# Axioms for a Heyting Algebra as a Set System (Partial Order Lattice Under Inclusion)

According to the corresponding section in Wikipedia:

An element $x$ of a Heyting algebra $H$ is called regular iff $x = \neg y$ for some $y \in H$.

Elements $x$ and $y$ of a Heyting algebra are called complements to each other iff $x \land y = 0$ and $x \lor y =1$. If it exists, any such $y$ must be unique and in fact equal to $\neg x$. If $x$ admits a complement then we call it complemented.

For any Heyting algebra, the following three conditions are equivalent:

1. $H$ is a Boolean algebra
2. Every element of $H$ is regular.
3. Every element of $H$ is complemented.

My question is the following:

If we consider this as a set system, then is a Heyting algebra the same as a Boolean algebra except that it does not necessarily need to be closed under set complements?

My reasoning is as follows:

1. The terminology complemented, complement, and set complement are quite similar.

2. If we change the lattice notation to the corresponding set notation ($\neg \iff ^c$, $\land \iff \cap$, $\lor \iff \cup$ ), then given a set $A$, and considering a Heyting algebra $H$ on its lattice of subsets, we have that $X \subset A$ is regular iff $X=Y^c$ for some $Y \in H, Y \subset A$, $X$ and $Y$ are complements if and only if $X \cap Y=\emptyset$, $X \cup Y=A \implies X = Y^c$. The above excerpt implies that if for every $X \in H$, $X=Y^c$ for some $Y\in H$ (i.e. $H$ is closed under set complements) then it is a Boolean algebra (since it is a lattice and hence closed under finite intersections and unions).

3. The same way that Boolean algebras are the algebraic model for classical logic, Heyting algebras are the algebraic model for intuitionistic logic, which rejects the Law of the Excluded Middle. Not necessarily being closed under set complements seemed to be the set-theoretic analog to the Law of the Excluded Middle not holding (because then $(X^c)^c$ is not necessarily defined, which means that we can not immediately conclude that $X=(X^c)^c$, the same way in intuitionistic logic we cannot always conclude that $\neg \neg x =x$).

• But you must bear in mind that the complement in BA, and also the pseudo-complement in HA are completely determined by the lattice structure, as $x' = \max \{ u : x \wedge u = 0 \}$. – amrsa Jul 25 '16 at 9:31

Yes and no.

Yes, in the sense that a complement of $x$ in a bounded lattice is a (necessarily unique) element $y$ with $x\wedge y = \bot$ and $x\vee y = \top$ and these obviously need not exist for all $x$ of some Heyting algebra.

No, in the sense that $\neg x$ in a Heyting algebra does not denote a complement of $x$, but a pseudocomplement, i.e. $\neg x := x \Rightarrow \bot$. The power set of a constructive set is surely closed under pseudocomplements. It just so happens that pseudocomplements are not necessarily complements, as $x\vee \neg x < \top$ may happen.

EDIT: Here is a classical counterexample (I think, I have not tried or seen this before):

Take the standard topology $T$ on $\mathbb{R}$. This is a Heyting algebra with "implication" $A \to B := \operatorname{int}(A^c \cup B)$, hence $\neg A = A \to \emptyset = \operatorname{int}(A^c)$. Let $A:=(0\,.. 1)$. Then:

$$\neg A = \operatorname{int}((-\infty\,.. 0] \cup [1\,.. \infty)) = (-\infty\,.. 0) \cup (1\,..\infty)$$

Hence:

$$A\cup \neg A = \mathbb{R} \setminus \{0,1\} \subsetneq \mathbb{R}$$

• This makes sense. Does $x\vee \neg x < \top$ ever happen for (truth) sets? I am struggling to think of an example, although probably because I am implicitly assuming the law of the excluded middle. – Chill2Macht Sep 12 '16 at 20:05
• @William I'm not sure what you mean. Do you mean powersets of constructive sets in some constructive theory? I'm afraid I can't help you with this in particular because I would need to explicitely find such a set in a theory, which contradicts LEM. -- Otherwise you can try to find such an example within classical mathematics by looking at various topologies I suppose. – Stefan Perko Sep 13 '16 at 12:28
• To be honest I probably can't clarify further -- I don't actually know what constructive sets or constructive theories are. Anyway I think I understand your point: $\neg$ corresponds to pseudocomplements, not necessarily to complements (only if the element $x$ is complemented), so considered as a set system, a Heyting algebra would be a "Boolean algebra except that some elements are not regular equivalently some elements are not complemented" so not being complemented might correspond to the set complement not being in the set, but not being regular might also just mean that the union of – Chill2Macht Sep 13 '16 at 15:58
• @William I've tried adding a classical example. Perhaps that helps. – Stefan Perko Sep 13 '16 at 19:02
• @William Sure, my pleasure ;) – Stefan Perko Sep 13 '16 at 19:15