Concentrated Gibbs Slices and Zeta Function Truncation: A Speculative Computational Analogy

Abstract

This essay examines the structural pattern of concentration-based truncation and high-dimensional expansion developed by Li and Yang for approximating antiferromagnetic Ising partition functions on random regular bipartite graphs.

Source Structure and Algorithmic Paradigm

The source paper by Li and Yang designs a Fully Polynomial Randomized Approximation Scheme (FPRAS) for the partition function Z_G(λ, β) of antiferromagnetic Ising models on random Δ-regular bipartite graphs. The key mathematical structures are:

• Exponential concentration: The Gibbs distribution exhibits exponential decay for configurations that are "large" on either side of the bipartition, quantified by Lemma 3.5 and Corollary 3.6.
• Slicing: The configuration space Ω = 2^V is decomposed into slices Ω_k = {S ⊆ V : |S| = k} (or conditioned on both sides). The partition function is approximated by summing over a truncated range of slices where concentration holds.
• High-dimensional expansion: On each slice, rapid mixing of down-up random walks is established via "trickle-down" theorems, yielding efficient samplers for the conditional distributions.

This creates a computational template: truncate using concentration bounds, then sample within slices using high-dimensional expanders.

Proposed Analogy to the Riemann Hypothesis

We speculate that this template suggests a computational framework for the Riemann zeta function ζ(s). The analogy maps:

• Ising configurations S ⊆ V ↔ Integers n ∈ ℕ (summation indices in Dirichlet series)
• Configuration size |S| ↔ Logarithmic height log n
• Partition function Z_G ↔ ζ(s) (or partial sums thereof)
• Gibbs measure concentration ↔ Decay of zeta tails or variance concentration in truncated explicit formulas
• High-dimensional expansion ↔ Spectral control of the de Bruijn–Newman heat flow or GUE statistics of zeros

The strength of this analogy is rated as a suggestive metaphor: the qualitative structure of truncation-and-sampling is shared, but the formal axioms (finite state space vs. infinite series, graph geometry vs. complex analysis) differ fundamentally.

Proposed Computational Experiments

We propose two unexecuted experiments:

1. Concentration testing: Compute variance statistics for truncated partial sums Σ_{n=N}^{2N} n^{−1/2−it} for random t, testing whether the variance exhibits exponential decay in the "slice" parameter log N analogous to the Ising concentration bounds.
2. Sliced heat flow integration: Implement a Monte Carlo evaluation of the de Bruijn–Newman heat flow H_t(x) using dyadic slicing of the integral, testing whether the error bounds mimic those predicted by high-dimensional expansion theory.

Obstructions and Honest Assessment

The analogy faces a critical obstruction: the Ising model has a finite configuration space permitting exact concentration inequalities, whereas ζ(s) involves an infinite Dirichlet series where truncation always incurs residual error. Furthermore, the integers lack the high-dimensional simplicial complex structure required for the trickle-down theorems. We conclude that while the algorithmic design pattern is suggestive, no structural isomorphism exists between the two domains.

This essay was produced by an automated research pipeline and has not been peer reviewed; conjectures herein are unproven.