Conditional Approaches to the Binary Goldbach Conjecture: Spectral Hypotheses and a Directional Spectral Non-Concentration Assumption
Year: 2026
Type: preprint
Abstract: We analyse the conditional spectral approach to the binary Goldbach conjecture and isolate the minimal additional hypothesis needed for the conclusion. On top of GRH, a recent preprint (Liu–Li, 2026) constructs a compact self-adjoint Hilbert–Schmidt operator T_{N,c} on a Hardy space with the exact identity r(N) = c_φ S(N) − G(N) = ⟨T_{N,c} v_{N,c}, v_{N,c}⟩, and proposes spectral hypotheses S and R yielding the Hardy–Littlewood asymptotic for every sufficiently large even N. With the framework inputs stated explicitly, we prove: the arithmetic identity separates from the encoding layer (an unconditional finite-DFT implementation reproduces Liu–Li's Appendix B numerics); the exact decomposition G(N) = Σ s_n² − Σ a_n² = (1/φ(N)) Σ_χ χ(−1)|P_χ|² locates the parity barrier in this subtraction, and G(N) > 0 only means "N is a sum of two prime powers", the last counterexample being N = 68; an explicit mean-value theorem, and the structural impossibility of closing the conjecture by finite verification. We introduce the directional spectral non-concentration hypothesis D, constraining only the maximal spectral weight ρ_N of the Riesz vector in an eigenbasis, and prove |r(N)| ≤ √ρ_N · Q_N · ‖T_N‖_HS. Hence ρ_N < ρ_req(N) implies the Goldbach conclusion, and ρ_N ≤ K(log N)^(−4−ε) implies the full asymptotic. D is strictly weaker than S and R with zero waste (61× and 1.37×10⁵× respectively too strong). Finally we characterise D itself: Q_N increases monotonically to the explicit constant Q_∞ = 5.94678142, giving the closed-form threshold ρ_req(N) ≍ K(log N)^(−4); D requires rank(T_N) > 2184·C_HS²·(log N)⁴ effective eigendirections; a 2×2 classification against R via four explicit finite-dimensional models shows D is sufficient but not necessary and not a repackaging of the conclusion; the participation ratio alone yields only ρ_N ≳ d_eff^(−1/2) against the required d_eff^(−1), the gap being precisely an eigenfunction anti-concentration (Type-II) inequality; and a c-optimised variant D* relaxes the threshold by at most e⁴ ≈ 54.6. A real-arithmetic Gram proxy (2880 odd characters mod 30030, CRT-FFT Dirichlet series) shows ρ decays only as m^(−0.52), so delocalisation cannot be manufactured by node count. All theorems are conditional on GRH and the stated hypotheses; this paper does not constitute a proof of the binary Goldbach conjecture. Keywords: Goldbach conjecture; Riemann hypothesis; Hardy space; Hilbert–Schmidt operator; spectral non-concentration; parity barrier; Type-II information; singular series; eigenfunction anti-concentration; conditional proof.
Author: Zeng, Chunping
Language: English
Subfield: Algebra and Number Theory
Field: Mathematics
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7237360935
Cites:
Cited by:
Related to:
