A Fixed-Prime Criterion for Reciprocals in… — Plain-Language Explanation

✨

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

Imagine you have a very special, infinite number line made of a specific base (like base 3, which uses only the digits 0, 1, and 2). Now, imagine you are only allowed to build numbers using a specific, restricted set of "bricks" (digits). For example, in the famous Cantor Set, you can only use the bricks 0 and 2. You are forbidden from using the brick 1.

This creates a "Missing-Digit Set." It's a collection of numbers that look like 0.202002... in base 3, but never contain a 1.

The paper asks a simple but tricky question: If you take a sequence of numbers that get huge very fast (like factorials, Fibonacci numbers, or products of polynomials), and you look at their reciprocals (1/1, 1/2, 1/6, 1/24...), will any of them ever land inside this restricted "Missing-Digit" club?

The author, Scott Duke Kominers, proves that for almost all these fast-growing sequences, the answer is no. There are only a tiny, finite number of exceptions.

Here is the breakdown of how he proves it, using some everyday analogies.

1. The "Digital DNA" of a Number

Every number has a "digital DNA"—its expansion in a specific base (like base 10 or base 3).

The Rule: If a number belongs to the Missing-Digit Set, its DNA must never contain a forbidden digit.
The Problem: Reciprocals of huge numbers (like $1/n!$ ) have very long, complex, and seemingly random DNA sequences. Checking them one by one is impossible.

2. The Detective's Two Tools

To solve this, the author uses two mathematical "detective tools" that look at the structure of the number rather than just its digits.

Tool A: The "Periodicity" Clue (Korobov's Estimate)

When you divide 1 by a number, the digits eventually start repeating in a pattern (like a song on a loop).

The Insight: If a number is missing a specific digit (like the digit 1), the length of this repeating loop cannot be just any length. It has to be "short" relative to the complexity of the number's prime factors.
The Metaphor: Imagine a clock. If the clock is missing the number "1" on its face, the hands can't spin in a way that requires a "1" to exist. This forces the clock's cycle to be much simpler than a normal clock.

Tool B: The "Prime Power" Pressure (p-adic Valuation)

Numbers are built from prime factors (like 2, 3, 5, 7).

The Insight: If a number has a huge power of a specific prime (like $2^{100}$ ), it forces the repeating pattern of its digits to be very long and complex.
The Metaphor: Think of a prime factor as a "weight." If you stack too many heavy weights (a high power of a prime) on a number, it forces the number's "digital DNA" to stretch out and become very long.

3. The Big Contradiction

The paper's main argument is a tug-of-war:

The Missing-Digit Rule says: "If you are in our club, your repeating pattern must be short and simple."
The Fast-Growing Sequence (like factorials) says: "I have so many heavy prime weights that my repeating pattern must be incredibly long."

The author proves that for sequences growing fast enough, the "length required by the weights" eventually becomes longer than the "maximum length allowed by the missing digits."

The Result: The number cannot exist in the club. It's like trying to fit a 10-foot tall giraffe into a 3-foot tall dog house. It's physically impossible. Therefore, after a certain point, none of the reciprocals can belong to the set.

4. Why This is a Big Deal (The "Structural" Twist)

Previous work could only prove this for factorials ($1, 2, 6, 24...$). The author realized that the "tug-of-war" logic works for any sequence, not just factorials.

He even found a case where the old method failed but his new method worked:

The Old Method: Looked at the largest prime factor. It was like saying, "The giraffe is too tall because its head is huge."
The New Method: Looks at the structure of the order. It says, "The giraffe is too tall because its entire skeleton forces it to be long, even if the head isn't the biggest part."

This allowed him to prove the rule works for:

Superfactorials (products of factorials).
Polynomial products (like $(1^2+1)(2^2+1)(3^2+1)...$ ).
Fibonacci products (products of the Fibonacci sequence).
Exponential products (like $(3^1-1)(3^2-1)(3^3-1)...$ ).

Summary

The paper is a masterclass in finding a structural rule that stops a number from having "missing digits."

It's like realizing that no matter how you try to build a tower out of specific blocks, if the tower gets too heavy (too many prime factors), it will inevitably collapse into a shape that includes a forbidden block. The author provides a universal blueprint to prove this collapse happens for almost any rapidly growing sequence of numbers.

The Bottom Line: If you take a number that grows super-fast and flip it upside down ( $1/n$ ), it will eventually become too "complex" to hide in a set that forbids certain digits. The universe of numbers is too messy to stay hidden forever.

1. Problem Statement

The paper addresses the problem of determining the finiteness of the intersection between a set of reciprocals of integers, $\{1/a_n : n \in \mathbb{N}\}$ , and a missing-digit set $K_{m,D}$ .

Missing-Digit Set ( $K_{m,D}$ ): Defined in base $m \ge 3$ using a digit set $D \subset \{0, 1, \dots, m-1\}$ where $1 < |D| < m$ . It consists of numbers whose base- $m$ expansion contains only digits from $D$ :
$K_{m,D} := \left\{ \sum_{j=1}^{\infty} \frac{a_j}{m^j} : a_j \in D \right\}.$
The Question: For a given sequence of integers $\{a_n\}$ , is the set $\{1/a_n\} \cap K_{m,D}$ finite?
Context: Previous work by Lin, Wu, and Yang (2026) established finiteness for the specific case where $a_n = n!$ (factorials) using a "fixed-prime valuation growth" argument. This paper aims to abstract that argument into a general structural criterion applicable to a much broader class of sequences.

2. Methodology

The author's approach relies on two primary mathematical ingredients combined into a new structural obstruction:

A. Korobov's Digit-Distribution Estimate

The paper utilizes a theorem by Korobov regarding the distribution of digits in purely periodic base- $m$ expansions of rational numbers.

Key Insight: If a rational number $r/Q$ (with $\gcd(Q, m)=1$ ) belongs to $K_{m,D}$ , the length of its period, $\text{ord}_Q(m)$ , is constrained by the length of the period of its radical, $\text{ord}_{\text{rad}(Q)}(m)$ .
Lemma 2.1: Establishes that $\text{ord}_Q(m) \le c_{m,D} \cdot \text{ord}_{\text{rad}(Q)}(m)$ , where $c_{m,D}$ is a constant depending only on $m$ and $D$ . This implies that the "complexity" of the period is controlled by the square-free part of the denominator.

B. Order-Lifting Formulas

The paper employs standard number-theoretic lemmas regarding the $p$ -adic valuation of multiplicative orders.

Mechanism: If a denominator $Q$ contains a high power of a fixed prime $p_0$ (i.e., $\nu_{p_0}(Q)$ is large), the multiplicative order $\text{ord}_Q(m)$ must also acquire a high $p_0$ -adic valuation.
Lifting: Specifically, $\nu_{p_0}(\text{ord}_{p_0^k}(m)) \ge k - t$ , where $t$ is a constant overhead depending on $m$ and $p_0$ .

C. The Structural Obstruction

By combining the above, the author derives a contradiction for sequences where the $p_0$ -adic valuation of the denominator grows too fast relative to the $p_0$ -adic valuation of the order of the radical.

The Inequality: For $r/Q \in K_{m,D}$ , the valuation must satisfy:
$\nu_{p_0}(Q) \le t + \nu_{p_0}(\text{ord}_{\text{rad}(Q)}(m)) + \log_{p_0}(c_{m,D}).$
If a sequence $\{a_n\}$ produces denominators where $\nu_{p_0}(Q_n)$ grows faster than the RHS, the intersection must be finite.

3. Key Contributions

1. The Fixed-Prime Criterion (Theorem 3.5)

The paper formalizes a general criterion for finiteness. For a sequence $\{a_n\}$ , let $Q_n$ be the part of $a_n$ coprime to $m$ . If there exists a fixed prime $p_0 \nmid m$ and functions $\alpha(n), \gamma(n)$ such that:

$\nu_{p_0}(Q_n) \ge \alpha(n)$ (valuation growth)
$\nu_{p_0}(\text{ord}_{\text{rad}(Q_n)}(m)) \le \gamma(n)$ (order growth)
And $\alpha(n) > t + \gamma(n) + \log_{p_0}(c_{m,D})$ for large $n$ .

Then, the intersection is finite. This separates the "denominator-level" arithmetic obstruction from the specific sequence properties.

2. The Largest-Prime-Factor Corollary (Corollary 3.6)

A more practical version of the theorem is provided, replacing the complex order term with the largest prime factor $P^+(Q_n)$ .

It states that if $\nu_{p_0}(Q_n)$ grows linearly (or sufficiently fast) while $P^+(Q_n)$ grows only polynomially, finiteness holds.
This recovers and generalizes the Lin-Wu-Yang result for factorials.

3. Structural vs. Coarse Bounds

A major contribution is demonstrating that the structural criterion (using $\text{ord}_{\text{rad}(Q)}$ ) is strictly stronger than the coarse criterion (using $P^+(Q)$ ).

In many cases, the largest prime factor grows exponentially, which would fail the coarse test.
However, the multiplicative order of the radical may grow only logarithmically, allowing the structural theorem to succeed where the coarse one fails.

4. Results and Applications

The paper applies the criterion to several families of sequences:

Factorials ( $n!$ ): Recovers the known finiteness result for $1/n! \in K_{m,D}$ .
Superfactorials ( $\prod_{k=1}^n k!$ ): Proves finiteness. The valuation grows quadratically, easily overwhelming the logarithmic order bound.
Products of Polynomial Values ( $\prod f(k)$ ): Proves finiteness for sequences defined by products of non-constant polynomials.
Products of Fibonacci Numbers ( $\prod F_k$ ): Proves finiteness. This is a "marginal" case where the largest prime factor grows exponentially ( $\approx \phi^n$ ), but the valuation of the product grows linearly ( $\approx \frac{5}{6}n$ ). The criterion holds because the linear growth of the valuation exceeds the logarithmic growth of the order term (scaled by $\log \phi$ ).
Products of $(m^k - 1)$ : This is the critical example distinguishing the two methods.
- Here, $P^+(Q_n)$ can be as large as $m^n - 1$ (exponential).
- However, $\nu_{p_0}(\text{ord}_{\text{rad}(Q_n)}(m))$ is bounded by $\log_{p_0} n$ (logarithmic).
- The coarse $P^+$ -based criterion fails (linear growth of valuation vs. linear growth of $\log P^+$ ), but the structural criterion succeeds because the order term is logarithmic. This proves finiteness for $1/\prod (m^k-1)$ .

5. Significance

Generalization: The paper moves beyond specific sequences (like factorials) to a structural framework applicable to any sequence of reciprocals, provided one can estimate the $p$ -adic valuation of the denominator and the order of the radical.
Refinement of Methods: It highlights that relying solely on the size of the largest prime factor is insufficient for certain rapidly growing sequences. The "radical-order" term is the correct invariant to measure the obstruction imposed by missing-digit sets.
Explicit Computations: The paper provides explicit cutoffs and finite verification procedures (using base- $m$ long division) to determine exactly which small $n$ satisfy the condition, offering concrete results for standard Cantor sets (e.g., identifying exactly which superfactorial reciprocals lie in the middle-third Cantor set).
Limitations: The paper also identifies "non-examples" (primorials and central binomial coefficients) where the fixed-prime method fails because the $p$ -adic valuation of the denominator does not grow sufficiently fast (staying bounded by 1 for square-free sequences or specific subsequences).

In summary, Kominers provides a powerful, reusable arithmetic tool that refines the understanding of rational points in fractal sets, demonstrating that the interplay between prime valuations and multiplicative orders is the decisive factor in determining finiteness.

A Fixed-Prime Criterion for Reciprocals in Missing-Digit Sets