Goodenough-JB: Physicist review (Step 2)

Reviewer role: PhD physicist (mainstream view). Document reviewed: Braga, Grundish, Murchison, Goodenough, "Alternative strategy for a safe rechargeable battery", Energy Environ. Sci. (2016), doi:10.1039/c6ee02888h. I re-read the full text layer myself [read].

Source gap (repeated from Step 1): The three foundational papers (Braga 2014 J. Mater. Chem. A; Braga 2016 Energy Environ. Sci. 9, 948; Braga 2017 J. Electrochem. Soc. 164, A1) were not provided and not read. Conductivity, wetting, 0.06 eV barrier and dipole/ferroelectric claims are asserted in this paper by citation only. Raw data and figure values were not available (figures were not rendered; I used text and captions). Where I rely on outside criticism or replication, it is tagged [web] with a URL; anything else is [memory] and marked.

Corrections to Step 1

1. Claim 4 voltage wording. claims.md says the redox center sets "about 2.5 to 3.1 V". The paper gives Li-S Vmax about 2.65 V, VOC 2.61 V, 2.55 V under LED load, and 3.08 V for MnO2 [read]. Ferrocene voltage is not stated in the text I read. facts.json's "about 2.5 V" for sulfur is a little low. Minor.
2. Claim 3 "cell voltage and its drift are explained by two capacitors in series". Accurate, but add that the paper itself calls the typical initial VOC about 1.65 V (Fig. 5), which is lower than the 2.5 to 3.5 V working voltages. The paper explains the later rise to Vmax by a drop in Cc, not by any chemical change [read].
3. Claim 5 "tens to hundreds of cycles". The abstract says "over 250" cycles for MnO2; the figure caption for Fig. 7 says the envelope of 170 cycles [read]. The text conflict (250 vs 170) is internal to the paper and should be noted. Li-S 46 cycles was "continued beyond 46 cycles despite an imperfect seal" [read].
4. Claim 5 "90.1%". The paper does not give the capacity in mAh or the Li mass; "90.1% of the lithium anode capacity" is therefore not checkable from the text. An outside critic computes about 10.8 times the theoretical sulfur capacity [web: https://lacey.se/2017/03/29/on-the-skepticism-surrounding-the-goodenough-battery].
5. facts.json mainstream_conflict, item (2) and the [memory] note on skepticism. Now web-verified, see below: public skepticism exists, and one independent group (Graz) characterised the glass material and disputed its identity. Step 1's "other groups publicly questioned ... on thermodynamic grounds" is supported by Lacey's blog (a single critic, a materials scientist writing a blog, not a peer-reviewed rebuttal) [web].
6. facts.json relevance "capacitor-based explanation of cell voltage is a departure from conventional electrochemistry" is correct and understated: see claim 3 below, where the equation is shown to be dimensionally fine but physically unable to supply the claimed energy.

No other errors found. The summary of the paper's methods and figures matches the text.

Web checks done (budget about 6)

• Lacey blog (skeptic, a materials chemist): no net chemical change on discharge yet about 2.5 V claimed; no EDS spectra of Li; max 250 cycles; 10.8x sulfur capacity; composition formula changed between papers; no independent replication of the cells. https://lacey.se/2017/03/29/on-the-skepticism-surrounding-the-goodenough-battery [web]
• Graz University of Technology group (Hanghofer, Redhammer, Rohde, Hanzu, Senyshyn, Wilkening, Rettenwander), Chem. Mater. 30, 8134 (2018): synthesised "Li3OCl" and Li2OHCl; air-exposed Li3OCl degrades quickly to Li2CO3 and amorphous LiCl·xH2O. https://pubs.acs.org/doi/10.1021/acs.chemmater.8b02568 [web]. The search summary I saw also says these authors reproduced Braga's synthesis and found water-containing LiCl-like material rather than Ba-doped Li3OCl; that statement came from a secondary summary (InsideEVs/Lacey pages) and I did not read the Graz paper's own text on that point, so treat it as [web, secondary]. Context: https://insideevs.com/news/404967/latest-braga-goodenough-paper-controversy/
• Not checked: whether the glass or the plating cells were later replicated by anyone else. I found no replication; absence in a short search is not proof of absence [memory, low].

---

Claim 1: Li/Na-O-Cl glass with sigma > 1e-2 S/cm at 25 C, barrier 0.06 eV, window > 9 eV

a. Math consistent? The paper gives no derivation, only citations. Arrhenius conduction sigma ~ exp(-DHm/kT) is standard. A barrier of 0.06 eV is only about 2.3 kT at 300 K (kT = 0.0259 eV), i.e. nearly barrier-free; best-in-class crystalline superionic conductors are about 0.2 to 0.3 eV (Li10GeP2S12 about 0.25 eV, sigma about 1e-2 S/cm) [memory, high]. Not inconsistent, but extraordinary. The "energy-gap window > 9 eV" is a statement about electronic gap, but stability against Li metal depends on the electrolyte's reduction potential (thermodynamic, set by chemistry), not the band gap. A wide gap does not by itself prevent reduction by Li; Li3OCl (if present) would be thermodynamically stable against Li only if Li2O and LiCl are its reduction products, in which case a passivating layer of those still forms (a kind of SEI). That is a mainstream objection to "no SEI because wide gap" [memory, medium; the general rule is standard textbook electrochemistry].

b. Consistent with experiment? Conductivity above 1e-2 S/cm with a 0.06 eV barrier is not what is found in crystalline Li3OCl antiperovskites (about 1e-3 S/cm or lower at room temperature, barrier about 0.3 eV, [memory, medium]). The Graz result [web, above] suggests the measured conduction in "Braga glass" may come from hydrated LiCl-type amorphous phases, which can be highly conductive, so the number itself is plausible for some material; its identity is the problem. Without the 2014/2016 papers I cannot evaluate their impedance spectra (and impedance on a 0.06 mm membrane with unknown geometry/pressure is easy to misread).

c. Tested? Authors' own group only. One independent group characterised the material (Graz) and disputed the composition [web]. Conductivity of an independently made batch: not found by me. Dismissed without testing? No; at least one independent synthesis was attempted.

d. If true: Solid-state ionics would need a new mechanism for near-barrierless Li+ hopping in an amorphous halide-oxide; the usual correlation of barrier with lattice/bottleneck size would break. Not impossible; would be a major discovery.

Verdict: Never properly tested for the stated composition by a neutral party; independent characterisation (Graz) points to a different material, so the specific composition claim is Tested, not supported (weak, one group, composition only). The conductivity number itself: Plausible, untested as a property of whatever material it is. [Split verdict.] Rests partly on [web] secondary reporting.

---

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

a. Math consistent? No equations; qualitative. Internally the paper's reasoning "wetted, therefore no dendrites" is an assertion. Dendrites in solid electrolytes arise from current focusing at defects, filament growth in cracks and grain boundaries; wetting helps contact but does not exclude it. The authors anneal to "reform the solid glass without grain boundaries", which is relevant, but the evidence shown is one symmetric Li//Li cell impedance trace (Fig. 9, by caption: "negligible" impedance) [read].

b. Consistent with experiment? Good Li wetting of solid electrolytes does reduce interfacial resistance (e.g. Li on Li7La3Zr2O12 after surface treatment; [memory, medium]), so the idea is not absurd. But Fig. 9 reports short, low-current chronopotentiometry; dendrite-free cycling at practical current density (several mA/cm2) and capacity (several mAh/cm2) is the accepted criterion, and the paper has none of that. "No SEI" is also at odds with thermodynamics if the glass contains reducible species such as Cl-, O or OH- against Li (would form Li2O, LiCl, LiH). With the water/hydroxide content suggested by Graz [web], reduction by Li is expected (LiOH + 2Li -> Li2O + LiH) [memory, high on the chemistry].

c. Tested? Only by the authors in this paper, short symmetric test; no independent test found. The "no Li remaining on anode after discharge" SEM/EDS (Fig. 2) is stated to show no Li left; EDS cannot detect Li (it is below the detection range of standard EDS, atomic number 3) [memory, high; standard EDS limitation]; Lacey also notes no EDS spectra of lithium [web]. So the Fig. 2 evidence is visual and image-based.

d. If true: Would remove the main obstacle to Li-metal batteries (dendrites and SEI); solid-state electrolyte theory (Monroe-Newman shear-modulus criterion) would need an exception.

Verdict: Never properly tested (no neutral controlled test; not dismissed without testing, but critics cite lack of data). The strong form "no SEI at all" is Plausibly contradicted by thermodynamics for a halide/oxide/hydroxide glass, so I would not call it Plausible; keep as Never properly tested.

---

Claim 3: Ferroelectric condensates; huge permittivity; EDLCs at both interfaces; Vcell = Q/Ceq

a. Math consistent? Equations (1) and (2) are mathematically fine: two capacitors in series, Ceq = CA CC/(CA+CC); Vcell = Q/Ceq. The reasoning issues are physical: - A capacitor's voltage reflects stored charge. Here Q is declared "a constant determined by that needed to equilibrate the electrochemical potentials", but then Vcell is said to be set by Q/Ceq, while the Fermi-level picture (Fig. 4) sets V by energy differences (mA = 3.05 V, mC = 0.40 V vs SHE) [read]. These are two different accounts and the paper does not reconcile them. - For fixed Q, V = Q/Ceq with Ceq <= min(CA, CC). The paper's Vmax/Vmin argument (CC << CA gives large V; CA = CC gives Ceq = C/2) is algebraically fine, but the size of the voltage change depends on capacitance values the paper never gives, so the model has free parameters and can fit nearly any V(t).

Energy check (my calculation). A capacitor stores E = (1/2) Q V, and discharges only the charge it holds. The paper claims a Li-S cell delivering 90.1% of the Li anode capacity (Li: 3860 mAh/g [memory, high; standard value from F/M(Li) = 96485/(6.94 g/mol)/3.6]). Suppose the cell supplies the sulfur-normalised multiple reported by a critic: 10.8 x the sulfur capacity (1675 mAh/g) [web: lacey.se]. Sulfur mass 1.99 mg [read] gives 3.33 mAh for S, so about 36 mAh delivered = 130 C. At about 2.6 V this is E = QV = 130 C x 2.6 V = 338 J (about 0.094 Wh). To hold 130 C at 2.6 V a capacitor needs C = Q/V = 50 F. A coin-cell interface of area A of about 1.5 cm2 (assumed; not stated in the text) with a good double layer of about 20 microfarad per cm2 gives about 30 microfarad, i.e. 1.7 million times too small. Even a parallel-plate capacitor across the whole 0.06 mm membrane [read] would need a relative permittivity eps_r = C d/(eps0 A) = 50 x 6e-5 / (8.85e-12 x 1.5e-4) = about 2e12. Glasses and normal ferroelectrics: eps_r about 10 to 1e5 (BaTiO3 ceramic about 1e3 to 1e4; colossal-permittivity oxides up to about 1e5 to 1e6) [memory, medium]. So "capacitance" cannot be the energy source for hours-to-weeks discharge. Capacitors can explain a transient voltage offset, not 28 days of current at 2.6 V. (Assumptions: area and Li mass not given in the paper; the argument holds for any plausible values within factors of 100.)

b. Consistent with experiment? In standard electrochemistry, two electrodes of the same material (Li metal anode; Li metal plated onto Cu) at equilibrium give open-circuit voltage 0 V (Nernst, both are the Li/Li+ couple). The paper's "unusual" explanation is outside textbook theory. Underpotential deposition of Li on Cu can give at most tens to hundreds of mV for a monolayer, and is exhausted after roughly one monolayer's worth of charge (mAh/cm2 scale: about 0.1 mC/cm2 per monolayer, tiny) [memory, medium]. Lacey makes the same point [web]. Real batteries with a Li anode and an oxide/sulfur cathode get voltage from the cathode reaction being consumed (insertion or conversion), which the paper says does not happen.

c. Tested? The capacitor model has not been tested independently, as far as I can find. A controlled test would be trivial (see below).

d. If true: Would violate energy conservation unless a real source of free energy exists (none identified); in particular a cell that moves Li from one Li electrode to another Li electrode and outputs 2.5 V power would be a perpetual-motion machine of the second kind (it needs a heat reservoir; a cell that converts ambient heat to work breaks the second law). The paper's own SEM evidence (Li plated at the cathode) and capacity (many times what sulfur can supply) show energy output without net chemistry; the alternative is that the charge comes from irreversible side reactions (electrolyte or impurities, including water), which makes the cell a primary or poorly reversible cell, and the "reversible" cycling a partly parasitic one. That is mainstream electrochemistry applied correctly and requires no new physics.

Verdict (split): - Capacitor-series equation as a description of transient voltages: Plausible, untested (mathematically fine; only transient). - Capacitor voltage as the source of sustained energy and "Vcell explained by Q/Ceq rather than electrode potentials": Refuted by calculation (energy budget above: needs C of about 50 F and eps_r of about 2e12 versus about 1e-4 F and 1e4 available) and conflicts with Nernst for a Li-vs-Li cell. Show-stopper unless the capacity figures are wrong, which is itself a measurement issue. - Ferroelectric condensates / giant permittivity in amorphous glass: Never properly tested (dielectric data only in unread cited papers; Lacey says the characterisation methods give no composition information [web]).

---

Claim 4: Plating cell: capacity set by anode metal; redox center sets V (about 2.5-3.1 V), not consumed; V about 3.5 V without it

a. Math consistent? Qualitative band diagram (Fig. 4). Energy bookkeeping: if Li is removed at the anode (EF = Li) and Li is deposited on Cu at the cathode, net Gibbs energy change is the difference between Li on Li and Li on Cu: near zero, plus nucleation and interface energy (Li on Cu shows overpotentials of order tens of mV [memory, medium]). The paper's own words: "the Fermi level of the lithium plated on the carbon-copper composite ... is determined by the Fermi level of the cathode current collector" [read]. Once Li metal exists on the Cu, the cathode surface potential is that of Li metal, so the 3.0 V difference vanishes: the plated Li is itself a Li electrode. A "catalytic relay" that is not consumed cannot change the thermodynamic open-circuit voltage; catalysts change kinetics, not the free-energy difference (second law). So the claim is internally inconsistent with the paper's own statement that cell voltage and Fermi levels determine energy.

b. Consistent with experiment? Textbook Li||Cu cells (e.g. coulombic efficiency tests) plate Li on Cu at about 0 V vs Li/Li+, negative of any cell with an anode at Li potential: discharging a Li anode into a Cu cathode is not a power source. The observation that these cells produce 2.5 to 3.5 V at all is therefore either a real new phenomenon or an artefact (reaction with impurities, hydroxide, Cl-, electrode-electrolyte reactions at a higher potential such as oxidation of some component, the sulfur itself being reduced at about 2.0 to 2.4 V vs Li/Li+ [memory, high], or a measurement issue like high-impedance voltmeter reading of a charged capacitor). The key observed fact, capacity above that of the cathode material (e.g. 3x MnO2 on the 34th cycle; about 10x S) [read, web], is exactly what you see when something else is being consumed, e.g. electrolyte decomposition or impurity redox.

c. Tested? Authors only, and "complete reaction not reversible after full discharge" [read]. No independent replication found. No control cell (no redox center, Cu//Li without sulfur) with full capacity data was shown for the 28-day discharge; the text says "the redox center could be removed" and gives about 3.5 V with none, with no figure for it that I could read. Also no check for gas, no mass balance of Li deposited vs capacity delivered (Fig. 2 is photos and SEM, EDS cannot see Li).

d. If true: Thermodynamics of electrochemical cells (free energy = nFE, Nernst) would have to be rewritten: a cell with identical electrode material both sides and a catalyst yielding work. Realistically it would have to be a different, unreported chemistry.

Verdict: The thermodynamic form of the claim (Li stripped from Li anode and plated on Cu delivers 2.5-3.5 V of sustained work with an inert redox "center"): Refuted by calculation (Nernst: identical Li/Li+ couple on both sides gives about 0 V; a catalyst cannot change an equilibrium potential). The empirical observation of long, high-capacity discharge: Never properly tested (needs a neutral controlled test; not dismissed, but not independently reproduced). Likely explanation per mainstream: undocumented side reactions.

---

Claim 5: Coin cells cycled tens to hundreds of times; 90.1% of Li capacity in 28 days; safe, low-cost, high-energy EV battery

a. Math consistent? Energy density quoted as 10.5 Wh per g of Li (8.5 above 2.34 V) [read]. Check: 3.86 Ah/g x about 2.7 V = 10.4 Wh/g, consistent with a full Li utilisation at about 2.7 V mean [calculation]. But the denominator is Li only. Cell-level energy density needs the whole cell mass (electrolyte, current collectors, Cu, case); the paper gives none [read, absent]. For comparison, the best Li-ion cells are about 0.25 to 0.30 Wh/g at cell level [memory, high]; Li-metal theoretical limit with an energy-dense cathode is about 0.4 to 0.5 Wh/g at cell level. A claim of 10.5 Wh/g (specific to the anode only) is not comparable. The abstract's "no lowering of the energy density" over 250 cycles vs the body's 170-cycle envelope is a mismatch [read].

b. Consistent with experiment? Cycling and LED demos (Fig. 8) show that the cells give some voltage and some current; this is consistent with a battery of some kind. The LED test shows a 2.55 V cell lighting an LED for some time; this does not distinguish a capacitor transient from a sustained source. "Safe" is not demonstrated: no abuse test (nail, short, overcharge, heating) in the paper, and metallic Li or Na next to a glass with Cl-/O species is not obviously safe [memory, medium on this last point]. Cycling at 40 mA/g and 25 mA/g, 10 to 20 hour half-cycles: far below EV rates (typical about C/3 to 1C, i.e. hundreds of mA/g). Seal imperfect for Li-S. Temperature range not shown.

c. Tested? Authors only. No independent replication of the cells found (as of Lacey 2017 [web]). The "1,200 cycles" press figure is not in this paper; I did not verify it and do not rely on it [web, secondary via Lacey].

d. If true: Would be a step change in EV batteries; mainstream battery science would need to accept a plating-cathode energy source, which, per claims 3 and 4, it cannot at present.

Verdict: The narrow observation (these coin cells gave voltage and many cycles at slow rate): Never properly tested (not neutral; slow rates; unreplicated). "Suitable for an EV; safe; high energy density": Not testable as stated at present (no cell-level data) and in the strong form follows claims 3-4, so Tested, not supported would be wrong; I classify it as Never properly tested, with the mechanism claimed Refuted by calculation (see claim 4).

---

Summary table

What single test would settle it

A neutral lab builds a Cu//Li cell with the glass, with a cell-level mass balance, then (i) runs the full 28-day discharge and weighs/titrates the deposited Li (for example by gasometry of H2 after water reaction, or by an independent titration method; not EDS), (ii) repeats with a rigorously dry, characterised electrolyte (XRD, NMR, Karl Fischer water) and a control with no redox center and one with an inert Cu//Cu dummy, (iii) measures electrolyte composition after discharge. If the delivered charge exceeds the Li inventory deposited, or if the control cells deliver the same capacity, side reactions explain it.