Gray-EV: Mathematician's review (Step 2)

Reviewer role: PhD mathematician. Checks derivations, units, hidden assumptions.

Source gap (repeated from Step 1). The patents are scanned images with no text layer. I re-read page images of the Crosby letter (full page), GB 2,030,801 p.1 (full spec and claims), US 3,890,548 pp.1, 7, 9 (background; drawing sheet 6), US 4,595,975 p.7 (claims). I did not read the middle description pages of US 3,890,548 or the description of US 4,595,975/4,661,747. The newspaper/magazine articles were not re-read; I rely on Step 1's summary of them [read: Step 1 facts.json, unverified by me]. No independent test report, waveform, or power-meter record exists in the folder. Conclusions on claims 4 and 5 rest on press text and [memory].

Web verification. WebSearch was refused (session budget exhausted). One WebFetch of Google Patents US3890548A succeeded: https://patents.google.com/patent/US3890548A/en. Web checks of the Gray prosecution history (claim 5) failed; the relevant statements remain [memory]/press-sourced and are marked as such.

Corrections to Step 1

1. The Crosby letter does NOT say "exceeds 100%". The page image reads: "The measured overall system efficiency exceeds 99%" [read: page image, crosby_letter.pdf]. The OCR "ggt" was "99%". Step 1 and claims.md claim 2 ("measured efficiency exceeded 100 percent") are wrong on this point. The letter is therefore internally inconsistent in a different way (see Claim 2): it states efficiency above 99% and, in the same letter, an output/input ratio of about 278.
2. Power figure is 26.8 W, not 26.9 W, in the letter. The image reads "9.75 watts; this equals 26.8 watts per hour". (The TATTLER reprint, per Step 1, gives 26.9; the two sources differ.) 5,454 / 26.8 = 203.5 h; 5,454 / 26.9 = 202.8 h. Both round to "203".
3. Battery figure is clear: "5,454 watts for one hour" (i.e. 5,454 Wh), not garbled. Signatory is J. A. Maize, Director of Engineering. The letter carries a notary seal (Carol L. Dye, 9/28/73). A notary seal attests the signature, not the measurements. The letter says "Tests reports supporting the above information is available at EvGray" (no report in the folder).
4. Step 1 said the letter's "9.75 watts" was for a 21-minute run. The letter writes "9.75 watts" for a 21-minute run, which is a unit error (watts are a rate; the quantity must be watt-hours). Read as 9.75 Wh in 21 min (0.35 h) the average power is 27.9 W, not 26.8 W (error about 4%). Read as 26.8 W x 0.35 h = 9.38 Wh. Neither matches exactly. [calculation]
5. Capacitor count / voltage (18 caps at 3,500 V "from the block diagram") could not be confirmed. Drawing sheet 6 (Figs 12-19) shows no such block diagram, and the Google Patents text (WebFetch) mentions nine stator electromagnet pairs with corresponding capacitors, three rotor pairs, and no numeric voltage ("charged until breakdown potential"). Treat "18 x 3,500 V" as unverified.
6. static_generator.pdf was identified by me: it is an EvGray sales sheet for "EVGRAY STATIC GENERATORS" (rotated page image): Generator-only model about 45 W from a 12 V car battery to run 110 V devices; battery model 30 W for 6 h; "No moving parts"; priced $50 and $75 [read: page image]. It is an inverter-type product sheet like ems_inverter.pdf, and says nothing about static electricity in the physics sense. It claims "approx. 50% brighter illumination" and longer battery life, with no data. Step 1 listed it as unidentified.
7. Step 1 correctly stated that the 1975 patent text itself invokes conservation of energy; I confirmed this on page 2 of US 3,890,548: "believed to operate on the principle of conservation of energy" [read: page image p.9] and GB 2,030,801: "an energy conserving system" [read: page image p.5]. Note this is stated as a belief.

---

Claim 1: Pulsed capacitor-discharge motor (US 3,890,548)

a. Internal consistency. The device description is mathematically unobjectionable. Energy stored in a capacitor is E = (1/2) C V^2 (C = capacitance, V = voltage). Discharging it through a coil of inductance L and resistance R gives a damped RLC (resistor-inductor-capacitor) transient; with a spark gap as a switch, the gap closes when V exceeds breakdown. A force between repelling same-polarity cores scales with coil current squared (or product of two currents), so an impulse is delivered each discharge. The patent's phrase "current is needed only at the instant of a capacitor discharge" is true of any pulsed circuit but is not a statement about energy: energy per pulse is the integral of V*I dt, however brief. The inference "short pulse, therefore little energy" is a hidden non-sequitur if intended as an efficiency argument. The patent's own text does not make the over-unity claim.

Worked scale check. Mechanical power P = 7,460 W at 1,100 rpm = 18.33 rev/s means 407 J per revolution at 100% conversion. At 3,500 V (if that figure is right) the capacitance needed to supply 407 J per revolution is C = 2E/V^2 = 66 microfarad per revolution. This is not absurd for pulse engineering, so the motor is mechanically feasible as a pulsed machine. The problem is the input accounting (claim 2), not the motor. [calculation]

b. Consistency with experiment. Pulsed-coil, capacitor-discharge and spark-gap switching are standard (coil guns, ignition systems, pulse-power) [memory: high]. Repulsion motors exist. Nothing here contradicts experiment as long as input energy is accounted for. No evidence in the folder that this motor ran continuously on a load.

c. Tested? The only test records are the promoter letter and press accounts. AP item (per Step 1): TRW's Donald Cronin saw a test where the motor ran about 10 s then stopped [read: Step 1 summary of various_articles.pdf; not re-read by me]. No independent, controlled test on record in the folder. I could not verify any other tests on the web (search unavailable).

d. What would have to change? Nothing; as a motor it is ordinary physics.

Verdict: Plausible, untested as a motor that turns (working pulsed repulsion machine, efficiency unknown). Not "Established": no data in the folder show it running a sustained load.

---

Claim 2: 10 hp out for about 26.9 W in; efficiency exceeded 100%; 203 h

a. Internal consistency. The letter fails its own arithmetic and definitions. I checked every number from the page image.

• Output: 48 ft-lb at 1,100 rpm gives P = torque x angular speed = 48 x 1.3558 N·m x 115.19 rad/s = 7,497 W = 10.05 hp. The stated 10 hp / 7,460 W is correct to within 0.5%.
• Efficiency, defined as output / input: 7,460 W / 26.8 W = 278 (that is, 27,800%). The letter instead says "exceeds 99%". If efficiency were 99%, input would be about 7,535 W. The two statements cannot both hold. Either "99%" is a typo for a much larger number, or it is a different quantity (e.g. "the battery is recycled to 99%"), but the letter does not define it. [calculation]
• Duration: 5,454 Wh / 26.8 W = 203.5 h, which matches the stated 203 h. But 5,454 Wh could deliver 7,460 W for only 5,454 / 7.460 = 0.731 h = 43.9 min if the output were taken from the batteries. The letter's "203 hours at 10 hp" implies 203 x 7.46 kW = 1,514 kWh of mechanical work from 5.454 kWh of stored chemical energy, a factor of 278.
• Units: "9.75 watts" for a 21-minute run is dimensionally a power, not an energy; "26.8 watts per hour" is not a unit of anything (watts per hour is a rate of change of power). The author clearly means watt-hours and watts, and the conversion 9.75 Wh / 0.35 h = 27.9 W does not give 26.8.
• Input power is not defined: the letter does not state where, with what instrument, and over what waveform it was measured. With 2-3 kV capacitor pulses and a vibrator/inverter charging stage, the battery current is a train of spikes; a DC ammeter or slow voltmeter-based estimate easily under-reads. Mean power needs the time-average of v(t) i(t) over complete cycles, measured with a wideband instrument. The letter gives none.
• "21 minute run" against "203 hours": the claim for 203 h is an extrapolation, not a measurement (the run measured is 21 min, or 0.17% of 203 h). Also the letter mixes the May 1-10 tests (Pan World Enterprises) and the Sep 25 "confirmation" without saying which measured what.

b. Consistency with experiment. A 278:1 output-to-input ratio with no external source contradicts energy conservation (first law) and every measured power balance of electric motors, which are 80-95% efficient at best. The extra input would have to be either unmeasured (pulse spikes, stored energy in capacitors at start and stop), or external (the "atmospheric static" suggestion; see claim 4). Mundane routes: (i) battery drain measured by DC meter; (ii) capacitors and flywheel inertia charged before the 21-min window; (iii) the dynamometer load or brake torque was not what was stated. I cannot say which; no data. [memory: medium on pulsed-measurement being a common failure; I did not verify this by web.]

c. Tested? By promoter-side engineers only. The letter is addressed to Larry Crosby by an institute named after him (the addressee's name and the institute's name coincide, 170 North Robertson Blvd appears as both the institute's and Mr. Crosby's address [read: page image]), so it is not independent. The AP item (per Step 1) gives TRW's Cronin seeing 10 s of running. No replication on record. Never independently measured with instrument records.

d. What would have to change? Conservation of energy would fail, or there would be an undiscovered energy input (the claim says "no external recharging generator" but that is not the same as no energy source).

Verdict (split). - As stated (10 hp from 26.8 W continuous): Refuted by calculation if taken as a true time-averaged electrical input from the batteries: output energy exceeds input by 278x, violating the first law of thermodynamics. Calculation above. - As a test report: Internally inconsistent (efficiency "exceeds 99%" vs 278:1; watts-per-hour units; 21-minute run extrapolated to 203 h; 9.75 vs 26.8). The document cannot be used as evidence of anything beyond a mismeasurement. - Whether a controlled test with proper power metering would show anything: Never properly tested. It was not dismissed without testing (Cronin and a Japanese group are said to have looked, per Step 1's summary of the AP item), but no instrumented neutral test is on record.

---

Claim 3: Back-EMF capture returns most of the energy to a battery or capacitor (GB 2,030,801; US 4,595,975; US 4,661,747)

a. Internal consistency. GB 2,030,801 is consistent as a circuit, and I checked the text page by page. Phase A: a 24 V battery charges a capacitor to about 2,000 V via a DC-DC converter. Phase B: closing the switch discharges the capacitor through inductive load L and a second battery "now closed switch number 2 ... through energy recovery battery number 2"; at capacitor voltage near zero, inertia of the inductor current would reverse-charge the capacitor ("overshoot"). Phase C: switch 1 opens so the current continues through L, switch 2, battery 2 and a diode path, a closed loop. This is the standard freewheel/flyback circuit. Energy bookkeeping: with E_C = (1/2)CV^2 delivered, E_C = W_mech + E_loss(R) + E_battery2 + E_L(final), and the energy that returns to battery 2 is bounded by E_C minus the work done and the losses. Recovery cannot raise the net: recovered energy was first taken from the source. The patent's wording is modest: "believes this to be a more efficient use of energy", "energy conserving system", and the only claims are "substantially as hereinbefore described" (claims 1-2 are by reference to the figures) [read: page image p.5]. In US 4,595,975 the claims (p.7) say "coupling a second capacitor to said load; and coupling said capacitor to said source", i.e. a second capacitor receives a signal from the load and returns it to the source (claims 2-4, 8). No quantity such as a recovery fraction is claimed. [read: page image p.7]

The claim as phrased in claims.md, "most of the energy is recycled", is not in the patent claims; it is a paraphrase of promoter statements. Mathematically, recycling a fraction f of the energy of each pulse gives an effective input per pulse of (1-f)E for an output of ηE... any 0 < f < 1 is consistent; only f giving output above input would be inconsistent.

b. Consistency with experiment. Flyback and regenerative recovery are routine: switch-mode power supplies, regenerative braking, resonant-charging inverters [memory: high]. Typical recovery is partial (tens of percent), with losses in the diode (about 0.7 V drop), the battery internal resistance, and charge acceptance of a lead-acid battery at high current (poor at high pulses). A pulse of high voltage into a 12 V battery has a very large V_in/V_battery ratio; most energy goes into the inductor current first, then into the battery; fine.

c. Tested? No test of the recovery circuit's efficiency is in the folder. Standard practice elsewhere shows partial recovery.

d. What would have to change? Nothing for partial recovery. For net gain from recovery alone, the first law.

Verdict (split). - Weak form (back EMF can be captured and partially returned to a battery): Established (standard power electronics; [memory: high]). - Strong form (so much is recycled that output exceeds input or that system efficiency exceeds the 99%/278:1 of claim 2): Refuted by calculation (a recovery path returns energy that was already supplied; it cannot create energy). - Gray's specific circuits' recovery fraction: Never properly tested (no data).

---

Claim 4: "Cold" or static electricity (possibly atmospheric) powers the motor

a. Internal consistency. "Cold electricity" has no definition in the sources. Electrical energy is carried by fields; V*I is power whether the pulse is "cold" or not. The press demonstrations (magnet "caught in a bare hand", per Step 1's summary of ems_electronic_power.pdf; I did not re-read it) show only that a magnet's kinetic energy was small: a 1.25 lb magnet (0.57 kg) moving a few m/s carries about (1/2)(0.57)(3^2) = 2.6 J, safely caught. The "runs cool" statement is unquantified: a motor at 10 hp with 99% efficiency would dissipate 75 W, which also runs cool, so "cool" does not discriminate. Statement is not falsifiable as stated because no measurable differs from conventional electricity is specified. The sales sheet (static_generator.pdf) uses "static" as a brand name for an inverter. "Atmospheric static electricity" energy: the fair-weather vertical field is about 100 V/m near the ground [memory: high], an energy density (1/2) eps0 E^2 = 4.4e-8 J/m^3; 7.46 kW from a collector would require an enormous collection region. [calculation: eps0 = 8.85e-12 F/m; E = 100 V/m: 0.5 x 8.85e-12 x 1e4 = 4.4e-8 J/m^3.] No such collector is in any patent I read.

b. Consistency with experiment. No experiment shows a "cold electricity" with distinct properties. The patent text claims none.

c. Tested? No test of any such power source or antenna is on record.

d. What would have to change? Electrodynamics would need a new energy source or a new sector; or atmospheric electricity would need to be collected at kilowatt levels, which known fields do not supply.

Verdict: Not testable as stated (undefined term). If read as an atmospheric energy supply: Refuted by calculation at the stated 7.46 kW (field-energy estimate above, order of magnitude), with a caveat that the claim is not in the patents. The patent text (3,890,548, p.2) attributes operation to conservation of energy, contradicting claim 4.

---

Claim 5: The LA DA prosecution (1974-76) was meant to suppress the invention

a. Internal consistency. Not a mathematical claim. The press narrative has no internal contradiction except that it requires motive; the same Step 1 summary gives the AP-reported DA reasoning that the device "would violate all the laws of physics" and that investors complained. Both versions are consistent with the same facts (raid, charges, eventual plea).

b. Consistency with experiment. N/A. Note from claim 2: if the motor's claimed output is not supported by the letter's own numbers, a fraud investigation (investors told 278:1 efficiency) is a plausible ordinary explanation. That is inference, not evidence of intent.

c. Tested? The facts (raid, eight counts dropped, two securities misdemeanours, 1976 plea) are as reported in the articles [read: Step 1 summary; tabloid/magazine sources, not independent]. I could not verify on the web (search budget refused). Motive cannot be tested from the documents. Court records not in the folder.

d. What would have to change? None for physics; it is a claim about officials' intent.

Verdict: Not testable as stated from the folder (intent); the outcome (plea to two securities counts) is press-reported [memory/press: medium, unverified on the web].

---

Summary table