Self-organized forest-fires near the critical time (Q865077)

The authors consider the following continuous-time version of the Drossel-Schwabl forest-fire model. Each site of the lattice \(\mathbb{Z}^d\) is either vacant or occupied by a tree. Vacant sites become occupied at rate 1, independently of anything else. Then sites are hit by lightning at rate \(\lambda\), the parameter of the model. When a site is hit by lightning, its entire occupied cluster instantaneously burns down, i.e. becomes vacant. Here it is considered the 2-dimensional case, i.e. the forest is represented by the square lattice. Instead of looking at the steady-state distribution the authors start with all sites vacant and look at the time \(t_c\) at which, in the modified model where there is only growth, but no ignition, an infinite cluster starts to from. This means for positive small \(\lambda\) the behaviour near the critical time \(t_c\) is defined by the relation \(1-\exp (-t_c)=p_c\), the critical probability for the site percolation. For every \(t>t_c\) the probability that the origin \(O\) burns before time \(t\) stays away from 0 as \(\lambda\) tends to 0. This intuitive reasoning leads to the conclusion that if \(m\) is taken sufficiently large and the above event is replaced by the event \(\{\)some vertex at distance \(\leq m\) from 0 burns before time \(t \}\), the corresponding probability will be close to 1 as \(\lambda\) tend to 0. This situation is related to problems which are closer to ordinary percolation. However it is shown that under a percolation-like assumption this intuition is false.