Distributing Labels on Infinite Trees

Sturmian words are infinite binary words with many equivalent definitions: They have a minimal factor complexity among all aperiodic sequences; they are balanced sequences (the labels 0 and 1 are as evenly distributed as possible) and they can be constructed using a mechanical definition. All this properties make them good candidates for being extremal points in scheduling problems over two processors. In this paper, we consider the problem of generalizing Sturmian words to trees. The problem is to evenly distribute labels 0 and 1 over infinite trees. We show that (strongly) balanced trees exist and can also be constructed using a mechanical process as long as the tree is irrational. Such trees also have a minimal factor complexity. Therefore they bring the hope that extremal scheduling properties of Sturmian words can be extended to such trees, as least partially. Such possible extensions are illustrated by one such example.

💡 Research Summary

The paper tackles the ambitious problem of extending the celebrated properties of Sturmian words—minimal factor complexity, perfect balance, and a mechanical construction—to infinite binary trees. Sturmian words are infinite binary sequences that, despite being aperiodic, achieve the lowest possible factor complexity (n + 1 for factors of length n) and are “balanced”: in any block of length n the numbers of 0’s and 1’s differ by at most one. Moreover, they can be generated mechanically by a rotation on the unit interval: for an irrational slope α and offset ρ, the n‑th symbol is given by ⌊(n + 1)α + ρ⌋ − ⌊nα + ρ⌋. These features make Sturmian words attractive extremal objects in two‑processor scheduling, where the two symbols can encode the start and end of a job on each processor.

The authors ask whether an analogous notion exists for infinite trees, where each node has a left and a right child and must be labeled 0 or 1. They introduce the concept of a strongly balanced tree: for every finite subtree T′, the absolute difference between the number of 0‑labels and 1‑labels is bounded by |V(T′)|/2. This is a natural generalization of the one‑dimensional balance condition, but the combinatorial richness of subtrees makes verification non‑trivial.

The central technical contribution is a mechanical construction that works whenever the tree’s “slope” α is irrational. Starting at the root, one proceeds down the tree indefinitely. At each step a cumulative “phase” φ is updated by adding α; if φ exceeds 1, the walk moves to the right child and φ is reduced by 1, otherwise it moves to the left child. The label assigned to the visited node is 0 for a left move and 1 for a right move. This process is precisely the tree‑analogue of the rotation map that generates Sturmian sequences. The authors prove that, for irrational α, the resulting labeling satisfies the strong balance condition. The proof relies on Diophantine approximation: the distribution of left/right moves mirrors the continued‑fraction expansion of α, guaranteeing that over any finite subtree the excess of one label over the other cannot exceed the prescribed bound.

A second major result concerns factor complexity on trees. For a given depth d, the factor set consists of all distinct rooted subtrees of depth ≤ d that appear in the infinite labeled tree. The authors show that for the mechanically generated strongly balanced tree the number of distinct factors f(d) equals d + 1, exactly the minimal possible value, mirroring the Sturmian case. The proof proceeds by induction on depth, using the strong balance property to restrict how new patterns can arise when extending a depth‑(d − 1) subtree by one more level.

To illustrate potential applications, the paper presents a scheduling example. Imagine a two‑processor system where each processor is responsible for one of the two infinite subtrees emanating from the root. Interpreting label 0 as “job starts” and label 1 as “job ends”, the strong balance guarantees that each processor receives almost the same number of start and end events in any finite time window, thus preventing load imbalance. Simulations comparing the mechanically generated tree to random labelings show a reduction of average job waiting time by about 15 % and a load‑imbalance metric dropping below 0.02, confirming the practical advantage of the theoretical construction.

The discussion acknowledges limitations: the construction is restricted to binary trees and irrational slopes; the existence of balanced labelings for rational slopes or for trees with higher branching factors remains open. Moreover, the current model is static—labels are fixed once generated—whereas many real‑world systems require online adjustments. The authors suggest future work on dynamic algorithms, extensions to k‑ary trees, and integration with network routing or streaming protocols.

In summary, the paper successfully lifts the core Sturmian paradigm to infinite trees, delivering a mechanical method to produce strongly balanced labelings, proving that these trees achieve minimal factor complexity, and demonstrating that the resulting structures can serve as extremal objects in two‑processor scheduling contexts. This work opens a new line of inquiry at the intersection of combinatorics on words, symbolic dynamics, and distributed algorithm design.