Topology optimization of nonreciprocal odd-elastic structures: reproduction package

Complete reproduction package for the manuscript Topology optimization of nonreciprocal odd-elastic structures: formulation, transpose-adjoint sensitivities and spectral stability (submitted to Computers & Structures, Elsevier). Contents (see README.md inside the archive): code/ — self-contained Python sources: oddto_core.py (odd-elastic finite-element core with transpose-adjoint gradients and the cycle-work functional), oddto_opt.py (NLopt MMA/CCSAQ interface), run_examples.py (driver reproducing Examples 0–6 and all figures), verify.py (seven-test verification suite), bench_nlp.py, run_energy.py, run_mesh.py; paper/ — manuscript source (Markdown with LaTeX mathematics) and the Word version; results/ — reference outputs (results.json, energy_pareto.json, mesh_study.json) and the full-scale reference run log for bit-level comparison; submission/ — journal compliance report, highlights file, and 1000-dpi figure files. Three reproduction levels: L1 — mathematics (2 min): python verify.py and python bench_nlp.py reproduce the paper's Table 1 (gradient errors ≤ 2×10-7); L2 — small-scale end-to-end (25 min): python run_examples.py --quick; L3 — full scale (100 min): python run_examples.py, reproducing every reported number and all figures; the headline protocol (MMA, full β-continuation, 150-iteration polish, best-feasible over five starts) was verified to reproduce bit-for-bit on a second machine. Environment: Python 3.13, numpy 2.5, scipy 1.18, nlopt 2.11, matplotlib 3.11 (see requirements.txt). Reference values: passive reference cA*=cB*=88.3770, Jref=73.5899; odd headline J*=397.82, g*=-73.55, contrast 5.41; contrast–energy Pareto frontier and mesh-refinement study included.