RankingsPrimerArbiterPhysicistMathematicianContrarianQEClaimsSources

Goodenough-JB: Science Contrarian review

Role: judge by experiment, separate "tested and failed" from "never properly tested" and "dismissed without testing", build the strongest honest case, and apply equal rigour to the researcher.

Source gap (repeated from Step 1): only one document was provided, Braga, Grundish, Murchison, Goodenough, "Alternative strategy for a safe rechargeable battery", Energy Environ. Sci. 2016, doi:10.1039/c6ee02888h. I read its text layer in full and rendered p.3 (Fig. 1) [read: page image p.3]. Refs 1-3 (the glass, conductivity and dipole papers) were not provided. Every statement about outside testing or critique rests on one web source (below) or on [memory]. My web budget was effectively exhausted: WebSearch refused (session limit), RSC page returned 403, only Wikipedia was fetched.

Web facts: [web] https://en.wikipedia.org/wiki/John_B._Goodenough says the 2017 glass-battery work "was met with widespread skepticism by the battery research community and remains controversial after several follow-up works", citing lack of comprehensive data, spurious interpretation of data, and that the proposed mechanism would violate the first law of thermodynamics; no independent replication is noted. That is a secondary summary; I did not see the primary critiques. Names of specific critics and their papers are [memory, low] and I do not rely on them. Check failed for: specific critique papers and any independent measurement of the glass conductivity.

Corrections to Step 1

  1. facts.json says figures were not rendered. I rendered Fig. 1. The discharge starts at 2.68 V (text says "Vmax ~ 2.65 V"), falls to 2.34 V at roughly 80% of Li capacity, and ends near 1.6 V at 90.1%. Rate is 30 mA/g [read: page image p.3]. The "28 days" and the sulfur axis (about 1070% of sulfur capacity) are consistent with each other (30 mA/g x 672 h = 20 Ah/g, about 1200% of 1675 mAh/g).
  2. Step 1 reports the energy density 10.5 Wh/g(Li) without checking it. It is inconsistent with the paper's own Fig. 1 (see claim 5).
  3. The Li-MnO2 cycle count: the abstract says "over 250 cycles", the Fig. 7 caption says the envelope of 170 cycles. Step 1 gave "170-250"; the paper contradicts itself and the discrepancy should be stated, not averaged. [read]
  4. Fig. 3 caption says charge was at constant voltage, while the text says 40 mAh/g for 10 h each way (constant current). Step 1 reported only the text version. [read]
  5. Step 1 key_claims says Li-S "about 2.5 V". The paper gives 2.68 V initial, 2.61 V open-circuit in Fig. 8b, but also "typical VOC 1.65 V" in Fig. 5 for a well-aged cell. These are different cells/states and the paper does not reconcile them. [read]
  6. Step 1 mainstream_conflict point (2) says a Li||Cu cell "normally sits near 0 V vs Li/Li+ at the plating step". That is correct and is the crux; it is not merely "a mechanism not given", it is a thermodynamic problem (claim 4).
  7. Step 1 tagged the "public skepticism" memory as unverified; it is now [web] (Wikipedia, above) at secondary level.

Claim 1: superionic glass (>1e-2 S/cm at 25 C, 0.06 eV, gap >9 eV)

a. Math: Nothing to check beyond Arrhenius form; the paper only asserts values via refs 1-2. A 9 eV gap is not credible as a thermodynamic stability window for a chloride-oxide glass in contact with Li metal: the usable window is set by decomposition (reduction of Li+ is the limit, oxide/chloride oxidation at the top), not by an optical-style band gap. The paper conflates the two. [memory, medium: standard solid-electrolyte interface-stability argument] b. Experiment: Known best solid electrolytes (sulfides such as Li10GeP2S12, about 1e-2 S/cm) reach this value but with barriers near 0.2-0.25 eV [memory, medium]. 0.06 eV with >1e-2 S/cm at room temperature in an oxychloride glass would be exceptional. Not contradicted by a specific experiment I can name; no raw impedance data is shown in this paper. c. Tested: By the authors' group only, in refs 1-3. [web] Wikipedia: no successful independent replication noted. Dismissed without testing? Unknown; I cannot show it was ignored. Status: tested by the claimants, not independently replicated as far as I can find. d. If true: a new class of solid electrolyte; would be a major advance but fits existing ion-transport theory (low barrier, glassy percolation). Little in physics must change.

Verdict: Never properly tested (no independent neutral test on record that I could find; the source gap prevents me from judging the primary data).

Strongest honest case: conductivity is a measurement that other labs can easily repeat; a low-barrier amorphous conductor is not forbidden. Against: the paper offers no impedance spectra, and the stability-window argument is wrong in kind.

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

a. Math: none. Qualitative. b. Experiment: Fig. 9 shows chronopotentiometry of a symmetric Li/glass/Li cell with "negligible" interface impedance [read]. A symmetric cell for a limited time does not show dendrite suppression at high current density or high capacity; no critical current density is given. Claim of "no SEI" conflicts with the usual thermodynamics: a chloride-oxide with Li+ should be stable against Li, which is possible, so an SEI-free interface is not impossible. Low interfacial impedance in a symmetric cell with a spring-loaded coin cell is plausible with soft Li foil. c. Tested: authors only (refs 2 and this paper). No cycling-to-failure, no post-mortem dendrite imaging beyond the SEM in Fig. 2. Not independently replicated per [web]. d. If true: modest; interface wetting is an active research topic.

Verdict: Plausible, untested (as an independent matter). The paper's strongest-supported piece.

Claim 3: dipole condensates, giant permittivity, voltage = Q/Ceq

a. Math: Eq. (1)-(2): Vcell = Q/Ceq, Ceq = CA CC/(CA+CC), "Q constant". For two capacitors in series sharing charge Q, the voltage across the pair is Q/Ceq, true as algebra. But then the text says: if CC << CA, Ceq ~ CC and V = Vmax; if CA = CC, Ceq = C/2 and V = Vmin. That is backwards or at least inconsistent with the "Q fixed" premise: Q/Ceq is largest when Ceq is smallest, fine, so Vmax occurs for CC << CA; the logic is consistent algebraically. The problem is the premise: Q is said to be "that needed to equilibrate the electrochemical potentials of the two electrodes", which would make Q depend on the electrode potential difference (Q = Ceq x delta-mu/e) and then V = delta-mu/e, a constant independent of C. The paper uses Q fixed and V varying with C, while also fixing the electrode potentials; both cannot hold. Also the EDL voltage measured across the cell is an electrode-potential difference; EDL capacitors do not set the cell emf, they only store a small charge in the interface. Also the numbers: mu_A = 3.05 V and mu_C = 0.40 V for sulfur gives a difference of 2.65 V, which equals the observed Vmax. That looks like an inserted fit, not an independent prediction. A Q = C V charge at say 100 uF/cm2 and 2 V is of order 0.2 mC/cm2 compared with the discharged charge of tens of mAh (about 100 C): the capacitor model cannot hold the capacity delivered. The charge delivered (about 36 mAh, about 130 C) is about 10^5 times what any double layer can hold. [memory, high: typical EDL capacity ~10-100 uF/cm2] b. Experiment: The VOC drift in Fig. 5 (rise then fall) is real data in their cells; nothing in the paper excludes ordinary explanations (interface layer growth, redox of impurities, ageing). No independent dielectric spectroscopy shown. c. Tested: no. d. If true: electrochemistry's relation between emf and electrochemical potentials of the electrodes would have to change; ferroelectric condensates in an amorphous ionic glass would be new solid-state physics.

Verdict: Internally inconsistent (the fixed-Q premise versus fixed electrode potentials; and a capacitor cannot account for the delivered charge, a calculation not a test). The dipole/permittivity part alone: Never properly tested.

Claim 4: plating cell; capacity set by anode metal; redox center sets V without being consumed; V ~3.5 V with none

a. Math/thermodynamics: If Li metal is plated on the cathode collector, the cathode half-cell is Li+/Li at 0 V vs Li/Li+, identical to the anode. Cell emf would be zero. Energy delivered would then come from nowhere, the first-law objection in [web] Wikipedia. The paper's argument is that the Fermi level of plated Li on the collector is "determined by the collector" and the voltage by a redox couple that is not reduced. A redox couple that is neither reduced nor oxidised cannot supply or store energy; a catalyst can speed a reaction but not shift the equilibrium potential. The claimed ~3.5 V with no redox center exceeds the 3.05 V difference between Li and SHE that the paper itself lists (3.05 V is the entire span to SHE), so it would mean a cathode at about +0.45 V vs SHE with Li deposited there: not Li deposition thermodynamics. b. Experiment: Fig. 2 EDS is offered as proof. Li (Z=3) is essentially invisible to standard EDS, and the paper's own caption says EDS "gave no evidence of Li left" on the anode, which is a negative that EDS cannot support for Li in either direction. Visual photos of "plated Li" on a Cu/carbon/sulfur/glass composite do not identify the phase (Li metal vs Li2S, Li2O, Li-carbon intercalate, electrolyte decomposition products). Alternative explanations for large capacity and 2.4-2.7 V plateaus include irreversible reduction of the glass or of carbon, sulfur conversion with the electrolyte acting as a Li source, or Li-ion intercalation in carbon; none is excluded. The sulfur axis shows capacity 10x the S capacity, so a 10-fold excess charge must come from something other than the sulfur; the claim that it is Li plating is the least-tested hypothesis. c. Tested: authors only. Not replicated independently per [web]; critics argue first-law violation, secondary source only. d. If true: either a free-energy source (violating the first law) or a different, unspecified chemistry. Under the paper's mechanism, the Nernst relation between electrode potentials and emf would have to be dropped.

Verdict: Refuted by calculation in its strong form (Li on both electrodes gives emf zero; a delivered energy of about 8-10 Wh/g Li moved with no net reaction violates energy conservation). Weak form (the cell does deliver capacity, via some undetermined chemistry involving the glass, carbon, and sulfur): Never properly tested, because no one has done a controlled mass-balance/post-mortem (titration of Li metal, XRD/NMR/XPS depth profile, gas analysis, isotope tracing). I cannot say it was dismissed without testing.

Calculation (units): sulfur 1.99 mg x 1675 mAh/g = 3.3 mAh; Fig. 1a shows about 1070% of sulfur capacity = about 36 mAh; at 3861 mAh/g Li and 90.1% this implies a Li anode of about 10 mg, consistent with a coin cell. Energy from the figure: 36 mAh x average about 2.4 V = about 86 mWh. Source of this energy under the plating mechanism: none, because the end state (Li on cathode side, no Li at anode) differs from the start state only by Li position, with equal chemical potential.

Claim 5: coin-cell cycling, 90.1% Li capacity, EV-suited battery

a. Math: Check of the energy density: 90.1% x 3.861 Ah/g = 3.479 Ah/g(Li). The stated 10.5 Wh/g (Li) then requires a mean voltage of 10.5/3.479 = 3.02 V. Fig. 1 never exceeds 2.68 V and averages roughly 2.4 V, giving about 8.3 Wh/g; even holding V at 2.68 V throughout gives 9.3 Wh/g. The paper's headline number is inflated by about 25% relative to its own figure. [read: page image p.3, plus arithmetic] The 8.5 Wh/g for V > 2.34 V is also close to the whole-curve figure, which is consistent with the first-guess only if the curve above 2.34 V holds about 80% of the capacity; I get 0.8 x 3.861 x 2.55 = 7.9 Wh/g. Modest mismatch. Further, "per gram Li" is not cell energy density: cathode composite (47:43:10 sulfur, glass, carbon), 0.06 mm electrolyte, current collectors and cell can are excluded; the paper says cell-level energy density, power, and temperature range were not reported. b. Experiment: Li-S coin cells with an unreported rate at 30 mA/g, 28-day discharge: about 1 mA/g-scale rates of a few hours per day have no relevance to vehicle power (EV needs of order 1C). Cycling: Li-S 46 cycles with an admitted imperfect seal; Na-ferrocene 200; Li-MnO2 170 vs 250 stated; abstract's "no lowering of energy density" is asserted, not tabulated. The cell was "no longer reversible" after the full discharge [read], which contradicts using the 10.5 Wh/g figure (which includes the irreversible region) as the cell's energy density; the usable reversible figure is the 8.5 Wh/g one. c. Tested: single lab, a handful of coin cells, no error bars, no number of cells reported. Replicated: no independent report found [web]. d. If true: practical solid-state Li-metal batteries; would not require new physics beyond claim 4's problem.

Verdict: Split. Reported coin-cell data themselves: Never properly tested (no independent test, no cell count, no mass balance). Headline energy density 10.5 Wh/g: Internally inconsistent with Fig. 1 (above). EV suitability: Not testable as stated (no cell-level metrics reported; "suitable" is advocacy).

Contrarian assessment: how the claim was received

Honest case for the claimants: a respected senior chemist with a record (LiCoO2 cathode, Nobel 2019 [memory, high]) proposes an unusual cell; coin cells that light LEDs for weeks are cheap to build; the Li-S capacity exceeding sulfur by 10x is an anomaly worth a real investigation, and a mass-balance experiment would settle it quickly. Critics' objections as summarised on Wikipedia were about data deficiency and thermodynamics, not about conduct; I found no evidence of blocked publication or suppressed data, so this does not fit the "dismissed without testing" pattern: the claim was published in a mainstream journal and was met with calculation-based scepticism.

Equal rigour on the researcher: the 10.5 Wh/g figure disagrees with its own plot; EDS is used to "prove" the presence or absence of Li; the 170 vs 250 cycle discrepancy; the capacitor model fixes both Q and electrode potentials; the 9 eV window conflates optical gap with electrochemical stability; the abstract's "commercial rates" is not supported by 30 mA/g and 25 mA/g tests; the 3.5 V no-redox-center value is stated with no data. The reputation of the lead author is not evidence for any of these.

Single decisive experiments

  1. Mass balance on a fully discharged Li-S cell: quantify metallic Li at the cathode (titration with H2O and measure H2, or solid-state 7Li NMR, XPS depth profile, neutron depth profiling) and compare with the 36 mAh delivered.
  2. Replace the redox center and glass by a plain Li/Cu cell in the same glass; measure whether any emf above 0 V appears with Li plated on Cu (the first-law test).
  3. Independent impedance spectroscopy and Arrhenius plots of the glass by an outside lab.

Verdict summary

  1. Superionic glass: Never properly tested (outside the authors).
  2. Wetting/no SEI/no dendrites: Plausible, untested.
  3. Dipole condensates plus Q/Ceq voltage: Internally inconsistent (voltage model); dipole part Never properly tested.
  4. Plating cell: strong form Refuted by calculation; weak form Never properly tested.
  5. Cycling/energy density/EV: Never properly tested; 10.5 Wh/g Internally inconsistent; EV suitability Not testable as stated.