03.12.2012 Views

PERCOLATION AND DISORDERED SYSTEMS Geoffrey GRIMMETT

PERCOLATION AND DISORDERED SYSTEMS Geoffrey GRIMMETT

PERCOLATION AND DISORDERED SYSTEMS Geoffrey GRIMMETT

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

13.1 Basic Properties<br />

Percolation and Disordered Systems 263<br />

13. R<strong>AND</strong>OM-CLUSTER MODELS<br />

First we summarise some useful properties of random-cluster measures. Let<br />

G = (V, E) be a finite graph, and write ΩE = {0, 1} E . The random-cluster<br />

measure on ΩE, with parameters p, q satisfying 0 ≤ p ≤ 1 and q > 0, is given<br />

by<br />

φp,q(ω) = 1<br />

�<br />

�<br />

p<br />

Z<br />

e∈E<br />

ω(e) (1 − p) 1−ω(e)<br />

�<br />

q k(ω) , ω ∈ ΩE<br />

where Z = ZG,p,q is a normalising constant, and k(ω) is the number of<br />

connected components of the graph (V, η(ω)), where η(ω) = {e : ω(e) = 1} is<br />

the set of ‘open’ edges.<br />

Theorem 13.1. The measure φp,q satisfies the FKG inequality if q ≥ 1.<br />

Proof. If p = 0, 1, the conclusion is obvious. Assume 0 < p < 1, and check<br />

the condition (5.2), which amounts to the assertion that<br />

k(ω ∨ ω ′ ) + k(ω ∧ ω ′ ) ≥ k(ω) + k(ω ′ ) for ω, ω ′ ∈ ΩE.<br />

This we leave as a graph-theoretic exercise. �<br />

Theorem 13.2 (Comparison Inequalities). We have that<br />

(13.3)<br />

(13.4)<br />

φp ′ ,q ′ ≤ φp,q if p ′ ≤ p, q ′ ≥ q, q ′ ≥ 1,<br />

φp ′ ,q ′ ≥ φp,q if<br />

p ′<br />

q ′ (1 − p ′ ) ≥<br />

p<br />

q(1 − p) , q′ ≥ q, q ′ ≥ 1.<br />

Proof. Use Holley’s Inequality (Theorem 5.5) after checking condition (5.6).<br />

�<br />

In the next theorem, the role of the graph G is emphasised in the use<br />

of the notation φG,p,q. The graph G\e (resp. G.e) is obtained from G by<br />

deleting (resp. contracting) the edge e.<br />

Theorem 13.5 (Tower Property). Let e ∈ E.<br />

(a) The conditional measure of φG,p,q given ω(e) = 0 is φ G\e,p,q.<br />

(b) The conditional measure of φG,p,q given ω(e) = 1 is φG.e,p,q.<br />

Proof. This is an elementary calculation of conditional probabilities. �<br />

More details of these facts may be found in [26, 159, 162]. Another comparison<br />

inequality may be found in [161].

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!