Numerical Relativistic Cosmology: Effective Backreaction Models and BSSN Comparison
Year: 2026
Type: preprint
Abstract: This repository contains the numerical and analysis framework accompanying the study: “Numerical Relativistic Cosmology: Effective Backreaction Models and BSSN Comparison” The project presents a comparative investigation of cosmological backreaction across three complementary approaches: • Full 3+1 numerical relativity using the BSSN formulation implemented in the Einstein Toolkit• Buchert scalar averaging formalism• GPU-accelerated phenomenological effective backreaction simulations inspired by Buchert averaging The work focuses on numerical consistency between effective averaging approaches and weakly perturbed full-GR evolutions rather than proposing a replacement for dark energy. Included in this repository: GPU simulation codes for effective backreaction evolution Numerical convergence datasets (128³, 256³, 512³ resolutions) BSSN ensemble analysis outputs Pantheon+ Type Ia supernova likelihood comparison scripts Data products used to generate figures and tables in the manuscript Key results: Stable BSSN ensemble backreaction:Q_D = (1.536 ± 0.014) × 10⁻⁴ GPU effective model convergence across all tested resolutions Empirical calibration factor:Z ≈ 5.47 × 10³used to map effective amplitudes to BSSN-derived backreaction values Pantheon+ comparison:χ² = 827.2 for the effective model versus χ² = 835.2 for flat ΛCDM The improvement Δχ² = −8.0 is not statistically significant Important note:The GPU effective model is phenomenological and does NOT constitute a direct numerical solution of Einstein’s equations. The calibration factor Z is setup-dependent and should not be interpreted as a universal physical constant. The present study should therefore be interpreted as a reproducible numerical comparison framework between effective cosmological averaging models and weakly perturbed general-relativistic simulations. Hardware and software: Einstein Toolkit (Cactus + ML_BSSN + Carpet) CUDA 13.0 NVIDIA GTX 1080 Ti MPI parallelization Grid resolutions up to 512³ cells Potential applications: Numerical relativity in cosmology Backreaction studies Effective averaging methods GPU cosmological simulations Phenomenological large-scale structure modeling Citation:If you use this repository, please cite the associated paper and Zenodo record. Author:Alexander TikhonovIndependent Researcher2026
Author: Tikhonov, Alexander
Language: English
Subfield: Astronomy and Astrophysics
Field: Physics and Astronomy
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7161472929
Cites:
Cited by:
Related to: