Naudin-JL: Arbiter (Step 3)

Role: weigh the physicist, mathematician and contrarian reviews, recompute where they differ, and give one final verdict per claim on the Step 2 scale. Tags: [read] from the project's documents (page, section), [web] with URL, [memory] with confidence, [calc] my own arithmetic (script arb.py in the session scratchpad, run in the project venv).

Terms used once: COP = coefficient of performance, here (useful output energy)/(energy put in). EHD = electrohydrodynamics, the push of a charged gas on its surroundings. F/I = force divided by current, in newtons per ampere. "Drift ceiling" = F = I d / mu, the most force that ions drifting through air can hand to the electrodes (I current, d gap, mu ion mobility, about 2e-4 m^2/(V s) for positive air ions).

Source gap (carried forward)

Naudin's pages were saved as text. Scope screens, graphs, videos, the 33-run CFR table, the MAHG per-run tables, Lifter v2/v3/Maximus pages and the Faraday-cage/enclosed-box pages are images or sub-pages that were not obtained. Also not obtained: the Gravitec/NSSTC vacuum video, Talley's Air Force reports, Stein's Purdue paper (a second web search this pass again found only citations), Brown's vacuum notes, the Dstl report, the USPTO/INPI/WIPO office actions and search reports. Verdicts on claims 4, 5 and 7 therefore rest partly on Naudin's description of a result and on the Investigator's recovered text pages (CFR calibration/low/medium runs, MAHG test method, EarthTech replication, Bishop's paper), not on the raw data. Where a conclusion rests on [memory] alone it is marked.

Stance on evidence (SOURCE HONESTY). Interested parties, reported as their characterisation: Naudin (hosts and promotes the ideas; no sale of devices found; co-inventor on two patents assigned to AES, a company formed to develop and sell electromagnetic generators; Horky (2009) says without a source that he joined Bearden's team; he calls his lab "my own and private lab"; see QE revision), Bearden/Magnetic Energy Ltd (MEG patent holders), Moller and Frolov/Faraday Lab (50-50 contract per Frolov in Kowalski), Bedini, Steorn, Gravitec (Serrano, patent holder, sells thrusters), Buhler/Exodus Propulsion, and Cornille/AES (the patent applicant). Bishop and the Sisyfos authors (Jelinek, Pekarek) are avowed sceptics; I weigh their arithmetic, not their tone. Single-source evidence is stated with its source and weighed, not labelled.

Verdict summary

Corrected claim list

Corrections to claims.md that the reviews establish (the claims, as the documents state them, with what a reader must not take at face value):

1. Lifter thrust. Naudin's frame is that the main thrust is "an electrokinetic effect on the medium", not ion wind. His own "true Lifter" page says the thrust needs the surrounding air ("a true Lifter fully enclosed in a box will not move") [read: _text/2003-Lifter-true-lifter, p.1], so the tube tests are not a vacuum-thrust claim. ARL did not "explain air thrust by ion drift": its abstract says the basis "is not understood" and its ballistic ion estimate (about 3 orders too small) is the collisionless model, not collisional EHD [read: ARL p.4]. Ianconescu, Sohar and Mudrik and the NASA/ISR report are the sources that claim an ion-drift explanation. The Dstl slide ("calculated ionic thrust more than an order of magnitude below observed") shows no method [read: Dstl slide 37].
2. Lifters. The Ianconescu match (about 6.5% on force) is at equal voltage; the model's current is 22 to 40% above the measured current, so thrust per ampere is not independently confirmed [read: deep/Ianconescu]. The 351 replications are Naudin's own self-registered list.
3. MEG. Naudin's two numbers are 6.76 W out against 3.25 W total DC input (COP 2.08) or against 1.5 W after he subtracts the control board's 1.75 W (COP 4.5) [calc]. Bearden's 7.3 is Bearden's number (2 Dec 2000 e-mail, via ZPEnergy), not on any Naudin page in the folder. Bearden's p.67 readings are "M RMS" (RMS of the power waveform), which Tektronix support calls meaningless; Naudin's later screens read "Mean" power, which is the right kind of quantity [read]. Naudin's own heat evidence (6 Dec 2000: "apparent" power greater than Joule heat in the load; "pessimistic about the calorimetric tests"; "measurement artifacts remain possible") belongs in the claim.
4. CFR. Add EarthTech's replication (COP2 average 0.98) and Kowalski, Little and Luce (COP 1.00 +/- 0.02) beside Kowalski's favourable 2003 comment. The 33-run table is still not in the folder.
5. MAHG. Add the control: Naudin's pure-DC blank returned 64%. The COP-21 page is titled "MAHG v2.0", filed under 2.1 on his index and logbook (Naudin's labelling is inconsistent). "10 MHz" is the pulse generator's rating, not the pulse frequency [read: pultests page].
6. Patents. FR 2 789 818 says "2.5 times" and "divided by 3" for 83 s versus 30 s (2.77 times). The US patent's own mechanical analogue says efficiency "cannot exceed 100%" [read: [0004]].
7. Other devices. On the Orbo page Naudin writes both "measured COP = 4.75" and "has demonstrated an overunity effect" (lead-in) and "the COP of this device is less than one. The factor 4.75 is NOT A COP" (comment); state both [read: deep/Orbo]. Naudin himself reports no overunity for the Bedini Magnetic Gate in a closed loop, "NO Overunity" for his magnetic-amplifier test, and "4.75 is NOT A COP" for Larsko's Orbo. These stay in the record as his own adverse results.

Claim 1: Lifter thrust is not simply ion wind; some force persists in vacuum

Agreement. All three reviewers: in air the thrust is ion drift with momentum handed to the air; Naudin gives no equation for a second mechanism; the vacuum-tube hover is ordinary electrostatics; the high-vacuum question is open. Disagreement is only in wording of verdicts.

Recomputed (script arb.py). Naudin's own four lifters, F/I against the drift ceiling of 150 N/A (d = 30 mm, mu = 2e-4):

Lifter 4 sits at the ceiling on a straight 30 mm path (ions follow longer curved field lines; ISR gives +38%). One caution neither physicist nor mathematician raised: Naudin's chart lists "weight 32 g, payload 4 g" for Lifter 4; if the 81.4 W hover carried 36 g, the ratio becomes 1.17 to 1.47 times the straight-path ceiling [calc], still inside the geometry allowance but with no margin. No point needs a second force. (Physicist's Lifter 1 "54 to 85 mN" should read 68 to 85 mN; the ratios he gives are right.)

Wire-in-vacuum-tube hover. Foil 200 x 20 mm x 12 um aluminium: 0.130 g, weight 1.27 mN [calc]. Two independent facts: (i) the measured current is about 1 uA (25 mW / 25 kV, confirmed), and the drift ceiling at 1 uA and 3 cm is 0.14 mN, nine times too small to lift the foil, so Naudin is right that ion drift from the wire does not hold it up; (ii) electrostatic attraction of a grounded foil by a 25 kV conductor, wire-over-plane image model restricted to the strip width, is 0.9 mN (bare 0.05 mm wire at 30 mm), 2 mN (0.5 mm), 2 to 5 mN at 20 mm and about 35 mN if the glass envelope carries the wire's potential on its outer surface [calc]. The reviewers' ranges (physicist 1 to 44, mathematician about 18 at 1 kV/mm, contrarian 0.7 to 4 mN) bracket this. The geometry (distance, charged glass) is not recorded, so the test cannot tell electrostatic pull from anything else; the force is of the weight's order or well above it. Both statements hold: Naudin excluded ion wind, and did not exclude the commonest static explanation. In the "true Lifter" version the plate and tube are rigidly joined (910 g on a torsion hanger): internal Coulomb forces cancel, so motion needs an external force (charged assembly against room surfaces, leads, or leakage); no angle, force, or control (tube removed, voltage off) is recorded. The unlinked version has the wire "handled by myself", so the reaction goes into his hand. Verdict: Refuted by calculation as evidence of a new effect.

Vacuum record, weighed (kept whatever the verdict). - Positive, all interested or undocumented: Brown's undated letter via Naudin (ARL Appendix B; notes "never published"); Gravitec/NSSTC video, 3 July 2003, 1.72e-6 Torr, 45 kV, rotation "weaker than at atmospheric pressure", held back until 2007 (Serrano, patent holder); Stein (Purdue 2000) at least 0.31 mN at 17 kV in 1.3e-3 Pa, known only through Losonc; Buhler/Exodus (US 11,511,891, commercial stake); Porcelli and dos Santos Filho 2016 (dielectric capacitors, not a vacuum lifter). - Null or reduced: Talley (Veritay, Air Force, 1988 and 1991: no thrust for static DC; a force only during breakdown with current), second-hand through NASA/ISR and ARL, and repeated in a Wikipedia summary [web: https://en.wikipedia.org/wiki/Biefeld%E2%80%93Brown_effect, via search result]; MSFC TM-2004-213283 "No NACAP performance was observed in vacuum" [read: research/files MSFC pdf, checked this pass]; ISR AIAA 2004 "no force was measured in a vacuum" [read, checked] with an apparatus that could not resolve less than one-sixth turn; Campbell/NASA US 6,775,123 "Thrust is substantially and significantly reduced in vacuum" and adds a gas supply for vacuum use [read, checked]; Tajmar and Schreiber 2020, no force to +/-0.3 mg up to 10 kV [web: https://www.sciencedirect.com/science/article/abs/pii/S0304388620300619; abstract as seen by the Investigator, page refused me with HTTP 403, so the check failed]; a builder in Naudin's own logbook (Xie Xuejun, 17 Jun 2006) saw the lifter "go down as air was pumped out" and only glow at low pressure. - ARL called vacuum "the most pressing question" in 2003. No neutral, calibrated, shielded, bake-out-controlled high-voltage vacuum test is on record. Ballistic electron emission at 45 kV and tens of microamps gives only 1e-8 to 1e-7 N and ballistic ions up to 1e-5 N [calc, physicist], so micro-newton rotation could come from leads, chamber walls, outgassing or field emission just as well as from a new force.

Verdicts. Air: Tested, not supported. Tube: Refuted by calculation (as evidence). Vacuum residual: Never properly tested (the contrarian's phrase "tested badly in both directions" is accurate; not dismissed without testing). Electro-gravity by elimination: Not testable as stated; Losonc's zero net force is a theorem for pairwise Coulomb forces on a closed system, and his own comparison sentence contradicts his numbers (20 mN is larger, not smaller, than 3.65 mN). What would settle it: one asymmetric capacitor and its mirror (plus a dummy of equal capacitance) on a torsion balance with shielded leads inside a grounded cage, 760 to 1e-6 Torr, 10 to 50 kV, with bake-out, residual-gas analysis, polarity reversal, force against pressure and current.

Claim 2: Lifters are real, reproducible and scale

Agreement on "real": ARL (three lifters and a wire, 30 to 37 kV), Ianconescu et al., Dstl (air and nitrogen, not helium), NASA/ISR, hundreds of builders. Disagreement on "scale": physicist says Refuted by calculation for practical craft, mathematician and contrarian say Plausible, untested. Recomputed: the ideal actuator-disc hover power for Lifter 4 (0.314 N over 0.574 m^2, 820 x 700 mm [read]) is 0.15 W against 81 W used, because ion drift speed is hundreds of m/s. One kilogram at Lifter 4's hover efficiency (3.86 mN/W) needs 2.5 kW, at its flight efficiency (2.36 mN/W) 4.2 kW [calc]. The physicist's "3.8 kW" is a slip; the correct range is 2.5 to 4.2 kW. That is an unfavourable energy budget, not a violation of a law, and Naudin's pages document scaling only to 194 to 250 g (Maximus; data pages not obtained). Stronger argument: the mathematician and contrarian, because "scales" as Naudin states it is a trend claim and the calculation shows cost, not impossibility. Verdicts: lifting in air Established; scaling Plausible, untested (Naudin's pages only) with the 2.5 kW/kg figure as the reader's caution. The 351-replication count is documented in one published source (Naudin's logbook), self-registered, no independent count. Settle: an independently weighed, powered flight of a 200 g or heavier lifter with measured power.

Claim 3: MEG apparent COP above 1

Where they agree: the Bishop input formula is wrong (constant voltage times AC amplitude over root 2; Smith correct); Bearden's "M RMS" readings are invalid; the radiation explanation for missing heat fails; no calorimetry and no closed loop exist; Jelinek-Pekarek's Poynting argument is sound in outline though Appendix 1 (magnet irrelevant) holds only for a linear core.

Disagreement: the strong form. Physicist and mathematician: Refuted by calculation. Contrarian: Never properly tested, because Jelinek-Pekarek's proof has errors in detail and the magnet's stored energy could give a transient gain. Weighing: the energy conclusion does not use Appendix 1; it needs only that all energy crosses the four electrical ports and that the core is single-valued. Under those assumptions mean output cannot exceed mean input. The one loophole, the magnet's depletable energy, is small. Recomputed: patent magnet 40 x 25.4 x 38.1 mm = 38.7 cm^3; (BH)max x V is 11.6 J for NdFeB (300 kJ/m^3) and 1.2 J for ferrite (30 kJ/m^3) [calc; (BH)max values from [memory], high]. An absolute upper bound, demagnetising a 1.3 T NdFeB magnet completely, Br^2/(2 mu0) x V is about 26 J [calc, Br value [memory], medium]. A gain of 3.5 W (6.76 minus 3.25) would use 11.6 J in 3.3 s, 7 W in 1.7 s [calc]. Smith's 30 J spin-precession figure is an unmeasured speculation and does not change this: 75 W would last 0.4 s. The contrarian's point survives in this form: no sustained gain can come from stored magnet energy, and no run long enough to show a decaying transient is reported. I adopt: strong form (sustained net gain from a passive magnet-biased core) Refuted by calculation, with the stated assumption. Bearden's vacuum source sits outside the assumption by definition: Not testable as stated (no coupling, no equation).

The "apparent COP" is the real dispute. Checks I reran: - Ohm's-law cross-check (physicist, Bishop): Naudin's stated 100 kohm "conditioned" load at 512 V amplitude dissipates 1.31 W; his screen shows 10.28 W, which needs 12.8 kohm. Bishop's 500 V / 44 mA gives 11.4 kohm and 11.0 W against 1.25 W (a factor 8.8) [calc]. Either the resistor was about 11 to 13 kohm under 1 kV at 20 kHz, in which case it would dissipate about 11 W and run hot, but Naudin reports the load "only a bit warm" (2008) and, in 2000, heat less than the apparent power; or the current channel over-reads. Smith's reply that the load was "conditioned" is circular, because it uses the disputed measurement to justify itself. - Phase sensitivity (mathematician): Naudin's screens give 255.3 V x 27.23 mA = 6.95 VA against 6.764 W (power factor 0.973). To make a true 3.25 W look like 6.76 W needs a phase error of 49 degrees, a 5.7 microsecond skew at 23.8 kHz [calc]; typical probe mismatch is tens of nanoseconds. So skew alone is not shown to explain it unless the circuit is strongly reactive, in which case a 1 microsecond skew can inflate readings by 1.2 to 1.7 times. The numbers are mutually consistent as measured; the open question is whether the channels are right. - Bearden's own figures disagree: 12.7 W / 3.588 W = 3.54 per coil, about 2.9 after the RMS correction, about 2.2 from the paper's own fitted curve at the same input [calc, read]. - Horky (2009, believer, school project): printed COP 1.57 to 2.89; on full supply power 0.69 to 1.26 for the one lamp measured and 1.37 to 2.53 with the second lamp's identical power doubled; the "I1-I2" row is I1 minus an undefined no-load current [read: deep/Horky]. - Radiation: a 1 A m^2 magnetic dipole at 20 kHz radiates 3e-13 W; the electric dipole fitting 1250 V/m at 50 cm radiates 8e-12 W [calc]. The missing heat cannot be radiated. - Flickinger: COP above 1 only with non-linear loads. A non-linear load cannot change the device's energy balance (the load is outside the surface) but does make power measurement of non-sinusoidal waveforms hard; this is a signature of a measurement problem, not a proof.

Which verdict for "apparent COP": the contrarian's Tested, not supported would require a controlled test; none exists. Naudin's own heat remark is a qualitative, uncontrolled observation (it differs between 2000 "warms up quickly" and 2008 "only a bit warm"). The physicist's and mathematician's Never properly tested is correct, with the weight of evidence stated: every internal check that can be run on the numbers on file points to a measurement error, and the inventors' own papers are not self-consistent. What would settle it: a calibrated power analyser or sample-by-sample mean(v x i) with probe deskew, calorimetry on the load against measured mean input, a linear resistive load as control, then a closed loop with a start battery for minutes with the battery voltage logged. The closed loop has been missing for about 26 years (Naudin, 13 Nov 2000: "I am trying to close the loop... have not yet succeed"; still "not yet" in 2008) and is a cheap test, so its absence is itself informative.

Absence of evidence (MEG). Bearden announced on 2 Dec 2000 a DoD-linked university replication and a certified laboratory test and said Naudin reached 7.3 in a third build [web: https://zpenergy.com/modules.php?name=News&file=print&sid=3918, via Investigator]; no report has been found. The patent was kept alive to 2013 and pledged to Colmac Holdings Ltd (Canada) in 2006 [read: Espacenet note]. Explanations: (a) never carried out (supported by: the names were withheld, no later citation); (b) carried out and negative, not published (supported by: no follow-up announcement, patent continued without a product); (c) carried out and confidential (supported by: the DoD link; weak, because no contract or report number appears anywhere). The record cannot choose. Deciding records: DTIC and university reports naming the MEG, the Correspondence pages at cheniere.org after 2001 (the site did not respond), Wayback copies of Naudin's v3.x pages, Colmac's registry file.

Claim 4: Mizuno-Ohmori plasma-electrolysis reactor (CFR)

Agreement on method: Naudin's arithmetic is right; the energy balance can be inflated by counting evaporation after power-off, and by counting droplets as vapour (carrying 84 J/g of sensible heat instead of 2260 J/g, a factor 27 [calc]). Disagreement: physicist adds Refuted by calculation for Naudin's accounting; mathematician and contrarian say Tested, not supported.

My checks. Calibration 56 525 J in, 55 700 J out (0.985); low power 74 081 J in, 73 181 J out (0.988); medium power 23 714 J in, 35 767 J out (1.51) [calc, all confirmed]. The printed ratios on his page are inverted (cosmetic). The mathematician's "170 g needs 2.6 times the input" for run 28 and the physicist's "at least 124 s" are the claimed COP 3093/1200 = 2.58 restated, not independent evidence, because the run duration is unknown and COP does not depend on it. Likewise "the input could evaporate only 4.7 g, not 10 g" in the medium run restates the claimed excess. They are descriptions of what the claim asserts, so I do not count them as refutation. The independent evidence is EarthTech's replication on Naudin's own method (Little and Luce, 15 Jul 2003): 15 runs in a Dewar flask, Clarke-Hess meter, 10 mg balance, COP2 average 0.98, and the column "COP with Cooldown", computed by adding the evaporation after power-off, runs 1.27 to 1.77 (I confirmed rows DEW1 to DEW9 at 1.27 to 1.62) [read: research/pages/et_replicationjln_2003.txt]. EarthTech says the cooldown evaporation is paid for by the heat already counted in E Heating, so adding it counts the same energy twice. The reason this is not yet Refuted by calculation for Naudin's own runs: EarthTech's reference for Naudin's use of the cooldown term is "private communication with J.L. Naudin" [read: ref 9], Naudin's page does not say when the 10 g or 26 g were weighed, and his low-power run closes at 0.988. The mechanism is shown on EarthTech's data; whether it explains Naudin's 82 s run is an inference. Kowalski, Little and Luce (ICCF-12, 2005): mean COP 1.00, standard deviation 0.02, 1.03 at 350 V, splashing and foam droplets named as the error source [read, checked in research/files]. Mizuno 2004: "few claimed replications ... In most cases, no excess heat or hydrogen is observed" [read, checked], which contradicts "fully replicable". Positive reports kept in the record: Mizuno and Ohmori (up to 200% in some runs, JJAP 2000, via Naudin's page), Cirillo, Dattilo and Iorio (ICCF-11 2004, COP 1.31 to 1.41), Mizuno and Toriyabe's single explosive event (estimated heat 800 times input), Kowalski's 2003 comment that Naudin's 33 runs were "reasonably reproducible" (before the EarthTech result, with the warning that doubts about Mizuno's data can be raised about Naudin's). None includes a control against droplet entrainment.

Verdict: Tested, not supported for Naudin's reactor (independent instrumented replication gave no excess and reproduced his level by his own accounting convention). General Mizuno claim: Never properly tested against droplet carry-over and post-run accounting. Naudin's reply to EarthTech has not been found. What settles it: condense and weigh all vapour and droplets (closed vessel), log mass and temperature through cooldown, run a resistive heater in the same cell at the same evaporation fraction, calibrate with the plasma instruments.

Claim 5: Moller Atomic Hydrogen Generator (MAHG)

Agreement: hydrogen dissociated on a hot tungsten filament returns exactly its dissociation energy (about 436 kJ/mol H2 [memory, high]) on recombination; the cycle moves heat, it does not create it. Naudin's blank (pure DC, runs 38 to 40) returned 64% while pulsed runs gave 224% (runs 11 to 37) and about 240% (2 h, 875 W out, net 515 W) [read]. Frolov's own earlier set-up gave 82% on smoothed DC and 124 to 250% with pulses, "with the remark that this gave reason to think that a possible mistake had been made" [read: mahg1 page].

Disagreement: the contrarian refuses "Refuted by calculation" until the input measurement is shown. The distinction is between the mechanism and the number. The mechanism (a closed H2/2H cycle as an energy source) is refuted by a state-function argument, and I adopt that. The number is not refuted.

The duty-cycle hazard (mathematician, contrarian, physicist, all agree on the algebra): for a pulse train of duty D, true mean power is D Vp Ip; RMS(V) x RMS(I) equals it when voltage and current have the same shape and, by the Cauchy-Schwarz inequality, is never below it, so RMS x RMS cannot inflate COP; mean(V) x mean(I) is D^2 Vp Ip, too small by D, so COP too high by 1/D. At D = 5%: 21 x 0.05 = 1.05 and 11.53 x 0.05 = 0.58 [calc]. What the pages say: the data-sheet columns are "Rms Voltage and Rms Current", while "the average voltage and current are also measured on the digital oscilloscope" without saying how they entered the efficiency [read: pultests and lrtests pages]. The long half-sine runs (50% duty, 224 to 243%) used RMS and are not exposed to this hazard. The 1/D pattern ("efficiency increases up to 1153% while the duty cycle decreases up to 5%") fits the hazard but also fits a real effect; the per-run table needed for the test (does COP x D stay constant?) is an image not in the folder. The 64% DC blank is a separate problem: if the bench loses 36% under DC it should lose the same under pulses, so the pulsed 224% would be at least 3.5 times larger than a DC-calibrated reading (224/64 = 3.5 [calc]). Naudin's filament-temperature formula (radiation only, S = 1e-3 m^2) gives 774 K for 20 W and 1590 K for 365 W, below the 2000 K he gives for dissociation [calc], so his own temperatures do not support the "1% to 7% atomic hydrogen" statement.

Verdicts: mechanism Refuted by calculation; COP 5 to 21 Never properly tested (originator's contractor and replicator only; Moller and Frolov hold a stake; the Australian test Frolov mentioned is not found). Settle: replace the filament with a pulsed resistive dummy in the same circuit and calorimeter (a 3.5-fold over-read means the bench, not hydrogen); tabulate COP x D for all pulsed runs; true sampled v x i mean on one record.

Absence of evidence (MAHG). Frolov's own report (New Energy Technologies no. 3, 2005), the Australian test, and any Moller company records were not reached. Explanations: quiet negative results, loss of interest after the DC-versus-pulsed discrepancy, no funding. No evidence for any; the record is simply not examined yet.

Claim 6: the Cornille-Naudin patents

FR 2 789 818: all agree. Resistive charging wastes half the energy; a switched inductor stage charges with 90% or more [memory, high]; stepwise charging with N equal steps has efficiency N/(N+1) [calc; web: https://staff.najah.edu/en/publications/11442/]. The patent claims faster charging (83 s to 30 s, a factor 2.77) with no efficiency definition; the examiner cited X/Y art against claims 1, 2, 4 and 5. Plausible, untested.

US 2005/0057116: all three agree the headline is an assumption and the multi-cycle formulas break charge conservation (the patent recharges C1 "with the voltage V2" with no source connected; conserving charge returns V1). Recomputed: 3.5 nF bank, E1 = 8.75e-4 J, beta 200, E2 = 0.175 J; efficiency 0.175/(2 x 8.75e-4 + 0.050) = 3.38, break-even beta = 2 + Es/E1 = 59.1, and beta = 10 (which the patent itself allows) gives 0.17 [calc, confirmed]. The mathematician's extra figure, alpha = 0.059 at the measured beta = 3.5, I reject: it mixes the 30 cm design's E1 and Es with the Fig. 8 trace of an unspecified device; nothing ties them together. The single measurement (500 V to 1750 V on opening K1, 1 gigaohm probe) is a constant-charge capacitance change, which is ordinary; it records no energy in or out.

Disagreement: mathematician adds Refuted by calculation once the work is counted; contrarian grants "Established" for the voltage gain and keeps "more than one" at Never properly tested. Physical point: raising the stored energy at constant charge needs equal work; in the patent's mechanical analogue a person pays it. In the plasma version the energy must come from the plasma's extinction, from the 50 mJ drive, or from elsewhere; with the patent's own numbers the rise (0.174 J) is more than three times the stated plasma energy (0.05 J), so something unmeasured must supply 0.12 J. That is a missing account, not a measured violation, and no one has measured either side. Physicist's added mechanism (induced surface charge on the glass stays after the plasma is off, so the voltage may not rise at all) is a documented barrier-discharge effect [web: hal.sorbonne-universite.fr/hal-03541456v1, abstracts only]. Three incompatible sources of the extra energy are given (Debye-length work, "surrounding magnetic ether", Earth's potential, formula 32 with 700 uF at "several million volts"; the accepted global potential is about 250 kV, which would store 22 MJ in 700 uF [calc; web: https://glossary.ametsoc.org/wiki/ionospheric-potential/]). The pressure example (136 Torr) reproduces as about 34 Torr (deep read; mathematician did not redo it).

Verdicts: FR Plausible, untested; US capacitance-switching voltage rise Established (as a capacitor effect); "efficiency more than one" Never properly tested; multi-cycle formulas and the unreconciled ledger Internally inconsistent. Settle: measured energy in (including 15 kV ignition and 2 kV bias) and out over 1000 cycles by integrated power or calorimetry on the Fig. 8 device.

Absence of evidence (AES and the patents). Record: FR 2 822 590 granted (B1); US application assigned to AES, abandoned 27 Jun 2006 for failure to respond to an Office action; EP 1389360 deemed withdrawn 2008 (renewal fee 2007 unpaid); AES (SIREN 419 785 985) last accounts 2015, no employees, struck off 8 Mar 2021 [read: Espacenet, EPO Register, Pappers notes]. Explanations: (a) cost of prosecution (supported by: renewal fee unpaid, small company; against: the French patent was granted and kept); (b) a substantive objection the applicants chose not to answer (supported by: abandonment for failure to respond to an Office action; against: no text of the action seen); (c) the applicants had no energy measurement to show (supported by: only one trace in the patent; against: none, but silence); (d) loss of interest or retirement of the principals (supported by: struck off 2021, Cornille president to 2021; neutral). Deciding records: USPTO file wrapper for 10/472,714, the PCT search report and opinion (WO02075912), the EP search report, INPI B1 claims. Not reached.

Claim 7: other hosted devices

Reviewers agree. Kept in the record: Bedini G-field "120% or better" and 12 W in, 72.5 W out for his own machine (Bedini's figures, no method; Not testable as stated); Bedini Magnetic Gate (Naudin's closed-loop test: "no overunity effect"; Tested, not supported, by the replicator himself); Mini-Romag (24 W, no data); Orbo (Naudin's build leaves the 400 mA input unchanged on loading, which is not a power statement without voltage and phase; Steorn's CEO called it "in no way a valid replication" (stake: licence sales); Steorn's jury of 22 experts led by Prof. Ian McDonald said in June 2009 that Steorn's attempts "have not shown the production of energy", and Steorn disputed this [web: https://theness.com/neurologicablog/the-jury-is-in-for-steorn-no-free-energy/, https://en.wikipedia.org/wiki/Orbo via search results; the jury's own posting was not opened]); 2SGen, KAPAGEN (2100 W of lamp light, input not given; lamp rating is not output power), GEGENE, Meyer-type cell, Hendershot (Epps found it "definitely not a nonreciprocal system" at higher voltage), Squires proposals (not built). Verdicts as in the summary table. Naudin's candour in this group (the Magnetic Gate correction, "NO Overunity", "4.75 is not a COP") is part of the record, and is to his credit.

What would settle it (one test per disagreement)

Open questions

1. Is there any residual static force in clean high vacuum at tens of kilovolts, once leads, walls and outgassing are controlled? (Gravitec video and Brown notes not obtained.)
2. Which of voltage-current skew, probe error or a truly non-linear load produces the MEG screens' 10.28 W into a stated 100 kohm?
3. Did Naudin weigh the 10 g and 26 g of CFR evaporation after cooldown, and what do the 33 run tables show?
4. Does MAHG COP times duty cycle stay constant, and how did the "average voltage and current" enter the 1153% runs?
5. What did the unanswered 2006 USPTO Office action say, and why did a patent that was granted in France lapse elsewhere?
6. What became of the DoD-linked MEG replication Bearden announced in December 2000?
7. Why do Gravitec/NSSTC (rotation in vacuum, 2003) and MSFC (no vacuum performance, 2004), both in Huntsville, differ? (A lead, not a finding.)
8. Naudin's later work (BiTT, ReGenX, GFS UAV) and Frolov's MAHG report were not reached.

Corrections to earlier documents

1. claims.md claim 1: ARL did not explain air thrust by ion drift (its abstract says the basis "is not understood"); the sentence should name Ianconescu and ISR. The Dstl slide shows no method.
2. claims.md claim 1: the tube pages say the surrounding air is required; they are not a vacuum-thrust claim.
3. claims.md claim 3: add Naudin's 6 Dec 2000 heat and EM-radiation text; separate Bearden's 7.3 from Naudin's 1.75/2.08/4.5; "RMS watts" applies to Bearden's p.67, not to Naudin's "Mean" screens; Smith is right on Bishop's input formula, Bishop right on the 9-fold load mismatch.
4. claims.md claim 4: add EarthTech (COP2 0.98) and Kowalski/Little/Luce (1.00 +/- 0.02); the page ratios "56525/55700" are inverted. claims.md claim 5: add the 64% DC blank; the COP-21 page is titled v2.0; "10 MHz" is an instrument rating. claims.md source gap: the CFR calibration/low/medium runs, MAHG method and blank, Bishop's paper and EarthTech were recovered by the Investigator; the gap is now the 33-run and MAHG per-run tables.
5. physicist.md: Lifter 1 ceiling "54 to 85 mN" should be 68 to 85; "1 kg needs 3.8 kW" should be 2.5 kW (hover) to 4.2 kW (flight); claim 4 "Refuted by calculation for Naudin's accounting" overstates (the cooldown use rests on Naudin's private communication, and the run-28 and medium-run "impossibilities" restate the claim).
6. mathematician.md: the claim-4 bounds (10.5 g; 66 g for run 28) restate the claimed COP; the alpha = 0.059 figure mixes devices; Lifter 2 and Lifter 4 both sit near the drift ceiling (the Lifter 1 points sit at a third).
7. contrarian.md: "Never properly tested" for the sustained MEG gain is replaced by Refuted by calculation under stated assumptions (the magnet's energy is 1 to 26 J); its verdict "Tested, not supported" for the apparent COP is replaced by Never properly tested because no controlled test exists.
8. Digest of Horky (p.18-20): already corrected in claims.md (artifact versus error; I1 minus I0).

Web verification (budget 8; used 4)

• Steorn jury, June 2009: retagged [web] (Neurologica, Wikipedia via search); wording "have not shown the production of energy"; jury's own posting not opened.
• Tajmar and Schreiber 2020: check failed (ScienceDirect HTTP 403; search returned no abstract); stays as the Investigator's second-hand report; not a verdict driver.
• Talley null: confirmed only through a Wikipedia summary and the NASA/ISR and ARL quotations; Talley's reports not found.
• Stein (Purdue) vacuum paper: check failed again; only citations found.
• Checked locally in saved copies: MSFC, ISR and Campbell vacuum statements, Mizuno 2004, Kowalski COP 1.00 +/- 0.02, EarthTech table (all [read]).
• Remaining [memory], none verdict-driving beyond noted: (BH)max values (high), Br of NdFeB (medium), H2 dissociation energy (high), switch-mode charger efficiency (high), type K thermocouple tolerance (medium, MAHG side remark only).

QE revision (2026-10-09)

Inputs: qe.md (verdict MINOR), the new deep reads in deep/ (Dstl slides pp.17 and 27-37, Horky pp.1-4, CFR index, Kowalski item 73, Orbo, ACT, GEGENE, Wikipedia), and three web checks this pass (budget 8; 7 used in all). Tags as before; [calc] is my arithmetic. Result: one verdict changes (claim 4, now split), one is added (1e), the rest stand with added record. No role removed material; the items below restore what the reviewers had and the first arbiter pass dropped.

1. Dstl check (QE defect S1; claim 1a)

The deep reader transcribed Dstl's charts [read: deep/Dstl, page images pp.17, 28-34]. p.17 is a photograph, not a table; the numbers are on pp.28-34. Using Dstl's own wire heights (the "d" of its p.21) and currents with F = I d / mu, mu = 2e-4: - IW8, 44 mm, 6.16 g, 1.07 mA: F = 0.235 N against weight 0.060 N, ratio 3.9 [calc, reproduced]. - IW8, 30 mm, 3.16 g and 4.22 g: ratios 2.2 and 2.7. IW3 (1.35 g) at 20 and 30 mm: 2.1 and 4.8. IW4 at 25 mm: about 3.0 [calc, from chart readings, +/-0.02 mA]. - So the ion bound is 2 to 5 times the weight lifted. Dstl's slide 37 says the calculated ionic thrust is "more than an order of magnitude less than that observed". For that to hold, d would have to be about 1 mm (0.0604 N / 10 needs I d / mu = 0.006 N, d = 1.1 mm at 1.07 mA [calc]); no slide supports that. Dstl shows no formula, mobility or gap, so its own calculation may differ; as far as the slides show, the statement is contradicted by Dstl's data. This is consistent with my Lifter 1 to 4 table (lifters at 0.3 to 1 of the ceiling) and removes the one institutional dissent kept against the ion-drift account on the numbers available. It does not remove Dstl's other observations, which stay in the record: effect "independent of voltage polarity" (p.37) beside "greatest force ... when small electrode is positive" (p.36), an internal tension with no polarity data shown; helium, no lift at 4.2 mA and 12.8 kV, nitrogen similar to air (p.28); "the details of the physics ... not understood" (p.36); an MoD horizon-scanning briefing (the stake is defence interest), with no error bars. - "Weight loss observed prior to lifting" refers to the p.34 balance chart: IW8 reads 2.94 g at 0 kV, about 2.9 g by 13 to 14 kV, 1.85 to 2.15 g at 18 kV where the data end; with 6.7 g added the reading falls to about 8.7 to 8.9 g (7 to 10 percent) at 22 kV with no lift [read]. Any upward force growing with voltage and current gives a gradual fall, so by itself this does not separate ion thrust from another cause. It would be anomalous only if the loss began before corona current; no current is shown at those voltages and no control (balance-pan electrostatics, ion wind on the balance) is described. Open, not a refutation. Verdict 1a stands: Tested, not supported (as an anomaly). Stronger than before.

2. Vacuum and related record added to claim 1c (QE items R2, R3, R5, 3.7)

Kept next to the earlier list, with source and weight: - Honda R&D (1996, via Musha): 62 g glass-disc capacitor immersed in oil, 8 kV AC pulses, weight loss 0.9 +/- 0.63 g, scatter to 1.9 g [read: deep/Musha]. One relayed measurement; errors as large as the effect; no controls described. A weak positive, not a test. - Brown's letter in ARL Appendix B [read: deep/ARL p.36-37]: a 4 ft rotor gave "virtually the same" torque in air and under transformer oil up to about 300 kV, "seeming to prove that aero-ionization was not wholly responsible"; Brown also reports years of torque logs correlated with solar, lunar and sidereal time, a Navy proposal declined on the expectation of no vacuum effect, and a 15 inch dome lifting "110% of its weight" (April 1973 reply). Brown is the originator and these are undated or recollected claims with no raw data (single source, quoted as the interested party's own account). Taken at face value the oil result would argue against air ionisation as the whole story. No test by others is on record. Naudin's own "TT.Brown EKP device in MINERAL OIL" page, which could bear on this, was not obtained. - NASA/ISR [read: deep/NASA-CR p.11-13]: the null is weaker than I wrote. The rig could not resolve less than one-sixth of a turn, the gaps were set for atmospheric pressure (so they arc at low pressure), maximum 50 kV, no torsion balance, no force bound given; one exception, Device 2 moved about one-eighth of a turn at 44 kV with "a large arc" at 5.5e-5 Torr, which the authors attribute to ejected material. This moves ISR from "no force was measured" to "a force below the rig's resolution would not have been seen". My earlier verdict (Never properly tested) already allowed for that. - Transdimensional Technologies (TdT), June 2001: Naudin's ACT page says the thruster "has been tested in June 2001 by Transdimensional Technologies in the vacuum chamber of the NASA Marshall Space Flight Center" with a photo link and nothing else: no pressure, voltage or result [read: deep/ACT]. Naudin relays it; TdT is a commercial party. A third Huntsville vacuum event, bearing on open question 7; result unknown. Naudin's own ACT run was in air (65 rpm at 1.22 W; no thrust in newtons). - Wikipedia (secondary, editor-written) says "Experiments have been conducted and found to both prove and disprove a force at very low pressure" and names none [read: deep/Wikipedia]. It cites Tajmar, Kossling and Neunzig, Sci. Rep. 14, 19427 (2024) only as support for "no replication / corona wind"; the paper's content was not read here, so it stays a lead for claim 1c. It dates Talley's measurement 1988 and his report 1990 (my table's "1988 and 1991" cites NASA/ISR and ARL; the dates should be reconciled when Talley's report is obtained). Verdict 1c stands: Never properly tested.

3. Cornille-Naudin Trouton-Noble anomaly (QE item R1; new claim 1e)

Naudin's own experimental work with his co-inventor, and the basis of Cornille's "stimulated forces" theory, was dropped from my first pass. Reported: a 500 pF Plexiglas capacitor on a nylon thread charged to about 70 kV by a Wimshurst machine "sought an East-West position and stayed locked" over dozens of repeats, 1996 to 1999 [web, via Investigator: https://www.longdom.org/open-access/a-review-on-electromagnetic-propulsion-by-stimulated-forces-53822.html]. The physicist: a reported anomaly, single source, "I do not rule it out"; mainstream expects a null (Trouton-Noble) [memory, high]. The contrarian: candidate for the classic artefacts of a charged body near grounded surfaces, leakage-current torque and air currents; none described as controlled. My weighing: documented in one published source (Cornille's 2020 review, an interested party as co-inventor and AES principal); no raw data and no second laboratory. East-West alignment is also what a thread-suspended charged body near a stray ambient field would do, so the observation does not select a theory. Verdict: Never properly tested. Settle: rotate the suspension and the room geometry, reverse polarity, use a screened enclosure and log the heading against sidereal time; compare with the same capacitor uncharged.

4. Claim 4: CNAM and other replications (QE items 3.1, 3.2, R4); verdict changed

What the deep reads establish. Naudin's CFR index page lists, in his words, "Full CFR replication at the CNAM in Paris" (no date on the page), "Successful" replications by Ronald Hohls (South Africa, 19 Sep 2006) and at Boulder, Colorado (15 Mar 2006), Morgan H. and Marissa C., "a French lab" and a Stirling engine run on the CFR; the linked pages and results are not in the folder [read: deep/CFR-index]. Correction to qe.md 3.1: the dates are Hohls 19 Sep 2006 and Boulder 15 Mar 2006; the CNAM item has none. The ICCF-12 paper linked from that page is by Dattilo, Iorio and Cirillo on transmutation, not a heat balance [read]. Naudin thanks Prof. Pierre Clauzon, Prof. Jean-Francois Fauvarque and Gerard Lalleve (Laboratoire d'Electrochimie Industrielle, CNAM) "for their scientific support", and writes "some French physicists work closely with me"; Clauzon's affiliation is not stated in the quote [read: deep/Kowalski item 73]. Naudin's own claims; no word from the three people in the folder.

New web check this pass (none of the primary reports were opened; these are Kowalski's pages, from a replicator with no stake in Naudin's work): - Fauvarque, Clauzon, Lalleve (2005), "Abnormal excess heat observed during Mizuno-type experiments", is cited by Kowalski with "COP, at 350 volts, approximately 1.35" and "COP's up to 1.41"; Kowalski also records Clauzon at ICCF-12 saying the cell voltage must be raised gradually (about 20 V per minute), the cathode was "preconditioned", and excess heat was not confirmed without preconditioning [web: https://msuweb.montclair.edu/~kowalskil/cf/270iccf12.html]. A search snippet places Clauzon at CNAM Electrochimie in a later paper by Biberian and others, which says Fauvarque et al. showed excess heat in light water near boiling at atmospheric pressure. This confirms that a CNAM group reported excess heat in the same cell type, on the same dates Naudin cites. It is not clear from these pages that the CNAM report controlled for droplet carry-over or counted cooldown differently. - Same Kowalski page: his Texas (EarthTech) table, nine values from 0.97 to 1.05 (about 1.00), and Colorado (Slaughter) mean 0.90 +/- 0.2 at 500 to 700 V; "Clearly we have not replicated such results". Droplets and cooldown are not discussed there; Slaughter's 30 later attempts "were not successful" (P.S. 3 Jan 2006). - A later Kowalski draft (dated 28 Jun 2006) calls Paris1 and "Colorado2" positive: 41 tests, mean COP 1.24, standard deviation 0.13, "seems to be real", origin "not clear to us"; it records the EarthTech Austin work as not confirming excess heat [web: https://msuweb.montclair.edu/~kowalskil/cf/300positive.html]. It is a draft and a replicator's working text; the final form was not checked. It means my statement of Kowalski's position (COP 1.00 +/- 0.02) is incomplete: his own view moved between late 2005 and mid 2006 and must be reported with both.

Weighing. Evidence against Naudin's own numbers: EarthTech on his method (COP2 average 0.98, excluding runs 9 to 11 with the stirrer off; the cooldown column that counts post-power-off evaporation gives 1.27 to 1.77) and Texas about 1.00. Evidence for excess heat in the same cell type: the CNAM group (1.35 to 1.41 with preconditioning), Cirillo, Dattilo and Iorio (1.31 to 1.41), Mizuno and Ohmori, Kowalski's Colorado2 draft (1.24), Naudin's 33 runs. None of the positive reports is shown here to include a droplet-entrainment control or a heater-in-same-cell control; the EarthTech mechanism (cooldown counted twice) is a specific, checkable explanation for Naudin's accounting but is not shown to cover the CNAM data. I do not have grounds to call the Mizuno-type effect refuted, and the record is not one-sided.

Verdict (replaces the earlier single verdict): Naudin's method as replicated by EarthTech and in the Texas cell: Tested, not supported. Naudin's own 33 runs: Never properly tested (table not obtained; the accounting issue is shown only on EarthTech's data). CNAM and the other positive reports: Never properly tested against droplet loss and post-run accounting (reports not read; CNAM's preconditioning requirement is a testable condition no neutral party has applied). The general Mizuno claim stays Never properly tested. Added settling record: Fauvarque, Clauzon, Lalleve (2005) and the Dattilo, Iorio, Cirillo ICCF-12 paper, Naudin's nfrcnam, Hohls and Boulder pages (Wayback), and Kowalski's final text. Same settling test as before (closed vessel, droplets, cooldown, resistive-heater control), run with CNAM's preconditioning protocol.

5. MEG: Horky pp.1-4 (QE defect S2; absence of evidence)

Horky (2009) states, with no source cited [read: deep/Horky p.2, translated]: Naudin "is now a member of Bearden's team" (while p.14 says Naudin took up MEG research "independently of Bearden's team", and that he "has so far not managed to build the MEG as an autonomous device"); Bearden and team moved to Hungary and a company Magnetic Energy Limited was founded, with a possible factory ("snad", "most probably" on the factory and its place only); Bearden concluded an agreement with a "National Matarials Science Laboratory" at the Academy of Sciences of a friendly country ("apparently Hungary", Horky's own inference); and Bearden's team "succeeded" in developing the MEG as an automotive closed-loop module of 2.5 kW, four synchronised for 10 kW, though "not a single piece has so far left the production line". No power in, efficiency, test conditions or tester is given. This is the only closed-loop claim in the paper and the only claim in it with no data and no source behind it; Horky's own replica ran from an external DC supply. Effect on the record: Naudin's stake is wider than I stated (see section 7). In the absence-of-evidence analysis for the MEG, explanation (c) "carried out and confidential" gains one unsourced claim of development; explanation (b) is unchanged. A claimed 2.5 kW closed-loop unit is exactly the test whose cheap form (a closed loop) Naudin has not achieved in 26 years; no record of it exists. Claim 3 verdicts stand (3a Refuted by calculation under stated assumptions; 3b Never properly tested; 3c Not testable as stated).

6. Claim 7 additions (QE items R6, 3.10, 3.13)

• Orbo: state both Naudin statements (see Corrected claim list 7). Larsko's 4.75 is 19 mW / 4 mW, a ratio of two small differences (input 4 V x 31 mA to 4 V x 32 mA, one meter step; output 7.72 V x 2.99 mA = 23 mW); output over total loaded input is 23 / 128 = 0.18 [calc, reproduced]; the motor slowed from 1551 to 1240 rpm under load, which supplies load energy without any gain. Naudin's own TEST 7 (400 mA unchanged when the LED panel is switched on, analog ammeter) gives no input voltage, so no power can be computed [read]. Verdict stands: Tested, not supported (Steorn's jury); Naudin's build and Larsko's data are Never properly tested as power statements.
• GEGENE (induction-cooker controller driving a Tesla bifilar coil; 18 numbered tests): the index page states no result for any test. Titles only: TEST #3 "calorimetric method", TEST #4 "first test of a closed loop GEGENE by Romero UK", TEST #9 "CLOSED-LOOP (Phase 2) via a Grid Tie Power Inverter", TEST #16 coil temperature with 4180 W of halogen lamps (a load size) [read: deep/GEGENE]. A Russian group's video title on the page ("IN 230 v / 4.3 a / 900 W OUT 2.3 KW") is that group's claim, no method. Results are on sub-pages not obtained; the absence of a figure here is not a negative result. Verdict: Never properly tested; the claimed closed-loop tests are listed as claims whose records are not obtained.
• 2SGen and DLE self-reports (DC input falls with the magnet; "input power drops dramatically and RPM doubled" under load) stay as Naudin's descriptions without voltage, phase or method: Not testable as stated. Home-page items (Zaev "COP up to 1.35"; Cornille's electrostatic pendulum "demystified") are listed, not weighed.

7. Corrections made in this file, from qe.md

1. "Unpaid hobbyist" had no source. Replaced above: no sale of devices found; co-inventor on two patents assigned to AES (formed "to develop and sell electromagnetic generators" [read: research/2026-10-09-pappers-societe.md]); one unopened snippet names him AES R&D director; Horky says he joined Bearden's team (no source); Naudin says "my own and private lab" [read: Kowalski item 73]. His stake in the MEG, MAHG hosting and AES patents is therefore commercial as well as promotional. The same words in contrarian.md section 2 are for the orchestrator to correct.
2. "Missing for 24 years" corrected to about 26 years (13 Nov 2000 to 2026). The same error in physicist.md claim 3 is for the orchestrator.
3. EarthTech's 0.98 excludes runs 9 to 11 (stirrer off); say so wherever it is quoted (arbiter claim 4 above, primer, physicist, mathematician).
4. The qe.md note 3.1 date reading (CNAM 19 Sep 2006) is itself wrong, as section 4 shows; the deep read of the HTML corrects it. For the orchestrator: qe.md and any digest that assigns 19 Sep 2006 to CNAM should say Hohls.
5. Orbo: arbiter claim 7 cited only the candid half; fixed above.
6. Kowalski's position on CFR: both sides stated (section 4).

8. Not changed and still open

Verdicts 1b, 1d, 2, 3, 5, 6 unchanged. Not obtained, in priority order: the CNAM and ICCF-12 reports, the Naudin nfrcnam, Hohls, Boulder and 33-run pages, the GEGENE results pages, Naudin's mineral-oil Lifter and unlinked-in-a-box pages, Tajmar, Kossling and Neunzig (2024), Talley's report, and Cornille's Trouton-Noble raw data. Web budget this arbiter pass: 7 of 8 used in total (4 earlier, 3 now); the primary CNAM report was not opened.