Using notation established in the previous subsection, let
be the number
of maximal proper sublattices of
and write
for the ratio
. Our first objective is to prove
The following corollary is immediate, settling the question posed at the end of subsection 3.1 (see also [2]):
Proposition 3.3.1 is proven by counting the large maximal sublattices of
type of
as provided by Corollary
3.2.10.
Suppose , k=2i>0 even. Denote by
the number of
distinct partitions of a k-element set A into 2-element subsets.
Proof:
This is obviously true for k=4. Suppose A' has k'=2i+2 elements
and the formula holds for k=2i. For any given 2-element subset
, there are
many partitions of the required type
including
. The number of such subsets is
, so we
obtain
suitable
partitions
of A'. But each of these has been counted k'/2 times (= the number of
blocks in any such partition), so
which simplifies to
as required.
Proof of Proposition 3.3.1:
Assume that ,
is odd. We count the number of large
maximal sublattices of type
of
: For each
, there are
as many of these as there are partitions of
into 2-element
subsets; hence their total number is
. We
conclude that
. On the other hand, we have
whenever
. Hence
Now whenever
. We infer that
for
. It is obvious that
is not bounded from above for any
fixed exponent k, completing the proof.
Next, we use Theorem 3.2.11 to shed some light on the so-called
sublattice spectrum question raised first by Birkhoff in the 1948 edition of
his ``Lattice Theory''. In the wording of [4, p. 19,], the question
reads as follows: ``Given n, what is the smallest integer such that
every lattice with order
contains a sublattice of exactly n
elements?''
It is shown by Havas and Ward in [27] that indeed exists for
any
(n>0) and that
for n>1. Not much seems to
be known about the values of
otherwise.
Theorem 3.2.11 provides the possible sizes of large sublattices of
for n>6: These are either of the form
or
,
with respective sizes
and
. Now
, so
contains no sublattices of size s for
and for
. It follows that
must be bigger than
.