Commit 699a256
committed
Phase F: predictor-corrector with radial return — first cut
Implements the predictor-corrector architecture from the web-advice doc
(vep_stress_update_full_latex.md §15 Stage 4) on isotropic VEP. Setup:
* Same harmonic shear box, RES=32
* Isotropic Maxwell rheology (no director, no rank-4 tensor)
* σ_y(x) from the same fault influence-function used in TI tests —
small (0.05) in a 0.06-wide layer at θ=15°, large (200) in the bulk
→ partial yielding, no global yield catastrophe
* Per-step trace files written EVERY step with flush, persist on kill
Configurations tested:
1. BDF-1 yield-in-residual (softmin in model) — baseline, works
2. ETD with yield_stress=∞ in model + post-solve J2 radial return on
psi_star (predictor-corrector single-shot)
3. ETD with yield_stress=∞ + radial return + outer Picard with σ
damping (ω_τ=0.5, max 6 iters)
Both 2 and 3 leave η_eff fixed at η_VE between Picard iters — i.e.
single Stokes solver per Picard step, just with iterated psi_star.
Killer-test results (θ=+15° fault layer, τ_y_fault=0.05, 1.5T):
| case | |σ_eq|_max end | |u_y|_max end |
|-----------------------------|----------------|---------------|
| BDF-1 baseline | 0.81 | 0.036 |
| ETD-1 + RR single-shot | 0.42 | 0.021 |
| ETD-1 + RR + Picard | 0.41 | 0.021 |
| ETD-2 + RR single-shot | 0.41 | 0.286 |
| ETD-2 + RR + Picard | runaway @ 60 | runaway |
Same first-order vs higher-order signature as before:
* ETD-1 + radial return is essentially identical to BDF-1, both
single-shot and Picard-equilibrated. The first-order numerical
dissipation (lesson #13) carries over to the predictor-corrector
architecture.
* ETD-2 + radial return drifts in the single-shot case (|u_y| 8× BDF
by end) and the outer Picard with σ-damping doesn't damp it — at
step 60 the case exits via the runaway guard with |σ|_eq = 18.2
and |u_y| = 10.2. Picard amplifies rather than equilibrates because
the σ-damping doesn't address the underlying lack of dissipation in
the linear-relaxation factor.
The single Stokes per Picard iter (with η_eff fixed at η_VE) is
NOT the full architecture from the web advice. The advice's Step 4
explicitly calls for *updating η_eff between Picard iters based on
σ_corrected*, which would require either (a) per-quad-point α,φ in
the constitutive model (we ruled out earlier — codegen explosion)
or (b) a second Stokes solve with viscosity-as-plastic-multiplier
treating σ_VE as fixed/explicit (the user's "two-Stokes" idea —
not implemented here).
Note: ETD-1/ETD-2 yield-in-residual baselines on this isotropic
setup hit a separate parent-class issue with spatial yield_stress
+ softmin (SNES line-search divergence). Filed as a follow-up;
sidestepped here by using yield_stress=∞ with external return mapping.
Per-step trace files (`_phase_f_<label>.trace.txt`) tracked in git
so the comparison plot regenerates from a fresh clone.
Underworld development team with AI support from Claude Code (https://claude.com/claude-code)1 parent 20db685 commit 699a256
7 files changed
Lines changed: 1101 additions & 0 deletions
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
| 1 | + | |
| 2 | + | |
| 3 | + | |
| 4 | + | |
| 5 | + | |
| 6 | + | |
| 7 | + | |
| 8 | + | |
| 9 | + | |
| 10 | + | |
| 11 | + | |
| 12 | + | |
| 13 | + | |
| 14 | + | |
| 15 | + | |
| 16 | + | |
| 17 | + | |
| 18 | + | |
| 19 | + | |
| 20 | + | |
| 21 | + | |
| 22 | + | |
| 23 | + | |
| 24 | + | |
| 25 | + | |
| 26 | + | |
| 27 | + | |
| 28 | + | |
| 29 | + | |
| 30 | + | |
| 31 | + | |
| 32 | + | |
| 33 | + | |
| 34 | + | |
| 35 | + | |
| 36 | + | |
| 37 | + | |
| 38 | + | |
| 39 | + | |
| 40 | + | |
| 41 | + | |
| 42 | + | |
| 43 | + | |
| 44 | + | |
| 45 | + | |
| 46 | + | |
| 47 | + | |
| 48 | + | |
| 49 | + | |
| 50 | + | |
| 51 | + | |
| 52 | + | |
| 53 | + | |
| 54 | + | |
| 55 | + | |
| 56 | + | |
| 57 | + | |
| 58 | + | |
| 59 | + | |
| 60 | + | |
| 61 | + | |
| 62 | + | |
| 63 | + | |
| 64 | + | |
| 65 | + | |
| 66 | + | |
| 67 | + | |
| 68 | + | |
| 69 | + | |
| 70 | + | |
| 71 | + | |
| 72 | + | |
| 73 | + | |
| 74 | + | |
| 75 | + | |
| 76 | + | |
| 77 | + | |
| 78 | + | |
| 79 | + | |
| 80 | + | |
| 81 | + | |
| 82 | + | |
| 83 | + | |
| 84 | + | |
| 85 | + | |
| 86 | + | |
| 87 | + | |
| 88 | + | |
| 89 | + | |
| 90 | + | |
| 91 | + | |
| 92 | + | |
| 93 | + | |
| 94 | + | |
| 95 | + | |
| 96 | + | |
| 97 | + | |
| 98 | + | |
| 99 | + | |
| 100 | + | |
| 101 | + | |
| 102 | + | |
| 103 | + | |
| 104 | + | |
| 105 | + | |
| 106 | + | |
| 107 | + | |
| 108 | + | |
| 109 | + | |
| 110 | + | |
| 111 | + | |
| 112 | + | |
| 113 | + | |
| 114 | + | |
| 115 | + | |
| 116 | + | |
| 117 | + | |
| 118 | + | |
| 119 | + | |
| 120 | + | |
| 121 | + | |
| 122 | + | |
| 123 | + | |
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
| 1 | + | |
| 2 | + | |
| 3 | + | |
| 4 | + | |
| 5 | + | |
| 6 | + | |
| 7 | + | |
| 8 | + | |
| 9 | + | |
| 10 | + | |
| 11 | + | |
| 12 | + | |
| 13 | + | |
| 14 | + | |
| 15 | + | |
| 16 | + | |
| 17 | + | |
| 18 | + | |
| 19 | + | |
| 20 | + | |
| 21 | + | |
| 22 | + | |
| 23 | + | |
| 24 | + | |
| 25 | + | |
| 26 | + | |
| 27 | + | |
| 28 | + | |
| 29 | + | |
| 30 | + | |
| 31 | + | |
| 32 | + | |
| 33 | + | |
| 34 | + | |
| 35 | + | |
| 36 | + | |
| 37 | + | |
| 38 | + | |
| 39 | + | |
| 40 | + | |
| 41 | + | |
| 42 | + | |
| 43 | + | |
| 44 | + | |
| 45 | + | |
| 46 | + | |
| 47 | + | |
| 48 | + | |
| 49 | + | |
| 50 | + | |
| 51 | + | |
| 52 | + | |
| 53 | + | |
| 54 | + | |
| 55 | + | |
| 56 | + | |
| 57 | + | |
| 58 | + | |
| 59 | + | |
| 60 | + | |
| 61 | + | |
| 62 | + | |
| 63 | + | |
| 64 | + | |
| 65 | + | |
| 66 | + | |
| 67 | + | |
| 68 | + | |
| 69 | + | |
| 70 | + | |
| 71 | + | |
| 72 | + | |
| 73 | + | |
| 74 | + | |
| 75 | + | |
| 76 | + | |
| 77 | + | |
| 78 | + | |
| 79 | + | |
| 80 | + | |
| 81 | + | |
| 82 | + | |
| 83 | + | |
| 84 | + | |
| 85 | + | |
| 86 | + | |
| 87 | + | |
| 88 | + | |
| 89 | + | |
| 90 | + | |
| 91 | + | |
| 92 | + | |
| 93 | + | |
| 94 | + | |
| 95 | + | |
| 96 | + | |
| 97 | + | |
| 98 | + | |
| 99 | + | |
| 100 | + | |
| 101 | + | |
| 102 | + | |
| 103 | + | |
| 104 | + | |
| 105 | + | |
| 106 | + | |
| 107 | + | |
| 108 | + | |
| 109 | + | |
| 110 | + | |
| 111 | + | |
| 112 | + | |
| 113 | + | |
| 114 | + | |
| 115 | + | |
| 116 | + | |
| 117 | + | |
| 118 | + | |
| 119 | + | |
| 120 | + | |
| 121 | + | |
| 122 | + | |
| 123 | + | |
0 commit comments