Gunderson: Physicist review (Step 2)

Reviewer: PhD physicist. Sources: facts.json, claims.md, deep/ reads, the patent texts in _text/ (read myself: US 2006/0163971 paras 0013-0015, 0039-0044, claims; US 8,093,869 cols 7-8 "[B]" and "[C]", FIG. 5/6 text on pp.19-21, claim 1), and the Investigator's sites.md and research/ notes. Tags: [read] document, [web] URL, [memory] recalled with confidence.

Source note: both documents are patents by the inventor's assignees (Magnetic Power Inc.; Chava Energy LLC), so they are the applicants' characterisation. Neither contains any measurement, prototype or energy balance [read]. The only numbers on performance come from later, non-patent sources (conference talks, a forum witness, assignee web pages), covered under "Investigator's findings" below.

Reader primer: the two physics ideas you need

1. Lenz's law and passivity. A load that draws power from a generator makes the generator work harder. In a transformer, a loaded secondary raises the primary current because the primary must now supply the output power as well as its own losses. Energy conservation for any passive linear network: real power in = power dissipated in the load + losses, never less.
2. Phase and "real" versus "reactive" reaction. Take a flux Phi(t) = Phi0 cos(wt) linking a closed loop of resistance R. Faraday: emf = -dPhi/dt = Phi0 w sin(wt). With a resistive loop, current i = emf/R is in phase with sin(wt). The loop's own reaction flux is proportional to i, so it is proportional to -dPhi/dt, which is 90 degrees away from Phi, not 180. Only a purely inductive loop (current = -Phi/L, no power consumed) gives a reaction flux exactly opposite (180 degrees) to Phi. So "180 degrees opposition" belongs to the reactive part (no power), and the power-consuming part is a quadrature, rate-proportional (damping-like) reaction. [read: this is derived above from Faraday's law; compare deep read note that the patent's 180 degree step is asserted]. This distinction is the heart of claim 3.

---

Claim 1: "virtual motion" of a permanent magnet's flux in a modulated soft core induces EMF with no moving parts (US 2006/0163971)

a. Mathematics. The document has no equations [read: deep read, pp.1-10]. Qualitatively it is internally consistent: a magnet injects flux into a soft core; an input coil changes the core's permeability or saturation state; the flux share going through different paths (around holes) changes in time; Faraday's law (emf = -dPhi/dt) then gives an emf in a wire threaded through the holes. The "sun seen from under agitated water" analogy [read: para 0014] is only an image; nothing is quantified.

b. Consistency with experiment. Fully consistent, and not new. Time-varying reluctance or permeability with a permanent-magnet bias is the principle of variable-reluctance pickups, fluxgates, magnetic amplifiers and parametric (pumped) transformers [memory, high]. The Patent Office citations (Sony US4103197, Magnetics Research US4639626 permanent-magnet variable-reluctance generator) and the "similar document" MEG US6362718 show the same family of ideas [web: https://patents.google.com/patent/US7830065B2/en via Investigator's notes]. Nothing says "virtual motion" needs to be a new kind of motion; it is flux redistribution.

c. Tested? No test is reported in the patent. The Investigator found no independent replication of the 2006 geometry. The granted version US 7,830,065 B2 (9 Nov 2010) added a resonant circuit with a square-wave source to claim 1 and a new geometry claim 11 [web: https://patents.google.com/patent/US7830065B2/en, checked by me; lapsed 9 Nov 2022, fee unpaid]. I confirmed the page keeps the "absence of electrical loading ... decreasing amounts of energy" passage. Grant shows the claimed structure is novel to an examiner; it is not a performance test.

d. If true. Nothing in mainstream physics changes for "EMF without motion by modulating permeability". It is standard.

Verdict: Established for the mechanism (flux modulation induces EMF with no moving parts); Plausible, untested as a practical generator (no output, efficiency or loss figures given). Which source supplies the energy is a separate question (claim 2).

Claim 2: the Lenz reaction opposes the magnets, aids the input, so more output needs less input, powered by the magnets until demagnetised

a. Mathematics. No equations, only flux-arrow bookkeeping (FIG. 3; paras 0039-0044). The argument adds fluxes in the core: on the magnet side of each hole, the reaction flux (c) cancels the magnet's flux (b); on the other side, (c) adds to input flux (a) [read: paras 0042-0043]. Flux superposition in a core is a statement about B at points; it is not an energy statement. Energy flow is given by the products emf x current in each winding. The document never computes the power at the input coil. Whether the output loop's flux "aids" the input flux does not determine the sign of power at the input; in a coupled system the input terminals see the reflected load through the mutual inductance, and for a passive network that reflected resistance is positive. So the inference "flux aids, therefore less input power" is not supported by the argument given. Internally the text is consistent in its own terms; the missing step is the energy accounting.

b. Consistency with experiment. - Energy conservation / parametric systems. A system whose permeability is pumped by the input coil is a parametric device. The Manley-Rowe relations [memory, high] say power delivered at the signal/output frequencies comes from the pump; a static bias magnet does not pump. So output power must be matched by input (pump) power plus losses, unless the magnet's state changes. - The magnet as fuel. Permanent magnets are not fuel in a cycling device. A magnet working on a reversible minor B-H loop returns to its starting state every cycle; its stored field energy has no net change. Only irreversible demagnetisation would consume it, and then the available energy is bounded by the energy product. Calculation (generous upper bound, treating the whole energy product (BH)max as extractable, which it is not): - sintered NdFeB (BH)max about 300-400 kJ/m^3 [memory, high; Wikipedia states NdFeB energy is "about 18 times" ferrite by volume, https://en.wikipedia.org/wiki/Neodymium_magnet, so ferrite is about 20-30 kJ/m^3, consistent with my memory]. Take 330 kJ/m^3. - A 10 cm^3 NdFeB magnet (1e-5 m^3): 330e3 x 1e-5 = 3.3 J. At 10 W that is 0.33 s of output. - A 10 W lamp for one hour is 36 kJ, which would need 36e3 / 330e3 = 0.11 m^3 of NdFeB (about 800 kg at 7,500 kg/m^3 [memory, high]). - One year at 10 W is 3.15e8 J, about 950 m^3 of NdFeB. So paragraph 0040's "very long periods" without demagnetisation is incompatible with the magnets being the energy source of a useful output; the energy either comes from the input source or from somewhere not identified. - Input lower as output rises. "Decreasing amounts of energy are generally required as increasing amounts are withdrawn" [read: para 0044] is the opposite of the loading behaviour of every measured transformer, generator or parametric amplifier. It is not outlawed by any single law if an unidentified source exists, but it is a perpetual-motion-class statement: with input approaching zero while output continues, the device is claimed to be a net energy source. The Patent Office classified the family under H02K 53/00 "alleged dynamo-electric perpetua mobilia" [web: Investigator, Google Patents]; that is the office's labelling, not a ruling on physics.

c. Tested? - In the patent: no. One unquantified sentence, "Experiments show ..." [read: para 0040]. - Inventor's later statements (priority 1): at the Energy Science and Technology Conference (ESTC) 2016 Gunderson described a "Magnetic Implosion Transformer" with a lamp "drawing 570% more energy than could be accounted for going into the input", input dialable to zero, "still in the process of validating", "not quite sure why it works" [web: https://emediapress.com/shop/magnetic-implosion-transformer/, the video seller has a commercial stake; description is signed by Gunderson]. His 2015 device was reported at COP about 1.02-1.03 ("barely overunity") [web: Investigator, energeticforum/rexresearch]. By 2018 he states that part of the force "is taken from the ambient energy" [web: https://web.archive.org/web/20180115005940/http://grahamgunderson.com/ou], whereas the 2006 patent said the magnets are the source. The stated energy source has changed. - Witness account (priority 2): Mark McKay, PE, a friend of about ten years, not a commercial party, reports two power analyzers and a Tektronix scope "agreed" at 1.53 W in, 9.43 W out (COP 6.16), and an hour later input 0 W with output 9.5 W [web: https://www.energeticforum.com/forum/energetic-forum-discussion/renewable-energy/13945-magnetic-implosion-transformer-replication]. A second attendee ("Reiyuki") recorded about 3.5 W in and 8.5 W out, with the result dependent on phase (110 degrees: 4 W in, 8 W out; 140 degrees: -1 W in, 3 W out). I weigh these as real reported instrument readings from an engineer, second-hand to me, with no waveforms, measurement points or raw data published. - Assignee statements: MPI's 24 Apr 2008 "POWERGENIE" report claims 2007 proof-of-concept prototypes with "reaction-less" behaviour; Chava's former web page claims "an early prototype ... more than 100 times the input" [web: Wayback magneticpowerinc.com and rexresearch.com/gunderson/gunderson.htm]. These come from the assignee, an interested party seeking investment; an anonymous "Physics Review Board" calls this fraud, an equally non-neutral critic. Both are the parties' characterisations. No data from either was found. The assignee offered NDAs, so data may exist privately. - Replication: none found, as of the Investigator's search.

What the reported measurements could mean (not an accusation; a list of what a test must exclude). The device is a resonant H-bridge at 20-50 kHz, 200-300 V in, a step-up to about 800 V peak, a synchronous rectifier "opened briefly near the phase transition" producing a 1-2 kV spike [web: same thread]. For such waveforms: 1. Real power is V x I x cos(phi). Near resonance the current and voltage are close to 90 degrees apart, so a small phase error in the current probe or analyzer channel skew gives a large error in watts, even a negative one. A negative "input power" at some phase settings (Reiyuki's -1 W at 140 degrees; Gunderson's -0.25 W) is exactly what a phase or bandwidth error looks like, though it is also what genuine energy return from stored reactive energy looks like for part of the cycle. 2. Narrow kV spikes need analyzer bandwidth far above the 50 kHz fundamental; under-sampling can hide energy on one side of the ledger. 3. "Input 0 W with output 9.5 W for an hour" is checkable without a power analyzer: the 12.7 Ah, 12 V battery [web: same thread] holds about 150 Wh. A 9.5 W load for one hour is 9.5 Wh. If the input is truly zero the battery voltage and charge would not fall at all; a coulomb count or a before/after capacity discharge test settles it. Output of 9.5 W for 1 h is 34 kJ; the magnets in the device (about 20 pieces of 2 x 3 x 0.5 in, about 1,000 cm^3 total [web: same thread]) have an upper-bound energy product of roughly 30 J if ferrite (25 kJ/m^3 x 1e-3 m^3) and 330 J if NdFeB, i.e. 100 to 1000 times too small, and they were not reported demagnetised. 4. A lamp's electrical output is inferred from the analyzer reading; no calorimetry or photometry. The reported hour-long zero input is the strongest statement in the record and the one a closed-loop (self-running) or coulomb-count test can decide. McKay himself says that if the instrumentation is right it "should be able to loop the system and have a self runner, but he ran out of time" [web: same thread].

d. If true. Energy conservation would have to be revised (or a new energy source identified); a coupled-inductor system would have to show negative reflected resistance at the input terminals while delivering power. That would be a net energy source, not a rearrangement of Lenz's law. Zero-point-energy language used by the assignee [web: MPI site, Chava page] offers no mechanism with a calculable power.

Verdict: Split. (i) Strong form (more output, less input, powered by magnets until demagnetised): Refuted by calculation for the magnets-as-fuel statement (energy product bound above) and for the passivity argument; for the experimental claims of COP 6 and zero input (2016, ESTC): Never properly tested (no neutral controlled test, no published waveforms or replication; not dismissed without testing either, since forum engineers examined it and no refutation has been published, but no independent lab has been involved). (ii) Weak form (flux superposition lowers the reflected reactance at some phases): Plausible, untested for a parametric core.

Claim 3: acoustic Lenz reaction returns to the source with 90 + 180 + 90 = 360 degrees, so it aids the drive and less drive energy is needed (US 8,093,869)

a. Mathematics. The arithmetic 90 + 180 + 90 = 360 is trivially right, and the underlying numbers check: wavelength of 125 kHz at 5000 m/s is 4 cm, so a quarter wave is 1 cm; the electromagnetic quarter-wave is about 600 m [read: deep read; I recomputed: c/125 kHz = 2398 m, a quarter is 599.6 m]. The inconsistency lies in the 180 degree step [read: cols 7, 20-21]. - The patent derives 180 degrees from "magnetic opposition (Lenz) is opposition, therefore antiphase". But Lenz opposition is to the rate of change of flux. As shown in the primer box, for a resistive (power-consuming) coil circuit the reaction is in quadrature with the flux (and so with the stress that produced it), and only the lossless inductive part is at 180 degrees. The patent itself concedes the phase "depends on inductance and tuning capacitor" [read: deep read]. - A resonator also has no single "arriving reaction wave" separate from the standing wave; the quarter-wave propagation delays are already part of how a standing wave is set up. - The decisive test is not phase bookkeeping but power: write the drive-point impedance Z = F/v at the source. Power into the source is 0.5 Re(F v*) = 0.5 |v|^2 Re(Z). For a linear, passive medium with a dissipative load, Re(Z) >= 0 at every frequency and power in equals power in the load plus other losses. An "aiding" reaction that reduced input below output would require Re(Z) < 0 (a negative resistance), which a passive coil and resistor cannot provide.

My numerical check (SymPy/NumPy venv; script at /tmp/gund/sim.py, not part of the record). A free-free elastic chain of 200 masses and springs (lossless apart from a tiny 1e-4 internal loss), force applied at one end, and a strain-rate-proportional damper (the analogue of a resistive coil: the magnetostrictive reaction stress proportional to current, which is proportional to strain rate) on the central spring (the nodal region) [memory: model is mine]. Over 4001 frequencies spanning the third odd resonance: power in minus power in the damper was >= 0 at every frequency (minimum 2.5e-6, i.e. zero within numerical error). At the resonance, damper power was 50%, 75% and 91% of drive power for damping strengths 0.01, 0.03 and 0.1. A purely elastic (180 degree, reactive) change of the central spring, +0.5 or -0.5 of its stiffness, drew essentially zero power (0.05 and 0.005 on a scale where loaded runs draw about 2000). The elastic reaction therefore shifts the resonance; only the damping-like reaction extracts power; and it does so at a cost to the source.

Careful reading of "requires less energy". The patent says the source needs "less acceleration input ... than was necessary before current was allowed to flow" [read: p.20]. With a constant-force source driven at resonance, adding any damping lowers the vibration amplitude and so can lower the power drawn from the source; that is ordinary resonator behaviour, not a gain. What the patent never shows (and I did not model beyond the identity above) is input power lower than output power. Claim 3 is therefore ambiguous: read as ordinary loading it is true and trivial; read as "the load reduces the source's work below the power delivered", it is refuted. The patent text contains no energy balance, so the stronger reading cannot be ruled in from the document.

b. Consistency with experiment. Magnetostrictive and piezo-magnetic transducers, gyrators and harvesters (Bell Labs US2101272, Butler US4443731, the Hayes patents cited on the face of US 8,093,869) work as passive reciprocal transducers [web: Investigator, Google Patents]. No transducer is known to return net energy to its source.

c. Tested? No test in the patent. The assignee's 2008 report claims "reaction-less" generation in 2007 prototypes [web: Wayback, MPI] without data. Later, mainstream groups cite the patent (US DOE US9960341B1, Oscilla Power, Sumitomo Riko, Mitsumi) as prior art on magnetostrictive harvesting [web: Investigator], which is citation as related technology, not replication of the "aiding" effect. Tae Chang N.E.T. "Sound wave resonance generator" has priority the same day (3 Dec 2008) [web: Investigator]; I have not read it.

d. If true. A passive coil-plus-magnetostrictor would have to present negative resistance to the drive, violating the second law/energy conservation for passive networks.

Verdict: Refuted by calculation for the strong form (aiding reaction, drive power below delivered power; passivity plus the quadrature argument). The weak form (load current changes the resonator's frequency or amplitude; magnetoelastic coupling modulates the drive) is Established as ordinary electromechanical coupling. The claim of an in-phase return was never tested; I note no experiment exists to contradict it, but the calculation is independent of any test.

Claim 4: opposite currents in neighbouring coils add magnetic opposition that increases delivered energy and cancels inductive reactance, "by analogy with a loaded transformer"

a. Mathematics. Two parts. (i) Reactance cancellation: two coupled coils in series-opposing have inductance L1 + L2 - 2M, so a net reactance reduction when coupling M is nonzero; the patent says this lets more current flow [read: pp.21-22]. That is correct circuit theory and can raise deliverable power for a given source emf (matching), without creating energy. (ii) "Each current acts as a transformer primary to the opposite turn, so the total power increases when both turns conduct" [read: p.22]. This misuses the transformer analogy. In a transformer a loaded secondary raises primary current because the primary source delivers more power. In the coil pair, both emfs come from the same acoustic stress wave; coupling between them is included in the circuit equations and the total is bounded by the acoustic power supplied. A mutual coupling that "reinforces" both currents would again be a negative-resistance circuit element.

b-c. Reactance cancellation by opposing windings is ordinary circuit practice [memory, high]. No test of the energy-increase statement exists in the documents.

d. For the strong form, energy conservation would have to fail.

Verdict: Split. Reactance cancellation / improved impedance matching: Established (circuit theory), small benefit, no data in the patent. "Increases total electrical energy" beyond the acoustic input: Refuted by calculation (passivity).

Claim 5: square-wave drive makes harmonic peaks coincide so power (I^2 R) scales as the square of the summed amplitudes

a. Mathematics. Correct about shape, wrong about energy. Peaks do coincide in a square wave (central node of a free-free rod is a displacement node for all odd modes, which I checked: mode cos(n pi x/L) is zero at x = L/2 for odd n). But the average power is what matters: the cross terms between different harmonics average to zero over a period (orthogonality, Parseval). Numerical check: unit-amplitude square wave, mean of s^2 = 1.00, sum of harmonic average powers 0.998 (with the odd harmonics up to n = 199). A sine at only the fundamental, with amplitude 4/pi, has average power 8/pi^2 = 0.81, so square-wave drive gives 1.23 times the fundamental-only power, which is exactly what the extra input energy in the harmonics supplies. The "square of the sum" is the instantaneous peak, at the moments of peak, balanced by the near-zero values elsewhere in the cycle.

b-c. The Parseval relation is basic Fourier analysis [memory, high]. No test is needed or reported. The patent presents no power measurement.

d. Nothing changes; the effect described is just a larger drive.

Verdict: Refuted by calculation for "substantially increases generated power" relative to the power put in; the peak-coincidence observation itself is Established but gives no net gain.

Claim 6: source-to-pickup spacing of at least 1/8 (preferably 1/4) acoustic wavelength distinguishes the device from prior gyrators and harvesters (claim 1 of US 8,093,869)

a. Mathematics. The geometry is consistent: the strain maximum (best pickup location) of a resonant bar lies a quarter wave from a free end; and a nonlinear time-changing load (semiconductor switches) is added to claim 1 but is not described in the text [read: claim 1; deep read]. The OCR renders 1/8 as "%"; I read the glyph from the text context ("not less than 1/8") and rely on the deep read's page images, tagged [read: page image p.22].

b. A pickup at the strain antinode is standard practice in resonant magnetostrictive devices [memory, medium-high; patent-family neighbours in the cited art]; whether the 1/8 separation is novel is a patent-law question that I do not judge (the Patent Office granted it, active to 2029; fees paid late in 2019 [web: Investigator]).

c. No performance comparison against prior devices appears in the patent.

d. Nothing in physics changes.

Verdict: Established that the geometry is a workable transducer arrangement; Never properly tested for any advantage over prior art; novelty is for patent law and is not a physics verdict.

---

Investigator's findings I kept in this review

Priority 1 (inventor): US 7,830,065 B2 grant and its changes (soft pedal on "natural resources", "believed to", deleted "persists until magnets are no longer magnetized" and "reduction of required input power as loads are applied"; para 0044 and 0040 kept) [web: Google Patents]; ESTC 2015 "Overunity Disclosure" and "Floyd Sweet's VTA"; ESTC 2016 MIT talk (570% claim) [web: emediapress]; grahamgunderson.com/ou (2018) naming ambient energy [web: Wayback]. The deleted phrases are worth the reader's attention: the applicant removed the most explicit over-unity wording before grant, but kept 0044. That is an observation about prosecution, not evidence about physics.

Priority 2 (support): McKay and Reiyuki accounts; MPI 2008 report; Chava "100 times" statement (assignee, interested); Gunderson's undocumented account of verifying a Sweet-VTA-like EV charger over a week. No data, no replication.

Priority 3 (related work): Sweet's Vacuum Triode Amplifier; MEG US6362718 (a permanent-magnet flux-switching core device) [web: Google similar documents]; E.V. Gray and Russian TK/Kapanadze circuits (named by forum attendees); mainstream magnetostrictive harvesting citing US 8,093,869. The mainstream literature builds on the transduction, not the over-unity reading. Mainstream consensus labels the zero-point-energy framing "inconsistent with thermodynamics" [web: naturalphilosophy wiki, a secondary characterisation].

Priority 4 (historical record, absence of evidence): Chava announced further prototypes, a "very broad" patent application and 30 more patents in three years; no further published families, papers or independent tests exist for Gunderson or Chava; US 7,830,065 lapsed 2022. Competing explanations the Investigator listed: (a) effects did not survive validation, (b) lapsed for cost, (c) trade secrecy/NDA, (d) funding or internal conflict (a commenter says Goldes was "ejected"), (e) suppression (no evidence found). My physics view on (a): the claim 2 energy analysis above and the measurement-artifact list make (a) the explanation most compatible with known physics; it does not make (b)-(d) false, and the record that would decide it (USPTO file wrappers, Chava filings, Gunderson's 2019 video, an open battery/closed-loop test) was not reached. I do not weigh (e): there is no evidence either way. Single-source items (McKay, Reiyuki, the 2008 report) are evidence and are quoted above, not dismissed.

What would settle it

1. A closed-loop test: feed the output back to the input and remove the battery. Run for hours, with the same circuit.
2. A battery coulomb count or full discharge before/after, with the "zero input" condition.
3. Measure input at the DC battery terminals (energy meter), not only at the AC side; use a bandwidth far above 50 kHz and a phase-calibrated probe set; calorimetry for the output.
4. For the acoustic patent: a driven magnetostrictive bar with measured force and velocity at the source and electrical power in the load; plot Re(Z) of the source. The prediction here is Re(Z) >= 0 with efficiency below 100%.

Corrections to Step 1

• facts.json for US 2006/0163971 says "Whether the application was granted is not stated." The Investigator found it was granted as US 7,830,065 B2 (9 Nov 2010) and lapsed 9 Nov 2022; I verified the Google Patents page [web]. facts.json and claims.md should record this. Claim 2's description in claims.md quotes the pre-grant text; the granted text keeps para 0044 and 0040 but removes the "persists until demagnetized" line (para 0048) and the "reduction of required input power" line (para 0049) per the Investigator's comparison (I checked 0044 only; the deletions are the Investigator's [web]).
• claims.md and facts.json describe the Lenz-aiding assertions as claims of the patents. Strictly: the aiding statements are in the description, not the claims, in both patents; the claims are structural (the facts correctly note this). I keep the claim numbering of claims.md.
• facts.json for US 8,093,869 says "(Faraday phase for a resistive circuit is 90 deg from the flux)". That is right, and it is the key to refuting claim 3; I add that the patent itself says the phase depends on tuning, so a resonant tuned-inductive circuit can produce 180 degrees only with no net power transfer.
• claims.md claim 6 says the spacing "is what distinguishes the device"; the text actually says the spacing is "not less than 1/8" for coils in the claim 1 (inductive elements from each other, and the source from the nearest conductor). Step 1 is accurate; the OCR "%"/"4" glyphs for 1/8 and 1/4 should be read as fractions (page-image check done in the deep read).
• Quote check passed: paras 0039-0044 and claim 1 of 8,093,869 match the quotes in claims.md.
• Gap in Step 1's evidence note: "Neither contains experimental data" is correct for the patents, but the project record includes later non-patent claims (ESTC 2015-16, witness posts, MPI/Chava web claims) that the claims list does not mention; these are included above.

Verdict summary

Web checks used: US7830065B2 (Google Patents), Neodymium magnet (Wikipedia); the (BH)max numbers for NdFeB and ferrite remain [memory, high] with the ratio checked; Manley-Rowe remained [memory, high] and was not re-checked on the web (budget).