Fixed-Perimeter Franklin Statistics and an Eventual Inequality

summary

The gist

Fixed-perimeter partition statistics are analyzed to prove an eventual inequality between two related counts for fixed-perimeter partitions…

In short

The paper investigates fixed-perimeter partition statistics to prove an eventual inequality between two counts: FDj,k(n) and FOj,k(n). It shows that for $k ext{ ≥ } 3$, the number of partitions with exactly $j$ part sizes of multiplicity at least $k$ eventually grows strictly faster than the number having exactly $j$ distinct part sizes divisible by $k$. This is proven by comparing the dominant poles of their generating functions.

Key concepts

FDj,k(n)
This statistic counts partitions of perimeter n that have exactly j part sizes where each size appears at least k times (multiplicity $\ge k$). It is one of the two quantities being compared in the study.
FOj,k(n)
This statistic counts partitions of perimeter n that have exactly j distinct part sizes which are all divisible by k. This is the other quantity used to establish the eventual inequality.
Dominant Pole Structure
The growth rate of both statistics is determined by the dominant pole (the root with the largest magnitude) of their respective bivariate generating functions. The paper shows that for $k \ge 3$, these poles, $\rho_k$ and $\sigma_k$, are positive real numbers, and $\rho_k < \sigma_k$. This difference in pole locations dictates the eventual inequality.
Profile Word Encoding
A unique way to represent a partition is by encoding it as a profile word. This involves listing the multiplicities of part sizes up to the largest part $\alpha$. The perimeter is calculated based on this encoding, and this representation is used to derive the compact generating functions for the statistics.

Terminology used across episodes

This episode discusses

The paper

Fixed-Perimeter Franklin Statistics and an Eventual Inequality · Read on arXiv

MARCUS MCCREA

Gray, Payne, Swisher, and Watson conjectured that, for fixed integers j at least 0 and k at least 2, the number FD j,k(n) of partitions of perimeter n having exactly j part sizes of multiplicity at least k is eventually at least the number FO j,k(n) having exactly j distinct occurring part sizes divisible by k. We derive bivariate generating functions for both statistics using the profile-word encoding of a partition. For k at least 3, coefficient extraction shows that fixing j changes the order of the dominant pole but not its location. The corresponding dominant singularities are positive real numbers ρ k and σ k, where ρ k satisfies ρ k+ρ k 2+ +ρ k k=1 and σ k satisfies σ k k=(1-σ k) k-1. We prove that ρ k<σ k for every k at least 3. Consequently, FD j,k(n)/FO j,k(n) to infinity as n to infinity, proving the conjecture and yielding a strict eventual inequality for k at least 3. The case k=2 recovers the known exact identity.

Transcript

Introduction to the show: ident: Mathematics. The day in mathematics: new theorems, new proofs, and what they open.

Ada: I'm Ada.

Emmy: And I'm Emmy, here with Kolya.

Ada: We're looking at the paper: "Fixed-Perimeter Franklin Statistics and an Eventual Inequality".

Emmy: Here's the gist of the paper. Fixed-perimeter partition statistics are analyzed to prove an eventual inequality between two related counts for fixed-perimeter partitions, which establishes a strict eventual inequality for certain parameters Pruning redundant reasoning tokens The conjecture states that for fixed integers j ≥ 0 and k ≥ 2, the number FDj,k(n) of partitions of perimeter n having exactly j part sizes of multiplicity at least k is eventually at least the number FOj,k(n) having exactly j distinct part sizes divisible by k The paper derives compact bivariate generating functions for both statistics from the profile-word encoding of a partition For k ≥ 3, coefficient extraction shows that fixing j changes only the order of the dominant pole, not its location The two dominant poles are positive real numbers ρk and σk determined by ρk + ρ 2 k + · · · + ρ k k = 1, σ k k = (1 − σ n k) k−1 The proof shows that ρk < σk for every k ≥ 3, which proves the conjecture and gives a strict eventual inequality for k ≥ 3 The case k = 2 recovers the known exact identity How the training signal is built The profile word encoding of a partition with largest part α can be written uniquely as π = 1m1 2 m2 · · ·(α − 1)mα−1α mα+1, m1,..., mα ≥ 0 Its profile word is ENm1ENm2 · · · ENmα N The perimeter is Γ(π) = α + X α i=1 mi In the generating functions below, a block ENm receives weight q(m+1) and the final N receives weight 1 The generating function for FDj,k(n) is given by Dk(z, q) = q(1 − (− z)q k−1 / 1 − 2q + (− z)q k+1 Where the gains come from The key compression in the generating functions is summing a complete period of k possible part sizes The formula for Ok(z, q) is given by Ok(z, q) = A + A squared + · · · + Ak−1 + zAk / 1 − Ak This structure arises from conditioning on the largest part α The generating function for FDj,k(q) is derived by expanding Dk(z, q) and extracting the coefficient of z j Pruning redundant reasoning tokens The dominant pole structure for both statistics is independent of j, reducing the conjecture to a comparison of two positive real roots The FD denominator factors as Hk(q) = (1 − q) 1 − q − q squared − · · · − q k The unique positive solution ρk of Hk(q) = 0 is the dominant pole for FDj,k(n) On the FO-side, Kk(q) = 1 − q k (1 − q)(k−1) The unique positive solution σk of Kk(q) = 0 is the dominant pole for FOj,k(n) Lemma 4.1 proves that ρk < σk for every k ≥ 3 Where the gains come from The coefficient growth estimates show that FDj,k(n) ∼ CD j,kn j / ρ(-n) k and FOj,k(n) ∼ CO j,kn j / σ(-n) k Dividing these asymptotic expressions yields the ratio FDj,k(n)/FOj,k(n) ∼ CD j,kn j / (CO j,kn j) sigma k / ρ k n Since σk/ρk > 1 by theorem 4.1, the ratio tends to +∞ proving the theorem The corollary confirms that for k ≥ 3, the inequality is strict for all sufficiently large n Pruning redundant reasoning tokens For k = 2, direct simplification of the generating functions recovers the exact identity FOj,2(n) = FDj,2(n) The proof isolates a simple structural principle where fixing j raises the order of the dominant pole by j but leaves its location unchanged This separation is controlled entirely by the two unmarked profile kernels The root gap ρk < σk then creates an exponential separation between the two fixed-j statistics This suggests that other fixed-perimeter statistics encoded by periodic profile weights may admit similar eventual comparisons through their zero-defect growth kernels The references are

1: S.

Ada: Fu and D.

Emmy: Tang, Partitions with fixed largest hook length, Ramanujan J.

Ada: 45 (2018), 375–390

1: G.

Paper summary: Ada: So wrapping up the paper "Fixed-Perimeter Franklin Statistics and an Eventual Inequality," they establish this eventual inequality for fixed integers j greater than or equal to zero and k greater than or equal to three. It proves that FDj,k(n) will be strictly greater than FOj,k(n) once n gets large enough.

Emmy: The paper is interesting because it takes a complex counting problem and reduces it down to comparing the locations of two dominant poles in generating functions, which shows a deep structural relationship between these two partition statistics. The authors use the profile-word encoding to get those simple rational bivariate generating functions, which is key for that reduction.

Kolya: For me, what I found most significant is the explicit comparison of rho k and sigma k and how they prove that one is smaller than the other for k greater than or equal to three. That mathematical proof of rho less than sigma really solidifies why this divergence happens eventually.

Ada: And it does matter because for anyone working on partition theory with fixed perimeter constraints, this gives them a concrete direction—they know that when k is three or higher, they can expect one count to grow faster than the other.

Emmy: It’s about showing that the structure of the profile word encoding dictates this separation, and it hints that similar eventual comparisons might be found in other fixed-perimeter statistics if they also have zero defect growth kernels.

Kolya: So basically, for listeners, it’s a result that shows an eventual gap between two types of partition counts based on how many parts repeat versus how distinct the parts are when they have some divisibility property. That's the main thing we got here.

Conclusion: Ada: So we've looked at how these partitions behave when you fix the perimeter, and now we need to wrap up what this paper is actually saying about those counts.

Emmy: The title is "Fixed-Perimeter Franklin Statistics and an Eventual Inequality." It sounds really technical, but it’s actually about proving that two different ways of counting partitions eventually separate themselves.

Ada: Exactly. It shows that for these specific statistics, FDj,k(n) will eventually be bigger than FOj,k(n) when you look at very large numbers n. And this happens for integers j greater than or equal to zero and k greater than or equal to three.

Kolya: So what does that mean in practice? It means if you're dealing with partitions where you restrict the total perimeter, and you're looking at how many parts have a certain multiplicity versus how many parts are distinct and divisible by some number, eventually one count will outpace the other consistently as n grows.

Emmy: The structure behind this is really neat. They used profile-word encoding to get these generating functions that let them compare the growth rates by looking at the dominant poles in those functions. It’s like finding two specific points on a map and proving one point is always further away than the other for large distances.

Ada: And for k greater than two, they prove that this separation isn't just an approximation; it's a strict eventual inequality. The core mechanism relies on showing that the root of one generating function is always smaller than the root of the other one across all relevant parameters.

Kolya: That comparison between rho k and sigma k is what really gives it weight. It’s not just a guess; they found a rigorous way to show that for every k three or higher, this relationship holds true, which means that as n gets huge, the ratio of FDj,k(n) over FOj,k(n) goes to infinity.

Emmy: That comparison between those two positive real numbers is where the algebraic geometry comes in. It’s showing you the underlying structure—that for these specific constraints, one counting mechanism inherently grows faster than the other as you scale up.

Ada: So they’ve established this eventual gap between these two fixed-perimeter counts for k greater than two. But what does this tell us about the whole world of partition theory? It suggests that even in highly constrained counting problems, there are inherent structural differences in how different properties manifest at large scales.

Kolya: It’s a nice result because it shows that constraints don't just flatten everything out; they can actually create predictable, long-term growth patterns between related statistics. This opens the door to seeing if we can find similar separation results in other types of partition problems using this same generating function approach.

More episodes

← Home