REPRESENTATION THEOREMS IN UNIVERSAL ALGEBRA AND ALGEBRAIC LOGIC
by
Martin Izak Pienaar
THESIS
presented in partial fulfillment of the requirements for the degree
MAGISTER SCIENTIAE
in
MATHEMATICS
in the
FACULTY OF SCIENCE
of the
RAND AFRIKAANS UNIVERSITY
Study leader: Prof. V. Goranko
NOVEMBER 1997 Baie dankie aan:
Prof. V. Goranko - vir leiding in hierdie skripsie.
RAU en SNO - finansiele bystand gedurende die studie.
my ouers - onbaatsugtige liefde tydens my studies.
Samantha Cambridge - vir die flink en netjiese tikwerk.
ons Hemelse Vader - vir krag en onderskraging in elke dag.
Vir my ouers en vriende - julie was die .inspirasie viz- my studies CONTENTS
1 INTRODUCTION TO LATTICES AND ORDER 1
1.1 Ordered Sets 1 1.2 Maps between ordered sets 2 1.3 The duality principle; down - sets and up - sets 3 1.4 Maximal and minimal elements; top and bottom 4 1.5 Lattices 4
2 IMPLICATIVE LATTICES, HEYTING ALGEBRAS AND BOOLEAN ALGEBRAS 8
2.1 Implicative lattices 8 2.2 Heyting algebras 9 2.3 Boolean algebras 10
3 REPRESENTATION THEORY INVOLVING BOOLEAN ALGEBRAS AND DISTRIBU- TIVE LATTICES: THE FINITE CASE 12
3:1 The representation of finite Boolean algebras 12 3.2 The representation of finite distributive lattices 15 3.3 Duality between finite distributive lattices and finite ordered sets 17
4 REPRESENTATION THEORY INVOLVING BOOLEAN. ALGEBRAS AND DISTRIBU- TIVE LATTICES: THE GENERAL CASE 20
4.1 Ideals and filters 20 4.2 Representation by lattices of sets 22 4.3 Duality 27
5 REPRESENTATION THEORY: A MORE GENERAL AND APPLICABLE VIEW 30
5.1 Analogue statements for the representation theorems of chapters 3 and 4 30 5.2 Introduction to relation algebras, finite - dimensional cylindric algebras, and their interconnections 32 5.3 Representation of relation algebras 35 5.4 Representation of finite - dimensional cylindric algebras 37
6 BOOLEAN ALGEBRAS WITH OPERATORS'- , 42
6.1 Definitions and notation 43 6.2 The so - called extension theorem for Boolean algebras with operators 44 6.3 A representation theorem for Boolean algebras with operators 45 6.4 Application of the results on Boolean algebras with operators to closure algebras and cylindric algebras 47
7 SOME APPLICATIONS OF REPRESENTATION THEOREMS IN UNIVERSAL AL- GEBRA AS WELL AS IN ALGEBRAIC LOGIC 51
7.1 Applications in universal algebra 51 7.2 The subdirect product representation theorem - an extension of Birkhoff's repre- sentation theorem for finite distributive lattices 53 7.3 Applications in algebraic logic 55
REFERENCES 59 OPSOMMING
In hierdie skripsie word die invloed van sg. "representation theorems" in universele algebra en algebraiese logika, op sekere abstrakte wiskundige strukture beskou. Stellings wat 'n sekere abstrakte wiskundige struktuur voorstel as 'n meer algemene wiskundige model, word ondersoek. Die skripsie is as volg georganiseer:
Hoofstuk 0 dien as 'n basiese inleiding tot "representation theorems". Dit verduidelik die idee agter "representation theorems" en verskaf 'n basiese "representation theorem".
.Hoofstuk 1 bestaan hoofsaaklik uit definisies wat van belang is vir die res van die skripsie.
In hoofstuk 2 word die wiskundige strukture, naamlik "implicative lattices", "heyting alge: bras", en "Boolean algebras" ondersoek.
Hoofstuk 3 verduidelik die rol van "representation theorems" in eindige "distributive lattices" en eindige "Boolean algebras".
In hoofstuk 4 word ekwivalente resultate as die van hoofstuk 3, vir arbitrere "Boolean alge- bras" en "distributive lattices" bespreek.
Hoofstuk 5 bespreek die idee van "representation theorems" in die algemeen. Twee verdere strukture, naamlik "relation algebras" en "finite - dimensional cylindric algebras" word be- spreek.
Hoofstuk 6 is 'n "uitbreiding" van vorige resultate. Die rol van "representation theorems" in arbitrere "Boolean algebras" word bespreek.
In hoofstuk 7 word sommige toepassings van "representation theorems" in universele algebra en algebraiese logika beskou.
-000- SUMMARY
The objective of this thesis is to discuss the essence and importance of representation the- orems in the area of universal algebra and algebraic lOgic, generally and by means of some fundamental examples.The thesis is organized as follows:
Chapter 0 is introductory. It gives an initial illustration of the topic by means of one of the historically first and most important examples of representation theorems in mathematics, viz. Cayley's representation theorem for groups.
Chapters 1 and 2 contain preliminary material on order and lattices as well as specific types of lattices, Heyting and Boolean algebras, which are important in algebraic logic. They provide the neccessary technical background for the thesis.
Chapters 3 to 7 are the core of the thesis.
Chapter 3 presents in detail the representation theorems for finite distributive lattices via duality with ordered sets and for Boolean algebras as fields of subsets of finite sets.
Chapter 4 presents the classical representation theorems(in the general case) for distributive lattices and Boolean algebras, due to M.Stone, G.Birkhoff and H.Priestley, and discuss the duality between Boolean algebras and Boolean topological spaces.
Chapter 5 mentions analogous results for implicative lattices and Heyting algebras due to Stone and Rasiowa-Sikorski. Further, it briefly introduces relation algebras and some basic results about them.
Chapter 6 is based on the classical result of Tarski and Jonsson providing a general represen- tation theorem for Boolean algebras with normal and additive operators, with applications to representation theorems for relation algebras and cylindric algebras.
Chapter 7 mentions some representation results in universal algebra., most importantly Birkhoff's subdirect reducibility theorem. It also briefly discusses an application of rep- resentation theorems in algebraic logic, illustrating the case of intuitionistic logic.
-000- Chapter 0
Introduction to Representation Theorems
Quite often, all the different fields in Mathematics seem to be independent with no inter- connection. In a specific field, for instance algebra, certain structures sometimes seem to be very abstract and it is difficult to express the global effect of the involved structure in that field. This "problem" is partly overcome by so - called representation theorems. By a representation theorem, the "representation" of an abstract mathematical structure as Xsometimes well - known) mathematical model is studied. This ensures that the involved structure as well as its applications become clearer. Because representation theorems are mostly based on isomorphisms between the involved structures, one can also conclude that properties which hold in the one structure must also hold (up to an isomorphism) in the other, less - involved, structure.
To explain the ideas mentioned in the above paragraph, it is good to investigate one of the historical first examples of a representation theorem. We are namely going to state and prove Cayley's well - known theorem which ensures that an arbitrary group is isomorphic to a group of permutations. This would mean that] properties which hold in a certain group, will also be true in the involved group of permutations (due to the isomorphism between the two structures).
Definitions:
A permutation of a set A is a function from A to A that is both one - to - one and onto.
A permutation group of a set A is a set of permutations of A that forms a group under function composition.
Informally, Cayley's theorem states that every group is isomorphic to some group consisting of permutations under permutation multiplication. This is a nice and intriguing result, and is a classic in group theory. At first glance, the theorem might seem to he a tool to answer all questions about groups. What it really shows is the generality of the groups of permutations. Examining subgroups of all permutation groups for a set of all sizes would be a tremendous task. Cayley's theorem does show that if a counterexample exists to some conjecture you have made about groups, then some group of permutations will provide the counterexample.
We next state and prove the actual theorem:
Theorem 0.0.1 Cayley's Theorem Every group is isomorphic to a group of permutations.
Proof:
Let G be a given group.
Step 1: Our first task is to find a set C' of permutations that is a candidate to form a group isomorphic to G. Think of G as just a set, and let SG be the collection of all permutations of G. Then SG is a group under permutation multiplication (this fact follows easily from basic definitions). Note that in the finite case if C has n elements, SG has n! elements. Thus, in general, SG is clearly too big to be isomorphic to G. We define a certain subset of SG. For a E C, let A, be the mapping of G into G given by Aa (x) = ax for x E G. (We can think of )1„ as left multiplication by a). If A a (s) = A a (y), then ax = ay so that x = y (by an easy exercise). Thus A, is a one - to - one function. Also, if y E G, then A a (a - ly) = a(a'y) = y, so )' a maps G onto G. Since Aa : G is both one - to - one and onto G, A, is a permutation of G; that is A, E SG. Let now G' = a E G}.
Step 2: We claim that G' is. a subgroup of SG. We must show that G' is closed under Permutation multiplication, contains the identity permutation, and contains an inverse for each of its elements. First, we claim that )t a \b = Aab• To show that these functions are the same, we must show that they have the same action on each x E G. Now (A,Ab)(x) = A a (Ab(x)) = A a(bx) = a(bx) = (ab)x = Aab(x). Thus Adtb = kb, so G' is closed under multiplication. Clearly for all x E G, ) e (x) = ex = x, where e is the identity element of C, so A, is the identity permutation i of SG and is in Since A„A b = A ab, we have A„A„-i = A aa -i = Ae , and also A„--1A„ = )' e . A: 1 , Hence (Aa)- so that (A,) -1 E Thus G' is a subgroup of SG. Step 3: It remains for us to prove now that G is isomorphic to this group G' that we have described. Define : G C' by 0(a) = Aa for a E G. If 0(a) 0(b), then .\a and )'b must be the same permutation of G. In particular, A a (e) = )t b (e), so ae = be and a = b. Thus 0 is one - to - one. It is immediate that 0 is onto G' by the definition of G'. Finally, 0(ab)= Aabi while q(a)q(b) = Aa Ab. But we saw in step 2 that )tab and )'a )b are the same permutation of G. Thus 0(ab) 85(a)0(b), so that 0 is an isomorphism from G onto G'. Therefore the theorem holds.
Note that, for the above proof, we could have considered equally well the permutations p a of G defined by pa(x) = xa for x E G. (We can think of p a as meaning right multiplication by a). These permutations form a subgroup G" of SG, again isomorphic to G, but under the map it : G G" defined by it(a) = pa-i. The group G' in the proof of Cayley's theorem is the left regular representation of G, and the group G" in the preceeding comment is the right regular representation of G. Chapter 1
Introduction to Lattices and Order
1.1 Ordered Sets
Definitions: Let P be a set. An order (or partial order) on P is a binary relation < on P such that, for all x, y, z, E P,
x 5_ x,
x < y and y < x imply x = y,
x < y and y < z imply x < z.
These conditions are referred to, respectively, as reflexivity, anti-symmetry and transitivity. A set P equipped with an order relation < is said to be an ordered set (or partially ordered set - also known as a "poset"). Where it is necessary to specify the order relation, we write (P; An order relation < on P gives rise to a relation < of strict inequality: x < y in P if and only if x < y and x y. It is possible to re-state conditions (i) - (iii) above in terms of <, and so to regard < rather than < as the basic relation.
Other notation associated with < is predictable. We use x < y and y > x interchangeably, and write x y to mean x < y is false, and so on. Less familiar is the symbol II used to denote non- comparability: we write x y if x y and y X x.
1.1.1 Chains and antichains
Let P be an ordered set. Then P is a chain if, for all x, y E P, either x < y or y < x. Alternative names for a chain are linearly ordered set and totally ordered set. The ordered set P is an antichain if x < y in P only if x = y. Clearly, with the induced order, any subset of
a chain (an antichain) is a chain (an antichain). Let P be the n-element set {0,1, ,n —1}. We write n to denote the chain obtained by giving P the order in which 0 < 1 < < n 1 and n for P regarded as an antichain.
1.1.2 The covering relation.
Let P be an ordered set and x, y E P. We say x is covered by y (or y covers x), and write x y or y F- x if x < y and x < z < y implies z = x.The latter condition is demanding that there be no element z of P with x < z < y.
Observe that, if P is finite, x < y if and only if there exists a finite sequence of covering relations x = xo H x i H ... = y. Thus, in the finite case, the order relation determines, and is determined by, the covering relation.
1.2 Maps between ordered sets
Definitions: Let P and Q be ordered sets. A map cp:P---)Qis said to be
order-preserving (or, alternatively, monotone) if x < y in P implies cp(s) < cp(y) in Q;
an order-embedding if x < y in P if and only if cp(x) < v(y) in Q;
an order-isomorphism if it is an order-embedding mapping P onto Q.
When cp : P Q is an order-embedding we write cp : Q. When there exists an order- isomorphism from P to Q, we say that P and Q are order-isomorphic and write P Q.
Remarks:
Let cp : P Q and : Q R be order- preserving maps. Then the composite map o v, given by (0 o c,9)(x) 1/)(yo(x)) for x E P, is order-preserving. More generally the composite of a finite number of order-preserving maps is order-preserving, if it is defined.
Let cp : P Q and let y(P) (defined to be {cp(x) x E P}) be the image of cp. Then V(P) P. Ordered sets P and Q are order-isomorphic if and only if there exist order-preserving . maps y : P Q and ?,/, : Q P such that cp. o 7b= idQ and o cp = idp (where ids : S S denotes the identity map on S given by ids (x) = x for all x E S).
9 1.3 The Duality Principle; down-sets and up-sets
1.3.1 The dual of an ordered set
Given any ordered set P we can form a new ordered set P8 (the dual of P) by defining x < y to hold in P6 if and only if y < x holds in P. For instance, if there exists a unique element in P which covers just 3 other elements, then there exists a unique element in Pb which is covered by just 3 other elements. This situation is illustrated in figure 1.1.
In general, given any statement (I) about ordered sets, we obtain the dual statement (D6 by replacing each occurrence of < by > and vice versa.
1.3.2 The Duality Principle
Given a statement (I) about ordered sets which is true in all ordered sets, then the dual statement 4)6 is true in all ordered sets.
1.3.3 Definitions and Remarks
Let P be an ordered set and Q C P.
Q is a down-set (alternative terms include decreasing set or order ideal) if; whenever xE Q,yEP and y < x, we have y E Q
Dually, Q is an up-set (alternative terms are increasing set or order filter) if, whenever xE Q,yEP and y > x, we have y E Q.
3 Besides being related by duality, clown-sets and up-sets are related by complementation: Q is a down-set if and only if P \ Q is an up-set.
Given an arbitrary subset Q of P and x E P, we define
Q = {yE P (3x E Q)Y < *x}; Q = {y E P (3x E Q)y > x}; x = {yEP I y5x} and x = {y€P I y>x}.
It is obvious that I Q is the smallest down-set containing Q and that Q is a down-set if and only if Q =j, Q, and dually for I Q. Clearly 1 {x} x, and dually for x. The family of all down-sets of P is denoted by•V(P). It is itself an ordered set, under the inclusion order, and plays a crucial role in the proofs of representation theorems involving lattices and order.
1.4 Maximal and minimal elements; top and bottom
1.4.1 Maximal and minimal elements
Let P be an ordered set and let Q C P. Then
a E Q is a maximal element of Q if a a E Q is the greatest (or maximum) element of Q if a > x for every x E Q, and in that case we write a = max Q. A minimal element of Q, the least (or minimum) element of Q and min Q are defined dually, that is by reversing the order. Note that the greatest (or least) element need not exist, but if it does it is unique. 1.4.2 Top and bottom Let P be an ordered set. The greatest element of P, if it exists, is called the top (element) of P and written T. Similarly, the least element of P, if it exists, is called the bottom (element) and denoted by 1. 1.5 Lattices Definitions: 4 (i) Let P be an ordered set and let S C P. An element x E P is an upper bound of S if s < x for all s E S. A lower hound is defined dually. The set of all upper bounds of S is denoted by Su and the set of all lower bounds by S'. Thus: Su := E P I (Vs E S)s x} and S I := E P I (Vs E S)s x}. (ii) Let (X, <) be a poset and x, y E X. The least upper bound of {x, y}, if it exists, is called the join of x and y, denoted by x V y; the greatest lower bound of {x, y}, if it exists, is called the meet of x and y, denoted by x A y. (iii) First notion of a lattice: A lattices is a non-empty poset (X, <) in which every two elements have join and meet. EXamples of lattices 1 are, for instance: Every linearly ordered set. (P(X),C) for any set X. (Ar+, I), where I is the divisibility relation in N.+. The family of subspaces of a linear space, ordered by C. The family of subgroups of a group, ordered by C. (iv) Second notion of a lattice: A lattice 2 is a non-empty set X endowed with two binary operations, _A and V, which satisfy the following identities for every x, y, z E X: xAx=x; xVx=x sAy=yAx; xVy=yVx x A (y A z) = (x A y) A z; x V (y V z) = (x V y) V z x V (x A y) = x; x A (x V y) = x The above two notions of a lattice are equivalent in the following sense: Proposition 1.5.1 (i) If A, V) is a lattice2 , then the relation < defined by x < y if and only if x A y = x is a partial order in X and (X, <) is a lattice' with a meet A and a join V. (ii) If (X, <) is a lattices , then the corresponding lattice2 (X, A, V) satisfies the property x < y if and only ifsAy=x if and only if x V y = y for every x,y E X, and thus determines a lattice' identical with (X, <). Proof: (i) Let (X, A, V) be a lattice2 . The relation < in X, defined by x < y if and only if x Ay = x, is a partial ordering: 5 For every x E X, x A x = x hence x < x, so that < is reflexive. Suppose x < y and y < z, for x, y, z E X. Thus x Ay = x and y A z = y. Hence xAz= (x A y) Az = x A (y A z) = x A y = x, so that x < z which implies that < is transitive. Suppose x < y and y < x, for x, y E X. Thus x A y = x and y A x = y, so that x = y which implies that < is antisymmetric. Furthermore, for every x,y E X, (xAy)Ax=xA(yAx)=xA(sAy)=(xAx)Ay=xAy. Likewise (x A y) A y = x A (y A y) = x A y, so that x A y is a lower bound for x and y. x A y is the greatest lower bound of x and y: Letz zA(sAy).(zAx)Ay=zAy=z which implies that z < x A y. Also, for every x,y E X, (xVy)Vx=xV(yVx)=xV(xVy)=(xVx)Vy=xVy. Likewise (x V y) Vy=xV (y V y) = x V y, so that x V y is an upper bound for x and y. x V y is the least upper bound of x and y: Let z > x and z > y, for z E X. This means that x < z and y< z, so that x A z = x and y A z= y. Therefore xVz =z and y V z = z. Then zV(xVy).(zVx)Vy=zVy=z which implies that z A (x V y) = x V y. Thus (x V y) Az = x V y, so that x V y < z. Thus: (X, <) is a non-empty poset in which every two elements have join and meet, so that (X, <) is a lattice'. (ii) We just give a brief summary of the proof of this part: For every x, y E X, if x < y then x is the greatest lower bound of x and y and y is the least upper bound of x and y. Conversely, if x Ay = x then x < y and so forth. Hereafter we identify the notions of lattice' and lattice 2 and call each of them a lattice. Definitions: 6 (i) A lattice (X, <) is modular if it satisfies the following modular law: For every x, y, z E X, x < y implies x V (y A z) = (x V y) A z. Examples of modular lattices are: Any linear ordering. (P(X), C ) for any set X. The lattice of subspaces of any linear space. (ii) A lattice (X, A, V) is distributive if it satisfies either one of the following distributive laws: For every x, y, z E X, x A (y V z) = (x A y) V (x A z) x V (y A z) = (x V y) A (s V z) Examples of distributive lattices are: Any linear ordering. (P(X), ) for any set X. (iii) A non-empty subset Y of a lattice X is a sublattice of X if it is closed under meets and joins, i.e. for every al, a2 E Y, a l V a2 E Y and al A a2 E Y. 1.5.1 Maps between lattices Let (X1 , A1, VI) and (X2, A2, V2) be two lattices. A mapping h : Xl —44 2 is called: A homomorphism (of lattices) if h(x = h(s) A2 h(y) and h(x V 1 y) = h(x) V2 h(y) for every xl, x2 E An isomorphic embedding if it is an injective homomorphism. An isomorphism if it is a bijective (i.e. both injective and surjective (onto)) homomor- phism. Chapter 2 Implicative Lattices, Heyting Algebras and Boolean Algebras In this chapter we begin the study of Boolean algebras qua lattices. The representation theory of chapters 2 and 3 carries this further. 2.1 Implicative Lattices Definitions: (i) Let (X, A, V) be a lattice and a, b, E X. An element c E X is called a pseudo-complement of a relative to b if c is the greatest element in X with the property c A a < b. We denote the pseudo - complement of a relative to b (if it exists) by a b. Note that: For every x E X,x < a --+ b if and only if x A a < b. a a exists if and only if the lattice has a top element T and then a a = T. (ii) A lattice (X, A, V) is an implicative lattice if a b exists for every a, b E X. The latter operator is called an implication. Note that every implicative lattice has a top element T (but not necessarily a bottom element 1) and hence can be regarded as a system (X, A, V, T). Examples of implicative lattices are: (P(X), C) for any set X. (A representative example) The lattice of open dense subsets of a topological space. For any two elements, say A and B of this lattice,A B = Int (A U B) where Int(A U B) denotes the interior of the set AU B and A denotes the complementary set of A in the involved topological space. 8 In later chapters we will look more carefully into representation theorems involving distribu- tive and implicative lattices specifically. The following proposition will play a central part in these chapters. Proposition 2.1.1 Every implicative lattice (X, A, V, T) is distributive. Proof: Take any elements x, y, z E X. Denote u = (x A y) V (x A z). This means that x A y < u andxAz