Goodenough-JB: Arbiter (Step 3)

Document under review: Braga, Grundish, Murchison, Goodenough, "Alternative strategy for a safe rechargeable battery", Energy & Environmental Science (2016), doi:10.1039/c6ee02888h. Inputs reconciled: facts.json, claims.md, physicist.md, mathematician.md, contrarian.md.

Source gap (repeated from Step 1): Only this one paper was provided (read in full from its text layer; Fig. 1 page rendered by two reviewers). The works carrying the conductivity, wetting and dipole evidence were NOT provided and NOT read: Braga et al. 2014 (J. Mater. Chem. A), Braga et al. 2016 (Energy Environ. Sci. 9, 948) and Braga et al. 2017 (J. Electrochem. Soc. 164, A1). Raw data, figure values and primary critiques/replications were not available. Conclusions on claims 1, 2 and the dipole part of 3 therefore rest on citation-only assertions plus [web]/[memory] context. Conclusions on claims 3-5 (energy budget, Nernst argument, internal consistency) rest on the paper's own numbers and arithmetic, not on memory.

Where the reviewers differed, and how I settled it

1. Is the 10.5 Wh per g of Li consistent with Fig. 1? (Mathematician and physicist said yes; contrarian said no)

Recomputed. Faraday: 96485 / (6.94 g/mol x 3.6) = 3862 mAh/g Li. - Paper: 90.1% of Li capacity delivered, so delivered capacity = 0.901 x 3.862 = 3.48 Ah/g Li. - 10.5 Wh/g divided by 3.48 Ah/g = 3.02 V mean voltage required. - Fig. 1 peaks at about 2.68 V (text: Vmax about 2.65 V) and ends near 1.6 V [read, p.3 page image per contrarian and mathematician]. Mean about 2.4 V gives 3.48 x 2.4 = 8.4 Wh/g; even a flat 2.68 V gives 9.3 Wh/g. - The mathematician and physicist got 10.5 only by using the full 3.86 Ah/g (100% utilisation) at 2.72 V. That mean voltage exceeds the plot's maximum (2.68 V), and 10.5 / 2.4 V = 4.4 Ah/g exceeds the theoretical 3.86. So the contrarian is right: the headline figure does not match the paper's own plot under any reading (about 25% too high). I did not see the underlying data; I cannot say how the authors computed it.

2. Size of the capacitor mismatch (all three agree it is enormous; numbers differ slightly)

Recomputed with the paper's 1.99 mg sulfur [read] and the Fig. 1 excess of about 10.8x sulfur capacity (1675 mAh/g) [web: https://lacey.se/2017/03/29/on-the-skepticism-surrounding-the-goodenough-battery, which quotes about 35.9 mAh, and the mathematician and contrarian read about 1050-1070% off Fig. 1]. - S capacity = 1.99 mg x 1.675 mAh/mg = 3.33 mAh. Delivered = 10.8 x 3.33 = 35.9 mAh = 129 C. Implied Li mass = 35.9 / (0.901 x 3862) = 10.3 mg (my inference; plausible for a coin cell). - Equivalent capacitance at 2.65 V: C = Q/V = 129 / 2.65 = 49 F. - Energy delivered about 129 C x 2.4 V = 310 J (0.086 Wh). A capacitor holding charge Q at V stores only (1/2)QV = 171 J, so even the whole capacitor emptied once would supply only 55% of what was delivered. And a capacitor's voltage falls linearly as charge is drawn; Fig. 1 shows a plateau. - Real double layer: about 20 uF/cm2 [memory, high; typical range 10-100] on an assumed 1.5 cm2 (area not stated in the paper) is about 3e-5 F, about 1.6 million times too small; at 100 uF/cm2 still about 3e5 times too small. As a bulk parallel plate across the 0.06 mm membrane [read], eps_r = C d/(eps0 A) = 49 x 6e-5 / (8.85e-12 x 1.5e-4) = about 2e12, against 1e1 to 1e5 for known dielectrics [memory, medium]. The verdict does not depend on the assumed area (a factor of 100 either way changes nothing). - The reviewers' figures (47-50 F, 166 vs 300 J, "1.8x") agree with mine to rounding.

3. Is "Q constant" vs "charge flows" a contradiction? (Mathematician, contrarian: yes; physicist: only physically unreconciled)

The paper's equations (1)-(2) are correct algebra for two capacitors in series sharing one charge Q [read]. But the paper also says Q is "that needed to equilibrate the electrochemical potentials" and sets the electrode energies from Fermi levels (mu_A 3.05 V, mu_C 0.40 V vs SHE) [read]. If Q equilibrates fixed potentials, then V is the potential difference and independent of C; if Q is fixed, V changes with C but then the electrode potentials are not fixed. Both cannot hold. I side with the mathematician and contrarian: this is an internal inconsistency, not just an unreconciled gap. I also confirm the observation (mathematician, contrarian) that 3.05 - 0.40 = 2.65 V equals the reported Vmax, which is the ordinary Li/S electrode-potential difference and so is the conventional explanation, not evidence for the capacitor model.

4. Verdict on claim 4: "Refuted by calculation" (physicist, mathematician, contrarian-strong-form) versus the empirical observation

All three agree. I keep a split: the thermodynamic statement is refuted by an accepted law; the observation that the cells delivered much more charge than the sulfur can supply is a real unexplained result in the authors' hands and is not refuted. - Nernst check (mathematician): for the same couple on both sides (Li to Li), E = (RT/F) ln(a1/a2). To get 2.65 V needs ln ratio = 2.65/0.0257 = 103, i.e. a ratio of 10^44.8. No two pieces of Li metal differ that way. I reproduce this (2.65/0.05916 = 44.8 decades). - The refutation assumes the cathode-side deposit really is Li metal. The paper's Fig. 2 does not show that: EDS cannot detect Li (Z = 3) [memory, high; Lacey makes the same point, web above], and the caption reports only "no Li left" at the anode. So the strong form fails by calculation if the paper's own description (Li plated on the Cu/C collector) is accepted; if the deposit is something else, the claim as written is wrong and the cell is some other chemistry. - A not-consumed redox center cannot shift an equilibrium potential (catalysts change rates only). The mathematician adds that the paper later says S8 IS reduced below 2.34 V, so it acts as an electrode material. Accepted.

5. Claim 1 verdict: "Tested, not supported (composition)" (physicist) versus "Never properly tested" (mathematician, contrarian)

I side with "Never properly tested". The physicist's basis is a Graz group paper (Hanghofer et al., Chem. Mater. 30, 8134, 2018). I could not open it (HTTP 403 from pubs.acs.org) and the InsideEVs page returned 502, so the statement that they reproduced Braga's synthesis and found hydrated LiCl-type material is a secondary report that I have not verified. Even if true, it is a disagreement about material identity, not a controlled test of the claimed conductivity in the claimed glass. Lacey (blog, not peer reviewed) separately says composition formulas differ between papers and water removal may be insufficient [web: lacey.se link above]. So: tested by the authors only; identity disputed; no neutral conductivity/Arrhenius test I can cite.

6. Claim 2 verdict: "Never properly tested" (physicist) versus "Plausible, untested" (contrarian, weak form by mathematician)

Split, as the mathematician did: weak form (low interface impedance in a spring-loaded symmetric cell, Fig. 9 by caption) is Plausible, untested; strong form (no SEI, dendrite-free at practical current density and capacity) is Never properly tested. I agree with the physicist and mathematician that "no SEI because the gap is above 9 eV" is a category error: a band gap is not an electrochemical stability window, and with reducible species (Cl-, O, any OH-/water) the thermodynamics expects Li2O/LiCl/LiH interphases [memory, high on general chemistry]. The contrarian's own text slips here ("a chloride-oxide with Li+ should be stable against Li, which is possible"); I discount that.

7. Smaller factual corrections the reviewers agreed on

• Voltages: Li-S Vmax about 2.65 V (Fig. 1 shows 2.68 V peak), Li-S VOC 2.61 V and 2.55 V under LED load, Li-MnO2 3.08 V. [Superseded by QE revision: Fig. 6 gives the ferrocene voltages.] The ferrocene voltage is not stated in the text. claims.md's "2.5 to 3.1 V" is a rough summary; ferrocene "about 2.5 V" is unsupported [read].
• [Superseded by QE revision: the 1.65 V vs 2.61 V item is a cell-age difference, not a contradiction.] The paper contradicts itself: MnO2 cycles "over 250" (abstract) versus 170 (Fig. 7 caption). State both; do not average.
• Fig. 3: text says "40 mA h g-1 for 10 h" (units of capacity used for a current) while the caption says discharge at 40 mA/g and charge at constant voltage. Not both constant-current.
• "No loss of energy density" over cycles has no tabulated retention data for the 46-, 170- or 200-cycle cells. Li-S used a seal admitted to be imperfect. Cycling rates (25-40 mA/g, 10-20 h half-cycles) are far below EV rates. Lacey says only about 2.8% of the discharge capacity was actually cycled [web: lacey.se link above].
• No cell count, error bars, mass balance, abuse (safety) test, temperature range, power or cell-level energy density is reported [read; absence].
• The "1,200 cycles" figure is a press claim not in this paper; Lacey notes the paper shows at most 250 [web: lacey.se link above].

Per-claim reconciliation

Claim 1: Superionic Li/Na-O-Cl glass (above 1e-2 S/cm, 0.06 eV, window above 9 eV)

• Agree: nothing can be checked mathematically here; values are citations only. 0.06 eV is about 2.3 kT at 300 K (kT = 0.0259 eV), nearly barrier-free, versus about 0.25 eV for the best sulfide Li10GeP2S12 [memory, medium: not web-verified, I could not reach a source]. A 9 eV gap is not an electrochemical window. No independent replication found.
• Disagree: physicist (composition "Tested, not supported") vs the other two (Never properly tested). Resolved above (section 5).
• Stronger argument: the "Never properly tested" camp, because the contrary evidence is secondary and about identity.
• Settles it: an outside lab makes the glass from the stated recipe, documents the water content (Karl Fischer), composition (XRD/NMR) and measures impedance with blocking electrodes at several temperatures (Arrhenius plot).
• Verdict: Never properly tested. Composition/identity is disputed [web, secondary]. Not shown to be dismissed without testing. Rests on citation, [memory] and secondary [web].

Claim 2: Wetting, no SEI, dendrite-free plating/stripping

• Agree: only the authors' short symmetric-cell chronopotentiometry (Fig. 9, "negligible" impedance by caption); no critical current density; spring pressure helps; no independent test.
• Disagree: physicist/mathematician-strong vs contrarian on label. Resolved (section 6).
• Verdict (split): weak form (low interface impedance in a lab cell): Plausible, untested. Strong form (no SEI, dendrite-free at practical rates/capacity): Never properly tested. Not dismissed without testing (published in a mainstream journal; critics asked for data).

Claim 3: Ferroelectric condensates, giant permittivity, voltage from series EDLCs (Vcell = Q/Ceq)

• Agree: equations (1)-(2) are fine algebra; no capacitance or dielectric spectrum is given in this paper; the voltage drift in Fig. 5 is assigned to capacitance changes with no capacitance measured (unfalsifiable as written); the energy budget fails by a factor of a million or more.
• Disagree: physicist/mathematician "Refuted by calculation" vs contrarian "Internally inconsistent". Both are supported; the energy-budget refutation is stronger because it needs only the paper's own capacity and voltage.
• Verdict (split): (i) capacitor-series model as the source of cell voltage and sustained energy: Refuted by calculation (about 49 F needed versus about 3e-5 F available; 310 J delivered versus at most 171 J storable; plateau versus linear capacitor decay), and also Internally inconsistent (fixed Q versus fixed electrode potentials). (ii) Series-capacitor picture as a description of small transient open-circuit offsets: Plausible, untested. (iii) Ferroelectric condensates and giant permittivity in an amorphous glass: Never properly tested (evidence sits in the unread 2017 paper).

Claim 4: Plating cell; capacity set by anode metal; unconsumed redox center sets 2.5-3.1 V; 3.5 V with none

• Agree: Li stripped from Li and plated on Li-covered Cu is the same couple, about 0 V; Li plating on Cu is near 0 V vs Li/Li+ [memory, high]; 3.5 V with no redox center has no data in the text and exceeds even the MnO2 value (3.08 V); a catalyst cannot shift an equilibrium potential; EDS cannot show Li.
• Disagree: none on the verdict; only on emphasis (energy conservation vs Nernst). The Nernst route (10^44.8 activity ratio) is cleaner and I use it.
• Verdict (split): the thermodynamic claim that Li-to-Li plating with an inert redox center delivers 2.5-3.5 V of sustained work: Refuted by calculation (Nernst plus energy budget; sections 2 and 4). The observation of about 10.8x sulfur capacity in a 28-day discharge: Never properly tested (no mass balance, no neutral replication; not shown to be dismissed). Mainstream default explanation is unreported side reactions (electrolyte or impurity or water redox, carbon, the imperfect seal), which is a hypothesis, not a demonstrated finding.

Claim 5: Cycling (46/170-250/200 cycles), 90.1% of Li capacity in 28 days, safe, EV-ready

• Agree: authors only; slow rates; no cell-level metrics; no safety test; cycle-count contradiction.
• Disagree: whether 10.5 Wh/g is consistent. Resolved: it is not (section 1). The contrarian's added point also holds: the paper says the cell is "no longer reversible" after full discharge, so the 10.5 Wh/g includes irreversible capacity; the 8.5 Wh/g (above 2.34 V) is the more defensible reversible figure, and my own check on it (0.8 x 3.862 x 2.55 is about 7.9 Wh/g, from the contrarian) is a modest mismatch only. Both are per gram of Li, not of cell; best Li-ion cells are about 0.25-0.30 Wh/g at cell level [memory, high], so they are not comparable.
• Verdict (split): reported coin-cell cycling and capacity: Never properly tested. Headline 10.5 Wh/g: Internally inconsistent with the paper's own Fig. 1. "Safe, low-cost, EV-suitable": Not testable as stated (no cell-level data or abuse test). Mechanism behind the cells: see claim 4 (Refuted by calculation).

The contrarian's distinction, kept visible

On how it was received: the contrarian found no evidence of blocked publication or suppressed data; the paper appeared in a mainstream RSC journal and drew calculation-based skepticism. Wikipedia (secondary) says the work "was met with widespread skepticism ... and remains controversial after several follow-up works" and reports objections of insufficient data, spurious interpretation and a first-law problem, with no independent replication noted [web: https://en.wikipedia.org/wiki/John_B._Goodenough, as quoted by two reviewers; I did not re-fetch it]. So this is not a "dismissed without testing" case; it is a "published, unreplicated, thermodynamically challenged, mass-balance never done" case. The researcher's reputation (LiCoO2 cathode; Nobel Prize in Chemistry 2019 [memory, high]) is not evidence for any of the claims.

Web checks (budget 8; 4 calls made, 2 succeeded)

• Lacey blog: succeeded; supports 10.8x sulfur capacity, about 35.9 mAh, EDS limitation, no independent confirmation, 250 not 1,200 cycles, composition inconsistency, suspected water/hydroxide chemistry. It is a single critic's blog, not peer reviewed. [web] https://lacey.se/2017/03/29/on-the-skepticism-surrounding-the-goodenough-battery
• Graz paper (Chem. Mater. 30, 8134): HTTP 403, not read. InsideEVs: HTTP 502. Web search: session limit reached. The Graz identity claim stays [web, secondary, unverified by me].
• Not web-verified, still [memory]: Li10GeP2S12 barrier about 0.25 eV (medium); Li3OCl crystalline antiperovskite conductivity about 1e-3 S/cm or lower with barrier about 0.3 eV (medium); EDL capacitance 10-100 uF/cm2 (high); permittivity ranges of dielectrics (medium); Li theoretical capacity (high, recomputed from F). No verdict depends on the medium ones except the "extraordinary" framing in claim 1.

Corrected claim list

1. A Li/Na-O-Cl glass is claimed to conduct the alkali ion above 1e-2 S/cm at 25 C (text) while the paper's own Fig. 9 is annotated 5-10 mS/cm (0.06 eV barrier; gap above 9 eV in the text, "E_g > 8 V" in Fig. 4; and thereby a wide "window"). The values are asserted by citation and are not even mutually consistent within the paper [QE revision]; the gap is not an electrochemical stability window; the material's identity is disputed by at least one outside group (secondary report). Never properly tested.
2. The glass is claimed to be wetted by Li/Na and to give no SEI and dendrite-free plating. Evidence here is the authors' one symmetric-cell chronopotentiometry run (Fig. 9: +-3 mA/cm2, 1 h half-cycles, about 79 cycles, about 200 h, overpotential about 5-6 mV, ASR about 2 ohm cm2), with no impedance spectrum or post-mortem; the voltage drop after a 75 mV spike is equally consistent with a soft short. Weak form (low interface resistance in a lab cell): Plausible, untested (verdict unchanged, reasoning revised). Strong form (no SEI, dendrite-free): Never properly tested.
3. Ageing dipoles are claimed to form ferroelectric condensates with giant permittivity, and the cell voltage is claimed to follow Vcell = Q/Ceq. As an energy/voltage source: Refuted by calculation and Internally inconsistent (about 49 F needed). Giant permittivity: Never properly tested. Transient offsets: Plausible, untested.
4. Li is claimed to plate on a Cu/C collector during discharge with capacity set by the anode, and an unconsumed redox center (S8, ferrocene, MnO2) sets about 2.65-3.08 V (Li-S 2.50-2.68 V; Na-ferrocene discharge about 1.9-2.1 V and charge 2.4-2.6 V vs Na [Fig. 6, digitised]; Li-MnO2 about 3.05-3.15 V); about 3.5 V with none (no data in the text). The paper's stated mechanism (voltage set by a Fermi-level difference between the two collectors) is a contact potential, which equalises on connection and cannot drive sustained current. Thermodynamic form: Refuted by calculation. The excess-capacity observation (about 10.8x sulfur capacity): Never properly tested. The MnO2 "3x" is only about 1.6x the one-electron theoretical capacity and only on cycle 34.
5. Coin cells reportedly cycled 46 (Li-S), 200 (Na-ferrocene) and 170 (caption), "over 250" (abstract) or 270 (Fig. 7 label) cycles (Li-MnO2) at 25-40 mA/g in fixed-time protocols that cannot show capacity fade, delivering 90.1% of the Li capacity in one 28-day discharge. The 10.5 Wh/g (Li) figure is inconsistent with Fig. 1 (needs 3.02 V mean; plot gives about 2.4 V and at most 2.68 V, so about 8.4 Wh/g at most about 9.3). Cycling data: Never properly tested. Energy density: Internally inconsistent. EV suitability and safety: Not testable as stated.

What would settle it

1. Mass balance on a fully discharged Li-S cell (the 28-day case): quantify metallic Li at the cathode side by water reaction and H2 gasometry, 7Li NMR, or neutron depth profiling, and compare with 129 C (35.9 mAh). Settles claims 3-5 (are Li atoms really moved, or is something else consumed?).
2. A control cell: Li / glass / Cu/C with no redox center, and a Cu//Cu dummy, discharged with the same protocol. Settles the "3.5 V with no redox center" claim and the first-law question. Under Nernst, a true Li-to-Li cell reads 0 V.
3. Open-circuit voltage reference-electrode measurement of each electrode versus Li/Li+ during ageing, plus measured capacitance (impedance spectroscopy) at each ageing time. Settles whether Fig. 5 is capacitance or ordinary interface evolution.
4. Independent synthesis with a dry, characterised recipe (Karl Fischer water, XRD, NMR) and impedance with Arrhenius plot. Settles claim 1.
5. Critical current density and long cycling in Li/glass/Li at mA/cm2 and mAh/cm2, plus post-mortem imaging. Settles claim 2.
6. Cell-level energy density, rate and abuse tests from an outside lab. Settles claim 5.

Open questions

• Where does the roughly 36 mAh come from: electrolyte decomposition, hydroxide/water chemistry, carbon, leakage through the admitted imperfect seal, or something not yet identified?
• Are the Li-S 2.65 V plateau and the mu_A - mu_C = 2.65 V coincidence just the normal Li/S couple (sulfur reduced, contrary to the paper's "not consumed" statement)? Not settled: the figure does not prove it, and the coincidence is partly circular (see QE revision).
• (Resolved by Fig. 5: the 1.65 V vs 2.61 V difference is cell age.) New: why does the terminal voltage under discharge rise about 1 V ABOVE the preceding open-circuit voltage (about 1.65 V to 2.68 V)?
• New: what is reduced at about 2 V vs Na in the Na-ferrocene cell, when ferrocene's usual couple is Fc/Fc+ at about 3.1 V vs Na?
• What do the 2014, 2016 and 2017 papers actually show for conductivity, dielectric response and wetting (not read here)?
• Have any independent groups reproduced the cells (not found in this check; absence in a limited search is not proof of absence)?
• How was the 10.5 Wh/g computed?
• What did the Graz group actually measure (their paper was not readable here)?

QE revision

Inputs: qe.md (MINOR), deep/Alt-Strat-for-Safe-Rech-Battery-d78469.md (Opus read of Figs. 2-9 from page images; digitised values good to about +-2 % of full scale, derived numbers marked [calc] there). Where the deep read corrects QE's own first reading, the deep read is used (item 3 below).

Headline: no central verdict changes. Claims 3 and 4 remain Refuted by calculation; claim 5 energy density remains Internally inconsistent; the excess-capacity and cycling results remain Never properly tested. Reasoning changes on claims 1, 2, 4 and 5 as follows.

New evidence and corrections

1. Fig. 9 (claim 2). The arbiter's "short trace, no current density" was wrong. The symmetric cell ran +-3 mA/cm2 (labelled; area about 0.43 cm2 implied by 1.28 mA, inferred), 1 h half-cycles (3 mAh/cm2, about 14.5 um of Li per half-cycle [calc]), about 79 full cycles over about 160 h of a 200 h trace, overpotential about 5-6 mV, so ASR about 2 ohm cm2 [calc]. This is an authors-only, single-current-level test, with no critical-current sweep, impedance spectrum or post-mortem. After a +75 mV spike near 40 h the overpotential fell and stayed low and flat; the figure cannot tell better wetting from a soft short through a dendrite [memory, medium]. Verdict: weak form stays Plausible, untested (ASR is low and the authors did run a practical-level current), but now for this reason, not for lack of a current density. Strong form (no SEI, dendrite-free) stays Never properly tested. "What would settle it" item 5 is partly done by the authors at one level; an independent run with post-mortem is still needed.
2. Claim 1 internal inconsistency. Text: sigma > 1e-2 S/cm. Fig. 9 annotation: sigma_optimized = 5-10 mS/cm at 25 C, i.e. at or below the text. Fig. 4: "E_g > 8 V" (a unit slip) versus "> 9 eV" in the text. The Fig. 9 resistance (about 5-6 mV at 3 mA/cm2) is compatible with 5-10 mS/cm for a 0.06 mm membrane (thickness for that cell not given) [calc]. Verdict unchanged: Never properly tested.
3. Fig. 6 (claim 4, 5). The Na-ferrocene discharge is about 1.9-2.1 V and charge about 2.4-2.6 V vs Na at 40 mA/g of ferrocene. The "not stated" claim is withdrawn. QE's reading that these cycles were minutes long is NOT supported: the time axis is per cycle and each half-cycle is about 2 h (about 80 mAh/g, 56 % of ferrocene's 144 mAh/g one-electron capacity [calc]). The charge-discharge gap grows from about 0.15-0.4 V to about 0.7 V by cycle 200, a degradation sign the paper does not mention. Puzzle: "ferrocene was not oxidized" is not informative on discharge (a reduction); ferrocene's usual couple is Fc/Fc+ near 3.1 V vs Na [memory, medium], about 1 V above where this cell works. What is reduced at about 2 V is unexplained.
4. Fig. 5 (open question answered, claim 3). 1.65 V is reached after about 4 weeks on open circuit (fit gives about 726 h); 2.61 V (Fig. 8b) is a cell of different age. The "paper contradicts itself" item is withdrawn. The fits (V = V0 + (Vmax - V0)(1 - exp(-t/94 h)); decay V = (Vmax/2)(1 + exp(-t/346 h)), decay clock starting at 240 h, asymptote 1.325 V) have about as many free parameters as points and an assumed halving of voltage; this does not test the capacitor model. They also disagree with Fig. 4's own "C_C << C_A" and aged 1.6 V: matching 2.65 to 1.6 V needs C_C about 0.43 C_A initially [calc], not much smaller. This strengthens Internally inconsistent for claim 3.
5. New anomaly (Fig. 5 + Fig. 1). Fig. 1 starts near 1.65 V (last open-circuit reading about 1.68 V) and jumps to 2.68 V once discharge current flows, about 1 V above open circuit; an ordinary cell sags below its open-circuit voltage under load [memory, high]. The paper attributes it to current reducing C_C. Ordinary candidates (surface film broken down by current, contact or instrument effect) are hypotheses only. This is an open item, not a verdict change.
6. Signs and the "2.65 V coincidence" (claim 3). The paper gives mu_A = -3.05 V and mu_C = -0.40 V vs SHE (the figure plots E_F = -mu); earlier files dropped the minus signs. The difference, 2.65 V, is unchanged. mu_A matches the textbook Li+/Li (about -3.04 V vs SHE [memory, high]). But no method is given for mu_C, and Fig. 4 itself lists V_cell about 2.65 V for the fresh cell, so "3.05 - 0.40 = 2.65 = Vmax" is partly circular. I therefore soften my section-3 statement: the 2.65 V being "the ordinary Li/S couple" is a reasonable but unproven reading. Conventional Li-S discharge plateaus are about 2.3-2.4 V and 2.1 V [memory, high], so a 2.65-2.68 V start does not match them cleanly either. The Nernst and energy-budget refutations of the capacitor mechanism do not depend on this point.
7. Contact-potential rebuttal of the paper's stated mechanism (claims 3, 4). The paper says the voltage is "limited by the energy difference between the Fermi levels of the anode and the cathode current collector" and reads about 3.5 V with no redox center. A Fermi-level (work-function) difference between two metals is a Volta (contact) potential; once they are connected, electrons flow until the Fermi levels equalise and no sustained current or work results [memory, high; standard solid-state physics]. A battery's voltage needs a chemical reaction whose free energy is released; an unconsumed catalyst or collector cannot supply it. Together with the Nernst figure (10^44.8 activity ratio for Li to Li), this is the cleanest statement of why the "3.5 V, no redox center" claim fails.
8. Electrostatic storage admission (claim 3). The paper says the voltage was limited "to avoid electrostatic energy storage at a high voltage, as has been demonstrated can occur (ref. 2)" [read p.4]. The authors thus acknowledge capacitor-like energy storage as a separate effect; this bears on the energy-budget argument (it cannot supply 129 C) but is not evidence for it.
9. Excess capacity, second instance (claim 4/5). MnO2 "3x" depends on basis: about 2.7x against the inset's measured basis (about 185 mAh/g), 5x against the caption's C/4 basis (100 mAh/g), but only about 1.6x against the 308 mAh/g one-electron theoretical capacity [calc]. The routine 10 h cycles (about 250 mAh/g by the figure's labelling) are within MnO2's own capacity; only the single 20 h cycle 34 exceeds it. This is a much weaker anomaly than the 10.8x sulfur case and is not independent evidence for the plating mechanism. Cycles 35-270 take about 10.3 h each [calc], so the per-cycle charge in the long run is probably about half of 10 h at 25 mA/g (inference). "Commercial rates" in the abstract is contradicted by C/4 to C/12 at 10-20 h per cycle.
10. Fig. 3 (claim 5). Window 2.50-2.67 V (never near 2.34 V); 40 mA/g x 10 h = 400 mAh/g per discharge, about 24 % of the sulfur's theoretical capacity per cycle if "g" is g of S (not stated) [calc]. So the 46 cycles do not require excess capacity and do not test the "sulfur not reduced above 2.34 V" claim outside a 0.17 V window. Cycling is by fixed time and current, so capacity fade cannot appear; only voltage stability is shown. The charge step is described three ways (text, caption, in-figure). Cycle counts on record: 46; 200; 170, "over 250" and 270.
11. Fig. 2 (claim 4 evidence). No EDS spectrum, map or percentages appear in the figure. Photo (d) shows a dark, crusty deposit, not bright metallic Li; panel (f) shows a dark patch remaining on the anode side after Au coating; (g) has no scale bar or data bar; the Li/S assignment of (h) and (i) is by caption only. Handling and air exposure are not described. EDS cannot detect Li [memory, high], so "no Li on the anode by EDS" is uninformative and EDS cannot confirm Li plating either. There was no XRD, XPS, NMR, titration, control cell or deposit mass. The visual evidence for "Li plated on the cathode" is weak; the argument rests on the capacity inference.
12. Candidate sources of the excess charge (claim 4 open question). In addition to water/hydroxide, the membrane is glass in recycled paper (cellulose) or fiberglass, cast from an ethanol slurry and heated only above 130 C [read p.3]. Residual ethanol, water and cellulose -OH groups react with Li metal [memory, high], and carbon, the admitted imperfect seal and electrolyte decomposition remain candidates. Listed as hypotheses, not findings.

Web check on the comment and reply

Recalled [memory, medium]: a comment on this paper arguing a first-law problem (Steingart and Viswanathan, Energy Environ. Sci., about 2018) and a Braga/Goodenough reply. I tried to verify it: WebSearch refused (session web search limit reached) and a WebFetch of an RSC page returned HTTP 403 (and that URL was itself a guess). Check failed; the comment and reply remain [memory, medium] and unverified. Authors, year, DOI and arguments are not stated here. The reception paragraph still rests on Wikipedia and the Lacey blog (secondary). A reader should look up the comment and reply directly on pubs.rsc.org.

Changed verdicts

None changed. Table of what moved:

Corrections to earlier reviewer text, now withdrawn: "no current density" (claim 2); "ferrocene voltage not stated"; "1.65 V vs 2.61 V contradiction"; positive signs of mu_A and mu_C; "Fig. 1 starts at 2.68 V" (it starts near 1.65 V and jumps).