Fix an arbitrary complete lattice $\mathfrak{A}$ with order $\sqsubseteq$. I call elements $a,b\in\mathfrak{A}$ intersecting and denote $a\not\asymp b$ iff there is a non-least element $c\in\mathfrak{A}$ such that $c\sqsubseteq a \wedge c\sqsubseteq b$.

I call full star of element $a$ the set $\star a=\{c\in\mathfrak{A} \,|\, c\not\asymp a\}$.

I call a complete free star such a subset $S\subseteq\mathfrak{A}$, not containing the least element, that $\bigsqcup T \in S \Leftrightarrow T \cap S \neq \emptyset$ for every set $T\subseteq\mathfrak{A}$ (by $\bigsqcup T$ I denote the join (=supremum) of the set $T$.)

Now we can define principal elements of the complete lattice $\mathfrak{A}$:

An element $a\in\mathfrak{A}$ is principal iff its full star $\star a$ is a complete free star.

Question: Has anyone researched principal elements of complete lattices before me today?

Note: I call such elements principal because for the lattice $\mathfrak{F}$ of filters on a set (ordered reverse to set-theoretic inclusion of the filters), this coincides with the customary definition of principal filters.

Moreover, my idea can be generalized from complete lattices to arbitrary posets just replacing the formula $\bigsqcup T \in S \Leftrightarrow T \cap S \neq \emptyset$ with more general formula $$\forall Z \in \mathfrak{A}: ( \forall X \in T : Z \sqsupseteq X \Rightarrow Z \in S) \Leftrightarrow T \cap S \neq \emptyset.$$

Has anyone researched this case (for arbitrary posets, not just complete lattices) before me?

Hm, maybe the set of principal elements of a (distributive) lattice is the same as the center of the lattice?

  • $\begingroup$ You don't use $b$ in the definition of "intersecting elements". $\endgroup$ – Olivier Bégassat Jun 21 '14 at 16:04
  • $\begingroup$ @OlivierBégassat: Thanks, corrected $\endgroup$ – porton Jun 21 '14 at 16:40
  • $\begingroup$ Isn't it the case that $a$ and $b$ are intersecting iff their meet is different from the bottom element? $\endgroup$ – Olivier Bégassat Jun 21 '14 at 16:46
  • $\begingroup$ @OlivierBégassat: Yes, but only for lattices with a bottom element, not for arbitrary posets. My definition is valid for any posets $\endgroup$ – porton Jun 21 '14 at 17:26
  • $\begingroup$ It's just that your question is phrased in terms of complete lattices, and those have top and bottom elements. $\endgroup$ – Olivier Bégassat Jun 21 '14 at 19:09

To avoid confusion, you should provide a definition of a filter. I'm using the more general definition of an order filter that is sometimes called an upset. But your suggestion looks not very promising to me as your can see below.

Here, a filter $F⊆\mathfrak A$ is always an order filter: $$x∈F⇒(∀y∈\mathfrak A:x≤y⇒y∈F)$$

In any lattice the equivalence $c\sqsubseteq c\wedge a ⇔ c \sqsubseteq a$ holds. Thus, $a≭b$ iff there exists a principal filter that contains $a$ and $b$ that is not generated by a minimal element. This is some kind of orthogonality relation. It reminds me to the orthogonality of $\ell$-groups and to residuated lattices.

$*a$ is the union of all principal filters of the atoms below $a$ (in case of their existence) or the infinitely descending chains just above the minimal elements.

A complete free star $S$ is an order filter as for $a,b∈S$ the relations $\{a,b\}⊆\mathfrak A$ and $a\vee b∈S$ hold (for arbitrary orders consider $a≤b$ and $a∈S$). Furthermore if $a\vee b∈S$ then $a∈S$ or $b∈S$ holds. This is a prime filter or – in the dual order – a prime ideal. Furthermore, it is related to irreducible elements. In a finite lattice the set $S$ is the union of principal filters of $\vee$-irreducible elements. So it is something like an orthogonality relation, again.

So in a lattice every element except $0$ is principal. In an arbitrary ordered set an element $x$ is principal if the filter $↑(↓x\setminus \min\mathfrak A)$ generated by its principal ideal $↓x$ without the minimal elements is something like a prime filter i.e. if each upper bound of any nontrivial subset of its complement $\mathfrak A\setminus \bigl(↑(↓x\setminus \min\mathfrak A)\bigr)$ is in that complement. For this consideration the don't play an important role. It's much easier if you keep them in the filters.

BTW.: Have you ever heard of Formal Concept Analysis? Their view of lattice theory might be helpful for you.

  • $\begingroup$ "So in a lattice every element except $0$ is principal." It is wrong: non-principal filters are non-principal in this sense. It seems you confuse for-all-filters with for-all-principal-filters or for-all-ultrafilters. I haven't check the details, but I am sure in the lattice of filters on a set, not only $0$ is principal $\endgroup$ – porton Jun 22 '14 at 20:27
  • $\begingroup$ In fact a non-principal filter may be non-principal in a certain lattice. But that's true for every element of any lattice and out of the scope of this question. If you have a counterexample then you should revise your definitions. You don't provide a definition of a filter. There is a common definition of ultrafilters in topology, but not in order theory. So you are probably the one who confuses different notions. Regarding your last sentence, $0$ is not principal by your definition. I came to the conclusion that all other elements are principal. So where is the contradiction? $\endgroup$ – Keinstein Jun 24 '14 at 20:37

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.