10
$\begingroup$

It is well known that a lattice is distributive iff it excludes as a sublattice $N_5$ (the pentagon) and $M_3$ (three unordered elements with a top and bottom). Further, a lattice that only excludes $N_5$ is modular, and consequently a lattice excluding $N_5$ but including $M_3$ is modular but not distributive.

What, if anything, is known about the class of lattices that excludes $M_3$? Does it relate to classes of structures of any particular interest? (Note this is not the same as asking about "nonmodular varieties", since those are varieties which include $N_5$).

$\endgroup$
1
  • $\begingroup$ I found this question while looking for any studied/natural classes of lattices which are not distributive, and which also do not contain $M_3$ (i.e. not necessarily the class all lattices which avoid $M_3$). For anyone else looking for this type of thing: two classes of lattices with this property are join semi-distributive lattices, and meet-distributive lattices. See Paul H. Edelman, Meet-Distributive Lattices and the Anti-Exchange Closure, or for more information one could start at arxiv.org/pdf/1810.01528.pdf $\endgroup$ Oct 12, 2023 at 4:56

2 Answers 2

15
$\begingroup$

This is a (slightly edited) copy of the answer I posted on math.SE to the question What do we call a lattice that does not have a sublattice the shape of the diamond $M_3$?:

Let $\mathbf K$ be the class of all lattices not containing $M_3$ (the $5$-element nondistributive modular lattice) as a sublattice; in other words, lattices in which every modular sublattice is distributive. It is easy to see that $\mathbf K$ can be characterized as the class of all lattices satisfying the following sentence $\varphi$:$$\forall u\forall v\forall a\forall b\forall c[(ab=ac=bc=u)\wedge(a+b=a+c=b+c=v)\rightarrow(u=v)]$$or equivalently$$\forall a\forall b\forall c[(ab=ac=bc)\wedge(a+b=a+c=b+c)\rightarrow(ab=a+b)]$$ or $$\forall a\forall b\forall c[(ab=ac=bc)\wedge(a+b=a+c=b+c)\rightarrow(a=b)].$$ Inasmuch as $\varphi$ is a universal Horn sentence, it follows that $\mathbf K$ is closed under taking sublattices, direct products and reduced products; it is a quasivariety.

On the other hand, $\mathbf K$ can not be characterized by identities, and is not closed under taking homomorphic images. Bjarni Jónsson [Sublattices of a free lattice, Canad. J. Math. 13 (1961), 256-264, Lemma 2.6(i)] observed that, as a corollary of P. M. Whitman's work, elements $u,a,b,c$ of a free lattice satisfy the condition: if $u=ab=ac$, then $u=a(b+c)$. It follows that $M_3$ is not a sublattice of a free lattice, i.e, the class $\mathbf K$ contains all free lattices. Thus every lattice is a homomorphic image of a member of $\mathbf K$, and every identity which holds in all $M_3$-free lattices is a consequence of the lattice axioms. In view of this, we are not likely to find a much simpler algebraic characterization of $M_3$-free lattices than the one in the previous paragraph.

$\endgroup$
2
  • $\begingroup$ This is very nice, thanks! I have another argument that follows from covers of varieties that shows this is not a variety, which is why I wrote "class" and not "variety". I was wondering if it had been considered in any literature nonetheless. I'll leave this open in case anyone has an answer to that, but I do appreciate this answer as nice and clean. $\endgroup$
    – Gershom B
    May 2, 2020 at 1:00
  • 2
    $\begingroup$ Not a variety, but it is a quasivariety. $\endgroup$
    – bof
    May 2, 2020 at 1:49
10
$\begingroup$

As bof says, the class $\mathcal K$ of lattices omitting $\mathbf M_3$ as a sublattice is a quasivariety that is not a variety. As bof also says, this implies that $\mathcal K$ is closed under the formation of sublattices, products, and reduced products, but the class is not closed under the formation of homomorphic images. But $\mathcal K$ is closed under some types of homomorphic images.

A. (Projection onto a direct factor) If $L\times L'\in \mathcal K$, then $L, L' \in \mathcal K$. Every quasivariety of lattices has this property, but there exist quasivarieties of algebraic structures that do not have this property.

B. (Quotients of finite members) If $L\in \mathcal K$ is finite, then any homomorphic image of $L$ is also in $\mathcal K$. This property typically fails for quasivarieties, even quasivarieties of lattices. It holds for $\mathcal K$ because $\mathbf M_3$ is projective in the class of finite lattices.

$\endgroup$
1
  • $\begingroup$ I guess A holds because $L$ and $L'$ are sublattices of $L\times L'$ because lattices have trivial sublattices. B seems harder. Is there a literature about $\mathcal K$? Does it have a name? $\endgroup$
    – bof
    May 2, 2020 at 3:02

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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