Regular ternary sums of generalized polygonal numbers

This paper establishes the existence of an explicit constant $C$ beyond which no regular ternary sum of generalized $m$-gonal numbers exists for any integer $m$.

The Gist

Imagine you are a master builder trying to construct a massive, infinite tower using only specific types of Lego bricks. These aren't just any bricks; they are "Polygonal Bricks."

In the world of numbers, a "Polygonal Number" is a number that can be arranged into a perfect geometric shape.

• Triangular numbers (like 1, 3, 6, 10) form triangles.
• Square numbers (1, 4, 9, 16) form squares.
• Pentagonal numbers form pentagons, and so on.

The ancient mathematician Fermat once claimed that if you have enough of these specific bricks, you can build any positive integer (1, 2, 3, 4...). This is true, but it might take you a lot of bricks.

The Big Question: The "Three-Brick" Limit

This paper asks a very specific, tricky question: Can we build every number using exactly three of these Polygonal Bricks?

Mathematicians call this a "ternary sum."

• If you use triangular bricks (3-sided), you can build everything with three. (This was proven by Gauss).
• But what if you use 100-sided bricks? Or 1,000-sided bricks?

The author, Mingyu Kim, is investigating whether there is a "tipping point." Is there a maximum size for the polygon (let's call it $m$) beyond which it becomes impossible to build every number using just three bricks?

The Detective Work: "Local" vs. "Global"

To solve this, the author uses a clever detective strategy involving two types of clues:

1. Local Clues (The Neighborhood Check): Imagine you are checking if a number can be built. You check if it can be built in "neighborhoods" (mathematical systems based on specific prime numbers like 2, 3, 5, etc.). If a number passes the test in every neighborhood, we say it is "locally represented."
2. Global Clues (The Whole Picture): This is the actual construction. Can you really build the number using the three bricks?

A "Regular" form is a magical set of three bricks that is so perfect that if a number passes the "Neighborhood Check" (Local), it is guaranteed to pass the "Whole Picture" check (Global).

The Problem: For very large polygons (huge $m$), the "Neighborhood Check" becomes too easy. You can find numbers that pass the local test but fail the global test. The author wants to prove that for polygons larger than a certain size, no such "perfect" set of three bricks exists.

The Analogy: The "Impossible Lock"

Think of the "Regular Ternary Sum" as a Master Key.

• For small polygons (like triangles or squares), the Master Key works perfectly. If a door looks unlocked from the outside (local), the key opens it (global).
• As the polygons get bigger and more complex (higher $m$), the locks get more complicated.
• The author proves that eventually, the locks get so weird that no single Master Key exists that can open every door that looks open from the outside.

The "Watson Transformation": The Magic Shrink Ray

How did the author prove this? They used a mathematical tool called a Watson Transformation.

Imagine you have a giant, messy pile of bricks that you think might be a Master Key. The Watson Transformation is like a Magic Shrink Ray.

1. It takes your big, messy pile of bricks.
2. It shrinks it down into a smaller, cleaner pile.
3. Crucially, if the big pile was a "perfect" key (regular), the small pile must also be a "perfect" key.

The author uses this ray repeatedly. They keep shrinking the problem until they reach a point where the math simply breaks. They show that if you assume a perfect key exists for a huge polygon, the shrinking process eventually forces the numbers to be so small and contradictory that the assumption must be false.

The Verdict: The "C" Constant

The paper concludes with a specific "Stop Sign." The author calculates a constant $C$.

• If you try to build a "perfect three-brick system" for a polygon with more than $C$ sides, it is mathematically impossible.
• The value of $C$ depends on the shape of the polygon.
  • For some shapes, the limit is around 35.
  • For others, it's around 147.
  • For the most complex shapes, the limit is 712.

In plain English:
If you try to make a rule that says "Any number that looks buildable locally is buildable globally" using three 713-sided polygonal numbers, you will fail. The universe of numbers is too complex for such a simple rule to work for shapes that big.

Why Does This Matter?

This isn't just about counting Lego bricks. It's about understanding the hidden structure of numbers.

• It tells us the limits of how "predictable" numbers can be.
• It connects geometry (shapes) with algebra (equations) in a deep way.
• It proves that while math is full of beautiful patterns, there are hard boundaries where those patterns break down.

The Takeaway:
Just because something looks like it should work in every little test (locally), it doesn't mean it works in the real world (globally). And for shapes with too many sides, this gap becomes unbridgeable when you only have three pieces to work with.

Technical Summary

1. Problem Statement

The paper addresses the classification of regular ternary sums of generalized $m$-gonal numbers.

• Definitions:
  • An integer $n$ is an $m$-gonal number if $n = P_m(u) = \frac{(m-2)u^2 - (m-4)u}{2}$ for some $u \in \mathbb{N}$.
  • A generalized $m$-gonal number allows $u \in \mathbb{Z}$.
  • A ternary $m$-gonal form is a polynomial $f(x_1, x_2, x_3) = a_1 P_m(x_1) + a_2 P_m(x_2) + a_3 P_m(x_3)$ with $a_i \in \mathbb{N}$.
• Regularity: The author adopts the definition that a form $f$ is regular if it represents every non-negative integer that is locally represented by $f$ (i.e., represented over $\mathbb{Z}_p$ for all primes $p$ and over $\mathbb{R}$).
• The Core Question: While it is known that for a fixed $m$, there are only finitely many regular ternary $m$-gonal forms, the existence of such forms for arbitrarily large $m$ was an open question. The paper aims to prove that there exists an absolute constant $C$ such that no regular ternary $m$-gonal form exists for any $m > C$.

2. Methodology

The proof relies on transforming the problem of representing polygonal numbers into the theory of integral quadratic forms and lattices, specifically utilizing Watson transformations and local-global principles.

A. Transformation to Quadratic Forms

The author establishes an equivalence between the representation of an integer $n$ by the $m$-gonal form $f$ and the representation of a transformed integer by a complete ternary quadratic polynomial $g$.

• By defining constants $\delta, c, d, \mu$ based on $m \pmod 4$ and $m \pmod 3$, the equation $f(\mathbf{x}) = n$ is equivalent to:
  $$g(\mathbf{y}) = \mu n + d^2(a_1+a_2+a_3)$$
  where $g(\mathbf{y}) = a_1(cy_1 - d)^2 + a_2(cy_2 - d)^2 + a_3(cy_3 - d)^2$.
• This transforms the problem into studying tight regular Z-cosets (lattices shifted by a vector). A form is regular if and only if the associated Z-coset is "tight regular" (represents all locally represented integers $\ge$ its minimum).

B. Watson Transformations

The proof utilizes Watson transformations ($\Lambda_p$ and $\lambda_p$) to manipulate the underlying lattice $L$.

• If a regular form exists, there exists a "primitive tight regular" Z-coset $L+v$ with conductor $c$.
• By applying Watson transformations modulo primes $p$ not dividing the conductor, the author constructs a new lattice $K$ that is $p$-stable (locally represents all units or specific classes) while preserving the conductor.
• This process allows the author to assume the existence of a "minimal" regular form with specific stability properties at all primes.

C. Counting Arguments and Bounds

The core of the proof involves deriving contradictions for large $m$ (and thus large conductor $c$) by comparing:

1. Lower Bounds: The number of integers locally represented by the form (derived using Lemma 2.7 and the function $\psi_p(n)$).
2. Upper Bounds: The number of integers globally represented by the form, constrained by the coefficients $a_1, a_2, a_3$.

• The author defines a function $\eta(n, s)$ representing the minimum number of integers in a sequence that are locally represented but potentially not globally represented if the coefficients are small.
• By assuming $c$ is large, the author shows that the form must represent a specific number of integers. However, if the coefficients $a_i$ are small (which they must be if the form is to be regular for large $c$), the form cannot represent that many distinct values, leading to a contradiction.

3. Key Contributions

1. Explicit Upper Bound: The paper provides an explicit constant $C$ such that no regular ternary $m$-gonal form exists for $m > C$. This resolves the finiteness question for the existence of such forms across all $m$.
2. Case-by-Case Analysis: The proof is divided into four cases based on the congruence of $m$ modulo 4 and 3. For each case, the author derives specific bounds for the conductor $c$ and the number of "bad" primes $t$ (primes where the lattice is anisotropic).
3. Refinement of Regularity: The paper clarifies the distinction between "strict" regularity (representing all locally represented integers, including negatives) and "non-negative" regularity. It argues that for $m$-gonal forms, the non-negative definition is the only viable one because these forms cannot represent negative integers globally, even if they do so locally.

4. Main Results (Theorem 1.1)

The paper proves that if a regular ternary $m$-gonal form exists, then $m$ must satisfy the following upper bounds:

Condition on $m$	Upper Bound for $m$
$m \equiv 1 \pmod 2$ and $m \equiv 2 \pmod 3$	35
$m \equiv 1 \pmod 2$ and $m \not\equiv 2 \pmod 3$	147
$m \equiv 2 \pmod 4$ and $m \equiv 2 \pmod 3$	38
$m \equiv 2 \pmod 4$ and $m \not\equiv 2 \pmod 3$	142
$m \equiv 0 \pmod 4$ and $m \equiv 2 \pmod 3$	188
$m \equiv 0 \pmod 4$ and $m \not\equiv 2 \pmod 3$	712

Consequently, no regular ternary $m$-gonal form exists for $m > 712$.

5. Significance

• Finiteness of Regular Forms: This result completes the classification program for regular ternary sums of polygonal numbers. It confirms that the set of all such forms is finite, not just for a fixed $m$, but globally across all $m$.
• Methodological Advancement: The paper successfully adapts the "Watson transformation" technique, traditionally used for quadratic forms, to the context of generalized polygonal numbers (which are quadratic forms with linear shifts). This bridges the gap between classical lattice theory and the study of polygonal number representations.
• Computational Feasibility: By establishing an explicit bound ($m \le 712$), the paper implies that the complete list of all regular ternary $m$-gonal forms is computationally enumerable, paving the way for future exhaustive classification projects.

In summary, Mingyu Kim's work provides a definitive theoretical limit on the existence of regular ternary sums of generalized polygonal numbers, utilizing deep tools from the arithmetic of quadratic forms to establish strict numerical bounds.