Title:Broadcasting colourings on trees. A combinatorial view

Abstract:The broadcasting models on a $d$-ary tree $T$ arise in many contexts such as biology, statistical physics and computer science. Here we consider the $k$-colouring model. A basic question is whether the information of the assignment of the root affects the distribution of the colourings of the leaves. This is the so-called {\em reconstruction/non-reconstruction problem}. It is well known that $d/\ln d$ poses a threshold, in the sense that
1) if $k>d/\ln d$, then the information of the colouring of the root vanishes as the height of the tree grows and it does not affect the distribution of the colourings of the leaves. 2) if $k<d/\ln d$, then the information of the colouring of the root affects the distribution of the colouring of the leaves regardless of the height of the tree.
Despite this threshold, there is no apparent combinatorial reason why such a result should be true. Searching for an explanation, we note that the above implies that for $k>d/\ln d$ and two broadcasting processes that assign the root different colours the following holds: There is a coupling such that the probability of having vertices which take different colour assignments in the two processes reduces as we move away from the root.
The description of such a coupling is a stronger statement than the reconstruction/non-reconstruction threshold. It would cast a light on the actual nature of this, very impressive, phenomenon. Devising one calls for understanding a complex combinatorial problem and it is a non-trivial task to achieve for any $k<d$.
In this work we provide a coupling which has the aforementioned properties for $k>2d/\ln d$ (much below $d$). Interestingly, the decisions it makes are based on simple local rules. Furthermore, we relate the existence of such a coupling with sampling $k$-colourings of sparse random graphs, with expected degree $d$, for $k< d$.