Meyers-Stan: Primer

Stanley A. Meyer's "Water Fuel Cell": electrolysis, resonance, and how to check whether a water capacitor can do what its patent says.

Primer author (Step 4), 2026-10-09; revised the same day after the QE (qe.md, verdict MINOR) and the arbiter's "QE revision" (arbiter.md section 14). Verdicts come from arbiter.md and are not changed here. Source tags: [read] = from a document in the project folder (page given where useful); [read: page image p.N] = read from a rendered page image; [web] = Internet source, with URL; [memory] = recalled from training, with a confidence level; [calc] = arithmetic redone by me and reproduced in example.py. Run it with:

.venv/bin/python review/Meyers-Stan/example.py

Source gap (repeated from claims.md and arbiter.md). Every Meyer document in the folder (patents, newsletters, the 222-page technical brief "The Birth of New Technology") is Meyer's own description. He had a stake: he held the patents and sold dealerships and "profit-sharing certificates". The folder holds no measurement of gas against electrical input, no test report and no independent witness statement about any of his devices. Its only named laboratory document is a 1983 sodium-and-potassium analysis of seven water samples (section 5), which says nothing about gas or energy. Meyer's own 1983 patent graphs (EP 0 111 573 A1) are the inventor's bench figures and are used in section 5, step 7. (Corrected after the QE: an earlier version of this note said the 1989 New Zealand video was "never transcribed" and that the photographs, Memos 425-430 and the appendices were unread. Text prep re-ran and produced a machine (whisper) transcript of the 2 h 38 min video, without speaker labels. That transcript, the 39 photographs, newsletters No. 4 and No. 8, the 1983 generator and injector patents, Birth pp.126-222 and Puharich's book pp.176-204 have now been deep-read; arbiter.md section 14 weighs them.) The best numbers about real cells (Meyer's 1995 energy table, the Nagypál and Perkins letters, the court docket) come from stanslegacy.com, an archive sympathetic to Meyer, as machine transcriptions of scans, marked "unverified" by the archive itself. Where a step below rests on them, I say so. The deciding government and court records (the NIST Office of Energy-Related Inventions evaluation, the UK DTI/Queen Mary College report, the 1996 Fayette County decision transcript) were not obtained.

New in this primer (no verdict changes). While re-deriving the worked example I found three things the reviews did not use; they are listed under "Corrections to earlier documents" at the end. (1) WO 92/07861 itself pairs "unipolar pulses from 0 to 650 volts" with a "1-2 milliamp minimum leakage" current; for the stated cell that pair needs water 2 to 4 times purer than water can be (section 5, step 5). (2) Nagypál's letter says that "between the plates appears only some millivolt potential difference, due to water's resistance, caused by contaminants": an experimenter's own observation that his tap-water cell did not hold a high voltage, which is what the calculation predicts. (3) Meyer's 1995 table contains two further arithmetic slips beyond the ones the reviewers found.

QE revision (no changes to existing verdicts). The QE found a central claim missing from the whole review: Meyer said in both 1989 talks that the system is self-sustaining. The arbiter added it as claim 7 (Refuted by calculation), and added claim 8 for peripheral claims in the brief. This primer now covers claim 7 in sections 1, 4, 5 (step 9) and 6. It also uses Meyer's own 1983 water analysis in place of a generic tap-water figure in section 5 (the conclusion is unchanged), adds Meyer's 1983 patent data to step 7, and restores to the record several items the QE found dropped (section 6, "Kept in the record").

---

1. The idea in one paragraph

Stanley Allen Meyer (1940-1998) of Grove City, Ohio, patented and promoted a "Water Fuel Cell" from the early 1980s until his death [read: patents; newsletters]. His central idea: put ordinary water between two stainless-steel tubes so that the water is the insulating layer of a capacitor; connect this "water capacitor" through coils ("resonant charging chokes") and a diode to a pulsed supply; tune the pulses to the circuit's resonance; then the voltage across the water climbs very high while the current stays tiny, and the electric field pulls the water molecules apart into hydrogen and oxygen gas "without electrolysis" and with far less energy than ordinary electrolysis needs [read: WO 92/07861 pp.3-4; Birth pp.7-13]. On top of this he claimed that the gas, specially treated, releases "atomic" energy far beyond ordinary burning (up to "2.5 million barrels" of oil per gallon of water), that cars, a dune buggy and aircraft could run on water through his injectors, and that a loop of magnetised gas could amplify electricity. In his 1989 talks he went further: once started from an outside source, the system would be "a self-sustaining oscillation system" that "maintains itself as long as there is water in the fuel cell", making electricity "without consuming the fuel source" [read: Einsiedeln and New Zealand talk transcripts]. The mainstream reading, which the arbiter accepts, is that the cell is an ordinary electrolysis cell driven by an unusual pulsed supply: it really does make hydrogen and oxygen, but water in such a cell behaves as a resistor, not a capacitor, so the resonant high voltage cannot build up, and no path of any kind can split water for less energy than burning the gas gives back, so a loop that splits water and feeds the gas's energy back can only run down. What was never done is the one cheap measurement that would settle it beyond argument: gas collected against true electrical input, with a plain-DC control, by someone without a stake.

2. What you need to know first

Electrolysis and Faraday's law. Pass a direct current (DC) through water containing some dissolved ions, between two metal electrodes, and hydrogen gas (H₂) bubbles off the negative electrode while oxygen (O₂) bubbles off the positive one. Each H₂ molecule needs 2 electrons delivered through the circuit. Faraday's law of electrolysis turns this into a rule: moles of H₂ = Q / (2F), where Q is the charge passed (coulombs, C; Q = current x time) and F = 96,485 C/mol is the Faraday constant (the charge on one mole of electrons). Each H₂ comes with half an O₂, so a mole of water gives 1.5 moles of mixed gas. Worked number: 1 ampere (A) for 1 hour is 3600 C, which makes 0.0187 mol H₂ plus 0.0093 mol O₂, i.e. about 685 cubic centimetres (cc) of mixed gas at 25 °C [calc]. That 685 cc per ampere-hour is the yardstick used all through this primer. Read: Atkins, Physical Chemistry, the electrochemistry chapters; Bard and Faulkner, Electrochemical Methods, chapter 1.

The energy cost of splitting water. Two numbers matter. The enthalpy of formation of liquid water, ΔH = 285.8 kJ/mol, is the heat released when hydrogen burns back to liquid water (the "higher heating value", HHV). The Gibbs free energy, ΔG = 237.1 kJ/mol, is the minimum electrical work needed to split one mole of water at 25 °C; the rest, TΔS = 48.7 kJ/mol (T the absolute temperature, ΔS the entropy change), can come in as heat from the surroundings [memory, high]. Divided by 2F these become voltages: 237.1 kJ / (2 x 96,485 C) = 1.23 V (the reversible voltage) and 285.8 kJ / (2F) = 1.48 V (the "thermoneutral" voltage, at which the cell neither warms nor cools). Real electrolysers run at 1.8-2.2 V per cell because of losses ("overpotentials") [memory, medium]. Two consequences: (i) gas energy out can exceed electrical energy in by at most ΔH/ΔG = 1.21 (121%), and only if the cell draws the difference as heat from its surroundings; (ii) these numbers depend only on the start state (liquid water) and the end state (H₂ and O₂ gas), not on the route. Energy is a state function. Any method of splitting, electrolysis or not, pays at least ΔG. Read: Schroeder, An Introduction to Thermal Physics, chapter 5 (free energy; it works the electrolysis of water as an example).

Capacitors and dielectrics. A capacitor is two conductors separated by an insulator (a dielectric). Its capacitance C (farads, F) is the charge it stores per volt. Filling the gap with a dielectric multiplies C by the relative permittivity εr (also called the dielectric constant, a pure number with no units). Water's εr is unusually large, about 78.5 at 25 °C and about 80 at 20 °C [memory, high]. For two concentric tubes, inner radius a, outer radius b, length ℓ: C = 2π ε₀ εr ℓ / ln(b/a), with ε₀ = 8.854 x 10⁻¹² F/m the permittivity of free space. Read: Griffiths, Introduction to Electrodynamics, chapter 4 (dielectrics) and chapter 2 (capacitance).

Conductors, and the lossy capacitor. Real water conducts, because it contains ions: hydrogen and hydroxide ions from water itself, plus dissolved salts. Its conductivity σ (siemens per metre, S/m) ranges from 5.5 x 10⁻⁶ S/m for theoretically pure water (limited by water's own self-ionisation) to about 0.01-0.05 S/m for tap water (100-500 μS/cm) [web: https://atlas-scientific.com/blog/water-conductivity-range/ as checked by the arbiter]. So a "water capacitor" is always a capacitor and a resistor side by side (in parallel). For the coaxial cell the resistance is R = ln(b/a) / (2π σ ℓ). Multiply R by C: the geometry cancels and RC = ε₀εr / σ. This product, τ = ε₀εr/σ, is the relaxation time: the time in which any charge you put on the plates leaks away through the water. Griffiths (chapter 7, section on Ohm's law) derives the same time for charge inside a conductor. If you drive the cell with angular frequency ω (radians per second, ω = 2πf, f in hertz), the ratio of capacitive current to resistive current is ωτ = ω ε₀εr/σ. When ωτ is much less than 1, the "capacitor" is really a resistor.

Series resonance and the quality factor Q. An inductor (coil) of inductance L (henries, H) has reactance X_L = ωL; a capacitor has reactance X_C = 1/(ωC) (both in ohms). In series they cancel at the resonant frequency f₀ = 1/(2π√(LC)). At resonance, with a small series resistance R_s, the current is limited only by R_s, and the voltage across the capacitor is Q times the drive voltage, where Q = (1/R_s)√(L/C) is the quality factor. Q is finite whenever there is loss: "infinite voltage at resonance" needs zero resistance. In a parallel tank (L and C side by side) the opposite happens: the current drawn from the supply is smallest at resonance, while a large current circulates inside the tank; the supply still has to replace whatever the tank's resistance dissipates. Read: Horowitz and Hill, The Art of Electronics, chapter 1; Griffiths, chapter 7 (inductance).

Measuring pulsed power. Power is P = vi only moment by moment. For pulses, the energy over a time T is E = ∫₀ᵀ v(t) i(t) dt. The average current Ī (what a cheap DC meter roughly reads, and what Faraday's law needs) is not the root-mean-square (RMS) current (what heats a resistor), and average-V times average-I is not average power. A 2025 study of pulsed electrolysis (Puteanus et al., International Journal of Hydrogen Energy) notes that some reported gains came from exactly this mistake [read: research/files/Puteanus-2025-IJHE-pulsed-electrolysis.pdf, introduction, as cited by the mathematician]. Read: Horowitz and Hill, chapter 1.

Ions drift, they do not fly. In liquid water an ion in a field E (volts per metre) moves at a steady drift speed v = μE, where μ is its mobility; for the hydrated proton μ ≈ 3.6 x 10⁻⁷ m²/(V s) [web: Mauerhofer et al. 2003, https://download.uni-mainz.de/fb09-ak-roesch/Roesch_Puplikationen/113 Mauerhofer-2003-PCCP.pdf, via arbiter]. The motion is like a bead sinking in honey: friction sets the speed, there is no build-up of momentum, and so no "bouncing" resonance.

Energy scales: chemical versus nuclear. Chemical bonds hold a few electronvolts (eV) per atom; nuclear reactions involve millions (MeV). Einstein's E = mc² sets the absolute ceiling for any mass m. A US gallon of water (3.785 kg) burns, as hydrogen, for about 60 MJ; its whole rest-mass energy is 3.4 x 10¹⁷ J [calc]. Read: any modern-physics text, e.g. Krane, Modern Physics, chapters on the nucleus.

3. The textbook view

3.1 What a water cell does. A pair of electrodes in water with a voltage above about 1.23 V (in practice about 1.5-2 V) between them makes H₂ and O₂ at the rate set by Faraday's law, moles H₂ per second = Ī / (2F), where Ī is the average (net) current. The energy cost per mole of water is the cell voltage times 2F, at least 237 kJ. Pulsed and alternating drives have been studied for decades; the recent literature finds no gain over steady DC for a standard cell model, while a minority of papers report modest gains over poor DC baselines, never below the thermodynamic minimum [read: Puteanus 2025, abstract; contrarian.md, section on claim 1, for the opposing papers].

3.2 The cell as a circuit element. Model the cell as C in parallel with R, with RC = τ = ε₀εr/σ. Two regimes: - ωτ ≫ 1 (very pure water, or very fast pulses): the cell behaves as a capacitor. It can be charged to a high voltage and will hold it for a time of order τ. - ωτ ≪ 1 (tap water at kHz): the cell behaves as a resistor. Whatever voltage V sits across it drives a current V/R and dissipates V²/R, and that current electrolyses the water.

For any drive the cell's own quality factor is Q_cell = ωτ = ω ε₀εr / σ. A lossless coil in series cannot raise the cell voltage above about Q_cell times the drive (and never below the trivial gain of 1 at DC). Section 5 shows this.

3.3 Fields and bonds. Field strength across a gap d at voltage V is E = V/d. Strong fields do help water dissociate into ions (the "second Wien effect"), noticeably at around 10⁷ V/m for weak electrolytes and above 10⁸ V/m for water itself [web: https://arxiv.org/pdf/2405.05882 ; https://arxiv.org/pdf/2208.10828, via arbiter]. Tearing water molecules apart by the field alone needs about 3.5 x 10⁹ V/m (0.35 V per ångström, in computer simulations of liquid water), with 15-20% of molecules dissociated at about 10¹⁰ V/m (about 5 eV across 0.1 nm) [web: https://arxiv.org/pdf/2506.11734, via arbiter section 14.11, from a search summary; earlier tagged memory, medium]. Dissociating into ions is not the same as releasing gas: gas appears only when electrons are exchanged at the electrodes, which is electrolysis again.

3.4 Energy accounting. Burning a mole of H₂ with half a mole of O₂ gives back ΔH = 285.8 kJ, never more. So a car "running on water" must get the splitting energy from somewhere else (its battery, recharged by the engine's alternator, which is run by the engine), and the loop loses energy at every stage. A stoichiometric oxygen-hydrogen flame is very hot, about 2,800 °C (about 5,000 °F) [memory, high], which is why it melts steel quickly at any flow rate.

3.5 Real engineering that sounds like Meyer. These are genuine, and they are what makes Meyer's story sound plausible: - Resonant charging chokes. In radar "line-type" pulse modulators (1940s onward) a DC supply charges a pulse-forming capacitor bank through a "charging choke" (an inductor) and a hold-off diode. Because the inductor keeps current flowing after the capacitor voltage passes the supply voltage, the capacitor ends at twice the supply voltage, and the diode keeps it there until the switch fires [memory, high; Glasoe and Lebacqz, Pulse Generators, MIT Radiation Laboratory Series vol. 5, 1948]. Meyer's term "resonant charging choke" and his diode come from this tradition. The doubling works only if the capacitor holds its charge between pulses, i.e. if its leakage time τ is long. - Water as a capacitor dielectric. Pulsed-power machines (for example the large water-insulated transmission lines of Sandia's accelerators) really do use water as a capacitor dielectric, because εr ≈ 80 stores a lot of energy per volume. They use deionised water and charge it in microseconds, short compared with τ, and they still lose charge through it [memory, high for the practice; medium for details]. This is the one regime in which a "water capacitor" works, and it agrees with the arithmetic of section 5. - Parallel resonance draws little supply current. Meyer's "at resonance the current draw is minimised" describes a real property of a parallel tank (section 2). It does not remove the energy dissipated inside the tank. - Hydrogen flame engineering. Diluting hydrogen with air and exhaust to slow its burning speed, and quenching passages to stop flashback, are real combustion techniques; Meyer's flame and injector patents work in this area [read: Birth pp.34-37; deep read].

4. Where the inventor departs

Meyer's starting points are real: series and parallel resonance, the resonant charging choke and diode, water's high dielectric constant, the electric force on charged parts of a molecule (Coulomb's law, which he quotes as Eq 13), and the fact that the cell makes gas. The departures, side by side:

The fair version of the claim. Stripped of the infinite-voltage story, one testable statement remains: this pulsed circuit makes more gas per joule than the same cell on plain DC. Thermodynamics allows that (a badly run DC cell can be beaten), up to the ΔH/ΔG = 121% ceiling with heat drawn from the room. It was never measured properly, and the arbiter rates it Never properly tested.

Puharich (the folder's related work). Andrija Puharich's US 4,394,230 (1983), which Meyer's US 4,798,661 cites, claims splitting water by audio-frequency "resonance" of "tetrahedral" water molecules at 80-100% efficiency and reports 114.9% energy return in one run [read: deep read]. Its arithmetic divides the energy per mole by 22,400 cc (one mole of gas) instead of 33,600 cc (the 1.5 moles of gas from one mole of water); corrected, the run gives 63% (against ΔG) and 77% (against ΔH) [calc, per arbiter section 8]. Water's molecular vibrations are in the terahertz-infrared range, not at kHz [memory, high].

5. Worked example: can the 3-inch water capacitor do what WO 92/07861 says?

WO 92/07861 (filed 1990) is the drive-circuit application for the water cell. It gives one concrete cell and one operating point [read: WO 92/07861 pp.9, 13-14]: - cell: "a 0.5 inch rod enclosed within a 0.75 inside diameter cylinder", "about 3.0 inches" long, stainless steel; - resonance "typically at a frequency of 5 KHz"; - "At start up ... current draw through the water cell will measure about 25 milliamp; however, when the circuit finds a tuned resonant condition, current drops to a 1-2 milliamp minimum leakage condition"; - "unipolar pulses from 0 to 650 volts", giving "conversion of about 5 gallons of water per hour into a fuel gas".

We test these numbers against each other, using only textbook laws. Every number here is printed by example.py.

Step 1. Capacitance and resistance of the cell. Radii a = 6.35 mm, b = 9.525 mm (gap 3.175 mm), length ℓ = 7.62 cm, so ln(b/a) = 0.4055. With εr = 78.5:

C = 2π ε₀ εr ℓ / ln(b/a) = 2π (8.854 x 10⁻¹²)(78.5)(0.0762) / 0.4055 = 821 pF.

The resistance R = ln(b/a)/(2πσℓ) depends on the water:

Meyer's own water. The two "Meyer" rows come from the inventor's own source, not a generic range. The Birth of New Technology prints an "Analysis of Water" chart (Fig. 7-15, p.163) credited to "R & D Laboratory, A Certified Biological Testing Laboratory, Columbus, Ohio 43204, C. Thomas Bell, Ph.D., Lab Director, Test Date: January 10, 1983, Test Method: Spectrophotometer Analysis, Contaminates: Sodium (Na) and Potassium (K)": distilled under 2 ppm, tap 11, rain 16.1, lake 20.2, drilled 24, river 45, ocean 26,100 ppm (ppm = parts per million by mass, i.e. milligrams per litre) [read: page image p.163]. The laboratory is credited only with this salt assay, not with any gas or energy test. To turn ppm into conductivity, divide by the ion's molar mass to get moles per litre and multiply by the salt's molar conductivity (how well one mole of it conducts in dilute solution: 126.4 S cm²/mol for sodium chloride, 149.9 for potassium chloride [memory, high]). Tap water's 11 ppm read as sodium chloride gives about 60 μS/cm, read as potassium chloride about 42 μS/cm; "under 2 ppm" gives about 11 μS/cm [calc]. Calcium and magnesium (water hardness) were not measured, so real tap water conducts more; these figures are a lower bound, which makes the conclusion below safer, not weaker. (The same chart labels water's dielectric constant "78.54Ω", the units slip in section 4. Newsletters 10 and 11a reprint the chart; their deep reads read the date as "1963", while Birth p.163 prints 1983.) The row "very well distilled, 1 μS/cm" is kept as a best case for good laboratory water; it is about ten times purer than Meyer's own distilled sample.

example.py also confirms symbolically that R x C = ε/σ: the size and shape of the cell drop out.

Step 2. The coil for 5 kHz. L = 1/((2πf)²C) = 1/((3.14 x 10⁴)² x 821 x 10⁻¹²) = 1.23 H, a big but buildable choke (Nagypál used about 0.8 H with a 10 nF cell [read: stanslegacy transcription of his letter]). The capacitor's reactance at 5 kHz is X_C = 1/(ωC) = 38.8 kΩ. So hitting a resonant frequency is easy. The question is the voltage gain.

Step 3. Capacitor or resistor? Compare the two currents through the cell: Q_cell = ωτ = ω ε₀εr/σ.

For tap water the current through the water's resistance is over 2,000 times the capacitive current. On Meyer's own figures the conclusion is the same: his tap water gives Q_cell ≈ 0.004 (the resistive current is about 280 times the capacitive one) and his distilled water 0.02 (50 times). The 5 kHz drive period is 200 μs; tap water loses any charge on the plates in about 14 ns, roughly 14,000 times faster. A diode "staircase" charge pump cannot build up voltage on a capacitor that empties itself between steps.

Step 4. The best gain a coil can give. Put an ideal, lossless inductor L in series with the cell (C parallel to R). The voltage divider gives

V_cell / V_drive = R / (R − ω²LRC + jωL),

where j = √−1. At the resonance ω²LC = 1 the first two terms cancel and the gain is R/(jωL) = −jωRC, whose size is ωRC = Q_cell (example.py derives this with SymPy). Scanning every frequency from 1 mHz to 100 MHz numerically: for tap and distilled water (including both of Meyer's own samples) the largest gain is 1.000, at DC (the supply simply passing through the coil); there is no resonant peak at all. For theoretically pure water the peak is 4.0, at 4.9 kHz. To turn 12 V into 20,000 V (a gain of about 1,700), the water would need Q_cell ≈ 1,700, i.e. a conductivity about 400 times below the self-ionisation limit of perfectly pure water. Real water cannot do it, whatever the coil.

Step 5. What it costs to hold a voltage on the cell. If the voltage is somehow there, the water conducts and dissipates P = V²/R:

Now the patent's own pair, 650 V with "1-2 milliamp": R = 650 V / 1.5 mA ≈ 325-650 kΩ, which for this cell means σ = 1.3-2.6 x 10⁻⁶ S/m, 2 to 4 times purer than theoretically pure water [calc]; against Meyer's own distilled water (under 2 ppm) the cell would draw 0.84 A at 650 V, about 560 times his 1.5 mA. Even granting that the pulses are "0 to 650 volts" and so not at 650 V all the time: with 500 μS/cm tap water the full voltage could be present for only about 0.004% of the time (4 x 10⁻⁵) to average 1.5 mA. (There is one way to read "1-2 milliamp" consistently: as the supply-side average of a pulsed, mostly reactive drive, measured with a meter that does not capture the cell current. That reading saves the patent's sentence at the cost of its claim, because then the 1-2 mA says nothing about the energy reaching the water.) The same arithmetic applies to Appendix A's "40,000 volts @ 1 ma = 40 watts" [read: Birth p.221]: it implies a 40 MΩ cell, 260 times the resistance this cell would have even when filled with perfectly pure water [calc]. Nagypál, the one outside builder whose letter gives numbers, wrote that with tap water "between the plates appeares only some millivolt potential difference, due to water's resistance, caused by contaminants" [read: stanslegacy transcription of his letter, 4 Jun 1991, binder p.66]. Whatever his meter's accuracy, the direction is what step 5 predicts: tap water would not hold the voltage.

Step 6. What "5 gallons per hour" would cost, by any route. 5 US gallons = 18.93 kg of water = 1,051 mol per hour = 0.292 mol/s. Because energy is a state function, the route does not matter: - minimum electrical power (ΔG): 0.292 x 237.1 kJ = 69 kW; - energy the gas releases when burned (ΔH): 0.292 x 285.8 kJ = 83 kW; - if it is electrolysis, the Faraday current: 0.292 x 2 x 96,485 = 56,000 A.

The stated drive, about 650 V x 1 mA = 0.65 W, is short by a factor of about 10⁵ in power and 4 x 10⁷ in current [calc]. A "non-electrolytic" route would remove the 56,000 A but not the 69 kW. The patent writes this throughput in the future tense ("will occur") and gives no measurement [read: deep read].

Step 7. The inventor's own measurements, checked the same way. The first set is in the project folder; the rest come from Meyer's 1995 binder via the stanslegacy machine transcription (unverified; the scans are the record). - Meyer's 1983 patent (EP 0 111 573 A1, "Hydrogen Generator System", filed 14 Dec 1982; deep read with page images, weighed in arbiter section 14.1). The text gives one operating point: "the source 30 supplied 12 volts at 1 ampere" [read: p.12]. Faraday's law gives 0.685 L/h of mixed gas for 1 A [calc]. The text says gas "is a linear function to magnitude of voltage applied", that the water heats as voltage and current rise until steam forms (Meyer calls the heating "a serious limitation"), and that gas depends on the type of water [read: pp.12-13]. Fig. 11 shows distilled water as a straight voltage-against-current line, about 87 V at 1 A (a resistor of about 85-90 Ω); Figs. 8 and 15 show the current rising 23-fold from distilled to ocean water. Gas proportional to current, heating, and current set by the dissolved salts are the signatures of ordinary electrolysis, drawn by Meyer himself. Only Fig. 10 (and Fig. 14, the same data per kWh) has gas numbers: 456 cubic inches per hour at 84 W, and 2,700 cubic inches per hour at 120 W. Converted [calc, per arbiter]: 7.5 L/h and 44.2 L/h, whose heat of combustion is 19% and 80% of the electrical input, inside the energy limit. They match Faraday's law at a cell voltage of 7.7 V and 1.86 V; no voltage or current is given for them, so they cannot be checked further. One point inside the patent contradicts itself: the text says pulsing "enhances considerably the splitting" (p.8), but Fig. 16 shows gas falling in a straight line as the pulse on-time is shortened, so steady DC gives the most gas (arbiter: Internally inconsistent). (The patent's "37%" figure, which the QE first read as an electrolysis efficiency, refers to photosynthesis, per the deep reader and arbiter.) The Canadian version (CA 1,234,774, granted 1988) adds a current "restricted ... to a negligible value" and "less than 1 ampere" to the claims; the 1982 text itself runs at 1 A and 84-120 W. - Meyer's 1995 table (binder pp.59-60): "12.5VDC @ 40 amps = 500 watts", "1 liter cavity @ 7lbs / min", "1 Liter Cavity = 1,000 cc of Gas Volume = 1lb Gas Pressure", hence 7,000 cc/min = 420,000 cc/h, compared against "Prior Art 'Electrolysis'" at "10cc Gas Production Rate per (1) amp / hr". Three checks [calc]: (i) Faraday gives 685 cc per ampere-hour, not 10, so the baseline is 68 times too low. (ii) A pound per square inch (psi) is a pressure, not 1,000 cc. If a 1-litre cavity gains 7 psi each minute, that is 7/14.7 = 0.476 litre of gas at atmospheric pressure per minute, 28.6 L/h. Faraday's law for 40 A gives 27.4 L/h. Read this way, Meyer's own 500 W test is ordinary electrolysis at 40 A, within 4% (assuming the transcription's "7lbs" and gauge pressure; the arbiter flags both). (iii) Taken at face value, 420 L/h of gas would hold 909 W of combustion energy from 500 W in, an energy gain of 1.8 that no electrical process allows, and 15 times what 40 A can make. - Nagypál (binder pp.66-68): about 20 mA, "about 0.5-1 cubic centimeter gas" per minute. Faraday for 20 mA: 0.23 cc/min. His rate is 2.2 to 4.4 times Faraday [calc]. This is the only number in the whole record above a Faraday estimate. It may be (a) a real excess, (b) a current read on a meter that misses the spiky part of a pulsed waveform, or (c) gas that is not electrolysis gas (dissolved air driven out, bubbles counted by eye at about 1 cc with "50%" of the gas still "on the plates"). Even at face value it is 0.1 W of gas energy, and his own comparison ("300%") is against a reference cell using 10-12 W, about 100 times worse than the thermodynamic minimum. It deserves the controlled repeat it never got.

Step 8 (side check, claim 2). One gallon's hydrogen burns for 60.1 MJ = 0.0098 barrel of oil (at 6.1 GJ per barrel [memory, high]). "2.5 million barrels" is 1.5 x 10¹⁶ J: 2.5 x 10⁸ times the chemical energy and 4.5% of the gallon's whole mc² [calc]. Fusing every hydrogen nucleus in the gallon to helium would give about 44,000 barrels, 57 times less (arbiter, section 2). No chemical or electrical process at a few kilovolts can reach nuclear energies.

Step 9 (side check, claim 7: the self-sustaining loop). Meyer's illustration is "100 volts times 100 amps ... 10 kilowatts" of electricity drawn from the water's electrons and fed back to keep the cell going. Electrons cannot be taken from water without splitting it: two electrons per molecule. 100 A is 100 C/s, which is 100/96,485 = 1.04 mmol of electrons per second, so 0.518 mmol of water per second, 33.6 g of water per hour. Burning all of that hydrogen gives back 0.518 x 10⁻³ x 285.8 kJ = 148 W, not 10 kW (Meyer's figure is 68 times too large), and splitting it costs at least 1.23 V x 100 A = 123 W [calc]. Run the loop on paper: the split costs at least 237 kJ/mol; the gas returns at most 237 kJ/mol as electricity (ideal fuel cell) or 286 kJ/mol as heat, of which a heat engine turns only part into electricity. At 100% everywhere the loop just breaks even; with a 70% electrolyser and a 55% fuel cell each joule fed in comes back as 0.38 J [calc; memory, medium for the two efficiencies]. It stops from losses long before the water runs out, and burning the gas gives the water back anyway.

What the example shows. Meyer's circuit formulas are right in form, and his parts (choke, diode, cell) are real engineering. The numbers fail at one specific place: water's conductivity. It makes the cell a resistor at kHz, even for the water Meyer's own 1983 analysis describes, caps the resonant gain at about 1 (tap) to 4 (perfectly pure), makes "650 V at 1-2 mA" impossible for any real water, and turns any voltage that does appear into ordinary electrolysis current. Separately, energy conservation alone, with no cell model at all, puts the claimed throughput 10⁵ times beyond the stated power. The one escape the mathematics leaves open: microsecond pulses shorter than τ on very pure water, the regime pulsed-power engineers actually use. Meyer's documents give no pulse lengths (the only time scale found, "within a millisecond or less" in Memo 425, is a voltage rise time to "up to and beyond 90,000 volts", a design statement, not a pulse length [read: Birth p.133, per arbiter 14.1]), so that escape cannot be tested from them, and even there the gas still costs at least 237 kJ/mol.

6. Evidence table

Verdicts are the arbiter's (arbiter.md section 11, including the QE revision in section 14). Claims 7 and 8 and the new sub-verdicts under claims 1 and 4 are the arbiter's own, from calculation; the three reviewers have not yet weighed them.

Records that would decide most of it (arbiter section 12): the USPTO file wrappers of US 4,936,961 and 5,149,407; the NIST Office of Energy-Related Inventions file (Lewett letter, 28 Jun 1991); the DTI/Queen Mary report and Laughton's report of 13 Apr 1992 (probably one document); the Fayette County decision transcript and Final Judgment Entry (case CVH19930292, judgment 24 Sep 1996, $40,515.72 garnished); the binder scans of pp.59-60 and 66-68; the probate valuation of "Water Fuel Cell: sole proprietorship" (1999); the Perkins Project TACO file; the 1985 television footage. Added after the QE: the "WFC Patent Validation Report dated January 14, 1983", cited once by Meyer (Birth p.190, Memo 429) as "WFC Test-Results" showing operability, with no author, laboratory or number, and not in the folder (it may be in the 1995 binder or the 1983 Canadian and European patent files); the surviving hardware, which could be run on a calorimeter with logged power (the "83 DuneBuggy" transferred in probate on 5 May 2000, and the cells, voltage-intensifier cylinders and control racks in the 39 photographs, which carry "Gries Petty Research / RWGresearch.com / AlexPetty.com" watermarks, chain of custody unknown, and show no operation); and Channel 4's "It Runs on Water" (17 Dec 1995).

Kept in the record whatever the verdict (SOURCE HONESTY): the Electronics World eyewitness account (Hindley, with Professor Laughton and Admiral Sir Anthony Griffin present, Oct 1990), including the magazine's statement that Section 101 patents depend on "a successful demonstration of the invention to a Patent Review Board" (the magazine's and Hindley's characterisation; the reviewers found that USPTO practice presumes utility rather than requiring a demonstration [web: https://www.uspto.gov/web/offices/pac/mpep/s2107.html]) and its editor's caveat that the technique "has more holes in it than a sieve"; Perkins's bench observation (dissociation from about 10 V, "less than 1 milliamp", room temperature) and the same firm's note that no water-fuelled engine was shown; the Sussex "breadboard rig" that "succeeded in producing hydrogen" (report not found); the Queen Mary/DTI experts who "were unable to reconcile" the descriptions with established science "however they left open the possibility" (per Perkins); Hans Nieper's 1996 letter on "the tested overunity effect" (unseen); Griffin's Warsash lecture (Sep 1993, Royal Institution of Naval Architects, as reported in Meyer's binder, unseen): "10 W in, gas at 10 psi within 10 s, vessel cold" (the arbiter's check: 10 W for 10 s is 100 J, enough for at most about 13 mL of gas at 1 atm, so a 10 psi rise needs a vessel of about 19 mL or less, and 100 J warms 100 mL of water by only 0.24 K, so "cold" is expected; the observation is neutral until the vessel volume is known; it is a single report by an interested admiral [calc, per arbiter 14.9]); the Dublin Institute of Technology explanation in the 1995 binder (not read by anyone); Lammertink's hypothesis (tuks.nl) that Meyer's first cell "worked then stopped" because of a scrap-steel electrode alloy, linked to Meyer's 1982 steel-analysis affidavits by Robert L. Queen (an explanation by an outside builder, never tested; it predicts an electrode-material effect and would be tested by changing electrodes); Gerald Pollack's "exclusion zone" proposal, in which water next to hydrophilic surfaces forms charge-separated layers hundreds of micrometres thick [memory, medium], which could matter in gaps of 0.25-1.6 mm (contested in mainstream science, not Meyer's own idea, and with no energy-balance prediction for this cell: the arbiter rates it Not testable as stated for Meyer's cell); the 1983 Bell water analysis (section 5), the one named laboratory in the folder; Meyer's statements in the 1989 New Zealand talk (a meeting about manufacturing and investment in New Zealand: "small amounts of money" to set up manufacturing; a production unit "in 18 to 24 months"; aircraft such as an F-16 or a T-39 Learjet to fly "off the atomic power of water"; contacts at the Pentagon and Wright-Patterson Air Force Base, identities unverified; one participant's remark that a demonstration model "doesn't need to be demonstrated it just needs to work", and Meyer's reply that a first model was needed "to generate interest"); an audience member's recollection in the same meeting of running a Ford six-cylinder engine on condensed and re-injected water that "started to continue on with less energy input" (not Meyer's claim; unquantified); Meyer's 1997 placard "Now, under WFC Industrial Contract: Europe" (his own marketing text; no party, date or country); the earlier investor suits in Franklin County, Ohio, 79CV004853 (Schuler and Leukart v. Meyer, 1979-81) and 93CV005477 (Schneider, Brooks and Gilvesy v. Meyer, settled 1994; Brooks invested over $300,000), beside the 1996 Fayette County case, a fifteen-year pattern of disputes with investors that bears on the stake and finance record (sites.md, docket indexes; outcomes beyond the settlement not read); the press and broadcast record, none of it seen by any role: Channel 4 "It Runs on Water" (17 Dec 1995), Ship & Boat International (RINA, Jun 1991), Asian Defence Journal (1992), IECEC 1993 (Bailey and Grotz), Air & Space Smithsonian (Oct/Nov 1991), Spy (Dec 1991), Fortean Times (1997); and other recorded talks (the 1992 Global Sciences Congress "Voltage Disassociation of Water", a 1993 interview, lecture videos on the Internet Archive); the NIST "net consumer" report (second-hand); the court case and Meyer's own account of it (the judge ruled the process "only electrolysis", court audio switched off: his allegation, no second source); the family's accounts of buy-out offers and of poisoning, beside the coroner's finding of cerebral aneurysm; the missing police tapes of the two Belgian investors; and the patents' lapse after the estate assigned them to Meyer's widow in 2001. The arbiter's section 10 sets out the competing explanations for the long silence (nothing remarkable existed; trade secrecy; unfavourable tests unpublished; instruments that could not read pulsed power; suppression) and the records that would decide between them.

7. What a learner should take away

Solid. Pulsed high-voltage cells do make hydrogen and oxygen from water. Series and parallel resonance, resonant charging through a choke and a diode, and water's high dielectric constant are all real. Meyer's textbook formulas (Eqs 2-4, 8) are correct in form, and some of his flame and injector engineering (diluting hydrogen to slow its burning) is ordinary combustion practice.

Wrong, and why the mistake is instructive. - A dielectric constant is not insulation. Water has a large εr and a conductivity. The ratio that decides which wins, ωε/σ, is independent of the cell's size and shape. This one dimensionless number, Q_cell, is the whole story of claim 1. It is a good habit for any "capacitor" made of a real material: compute ωτ first. - "Infinite at resonance" is a lossless idealisation. Every real resonance has a Q, and here the load itself sets it. - Energy is a state function. Calling a process "not electrolysis" does not change the price of splitting water: at least 237 kJ/mol whatever the route, and burning returns at most 286 kJ/mol. This alone disposes of water as a net fuel, of the 5-gallon-per-hour throughput and of the self-sustaining loop (claim 7), without any model of the cell. Water is the ash of burning hydrogen: a loop that splits it and recombines the gas can at best break even, and in practice returns about 38% of what is fed in. - Units catch errors. A dielectric constant added to ohms, amperes squared called watts, "lbs" of pressure called 1,000 cc of gas, a baseline of 10 cc per ampere-hour against Faraday's 685: each slip, found by checking units, removes a claimed gain. Meyer's own 500 W test, read with correct units, is ordinary electrolysis. - Read the inventor's own data first. Meyer's 1983 patent graphs (gas linear in current, water heating, current set by dissolved salts) are a textbook picture of electrolysis, and its Fig. 16 shows DC beating pulses, against its own text. - Meters on pulsed circuits mislead. "Milliamps on the ammeter" with tens of kilovolts "apparently" pulsed is not a power measurement; only ∫v i dt is.

Genuinely open. - No one ever published the 100-second test: true input energy, net charge and gas volume together, with a plain-DC control in the same cell. Whether Meyer's pulsed drive makes gas more efficiently than a poor DC drive of the same cell (which thermodynamics allows, up to 121%) is unknown. - Nagypál's 2-4 x Faraday excess, and Perkins's sustained flame at "less than 1 milliamp", were never repeated with the net current measured. - What the actual waveform and pulse length at the cell were. Only pulses shorter than τ (microseconds, on very pure water) would let the cell act as a capacitor. - The history: what NIST, the DTI/Queen Mary experts and the court's three experts actually measured; whether Laughton's view changed between 1990 and 1996; why Project TACO stops in the record in February 1993; what the "Patent Validation Report" of 14 Jan 1983 contained; whether the 1997 "Industrial Contract: Europe" was real; and why the technology went quiet after 1998. Surviving hardware (the buggy, the photographed cells) could still be measured. The records are listed in section 6.

8. Check yourself

1. A cell made of two flat plates 2 cm x 5 cm, 1.5 mm apart, is filled with tap water of 300 μS/cm. Without computing its capacitance or resistance separately, find its relaxation time τ and its Q_cell at 10 kHz. Does making the plates larger help?
2. Meyer writes that at resonance "the current draw ... is minimized". Name the kind of circuit in which this is true, and explain why it does not mean the power drawn is small.
3. A builder reports 3 cc of mixed gas per minute at a meter reading of 50 mA. How many times the Faraday value is this? List two ordinary explanations to rule out before calling it anomalous.
4. Why can no process, electrolytic or not, split one mole of liquid water at 25 °C with less than 237 kJ of work? What extra 49 kJ can come from somewhere else, and from where?
5. Puharich's patent divides 286,021 J/mol by 22,400 cc to get 12.77 J per cc of gas. What is wrong, and what is the correct figure?

9. Further reading

Mainstream: - D. J. Griffiths, Introduction to Electrodynamics: chapter 4 (dielectrics, permittivity), chapter 7 (Ohm's law, charge relaxation in conductors, inductance). - P. Horowitz and W. Hill, The Art of Electronics, chapter 1 (RLC circuits, resonance, RMS versus average). - D. V. Schroeder, An Introduction to Thermal Physics, chapter 5 (free energy; electrolysis of water and fuel cells). - A. J. Bard and L. R. Faulkner, Electrochemical Methods, chapter 1 (Faraday's law, overpotential). - P. Atkins and J. de Paula, Physical Chemistry, the electrochemistry and ion-mobility chapters. - G. N. Glasoe and J. V. Lebacqz (eds.), Pulse Generators, MIT Radiation Laboratory Series vol. 5 (1948): resonant charging chokes and line-type modulators [memory, high for the book; chapters not checked]. - Puteanus et al., "Efficiency improvement by pulsed water electrolysis: An unjustified hope", International Journal of Hydrogen Energy (2025) [read: research/files/Puteanus-2025-IJHE-pulsed-electrolysis.pdf].

Related work by others (discussed by the reviewers; not read in full by any role): - N. Shimizu and co-workers, papers on ultra-short-pulse water electrolysis: the opposing camp to Puteanus et al. (cited there). - P. Graneau and co-authors (with Azevedo and N. Graneau), electrically induced water explosions, Physics Letters A (1986); Gary L. Johnson (Kansas State University), "Electrically Induced Explosion in Water". Meyer files these in his 1995 binder, and Appendix A of the brief cites a Northeastern "Powerful Water-Plasma Explosion" report and the Kansas State report [read: page image p.221]. The reviewers note they describe discharge energy, not nuclear release. - I. Kucherov, US 5,632,870 (1997): an electrolytic cell said to draw ambient heat into the split; cites Meyer. Its claim is bounded by the 121% (ΔH/ΔG) ceiling of section 2. - Gerald Pollack's exclusion-zone water and Lammertink's tuks.nl pages (section 6, "Kept in the record").

The inventor's own key works (all in the project folder unless stated): - US 4,936,961, "Method for the Production of a Fuel Gas" (26 Jun 1990): the water capacitor cell. - WO 92/07861, "Control and Driver Circuits for a Hydrogen Gas Fuel Producing Cell" (1992): the cell numbers used in section 5. - US 5,149,407 / WO 89/12704, "Process and Apparatus for the Production of Fuel Gas and the Enhanced Release of Thermal Energy" (the "fracturing" claims). - EP 0 103 656 / CA 1,234,773 / US 4,798,661 (1982 family): gap resonance. - WO 92/22679 and WO 92/08046 / US 5,293,857 (the last in research/files/): injection and fuel management. - CA 1,213,671 / EP 0 098 897: the Electrical Particle Generator. - The Birth of New Technology (technical brief, c.1996; also a 233-page copy in research/files/ from viXra 0811.0001). - EP 0 111 573 A1 / CA 1,234,774, "Hydrogen Generator System" (filed 1982): Meyer's own 1983 operating point and graphs (section 5, step 7). - WFC newsletters Nos. 3, 4, 8, 10, 11a; the 1989 Einsiedeln conference talk and the 1989 New Zealand house meeting (both machine-transcribed; the New Zealand transcript has no speaker labels, and some figures in it are garbled). - Electronics World + Wireless World, January 1991, pp.10-12 [read: research/files/Electronics-World-1991-01.pdf]: the only contemporary magazine account in hand. - Meyer's 1995 "WFC International Independent Test-Evaluation Report" binder, transcribed at https://stanslegacy.com/documents/ [web; machine transcription, unverified]. - Related: A. Puharich, US 4,394,230 (1983), and his book on the ELF magnetic model of matter and mind (the "PHASER" chapter, pp.176-204, in the folder).

Later accounts and media (not seen by any role; listed so the reader can look for them): Channel 4, "It Runs on Water" (17 Dec 1995); Air & Space Smithsonian (Oct/Nov 1991); Spy (Dec 1991); Ship & Boat International (RINA, Jun 1991); Asian Defence Journal (1992); IECEC 1993 (Bailey and Grotz); Fortean Times (1997).

Answers

1. τ = ε₀εr/σ = (8.854 x 10⁻¹² x 78.5)/(0.03 S/m) = 2.3 x 10⁻⁸ s (23 ns). Q_cell = ωτ = 2π x 10⁴ x 2.3 x 10⁻⁸ = 0.0015. Larger plates do not help: C and 1/R both scale with area/gap, so their ratio, and hence τ and Q_cell, depend only on the water.
2. A parallel LC tank (the coil and the cell side by side, as Electronics World described it). At resonance the supply sees a high impedance and draws little current, while a large current circulates between L and C. The power drawn is still whatever the tank dissipates, mostly V²/R in the water, so a small supply current does not mean a small loss unless the water's resistance is very high.
3. Faraday: 0.05 A x 60 s = 3 C per minute; 3/(2 x 96,485) = 1.55 x 10⁻⁵ mol H₂; x 1.5 x 24.5 L/mol = 0.57 cc/min. 3 cc/min is about 5 times that. Rule out: (a) the meter reads an average or misses short current spikes, so the net charge was larger; (b) the gas is partly dissolved air or water vapour released by heating or the field; also (c) volume read inaccurately, or not at the stated temperature and pressure.
4. Energy (strictly, the Gibbs free energy at fixed temperature and pressure) depends only on the initial and final states, not on the path, so any process from liquid water to H₂ + ½O₂ at 25 °C needs at least ΔG = 237 kJ/mol of work. The remaining TΔS ≈ 49 kJ/mol (up to ΔH = 286 kJ/mol) can be heat drawn from the surroundings, so the gas can hold up to 121% of the electrical work put in, with the room supplying the difference.
5. One mole of liquid water gives 1.5 moles of gas (1 H₂ + ½ O₂), which is 33,600 cc at STP (standard temperature and pressure: 0 °C and 1 atmosphere, where one mole of gas fills 22.4 litres), not 22,400. The correct figure is 286,021/33,600 = 8.51 J per cc of mixed gas (7.05 J/cc on the Gibbs basis). With it, the patent's 114.9% becomes 77%.

Corrections to earlier documents (for the orchestrator)

1. WO 92/07861's own operating point (new; supports, does not change, arbiter claim 1b). The patent pairs "unipolar pulses from 0 to 650 volts" [read: p.13] with "a 1-2 milliamp minimum leakage condition" at resonance in the 3-inch cell [read: p.9]. For that cell, 650 V at 1-2 mA requires σ = 1.3-2.6 x 10⁻⁶ S/m, 2 to 4 times below the conductivity of theoretically pure water [calc: example.py step 5]. The reviewers checked the 40 kV / 1 mA figure from Appendix A but not this one. Add it to arbiter.md section 3 beside the 40 MΩ point.
2. Nagypál's millivolt observation (missed by all reviews and the arbiter). The transcribed letter (binder p.66) says: "I got a rising step-waveformed current-pulses, between the plates appeares only some millivolt potential difference, due to water's resistance, caused by contaminants." It also records a 1 kΩ resistor in series with the secondary and 1 kΩ resistors with the chokes. This is a builder's direct statement that his tap-water cell did not hold the high voltage, consistent with Q_cell ≪ 1; it also makes his 20 mA reading harder to interpret (a few millivolts cannot drive electrolysis, so either the voltage reading or the current path is not what it seems). The arbiter's Nagypál paragraph (section 3, item 3) and Correction 7 should quote it alongside "your theory works" and "can't reproduce".
3. Meyer's 1995 table, two further slips (binder pp.59-61, stanslegacy transcription, unverified). (a) Line 10 computes the "prior art" yield as "1cc x 4.4 amps / hr = 4.4 cc/hr" although line 8 gives 10 cc per ampere-hour; consistent use gives 44 cc/h, and the "27.5 cc/hr" baseline would be 275, cutting the "Differential Efficiency Factor" from 1,696 to about 170 even on Meyer's own baseline. (b) Line 14 uses "Mass Unit of Water Molecule = (2H x 1Mu) + (1 Oxy x 8 Mu) = 10 Mu's" to say the gas is 20% hydrogen; oxygen's mass number is 16 (8 is its atomic number), so water is 18 mass units and 11% hydrogen by mass, and by volume electrolysis gas is 67% hydrogen. The reviewers found the 68x baseline and the psi-as-volume errors but not these. Add to physicist/arbiter notes on the table; neither changes a verdict.
4. Who said patents require a demonstration. The arbiter (section 3, item 5) and physicist attribute the patents-as-confirmation argument to Hindley. The Electronics World article itself (editorial text, not only Hindley's quote) states that a Section 101 grant "is dependent on a successful demonstration of the invention to a Patent Review Board" [read: research/files/Electronics-World-1991-01.pdf, pdftotext]. Attribute it to the magazine and to Hindley; the reviewers' MPEP 2107 check stands as the counter-evidence.
5. "Parallel resonant circuit". The same article describes the coil and cell as forming "a parallel resonant circuit" excited through a rectifier diode as a charge pump. This supports the mathematician's correction to claims.md ("current minimised at resonance" is a parallel-tank property), and should be cited there.