You are on page 1of 6

What is a Lattice?

Null Void
Summer 2015

It is impossible for a man to learn what he thinks he


already knows.
Epictetus

It seems to me that over the course of my mathematical education, there has been talk of a certain (usually algebraic)
object being a lattice. This has always been very unclear to me, and I ended up just thinking it was among the
many other english words of which I did not know the precise meaning1 . This often lead to statements being very
unclear, when they should have been precise. Most notably, in [1]: when they define the sum of a family of ideals and
assert that it is the smallest ideal in the ring that contains each. They go on and state that the intersection of ideals
is an ideal. Then say
Thus the ideals of A form a complete lattice with respect to inclusion.
This was the point at which I accepted (i.e. realized) that there was something that I had simply missed somewhere
along the way. After quietly making a mental note, I put this on the back burner. Now that the semester is over, I
am revisiting this. Admittedly, I only remembered because the math project (research but we will see how much I
am able to research, in the precise sense of that word) seems lend itself nicely to a description by lattices.

What is a Poset?

The first place to start is with the definition of a partially ordered set, or poset (adapted from [2]).
Consider a set X with a relation binary relation . If is reflexive, anti-symmetric2 , and transitive, then is called
a partial order and the pair (X, ) is called a poset.
There are a lot of different examples of posets which are very informative to their structure. One example is (alluded
to above) that any collection of sets is a poset under inclusion.
Now, you may be asking yourself, Is that strict inclusion, or otherwise? If it is not a strict inclusion, then we have
a partial order as defined above. However, if you insist that it is strict, then you get what is called a strict partial
ordering . That is, there is a relation on your set that is anti-reflexive 3 and transitive. These conditions imply
asymmetry 4 as well.
There is another example that I found interesting which I came across while doing my algebra project. That is, that
every poset X is also a category CX . The objects of the category are the elements of X and there is an arrow from a
to b if a b. Note that our order relation cannot be a strict partial order, otherwise we would not have the identity
morphism5 .

1.1

Constructions from Posets

Now we may ask if there are any morphisms of posets. Indeed, there are.
1 This

is a symptom of my complete disregard for any non-technical study until I began writing proofs.
anti-symmetric relation , is such that: a b and b a = a = b.
3 An anti-reflexive relation on a set, is one where no element is related to itself. Observe that this is a property that depends on the
set.
4 An asymmetric relation is such that: a b = b  a.
5 Try to come up with (with proof!) some other posets, strict and otherwise.
2A

Given two posets, (X, ) and (X 0 , 0 ), a morphism f : (X, ) (X 0 , 0 ) of posets is a function f : X X 0 such
that
x y = f (x) f (y), for all x, y X.
Note that if we view X and X 0 as categories, then the morphism f is actually a functor ! In fact, this is actually one
of the easier ways to remember exactly what a functor is.
Keeping up with the category theory, we note that if we have a poset (X, ), then we actually get another poset
(X, ) which is defined by the rule: a b if b a. This is the same as taking the dual of the category CX which
reverses all the arrows. The fact that this relation gives another poset is called the duality principle and the new space
is called the dual of (X, )
If it is also the case that (X, ) is simply ordered (or sometimes called a chain), that is to say that for every
a, b X, either a b or b a. This is sometimes expressed as saying that every element is comparable.
Since part of our regimen is trying to find unique elements, we may ask if there is a largest or smallest element in our
poset. However, upon more careful thought6 we see that this is potentially ambiguous. So we clarify and ask if there is
an element, which we now call , which has nothing larger? I.e. < x for no x X. If there is such an element, then
it is called a maximal element of the poset (minimal element is defined dually). As mentioned, there is another
related notion. Suppose for all x X, we have that x . Then, not only is comparable to everything, it is also
larger. This is a much stronger statement than the first! If such an element exists, then it is called the maximum
(resp. minimum).
Also when we say that an element of a poset, a, covers another element, b, this means that b a and {x | a <
x < b} = . Sometimes this is expressed by saying that a is the immediate successor to b (or b is the immediate
predecessor of a).

What is a Lattice?

As this paper is concerned with lattices, we need to introduce the two operations: =meet and =joint7 , which
are defined (when they are defined at all!) on subsets S of a poset (X, ). The meet of a (sub)set S (of X) is the
infimum (or greatest lower bound ) of S, and is denoted S. The joint of S is (surprise!) the supremum (or least upper
bound ) of S, denoted S. Thus, meet (resp. joint) is only defined when S has a infimum (resp. supremum).
However, this construction of the meet and joint is more general than we will usually need. The important case is
when our set has two elements. Suppose S = {a, b} (i.e. has two elements). Then we denote S as simply a b and
S as a b. We get the nice little picture (actually called a Hasse diagram) to help us remember.
(L.U.B. =)

ab

(= Join)

(G.L.B. =)

a b (= meet)

Now, with this, we are equipped to define a lattice.


Suppose (L, ) is a poset and for every pair of elements a, b in L, the meet and joint of a and b are defined (i.e. a b
and a b exist), then L is said to be a lattice with respect to .
6 In particular, there is more than one way to define the term largest or smallest as it is used in everyday language. Focusing on
larger, does one mean: There is nothing larger? Or that everything is smaller? The second has the implicit assumption that everything
can be compared to this element. But we could have an element that has nothing larger and still not be the largest as there may be
something that is not comparable.
7 To remember which is which: meet symbol is half of the letter M while joint is half of a (rotated) J. Alternatively, just think that
if you join someone, then you are in union ( ' ) with them; if you look at where you meet someone, its is where you intersect ( ' ).
Further, to recall which goes on top and on bottom (in the following figure) just think that the symbols are often representing union
and intersection and each looks its corresponding symbol for meet and join. Since taking unions makes things bigger, and intersections
smaller...

Some immediate observations (that are important): If L is a lattice, then we can consider the meet and joint as binary
operations on S:
:LLL
and
: L L L.
Further, it is obvious that the following properties follow directly from definition:
L.1

x x = x,

xx=x

L.2

x y = y x,

L.3

x (y z) = (x y) z,

L.4

x (x y) = x (x y) = x

xy =yx
x (y z) = (x y) z

And notice
x y x y = x x y = y.
It should be noted that it is possible to define a lattice entirely in terms of these properties. That is, define it as a set
for which there are two binary operations, that satisfy L.1-L.4. This is similar to how a group or ring is defined.

2.1

Basic Properties

With this definition in hand, we can set out and explore and see what can be said about lattices!
First, it seems as if we are almost being asked to check how the meet and join interact with the order relation in a
lattice. We find that these operations are isotone, that is: If a b, then (x a) (x b) and (x a) (x b).
We should check how meet and join interact with each other. We get the distributive inequalities:
x (y z) (x y) (x z)

and x (y z) (x y) (x z).

Further, we have that elements in a lattice satisfy the modular inequality:


a b = a (x b) (a x) b.
We bring our attention back to the way that we originally defined the meet and join in a poset: being the infimum
and supremum of a subset, respectively. This formulation has not been abandoned. For if we have a lattice in which
every subset has an infimum and a supremum, then the lattice is said to be a complete lattice.

Constructions

Now you may say We have to have TWO whole operations to have a lattice! I want to play with lattices but I only
have ONE operation! Well, youre in luck because there is just such a notion! Well almost, it is that of a semilattice.
That is, a semilattice is a set with a single binary operation which satisfies (half of) properties L.1-L.4
We may also examine the ideas of morphisms of lattices. In particular, a morphism of lattices f : L1 L2 is such
that
f (a b) = f (a) f (b) and f (a b) = f (a) f (b)
for all a, b L1 . This is very familiar in that its structure resembles closely the structure of a ring homomorphism.
We also have the notion of a sublattice. That is S is a sublattice of L if S L and S is a lattice under the meet and
join of L, when restricted to S.
We now explore what can be said about the (pre-) image of a sublattice under a morphism. If S is a sublattice of L1 ,
and f : L1 L2 is a morphism of lattices, then we claim that f (S) is a sublattice of L2 . To show this, we must check
properties L.1-L.4. But this is annoying, and we can simply notice that in the definition of a morphism of a lattice,
we basically inherit the lattice structure. Consider what would make f (S) fail to be a lattice: one of the properties
would fail. But failure of a property in L2 would imply a failure in L1 .8
8I

realize this argument is a bit sketchy. However, it is important to try to find the tricks and shortcuts as they force deep understanding.
Only with understanding can we be lazy. So, when looking back over this, try to think why it is impossible to have any other result,
and why we dont really need to check each property.
And Im done.

Similarly, if we have that S 0 is a sublattice of L2 , then f 1 (S 0 ) is a sublattice of L1 . Again, think about what could
cause failures.
We must be careful. For if you have been reading carefully, you may have noticed that a subset S of a lattice L may
be a lattice in its own right under the inherited order relation, yet fail to be a sublattice of L. This is something to
think more about. A hint in the footnote9 .

3.1

Comparing Lattices to Things Known

As previously noted, we want to keep following our nose with the motivating example of ideals of rings. For that
reason, we try to find a concept analogous to that of an ideal in a ring. And if there was not such a concept, would I
have typed this all out? No. An ideal of lattice L is a nonempty set J such that: (a J
x a = x J,

and a, b J = a b J.

As with rings, we have that the inverse image (of some morphism of lattices) of an ideal is itself an ideal, but the
image of an ideal need not be an ideal. To even further drive the point home, we have that the set J(a) = a L of all
x a in L is an ideal and is called a principle ideal .
Now lets take a step back further. We have this nice notion that is very similar to that of a ring. If we ask what was
so nice about ideals of rings, at least one of things we can say is that it has the super closure property. Exploring
this more, we get the idea of a closure property. That is, suppose T is some set. a property of the subsets of T is called
a closure property if T has the property, and any intersection of subsets with the property, retains the property.

3.2

Products of Lattices

We now want to construct new lattices from old. We first do this by defining what the product of posets:
Given two posets P, Q, the product, denoted P Q, will be a poset which is the set of all pairs (p, q) and is partially
ordered by
(p1 , q1 ) (p2 , q2 ) p1 P p2 and q1 Q q2 .

Theorem.

Now we claim that given two lattices L, M that their product, defined in this way, will be a lattice.

Proof. Why should this be true? Well we need to check that each element has a meet and join. But we notice that
(x1 x2 , y1 y2 ) contains (i.e. is larger than) both (x1 , y1 ) and (x2 , y2 ). Thus it is an upper bound for them. To see
that it is indeed the least upper bound, it is sufficient to consider that if there were a smaller element, then (this may
be sketchy but I think this works) if we project the product onto its components, this implies that our candidate (for
a smaller element) would have to be a smaller upper bound in at least one of the components. This means it was not
a new element and our claim follows.

Special Types of Lattices

The analogy given in [2] is that we do not typically study all rings but rather, at least in my (interests) case,
commutative rings. Analogously, we dont typically study all the lattices, but rather a subset of the interesting
types. Of those there are two of particular interest that we will briefly discuss here.

4.1

Distributive Lattices

If a lattice is distributive, then it satisfies the distributive law for both operations, i.e. for all x, y, z L
L.5
L.5

x (y z) = (x y) (x z)
x (y z) = (x y) (x z).

9 Let G be a group. Consider the lattice of set-theoretic subset of G, and also the lattice of subgroups of G. Then the join in the first
case is simply the union. So what is the join in the second case?

As it turns out L.5 is equivalent to L.5 and thus we just refer to them as L.5.
Here

are some examples that were stumbled upon while looking through [3]
The power set P(S) of a set S is a distributive lattice under intersection and union.
Any totally ordered set is a distributive lattice.
The positive integers ordered by divisibility (the meet is the GCD and the join is the LCM, or vice versa via
duality) is a distributive lattice.

However, the following types of lattices are typically the ones of most interest as many of the objects of central concern
give rise to examples of these.

4.2

Modular Lattices

A lattice is called modular when it satisfies the modular equality10 i.e.


L.6

a b = a (x b) = (a x) b.

This is a weaker condition that that of distributivity. This is because b = a b and so we have that L.6 says
a b = a (x b) = (a x) (a b).
So we have additional restrictions to get the distributive condition. However, it turns out that this is where a lot of
the interesting cases pop up.
Among the examples of modular lattices, we have normal subgroups of a group G form a modular lattice with the join
and meet conventionally defined.
Further, the submodule structure of a module over any ring R forms a modular lattice.
Now modular lattices are quite important, so it is valuable to know when a lattice is not modular. First, try to see
why the following lattice (defined in graph form) is not modular.

b
x
a

This lattice, referred to as N5 , is not modular and as it turns out, any lattice that fails to be modular has this N5 as
a sublattice!
Here is another way that we can characterize a modular lattice:
Theorem.
A lattice L is modular Whenever:
x L, then a = b.

(a) a b, (b) a x = b x, (c) a x = b x, for some

Proof. There is really nothing tricky. Just a straightforward verification. Good for practicing with the definition of
modular, if that kind of thing is needed.
Now we present a theorem that will have a lot of consequences that are not readily apparent.
Theorem.
The statement says that if a b, a, b L a modular lattice, then the map which sends x 7 x b is
an isomorphism of the interval I[a, a b] and I[a b, b] with inverse given by y 7 y a.
Proof. If we just write down what each statement of the proof is saying, it becomes obvious (but not until you write
it down and draw a picture! Use a real number line to start).
10 I got quite confused for a while how this was different from the modular property above. Notice that this is equality where that is an
inequality.

Administrative Remarks

There is quite a bit more theory regarding modular lattices. However, that is quite a bit more in-depth. So I will close
here (until possibly later if I do end up needing to know about modular lattices).

References
[1] M. Atiyah and I. MacDonald, Introduction to Commutative Algebra, Westview Press, Oxford, 1969.
[2] S. MacLane and G. Birkhoff, Algebra, The MacMillan Company, London, 1967.
[3] N. Jacoboson, Basic Algebra I, W. H. Freeman and Company,

You might also like