Lattices: An Introduction

Felix Gotti
[email protected]

UC Berkeley

Student Combinatorics Seminar

Felix Gotti [email protected] Lattices: An Introduction

Outline

General Lattices

Modular Lattices

Distributive Lattices

Felix Gotti [email protected] Lattices: An Introduction

Joins and Meets
Definition (Joins and Meets)
Let P be a poset.
I If S ⊆ P a join (or supremum) of S, denoted by
_
s,
s∈S

is an element of u ∈ P that is an upper bound of S satisfying

that if u 0 is any other upper bound of S, then u ≤ u 0 .
I The definition of a meet (or infimum) of S ⊆ P, denoted by
^
s,
s∈S

is dual to the definition of join.

Remark: Note that if a join (resp., meet) exists then it is unique.
Felix Gotti [email protected] Lattices: An Introduction
Definition of Lattice
Definition
I A join-semilattice (resp., meet-semilattice) is a poset such
that any pair of elements have a join (resp., meet).
I A lattice is a poset that is both a join-semilattice and a
meet-semilattice.
I If L is a lattice and S ⊂ L such that r ∨ s, r ∧ s ∈ S for all
r , s ∈ S, we say that S is a sublattice of L.
Example of lattices:
I Every totally ordered set is a lattice.
I If L and M are lattices, so are L∗ , L ⊕ M, and L × M. While
L + M is not a lattice, at least L or M is empty,
(L + M) ∪ {0̂, 1̂} is always a lattice.
I The lattices L and M are sublattices of L ⊕ M and
(L + M) ∪ {0̂, 1̂}.
Felix Gotti [email protected] Lattices: An Introduction
Complete Lattices

Theorem
If L is a lattice, the join (resp., the meet) of any finite subset of L
exists.
Proof: It follows by induction.

Remark:
I If L is a lattice and S ⊆ L is an infinite subset, the join (resp.,
meet) of S might not exists. Consider an open interval of R.
I A finite lattice always contains 0̂ and 1̂.

Definition
A lattice L is said to be complete if every subset of L has a join
and a meet.

Felix Gotti [email protected] Lattices: An Introduction

Lattices (continuation)

Theorem
A finite join-semilattice containing 0̂ is a lattice.
Sketch of proof: Let L be a join-semilattice with 0̂. For u, v ∈ L
consider the set S = { s ∈ L | s ≤ u and s ≤ v }. Take m to be the
join of the finitely many elements of S. Check that m = u ∧ v .

Remark: The above theorem fails when the join-semilattice is not

finite. Example?

Felix Gotti [email protected] Lattices: An Introduction

Morphism of Lattices

Definition
Let ϕ : L → L0 be map between lattices.
I ϕ is a lattice homomorphism if ϕ(r ∨ s) = ϕ(r ) ∨ ϕ(s) and
ϕ(r ∧ s) = ϕ(r ) ∧ ϕ(s) for all r , s ∈ L.
I A lattice homomorphism ϕ : L → L0 is said to be an
isomorphism if it is a bijection. In this case, we say that the
lattices L and L0 are isomorphic.
Remarks:
I It is easy to check that a homomorphism of lattices is an
order-preserving map.
I A homomorphism of lattices ϕ : L → L0 is an isomorphism iff
there exists a homomorphism of lattices ψ : L0 → L such that
ϕ ◦ ψ = IdL0 and ψ ◦ ϕ = IdL .

Felix Gotti [email protected] Lattices: An Introduction

Finite Semimodular Lattices
Theorem
For a finite lattice L, the following conditions are equivalent:
1. If r , s ∈ L both cover r ∧ s then r ∨ s covers both r and s.
2. L is graded, and the rank function ρ of L satisfies
ρ(r ) + ρ(s) ≥ ρ(r ∨ s) + ρ(r ∧ s) for all r , s ∈ L.
Sketch of proof: (2) =⇒ (1): If r and s cover r ∧ s then
ρ(r ) = ρ(s) = ρ(r ∧ s) + 1 and ρ(r ∨ s) > ρ(r ) = ρ(s). Apply
inequality (2).
(1) =⇒ (2): If L is not graded there is a nongraded interval [u, v ]
with minimal length. Take r1 , r2 covering u such that [r1 , v ] and
[r2 , v ] are both graded with different lengths. The saturated chains
ri < r1 ∨ r2 = t1 < · · · < tn = v have lengths n. Contradiction.
If the inequality in (2) does not hold, take r , s ∈ L with
(`(r ∧ s, r ∨ s), ρ(r ) + ρ(s)) minimal (lexicographically) such that
ρ(r ) + ρ(s) < ρ(r ∧ s) + ρ(r ∨ s). If s ∧ r < s 0 < s, take the pair
(R, S) = (s 0 ∨ r , s) to contradict the minimality of the pair (r , s).
Felix Gotti [email protected] Lattices: An Introduction
Finite Semimodular Lattices

Definition (Modular Lattice)

I A lattice satisfying any of the above conditions is said to be

finite upper semimodular. A finite lattice is called lower
semimodular if its dual is upper semimodular.
I A lattice L is modular if it is upper and lower semimodular.
Example of modular lattices:
1. For every n ∈ N, the lattice [n] is modular.
2. If S is finite then P(S) is modular.
Q
3. For which sets S is the lattice S modular?
4. Give an example of lower semimodular lattice that is not
upper semimodular.

Felix Gotti [email protected] Lattices: An Introduction

A Characterization of Modular Lattices
Theorem (Characterization of Modular Lattices)
A lattice L is modular iff r ∨ (t ∧ s) = (r ∨ t) ∧ s for all r , s, t ∈ L
such that r ≤ s.
Sketch of proof: (Sufficiency) Since r ≤ s we have
r ∨ (t ∧ s) ≤ (r ∨ t) ∧ s. Using the modularity condition we can
verify that ρ(r ∨ (t ∧ s)) = ρ((r ∨ t) ∧ s). Hence
r ∨ (t ∧ s) = (r ∨ t) ∧ s, as desired.
(Necessity) Take r , s ∈ L both covering r ∧ s. Take u ∈ L such that
r ≤ u < r ∨ s, and let v = u ∧ s. Since r ∧ s ≤ v ≤ s and v 6= s
(otherwise r ∨ s ≤ u), we have v = r ∧ s. Using r ≤ u, we get

r = r ∨ (r ∧ s) = r ∨ (s ∧ u) = (r ∨ s) ∧ u = u.

Hence L is upper semimodular. Dualizing we have

r ∨∗ (t ∧∗ s) = (r ∨∗ t) ∧∗ s for all r , s, t ∈ L∗ with r ≤∗ s. Hence
L is also lower semimodular, and so a modular lattice.
Felix Gotti [email protected] Lattices: An Introduction
Complemented and Atomic Lattices
Definition (Complemented Lattice)
A lattice L having 0̂ and 1̂ is said to be complemented if for all
r ∈ L there exists a complement s ∈ L, meaning r ∧ s = 0̂ and
r ∨ s = 1̂. If each t ∈ L has a unique complement we say that L is
uniquely complemented.
Example of complemented lattices: For any set S its power is
(uniquely) complemented. A totally ordered set T is
complemented iff |T | ≤ 2. When is Dn complemented?
Definition
If L is a finite lattice with 0̂, an element r ∈ L is an atom if it
covers 0̂. If every element of L is the join of atoms, then L is said
to be atomic. Dually, we can define coatom and coatomic lattice.
Example of atomic lattices: P(S) is atomic. [n] is atomic iff
n ≤ 2.
Felix Gotti [email protected] Lattices: An Introduction
Finite Geometric Lattice
Definition
If for all r , s ∈ L such that r ≤ s, the interval [r , s] is itself
complemented, we say that L is relatively complemented.

Theorem
For a finite upper semimodular lattice L the following conditions
are equivalent.
1. L is atomic.
2. L is relatively complemented.
Proof: Omitted.

Definition (Finite Geometric Lattice)

A finite semimodular lattice satisfying the above conditions is
called finite geometric lattice.
Felix Gotti [email protected] Lattices: An Introduction
Distributive Lattices

Definition (Distributive Lattice)

A lattice L is said to be distributive if the following conditions
hold:
I r ∧ (s ∨ t) = (r ∧ s) ∨ (r ∧ t) for all r , s, t ∈ L;
I r ∨ (s ∧ t) = (r ∨ s) ∧ (r ∨ t) for all r , s, t ∈ L.

Remark: Both conditions in the above definition are equivalent.

Felix Gotti [email protected] Lattices: An Introduction

Distributive Lattices (Examples)

Examples of distributive lattices:

1. Every totally ordered set is a distributive lattice.
2. For a set S its power set, P(S) is a distributive lattice.
3. For n ∈ N, the lattice Dn is distributive.
Q
4. If n ≥ 3 the lattice n is NOT distributive.

Felix Gotti [email protected] Lattices: An Introduction

Distributive Lattices (continuation)
Theorem
For a distributive lattice L, the following conditions hold.
1. L is modular.
2. The complement of each element, if exists, must be unique.
Sketch of proof:
1. For r , t, s ∈ L such that r ≤ s we have

r ∨ (t ∧ s) = (r ∨ t) ∧ (r ∨ s) = (r ∨ t) ∧ s.

Therefore, by the characterization theorem of modular

lattices, part (1) follows.
2. Let s such that r , r 0 ∈ P are two complements of s. Then

r = r ∨((s∧r )∨(s∧r 0 )) = r ∨(s∧(r ∨r 0 )) = (r ∨s)∧(r ∨r 0 ) = r ∨r 0 ,

and so r 0 ≤ r . Similarly r ≤ r 0 . Hence r = r 0 .

Felix Gotti [email protected] Lattices: An Introduction

Antichains and Order Ideals

Definition
Let P be a poset.
1. A subset A of P is an antichain if any two distinct elements of
A are incomparable.
2. A subset I of P is an order ideal if s ∈ I and r ≤ s implies
that r ∈ I .

Theorem
Let P be a finite poset. There is a bijection between the set of
antichain and the set of order ideals of P.
Sketch of proof: Assign to the order ideal I the antichain AI
consisting of all maximal elements of I . Conversely, assign to the
antichain A the order ideal IA := { s ∈ P | s ≤ a for some a ∈ A }.

Felix Gotti [email protected] Lattices: An Introduction

The Lattice of Order Ideals

Definition
Let P be a poset.
1. An order ideal I is generated by the antichain A, if I = IA .
2. If I is generated by {s} it is called principal and denoted by Λs .
3. The set of all order ideals of P is denoted by J(P).

Theorem
If P is a poset J(P) is a distributive lattice.
Sketch of proof: J(P) is a poset under inclusion. If I and J are
order ideals of P then so are I ∩ J and I ∪ J; therefore J(P) is a
lattice. Since intersection and union of sets distribute with each
other, J(P) is distributive.

Felix Gotti [email protected] Lattices: An Introduction

Join-irreducible Elements

Definition
Let L be a lattice and s ∈ L. We call s join-irreducible if s 6= 0̂ and
s is not the join of two strictly smaller elements.

Theorem
Let P be a finite poset.
1. An order ideal I of P is join-irreducible in J(P) iff it is
principal.
2. The set of join-irreducible of J(P), considered as a subposet
of J(P), is isomorphic to P. Hence J(P) ∼ = J(Q) iff P ∼
= Q.
Proof: Straightforward.

Felix Gotti [email protected] Lattices: An Introduction

Fundamental Theorem of Finite Distributive Lattices

Theorem (Fundamental Theorem of FDL)

Let L be a finite distributive lattice (FDL). Then, up to
isomorphism, there is a unique poset P such that L ∼ = J(P).
Sketch of proof: Let P be the set of join-irreducibles of L. For
t ∈ L set It = { s ∈ P | s ≤ t }. Define φ : L → J(P) by φ(t) = It .
Since It is an order ideal for each t ∈ L, the map φ is well defined.
The fact that J(P) is a lattice implies that φ is injective. To show
that φ is surjective, take I ∈ J(P) and check that φ(t) = I , where
t = ∨s∈I s. Check that I = It . The inclusion I ⊆ It follows
immediately. Conversely, take u ∈ It . Since ∨s∈I s = ∨s∈It s = t we
have
u = ∨s∈It u ∧ s = ∨s∈I u ∧ s.
Since u is join-irreducible u ∧ s = u for some s ∈ I . Then u ≤ s,
which means that u ∈ I . Therefore It = I and so φ is onto.

Felix Gotti [email protected] Lattices: An Introduction

FTFDL (consequences)

Theorem
If P is a poset of order n then J(P) is graded of rank n.
Furthermore, if I ∈ J(P) then ρ(I ) = |I |.
Proof: Exercise.
Theorem
If L is a FDL, the following conditions are equivalent.
1. L is complemented.
2. L is relatively complemented.
3. L is atomic.
4. 1̂ is a join of atoms.
Proof: Exercise.

Felix Gotti [email protected] Lattices: An Introduction

References

M. Bona. A Walk Through Combinatorics. World Scientific,

New Jersey, 2012.
G. Grätzer. Lattice Theory. Dover Publication, Inc., New
Jersey, 2009.
R. Stanley. Enumerative Combinatorics. Cambridge University
Press, New York, 2012.
J. H. van Lint and R. M. Wilson. A Course in Combinatorics.
Cambridge University Press, Cambridge, 1994.

Felix Gotti [email protected] Lattices: An Introduction