Academia.eduAcademia.edu

Minimal Clans

2015

A minimal clan is a lattice-ordered partial semigroup having the cancellation property and difference property. Wyler (1966 [15]) proved every Abelian clan could be embedded into an Abelian lattice-ordered group. Conrad (1990) showed that every clan could be embedded into a lattice-ordered group. There are two main results in this paper. First theorem said a clan C is complete if and only if the lattice-ordered group generated by C is complete. The second theorem said every clan contains a greatest generalized Boolean algebra. That answered Schmit's question ([16]). The second part of this paper is to that if an l-group G has minimal generating clan such that {0} is the only subsemigroup then G is a Specker group. This theorem makes a connection between Specker groups and Clans.

International Journal of Engineering Technology, Management and Applied Sciences www.ijetmas.com November 2015, Volume 3, Issue 11, ISSN 2349-4476 Minimal Clans Shinemin Lin, Ph.D. Savannah State University ABSTRACT A minimal clan is a lattice-ordered partial semigroup having the cancellation property and difference property. Wyler (1966 [15]) proved every Abelian clan could be embedded into an Abelian latticeordered group. Conrad (1990) showed that every clan could be embedded into a lattice-ordered group. There are two main results in this paper. First theorem said a clan C is complete if and only if the latticeordered group generated by C is complete. The second theorem said every clan contains a greatest generalized Boolean algebra. That answered Schmit’s question ([16]). The second part of this paper is to that if an l-group G has minimal generating clan such that {0} is the only subsemigroup then G is a Specker group. This theorem makes a connection between Specker groups and Clans. Keywords Clan, Groupoid, l-group, l-ideal, Specker group, subsemigroup, Free generating Boolean Algebra, eequivalent. INTRODUCTION I gather some terms and fundamental results about l-groups. For additional background, refer to [1], [4], and [6]. I follow tradition and use additive notation for the group operation even though most of groups may not be abelian. A partially ordered group (po-group) is a group (G, +) that has a partially order, ≤, defined on it such that a ≤ b implies x + a ≤ x + b and a + x ≤ b + x for all a, b, x of G. If the order is also a lattice order or a total order, then G is called a lattice-ordered group (l-group) or a totally-ordered group (o-group), respectively. The symbols , and  are used to denote the greatest lower bound and the least upper bound. A subgroup A of G is an l-subgroup if A is also a sub-lattice of G. An l-subgroup A is convex, of 0 ≤ g ≤ q ∈ A implies g ∈ A. Suppose H is an l-subgroup of an l-group G, H is large in G or G is an essential extension of H if for each convex l-subgroup L≠ 0 of G, L ∩ H ≠ 0. An l-group G is Archimedean if 0 ≤ ng < h for all n ∈ N implies g = 0. For an Archimedean l-group G, if G admits no proper essential extension then G is essentially closed. An essentially closed essential extension of l-group G is called essential closure of G and is denoted by Ge. An l-group G is complete (laterally complete), if every bounded (disjoint) subset M of G+, ∨ M ∈ G. An l-group H is a completion (lateral completion) of G if H is complete (lateral complete), and G is l-isomorphically dense in H, and no proper l-subgroup of H containing an l-isomorphic copy of G is complete (lateral complete). The completion (laterally completion, divisible hull) of G will be denoted by G^ (GL, Gd). An element h ≤ g is called a component of g if (g – h) ∧ h = 0; an element g > 0 of G is called singular if for any 0 < x ≤ g then x is a component of g. An l-subgroup H of G is saturated if for all h of H, the components of h belong to H. For g ∈ G, g+ = g ∨ 0 (g- = (-g) ∨ 0) is called the positive (negative) part of g, and |g| = g+ + g- is called the absolute value of g. The principal convex l-subgroup of G that is generated by g is denoted by G(g), which is the subset G(g) = {x ∈ G| 0 ≤ |g|≤ n|g| for some 0 < n ∈ N}. A partial groupid is a set E with a relation S  E x E and a map + : S  E such that there exists an element 0 ϵ E satisfying (0,x) ϵ S, (x,0) ϵ S and 0 + x = x+ 0 for all x ϵ E. A partial semigroup is a partial groupoid (E, S, +) which is associative. 33 Shinemin Lin International Journal of Engineering Technology, Management and Applied Sciences www.ijetmas.com November 2015, Volume 3, Issue 11, ISSN 2349-4476 An ordered partial semigroup is a partial semigroup (E, S, +) with ordered relation ≤ such that µ + x + v ≤ v + x + µ holds for all x, y ϵ E satisfying x ≤ y and for all µ, v ϵ E satisfying (µ, x) ϵ S as well as (µ,y) ϵ S and (µ+y, v) ϵ S. An ordered partial semigroup has lattice property if (E, ≤) is a lattice. It has difference property if for all x, y ϵ E having a x  y, and a x  y, there exist µ and v ϵ E+ satisfying µ + x = x  y = x + v and µ + x  y = y = x  y + v. It has cancellation property if x = y holds for all x, y ϵ E satisfying µ + x + v = µ + y + v for some u, v ϵ E. A minimal clan (or a clan) is a lattice-ordered partial semigroup having the cancellation property and difference property. A minimal clan E is called positive if E = E+. An abelian clan C is Archimedean if a ϵ C is such that na ≤ b in [C], for all natural number n and a fixed element b of C, then a ≤ 0. Remarks: 1. (Wyler) Every abelian clan C can be embedded into an abelian I-group G (= [C]). Furthermore, C is Archimedean if and only if [C] is Archimedean. 2. (Conrad) Every clan C can be embedded into an l-group G (=[C]). SECTION 1 Definition 1.1 1. A clan C is called a generating clan of an I-group G, if C is a convex sublattice of G such that for every element g of G, g=c1 + c2 + … + cn, where c 1 ϵ C for all I ϵ {1, 2, 3, …, n}. We write G = [C]. 2. A subset D of C is called a subclan of C if D is a sublattice of C and for a, b ϵ D such that a ≤ b there exists c ϵ D such that a + c = b. If D is also convex then D is called a convex subclan of C. Lemma 1.2. Let C be a positive clan, and D be a subset of C then D’ = {x ϵ C| x  d = 0 for all d ϵ D} is a convex subclan. Proof. Clearly, D’ is a sublattice of C. If a ≤ b, a, b ϵ D’  C, then there is c ϵ C satisfying a + c = b. For all d ϵ D, 0 ≤ x  d ≤ (a + c)  d = b  d = 0; therefore, c  d = 0 which implies c ϵ D. Definition 1.3. D’ is called a polar subclan (or polar) of C. Lemma 1.4. Let R(C) be the set of all polars of C. Then R(C) is a complete Boolean algebra. Proof. The proof is similar to the proof of the theorem that the set of all polars of an I-group is a complete Boolean algebra ([4]). Proposition 1.5. Let C be a positive abelian clan. Then G = [C] is a free abelian I-group over C. Proof. C g [C]=G F F(C) f f* [f(C)] H, any abelian I-group Let f be an additive lattice homomorphism from C into any abelian I-group H, and g be the natural embedding of C into [C]. Then f(C) is an abelian clan that is in H. [f(C)] is an l-ideal of H. Now define f*: g1+ g2+… +gn  f(g1) + f(g2) + … + f(gn), then f* is an l-homomorphism from G into H. Lemma 1.6. A clan C freely generates an abelian l-group G. Then C is a generating clan of G. 34 Shinemin Lin International Journal of Engineering Technology, Management and Applied Sciences www.ijetmas.com November 2015, Volume 3, Issue 11, ISSN 2349-4476 Proof. By Wyler’s construction, G is a free abelian l-group on C, it suffices to show that C is a convex sublattice of G. Clearly C is a sublattice of G. Now suppose 0 ≤ g ≤ c, where c ϵ C, and g ϵ G. Then g = g1+ g2+… +gn , where g 1 ϵ C for all i’s and there is 0 ≤ h = h1+ h2+… + hm ∈ G, hi∈ C such that g + h = c. This implies the addition of g1+ g2+… + gn is defined in C hence g ϵ C. Definition 1.7. A clan C is complete if T= { 0 ≤ ci ϵ C | i ∈ I} is a bounded subset of C then  T ϵ C. Proposition 1.8. Let C be a complete clan that generates the l-group G, and let T = { 0 ≤ ci ϵ C | i ∈ I} be a bounded subset of C. Then ˅GT=˅CT. Proof. Let v= v1ci be the join of T in C. Then v ≥ ci in G for all i ϵ I. Suppose u ∈ G satisfying u ≥ ci in G for all i ϵ I, then u ≥ ˅GT. Now v ≥ v  u = (˅Ici)  u = t ≥ 0. Therefore t ϵ C, and v ≥ ˅I(u  ci) = t  t ≥ ci  u = ci for all i ∈ I. That implies t ≥ ˅Ici= v, and hence t = v. Since u  (˅Ici) = t, u ≥ t = v. Therefore v = ˅GT. Lemma 1.9 ([4]) Let G be an l-group and let {gi | i ∈ I} be a subset of G and g ∈ G. Then ˅I(gi + g) exists if and only if ˅Igi exists. If this is the case then ˅I(gi + g) = ˅Igi + g. Theorem 1.10. A Clan C is complete if and only if G = [C] is complete.  T = {c i ϵ C | i ϵ I} ϵ  Proof. ( ) Let be a bounded subset of C, say bounded by g C G. Then, in G, ˅GT = v ≤ g exists. By convexity of C, v ϵ C. Let s be an element of C  G such that s ≥ ci for all i ϵ I. ˅ T ˅ T Then s ≥ G = v. Hence v = C . ( ) Let T = {a i ϵ C | i ϵ I} be a bounded subset of G, say bounded by g = c1 +c2 +…. + cn> 0, where a ϵ C for all i ϵ I. We apply induction on n. If n =1, then g = c1 ϵ C implying ˅I ai exists. Now suppose that T is bounded by g = c1 + c2 + …. +cn-1, then ˅ai exists in G. If T is bounded by g = (c1 +c2+….+cn-1) + cn, cn ≥ 0, then ai - cn ≤c1 +c2 + …. + cn-1. By induction hypothesis, ˅I(ai - cn) exists in G and is equal to (˅ai) – an. Therefore ˅GT exists. i Remark. Let G = [C], C is literally complete but [C] is not necessarily laterally complete. Example. Let C be a complete Boolean algebra of all subset of natural numbers. Then C is laterally complete. But [C] = set of all bounded sequences of integers, which is not laterally complete. Corollary 1.11. Let C be a positive clan, then C is an Archimedean Clan. Proof. C is complete  [C] is complete  [C] is Archimedean  C is Archimedean. Corollary 1.12. If C is a positive Archimedean clan then there exists a complete clan C^ such that C is dense in C^. Proof. C is a convex sublattice of G = [C] which is Archimedean, so G has a unique completion G^. Let C^ = (G^)+ then C  C^ and C is dense in C^. Theorem 1.13. ([16]) Let C be a clan. If x ≤ 𝑛𝑖=1 yi, then there exists xi ϵ C satisfying x = 𝑛𝑖=1 xi and xi ≤ yi for all i ϵ {1, 2,…, n}. Moreover, if x ≥ 0 and y1,….,yn ≥ 0 then xi can be chosen such that 0 ≤ xi ≤ x  yi hold for all i ϵ {1, 2,….,n}. 35 Shinemin Lin International Journal of Engineering Technology, Management and Applied Sciences www.ijetmas.com November 2015, Volume 3, Issue 11, ISSN 2349-4476 Let C be a positive clan and let &(C) be the set of all convex subclans of C. For A, B ϵ &(C), let A + B denote the set {a +b | if a + b is defined, a ϵ A, b ϵ B}. Proposition 1.14. If D ϵ &(C) then [D] is a convex 1-subgroup of [C]. Proof. It is sufficient to show that [D] is convex in [C]. Let g be an element of [C] such that 0 < g < d ϵ [D]. We have 0 < g < d = d1 + d2 + ….+ dn, where di ϵ D for all i = 1, 2, ..., n. By the theorem 1.13, g = g1 +g2 + …. + gn with 0 ≤ gi ≤ di for all i = 1, 2, …, n. Since D is convex, gi ϵ D for all i = 1, 2, … , n. This shows g ϵ [D]. Proposition 1.15. If H is a convex 1-subgroup of [C] and D is a generating clan of H, then D  C ϵ &(C) and [D  C] = H. Proof. It is sufficient to show that D  C is convex in C. Let c be an element of C such that 0 < c < d ϵ D  C  [D] = H. Since [D] is convex, c ϵ [D], we can write c = c1 + c2 + ...+ cn, where ci ϵ D for all i = 1, 2, … , n. Hence c ϵ C  D. Clearly, H  [D  C]. Now if 0 < h ϵ H  [C], then h= h1+h2+…+ hn, where hi ϵ D for all i ϵ {1, 2,….,n}. But hi ϵ [C] for all i ϵ {1, 2, … , n}, which implies hi = hi1 +….+ Him, where hij ϵ C for all j ϵ {1, 2,…,m}. We have hi ϵ D  C. This shows h ϵ [C  D]. Therefore, H= [C  D]. Schmit asked in [16] whether each minimal clan contains a greatest generalized Boolean algebra. Here we answer this question. Proposition 1.16. Suppose C is a clan. Then C contains a greatest generalized Boolean algebra. Proof. Let G be the 1-group generated by C. i.e. G = [C]. Let H be the Specker kernel of G and X be the set of singular elements of H. Then D = X  C is a convex subclan of C that generates H. Hence D is a generalized Boolean algebra that is contained in C. Since H is the Specker kernel of G, D is the greatest generalized Boolean algebra that is contained in C. Proposition 1.17. Let A, B be the convex subclans of C, then A + B = B + A if and only if A + B ϵ &(C) and B + A ϵ &(C). Proof. ( ) a) If a ϵ A and b ϵ B such that a+b is defined in C, then a = a1+b1 and b= a2+b2 where ai ϵ A and bi ϵB for i= 1, 2. a+b = (a1+b1) + (a2+b2) = a1 + (b1+a) + b2 = a1 + (a2’+b1’) +b2 = (a1 +a2’) + (b1’ + b2) ϵ A+B. b) If a, b ϵ A+B  C then there exist c1 and c2 ϵ C such that a +c1 = a˅b= c2 + a and c1 + a  b = b = a  b + c2. If a  b=0 then a˅b = a + b, so a˅b ϵ A +B. If a  b > 0 then 0 < c1< b, so c1 ϵ B. Therefore, a ˅ b = a + c1 ϵ A+B. Since 0 < a  b < a ϵ A+B, a  b ϵ A  A+B. This proves that A+B is a lattice. c) If g ϵ A+B and 0 < x <g then 0 < x < a+b = g. By theorem 1.13, x= a1+b1, where 0< a1 < a and 0 < b1< b. This shows that x ϵ A + B. Therefore, A + B is convex in C. Likewise, we can prove B+A is a convex subclan of C. (  ) Let A˅B denote the convex subclan of C that is generated by the set {A, B}. Clearly A˅B  A+B but A  A˅B and B  A˅B  A+B  A˅B, so A+B = A˅B = B˅A = B+A. Definition 1.18. A positive clan is normal valued if for all A, B ϵ (C), we have A+B = B+A. 36 Shinemin Lin International Journal of Engineering Technology, Management and Applied Sciences www.ijetmas.com November 2015, Volume 3, Issue 11, ISSN 2349-4476 Proposition 1.19. Let C be a generating clan of an 1-group G. Then G is normal valued if and only if C is normal valued. Proof. (  ) If A, B ϵ (C) then [A] + [B] = [B] + [A] If 0 < x ϵ A +B, then x=a +b ϵ [A] + [B] = [B] + [A], and hence 0 < x = b1 +a1 = (b1 +a1) +< b1+ + a1+ that implies that there exists 0 < a2 ϵ [A], 0 < b2 ϵ [B], such that x = b2 + a2 , this implies a2, b2 ϵ C. Therefore, a2 ∈ A and b2 ∈ B. This proves A+B  B+A. Likewise, we can show that B+A  A+B. (  ) If A, B ϵ (G) then [A  C] = A and [B  C] = B. For any a ϵA, and b ϵB, a + b = (a1+….an) + (b1+…..bm), where ai’s ϵ A  C and bj’s ϵ B  C. We know that (A  C) + (B  C) = (B  C) + (A  C). Therefore (a1+...an) + (b1+….bm) ϵ B+A, which implies A+B  B+A. Likewise, we can show B+A  A+B and hence A + B = B+ A. Proposition 1.20. Suppose C is a generating clan of an l-group G. If for each G(g) there exists a c ϵ C such that G(c) = G(g), then C =  [0,c] and G =  G(c). Proof. 0< g ϵ G implies g ∈ [C]+, the semigroup generated by C.  g = c1 + c2 + …. + cn, ci’s ∈ C+  G(g) = G (c1+c2+….+cn) = G(c1 + c2…..+ cn) = G(c1  c2  c3  …  cn) = G(c), where c = c1  c2  …..  cn∈ C. SECTION 2 From now on, let C be a positive abelian clan. Thus C is a convex sublattice of the positive cone G of an abelian l-group G. Moreover, the lattice order and the partial addition on C is that induced by G. Let S be a subsemigroup of C and let SC = {x ∈C | x ≥ s for some s ∈S} denotes the convex hull of S in C. + Lemma 2.1. SC is a subsemigroup of C. Thus each maximal subsemigroup of C contains 0 and is convex. Proof. For x, y ∈ SC  x < s and y < t for some s, t ∈ S  x+y is defined in C and x+ y  s + t so x + y ϵ SC. Lemma 2.2. If G(a)+  G(b)+  C then G(a˅b)+  C. Proof. For all n ϵ N, na˅nb = n( a˅b). Since a + b = b + a and since na, nb ϵ C, we have n(a˅b)= na˅nb ϵ C. Let K= {a ϵ C | G(a)+  C}. Then by above lemma {G(a)+ | a ϵK} is a directed set of subsets of C. Lemma 2.3. S =  {G(a)+ | a ϵ K} is the largest subsemigroup of C. Proof. Clearly S is a subsemigroup of C and, of course S is convex and contains 0. Now suppose B is a subsemigroup of C. Then, without loss of generality, we may assume 0 ϵ B. B is convex in C. Thus if 0 < b ϵ B then G(b)+  B  C, so b ϵ B and hence B  S. In particular [S] is an l-ideal of G = [C]. If C is not an abelian clan then C may not have a largest subsemigroup. The next example is provided by Dr. David Nelson. Example. Let G = A(R) be the group of all o-permutations of the real numbers and let α(x) = x+1 for all x ϵ R. Then [e, α] is a clan with maximal convex subsemigroups. For let β ϵ G with orbit < 1 then [e, α]  G(β)+, but the product of two such semigroups need not belong to [e, α]. Thus for a nonabelian clan need not have a largest subsemigroup. 37 Shinemin Lin International Journal of Engineering Technology, Management and Applied Sciences www.ijetmas.com November 2015, Volume 3, Issue 11, ISSN 2349-4476 Definition 2.4. A clan C is called a principal clan if C = [0,g] for some g ϵ [C]. Let G be an l-group that is generated by a principal clan C and let M be the intersection of all maximal l-ideas of G(g) (i.e. all the values of g in G(g)). Lemma 2.5. M+ is the greatest subsemigroup of C = [0, g]. Proof. For a, b ∈ M+ we have a + b ∈ M+ which implies a + b << g implies a + b ∈ C. Therefore, M+ is a subsemigroup of C. If B is a subsemigroup of C then, without loss of generality, we may assume 0 ∈ B and B is convex in C. Thus [B] is an l-ideal of G with positive cone B. Suppose, by way of contradiction, that x ∈ B\M+ then x does not belong to Gγ, a value of g, so nx is not comparable with g, which is a contradiction. Thus B  M+. Note. G/[M] is Archimedean. In fact it is a subdirect product of real numbers, so C is an extension of the largest subsemigroup M+ be an Archimedean principal clan. Once again let C be an arbitrary positive abelian clan. Then for each a ∈ C, [0, a] is a principal subclan. Let M(a) be the largest subsemigroup of [0, a], then a < b implies [0, a]  [0, b] implies G(a)  G(b). Also M(a) = {b ∈ G+| g << a} so M(a)  M(b). Lemma 2.6. M =  {M(a)| a ∈ C} is a convex subsemigroup of C that contains 0. Proof. a, b ∈ C implies M(a)  M(B)  M(a  b), so we have an upper directed set. Note. M need not to be the largest subsemigroup of C.   Example. Let G = R  R , and C = G+ then M = R  0 Lemma 2.7. If C is a positive abelian clan with {0} the only subsemigroup, then [C] is Archimedean. Proof. If [C] is not Archimedean, then we can pick 0 < a << b in C. Then 0  a ∈ M(a), which is a semigroup. Let S =  {G(a)+ | a ∈ C} be the largest subsemigroup of the clan C. Lemma 2.8. [C]/[S] =  {([S] + G©)/[S] | c ∈ C}. Proof. x∈ ([C]/[S])+ implies x = [S] + g, g >0, but g ∈ G(c ) for some 0 < c ∈ C so x ∈ ([S] + G(c))/[S]. Lemma 2.9. [C]/[S] is the join of convex l-subgroup ([S]+G(c ))/[S]  G(c )/([S]  G(c )) Proposition 2.10. Let G =[X] be a Specker group with the set of all singular elements X. Then X is the smallest generating clan of G. Proof. Suppose Y is a subclan of X and x ∈ X. If Y generates G then it must have an element y that is a-equivalent to x. But then y and x are singular that implies x = y. Now we need to show that any generating clan Y must contains X. We may assume, without loss of generality, G    Z  and X is the set of all characteristic elements of G. Since Y is convex, for each y ∈ Y the characteristic function on the support of y belongs to Y. Therefore, Y  X generates G as a group and it is a convex sublattice of G. Therefore, Y  X is a generating clan so Y  X = X and hence Y  X. 38 Shinemin Lin International Journal of Engineering Technology, Management and Applied Sciences www.ijetmas.com November 2015, Volume 3, Issue 11, ISSN 2349-4476 The converse is not true. For it we let G be a totally ordered abelian group with no maximal convex subgroup, then G+ is the only generating clan and hence the smallest generating clan. But G is not a Specker group. Proposition 2.11. Let C1 and C2 be generating clans for the representable 1-group G. Then C1  C2 is a generating clan. Proof. We know that [C1] = [C2] = G and want to show that [C1  C2] = G. Let ~ denote a-equivalence. 0 < g ∈ G g ~ a ∈ C1 and g~ b ∈ C2. There is an integer n such that ng > a and b and na, nb > g  ng > a  b and nan b = n(a  b) > b. Therefore, b  a ~ g. Remark. Suppose G is an o-group. Then, since the subclans are the convex subset of G+, they form a chain. So if G does not have a largest proper convex subgroup then G+ is the only generating clan. Suppose M is the largest convex subgroup of G then the generating clans are of the form [0,g], where g G+ \M. Corollary 2.12. If G = [X] is Specker with X the generalized Boolean algebra of singular elements, then G is the free I-group over X. Proof. X is a generating clan of G. Corollary 2.13. If C = [X] is Specker with generating clan X of singular elements, then any two addition on G so that it is an l-group are connected by an isomorphism t that is an identity on X. Proof. The only addition on X is the join of disjoin elements. So each addition of G must agree on X. Definition 2.14. A  &(C) is regular if A is maximal with respect to not containing c ∈ C. Proposition 2.15. A  &(C) is regular if and only if [A] is regular in G = [C]. Proof: (  ) Let Gα be a value of g  G = [C]. Then there is a element c  C such that c ~ g since G(g) = G[c] for some c  C. Therefore c  Gα hence c  Gα  C = A  &[C]. If B  A such that B  &[C] then [A]  [B]. Therefore, c ~ b for some b  [B]. Therefore, c  [B] hence c  [B]  C = B, Therefore A is maximal with respect to not containing c. (  ) Suppose A  &[C] is maximal with respect to not containing c  C. Then pick g  G such that g ~ c then g  [A] for if g  [A] implies c  [A] which implies c  A which contradicts our assumption. Now for any X  &(G) such that [A]  X then A  X  C  &(C)  c  X  C  g  [X  C] = X. Therefore, [A] is a value of g. Lemma 2.16. Let N be a convex sublattice of C, then [N] is an l-ideal of [C]. Consider the subset T of the l-group [C]/[N], where T={c + [N] | c ∈ C} then (1) T is a convex sublattice of [C]/[N] and hence a clan. (2) [T] = [C]/[N]. Proof. (1) T is a lattice, since (c + [N])  (d + [N]) = c  d + [N]  T, and (c + [N]  (d + [N]) = c  d + [N]  T, Now we want to show that T is convex. If 0 < g + [N]  d + [N]  T, then there is m  [N]+ such that g  d + m. Since 0 < g + [N] so 0 < (g+ - g-) + [N] = (g+ + [N]) – (g- + [N]) so 0 < g- + [N] < g++ [N] implies g- + [N] = 0 hence g-  [N]. Without loss of generality, we may assume 0 < g  G = [C].  g = d1 +m1 where 0< d1< d and 0< m1< m. g + [N] = d1 +m1 + [N] = d1 + [N]  T. To show (2) Suppose 0 < s  [C]/[N]. Then S = c + [N]/[C], where c  [C] 39 Shinemin Lin International Journal of Engineering Technology, Management and Applied Sciences www.ijetmas.com November 2015, Volume 3, Issue 11, ISSN 2349-4476 = c1 + c2 + … + cn + [N], ci  C for i=1,….,n = (c1+ [N]) +….+ (cn + [N])  [T] On the other hand, suppose 0 < t  [T]. Then t = (t1 + [N]) +……+ (tn + [N]), ti  C for i = 0,…..,n. = (t1 + t2 + … + tn) + [N]  [C]/[N]. Proposition 2.17. Let N be a convex subsemigroup of C, hence a convex sublattice. Then there is a one to one correspondence between the set of all convex subsemigroups C(&[C]) of C that contain N and the set of all convex subsemigroups of T, where T = [c +[N]| c  C]. Proof. (1) Suppose M is a convex subsemigroup of T. Let α : [C]  [C]/[N] be the natural l-homomorphism. Then α maps C onto T. Let M* = [c  C| cα  M]. Then (a) For all n  N, nα = 0 hence N  M*. (b) If h1, h2  M* then h1α, h2α  M, h1α+h2α  M T. h1α + h2α = (h1 + h2)α . Hence h1 + h2 is defined in C. Hence h1 + h2  M*. (c ) If 0 < h  h1  M* then 0 <hα  h1α  M*  hα  M  h  M*. Therefore M* is a convex subsemigroup of C that contains N and (M*)α = M. For if m*  M*, then m*α  M and if m  M  T, there is c  C such that cα = m, so c  M*. (2)If J is a convex subsemigroup of C that contains N then J’ = Jα = [j + [N] | j  J] is a convex subsemigroup of T and [J’] * = J. For if c  (J’)* then cα  J’. There is j  J such that cα = jα  c – j  N  c = j + m  J. If j  J then jα  J’  j  (J’)*. Now define f : C(&[C])  T by f(J) = J’ and g : T  C(&[C]) by g(M) = M*, then by (2) g(f(J)) = J and by (1) f(g(M)) = m. Corollary 2.18. If S is a maximal subsemigroup of C, then T = [c + [S] | c  C] has no proper nonzero convex subsemigroup. Lemma 2.19. Let S be the largest subsemigroup of C. Then [S] = G = [C] if and only if S = G + (if and only if C= G+). Proof. (  ) Clear. ( ) Suppose, by way of contradiction, that S is a proper subset of G+. Hen pick g ∈ G+\S then there is a value Gα of g such that S  Gα hence [S]  G which is a contradiction. Proposition 2.20. Let C be a proper generating clan of G. S be a maximal subsemigroup of C, and P is a value of G such that P  [S]. Then 1) G/P is an Archimedean o-group 2) G/[S] is Archimedean. Proof. 1) G/P is generated by {c + P | c ∈ C}, which is a convex sublattice of G/P and hence is a clan. If {c + P| c ∈ C} is not Archimedean then there is 0 < x + P << y + P. Let K = {c ∈ C | c + P << y + P}. Then K  C and K = [0]. If k1 + P << y + P and k2 +P << y + P then (k1+k2) + P << y +P hence k1 + k2 + P << y + P and hence k1 + k2∈ K. So K is a subsemigroup of C and for all s  S, s + P = 0 + P << y + P hence S is properly contained in K which contradicts that S is maximal subsemigroup. Therefore [c + P | c ∈ C] is Archimedean so G/P is an Archimedean o-group. {G / P | [ S ]  P and P is a value of G} by 2) Now define  : G   40 Shinemin Lin International Journal of Engineering Technology, Management and Applied Sciences www.ijetmas.com November 2015, Volume 3, Issue 11, ISSN 2349-4476  (g)      (G)  {G / P | [S ]  P and P is a value of G} . Clearly [S]   {P  | [S] c P  and P  is a value of G}. If g  [S] then there is a value P of g such that [S]  P g   {P  | [S] c P  and P  is a value of G}  [S]   {P  | [S] c P  and P  is a value of G}. Therefore ker  =  {P  | [S] c P  and P  is a value of G} = S.  G/[S]   (G)  {G / P | [ S ]  P and P is a value of G} . (….g +P,….) then is an l-homomorphism and G/ker Therefore, G/[S] is a subdirect product of Archimedean o-group hence Archimedean. Corollary 2.21. M = {M  |  ∈  , M  is a maximal 1-ideal of G} then  {M  | M  ∈ M }  C. Proposition 2.22. For an abelian I-group G. The following are equivalent. (1) G has proper generating clan. (2) G has maximal I-ideal. Proof. (2)  (1) Let M be a maximal 1-ideal of G. Then G/M is an archimedian o-group. Pick 0 < g  M then [M, r+M] generates G/M, so C =  [0, g  m] | m ∈ M] is a proper generating clan of G. (1)  (2) Let C be a proper generating clan of G. Since G is abelian, G is a subdirect product of O-groups.  ,11l momomorphism  Gi, Gi is a 0-group for all i ∈ I. C  G    pi, a projection map Gi Identify C as a subset {<c> | c ∈ C} of G. Let Ci = pi(  (C )), then for 0 < gi∈ Gi there is g ∈ G such that pi(  (g )) = gi, and since C generates G there is c ∈ C such that c is e-equivlent to g. Hence g  nc for some n ∈ N. Therefore 0 < gi = pi(  (g ))  pi(  (nc )) = npi(  (c )), which implies gi is a-equivalent to pi(  (c )), so Ci generates Gi. Similarly  (C ) generates  (G) properly since  is one to one. Gi   (C )   (G ). Let [S] =  (G) Now, it Ci = Gi for all i ∈ I, then Gi = Ci   (C )   then  (G ) = S =  (C ), which is a contradiction. Therefore [S] is a proper subset of  (G) = [  (C )]. Then pick 0 < g ∈  (G)\[S]. Let P be a value of g such that S  [S]  P, then  (G)/P is an 0-groupp generated by {  (c ) + P| c ∈ C} = T. If there is  (c1) + P <<  (c2) + P ∈ T, then let K = {  (c)|  (c) <<  (c2)}  0. Claim K is a subsemigroup of  (C ). For if  (a),  (b) ∈ K, then  (a) + P<<  (c2) + P,  (b) + P <<  (c2) + P, so  (a)+  (b) + P <<  (c2) + P, so S  K which contract to that C/P is archimedean. It follows that T is an Archimedean clan, hence  (G)/P is an Archimedean o-group. Therefore P is a maximal l-idea of  (G). Let M = {x ∈ G |  (x) ∈ P}, then M is an l-idea of G. Suppose M  J, an l-ideal of G, then P =  (M)   (J)   (G) so  (M) =  (J), hence M = J. This proves that M is a maximal l-ideal of G. If there is i ∈ I such that Ci  Gi, since Gi is an o-group, Gi has maximal l-ideal Mi. Let M = {g ∈ G| pi(  (g)) ∈ Mi} then Mi is an l-ideal of G. Define  : G/M  Gi/Mi by g + M  pi(  (g))+ Mi. Then  is onto and ker  = {g + M | pi(  (g)) ∈ Mi} = {g + M | g ∈ M} = 0 so  is one to one. So G/M is an Archimedean o-group. That means M is a + 41 Shinemin Lin International Journal of Engineering Technology, Management and Applied Sciences www.ijetmas.com November 2015, Volume 3, Issue 11, ISSN 2349-4476 maximal l-ideal of G. In particular, if S = [0] then all values of G are maximal l-ideal, so G is an Archimedean l-group. Theorem 2.23. G admits min generating clan C in which {0} is the only subsemigroup then G is Specker. Proof. (1). G is Archimedean and hence abelian. (2). G+ is not the only generating clan so G has maximal l-ideals. (3). Let M = {M  |  ∈  } be the set of all maximal l-ideal. For each M  , T  = {c + M  |c ∈ C} generates G/M  and T  is a minimal generating clan of G/M  . For if there is H* = {h + M  | h ∈ H}  T  such that H* is also a generating clan of G/M  , then N = M  C, then N  H. For g ∈ G if and only if g ∈ M   g ∈ [N]  [H]. If g  M  , g + M  = n(h + M  )  g = nh + m ∈ [H], m ∈ M   H is properly contained in C and H generates G. This is a contradiction. So G/M  has least element d + M  and hence G/M   Z. we may assume, without loss of generality, T  [0, 1] as clans. (4).  {M  |  ∈  }  C, hence  {M  |  ∈  } = 0. If g  C then there is a value Gα of g such that g  Gα. Let Cα =Gα  C, then Cα is a value of C, hence a convex sublattice of C. Let Tα = {c + [Cα]| c ∈ C} = {c + Gα | c ∈ C} then Tα is a convex sublattice of G/Gα and generates G/Gα. If [C]/Gα is not Archimedean then [c +G| c ∈ C] is not an Archimedean clan. Then there exists 0 < c + Gα << d + Gα] in T. Let K ={c ∈ C | c + Gα << d + Gα} then K is a nonzero subsemigroup of C which is a contradiction. Therefore T is an Archimedean clan. Therefore, Gα is a maximal l-ideal. Hence g   all maximal l-ideals   all maximal l-ideals  C. (G/M  )  Z  by α(g) = (…g  +M  ….). Then ker α =  {M  |  ∈ (5). Define α : G     }= 0, so α is a one to one l-isomorphism and α|C : c   I  where I  = [0,1]. Therefore C is the set of all singular elements of G and hence G is a Specker group. REFERENCES [1] M. Anderson, and T. Feil, Lattice-Ordered Groups: An Introduction, D. Reidel, 1987 [2] Bigard, A., Keimel, K., and Wolfenstein, S., Groupes et Anneaux Reticules. Lecture Notes in Mathematics, 608. Ed. A. Dold an B. Eckmann. Berlin: Springer-Verlag, 1977. [3] Birkhoff, G., Lattice Theory, Third Edition, Amer.Math. Soc. Coll. Publ. 25, New York, 1968 [4] Darnel, M., Lattice-Ordered Groups, Preprint. [5] Fuchs, L., Infinite Abelian Groups, 2 vols. Academic, New York, 1970/3 [6] Conrad, P., Lattice-Ordered Groups, Tulane Lecture Notes, 1970 [7] Conrad, P., Epi-Archmedean Groups, Czech. Math. Jour., 24 (99) 1974, 192-218 [8] Conrad, P., The Lateral Completion of a Lattice-Ordered Group, Proc. London Math. Soc. (3) 19, 1969, 444-480. [9] Conrad, P., The Essential Closure of an Archimedean Lattice-Ordered Group, Duke Math. Jour. 38 !971, 151160. [10] Conrad, P., Harvey, J., Holland, W.C., The Hahn Embedding Theorem for Lattice-Ordered Groups, Trans. Amer. Math. Soc. 108 1963, 143 – 169. [11] Conrad, P., Torsion Radicals of Lattice Ordered Groups, Symposia Math., 21 Academic Press, (1977), 479-513. [12] Conrad, P., The Hulls of Semiprime Rings. Czech. Math. Jour., 28 (103) (1978) Praha. [13]Conrad, P., Boundable Sets in Lattice Ordered Groups and Semiprime Rings, Preprint [14] Martinez, J., Torsion Theory for Latice-Ordered Ggroups I, II. Czech. Math. Jour. 25 (1975) 284-299; 26 (1976) 93-100. [15] Wyler O., CLANS, COMP. Math 17 (1966), 172 – 189 [16]Schmit K. D., Minimal Clans: A Class of ordered Partial Semigroups Including Boolean Rings and LatticeOrdered Groups., Statistik, Universitat Mannheim, A5 42 Shinemin Lin International Journal of Engineering Technology, Management and Applied Sciences www.ijetmas.com November 2015, Volume 3, Issue 11, ISSN 2349-4476 [17] Lin, S., Structure Theorems of Specker Groups (I), International Journal for Innovation Education and Research vol 3-8 2015 [18] Lin, S., Specker Rings and Semiprime Rings, International Journal of Engineering Technology, Management and Applied Scienced, Vol 3-9 83-88, 2015 [19 ] Bowman, M., Debray, S. K., and Peterson, L. L. 1993. Reasoning about naming systems. . [20] Ding, W. and Marchionini, G. 1997 A Study on Video Browsing Strategies. Technical Report. University of Maryland at College Park. [21] Fröhlich, B. and Plate, J. 2000. The cubic mouse: a new device for three-dimensional input. In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems [22] Tavel, P. 2007 Modeling and Simulation Design. AK Peters Ltd. [23] Sannella, M. J. 1994 Constraint Satisfaction and Debugging for Interactive User Interfaces. Doctoral Thesis. UMI Order Number: UMI Order No. GAX95-09398., University of Washington. [24] Forman, G. 2003. An extensive empirical study of feature selection metrics for text classification. J. Mach. Learn. Res. 3 (Mar. 2003), 1289-1305. [25] Brown, L. D., Hua, H., and Gao, C. 2003. A widget framework for augmented interaction in SCAPE. [26] Y.T. Yu, M.F. Lau, "A comparison of MC/DC, MUMCUT and several other coverage criteria for logical decisions", Journal of Systems and Software, 2005, in press. 43 Shinemin Lin