Canonical Boolean Algebra

• Canonical Boolean Algebra is defined as the unique dense and compact completion of a Boolean algebra, resulting in an atomic, complete structure via ultrafilter powerset representation.
• It extends Boolean homomorphisms to complete lattice homomorphisms using Stone duality and the Stone–Čech compactification, thereby preserving injectivity, surjectivity, and isomorphism.
• The framework offers both concrete and point-free constructions, enabling generalizations to l-algebras and inspiring canonical normal forms in emerging hardware applications.

Searching arXiv for relevant papers on canonical extensions of Boolean algebras and related perspectives. Canonical Boolean algebra, in the sense of canonical extension theory, denotes the unique dense and compact completion of a Boolean algebra $B$. For Boolean algebras this completion, written $B^\sigma$, is an atomic, complete Boolean algebra, and it admits a concrete representation by the powerset of the ultrafilter space of $B$. The same theory extends Boolean homomorphisms to complete-lattice homomorphisms, identifies these extensions through Stone duality and the Stone–Čech compactification, and also has a choice-free, point-free formulation via Booleanization of a free frame (González, 2017, Bezhanishvili et al., 2021).

1. Dense and compact completion

Let $B=(B,\land,\lor,\neg,0,1)$ be a Boolean algebra. A completion of the lattice $B$ is a pair $(C,e)$ in which $C$ is a complete lattice and $e\colon B\hookrightarrow C$ is a lattice embedding. The completion is called dense when every $c\in C$ can be written both as

$$c=\bigvee\{\,e(b)\mid b\in F\}, \qquad c=\bigwedge\{\,e(a)\mid a\in I\},$$

where $B^\sigma$0 ranges over filters of $B^\sigma$1 with $B^\sigma$2, and $B^\sigma$3 ranges over ideals with $B^\sigma$4. It is called compact when, whenever $B^\sigma$5 is a filter and $B^\sigma$6 an ideal of $B^\sigma$7 such that

$B^\sigma$8

then $B^\sigma$9 (González, 2017).

By the Jónsson–Tarski and Gehrke–Harding theorem, every bounded lattice has, up to unique isomorphism, a dense and compact completion. The unique dense and compact completion of $B$0 is called its canonical extension and is denoted $B$1. In the Boolean-algebra case, $B$2 is in fact an atomic, complete Boolean algebra (González, 2017).

This formulation isolates the canonical extension by two order-theoretic constraints rather than by a particular representation. A plausible implication is that the term “canonical” refers not to a preferred presentation but to uniqueness up to isomorphism under density and compactness.

2. Ultrafilter representation and extension of homomorphisms

For a Boolean algebra $B$3, let

$B$4

and define

$B$5

Here $B$6 is a complete atomic Boolean algebra, $B$7 is an embedding, and $B$8 is dense and compact. Hence

$B$9

in the Boolean case (González, 2017).

If $B=(B,\land,\lor,\neg,0,1)$0 is a Boolean homomorphism, identify

$B=(B,\land,\lor,\neg,0,1)$1

Its canonical extension $B=(B,\land,\lor,\neg,0,1)$2 is then defined for each $B=(B,\land,\lor,\neg,0,1)$3 by

$B=(B,\land,\lor,\neg,0,1)$4

This map is the unique complete-lattice homomorphism

$B=(B,\land,\lor,\neg,0,1)$5

that extends $B=(B,\land,\lor,\neg,0,1)$6, in the sense that $B=(B,\land,\lor,\neg,0,1)$7 (González, 2017).

The representation $B=(B,\land,\lor,\neg,0,1)$8 makes the canonical extension concrete: the completion is not merely abstractly complete, but explicitly realized as a powerset algebra over the Stone space of ultrafilters.

3. Stone duality and the Stone–Čech characterization

Under Stone duality, a Boolean homomorphism $B=(B,\land,\lor,\neg,0,1)$9 corresponds to the continuous map

$B$0

Because $B$1 and $B$2 are discrete, one may form their Stone–Čech compactifications

$B$3

By the universal property of $B$4, the map $B$5 extends uniquely to

$B$6

Since each $B$7 is Stone dual to $B$8, one may dualize once more to obtain

$B$9

The main characterization theorem states that

$(C,e)$0

Thus the canonical extension $(C,e)$1 of $(C,e)$2 coincides with the Boolean-dual of the continuous extension of the Stone-dual $(C,e)$3 (González, 2017).

Concretely, points of $(C,e)$4 are ultrafilters $(C,e)$5, and

$(C,e)$6

The proof proceeds by showing that both $(C,e)$7 and $(C,e)$8 are complete lattice homomorphisms extending $(C,e)$9, after which uniqueness yields equality (González, 2017).

This characterization gives a topological reading of canonical extension. The order-theoretic extension of a Boolean homomorphism is identified with a compactification-theoretic construction on the dual discrete ultrafilter spaces.

4. Preservation properties and illustrative cases

If $C$0 is a Boolean homomorphism, then the canonical extension preserves the basic categorical properties of injectivity and surjectivity. Specifically:

• If $C$1 is injective, then $C$2 is injective.
• If $C$3 is surjective, then $C$4 is surjective.
• Consequently, if $C$5 is an isomorphism then so is $C$6 (González, 2017).

The proof uses the dual diagram involving $C$7 and $C$8, together with the fact that a one-to-one, respectively onto, map between discrete spaces extends to a one-to-one, respectively onto, map under Stone–Čech compactification (González, 2017).

Several examples clarify the scope of the construction. If $C$9 is finite, then $e\colon B\hookrightarrow C$0 is finite and discrete, so $e\colon B\hookrightarrow C$1. Hence

$e\colon B\hookrightarrow C$2

and similarly every $e\colon B\hookrightarrow C$3. For finite algebras, canonical extension is therefore trivial (González, 2017).

If $e\colon B\hookrightarrow C$4 is an infinite Boolean algebra and $e\colon B\hookrightarrow C$5, then the inclusion $e\colon B\hookrightarrow C$6 induces a restriction map $e\colon B\hookrightarrow C$7 that is a dense embedding of discrete spaces. The corresponding map $e\colon B\hookrightarrow C$8 is an isomorphism, recovering the fact that completing twice does nothing new (González, 2017).

A further example takes $e\colon B\hookrightarrow C$9 and lets $c\in C$0 “project away” one point. Then $c\in C$1 forgets one ultrafilter point of the discrete space $c\in C$2, and $c\in C$3 recovers the usual extension on the power-set algebras (González, 2017).

5. Point-free construction by free frames and Booleanization

A point-free approach begins with the poset

$c\in C$4

ordered by inclusion and equipped with the Alexandroff topology, in which

$c\in C$5

In parallel, one considers

$c\in C$6

ordered by reverse inclusion: $c\in C$7 With this order, $c\in C$8 is a bounded meet-semilattice (Bezhanishvili et al., 2021).

Let

$c\in C$9

The free frame on a meet-semilattice with top is isomorphic to the frame of down-sets of $$c=\bigvee\{\,e(b)\mid b\in F\}, \qquad c=\bigwedge\{\,e(a)\mid a\in I\},$$0. Its Booleanization is

$$c=\bigvee\{\,e(b)\mid b\in F\}, \qquad c=\bigwedge\{\,e(a)\mid a\in I\},$$1

where

$$c=\bigvee\{\,e(b)\mid b\in F\}, \qquad c=\bigwedge\{\,e(a)\mid a\in I\},$$2

The Booleanization $$c=\bigvee\{\,e(b)\mid b\in F\}, \qquad c=\bigwedge\{\,e(a)\mid a\in I\},$$3 is a complete Boolean algebra, and the inclusion $$c=\bigvee\{\,e(b)\mid b\in F\}, \qquad c=\bigwedge\{\,e(a)\mid a\in I\},$$4 preserves all meets and joins (Bezhanishvili et al., 2021).

If $$c=\bigvee\{\,e(b)\mid b\in F\}, \qquad c=\bigwedge\{\,e(a)\mid a\in I\},$$5 denotes the canonical insertion of generators and

$$c=\bigvee\{\,e(b)\mid b\in F\}, \qquad c=\bigwedge\{\,e(a)\mid a\in I\},$$6

then

$$c=\bigvee\{\,e(b)\mid b\in F\}, \qquad c=\bigwedge\{\,e(a)\mid a\in I\},$$7

Moreover,

$$c=\bigvee\{\,e(b)\mid b\in F\}, \qquad c=\bigwedge\{\,e(a)\mid a\in I\},$$8

so $$c=\bigvee\{\,e(b)\mid b\in F\}, \qquad c=\bigwedge\{\,e(a)\mid a\in I\},$$9 is already complemented in $B^\sigma$00, and $B^\sigma$01 is well-defined (Bezhanishvili et al., 2021).

The resulting theorem states that $B^\sigma$02 is a canonical extension of $B^\sigma$03: $B^\sigma$04 is a Boolean-algebra embedding, every element of $B^\sigma$05 is a join of meets of elements from $B^\sigma$06, and compactness holds in the finitary form that if $B^\sigma$07, then already $B^\sigma$08 for some finite $B^\sigma$09 (Bezhanishvili et al., 2021).

This construction is point-free because it lives entirely in the world of frames and Booleanizations. It is also identified with the regular-open presentation: $B^\sigma$10 and the map $B^\sigma$11 coincides with

$B^\sigma$12

viewed as a regular open set. In this form, one recovers the choice-free construction of Gehrke–Holliday (Bezhanishvili et al., 2021).

6. Generalizations and distinct uses of canonical form

The point-free pattern extends beyond Boolean algebras to the category $B^\sigma$13 of bounded archimedean $B^\sigma$14-algebras. For $B^\sigma$15, let $B^\sigma$16 be the set of archimedean $B^\sigma$17-ideals of $B^\sigma$18, ordered by inclusion. Banaschewski’s theorem shows that $B^\sigma$19 is a compact regular frame. Writing

$B^\sigma$20

forming $B^\sigma$21, and then passing to the Specker algebra $B^\sigma$22 and its Dedekind completion, one obtains a natural embedding

$B^\sigma$23

and

$B^\sigma$24

is a canonical extension of $B^\sigma$25 in the category of bounded archimedean $B^\sigma$26-algebras (Bezhanishvili et al., 2021).

A terminological distinction is also necessary. In a different line of work, “canonical” refers not to canonical extension but to canonical normal forms for asymmetric basis logic. There, the primitive operations are IAND,

$B^\sigma$27

and IMPLY,

$B^\sigma$28

Because $B^\sigma$29 and $B^\sigma$30 are each functionally complete, any Boolean function can be cast in canonical forms such as sum-of-IANDs (SOI), canonical NAND of IMPLYs (NOI), IAND-of-ORs (IOS), and IMPLY-of-NANDs (ION) (Vyas et al., 2024).

This is a different use of canonical structure. In canonical extension theory, the central object is the unique dense and compact completion of an algebra; in asymmetric basis logic, the central object is a canonical form for logic synthesis in emerging memristive and spintronic hardware. The latter framework introduces fundamental identities, theorems, and canonical normal forms tailored to IAND and IMPLY, and a previously proposed modified Karnaugh map method based on a subset of those principles demonstrated a 28% reduction in computational steps for an algorithmically designed memristive full adder (Vyas et al., 2024).