Let $X$ be a set and let $\mathcal A$ be a non-empty collection of subsets of $X$.

For every $x\in X$ define $\mathcal A_x:=\{A\in\mathcal A\mid x\in A\}$.

Now let it be that collection $\{\mathcal A_x\mid x\in X\}\subseteq\mathcal P(\mathcal A)$ is an algebra on $\mathcal A$ in the sense that the collection is closed under intersection and complementation.

Also let it be that the function $\phi:X\to\mathcal P(\mathcal A)$ prescribed by $x\mapsto \mathcal A_x$ is injective.

Then $(X,\leq)$ where by definition $x\leq y\iff \mathcal A_x\subseteq\mathcal A_y$ can be recognized as a Boolean algebra and function $\phi$ mentioned above as an isomorphism. Also every $A\in\mathcal A$ appears to be an ultrafilter of $(X,\leq)$.

Now my question:

Is $\mathcal A$ necessarily the collection of all ultrafilters of $(X,\leq)$?

Remark1: it is common to start with a Boolean algebra in order to show a correspondence between this algebra and its set of ultrafilters. In the setup above things are turned around.

Remark2: I added tags "logic" and "predicate-logic" because actually this question arose when I tried to construct a Boolean algebra on (equivalence classes of) $\mathcal L$-formulas by the use of functions that split up all $\mathcal L$-formulas in true formulas and false formulas. These functions correspond with pairs $(\mathfrak A,\sigma)$ where $\mathfrak A$ is an $\mathcal L$-structure and $\sigma$ is an assignment. They provide ultrafilters but I want to know: do they provide all ultrafilters?

  • $\begingroup$ The statement that all ultrafilters on $\mathcal{L}$-formulas come from a pair $(\mathfrak A,\sigma)$ is just a rephrasing of the completeness theorem. (It's just saying every consistent theory has a model, where you enlarge the language by adding constants for each variable you use in your formulas.) $\endgroup$ – Eric Wofsey Aug 30 '20 at 14:24

No. For instance, let $X$ be any Boolean algebra, let $S$ be the Stone space of $X$, and let $\mathcal{A}\subseteq S$. Your map $\phi:X\to\mathcal{P}(\mathcal{A})$ is then a Boolean homomorphism, and identifying $X$ with the algebra of clopen subsets of $S$, it is just the map that takes a clopen subset of $S$ and intersects it with $\mathcal{A}$. This makes it clear that $\phi$ is injective iff $\mathcal{A}$ is dense in $S$, so that it intersects every nonempty clopen set. So, $\mathcal{A}$ need not be all the ultrafilters on $X$, but only a dense subset of them.

(Note that every example is in fact isomorphic to one of the form above, since instead of starting with $\mathcal{A}$, you can start with the induced Boolean algebra structure on $X$ and then take $\mathcal{A}$ as a set of ultrafilters on $X$.)

For a very explicit example, you could take $X$ to be the power set of an infinite set $Y$, and $\mathcal{A}$ to be just the principal ultrafilters on $Y$. Then $\phi:\mathcal{P}(Y)\to\mathcal{P}(\mathcal{A})$ is just the isomorphism induced by the obvious bijection between $Y$ and $\mathcal{A}$.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.