Broadcasting colourings on trees. A combinatorial view

The broadcasting models on a d-ary tree T arise in many contexts such as biology, information theory, statistical physics and computer science. We consider the k-colouring model, i.e. the root of T is assigned an arbitrary colour and, conditional on this assignment, we take a random colouring of T. A basic question here is whether the information of the assignment at the root affects the distribution of the colourings at the leaves. This is the so-called reconstruction/non-reconstruction problem. It is well known that d/ln d is a threshold function for this problem, i.e. * if k \geq (1+\eps)d/ln d, then the colouring of the root has a vanishing effect on the distribution of the colourings at the leaves, as the height of the tree grows * if $k\leq (1-\eps)d/ln d, then the colouring of the root biases the distribution of the colouring of the leaves regardless of the height of the tree. There is no apparent combinatorial reason why such a result should be true. When k\geq (1+\eps)d/ ln d, the threshold implies the following: We can couple two broadcasting processes that assign the root different colours such that the probability of having disagreement at the leaves reduces with their distance from the root. It is natural to perceive such coupling as a mapping from the colouring of the first broadcasting process to the colouring of the second one. Here, we study how can we have such a mapping “combinatorially”. Devising a mapping where the disagreements vanish as we move away from the root turns out to be a non-trivial task to accomplish for any k \leq d. In this work we obtain a coupling which has the aforementioned property for any k>3d/ln d, i.e. much smaller than d. Interestingly enough, the decisions that we make in the coupling are local. We relate our result to sampling k-colourings of sparse random graphs, with expected degree d and k<d.

💡 Research Summary

The paper investigates the broadcasting k‑colouring model on a d‑ary tree T, a paradigm that appears in biology, information theory, statistical physics, and computer science. In this model the root receives an arbitrary colour and, conditioned on this assignment, each vertex independently chooses a colour among the k possibilities, respecting the proper‑colouring constraint. The central question is the reconstruction problem: does the colour at the root leave a detectable trace in the distribution of colours observed at the leaves as the tree height grows?

It is well‑known from earlier probabilistic analyses that the threshold for reconstruction is around d / ln d. More precisely, if k ≥ (1 + ε)d / ln d then the influence of the root vanishes (non‑reconstruction), whereas if k ≤ (1 – ε)d / ln d the root’s colour biases the leaf distribution for any height (reconstruction). These results, however, rely on information‑theoretic or statistical‑physics techniques and offer little combinatorial insight.

The authors address this gap by constructing an explicit coupling between two broadcasting processes that start from different root colours. A coupling is a joint probability space in which the two colourings are generated simultaneously; the goal is to make the probability of disagreement at a leaf decrease with its distance from the root. Such a coupling can be viewed as a deterministic mapping from the configuration of the first process to that of the second, provided the mapping respects the local constraints of proper colourings.

The main contribution is a coupling that works for any k > 3d / ln d, a regime far below the trivial bound k = d. The construction is purely local: the decision made at a vertex depends only on the colours of that vertex and its immediate children. The algorithm proceeds in two phases.

Safe‑zone phase (top half of the tree). Starting from the root, the first ⌊h/2⌋ levels are treated as a “safe zone”. For each vertex in this zone the set of admissible colours is restricted so that, with high probability, no conflict arises when the coupling is applied later. This restriction is achieved by a simple greedy rule that discards colours already used by the sibling in the opposite process.
Re‑assignment phase (bottom half). Below the safe zone the coupling performs a re‑assignment of colours. When a vertex needs to choose a colour for its children, it looks at the colour already assigned to the corresponding child in the other process and selects a colour that avoids a clash while keeping the marginal distribution uniform. Because the decision uses only the local neighbourhood, the coupling remains tractable and can be implemented in linear time with respect to the number of vertices.

The authors prove that under the condition k > 3d / ln d the probability that a leaf disagrees in the two processes decays exponentially with its depth. The proof relies on a contraction argument for the associated Markov chain: the expected Hamming distance between the two colourings shrinks by a factor strictly smaller than one at each level. Consequently, the root’s information is “washed out” as it propagates down the tree, establishing non‑reconstruction in a combinatorial fashion.

Beyond the tree setting, the paper connects the coupling to sampling k‑colourings of sparse random graphs G(n, d/n) with expected degree d and k < d. Locally, such graphs resemble d‑ary trees, so the same coupling can be embedded into a Markov chain that samples colourings. The authors show that the chain mixes rapidly when k > 3d / ln d, providing a simple and efficient algorithm for approximate uniform sampling in a regime where previous algorithms required stronger conditions (often k ≥ d).

The structure of the paper is as follows:

Section 1 introduces the broadcasting model, defines reconstruction, and reviews known threshold results.
Section 2 formalises the notion of a coupling and explains why a local, deterministic mapping is desirable from a combinatorial perspective.
Section 3 presents the coupling algorithm, detailing the safe‑zone construction and the re‑assignment rule, and proves its correctness.
Section 4 contains the contraction analysis, establishing that the expected disagreement probability contracts by a factor < 1 whenever k > 3d / ln d.
Section 5 extends the technique to sparse random graphs, describes the induced Markov chain, and proves rapid mixing.
Section 6 discusses implications, possible extensions to other broadcast models (e.g., Potts or Ising), and open problems such as tightening the constant 3 or handling irregular trees.

In summary, the paper delivers a purely combinatorial proof that the reconstruction threshold on d‑ary trees lies at Θ(d / ln d) by exhibiting an explicit, locally defined coupling that works for k > 3d / ln d. This bridges the gap between probabilistic threshold results and constructive algorithmic methods, and it yields practical sampling algorithms for colourings of sparse random graphs in a previously inaccessible parameter range.