The Simplicial Geometry of Integer Partitions: An Exact $O(1)$ Formula via $A_{k-1}$ Root Systems
Antonio Bonelli
TL;DR
The paper tackles the exact evaluation of the partition function $p_k(n)$ by embedding the problem in the geometry of the $A_{k-1}$ Weyl chamber and applying a Simplicial Successive Decomposition (SSD). It establishes a unimodular triangulation into $N_k=\binom{k}{2}$ simplices, and derives an exact $O(1)$ closed-form formula using Brion's localization and a Betke-Kneser valuation framework, with residue shifts from a Residue Matrix $\mathcal{R}$. Key contributions include the Triangular Cardinality theorem, the SSD closed-form, Ehrhart–Macdonald reciprocity, and the Core Collapse regime, all supported by both theoretical proof and a $k=12$ experimental validation. The work provides a geometric, deterministic mechanism for partition enumeration, reducing the apparent combinatorial complexity to a fixed simplicial basis tied to the root system $A_{k-1}$, with practical implications for exact counting in high-dimension regimes where continuous approximations fail.
Abstract
We present a structural resolution to the exact evaluation of the partition function $p_k(n)$, addressing the limitations of traditional recursive and asymptotic methods. By introducing the Simplicial Successive Decomposition (SSD) framework, we demonstrate that the partition polytope $\mathcal{P}_{n,k}$ is not an arbitrary geometric object, but admits a rigid minimal unimodular triangulation into exactly $N_k = \binom{k}{2}$ simplices. This cardinality is determined by the positive root system of the $A_{k-1}$ Weyl chamber.We decompose Euler's generating function into a finite sum of simplicial rational transforms. By applying Brion's localization theorem and the negative binomial expansion, we derive an exact closed-form formula with $O(1)$ computational complexity. The validity of the model is confirmed through Ehrhart-Macdonald reciprocity, ensuring accuracy in the "Core Collapse" regime where the polytope's interior is empty and continuous volume approximations are inapplicable.
