Home » Markov chain mixing » Maximizing total variation distance for the Ising model

Maximizing total variation distance for the Ising model

Let $G=(V,E)$ be a finite connected graph with $n$ nodes. Let $P$ denote the transition matrix representing Glauber dynamics for the Ising model on $G$ ., and fix a time $t$ . Which two initial configurations $x,y$ maximize the total variation distance $\|P^t(x,\cdot)-P^t(y,\cdot)\|$ ?

The obvious guess is that $x,y$ should be the all $+1$ and all $-1$ states, respectively. The challenge is to prove (or disprove) this guess.

REFERENCES
[1] Markov Chains and Mixing Times

Maximizing total variation distance for the Ising model

Send me questions about this open problem

Contact Me