[petlenz]

Closes #113. Part of the uvwmaterial port (#94); stacked on #140.

materials/hencky_elasticity.h (hencky_elasticity): ε = ln V = ½ ln b with b = FFᵀ, τ = C : (ε − ε_eig), P = τ F⁻ᵀ, dP/dF = (C : d ln V/dF)·F⁻ᵀ − P ⊠̄ F⁻¹ with d ln V/dF = ½ (d ln b/db) : (I⊠F + F⊠̄I); the stiffness a consumed property (tangent_source, e.g. an isotropic_tangent of produced K and G — so cure-dependent moduli are a wiring), optional eigenstrain_source (thermal or chemical eigenstrain in the log strain), output log_strain. Dim generic. Replaces hencky_strains_linear_elastic and its thermo / thermo-chemo variants (the latter are hencky_elasticity plus thermal_eigenstrain).

tmech pin → main 46d8fdd (tmech#38). The report of this PR found that tmech's log and polar_decomposition were wrong at near-repeated eigenvalues (petlenz/tmech#37); the first version of this PR carried an interim core/log_strain.h on Eigen. With the fix merged the material uses tmech::log(b) and .derivative() directly, and the interim file, its tests and the Eigen include are gone. A tag would be preferable to the commit; FIND_PACKAGE_ARGS 1.2 still accepts an installed 1.2.0, which lacks the fix — the next tmech release should raise it.

Reports: load_case::fd_step is std::optional; a case may set it, otherwise the driver default applies instead of being overwritten. The finite-strain standard's fd_step: 1e-6 workaround is dropped; report_hencky_elasticity, report_neo_hooke and report_saint_venant_kirchhoff pass at 1e-7. docs/materials/hencky_elasticity/report.json: the finite-strain standard plus the uniaxial, volumetric and simple-shear closed forms.

Tests (test_hencky_elasticity, 7): dilatation P = 3K ln λ/λ exactly; the small-strain limit is linear elasticity; under uniaxial stress ln F₂₂ = −ν ln F₁₁ exactly and P₁₁ = E ln λ/λ, tangent against central differences at λ = 1.1, 1.5, 0.7 and in simple shear; the tangent at exactly and nearly repeated stretches (the tmech#37 symptom: (1.15, 1.15, 1), (1.2, 1.2, 1.2), (1.15, 1.15 + 1e-9, 1), (1.204, 0.9458, 0.9458)); objectivity P(RU) = R P(U) and P(R) = 0; constrained heating and free expansion with an eigenstrain; JSON.

Docs: docs/uvwmaterial-mapping.md row for the Hencky family; the card converter points hencky_strains_linear_* at hencky_elasticity (like saint_venant_kirchhoff, a finite-strain card needs an external_deformation_gradient_source written by hand).

907/907 tests pass (GCC 14, reports on).

JSON input

New material types: hencky_elasticity. A json_model document that loads (the FromJson test of the PR):

{"materials": [
  {"type": "external_deformation_gradient_source", "name": "F"},
  {"type": "constant_scalar", "name": "K", "value": 166.67},
  {"type": "constant_scalar", "name": "G", "value": 76.92},
  {"type": "isotropic_tangent", "name": "stiffness", "K_source": "K", "G_source": "G"},
  {"type": "hencky_elasticity", "name": "h", "deformation_gradient_source": "F", "tangent_source": "stiffness"}
]}

tangent_source a produced small-strain stiffness (cure-dependent moduli are a wiring); optional eigenstrain_source (a thermal_eigenstrain); outputs stress (P), tangent (dP/dF), log_strain.

[petlenz]

The cause behind core/log_strain.h is fixed in tmech: petlenz/tmech#38 (Jacobi eigenpairs, projector eigenbases, closes petlenz/tmech#37). With that tmech, tmech::log(polar_decomposition(F).V()) and its derivative chain are exact at repeated stretches (central difference with h = 1e-7: 2e-9 instead of 0.46).

Plan for this PR once tmech#38 is merged and tagged: bump the tmech pin, drop core/log_strain.h and test_log_strain, and let hencky_elasticity use tmech::log(V) with log_V.derivative() : V().derivative() directly (the first version of this branch). Until then the Eigen-based evaluation keeps the material correct against the pinned tmech.

[petlenz]

Review findings (design, tmech expressions instead of component loops, modern C++, comments, tests, correctness). Severity: must = wrong result, hang, build break or unreachable feature; should = design, duplication, accuracy; nit = style.

hencky_elasticity — needs changes

• [must] core/log_strain.h:17-23 hard-codes tensor<T,3,2>; hencky_elasticity<Traits> passes Dim → ill-formed for Dim ≠ 2… i.e. only 3D compiles while neo_hooke/SVK are Dim-generic. Template on Dim.
• [should] component loops marshalling to/from Eigen → Eigen::Map<Matrix<T,3,3> const>(b.raw_data()) (precedent unknown_layout.h:307) and tmech::adaptor<T const,3,1,full<3>> on eigenvector columns + otimes(n,n).
• [should] divided-difference switch at 1e-8·mean sits where the loss is worst; use sqrt(eps) or log1p(d/λ)/(2d) for every d.
• [should] mapping doc and converter still say Hencky is unported (hencky_elasticity: linear elasticity in logarithmic strain with thermal and chemical eigenstrain #113).
• [should] case_runner.h:123 opts.fd_step = c.fd_step discards a caller's option → std::optional + value_or.
• [should] swap plan: simpler than log(V) + polar derivative is ln_b = tmech::log(b); strain = ½ ln_b; derivative = ½ dcontract(ln_b.derivative(), db_dF) (what uvwmaterial did; db_dF is already a tmech expression).
• [should] no objectivity test (P(RU) = R P(U), P(R) = 0).
• [nit] deformation_gradient_source fetched twice; <array> missing; [[nodiscard]]; 6-line header; fd_step 1e-6 rationale.

[petlenz]

Review findings addressed in the last commit (tmech#38 is merged, pin → main 46d8fdd):

• must, Dim 3 only: core/log_strain.h is deleted; the material uses tmech::log(b) and .derivative(), which exist for Dim 2 and 3, and has no 3D code left.
• should, Eigen marshalling loops: gone with the file.
• should, divided-difference switch at 1e-8: gone with the file; tmech decides by the cancellation of f(λᵢ) − f(λⱼ) (tmech#38).
• should, mapping doc / converter: row for hencky_strains_linear_elastic and the thermo variants → hencky_elasticity (+ thermal_eigenstrain); the converter raises the same hand-written hint as for saint_venant_kirchhoff.
• should, opts.fd_step overwritten: load_case::fd_step is std::optional<T>, value_or(opts.fd_step); test_report_case covers unset and set.
• should, swap plan: done as suggested, ln_b = tmech::log(b), strain = ½ ln_b, derivative = ½ (ln_b.derivative() : db/dF).
• should, objectivity test: Hencky.IsObjective (P(RU) = R P(U) to 1e-12, P(R) = 0); plus TangentIsConsistentAtRepeatedStretches for the tmech#37 cases.
• nit, source fetched twice: m_F_name like neo_hooke; update → compute; header comment 2 lines; fd_step 1e-6 removed from the finite-strain standard (all three finite-strain reports pass at 1e-7).

Not done here: a 2D runtime test — external_deformation_gradient_source is 3D only (static_assert at external_state_source.h:97), so a 2D Hencky point has no F provider yet.

[petlenz]

Producer/consumer pass: hencky_elasticity consumes its stiffness (tangent_source, tangent_property; an isotropic_tangent of produced K and G in the tests and the report) instead of carrying K/G parameters — the chemo variant of uvwmaterial with cure-dependent moduli is then a wiring, not a new material. The thermal eigenstrain in the fixture consumes alpha_source. Mapping row and converter hint updated. 929/929.