Research programs

Supporting validation dossier · P1A/P1B companion

Reproducible Cosmological Proxy and Pipeline Checks: Stock-CAMB ΛCDM+ΔN_eff MCMC, Synthetic NaMaster Recovery, and a Generic Spectator-ALP Birefringence Fit

The MCMC validation dossier behind the bounce-theory papers — chains, configs, and controls

Three self-contained reproducibility checks, none of which tests or measures torsion: a stock-CAMB/Cobaya ΛCDM+ΔN_eff MCMC proxy (309,189 frozen samples, ΔN_eff consistent with zero), a 500-realization synthetic NaMaster E→B pseudo-Cℓ pipeline-recovery validation, and a generic spectator-ALP birefringence consistency fit with a quantified prior-predictive accommodation cost. Frozen chains, configs, and scripts are publicly archived.

In preparation · structural no-go survey (14-barrier catalog)

A Structural No-Go Survey of Minimal Spin-Torsion Routes to Dark Energy and Bounce Phenomenology

Which minimal spin-torsion routes to dark energy or a bounce are structurally closed — a 14-barrier survey (draft)

A systematic survey of 7 foundation mechanism classes and 6 observational branches, extracted from the retired P1U barrier catalog, giving 14 distinct mechanism-class constraints across 14 catalog entries and closing four candidate dark-energy routes (R1-R4). Draft v1C.0.15, compiled 25 pp, 0 undefined refs, 0 overfull boxes — closes the R12 correctness-convergence board on the exact v1C.0.14 PDF (sha256 9dd5c708..., 24 pp). Four legs: Claude Opus INT MAJOR REVISIONS (2 MAJOR / 9 MINOR, five candidate findings withdrawn after 300-DPI re-render or artifact cross-check), Grok grok-4.3 REJECT (3 ESSENTIAL / 3 MAJOR / 2 NIT, all scope, self-containment, and length, none computational), and Gemini gemini-3.1-pro-preview ACCEPT WITH MINOR CORRECTIONS (1 MINOR + 2 NIT) — Gemini's second ACCEPT-class verdict on P1C after its R4 ACCEPT, and the board's third ACCEPT-class verdict overall (Gemini R4, Claude R5, Gemini R12); the Perplexity leg FAILED and is recorded as failed, never a verdict. The headline of the round: both of Claude's MAJORs were CONFIRMED CORRECT by an independent computation that solved the Einstein-Cartan-Holst connection equation directly (research/theory_audit/ech_torsion_onshell_2026_08_08.py and .md) — varying the first-order ECH action with respect to all 24 independent contorsion components with no irrep ansatz, cross-checked against an independent differential-form route, gives on-shell torsion T_abc = alpha*eps_abcd*J5^d + beta*(eta_ab*J5_c - eta_ac*J5_b) with beta/alpha = 1/(2*gamma); both the axial and the trace-vector irreps are nonzero at every finite nonzero gamma, and the tensor irrep is identically zero. Pure axiality is only the gamma -> infinity Einstein-Cartan limit; at the LQG value gamma = 0.2375 the trace-vector coefficient is 2.11x the axial one. The referee was right, and the correction overturns a result one of this repository's own prior released artifacts had asserted as pure-axial torsion; that artifact now carries a dated erratum addendum stating the corrected result while leaving its original text and provenance intact. Consequences landed in v1C.0.15: O4 is NOT identically zero on shell - O4(bare) = -24*alpha*beta*(J5.J5) = -192*pi^2*G^2*gamma^3/(1+gamma^2)^2*(J5.J5), i.e. O4^[4] = -3*kappa*gamma^3/(1+gamma^2)^2*(J5.J5), so the paper's 'strictly stronger disposal' claim is WITHDRAWN; O1 = O6 = -O2 + (1/2)O4 on shell, so O1 and O6 are NOT exact total derivatives on the ECH branch; O1 = O6 and the Nieh-Yan relation 2*O1 + 2*O2 - O4 = 0 both SURVIVE, re-verified at finite gamma on six curved on-shell ECH configurations at gamma in {19/80, 1, 3}, while O1 = -O2 FAILS. The physics conclusion survives intact: O1, O4 and O6 join O5 in the kappa-suppressed Fierz-closed (J5.J5) disposal class, with O4^[4]/O5^[4] = gamma/(1+gamma^2) ~ 0.22 at gamma = 0.2375 — no new light scale, and the 'no (meV)^4 vacuum energy without a new light scale' conclusion is unchanged. A genuine manuscript convention inconsistency was also fixed: Sec. II's T = kappa*S and App. E's Eq. (E2) normalizations differed by a factor of two in torsion amplitude; the survey now uses Eq. (E2)'s normalization (the Freidel-Minic-Takeuchi solution of the connection equation) throughout, stated as such in both Sec. II and App. E, and O5 reduces to -3*kappa*[gamma^2/(1+gamma^2)]*(J5.J5). Appendix C's claim that the trace-vector and tensor torsion irreps 'appear only when the minimal coupling assumption is relaxed' was FALSE for the trace-vector and is corrected: only the tensor irrep is outside minimal coupling, while the trace-vector irrep is generated by the Holst term under minimal coupling and is inside the Fierz lemma's reach via O4. Data and Code Availability now cites two additional artifacts via the \artifact convention (research/theory_audit/ech_torsion_onshell_2026_08_08.py and .md), with the frozen-commit provenance sentence restated accordingly. Also closed this round: Levi-Civita 'symbol' -> 'tensor' with eps_0123 = -1; the form-to-density Nieh-Yan conversion given in one line so Eq. (11)'s coefficients are reproducible from the quoted identity; M_Pl -> reduced M-bar_Pl inside the App. A 1 identity; the abstract gained a branch-vs-channel qualifier, a Tier-III qualifier on the 61-67 orders claim, and a softened independence claim; Table II's R2 birefringence register reconciled with the abstract's headline; App. A Case I's dimension-(+1) referent named; the Shapiro-Teixeira arXiv-version parenthetical explained; App. A density symbols glossed; and the abstract's long spanning-list sentence split. Deferred-genuine, carried unchanged: the frozen-release Zenodo DOI for this survey's own verification scripts (P-round packaging item the paper already discloses as planned). Page count moved 24 -> 25 because the correction required new text, reported rather than smoothed over; Grok's <=15 pp target remains unmet. DRAFT: R-phase NOT converged at R12 - two correctness-grade MAJORs plus several correctness-grade minors were found and closed, and the round overturned a result one of the paper's own released artifacts had asserted, so R13 on the exact v1C.0.15 PDF is the next correctness-convergence check; no readiness score has been computed and no venue/Zenodo kit exists yet.