Groupoids on a Skew Lattice of Objects
Total Page:16
File Type:pdf, Size:1020Kb
ISSN 2590-9770 The Art of Discrete and Applied Mathematics 2 (2019) #P2.03 https://doi.org/10.26493/2590-9770.1342.109 (Also available at http://adam-journal.eu) Groupoids on a skew lattice of objects Desmond G. FitzGerald ∗ School of Mathematics and Physics, University of Tasmania, Private Bag 37, Hobart 7001, Australia To Jonathan Leech—Na zdravje! Received 16 October 2018, accepted 11 December 2019, published online 27 December 2019 Abstract Motivated by some alternatives to the classical logical model of boolean algebra, this paper deals with algebraic structures which extend skew lattices by locally invertible ele- ments. Following the meme of the Ehresmann-Schein-Nambooripad theorem, we consider a groupoid (small category of isomorphisms) in which the set of objects carries the struc- ture of a skew lattice. The objects act on the morphisms by left and right restriction and extension mappings of the morphisms, imitating those of an inductive groupoid. Condi- tions are placed on the actions, from which pseudoproducts may be defined. This gives an algebra of signature (2; 2; 1), in which each binary operation has the structure of an orthodox semigroup. In the reverse direction, a groupoid of the kind described may be reconstructed from the algebra. Keywords: Inductive groupoids, skew lattices, orthodox semigroups. Math. Subj. Class.: 20L05, 20M19, 06B75 1 Non-commutative and non-idempotent lattice analogues As non-classical logics have been developed for various knowledge domains, so various algebras have been proposed as extensions or alternatives to the classical model of boolean algebra. A significant one for this paper is the theory of skew lattices; for a contemporary account, see Leech’s surveys [10, 11]. We provide the details of our notation in Section3. Another proposal is that of MV-algebras, and their coordinatisation¨ via inverse semigroups as described by Lawson and Scott [9]. Thus one theme is to allow sequential operations and hence non-commutative logical connectives, and another introduces non-idempotent ∗The author thanks Vicky Gould, Michael Kinyon and Anja Kudryavtseva for their interest and input in discus- sions, both in person and by email, which helped convert a half-baked workshop idea into an interesting project. E-mail address: d.fi[email protected] (Desmond G. FitzGerald) cb This work is licensed under https://creativecommons.org/licenses/by/4.0/ 2 Art Discrete Appl. Math. 2 (2019) #P2.03 connectives. This note will consider a combination of these themes, by seeking reasonable structures which extend skew lattices by locally invertible elements. The principal tool in this construction is based on the ideas behind the Ehresmann- Schein-Nambooripad (ESN) theorem, of which a full account is given in Chapter 4 of Lawson’s book [8]: we consider a small category of isomorphisms in which the set of objects carries the structure of a skew lattice. We postulate that the objects act (partially) on the morphisms by left and right restriction and extension mappings of the morphisms, imitating those of an inductive groupoid. Certain reasonable conditions are postulated, and from these a suitable pseudoproduct is defined, much as in the inverse semigroup case, for each skew lattice operation (a non-commutative “meet” and “join”). This results in a total algebra involving two orthodox semigroups with a common set of idempotents isomorphic to the given skew lattice. Because of the complexity involved in having two operations, we begin by considering a groupoid over a set of objects with a single band operation. A much more general situation has been studied, under the name of weakly B-orthodox semigroups, by Gould and Wang in [2], but because the present author has been unable to find this special case treated in the literature, a detailed account will be given here. Later sections deal with the pair of linked band operations, construct the total algebra described above, and show how the original groupoid may be recovered from the algebra. Aspects of the constructions which need further elaboration are noted in the final section. 2 Groupoids on a band of objects Let us recall from [8] that an inverse semigroup is equivalent to an inductive groupoid, i.e., • a (small) category of isomorphisms with • a meet operation on objects and • a notion of restriction of a morphism to any of its subdomains. We attempt something similar here, but changing the conditions on the set of objects. Let G be a groupoid with composition ◦ and B its set of objects, endowed with an associa- tive and idempotent operation ^. Then (B; ^) is known as a lower band, and possesses a pair of natural preorders: we write • a ≤L b if and only if a = a ^ b, and a ≤R b if and only if a = b ^ a. As usual, we may identify each object b with its identity ib, and write dg and rg for the domain and range maps in G , thus: dg = g ◦ g−1; rg = g−1 ◦ g. Suppose too that for each a 2 B there are left and right restriction (partial) operations aj; ja : G ! G such that: • ajg is defined whenever a ≤L dg, with ajg : a ! r(ajg) ≤L rg; and (lateral-) dually, • gja is defined whenever a ≤R rg, with gja : d(gja) ! a; d(gja) ≤L dg. Figure1 shows the left and right restrictions. (There are analogous (in fact, vertically dual) requirements for extension operators, which will be dealt with more explicitly in Section3.) Certain sensible axioms must be satisfied: D. G. FitzGerald: Groupoids on a skew lattice of objects 3 g g c d c d a ≤L c a ≤R d a r(ajg) d(gja) a ajg gja Figure 1: Left and right restriction operators. (i) (identities) d(g)jg = g; (ii) (preorders) if a ≤L b, then ajib = ia; (iii) (transitivity) if a ≤L b ≤L dg, then ajg = a^bjg = aj(bjg); (iv) (composition) if f ◦ g is defined (so that rf = dg), then aj(f ◦ g) = (ajf) ◦ (r(ajf)jg); the right-hand composite being defined because r(ajf) ≤L rf = dg; see Figure2. f g b c d ≤L ≤L ajf r(ajf)jg a r(ajf) r(ajf ◦ g) aj(f ◦ g) Figure 2: Restriction of a composite morphism. 2.1 Actions and conjugacy g g Let us write, without prejudice, a as an alternative for r(ajg), and a for d(gja). This lightens the notation, and emphasises the similarity to actions and conjugates. Caution: f −1 However, it should not be taken to mean that anything like a = (f ja) ◦ (ajf), or f f −1 a = a, or ajf = fjaf necessarily hold: in general, fjaf = fjrf^af . What we do have, following from aj(f ◦ g) = (ajf) ◦ (r(ajf)jg), is that −1 ajib = ia = (ajf) ◦ (af jf ); 4 Art Discrete Appl. Math. 2 (2019) #P2.03 so −1 −1 ib f −1 (ajf) = af jf ; a = a; and a = (af jf ) ◦ (ajf); moreover, af◦g = (af )g. We want to link right and left “actions” by (ajf)jb = aj(fjb), but there is a little problem here, since one of the two sides of that equation may fail to be defined while the other is defined. We therefore seek to extend the conditionally-defined restrictions to total maps by the following device, based on the pseudoproduct construction familiar from the ESN theorem: For g 2 G and any a 2 B, define ajg := a^dgjg, the right hand side being meaning- ful since a ^ dg ≤L dg. (Note that if a ≤ dg already, the notations agree.) The next figure shows the situation (where, also by extension, we write ag for the already-defined (a ^ dg)g). g a b c ≤L a ^ b (a ^ b)g = ag ajg Figure 3: Generalised restriction of g to object a and action of g on a. Then if g = ib, we have a ^ ib = a^bjib = ia^b = ia ^ ib = ia ^ b = iaja^b, and we may write a ^ g for ajg without conflict. A little re-writing of definitions shows that (a ^ b) ^ g = a^bjg = aj(bjg) = a ^ (b ^ g) (2.1) and a ^ dg = d(a ^ g) = (a ^ g) ◦ (a ^ g)−1: (2.2) We complete our list of postulates with the previously-mentioned (ajf)jb = aj(fjb), which we now write as • (a ^ f) ^ b = a ^ (f ^ b); for all a; b 2 B and f 2 G . (More fully, this is (a^df jf)jaf ^b = a^f bj(fjb^rf ).) Next, we may extend the composition further, to a pseudoproduct ⊗: when f : z ! a and g : b ! c, we define f ⊗ g := (fja^b) ◦ (a^bjg) = (f ^ (a ^ b)) ◦ ((a ^ b) ^ g): The pseudoproduct is defined for all pairs f; g. Then a ^ f is actually just ia ⊗ f. This is indeed an extension of meaning: when f ◦ g is defined, f ⊗ g = f ◦ g, and when f = ia and g = ib, f ⊗ g = ia ⊗ ib = ia^b = ia ^ ib; so we may as well use just the one symbol ^ for ⊗, as it extends ◦ and the restrictions, as well as the original ^ on B. Let us check remaining non-trivial cases for associativity. D. G. FitzGerald: Groupoids on a skew lattice of objects 5 f g z a b c ≤R ≤L f (a ^ b) a ^ b (a ^ b)g fja^b a^bjg Figure 4: Diagram illustrating the pseudoproduct. Lemma 2.1. For all f; g 2 G and e 2 B, with f : df ! a and g : b ! rg, (i) (f ^ e) ^ g = f ^ (e ^ g), (ii) ef^g = (ef )g, and (iii) ej(f ^ g) = (ejf) ^ g. Proof. (i): By definition, (f ^ e) ^ g = (fje) ^ g = (fja^e)jb ◦ a^ejg = (fja^e^b) ◦ (a^e^bjg); while f ^ (e ^ g) = fje^b ^ e^bjg = f ^ (e ^ g).