Professional Documents
Culture Documents
Probability
Research Seminar
Gabriel Baptista
Insituto Nacional de Matem
atica Pura e Aplicada
IMPA
Gabriel Baptista
Outline
Percolation
Continuity
Topology
Schramms Conjecture
Cheeger Constant
Intuition
Configuration
Result on r-regular graphs
Gabriel Baptista
Percolation
F = (cilinders)
Y
Pp =
Ber (p)
eE
Connected component:
Cx = {y V (G ) : x y }
Where x y open path joining x to y
Gabriel Baptista
Percolation
Definition
The percolation probability on G is G (p) = Pp (x : |C (x)| = )
Obs: G (p) increasing on p.
Definition
Critical percolation probability pc of a graph G :
pc (G ) = sup{p [0, 1] : G (0) = 0} = inf {p [0, 1] : G (0) > 0}
Gabriel Baptista
Continuity of pc
Gabriel Baptista
Continuity of pc
Example
Let S(k) = {x Z3 : x3 [0, k]}:
d((S(k), 0), (Z3 , 0)) = 2k (S(k), 0) (Z3 , 0)
On G = {G F : G is transitive} d is a metric.
Example
d(Z2 Zk , Z3 ) = 2k Z2 Zk Z3
d(Z Zk , Z2 ) = 2k Z Zk Z2
Gabriel Baptista
Continuity of pc
Gabriel Baptista
Cheeger Constant
Gabriel Baptista
Cheeger Constant
Gabriel Baptista
Cheeger Constant
Good properties:
Easier to count events on (space of configurations):
2
2m 2m2
... 2
2m!
2
|| = 2
= m
=: N(m)
m!
2 m!
Construct a graph from half-edges
P ((F ) is a simple graph) exp
Canfield in [3])
Gabriel Baptista
1d 2
4
(Bender and
Cheeger Constant
Theorem
Let r 3 and 0 < < 1 be such that:
24/r < (1 )1 (1 + )1+
Then a.e. r-regular graph G has h(G )
(1)r
2
Definition
Given a configuration Fn we define its Cheeger constant to be:
|E (A, V \ A)|
n
h(Fn ) = min
: 0 < |A|
AV
|A|
2
Where E (A, V \ A) is the set of edges wih exactly one endvertex in
jA Wj
Gabriel Baptista
Cheeger Constant
Proposition
Almost all configurations have a Cheeger constant bounded below
by (1)r
2 .
Proof.
U V and s, u N:
k(U) = |{k : |Ak (
jU
Wj )| = 1}|
A(U, s) = {F : k(U) = s}
S
P(u, s) = P( |U|=u A(U, s))
Suppose that a configuration F has h(F ) <
A V ;
(1)r
2 .
Then:
|E (A, V \ A)|
(1 )r
<
|A|
2
Gabriel Baptista
Cheeger Constant
Proof (Cont.)
Then:
|E (A, V \ A)| <
(1 )r |A|
P(F ) P(|A|, |E (A, V \ A)|)
2
(1)ru
2
such that
1un/2
0ss(u)
Gabriel Baptista
Cheeger Constant
Proof (Cont.)
Given U0 V , estimate P(A(U0 , s)) counting the ways A(U0 , s)
can occur. Let u = |U0 |:
S
S
s vertices from jU0 Wj and s from j U
/ Wj to form the
ru r (nu)
edges: s
;
s
s! different ways of connect them;
On one side ru s vertices connect among them, on the other
r (nu)s
r (n u) s: N( rus
)
2 )N(
2
Gabriel Baptista
Cheeger Constant
Proof (Cont.)
P(A(U, s)) only depends on |U|:
P(u, s)
X
|U|=u
n
ru
r (nu)
s
n
P(A(U, s)) =
P(A(U0 , s))
u
r (nu)s
)
s!N( rus
2 )N(
2
=: P0 (u, s)
N( rn
)
2
Gabriel Baptista
Cheeger Constant
Proof (Cont.)
Remember s(u) <
(1)ru
:
2
n
P0 (u, s) r
1un/2
0ss(u)
And:
b2c
X
u=1
b2c
X
u=1
Cheeger Constant
Proof (Cont.)
The first follow from bounds on the binomial formula
Note that:
n
n/2
P0 (n/2, s(u)) =
rn/22
s!N(
s(u)
rn
2
2
)2
N(rn/2)
Gabriel Baptista
References
Gabriel Baptista
Gabriel Baptista
Appendix
Define: p as:
p =
1 : Xe < p
0 : Xe p
Then Pp = p P.
Definition
We say that a random variable N on (, F) is increasing if
N(w ) N(w 0 ) for any w w 0 . An event A F is increasinf if IA
is an increasing random variable.
Gabriel Baptista
Appendix
Proposition
If N is an increasing random variable on (, F) and p1 p2 then:
Ep1 (N) Ep1 (N)
Proof.
We can see from the definition that if p1 p2 then p1 p2 .
Since N is increasing, we have N(p1 ) N(p2 ). Integrating we
get Ep1 (N) Ep2 (N)
Gabriel Baptista