M01 volumetric FEM: current corrected evidence edition
=====================================================

Current results are in results-v2/. The original results/ directory is a preserved
failed evidence-writing attempt. The repair changed NumPy-scalar JSON serialization
and report bookkeeping, not the equations, mesh, material, frequency or thresholds.
The old source/progress and its disposition are in failed-attempt-v1/. Its progress
count24 double-counted records after report-write exceptions; it is NOT24distinct
eigensolutions or independent physical experiments.

Current study:12 mesh/material eigensolutions,18 frequency-response curves and12
direct complex reference solves. There are15 author tests, zero new ODE integrations
and zero physical trials. The separate reviewer reconstructed the matrices and
reported227 passing consistency checks without waiving adequacy failures.

All numerical identities passed. All3 first-six-frequency mesh screens and all3
whole-band mesh screens failed. All12 comparisons of24versus48retained modes failed
the whole-band screen. Five of12direct-vs-modal displacement comparisons failed.
The three20kHz mesh compliance and three volume/time RMS stress comparisons passed.
These are unconverged resonance-band outputs, not calibrated material predictions.

The result contains actual volumetric geometry:729nodes and3072tetrahedra at the
finest mesh. Animation fields derive from the computed complex displacement/stress.
Use Re(U*exp(i*theta))=Ureal*cos(theta)-Uimag*sin(theta). Display amplification and
slowed playback are not physical amplitude/time. Each material's selected frequency
is a finite-grid peak under1N peak top traction with an ideal fixed bottom.

Current figures are figures/M01_material_response.png and M01_mesh_convergence.png,
plus SVG and figure-data CSV. The volume float64 data are separate NPZ downloads
listed by publication/VOLUME-INDEX.json; they are excluded from the core ZIP to
avoid duplicating large compressed arrays. The complete exact public allowlist is
publication/PUBLIC-FILES.json. A core-only package verifier reports those omissions
explicitly; use --require-volumes once the three NPZ files are present.

Portable commands:
  python -m unittest -v test_fem3d.py
  python publication/verify_package.py
  python run_study.py --output fresh-results
  python make_figures.py --results fresh-results --output fresh-figures

The full source and declared inputs reproduce the calculations; use a fresh output
directory. Free-vibration bases within repeated eigenvalue clusters can rotate or
change sign between numerical libraries. The physical subspace and forced response
are the meaningful comparison, not a unique named eigenvector animation.

The materials are rounded catalog baselines: structural steel,6061 aluminum and
amorphous fused silica. Crystalline quartz/diamond, hardness, cracks, fracture,
plasticity, real boulder geometry and actual load coupling are not modeled.
No new blanket reuse licence or hardware-capability claim is applied.
