Structural Obstructions in Fixed-Shift Prime Correlations via Mellin-Laplace Kernels
Yung-Hua Chen
TL;DR
The paper develops a Mellin-Laplace analytic framework to study fixed-shift prime correlations r_h(n)=Λ(n)Λ(n+h), demonstrating that the associated Dirichlet series R_h(s) lacks multiplicativity, Euler products, and a pole at s=1. By smoothing with a compactly supported kernel W and working in Re(s)>1, the authors obtain an absolutely convergent Mellin representation for the smoothed sum S_{W,h}(N) and reveal a structural boundary obstruction: the boundary integral on Re(s)=1+ε cannot be decomposed into a dominant main term plus a smaller error term, with the oscillatory part carrying an unavoidable N^{1+ε}log^{2}N growth. This obstruction is intrinsic to the analytic domain and not a technical artifact of kernel choice, explaining why standard contour methods fail to yield a main term for fixed-shift correlations like twin primes. The framework isolates the analytic ingredients available in Re(s)>1 and clarifies what additional arithmetic, spectral, or probabilistic input would be required to overcome the barrier in future work.
Abstract
This paper develops a Mellin-Laplace analytic framework for the fixed-shift prime correlation r_h(n) = Lambda(n) Lambda(n+h) for h not equal to 0. This sequence has no multiplicative structure, no Euler product, and no singularity at s = 1. For every compactly supported Mellin-Laplace admissible kernel W, the smoothed shifted sum S_{W,h}(N) admits an absolutely convergent Mellin representation that holds entirely in the half-plane Re(s) > 1, with no use of analytic continuation. The Mellin transform of W provides quantitative vertical decay, enabling full contour control on the boundary line Re(s) = 1 + eps. A Tauberian boundary analysis shows that both components of the boundary integral grow like N^{1+eps}, while the oscillatory part contributes an unavoidable N^{1+eps} (log N)^2 term. As a result, the boundary integral cannot be decomposed into a dominant main term plus a smaller error term, revealing a structural obstruction to main-term extraction for fixed-shift correlations. These results give a complete analytic description of shifted prime correlations in their natural domain of convergence and clarify the analytic difficulties underlying problems such as the twin prime conjecture.
