11 The threshold for a Hamilton cycle
Theorem 11.1 (Pósa).
If
and
then
with high probability.
Let be a path
in a graph .
Let . Define
the path by
We call this the rotation of
through the edge .
We write for on the
same vertex set if I can
obtain by applying a
sequence of rotations to .
We define .
Define
|
|
and
.
|
|
Lemma 11.2.
Let
be a graph, and let
be a path. Then
Proof.
Let
be . Let
and let
and
. We
show .
Let be
a path with
endpoint .
Let .
Case 1: and
. We rotate
through the edge
from to form
a path . Of
course . Note
in this path ,
either
or . This
means .
Case 2: If either
or then there was
a first path , when we
were rotating from
to , where we
removed one of
or .
Note carefully that this means that in ,
one of is an
end of .
So .
□
Definition 11.3 (Expander).
Say an -vertex
graph
is an expander if for all
disjoint with ,
,
we have .
Lemma 11.4.
Let
be an expander on
vertices. Let be a
longest path. Then
(note ).
Proof.
If is a
longest path in ,
we have that
otherwise we could rotate and extend ,
contradicting maximality of .
By Lemma 11.2,
So
Note
so use expander condition to see
Lemma 11.5.
Let ,
. Then
is an expander
with probability
for large enough .
Proof.
Lemma 11.6.
If ,
. Then
with
high probability.
Proof.
Let be
the event that
is an expander.
Let
|
|
Then
Since a longest path
in
has
. So if
is to hold, we cannot
include any edges between
and
. To
finish,
11.1 Sprinkling
Let ,
on the same vertex
set. Consider .
Then ,
where
for
small.
Proof of when .
Let
where ,
where .
Let :
then
where .
Define to be the
event that is an
expander. Define to
be the event that .
Then
Let be a
Hamiltonian path in .
Since is an
expander, we have .
So there are
edges that, if included, will complete a Hamiltonian cycle by Pósa rotation lemma. So
|
|