ωr,ωl,τ) is a biordered set if ωr, ωl and τ satisfy the following axioms.

  1. fωre fτr(e)rf
  2. gωrfωre gτr(f) = (gτr(e))τr(f).
  3. f,g ωr(e) andgωlf imply gτr(e)ωlfτr(e) and (gτl(f))τr(e) = (gτr(e))τl(fτr(e)).

  4. Let
    M(e,f) = ({(gτl(e),gτr(f))|g ωl(e) ωr(f)},ωr × ωl). (2.4)

    Then there exists a subset S(e,f) of ωl(e) ωr(f) such that for every g S(e,f), (gτl(e),gτr(f)) is universally maximal in M(e,f). Further f,g ωr(e) imply

    S(f,g)τr(e) = S(fτr(e),gτr(e)). (2.5)
  5. Let g1, g2 ωl(e) ωr(f). Then if (g2τl(e),g2τr(f))ωr × ωl(g1,τl(e),g2τr(f)) there exist g2 and g2 in E satisfying the following conditions:
    1. g2lg2rg2,g2ωrg1,g2ωlg1.
    2. g2τl(e) = g2τl(e),g2τr(f) = g2τr(f)
    3. τl(g2,g2) τr(g1) = τr(g2,g2) τl(g1),

    where τl(g2,g2) = τl(g2)|ω(g2) and τr(g2,g2) = τr(g2)|ω(g2).

In a biordered set (E,ωr,ωl,τ), we call the quasi-order ωr(ωl) the right (left) quasi-order, the partial transformations τr(e)(τl(e)), e E, the partial right (left) translation (or r(l)-translation) corresponding to e and the set S(e,f), the sandwich set for e and f.

When there is no confusion we denote the biordered set (E,ωr,ωl,τ) by E. Also if E and E are two biordered sets, for convenience we denote quasi-orders, translations etc. by the same symbols if the context makes the meaning clear. Thus eωrf means e and f are ωr-related in E if e,f E and they are so related (under this order) in E if e,f E. We adopt the same convention to translations, sandwich sets etc.

As observed earlier, axioms for biordered sets include (Bi),i = 1,2,3,4,5 as well as their duals. The dual (Bi) of (Bi) can be obtained by interchanging r and l in (Bi). It may be noted that axiom (B5) is essentially self dual.

By (2.1), ω is a partial order so that

e,f E,ω(e) = ω(f) e = f;

and consequently

τr(e) = τr(f) e = f.

Equality of two partial transformations α and β of E is to be understood in the sense of equality of the associated subsets of E × E. Since this idea is used frequently in this work, we state the necessary and sufficient condition for this equality explicitly as follows.

α = β  dom α =  dom β, and for all e  dom α,eα = eβ. (2.6)

Remark 2.1: The structure of a biordered set depends not only its quasi-orders, but also on the set of partial translations. For, it is possible to have two different biordered sets with the same underlying set and the same left and right quasi-orders. This is shown by the following example.

Example 2.1: Let E = {e,f1,f2,f3). Define ωr = (E ×{e}) (E)2 and ωl = {(f2,e),(f3,e)} IE where E = {f1,f2,f3}. Then ωr and ωl are quasi-orders on E such that ωl ωr and

ωr (ωl)1 = (ωr)1 ωl = I E,

where E = {f1,f2,f3}. Then ωr and ωl are quasi-orders on E such that ωl ωr and

ωr (ωl)1 = (ωr)1 ωl = I E.

It is clear that ωr(e) = E,ωr(fi) = E(i = 1,2,3), ωl(e) = ω(e) = {e,f2,f3} and ωl(fi) = ω(fi) = {fi}(i = 1,2,3). Let τr(e) denote the idempotent mappings of E onto ω(e) such that f,τr(e) = f2. Define τr(fi) by fjτr(fi) = fi for i,j = 1,2,3. Further let τl(x) denote identify map on ωl(x) = ω(x). If

τ = {τr(x)|x E}{τl(x)|x E}

then (E,ωr,ωl,τ) is a biordered set. Further if τ1r(e) denotes another idempotent mapping of E onto ω(e) such that f,τr(e) = f3 and if

τ1 = {τr(x)|x E}{τ 1r(e)}{τl(x)|x E}

then (E,ωr,ωl,τ1) is also a biordered set. These two differ only with respect to translations.

Remark 2.2: Let (E,ω) be a semilattice. If

I = {Iω(e)|e E},

(where each Iω(e) denotes identity map on the ω-ideal ω(e) of E), then (E,ω,ω,I) is a biordered set. Since for each e E, τr(e) and τl(e) are identity maps, axioms (B1),(B2),(B3) and their duals are obviously satisfied. Axiom (B4) (and (B4)) is satsified with S(e,f) = {ef} where ef denotes the product of e and f in (E,ω) and (B5) is satsified with g2 = g2 = g2.

On the other hand if a biordered set E satsifies the condition that

ωr = ωl, (2.7)

then τ must coincide with I defined above. For e,f E, h S(e,f), implies that

S(e,f) = S(f,e) = {h}

and h is the greatest member of ω(e) ω(f). Thus (E,ω) is a semilattice.

We say that the biordered set satisfying the condition (2.7) is a semilattice. In the following we do not distinguish between the biordered set (E,ω,ω,I) and the semilattice (E,ω).

It may be noted that in a biordered set (E,ωr,ωl,τ) one or more of the quasi-ordered sets (E,ωr), (E,ωl) or (E,ω) (but not all) may be semilattices without the biordered set satisfying condition (2.7). This is shown by an example given below. Clearly if (E,ωr)((E,ωl)) is a semilattice, ωr = ω(ωl = ω).

Example 2.2: Let (E,ωr,ωl,τ) be a biordered set and let 1 denote a symbol not representing any element of E. Let E1 = E {1}. Extend ωr and ωl to E1 by defining

eω1

for every e E, and τ by defining

τr(1) = τl(1) = I E.

With these extended quasi-orders and translations, E1 becomes a biordered set.

We say that e E is the identity (the zero) of the biordered set E if and only if for every f E

fωe(eωf).

Let E1 stands for E if E has identity, and it denotes the biordered set constructed above if E has no identity. Similarly E0 stands for E if E has a zero. If not, it denotes the biordered set obtained by adjoining 0 to E, where 0ωe, 0τr(e) = 0τl(e) = 0 for every e E {0}.

Example 2.3: Let (E,) be a semilattice and X,Y denote arbitrary non-null sets. Let E = X × E × Y and define ωr and ωl in E as follows.

(x,e,y)ωr(x,e,y) e e,y = y; (x,e,y)ωl(x,e,y) x = x,e e.

Then ωr and ωl are quasi-orders such that

ωr (ωl)1 = (ωr)1 ωl = I E.

Also

ωr(x,e,y) = {(x,e,y)|x X,e e} ωl(x,e,y) = {(x,e,y)|e e,y Y }

and

ω(x,e,y) = {(x,e,y)|e e}.

Define τr(x,e,y) and τl(x,e,y) by

(x,e,y)τr(x,e,y) = (x,e,y)

and

(x,e,y)τl(x,e,y) = (x,e,y).

Then it is clear that τr(x,e,y) and τl(x,e,y) are idempotent mappings of ωr(x,e,y) and ωl(x,e,y) respectively onto ω(x,e,y). Let

τ = {τr(x,e,y)|(x,e,y) E}{τl(x,e,y)|(x,e,y) E}.

Then (E,ωr,ωl,τ) is biordered set.

Axioms (B1),(B2),(B3) and their duals are easy to verify. Now

ωl(x,e,y) ωr(x 1,e1,y1) = ω(x,ee1,y).

Hence for any (x,g,y) ωl(x,e,y) ωr(x1,e1,y1) we have x = x,g ee1,y = y1,

(x,g,y1)τl(x,e,y) = (x,g,y)ωr(x,ee 1,y1)τl(x,e,y) (x,g,y1)τr(x 1,e1,y1) = (x1,g,y1) = ω(x,ee1,y1)τr(x 1,e1,y1)

and consequently

(x,ee1,y1) S((x,e,y),(x1,e1,y1)).

If (x,g,y1) S((x,e,y),(x1,e1,y1)), then from the fact that ((x,g,y),(x1,g,y1)) and ((x,ee1,y),(x1,ee1,y1)) are universally maximal in M((x,e,y),(x1,e1,y1)), we have g = ee1, and

S((x,e,y),(x1,e1,y1)) = {(x,ee1,y1)}.

If (x,e,y),(x1,e1,y1) ωr(x,e,y), then

y = y1 = y, S((x,e,y)τr(x,e,y),(x 1,e1,y1)τr(x,e,y)) = S((x,e,y),(x,e 1,y)) = {(x,ee 1,y)}

and

S((x,e,y),(x1,e1,y1))τr(x,e,y) = (x,ee 1,y1)τr(x,e,y) = {(x,ee 1,y)}.

From this axiom (B4) follows, Axiom (B4) may be verified similarly.

If (x,g1,y1),(x,g2,y1) ωl(x,e,y) ωr(x1,e1,y1) then hypothesis of (B5) implies that

(x,g2,y1)ω(x,g1,y1)

and hence axiom (B5) is satisfied. (In this case, g2,g2,g2 of axiom (B5) coincide.)

Note: In this example

S((x,e,y),(x1,e1,y1)) = {(x,ee1,y1)}

and

S((x1,e1,y1),(x,e,y)) = {(x1,ee1,y)}.

Hence unless x = x1 and y = y1

S((x,e,y),(x1,e1,y1))S((x1,e1,y1),(x,e,y)).

Thus in a biordered set, in general, S(e,f)S(f,e).

Example 2.4: Let E denote the biordered set constructed in 2.3. Then if both X and Y contain more than one element, ωrωlω and if X(Y ) contains only one element, ωr ωl(ωl ωr).

Let (E)0 denote the biordered set obtained by adjoining 0 to E. Then ((E)0,ω) is a semilattice.

When both X and Y contain more than one element, this gives an example of a biordered set E such that ωrωlω and (E,ω) is a semilattice.

Following are certain consequence of the definition of biordered sets. (E stands for a biordered set, e, f etc. are its elements.)

Lemma 2.3. erf fτr(e) = e.

Proof. If fτr(e) = e,erf by (B1). Conversely if erf, fτr(e) ω(e) and by (B1) fτr(e)rfre. Hence by (2.1)

(fτr(e),e) ω (r) ωl (ωr)1 = I E.

Lemma 2.4. elgrf S(e,f) = {g}.

Proof. Let h S(e,f). Then by (B4) and Lemma 2.3* we have

e = gτl(e)ωrhτl(e)ωe,

and thus, hlg. Dually we can show that hrg.

Lemma 2.5. eωrf e S(e,f) and eωlf eτl(f) S(e,f).

Proof. In the first case, if h S(e,f)

e = eτl(e)ωrhτl(e)ωe

so that elh. Since e,h ωr(f), by (B3), eτr(f)lhτr(f). Hence

((eτl(e))),(eτr(f))(r ×l),(hτl(e)),(hτr(f))

and therefore (eτl(e),eτr(f)) is universally maximal in M(e,f).

In the second case, we have by (B2) that (eτl(f))τl(e) = e and hence

(eτl(f))τl(e) = eωrhτl(e)ωe.

Thus (eτl(f))τl(e) = hτl(e) and by (B3), (eτl(f))τr(f)lhτr(f). It now follows that ((eτl(f))τl(e),(eτl(f))τr(f)) is universally maximal in M(e,f).

Lemma 2.6. Let ele and frf. For g ωl(e) ωr(f), define

(gτl(e),gτr(f))m(e,g) = (gτl(e),gτr(f)). (2.8)

Then m(e,f) is an order isomorphism of M(e,f) on to M(e,f). In particular

S(e,f) = S(e,f).

Proof. For g ωl(e) ωrf, by axioms (B2) and (B2),

(gτl(e),gτr(f))m(e,f) = ((gτl(e))τl(e),(gτr(f))τr(f)).

If

(gτl(e),gτr(f))m(e,f) = (gτl(e),gτr(f)),

then

m(e,f) m(e,f) = I M(e,f)

and

m(e,f) m(e,f) = I M(e,f).

Cosequently m(e,f) is a bijection of M(e,f) onto M(e,f) and

m(e,f) = (m(e,f))1.

Now let

(g1τl(e),g 1τr(f))(ωr × ωl)(g 2τl(e),g 2τr(f)).

Then g1τl(e)ωrg2τl(e) ωl(e) and hence by (B3)

g1τl(e) = (g 1τl(e))τl(e)ωr(g 2τl(e))τl(e) = g 2τl(e).

Dually we also have g1τr(f)ωlg2τr(f). Thus m(e,f) is order preserving and so also is m(e,f). Consequently m(e,f) is an order isomorphism.

The equality S(e,f) = S(e,f) follows from the fact that order isomorphisms preserve universally maximal elements.

Lemma 2.7. Let e,f,g1,g2,g2,g2 be as in (B5). Then we have the following:

  1. If g1 = g2τr(g1), then g1 = g2τl(g1).
  2. g2lg2rg1lg2rg2.
    1. g2τl(e) = g2τl(e)rg1τl(e) = g2τl(e),
    2. g2τr(f) = g2τr(f)lg1τr(f) = g2τr(f).

Furthermore,

  1. g1 = g1 g2τl(e)rg1τl(e) (or g2τr(f)lg1τr(f)).

In particular if h1,h2 S(e,f), there exist h2 and h2 in S(e,f) such that

  1. h2lh2rh1lh2rh2.

Proof. Statement (a) follows from (a) and (c) of (B5) and statement (b) from (a) of this lemma and (a) of (B5). From statement (b) of this lemma, (B1), (B3) and (B5) we have

g2τl(e)rg 1τl(e)lg 2τl(e),

and

g1τl(e)rg 2τl(e) = g 2τl(e)rg 2τl(e).

We now obtain from (b) of (B5), the statement (c)(i). The second relation in (c) can be proved dually.

(d) If g1 = g1, clearly, g2τl(e)rg1τl(e) and g2τr(f)lg1τr(f). Conversely assume that g2τl(e)rg1τl(e). The fact that g1τl(e)rg1τl(e) and g1 ω(g1) together give g1τl(e) = g1τl(e). In other words g1lg1 and consequently g1 = g1.

If we assume that g2τr(f)lg1τr(f), we arrive at the same conclusion.

(e) If h1,h2 S(e,f)

(h1τl(e),h 1τr(f))(r ×l)(h 2τl(e),h 2τr(f)).

Thus h1 and h2 satisfy the hypothesis of (B5), and therefore there exist h2 and h2 in E such that

h2lh2rh 1lh 2rh 2,

where h1 = h2τl(h1). By statement (c) of this lemma, we see that h2,h2S(e,f) and since by (d), h2τl(e)rh1τl(e),h1 = h; which proves (e).

Definition 2.2. A mapping θ : E E of a biordered set E into biordered set E is a bimorphism if it satisfies the following axioms and their duals.

  1. Let e1,e2 E and e1ωre2. Then e1θωre2θ and (e1τr(e2))θ = e1θτr(e2θ).
  2. e1,e2 E S(e1,e2)θ S(e1θ,e2θ).

A bimorphism θ of E in to E is a retraction (co-retraction) if it satisfies the following axiom.

  1. There exists a bimorphism θ : E E such that θ θ(θ θ) is identity on E(E).

θ is an isomorphism if and only if θ is both a retraction and a co-retraction. (The fact that θ is an isomorphism of E onto E may be expressed by writing θ : E E.)

Identity map IE of E is evidently an isomorphism. Also all retractions (co-retractions) are surjections (injections) and therefore isomorphisms are bijections. Axiom (m3) gives that if θ is a retraction (co-retraction), the bimorphism θ associated with θ is a co-retraction (retraction). If θ is a one-to-one retraction (onto co-retraction), we can see that θ θ = IE(θ θ = IE) so that θ = θ1 . Hence in this case both θ and θ1 are isomorphisms and clearly this holds for all isomorphisms.

It is not difficult to verify that if

θ : E E, andθ : E E

are bimorphism then

θ θ : E E

is also a bimorphism.

For, let e1,e2 E. Then

e1ωre 2 e1θωre 2θ by (m1) (e1θ)θωr(e 2θ)θ by (m 1) (e1)θ θωr(e 2)θ θ

and

(e1τr(e 2))θ θ = ((e 1τr(e 2))θ)θ = (e1θτr(e 2θ))θ by ( m 1) = (e1θ)θτr((e 2θ)θ) by ( m 1) = ((e1)θ θ)τr((e 2)θ θ).

Thus θ θ satisfy (m1) (m1) is verified dually.

Now let h S(e1,e2). Thus by (m2), hθ S(e1θ,e2θ) and (hθ)θ S((e1,θ)θ, (e2,θ)θ). That is (h)θ θS((e1)θ θ,(e2)θ θ). Thus θ θ satisfies (m2) also.

Definition 2.3. Let E be a biordered set and E E. Define

ω1r = ωr (E)2,ω 1l = ωl (E)2 and τ1 = τr(e)|E|e E{τl(e)|E|e E}. (2.9)

Then (E,ω1r,ω1l,τ1) is a biordered subset of E if the following conditions are satisfied.

  1. E = (E,ω1r,ω1l,τ1) is a bordered set.
  2. If e,f E and if S1(e,f) and S(e,f) denote sandwich sets of e and f in E and E respectively, then
    S1(e,f) S(e,f).

Clearly E is a biordered subset of itself. Axiom (BS1) implies in particular that for each e E, τr(e)|E and τl(e)|E are partial transformations of E so that

E ωr(e)τr(e) E (2.10)

and that S1(e,f)(e,f E). Clearly S1(e,f) E. Hence by (BS2), S1(e,f) S(e,f) E. If M1(e,f) and M(e,f) denote the sets defined by (2.4) for E and E respectively, then by (2.10), M1(e,f) M(e,f). If h S(e,f) E,hτl(e) and hτr(f) are members of E and therefore

(hτl(e),hτr(f)) M 1(e,f).

Since (hτl(e),hτr(f)) is universally maximal in M(e,f) it must be universally maximal in M1(e,f) also. Thus h S1(e,f) or

S1(e,f) = S(e,f) E. (2.11)

This, together with (2.5) (of axiom (B4)), shows that for all e1,e2 in E such that e1,e2 ωr(e) ,

S(e1,e2) Eτr(e) = S(e 1,e2)τr(e) E. (2.12)

Lemma 2.8. Let E E, where E is a biordered set. Let ω1r and ω1l be defined by (2.9) and let τ1 be a family of partial transformations of E such that E = (E,ω1r,ω1l,τ1) is a biordered set. Then E is a biordered subset of E if and only if the identity mapping IE of E into E is a bimorphism.

If E is biordered subset of E and if θ : E E1 is a bimorphism, then θ|E is bimorphism of E into E1. Further, if E is a biordered subset of E it is also a biordered subset of E.

Proof. Let E be a biordered subset of E. Then definitions of ω1r, ω1l, τ1 and axiom (BS1) together imply that the mapping IE of E into E satisfies axiom (m1) of Definition 2.2. Axiom (BS2) is the same as axiom (m2) with IE in plane of θ. Hence IE is a bimorphism.

On the other hand if E = (E,ω1r,ω1l,τ1) is a biordered set such that ω1r and ω1l are quasi-orders defined by (2.9) and IE is a bimorphism, then axiom (m1) shows that for every f E , e ω1r(f) and e ω1l(f)

eτ1r(f) = eτr(f) andeτ 1l(f) = eτl(f).

Hence τ1 is the same as the family defined by (2.9) and as a consequence E satisfies (BS1). Statement (BS2) follows from axiom (m2). Therefore E is a biordered subset of E.

The remaining statements of Lemma 2.8 are consequences of the fact that if θ1 : E1 E2 and θ2 : E2 E3 are bimorphism, then θ1 θ2 is a bimorphism of E1 into E3.

Lemma 2.9. Let E be a subset of a biordered set E. Then E is a biordered subset if and only if the following conditions and their duals are satisfied.

  1. For every e E
    E ωr(e)τr(e) E.
  2. e1,e2 ES(e1,e2) E
  3. e E and e1,e2 ωr(e) E imply that (S(e1,e2) E)τr(e) = S(e 1,e2)τr(e) E.

Proof. If E is biordered subset, (a) and (c) are the same as (2.10) and (2.12). (b) follows from (2.11).

Now suppose E satisfied (a), (b) and (c). Define ω1r,ω1l and τ1 by (2.9). Then by (a), τ1 is a family of partial transformations of E. The fact that (E,ω1r,ω1l,τ1) are satisfied axioms (B1), (B2), (B3) and their duals follows from the definitions of ω1r,ω1l and τ1. For e,f E, let M(e,f) and M1(e,f) denote quasi-ordered sets (Definition 2.4) of E and E respectively. Then M1(e,f) M(e,f) and by (b), at least one universally maximal member of M(e,f) belongs to M1(e,f). Since the quasi-order of M1(e,f) is the restriction of ωr × ωl to the subset M1(e,f) of M(e,f) the set of universally maximal members of M1(e,f) is non-empty and these members are universally maximal in M(e,f) also. Thus (hτ1l(e),hτ1r(f)) is universally maximal in M1(e,f) if and only if h S(e,f) E = S1(e,f). Now if e E and e1,e2 ω1r(e) then e1,e2 ωr(e) E and therefore

S1(e1,e2)τ1r(e) = S(e 1,e2) Eτr(e) = S(e1,e2)τr(e) E = S(e1τr(e),e 2τr(e)) E = S1(e1τ1r(e),e 2τ1r(e)).

Dual statement of (2.5) can be verified similarly. Hence E satisfies axioms (B4) and (B4). To assert that E satisfies (BS1), we have to show that E satisfies axioms (B5) also. Let g1,g2 ω1r(e) ω1r(f) where e,f E and

(g2τ1l(e),g 2τ1r(f))ω 1r × ω 1l(g 1τ1l(e),g 1τ1r(f)).

Then g1 and g2 satisfy this condition in E also and therefore by (B5) there exist g2 and g2 in E satisfying (a), (b) and (c) of (B5). Now it is enough to establish that g2,g2 E.

Since g1,g2 E, S1(g2,g1). Choose h S1(g2,g1) and suppose that h1 = hτr(g1) = hτr(g1). Then by (a), h1 E and as a consequence of Lemmass 2.5 and 2.6,

g2S(g 2,g 1) = S(g2,g1).

Hence h,g2S(g2,g1) and therefore h1lg2τr(g1) = g1. By Lemma 2.7(a) g1 = g2τl(g1) and we can prove similarly that there exists h2 E such that h2rg1. But since h1,h2 E, statement (b) implies that

S1(h1,h2) = S(h1,h2) E = {g 1} E,

and consequently g E. Thus g2,g1 E and S1(g2,g1) = S(g2,g1) E = {g2} E. This leads to the conclusion that g2 E. Dually we can show that g2 E. Thus we observe that (E,ω1r,ω1l,τ1) is a biordered set and that it satisfies (BS1). For e,f E, S1(e,f) = S(e,f) E is the sandwich set of e and f in E and therefore (BS2) is also satisfied.

Corollary 2.10. Let E be a biordered set. Then for every e E, ωr(e), ωl(e) and ω(e) are biordered subsets of E.

Proof. Given e E, for all g ωr(e), ωr(g) ωr(e). Further given g,g ωr(e), if gωlg, then gτl(g)ωgωre. Hence condition (a) of Lemma 2.9 is satisfied. If g1,g2 ωr(e) and h S(g1,g2) then hωrg2 and consequently h ωr(g2) ωr(e). Thus S(g1,g2) ωr(e). Hence ωr(e) satisfies (b) of Lemma 2.9 also. Statement (c) is obvious, since for g1,g2 ωr(g) ωr(e) where g ωr(e), S(g1,g2) ωr(g) ωr(e). This proves that ωr(e) is a biordered subset.

Similarly using Lemma 2.9 we can prove that ωl(e) and ω(e) are also biordered subsets of E.

Theorem 2.11. Let E be a biordered set. Then for e E, τr(e) is a bimorphism of ωr(e) onto ω(e) and if ere,

τr(e,e) = τr(e)|ω(e) (2.13)

is an isomorphism of ω(e) onto ω(e).

Proof. Let e E. Then axioms (B1) and (B2) show that τr(e) preserves ωr and ωl. Let e,ωre2ωre. Then e1ωre2τr(e)ωe and hence by (B2).

(e1τr(e))τr(e 2τr(e)) = e 1τr(e 2τr(e)) = (e1τr(e 2))τr(e 2τr(e)). (2.14)

Also e1τr(e)re1re1τr(e2)ωe2re2τr(e) and therefore (e1τr(e2))τr(e)ωe2τr(e). Hence by axiom (B2) we get

(e1τr(e 2))τr(e 2τr(e)) = ((e 1τr(e 2))τr(e))τr(e 2τr(e)) = (e1τr(e 2))τr(e). (2.15)

From (2.14) and (2.15) we get

(e1τr(e 2))τr(e) = (e 1τr(e))τr(e 2τr(e)),

and hence τr(e) satisfies axiom (m1) of Definition 2.2. Axiom (m1) follows from axiom (B3) and axiom (m2) follows from (2.5).

If ere, it can be similarly proved that τr(e,e) is a bimoprhism of ω(e) into ω(e) and τr(e,e) is a bimorphism of ω(e) into ω(e). But for e ω(e) by axiom (B2),

(e)τr(e,e) τr(e,e) = (eτr(e))τr(e) = e,

and similarly we can prove that

τr(e,e) τr(e,e) = I ω(e).

Hence by the definition of isomorphism, τr(e,e) is an isomorphism of ω(e) onto ω(e).

NOTE. clearly τr(e,e) is identity map Iω(e). Also we get by axiom (B2) that for all erere,

τr(e,e) τr(e,e) = τr(e,e). (2.16)

Dually τl(e,e) is also Iω(e) and the two symbols τr(e,e) and τl(e,e) may be used according to convenience.

Remark 2.12: We have already observed that (Remark 2.2) the concept of a biordered set is a generalisation of the concept of a semilattice. It can be verified without difficulty that a subset E of a semilattice E is a subsemilattice of E if and only if it is a biordered subset of E. Further a mapping θ of semilattice E1 into semilattice E2 is a homomorphism of E1 into E2 (i.e. a mapping such that (e1,e2)θ = e1θ e2θ for all e1,e2 E1) if and only if it is a bimorphism of E1 into E2.