E. What we must replicate for a like-for-like comparison
E. What we must replicate for a like-for-like comparison¶
Ordered by expected effect on the ionic-current QoI.
| # | Setting to match | Value | Why it matters |
|---|---|---|---|
| 1 | TDS stabilisation | streamline + crosswind on, crosswind = Do Carmo and Galeão, approximate residual, isotropic off | Adds a residual-based term to the transport operator that our unstabilised solve does not have; biases the current on any practical mesh |
| 2 | Flow element pair + stabilisation | P1+P1 with streamline + crosswind | Their pressure stabilisation is what makes equal-order legal; a Taylor-Hood P2/P1 solve is a different discretisation of the same PDE |
| 3 | Convective term form | Conservative form | Changes the assembled operator and the mass balance at coarse resolution |
| 4 | Discretisation orders | V quadratic, c quadratic, u/w/p linear | Directly sets the EDL resolution per element |
| 5 | Wall-distance definition | closest-point general extrusion onto the pore surface only, membrane excluded, Euclidean | Sets every wall correction; including the membrane would change EOF and current near the mouth |
| 6 | Permittivity parameters | epsr_ms = 30.08, epsr_alpha = 11.5, epsr0 = 78.15 |
See D3 — this is what was solved, whatever the ruling says |
| 7 | Concentration clamps | cdf argument clamped to [10⁻⁶ M, 5.3 M]; charge density clamped to c ≥ 0 |
Without them the fits extrapolate and Newton diverges at 5 M |
| 8 | ε values | pore 20, membrane 3.2 | Sets the fixed-charge screening |
| 9 | Mesh resolution | 0.05 nm max on the pore wall, growth 1.05, 0.1 nm in the pore domain, 0.001 nm global min, ~1.2 × 10⁵ triangles, min quality ≥ 0.64 | Comparable resolution, not identical elements |
| 10 | Nonlinear strategy | fully coupled, damping 0.2 → min 0.01, ≤ 100 iterations, rel. tol. 1e-6 | Same converged state; different paths are fine as long as the tolerance matches |
| 11 | Continuation | ramp V_bias, reuse previous solution |
Determines which branch is found at high bias / low salt |
| 12 | Current extraction | variational boundary flux, integration order 4 | Cross-section integrals of the CG flux disagree at the percent level |
| 13 | Steric flux weak form | as printed in C.6, α_i = k_i/(1 − Σ c_j v_j) |
Sign and prefactor as given; trap 2 of knowledge/00-orientation/02-the-five-things-most-likely-to-trip-you.md |