Bearden: Primer
Project: Thomas E. Bearden, "The Final Secret of Free Energy" (1993, with updates of February and 13 May 1994) and the Motionless Electromagnetic Generator patent, US 6,362,718 (Patrick, Bearden, Hayes, Moore, Kenny; granted 26 March 2002).
Inputs: arbiter.md (authoritative verdicts, including its "QE revision", "Round 2", "Evidence-rule revision (2026-10-05)" and "Investigator findings (2026-10-05)" sections), physicist.md, mathematician.md, contrarian.md, the Investigator's sites.md and research/ notes (2026-10-05), the deep reads in deep/ (the 15 Feb 1994 update, pp.28-41; the 13 May 1994 paper, pp.42-49; the patent), and my own reading of both source PDFs. Companion script: example.py (run command in its header; it runs cleanly and reproduces every number in Section 5, including Part F for the Investigator findings in Section 5c).
Source tags: [read] = from the source documents; [read: page image p.N] = read from a rendered page image; [web] = Internet source, URL given; [memory] = recalled, with confidence. [calc] = arithmetic done here and in example.py.
Source gap note (repeated from earlier steps). Both documents have a usable text layer, so this is not a [memory]-only review. Remaining gaps: the two 1994 updates (15 Feb, pp.28-41; 13 May, pp.42-49) have now been deep-read in full, and neither contains a single measurement (no voltage, current, power or waveform) [read, per deep read]; no test, replication or suppression claim comes from the documents themselves, only from web sources named in arbiter.md. Each is given with its stake: Bearden himself (a co-inventor of US 6,362,718, which lists five individual inventors and no assignee [read: patent front page], and spokesman for Magnetic Energy Ltd, the private company he describes as developing the MEG, so he has a commercial stake); J. L. Naudin (an independent private builder with no stated commercial stake); and archive and news sites (ZPEnergy, rexresearch, amasci). New in this primer: earlier steps had not read the patent's Figs. 6-8 as images. I rendered them (PDF pages 3 and 4) and read the curves by eye; values quoted from them are approximate (about +/-5% of full scale) and tagged [read: page image].
Investigator findings (2026-10-05) and their gap. The Investigator searched patent offices, the Internet Archive's copy of the USPTO (United States Patent and Trademark Office) file of the MEG continuation application, the EPO (European Patent Office) Register, J. L. Naudin's JLN Labs pages, a mirror of Bearden's cheniere.org site and company registers. Nothing was downloaded, so every item from that run is the Investigator's reading of a web page, carried at second hand by the reviewers, the arbiter and this primer [arbiter]. Gaps that remain: only the first 50,000 of 85,447 characters of the applicants' 2004 remarks to the USPTO were read; Bearden's declaration filed with them, the abandonment notice, the EPO annex pp.4-5, the applicants' 2007/2008/2009 EPO papers and the 2013 minutes were not read. What Bearden told the two patent offices is therefore known from partial reads.
1. The idea in one paragraph
Bearden argued that electrical "potential" (voltage) is not just a bookkeeping number but an ordered flow of energy from the vacuum, and that every charge is constantly fed by this flow from the whole universe. If that were so, a battery would not be an energy store but a "gate" that sets up a potential, and you could take energy out of that potential for free, as long as you did it without letting current flow back through the battery (which, he said, is what "kills" the source). His recipe: use the battery only to charge a separate "collector" for a very brief moment before any current can flow, disconnect it, and let the collector power the load; repeat. He also co-patented a solid-state device, the Motionless Electromagnetic Generator (MEG), a magnetic core with a permanent magnet and switched coils, whose inventors reported more electrical power out than in. The physics questions are: is "potential = energy" correct, can a capacitor be charged "for free", and do the MEG measurements show real excess power?
2. What you need to know first
2.1 Electric potential and potential energy. The electric potential phi (Greek "phi", unit volt, V = joule per coulomb) at a point is the work per unit charge needed to bring a small test charge there from a reference point. A charge q placed at potential phi has potential energy U = q phi (joules). So potential is energy per charge, not energy. A "voltage" V between two points is a difference of potential. Read: Griffiths, Introduction to Electrodynamics, Chapter 2 (Electrostatics), sections on the potential and on work and energy [memory, high].
2.2 Capacitors and their stored energy. A capacitor is two conductors separated by an insulator. Its capacitance C (farad, F) is the charge Q (coulomb, C) stored per volt: Q = C V. Charging it is a gradual process: early charge is moved through a small voltage, later charge through a large one. Adding up the work gives the stored energy W = (1/2) C V^2 = Q V / 2, not Q V. This factor of one half is the heart of this project. Read: Griffiths Chapter 2 (capacitors) and any introductory circuits text on RC circuits [memory, high].
2.3 RC circuits. When a battery of voltage V charges a capacitor C through a resistance R (ohm), the current is i(t) = (V/R) e^(-t/RC). The product RC (seconds) is the time constant: it sets how fast the capacitor charges. A first-order linear ODE (ordinary differential equation) you already know how to solve. Read: Horowitz and Hill, The Art of Electronics, Chapter 1 [memory, high].
2.4 Maxwell's equations, potentials and Poynting's theorem. Maxwell's four equations relate the electric field E and magnetic field B to charges and currents. The fields can be written in terms of a scalar potential phi and a vector potential A. Poynting's theorem is the energy-conservation law that follows from Maxwell's equations: energy in a region changes only by energy flowing across its boundary or work done on charges. Read: Griffiths Chapters 7 (Electrodynamics), 8 (Conservation Laws) and 10 (Potentials and Fields) [memory, high].
2.5 Electromagnetic waves. In empty space Maxwell's equations give waves travelling at the speed of light with E and B perpendicular to the direction of travel ("transverse"). Polarisers work because of this. Read: Griffiths Chapter 9 [memory, high].
2.6 Transformers and AC power. A transformer couples two coils through a magnetic core; ideally the voltage ratio equals the turns ratio, and power out equals power in (minus losses). For alternating or pulsed signals, the real power is the time average of v(t) x i(t). If current and voltage are out of step (a phase shift), the product of their sizes overstates the power. Read: Horowitz and Hill Chapter 1; Purcell and Morin, Electricity and Magnetism, chapter on AC circuits [memory, high].
2.7 Charge relaxation time. If you somehow put extra charge inside a conductor, it flows to the surface with a time constant tau = epsilon / sigma, where epsilon is the material's permittivity (F/m) and sigma its conductivity (siemens per metre, S/m). Read: Griffiths Chapter 7 (Ohm's law section and its problems) [memory, medium for exact location].
3. The textbook view
Symbols used here: q, Q charge (C); phi scalar potential (V); A vector potential (V s/m); E electric field (V/m); B magnetic field (tesla, T); rho charge density (C/m^3); J current density (A/m^2); epsilon_0 = 8.854e-12 F/m, the permittivity of free space; mu_0 = 4 pi x 1e-7 H/m, the permeability of free space; c speed of light; t time; div, grad, curl the usual vector-calculus operators.
3.1 Potential versus energy.
U = q phi (potential energy of charge q at potential phi; joules = coulombs x volts)
phi = U / q (potential is energy per unit charge)
They have different units, so they cannot be "the same thing".
3.2 Fields from potentials.
E = -grad phi - dA/dt
B = curl A
You can change phi and A by a "gauge transformation" (for example add a constant to phi) without changing E or B. In classical physics only E and B produce forces (F = q(E + v x B), with v the charge's velocity). In quantum mechanics the potentials do more: the Aharonov-Bohm (AB) effect (predicted 1959) shows an electron's quantum phase is shifted by A in a region where B = 0 [memory, high]. That is a real point in favour of "potentials are fundamental", but it shifts phases; it is not an energy source.
3.3 Field energy is finite and local. The energy stored in an electric field is
U_E = (epsilon_0 / 2) * integral of E^2 dV (dV = volume element)
For a capacitor this gives W = (1/2) C V^2. For a point charge, the energy outside any radius R is q^2 / (8 pi epsilon_0 R): finite.
3.4 Poynting's theorem (energy conservation for EM).
d/dt (energy in fields inside volume) = - (flow of S through the surface) - (work on charges inside)
S = (1/mu_0) E x B (Poynting vector, W/m^2)
Consequence: a closed system described by Maxwell's equations cannot put out more energy than it takes in. Every reviewer and the arbiter agree this part of Bearden's reasoning is correct [arbiter, Claim 4].
3.5 Waves in vacuum are transverse. In empty space div E = 0. For a plane wave E = E_0 exp(i(k.r - omega t)) (k is the wave vector pointing along travel, omega the angular frequency, r position), div E = i k.E = 0, so E is perpendicular to k. A longitudinal electromagnetic wave in vacuum is ruled out by two lines of algebra [arbiter, Claim 2].
3.6 Charging a capacitor costs Q V; storing it keeps Q V / 2. A battery of voltage V that pushes charge Q through itself does work Q V. If that charge ends up on a capacitor, the capacitor holds Q V / 2; the other half is turned into heat (and some radiation) in the wires and switch, and this is true for any resistance, even a tiny one. (Section 5 derives it.)
3.7 Relaxation time. tau = epsilon / sigma describes how fast excess charge inside a material disappears to its surface. It is not a delay before a circuit's current starts. The delay of a circuit is its RC time constant (and, at the shortest times, the light travel time along the wires).
3.8 Transformer limits and AC power. For an ideal transformer, V_out / V_in = N_out / N_in (N = number of turns), and P_out = P_in. Resonance (coil capacitance plus inductance ringing) can make peak voltages higher than the turns ratio, but it stores and returns energy each cycle; it does not create average power. Real power:
P = (1/T) * integral over one period T of v(t) i(t) dt
For sine waves: P = V_rms * I_rms * cos(phase angle)
V_rms and I_rms are root-mean-square values. "Average volts times average amps" is not power. For any waveforms, V_rms x I_rms is only an upper bound on the true power (it follows from the Cauchy-Schwarz inequality, the integral version of |a.b| <= |a||b|); the two are equal only when v(t) and i(t) are proportional (same shape, in phase). Example: v = 10 sin(wt) V and i = 2 sin(3wt) A (w the angular frequency) give V_rms x I_rms = 7.07 x 1.41 = 10 VA but zero true power, because different frequencies average to zero over a cycle [arbiter, recomputed; example.py Part F1]. Pulsed and spiky waveforms are full of harmonics, so this matters for the MEG.
4. Where Bearden departs
| Topic | Textbook | Bearden | Comment |
|---|---|---|---|
| Energy vs potential | U = q phi; phi is energy per charge (J/C) | "Energy and potential are identically" the same; "exactly the same definition" [read, 1993 paper] | Units differ. He concedes circuit theory uses potential as "excess energy per coulomb" [read] and then writes K_T = V_T x Coul_T [read, eq. 6], which silently switches back to energy. Verdict: Internally inconsistent [arbiter]. |
| Static potential | A solution of Laplace's equation; finite field energy | Contains an infinite series of wave pairs (Whittaker 1903/04, Ziolkowski) carrying energy in from the whole universe [read] | Whittaker's expansion is correct mathematics: any harmonic function can be written as a sum of plane waves in complex directions [web: https://arxiv.org/html/2205.08309v2, per arbiter]. An expansion is a way of writing a function; it adds no energy. Verdict: math Established; energy reading Refuted (stored-energy reading) or Not testable (reservoir reading) [arbiter]. |
| Fields in vacuum | E and B exist and carry energy and momentum | No force fields in vacuum, only potentials; vacuum EM waves are longitudinal [read] | div E = 0 forces transverse waves (3.5); polarisers confirm. Verdict: Tested, not supported and Refuted by calculation [arbiter]. |
| Charge | e is fixed; matter is neutral to about 1e-21 e | Charge is not quantized; it varies with ambient potential [read, 13 May 1994] | Neutrality bound [web: https://arxiv.org/abs/1102.2766]. Also a calculus slip: he writes d/dt[(phi)(m)] = (dphi/dt)(dm/dt) [read, 1994 update], but the product rule is d(fg)/dt = f'g + fg'. |
| Charging a capacitor | Battery spends Q V; capacitor stores Q V / 2 | During the relaxation time the source delivers "current-free potential", so the energy in the collector is free [read, collector section] | Section 5: the "free" K_T = Q V is exactly what the battery spent. Verdict: Refuted by calculation [arbiter]. |
| Relaxation time | tau = epsilon/sigma, charge decay inside a material | The time "for free electrons to reach the skin of the wire after potential is applied"; during it no current flows; dope copper to make tau about 1 ms while keeping a good conductor [read] | A 1 ms tau requires sigma = 8.85e-9 S/m, an insulator [calc]. Verdict: Internally inconsistent [arbiter]. |
| Conservation of energy | Closed systems conserve energy (Poynting) | Closed CEM cannot be overunity, so working devices must be open to a hidden vacuum source; "we are easily permitted to ... violate the 'local energy conservation law'" (fn. 25) [read]; elsewhere "does not violate the conservation of energy law" [read] | The conditional is correct; the hidden source is given no formula or signature. Verdict: conditional Established; hidden source Not testable as stated [arbiter]. |
| MEG | A transformer with a magnet: P_out <= P_in on average | Measured output "substantially higher" than input at every point; extrapolated COP 3.44 per coil at 100 V; can self-run [read, patent] | See Sections 5b and 5c. The patent itself never mentions the vacuum or Whittaker [read; arbiter]; the link is made only in the applicants' 2004 remarks to the USPTO (below). |
| MEG mechanism (2004) | Changing the vector potential A outside a coil core gives E = -dA/dt; the EMF round a coil is set by the changing flux it encloses (Faraday) | The magnet is a "broken-symmetry dipole" absorbing energy from the vacuum; B is confined to the core and a curl-free A outside, struck by sharp pulses, gives "strong E-field energy pulses, by the well-known equation dA/dt = -E", the free input [web: USPTO file, applicants' remarks of 24 May 2004, per arbiter] | E = -dA/dt is correct and is ordinary induction. Outside the core B = curl A = 0, so locally A = grad(chi) for some function chi; its time change gives an E that is a pure gradient, with zero EMF round any loop that does not enclose the changing flux. No extra energy appears [arbiter, verified]. Verdict: Not testable as stated (no formula for the power the vacuum supplies) [arbiter]. |
| Capacitor across a magnet (2002) | Static E and H with no free current: div(E x H) = 0, so the Poynting flux circulates and delivers nothing | A charged capacitor or electret laid across a magnet, E perpendicular to H, is a "proven free energy generator" [web: Bearden letter, 15 Aug 2002, per arbiter] | div(E x H) = H . curl E - E . curl H = 0 for static fields (curl E = 0 and curl H = 0 there). This is the textbook "hidden momentum" / Feynman-disk case. No device, load or measurement is named. Verdict: Not testable as stated [arbiter]. |
The exact point of departure for the practical claim is a single line: Bearden's eq. [6], K_T = V_T x Coul_T, labelled "amount of trapped energy, each cycle" [read, eqs. 6-8]. In textbook terms that is the energy the battery gives up, not the energy that ends up in the collector and not a gift from the vacuum.
The 1994 charge split and the capacitor formula. In the 13 May 1994 paper Bearden splits ordinary current dq/dt into a massless "displacement current" dPhi/dt (Phi his symbol for potential) that carries the energy, and a "mass current" dm/dt that does not [read: page image p.44]. Units alone rule this out: dPhi/dt is volts per second, not amperes, and Phi x m would be volt-kilograms, not coulombs [read: deep read, sympy check]. Maxwell's displacement current is epsilon_0 dPhi_E/dt with Phi_E the electric flux (integral of E over an area, unit V m), not the potential. As support he cites SQUID (Superconducting QUantum Interference Device, a very sensitive magnetometer) detections of the magnetic field between the plates of a charging capacitor, with no reference [read: p.44]. That field is real and has been measured; the measurements confirm Maxwell's displacement current, which equals the conduction current in the leads, not Bearden's split of charge into massless dPhi/dt and mass dm/dt parts [memory, medium]. The same page offers the capacitor energy as "K = 1/2 (CV)^2" [read: page image p.44]. (CV)^2 = Q^2 has units of C^2, not joules; the correct form is K = 1/2 C V^2 = Q V / 2 (Section 2.2). This is probably a typo, but it is the one formula he offers as proof.
The 2004 remarks: the missing bridge from theory to device. The arbiter had noted that the patent never mentions the vacuum, so the theory (claims 1-4) and the only device with data (the MEG) were not linked in the source documents. The Investigator found the link in the USPTO file of the continuation application 10/081,847 (filed 26 Feb 2002). After the examiner rejected it, the applicants' remarks of 24 May 2004, filed with a declaration by co-inventor Bearden, set out his theory as the MEG's mechanism: the magnet as a broken-symmetry dipole absorbing vacuum energy (citing Lee and Yang 1957), the static field as a Whittaker set of energy flows, the Aharonov-Bohm localisation of B, E = -dA/dt pulses, a heat-pump analogy for COP above 1, a worked feedback example (COP 5 open loop, part fed back, "COP = oo"), Kron's "negative resistor" and non-conservation of energy in general relativity [web: https://archive.org/details/gov.uspto.patents.application.10081847, per arbiter; first 50,000 of 85,447 characters read]. These are the applicants' legal argument, written to obtain a patent. In the part read there is no calculation of how much power is absorbed.
The same remarks, as summarised by the Investigator, make further arguments, kept here for the record (applicants' legal argument and co-inventor's declaration; [web: https://archive.org/stream/gov.uspto.patents.application.10081847/10081847-image_file_wrapper/10081847-2004-05-24-00003-REM_djvu.txt, per arbiter]; none changes a verdict): - Negative energy from "holes" in the Dirac sea (Dirac's picture of the vacuum as a filled sea of negative-energy electrons, in which a missing electron behaves as a positron), with a reference to Bedini (which Bedini work is not recorded). - "Time-energy": E = t c^2 (t a time, c the speed of light), energy said to be converted from time into ordinary spatial energy in sharp gradients. Units check on the formula as summarised: t c^2 has units s x m^2/s^2 = m^2/s, but a joule is kg m^2/s^2, so as written it is not an energy; a factor in kg/s would be needed and the summary shows none [arbiter's calculation]. - The static field as a Prigogine dissipative structure (Ilya Prigogine's term for an ordered system kept far from equilibrium by a through-flow of energy). - Local versus global energy conservation in general relativity, citing Hilbert (1917), Logunov and Loskutov, and Kondepudi and Prigogine. - The heat-pump analogy with figures: COP 3-4 for real heat pumps and a "theoretical maximum ... about COP = 9.22", cited to Halliday and Resnick. A heat pump's COP above 1 is real because the extra heat is drawn from a cold reservoir and that heat flow can be metered; the analogy only works if the MEG's extra input is named and metered the same way, which is what the examiner asked for [arbiter].
One tension must be stated plainly: the patentees describe the energy source two ways. The 2000 patent text names the permanent magnet's flux as the fuel and concedes the magnet demagnetises [read]; the energy-budget bound in Section 5b applies to that version. The 2004 remarks name the vacuum as the source, which that bound does not address. The examiner had asked the applicants to show "how the magnetic flux stored within the permanent magnet is sufficient". Both descriptions are the patentees' own; neither has a measurement behind it [arbiter].
5. Worked example: the collector energy budget
This is Bearden's own scheme, with the simplest real parts. Every number is reproduced by Part A and B of example.py.
Setup. A battery with EMF (electromotive force, its open-circuit voltage) V = 12 V charges a collector capacitor C = 1 microfarad (1e-6 F) through whatever resistance R the wires and switch have. Then the battery is disconnected and the capacitor discharges into a load resistor.
Step 1: write the circuit equation. Kirchhoff's voltage law (voltages around a loop add to zero), with q(t) the capacitor charge and i = dq/dt the current:
V = i R + q / C
Step 2: solve it. With q(0) = 0 the solution is
q(t) = C V (1 - e^(-t/RC))
i(t) = (V/R) e^(-t/RC)
Current starts immediately at V/R and decays with time constant RC. Nothing about epsilon/sigma enters.
Step 3: energy the battery gives up. The battery pushes current i against its EMF V:
W_batt = integral_0^inf V i dt = V * (total charge) = V Q = C V^2
= 1e-6 x 12^2 = 144 microjoules
Q = C V = 12 microcoulombs. Note: 144 microjoules is exactly Bearden's K_T = V_T x Coul_T.
Step 4: energy turned to heat in R.
W_heat = integral_0^inf i^2 R dt = (V^2/R) * integral e^(-2t/RC) dt = (V^2/R)(RC/2) = (1/2) C V^2
= 72 microjoules
R cancels. Whether R is 0.01 ohm or 10,000 ohm, the loss is 72 microjoules; a smaller R just makes it happen faster (example.py checks R = 0.01, 1, 100 and 10,000 ohm numerically and gets 72.00 uJ each time). Good copper does not cause the loss and "doped" wire does not prevent it.
Step 5: energy stored.
W_cap = Q^2 / (2C) = (1/2) C V^2 = 72 microjoules
Check: 144 = 72 + 72. Energy is conserved.
(Note: Bearden's 13 May 1994 paper prints the stored energy as "K = 1/2 (CV)^2" [read: page image p.44]. That has units of C^2, not J; the correct form is 1/2 C V^2, as used here. Probably a typo.)
Step 6: discharge into the load. The capacitor can deliver at most what it holds: 72 microjoules (example.py: 72.00 uJ into a 50 ohm load).
Step 7: the score. Coefficient of performance (COP, energy delivered to the load divided by energy taken from the source) = 72 / 144 = 0.5 per cycle. After one million cycles the battery has passed 12 coulombs and given up 144 J, and the loads have received 72 J. A car battery of, say, 50 ampere-hours (180,000 C) would be flat after about 15 billion such cycles, exactly as fast as its charge count predicts [calc].
Step 8: the fair "weak form". Is there anything to Bearden's intuition that the charging step wastes energy? Yes. If you raise the supply in N equal small steps instead of one jump, each step loses only (1/2) C (V/N)^2, so the total loss is C V^2 / (2N):
| N steps | supplied (uJ) | stored (uJ) | lost (uJ) | efficiency |
|---|---|---|---|---|
| 1 | 144.00 | 72.0 | 72.00 | 0.500 |
| 2 | 108.00 | 72.0 | 36.00 | 0.667 |
| 10 | 79.20 | 72.0 | 7.20 | 0.909 |
| 100 | 72.72 | 72.0 | 0.72 | 0.990 |
This is "adiabatic charging", used in low-power electronics [memory, high for the principle]. Efficiency climbs toward 1 and never passes it.
Bearden makes the same point himself in the 13 May 1994 paper: "With ordinary capacitors ... one can readily demonstrate that the efficiency can approach 1.0 rather than 0.50 as expected" [read: p.45]. That is the textbook adiabatic (step) charging result that PART B of example.py reproduces: efficiency N/(N+1) with N steps (counting energy drawn as (1/2) C V^2 (1 + 1/N)), approaching 1 from below and never giving a gain above 1 [calc]. (On p.43 the same paper says an ordinary capacitor dissipates as much as it stores, which contradicts p.45 [read].) His 15 Feb 1994 update proposes "several hundred" steps [read: p.30]; at N = 300 the efficiency is 0.997 [calc]. Loss reduction is real engineering; gain is not available.
Step 9: the relaxation time numbers. Bearden's own figures are correct arithmetic but mean something else: - Copper: tau = 8.854e-12 / 5.96e7 = 1.49e-19 s (he says 1.5e-19 s) [calc; read]. - Fused quartz, sigma about 1e-17 S/m: tau = 8.9e5 s = 10 days using epsilon_0, or 46 days using quartz's relative permittivity of about 4.5 [calc; conductivity is an order-of-magnitude value, memory medium]. His "10 days" is the right order. - To get tau = 1 ms you need sigma = epsilon_0 / tau = 8.85e-9 S/m, 6.7e15 times less conductive than copper: an insulator, which contradicts his "still a good conductor" [calc]. - The real charging delay for 1 ohm and 1 microfarad is RC = 1 microsecond.
5b. A second check: what the MEG patent's own numbers say
(Parts D and E of example.py.)
The headline arithmetic [read, patent; calc]: - Input at 100 V: 100 V x 0.140 A = 14.0 W. - Output: 4000 V x 0.012 A = 48.0 W "for each of the output coils". - 48 / 14 = 3.4286. The patent prints 3.44. - If both coils really gave 48 W the total would be 96 W, COP 6.86; the patent does not say this. - The patent states plainly that "the data was extrapolated" to 100 V; the switching circuit limited measurements to 75 V [read].
What Figs. 7 and 8 actually show [read: page image p.3]: measured points from about 10 V to 75 V input. Output power (Fig. 7) rises from about 1 W to about 22 W; the COP points (Fig. 8) rise from about 1.7 at 10 V to about 2.6 at 75 V. The solid curve is a least-squares polynomial fit [read, text] and is drawn on to 100 V, where it reaches about 48 W and COP about 3.4. So the 3.43 figure is the fitted curve carried 25 V past the last data point; the largest measured COP is about 2.6. The figures do not say whether Fig. 7's output power is per coil or total.
What Fig. 6 shows (75 V input) [read: page image p.4; values approximate]: - Each output coil voltage (Figs. 6E, 6F) is a roughly sinusoidal AC wave of about +/-1500 V peak (3000 V peak-to-peak), period about 23 microseconds (two 11.45 microsecond drive pulses). - Drafting note: the description gives 11.45 microseconds per pulse, which matches 87.5 kHz (1/87.5 kHz = 11.43 us), but claims 10 and 26 of the granted patent say "approximately 11.5 milliseconds", 1000 times longer (about 87 Hz) [read: patent pp.14-15; calc]. This is almost certainly a unit slip, and an example of imprecise patent drafting. - Output currents (6G, 6H) swing about +/-40 mA with sharp spikes at each switching edge, and their shape does not track the voltage wave closely, so there is a phase shift between them. - Input current from the battery (6D) has spikes to about 1.2 A at each switching edge and spends much of the time near zero or slightly negative (down to about -0.3 A).
Three consequences, each a simple calculation [calc; interpretation is inference]: 1. Turns ratio. 450/40 = 11.25. An ideal transformer at 75 V gives 844 V; Fig. 6 shows about 1500 V peak, 1.8 times that. At 100 V the ratio gives 1125 V, and 4000 V is 3.6 times that. Ringing or resonance explains high peaks, but resonance stores and returns energy; it does not add average power. Note also that an AC wave symmetric about zero has an average voltage of zero, so "4000 volts average" [read] cannot be a plain average. Linearly scaling Fig. 6's 3000 V peak-to-peak from 75 V to 100 V gives 4000 V; whether that is what the patent meant is not stated (inference, unconfirmed). 2. Phase matters. Take Fig. 6's rough values, 1500 V and 40 mA peak sine waves. Real power per coil is (1/2)(1500)(0.040) cos(phase) = 30 W x cos(phase). At 0 degrees phase, 30 W; at 45 degrees, 21.2 W; at 80 degrees, 5.2 W; at 90 degrees, zero. "V_rms x I_rms" stays at 30 VA (volt-amperes) for every phase, and "average |v| x average |i|" stays at 24.3. Only the time average of v(t) i(t), sampled fast and simultaneously on both channels, gives power. The patent says its computed real power "agree[s] with RMS power measured using a Tektronix THS730 digital oscilloscope" [read]; that is a check of arithmetic on the same probes, not an independent instrument [arbiter]. The instrument maker has said what it thinks of that cross-check: a Tektronix customer-support reply, forwarded by the MEG builder Jon Flickinger and posted by Naudin in 2002, states that "the RMS measurement on a math waveform is bogus and has no value" for non-sinusoidal waveforms and that true power is the mean of the point-by-point product v(t) i(t) [web: http://jnaudin.free.fr/meg/ths720p.htm, Investigator's reading; the arbiter's re-fetch was refused]. That is one support reply relayed by a builder; whether it describes the patentees' exact THS730 procedure is inference. It removes the cross-check as support, but it does not show the patent's primary calculation (real power from sampled V and I, averaged over the period) was wrong, and the patent gives no raw data to show how that calculation was done [arbiter]. 3. Input sensitivity. Fig. 7 and 8 together imply an input of about 22 W / 2.6 = 8.4 W at 75 V, i.e. an average battery current of only about 0.11 A, made of 1.2 A spikes and negative stretches. A 20 mA offset in the current probe would change the input power by 18 percent, and a 50 mA offset by 45 percent. A small average built from large positive and negative pieces is exactly where measurement error concentrates.
The patent names the magnet as the fuel. This is the patentees' own characterization of their device, written to obtain a patent; it is not an outside measurement. In its own words [read: patent p.12]: the generator is "an open dissipative system, receiving, collecting, and dissipating energy from its environment; in this case, from the magnetic flux stored within the permanent magnet. Continued operation of the electromagnetic generator 10 causes demagnetization of the permanent magnet." Rare-earth magnets (samarium cobalt or neodymium-iron-boron) are preferred "since such a magnetic material has a relatively long life in this application." It then compares the device to "a nuclear reactor, in which a number of fuel rods radiate energy which is used to keep the chain reaction going". No demagnetisation rate, magnet lifetime or before/after magnet measurement is given [read; deep read].
Worked question: how long would the magnet last? [read dimensions; calc] - Magnet volume: 40 x 25.4 x 38.1 mm = 38.7 cm^3 = 3.87e-5 m^3. - Stored energy, two estimates: (BH)max x volume, where (BH)max (the maximum energy product, a magnet-grade figure in kJ/m^3) for sintered NdFeB spans about 190-450 kJ/m^3 [web: https://en.wikipedia.org/wiki/Maximum_energy_product and https://www.samaterials.com/content/maximum-energy-product-in-magnetic-materials.html, search summaries, pages not opened, per arbiter], gives about 7 J to 17 J; B^2/(2 mu_0) x volume with B = 1.2 T gives about 22 J as an upper figure. So roughly 7-17 J, to 22 J [calc; example.py Part D]. The patent prefers rare-earth magnets but does not say which material the tested unit used; a ferrite magnet would store less [memory, medium, per arbiter], so the bound only tightens. - Claimed excess at 100 V: 2 x 48 W - 14 W = 82 W. Time to use 7-22 J at 82 W: 0.09 to 0.27 s, i.e. well under 1 s. One coil only (34 W excess): 0.22 to 0.65 s. Even the measured 75 V case (about 14 W excess) would use it up in 0.5-1.6 s. - A factor-of-ten error (anywhere from about 0.7 J to about 220 J) still gives at most 2.7 s at 82 W, not "a relatively long life". - The reviewers' earlier figures of 1.5-4 J rested on guessed magnet volumes (5-10 cm^3); the figures here use the patent's own 38.7 cm^3 [arbiter].
So the patent's own stated energy source cannot supply its own claimed output for more than a fraction of a second. (Also, a magnet whose flux is merely switched back and forth returns to the same state each cycle and supplies no average power, so this bound applies to the patent's stated mechanism; the closed-cycle energy argument is the decisive one [arbiter].)
None of this proves the inventors' measurements wrong. It shows that the published numbers are not enough to establish COP > 1, and that the measurement is of the kind where errors are easy. That a measurement artefact is the likely explanation is the arbiter's inference from these numbers, now "better supported inference" after the Investigator findings (Section 5c), but still not an established finding [arbiter].
5c. What the builders and the patent offices add (Investigator findings, 2026-10-05)
(Part F of example.py.) All items are the Investigator's readings of web pages, carried by the arbiter [arbiter].
Naudin's v2.1 numbers: where you draw the boundary decides the ratio. Naudin's status of 6 Dec 2000 for his MEG v2.1 (ferrite magnets, Metglas cores, a driver at about 16-20 kHz, a 10 ohm current shunt on a Tektronix THS720P scope) gives [web: http://jnaudin.free.fr/meg/megv21.htm, per arbiter]: - the control board alone draws 1.75 W with no load; with a 9 W lamp as load the DC input is 3.25 W; he credits 1.5 W of that to the lamp, against "a measured OUTPUT = 6.76 Watts"; - 6.76 / 1.5 = 4.5, but 6.76 / 3.25 = 2.08 if the board is inside the boundary, as it must be for a self-running device [calc; arbiter]. This is the same question as whether the patent's 14 W includes the oscillator supply; - he states the output is "really measured" by several instruments "but unfortunately measurement artifacts remain possible"; a "conditioned" carbon resistor or MOV (metal oxide varistor, a nonlinear voltage-dependent resistor) load "is REQUIRED" to get the output figures; "in most of cases the 'apparent' power measured seems greater than the heat dissipated by Joule's effect" in the load; calorimetry and closing the loop "has not yet been done".
Three different "1.75" values appear in this record and must not be confused: the v2.0 page's "measured COP is 1.75" (7 Nov 2000); the v2.1 control board's no-load draw of 1.75 W; and Flickinger's "COP 1.75 to 5" on MOV loads [arbiter].
Can radiation explain Naudin's gap? Naudin attributes the difference between apparent power and load heat to electromagnetic radiation (2.8 mT near the load at 16 kHz; 1250 V/m at 50 cm). A small loop of N turns and area A radiates as if it had a radiation resistance R_rad = 31,171 (N A / lambda^2)^2 ohm, where lambda = c/f is the wavelength (18.75 km at 16 kHz) [memory, high, formula]. For 1 turn of 0.03 m^2 this is about 2e-16 ohm; for 100 turns of 0.01 m^2, about 2.5e-13 ohm [calc; example.py Part F3]. With amperes of current that radiates nothing measurable. And 1250 V/m, if it were a travelling wave, would carry E^2/377 = 4.1 kW/m^2 (377 ohm is the impedance of free space), far more than the device draws; so it must be a stored (reactive) near field, not radiated power [calc]. Radiation cannot account for the gap. How Naudin measured the 2.8 mT and 1250 V/m is not known.
Linear versus nonlinear loads. Jon Flickinger, another independent builder, writes in his undated "MEG Notes" that with linear resistive loads "it is a waste of time ... if one expects to see any excess energy"; COP above 1 appears only with nonlinear MOV loads, "COP's ranging from 1.75 to 5"; and he asks: "If anyone should experience valid COP's >1 with standard linear loads, please speak out!" [web: http://jnaudin.free.fr/meg/megnot01.htm, per arbiter]. This is evidence against excess power on linear loads, from one uncontrolled report. It fits what an artefact predicts: a nonlinear load makes the current waveform non-sinusoidal (harmonic-rich), which is exactly where V_rms x I_rms overstates true power (Section 3.8).
The self-powering condition. Let c be the open-loop COP and f the fraction of the output fed back to drive the input. The loop sustains itself only if f c >= 1 [arbiter, from the mathematician]. At f = 0.5 that needs c >= 2, which the patent's measured range (about 1.7 to 2.6) straddles; c = 1.7 would need 59% of the output fed back, c = 2.6 only 38% [calc; example.py Part F4]. So closing the loop adds no new physics: it restates c > 1. Its value is practical: a run with the supply removed either keeps going or stops, whatever instruments are used.
What the patent offices did. - USPTO. The continuation 10/081,847 was rejected on 23 Jan 2004 by examiner Shawn Riley under 35 USC 101 (lack of utility), as violating the first and second laws of thermodynamics, without any test. The same examiner required a model or demonstration, writing that the applicant must show "how the magnetic flux stored within the permanent magnet is sufficient to contribute usable electrical energy". The applicants asked to defer that (24 May 2004); he refused in the final rejection (26 Jul 2004); the application was abandoned on 16 May 2005 for failure to respond, with no appeal in the file [web: archive.org USPTO file; PlainSite docket, per arbiter]. The arbiter's reading: rejected on theory by one examiner, a demonstration invited and not supplied, and the file does not say why. That is not a blocked test and not "tested and failed". The examiner (the agency's voice, with a stake in the rejection) argued from a windmill that COP cannot exceed 1, which confuses COP (output over operator-paid input) with efficiency (output over all input); the applicants' distinction was fair, but the core of his demand, a source large enough, is sound [arbiter]. The examiner's other grounds in the same 2004 rejection [web: https://archive.org/stream/gov.uspto.patents.application.10081847/10081847-image_file_wrapper/10081847-2004-01-23-00002-CTNF_djvu.txt, per arbiter]: rejection under 35 USC 112 (first and second paragraphs: the description and claims must enable and define the invention); "obviousness-type double patenting" over claims 1-29 of the parent US 6,362,718 (the new claims were not patentably distinct from the granted ones); drawings objected to ("when the input power source (e.g., 38) is taken away, the remaining circuit becomes inoperable", and the COP figures as "violating at least the first and second laws of thermodynamics"); the title, which "is seen as a transformer which relies on the concept of perpetual motion"; page 2 of the specification missing; and the court case Newman v. Quigg (1988) cited for the demonstration requirement. The double-patenting ground matters for the abandonment: if the continuation added little beyond the parent, dropping it cost the applicants little, since they already held the parent patent. That is one documented explanation (a redundant continuation) for why the demonstration was not supplied and the application lapsed; it supports that reading without excluding the others in "Where the record goes silent" [arbiter]. The parent patent (application 09/656,313) was granted on 26 March 2002 [read: patent front page]. Whether its examiner raised a utility objection or asked for a demonstration is unknown: no such demand is known, and the parent file was not obtained (archive.org returned "Item cannot be found") [web: https://archive.org/details/gov.uspto.patents.application.09656313, per arbiter]. - EPO. EP 1 446 862 B1 was granted on 23 Oct 2013 after a 12-year prosecution (two "deemed withdrawn" episodes, both cured). Its claim 1 requires non-saturating switching and feeding output power back to drive the switching: the self-powering loop. The EPO found the device's structure anticipated by an older flux-switching patent, US 4,006,401 (deRivas, 1977); the only novelty is the self-powering feedback, which is also the one feature never demonstrated [web: https://patents.google.com/patent/EP1446862B1/en; EPO Register, per arbiter]. A grant is not a performance test. - The European text. Compared with the US patent, the granted European description keeps the extrapolated "3.44" and the demagnetisation paragraphs, but lacks the COP-versus-voltage plot and the sentences that output was "substantially higher than the input", the THS730 agreement, and "all ... significantly greater than 1" [web: Google Patents texts compared by the Investigator, per arbiter]. Competing readings: trimming of unsupported performance statements under examiner pressure (clarity objections were raised in 2006); routine redrafting; or a choice to carry less checkable evidence. The 2007/2008 replies and 2013 minutes would decide. No intent is inferred. The effect is that the 3.44 now stands without the data it was extrapolated from.
Bearden's 15 Aug 2002 letter (co-inventor and company spokesman) describes "successful little experimental devices", about $9M per year for three years sought, demonstrations given only to prospective investors, and planned 2.5 kW and 10 kW units [web: https://billstclair.com/www.cheniere.org/correspondence/081502.htm, per arbiter]. No investor's measurement report has been found. The letter also says that "every previous legitimate COP>1.0 system ... has been firmly defeated right at this same 'vertical cliff'" of scale-up funding; this is Bearden's characterization, and no system or record is named in the letter as summarised. It reserves 10% of sales in poor countries for wells, schools and clinics; announces his book Energy from the Vacuum: Concepts and Principles (2002), "nearly 1,000 pages"; and says "Our first patent has issued, and two others are in various stages", which the Investigator matches to the US continuation (abandoned 2005) and the European/PCT (Patent Cooperation Treaty) family (granted 2013) [web, same letter, per arbiter].
Company-records leads (none opened). A Google search for Magnetic Energy Ltd, Huntsville, Alabama, returned [web: https://www.google.com/search?q=%22Magnetic+Energy+Ltd%22+Huntsville+Alabama, per arbiter]: a Manta directory listing of the company as "Noncommercial research organizations"; a ChamberofCommerce.com listing in Huntsville; a Facebook page "Magnetic Energy Ltd (@MagneticEngergyLtd)" describing MEG technology; a Bearden paper on pdfcoffee giving the address "Magnetic Energy Ltd. P.O. Box 1472 Huntsville, AL 35807"; and a ZPEnergy "Letter to JLN Laboratories" in which a writer from the AIAS group (the Alpha Institute for Advanced Studies, M. W. Evans's group, behind the Evans et al. paper [memory, medium]) offers to work "with your company and Magnetic Energy Ltd. to develop the MEG commercially". These are directory traces of the company, not registration records; the ZPEnergy letter is a lead on the AIAS group's commercial interest, which bears on the stake behind the Evans et al. paper (see "Who is speaking") [arbiter].
Related prior art. Flux-switching "static generators" before the MEG include Starr US 2,802,170 (1957), Toffolo US 3,087,108 (1963), Garron US 3,368,141 (1968), Galley FR 2,312,135 (1975), deRivas US 4,006,401 (1977), Richardson US 4,077,001 (1978) and Flynn WO 00/07285 (Parallel Path Technology) [web: EPO annex and http://jnaudin.free.fr/meg/meg.htm, per arbiter]. None has been read by any reviewer; which of them claim COP > 1 is unknown.
6. Evidence table
All verdicts are the arbiter's final verdicts.
| Claim | Tested? | Result | Replicated? | Verdict |
|---|---|---|---|---|
| 1a. Whittaker/Ziolkowski expansion of a potential into waves | Mathematics, checked | Correct theorem; no energy content | n/a | Established |
| 1b. "Energy = potential" | n/a | Unit error (J vs J/C) | n/a | Internally inconsistent |
| 1c. Each potential holds infinite energy from the universe | No experiment proposed | Stored field energy is finite and computable | n/a | Refuted by calculation (stored-energy reading); Not testable as stated (reservoir reading) |
| 2a. Vacuum EM waves are longitudinal | Yes, by a century of polarisation and radio experiments (not aimed at Bearden) | Transverse; div E = 0 | Many times | Tested, not supported; Refuted by calculation |
| 2b. Charge varies with ambient potential | Yes, indirectly: neutrality of matter to ~1e-21 e [web: https://arxiv.org/abs/1102.2766]; constant e in accelerators | Not supported | Yes | Tested, not supported |
| 2c. d/dt[(phi)(m)] = (dphi/dt)(dm/dt) | n/a | Product rule violated; units wrong | n/a | Internally inconsistent |
| 2d. EM wave = oscillation of the rate of time flow; photon-antiphoton to graviton | No | No measurable prediction | n/a | Not testable as stated |
| 3a. Collector: battery never depletes; runs a car or grid | Never built with data by Bearden, who released the scheme openly ("I am freely giving away what required me an arduous 30 years of my life to discover" [read: p.4]); no published build by anyone found; ordinary capacitor charging measured countless times (see "Where the record goes silent") | Energy budget gives COP <= 0.5 per cycle | n/a | Refuted by calculation |
| 3b. "Current-free" charging via relaxation time | n/a | tau = 1 ms implies an insulator | n/a | Internally inconsistent |
| 3c. Lower-loss charging (weak form) | Not by Bearden; principle used in adiabatic circuits | Reduces loss; no gain | n/a | Plausible, untested (as Bearden's scheme); no overunity content |
| 3d. Step-charged collector, "several hundred" steps (15 Feb 1994) | No measurement reported [read: pp.28-41] | Efficiency N/(N+1), 0.997 at N = 300; never above 1 | n/a | Plausible, untested as loss reduction; no gain |
| 3e. Source "can operate at an efficiency greater than unity", the vacuum making up the difference; device "can conceivably run itself" (15 Feb 1994) | No input or output number given [read: pp.28, 30, 32] | Permissibility argument only | n/a | Not testable as stated; if it means sustained self-running, Refuted by calculation |
| 3f. Operational efficiency alpha > 1 as "overunity"; "conventional" alpha "always underunity" (15 Feb 1994) | Ohm's law | Alpha > 1 is routine (12 V, 0.1 ohm source, 10 ohm load: alpha = 100, true efficiency 0.99) | n/a | Alpha > 1 Established as ordinary; "always underunity" Refuted by calculation |
| 3g. Fogal charge-blocking semiconductor (CBS) "negative resistor" (13 May 1994, Figs. 1-2): blocks dm/dt, passes dPhi/dt into a step-up transformer; "very high variable power gain" | No data of any kind given; no structure, ratings or source for the device in the paper [read: pp.44-46]. Outside the paper, William J. Fogal's own patents (US 5,196,809; US 5,430,413, "High gain, low distortion, faster switching transistor") describe a transistor with a 0.2-100 uF electrolytic capacitor from emitter to ground, claiming gain "at least twice" the bare transistor's and no output above input or vacuum energy [web: https://patents.google.com/patent/US5430413A/en, fetched summary, per arbiter] | Two characterizations: Fogal's (a higher-gain amplifier) and Bearden's (an overunity dm/dt blocker). As Bearden describes it, "blocks electrons, passes displacement current" is what a series capacitor does, and Fogal's capacitor-at-the-emitter structure fits that reading; Bearden concedes ordinary meters would show energy conserved | No measurement of gain above input found | Reported gain: Never properly tested (no data). Fig. 2 self-powered loop (supply removed after start-up): Refuted by calculation (closed-loop energy budget, as claims 3 and 5) |
| 4a. Closed classical EM cannot be overunity | Poynting's theorem; all circuit measurements | Correct | Yes | Established |
| 4b. Overunity devices tap a hidden vacuum source | No formula or signature given | Untestable | n/a | Not testable as stated |
| 4c. Sweet's vacuum triode amplifier (VTA) | No independent test. The account: Sweet reported about 330 microwatts in and over 500 W out, kept the device and the magnet-conditioning process from independent test, and (per Bearden, his collaborator and promoter) feared for his life [web, search summaries of http://amasci.com/freenrg/sweetvta.html and http://www.rexresearch.com/sweet/1nothing.htm, per arbiter; pages not opened]. Documented in Sweet's and Bearden's accounts; no second source found yet for the stated reason | No measurement by a third party in either direction | No | Never properly tested |
| 4d. Maxwell's quaternion theory had gravity-EM unification that Heaviside removed | Historical record | Maxwell's own 1865 gravity attempt failed on negative energy [web, per arbiter] | n/a | Tested, not supported (judged on the historical evidence in the Result column; the earlier label "Refuted by the historical record" was not on the verdict scale) |
| 4e. Quaternion vs vector notation | Historical | Equivalent for classical EM | n/a | Historically superseded but instructive |
| 4f. MEG mechanism in the 2004 USPTO remarks (magnet absorbs vacuum energy; curl-free A outside the core; E = -dA/dt pulses as free input) | No; no formula for the power supplied in the part read (50,000 of 85,447 characters) | E = -dA/dt is ordinary induction; a curl-free A gives no EMF round a loop; no extra energy | n/a | Not testable as stated |
| 4g. The MEG self-powered run as the test of a vacuum source (Bearden names the MEG and the vacuum in 2004; EP claim 1 requires the self-powering loop) | No run with the supply removed on record; the USPTO examiner's demonstration request (2004) was not answered | None | No | Never properly tested |
| 4h. Capacitor or electret across a magnet as a "proven free energy generator" (Bearden letter, 15 Aug 2002) | No device, load or measurement named | Static div(E x H) = 0: the energy flow circulates, no net power | n/a | Not testable as stated |
| 5a. MEG measured COP > 1 at 10-75 V (measured about 1.7-2.6 per Fig. 8) | Accounts, each with its source and stake: (i) the patentees, COP about 1.7-2.6 measured [read]; (ii) Magnetic Energy Ltd builds per Bearden (e-mail, 2 Dec 2000): "COP = 17 to 20, higher in some experimental build-ups", one unit "pushed ... to destruction of the output coil" at "COP>100" [web: https://billstclair.com/www.cheniere.org/correspondence/120200.htm, per arbiter]; (iii) J. L. Naudin, independent private builder, own instruments: "The measured COP is 1.75" (v2.0, 7 Nov 2000) [web: https://www.novakcorp.com/energy/experiments/megv2.htm, per arbiter], and later builds at COP 1.3 and 7.3 as reported by Bearden (Naudin's own pages for those did not load); (iv) a university replication "under U.S. Department of Defense auspices", arranged per Bearden, names withheld; the only account of its end (two search summaries, source page not found) says the administration "refused to sign a noncircumvention agreement, and consequently the process was stopped" before any measurement [web, per arbiter]; (v) Naudin v2.1 (6 Dec 2000): 6.76 W apparent output on 3.25 W total input, ratio 2.08, only with nonlinear loads, apparent power above load heat, "measurement artifacts remain possible", no calorimetry or loop closure [web, per arbiter]; (vi) Flickinger, independent builder: COP 1.75-5 with MOV loads, nothing with linear loads [web, per arbiter]; (vii) Tektronix support (relayed by Flickinger): the scope-math "RMS power" the patent cites as its cross-check is invalid for non-sinusoidal waveforms [web, per arbiter]; (viii) USPTO continuation rejected on theory (2004), demonstration requested and not supplied, abandoned 2005 [web, per arbiter] | No neutral, calorimetric test on record; no true v(t) i(t) integral or closed-loop run reported by anyone, either way. Builder reports find excess only on nonlinear loads (uncontrolled, single sources) | No independent confirmation | Never properly tested (sub-findings: the patent's cross-check method is one the instrument maker calls invalid; excess only on nonlinear loads in builder reports; no calorimetry) |
| 5b. COP 3.44 at 100 V | No; extrapolated from a curve fit | 48/14 = 3.43; voltage 3.6x turns ratio unexplained; the European text keeps 3.44 but drops the plot it was extrapolated from | No | Not testable as stated |
| 5c. Self-powered sustained output | No self-running trial reported in the patent [read]. The one recorded attempt: Naudin, "I am trying to close the loop, but at this moment, I have not yet succeed ... the device stops after a while", which he attributes to step-down losses [web: novakcorp megv2 page, per arbiter] | Excluded by energy conservation for a closed device. Naudin's attempt agrees with this, but the verdict rests on the energy budget, not on one builder's attempt. Running needs f c >= 1 (Section 5c) | No | Refuted by calculation (conditional on a closed system and on the patent's own naming of the magnet as the fuel; the vacuum-source version is row 4g) |
| 5d. The ordinary flux-switching circuit as a device (no overunity claim) | EPO: structure anticipated by US 4,006,401 (1977) | No performance data for it as a sub-unity generator in the sources | n/a | Plausible, untested as stated |
Who is speaking. Everything known about MEG performance comes from people with a stake in it or close to it. The mechanism and COP figures in the patent are the five patentees' description, written to obtain a patent. Bearden's later figures (COP 17-20, >100) are a co-inventor's, speaking for the company developing the device. Naudin and Flickinger are outside builders, using their own instruments with no neutral observer; Naudin's page shows an interest in the MEG succeeding, and he himself says artefacts remain possible. The 2004 USPTO remarks are the applicants' legal argument for a patent. The Tektronix statement comes from the instrument maker, with no stake in the MEG, but as one support reply relayed by a builder and not as a test of a MEG. The USPTO examiner and the EPO speak for agencies; the examiner's rejection is the agency's characterization, and patent grants (US 2002, EP 2013) do not require a performance test [memory, high, per arbiter]. The Evans et al. (2001) theory paper takes COP > 1 as given; its author list, as shown on Naudin's index page, includes T. E. Bearden [web: http://jnaudin.free.fr/meg/meg.htm, per arbiter], so it is a co-inventor's own account, not outside support.
On "dismissed without testing": the arbiter found no documented case of blocked publication for these claims. On the US continuation, the examiner rejected the claims on theory without a test, but he also asked for a demonstration, which the applicants did not supply before letting the application lapse; the file does not say why. A test was invited, not blocked. The one independent MEG replication that was arranged stopped, on the only account found, over a non-circumvention agreement demanded by the developers, before anything was measured; that is neither a failed test nor a refusal to look on technical grounds. Bearden's COP 17-20 and >100 figures come with no waveforms or input/output values, so they do not change the verdict. The physicist argued that the builders' reports put the strong form at "Tested, not supported" at an informal level; the arbiter rejected this, because the scale requires controlled experiments and these reports are uncontrolled; Flickinger's linear-load result is stated in full as evidence against excess, without being treated as a controlled refutation. The honest status of the MEG is "nobody neutral has tested it", not "tested and failed" and not "suppressed" [arbiter].
Where the record goes silent
In four places the record stops where you would expect it to continue. A silence like this is something to explain, not a neutral gap. For each, the arbiter sets out the competing explanations, the evidence for and against each, and the records that would decide [arbiter, Claims 3-5, QE revision item 5 and Investigator findings]. None of these analyses changes a verdict. The Investigator's 2026-10-05 run added records to item 1 only; it did not reach the Sweet, Fogal or collector records (rexresearch and amasci were not searched), so items 2-4 are unchanged.
- The MEG after 2001. In 2000-2001 Bearden reported COP 17-20 builds, an arranged replication under DoD (US Department of Defense) auspices, and Magnetic Energy Ltd "in intense negotiations with several large financial enterprises, for the capitalization necessary" (status page, 11 April 2001) [web: https://www.rootaction.net/~jay/science/www.cheniere.org/megstatus.htm, per arbiter]. Twenty years later a ZPEnergy article (A. Craddock, "Twenty years after Tom Bearden's MEG", 1 Jan 2021) repeats Bearden's 2000 statements and records no later test, product or data [web: https://www.zpenergy.com/modules.php?name=News&file=article&sid=3918, per arbiter]. The Investigator found that the silence has a documented trace: the US patent stood in the inventors' names until a 2006 assignment to Magnetic Energy, Ltd. (Alabama), with a 2006 security agreement in favour of Colmac Holdings Limited (Canada), a lapse and reinstatement in 2006 and fees paid to year 12 (2013); the European patent was prosecuted for 12 years to grant in 2013; and a search of the Alabama Secretary of State register found no entity named "Magnetic Energy" (one form search; directory listings in Huntsville not opened) [web, per arbiter]. Possible explanations: - Measurement artefact (reactive power, an uncounted drive supply, ringing). For: the 1500-4000 V output against the turns ratio, no power factor given, Naudin's failed attempt to close the loop; and now the Tektronix rule against the patent's cross-check, excess appearing only with nonlinear loads, apparent output above load heat in Naudin's v2.1, and the removal of the measured plot from the European text (reason unknown). Against: no calorimetry or full-waveform analysis of any MEG has been published either way; the patent's primary method may have been valid; the substance of the investor demonstrations is unknown. - Commercial failure (no capital raised). For: the 2001 "intense negotiations" with no announced result; the 2002 letter's "successful little experimental devices" with funding still sought; the 2006 security agreement with a Canadian holding company; a late-paid 2010 fee; no product; no Alabama entity found. Against the simple story that the project vanished in 2002: fees paid to 2013 and a 12-year European prosecution show sustained spending. - Trade secrecy. For: the non-circumvention demand and the withheld names of the university and agency (2000); Bearden's stated policy of demonstrations only to prospective financiers (2002); the demonstration request left unanswered (2004). Against: the structure is published in two patent families, so what was kept private is the measurements, not the design. And the unanswered 2004 demonstration request has another documented reading: the examiner also rejected the continuation for double patenting over the parent's claims 1-29, so it may simply have been a redundant continuation not worth defending (Section 5c). - Absorption into another firm or programme under a new name. The Colmac security agreement is the only lead; Colmac is otherwise unidentified. No evidence for or against. - Suppression. Bearden cites the establishment's "fierce dogmatic hassle" over cold fusion as his reason for withholding names [web, Bearden e-mail, per arbiter], but that is a reason for secrecy, not evidence that anyone blocked the MEG. What was found is a theory-based rejection by one examiner, a granted European patent, and a demonstration invited and not given. Nothing found supports suppression; nothing excludes it. The university non-circumvention account still rests on search summaries only. - What would decide: the Colmac security agreement and any release (USPTO assignment records, reel/frame 017759/0537, also 017706/0167 and 017746/0237) and Colmac's record in Corporations Canada; Magnetic Energy Ltd's company record; the rest of the 2004 remarks and Bearden's declaration; the EPO 2006 annex pp.4-5, the 2007 and 2009 replies, the 2008 annex and the 2013 minutes; the parent US file 09/656,313; any investor-demonstration measurement report; the identity of the university and DoD programme (and the primary source for the non-circumvention account); and Naudin's v3.0/v4.0 pages and the COP 1.3 and 7.3 builds.
- Sweet's VTA. No independent measurement, no surviving unit in a lab and no published conditioning process. Possible explanations: a measurement artefact that withholding the device kept anyone from finding (consistent with the lack of a test, no direct evidence); trade secrecy over an unpatented process (consistent with the reported secrecy of the conditioning step); fear of harm, as Bearden states (his account only, no corroborating record found yet); the process lost with the inventor (Sweet's death in the mid-1990s is [memory, medium], not checked). What would decide: any surviving unit or Sweet's notes, and any contemporary third-party measurement (the rexresearch and amasci compilations, not opened in this review).
- The Fogal CBS after 1994. Bearden presented it as making overunity "readily obtained"; no measurement of power gain above input, no build of his Figs. 1 or 2, and no later Fogal patent claiming gain above input were found. Possible explanations: the gain was a meter artefact of the kind Bearden himself describes (an ammeter that reads dPhi/dt as ordinary current), so careful builders found nothing to report; commercial or trade secrecy on Fogal's side; or builds that are not recorded in the sources searched. What would decide: the rexresearch Fogal compilation (not opened), contemporary press reports, and any bench data on the Fogal transistor.
- The collector. Bearden gave the scheme away so that it would "trigger a thousand, or even ten thousand, scientists and engineers to develop overunity energy devices" [read: p.4], so secrecy cannot explain the silence. Possible explanations: builders found no gain and did not publish null results (consistent with the energy budget in Section 5; no record of such attempts found); the scheme was too under-specified to build (no component values, and the 1 ms "good conductor" target is an insulator); or builds exist that are not recorded in the sources searched. What would decide: a search of free-energy forums and archives for collector builds with measurements.
7. What a learner should take away
What is solid. - Bearden is right that closed classical electromagnetism forbids overunity: if a device really gave more out than in, something outside the usual accounting would have to be feeding it. This is a correct reading of Poynting's theorem. - He is right that potentials are physically important. The Aharonov-Bohm effect shows potentials affect quantum phase where fields vanish, and in modern physics potentials are the basic variables. This was a minority emphasis in older textbooks, and it is a legitimate point. - His relaxation-time numbers for copper and quartz are correctly computed. Whittaker's 1903/1904 theorems are real and elegant mathematics. - Charging losses can be reduced by stepping the voltage (adiabatic charging). His sense that the charging step wastes energy is correct. - E = -dA/dt, which Bearden's 2004 remarks lean on, is a correct equation. It is ordinary induction. - The Tektronix rule is correct and worth memorising: true average power is the mean of v(t) i(t); V_rms x I_rms is an upper bound, equal to it only for same-shape, in-phase waveforms (Section 3.8).
What is wrong, and why the mistake is instructive. - Units. Calling potential "energy" mixes J/C with J. Checking units is the fastest way to catch an argument that has quietly changed the meaning of a word. - The factor of one half. K_T = V Q is the energy the battery spends. The capacitor keeps half, whatever the resistance. The "free" energy in the scheme is the source's own expenditure, moved and partly lost. Many inventors have been caught by this. It is a two-capacitor-paradox cousin: the loss does not depend on R, so better or worse wires cannot fix it. - The "suicide circuit". In the 15 Feb 1994 update Bearden says an ordinary closed circuit sends "one half" of the collected energy back through the source, which he calls a "suicide circuit" [read: pp.31-33, per arbiter]. The one half is asserted, not derived; it is true only for a matched load (load resistance equal to internal resistance, 50% efficiency), and a 12 V battery with 0.1 ohm inside driving 10 ohm delivers 99% [calc, per arbiter]. He then writes: "this 'suicide circuit' practice is guaranteed to keep the power meter on your home or business, and to keep the meter on the gas pump ... Perhaps one may be forgiven for suspecting that, at the deepest levels of financial control, this may be the real purpose in seeing that the existing interpretation of classical EM stays 'as is.'" [read: pp.32-33]. He states this as a suspicion; no document, person or event is offered in support, and the arbiter found none. - Borrowed quantities. tau = epsilon/sigma is about charge inside a material, not about when current starts in a circuit. Taking a correct number from one context and applying it in another is a common route to wrong conclusions. - Mathematical expansions are not physical contents. Writing a static potential as a sum of waves does not mean waves are flowing; a Fourier series of a square wave does not mean a square wave "contains" infinitely many oscillators drawing energy from somewhere. - Two different "efficiencies". In the 15 Feb 1994 update Bearden defines "operational efficiency Alpha" as "the average power expended in the load to power it, divided by the average power expended inside the source to dissipate its dipolarity" [read: pp.31-32], i.e. alpha = (average load power) / (average power dissipated inside the source). That is not the true efficiency eta = (load power) / (total power drawn from the source). The two are linked by alpha = eta / (1 - eta), or eta = alpha / (1 + alpha) [calc]. Example: a source that dissipates 1 W internally while delivering 2 W to the load has alpha = 2/1 = 2, but it draws at least 1 + 2 = 3 W in total, so eta is at most 2/3. However large alpha gets, eta = alpha/(1 + alpha) stays below 1. Alpha > 1 is the normal case for any source whose internal resistance is smaller than the load's (a 12 V battery with 0.1 ohm inside driving 10 ohm has alpha = 100, eta = 0.99 [calc]), so Bearden's statement that conventional operational efficiency "is always underunity" [read: p.32] is wrong by his own definition, and alpha > 1 does not mean more out than in. Watch for any ratio that leaves part of the input out of the denominator. - A device described only in words. The Fogal charge-blocking semiconductor of the 13 May 1994 paper "blocks passage of electrons ... but passes displacement current" [read: p.45]: that is the textbook description of a capacitor. Bearden concedes that ordinary meters on it would show energy conserved, and he reports no measurement [read: pp.44-46]. Fogal's own patents describe something more modest: a transistor with an electrolytic capacitor from emitter to ground, claiming higher gain, with no claim of output above input and nothing about the vacuum [web, per arbiter]. So the overunity reading is Bearden's interpretation of another inventor's device, and the inventor's own description neither supports nor contradicts it; it simply does not make the claim. A claimed gain that no instrument can see is not yet a claim that can be tested. - Pulsed power measurement. Average-of-product versus product-of-averages, phase, probe offsets and bandwidth decide whether a COP above 1 is real. The MEG patent's published data do not let anyone check these. The builders' reports show the pattern to watch for: excess appears only with nonlinear loads (harmonic-rich currents, where V_rms x I_rms overstates power), and the "apparent" output exceeds the heat in the load. Naudin's v2.1 also shows how the boundary decides the answer: 6.76 W out is 4.5 times the lamp-side input but 2.08 times the total input including the control board. - Correct equation, wrong energy accounting. E = -dA/dt is right, but outside a core where B = 0 the vector potential is a gradient, so its change gives no EMF round a loop; the coil sees only the Faraday EMF of the flux it encloses. Likewise, crossed static E and H fields (the capacitor on a magnet) have a circulating Poynting vector with zero divergence: energy flows round in a loop and none is delivered. A nonzero Poynting vector is not a power source. - Two descriptions of one fuel. The patent names the magnet's flux as the source (finite, about 7-22 J, gone in under a second at the claimed surplus); the 2004 remarks name the vacuum. When a claim's stated energy source changes between documents, ask which version a given calculation or test addresses.
What is genuinely open. - No neutral laboratory has published a full-waveform, calorimetric test of a MEG built to the patent. The test is cheap (a few thousand dollars of instruments, per arbiter) and would settle the measurement question. The arbiter's specification: record v(t) and i(t) at the supply (including the oscillator and logic supply) and at each output, integrate the point-by-point product, and put first a linear (resistive) load and then a nonlinear (MOV) load into a calorimeter; an artefact shows excess only in the scope-math number, not in the heat. Then the supply-removed run (it needs f c >= 1, and it is exactly the granted European claim 1): it runs or it stops, whatever the instruments. If the magnet is the fuel, measure its flux before and after hours of running. The one independent replication that was arranged (university, DoD auspices, per Bearden) reportedly stopped over a non-circumvention agreement before any measurement; which university and programme, and the primary source for how it ended, are unknown. - Why the MEG record goes silent after 2001 (artefact, commercial failure, trade secrecy, absorption under a new name, or suppression) is unresolved; Section 6, "Where the record goes silent", lists the evidence for and against each and the records that would decide (the Colmac security agreement, the full EPO and USPTO files, investor-demonstration reports, Naudin's later builds). - Why the applicants did not answer the USPTO's 2004 demonstration request, and how they answered the EPO on industrial applicability, are not in the pages read. Why the COP plot and the "all COP > 1" sentences are absent from the European text is unknown. - Whether investor representatives ever measured a MEG, and whether Naudin ever did calorimetry or closed the loop in later builds (COP 1.3 and 7.3 as reported by Bearden; v3.0/v4.0), is unknown. - Which of the older flux-switching patents (Starr 1957 to Flynn 2000) claimed COP > 1 is unknown; none has been read. - Whether the patent's 14 W input includes the oscillator and flip-flop supply is not stated [arbiter]. - What produced the patent's high output voltages (ringing, resonance, spikes) and the phase between output voltage and current are not quantified in the patent. - Sweet's VTA cannot be assessed without the device or its process; a surviving unit, Sweet's notes or a contemporary third-party measurement would change that. - What the Fogal CBS is now has a partial answer (Fogal's patented transistor exists); no measurement of gain above input has been found, and the rexresearch Fogal compilation has not been read. If Bearden's Fig. 2 loop is ever built, the decisive test is to run it with the DC supply disconnected, driving a resistive load for hours, as for the MEG [arbiter]. - Whether any measurable prediction can be drawn from Bearden's scalar-potential version of the AB effect; he states one, untested [read, per contrarian].
8. Check yourself
- A battery of 9 V charges a 10 microfarad capacitor through a 1000 ohm resistor. How much energy leaves the battery, how much is stored, and how much heats the resistor? What changes if the resistor is 1 ohm?
- Starting from div E = 0 in empty space and a plane wave E = E_0 exp(i(k.r - omega t)), show that E must be perpendicular to k. What experiment shows this directly?
- A material has sigma = 1e-3 S/m and relative permittivity 10. Compute its relaxation time. Is that the time a 1 microfarad capacitor connected through a 100 ohm resistor takes to charge? If not, what is?
- A MEG output coil shows a 1500 V peak sine voltage and a 40 mA peak sine current. The input is 8.4 W. For what phase angle between them does the true COP fall below 1?
- Why is a closed-loop self-running test (output feeding the input, start-up battery removed, resistive load, run for hours) a better test of the MEG than any amount of oscilloscope power arithmetic?
9. Further reading
Mainstream. - D. J. Griffiths, Introduction to Electrodynamics: Ch. 2 (electrostatics, potential, capacitor energy), Ch. 7 (Ohm's law, Maxwell's equations), Ch. 8 (Poynting's theorem, field momentum, and the static E x B "hidden momentum" examples relevant to the capacitor-on-a-magnet remark), Ch. 9 (EM waves, transversality), Ch. 10 (potentials and gauge) [memory, high]. - R. P. Feynman, The Feynman Lectures on Physics, Vol. II, Ch. 27 "Field Energy and Field Momentum" (the third-law argument and why fields carry momentum) and the vector-potential chapter discussing the AB effect [memory, high for Ch. 27; medium for the AB chapter number, so not given]. - E. M. Purcell and D. J. Morin, Electricity and Magnetism (capacitors, AC circuits) [memory, high]. - P. Horowitz and W. Hill, The Art of Electronics, Ch. 1 (RC circuits, power) [memory, high]. - Y. Aharonov and D. Bohm, "Significance of electromagnetic potentials in the quantum theory", Phys. Rev. 115, 485 (1959) [memory, high]. - G. Bressi et al., "Testing the neutrality of matter by acoustic means in a spherical resonator", Phys. Rev. A 83, 052101 (2011) [web: https://arxiv.org/abs/1102.2766]. - Whittaker's 1903 plane-wave representation explained: [web: https://arxiv.org/html/2205.08309v2]. - Maxwell's 1865 attempt at gravity and why he abandoned it: commentary at [web: https://royalsocietypublishing.org/rsta/article/373/2039/20140473/114863/a-paper-I-hold-to-be-great-guns-a-commentary-on].
Bearden and the MEG. - T. E. Bearden, "The Final Secret of Free Energy" (9 Feb 1993, revised 23 Mar 1993), with "Additional Comments" (12 Mar 1993) and updates (Feb 1994, 13 May 1994) [read; source folder]. - S. L. Patrick, T. E. Bearden, J. C. Hayes, K. D. Moore, J. L. Kenny, "Motionless Electromagnetic Generator", US 6,362,718 B1 (2002) [read; source folder]. Read the experiment section and Figs. 6-8 together. - J. L. Naudin's MEG v2.0 page (independent private builder; "The measured COP is 1.75"; coil failure above 3 kV; unsuccessful attempt to close the loop): [web: https://www.novakcorp.com/energy/experiments/megv2.htm, per arbiter]. - T. E. Bearden, e-mail of 2 Dec 2000 (COP 17-20 and >100 claims; the DoD-auspices replication; Naudin's later builds): [web: https://billstclair.com/www.cheniere.org/correspondence/120200.htm, per arbiter]. MEG status page, 11 April 2001: [web: https://www.rootaction.net/~jay/science/www.cheniere.org/megstatus.htm]. - A. Craddock, "Twenty years after Tom Bearden's MEG", ZPEnergy, 1 Jan 2021: [web: https://www.zpenergy.com/modules.php?name=News&file=article&sid=3918]. - J. L. Naudin, MEG v2.1 status (6 Dec 2000; input accounting, nonlinear loads, "measurement artifacts remain possible"): [web: http://jnaudin.free.fr/meg/megv21.htm]. Jon Flickinger, "The MEG Notes" (linear vs nonlinear loads): [web: http://jnaudin.free.fr/meg/megnot01.htm]. The Tektronix note on scope "RMS power" (posted 2002): [web: http://jnaudin.free.fr/meg/ths720p.htm]. Naudin's MEG index with the related-patent list: [web: http://jnaudin.free.fr/meg/meg.htm]. - USPTO file of continuation 10/081,847 (examiner's rejections of 2004 and the applicants' remarks with Bearden's mechanism; abandoned 2005): [web: https://archive.org/details/gov.uspto.patents.application.10081847]. Read the 24 May 2004 remarks for Bearden's own link between his theory and the MEG. - EP 1 446 862 B1 (granted 23 Oct 2013; claim 1 requires the self-powering loop): [web: https://patents.google.com/patent/EP1446862B1/en]; file history on the EPO Register: [web: https://register.epo.org/application?number=EP01988155&lng=en&tab=doclist]. Compare its description with US 6,362,718. - T. E. Bearden, letter of 15 Aug 2002 (company status, funding sought, investor-only demonstrations, the capacitor-on-a-magnet remark, the "vertical cliff"): [web: https://billstclair.com/www.cheniere.org/correspondence/081502.htm]. - T. E. Bearden, Energy from the Vacuum: Concepts and Principles (2002), his book-length statement of the theory, announced in the 15 Aug 2002 letter [web, per arbiter]; not read in this review. - J. deRivas, US 4,006,401 (1977), the flux-switching patent the EPO found anticipates the MEG's structure; other prior art in Section 5c (none read by this review). - Evans et al., "Explanation of the Motionless Electromagnetic Generator with O(3) Electrodynamics", Found. Phys. Lett. 14(1) (2001): a theory paper that assumes COP > 1, not a test; Bearden is a co-author per the author list on Naudin's index page [web: http://jnaudin.free.fr/meg/meg.htm] [web: https://link.springer.com/article/10.1023/A:1012085429802]. - W. J. Fogal, US 5,196,809 and US 5,430,413, "High gain, low distortion, faster switching transistor" (the inventor's own description of the device Bearden calls the CBS): [web: https://patents.google.com/patent/US5430413A/en].
Answers
- Q = C V = 10e-6 x 9 = 90 microcoulombs. Battery: Q V = 810 microjoules. Stored: (1/2) C V^2 = 405 microjoules. Resistor: 405 microjoules. With 1 ohm the split is identical; only the time constant changes (RC = 10 ms with 1000 ohm, 10 microseconds with 1 ohm).
- div E = i k.E_0 exp(i(k.r - omega t)) = 0 for all r and t, so k.E_0 = 0: E is perpendicular to the direction of travel. Crossed polarisers (Malus's law, intensity proportional to cos^2 of the angle) show it directly; a longitudinal wave would pass a polariser at every angle equally.
- tau = 10 x 8.854e-12 / 1e-3 = 8.9e-8 s (89 ns). No: the capacitor charges with time constant RC = 100 x 1e-6 = 100 microseconds, set by the circuit, not by the material's relaxation time.
- Real power = (1/2)(1500)(0.040) cos(phase) = 30 cos(phase) W. COP < 1 when 30 cos(phase) < 8.4, i.e. cos(phase) < 0.28, phase greater than about 74 degrees. A strongly reactive (nearly 90 degree) output would look large in volt-amperes yet deliver little power, which is why the phase must be measured.
- In a closed loop there is no input-power measurement to get wrong: either the load stays powered for hours with nothing connected, or the device stops. It removes phase errors, probe offsets, bandwidth limits and the question of which supplies were counted. A load in a calorimeter adds an independent check on output energy.