Goodenough-JB: Mathematician review (Step 2)

Reviewer role: PhD mathematician (derivations, units, hidden assumptions).

Source gap (repeated from Step 1): Only one document exists: Braga, Grundish, Murchison, Goodenough, "Alternative strategy for a safe rechargeable battery", Energy Environ. Sci. (2016), doi:10.1039/c6ee02888h. I read its full text layer [read] and rendered the Fig. 1 page [read: page image p.3]. The three papers that carry the conductivity, wetting and dipole claims (Braga 2014 J. Mater. Chem. A 2, 5470; Braga 2016 EES 9, 948; Braga 2017 J. Electrochem. Soc. 164, A1) were not provided and not read. Claims 1-2 and the dipole part of 3 therefore cannot be checked mathematically here. Raw data are absent. Conclusions about the voltage model and energy budget (claims 3-5) rest on the paper's own text and equations plus arithmetic below, not on [memory].

Corrections to Step 1

1. facts.json says Fig. 3 cycled "10 h each way at 40 mA/g". The paper's text says "40 mA h g-1 for 10 h each way" while the Fig. 3 caption says discharged at 40 mA/g and "charged at constant voltage". The units are inconsistent inside the paper (mA h/g is not a current), and charge was potentiostatic, not galvanostatic [read]. Not a Step 1 error as such, but the summary should not call both legs constant current.
2. facts.json (mainstream_conflict, point 4) and claims.md say capacity beyond sulfur "could equally be explained by side reactions". That is fair, but Step 1 missed the sharper internal point: the paper itself gives chemical potentials mu_A = 3.05 V and mu_C = 0.40 V vs SHE and Vmax = 2.65 V, and 3.05 - 0.40 = 2.65. The observed maximum voltage is exactly the ordinary Li/S electrode-potential difference, which is the conventional explanation, not a capacitor effect. (Fig. 1 itself shows 2.68 V peak [read: page image p.3].)
3. Step 1 reports "typical initial VOC about 1.65 V" (correct, [read]) but did not note that Fig. 8b reports a Li-S VOC of 2.61 V, and the text calls 1.65 V "typical of a well-aged cell". Those two numbers are in tension with no explanation given.
4. claims.md claim 4 gives "about 2.5 to 3.1 V". The paper gives S about 2.65 V (Vmax), 2.55-2.61 V in Fig. 8b, and 3.08 V for MnO2; ferrocene voltage is not stated in the text I read. Treat "2.5" for ferrocene as unsupported.
5. Step 1 correctly flagged that the abstract's ">250 cycles" for Li-MnO2 conflicts with the Fig. 7 caption's "170 cycles" [read]. Confirmed; I add that the 46-cycle Li-S and 200-cycle Na cells have no capacity-retention numbers anywhere, so "no loss of energy density" is asserted, not shown.
6. Step 1's mainstream_conflict (2) is right in spirit; this review makes it quantitative (below).

Key quantities I derived (units checked)

Symbols: F = Faraday constant 96485 C/mol; R = gas constant; T = 298 K; RT/F = 25.7 mV; a = activity of Li+ (effective concentration); mu = chemical potential; Q = charge (coulombs); C = capacitance (farads); V = volts.

A. Same metal at both electrodes. The paper says Li metal is at the anode and Li metal is plated on the cathode collector (Fig. 2, Fig. 4) [read]. The cell reaction is then Li(anode) -> Li(cathode), same element, same phase, same electrolyte. The standard free-energy change is zero, so the equilibrium cell voltage is E = (RT/F) ln(a_anode/a_cathode) ~ 0 (Nernst equation: voltage from a concentration or activity difference). To get E = 2.65 V you need ln(ratio) = 2.65/0.0257 = 103, i.e. a ratio of 10^44.8 (check: 2.65/0.05916 = 44.8 decades). No such activity difference exists between two pieces of lithium metal. If the cathode surface holds Li0 at 0 V vs Li+/Li, the cell reads 0 V, and a reading of 2.65 V means the cathode sits at +2.65 V vs Li+/Li, where Li+ cannot be reduced to Li0 at all (that needs 0 V or below).

B. Energy budget of Fig. 1. Fig. 1b: about 2.7 V average, 90.1% of the Li anode capacity [read: p.3]. Energy per gram of Li = theoretical capacity x average V = 3.86 Ah/g x ~2.72 V = 10.5 Wh/g, which reproduces the paper's 10.5 Wh/g. So the number is internally consistent arithmetic. It is the energy of an ordinary Li -> Li+ at 2.7 V couple. But moving Li0 to Li0 releases no chemical energy (point A). The energy must come from somewhere: the sulfur (1.99 mg x 1675 mAh/g = 3.3 mAh, only about a tenth of the delivered capacity, by the paper's own Fig. 1a, ~1050% of S capacity), side reactions with the electrolyte, or stored electrostatic energy.

C. Capacitor model, quantitatively (assumptions flagged). Fig. 1 reads ~1050% of S capacity; 1.99 mg S -> 3.33 mAh S -> ~35 mAh delivered -> Q ~ 125 C (this implies roughly 10 mg of Li, which is my inference, not stated in the paper). Take V ~ 2.65 V, so the equivalent capacitance needed is C = Q/V ~ 47 F. For a plausible coin-cell area A ~ 1.5 cm^2 (assumed) and membrane thickness d = 0.06 mm [read], parallel-plate C = eps_r eps_0 A / d gives eps_r = C d / (eps_0 A) = 47 x 6e-5 / (8.85e-12 x 1.5e-4) ~ 2e12. Real double layers have nanometre gaps, which is why supercapacitors reach tens of F/g, but then the model needs the EDL area, not the bulk dielectric constant, and the paper gives no capacitance measurement. Energy test: a capacitor at constant Q has stored energy (1/2) Q V = 0.5 x 125 x 2.65 ~ 166 J. Energy delivered is about Q x V_avg ~ 125 x 2.4 ~ 300 J (about 0.084 Wh). The delivered energy exceeds the electrostatic maximum by about 1.8x, even if the two capacitors held all of it. And a capacitor discharging obeys V = Q/C falling linearly to zero as Q is drawn off, not a flat 2.7 V plateau held while Q (assumed "constant") passes through the external circuit. The assumption "Q constant" and "current of 125 C flows" contradict each other.

D. Eq. (1)-(2) in isolation. Vcell = Q/Ceq with Ceq = CA CC/(CA + CC) is valid algebra for two capacitors in series that share one charge Q [checked]. Limits check: if CC << CA then Ceq ~ CC and V = Q/CC is large (Vmax); if CA = CC then Ceq = C/2 and V = 2Q/C; as CC grows toward CA, Ceq changes and V changes accordingly. The series algebra is fine. The problems are elsewhere: (i) "Q is a constant determined by that needed to equilibrate the electrochemical potentials" is not derived; equilibration of two different electrodes through an electrolyte is set by their potential difference, not by a fixed charge, and in the standard treatment the electrode potentials themselves, not a charge, are fixed. (ii) Q is not constant during open-circuit ageing (Fig. 5) if capacitance changes, unless the insulating-layer story conserves charge; none of this is calculated. (iii) A series pair of capacitors across a cell stores energy but cannot be a source of steady current without a Faradaic reaction.

E. Fig. 5 argument. Voc rises then falls, explained by CC first decreasing (insulating layer) then increasing (dipoles). With Q constant, V = Q/Ceq; the claim is unfalsifiable as given, since either direction of voltage change is assigned to a capacitance change with no capacitance measured. The same phenomenon is generic to interface evolution in any cell.

F. Fermi-level statement. The text says the Li plated on the collector has the Fermi level of the collector while the voltage is set by the redox couple of an unreduced redox center. A redox center that is neither reduced nor oxidised does not contribute a thermodynamic potential in the Nernst sense; a potential belongs to a couple that actually exchanges electrons. And the S8 center is later said to be reduced below 2.34 V, which means it does act as an electrode material. No equation connects the redox energy to Vcell.

Claim 1: Li/Na glass, sigma > 1e-2 S/cm, DeltaHm 0.06 eV, window > 9 eV

a. Mathematics: Only quoted values, no derivation here [read]. Arrhenius: sigma T = A exp(-DeltaH/kT) with DeltaH = 0.06 eV gives a conductivity change of exp(0.06/0.0257 x ...) ~ 10x between 25 C and about -30 C only, so "promises acceptable operation at lower temperatures" is arithmetically reasonable if the barrier is real. Window "Eg > 9 eV" is a band-gap, not an electrochemical window; an Li-O-Cl glass cannot have a thermodynamic window of 9 V against Li metal, since Li metal at 0 V vs Li+/Li would reduce it if the glass contains reducible species. A band gap sets a limit on what electron transfer is possible; it does not guarantee kinetic stability. A conflation of two different quantities. b. Consistency with experiment: Cannot be checked from this document; cited papers not read. [memory, medium] Li3ClO-type antiperovskite conductivities are generally reported around 1e-3 S/cm or lower at room temperature in other labs; I did not web-verify this specific figure. Unverified here. c. Tests: The paper reports none directly, only symmetric-cell chronopotentiometry (Fig. 9, whose data I did not read in image form). No independent replication cited in the paper. [web] https://en.wikipedia.org/wiki/John_B._Goodenough states the work "remains controversial after several follow-up works"; I found no source for a successful independent replication of the conductivity. d. If true: a record solid electrolyte, room-temperature conductivity 10x above most sulfide benchmarks, would require a new conduction mechanism in amorphous glass. Verdict: Never properly tested (no neutral third-party test on record that I could find; not shown to be dismissed without testing). Rests on citation only; cited works not read.

Claim 2: wetted by Li/Na, no SEI, dendrite-free plating/stripping

a. Mathematics: Not a mathematical claim. Hidden assumption: a Li-O-Cl glass is thermodynamically stable against Li metal. Chloride and oxide glasses containing O and Cl are not obviously stable against Li; formation of LiCl/Li2O interphases would be expected. The claim "no SEI" is not shown, only that Li was not visible by SEM/EDS at the anode (Fig. 2) [read]. b. Consistency with experiment: Fig. 9 symmetric cell shows low impedance (text only) [read]; Li dendrite penetration in ceramics at high current density is a widely observed phenomenon [memory, high], but a paper that reports no current density, cycle count or voltage in Fig. 9 text cannot support "dendrite-free" in a quantified way. Spring pressure is used [read], which also suppresses voids and shorts. c. Tests: authors' own symmetric cells only. No independent test found. d. If true: one of the central open problems of solid-state batteries (dendrites in solid electrolytes) would be solved; no change to basic theory, but large change to engineering consensus. Verdict: Plausible, untested in the weak form (low interface impedance in a lab cell); Never properly tested in the strong form (dendrite-free at EV-relevant current density and capacity).

Claim 3: ferroelectric condensates, huge permittivity, EDLCs; Vcell = Q/Ceq

a. Mathematics: Eqs (1)-(2) algebraically fine (D above), but the physical model is internally inconsistent: constant Q vs current passing (C), no stated capacitance, no charge balance, no definition of CA, CC in farads; units of "mA h g-1" used for a current (Fig. 3 text). Mixed use of Fermi level (EF = -mu) and chemical potential in Fig. 4 without sign convention beyond one sentence. The claimed dependence of capacitance on dipole ageing is qualitative only. b. Consistency: A plateau at constant ~2.7 V while 35 mAh is withdrawn (Fig. 1) contradicts V = Q/Ceq with fixed Q. No measured permittivity or EIS in this paper. Ferroelectric order in amorphous glass at room temperature is unusual; evidence is only in the unread ref. 3. c. Tests: none independent. Dielectric spectroscopy would be a simple test (ref. 3 may contain some; unread). d. If true: that a capacitive double layer can deliver sustained, plateau-voltage energy from 3.86 Ah/g of metal without a chemical driving force would violate energy conservation (first law); it would not be a revision of electrochemistry but of thermodynamics. The critique reported in [web] https://en.wikipedia.org/wiki/John_B._Goodenough is exactly this: that the mechanism "would violate the first law of thermodynamics". Verdict: Refuted by calculation (strong form: capacitor series model explains the cell voltage and energy; see B, C above, using the paper's own numbers; the ~1.8x shortfall and the Q-constant contradiction) and Internally inconsistent (Q constant versus Q flowing; mu_A - mu_C = Vmax coincidence ignored). Weak form (ageing changes interfacial impedance and hence open-circuit voltage readings): Plausible, untested.

Claim 4: plating cathode, redox center sets voltage unconsumed; 3.5 V with none

a. Mathematics: Thermodynamically inconsistent as stated (A above): Li0 on both sides gives zero equilibrium voltage. A cell with Cu/C at the "cathode" and no redox center cannot show 3.5 V from the quoted couple: the Li/Cu couple gives about 0 V (plating) and no Li+ -> Li0 reduction can happen at 3.5 V vs Li. Voltage of 3.5 V with no redox center exceeds even the highest-energy couple the paper identifies (MnO2, 3.08 V). b. Consistency with experiment: Known: Li plating on Cu in a Li||Cu cell occurs near 0 V vs Li+/Li [memory, high]; the paper also shows Li "plated" at 2.65 V per its own picture (Fig. 2d). The observed cathode-side Li-looking deposit could be a different phase (e.g. reduction product of the electrolyte or sulfur species, plus Au coating artefacts); SEM-EDS cannot detect Li (light element, EDS limited) [memory, high], so "Li plated" is not demonstrated by Fig. 2 [read: caption says "no evidence of Li left", i.e., absence of evidence, not presence]. c. Tests: authors only. No third-party measurement (reference electrode, XPS, titration) reported. d. If true: electrochemical potential would no longer be set by electrode couples; the Nernst equation and free-energy relation Delta G = -nFE would need replacement. Verdict: Refuted by calculation (strong form: Li metal plated at the cathode while a cell voltage of 2.5-3.5 V is delivered; Nernst and energy budget, A-C above). The weaker observation (capacity exceeding sulfur's) is Plausible, untested as an unrecognised side-reaction effect (electrolyte decomposition, etc.).

Claim 5: coin cells cycle tens to hundreds of times; 90.1% of Li capacity in 28 days; safe, low-cost, high energy density for EVs

a. Mathematics: Energy density 10.5 Wh/g_Li is per gram of Li only, not cell level; it ignores electrolyte, current collector, S and carbon, casing (47:43:10 mix is mostly non-lithium). Unit slip in text ("40 mA h g-1" as a current). Cycle numbers: 46, 170, 200, "over 250" differ between caption and abstract. "No lowering of energy density" has no data in the text I read. b. Consistency with experiment: Cycling a cell over a plateau of 2.5-3 V is not by itself inconsistent with experiment; a 2.65 V Li-S-like plateau with sulfur inactive is, per A. Alternative reading (my inference, not demonstrated): the capacity could come from reversible electrolyte or redox-center chemistry plus Li, and/or a leaking seal ("imperfect seal" admitted [read], which would admit air and moisture and allow other redox reactions). Charge capacity above the active material with an imperfect seal is a classic artefact pathway. c. Tests: authors' coin cells only; no rate data, no energy density of the whole cell, no temperature range, no safety test (nail, short, thermal) reported [read]. No independent replication found [web: Wikipedia, above]. d. If true: a single-lithium-cell energy density 3x or more over present Li-ion at the cell level would upend EV battery roadmaps. But the claim is not established at that level. Verdict: Never properly tested (the cycling and capacity are reported by the authors only; "safe" and "EV-suitable" parts are Not testable as stated since no cell-level metrics or safety tests are given).

Summary table

Source tags used

• [read] full PDF text layer and page image p.3 (Fig. 1); remaining figures not read as images.
• [web] https://en.wikipedia.org/wiki/John_B._Goodenough: the glass battery was criticised for incomplete data, spurious interpretation of data, and (per skeptics) violating the first law of thermodynamics; "remains controversial after several follow-up works". A second web search was refused (search budget exhausted), so the primary critique papers and any specific replication or failure-to-replicate studies were not checked. I make no claim that a specific replication was or was not done.
• [memory, high] Li plating on Cu occurs near 0 V vs Li+/Li; EDS is poor for Li; Li dendrites in ceramic electrolytes.
• [memory, medium, not web-verified] antiperovskite Li3ClO room-temperature conductivities reported elsewhere around 1e-3 S/cm or lower; not relied upon for any verdict.
• Arithmetic by hand: C (47 F, eps_r ~2e12) assumes an electrode area of 1.5 cm^2 and ~10 mg Li, both my inferences from Fig. 1; the verdict does not depend on them, because the energy check (300 J delivered vs 166 J maximum stored) and the Nernst argument use only the paper's own voltage and capacity.