PERCOLATION AND DISORDERED SYSTEMS Geoffrey GRIMMETT
PERCOLATION AND DISORDERED SYSTEMS Geoffrey GRIMMETT
PERCOLATION AND DISORDERED SYSTEMS Geoffrey GRIMMETT
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
Percolation and Disordered Systems 249<br />
satisfies π(ρ; α, β) > 0 if 0 < α, β < 1, and furthermore<br />
Also,<br />
π(+; α, β) = β d (1 − α) (2d<br />
2 )−d .<br />
π(∅; α, β) = Pα,β(ρ0 = ∅) = 1 − �<br />
π(ρ; α, β).<br />
Let prw, p+ satisfy prw, p+ > 0, prw + p+ ≤ 1. We pick α, β such that<br />
0 < α, β < 1, β > βc(α, 2) and<br />
ρ∈R<br />
(10.20) π(+; α, β) ≥ 1 − prw (≥ p+).<br />
(That this may be done is a consequence of the fact that βc(α, 2) < 1 for all<br />
α > 0; cf. Lemma 10.18.)<br />
With this choice of α, β, let<br />
� �<br />
π(ρ; α, β)<br />
A = min : ρ �= +, ρ ∈ R<br />
π(ρ)<br />
with the convention that 1/0 = ∞. (Thus defined, A depends on π as well as<br />
on prw. If we set A = min � π(ρ; α, β) : ρ �= +, ρ ∈ R � , we obtain a (smaller)<br />
constant which is independent of π, and we may work with this definition<br />
instead.) Then<br />
π(ρ; α, β) ≥ Aπ(ρ) for all ρ �= +,<br />
and in particular<br />
(10.21) π(ρ; α, β) ≥ (1 − prw − p+)π(ρ) if ρ �= +<br />
so long as p+ satisfies 1 − prw − p+ < A. Note that A = A(α, β) > 0.<br />
We have from the fact that β > βc(α, 2) that the percolation process ω<br />
a.s. contains an infinite open cluster. It follows that there exists a.s. a rw<br />
point in L(ω) which is L(ω)-non-localised. The labyrinth Z of the theorem<br />
may be obtained (in distribution) from L as follows. Having sampled L(ω),<br />
we replace any crossing (resp. reflector ρ (�= +)) by a rw point with probability<br />
π(+; α, β) − p+ (resp. π(ρ; α, β) − (1 − prw − p+)π(ρ)); cf. (10.20) and<br />
(10.21). The ensuing labyrinth L ′ (ω) has the same probability distribution<br />
as Z. Furthermore, if L(ω) is non-localised, then so is L ′ (ω).<br />
The first part of Theorem 10.17 has therefore been proved. Assume henceforth<br />
that d ≥ 3, and consider part (b).<br />
Now consider a labyrinth defined by prw, p+, π(·). Let e be an edge of Z d .<br />
Either e lies in a unique path joining two rw points (but no other rw point)<br />
of some length l(e), or it does not (in which case we set l(e) = 0). Now,<br />
the random walk in this labyrinth induces an embedded Markov chain on the<br />
set of rw points. This chain corresponds to an electrical network obtained<br />
by placing an electrical resistor at each edge e having resistance l(e) −1 . We