CIIL v7 (Computational Investigation of Inaccessibility Landscapes, version 7) is a computational framework for mapping the accessible feature-space geometry of driven semiclassical optomechanical systems under known and generative drive protocols. This archive accompanies the paper “Accessibility Geometry and Generative Regime Search in Driven Optomechanical Systems. ” The archive contains the complete CIIL v7 simulation and analysis package, including simulation code, feature matrices, observables tables, generated figures, reproducibility scripts, parameter definitions, analysis outputs, and the manuscript PDF. The framework integrates the driven semiclassical optomechanical equations over a 5 × 5 × 4 parameter grid spanning optomechanical coupling g0 ∈ 0. 005, 0. 07, cavity linewidth κ ∈ 0. 01, 0. 50, and detuning Δ ∈ 0. 5, 2. 0ωm. The system combines known drive atlas construction, bounded generative drive search, feature space embedding analysis, directed transition-graph construction, and a six stage irreducibility pipeline for candidate regime validation. Three bounded generative drive families are tested: Fourier superposition drives, burst/pulse drives, and coloured stochastic drives. The analysis extracts a 31 dimensional feature representation spanning mechanical observables, cavity observables, coupling transfer observables, temporal and dynamical observables, modal participation geometry, and semiclassical proxy indicators. Across 1, 126 valid simulations, the framework identifies strongly anisotropic feature-space expansion under burst type generators while coupling-transfer and temporal sectors remain highly rigid. Burst generators expand the cavity observable sector to 7. 74× the known-drive spread (95% bootstrap CI: 2. 17, 14. 98) while coupling-transfer and temporal sectors remain below 0. 03×, revealing approximately 370-fold anisotropy in accessible feature-space expansion. No generated drive passed the strict combined novelty criterion consisting of both geometric separation and density based outlier requirements. All generated candidates that advanced to the irreducibility stage remained continuously connected to the known drive manifold within the tested feature representation and parameter ranges. The result is explicitly bounded: it applies only to the tested generator families, sampled parameter region, and 31-dimensional observable representation, and does not claim formal topological equivalence or exclusion of regimes reachable through untested generators. The framework additionally quantifies embedding dependence, demonstrating that UMAP recovered substantially more true nearest-neighbour relationships than PCA-2 for this dataset, making embedding choice consequential for regime interpretation. Transition-graph analysis identified a densely connected reversible regime structure across the seven known drive families. The simulations were executed under Python 3. 12 using NumPy, SciPy, scikit-learn, NetworkX, Matplotlib, Pillow, and UMAP learn with a fixed global random seed of 42 for reproducibility. The full simulation suite required approximately 1. 6 CPU-hours with a wall time of approximately 20 minutes on five parallel cores. The primary reproduction command is: python ciilᵥ7. py --cores 5 --n-gen 50 --output ciilᵥ7ₚaper Recommended keywords: optomechanics, nonlinear dynamics, computational physics, feature space geometry, regime discovery, generative search, semiclassical dynamics, accessibility geometry, transition graphs, manifold analysis, and CIIL.
Kearon Allen (Thu,) studied this question.