Appendix D: Computational Provenance, Reproducibility, and Auditability

This appendix documents the computational architecture, execution discipline, and provenance guarantees underlying all numerical results reported in this work. Its purpose is to ensure full transparency, independent reproducibility, and auditability at every level of the numerical pipeline.

Unlike minimal reproducibility appendices that describe isolated scripts, this appendix records a complete computational research framework spanning theory development, simulation, validation, falsification, and publication.

D.0 Scope of the Computational Framework

All numerical results in this manuscript are produced within a multi-layered computational research system comprising theoretical modules, numerical solvers, diagnostic pipelines, large-scale validation engines, and automated publication workflows.

The codebase supporting this work contains over 800 tracked files across more than 150 directories and includes:

• Multiple independent physical subprojects,
• A shared numerical and diagnostic core,
• Automated large-scale parameter scans,
• Horizon-scaling and history-dependence tests,
• End-to-end regeneration of figures, tables, and manuscript assets.

This appendix specifies the principles and structure that guarantee reproducibility across this full computational stack.

D.1 Design Principles

The numerical framework was developed under the following non-negotiable constraints:

• No hidden parameters, adaptive tuning, or manual intervention,
• Deterministic execution under fixed random seeds,
• Explicit separation between dynamics, diagnostics, and validation,
• Complete logging of inputs, outputs, and derived metrics,
• Immutable raw data once generated.

No result reported in this manuscript relies on interactive decisions, visual inspection, or post-hoc filtering.

D.2 Repository Architecture

The repository is organized hierarchically to separate concerns across theoretical development, numerical execution, validation, and publication.

At the highest level, the project consists of:

• Standalone theoretical manuscripts,
• A shared numerical simulation and diagnostics engine,
• Independent experimental pipelines,
• Automated validation and audit subsystems,
• Fully scripted publication workflows.

Within the ‘PrePhysicalSelection_EmergentReality‘ module, the structure is further divided into:

• ‘src/‘: core physics, numerics, diagnostics, and stability analysis,
• ‘experiments/‘: isolated experiment definitions,
• ‘scripts/‘: execution, validation, and reproduction tools,
• ‘outputs/‘ and ‘validation_results_big/‘: immutable result artifacts,
• ‘latex/‘: manuscript and appendix sources,
• ‘pdf/‘: compiled publication outputs.

This strict separation prevents contamination between theory, simulation, analysis, and reporting stages.

D.3 Simulation Execution

All simulations are executed exclusively via dedicated experiment drivers, most notably:

This script:

• Solves the inertial emergent gravity equations,
• Evolves full spatiotemporal fields,
• Saves complete field histories to ‘.npz‘ files,
• Records all runtime parameters in file metadata.

No diagnostics, classification, or visualization occur during simulation.

D.4 Analysis Pipeline

Post-processing is handled solely by analysis scripts such as:

These scripts:

• Load raw field data without modification,
• Compute derived observables (centroids, $d(t)$, $L(t)$),
• Evaluate orbit counts and phase indicators,
• Write structured CSV and JSON logs,
• Generate figures using a non-interactive rendering backend.

The non-interactive backend ensures fully automated, batch-safe execution.

D.5 Large-Scale Validation Infrastructure

Systematic parameter scans and robustness tests are orchestrated by:

For each damping value $\gamma$, this validator performs:

• Multiple independent random seeds,
• Multiple temporal horizons ($T$, $2T$, $4T$),
• Full logging of standard output and error streams,
• Automatic aggregation into summary tables and reports.

The validator introduces no new dynamics and alters no model behavior.

D.6 Determinism and Reproducibility Guarantees

Reproducibility is ensured through:

• Explicit random seeds passed via command-line arguments,
• Fixed grid resolution and timestep across all runs,
• Deterministic numerical integration schemes,
• Version-locked software dependencies.

Given identical inputs, all extracted metrics are reproducible up to floating-point roundoff.

D.7 Validation Philosophy

No result is accepted unless it satisfies all of the following:

• Consistency across independent seeds,
• Stability under temporal horizon extension,
• Independence from visualization or plotting choices,
• Agreement between raw diagnostics and aggregated statistics.

This ensures that observed orbital behavior is structural rather than numerical or procedural.

D.8 End-to-End Regeneration

All results can be regenerated from raw simulation data by executing a single, non-interactive reproduction pipeline that rebuilds:

• Figures,
• Tables,
• Diagnostic summaries,
• Manuscript-ready assets.

This establishes the manuscript itself as a reproducible computational artifact.

D.9 Computational Audit Trail

Every numerical claim reported in this work is associated with:

• A raw binary state file (‘.npz‘),
• A corresponding analysis log,
• A deterministic diagnostic output,
• A reproducible figure or table artifact.

This creates a complete audit trail from physical claim to executable evidence.

D.10 Reproducibility Statement

All numerical results reported in this manuscript are fully reproducible by an independent researcher with access to the codebase and a standard Python numerical environment. No manual tuning, selective filtering, or interactive decision-making is involved at any stage.

Philosophical note. Reproducibility is treated here not as a supplementary requirement, but as a structural constraint on the physical claims themselves. The existence criteria proposed in this work are defined only insofar as they survive repeated, history-dependent numerical verification.