Naudin-JL: Primer

Jean-Louis Naudin's JLN Labs replications (1996-2013): the "Lifter" asymmetrical capacitor, the MEG transformer, the plasma-electrolysis and atomic-hydrogen reactors, and two Cornille-Naudin patents. How to tell which forces and which energy numbers are real.

Source tags: [read] = from a document in the project folder (the _text/ extraction, a deep read in deep/, or a file the Investigator saved in research/); [web] = Internet source, URL given; [memory] = recalled from training, with a confidence level; [calc] = arithmetic redone for this primer and reproduced by example.py. Verdicts come from arbiter.md, including its "QE revision (2026-10-09)" section; this primer adds none.

Terms and acronyms used below. COP (coefficient of performance) = useful energy out ÷ energy put in, both measured over the same run; a COP above 1 ("over-unity") would mean more energy out than in. "Apparent COP" is a COP computed from instrument readings that have not been cross-checked. EHD = electrohydrodynamics. ARL = US Army Research Laboratory. Dstl = UK Defence Science and Technology Laboratory (Ministry of Defence). NASA = US National Aeronautics and Space Administration; MSFC = NASA's Marshall Space Flight Center (Huntsville, Alabama); NSSTC = National Space Science and Technology Center (also Huntsville); ISR = Institute for Scientific Research, NASA's contractor for the 2004 capacitor-thruster report; TM = Technical Memorandum (a NASA report series); TdT = Transdimensional Technologies, a company. DoD = US Department of Defense. VTOL = vertical take-off and landing. MEG = Motionless Electromagnetic Generator (Bearden). CFR = Naudin's "Cold Fusion Reactor", his build of the Mizuno-Ohmori plasma-electrolysis cell. ICCF = International Conference on Cold Fusion (ICCF-12: Yokohama, 2005). CNAM = Conservatoire National des Arts et Métiers, Paris. MAHG = Moller Atomic Hydrogen Generator. AES = Advanced Electromagnetic Systems S.A., the French company that filed the Cornille-Naudin patents. RMS = root mean square (section 2).

Source gap (repeated from claims.md and arbiter.md). Naudin's pages were saved as text. The scope screens, graphs, photographs and videos they point to are mostly missing, and so are many sub-pages that hold the numbers: the 33-run table of the plasma-electrolysis reactor, the per-run tables of the atomic-hydrogen generator, the Lifter v2/v3/"Maximus" pages, the Faraday-cage and enclosed-box Lifter pages, and Naudin's "TT.Brown EKP device in MINERAL OIL" page. Also not obtained: the Gravitec/NSSTC vacuum video, Talley's Air Force reports (AFAL-TR-88-031, PL-TR-91-3009), Stein's Purdue vacuum paper, Brown's vacuum notes, the full Dstl Lifter report (its 2005 slides were read, charts as page images), the US and European patent office actions, the CNAM and ICCF-12 plasma-electrolysis reports, Naudin's pages for the CNAM, Hohls and Boulder replications, the GEGENE test-result pages, Tajmar, Kößling and Neunzig (2024), and Cornille's raw data for the Trouton-Noble experiment. The Investigator recovered some missing pages (reactor calibration and low/medium-power runs, the atomic-hydrogen test method, EarthTech's replication table, Bishop's MEG critique, Kowalski's ICCF-12 notes). Where a conclusion below rests on Naudin's description of a result rather than the result itself, it says so.

Who is an interested party. Naudin promoted these ideas on his site. No sale of devices by him was found. He is co-inventor on two patents assigned to AES, a company formed "to develop and sell electromagnetic generators" [read: research/2026-10-09-pappers-societe.md]; one unopened web snippet calls him its R&D director; Horky (2009) states, citing no source, that Naudin "is now a member of Bearden's team" [read: deep/Horky p.2]; Naudin himself describes "my own and private lab" [read: deep/Kowalski item 73]. His stake is therefore commercial as well as promotional [arbiter.md, QE revision s.7]. Bearden and Magnetic Energy Ltd hold the MEG patent. Moller and Frolov (Faraday Lab) had a contract on the atomic-hydrogen generator. Gravitec (Serrano) and Exodus Propulsion (Buhler) sell or patent capacitor thrusters. Cornille and AES were the applicants of the Cornille-Naudin patents. Their descriptions below are reported as theirs.

---

1. The idea in one paragraph

Jean-Louis Naudin, a French experimenter, ran a web laboratory ("JLN Labs") from 1996 to about 2013 where he rebuilt, photographed and published almost every "new energy" or "antigravity" device that was being discussed, and invited others to copy him [read: page dates across _text/]. His best-known build is the Lifter: a feather-light triangle or hexagon of balsa and aluminium foil with a thin wire strung 3 cm above the foil. Put 30 to 45 thousand volts between the wire and the foil and the whole thing rises silently off the bench. Naudin, following T. Townsend Brown, held that this is not "simple ion wind" but an "electrokinetic effect on the medium", with a small extra thrust that would survive even in vacuum and could push spacecraft [read: _text/2001-Lifter-full-story, p.1]. He also replicated Tom Bearden's Motionless Electromagnetic Generator (MEG), reporting that the scope showed more power out than in; Mizuno and Ohmori's plasma-electrolysis "cold fusion" cell, reporting up to about two and a half times more heat than electricity; and Moller's atomic-hydrogen generator, reporting up to 21 times. The mainstream reading is that the Lifter is real and flies by an ordinary force (charged air molecules pushed by the field and dragging neutral air with them) whose size Naudin's own numbers match, while the over-unity results come from measurement pitfalls that are worth learning to recognise: how pulsed power is computed on a scope, where boiled-off water went, and what a "blank" run is for. Naudin deserves credit for publishing his own negative results too.

1a. Naudin's research projects at a glance

Projects with real results or notable claims, oldest first. Figures are Naudin's own unless a source is named; verdicts are the arbiter's (including its QE revision).

Notes on outcomes.

• Trouton-Noble. Cornille and Naudin report a 500 pF capacitor at about 70 kV that "sought an East-West position and stayed locked" over dozens of repeats. Documented in one published source (Cornille's 2020 review; Cornille is co-inventor and AES principal); no raw data or described controls. Mainstream physics expects a null.
• Bedini Magnetic Gate. Naudin's own closed loop: "no overunity effect".
• FR patent. 1.3 F charged in 30 s instead of 83 s, "over 90%" efficiency, no method.
• MEG. Naudin's scope read 6.76 W out against 3.25 W in (COP 2.08; 4.5 after he subtracts the control board). Checks on file point to measurement error: the stated 100 kΩ load should take 1.31 W, not the 10.28 W shown, and Naudin wrote that the load warmed less than the apparent power. No calorimetry; the loop was never closed.
• US patent. One voltage trace (500 V to 1750 V at constant charge), no energy measurement, three unreconciled sources of the extra energy. Granted in France; the US application was abandoned in 2006 after an unanswered Office action.
• Lifter. His four lifters hover at the thrust per watt ion drift predicts; ARL, Dstl, NASA/ISR and Ianconescu et al. flew lifters too. Built to about 194-250 g; lifting 1 kg would take 2.5-4.2 kW. The 351 replications are Naudin's self-registered log.
• Vacuum. Ion drift could not hold up the tube-test foil, as Naudin said, but ordinary electrostatic pull is of the foil's weight or more. In high vacuum the positive reports come from interested parties or were not obtained (Gravitec, a patent holder; Brown's unpublished notes); the nulls (MSFC, ISR, Talley second-hand) have limited sensitivity. A June 2001 test of Naudin's ACT at MSFC by TdT (a company) is relayed by Naudin with no result.
• CFR. 33 runs (table not obtained); calibration 0.985, low power 0.988, medium power 1.51. EarthTech's replication of his method averaged 0.98 (runs 9-11 excluded, stirrer off), and 1.27-1.77 only when post-power-off evaporation is counted; Kowalski, Little and Luce's Texas cell gave 1.00 ± 0.02. Positive reports: the CNAM Paris group (1.35-1.41 after cathode preconditioning, via Kowalski's pages) and Kowalski's 2006 draft (mean 1.24). Verdict: Naudin's method as replicated, Tested, not supported; his own 33 runs and the positive reports, Never properly tested (against droplet loss and post-run accounting).
• MAHG. Reported COP 2.43, 5.26 and 21 (5% duty cycle); his pure-DC blank returned 64%. Moller and Frolov hold a stake. If mean(V) × mean(I) was used, COP 21 becomes about 1.05 (not proven).
• Orbo. Naudin's page says both "measured COP = 4.75" and "4.75 is NOT A COP". Steorn's own jury (2009) said its attempts "have not shown the production of energy"; Steorn, which sold licences, disputed it.
• GEGENE. 18 tests, including calorimetric and closed-loop runs, listed by title only; result pages not obtained (not a negative result).

2. What you need to know first

Electric force on charge, and force on a conductor. A charge q in an electric field E feels F = qE. A cloud of charge with density ρ (coulombs per cubic metre, C/m³) feels a force per unit volume f = ρE (newtons per cubic metre). A charged wire near a grounded metal sheet pulls on it: the wire induces opposite charge on the sheet's surface, with surface density σ (C/m²), and the sheet feels an outward electrostatic "pressure" σ²/(2ε₀) toward the charge, where ε₀ = 8.854 × 10⁻¹² F/m is the permittivity of free space. The method of images gets σ quickly: a line charge λ (C/m) at height a above a grounded plane acts, above the plane, like λ together with a mirror charge −λ at depth a. Read: Griffiths, Introduction to Electrodynamics, chapter 2 (electrostatics, surface charge and the force on a conductor) and chapter 3 (method of images).

Corona and ion drift. Near a very thin wire at tens of kilovolts the field exceeds air's breakdown strength (about 3 MV/m [memory, high]) in a thin sheath, so the air there ionises: this is the faint glow and hiss called corona. Ions of the wire's sign are repelled into the gap. In air they cannot accelerate freely; they collide with neutral molecules about 10¹⁰ times a second [read: NASA/CR-2004-213312 p.16, via physicist.md] and settle at a drift velocity v = μE, where μ is the ion mobility (about 2 × 10⁻⁴ m²/(V·s) for positive air ions [read: Ianconescu et al. p.7, via deep/]). Every collision hands momentum to the neutral air, which is set moving: that moving air is the ion wind. Read: Raizer, Gas Discharge Physics (corona and ion mobility; chapter numbers not given here).

Momentum conservation and Newton's third law. For any isolated system the total momentum (including momentum stored in the electromagnetic field) is constant, so the internal forces between its parts sum to zero. A device can only accelerate if it pushes on something outside itself: air, the bench, the walls, its power leads, or the hand holding it. Read: Taylor, Classical Mechanics, chapter 1 and the chapter on momentum; Griffiths, chapter 8 (momentum in electromagnetic fields).

Thrust, power and why fast exhaust is expensive. A thruster that pushes air mass at rate ṁ (kg/s) to speed v makes thrust T = ṁv and spends kinetic power ½ṁv² = ½Tv. For a given thrust, slow exhaust over a big area is cheap; fast exhaust is costly. Helicopter "momentum theory" gives the least power to hover by blowing air down through an area S: P_ideal = T^(3/2)/√(2ρ_air S), with ρ_air ≈ 1.2 kg/m³ the density of air. Read: Leishman, Principles of Helicopter Aerodynamics, the chapter on momentum (actuator-disc) theory.

Power in pulsed and AC circuits. Instantaneous power is p(t) = v(t)·i(t). The only correct average power is the time average of the product, P = (1/T)∫₀ᵀ v i dt, written mean(v·i). For sinusoids this equals V_rms I_rms cos φ, where V_rms and I_rms are root-mean-square values (√ of the mean of the square) and φ is the phase angle between voltage and current; cos φ is the power factor. Two traps: (i) the RMS of the power waveform is not the mean power; (ii) for pulses, mean(v) × mean(i) is not mean(v·i). A timing mismatch (skew) between the voltage and current probes shifts φ and corrupts the product. Read: Horowitz and Hill, The Art of Electronics, chapter 1 (RMS, power, reactance).

Energy conservation for an electromagnetic device (Poynting's theorem). For any closed surface around a device, the electromagnetic energy flowing in through the surface equals the rise in stored energy inside plus what is dissipated. Average over many cycles of a repeating operation and the stored energy returns to the same value, so mean power out = mean power in − losses. A permanent magnet is a store of energy, not a source: the most it can give up is of order (BH)_max × volume, where (BH)_max (J/m³) is the magnet's "energy product". Read: Griffiths, chapter 8 (Poynting's theorem).

Heat, phase change and calorimetry. Heating water from 20 °C to 100 °C takes about 4.18 J/(g·K) × 80 K ≈ 335 J/g; turning it into steam at 100 °C takes a further 2260 J/g (the latent heat of vaporisation) [memory, high]. Water that leaves a boiling cell as liquid droplets carries only the sensible heat, not the 2260 J/g. A blank or calibration run with an ordinary resistive heater in the same cell tells you how much of the counted heat your method recovers. Read: Schroeder, An Introduction to Thermal Physics, chapter 1 (heat capacity and latent heat).

Bond energies and state functions. Splitting a hydrogen molecule into two atoms costs about 436 kJ per mole of H₂; letting them recombine releases exactly the same amount [memory, high]. Energy is a state function: any cycle that returns hydrogen to where it started releases zero net energy, whatever happens in between. Read: Atkins, Physical Chemistry, the chapters on thermochemistry.

Capacitor energy at constant charge. A capacitor of capacitance C holding charge Q has voltage V = Q/C and stored energy E = Q²/(2C) = ½CV². If C is reduced while Q stays fixed (plates pulled apart, or a conducting layer removed), V and E both rise in proportion to 1/C, and the extra energy must be supplied as work by whatever reduced C. Read: Griffiths, chapter 2 (energy of a charge distribution) and chapter 4 (capacitors with dielectrics).

3. The textbook view

3.1 The Lifter as an electrohydrodynamic (EHD) thruster

Electrohydrodynamics (EHD) is the study of forces exerted by electric fields on gases or liquids that carry some charge. The mainstream account of the Lifter [read: physicist.md "Mainstream view"; ARL-TR-3005 eq. 14; Ianconescu et al.; NASA/CR-2004-213312]:

1. The thin wire (radius r ~ 0.05 mm) at voltage V ~ 40 kV ionises the air in a thin sheath.
2. Ions of the wire's sign drift across the gap d (30 mm in Naudin's lifters) toward the foil at v = μE, a few hundred metres per second.
3. The field pushes on the ion cloud with total force ∫ρE dV. Because the ions are in a steady drift, that force is passed by collisions to the neutral air, which is pushed downward. The equal and opposite reaction acts on the electrodes (as the foil's attraction toward the ion cloud and the wire's repulsion from it). "Electrostatic pull on the space charge", "ion drift" and "ion wind" are one force seen three ways.

The size of that force follows in two lines (derived in section 5):

F = I d / μ

where I is the current (A), d the drift distance (m) and μ the mobility. The power drawn is P = VI, so the thrust per watt is

F/P = d/(μV) = 1/v_drift,

the inverse of the ion drift speed. This predicts: thrust grows with current and gap; thrust needs a gas (no gas, no collisions, no momentum passed on); thrust follows the electrode asymmetry, not the polarity; it changes with the gas (mobility differs); and an enclosed lifter does not move, because the air it pushes pushes back on the box. All of these are observed [read: NASA/CR p.11-13, no rotation resolved below about 300 Torr on an apparatus that could not resolve less than one-sixth of a turn, with gaps not designed for low pressure, and one exception: an eighth-turn movement together with a large arc at 5.5 × 10⁻⁵ Torr and 44 kV, which the authors attribute to ejected material; Dstl slides, lift in air and nitrogen, none in helium at the test voltage; ARL and ISR, polarity; Tajmar sealed box, null to about 10 µN; physicist.md].

Two different calculations go by the name "ion wind" and must be kept apart. The ballistic model (ions fly freely and hit the foil) gives a force orders of magnitude too small; ARL computed this and called the Lifter's basis "not understood" [read: ARL abstract, p.4]. The collisional drift model above gives the right size. Statements such as "ion wind is three orders of magnitude too small" refer to the ballistic model, not to the collisional one [read: arbiter.md, corrected claim 1; deep/ARL].

3.2 Can any force survive in high vacuum?

In vacuum there is no air to push. Ions or electrons that leave the electrodes ballistically can only give a reaction force equal to the momentum they carry: at 45 kV and tens of microamps, about 10⁻⁸ to 10⁻⁷ N for electrons and up to 10⁻⁵ N for ions [calc: physicist.md, via arbiter.md]. An isolated charged system feels no net force from its own pairwise Coulomb forces (Newton's third law). So a static DC force in clean high vacuum would need either an external body to push on (chamber walls, leads, residual gas from outgassing, field-emission currents) or new physics coupling electric fields to gravity or inertia. Mainstream physics expects zero, but this is the one point in the Lifter story that has not been tested cleanly (section 6).

3.3 Energy accounting for the MEG, the reactors and the patents

• MEG. A permanent magnet in a transformer core is an old idea: the magnetic amplifier and flux-switching devices use it. Over many cycles Poynting's theorem gives mean output ≤ mean input, as long as all energy crosses the four electrical terminals. The magnet in Bearden's patent (40 × 25.4 × 38.1 mm = 38.7 cm³) can give up at most about 12 J if it is NdFeB ((BH)_max ≈ 300 kJ/m³ [memory, high]), about 1 J if ferrite, and about 26 J even if it were completely demagnetised [calc: arbiter.md]. A sustained excess of a few watts would empty that in seconds.
• Plasma electrolysis (CFR). When a cathode in an electrolyte glows with plasma, water boils violently and spray leaves with the steam. Counting all lost mass as steam credits each gram of spray with 2260 J it never took. That is a factor of 2260/84 ≈ 27 on the droplet mass (84 J/g is sensible heat from about 80 °C to 100 °C; arbiter's figure) [calc: arbiter.md].
• Atomic hydrogen (MAHG). Langmuir's atomic-hydrogen torch (1920s) dissociates H₂ on a hot tungsten surface and gets very high flame temperatures from recombination [memory, high]. It moves energy from the electric supply to the flame; it does not create it, because the H₂ → 2H → H₂ cycle returns to its start (a state function).
• Capacitor patents. Charging a capacitor from a fixed-voltage source through a resistor loses half the energy in the resistor, whatever the resistance. Switched-inductor ("switch-mode") chargers reach 90% or better [memory, high], and charging in N equal voltage steps gives efficiency N/(N+1) [web: https://staff.najah.edu/en/publications/11442/, via arbiter.md]. At constant charge, lowering C raises the stored energy only at the cost of equal work.

4. Where the inventor departs

Naudin's starting points are real. Brown's 1960 patent, which Naudin quotes, describes "electrodes ... immersed in a dielectric medium" producing "relative motion between a structure and the surrounding medium" [read: Brown US 2,949,550, as quoted on _text/2001-Lifter-v1-Lifter1-experiment, p.1]: a statement that is fully compatible with EHD. Naudin's own "true Lifter" page concludes that "the surrounding medium is therefore required for the motion ... a true Lifter fully enclosed in a box will not move ... A polarisable dielectric medium is required to get a thrust" [read: _text/2003-Lifter-true-lifter-wire-in-vacuum-tube, p.1]. That is what the mainstream says too. The departures are narrower than they first appear.

A short side check that underlies the MAHG row (example.py step 6). For a pulse train that is on for a fraction D of each period at voltage V_p and current I_p, the true mean power is D·V_p·I_p. The product of the separate means is (D V_p)(D I_p) = D²V_pI_p, too small by a factor D. Read that way at D = 5%, a reported COP of 21 becomes 21 × 0.05 = 1.05, and 11.53 becomes 0.58 [calc]. RMS × RMS cannot inflate COP this way (it equals D V_pI_p for rectangular pulses and is never below the true mean). Naudin's data sheets say inputs were "Rms Voltage and Rms Current", while "the average voltage and current are also measured on the digital oscilloscope" without saying which entered the high-COP runs [read: pultests and lrtests pages via arbiter.md]. So the hazard fits the pattern but is not proven to be the cause; the per-run table that would show whether COP × D stays constant is an image not in the folder.

5. Worked example: how hard can ions push, and what held the foil up?

This uses only Naudin's own published numbers [read: Lifters comparison chart, _text/2001-Lifter-v4-experiment, p.1; Lifter1 page; vacuum-tube page]. Every number is reproduced by example.py.

Step 1: derive the ion-drift force

Take a slab of the gap, cross-section A, length d, filled with ions of charge density ρ in field E.

• Force per unit volume on the ions: f = ρE.
• The ions drift at v = μE, so the current density is J = ρv = ρμE, which gives ρ = J/(μE).
• Substitute: f = (J/(μE))·E = J/μ. The field cancels.
• Total force: F = f × (A d) = (J A) d/μ = I d/μ.

Units: A·m ÷ (m²/(V·s)) = A·V·s/m = J/m = N. The force is the most that drifting ions can pass to the air; field lines longer than d raise it, losses lower it. NASA's contractor (ISR) estimates the curved paths add about 38% [read: NASA/CR p.16-18, via physicist.md].

Step 2: apply it to Naudin's four Lifters

With d = 30 mm and μ = 2.0 to 2.5 × 10⁻⁴ m²/(V·s), weight = mass × 9.81 m/s²:

[calc]. Three of four lifters sit comfortably under the ceiling. Lifter 4 at hover sits at it on a straight 30 mm path: it needs an effective ion path of 31 to 39 mm, which the 40 mm-high foil skirt and curved field lines allow. One caution from the arbiter: the chart lists "payload 4 g" for Lifter 4, and Naudin's Lifter 1 page ties the second power column to flight with the payload; if the 81.4 W hover also carried 4 g, the need rises to 1.17 to 1.47 times the straight-path ceiling, still within the geometry allowance but with no margin [read: arbiter.md, claim 1]. Nothing in the table needs a second force; equally, the table cannot rule out a small one.

The same check works on Dstl's data. Dstl's 2005 slides say "calculated ionic thrust more than an order of magnitude less than that observed" without showing the calculation. Its own charts [read: deep/Dstl, page images pp.28-34] give lifter IW8 carrying 6.16 g (0.0604 N) at 1.07 mA with the wire 44 mm above the foil: Id/μ = 1.07 × 10⁻³ × 0.044 / 2 × 10⁻⁴ = 0.235 N, 3.9 times the weight. For the ion force to be a tenth of the weight the gap would have to be about 1.1 mm [calc: example.py step 2b]. Across Dstl's lifters the ion bound is 2 to 5 times the weight [arbiter.md, QE revision s.1]. So, as far as the slides show, Dstl's sentence is contradicted by Dstl's own numbers (its method may differ; it is not shown).

Step 3: the sharpest test is thrust per watt

Because F = Id/μ and P = VI, the predicted maximum thrust per watt is d/(μV) = 1/v_drift. Compare with what Naudin measured (weight ÷ power):

[calc]. This comparison uses no adjustable constants except a textbook mobility. Naudin's measured values all fall at or below the ion-drift prediction except Lifter 4 hover, which sits 4 to 30% above the straight-path value, inside the curved-path allowance. Had a second, non-EHD force been doing much of the lifting, measured thrust per watt would stand well above this line. It does not.

Step 4: why lifting a kilogram is so expensive

Lifter 4 pushes 0.314 N over 0.820 × 0.700 m = 0.574 m². An ideal rotor of that area would hover with P = T^(3/2)/√(2ρ_air S) = 0.15 W. The Lifter uses 81.4 W, about 540 times more [calc], because it accelerates a little air through ions moving at about 300 m/s instead of a lot of air slowly. At Lifter 4's best efficiency (3.86 mN/W hover; 2.36 mN/W flight), one kilogram needs 9.81 N ÷ 3.86 mN/W = 2.5 kW to hover, or 4.2 kW in flight [calc]. That is a cost, not a law against it; MIT's 2018 solid-state EHD aeroplane [memory, high: Xu et al., Nature 563, 2018] flew on the same ion-drift physics; it did not escape the thrust-per-watt limit, it was designed around it.

Step 5: the wire in a vacuum tube

Naudin's set-up: three 38 mm glass bulbs with tungsten filaments joined in a triangle at +25 kV, held by hand above a grounded aluminium foil 200 × 20 mm, 12 µm thick, "about 1 µA" [read: unlinked vacuum-tube page].

• Foil mass = 2700 kg/m³ × 0.200 × 0.020 × 12 × 10⁻⁶ m³ = 0.130 g; weight 1.27 mN.
• (a) Ion drift, which Naudin set out to exclude: at 1 µA over 30 mm, F ≤ Id/μ = 10⁻⁶ × 0.03 / 2 × 10⁻⁴ = 0.15 mN, about 0.12 of the weight [calc]. Naudin is right: ion drift from the wire cannot hold the foil up.
• (b) Static attraction, which the test did not exclude. Model the charged conductor as a long cylinder of radius r with its axis at height h over a grounded plane. Its outside field equals that of a line charge λ = 2πε₀V/arccosh(h/r) at height a = √(h² − r²). The induced charge on the plane is σ(x) = −λa/(π(x² + a²)), and the plane is pulled up with pressure σ²/(2ε₀). Integrating over the foil's 20 mm width only (|x| ≤ 10 mm) and multiplying by its 200 mm length:

[calc]. In every geometry the static pull is of the order of the foil's weight or many times it. The reading Naudin himself first set out to check ("to check if the thrust observed previously was simply an electrostatic attraction" [read: true-Lifter page]) remains the simplest one. What the test shows: ion drift is not what lifts the foil. What it does not show: that anything other than ordinary electrostatic attraction does.

What the example teaches. Naudin's air data are a good confirmation of collisional EHD, not a counter-example: the ion-drift formula, with one textbook constant, predicts his thrust per watt within the geometry allowance. His vacuum-tube experiment removes one explanation correctly and leaves a second, ordinary one standing. The general lesson is to calculate every conventional force before attributing a residual to a new one, and to record the distances that the calculation needs.

6. Evidence table

Sources for the table: arbiter.md (verdict summary and claims 1-7), which cites the documents and research files by page.

7. What a learner should take away

Solid. The Lifter is real, cheap to build and a good teacher of gas-discharge physics. Its thrust is a textbook force with a two-line derivation, and Naudin's carefully published voltages and currents let you check that derivation yourself. His FR fast-charger patent rests on ordinary switch-mode engineering. His calibration runs on the plasma reactor (0.985 and 0.988) show he knew how to close an energy balance when the cell was quiet. And he published results that went against him: the Magnetic Gate, the magnetic amplifier, Larsko's Orbo number. The arbiter rightly counts that candour to his credit, with one balance: the same Orbo page also says the loading test "has demonstrated an overunity effect" [read: deep/Orbo], so both statements belong in the record.

Wrong, and why each mistake is instructive. - Excluding one mechanism is not identifying another. The vacuum-tube test excluded ion drift and then attributed the remaining force to something new without calculating the static pull. Always list and size every ordinary force first, and record the geometry the calculation needs. - Two models can share a name. "Ion wind is far too small" was true of the ballistic model and false of the collisional one. Ask which model a number comes from before quoting it. - A scope's product is only as good as its channels. The MEG numbers multiply correctly; the doubt is in the current channel and the probe timing. An independent check (Ohm's law on the load, heat in the load) catches what the screen cannot. RMS of a power trace is not power, and mean(V) × mean(I) is not power for pulses. - Blanks are there to be believed. A DC blank at 64% means the bench under-reads or over-reads by a known factor; a pulsed run must be read through the same correction, not beside it. - Count phase changes honestly. Spray is not steam, and evaporation after the power is off was paid for by heat already counted.

Genuinely open. - Whether any static force exists on an asymmetric capacitor in clean high vacuum at tens of kilovolts. ARL called it "the most pressing question" in 2003 [read: deep/ARL], and it is still untested with a calibrated, shielded torsion balance, bake-out, residual-gas analysis, polarity reversal and a mirror-image control. The Gravitec/NSSTC (rotation, 2003) and MSFC (no performance, 2004) results, both from Huntsville, have not been reconciled in any record found, and a third Huntsville event, TdT's June 2001 test of Naudin's ACT in the MSFC vacuum chamber, has no recorded result. Brown's air-versus-oil torque claim, if it held, would argue against air ionisation as the whole story; no one else is on record testing it, and Naudin's own "Brown device in mineral oil" page was not obtained. - The Cornille-Naudin Trouton-Noble result: a charged capacitor on a thread reported to lock East-West over dozens of repeats (1996-99). It is Naudin's own experimental work and the basis of his co-inventor's "stimulated forces" theory. Caveats from the arbiter: one published source (Cornille, an interested party), no raw data, controls not described, and an East-West alignment is also what a thread-hung charged body does in a stray ambient field. What would settle it: rotate the suspension and the room geometry, reverse polarity, screen the enclosure, log the heading against sidereal time, and compare with the same capacitor uncharged [arbiter.md, QE revision s.3]. - Which of probe skew, a mis-scaled current channel or a truly non-linear load produced the MEG's 10.28 W into a stated 100 kΩ; and what became of the DoD-linked university replication Bearden announced in December 2000 (no report found; possible readings: never done, done and negative, done and confidential; the record cannot choose) [read: arbiter.md]. Horky (2009) adds, citing no source, that Bearden's team developed 2.5 kW closed-loop automotive MEG modules (four for 10 kW) of which "not a single piece has so far left the production line" [read: deep/Horky p.2]: a claim of development with no data, beside a closed loop that Naudin has not achieved in about 26 years. - Whether the CNAM group's excess heat (COP about 1.35 to 1.41 after preconditioning) survives a droplet and same-cell-heater control. It is the most neutral plasma-electrolysis test connected to Naudin, and its report was not read here. - Whether Naudin weighed the CFR's evaporated water after cooldown, what his 33-run table shows, and whether MAHG COP × duty cycle stays constant across runs. - Why the US Cornille-Naudin patent was abandoned after an unanswered Office action while its French sibling was granted (cost, a substantive objection, no measurement to offer, or retirement; the file wrapper would decide) [read: arbiter.md, claim 6]. - Naudin's later work (BiTT, ReGenX, GFS UAV) and Frolov's MAHG report were not reached.

8. Check yourself

1. A lifter draws 1.5 mA at 35 kV with a 40 mm wire-to-foil gap. Using μ = 2.0 × 10⁻⁴ m²/(V·s), what is the most mass ion drift could hold up, and what is the predicted thrust per watt?
2. A builder wants a lifter to carry a 100 g camera. At Naudin's best measured 3.86 mN/W, how much power does that take, and what does the momentum-theory ideal suggest about where the rest goes?
3. A scope shows mean(V) = 10 V and mean(I) = 0.2 A for a rectangular pulse train at 10% duty, V_p = 100 V, I_p = 2 A. What is the true input power, and by what factor would mean × mean understate it?
4. In a plasma-electrolysis run, 10 g of water leaves the cell as spray at 100 °C, but the experimenter counts it as steam. By how many joules is the heat output overstated?
5. Bearden's MEG magnet is 38.7 cm³ of NdFeB with (BH)_max ≈ 300 kJ/m³. If the device really delivered 3.5 W more than its input, how long could the magnet sustain it, and what simple experiment would therefore settle the question?

9. Further reading

Mainstream: - Griffiths, D. J., Introduction to Electrodynamics: chapter 2 (electrostatics, force on a conductor), chapter 3 (method of images), chapter 8 (Poynting's theorem, field momentum). - Raizer, Yu. P., Gas Discharge Physics (Springer): corona discharge and ion mobility. - Leishman, J. G., Principles of Helicopter Aerodynamics: momentum (actuator-disc) theory of hover power. - Horowitz, P. and Hill, W., The Art of Electronics, chapter 1: RMS, average and real power. - Schroeder, D. V., An Introduction to Thermal Physics, chapter 1: heat capacity, latent heat. - Bahder, T. B. and Fazi, C., Force on an Asymmetric Capacitor, ARL-TR-3005 (2003) [read: in the folder; see arbiter.md for the corrected copper-ion figure]. - Canning, F. X., Melcher, C. and Winet, E., Asymmetrical Capacitors for Propulsion, NASA/CR-2004-213312 (2004) [read: in the folder]. - Tajmar, M., "Biefeld-Brown effect: misinterpretation of corona wind phenomena" (2004) [read: in the folder]. - Ianconescu, R., Sohar, D. and Mudrik, M., analysis of the Brown-Biefeld effect [read: in the folder, filed under 2012]. - Masuyama, K. and Barrett, S. R. H., "On the performance of electrohydrodynamic propulsion", Proc. R. Soc. A 469 (2013) [memory, high]; Xu, H. et al., "Flight of an aeroplane with solid-state propulsion", Nature 563 (2018) [memory, high]. - Little, S. R. and Luce, G. (EarthTech), replication of Naudin's plasma-electrolysis method, 15 Jul 2003 [read: research/pages/et_replicationjln_2003.txt]; Kowalski, L., Little, S. R. and Luce, G., ICCF-12 (2005) [read: research/files]; Kowalski's notes on ICCF-12 and his 2006 draft on positive results [web: https://msuweb.montclair.edu/~kowalskil/cf/270iccf12.html; https://msuweb.montclair.edu/~kowalskil/cf/300positive.html]. - Fauvarque, J.-F., Clauzon, P. and Lallevé, G., "Abnormal excess heat observed during Mizuno-type experiments" (2005) [cited by Kowalski; not read]. - Tajmar, M., Kößling, M. and Neunzig, O., "In-depth experimental search for a coupling between gravity and electromagnetism with steady fields", Scientific Reports 14, 19427 (2024) [cited in the folder's Wikipedia copy; not read]. - Jelinek and Pekarek (Sisyfos), Poynting-theorem critique of the MEG (2010, Czech) [read: in the folder]; Bishop, S., MEG critique (2002) [read: research/files].

The inventor and those he worked with: - Naudin, J.-L., JLN Labs pages, 1996-2013 (main site jnaudin.free.fr; the CFR and MAHG pages are on jlnlabs.online.fr) [read: README p.1]: Lifter project (Lifter1, v4, full story, wire-in-vacuum-tube tests, worldwide replications log), MEG v2.0/v2.1 tests, CFR plasma-electrolysis reactor, MAHG, and the hosted devices [read: _text/]. - Cornille, P. and Naudin, J.-L., FR 2 789 818 A1 (fast battery charger, 1999/2000) and US 2005/0057116 A1 (electrical power source, priority 2001) [read: deep/]. - Brown, T. T., US Patent 2,949,550, "Electrokinetic Apparatus" (1960) [read: quoted on the Lifter1 page]. - Bearden, T. E., Hayes, Kenny, Moore and Patrick, MEG paper (2000) [read: in the folder; interested party]. - Horky, MEG replication from Naudin's plans (2009, Czech) [read: deep/]. - Losonc, lifter-in-vacuum charge-force analysis (undated) [read: in the folder; formulas partly missing]. - Cornille, P., "A review on electromagnetic propulsion by stimulated forces" (2020), including the Trouton-Noble experiment with Naudin [web: https://www.longdom.org/open-access/a-review-on-electromagnetic-propulsion-by-stimulated-forces-53822.html, via the Investigator; interested party]. - Brown's letter, reproduced in ARL-TR-3005 Appendix B [read: deep/ARL p.36-37].

Corrections to earlier documents

Found while re-deriving the worked example (for the orchestrator; no verdict changes): 1. physicist.md, claim 1, wire-in-vacuum-tube table: the force-per-length formula is printed as λ²/(4πε₀·2h). The image force on the line charge, which equals the total pull on an infinite grounded plane, is λ²/(4πε₀h), twice that. The physicist's "1.2 mN" for a 0.05 mm wire at 30 mm is half the full-plane value (2.3 mN); restricting the pull to the 20 mm foil strip, as the arbiter does, gives 0.91 mN, matching the arbiter's 0.9 mN [calc: example.py step 5]. The conclusion (pull of the order of the weight) is unchanged. 2. arbiter.md, claim 1: "about 35 mN if the glass envelope carries the wire's potential". With the strip-restricted model the result depends on which distance is 30 mm: 8.2 mN for a 30 mm gap to the glass surface, 69 mN for the tube axis at 30 mm [calc: example.py step 5]. The arbiter's figure lies inside that range; Naudin did not record the distance. No verdict depends on it.

---

Answers

1. F ≤ Id/μ = 1.5 × 10⁻³ × 0.040 / 2.0 × 10⁻⁴ = 0.30 N, which holds up 0.30/9.81 = 30.6 g. Thrust per watt = d/(μV) = 0.040/(2.0 × 10⁻⁴ × 35 000) = 5.7 mN/W (equivalently 0.30 N ÷ 52.5 W). The drift speed is μV/d = 175 m/s.
2. 0.981 N ÷ 3.86 mN/W = 254 W (plus the lifter's own weight). The ideal-rotor power for the same thrust over a similar area is a fraction of a watt; nearly all of the rest goes into ions colliding with air at hundreds of metres per second, i.e. heat and a thin fast jet. Thrust per watt is 1/v_drift, so slow, wide flows are cheap and fast, narrow ones expensive.
3. True mean power = D V_p I_p = 0.1 × 100 × 2 = 20 W. mean(V) × mean(I) = 10 × 0.2 = 2 W, too small by the factor D = 0.1, so any COP computed from it is ten times too high.
4. Counted as steam it carries 2260 J/g of latent heat that spray never took: 10 g × 2260 J/g = 22.6 kJ overstated (the sensible heat to 100 °C is real in both cases). Over an 82 s run that is about 275 W of phantom output.
5. Stored energy ≈ 300 000 J/m³ × 38.7 × 10⁻⁶ m³ = 11.6 J, which 3.5 W would empty in 3.3 s (even total demagnetisation, about 26 J, lasts 7.4 s). So closing the loop (output feeding input, with a start battery whose voltage is logged) for a few minutes would settle it: a real source keeps running, a measurement artefact stops. That test was attempted and not achieved ("have not yet succeed", Naudin, 13 Nov 2000).