Curvature-Bound Bounce Cosmology: Numerical Verification

Reproducibility package for:

Numerical Verification of a Curvature-Bound Bounce Cosmology: Effective-Field-Theory Control Pushes the Transition Below the Planck Scale

Author: Ho Hyung Kim, Independent Researcher, Seoul, Republic of Korea.

Latest archived preprint (v1.3): 10.5281/zenodo.21819281

Scope

This is a numerical audit of selected consequences of a phenomenological bounce ansatz. It is not a validated quantum-gravity theory, a derivation of a CRBC-specific covariant action, or an analysis of observed CMB maps.

The package documents:

• reproduction of a constant-$p$ background correspondence to the published Ye--Piao $c_T=1$ beyond-Horndeski family;
• coefficient-gate checks for a supplied phenomenological $w(t)$ profile;
• exclusion of a minimal single-field k-essence route;
• a blue single-field adiabatic spectrum and the requirements of an added entropic sector;
• re-integration of the scalar and entropy spectra on a derived energy-transfer background;
• explicit negative results for the tested thermal Hagedorn, conformal-coupling, and background-reconstruction routes; and
• a kinetic-dependent derivative-coupling proxy, retained only as a candidate pending a variational derivation and a complete perturbative stability audit;
• simulation-only tests of a perturbative Bianchi-I quadrupole estimator.

What remains required

1. Derive a covariant, degenerate action for the specific $w(t)$ profile by varying the full action, rather than reconstructing a supplied background.
2. Establish DHOST degeneracy and derive the scalar and tensor quadratic actions, including $Q_s$, $c_s^2$, $Q_T$, and $c_T^2$.
3. Derive the EFT cutoff and test strong coupling where the candidate proxy has $1+\nu\rho_2\sim10^3$, as well as non-linear Bianchi-I stability.
4. Supply reheating, amplitude normalization, and any black-hole-interior matching conditions.
5. Run a blinded, likelihood-based analysis of actual CMB maps after completing realistic foreground, mask, and polarization validation.

Version 1.3 — derived-background and covariant-action audit

Version 1.3 makes the manuscript more falsifiable by preserving negative and inconclusive results instead of treating the supplied fluid profile as a field theory. On the derived energy-transfer background, the adiabatic result becomes more strongly blue ($n_s=4.54$, versus $3.34$ for the earlier tanh background). The entropy mechanism remains numerically available, but its central tuning must be recalibrated for that background. The tested local thermal Hagedorn production route, a canonical-scalar conformal coupling, and a beyond-Horndeski background reconstruction do not provide a covariant origin for the transition. The remaining derivative-coupling calculation is a background-level proxy only: its equations have not been obtained by varying a complete covariant action, and its DHOST, perturbative-stability, and strong-coupling tests have deliberately not been claimed.

See the v1.3 English revision record and its Korean counterpart.

Version 1.2 — sub-Planck transition scale

The direct Hagedorn/string-density identification is rejected by the declared EFT-control gate: (\max(E_{\rm char}/\Lambda)=0.519), above the 0.1 criterion. Control requires (\rho_H/\rho_c>725); at (10^4), the independent CPU check gives 0.05188 with no gate violations. Consequently, the dimensionless curvature-bound coefficient is not assumed to be unity: the manuscript derives (\eta=\mathcal K_{\max}\ell_P^4\sim10^{-16})--(10^{-6}). The transition is therefore six to sixteen orders below Planck curvature. See Korean revision record and English revision record.

Contents

• paper/ — English manuscript PDF and LaTeX source.
• code/ — selected Python scripts for the numerical gates and simulations.
• outputs/ — trajectory and JSON reports used by the manuscript.
• docs/ — Korean audit record, EFT coefficient contract, and Planck-analysis preregistration.

GPU null campaign (preregistration §3.4)

The 10^5-null campaign of the Planck-quadrupole preregistration runs on a single CUDA GPU through a torch reimplementation of the two healpy transforms the pipeline uses (code/crbc_gpu_sht.py; ~0.2 s per realization at nside 1024, lmax 1000 on an RTX A2000, versus ~16 s per realization on one CPU node). Static healpy/CAMB inputs are cached once by code/crbc_gpu_prepare_inputs.py (run it under a Python with healpy + camb, e.g. WSL); the campaign itself is

python code/crbc_planck_null_campaign_gpu.py            # 10^5 nulls, checkpointed
python code/crbc_gpu_sht.py                             # SHT self-test vs healpy

Per-seed parity against the healpy pipeline is 1.1e-5 in the estimator channels (code/crbc_gpu_reference_check.py); the audit record is §16 of the Korean preregistration document.

Reproduction

Install the dependencies listed in code/requirements-gpu.txt. CPU execution is supported for small checks; CUDA is required to reproduce the archived GPU performance results. Begin with:

python code/crbc_background_scan.py --device cpu --points 128 --time-steps 257
python code/crbc_kessence_no_go_scan.py --device cpu --points 10000
python code/ye_piao_2019_corrected_reproduction.py --device cpu --points 3001 --extent 20

The published background correspondence is based on Ye and Piao, arXiv:1901.02202. The Planck tilt benchmark is from Planck 2018 X.

Citation and license

This release is licensed under CC BY 4.0. Cite v1.3, 10.5281/zenodo.21819281.