Some Boolean layer cakes are lattices under their induced order. Indeed, we have the following
Proof:
Suppose P is realized on X and contains layers k and k+s (0<k<k+s<n,
) but none of the layers j with k<j<k+s. Pick a j-set
and write it as the union of k-sets
.
has no
least upper bound within P since any two k+s-sets containing Y provide
incomparable minimal upper bounds.
For the remainder of this paper, we will deal almost exclusively with layer cakes consisting of levels 0,2,3,n. We introduce the following conventions to smoothen the presentation:
We use as a shorthand for the lattice P(0,2,3,n;n). Since we will be
dealing with different realizations of
at the same time, we
adopt the following abbreviations: Given any (finite) set X with
,
T(X) stands for the the ordered set of
all 2-subsets and all 3-subsets of X, and L(X) for the lattice obtained
from T(X) by adjoining top and bottom. Sublattice will always mean
0-1-sublattice, so L(Y) is a sublattice of L(X) whenever
.
will
denote a set with n elements, hence
. For
, let
.
We call a sublattice S of large iff its size exeeds that of the
next smaller layer cake of the type considered here, that is, if
. The purpose of this subsection is to classify all
large
sublattices of
and to show that they all contain a copy of
as a
sublattice (Theorem 3.2.11).
Let S be any sublattice of L(X). Define
,
and
. Also,
,
and
. Thus |S| = d(S)+2.
Further, an S-quadrangle is any set of
four distinct points such that
,
,
and
all belong to S. The basic fact about S-quadrangles
is
Proof:
by the arithmetic
of L(X), and analogously for
. The four 3-subsets of
must be in S as least upper bounds of 2-subsets.
Proof:
Let and define
similarly. Assume n=2k. If
, then
and
analogously for
. It follows that
, so pick two
points
in
. But then
is an S-quadrangle
and by 3.2.2 we have
as claimed. The proof for the case
n=2k+1 is similar.
The main consequence of Lemma 3.2.3 is the existence of points with
low value of
whenever S is a proper sublattice of L(X).
Proof:
Assume n=2k and suppose for all
. By 3.2.3 S
must then contain all 2-subsets of
and so
. The proof for
n=2k+1 is analogous.
The bounds k-1 respectively k provided by 3.2.4 are sharp in the
following sense: There exist proper sublattices such that
for all
(if n=2k) respectively
for
all
(if n=2k+1). Indeed, let
with |Y|=|Z| and
(if n=2k) respectively
(if n=2k+1) and let
S be
together with top and bottom.
The following lemma relates the values of and
:
Proof:
If w=0, any two 3-subsets of containing p must be disjoint within
, for otherwise their meet would be a 2-set containing p.
Their number is thus at most [(n-1)/2]. Suppose
and let
. There are
3-sets
with
and
. Every point
can belong to at most one 3-set containing p, since
otherwise
would arise as meet. So there are at most n-w-1
such 3-sets not contained in
(this number being actually attained for
suitable S).
Combining 3.2.4 and 3.2.5, we obtain
Proof:
Define . g is monotonic in u for
.
Suppose n=2k. Since , we have
.
Choose
as provided by 3.2.4.
By 3.2.4, we infer
. If
,
monotonicity of g and 3.2.5 imply
,
thus
as desired. If w=0,
by
3.2.5(ii), thus
. If w=1,
by 3.2.5(i), thus
. Now
, so the desired inequality holds also for w=0,1.
Suppose n=2k+1. Since , we have
.
Choose
as provided by 3.2.4. By 3.2.4, we
infer
. For
, proceed as above. For w=0, we have
, thus
. If w=1, we obtain
, thus
. Now
, giving the desired inequality also for w=0,1.
For a given sublattice S of L(X) and any point , we define
together with top and bottom.
S[p] is a sublattice of L(X), the p-trace of S.
3.2.6 implies that any proper sublattice
has some
relatively big p-traces:
Proof:
|S[p]|=|S|-d(S,p).
The following proposition provides the base for the inductive proof of the classification theorem 3.2.11:
Proof:
If , then
. So assume S
is a proper sublattice and choose
as provided by 3.2.7.
We must show that
implies
. Now
|S[p]|=|S|-d(S,p), so we are done if we can show that
. By direct calculation,
.
Assume first n=2k, thus .
Now
by 3.2.6(i), so the desired
inequality boils down to
. Replacing
n by 2k, this is seen to hold provided
, which in turn is
true iff
, a condition satisfied in our case.
Assume now n=2k+1, thus . In this case
by 3.2.6(ii), so
must be
shown. Replacing n by 2k+1, this is equivalent to
, which
is true whenever
, satisfied in our case.
We will be concerned with the following particular sublattices of L(X) in the sequel:
Choose
two distinct points p and q in X. S(X,p,q) is the sublattice with
elements together with top and bottom.
Let be any nonempty collection
of pairwise disjoint 2-subsets of
.
is the
sublattice with elements
together with top and bottom.
It is easy to
see that and
, so all these lattices are large.
Proof:
Suppose . Any 2-subset of X belonging to S but not to
S(X,p,q) must be of the form
with
. If there is any
, y different from p,q,x, then
is an S-quadrangle and
thus
. In any case, S contains all 2-subsets of X, contradicting
S proper. Hence S contains no 2-sets outside S(X,p,q). Suppose there is a
3-set in S but not in S(X,p.q), necessarily of the form
with
x,y both different from q. But then
would
be a 2-set in S but not in S(X,p,q) which we just saw is impossible. We
conclude that S=S(X,p,q).
Let now and assume that
.
Then
and thus S=S(X,p,q) as S is proper, by
the first half of this proof. So we may assume that
consists exclusively of 3-sets, necessarily of the form
. Any
two 3-sets in S must be disjoint within
for otherwise their
meet would produce a 2-set in
, contrary to
assumption. We conclude that
for some collection
.
We are now ready to state the main theorem of this subsection:
Proof:
We use induction on n and thus assume that and the theorem holds
for all n' with
. Let S be a proper large sublattice of
and use Prop. 3.2.8 to choose a point
such that
S[p] is large in
.
All lattices under consideration
will be regarded as sublattices of
. There are three
cases: S[p] may be (i)
itself, or (ii)
for some
in
, or
(iii)
for some
and
collection
of pairwise disjoint subsets of
.
(i) is easy; If ,
consists of 3-sets exclusively
which must be disjoint within
, and thus
for some
. If
, then
for some
. Since S is proper, we infer
by Corollary 3.2.10.
For (ii), we first show that is not possible. Indeed, if
, then
by 3.2.5 and thus
. Now
, whence
. We will show
that
, a contradiction since S is large.
The required inequality is equivalent to
.
Now
and our inequality reduces to
which is satisfied whenever
. We conclude that
. Hence there are
in
such that
and
belong to S. It follows that for any further
, the set
is an S-quadrangle and thus
. Consequently,
and thus
as in (i).
For (iii), observe that , so the same argument as in (ii) shows
that
. Also the same quadrangle argument applies and shows that
S includes every 2-set and 3-set contained in
. Since
is nonempty, we find
, x,y both different
from p,q. This excludes the possibility that
for otherwise
, say, and thus
, a
contradiction. Consequently,
, finishing the inductive
step of the proof of 3.2.11.
It remains to show that 3.2.11 is true for n=6. Assume S is a
proper large
sublattice of , thus
. By 3.2.6
we find
such that
and
. Working within
, we find
such that
and
. It follows that
. Now if
, then
by 3.2.5 and thus
. But this would imply that
, contradicting
. Hence
and by the usual quadrangle argument we obtain that
. If
, then we must have
by 3.2.5 in order to obtain
. It follows that
with
. If
, then S includes
for some
and so equals
by
3.2.10 as S is proper.