fee1d4d8c521bcc8.tex
1: \begin{abstract}
2:   We consider homological edge percolation on a sequence
3:   $(\mathcal{G}_t)_t$ of finite graphs covered by an infinite
4:   (quasi)transitive graph $\mathcal{H}$, and weakly convergent to
5:   $\mathcal{H}$.  Namely, we use the covering maps to classify
6:   $1$-cycles on graphs $\mathcal{G}_t$ as homologically trivial or
7:   non-trivial, and define several thresholds associated with the rank
8:   of thus defined first homology group on the open subgraphs.  We
9:   identify the growth of the homological distance $d_t$, the smallest
10:   size of a non-trivial cycle on $\mathcal{G}_t$, as the main factor
11:   determining the location of homology-changing thresholds.  In
12:   particular, we show that the giant cycle erasure threshold $p_E^0$
13:   (related to the conventional erasure threshold for the corresponding
14:   sequence of generalized toric codes) coincides with the edge
15:   percolation threshold $p_{\rm c}(\mathcal{H})$ if the ratio
16:   $d_t/\ln n_t$ diverges, where $n_t$ is the number of edges of
17:   $\mathcal{G}_t$, and we give evidence that
18:   $p_E^0<p_{\rm c}(\mathcal{H})$ in several cases where this ratio
19:   remains bounded, which is necessarily the case if $\mathcal{H}$ is
20:   non-amenable.
21: \end{abstract}