Finite Spectral Contracts for Riemann Hypothesis Approaches: The Bass–Ihara Ramanujan Dichotomy (5-Graph Tower & 14-Vertex Non-Ramanujan Witness), Jensen–Pólya Transverse QFIM Stiffness, Mpemba–Trajectory–dBN Synthesis, and High-Precision Euler–Maclaurin / Riemann–Siegel \(C_4(p)\) Jet Audits in Lean 4, Rust/PyO3, and SymPy¶
Authors: Wes Turner
Subjects:
Number Theory (
math.NT); Spectral Theory (math.SP); Combinatorics (math.CO); Mathematical Physics (math-ph); Logic & Formal Verification (cs.LO)
Epistemic Status & Scope Declaration:
Solved & Machine-Checked (Finite Graph RH): Both directions of the finite-graph Bass–Ihara Ramanujan dichotomy [Bass, 1992] are proved in Lean 4 (
Signals.GrassmannianEBR.ihara_ramanujan_critical_circle_modulus_sqandSignals.RiemannSpectral.ihara_non_ramanujan_splits_off_critical_circle), including an existential, floating-point-free Lean 4 proof that the 14-vertex 3-regular connected Barbell graph \(G_{14}\) (\(q = 2\)) exceeds the Ramanujan bound \(2\sqrt{2}\) and splits two real Bass–Ihara roots strictly off the critical circle \(\lvert u \rvert^2 = 1/2\), contrasted across a 5-graph 3-regular tower (\(K_{3,3}\), Petersen, Heawood, Möbius–Kantor \(GP(8,3)\), and Barbell \(G_{14}\)).Open Millennium Problem (Classical RH for \(\zeta(s)\)): This paper does not prove or disprove the classical Riemann Hypothesis for \(\zeta(s)\) [Bombieri, 2000]. Instead,
Signals.Pending.RiemannHypothesisPendingBoundaryandSignals.Pending.finite_zero_audit_does_not_close_rh_bridgesformally isolate the four open infinite-dimensional analytical bridges (de Bruijn–Newman \(\Lambda_{\mathrm{dBN}} \le 0\) [Rodgers and Tao, 2020], adelic Sato trace-norm convergence, global Li/Jensen–Pólya/Weil positivity [Weil, 1952, Pólya, 1927, Li, 1997], and Berry–Keating/Connes self-adjoint operator closure) and prove that no finite zero scan closes them.
Source Provenance & Verification Artifacts:
AI Chat Source (design provenance only; audited and corrected below):
data/chats/Gemini_Quantum-Causal-Grassmannian-Monograph-Review_20261118-1629.md(lines2348–3600,4433–4946,4979–5728)Companion Papers & Engineering Plan:
Written Plan: Error Minimization, Mpemba–Trajectory Synthesis, and Phased Workstreams (
RB-1–RB-6)Verified Finite Contracts for Grassmannian EBR, All 80 Crystallographic Layer Groups, PSD Slavnov–Gram DPPs, Multi-Reference NOCI, Tequila-Compiled Grid DSLs, and Finite-Graph Ihara–Selberg Zeta Functions and Grassmannian EBR Chat-Code Porting Ledger
SignalsLean 4 Library (leanprover/lean4:v4.35.0-rc3, zerosorryaxioms):src/signals/Signals/RiemannSpectral.lean(Sections 1–10)src/signals/Signals/GrassmannianEBR.lean&src/signals/Signals/QuantumCausalInference.leansrc/signals/Signals/Pending.lean(RiemannHypothesisPendingBoundary,finite_zero_audit_does_not_close_rh_bridges)src/signals/SignalsTests.lean&src/signals/SignalsPendingTests.lean
signals-numerical-adapterRust + PyO3 (maturin) Extension (126passing unit tests):Python + SymPy + 35–80 digit
mpmath+ PyO3 Verification Suite:
Abstract¶
We audit all five Riemann Hypothesis (RH) approaches explored in the Grassmannian monograph dialogue (data/chats/Gemini_Quantum-Causal-Grassmannian-Monograph-Review_20261118-1629.md, lines 2348–5728) and consolidate every defensible finite algebraic, spectral, and numerical contract across Lean 4, Rust/PyO3 (maturin), and Python/SymPy + 35–80 digit mpmath.
Complete Finite-Graph Bass–Ihara Ramanujan Dichotomy Across a 5-Graph Tower (Solved in Lean 4, SymPy, and Rust): Pairing
Signals.GrassmannianEBR.ihara_ramanujan_critical_circle_modulus_sq(\(\lambda^2 \le 4q \implies \lvert u \rvert^2 = 1/q\)) withSignals.RiemannSpectral.ihara_non_ramanujan_splits_off_critical_circle(\(\lambda^2 > 4q \implies\) two real roots \(u_+ \ne u_-\) with \(u_+ u_- = 1/q\) and \(u_\pm^2 \ne 1/q\)) [Bass, 1992], we construct an explicit 14-vertex 3-regular connected non-Ramanujan Barbell graph \(G_{14}\) (\(q = 2\)) whose \(3 \times 3\) antisymmetric equitable quotient matrix has exact characteristic polynomial \(p(\lambda) = \lambda^3 - \lambda^2 - 6\lambda + 2\). Using Lean 4’s Intermediate Value Theorem (nonRamanujan14_exists_super_ramanujan_off_circle_roots), we prove without floating-point assumptions that \(G_{14}\) possesses an eigenvalue \(\lambda_2 \in (57/20, 143/50)\) (\(\lambda_2 \approx 2.8557725066 > 2\sqrt{2} \approx 2.8284271247\)) whose Bass–Ihara roots \(u_+ \approx 0.8125068\) and \(u_- \approx 0.6153795\) split strictly off the critical circle \(\lvert u \rvert = 1/\sqrt{2}\), while simultaneously collapsing the Markov relaxation gap from \(\gamma_{\mathrm{gap}} \in [0.3333, 1.0000]\) across the Ramanujan tower (\(K_{3,3}\), Petersen, Heawood, Möbius–Kantor \(GP(8,3)\)) down to \(\gamma_{\mathrm{gap}} = 0.0481\).High-Precision Rust/PyO3 Evaluators (Euler–Maclaurin, Riemann–Siegel \(C_0\text{..}C_4\) with Exact
Jet2/CJet2Analytic Derivatives, and Borwein Diagnosis): We implement self-containedf64Euler–Maclaurin summation, singularity-free Chebyshev Riemann–Siegel remainder terms \(C_0(p)\) through \(C_4(p)\) [Gabcke, 1979], 2nd-order forward-mode automatic differentiation (Jet2/CJet2) for exact \(Z'(t), Z''(t)\), and Borwein Algorithm 2 [Borwein, 2000] inriemann_spectral.rs. Cross-checked against 35–45 digitmpmath, Euler–Maclaurin locates all first 30 nontrivial zeros with maximum height error \(2.84 \times 10^{-14}\), extending Riemann–Siegel from \(C_2\) to \(C_4\) reduces Lehmer-pair evaluation error by \(198.5\times\) (1.25e-09\(\to\)6.31e-12, reaching the IEEE-754f64phase-rounding floor), and replacing finite differences withJet2analytic differentiation reduces \(\mathcal{J}(g_{125})\) error by \(2.87 \times 10^7\times\) (1.93e-04\(\to\)6.72e-12).Correction of Chat Check Figures for Gram’s Law Violations (\(g_{126}, g_{134}\)) and Lehmer’s 1956 Pair (#6709–#6710): Auditing the chat’s uncomputed check-figure table (
lines 5718–5725) against 35-digitmpmathand Rust Riemann–Siegel (\(C_0\text{..}C_4\)) exposes substantial numerical discrepancies in the chat table—including a sign reversal at Lehmer’s 1956 pair extremum \(\gamma_* \approx 7005.0818551\) [Lehmer, 1956, Csordas et al., 1994], where the true value is a positive peak \(Z(\gamma_*) = +0.0039673575\) with \(Z''(\gamma_*) = -22.346154\) (giving Channel B Bohm curvature rescue \(J_{\mathrm{bohm}}(\gamma_*) = \mathcal{J}(\gamma_*) = +0.0886552 > 0\), \(\eta_{\mathrm{Bohm}} = 1.0000\), directly dual to the osmotic/Bohm velocity split of [Sharoglazova et al., 2025]), and negative \(J_{\mathrm{bohm}}\) at \(g_{126}\) and \(g_{134}\) themselves (where osmotic \((Z')^2\) keeps \(\mathcal{J} > 0\)).Mpemba–Trajectory–dBN Synthesis, Analytical-Bridge Probes, & Strict Epistemic Boundary: Extending
Signals.RiemannSpectral(Sections 7–10) andriemann_rh_bridges.py, we prove exact quartic de Bruijn–Newman backward-heat closure (dbnQuarticHeat_pde_and_shift), 4-zero/10-zero Calogero–Moser neighbor squeezing (dbnFourZero_symmetric_neighbor_squeeze, [Drori et al., 2023, McGuire, 1964]), and Strong Mpemba slow-mode annihilation (mpemba_projection_eliminates_slow_weil_mode, [Holtzman and Raz, 2022, Summer et al., 2026])—which collapses the \(6\times 6\) orthonormal Hermite–Gaussian Weil Gram condition number by \(429.4\times\) (\(\kappa = 1{,}389.36 \to 3.24\)) with \(< 3.35 \times 10^{-50}\) prime-vs-zero discrepancy [Weil, 1952, Bombieri, 2000], alongside strict real-rootedness of Griffin–Ono–Rolen–Zagier Jensen polynomials for \(d = 2\text{..}8\) [Pólya, 1927, Griffin et al., 2019] and exact infinite-zero Keiper–Li positivity \(\lambda_1\text{..}\lambda_{25} > 0\) [Li, 1997, Keiper, 1992, Bombieri and Lagarias, 1999]. All four open analytical bridges for classical RH remain explicitly gated inSignals.Pending.RiemannHypothesisPendingBoundary.Finite Grassmannian Lift of the Quartic Heat Jet (Lean 4 and Rust): The even quartic coefficient state \((a_0,a_2,a_4)\) is augmented by a fixed reference direction to give a real \(\mathrm{Gr}_{\mathbb R}(2,4)\) chart with Plücker coordinates \((0,0,a_0,0,a_2,a_4)\). Lean proves the Klein relation and the additive-time ambient shear; Rust cross-checks the same finite formulas in
f64. This is a finite coefficient-plane encoding, not a canonical Hilbert-space embedding, or a closure of any global RH bridge.
1. Architecture & Epistemic Triage of the Five Chat RH Approaches¶
flowchart TB
subgraph Chat["Chat Provenance (Gemini_Quantum-Causal-Grassmannian-Monograph-Review, L2348-5728)"]
C1["1. Critical-Line Reflection &<br/>Coulson-Fischer Bifurcation"]
C2["2. Finite-Graph Bass-Ihara<br/>Plabic VRC & Ramanujan Bound"]
C3["3. Semi-Dirac Confinement,<br/>Slavnov/GUE & Sato Hirota"]
C4["4. Jensen-Polya QFIM Split,<br/>Gram Violations & Lehmer/dBN"]
C5["5. Li Unit-Circle Positivity<br/>& Off-Line Quartet Splitting"]
end
subgraph Lean["Machine-Checked Finite Lean 4 (Signals.RiemannSpectral + GrassmannianEBR)"]
L1["negative_transverse_stiffness_bifurcates_off_line<br/>ihara_non_ramanujan_splits_off_critical_circle"]
L2["nonRamanujan14_exists_super_ramanujan_off_circle_roots<br/>heawood_graph_ramanujan_critical_circle"]
L3["semi_dirac_transverse_hessian_pos<br/>slavnov_gue_2x2 & vandermondeDet3 & sato_hirota"]
L4["jensenPolyaDiscriminant_mul_product_rule<br/>lehmer_pair_failure & dbnTwoZero/ThreeZero"]
L5["liTransformModulusSq_eq_one_iff_critical_line<br/>liTransformModulusSq_off_line_splits"]
end
subgraph Rust["Rust/PyO3 + SymPy + 35-Digit mpmath (riemann_spectral.rs)"]
R1["Petersen (10v), Heawood (14v),<br/>Barbell G_14 (14v Non-Ramanujan)"]
R2["Euler-Maclaurin (<3e-14) + RS C0..C2<br/>+ Borwein Conditioning Diagnosis"]
R3["First 30 Zeros, Gram g_126/g_134,<br/>Lehmer 1956 (#6709/#6710), Li Sums"]
end
subgraph Pending["Epistemic Gate (Signals.Pending.RiemannHypothesisPendingBoundary)"]
P1["Open Bridge 1: de Bruijn-Newman Lambda_dBN <= 0<br/>Open Bridge 2: Adelic Sato Trace-Norm Limit<br/>Open Bridge 3: Global Li & Jensen-Polya Positivity<br/>Open Bridge 4: Self-Adjoint Spectral Operator Closure"]
end
C1 --> L1
C2 --> L2
C3 --> L3
C4 --> L4
C5 --> L5
L2 --> R1
L4 --> R2
L5 --> R3
L1 -. "Does NOT discharge" .-> P1
R3 -. "finite_zero_audit_does_not_close_rh_bridges" .-> P1
Approach |
Chat Source Lines |
Proved Finite Contract ( |
Rust/PyO3 & SymPy/mpmath Verification |
Open Analytical Bridge ( |
|---|---|---|---|---|
1. Reflection & Bifurcation |
|
|
Symbolic Cauchy–Riemann & quartic test |
Global absence of negative stiffness (\(\forall \gamma \in \mathbb{R}\)) |
2. Bass–Ihara Graph Zeta |
|
|
Petersen ( |
Trace-norm adelic limit \(p_{\max} \to \infty\) on \(\mathrm{Gr}_{\mathrm{Sato}}(\mathbb{A}_{\mathbb{Q}}/\mathbb{Q}^\times)\) |
3. Semi-Dirac, Slavnov & Sato |
|
|
Exact SymPy Wronskian & Vandermonde checks |
Self-adjoint Berry–Keating / Connes operator closure |
4. Jensen–Pólya, Lehmer & dBN |
|
|
30 zeros, Gram \(g_{125}\text{–}g_{135}\), Lehmer #6709–#6710 audited vs. 35-digit |
de Bruijn–Newman upper bound \(\Lambda_{\mathrm{dBN}} \le 0\) and global \(\mathcal{J}(\gamma) \ge 0\) |
5. Li’s Positivity Criterion |
|
|
\(\lambda_n^{(30)} > 0\) (\(n \le 100\)); off-line quartet turns negative at \(n = 3286\) |
Global positivity \(\lambda_n \ge 0\) for all \(n \ge 1\) across all infinitely many zeros |
2. Solved Finite Problem: The Bass–Ihara Ramanujan Dichotomy and the 14-Vertex Non-Ramanujan Barbell Graph \(G_{14}\)¶
2.1 Both Directions Proved in Lean 4¶
For a finite connected \((q + 1)\)-regular graph \(G = (V, E)\) with adjacency matrix \(A\), the reciprocal Bass–Ihara graph zeta function factorizes over the adjacency eigenvalues \(\lambda \in \operatorname{Spec}(A)\) as
The Graph Riemann Hypothesis asks whether every non-trivial zero of \(\zeta_G(u)^{-1}\) (corresponding to \(\lvert \lambda \rvert < q + 1\)) lies on the critical circle \(\lvert u \rvert = q^{-1/2}\) (\(\operatorname{Re}(s) = 1/2\) under \(u = q^{-s}\)). Lean 4 proves both directions cleanly:
Ramanujan Locking (
Signals.GrassmannianEBR.ihara_ramanujan_critical_circle_modulus_sq): Whenever \(q > 0\) and \(\lambda^2 \le 4q\), the roots \(u = \frac{\lambda \pm i\sqrt{4q - \lambda^2}}{2q}\) satisfy \(\lvert u \rvert^2 = \frac{1}{q}\).Non-Ramanujan Off-Circle Splitting Converse (
Signals.RiemannSpectral.ihara_non_ramanujan_splits_off_critical_circle): Whenever \(q > 0\) and \(\lambda^2 > 4q\) (with \(d = \sqrt{\lambda^2 - 4q} > 0\)), the two real roots \(u_\pm = \frac{\lambda \pm d}{2q}\) satisfy \(q u_\pm^2 - \lambda u_\pm + 1 = 0\), \(u_+ u_- = \frac{1}{q}\), \(u_+ \ne u_-\), and strictly leave the critical circle: \(u_+^2 \ne \frac{1}{q}\) and \(u_-^2 \ne \frac{1}{q}\).
2.2 Explicit 14-Vertex 3-Regular Connected Non-Ramanujan Witness (\(G_{14}\)) vs. Petersen and Heawood Graphs¶
Take two disjoint copies of the 7-vertex gadget \(H\) on vertices \(\{0, 1, 2, 3, 4, 5, 6\}\) with edges \((0,1), (0,2), (1,3), (1,4), (2,5), (2,6), (3,4), (3,5), (4,6), (5,6)\) (degree 2 at vertex \(0\), degree 3 at vertices \(1\text{–}6\)) and connect their degree-2 vertices by a single bridge edge \((0, 7)\) to obtain a 14-vertex 3-regular connected simple graph \(G_{14}\) (\(q = 2\), Ramanujan threshold \(2\sqrt{2} \approx 2.8284271247\)).
Because \(G_{14}\) has an exact 3-orbit equitable partition \((\{0\}, \{1, 2\}, \{3, 4, 5, 6\})\) in each half, its antisymmetric bridge sector (\(v_{7+i} = -v_i\)) reduces exactly to the \(3 \times 3\) quotient matrix
SymPy computes the exact factored characteristic polynomials across the 5-graph 3-regular (\(q = 2\)) tower (RB-4):
In Lean 4, nonRamanujan14_equitable_eigenvector verifies the exact eigenvector \((a, b, c) = (2, \lambda + 1, (\lambda^2 + \lambda - 2)/2)\) for any root of \(p(\lambda) = \lambda^3 - \lambda^2 - 6\lambda + 2\), nonRamanujan14_bracket_sign_change evaluates \(p(57/20) = -587/8000 < 0 < 6757/125000 = p(143/50)\) with \((57/20)^2 = 3249/400 > 8\), and nonRamanujan14_exists_super_ramanujan_off_circle_roots invokes intermediate_value_Icc to prove the existence of \(\lambda_2 \in (2.85, 2.86)\) strictly exceeding \(2\sqrt{2}\) with two real Bass–Ihara roots splitting off \(\lvert u \rvert^2 = 1/2\) [Bass, 1992]. Simultaneously, py_audit_regular_graphs_bass_ihara() links the Graph RH directly to Markov/Mpemba relaxation speed via \(\gamma_{\mathrm{gap}} = 1 - \max\lvert\lambda_{\mathrm{nt}}\rvert / 3\) and the Ramanujan margin \(2\sqrt{2} - \max\lvert\lambda_{\mathrm{nt}}\rvert\) [Holtzman and Raz, 2022]:
3-Regular Connected Graph (\(q = 2\)) |
Vertices |
\(\max \lvert \lambda_{\mathrm{nt}} \rvert\) |
Markov Gap \(\gamma_{\mathrm{gap}}\) |
Margin \(2\sqrt{2} - \max\lvert\lambda_{\mathrm{nt}}\rvert\) |
Ramanujan? |
Off-Circle Roots |
Real Off-Circle Roots \((u_+, u_-)\) |
|---|---|---|---|---|---|---|---|
Complete Bipartite \(K_{3,3}\) |
6 |
|
|
|
Yes |
|
None (\(\lvert u \rvert^2 = 0.5\)) |
Petersen |
10 |
|
|
|
Yes |
|
None (all 18 non-trivial roots have \(\lvert u \rvert^2 = 0.5\)) |
Heawood (Fano Levi) |
14 |
|
|
|
Yes |
|
None (all 24 non-trivial roots have \(\lvert u \rvert^2 = 0.5\)) |
Möbius–Kantor \(GP(8,3)\) |
16 |
|
|
|
Yes |
|
None (all 28 non-trivial roots have \(\lvert u \rvert^2 = 0.5\)) |
Barbell \(G_{14}\) |
14 |
|
|
|
No |
|
\(u_+ = 0.8125068\) (\(u_+^2 = 0.6601672\)), \(u_- = 0.6153795\) (\(u_-^2 = 0.3786919\)) |
In spectral \(s\)-coordinates (\(u = 2^{-s}\)), the two non-Ramanujan roots of \(G_{14}\) sit symmetrically off the critical line at \(s_- = -\log_2 u_+ \approx 0.299548\) and \(s_+ = -\log_2 u_- \approx 0.700452\) (\(s_+ + s_- = 1\), \(\sigma_{\mathrm{dev}} = \pm 0.200452\)), providing a concrete finite-graph realization of a Coulson–Fischer off-axis bifurcation alongside a \(10.99\times\) collapse in Markov relaxation rate relative to the 14-vertex Ramanujan Heawood graph.
3. High-Precision Rust/PyO3 Evaluators & RB-1 Error Minimization (Jet2/CJet2 + Riemann–Siegel \(C_0\text{..}C_4\))¶
3.1 Algorithmic Design (src/signals/numerical_adapter/src/riemann_spectral.rs)¶
Three complementary evaluators and a 2nd-order forward-mode automatic differentiation algebra (Jet2 / CJet2) are implemented in pure Rust (std only) and exposed via PyO3 (signals_numerical_adapter):
Euler–Maclaurin Analytic Continuation + Exact Complex 2-Jets (
zeta_euler_maclaurin,hardy_z_jet): Uses adaptive cutoff \(N = \max(18, \lceil 1.2 \lvert t \rvert + 18 \rceil)\) with 9 Bernoulli ratios \(B_{2k}/(2k)!\) (\(k = 1\text{..}9\)). PropagatingCJet2through \(n^{-s}\), the Euler–Maclaurin tail, and \(\vartheta(t)\) delivers exact machine-precision \((Z(t), Z'(t), Z''(t))\) with zero finite-difference step-size cancellation.Riemann–Siegel with Exact Chebyshev \(C_0(p)\) Through \(C_4(p)\) (
hardy_z_riemann_siegel_c4_jet): Evaluates the Stirling-shifted phase \(\vartheta(t)\) (\(< 10^{-15}\) error) plus the \(O(\sqrt{t})\) main sum and five Gabcke remainder terms [Gabcke, 1979]:
Because \(C_0(p)\text{..}C_4(p)\) are entire functions whose apparent poles at \(p = 1/4, 3/4\) cancel analytically, expanding \(C_0, C_2, C_4\) in even Chebyshev polynomials \(T_{2k}(2p - 1)\) and \(C_1, C_3\) in odd Chebyshev polynomials \(T_{2k+1}(2p - 1)\) (derived at 120-digit mpmath precision) and propagating Jet2 through Clenshaw recurrence eliminates both removable-singularity cancellation and finite-difference noise across \([0, 1]\).
3. Borwein Algorithm 2 (zeta_borwein): Included explicitly to diagnose why f64 Borwein (\(n = 22\)) [Borwein, 2000] matches \(\zeta(2) = \pi^2/6\) to \(< 10^{-13}\) and reaches \(2.98 \times 10^{-9}\) at \(\gamma_1 \approx 14.1347\), yet degrades to \(2.83 \times 10^{-1}\) at \(\gamma_{10} \approx 49.7738\) and \(1.07\) at \(\gamma_{30} \approx 101.3179\): its error scales as \(\cosh(\pi t / 2) / (3 + 2\sqrt{2})^n\), requiring \(n \gtrsim 100\) (and \(\sim 80\)-digit arbitrary precision) at \(t = 101.3\).
3.2 RB-1 Before/After Error-Minimization Benchmarks (py_compare_rs_c2_vs_c4)¶
Upgrading from \(C_2(p)\) + 4th-order finite differences (\(h = 5\times 10^{-4}\)) to \(C_4(p)\) + exact Jet2/CJet2 automatic differentiation drives every f64 observable down to the fundamental IEEE-754 phase-argument rounding floor \(\varepsilon_{\mathrm{phase}} \sim t \ln(\lfloor a \rfloor) \cdot \varepsilon_{\mathrm{mach}} \approx 6 \times 10^{-12}\) at \(t \approx 7{,}005\):
Benchmark Observable |
Baseline ( |
Upgraded ( |
Error Reduction Factor |
Limiting |
|---|---|---|---|---|
\(Z(\gamma_{30})\) at \(\gamma_{30} = 101.31785101\) |
|
|
|
Asymptotic \(O(a^{-11/2})\) at small \(a = 4.016\) (Euler–Maclaurin gives |
\(Z(\gamma_*)\) at Lehmer \(\gamma_* = 7005.08185510\) |
|
|
|
IEEE-754 |
\(Z''(\gamma_*)\) at Lehmer \(\gamma_* = 7005.08185510\) |
|
|
|
Exact |
\(\mathcal{J}(g_{125})\) at \(g_{125} = 280.80242938\) |
|
|
|
Zero step-size cancellation via |
3.3 First 30 Nontrivial Zeros Cross-Checked Against 35-Digit mpmath¶
Across all 30 first nontrivial zeros (py_find_first_n_riemann_zeros(30) vs. 35-digit mpmath.zetazero), the Rust Euler–Maclaurin root finder achieves maximum zero-height error \(2.84 \times 10^{-14}\) and maximum \(\lvert \zeta(\tfrac{1}{2} + i\gamma_k) \rvert = 5.94 \times 10^{-14}\):
Zero Index \(k\) |
Rust Height \(\gamma_k\) |
Error vs. 35-Digit |
\(\lvert \zeta_{\mathrm{EM}}(\tfrac{1}{2} + i\gamma_k) \rvert\) |
\(\lvert Z_{\mathrm{RS}, C_2}(\gamma_k) \rvert\) |
\(\lvert Z_{\mathrm{RS}, C_4}(\gamma_k) \rvert\) |
\(\lvert \zeta_{\mathrm{Borwein}, 22} \rvert\) |
\(\mathcal{J}(\gamma_k) = (Z'(\gamma_k))^2\) |
|---|---|---|---|---|---|---|---|
1 |
|
|
|
|
|
|
|
2 |
|
|
|
|
|
|
|
3 |
|
|
|
|
|
|
|
10 |
|
|
|
|
|
|
|
20 |
|
|
|
|
|
|
|
27 |
|
|
|
|
|
|
|
28 |
|
|
|
|
|
|
|
30 |
|
|
|
|
|
|
|
4. Jensen–Pólya Transverse QFIM Stiffness, Gram’s Law Violations, and Lehmer’s 1956 Pair (Correcting the Chat Table)¶
4.1 Exact Algebraic Identities Proved in Lean 4¶
Along the critical strip \(s = (\tfrac{1}{2} + \sigma_{\mathrm{dev}}) + i\gamma\), the second-order Cauchy–Riemann expansion of \(E(\sigma_{\mathrm{dev}}) = \lvert W(\gamma - i\sigma_{\mathrm{dev}}) \rvert^2\) around a real critical-line amplitude \(Z(\gamma)\) satisfies (Signals.GrassmannianEBR.criticalStripPolynomialEnergy_eq_jensenPolya and Signals.RiemannSpectral.jensen_polya_madelung_bohm_split_and_extremum_rescue):
Moreover, Signals.RiemannSpectral proves:
Laguerre–Pólya Multiplicative Closure (
jensenPolyaDiscriminant_mul_product_rule): \(\mathcal{J}(fg) = g^2 \mathcal{J}(f) + f^2 \mathcal{J}(g)\), so any finite product of real-rooted factors preserves \(\mathcal{J} \ge 0\).Lehmer Bifurcation Witness (
lehmer_pair_failure_destabilizes_and_splits): If \(Z'(\gamma_*) = 0\) and \(Z(\gamma_*) Z''(\gamma_*) > 0\), then \(\mathcal{J}(\gamma_*) < 0\) and the quadratic-truncated critical-strip energy vanishes off the critical line at \(\sigma_*^2 = 2 Z(\gamma_*) / Z''(\gamma_*) > 0\).De Bruijn–Newman / Calogero–Moser Flow (
dbnTwoZeroForwardSeparationRate_pos,dbnThreeZero_center_of_mass_conserved): Under \(\dot{z}_k = \sum_{j \ne k} 2/(z_j - z_k)\), two real zeros separate at rate \(4/(z_2 - z_1) > 0\), and the three-zero velocity sum vanishes identically.
4.2 Exact 35-Digit mpmath & Rust Audit vs. Uncomputed Chat Table¶
Auditing the chat’s uncomputed check-figure table (lines 5718–5725) against Rust py_audit_gram_law_and_lehmer() and 35-digit mpmath.siegelz reveals two key factual corrections:
At the Gram-violation points \(g_{126}\) and \(g_{134}\) themselves, \(Z(g_m)\) and \(Z''(g_m)\) are both negative, so \(J_{\mathrm{bohm}}(g_m) = -Z Z''\) is slightly negative (
-0.2103and-0.0923), while the osmotic Fisher slope \(J_{\mathrm{osm}} = (Z')^2\) dominates (+7.2783and+2.6105) to maintain \(\mathcal{J} > 0\). The pure Bohm extremum rescue (\(\eta_{\mathrm{Bohm}} = 1.0000\)) occurs at the displaced positive peak \(\gamma_* = 282.814083516\) between the two zeros inside \([g_{126}, g_{127}]\).At Lehmer’s 1956 near-miss pair (Zeros #6709 at
7005.0628661779and #6710 at7005.1005646697, gap \(\Delta\gamma = 0.0376984918\)), the extremum at \(\gamma_* = 7005.0818550970\) is a positive peak \(Z(\gamma_*) = +0.0039673563\) with \(Z''(\gamma_*) = -22.346182\) (opposite signs to the chat’s uncomputed table), yielding \(J_{\mathrm{bohm}}(\gamma_*) = \mathcal{J}(\gamma_*) = +0.08865527 > 0\) and \(\eta_{\mathrm{Bohm}} = 1.000000\).
Spectral Locus |
Height \(\gamma\) |
\(Z(\gamma)\) |
\(Z'(\gamma)\) |
\(Z''(\gamma)\) |
\(J_{\mathrm{osm}} = (Z')^2\) |
\(J_{\mathrm{bohm}} = -Z Z''\) |
\(\mathcal{J}(\gamma)\) |
\(\eta_{\mathrm{Bohm}}\) |
Gram Holds? |
|---|---|---|---|---|---|---|---|---|---|
\(g_{125}\) |
|
|
|
|
|
|
|
|
Yes |
\(g_{126}\) (1st Violation) |
|
|
|
|
|
|
|
|
No |
\(g_{126}\) Peak \(\gamma_*\) |
|
|
|
|
|
|
|
|
Extremum |
\(g_{127}\) |
|
|
|
|
|
|
|
|
Yes |
\(g_{133}\) |
|
|
|
|
|
|
|
|
Yes |
\(g_{134}\) (2nd Violation) |
|
|
|
|
|
|
|
|
No |
\(g_{135}\) |
|
|
|
|
|
|
|
|
Yes |
Lehmer 1956 (#6709–#6710) \(\gamma_*\) |
|
|
|
|
|
|
|
|
Extremum |
5. Li’s Unit-Circle Criterion and Why Finite Zero Audits Do Not Close Classical RH¶
5.1 Machine-Checked Unit-Circle Equivalence & Off-Line Splitting¶
Under Li’s Möbius map \(w(\rho) = 1 - 1/\rho\) for \(\rho = (\tfrac{1}{2} + \sigma_{\mathrm{dev}}) + i\gamma\), Signals.RiemannSpectral.liTransformModulusSq_eq_one_iff_critical_line proves
while liTransformModulusSq_off_line_splits proves that for any \(\sigma_{\mathrm{dev}} > 0\), \(\lvert 1 - 1/\rho \rvert^2 < 1 < \lvert 1 - 1/(1 - \bar{\rho}) \rvert^2\), placing one member of every off-line functional-equation quartet strictly outside the unit circle. In Rust/PyO3 (py_li_partial_coefficients), the partial sums \(\lambda_n^{(30)} = \sum_{k=1}^{30} 2\operatorname{Re}[1 - (1 - 1/\rho_k)^n]\) remain strictly positive for all \(n \in [1, 100]\), whereas injecting a single hypothetical off-line quartet \((\tfrac{1}{2} \pm 0.20) \pm 14.1347i\) produces \(\lvert 1 - 1/\rho \rvert \approx 1.0010 > 1\) and drives \(\lambda_n\) strictly negative starting at \(n = 3{,}286\) (\(\lambda_{3286} = -0.0744\), dipping to \(-235.87\) by \(n = 5{,}000\)).
5.2 Formal Pending Boundary (Signals.Pending.RiemannHypothesisPendingBoundary)¶
No finite check above can prove the classical Riemann Hypothesis. Signals.Pending.finite_zero_audit_does_not_close_rh_bridges formally encodes this boundary in Lean 4: a finite audit record with arbitrarily high verified height \(T_{\max} > 0\), audited zero count, and audited Gram violations still fails allFourBridgesClosed unless all four infinite-dimensional analytical bridges are established:
de Bruijn–Newman upper bound: \(\Lambda_{\mathrm{dBN}} \le 0\) (paired with Rodgers–Tao 2020’s \(\Lambda_{\mathrm{dBN}} \ge 0\)).
Adelic Sato trace-norm limit: convergence of finite Plabic VRC / Bass–Ihara transfers as \(p_{\max} \to \infty\) on \(\mathrm{Gr}_{\mathrm{Sato}}(\mathbb{A}_{\mathbb{Q}}/\mathbb{Q}^\times)\).
Global Li / Jensen–Pólya positivity: \(\lambda_n \ge 0\) for all \(n \ge 1\) and \(\mathcal{J}(\gamma) \ge 0\) for all \(\gamma \in \mathbb{R}\).
Self-adjoint spectral operator closure: an intrinsically self-adjoint archimedean + non-archimedean domain closure for a Hilbert–Pólya / Connes generator.
6. SymPy & Arbitrary-Precision Analytical-Bridge Probes (riemann_rh_bridges.py & Signals.RiemannSpectral Sections 7–9)¶
To push directly on the four open analytical bridges without overstating finite checks as a global proof, sustainablefactory/riemann_rh_bridges.py pairs seven symbolic SymPy proofs and four arbitrary-precision (50–80 digit mpmath) numerical probes with new machine-checked Lean 4 theorems in Sections 7–9 of src/signals/Signals/RiemannSpectral.lean.
6.1 Griffin–Ono–Rolen–Zagier (PNAS 2019) Positive-Integral Jensen Polynomials (\(d = 2\text{..}6\))¶
Pólya (1927) proved that the Riemann Hypothesis is equivalent to the hyperbolicity (all roots real) of every Jensen polynomial \(J^{d,n}(X) = \sum_{j=0}^d \binom{d}{j}\gamma_{n+j} X^j\) (\(d \ge 1, n \ge 0\)) built from the Taylor moments \(\gamma_n\) of \(\Xi(z) = \xi(\tfrac{1}{2} + iz)\). Griffin, Ono, Rolen, and Zagier (Griffin et al. 2019) proved hyperbolicity for every fixed degree \(d \ge 1\) and sufficiently large shift \(n \ge N(d)\) by showing Hermite convergence to \(He_d(X)\).
Lean 4 (
Signals.RiemannSpectralSection 7):jensenCubicDiscriminant_factored_rootsproves \(\Delta_3(a(X-r_1)(X-r_2)(X-r_3)) = a^4(r_1-r_2)^2(r_1-r_3)^2(r_2-r_3)^2 > 0\), anddepressedCubic_hermite3_discriminantproves \(\Delta_3(Z^3 - 3Z + c) = 27(4 - c^2) > 0 \iff c^2 < 4\) (\(\Delta_3 = 108\) at \(He_3(Z) = Z^3 - 3Z\)).Zero-Cancellation Positive Integral (
compute_gorz_jensen_moments): Using Riemann’s strictly positive kernel \(\Phi(u) > 0\) (\(\Xi(z) = 4 \int_0^\infty \Phi(u)\cos(zu)\,du\)) eliminates the catastrophic cancellation of circle Cauchy integration at high shifts (\(n > 20\)), matching \(\gamma_0 = \xi(\tfrac{1}{2})\) to \(< 10^{-60}\). Across all degrees \(d \in \{2, 3, 4, 5, 6\}\) and shifts \(n \in [0, 25]\), every computed root of \(J^{d,n}(X)\) is strictly real (\(\max \lvert \operatorname{Im}(r) \rvert = 0\)) with Hermite \(He_3\) depressed residual \(c_n \in [+0.0608, +0.1435] \ll 2\).
Jensen Degree \(d\) |
Shifts Audited \(n\) |
All Roots Real? |
\(\max \lvert \operatorname{Im}(\text{root}) \rvert\) |
Minimum Root Gap |
|---|---|---|---|---|
2 |
|
Yes |
|
|
3 |
|
Yes |
|
|
4 |
|
Yes |
|
|
5 |
|
Yes |
|
|
6 |
|
Yes |
|
|
6.2 Exact Infinite-Zero Keiper–Li Coefficients via \(\ln(2\xi(1+z))\)¶
Whereas Section 5.1 summed over the first 30 zeros, expanding \(\ln(2\xi(1+z)) = \sum_{k=1}^\infty b_k z^k\) on \(\lvert z \rvert = 2\) (strictly inside the zero-free disk \(\lvert z \rvert < 14.14\)) captures all infinitely many nontrivial zeros simultaneously via \(\lambda_n = n \sum_{k=1}^n \binom{n-1}{k-1} b_k\) (Signals.RiemannSpectral.keiperLi_binomial_inversion_123). At 65-digit precision (compute_exact_keiper_li_coefficients), \(\lambda_1\) matches the exact Bombieri–Lagarias identity \(1 + \gamma_E/2 - \ln(4\pi)/2 = 0.023095708966121034\dots\) to \(1.08 \times 10^{-66}\), and all \(\lambda_1\text{..}\lambda_{20}\) are strictly positive (\(\lambda_2 = 0.0923457\), \(\lambda_5 = 0.5755427\), \(\lambda_{10} = 2.2793394\), \(\lambda_{20} = 8.7692769\)).
6.3 50-Digit Weil Explicit-Formula Gaussian Gram Matrix (Prime + Archimedean vs. Zero Side)¶
Under Weil’s positivity criterion (Bombieri 2000), RH holds if and only if the symmetric Weil distribution \(W(g \ast \tilde{g})\) is positive semidefinite on all admissible test functions. Lean 4 proves both the finite Sylvester positivity gate (weilQuadForm2_pos_of_sylvester) and the explicit negative-determinant disproof witness \(v = (-w_{01}, w_{00})\) (weilQuadForm2_neg_witness_of_neg_det).
On the 4-dimensional Gaussian basis \(g_i(x) = e^{-\alpha_i x^2}\) with \(\alpha \in \{40, 80, 160, 320\}\), audit_weil_gaussian_gram_matrix evaluates every matrix entry independently via the Prime + Archimedean side (poles at \(s=0,1\), von Mangoldt sum \(-2\sum \Lambda(n) n^{-1/2} F(\ln n)\), and digamma integral) and via the Spectral Zero side:
Maximum Prime-vs-Zero Matrix Discrepancy (50-digit
mpmath):8.35e-52.Exact Symmetric Eigenvalues of \(W_{4\times 4}\):
3.4629e-05,7.0233e-04,9.9412e-03,1.1917e-01(all strictly positive, confirming Weil positivity on this 4D Gaussian subspace).
6.4 De Bruijn–Newman Backward Heat-Flow Collision Time on Lehmer’s 1956 Pair¶
Under the de Bruijn–Newman heat PDE \(\partial_t H_t = -\partial_z^2 H_t\), Signals.RiemannSpectral.dbnQuadraticHeat_collides_and_leaves_real_axis proves that near any real extremum peak (\(z_0 > 0, z_2 < 0\)), the quadratic solution \(H_t(\delta z) = (z_0 - z_2 t) + \tfrac{1}{2} z_2 \delta z^2\) hits a double root at \(t_{\mathrm{crit}} = z_0 / z_2 < 0\) and leaves the real axis strictly for all \(t < t_{\mathrm{crit}}\). On Lehmer’s 1956 near-miss pair (#6709 & #6710, \(\Delta\gamma = 0.0376984977\)), audit_lehmer_dbn_backward_heat finds
confirming to \(0.06\%\) that local backward heat collision matches the 2-zero Calogero–Moser exact merge time and quantitatively illustrating why Rodgers–Tao (2020) proved \(\Lambda_{\mathrm{dBN}} \ge 0\) while the remaining open half of RH is \(\Lambda_{\mathrm{dBN}} \le 0\).
6.5 Finite Grassmannian Lift of the Quartic de Bruijn–Newman Heat Jet¶
The quartic truncation has coefficient state \(a(t)=(a_0(t),a_2(t),a_4(t))\) for \(H_t(\Delta z)=a_0(t)+a_2(t)(\Delta z)^2+a_4(t)(\Delta z)^4\). Its finite heat step is
This makes the coefficient orbit a three-dimensional linear representation of the additive heat-time parameter. To obtain a Grassmannian point, augment the coefficient row by a fixed reference direction \(e_4\):
Interpret \((a_0,a_2,a_4)\) as coordinates in the finite Euclidean coefficient Hilbert space \(\mathbb R^3\) with the selected basis \(\{1,(\Delta z)^2,(\Delta z)^4\}\). A nonzero vector determines a ray in \(\mathrm{Gr}_{\mathbb R}(1,3)\cong\mathbb{RP}^2\); adjoining the fixed reference direction \(e_4\) embeds that ray as the 2-plane
The Lean and Rust implementations store the finite coefficient vector and this explicit frame; they do not supply a continuum \(L^2\) basis or a canonical map from an unspecified Hilbert space. For \(a\ne0\), the row span is a real 2-plane in \(\mathrm{Gr}_{\mathbb R}(2,4)\), hence in its complexification \(\mathrm{Gr}_{\mathbb C}(2,4)\). The Plücker relation is automatic for this decomposable bivector. The ambient shear
is lower triangular with unit diagonal, fixes the reference row, has reverse-time inverse \(U_{-h}\), and composes as \(U_{h_1}U_{h_2}=U_{h_1+h_2}\). More explicitly, \(U_h=I+hN+\frac{h^2}{2}N^2\) for a fixed nilpotent generator \(N\) with \(N^3=0\); \(N\) is not symmetric in the coefficient-space Euclidean inner product. Thus this finite flow generator is not itself a self-adjoint Hilbert–Pólya operator. The Lean implementation (DBNQuarticHeatGr24State, dbnQuarticHeatGr24Generator, dbnQuarticHeatGr24AmbientAction, and dbnQuarticHeatGr24GeneratorMatrix_heatStep) proves these finite polynomial identities and the nonzero-minor condition; the Rust implementation (DbnQuarticHeatGr24State) checks the same formulas numerically in f64. The zero coefficient vector maps to the zero Plücker vector and is not a point of the projective Grassmannian.
This construction is a chosen coefficient chart, not a canonical embedding of an infinite-dimensional Hilbert space and not a spectral model for the zeta zeros. A generic finite Hilbert-to-Grassmannian construction starts with an isometric frame \(E_N:\mathbb{C}^N\to\mathcal{H}\) whose range lies in the domain of a specified operator \(A\). Compress to \(A_N=E_N^*AE_N\), choose a full-rank \(N\times k\) frame \(X_0\), and evolve
Here \(p_t\) is the vector of \(k\times k\) minors, so it represents the transported \(k\)-plane. The quartic implementation is this recipe in a selected finite coefficient chart, with a row-frame convention. A dBN-to-Hilbert bridge would still need a justified Hilbert space, the unbounded generator \(A=-\partial_z^2\) with domain and boundary conditions, a convergent sequence of finite sections, and a proof that the limiting spectral data are the zeta-zero ordinates. None of those infinite-dimensional ingredients is supplied by the quartic jet.
Native Rust f64 Cross-Check Plan¶
Separately, Signals.RiemannSpectral defines the affine normalization
riemannSpectralNormalizedDimension(s) = 2s and the parameter and dimension
shadow reflections. It proves that both reflections are involutions, that
\(2-\Delta(s)=\Delta(1-s)\), and that
\(\operatorname{Re}(\Delta(s))=1\) iff \(\operatorname{Re}(s)=1/2\). These are
exact algebraic normalization identities only: they do not establish a
functional equation for \(\zeta\), identify any zeros, or imply self-adjointness
or the Riemann Hypothesis.
audit_dbn_quartic_heat_gr24 accepts \((z_0,z_2,z_4,t,h_1,h_2,\Delta z)\) and returns finite_inputs, chart_has_nonzero_minor, heat_step_max_abs, polynomial_reconstruction_abs, generator_intertwining_max_abs, plucker_relation_max_abs, ambient_semigroup_max_abs, and ambient_inverse_max_abs; max_abs_residual() aggregates the six numeric residuals. The native test checks a deterministic nonzero fixture with finite_inputs = true, chart_has_nonzero_minor = true, and max_abs_residual() < 10^{-10}. A second all-zero fixture must have chart_has_nonzero_minor = false and residual below \(10^{-12}\). The test also checks the generator update against right multiplication by \(U_h\).
This is a numerical consistency audit in IEEE-754 f64, not interval certification: the thresholds validate the implementation on the selected fixture, not a uniform floating-point error bound over all inputs. A proof-grade numerical certificate would need outward-rounded intervals and explicit input ranges. Neither that certificate nor the analytic Taylor remainder is part of the current Lean/Rust contract.
Model |
Grassmannian state |
Lean 4 implementation |
Rust implementation |
Status relative to Bridge 4 |
|---|---|---|---|---|
Finite quartic dBN chart |
\(\mathrm{Gr}_{\mathbb R}(2,4)\subset\mathrm{Gr}_{\mathbb C}(2,4)\); coefficient ray plus fixed \(e_4\) |
|
|
Implemented finite polynomial model; no unbounded domain, Hilbert-space closure, or global-\(\Lambda\) conclusion |
Generic Hilbert finite section |
\(\mathrm{Gr}_{\mathbb C}(k,N)\) from a rank-\(k\) subspace of a finite projection |
Not implemented: the operator/domain and convergence theorem are unspecified |
Not implemented for dBN |
Requires a defined candidate operator and a proved \(N\to\infty\) spectral-convergence result |
G-MGS superconductivity, the Turner dual-Grassmannian/iGPE product, and DDF second sound are omitted from this comparison because no state map from those models to \(H_t(z)\) has been specified. In particular, an ideal superconductor does not realize the auxiliary de Bruijn–Newman heat flow.
7. Reproducible Tri-Language Verification Commands¶
# 1. Lean 4 verification (Signals.RiemannSpectral + Signals.Pending + Tests)
make signals_build
# 2. Rust native tests (126 passing) + PyO3/maturin editable release wheel
cd src/signals/numerical_adapter
cargo test
TMPDIR=/workspaces/sustainablefactory/.tmp maturin develop --release
# 3. Python + SymPy + 50-80 digit mpmath + PyO3 integration suite
pytest -q tests/test_riemann_spectral_sympy_pyo3.py
Scholarly References¶
Bass, H. (1992). “The Ihara–Selberg zeta function of a tree lattice.” International Journal of Mathematics, 3(6), 717–797.
Bombieri, E. (2000). “The Riemann Hypothesis.” Official Problem Description, Clay Mathematics Institute.
Borwein, P. (2000). “An efficient algorithm for the Riemann zeta function.” Constructive Experimental and Nonlinear Analysis, CMS Conf. Proc. 27, 29–34.
Gabcke, W. (1979). Neue Herleitung und explizite Restabschätzung der Riemann–Siegel-Formel. PhD thesis, Universität Göttingen.
Griffin, M., Ono, K., Rolen, L., & Zagier, D. (2019). “Jensen polynomials for the Riemann zeta function and other sequences.” Proceedings of the National Academy of Sciences, 116(23), 11103–11110.
Lehmer, D. H. (1956). “Extended computation of the Riemann zeta-function.” Mathematika, 3(2), 102–108.
Li, X.-J. (1997). “The positivity of a sequence of numbers and the Riemann hypothesis.” Journal of Number Theory, 65(2), 325–333.
Rodgers, B., & Tao, T. (2020). “The de Bruijn–Newman constant is non-negative.” Forum of Mathematics, Pi, 8, e6.
8. Provenance Metadata (BibTeX & JSON-LD)¶
@misc{chat_rh_spectral_audit_2026,
author = {Turner, Wesley and Google Gemini},
title = {Quantum-Causal Grassmannian Monograph Review: Riemann Hypothesis Spectral Approaches, Bass-Ihara Graph Zeta, Jensen-Polya QFIM, and Lehmer Pair Audits},
year = {2026},
howpublished = {AI chat log (design provenance only; audited and corrected against 35-digit mpmath)},
note = {Source: data/chats/Gemini_Quantum-Causal-Grassmannian-Monograph-Review_20261118-1629.md, lines 2348-5728}
}
{
"@context": "https://schema.org",
"@type": "ScholarlyArticle",
"headline": "Finite Spectral Contracts for Riemann Hypothesis Approaches: The Bass-Ihara Ramanujan Dichotomy (14-Vertex Non-Ramanujan Witness), Jensen-Polya Transverse QFIM Stiffness, Li's Unit-Circle Criterion, and High-Precision Euler-Maclaurin / Riemann-Siegel C_2(p) Audits in Lean 4, Rust/PyO3, and SymPy",
"isBasedOn": [
"data/chats/Gemini_Quantum-Causal-Grassmannian-Monograph-Review_20261118-1629.md",
"src/signals/Signals/RiemannSpectral.lean",
"src/signals/numerical_adapter/src/riemann_spectral.rs"
],
"dateModified": "2026-11-19"
}