Consider the 1-dimensional Ising model HN=−JN∑i=1σiσi+1−HN∑i=1σiσi=±1
with periodic boundary conditions σN+1=σ1 and magnetic field H. Let Ω be the set of the one-sided infinite sequences ω=(σ1,σ2,…)∈Ω. Consider the map x:Ω↦{−1,1}, with coordinate function xn(ω)=σn and the shift transformation T:Ω↦Ω, with xn(Tω)=xn+1(ω)=σn+1. Let further F be a σ-algebra of subsets of Ω, which is generated by the sets (cylinders) of the form {ω∈Ω:(σi,…,σi+n−1)∈E}, while E is a subset of {−1,1}n. The σ-fields are generated by what is called "thin" cylinders, that is, sets of the form {ω∈Ω:xl(ω)=σl,i≤l≤i+n−1}
Since the finite disjoint unions of thin cylinders form a field which generates F,
a measure P on F is uniquely determined by the values
pn(si,…,si+n−1)=P{ω∈Ω:xi(ω)=si,…,xi+n−1(ω)=si+n−1}
for i≥1. Now, for any finite n,N, 1≤n≤N and 1≤i≤N−n+1, one can define the partial trace:
ZN(i,n):=∑σ1,…,σi−1,σi+n,…,σNe−βH(σ1,…,σN)
and the ordinary trace:
ZN:=∑σ1,…,σNe−βH(σ1,…,σN).
Then, consider the fraction p(N)n(si,…,si+n−1)=ZN(i,n)ZN.
From the latter, we can derive a probability measure in the following sense
pn(si,…,si+n−1)=limN→∞p(N)n(si,…,si+n−1),
and the associated transition probability is simply
p(s1,…,sn|sn+1)=pn+1(s1,…,sn+1)pn(s1,…,sn).
By substitution of the traces and some further algebraic manipulations one can show that the latter is equal to a stochastic and irreducible transition matrix. This implies that the transition probability is associated with an ergodic Markov shift.