Dyson’s partition ranks and their multiplicative extensions

Elaine Hou, Meena JagadeesanThe Ramanujan Journal 2017 · DOI

TL;DR

This paper investigates the "rank" of integer partitions, a statistic introduced by Dyson to explain certain congruences in partition theory. The authors extend this rank concept multiplicatively and prove a "convexity" property for these extended ranks, showing that the product of ranks for smaller partitions is greater than the rank of their combined partition. This work provides a deeper combinatorial understanding of partition properties and extends known results about the partition function.

What Problem Does This Paper Address?

The study of integer partitions, which are ways to write a positive integer as a sum of positive integers, has a rich history. Ramanujan famously discovered congruences for the number of partitions of certain integers, like $p(5n+4) \equiv 0 \pmod{5}$. Freeman Dyson sought a combinatorial explanation for these, introducing the "rank" of a partition as the largest part minus the number of parts. He conjectured that partitions could be grouped by their rank modulo $t$ to explain these congruences. While Dyson's rank successfully explained congruences modulo 5 and 7, it didn't fully explain those modulo 11.

This paper builds on Dyson's work by studying the "rank modulo $t$" function, denoted $N(r, t; n)$, which counts partitions of $n$ with rank congruent to $r$ modulo $t$. A key question is whether these rank functions exhibit similar properties to the partition function $p(n)$ itself. Specifically, the paper addresses:

  1. Convexity of Rank Functions: The partition function $p(n)$ satisfies a convexity property: $p(a)p(b) > p(a+b)$ for large enough $a, b$. The authors investigate if $N(r, t; n)$ also satisfies this property.
  2. Multiplicative Extension: The paper extends the concept of $N(r, t; n)$ from individual integers to partitions themselves. This involves defining a multiplicative extension of $N(r, t; \lambda)$ for a partition $\lambda = (\lambda_1, \dots, \lambda_k)$ as $\prod_{j=1}^k N(r, t; \lambda_j)$.
  3. Maximizing the Multiplicative Extension: For a given integer $n$, the authors aim to find the maximum value of this multiplicative extension over all partitions of $n$, and to identify which partitions achieve this maximum.

Methods

The paper employs a combination of analytic number theory, combinatorial arguments, and computational verification.

  • Analytic Bounds: The authors utilize explicit bounds for the partition function $p(n)$ derived by Lehmer and Bringmann. These bounds are crucial for establishing inequalities involving $N(r, t; n)$.
  • Error Term Analysis: A significant portion of the technical work involves bounding various "error terms" ($E_i(n)$) that arise in the analytic formulas for $p(n)$ and related functions. These bounds are essential for proving the convexity property.
  • Convexity Proof: Theorem 1.1, which establishes the convexity of $N(r, 3; n)$, is proven by first deriving explicit upper and lower bounds for $N(r, 3; n)$ using the analytic estimates. Then, Lemma 2.8 provides a general framework for proving $N(r, t; a)N(r, t; b) > N(r, t; a + b)$ for sufficiently large $a, b$, which is then applied to the specific case of $t=3$. The remaining cases for smaller $a, b$ are verified computationally.
  • Combinatorial Analysis for Maximum: To find the maximum of the multiplicative extension $N(r, 3; \lambda)$, the authors use a detailed combinatorial approach. This involves:
  • Bounding Partition Parts: Propositions are used to bound the size of the largest part ($\lambda_1$) and the multiplicities of other parts in partitions that achieve the maximum.
  • Case Analysis: The problem is broken down into cases based on the value of $r$ (0, or 1 and 2) and the residue of $n$ modulo $t$.
  • Recursive Structure: The analysis often involves considering partitions obtained by removing specific parts (like 7 or 14) and then re-appending them or their sums, using the results from Theorem 1.1 and computational data.
  • Computational Verification: Extensive use of SageMath is made to compute values of $N(r, t; n)$ for small $n$ and to verify specific inequalities and partition properties.

Datasets

The paper does not introduce new datasets. Instead, it relies on:

  • Pre-computed values of $N(r, t; n)$: The authors use values of $N(0, 3; n)$ and $N(1, 3; n)$ for small $n$ (up to $n=32$ and $n=21$ respectively) computed using SageMath. These are presented in Tables 3.1 and 3.2.

Key Insights and Ideas

  • Convexity of Rank Functions: A significant finding is that Dyson's rank functions $N(r, t; n)$ exhibit a convexity property similar to the partition function $p(n)$. This suggests a deeper structural similarity between these functions and implies that the "rank" statistic, when considered modulo $t$, behaves in a predictable, "well-behaved" manner for large inputs.
  • Multiplicative Extension as a Measure of "Ranked Partition Structure": Extending $N(r, t; n)$ multiplicatively to partitions $\lambda$ as $\prod N(r, t; \lambda_j)$ provides a way to quantify the "rank structure" of a partition itself. Maximizing this product reveals partitions that are "most representative" of a particular rank modulo $t$.
  • Explicit Formulas for Maximum Rank Values: The paper provides explicit, albeit complex, formulas for the maximum value of $N(r, 3; n)$ for different residues of $n$ modulo 7 and 14. This is a concrete result that characterizes the peak behavior of these rank functions.
  • Identification of Maximizing Partitions: Crucially, the paper not only finds the maximum values but also identifies the specific partitions that achieve these maxima. These partitions often exhibit regular structures, like repeated parts (e.g., $(7,7,\dots,7)$ or $(14,14,\dots,14)$), which is a testament to the underlying order in partition theory.
  • Bridging Analytic and Combinatorial Methods: The work effectively bridges analytic number theory (using bounds and asymptotic formulas) with combinatorial arguments and computational verification to prove its theorems.

Key Results

  • Theorem 1.1: Proves that for $t=3$, the rank functions $N(r, 3; n)$ satisfy the convexity property $N(r, 3; a)N(r, 3; b) > N(r, 3; a + b)$ for sufficiently large $a, b$ (specific bounds are given for $r=0, 1, 2$).
  • Theorem 1.2: Provides explicit formulas for the maximum value of the multiplicative extension $maxN(r, 3; n)$ for $r=0$ (modulo 7) and $r=1, 2$ (modulo 14). It also identifies the unique partitions that achieve these maximum values. For example, for $r=0$ and $n \ge 33$, $maxN(0, 3; n)$ depends on $n \pmod{7}$ and is achieved by partitions like $(7,7,\dots,7)$ or $(13,13,10,7,\dots,7)$.
  • Explicit Bounds for $N(r, 3; n)$: Proposition 2.7 provides bounds for $N(r, 3; n)$ in terms of $p(n)$, showing $N(r, 3; n)$ is roughly one-third of $p(n)$ for large $n$.

Limitations and Future Work

  • Scope of Convexity: The convexity result (Theorem 1.1) is proven for $t=3$ and for sufficiently large $a, b$. The paper notes that the remaining cases for smaller $a, b$ were verified computationally. Extending this to arbitrary $t$ and smaller $a, b$ remains an open problem.
  • Generalization to other $t$: The paper discusses the difficulty of obtaining effective asymptotics and bounds for $N(r, t; n)$ for general $t$. The authors suggest that similar methods might be used to prove convexity for $t=2$ (Conjecture 4.1) and for general $r, t$ (Conjecture 4.3).
  • Computational Verification: While computational verification is a powerful tool, it inherently limits the generality of the results to the range tested.
  • Ramanujan's Modulo 11 Congruence: The paper notes that Dyson's rank does not fully explain Ramanujan's congruence modulo 11, suggesting that further combinatorial statistics or extensions might be needed for a complete explanation.
  • Future Work: The authors explicitly mention exploring extensions to other values of $t$ and proving the conjectured convexity results for $t=2$ and general $t$. They also suggest that their methods could be applied to study other related partition functions.

Summary updated

BibTeX
@article{hou2017dysons,
  author = {Elaine Hou and Meena Jagadeesan},
  title = {{Dyson’s partition ranks and their multiplicative extensions}},
  journal = {The Ramanujan Journal},
  year = {2017},
  doi = {10.1007/s11139-016-9881-2},
}
Figure