This paper was converted on www.awesomepapers.org from LaTeX by an anonymous user.
Want to know more? Visit the Converter page.

G. Grätzer]http://server.maths.umanitoba.ca/homepages/gratzer/

Atom-generated planar lattices

G. Grätzer [email protected] [
(Date: Submitted Jan. 14, 2020; revised Sept. 3, 2020)
Abstract.

In this note, we discuss planar lattices generated by their atoms. We prove that if LL is a planar lattice generated by nn atoms, then both the left and the right boundaries of LL have at most n+1n+1 elements.

On the other hand, LL can be arbitrarily large. For every k>1k>1, we construct a planar lattice LL generated by 44 atoms such that LL has more than kk elements.

Key words and phrases:
lattice, planar, atom-generated.
2010 Mathematics Subject Classification:
Primary: 06B05.

1. Introduction

In this note, we deal with atom-generated planar lattices (AGP lattices, for short), that is, planar lattices (finite, by definition) generated by their sets of atoms; see Figure 1 for some small examples and Figures 24 for some larger ones.

Refer to caption

Figure 1. 33-generated AGP lattices

We have two results.

Theorem 1.

Let LL be a planar lattice generated its atoms. If LL has nn atoms, then both the left and the right boundaries of LL have at most n+1n+1 elements.

Theorem 2.

For every pair of integers k,nk,n with k>1k>1 and n>3n>3, there is a planar lattice LL with the following two properties:

  1. (i)

    LL is generated by its nn​ atoms;

  2. (ii)

    LL has more than kk​ elements.

22- and 33-generated AGP lattices are enumerated in Section 2. We prove Theorem 1 in Section 3 and Theorem 2 in Section 4.

For the basic concept and notation, see my books [6] and [7].

As usual, a planar lattice (finite, by definition) is a lattice with a planar diagram understood but not specified. For a more precise approach, see G. Czédli and G. Grätzer [5] and K. A. Baker and G. Grätzer [2].

Acknowledgement

I would like to thank Kirby Baker for his help and his contributions to this topic.

2. Small AGP lattices

We will utilize the following statement, see G. Czédli [4, 3.13A].

Lemma 3 (Left-right Inequality).

If LL is a planar lattice and a<lrba<_{\textup{lr}}b and b<lrcb<_{\textup{lr}}c in LL, then b<acb<a\vee c.

If the AGP lattice LL has 22 atoms, then LL must be isomorphic to 𝖢22=𝖡2\mathsf{C}_{2}^{2}=\mathsf{B}_{2}. The following lemma describes an AGP lattice with 33 atoms.

Lemma 4.

Let LL be an AGP lattice with 33 atoms, a1,a2,a3a_{1},a_{2},a_{3}, enumerated from left to right. Then a1a3a_{1}\vee a_{3} is the unit element of LL.

Proof.

Since LL is an AGP lattice with 33 atoms, it follows that 1=a1a2a31=a_{1}\vee a_{2}\vee a_{3}. So we obtain that 1=a1a31=a_{1}\vee a_{3} by the Left-right Inequality. ∎

If LL has 33 atoms, then by Lemma 4, there are four possibilities, up to isomorphism, see Figure 1. We see that Theorem 1 holds in all four cases.

3. Boundary chains

To prepare for the proof of Theorem 1, we need some preliminary results. In this section, let LL be an AGP lattice with its atoms a1,,ana_{1},\dots,a_{n} enumerated from left to right. For uLu\in L, define

(1) At(u)={iaiu}.\operatorname{At}(u)=\{\,i\mid a_{i}\leq u\,\}.

Let

(2) 𝖢n={12n}\mathsf{C}_{n}=\{1\prec 2\prec\dots\prec n\}

be the nn element chain.

Lemma 5.

The set At(u)\operatorname{At}(u) is an interval of 𝖢n\mathsf{C}_{n} for any uLu\in L. Moreover,

(3) At(x)At(y)\displaystyle\operatorname{At}(x)\subseteq\operatorname{At}(y)  for all xyLx\leq y\in L,
(4) At(uv)=At(u)At(v)\displaystyle\operatorname{At}(u\wedge v)=\operatorname{At}(u)\cap\operatorname{At}(v)  for all u,vLu,v\in L.
Proof.

Let uLu\in L and let 1i<jn1\leq i<j\leq n with i,jAt(u)i,j\in\operatorname{At}(u). By the Left-right Inequality, kAt(u)k\in\operatorname{At}(u) for any i<k<ji<k<j. Therefore, At(u)\operatorname{At}(u) is an interval. The other two statements are also obvious. ∎

Lemma 6.

For k{1,,n}k\in\{1,\dots,n\}, let

(5) Mk=(aikin){0}.M_{k}=\bigcup(\,\,\uparrow\!a_{i}\mid k\leq i\leq n\,)\cup\{0\}.

Then MkM_{k} is a sublattice of LL.

Proof.

By construction, MkM_{k} is closed under joins. To show that MkM_{k} is closed under meets, let x,yMkx,y\in M_{k}. Assume that xyMkx\wedge y\notin M_{k}. Since 0Mk0\in M_{k} but xyMkx\wedge y\notin M_{k}, there exists an atom ai(xy)a_{i}\in\,\downarrow\!(x\wedge y). Clearly, i<ki<k since otherwise xyaiMkx\wedge y\in\,\uparrow\!a_{i}\subseteq M_{k} would contradict that xyMkx\wedge y\notin M_{k}. By definition, we can pick jx,jy[k,,n]j_{x},j_{y}\in[k,\dots,n] such that jxAt(x))j_{x}\subseteq\operatorname{At}(x)) and jyAt(y)j_{y}\in\operatorname{At}(y). Using equation (3) of Lemma 5, we have that iAt(xy)At(x)i\in\operatorname{At}(x\wedge y)\subseteq\operatorname{At}(x) and iAt(xy)At(y)i\in\operatorname{At}(x\wedge y)\subseteq\operatorname{At}(y). Since i,jxAt(x)i,j_{x}\in\operatorname{At}(x) and i<kjxi<k\leq j_{x}, Lemma 5 gives that kAt(x)k\in\operatorname{At}(x). Similarly, kAt(y)k\in\operatorname{At}(y). Equation (4) of Lemma 5 yields that kAt(x)At(y)=At(xy)k\in\operatorname{At}(x)\cap\operatorname{At}(y)=\operatorname{At}(x\wedge y), that is, xyakMkx\wedge y\in\,\uparrow\!a_{k}\subseteq M_{k}, contradicting the assumption that xyMkx\wedge y\in M_{k}. ∎

Note an easy consequence of Lemma 6.

Corollary 7.

For k<k{1,,n}k<k^{\prime}\in\{1,\dots,n\}, let

(6) Mk,k=(aikik){0}.M_{k,k^{\prime}}=\bigcup(\,\,\uparrow\!a_{i}\mid k\leq i\leq k^{\prime}\,)\cup\{0\}.

Then Mk,kM_{k,k^{\prime}} is a sublattice of LL.

Proof.

Intersect the sublattice of Lemma 6 and its left-right reverse. ∎

Lemma 8.

Define the principal ideals

(7) Jk=(a1ak1)J_{k}=\,\downarrow\!(a_{1}\vee\cdots\vee a_{k-1})

for k=1,,nk=1,\dots,n. Then L=JkMkL=J_{k}\cup M_{k} for k=1,,nk=1,\dots,n.

Proof.

Let S=JkMkS=J_{k}\cup M_{k}. Observe that S=Jk(Mk{0})S=J_{k}\cup(M_{k}-\{0\}). Using that JkJ_{k} is a sublattice and a down-set, and also that Mk{0}M_{k}-\{0\} is an up-set, it is easy to see that SS is a sublattice of LL. ∎

Now we prove Theorem 1 in the following form.

Theorem 9.

Let LL be an AGP lattice with all its atoms a1,,ana_{1},\dots,a_{n} enumerated from left to right. Define the elements cl,1=a1c_{\textup{l},1}=a_{1}, cl,2=a1a2c_{\textup{l},2}=a_{1}\vee a_{2}, …, cl,n=a1anc_{\textup{l},n}=a_{1}\vee a_{n}. Then the chain CC:

(8) 0<cl,1cl,2cl,n=10<c_{\textup{l},1}\leq c_{\textup{l},2}\leq\dots\leq c_{\textup{l},n}=1

is a maximal chain in LL, in fact, the left boundary of LL.

Proof.

The inequalities in (8) follow from the Left-right Inequality. Let LL be as in the theorem and let us assume that CC is not maximal, that is,

(9) a1ak1<x<a1aka_{1}\vee\cdots\vee a_{k-1}<x<a_{1}\vee\cdots\vee a_{k}

for some xLx\in L. By Lemma 8, we can decompose LL as JkMkJ_{k}\cup M_{k}. By (9), xMkx\in M_{k} holds, and so xajx\geq a_{j} for some j{k,,n}j\in\{k,\dots,n\}. Therefore, [1,k][1,j]At(x)[1,k]\subseteq[1,j]\subseteq\operatorname{At}(x), and so a1akxa_{1}\vee\cdots\vee a_{k}\leq x, contradicting (9), so CC is maximal.

Since all the elements of LL on or to the right of CC form a sublattice containing all its atoms a1,,ana_{1},\dots,a_{n} generating LL (see for instance, David Kelly and Ivan Rival [8]), it follows that CC is on the left boundary of LL. ∎

Theorem 9 implies Theorem 1 by left-right symmetry.

4. Large AGP lattices with 44 atoms

Refer to caption

Figure 2. The lattice 𝖪1\mathsf{K}_{1}

Refer to caption

Figure 3. The lattice 𝖪2\mathsf{K}_{2}

Refer to caption

Figure 4. The lattice 𝖪n\mathsf{K}_{n}

To prove Theorem 2, we have to construct arbitrarily large AGP lattices with 44 atoms.

We start with the lattice 𝖪1\mathsf{K}_{1}, the ten-element lattice of Figure 2. It is planar. It is generated by its atoms; in fact, all but two of its elements are joins of atoms. The exceptions are b1b_{1} and b2b_{2}. However, b1=(a1a2)(a2a3)b_{1}=(a_{1}\vee a_{2})\wedge(a_{2}\vee a_{3}) and symmetrically for b2b_{2}.

We construct the lattice 𝖪n\mathsf{K}_{n} inductively. To illustrate the inductive step, we first construct the lattice 𝖪2\mathsf{K}_{2} of Figure 4.

We obtain 𝖪2\mathsf{K}_{2} from 𝖪1\mathsf{K}_{1} by adding six elements, α1,β1,γ1,δ1,ε1,φ1\alpha_{1},\beta_{1},\gamma_{1},\delta_{1},\varepsilon_{1},\varphi_{1}; see the black filled elements of Figure 4. Observe that 𝖪1\mathsf{K}_{1} is a meet-subsemilattice of 𝖪2\mathsf{K}_{2} and it is almost a join-subsemilattice (hence a sublattice) of 𝖪2\mathsf{K}_{2}. The exceptions are a1ca_{1}\vee c and d1cd_{1}\vee c (they both equal ii in 𝖪1\mathsf{K}_{1} and they both equal α1\alpha_{1} in 𝖪2\mathsf{K}_{2}) and symmetrically.

For the new elements, we have the following order relations:

(10) b1\displaystyle b_{1} <ε1<d1,\displaystyle<\varepsilon_{1}<d_{1}, b2\displaystyle b_{2} <φ1<d2,\displaystyle<\varphi_{1}<d_{2}, d1\displaystyle d_{1} <α1<i,\displaystyle<\alpha_{1}<i,
d2\displaystyle d_{2} <β1<i,\displaystyle<\beta_{1}<i, δ1 <φ1,\displaystyle<\varphi_{1},
and the following joins and meets:
(11) γ1 =α1β1,\displaystyle=\alpha_{1}\wedge\beta_{1}, α1 =d1c,\displaystyle=d_{1}\vee c, β1 =d2c,\displaystyle=d_{2}\vee c, γ1 =α1β1\displaystyle=\alpha_{1}\wedge\beta_{1}
ε1 =d1γ1\displaystyle=d_{1}\wedge\gamma_{1} φ1 =γ1d1\displaystyle=\gamma_{1}\wedge d_{1} 1\displaystyle 1 =ε1φ1.\displaystyle=\varepsilon_{1}\wedge\varphi_{1}.

Since 𝖪2\mathsf{K}_{2} is planar ordered set with bounds, it is a lattice.

To see that 𝖪2\mathsf{K}_{2} is generated by its atoms, first observe that the 44 changed joins in 𝖪1\mathsf{K}_{1} participate only in verifying that ii is generated by the atoms, but, of course, we have i=a1a2a3a4i=a_{1}\vee a_{2}\vee a_{3}\vee a_{4}. So all elements of 𝖪2\mathsf{K}_{2} are generated by the atoms in 𝖪2\mathsf{K}_{2}.

Now the induction step is very similar. We start with the lattice 𝖪n1\mathsf{K}_{n-1}—as illustrated by Figure 4—and extend it to the lattice 𝖪n\mathsf{K}_{n} by adding the six elements αn1,βn1,γn1,δn1,εn1,φn1\alpha_{n-1},\beta_{n-1},\gamma_{n-1},\delta_{n-1},\varepsilon_{n-1},\varphi_{n-1}, black filled in Figure 4. With obvious changes in (10) and (11), we describe 𝖪n\mathsf{K}_{n} and verify its properties.

References

  • [1] K. A. Baker, P. C. Fishburn, and F. S. Roberts, Partial orders of dimension 22. Networks 2 (1972), 11–28.
  • [2] K. A. Baker and G. Grätzer, Zilber’s Theorem for planar lattices, revisited. Acta Sci. Math. (Szeged).
  • [3] G. Birkhoff, Lattice theory. Third edition. American Mathematical Society Colloquium Publications, Vol. XXV. American Mathematical Society, Providence, R.I. 1967.
  • [4] G. Czédli, Finite convex geometries of circles. Discrete Mathematics 330 (2014), 61–75.
  • [5] G. Czédli and G. Grätzer, Planar Semimodular Lattices: Structure and Diagrams. Chapter 3 in [6].
  • [6] G. Grätzer, Lattice Theory: Foundation. Birkhäuser Verlag, Basel (2011)
  • [7] G. Grätzer, The Congruences of a Finite Lattice, A Proof-by-Picture Approach, second edition. Birkhäuser, 2016.
  • [8] David Kelly and Ivan Rival, Planar lattices. Canadian J. Math. 27 (1975), 636–665.