orthoposets: a construction of quantifiers fileoverwiev 1 quantiflers on a boolean algebra: basic...
TRANSCRIPT
IEGULDIJUMS TAVA NAKOTNE
Projekts Nr. 2009/0216/1DP/1.1.1.2.0/09/APIA/VIAA/044
ORTHOPOSETS: A CONSTRUCTION
OF QUANTIFIERS
Janis CırulisUniversity of Latvia
email: [email protected]
AAA 81
Salzburg, February 4–6, 2011
Thesis: A quantifier on an ordered algebra A is a closure re-
traction (i.e., a closure operator whose range is a subalgebra of A).
(Need not be a homomorphism!)
Aim: We are interested in systems of quantifiers – indexed
families (∇t: t ∈ T ) of quantifiers on A, where¦ T is a (meet) semilattice,¦ ∇s∇t = ∇s∧t,¦ every element of A belongs to the range of some ∇t.
1
OVERWIEV
1 Quantifiers on a Boolean algebra: basic facts
2 What do hold true in orthoposets?
3 Maximal orthogonal subsets in an orthoposet
4 Quantifiers induced by maximal orthogonal subsets
2
1. QUANTIFIERS ON A BOOLEAN ALGEBRA
1.1 Standard quantifier axioms (A.Tarski & F.B.Tompson
1952, P.Halmos 1955)
A quantifier on a Boolean algebra B is a unary operation ∇ such
that¦ ∇0 = 0,¦ a ≤ ∇a,¦ ∇(a ∧∇b) = ∇a ∧∇b.
Proposition. Every quantifier is an additive (even completely
additive) closure operator.
3
1.2 Quantifier axioms: another (equivalent) version (Ch. Davis
1954)
A quantifier on a Boolean algebra B is a unary operation ∇ such
that¦ a ≤ ∇a,¦ if a ≤ b, then ∇a ≤ ∇b,¦ ∇(∼∇a) = ∼∇a.
Origin: modal S5 operators.
Another name: a symmetric closure operator.
4
1.3 Quantifiers are closure retractions
Proposition (P. Halmos 1955).
An operation ∇ is a quantifier on a Boolean algebra B iff it is a
closure operator whose range is a subalgebra of B.
A subset M of B is the range of a closure operator C iff, for every p ∈ B,
C(p) = min{x ∈ M : p ≤ x}.
5
1.4 Quantifiers and partitions
Proposition (Ch. Davis 1950).
Let X be any set. An operation ∇ on P(X) is a quantifier iff
there is a partition U of X such that
∇(A) =⋃(Y ∈ U : Y ∩A 6= ∅)
for every A ∈ P(X).
(if this is the case, then ∇(A) =⋃
(∇({x}): x ∈ A))
6
2. QUANTIFIERS ON ORTHOPOSETS
2.1 Preliminaries: orthoposets
An orthoposet (or orthocomplemented poset) is a system
(P,≤,∼ ,1), where¦ (P,≤ 1) is a poset with the greatest element,¦ ∼ is a unary operation on P such that
¦ p ≤ q implies that ∼ q ≤ ∼ p,¦ ∼∼ p = p,¦ 1 = p ∨ ∼ p.
Let 0 := ∼1; then 0 is the least element of P and¦ 0 = p ∧ ∼ p.
7
Elements p and q of P are orthogonal (in symbols, p ⊥ q) if
p ≤ ∼ q.
Properties of the orthogonality relation:
¦ if p ⊥ q, then q ⊥ p,¦ p ⊥ p iff p = 0,¦ if p ≤ q and q ⊥ r, then p ⊥ q,¦ p ≤ q iff p ⊥ r for all r with q ⊥ r,¦ if P0 ⊆ P and
∨P0 exists, then
r ⊥ ∨P0 iff r ⊥ p for every p ∈ P0.
8
P – an orthoposet.
A suborthoposet of P is any subset P0 of P containing 1 and
closed under ∼ .
We may view P as a partial ortholattice (P,∨,∧,∼ ,1).
A suborthoposet of P is called a partial subortholattice if it is
closed also under existing joins and, hence, meets.
9
2.2 Quantifiers
Proposition (M.F. Janowitz, 1963).
On an ortomodular lattice L,
(a) every standard quantifier is a symmetric closure operator,
(b) every center-valued symmetric closure operator is a standard
quantifier,
(c) there are symmetric closure operators that are not standard
quantifiers.
However,¦ a closure retract on an orthomodular lattice need not be a
standard quantifier,¦ already in an ortholattice, a standard quantifier need not be
a symmetric closure operator.
10
By a quantifier on an orthoposet P we shall mean a symmetric
closure operator
¦ a ≤ ∇a, ¦ if a ≤ b, then ∇a ≤ ∇b, ¦ ∇(∼∇a) = ∼∇a.
On ortholattices: I. Chajda & H. Langer 2009.
Lemma (after Ch.Davis, 1954).
Every quantifier has the following properties:¦ ∇1 = 1, ∇0 = 0,¦ ∇∇p = ∇p,¦ p ≤ ∇p iff ∃p ≤ ∇q,¦ the range of ∇ is closed under existing meets and joins,¦ if p ∨ q exists, then ∇(p ∨ q) = ∇(p) ∨∇(q).
Corollary. An operation on P is a quantifier iff it is a closure
retraction. The range of a quantifier is even a partial subortho-
lattice of P .11
3 MAXIMAL ORTHOGONAL SETS
3.1 Some definitions
P – an orthoposet.
A subset P0 ⊆ P is¦ orthogonal if it is empty or either its elements differ from 0
and are pairwise orthogonal,¦ orthocomplete, if every its subset has a l.u.b.
Every orthogonal subset is included in a maximal one.
Let V stand for the set of all maximal orthogonal subsets of P .
12
For V ∈ V, let
V ∗:= {p ∈ P : for every v ∈ V , either v ≤ p or v ⊥ p}.
Lemma.
(a) An element belongs to V ∗ iff it is the join of some (uniquely
defined) subset of V ,
(namely, p =∨
(v ∈ V : v ≤ p))
(b) the relation ≤ on V defined by
U ≤ V iff U ⊆ V ∗ (iff U∗ ⊆ V ∗)is an ordering with the least element ⊥ := {1},(c) V has the greatest element iff P = V ∗ for some V ∈ V.
If V is orthocomplete, then V ∗ ' P(V ).
13
3.2 Quantifiers induced by subsets from V
Theorem 1.
Suppose that P is an orthoposet and V is a maximal orthogonal
subset in it. Then
(a) the mapping
∇V : p 7→ ∨(v ∈ V : v 6⊥ p)
is a quantifier with range V ∗,(b) a mapping ∇: P → V ∗ is a quantifier iff ∇ = ∇U for some
U ≤ V
Quantifiers of this type will be called V-quantifiers.
Warning: the needed suprema in (a) may not exist for some p
(i.e., the set V induces a totally defined quantifier only if it is
”sufficiently complete”).
14
Once more the theorem:
Suppose that V ∈ V. Then
(a) the mapping ∇V defined by
∇V (p) :=∨(v ∈ V : v 6⊥ p)
is a quantifier with range V ∗,(b) a mapping ∇: P → V ∗ is a quantifier iff ∇ = ∇U for some
U ≤ V .
Recall:
Proposition.
Let X be any set. An operation ∇ on P(X) is a quantifier iff there is a
partition U of X such that
∇(A) =⋃
(Y ∈ U : Y ∩A 6= ∅)
for every A ∈ P(X).
15
Examples.
(a) The simple quantifier ∇⊥:
∇⊥(p) =
{1 if p = 1,0 otherwise.
(b) The discrete quantifier Υ defined by
Υ(p) = p for all p
is a V-quantifier iff V has the greatest element.
(c) If V := {p,∼ p}, then V ∗ = {0, p,∼ p,1} and
∇V (q) =
0 if q = 0,p if 0 6= q ≤ p,
∼ p if 0 6= q ⊥ p,1 otherwise.
16
4 THE FAMILY OF V-QUANTIFIERS
Recall:
We are interested in systems of quantifiers – indexed families (∇t: t ∈ T ) of
quantifiers on A, where¦ T is a (meet) semilattice,¦ ∇s∇t = ∇s∧t,¦ every element of A belongs to the range of some ∇t.
A := P, T := V.
Questions: Is it the case that
(a) the poset V is a semilattice?(b) ∇U∇V = ∇U∧V ?(c) every element of P is in the range of some V-quantifier?
(a) – yes, if P is orthocomplete,(b) – not for all U and V ,(c) – yes (evidently).
17
4.1 Structure of V
Recall:
the relation ≤ on V defined by
U ≤ V iff U ⊆ V ∗ (iff U∗ ⊆ V ∗)
is an ordering with the least element ⊥ := {1}.
Theorem 2. If P is orthocomplete, then the poset V is bounded
complete
i.e., every subset of V having an upper bound has the l.u.b.;
equivalently, every non-empty subset of V has the g.l.b..
In particular, V is then a meet semilattice (V,∧,⊥).
18
Let U, V, W ∈ V and U, V ≤ W .
For u ∈ U , v ∈ V , let Wu,v := {w ∈ W : w ≤ u, v}.Then U ∨ V = {∨ Wu,v: u ∈ U, v ∈ V and Wu,v 6= ∅}.
19
4.2 Permutation
Theorem 3. Suppose that quantifiers ∇U un ∇V are defined.
Then the following conditions are equivalent:¦ U ≤ V ,¦ ∇U∇V = ∇U ,¦ ∇V∇U = ∇U¦ ran∇U ⊆ ran∇V ,¦ ∇U ≥ ∇V (i.e., ∇U(p) ≥ ∇V (p) for every p ∈ P ).
20
Let us call two sets U, V ∈ V correlated if, for all u ∈ U, v ∈ V ,
u ⊥ v iff u and v have no upper bound in U ∧ V .
It is the case, for instance, when U ≤ V or when u ⊥ v never
hold.
Theorem 4. Suppose that U, V ∈ V and that both quantifiers
∇U and ∇V exist. Then the following are equivalent:¦ ∇U∧V exists, and ∇U∇V = ∇U∧V ,¦ U and V are correlated.
21
Once more:
Let us call two sets U, V ∈ V correlated if, for all u ∈ U, v ∈ V ,
u ⊥ v iff u and v have no upper bound in U ∧ V .
Example. Suppose that U = {p,∼ p}, V = {q,∼ q} and p < q.
Then ∼ q < ∼ p, and¦ U ∧ V = ⊥,¦ U and V are not correlated, and¦ ∇U∇V (p) = ∇U(q) = 1, but ∇V∇U(p) = ∇V (p) = q.
Example. Suppose that U = {p,∼ p}, V = {q,∼ q}, p 6⊥ q and
p 6≤ q. Then¦ U ∧ V = ⊥,¦ U and V are correlated, and¦ ∇U∇V = ∇U∧V = ∇V∇U .
22