Professional Documents
Culture Documents
Lecture 1
Some properties of the logical connectives in PL
1
Let us choose and fix any tautology in the set F of all wffs of PL (over
some set P V of propositional variables) and denote it by >. So for every
tautology α, we have α ≡ >. Similarly, let us fix the notation ⊥ for a
contradiction.
Let us recall the following properties of the logical connectives ∧, ∨, ¬
that we mentioned in Section 4 of Lecture 3, Module I.
1. α ∧ β ≡ β ∧ α
2. α ∨ β ≡ β ∨ α
3. α ∧ (β ∧ γ) ≡ (α ∧ β) ∧ γ
4. α ∨ (β ∨ γ) ≡ (α ∨ β) ∨ γ
5. α ∧ (β ∨ γ) ≡ (α ∧ β) ∨ (α ∧ γ)
6. α ∨ (β ∧ γ) ≡ (α ∨ β) ∧ (α ∨ γ)
7. α ∧ > ≡ α
8. α ∨ > ≡ >
9. α ∧ ⊥ ≡ ⊥
10. α ∨ ⊥ ≡ α
11. ¬¬α ≡ α
12. ¬> ≡ ⊥
13. ¬⊥ ≡ >
14. ¬(α ∧ β) ≡ ¬α ∨ ¬β
The above seem to indicate some familiar properties. For instance, 1-4
appear to refer to commutativity and associativity of the ‘operators’ ∧, ∨
in the set F, while 5-6 indicate distributivity of ∧, ∨ over each other. 11
indicates that ¬ has the property of involution, and the last pair says that
∧, ∨, ¬ satisfy the DeMorgan laws. The only catch is that the above are
not equations in F, but logical equivalences between wffs in F. Indeed, it
is incorrect to say, for example, that α ∧ β = β ∧ α, because that would
amount to asserting that, as strings, α ∧ β and β ∧ α are identical – which
is certainly not the case.
Can we turn all the above logical equivalences into equations in some
set? Indeed, we can.
2
1 Operators in the quotient set F/ ≡
Consider the quotient set F/ ≡ discussed in Section 4.1 of Lecture 3, Module
I. Define the following operators in F/ ≡. Let [α], [β] be any elements in
F/ ≡.
Definition 1.1.
(i) [α] ∧ [β] := [α ∧ β].
(ii) [α] ∨ [β] := [α ∨ β].
(iii) ¬[α] := [¬α].
Exercise 1.1.
(i) [α] ∧ [β] = [α0 ] ∧ [β 0 ],
(ii) [α] ∨ [β] = [α0 ] ∨ [β 0 ],
(iii) ¬[α] = ¬[α0 ],
for any α0 ∈ [α], β 0 ∈ [β] (i.e. α ≡ α0 , β ≡ β 0 ).
We can now show that, all the logical equivalences listed at the beginning
of this section, translate into equations in the set F/ ≡. For instance,
[α] ∧ [β] = [β] ∧ [α], [α] ∧ [>] = [α], or ¬([α] ∧ [β]) = ¬[α] ∨ ¬[β]. This is left
as an exercise.
Moreover, as noted in Exercise 2.2 of Lecture 5, Module I, one can show
that (i) the equivalence class [>] is the set of exactly the tautologies, and
(ii) [⊥] is the set of exactly the contradictions.
Notation. We reserve the notation ‘1’ for [>], and ‘0’ for [⊥]. The reason
will be clear in the next lecture.
α ∼ β if and only if ` α ↔ β,
3
F/ ∼. As observed in Corollary 0.5, Lecture 15, Module I, the quotient
sets F/ ≡ and F/ ∼ are identical, but nevertheless, ∼ allows us to take a
syntactic look at the equivalence classes of wffs, as opposed to the semantic
look imparted by ≡.
The above deliberations on F/ ≡ (or F/ ∼) appear to suggest an alge-
braic structure for the set. Let us investigate this now. We shall find that
F/ ≡ forms what is called a Boolean algebra, and we will finally be led to
an intimate connection between the class of all Boolean algebras, and PL.
4
Module II
Lecture 2
Boolean algebras as lattices
1
We begin with the definition of a Boolean algebra viewed as an ordered
structure. In the next lecture, we shall see that there is an alternate, equiv-
alent definition of a Boolean algebra, viewed as rings.
Recall that a partial order ≤ on a set B is a binary relation that is
reflexive (x ≤ x, x ∈ B), transitive (if x ≤ y and y ≤ z then x ≤ z,
x, y, z ∈ B) and antisymmetric (if x ≤ y and y ≤ x then x = y, x, y ∈ B).
(B, ≤) forms a lattice, provided
(a) (B, ≤) is a partially ordered set,
(b) any two elements x and y of B have a least upper bound x ∨ y and a
greatest lower bound x ∧ y.
We denote the lattice by the tuple (B, ≤, ∨, ∧).
If the operations ∧ and ∨ in the lattice distribute over each other, the lattice
is said to be distributive.
If there is a least element 0 and a greatest element 1 in a lattice, it is said
to be bounded, and moreover,
if every element x of a bounded lattice has a unique complement xc such
that x ∨ xc = 1 and x ∧ xc = 0, the lattice is called complemented.
Example 0.1.
3. We also have the Boolean algebra on the set 2:= {0, 1}, viz. the struc-
ture ({0, 1}, ≤, ∨, ∧, ¬, 0, 1), or equivalently, ({F, T }, ≤, ∨, ∧, ¬, F, T ),
where F, T are the two truth values in PL and ∨, ∧, ¬ are the opera-
tions defining the truth tables.
2
In the last lecture of this module, we shall see that the Lindenbaum-
Tarski algebra and the Boolean algebra on 2 play an important role in
giving a close connection between PL and the class of all Boolean algebras.
Note. There is also the degenerate Boolean algebra over a singleton, but we
shall not consider it in this discourse.
Exercise 0.1. Prove that in each of the above examples, the structure
mentioned is indeed a Boolean algebra.
Exercise 0.2.
(i) The glb operation (∧) and lub operation (∨) are associative and
commutative.
(ii) x ∨ 0 = x, x ∧ 0 = 0, x ∨ 1 = 1, and x ∧ 1 = x.
(iii) Any non-empty finite subset {x1 , . . . , xk } of B has a glb and a
lub in B.
(iv) x ≤ y, u ≤ v imply x ∨ u ≤ y ∨ v as well as x ∧ u ≤ y ∧ v.
Proposition 0.1.
(i) De Morgan’s laws: (x ∧ y)c = xc ∨ y c ; (x ∨ y)c = xc ∧ y c .
(ii) x ≤ y if and only if y c ≤ xc .
(iii) x ≤ y c if and only if x ∧ y = 0.
3
Module II
Lecture 3
Boolean algebras as rings
1
Let us give an alternative definition of a Boolean algebra, viewed as
rings.
Example 0.1.
Definition 0.2. [α] + [β] := [¬(α ↔ β)] = [(α ∧ ¬β) ∨ (¬α ∧ β)].
2. The power set P(U ) of any non-empty set U forms a Boolean alge-
bra in the above sense: consider (P(U ), ∆, ∩, ∅, U ), where ∩ denotes
intersection in P(X), while ∆ is the operation of symmetric difference:
X∆Y := (X ∩ Y c ) ∪ (X c ∩ Y ), X, Y ⊆ U.
Exercise 0.1. Prove that in each of the above examples, the structure
mentioned is indeed a Boolean algebra.
2
Proposition 0.1.
(i) Each element is its own additive inverse: x + x = 0, for any x ∈ B.
(ii) The operation ∗ is commutative: x ∗ y = y ∗ x, for any x, y ∈ B.
Proof. Using idempotence of ∗ and group properties, we get the following.
For any x, y ∈ B, (x + y)2 = x2 + xy + yx + y 2 . So x + y = x + xy + yx + y,
which means xy + yx = 0. Now putting y = 1 we get x + x = 0, giving (i).
But then for the element xy of B we have (xy) + (xy) = 0. Uniqueness of
inverse in a group gives xy = yx.
1 An order relation
We now define a binary relation ≤ on any Boolean algebra (B, +, ∗, 0, 1)
that turns out to be a partial order on B.
Definition 1.1. For any x and y in B, x ≤ y if and only if xy = x.
Exercise 1.1. ≤ is a partial order.
Observe that in the Boolean algebra (F/ ≡, +, ∧, 0, 1), ≤ means the
following relation:
[α] ≤ [β] if and only if [α ∧ β] = [α], which is if and only if |= α → β.
Proposition 1.1.
(i) 0 is the least element and 1 is the greatest element in B with
respect to the partial order ≤, i.e. 0 ≤ x, x ≤ 1, for all x ∈ B.
(ii) Any two elements x and y of B have a glb (greatest lower bound)
in B, viz. the element xy. The glb is denoted x ∧ y.
(iii) Any two elements x and y of B also have a lub (least upper
bound) in B, viz. x + y + xy. It is denoted x ∨ y.
Exercise 1.2.
(i) ∧ and ∨ distribute over each other.
(ii) x ≤ y, u ≤ v imply xu ≤ yv. However, this property does not
hold for +, i.e. there may be x, y, u, v (not necessarily distinct!) in a
Boolean algebra (B, +, ∗, 0, 1) such that x ≤ y, u ≤ v, but x+u 6≤ y+v.
Proposition 1.2. Every element x in a Boolean algebra (B, +, ∗, 0, 1) has
a unique complement in B, viz. an element xc such that x ∨ xc = 1 and
x ∧ xc = 0.
Proof. We check which element y in B satisfies the two properties of a
complement, and then prove its uniqueness.
So assume x ∨ y = x + y + xy = 1 and xy = 0. This gives x + y = 1, so that
y = 1 + x. It is also easy to see that x ∨ (1 + x) = 1 and x(1 + x) = 0.
Thus y = 1 + x is unique, and denoted xc .
3
2 Equivalence of the two definitions of a Boolean
algebra
We now observe that the lattice-based definition of a Boolean algebra given
in the previous lecture and the ring-based definition presented here are, in
fact, equivalent.
Proof. We get the result from ≤ of Definition 1.1, Proposition 1.1, Exercise
1.2 and Proposition 1.2.
4
Module II
Lecture 4
Atoms
1
We now proceed towards proving a representation theorem for the class of
Boolean algebras. For this, let us go through the basic notion and properties
of atoms in these structures.
Example 0.1.
2
3 requires the following to be proved. For each wff α 6≡ ⊥, there is a wff β
such that (a) β 6≡ ⊥, (b) |= β → α, but (c) β 6≡ α. As α is a finite string,
it can only involve a finite number of propositional variables from P V . We
consider β := α ∧ p, where p ∈ P V does not occur in α. We have a choice
for p, as P V is infinite in this case. It is straightforward to verify (a)-(c) for
such a β.
Observe that we could use Proposition 0.1 to conclude that the algebra in
Example 0.1(2) above is atomic: we know that for finite P V , the quotient
set F/ ≡ is finite.
The following gives a special property of any atom in a Boolean algebra
(B, ≤, ∨, ∧,c , 0, 1), and we shall use it in the next lecture, in the proof of the
representation theorem.
3
Module II
Lecture 5
Homomorphisms and
Representation Theorem for finite Boolean algebras
1
1 Homomorphisms
We now move to structure-preserving maps between Boolean algebras.
Definition 1.1. (Homomorphism) Consider the Boolean algebras B :=
(B, ≤, ∨, ∧,c , 0, 1) and B 0 := (B 0 , ≤, ∨, ∧,c , 0, 1) and a map h from B to B 0 .
h is a homomorphism from B to B 0 if and only if for all x, y in B,
(i) h(x ∧ y) = h(x) ∧ h(y) and
(ii) h(xc ) = (h(x))c .
A homomorphism h is an isomorphism, if it is a bijection.
Note. We have used the same notation for the operations in B and B 0 . The
context should make the domain of the operations clear.
Exercise 1.1. For any homomorphism h from B to B 0 , prove that
h(x ∨ y) = h(x) ∨ h(y), h(0) = 0 and h(1) = 1.
Now suppose we begin with the ring-based definition of Boolean algebras
instead. What would be a characterization of homomorphisms?
Proposition 1.1. Consider a map h from B to B 0 , where B := (B, +, ∗, 0, 1)
and B 0 := (B 0 , +, ∗, 0, 1) are Boolean algebras. h is a homomorphism from
B to B 0 if and only if, for all elements x, y in B,
(i) h(x + y) = h(x) + h(y),
(ii) h(x ∗ y) = h(x) ∗ h(y), and
(iii) h(1) = 1.
Corollary 1.2. Prove that h(0) = 0.
It may be interesting to note that, in contrast, h(1) = 1 cannot be derived
from just (i) and (ii) of Proposition 1.1. Consider the power set Boolean
algebras of the sets N and Z of natural numbers and integers respectively,
and the inclusion map h from P(N ) into P(Z). h satisfies (i) and (ii), but
clearly not (iii).
Note. If we begin with the ring-based definition of Boolean algebras, a ho-
momorphism between two such Boolean algebras may be defined as a map
satisfying the conditions (i), (ii) and (iii) of Proposition 1.1.
Proposition 1.3. Let h be an onto map from B to B 0 , where B := (B, ≤,∨,
∧,c , 0, 1) and B 0 := (B 0 , ≤, ∨, ∧,c , 0, 1) are Boolean algebras.
h is an isomorphism from B to B 0 if and only if for all x, y ∈ B,
2
2 Representation theorem for finite Boolean alge-
bras
We now address representation results for Boolean algebras. These demon-
strate that the power set algebra is indeed a prototypical example of a
Boolean algebra. We prove the result for the finite case here.
3
Module II
Lecture 6
Stone’s Representation Theorem
1
Before we address the representation theorem in the general case, let
us define a couple more of basic notions in Boolean algebras. Consider a
Boolean algebra B := (B, ≤, ∨, ∧,c , 0, 1).
Definition 0.1. (Subalgebra) A subalgebra C := (C, ≤, ∨, ∧,c , 0, 1) of B
consists of a subset C of B such that C forms a Boolean algebra with respect
to the same operations and least and top elements as in B.
Clearly, ({0, 1}, ≤, ∨, ∧,c , 0, 1) and B itself, are trivial subalgebras of B. It
can be proved that a subset C of B forms a Boolean subalgebra of B, if and
only if C is closed under the operations c , ∧ and contains 0. An example
of a Boolean subalgebra of a power set Boolean algebra (P(U ), ∪, ∩,c , ∅, U )
with U infinite, is the one formed by the set of all finite and cofinite subsets
(i.e. those whose complements are finite) of U .
Definition 0.2. (Filter) Let F be a subset of B. F is a filter, if it satisfies
the following properties.
(i) F 6= ∅.
(ii) If a, b ∈ F then a ∧ b ∈ F .
(iii) If a ∈ F, a ≤ b then b ∈ F .
B is trivially a filter of itself. So we define F to be a proper filter, if it is a
proper subset of B or equivalently, if 0 does not belong to F .
F is a prime filter if it is proper and if a ∨ b ∈ F implies that either a or b
belongs to F .
By F (a), we denote the smallest filter in B that contains a, a ∈ B. It is
called the principal filter generated by a.
It can be shown that F (a) = {x ∈ B : a ≤ x}. If a 6= 0, F (a) is a proper
filter. Thus every element of a Boolean algebra gives rise to a (proper) filter.
Observe also that if F is a prime filter, then for any element a ∈ B, exactly
one of a and ac is in F .
We are now ready to present the Stone’s representation theorem – we
shall only give the main ideas of the proof. Two fundamental results about
filters in a Boolean algebra will be assumed.
(a) There is a prime filter containing any proper filter. In particular, there
is a prime filter containing F (a), for any a(6= 0) ∈ B.
(b) For any prime filter F , there is a homomorphism h : B −→ {0, 1} such
that F = {x ∈ B : h(x) = 1}.
Theorem 0.1. (Stone’s Representation Theorem) Every Boolean al-
gebra is isomorphic to a subalgebra of the power set Boolean algebra of some
set.
Proof. Consider the Boolean algebra 2 := ({0, 1}, ≤, ∨, ∧, ¬, 0, 1), and let
B := (B, ≤, ∨, ∧,c , 0, 1) be any Boolean algebra. Let H(B) denote the set
2
of all homomorphisms from B to 2. Define a map H : B −→ P(H(B)) as
follows. For any b ∈ B,
3
Module II
Lecture 7
Completeness of PL with respect to the class of
Boolean algebras
1
We have seen that 2 := ({0, 1}, ≤, ∨, ∧, ¬, 0, 1) or equivalently, ({F, T }, ≤
, ∨, ∧, ¬, F, T ), is an instance of a Boolean algebra. Moreover, for any set U ,
we know that the power set P(U ) is in bijection with the set 2U of all maps
from U to 2 := {0, 1}. One may extend the bijection to an isomorphism,
considering the power set Boolean algebra on U on the one hand, and on the
other, the Boolean algebra 2U := (2U , ≤, ∨, ∧, ¬, 0, 1). The operations in the
latter are induced by those in 2, e.g. for f, g : U −→ 2, f ∧ g is defined to
be a map in 2U such that (f ∧ g)(x) := f (x) ∧ g(x), x ∈ U , where ∧ on the
right hand side is the glb operation in 2. Given the representation theorems
presented in the previous lectures, 2 is therefore, a special Boolean algebra.
In this lecture, we shall see its importance from the standpoint of PL.
Recall the notion of semantic consequence |= in PL. A valuation is a
map from P V to 2 (equivalently, to {T, F }). 2 can now be replaced by any
Boolean algebra B, and we get the following generalized notion of semantic
consequence, with respect to the class B of all Boolean algebras.
Definition 0.1.
(i) A valuation in a Boolean algebra B := (B, ≤, ∨, ∧,c , 0, 1) is a map
v : P V −→ B. We write v |=B α to mean v(α) = 1 in B, and say v is
a model for α in B.
(ii) A wff α is true in B if every valuation v in B is a model for α in
B. We write B |= α.
(iii) α is said to be valid in the class B of all Boolean algebras, if α is
true in every Boolean algebra B of B. We write |=B α.
Using this definition of →, one may easily verify the following for any a, b, c ∈
B.
Proposition 0.1.
(i) a → (b → a) = 1.
(ii) (a → (b → c)) → ((a → b) → (a → c)) = 1.
(iii) (bc → ac ) → (a → b) = 1.
Do the expressions on the left hand sides look familiar? Indeed, these look
like the axioms A1-A3 of PL. We have, in fact, arrived at one of the main
results of this lecture, viz. soundness of PL with respect to the class B of
all Boolean algebras.
2
Theorem 0.2. (Soundness) For any wff α, if ` α then |=B α.
Proof. Let |=B α. Then α is true in every Boolean algebra, and in particular,
in 2, i.e. 2 |= α. As observed in the remark after Definition 0.1, this simply
means |= α. By completeness of PL, we have ` α.
v0 (p) := [p],
for any p ∈ P V . One can then prove by induction on the number of con-
nectives in any wff β that
v0 (β) = [β].
As v0 is also a model of α, we thus get 1 = v0 (α) = [α], i.e. [α] is the class
of all PL theorems. In other words, ` α.