March 2026 · Novelty Accounting
What's New in This Research
Every contribution is labeled honestly: NOVEL (we did this first), VERIFICATION (we confirmed someone else's result), or FRAMEWORK (we assembled existing pieces into a new structure). Full credit to prior art throughout.
Houston Golden · Independent Researcher
How We Label Each Contribution
Every result is labeled with exactly what we did. "Derived" is ambiguous — so we don't use it. Instead, each contribution gets one of these clear labels:
| Level | Label | What it means |
|---|---|---|
| N3 | NOVEL CONTRIBUTION | We did this first. New theorem, new method, new measurement, or first comprehensive mapping that goes substantially beyond all prior work. |
| N2 | FIRST COMPUTATION | New quantitative result within a known framework. We applied existing tools in a new way, or computed something specific for the first time, or systematically explored a known problem. |
| N1 | INDEPENDENT VERIFICATION | We confirmed someone else's result. Independent re-derivation or numerical check. Confirms prior work but does not extend it. |
| N0 | PRIOR ART | Not ours. Established result from the literature that we use as a foundation. No novelty claimed. |
No result is claimed as a breakthrough (N4). The strongest claims are "Novel contribution" — new theorems or methods. We would rather underrate ourselves than overclaim. This page proves we know what is ours and what is not.
Novel Contributions — We Did This First (3 Results) + 1 Framework Synthesis
These are results where our contribution goes substantially beyond all prior work. Each card details exactly what existed before, what we added, why it matters, and how to verify.
1. Perturbation-Transparency Theorem
NOVEL CONTRIBUTION · Paper 1, §12 · Our proof, assembling known ingredients into a new theorem
Why it matters
This theorem tells you what the bounce CAN'T do. It proves that the bounce mechanism itself is invisible to our telescopes. Any testable prediction must come from the contraction dynamics before the bounce, not from the bounce mechanism. This saved years of searching in the wrong place and redirected the entire research program.
What it is
A formal all-orders proof that the Barbero-Immirzi parameter \(\gamma\), which controls the bounce dynamics in Einstein-Cartan-Holst (ECH) cosmology, is completely invisible in all perturbative observables — scalar, vector, and tensor. The bounce mechanism leaves no fingerprint in the CMB or galaxy surveys. This is a 5-step proof chain: zero spin density for scalar matter → zero torsion → Levi-Civita connection at all perturbative orders → Holst term reduces to topological Nieh-Yan invariant → no perturbative dynamics from \(\gamma\).
What existed before
The general understanding that ECH with scalar matter has zero torsion is implicit in Hehl et al. (1976) and the Einstein-Cartan literature broadly. Freidel, Minic & Takeuchi (2005) established the parity-violating coupling structure of the Holst term. The connection between zero scalar spin density and vanishing torsion is a standard result in EC theory. Calcagni & Mercuri (2009) and Mercuri (2009) discussed the topological nature of the Holst term in various matter-coupling scenarios. de Berredo-Peixoto et al. (2012) noted that in GR the Holst term does not affect dynamics because the dual Riemann contraction vanishes when torsion is absent. Långvik et al. stated that for minimally coupled matter the connection reduces to Levi-Civita, noting that nonminimal scalar couplings could restore Holst dynamics.
What we did (and didn't do)
The individual ingredients above were all known. Our contribution is the assembly: no prior work combined them into an explicit all-orders perturbation theorem for minimally coupled scalar matter in ECH. We formalized this as an explicit theorem with five numbered steps, proved it extends to tensor perturbations (not just scalar), and verified numerically that the topological identity holds to machine precision. The result has structural consequences: it means every surviving testable prediction from the bounce must be mechanism-independent (depending only on the contraction dynamics, not on what causes the bounce).
Computational verification
Numerical test: \(|\varepsilon^{\mu\nu\rho\sigma} R_{\mu\nu\rho\sigma}| < 10^{-15}\) across 1000 random Riemann tensors with the symmetries of FRW perturbations. This confirms the Holst term vanishes identically as a topological invariant.
How to check
Paper: §12, 5-step proof chain. Script: verify_holst_vanishing.py
2. 14-Barrier Systematic Closure Map
NOVEL CONTRIBUTION · Paper 1, §11 · Nobody mapped all 14 before us
Why it matters
This map proves that dark energy cannot come from the bounce — period. Instead of testing one or two mechanisms and hoping, we exhaustively closed every standard route. This is the most comprehensive negative result in ECH cosmology and it tells future researchers exactly where NOT to look, saving significant effort.
What it is
A complete catalog of 14 independent structural barriers that close all standard mechanism classes tested from a nonsingular quantum bounce to late-time dark energy within the ECH framework. Each barrier is named, quantified with equations, and cross-referenced against the specific foundation or branch study that established it. The map covers 7 foundation studies (A–G) and 17 research branches (H–W).
What existed before
Individual barriers are known in various contexts. Mass-coupling issues in Poincaré gauge theory are discussed in Blagojević & Hehl (2013). Scale separation and fine-tuning problems are standard in the cosmological constant literature. 't Hooft (1979) established technical naturalness as a criterion. Weinberg (1989) cataloged the cosmological constant problem. Various groups have noted specific difficulties with torsion-based dark energy, including Shie, Nester & Yo (2008).
What we did (and didn't do)
The individual barriers were mostly known or expected. Our contribution is the systematic exhaustion: no prior work tested and closed all standard mechanism classes for connecting a nonsingular bounce to late-time dark energy within a single theoretical framework. The 14 barriers are organized into a complete closure argument: mass-coupling lock (Barrier 1), Topological-Shift Duality (Barrier 2, itself an original theorem — see #4 below), scalar-tensor universality (Barrier 3), Planck suppression (Barrier 4), attractor-sensitivity dilemma (Barrier 5), parameter immunity (Barrier 6), Liouville conservation (Barrier 7), and seven more. The systematic exhaustion across 7 foundations and 17 branches — with each branch opened only after passing a 4-question filter — is unprecedented in scope for this class of problem.
Computational verification
Each barrier is backed by its own derivation in the corresponding research directory. Fine-tuning numbers (e.g., \(10^{-122}\) for Planck suppression, \(10^{-57}\) for graviton-loop fine-tuning in PGT) are computed analytically with dimensional analysis cross-checks.
How to check
Paper: §11, Table of 14 barriers with equations. Research files: research/foundation_A_pgt/ through research/foundation_G_bounce_vacuum_selection/ (7 foundations) and research/branch_H_bounce_only/ through research/branch_W_*/ (17 branches).
3. \(f_{\rm NL} = -35/8\) Forecast Package
NOVEL CONTRIBUTION · Paper 2 · First comprehensive forecast for testing Cai et al.'s prediction with upcoming surveys
Why it matters
This is what makes the bounce hypothesis testable. Cai et al. derived the prediction fNL = -35/8 in 2009, but nobody built the full machinery to test it with actual upcoming surveys. We did: SPHEREx sensitivity, Bayesian model comparison, template mismatch quantification, and robustness against every known systematic. SPHEREx data (~2028) will either confirm or kill this hypothesis at ~5σ.
What it is
The first comprehensive forecast package for testing the matter-bounce non-Gaussianity prediction with upcoming surveys. This combines a parameter-free theoretical prediction with instrument-specific sensitivity calculations, Bayesian model comparison, and a systematic robustness analysis across multiple potential failure modes. The prediction is \(f_{\rm NL}^{\rm local} = -35/8 = -4.375\), which is 300 times larger than standard inflation and opposite in sign.
What existed before (and what's ours)
Not ours: Cai, Xue, Brandenberger & Zhang (2009) derived \(f_{\rm NL} = -35/8\) for exact matter-dominated contraction. Heinrich, Doré & Krause (2023) published SPHEREx bispectrum forecasts for generic \(f_{\rm NL}^{\rm local}\). Dalal et al. (2008) established scale-dependent bias as a complementary \(f_{\rm NL}\) probe. Li & Brandenberger (2014) re-derived a related result with a factor-of-2 difference in convention.
Ours: To our knowledge, no prior work combines all of the following into a single matter-bounce-specific forecast: (a) SPHEREx and MegaMapper sensitivity forecasts specific to the bounce prediction; (b) Bayesian model comparison with 600,000+ Monte Carlo realizations showing bounce favored at ~8-17:1 over tuned multifield; (c) GR-projection robustness analysis; (d) template-mismatch quantification (\(r \approx 0.85\text{--}0.90\)); (e) \(\varepsilon\) correction bounded [1–8%]; (f) cubic bounce transmission estimate (0.024% correction); (g) Li-Brandenberger convention resolution establishing that even the worst-case value is detectable at \(4.4\sigma\) by MegaMapper.
Computational verification
Eight independent computational programs back this forecast:
research/fisher_robustness_surface/— Fisher matrix sensitivity surfaceresearch/bayesian_discrimination_program/— 600K+ MC Bayesian model comparisonresearch/fnl_epsilon_correction/— \(\varepsilon\) correction bounded [1–8%]algebraic_commutator_proof.py— Exact-arithmetic proof: 2×(Eqs. 34–36) = Eq. 37phase3_fisher_overlap.py— \(\ell\)-space Fisher template overlap (CAMB + Planck noise)f1_injection_recovery.py— 200-realization MC injection recoveryF3_cmb_residuals_eb/scripts/— NaMaster EB analysis + injection + frequency consistencyresearch/cubic_bounce_transmission/— Cubic bounce bispectrum transmission
How to check
Paper 2 (v1.6.0): Full forecast sections with SPHEREx/MegaMapper projections, Bayesian comparison figures, and robustness tables. Key equation: \(f_{\rm NL}^{\rm local} = -\frac{35}{8}\) with physics-derived polynomial (6,2,−18,10,−66,18), verified via algebraic commutator proof and injection recovery.
5. Physics-Derived Full-Commutator Polynomial
NOVEL CONTRIBUTION · Paper 2 · Our algebraic derivation resolving a 15-year ambiguity
Why it matters
This resolves a factor-of-2 ambiguity in the literature. Two groups (Cai et al. 2009 and Li et al. 2017) published different answers, and nobody could tell who was right. We traced the discrepancy to the in-in commutator, derived the full polynomial from the original paper's own equations, and proved both groups are correct at their respective levels. This matters because the exact polynomial determines how well SPHEREx can actually detect the signal.
What it is
Using Cai et al.’s own intermediate vertex contributions (Eqs. 34–36), we derive the full-commutator shape polynomial algebraically: (6, 2, −18, 10, −66, 18). This is proven by exact rational arithmetic (Python Fraction) at multiple configurations. The published Eq. 37 coefficients (3, 1, −9, 5, −66, 9) are the single-time-ordering values. The factor of 2 is the in-in commutator: \(i\langle[\zeta^3, L]\rangle = -2\,\mathrm{Im}\,\langle\zeta^3 L\rangle\).
What existed before (and what's ours)
Not ours: Cai et al. (2009) published Eq. 37 with the single-ordering coefficients. Li et al. (2017) obtained a factor-of-2 different result.
Ours: No prior work traced the factor-of-2 to the commutator, identified the printed coefficients as single-ordering values, or derived the full-commutator polynomial from the paper’s own intermediate equations. A literature search (2009–2024) found no independent numerical verification of \(-35/8\).
What we found
- Algebraic proof: \(2 \times (\text{Eqs. 34+35+36}) = \text{Eq. 37}\) at equilateral (exact) and folded (exact)
- Physics-derived polynomial: (6,2,−18,10,−66,18) — the full-commutator result
- Published (3,1,−9,5,−66,9) identified as single-ordering coefficients
- Template overlap with true polynomial: \(r \approx 0.85\text{--}0.90\) — first explicit measurement of template mismatch
How to check
algebraic_commutator_proof.py — Run with Python 3. Uses exact Fraction arithmetic. Reproduces all benchmarks.
6. Independent EB Birefringence Analysis
FIRST COMPUTATION · Paper 2 · We applied a standard method to test our specific prediction
Why it matters
We put our money where our mouth is. Rather than just making a birefringence prediction and waiting, we ran our own independent analysis on Planck data. The result is consistent with our prediction — but we are transparent about the limitations (resolution dependence, miscalibration uncertainty). This is supporting evidence, not primary proof.
What it is
An independent measurement of cosmic birefringence \(\beta\) from Planck SMICA maps using NaMaster with B-mode purification. Includes a resolution ladder (NSIDE 256–2048), frequency consistency test (100/143/217 GHz), injection recovery, and miscalibration marginalization.
Key results
- Lead result: \(\beta = 0.19 \pm 0.03°\) at NSIDE=1024
- Frequency consistency: All 3 HFI channels positive, spread < 3σ
- Injection recovery: All levels pass (0–0.5°), including \(\beta = 0.27°\) with zero bias
- Resolution stress test: NSIDE=2048 gives \(\beta = 0.07 \pm 0.02°\) (high-\(\ell\) instability)
- Miscalibration marginalization: \(\beta = 0.16 \pm 0.09°\); prediction (0.27°) at 1.2σ
What existed before (and what's ours)
Not ours: NaMaster and the birefringence estimator methodology. The Minami & Komatsu (2020) detection. The ALP birefringence model class (Fujita et al. 2021).
Ours: The resolution ladder and frequency consistency test applied specifically to the matter-bounce ALP prediction. The NSIDE=2048 stress test revealing high-\(\ell\) instability. Injection recovery validating the NaMaster estimator for our science range.
4. Topological-Shift Duality
FRAMEWORK · Paper 1, §11.2 (Barrier 2) · We formalized a known tension as a precise theorem
Why it matters
This closes a loophole that keeps coming up. People keep trying to use the Nieh-Yan term to build light pseudoscalars for dark energy. This theorem proves it can't work: you can have mass protection OR geometric dynamics, but never both. It applies beyond our framework — it constrains any attempt to use topological gravity terms as sources for light pseudoscalars.
What it is
A theorem proving that mass protection and geometric content are mutually exclusive for pseudoscalar fields coupled to the Nieh-Yan 4-form. If the Nieh-Yan term is topological (as in standard EC theory), the pseudoscalar mass is shift-symmetry protected but the coupling is a total derivative with no dynamics. If the Nieh-Yan term is non-topological (as in metric-affine gravity), the coupling generates dynamics but the shift symmetry is broken and no mass protection exists. You cannot have both simultaneously.
What existed before (and what's ours)
Not ours: The Nieh-Yan term and its topological properties (Nieh & Yan, 1982). Metric-affine gravity with non-topological Nieh-Yan (Obukhov et al.). The general tension between shift symmetry and non-trivial dynamics is recognized informally in the axion literature. Bombacigno et al. (2021) showed the Nieh-Yan term loses topological character with nonmetricity. Karananas et al. (2025) argued the relevant operator cannot generate a potential.
Ours: We formulated the known tension as a precise biconditional duality, synthesizing the insights of Bombacigno et al. and Karananas et al. into a named structural result. The two desirable properties (mass protection via shift symmetry, and non-trivial geometric content) are provably structurally incompatible.
How to check
Paper: §11.2, Barrier 2. Research files: research/foundation_B_lock_breaking/ (Phase 1 and Phase 2 analyses).
First Computations — New Quantitative Results Within Known Frameworks (13 Results)
These results apply existing tools in new ways, compute specific quantities for the first time, or systematically explore known problems with original quantitative estimates. Each adds something beyond prior art, but none rises to a new theorem or method. FIRST COMPUTATION for all entries.
| # | Result | Paper | Prior Art (not ours) | What We Did | Why It Matters | Code |
|---|---|---|---|---|---|---|
| 1 | ALP Birefringence Consistency | §11.5 | Fujita et al. (2021) established the ALP birefringence model class. Minami & Komatsu (2020) reported the 3.6σ detection. Eskilt et al. provided the joint Planck+ACT constraint. | ECH parity structure motivates ALP coupling. Numerical ΛCDM field evolution gives Δφ/fa = 0.65. Predicted β = 0.27° matches observed 0.342 ± 0.094° within 1σ. Caveat: not unique to ECH — standard ALP formula. | Our prediction matches an actual observation. LiteBIRD will test at 9σ in early 2030s. | alp_field_evolution/ |
| 2 | Mass-Coupling Lock | §11.1 | Known PGT mass issues discussed in Blagojević & Hehl (2013). Fine-tuning in gravitational theories widely discussed. | Quantified precisely: geff ~ 10-61. Fine-tuning at 10-122 (graviton loop). Named and cataloged as Barrier 1. | Puts an exact number on why torsion-mediated dark energy requires absurd fine-tuning. | foundation_A_pgt/ |
| 3 | Scalar-Tensor Universality | §11.3 | Known FRW symmetry constraints. Standard result that torsion vanishes for scalar matter in EC theory. | Proved T0 = Q0 = 0 exactly for scalar matter on FRW. Environmental mass evades the lock but reduces to standard scalar-tensor gravity. | Closes a "clever workaround" — even if you dodge one barrier, the physics collapses to known territory. | foundation_C/ |
| 4 | Planck Suppression | §11.4 | Known Planck-scale suppression in EFT. Appelquist-Carazzone decoupling theorem. | Quantified for disformal couplings in ECH: all distinctive geometric effects suppressed by k²/MPl² ~ 10-122 at cosmological scales. | Any "geometric dark energy" from torsion is 122 orders of magnitude too weak to observe. | foundation_D/ |
| 5 | Attractor-Sensitivity Dilemma | §11.5 | Known attractor arguments in dynamical systems. Penrose (1979) on gravitational entropy. | Formalized as a precise dilemma: attractor erases info needed for DE; no attractor requires fine-tuned initial conditions. Either way, bounce-to-DE fails. | A catch-22 that kills all "bounce remembers initial conditions" proposals. | foundation_F/ |
| 6 | Parameter Immunity | §11.6 | Known cyclic cosmology issues. Steinhardt & Turok (2002) on entropy problems in cyclic models. | Quantified: any pre-bounce parameter influence is exponentially suppressed by e-MPl/σ. Closes Foundation G. | The bounce erases all memory of pre-bounce physics. Nothing useful survives the crossing. | foundation_G/ |
| 7 | Liouville Conservation | §11.7 | Liouville's theorem is standard classical mechanics. Entropy production in bouncing cosmologies discussed by various authors. | Applied to bounce state selection: the bounce cannot preferentially select low-entropy states without violating unitarity. | Kills "the bounce selects the right vacuum" proposals — the bounce is thermodynamically blind. | foundation_G/ |
| 8 | UV→IR Specificity Dilemma | §11.8 | EFT hierarchy is standard ('t Hooft 1979). UV/IR mixing discussed in string theory and condensed matter. | Any mechanism transferring Planck-scale bounce info to the DE scale must violate technical naturalness — and no protecting symmetry exists in minimal ECH. | You can't bridge 122 orders of magnitude without fine-tuning. Period. | foundation_E/ |
| 9 | Decoupling Universality | §11.9 | Appelquist-Carazzone theorem (1975) is standard QFT. Application to gravity discussed in EFT literature. | Applied with 10-122 suppression to all torsion-mediated effects at cosmological scales. Universal: no torsion mechanism produces observable low-energy effects without fine-tuning. | A blanket impossibility result covering all torsion-based dark energy proposals. | foundation_D/, branch_L/ |
| 10 | Gravitational Democracy | §11.10 | Known democratic coupling in GR (equivalence principle). Poplawski (2010) on torsion-mediated baryogenesis. | Applied to bounce baryogenesis: democratic gravitational coupling prevents the bounce from selectively producing any particular particle species. | The bounce treats all particles equally — you can't get matter-antimatter asymmetry from it. | branch_N/ |
| 11 | IR Vacuum 4-Route Closure | §11.11 | Vacuum energy calculations (Birrell & Davies 1982). Sequestering proposals (Kaloper & Padilla 2014). | All 4 independent routes to deriving w = -1 from the bounce closed within minimal ECH. Each route fails for a distinct structural reason. | Proves you can't derive the cosmological constant from the bounce, no matter which approach you try. | ir_vacuum_program/ |
| 12 | MCMC Verification Infrastructure | §9 | Cobaya (Torrado & Lewis 2021) and CAMB (Lewis & Challinor 2000) are existing tools. MCMC methodology is standard. | 424K posterior samples. Honest null: ΔNeff ≈ 0, H0 = 67.68 (ΛCDM). Corrected our own earlier artifact (H0 = 69.2 was a SH0ES prior artifact). | We found our own error and reported it. This is what honest science looks like. | reproducibility/cosmology/ |
| 13 | Hybrid-DE Loophole Rejection | §11.12 | w0wa parametrization (Chevallier-Polarski-Linder) is standard. Dark energy model space well-explored. | 7 disguised forms of ECH dark energy explored and rejected. Each reduces to one of the 14 barriers. | Closes "what if you dress it up differently?" loopholes. All roads lead to the same barriers. | branch_O/ through branch_U/ |
Independent Verifications — We Confirmed Someone Else's Result
Several results in this research program are independent verifications of established physics. We claim NO novelty for these foundations. Our novel contributions are the structural results (above) that emerged from systematically stress-testing them. INDEPENDENT VERIFICATION for all entries.
ECH Action & Field Equations VERIFICATION
Original authors: Hehl et al. (1976), Holst (1996), and the EC/LQC community. The Einstein-Cartan-Holst action with the Barbero-Immirzi parameter and the resulting modified Friedmann equation are standard results. We re-derived them as a foundation for our analysis but do not claim this derivation as novel. Our contribution begins with what we did with these equations.
Modified Friedmann Equation from LQC VERIFICATION
Original authors: Ashtekar & Singh (2011) and the LQC community. The bounce-modified Friedmann equation \(H^2 = \frac{8\pi G}{3}\rho\left(1 - \frac{\rho}{\rho_c}\right)\) is their result. We use this as an input, not a result.
Four-Fermion Interaction from Torsion VERIFICATION
Original authors: Hehl et al. (1976) and Kibble (1961). Integrating out torsion in EC theory produces a four-fermion contact interaction. This is textbook material. We verified it as part of Foundation A.
fNL = -35/8 Prediction VERIFICATION
Original authors: Cai, Xue, Brandenberger & Zhang (2009). The core prediction \(f_{\rm NL} = -35/8\) for exact matter-dominated contraction is their result. We verified it algebraically and numerically, but the derivation itself is prior art. Our novel contribution is the forecast package built around it (see #3 above), not the prediction itself.
Prior Art — Not Ours, But We Build On It
These are the foundational results from the literature on which our entire research program rests. We rely on them, build on them, and cite them throughout — but we claim zero novelty for any of them. This list is not exhaustive; the full bibliography contains 57+ entries. PRIOR ART for all entries.
| Authors & Year | Result | How We Use It |
|---|---|---|
| Hehl, von der Heyde, Kerlick & Nester (1976) | Einstein-Cartan theory: general relativity with torsion, four-fermion contact interaction | Foundation of the ECH framework. Torsion field equations, spin-density coupling, algebraic (non-propagating) torsion. |
| Ashtekar & Singh (2011) | Loop quantum cosmology bounce: \(\rho_c \approx 0.41\rho_{\rm Pl}\) | Provides the bounce mechanism and modified Friedmann equation used throughout Papers 1 and 2. |
| Freidel, Minic & Takeuchi (2005) | Barbero-Immirzi parameter as parity-violating coupling in the Holst action | Establishes the parity structure that motivates the ALP coupling and underlies the perturbation-transparency theorem. |
| Poplawski (2010–2019) | Cosmology in EC gravity: black hole universe, torsion-mediated bounce, big bounce cosmology | Motivates the ECH bounce scenario. Specific models serve as comparison cases for our structural analysis. |
| Cai, Xue, Brandenberger & Zhang (2009) | \(f_{\rm NL} = -35/8\) for matter-dominated contraction | The core prediction of Paper 2. We verified and built a forecast package around it but did not derive it. |
| Nieh & Yan (1982) | Nieh-Yan topological invariant in Riemann-Cartan geometry | The topological term whose properties underlie the Topological-Shift Duality theorem (N3-4). |
| Minami & Komatsu (2020) | Cosmic birefringence detection: β = 0.35 ± 0.14° (2.4σ) | Observational target for the ALP birefringence prediction. Updated by Eskilt et al. to 3.6σ. |
| Heinrich, Doré & Krause (2023) | SPHEREx bispectrum forecast: σ(fNL) ~ 0.46 | SPHEREx sensitivity numbers used in our forecast package. We applied their Fisher matrix projections to the bounce-specific prediction. |
| Dalal, Doré, Strauss & Tegmark (2008) | Scale-dependent bias as fNL probe | Provides the complementary large-scale structure probe used in the MegaMapper forecast component of Paper 2. |
| 't Hooft (1979) | Technical naturalness criterion | Standard against which we test every proposed mass hierarchy. Violations of technical naturalness close multiple barrier routes. |
Computation Scripts
Every quantitative claim in the research program is backed by reproducible computation. Below are the key scripts, what they compute, and which paper claims they support.
| # | Script / Directory | What It Computes | Backs This Claim |
|---|---|---|---|
| 1 | compute_alp_birefringence.py |
ALP field evolution in ΛCDM background. Computes Δφ/fa = 0.65 and predicted birefringence angle β = 0.27°. | Paper 1 §11.5 |
| 2 | compute_fnl_epsilon.py |
O(ε) fNL correction for w = −0.003 departure from exact matter domination. Result: δfNL = 0.6%. | Paper 2 §2.3 |
| 3 | verify_holst_vanishing.py |
SymPy + numerical verification of the perturbation-transparency theorem. Tests |εμνρσRμνρσ| across random Riemann tensors to machine precision |ε R| < 10−15. | Paper 1 §12 |
| 4 | compute_cosine.py |
Template overlap between local and matter-bounce bispectrum. CMB Fisher: r = 0.90; LSS/SDB: r = 0.85. Consolidated: r ≈ 0.85–0.90. | Paper 2 §3.2 |
| 5 | compute_triangle_diagram.py |
One-loop photon-torsion triangle vertex computation. Result: gaγ = 7×10−26 GeV−1. | Paper 1 §14 |
| 6 | compute_parity_odd_tidal.py |
Post-Newtonian parity-odd tidal torque amplitude. Result: A0 ~ 10−128 (125 OOM gap below detectability). | Paper 1 §5 |
| 7 | compute_mukhanov_sasaki.py |
Mukhanov-Sasaki equation through the LQC bounce. Computes spectral index (ns = 1.31) and bounce features in the primordial power spectrum. | Paper 1 §15 |
| 8 | compute_bispectrum_transmission.py |
Cubic bounce bispectrum transmission coefficient. Result: δfNL = 0.024% correction, negligible. | Paper 2 §2.3 |
| 9 | 02_compute_fisher_robustness.py |
Baseline Fisher forecast + kmin scan for SPHEREx and MegaMapper fNL detection, including GR-projection marginalization. | Paper 2 §6 |
| 10 | compute_photoz_degraded_forecast.py |
Photo-z catastrophic outlier degradation analysis. Shows bispectrum fNL forecast is robust to 10% outlier contamination. | Paper 2 §7.3 |
All scripts are available in the GitHub repository. MCMC chains run on RunPod GPU infrastructure (Cobaya v3.6.1 + CAMB v1.6.5). Known gaps are documented in KNOWN_GAPS.md.
AI Discovery Pipelines — Data Products (4 Active)
Beyond the theoretical research program, we operate AI-enhanced discovery pipelines that produce novel observational catalogs. These are independent data products — each labeled below as NOVEL or VERIFICATION. The anomaly catalog is the subject of Paper 3 (draft, targeting ApJS).
DESI DR1 Spectral Anomaly Catalog
NOVEL CONTRIBUTION · Pipeline B · Paper 3 (draft) · ~90x larger than any prior DESI anomaly search
Why it matters
195,829 previously uncharacterized objects in a single catalog. The unexpected finding that galaxies are 20x more anomalous than QSOs was not reported by anyone before us. These objects are waiting for follow-up spectroscopy and could contain new classes of astrophysical transients, rare emission-line objects, or systematic effects that contaminate cosmological surveys.
What it is
195,829 spectral anomalies identified from the full DESI DR1 Main Survey catalog (~18M spectra) using a spectral autoencoder running on an H200 GPU at 896 spectra/sec. This is the first full-DR1-scale autoencoder anomaly search, ~90x larger than prior EDR work (Liang et al. 2023; Nicolaou et al. 2026). Target journal: ApJS.
Key findings
- Galaxies are 20x more anomalous than QSOs (0.75% anomaly rate vs 0.037%) — nobody reported this before
- 6-database cross-match: SIMBAD 0.2%, NED 12.7%, AllWISE 1.5%, Milliquas 0%, Gaia 0.6% — 99.8% uncataloged in SIMBAD
- 200/200 artifact-free visual verification
- 151,244 multi-band anomalies; 44,436 B-dominant anomalies; 96 artifact suspects flagged
What existed before (and what's ours)
Not ours: Autoencoder anomaly detection methods. Liang et al. (2023) applied autoencoder + normalizing flow to ~250K DESI EDR spectra. Nicolaou et al. (2026) applied VAE + Astronomaly to ~208K DESI EDR spectra.
Ours: The ~90x scale jump (DR1 vs EDR), the 20x galaxy anomaly rate discovery, the 6-database cross-match characterization, and the 200/200 artifact-free verification.
Enhanced 18M DESI DR1 Catalog
NOVEL CONTRIBUTION · Pipeline B · In Progress (36%) · First catalog of its kind
Why it matters
Nobody has done this before. A uniform ML characterization of every spectrum in DESI DR1 with 45 computed columns. When complete, any astronomer can query "show me the most unusual galaxies in this sky region" instantly.
What it is
A uniform machine-learning characterization of every spectrum in DESI DR1, with 45 computed columns per object: anomaly scores, spectral feature measurements, cross-match results, quality flags, and derived quantities. 6.5M of 17.9M spectra processed (36%), running on H200.
What existed before
No equivalent product exists. DESI's own pipeline provides redshifts and classifications, but not autoencoder-derived anomaly scores or the additional 45-column feature set.
w0-wa MCMC Quintom Result
INDEPENDENT VERIFICATION · Supporting Contribution · Our independent MCMC run confirms DESI DR2 findings
Why it matters
We independently confirmed DESI's quintom-crossing signal. This matters because if dark energy is evolving (w crosses -1), it would be consistent with certain bounce cosmology scenarios. Our independent run using our own infrastructure gives P(quintom-B) = 98%, matching DESI's result.
What it is
Converged w0-wa MCMC analysis: w0 = −0.871 ± 0.061, wa = −0.542 ± 0.247, with P(quintom-B) = 98% and R−1 = 0.009.
What existed before (and what's ours)
Not ours: DESI DR2 collaboration reported w0-wa constraints with similar quintom-crossing preference (2.8–4.2σ, dataset-dependent).
Ours: Independent verification using our own MCMC infrastructure with the same datasets.
Galaxy Chirality Catalog
NOVEL CONTRIBUTION · Pipeline A · 82% Complete · Will be ~40x larger than any prior catalog
Why it matters
Tests cosmic parity violation at unprecedented scale. If the universe has a preferred handedness (more clockwise than counterclockwise spiral galaxies), it would be a smoking gun for new physics. Previous catalogs were too small to be conclusive. At 8.47M classifications, this will be 40x larger than anything before it.
What it is
8.47M galaxy chirality classifications from an equivariant CNN (93.7% 3-class accuracy, 8/8 bias tests passed, CW = 0.5012). Running on H100 GPU, 82% complete. When finished, this will be the largest galaxy chirality catalog ever produced, enabling population-level tests of cosmic parity violation.
What existed before (and what's ours)
Not ours: Shamir (2020, 2022) published chirality catalogs of ~100K–200K galaxies using Galaxy Zoo morphologies.
Ours: ~40x scale increase, purpose-built equivariant classifier (not repurposed Galaxy Zoo data), and comprehensive 8/8 bias testing suite.
Summary
This page exists because intellectual honesty matters more than impression management. Every claim is labeled: what we did first, what we verified, what we assembled, and what belongs to someone else. If we have erred, it is on the side of understatement.
Houston Golden · houston@hubify.com
Papers · GitHub · Project Dossier · Research Overview
Full bibliography: 57+ entries in references.bib. Known gaps documented in KNOWN_GAPS.md.