On the maximal run-length function in the Lüroth expansion

This paper investigates the multifractal properties of the exceptional set in the Lüroth expansion where the maximal run-length function exhibits linear growth, specifically establishing the Hausdorff dimension for sets defined by prescribed lower and upper limits of the ratio between the run-length and the number of digits.

The Gist

Here is an explanation of the paper, translated from complex mathematics into a story about patterns, races, and the hidden structure of numbers.

The Big Picture: A Game of "Run-Length"

Imagine you have a magical machine that turns any number between 0 and 1 into a long, endless string of numbers (digits). This is called a Lüroth expansion. Think of it like a barcode for a number, but instead of black and white bars, it uses integers like 2, 3, 4, 5, and so on.

For example, a number might look like this:
[2, 2, 2, 5, 3, 3, 3, 3, 3, 8, ...]

Now, imagine you are a detective looking for streaks.

• In the sequence above, you see three 2s in a row. That's a streak of length 3.
• Then you see five 3s in a row. That's a streak of length 5.

The Run-Length Function ($\ell_n$) is simply the length of the longest streak you find if you look at the first $n$ digits of the number.

The Normal Behavior: The "Logarithmic" Rule

For almost every number you pick (like 0.12345...), these streaks behave in a very predictable, boring way. They grow, but very slowly.

• If you look at the first 1,000 digits, the longest streak might be around 10.
• If you look at the first 1,000,000 digits, the longest streak might be around 20.

Mathematicians call this logarithmic growth. It's like a snail that gets a little faster every day but never really runs a marathon. The paper acknowledges this: "For almost all numbers, the longest streak grows like the logarithm of the total digits."

The Exceptional Set: The "Linear" Rebels

But what if you want to find the rebels? What if you are looking for the rare, weird numbers where the streaks grow fast?

Imagine a race.

• Normal numbers: The streak grows slowly, like a snail.
• Rebel numbers: The streak grows linearly. This means if you look at 1,000 digits, the streak is 100 long. If you look at 1,000,000 digits, the streak is 100,000 long. The streak is a fixed percentage of the total length.

The paper asks: "How many of these rebel numbers exist?"

In math, "how many" isn't just about counting (1, 2, 3...). Since there are infinite numbers, we measure their "size" using something called Hausdorff Dimension.

• Think of a line as having dimension 1.
• Think of a single point as having dimension 0.
• Think of a "dust" of points scattered everywhere as having a dimension between 0 and 1.

The authors want to know the "dimension" (the size/weight) of the set of numbers where the streaks grow at specific speeds.

The Three Scenarios

The paper breaks down the rebels into three categories based on two numbers, $\alpha$ and $\beta$.

• $\alpha$ is the slowest speed the streak ever settles into (the floor).
• $\beta$ is the fastest speed the streak ever reaches (the ceiling).

1. The "Ghost" Case ($\beta = 0$)

If the fastest speed is 0, it means the streaks never grow linearly at all. They stay tiny.

• Result: This is actually the "normal" behavior. Almost all numbers are here. The "size" (dimension) is 1 (the full size of the number line).

2. The "Impossible" Case ($\alpha > \frac{\beta}{1+\beta}$)

The authors discovered a mathematical law of physics for these streaks. You cannot just pick any two speeds you want.

• If you demand that the streaks are always very fast (high $\alpha$) but also sometimes even faster (high $\beta$), there is a limit.
• If your requirements are too strict (specifically, if the minimum speed is too high relative to the maximum), no such numbers exist.
• Result: The set is empty (or just a few isolated points). The dimension is 0.

3. The "Goldilocks" Case (The Sweet Spot)

This is the main discovery. If you pick speeds that are physically possible (where the minimum isn't too high compared to the maximum), there are numbers that fit this description.

• Result: These numbers form a "fractal dust." They are rare, but they are everywhere in a complex, self-repeating pattern.
• The paper provides a precise formula to calculate the dimension of this dust. It's a specific number between 0 and 1, calculated using a complex equation involving the speeds $\alpha$ and $\beta$.

The Analogy: The "Streaky" Cookie Jar

Imagine a giant jar of cookies.

• Normal Cookies: Most cookies have a few chocolate chips scattered randomly. If you look at a huge cookie, the longest line of chips is short.
• Rebel Cookies: You are looking for cookies where the chips are arranged in massive, long lines.
  • Some cookies have lines that are 1% of the cookie's width.
  • Some have lines that are 50% of the width.

The paper asks: "If I tell you I want cookies where the longest line of chips is at least 10% of the width but at most 20% of the width, how many such cookies are there?"

The answer isn't "zero" or "infinity." The answer is a specific fractal dimension. It's like asking, "How much space does this specific type of cookie dust take up?" The paper gives you the ruler to measure that dust.

Why Does This Matter?

1. Understanding Randomness: It helps us understand the difference between "random" noise and "structured" chaos. Even in a system that looks random (like the digits of a number), there are hidden rules about how long patterns can last.
2. Fractal Geometry: It adds a new piece to the puzzle of how complex shapes (fractals) are built. It shows that even within the infinite complexity of numbers, there are precise layers of "size" for different behaviors.
3. The "Exceptional" is Everywhere: It proves that while these "fast-streak" numbers are rare (they have dimension less than 1), they are not just a few oddballs. They form a rich, intricate structure that fills a specific amount of space in the mathematical universe.

Summary in One Sentence

This paper calculates the exact "size" of the mysterious group of numbers where the longest repeating patterns grow at a steady, linear speed, revealing that while these numbers are rare, they form a beautiful, complex fractal structure with a precise mathematical dimension.

Technical Summary

Here is a detailed technical summary of the paper "On the maximal run-length function in the Lüroth expansion" by Dingding Yu.

1. Problem Statement and Context

The Lüroth Expansion:
The paper investigates the metric properties of the Lüroth expansion, a representation of real numbers $x \in (0, 1]$ similar to continued fractions or dyadic expansions. For a given $x$, the expansion is defined by a sequence of integers $d_n(x) \ge 2$ such that:
$$x = \frac{1}{d_1(x)} + \sum_{n=2}^{\infty} \frac{1}{d_1(x)(d_1(x)-1)\cdots d_{n-1}(x)(d_{n-1}(x)-1)d_n(x)}$$
The transformation generating these digits is the Lüroth map $T(x)$.

The Maximal Run-Length Function:
The central object of study is the maximal run-length function, denoted $\ell_n(x)$. For the first $n$ digits of the expansion of $x$, $\ell_n(x)$ is defined as the length of the longest consecutive block of identical digits:
$$\ell_n(x) = \max \{k \ge 1 : d_{j+1}(x) = \dots = d_{j+k}(x) \text{ for some } 0 \le j \le n-k\}$$

Previous Results:

• Law of Large Numbers: Sun and Xu [12] established that for almost all $x$, $\ell_n(x) \sim \log_2 n$.
• Multifractal Analysis (Logarithmic Scale): The set of points where $\ell_n(x)$ grows at a rate different from $\log_2 n$ (e.g., $\lim \ell_n(x)/\log_2 n = \alpha$) was shown to have full Hausdorff dimension (dimension 1).
• Gap in Literature: While the behavior relative to $\log n$ is well-understood, the behavior relative to the linear scale $n$ (i.e., when $\ell_n(x)$ grows linearly) had not been fully characterized for the Lüroth expansion.

The Specific Problem:
The paper aims to determine the Hausdorff dimension of the "exceptional set" $E(\alpha, \beta)$ where the maximal run-length exhibits linear growth with specific lower and upper limits:
$$E(\alpha, \beta) = \left\{ x \in (0, 1] : \liminf_{n \to \infty} \frac{\ell_n(x)}{n} = \alpha, \quad \limsup_{n \to \infty} \frac{\ell_n(x)}{n} = \beta \right\}$$
where $0 \le \alpha \le \beta \le 1$.

---

2. Methodology

The author employs a combination of metric number theory, fractal geometry, and probabilistic methods to solve the problem.

A. Upper Bound Estimation

To establish the upper bound for the Hausdorff dimension, the author uses a covering argument:

1. Decomposition: The set $E(\alpha, \beta)$ is decomposed based on the occurrence of long runs of identical digits.
2. Cylinder Sets: The space is covered by cylinder sets $I_n(d_1, \dots, d_n)$, which correspond to specific initial digit sequences. The length of these intervals is explicitly calculated using the properties of the Lüroth map.
3. Counting: The author counts the number of such cylinders that satisfy the condition of having a run of length approximately $\beta n$ (or $\alpha n$) within the first $n$ digits.
4. Mass Distribution/Summation: By summing the $s$-th powers of the lengths of these covering intervals, the author derives conditions under which the $s$-dimensional Hausdorff measure is finite. This leads to an equation involving the parameter $s$ that must be satisfied.

B. Lower Bound Estimation

To establish the lower bound, the author constructs a specific subset of $E(\alpha, \beta)$ and applies the Mass Distribution Principle:

1. Construction of a Cantor-like Set: A subset $G(M)$ is constructed by fixing the digits of $x$ in a specific pattern:
   • Long blocks of identical digits (specifically the digit 2) of length $m_k$ are inserted at specific positions $n_k$.
   • The gaps between these blocks are filled with digits chosen from a bounded set $\{2, \dots, M\}$.
   • The sequences $\{n_k\}$ and $\{m_k\}$ are carefully chosen to ensure the resulting set lies within $E(\alpha, \beta)$.
2. Mapping and Hölder Continuity: A map $f$ is defined that removes the "special" positions (where the long runs occur). The author proves that this map is Hölder continuous. This allows the transfer of dimension bounds from the image set $f(G(M))$ back to the original set.
3. Measure Construction: A probability measure $\mu$ is defined on the symbolic space of the constructed set. The measure is distributed such that the mass of a cylinder decays at a rate related to the target dimension.
4. Gap Analysis: The author proves that the fundamental intervals (cylinders) in the constructed set are well-separated (have a "gap" proportional to their length). This ensures the measure satisfies the mass distribution principle condition: $\mu(B(x, r)) \le C r^s$.

C. Key Mathematical Tools

• Moran-type Formulas: Used to calculate dimensions of sets with bounded digits.
• Implicit Function Theorem: Used to analyze the properties of the dimension function $s(u)$ defined by the equation $\sum (2^{-u}/(t(t-1)))^s = 1$.
• Asymptotic Analysis: Precise estimation of the ratios $m_k/n_k$ to match the target $\alpha$ and $\beta$.

---

3. Key Contributions and Results

The main result is Theorem 1.2, which provides the exact Hausdorff dimension of the set $E(\alpha, \beta)$.

Let $s(u)$ be the unique solution to the equation:
$$\sum_{t=2}^{\infty} \left( \frac{1}{2^u t(t-1)} \right)^s = 1$$

The dimension is given by:
$$\dim_H E(\alpha, \beta) = \begin{cases} 1 & \text{if } \beta = 0 \quad (\text{implies } \alpha=0) \\ s\left( \frac{\beta^2(1-\alpha)}{(1-\beta)[\beta - \alpha(1+\beta)]} \right) & \text{if } 0 \le \alpha < \frac{\beta}{1+\beta} < \beta < 1 \\ 0 & \text{otherwise} \end{cases}$$

Key Insights from the Result:

1. Trivial Cases:

   • If $\beta = 0$, the set has full dimension (1). This aligns with the fact that for almost all $x$, $\ell_n(x) \sim \log n$, so $\ell_n(x)/n \to 0$.
   • If $\alpha > \frac{\beta}{1+\beta}$, the set is empty (or countable), hence dimension 0. This is a structural constraint: it is impossible for the lower limit of the run-length ratio to exceed $\frac{\beta}{1+\beta}$ given an upper limit $\beta$. This arises from the geometric constraints of how runs can be packed into the expansion.

2. Non-Trivial Regime:

   • When $0 \le \alpha < \frac{\beta}{1+\beta} < \beta < 1$, the dimension is strictly between 0 and 1.
   • The dimension is determined by the function $s(u)$ evaluated at a specific argument derived from $\alpha$ and $\beta$. The argument represents the "density" or "frequency" of the long runs required to satisfy the liminf and limsup conditions.

3. Comparison with Other Expansions:
   The paper notes that similar problems for dyadic expansions (Erdős and Rényi) and continued fractions have been studied. This work extends the theory to the Lüroth expansion, which has a different probabilistic structure (digits are i.i.d. with a specific distribution) compared to continued fractions (where digits are not independent).

---

4. Significance

1. Completing the Multifractal Spectrum: The paper fills a critical gap in the multifractal analysis of the Lüroth expansion. While the logarithmic growth regime was known, the linear growth regime (which corresponds to "large deviations" or rare events where digits repeat excessively) was uncharacterized.
2. Structural Constraints: The discovery of the constraint $\alpha \le \frac{\beta}{1+\beta}$ reveals a fundamental limitation on the asymptotic behavior of run-lengths in this specific dynamical system. It highlights how the geometry of the expansion restricts the possible combinations of liminf and limsup.
3. Methodological Advancement: The construction of the lower bound using a carefully designed Cantor set with variable block lengths and the subsequent use of a Hölder continuous map to handle the "deleted" positions demonstrates a sophisticated technique applicable to other non-standard expansions.
4. Connection to Thermodynamic Formalism: The function $s(u)$ appearing in the result is related to the pressure function in thermodynamic formalism, linking the geometric dimension of the exceptional set to the statistical properties of the underlying dynamical system.

In summary, Dingding Yu provides a complete characterization of the Hausdorff dimension for sets of numbers with prescribed linear growth rates of maximal run-lengths in the Lüroth expansion, establishing both the existence of such sets and their precise fractal dimensions.