🔢 mathematics

Balanced rectangles over Sturmian words and minimal discrepancy intervals

This paper characterizes the balance properties of $m \times n$ rectangular matrices formed from Sturmian words with arbitrary slope $\alpha$ using Ostrowski representations and the distribution of $n\alpha \bmod 1$ , thereby generalizing previous results that were limited to quadratic irrational slopes.

Original authors: Ingrid Vukusic

Published 2026-04-20

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Ingrid Vukusic

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

The Big Picture: Tiling a Floor with a Strange Pattern

Imagine you have a very long, infinite strip of floor tiles. These tiles are either Black or White. But this isn't a random pattern; it's a specific, mathematical rhythm called a Sturmian word.

Think of this pattern like a musical rhythm that never repeats exactly but follows a strict rule based on an "irrational" number (like $\pi$ or the Golden Ratio). A famous example is the Fibonacci word, which looks like 0100101001....

The Problem:
In a normal, repeating pattern (like a checkerboard), if you take any square piece of the floor, you can predict exactly how many black and white tiles are inside. But in this strange, non-repeating pattern, things get tricky.

The author, Ingrid Vukusic, asks a specific question: If I cut out a rectangular piece of this floor (say, 2 rows by 3 columns), will the number of black tiles always be the same, or will it vary wildly?

She defines a rectangle as "Balanced" if the number of black tiles inside it is always either $C$ or $C+1$ . It's never $C-1$ or $C+2$ . It's "fair."

The Analogy: The "Fair Share" of Points

To solve this, the paper uses a clever trick. Instead of looking at the tiles directly, it imagines a clock face (a circle from 0 to 1).

The Dots: Imagine dropping dots on this clock face at positions $0, \alpha, 2\alpha, 3\alpha, \dots$ (where $\alpha$ is the rule that generates the pattern).
The Slice: Now, imagine taking a slice of the clock (an interval) of a specific length.
The Question: If you slide this slice around the clock, does it always catch roughly the same number of dots?

If the slice catches either $k$ or $k+1$ dots no matter where you put it, the slice is "Balanced."

The paper proves a magical connection: The rectangular tiles are balanced if and only if these clock slices are balanced.

The Solution: The "Ostrowski Recipe"

So, how do you know if a specific rectangle (like a $2 \times 10$ grid) is balanced? You can't just count them all; the pattern goes on forever.

The author provides a "recipe" based on something called Ostrowski Representation.

The Metaphor: The Currency System
Imagine you have a strange currency system where the bills aren't 1, 10, 100. Instead, the bills are special numbers derived from the rule $\alpha$ (let's call them $q_0, q_1, q_2, \dots$ ).

$q_0 = 1$
$q_1 = \text{something else}$
$q_2 = \text{something else}$

Every number (like your rectangle width $m$ or height $n$ ) can be built by adding up these special bills. But there's a rule: You can't use too many of the same bill, and you can't use two specific bills right next to each other. This is the Ostrowski Representation.

The "Balanced" Rule:
The paper says a rectangle is balanced only if the "bills" used to build the width ( $m$ ) and the height ( $n$ ) fit into one of four specific "shapes" or patterns.

Think of it like a lock and key:

The Lock: The rectangle dimensions ( $m$ and $n$ ).
The Key: The Ostrowski representation (how you build $m$ and $n$ using the special bills).
The Result: The lock only opens (the rectangle is balanced) if the key fits one of the four specific shapes described in the paper.

Why This Matters (The "So What?")

Generalizing the Past: Previous researchers had solved this puzzle, but only for very specific, "nice" numbers (like the Golden Ratio). They used computer programs to check them. This paper solves it for every irrational number, not just the "nice" ones.
The Method: Instead of using a computer to brute-force the answer, the author used deep math about how numbers are distributed on a circle (Discrepancy Theory). It's like figuring out the rules of a game by understanding the physics of the ball, rather than just playing the game a million times.
The "Split" and the "Parity":
- Sometimes, the "bills" for $m$ and $n$ are completely separated (like $m$ uses small bills and $n$ uses huge bills). This is usually balanced.
- Sometimes, $m$ is a specific "convergent" (a very good approximation of the rule), and $n$ has to follow a strict "even/odd" rule with its bills. If the parity is wrong, the rectangle is unbalanced.

Summary in One Sentence

This paper discovers the exact mathematical "recipe" (based on how you build numbers using special bills) that tells you whether a rectangular chunk of a complex, non-repeating pattern will have a fair distribution of colors, solving a puzzle that was previously only partially understood.

The "Takeaway" for a General Audience

Imagine you are a chef trying to cut a cake that has a weird, non-repeating swirl of chocolate and vanilla. You want to cut a rectangular slice that has a perfect 50/50 mix (or as close as possible).

This paper gives you a calculator. You type in the size of your slice (width and height), and the calculator checks the "mathematical DNA" of those numbers. If the DNA matches one of four secret patterns, your slice will be perfectly balanced. If not, your slice will be lopsided. The author figured out exactly what those four patterns are for any cake, not just the famous ones.

1. Problem Statement

The paper addresses the 2-dimensional balance problem for Sturmian words.

Context: Sturmian words are infinite binary sequences defined by an irrational slope $\alpha$ and an intercept $\theta$ . They are known to be "balanced" in one dimension, meaning the number of 1s (weight) in any two factors of the same length differs by at most 1.
Extension: The authors consider $m \times n$ rectangular submatrices (rectangles) extracted from the infinite Hankel matrix of a Sturmian word. A rectangle is defined as balanced if the sum of its entries (total weight) takes at most two consecutive integer values for all possible starting positions.
Gap: Previous work (Anselmo et al., Shallit & Vukusic) characterized balanced rectangles only for quadratic irrational slopes (using Zeckendorf representations and automata like Walnut). The open problem was to provide a full characterization for all irrational slopes $\alpha$ .

2. Methodology

The paper employs a multi-step approach combining combinatorics on words, Diophantine approximation, and discrepancy theory.

A. Reduction to Interval Discrepancy

The core insight (Theorem 3.1) is an equivalence between the balance of $m \times n$ rectangles and the balance of intervals in the distribution of the sequence $\{k\alpha\} \pmod 1$ .

The weight of a rectangle is shown to depend on the number of points $\{ (i+\ell)\alpha \}$ falling into a specific interval of length $\{n\alpha\}$ .
Theorem 3.1: The $m \times n$ rectangles are balanced if and only if intervals of length $\delta = \{n\alpha\}$ are "balanced" with respect to the set of points $B = \{0, \alpha, \dots, (m-1)\alpha\}$ .
Definition of Balanced Intervals: An interval length $\delta$ is balanced with respect to a set $B$ if, as the interval slides over the torus $\mathbb{R}/\mathbb{Z}$ , the number of points it contains takes at most two values ( $c$ or $c+1$ ).

B. Bijectivity of Maps

To determine when intervals are balanced, the author introduces a combinatorial criterion based on bijective maps (Lemma 4.3).

For a set of points $B$ and interval length $\delta$ , two functions $f_{left}$ and $f_{right}$ are defined, mapping indices of points to the indices of the nearest neighbors to the left and right of the interval's endpoint.
Key Lemma: The intervals are balanced if and only if these maps are bijective on the index set $\{0, \dots, m-1\}$ .
This transforms the geometric problem into an arithmetic problem involving the ordering of $\{x\alpha\}$ values.

C. Ostrowski Representations

The analysis relies heavily on the Ostrowski representation of integers with respect to $\alpha$ .

Any integer $n$ can be uniquely written as $n = \sum b_k q_k$ , where $q_k$ are the denominators of the convergents of $\alpha$ , and the digits $b_k$ satisfy specific constraints.
The paper utilizes deep properties of these representations (Lemmas 5.1–5.17) to estimate the distance $\|n\alpha\|$ and determine the ordering of fractional parts $\{x\alpha\}$ .
The proof involves analyzing the "small" and "large" parts of the Ostrowski representation to determine which integers minimize the distance to the nearest integer in specific ranges.

3. Key Contributions and Results

Main Theorem (Theorem 2.1)

The paper provides a complete characterization of balanced $m \times n$ rectangles for any irrational slope $\alpha$ . The rectangles are balanced if and only if the Ostrowski representations of $m$ and $n$ (with respect to $\alpha$ ) satisfy one of four specific structural conditions:

Split Representation (Small $m$ , Large $n$ ):
- $m$ uses only "small" indices in the Ostrowski expansion, and $n$ uses only "large" indices (disjoint sets of indices).
- Intuition: The interval length $\{n\alpha\}$ is extremely small (or large), containing at most one point from the set $B$ .
Convergent/Semi-Convergent Cases:
- $m$ is the denominator of a convergent ( $q_M$ ) or a semi-convergent ( $q_{M-1} + a q_M$ ).
- $n$ has a specific structure where its "large" digits appear at indices with specific parity relative to $m$ .
- Intuition: The points in $B$ are extremely evenly distributed (approximating a lattice), and the interval length aligns perfectly with the gaps between these points.

Technical Innovations

Generalization: Extends results from quadratic irrationals (where continued fractions are periodic) to all irrationals.
Uniform Approach: Instead of using automata (which fail for non-quadratic irrationals), the proof uses Diophantine approximation and the properties of the Ostrowski expansion.
Symmetry Handling: The paper rigorously handles the symmetry between $m$ and $n$ and the relationship between $\alpha$ and $1-\alpha$ .

4. Significance

Theoretical Completeness: This work solves the 2D balance problem for Sturmian words in full generality, removing the restriction to quadratic slopes.
Methodological Shift: It demonstrates that the balance property is fundamentally linked to the distribution modulo 1 of the sequence $n\alpha$ . The characterization via Ostrowski representations provides a number-theoretic lens for understanding combinatorial properties of words.
Connection to Discrepancy Theory: The paper bridges the gap between combinatorics on words and the theory of bounded remainder sets and minimal discrepancy intervals. It shows that "balanced rectangles" are equivalent to intervals having "minimal discrepancy" (error $< 1$ ) for a fixed number of points.
Algorithmic Implications: While the previous results relied on the software Walnut (limited to quadratic irrationals), this paper provides a theoretical framework that could potentially be adapted for algorithmic verification for specific classes of irrationals, though the general case remains non-computable in the sense of finite automata.

Summary Conclusion

Ingrid Vukusic's paper establishes that the balance of $m \times n$ rectangles in Sturmian words is entirely determined by the Ostrowski representations of $m$ and $n$ . The rectangles are balanced if and only if the representations exhibit a "split" structure (separating small and large digits) or if $m$ corresponds to a convergent/semi-convergent with specific parity constraints on $n$ . This result unifies previous findings and deepens the understanding of the interplay between Diophantine approximation and the combinatorial structure of Sturmian sequences.