OPEN WORKSTREAM — Scotchn/Astra WFC_Behave.sub cell model: state of the argument, conceded errors, and the four unresolved questions (handoff 2026-09-15)

HANDOFF RECORD. Scotchn is relaying between StanBot and Astra; this entry is the shared state so the thread can resume days or weeks later. Read with #3392 (edge-only differential drive) and #3393 (EIS reference pointer).

=== MODEL AS IT STANDS (WFC_Behave.sub, version "stable bounded kinetics") ===
Cfast=98pF, Cdl=1uF, Rion=20k, Rpar=1G, I0=1nA, Ilim=0.5A, b=80mV (e-fold scale, NOT per decade; per-decade = bln10 = 184mV), BbatEnable=0, exponent guard limit(v/b,-40,40) retained for behavioural equivalence.
Kinetics now evaluated in the log domain: z=ln|K|/L, u=exp(-|z|), r=eps/L, s=sign(K); two branches meeting in value and derivative. VERIFIED by me: both reduce algebraically to I = K/(1+sqrt(K^2+eps^2)/L). Max abs difference vs prior form 4.44e-16 A on a 0.5A scale = 2DBL_EPSILON, i.e. bit-identical. The Ik node (which materialised ~2.35e8 A into 1 ohm => 2.35e8 V) was REMOVED. This was the single best change made all session.

=== ESTABLISHED RESULT THAT SURVIVES EVERYTHING ===
The Ps lag is pure RC and independent of all chemical parameters. At a 2V ramp endpoint, Ps = 1.800 / 0.73575 / 0.096748 / 0.0099667 V for 10 / 100 / 1k / 10k V/s. All four match the closed-form ramp-into-RC response V_c(T) = r*[T - tau*(1 - exp(-T/tau))] with tau = Rion*Cdl = 20 ms to five significant figures. CONCLUSION: edge rate demonstrably controls interfacial potential, attenuating it ~180x at 10kV/s. This is robust. How much CURRENT that lag gates is undetermined.

=== ERRORS I CONCEDED (record them so they are not repeated) ===

1. Randles-Sevcik misapplied, TWICE. i_f = k*sqrt(v) is a DIFFUSION law where v is scan rate (V/s), not applied voltage. Astra correctly ruled it out for this model, which has no concentration/depletion/diffusion dynamics. The intercept of an i/sqrt(v) plot identifies nothing physical here. My "run a scan-rate sweep and read the intercept" was the wrong instrument.
2. OVERCLAIM ON THE 461mA vs 100uA SPLIT. I called it "the VIC thesis in circuit form". Astra's counter is correct and stronger than "it reduces the ratio": with a physical Ilim (~21uA, see below) the ORDERING REVERSES. Rough figures: 10 V/s -> Bgas capped ~21uA + Bdir 9.1uA + Cdl 10uA ~= 41uA; 10kV/s -> Cdl 99.5uA + Cfast 0.98uA ~= 110uA. Fast would draw MORE than slow. The 4600x factor was an artefact of a borrowed 0.5A ceiling.
3. Bookkeeping: 21uA is MY estimate, never adopted into the model. I spoke as though it were. It is not.
4. "Butler-Velmer" typo; it is Butler-VOLMER (Max Volmer).

=== ASTRA'S POINTS I ACCEPT IN FULL ===

• Bdir resting conductance 5.05uS = 198.0 kOhm. With Ith = 2mA and the largest fast-capacitor current only 0.98uA, the tanh blend never leaves 0.5 in any test run. Bdir behaves as a FIXED 198k SHUNT throughout. At 0.2V it contributes 1.01uA of a 1.023uA total, i.e. it IS the model at low voltage. Needs measured justification before any of it is interpreted as chemistry.
• OCV reflects a MIXED potential and does not identify a single reaction's thermodynamic equilibrium potential.
• Ps (two-terminal internal node) is not automatically an electrode overpotential relative to a specified reference.
• Empirical terminal model vs electrode-reaction model are different projects with different meanings for a fitted offset.
• One should not choose an equilibrium potential merely to force zero current at zero applied voltage. (My motive was partly cosmetic. Conceded.)

=== MY POSITION THAT I STILL DEFEND ===
1.229 V is not a fitted offset. It is dG0/nF = 237.13 kJ/mol / (2 * 96485 C/mol) = 1.2288 V, i.e. thermodynamics. The model currently reports Bgas = 268.1 uA at 1.0 V — a branch named for gas production evolving gas 230 mV BELOW the reversible potential for water splitting. That is not cosmetic; it is forbidden. Whatever else is unresolved, that number cannot stand.

=== PROPOSED RESOLUTION (my answer to Astra's "what interface?") ===
The defect is that ONE branch is doing duty for SEVERAL reactions. Split it:
(a) a low-threshold MIXED/CORROSION branch (oxygen reduction, Fe/Cr oxide film processes at stainless in aerated distilled water) with its own small I0 and NO 1.229V threshold — this is real measurable sub-threshold current, correctly having no water-splitting gate;
(b) a WATER-SPLITTING branch gated at 1.229 V.
Then sub-threshold current is honestly labelled corrosion rather than mislabelled as gas, and the gas branch obeys thermodynamics. Each has parameters a real polarisation curve can identify separately.
On the two-terminal objection: if Ps is the TOTAL interfacial drop across both electrodes in series, 1.229 V is the correct full-cell number (the reversible potential already sums both half-reactions). If Ps is meant to be ONE electrode, it is not, and the model needs two Ps nodes. The model should state which, explicitly.

=== ESTIMATE UNDER DISPUTE: Ilim ===
i_lim = nFD*C/delta with n=1, F=96485, D=9.3e-9 m2/s (H+), delta=100um, C=1e-7 mol/L = 1e-4 mol/m3 (water autoionisation) => 0.897 mA/m2. For ~0.0236 m2 tube pair (25mm x 300mm) => ~21 uA. Model carries 0.5 A, i.e. ~23,600x larger. In distilled water the mass-transport ceiling is MICROAMPS because there is nothing to transport. NOT ADOPTED — an estimate awaiting measurement. Note the coincidence worth testing: with equilibrium referenced to 1.229V, Tafel at Ps=2V gives Ik = 15.3 uA against a transport ceiling of ~21 uA — same order, from independent physics.

=== FOUR OPEN QUESTIONS ===
Q1. Which current does the Bdir sense filter actually watch? If Cfast only, the blend is structurally dead in all realistic operation and Bdir should be DECLARED a fixed shunt. If total current, the 10 V/s run at 461 mA should have tripped Ith=2mA and did not — that would be a bug. UNRESOLVED, needs a look at the .sub.
Q2. Is Ps the total two-electrode interfacial drop or a single electrode? Determines whether 1.229 V or a half-reaction potential applies, and whether the model needs two Ps nodes.
Q3. Should the kinetic branch be split into mixed-potential + water-splitting (my proposal) or kept as one empirical branch with no thermodynamic claim (Astra's implied alternative)? Decidable by measurement — see M3.
Q4. Real values of tau, Rion, Cdl, I0, Ilim. Everything above is conditional on these.

=== AGREED MEASUREMENT LIST (Astra's, plus my three additions) ===
M1. 0.1 V step response: RETAIN step amplitude, initial current, decay constant, final current, AND resting voltage — not just tau. Plus water conductivity and cell geometry at the same moment. Keep amplitude under 0.3V so kinetics stay asleep (Tafel does not reach 1uA until ~0.55V with present I0/b). Note the measurement sees TWO interfaces in series, so extracted C = C1C2/(C1+C2).
M2. EIS sweep 0.1 Hz - 100 kHz, <10 mV amplitude. Nyquist: HF real-axis intercept = Rion; semicircle diameter = R_ct; f at semicircle apex gives Cdl via f_max = 1/(2piR_ctCdl). A squashed/tilted arc means constant-phase element, i.e. the single-tau RC model is an approximation whose error must be stated. Suspected 8 Hz corner and the 2.51-3.16 kHz operating points must both be inside the window.
M3. SLOW POLARISATION CURVE, 0 to 2.5 V at ~1 mV/s, current on a LOG axis. A knee near 1.23 V confirms the two-branch structure by measurement rather than assertion. THIS IS THE EXPERIMENT THAT SETTLES Q3.
M4 (added). Resting voltage before AND after the run — drift indicates the mixed potential settling, and how fast.
M5 (added). OCV decay logged for 1-2 minutes after switch-off — free read on adsorbed species and on whether anything is stored.

=== NEXT MODEL EDITS ALREADY AGREED, IN ORDER ===

1. Combiner already made odd-symmetric / pole-free (done, verified bit-identical).
2. Then bidirectional Butler-Volmer driver — but ONLY after Q2 is answered, since it needs a justified equilibrium reference.
3. Then revisit Bdir per Q1. Do NOT swap in the bidirectional law before the combiner fix, or the old pole at Ik = -Ilim (reachable at eta = -b*ln(Ilim/I0) = -1.60 V) becomes live.

STATUS: Scotchn resuming in the morning with the physical cell cleaned and fresh distilled water. Astra continuing the model argument. This entry is the resume point.