Skip to content

QF Solver 0.2.5a0 V&V matrix

Rows marked OBSERVED_INTERNAL record exploratory evidence from the incremental implementation. The G01 rows below are promoted to PASS_INTERNAL only when the g01_latest manifest records the final clean source SHA. PASS_INTERNAL or PASS_EXTERNAL_CORRELATION_BOUNDED describes the evidence result; it does not by itself close a gate when an acceptance band still requires Owner approval. owner_accepted_experimental_bounded_use records an explicit Owner decision that accepts evidence only within the stated domain. Element families marked SHOULD do not block release unless promoted.

Final Owner scope disposition — 2026-08-28

The Owner-approved scope revision is recorded in 0_2_5_owner_scope_revision.md. The rows below retain their measured evidence labels and historical gate links; the disposition determines whether they are part of the qualified 0.2.5a0 claim.

Evidence group Final treatment
G01, G02, G03, G05, G08 and G09 PASS rows retained as qualified bounded evidence, subject to each gate's stated envelope
G04 rows (VV-016, VV-017, VV-039, VV-052, VV-060, VV-064, VV-072, VV-074–VV-076) retained as internal/research or observed evidence; excluded from qualified release claims; G04 is NOT_IN_RELEASE_SCOPE
G06 rows (VV-022–VV-024, VV-034, VV-035, VV-041–VV-043, VV-059, VV-062–VV-065, VV-083–VV-089) retained with their actual open/research/blocked labels; excluded from qualified release claims; G06 is NOT_IN_RELEASE_SCOPE and qualification is deferred
G07 friction rows optional and excluded from the release

No G04 or G06 observation is promoted by this table. G10 evaluates only the remaining external MUST cells in the final approved scope.

V&V ID Scope Model/path Metrics Reference Status Gate
VV-001 J2 constitutive elastic/yield/uniaxial stress, alpha, tangent analytical PASS_INTERNAL G01
VV-002 J2 constitutive unload/reload/reversal cycles full stress-strain path analytical law PASS_INTERNAL G01
VV-003 J2 constitutive shear/hydrostatic/non-proportional yield, flow, PEEQ invariants/independent calc PASS_INTERNAL G01
VV-004 J2 tangent eight elastic/near-yield/plastic, cyclic and non-proportional states with FD-step sweep max/RMS relative derivative error centered FD PASS_INTERNAL G01
VV-005 state true failed increment state digests, retry result small-step reference PASS_INTERNAL G01/G09
VV-006 J2 elements common multi-element mesh, 4 families F-u, reactions, VM, PEEQ, energy converged mesh/external PASS_INTERNAL G01
VV-007 J2 load-step sensitivity coarse/reference/refined on connected mesh solution, PEEQ, dissipation and cost trends step-sensitivity study PASS_INTERNAL G01
VV-008 objectivity rigid translation/rotation and translation+rotation, four families spurious strain/stress/force/energy, det(F) exact invariant PASS_INTERNAL G02
VV-009 geometric element TET4 TL force/tangent/energy FD numerical derivative PASS_INTERNAL G02
VV-010 geometric element HEX8 force/tangent/energy FD numerical derivative PASS_INTERNAL G02
VV-011 geometric global TET4/HEX8 large-rotation bounded dead-load path load-displacement, reactions, energy, det(F), Newton residuals internal bounded path; post-limit is G04 owner_accepted_experimental_bounded_use G02
VV-012 geometric distortion TET4/HEX8 four-level pre-limit refinement robustness, Jacobian, displacement, reaction, energy, stress/strain trends bounded mesh-trend contract; Owner scope decision owner_accepted_experimental_bounded_use G02
VV-013 high-order geometric TET10/HEX20 common TL J2 residual/state/recovery smoke numerical/external OBSERVED_INTERNAL_RESEARCH G02
VV-014 buckling TET4 total-Lagrangian Euler column, four structured levels 16x4x4 -> 40x10x10 critical factor, mode, Euler error, refinement change, bracket, det(F) analytical Euler PASS_INTERNAL G03
VV-015 buckling constrained 5-TET4 unit-block solid probe critical factor, mode residual, mode normalization/MAC Code_Aster 18.1.0 pinned execution PASS_EXTERNAL_CORRELATION_BOUNDED G03
VV-016 arc-length reduced shallow-arch equilibrium lambda = u - u^3 lambda-u branch, limit point, equilibrium residual analytical reduced-order reference OBSERVED_INTERNAL_RESEARCH G04
VV-017 arc-length restart from intermediate checkpoint endpoint displacement, load-factor continuation state continuous arc-length run OBSERVED_INTERNAL_RESEARCH G04
VV-018 contact local gap/projection/normal tangent residual/tangent FD analytical/FD OBSERVED_INTERNAL_RESEARCH G05
VV-019 contact global block-plane open/close/recontact reaction, gap, active set analytical/external PASS_EXTERNAL_CORRELATION_BOUNDED G05
VV-020 contact sliding bounded faceted/finite-sliding projection path, penetration, reactions internal bounded contract + compatible external contact histories OBSERVED_INTERNAL_RESEARCH G05/G10
VV-021 contact rollback injected contact assembly failure with adaptive retry rejection reason, retry increment, committed-step history common transaction contract OBSERVED_INTERNAL_RESEARCH G05/G09
VV-022 coupled J2 + geometry F-u, VM, PEEQ, energy, tangent approved model/external OBSERVED_INTERNAL_RESEARCH G06
VV-023 coupled geometry + contact load-gap-u, reactions, energy external OBSERVED_INTERNAL_RESEARCH G06/G10
VV-024 coupled SHOULD J2 + geometry + contact complete histories external OBSERVED_INTERNAL_RESEARCH G06/G10
VV-025 friction optional stick/slip/cycle traction, slip, dissipation analytical/external PLANNED COULD G07
VV-026 Newton consistent tangent rate residual histories, reduction ratios, observed-order estimates, iterations expected local behavior OBSERVED_INTERNAL G01/G02/G06
VV-027 failures full taxonomy matrix reason, rollback, converged=false contract PASS_INTERNAL G09
VV-028 regression complete 0.2.4 + 0.2.5 suite tests, skips, coverage, artifacts frozen baseline PLANNED G11
VV-029 release clean docs/build/smoke hashes and imports clean environment PLANNED G12
VV-030 J2 mesh refinement regular block levels 1/2/4, four families displacement, reactions, VM, PEEQ, energy, cost internal trend study PASS_INTERNAL G01
VV-031 J2 cyclic path load/unload/reversal/reload, four families PEEQ, dissipation, residual, iterations constitutive limitations PASS_INTERNAL G01
VV-032 Nonlinear phase timing regular two-cell shared J2 TET4/HEX8 smoke assembly, sparse linear solve, line-search wall time reproducible benchmark telemetry PASS_INTERNAL G08
VV-033 J2 external multi-element regular two-cell shared TET4/TET10/HEX8/HEX20 UX, reaction, SXX, PEEQ histories Code_Aster Docker PASS_EXTERNAL_CORRELATION_BOUNDED G01/G10
VV-034 J2 finite-kinematic candidate TET4/TET10/HEX8/HEX20, kinematics=total_lagrangian_j2 objectivity, element tangent FD, residual and state recovery internal research contract OBSERVED_INTERNAL_RESEARCH G02/G06
VV-035 Common contact contribution TET4 + initial node-to-triangle penalty contact sparse contact residual/tangent, unilateral activation, common Newton internal research contract OBSERVED_INTERNAL_RESEARCH G05/G06
VV-036 J2 energy balance connected two-element mesh, ten monotonic increments, four families Wext, Ue, Dp, signed/relative imbalance, point dissipation work-energy identity and non-negative dissipation PASS_INTERNAL G01
VV-037 J2 rollback adversarial TET4 connected mesh, injected rejected trial, cutback/retry state digest, clean displacement, retry log, final reference difference fixed-step reference PASS_INTERNAL G01/G09
VV-038 Linear buckling bounded homogeneous TET4/HEX8 preload with sparse generalized eigsh(K, M=-Kg) when positive definite, shift-invert eigs after a sparse bracket when indefinite, and diagnosed bracket fallback otherwise critical factor, formulation, bracket/fallback reason, tangent nnz, preload residual, critical-mode residual internal sparse contract PASS_INTERNAL G03
VV-039 Arc-length continuation proportional nonlinear TET4 path to target factor load-factor path, monotonicity, residual histories, endpoint comparison fixed load-control path OBSERVED_INTERNAL_RESEARCH G04
VV-040 Common frictionless contact TET4 penalty contact open/penetrating plus initial/updated common Newton active set, sparse tangent, search mode, residual unilateral contact contract OBSERVED_INTERNAL_RESEARCH G05/G06
VV-041 Coupled material/geometric path connected two-element TET4 total_lagrangian_j2 through common Newton kinematics, residual history, PEEQ, iterations common-driver contract OBSERVED_INTERNAL_RESEARCH G06
VV-042 Coupled geometric/contact path connected two-element TET4 Total-Lagrangian plus initial penalty contact contact mode, residual, displacement, iterations common-driver contract OBSERVED_INTERNAL_RESEARCH G06
VV-043 Coupled triple path connected two-element TET4 J2 + Total-Lagrangian + updated penalty contact shared driver, residual/tangent path, contact search mode, PEEQ common-driver contract OBSERVED_INTERNAL_RESEARCH G06
VV-044 Nonlinear path profiling TET4 load-control, arc-length, contact and coupled paths elapsed time, phase timings, DOF, Newton iterations, Python peak allocation, RSS when available reproducible benchmark telemetry PASS_INTERNAL G08
VV-045 External unilateral contact oracle scalar spring-supported point, compression and separation branches QF/Code_Aster displacement, gap closure/opening, active-set branch Code_Aster 18.1.0 Docker digest PASS_EXTERNAL_CORRELATION_BOUNDED G05/G10
VV-046 External TET4-TL buckling four structured TET4 column levels QF/CalculiX critical factor, Euler error, homogeneous stress patch CalculiX 2.20 Docker campaign PASS_EXTERNAL_CORRELATION_BOUNDED G03/G10
VV-070 External solid-family buckling probe shared one-cell TET4/TET10/HEX8/HEX20 decks, first factor factor, relative difference, execution diagnostics CalculiX 2.20 Docker campaign BLOCKED_EXTERNAL_TOOL G03/G10
VV-071 Bounded finite-sliding projection common penalty path with updated search and slave projection outside the current triangle clamped barycentrics, gap, active contact, sparse tangent and serialized projection mode internal contact contract OBSERVED_INTERNAL_RESEARCH G05
VV-047 Contact recontact path TET4 load path open/close/reopen/reclose active-set sequence, gaps, residual per load step common penalty driver OBSERVED_INTERNAL_RESEARCH G05/G09
VV-053 Contact penalty sensitivity TET4 node-to-triangle penalty sweep 1e2..1e6 convergence, penetration trend, contact tangent nnz, residual local penalty asymptotic trend OBSERVED_INTERNAL_RESEARCH G05/G08
VV-055 Multi-face contact search two-face planar master surface with two compatible slave positions selected face index, face count, gap, active set bounded geometry-search contract OBSERVED_INTERNAL_RESEARCH G05
VV-056 Updated-contact face crossing two connected TET4 elements with tangential/normal load and two-face master surface face sequence, switch count, gaps, residual history, common-driver status bounded updated-search contract OBSERVED_INTERNAL_RESEARCH G05/G06
VV-057 Large-deflection geometric smoke unit-block TET4/HEX8 transverse dead load end-line angle, detF, strain energy, residual history internal geometric research contract PASS_INTERNAL_RESEARCH G02
VV-066 Large-deflection mesh sensitivity regular TET4/HEX8 block levels 1/2/3/4 at pre-limit load scale 0.2 displacement, end-line angle, detF, energy, stress/strain, residual and Newton cost bounded internal geometric mesh-trend contract PASS_INTERNAL_RESEARCH G02
VV-067 High-order geometric mesh sensitivity regular TET10/HEX20 block levels 1/2 at load scale 0.25 displacement, detF, energy, residual and Newton cost internal low-load geometric mesh-trend contract OBSERVED_INTERNAL_RESEARCH G02
VV-058 Buckling mesh sensitivity assembled homogeneous levels 1/2 on TET4/TET10/HEX8/HEX20 critical factor, bracket, DDL, tangent nnz, preload residual, critical-mode residual, coarse/medium trend internal sparse mesh-trend contract OBSERVED_INTERNAL_RESEARCH G03
VV-059 Multi-family J2/geometric coupling connected two-element TET4/TET10/HEX8/HEX20 meshes shared driver, kinematics, residual, iterations, PEEQ internal common-driver contract OBSERVED_INTERNAL_RESEARCH G02/G06
VV-060 FEM arc-length path imperfect TET4 total-Lagrangian cantilever, 24 sparse continuation steps load-factor path, residual, detF, tip response existing sparse FEM arc-length contract OBSERVED_INTERNAL_RESEARCH G04
VV-061 Nonlinear component profiling regular two-cell J2 load-control, TET4/TET10/HEX8/HEX20 element setup/kernel/scatter, sparse conversion, solve time, nnz, cache hits/misses, RSS/allocation observations reproducible phase telemetry PASS_INTERNAL G08
VV-062 Multi-family J2/geometric/contact coupling one regular block per TET4/TET10/HEX8/HEX20 with active small-strain J2 and updated penalty contact crossing from open to active shared driver, active-step transition, initial/final gap, penetration, PEEQ, residual and Newton iterations bounded common-driver contact contract OBSERVED_INTERNAL_RESEARCH G05/G06
VV-063 Finite-kinematic limit recovery common small-load TET4/TET10/HEX8/HEX20 comparison between small-strain and total_lagrangian_j2 displacement recovery, residual, zero plastic strain small-strain regime consistency OBSERVED_INTERNAL_RESEARCH G02/G06
VV-064 Finite-kinematic arc-length common nonlinear driver on homogeneous TET4/TET10/HEX8/HEX20 total_lagrangian_j2 paths to signed load factor 0.5 with bounded adaptive radius load-factor path, final PEEQ, residual history, step count, radius controls and common-driver contract internal bounded monotone continuation contract OBSERVED_INTERNAL_RESEARCH G04/G06
VV-065 Geometric/contact composition Total-Lagrangian elastic TET4 plus fixed master patch through the common sparse penalty contribution active contact, final gap, penetration, residual, detF, strain energy bounded geometric/contact composition contract OBSERVED_INTERNAL_RESEARCH G02/G05/G06
VV-083 Coupled global tangent FD J2+geometry on TET4/TET10/HEX8/HEX20 and geometry+contact TET4 with fixed committed states centered global internal-force derivative, sparse tangent, maximum/mean relative error independent internal FD contract PASS_INTERNAL G06
VV-084 Geometry/contact mesh replay bounded TET4 geometry plus fixed node-to-triangle penalty contact at levels 1/2/4 active contact, gap/penetration, detF, residual and displacement trend bounded internal research mesh contract PASS_INTERNAL_RESEARCH G06
VV-085 J2+geometry external replay Code_Aster GDEF_LOG regular two-cell TET4/TET10/HEX8/HEX20, aligned load histories complete displacement/reaction/stress/PEEQ histories and external execution status pinned Code_Aster Docker replay; convention review required OPEN G06/G10
VV-086 J2+geometry Green-Lagrange replay Code_Aster GREEN_LAGRANGE regular two-cell TET4/TET10/HEX8/HEX20, aligned 16-point histories execution, displacement/reaction/stress/PEEQ histories and measured deviations pinned Code_Aster Docker replay; TET10 non-convergence and convention review remain OPEN G06/G10
VV-087 External CONTINUE contact oracle Code_Aster DKT surface pair with penalty contact and discrete support, independent of QF solid composition external execution, surface-contact response and limitation classification pinned Code_Aster Docker replay; bounded small-kinematic supporting oracle PASS_EXTERNAL_CORRELATION_BOUNDED G06/G10
VV-088 Finite-kinematic contact probe Same external contact oracle with GREEN_LAGRANGE requested fail-closed model capability diagnostic Code_Aster model constraint BLOCKED_EXTERNAL_MODEL G06/G10
VV-089 TET4 finite-kinematic contact comparison Code_Aster TET4 GREEN_LAGRANGE unilateral plane versus QF static active-set node-to-triangle path, ten load points displacement, gap and contact-reaction histories bounded observable mapping; reaction deviation requires review OPEN G06/G10
VV-048 High-order geometric path TET10/HEX20 common TL adapter residual, detF, energy, state recovery and tangent contract internal bounded research contract OBSERVED_INTERNAL_RESEARCH G02
VV-049 High-order linear buckling homogeneous TET10/HEX20 preload and sparse generalized/fallback tangent path, including indefinite-mass shift-invert refinement critical factor, formulation, fallback reason, bracket width, tangent nnz, preload residual internal sparse research contract OBSERVED_INTERNAL_RESEARCH G03
VV-050 Path failure taxonomy capped arc-length continuation and unbracketed buckling factor reason, converged flag, solver diagnostics, no partial success structured failure contract PASS_INTERNAL G09
VV-051 Multi-step rollback injected failure after one accepted adaptive increment committed prefix, cutback factor, retry path, final convergence fixed-step/adaptive transaction contract PASS_INTERNAL G01/G09
VV-054 Contact penetration failure configured penalty penetration limit exceeded on a trial typed reason, converged=false, penetration diagnostics failure contract PASS_INTERNAL G05/G09
VV-068 Contact penetration cutback/retry regular multi-element TET4 contact guard with adaptive load control typed guard failures, committed factors, retry log, final displacement/reaction/gap versus fine-step reference fixed small-step reference PASS_INTERNAL G05/G09
VV-052 Arc-length limit point reduced shallow-arch branch through the analytical turning point branch turn, reference curve error, residual history analytical reduced-order reference OBSERVED_INTERNAL_RESEARCH G04
VV-069 State transaction corruption guard committed material/contact state mutated during a detached trial typed STATE_CORRUPTION, before/after digests, no silent rollback transaction invariant PASS_INTERNAL G01/G09
VV-072 Arc-length retry telemetry injected failed continuation trial followed by radius cutback/retry failure reason, failed radius, retry radius, failure diagnostics, rollback flag and retry result common transaction and continuation contract PASS_INTERNAL G04/G09
VV-073 Non-finite correction classification sparse correction NaN, +Inf, -Inf and mixed payload policy deterministic reason, converged=false, structured Full Newton diagnostics non-finite failure contract PASS_INTERNAL G09
VV-074 Common-driver FEM snap-through two-element TET4 total-Lagrangian model with signed arc-length continuation through a load-factor turning point signed load-factor path, control displacement, predictor/branch diagnostics, residual, det(F) bounded internal branch-tracking contract OBSERVED_INTERNAL_RESEARCH G04
VV-075 Common-driver FEM arc-length restart restart from the checkpoint immediately before the observed turning point continuation suffix, restart metadata, endpoint displacement and state equality continuous-run suffix comparison OBSERVED_INTERNAL_RESEARCH G04
VV-076 Common-driver FEM arc-length rollback controlled Newton failure after two corrections near the turning point failure classification, radius cutback, clean retry state, committed step history common transaction/retry contract OBSERVED_INTERNAL_RESEARCH G04/G09
VV-077 Contact slave patch two slave nodes against a two-face triangulated master surface round-trip schema, active contacts, selected facets, sparse tangent bounded node-patch contact contract OBSERVED_INTERNAL_RESEARCH G05
VV-078 Contact three-facet sliding one slave node traversing three connected master facets face sequence, facet switches, gap, active state, sparse tangent bounded updated-search contract OBSERVED_INTERNAL_RESEARCH G05
VV-079 Contact updated normals deformed master facet under updated finite-sliding search finite unit normal, updated orientation, active contact current-configuration geometry contract OBSERVED_INTERNAL_RESEARCH G05
VV-080 Contact facet-transition rollback failed trial after a facet transition, followed by replay and retry failure reason, committed facet, rollback facet, retry facet, deterministic force common contact transaction contract OBSERVED_INTERNAL_RESEARCH G05/G09
VV-081 Geometric small-strain limit elastic Total-Lagrangian versus small-strain static, TET4/TET10/HEX8/HEX20 at 1e-2/1e-3/1e-4 load factors relative displacement error and positive det(F) trend finite-kinematic limit consistency PASS_INTERNAL G02
VV-082 Geometric external correlation TET4/HEX8, twelve-step Green-Lagrange Code_Aster ramp with initial state full tip displacement, reactions, SIEF_ELGA stress and EPSI_ELGA strain histories pinned Code_Aster 18.1 Docker image PASS_EXTERNAL_CORRELATION_BOUNDED G02/G10

VV-033 uses the explicit QF parameter tet10_nonlinear_quadrature=code_aster_5 for TET10. This matches the five ELGA values per TETRA10 element exposed by the pinned Code_Aster run. The legacy Hammer four-point rule remains the default for existing models. The bounded external row is therefore convention-matched, but it does not alone close G01/G10 because those gates also require energy, mesh, cyclic, rollback, threshold and final-SHA evidence.

G04 Owner audit outcome

The controlled G04 audit is archived under results/vnv_0_2_5/g04_latest/. VV-074, VV-075 and VV-076 remain internal research evidence: the common-driver FEM path crosses a signed load-factor turn, restart reproduces the continuation suffix, and rollback records a radius cutback and clean retry. The required arc-length mesh study is absent.

The historical monotone Code_Aster path was diagnosed as a configuration mismatch, not a numerical deviation: it applied the opposite physical load and used mean crown displacement instead of QF's apex control quantity. The corrected pinned replay reproduces a single turning branch, with maximum load-factor difference 4.8719e-07 when parameterized by apex displacement. The detailed bounded diagnostic is DOC-NL-025-030.

These observations do not satisfy G04 closure. The external discrepancy is resolved, but no exact published FEM reference and no required four-level arc-length mesh study are available. No arc-length production claim is promoted.

Controlled G02 evidence

The dedicated controlled pack is archived under results/vnv_0_2_5/g02_latest/. Its summary.json, report.md, gate_decision.json and evidence_manifest.json record the same source SHA, dirty=false, UTC timestamp, command, environment and SHA-256 digests. It contains four-family objectivity, sparse tangent finite-difference, bounded TET4/HEX8 large-rotation, four-level pre-limit mesh, four-family small-strain limit and pinned Code_Aster TET4/HEX8 correlation evidence.

The numerical source pack deliberately reports 025-G02 = OPEN: it is the immutable pre-Owner evidence record. The Owner decision in 0_2_5_g02_owner_review.md closes G02 for the bounded elastic TET4/HEX8 scope only, with APPROVED_BOUNDED_REFINEMENT. The Code_Aster correlation remains bounded numerical code-to-code evidence, not physical validation. This decision does not qualify total_lagrangian_j2, high-order plastic finite-kinematic paths, post-limit load control, buckling, arc-length, contact or coupling, and it does not alter G03-G06.

Historical and exploratory evidence

The controlled G09 evidence is archived under results/vnv_0_2_5/g09_latest/. Its manifest is authoritative for the exact candidate SHA and reports dirty=false; the report contains 22 cases, all passed. In particular, the correction path now distinguishes NaN from both signs of Inf, and mixed arrays give INF_DETECTED deterministically when an infinite value is present. This closure is limited to failure handling and does not alter the status of functional gates.

The observed rows are backed by the following implementation and test artifacts. They remain provisional until regenerated and attached to a final release SHA:

  • J2 constitutive, tangent and transaction paths: tests/unit/test_nonlinear_constitutive_vv.py, tests/unit/test_robustness_tangent_fd.py, tests/unit/test_nonlinear_state_transaction_contract.py;
  • four-family connected mesh and load-step sensitivity: tests/unit/test_nonlinear_multielement.py, tests/unit/test_nonlinear_load_step_sensitivity.py;
  • four-family mesh refinement and cyclic paths: tests/unit/test_nonlinear_multielement.py, tests/unit/test_nonlinear_cyclic.py, src/solveur/verification/robustness_nonlinear_solids.py;
  • four-family energy balance: tests/unit/test_nonlinear_multielement.py, src/solveur/verification/robustness_nonlinear_solids.py;
  • adversarial rollback and fixed-step comparison: tests/unit/test_nonlinear_multielement.py, src/solveur/verification/robustness_nonlinear_solids.py;
  • bounded buckling, arc-length and common contact observations: tests/unit/test_nonlinear_multielement.py, src/solveur/verification/robustness_nonlinear_solids.py;
  • finite-kinematic arc-length bounded four-family path with adaptive-radius contract: tests/unit/test_nonlinear_multielement.py, tests/unit/test_total_lagrangian_j2.py, src/solveur/core/nonlinear.py, src/solveur/core/nonlinear_controls.py, src/solveur/verification/robustness_nonlinear_solids.py, results/vnv_0_2_5/robustness_high_order_latest/finite_kinematic_arc_length.png, results/benchmark_0_2_5/arc_length_finite_kinematic_latest.json;
  • targeted performance replays for the bounded geometric, contact and coupled paths: results/benchmark_0_2_5/geometric_static_all_families_latest.json, results/benchmark_0_2_5/contact_tet4_latest.json, results/benchmark_0_2_5/coupled_tet4_latest.json; these are single-run dirty-worktree observations and remain non-qualifying for G02/G05/G06/G08;
  • geometric Total-Lagrangian plus penalty contact composition: tests/unit/test_nonlinear_multielement.py, src/solveur/core/geometric_nonlinear.py, src/solveur/mesh/contact_validation.py, src/solveur/verification/robustness_nonlinear_solids.py;
  • reduced shallow-arch arc-length branch-following evidence: tests/unit/test_nonlinear_multielement.py, results/vnv_0_2_5/robustness_high_order_latest/summary.json, results/vnv_0_2_5/robustness_high_order_latest/shallow_arch_arc_length.png;
  • bounded analytical Euler buckling evidence: results/vnv_0_2_5/robustness_high_order_latest/euler_buckling/summary.json, results/vnv_0_2_5/robustness_high_order_latest/euler_buckling/report.md, src/solveur/verification/tet4_total_lagrangian_buckling.py;
  • high-order geometric and buckling research observations: tests/unit/test_geometric_nonlinear_public.py, tests/unit/test_linear_buckling.py, tests/unit/test_nonlinear_multielement.py, results/vnv_0_2_5/robustness_high_order_geometric_latest/summary.json, results/vnv_0_2_5/robustness_high_order_buckling_latest/summary.json;
  • large-deflection geometric smoke VV-057: src/solveur/verification/robustness_nonlinear_solids.py, tests/unit/test_nonlinear_multielement.py, results/vnv_0_2_5/robustness_high_order_latest/summary.json;
  • large-deflection mesh sensitivity VV-066: src/solveur/verification/robustness_nonlinear_solids.py, tests/unit/test_nonlinear_multielement.py, results/vnv_0_2_5/robustness_high_order_latest/summary.json;
  • high-order low-load mesh sensitivity VV-067: src/solveur/verification/robustness_nonlinear_solids.py, tests/unit/test_nonlinear_multielement.py, results/vnv_0_2_5/robustness_high_order_latest/summary.json;
  • assembled buckling mesh sensitivity VV-058: src/solveur/verification/robustness_nonlinear_solids.py, tests/unit/test_nonlinear_multielement.py, results/vnv_0_2_5/robustness_high_order_latest/summary.json; the sparse buckling rows now also record the normalized critical-mode residual at the reported factor; this is a diagnostic, not a qualification threshold or a closure of 025-G03.
  • multi-family J2/geometric coupling VV-059: src/solveur/verification/robustness_nonlinear_solids.py, tests/unit/test_nonlinear_multielement.py, results/vnv_0_2_5/robustness_high_order_latest/summary.json;
  • FEM arc-length path VV-060: src/solveur/verification/robustness_nonlinear_solids.py, tests/unit/test_nonlinear_multielement.py, results/vnv_0_2_5/robustness_high_order_latest/summary.json, src/solveur/verification/total_lagrangian_structural.py;
  • component-level nonlinear profiling VV-061: src/solveur/core/nonlinear_assembly.py, src/solveur/core/nonlinear_controls.py, scripts/benchmark_nonlinear_025.py, results/benchmark_0_2_5/nonlinear_load_control_component_profile_latest.json;
  • reusable nonlinear assembly-plan contract and replay extension of VV-061: src/solveur/core/nonlinear_assembly.py, src/solveur/core/nonlinear.py, tests/unit/test_nonlinear_assembly_plan.py, results/benchmark_0_2_5/nonlinear_load_control_cache_latest.json;
  • Newton residual-rate characterization VV-026: src/solveur/verification/robustness_nonlinear_solids.py, tests/verification/test_robustness_newton_rate_vnv.py, results/vnv_0_2_5/robustness_high_order_latest/summary.json;
  • four-family finite-kinematic contact composition VV-062: src/solveur/verification/robustness_nonlinear_solids.py, tests/unit/test_nonlinear_multielement.py, results/vnv_0_2_5/robustness_high_order_latest/summary.json;
  • finite-kinematic small-strain limit recovery VV-063: src/solveur/verification/robustness_nonlinear_solids.py, tests/unit/test_nonlinear_multielement.py, results/vnv_0_2_5/robustness_high_order_latest/summary.json;
  • controlled G06 coupled replay (VV-022, VV-023, VV-024, VV-059 and VV-062 observations): results/vnv_0_2_5/g06_latest/summary.json, results/vnv_0_2_5/g06_latest/report.md and results/vnv_0_2_5/g06_latest/evidence_manifest.json, plus the complete results/vnv_0_2_5/g06_latest/coupled_histories.json capture. The replay is internal research evidence only and does not satisfy the missing external MUST correlation. The companion results/vnv_0_2_5/g06_diagnostic/summary.json records an independent global coupled tangent FD check, with maximum relative errors below 2.4e-9 for J2+geometry and 1.2e-10 for geometry+contact. The companion results/vnv_0_2_5/g06_geometry_contact_mesh/summary.json records the bounded three-level geometry/contact mesh replay. Neither sub-proof closes the Code_Aster MUST or promotes finite-kinematic J2 beyond research.
  • external G06 probes VV-086, VV-087, VV-088 and VV-089: results/vnv_0_2_5/g06_j2_geometry_code_aster/green_lagrange/green_comparison.json, results/vnv_0_2_5/g06_geometry_contact_code_aster/summary_linear.json, results/vnv_0_2_5/g06_geometry_contact_code_aster/summary.json and results/vnv_0_2_5/g06_geometry_contact_code_aster/tet4_green_lagrange/comparison.json. The Green-Lagrange replay completes TET4, HEX8 and HEX20 but TET10 fails at the first load point; its measured deviations remain under convention review. The independent CONTINUE contact oracle passes only in its required small- kinematic setting. The deformable TET4 Green-Lagrange contact probe reaches ten load points and agrees in displacement/gap after activation, but its mapped reaction history differs by up to 76.7% and is not an identical multiplier observable. QF uses its historical exact active-set multiplier path for this static comparison; the penalty formulation belongs to the separate bounded research composition. These results strengthen the evidence record but do not close the finite-kinematic solid MUST.
  • aggregated adversarial failure contract VV-027, VV-050, VV-051 and VV-054: src/solveur/verification/nonlinear_failure_campaign.py, tests/unit/test_nonlinear_failure_campaign.py, results/vnv_0_2_5/robustness_high_order_latest/summary.json;
  • bounded coupling composition observations (J2/geometric/contact): tests/unit/test_nonlinear_multielement.py, src/solveur/verification/robustness_nonlinear_solids.py;
  • repeated all-family performance characterization: scripts/benchmark_nonlinear_025.py, results/benchmark_0_2_5/nonlinear_load_control_all_families_repeats2_latest.json;
  • bounded external unilateral contact oracle: results/vnv_0_2_5/contact_code_aster_liaison_unil/summary.json, results/vnv_0_2_5/contact_code_aster_liaison_unil/report.md, tests/verification/test_code_aster_contact_vnv.py;
  • bounded external TET4-TL stress/buckling campaign: results/vnv_0_2_5/calculix_tl_structural/summary.json, results/vnv_0_2_5/calculix_tl_structural/report.md, qualification/vnv/external/calculix_tl_structural/reference/summary.json;
  • external solid-family buckling probe VV-070: src/solveur/verification/calculix_buckling_025.py, scripts/run_calculix_buckling_025.py, tests/unit/test_calculix_buckling_025.py, results/vnv_0_2_5/calculix_buckling_solid_families_mode1_recorded/summary.json; the deck applies the fixed boundary set, requests only the first factor and bounds the Lanczos subspace to the free-equation count. The current CalculiX replay records relative differences of about 24.6 % (TET4), 45.1 % (TET10) and 13.4 % (HEX8), plus a native C3D20 execution stop. These are explicit external blockers, not PASS evidence and do not close G03/G10.
  • common contact recontact path: src/solveur/verification/robustness_nonlinear_solids.py, tests/unit/test_nonlinear_contact_composition.py, tmp/vnv_0_2_5_recontact_latest/summary.json;
  • updated-contact face crossing VV-056: src/solveur/verification/robustness_nonlinear_solids.py, tests/unit/test_nonlinear_contact_composition.py, results/vnv_0_2_5/robustness_high_order_latest/summary.json;
  • bounded finite-sliding projection VV-071: src/solveur/contact/entities.py, src/solveur/contact/solver.py, src/solveur/mesh/contact_validation.py, src/solveur/verification/robustness_nonlinear_solids.py and tests/unit/test_contact_finite_sliding.py; the opt-in path is restricted to updated frictionless penalty contact, records a deterministic two-face clamped-projection crossing, and remains non-qualifying for a general surface-to-surface claim. The common Newton step now also exposes the finite-sliding flag, clamped-projection flags, closest distances and the exact/bounded projection mode in its serialized increment diagnostics;
  • local penalty contact tangent finite-difference contract: src/solveur/verification/robustness_nonlinear_solids.py, tests/unit/test_nonlinear_multielement.py, results/vnv_0_2_5/robustness_high_order_latest/summary.json; the fixed-active node-to-triangle tangent is compared with centered finite differences over three perturbation sizes. The result remains internal research evidence and excludes active-set transitions;
  • minimum-increment failure contract: tests/unit/test_nonlinear_failure_campaign.py, src/solveur/verification/nonlinear_failure_campaign.py;
  • sparse backend failure classification: tests/unit/test_nonlinear_failure_modes.py verifies that a runtime failure from the sparse factorization backend is reported as LINEAR_SOLVER_FAILURE, distinct from SINGULAR_TANGENT;
  • adversarial campaign aggregation: tests/unit/test_nonlinear_failure_campaign.py verifies the same typed backend failure in the campaign report, alongside the injected contact cutback/rollback case;
  • contact assembly failure with adaptive cutback/retry: tests/unit/test_nonlinear_failure_campaign.py, src/solveur/verification/nonlinear_failure_campaign.py;
  • real contact penetration guard cutback/retry VV-068: tests/unit/test_nonlinear_failure_campaign.py, src/solveur/verification/nonlinear_failure_campaign.py; the guarded multi-element TET4 result follows [0.5, 0.75, 1.0] after two typed penetration failures and matches an eight-step reference within the recorded displacement/reaction/gap limits;
  • committed-state corruption guard VV-069: src/solveur/core/material_state.py checks a digest captured at begin_trial() before either commit or rollback, and raises a structured STATE_CORRUPTION error if the committed object was mutated in place; tests/unit/test_nonlinear_state_transaction_contract.py covers both the generic contact-style transaction and the material-state session;
  • multi-step adaptive rollback after a previously committed increment: tests/unit/test_nonlinear_failure_campaign.py, src/solveur/verification/nonlinear_failure_campaign.py;
  • arc-length checkpoint/restart contract: tests/unit/test_nonlinear_checkpoint.py, src/solveur/core/nonlinear_checkpoint.py, src/solveur/io/nonlinear_checkpoint.py;
  • Total-Lagrangian objectivity/tangent: tests/unit/test_total_lagrangian_hex8.py, tests/unit/test_geometric_nonlinear_public.py;
  • high-order Total-Lagrangian J2 integration and post-processing: tests/unit/test_total_lagrangian_j2.py covers TET10 and HEX20 through the common nonlinear driver, including residual, material state and integration-point recovery. The bounded finite-kinematic campaign now also records TET10 and HEX20 rows with rigid-rotation and tangent-FD checks. This is an internal research observation only. The controlled campaign replay is archived at results/vnv_0_2_5/robustness_high_order_latest/summary.json and remains non-qualifying for G02/G06.
  • failure contracts and sparse continuation: tests/unit/test_nonlinear_failure_modes.py, tests/unit/test_nonlinear_iteration_sparse.py, tests/unit/test_nonlinear_failure_campaign.py, src/solveur/verification/nonlinear_failure_campaign.py;
  • internal campaign runner: src/solveur/verification/robustness_nonlinear_solids.py.

G03 controlled closure update

The earlier buckling observations in this document are historical and are superseded for G03 by the final bounded evidence in results/vnv_0_2_5/g03_euler_final/ and results/vnv_0_2_5/g03_final/. The current qualified source is 85c75d06955976251dd54ad782f57f1eb5a7f8f4; the source worktree was clean when the evidence was generated. The corrected path isolates the initial-stress geometric tangent from the material tangent. It passes the four-level Euler study, the exact TET4 Code_Aster factor/mode probe and the buckling failure contract. G03 is therefore PASS for its bounded TET4-first-instability scope. TET10/HEX8/HEX20 external buckling, post-buckling and imperfection-sensitive collapse remain research/unqualified, and CalculiX remains a SHOULD comparison with retained negative high-order evidence.