Naudin-JL: Physicist review (Step 2)

Reviewer role: PhD physicist. I state the mainstream view, judge each central claim from claims.md, and keep the Investigator's findings in the record whatever my verdict. Tags: [read] = from a document in the folder or the Investigator's saved files; [web] = Internet source with URL; [memory] = recalled, with confidence; [calc] = my own arithmetic from [read] inputs (script at the session scratchpad, calc.py).

Source gap carried forward (from claims.md): Naudin's scope captures, graphs, videos and many sub-pages were not saved, so several verdicts rest on his text description of a result, not on the result. The Investigator has since recovered some of the missing pages (CFR calibration and low/medium-power runs, MAHG test method, EarthTech's replication table, Bishop's MEG paper); I used them. Still missing: the 33 CFR run data, MAHG per-run tables, the Gravitec vacuum video, Talley's reports (I rely on the NASA/ISR quotation of Talley), Stein's Purdue paper, the Dstl Lifter report, all USPTO/EPO office actions.

Stance on evidence (SOURCE HONESTY): Naudin, Bearden, Bedini, Steorn, Gravitec, Buhler and the patent applicants are interested parties; their descriptions are reported as theirs. Bishop and the Sisyfos authors are openly hostile critics; their tone is theirs, and I weigh their arithmetic, not their adjectives.

Verdict summary

---

Claim 1: Lifter thrust is not simply ion wind

Mainstream view, stated fairly

The Lifter is a corona-driven electrohydrodynamic (EHD) thruster. A thin wire at several tens of kV ionises air in a thin layer; ions of one sign then drift through the gap toward the foil, pushed by the electric field. They collide with air molecules about 10^10 times a second [read: NASA/CR-2004-213312 p.16] and hand the momentum to the air, which leaves as a jet; the electrodes feel the equal and opposite reaction. The force on the ion cloud, the electrostatic attraction of the foil toward that charge, and the "ion wind" are one thing seen three ways: the electric force on the space charge, transmitted to the gas. It is a one-dimensional formula, F = I d / mu, where I is the current, d the drift distance and mu the ion mobility (about 2e-4 m^2/(V s) for positive air ions [read: Ianconescu p.7]). ARL derives the same expression (its eq. 14) [read: ARL p.24]. This is standard EHD theory [memory, high confidence]. Because it needs a gas to push on, it predicts no thrust in vacuum.

An important qualification that Step 1 under-weights: the "ballistic ion" calculation (ions fly freely, give their momentum to the electrode) is a different and much weaker model, and it is the one ARL ruled out. "Ion wind is 3 orders too small" (ARL) and "calculated ionic thrust is more than an order of magnitude below the observed lift" (a Dstl slide, no method given) refer to that ballistic picture or to an unstated one. They are not tests of collisional EHD.

a. Is the mathematics internally consistent?

Naudin offers no mathematics for the non-ion-wind account, only verbal statements ("attracted toward the polarised dielectric medium"). The mathematics of the mainstream account is consistent: Ianconescu, Sohar and Mudrik solve Laplace's equation for the wire and foil, add space charge, and obtain the lift of a 7 g triangle within an average 6.5% over four points [read: deep/Ianconescu]. Their method uses the Deutsch assumption, which the authors themselves say "violates some laws of physics" in general; it works for the thrust but is not a first-principles proof. ARL's drift scaling gives 16 g for a ~5 g lifter (right order) and says "it is difficult to see how ionic drift enters into the theory" [read: ARL abstract]. The ARL copper-ion figure (eq. 6) is miscalculated: I recompute 0.023 g, not 0.002 g, with the report's own inputs [calc]; the electron figure and both drift figures check. See Corrections.

Checking Naudin's own lift data against the EHD formula [calc, mu = 2.0 to 2.5e-4 m^2/(V s), d = 30 mm = the wire-to-foil distance he states]:

Lifter1 sits below the ion-drift ceiling with room to spare. Lifter v4 sits at the ceiling when only the straight 30 mm gap is used; ions follow curved field lines longer than 30 mm (ARL and ISR both note the geometry factor, ISR +38% [read: NASA/CR p.16-18]), so the real ceiling is higher and the data show no force beyond the ion-drift ceiling. Thrust per watt is 1.4 mN/W (Lifter1) and 3.9 mN/W (v4), i.e. an effective ion drift speed of the order of 250 to 700 m/s, a sensible number for mu*E in a megavolt-per-metre field. Nothing in Naudin's air data requires a new force.

b. Consistent with established experiments?

Air: yes, with ion-drift EHD. Consistent with it: thrust falls with pressure and vanishes below about 300 Torr [read: NASA/CR p.11]; thrust persists in nitrogen and argon, not in helium at the Dstl current and voltage [read: Dstl slides]; the force direction follows the asymmetry and not the polarity (ARL, ISR); an enclosed lifter does not move (Tajmar's sealed box, and Naudin's own statement "a true Lifter fully enclosed in a box will not move" [read: Naudin true-lifter page]). Xie Xuejun's lifter "went down as air was pumped out" [read: replications log], which is what EHD predicts. Naudin's "enclosed box" sentence is the mainstream prediction.

The wire-in-vacuum-tube tests (6 and 7 March 2003) deserve a calculation, because Naudin presents them as proof that "ion wind is not responsible". The setup: a tungsten wire inside a glass bulb at 25 kV, a free aluminium foil 200 x 20 mm x 12 um, grounded. Foil mass = 2700 kg/m^3 x 0.2 x 0.02 x 12e-6 = 0.13 g, weight 1.27 mN [calc]. A charged wire above a grounded conductor attracts it, with no current and no ion wind needed. Line charge of a wire radius r at height h over a grounded plane: lambda = 2 pi eps0 V / ln(2h/r); force per length = lambda^2 / (4 pi eps0 * 2h); times 0.2 m [calc]:

The electrostatic pull equals or exceeds the foil's weight (1.27 mN) for every plausible case. A micro-ampere leakage current at 25 kV is 25 mW, which is just the supply's dissipation. So "the Lifter hovers with only 25 mW" is the statement that a 0.13 g foil can be held by a static force; it is not evidence of a non-ion-wind thrust. Naudin's conclusion that "the attraction force of the plate towards the wire is stronger than the attraction of the wire towards the plate" contradicts Newton's third law for a pair of electrostatic forces; the wire apparatus was "handled by myself", so the reaction went into his hand. In the linked version (910 g on a torsion suspension) no force, torque or angle is recorded [read: page]. The sealed tube does not remove the other sources: the exposed HV leads and tube ends at 25 kV (corona is audible only above a threshold; "no hissing" does not exclude micro-corona), and the net charge on the whole assembly attracted to grounded surroundings. NASA's contractor report treats the image-charge force explicitly [read: NASA/CR p.14-15]. The page's inference that "the surrounding dielectric is required" and the lab is the "external referential" is, in effect, momentum exchange with the outside world, which the mainstream also says; it does not identify a new force.

Vacuum: the record is this. - Talley (Veritay for the Air Force, PL-TR-91-3009): "under high vacuum conditions ... no detectable propulsive force was electrostatically induced"; he attributed the force he did see to electrostatic interaction with the chamber [read: quoted in NASA/CR p.6]. Report itself not obtained. - MSFC (TM-2004-213283): "No NACAP performance was observed in vacuum" [read: Investigator note]; ISR (AIAA 2004): "no force was measured in a vacuum" [read: Investigator note], although the ISR rig could not resolve less than one-sixth of a turn and had no torsion balance [read: NASA/CR p.11], so its null has no stated upper limit. - Campbell's own US 6,775,123 says thrust "is substantially and significantly reduced in vacuum" and adds a gas supply for vacuum use [read: Investigator note]. - 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]; pressure not confirmed. - Positive: Brown's letter (undated text, quoted by Naudin to ARL, Appendix B): vacuum above 10^-6 Torr, "thrust (towards the positive)" at 70 to 220 kV, lab notes "never published and are not available to me now" [read: ARL p.35-37]. The letter is by the inventor, transmitted by Naudin, unverified by ARL, and the notes and a film it mentions have not been located. Gravitec/NSSTC video (3 July 2003, held back "for confidential reasons" until 2007): two asymmetrical capacitors on a rotating arm at 1.72e-6 Torr, 45 kV, rotation "weaker than at atmospheric pressure" [read: Gravitec page]; no force, angle or current is given and the video is not in the folder. Stein (Purdue) at least 0.31 mN at 17 kV in 1.3e-3 Pa, as quoted by Losonc [read: deep/Losonc]; Stein's paper was not obtained. Buhler/Exodus (US 11,511,891 B2) and Porcelli (dielectric capacitor, force toward the small electrode) are other positive reports by others [read: Investigator notes]; the first is by a party with a commercial stake.

What a clean vacuum null needs to beat. Ballistic electron emission at 45 kV and 10 to 100 uA gives only 7e-9 to 7e-8 N; ballistic nitrogen ions give 1.6e-6 to 1.6e-5 N [calc]. A torsion arm resolves micro-newtons, so an arm rotating at 10^-6 Torr cannot by itself separate those from (i) electrostatic torque between the HV leads, the arm and the chamber wall (Talley's finding), (ii) outgassing, which keeps a corona alive near 10^-6 Torr (Tajmar's hypothesis, unmeasured), and (iii) field-emission currents. The Gravitec description has none of the controls (current, gas load, leads shielded, polarity reversal, shielding the HV feeds).

Gravitec and MSFC both worked in Huntsville in 2003-2004; their results (rotation vs no performance) have not been reconciled in any record I have. This is a lead, not a finding.

c. Has it been tested?

Air thrust: tested often, by ARL (3 lifters and a wire, 30 to 37 kV [read: ARL]), Dstl, ISR, MSFC, Ianconescu et al. (model and experiment), Naudin and his 351 self-registered builders. No test of the "not ion wind" statement has found a residual after the ion-drift estimate; ISR reports the model reaches 77% of the measured force for its best device and flags that its currents, force and gap cannot be reconciled from the report [read: deep/NASA]. Replication of Naudin's own vacuum-tube tests by others: none on record. Vacuum thrust: see above.

d. If true, what would change?

A force on an isolated, electrically driven body in vacuum with no exhaust would violate conservation of momentum unless momentum is stored in a field or exchanged with a distant reservoir. Pulling that off needs a new coupling of electromagnetism to inertia or gravity (Musha's zero-point-field proposal, Brown's "electrogravity", Williams' 5-D theory cited by Tajmar). Musha's paper contains one relayed Honda R&D measurement (0.9 +/- 0.63 g weight loss, scatter up to 1.9 g, in oil, AC pulses) [read: deep/Musha]; it is a single relayed source, with large scatter, and the errors are as large as the effect. I do not consider it a test. Tajmar's own 5-D box test found no force above about 10 uN in air [read: deep/Tajmar].

Verdict, Claim 1

• Air, "main thrust is an electrokinetic effect on the medium": this is the EHD account; as stated by Naudin ("not a simple ionic wind") the contrast is largely semantic; the claim that it is not ion drift is Tested, not supported.
• Wire-in-vacuum-tube, "proves ion wind is not responsible": Refuted by calculation as evidence (static attraction 1 to 44 mN vs 1.27 mN weight, [calc]).
• Vacuum Biefeld-Brown thrust at 45 kV: Never properly tested with a calibrated, shielded, cleanly outgassed balance; where tested (Talley, MSFC/ISR, Tajmar-Schreiber at lower voltage) it was Tested, not supported. The positive reports are single-source or unreleased and are kept in the record.
• Losonc (net electrostatic force zero in vacuum for Naudin's lifter 1, then "electro-gravity" by elimination): the calculation is plausible and, by Newton's third law, close to guaranteed for pairwise Coulomb forces on a closed system (deep read inference, which I agree with); the inference to "electro-gravity" does not follow, since vacuum currents, emission and leads are not modelled, and his scaling sentence contradicts his own numbers (F2 ~20 mN is larger, not smaller, than 3.6 mN). Not testable as stated for the elimination step.

---

Claim 2: Lifters are real, reproducible and scale

Mainstream: real and reproducible; the same physics as ionocraft and EHD thrusters (de Seversky 1964 patent; Brown 1928-1960 patents) [read: ARL]. ARL built three lifters and a wire that lifted at 30 to 37 kV [read: ARL p.15-19]; Ianconescu et al. matched lift measured to 6.5%; Dstl lifted in air and nitrogen, not helium [read: Dstl]. The folder's list of 351 builders is self-registered, with no independent verification (SOURCE HONESTY: one self-reported list, not nothing).

a. Mathematics: Naudin's claims are numbers, not derivations; arithmetic checks: Lifter1 23.9 W = 41.9 kV x 570 uA ✓ [calc]; Lifter v4 81.4 W = 40.5 kV x 2.01 mA ✓. b. Consistent: yes (table above). c. Tested/replicated: yes, by ARL, Dstl, Ianconescu, ISR (similar devices) and many builders. d. If it were true: nothing changes; it is mainstream EHD.

Scaling is where the claim fails. Naudin's own table gives 4 to 13 g of device for 24 to 133 W. The ideal momentum-theory (actuator disc) hover power for 0.314 N over his 0.57 m^2 footprint is T^1.5 / sqrt(2 rho A) = 0.15 W [calc]; the lifter uses 81 W, about 500 times more, because the ion drift speed (hundreds of m/s) is high and the efficiency limit is thrust/power = 1/v_drift. Lifting 1 kg would take about 3.8 kW at v4's 3.9 mN/W [calc]. Static-lift craft cannot scale. The "VTOL craft" proposal is therefore Refuted by calculation for practical payload.

Verdict, Claim 2: real and reproducible, Established. "Scales": Established only for small devices; Refuted by calculation as a route to a practical aircraft.

---

Claim 3: Naudin's MEG replication gives an apparent COP above 1

Mainstream view

The MEG (Bearden et al., US 6,362,718) is a permanent magnet in a ferromagnetic core with two input and two output coils; the inputs are pulsed alternately. A transformer with a DC magnetic bias is ordinary engineering (magnetic amplifier, flux-switching). Energy conservation for the whole device (Poynting's theorem over a closed surface) says that the average power out equals the average power in, plus the rate of change of stored energy. The only reservoir that could provide net energy is the magnet. It is small: for Bearden's patent magnet, 40 x 25.4 x 38.1 mm = 38.7 cm^3, the available field energy is of the order (BH)max x V: about 12 J for NdFeB (300 kJ/m^3, [memory, high confidence]) and about 1 J for ferrite (about 30 kJ/m^3, [memory, high]) [calc]. Naudin used ferrite magnets [read: v2.1 page]. A claimed net gain of 3.5 W (his 6.76 W out vs 3.25 W in) would exhaust 12 J in 3 s; 7 W (run #2 as Horký reads it) in 1.6 s [calc]. Jelinek and Pekarek reach the same order ("a few tens of joules" / "tens of seconds" at 5 W) [read: deep/Jelinek]. Smith's 30 J estimate from electron-spin precession is an unmeasured speculation [read: Smith page]; even taken at 30 J, 7 W lasts 4 s. Nothing in the folder reports a run long enough to reach that energy.

a. Is the mathematics internally consistent?

Three internal checks on the numbers.

1. Power from "RMS watts". Bearden's p.67 scope numbers are "M RMS 3.588 W" in and "M RMS 12.7 W" out. The RMS of the instantaneous-power trace is not the mean power. For a sinusoid into a resistor p(t) = P(1 - cos 2wt), whose RMS is sqrt(1.5) P = 1.22 P [calc]; correcting only the output gives 10.4 W, a ratio of 2.9 per coil, not 3.5 [read: deep/Bearden]. Tektronix support (via Flickinger, 2002) calls the RMS of a math waveform "bogus"; Naudin put a warning about the THS720P on his own page [read]. The 1996-style mean displays ("PWR OUT M Mean 6.764 W", "cMean 10.28 W") in Naudin's later screens avoid that particular error [read: deep/Horky].

2. Ohm's law cross-check on the load. This is the check that decides the matter, and Bishop's paper (now obtained) contains it, separate from his input formula. Naudin states the load is a "conditioned" 100 kohm carbon resistor [read: v2.1 page]. In the v3.0 test run #2 (26 Nov 2000) the output is 1.024 kV peak-to-peak (amplitude 512 V). Power into 100 kohm is V_rms^2/R = (362 V)^2 / 1e5 = 1.31 W [calc]. Naudin's screen shows 10.28 W, which needs R_eff = 12.8 kohm. Bishop's reading of the v3.0 screen is a 500 V amplitude and a 44 mA amplitude; V/I = 11.4 kohm, power 11 W against 1.25 W for 100 kohm: a factor 8.8 [calc; read: Bishop p.9]. Bearden's own figure gives V_rms/I_rms = 693 V / 15.44 mA = 45 kohm [calc], load not stated. So either the "conditioned" resistor had an effective value near 10 kohm under 1 kV at 20 kHz (no one measured it), or the current channel (a 10 ohm shunt read through a 1:10 probe, at 0.1 V for 10 mA) over-reads by about 9 times. Either explanation can be tested in an afternoon by measuring the resistor at operating conditions and its heating.

3. Input side. Bishop's input formula (constant 28 V times I_0/sqrt 2 = 2.47 W) is wrong: a constant voltage times a symmetric AC current averages to zero unless the current has a DC part. Smith is right on that point [read: Smith page], and Bishop's input figure should not be cited. The error is separate from the output-current mismatch above, which stands. Horký's run supports neither side cleanly: his "input power" subtracts an undefined no-load current I_0 [read: deep/Horky], and on full supply power his own table gives COP 2.53 to 1.37 for both lamps, 1.26 to 0.69 for the one lamp actually measured (the second lamp's value is identical, so it was doubled, not measured) [read: deep/Horky; calc]. His output is V_rms x I_rms in volt-amperes, with no power factor.

b. Consistent with established experiments?

A passive magnet-biased transformer driven by a switching stage cannot give net power on average. Naudin's own observations point the same way. In his status note of 6 Dec 2000: "in most of cases the 'apparent' power measured seems greater than the heat dissipated by Joule's effect in the RLoad" and "this makes me pessimistic about the calorimetric tests results"; his 2008 e-mail adds "the RLoad resistor doesn't heat up too much, it is only a bit warm ... a possibility of a measurement artifact" [read]. For a resistor all electrical power becomes heat; heat below the electrical reading means the reading is too high. His explanation that "most of the power is radiated in EM form" cannot hold: radiated power from a magnetic dipole of 1 A m^2 (generous for a 10 cm device) at 20 kHz is mu0 w^4 m^2 / (12 pi c^3) = 3e-13 W [calc]. A 2.8 mT field close to the load and 1250 V/m at 50 cm are near-field (reactive) fields that carry no net power outward. Flickinger's observation (that COP>1 appears only with non-linear loads such as varistors) [read: research note] and Horký's ("linear load gives very low output; non-linear load, output rises") fit a measurement problem with non-sinusoidal waveforms, as Sisyfos (Jelinek and Pekarek) argue [read: deep/Jelinek], but a non-linear load can also change the real waveform, so the pattern alone does not prove artifact.

Jelinek and Pekarek's Appendix 1 (the magnet does not matter because the permanent-magnet flux is constant) holds only for a linear core, as the deep read finds; a saturable core biased by a magnet is a magnetic amplifier. Their Appendix 2 (Poynting theorem: mean out equals mean in) is sound in outline: the energy bookkeeping does not depend on the core being linear, only on the stored energy being periodic. That is the argument that carries the weight, together with the magnet's small energy.

c. Has it been tested?

• Bearden et al. (interested parties): scope readings and plots, open loop; they state their own build-up was the replication [read: deep/Bearden p.1]. A "COP 7.3 in a third build" appears in Bearden's e-mail of 2 Dec 2000 [web: https://zpenergy.com/modules.php?name=News&file=print&sid=3918 via Investigator]; no Naudin page in the folder shows 7.3 (they show 1.75, and screens with 3.2 to 3.5 by arithmetic), so it is Bearden's figure, not found in the primary pages here.
• Naudin: waveforms matched; COP 1.75 (7 Nov 2000); loop not closed ("the device stops after a while" [read]); calorimetry not supporting.
• Horký (2009, school project, believer, not peer-reviewed): COP 1.57 to 2.89 as printed, 0.69 to 1.26 on one lamp and full supply power (my recomputation from his numbers); no heat measurement.
• Bishop, Sisyfos: analyses of Naudin's and Horký's data, not experiments.
• Promised DoD-linked university replication and a certified-laboratory test (Bearden, December 2000): not found in any record read. Competing explanations: never happened; happened and negative; happened and kept confidential. Deciding records: DTIC or university reports naming the MEG; Bearden's "Correspondence" pages after 2001. Patent kept alive to term (fees paid to 2013) and pledged to Colmac Holdings Ltd (Canada) in 2006 [read: Espacenet note], which shows continued commercial interest but no product, licence or test report.

I count this as never completed: no calorimetry, no closed loop, no independent power analyser measurement.

d. If true?

A net energy gain from a magnet-biased transformer would require a new energy source (Bearden: "active vacuum", Whittaker potentials, "regauging"). It would contradict energy conservation for a closed electromagnetic system, which is checked in every transformer ever built to better than a percent. The Poynting-vector argument (Bearden's "Heaviside vs Poynting" energy-flow papers, 10-trillion-times figure in the Fig. 5 caption [read: deep/Bearden]) does not change the integral of S over a closed surface.

Verdict, Claim 3. Split: (i) "an apparent COP above 1 on a scope": reproduced as readings, but Never properly tested for being real power; Naudin's own heat observation and the load-resistance mismatch point to measurement error; the decisive tests (calorimetry on the load, mean vi from a power analyser with probe deskew, closed loop) were not done by anyone on record. (ii) Real net energy gain: Refuted by calculation* (the magnet's energy of 1 to 12 J against a claimed several watts). A closed loop would settle it in seconds; its absence 24 years after the first claims is itself evidence (it was dismissed by many, but also never achieved by those who believe in it).

---

Claim 4: The Mizuno-Ohmori plasma-electrolysis "Cold Fusion Reactor" gives excess heat

Mainstream view

A tungsten cathode in hot K2CO3 solution at 150 to 300 V forms a vapour sheath with a glow discharge. Heat comes from the electric power; measuring it needs the vaporised water, the water heated, and the losses. Ordinary calorimetric traps in this arrangement: (i) mass lost by splashing and aerosol, which carries only about 84 J/g of sensible heat (4.18 J/g/K x 20 K) rather than 2260 J/g latent heat, a factor 27 [calc], so counting every gram lost as evaporated overstates the output; (ii) energy counted twice; (iii) incomplete accounting of the H2 and O2 leaving. No known chemical or nuclear process explains a reproducible 1.5 to 2.5 power gain at 100 W scale without radiation or products; Mizuno himself writes that the effect is rarely replicated: "In most cases, no excess heat or hydrogen is observed" [read: Mizuno 2004, via Investigator].

a. Mathematics

I verified Naudin's arithmetic from his pages [read: Investigator note, jlnlabs.online.fr/cfr/html]: - Calibration (resistive heater, same beaker): 297.5 W x 190 s = 56,525 J in; 8 g x 2260 + 600 g x 15 K x 4.18 = 55,700 J out. Ratio out/in 0.985; the page prints "56525 / 55700 = 0.985", which is in/out (1.015) [calc]. Trivial slip. The calibration validates the method with a heater, not with the plasma running. - Low power (glow only): 148.4 W x 499.2 s = 74,081 J in; 26 g x 2260 + 150 x 4.18 x 23 = 73,181 J out: 0.988 [calc ✓]. - Medium power (incandescent cathode): 289.2 W x 82 s = 23,714 J in; 10 g x 2260 + 150 x 4.18 x 21 = 35,767 J out: 1.51 [calc ✓]. Notice what the input could do by itself: after the sensible heating (13,167 J) the input leaves 10,547 J, enough to evaporate 4.7 g, not 10 g [calc]. The "excess" is therefore five-plus grams of mass loss in 82 s, with no way, in this page, to say whether it left as vapour or as droplets. - Run 28 (Kowalski quoting Naudin): 170 g evaporated, 3093 W of heat from 1200 W: 170 g x 2260 = 384 kJ requires at least 124 s at 3093 W [calc]. Run duration is not in the folder.

b. Consistent with established experiments?

EarthTech's instrumented replication of Naudin's reactor (Little and Luce, 15 July 2003) is the key source [web: https://web.archive.org/web/20030810115339/http://www.earthtech.org:80/experiments/Inc-W/2003/replicationjln.htm; saved in research/files]. Their final COP2 averages 0.98 over the runs with stirring [read]; the table includes a column "COP with Cooldown" computed Naudin's way, which includes the energy of evaporation after power-off. I averaged that column over the 15 runs: 1.50 (1.27 to 1.77) [calc]. That is Naudin's reported level ("typically around 1.6, sometimes exceeding 2.5" [read: EarthTech intro]). With the same data the whole excess disappears when the cooldown evaporation is dropped, which is correct on first principles: the heat that evaporates water after power-off is the heat already stored in the solution (the E heating term), so counting it again double-counts. EarthTech state this and note Naudin's justification was not understood [read]. Kowalski, Little and Luce (ICCF-12, 2005): mean COP 1.00 +/- 0.02 in a Mizuno-type cell; splashing and foam droplets named as error sources [web: https://lenr-canr.org/acrobat/KowalskiLsearchingfa.pdf, per Investigator].

Naudin's own calibration did not include a plasma-mode blank. His page states the incandescent cathode condition for excess heat, "only at this moment, the heat produced is greater than the electrical power input". The fraction of the input evaporated jumps in that regime: in the low-power run 79% of the input went into evaporation, in the medium run 95% of the input equivalent (22,600 J of 23,714 J) while the sensible heating is not small either. A transition to violent boiling and spray is the regime where droplets are expected.

c. Tested?

Yes. Independent: EarthTech (instrumented, balance read to 10 mg, power meter, stirrer, thermistor): no excess. Kowalski: Naudin's 33 runs were called "reasonably reproducible" in his 2003 comment, with the warning that doubts raised about Mizuno's data "can also be raised about Naudin's data" [read: Kowalski item 73]; he later co-authored the null result above. Naudin lists replications at CNAM (Paris, 2006), South Africa and Boulder (self-reported on his index; no report in the folder). Mizuno's own later statement is that replications are few. Kowalski also took part in the null.

d. If true?

Reproducible 1.5 to 3 power gain from light water and tungsten without radiation or products would need a new heat source (Naudin: atomic hydrogen recombining on tungsten, "a real cold fusion effect"); the first term is chemistry that returns the energy spent to dissociate (see claim 5); the second would require nuclear reactions with products and radiation that no one has measured here (Naudin measured none for this device in the folder). The atomic-hydrogen account cannot give energy beyond the dissociation energy supplied.

Verdict, Claim 4: Tested, not supported. The specific source of Naudin's excess is identified (cooldown double-counting reproduces about 1.5 from EarthTech's own data, plus unquantified droplet loss), so for Naudin's accounting it is Refuted by calculation. This does not rule out Mizuno's original work, which is not before me; Mizuno reports it as rarely replicated.

---

Claim 5: Moller's Atomic Hydrogen Generator (MAHG), COP 5 to 21

Mainstream view

Langmuir's atomic-hydrogen torch dissociates H2 on a hot tungsten filament (about 4.5 eV, 436 kJ/mol [memory, high]) and recovers the same energy where the atoms recombine on a cooler surface. It transports heat; it does not make it. Hydrogen is recycled, so no fuel is consumed, but nothing net is created: the electrical input is all there is. To deliver 875 W (Naudin's long run) by recombination alone, 2 mmol/s of H2 (4 mg/s) must be dissociated and recombined; that is where the heat is moved, not made [calc].

a. Mathematics

The calorimeter formula is fine: P_out = flow (m^3/h) x dT x 1161 W per m^3/h K [calc: 4.18e6/3600 = 1161]. The input definition varies: the data sheets list "Rms Voltage and Rms Current" [read: pultests page]; for a purely resistive filament RMS V x RMS I equals the mean power, whatever the pulse shape [calc, algebra]. For the 5% duty-cycle runs with a battery and MOSFET switch the page adds that the "average voltage and current" were also read on the scope but does not say how they entered the efficiency. If input were taken as (mean V) x (mean I), a pulse train of duty cycle D understates input power by the factor D, and apparent COP is inflated by 1/D: 20 at 5%, 10 at 10% [calc]. Naudin reports COP 21 at 5% duty and "up to 1153%" at lower duty and says "efficiency increases ... while the duty cycle decreases" [read]. The match with 1/D is striking but I cannot prove it; the page does not state the method. I record it as the first thing a reader should check from the raw scope data.

b. Consistent?

Naudin's own control contradicts a clean calorimeter: pure-DC "blank" runs (38 to 40) returned 64% (he attributes 36% to losses) [read: Investigator note]. A bench that recovers 64% of the DC input is not calibrated to read 100%; against that baseline the pulsed 224% would be 3.5, and Frolov's earlier set-up gave 82% with smoothed DC but 124 to 250% with pulses (he flagged "a possible mistake") [read: mahg1 page]. The same pattern, honest DC below 100% and pulsed above, appears in two independent set-ups, both from interested parties and both with the pulsed input measurement as the likely weak point. Calorimeter issues at low input: 10 W input with COP 11 means 110 W out, only 0.95 K of rise at 0.1 m^3/h [calc]; a 0.5 K thermocouple offset drift (type K standard tolerance is of that order [memory, medium: ±2.2 K or 0.75%; not re-checked]) equals 58 W at that flow. Naudin notes warm-up transients ("the power output was stable after 2 hours").

c. Tested?

Run by Naudin and by Frolov (Faraday Lab, the originator, hired by Moller, who had a stake). No independent replication found (the Investigator's search for a Frolov "FMAHG" report in New Energy Technologies no. 3 (2005) is unreached). Kowalski comments but does not accept or reject: "a chemist would say the dissociation-recombination cycle cannot give net energy" [read: Kowalski item 289]. Naudin's Geiger check during run 87 found "no anomalous radiations" [read].

d. If true?

A closed hydrogen-tungsten cycle yielding net energy would violate the first law, unless hydrogen on hot tungsten has some non-chemical reaction; nothing in the folder supports one.

Verdict, Claim 5: mechanism as an energy source, Refuted by calculation; the reported COP numbers, Never properly tested by a neutral party, with the likely error sources (pulsed input measurement, thermal transients, an uncalibrated bench: DC control 64%) named above.

---

Claim 6: The Cornille-Naudin patents

FR 2 789 818 A1 (1999), fast charger

Mainstream: charging a capacitor from a fixed voltage source through a resistor wastes half the energy, whatever R (the patent states this correctly in energy terms; its "half the charge" wording is loose [read: deep/FR patent]). Wasting half is a property of resistive charging, not of capacitors. A switch-mode converter (inductor or transformer, switch, current limit) charges a capacitor to a target voltage with efficiency of 90% or more [memory, high confidence], which is what this patent describes (a self-switching chopper driven from its own charge current). A charge time of 30 s versus 83 s (2.77 times; the patent says 2.5 and "divided by 3") for a 1.3 F bank [read] simply reflects a higher average current. No claim of overunity. The preliminary search report cited prior art as X/Y against claims 1, 2, 4, 5 [read]. Efficiency definition and method are not given in the patent. Verdict: Plausible, untested for the numbers; Established in principle (not novel in principle).

US 2005/0057116 A1 (priority 2001): "efficiency more than one"

What it claims [read: deep/US patent]: charge a plate capacitor; the plates face plasma-filled tubes; when the plasma is conducting the capacitor is two thin glass capacitors in series (C1); when the plasma is switched off the capacitance falls to C2 = C1/beta (beta = 590 in the 30 cm example); at constant charge V2 = beta V1 and stored energy E2 = beta E1. In the mechanical analogue (pulling the inner plates out) the patent itself says the observer's work pays for the gain and "the efficiency of the operation cannot exceed 100%" [0004]. In the plasma version it says the work is "almost free" because Coulomb forces act over a few Debye lengths [0031]-[0032].

a. Mathematics. The patent's arithmetic checks: C = 3.5 nF, E1 = 8.75e-4 J, E2 = 0.175 J, 0.175 / (2 x 8.75e-4 + 0.050) = 3.38, i.e. 338% [calc]. This is a calculation from an assumed multiplier of 200 (reasonably taken, in its words) and an assumed 50 W plasma for 1 ms. With beta = 10, which the patent itself calls a possible real value, the same formula gives 0.17, below 1 [calc; read: [0064]]. The pressure example (136 Torr) does not reproduce; the printed inputs give about 34 Torr [read: deep]. Multi-cycle growth beta^(2n-1) (formulas 22-28) requires the capacitor to be "charged with the voltage V2" after relighting the plasma; with the source disconnected, charge conservation gives V1 back, not V2; the extra charge has no source [read: deep]. Three different accounts of where the energy comes from, none reconciled, none in the claims: work over Debye lengths; "the surrounding magnetic ether"; the Earth's electrical potential energy (formula 32, 700 uF x several million volts). With the accepted global-circuit potential of about 250 kV (range 200 to 300 kV) [web: https://glossary.ametsoc.org/wiki/ionospheric-potential/] the stored energy 1/2 C V^2 is 22 to 32 MJ for 700 uF [calc], but nothing connects a laboratory capacitor to it with a source of that size (fair-weather current is picoamps per square metre [memory, medium]). Internally inconsistent.

b. Physics. The charges induced on the plasma side of the glass are held against the plates by their mutual attraction. When the ionising supply is removed, those layers do not leave by themselves; surface charge on dielectric walls after a discharge is a well-documented memory effect in barrier discharges [web: search results, e.g. https://hal.sorbonne-universite.fr/hal-03541456v1/document, and the Greifswald work listed there]. If they stay, the field in the middle gap stays near zero and the voltage does not rise. For V to rise beta-fold the induced charges must be physically removed from the plates' vicinity, and that costs the energy that appears as the increase; the patent assigns that cost to nothing measured. This is the central physical objection: the energy has to come from somewhere, and the work of detaching the induced charge is an input.

c. Tested? One trace only (Fig. 8): 500 V rising to 1750 V when switch K1 opens, read on a 1 gigaohm probe, decaying over some hundreds of ms [read: deep]. No input energy, no output energy, no capacitance, no repeats. My inference: a decay of ~100 ms through 1 Gohm suggests about 100 pF, and about 0.15 mJ at 1750 V [calc, inference]; the pre-jump decay (2000 V to 500 V in 70 ms) is as fast, so a leakage path besides the probe is present. A 3.5-fold voltage rise at constant charge is exactly what opening a switch that removes a series capacitor does; it is not an energy gain.

Record: FR2822590 granted in France (B1); US application abandoned 27 June 2006 "failure to respond to an Office action"; EP1389360 deemed withdrawn 2008 for non-payment of the 2007 renewal fee; the applicant AES (SIREN 419 785 985) filed last accounts for 2015, 0 employees, struck off 8 March 2021 [read: Investigator notes]. Competing explanations (Absence of evidence): cost; a substantive objection the applicants chose not to contest; loss of interest or retirement. Deciding records: USPTO file wrapper for 10/472,714; the EP search report; the PCT international search report (WO02075912). Not reached.

d. If true: a device producing more electrical energy than it consumes in a closed cycle; energy conservation would need a new reservoir. Co-inventor Cornille's own later "stimulated forces" theory (absolute motion relative to the vacuum, violation of Newton's third law by the magnetic force [read: Cornille 2020 review via Investigator]) shows the same departure: it would change electrodynamics' treatment of Newton's third law and Lorentz invariance (the Trouton-Noble experiment is a null in mainstream physics [memory, high]). The Cornille-Naudin Trouton-Noble capacitor "locked East-West" over dozens of repeats (1996-1999) [read, Cornille 2020] is a reported anomaly, single-source, for which the folder has no raw data; an East-West alignment is also what a capacitor suspended in a stray ambient field or by a twisted thread would do, and the control (rotated suspension) is not described in what I read. I do not rule it out; it is untested by others.

Verdict, Claim 6: FR 2 789 818: Plausible, untested (and ordinary). US 2005/0057116: Internally inconsistent (unreconciled energy sources; charge growth in the multi-cycle formulas); the energy claim Never properly tested (no energy measurement).

---

Claim 7: Other devices presented as overunity or "regauging" machines

Mainstream view for all: a system with an energy input that is measured, a load whose power is measured, and no hidden storage must satisfy output <= input over a cycle; claims of more need an unaccounted input or a measurement error. Where the folder has no input/output numbers, there is nothing to test.

Interested-party statements (Bedini, Steorn's CEO, Gravitec's Serrano, Bearden) are kept as theirs. Naudin's own pages in this group are mixed: sometimes (Magnetic Gate, "4.75 is not a COP", MEG "no overunity" in the Magnetic Amplifier Experiment) he states there is no overunity, and he labels measurement artifacts possible. That candour is part of the record and is to his credit.

---

Investigator's findings I weighed (kept whatever my verdict)

1. Bearden's e-mail of 2 Dec 2000 (COP 7.3, DoD-linked university replication and certified laboratory test promised; names withheld): kept; no corresponding record found; interested party. Bears on claim 3.
2. Cheniere/Bearden status page, Craddock 2021 ("replicated by a number of researchers", DOE website once listed it): kept; the replication list was not obtained.
3. MEG patent: lapsed 2006 and reinstated; security agreement to Colmac Holdings Ltd (Canada) 2006; fees paid through 2013. Kept; shows continued interest, not test results.
4. Bishop's paper (obtained): input formula wrong (Smith correct); output-current mismatch right.
5. Flickinger (non-linear loads only).
6. EarthTech replication of the CFR and Kowalski et al. (claim 4): decisive for claim 4; Naudin's reply, if any, is not found.
7. Mizuno 2004 (rare replication), Mizuno and Toriyabe 2005 (an explosive event, estimated heat "800 times higher than input", in a conventional cell): the latter is an unreplicated anomalous event, kept; not a test of Naudin's CFR.
8. Naudin's MAHG method and the DC-blank result, Frolov's smoothed-vs-pulsed pattern: weighed in claim 5.
9. NASA/ISR/Campbell vacuum records, Tajmar and Schreiber 2020, Porcelli 2016, Buhler/Exodus (US 11,511,891), Cornille 2020 review: weighed in claims 1 and 6. Porcelli and Buhler are positive reports by others (the second with a commercial stake); neither is a test of Naudin's device.
10. Patent and company records (FR2822590 granted; US abandoned 2006; EP lapsed 2008; AES struck off 2021): kept under claim 6, with the competing explanations.
11. Later Naudin work (BiTT, ReGenX, GFS UAV; Slideshare): not reached; not judged.
12. Steorn's jury verdict (web-checked).

Web verification (budget 6; used 3)

• Steorn jury June 2009: retagged [web] (Wikipedia, Neurologica); the original jury statement itself was not opened.
• Earth-ionosphere potential about 250 kV (range 200 to 300 kV), not "millions of volts": [web: https://glossary.ametsoc.org/wiki/ionospheric-potential/].
• Wall-charge memory in barrier discharges: [web: hal.sorbonne-universite.fr result list]; abstracts only, not full papers.
• Not re-checked and still [memory]: (BH)max values (high); H2 dissociation 436 kJ/mol (high); type K thermocouple tolerance (medium); switch-mode charger efficiency (high); fair-weather current density (medium).

Corrections to Step 1

1. claims.md claim 1 and facts.json: ARL did not "explain air thrust by ion drift and momentum transfer to air." ARL's own abstract says the physical basis "is not understood"; its drift scaling gives the right order and it says "it is difficult to see how ionic drift enters into the theory" [read: ARL p.4]. Ianconescu et al. and ISR (NASA/CR-2004-213312) are the sources that claim an ion-drift explanation. Martins and Pinheiro (simulation) say the main force is electrostatic on the electrodes and the ion wind is its reaction (94 to 98% electrostatic [read: deep]); this is the same EHD force described from the other side, not a rejection of EHD.
2. claims.md claim 1: "Dstl: calculated ionic thrust ... more than an order of magnitude below the observed lift" is a slide statement with no method shown; it should not be set beside Ianconescu's 6.5% match without saying so.
3. claims.md claim 3, Smith's reply: "Bishop used a wrong power formula" is half right. Bishop's input formula is wrong (constant voltage times AC amplitude / sqrt 2; Smith correct). Bishop's other point, that Naudin's output current (44 mA) is about 9 times what Ohm's law gives for the stated 100 kohm load (5 mA), is independent, arithmetically correct on the figures in Bishop's paper, and not answered in what is in the folder. Smith's rebuttal ("the load was conditioned") is a possible answer but means the load was not 100 kohm.
4. claims.md claim 3: Horký. The deep read already corrects the digest; I add that on full supply power Horký's one measured lamp gives COP 0.69 to 1.26, and the second lamp's power is a copy of the first (identical to the last digit), so total COP 1.37 to 2.53 on full supply power is an assumed doubling [calc from deep/Horky Tab. 1].
5. claims.md Source gap: Several pages named as unobtained were recovered by the Investigator (CFR calibration, low/medium-power run numbers, MAHG test method including the DC blank, Bishop's paper, the EarthTech replication). The gap now is the CFR 33-run table and the MAHG run tables.
6. claims.md claim 4: the sentence "EarthTech" should be added: independent instrumented replication of Naudin's own reactor found COP2 0.98 and reproduced his ~1.5 only by his cooldown addition. Claims.md lists only Kowalski's web commentary, which is favourable.
7. claims.md claim 5: the Investigator's finding that the MAHG blank (pure DC) gave 64% and that the 1153% runs used a different supply and a different input measurement should be in the claim text; as written the reader sees 2.43, 5.26, 21 without the control.
8. facts.json, Lifter pages: Naudin's "unlinked Lifter" page prints the current as "1A"; as printed the figure is a lost micro sign (microamps); with 25 kV and 25 mW the current is 1 uA. Step 1 noted this ("evidently microamps"); I confirm it by arithmetic: 25 mW / 25 kV = 1 uA.
9. Bearden's "COP 7.3" (Investigator, via ZPEnergy) is a figure in Bearden's e-mail and in no page of Naudin's in the folder; claims.md's list of Naudin's numbers (1.75; 6.76 W / 3.25 W) is correct. Fig. 12-style screens give 3.2 by arithmetic (run #2: 10.28 W / 3.179 W [read: deep/Horky]); I computed that, Naudin does not state it.
10. ARL eq. 6 (0.002 g): confirmed miscalculated; correct value 0.023 g [calc].