Arbiter: Haisch-Bernard (US 7,379,286 "Quantum Vacuum Energy Extraction"; Forward 1984)

Role: Step 3 arbiter. I read claims.md, facts.json, physicist.md, mathematician.md, contrarian.md, sites.md and the research notes, and recomputed where the reviewers differed. Tags: [read] project documents as quoted by reviewers (I did not re-read the patent pages myself, except where stated), [web] with URL, [memory] with confidence, [calc] my own arithmetic (Python, 2026-10-04).

No source gap: all three project documents have readable text (patent and Forward are OCR, equations checked by reviewers against page images).

Interested parties: Haisch and Moddel (inventors; assignee Jovion Corporation; Moddel declares a Jovion stockholding in the 2019 paper) describe their own scheme; their statements are their characterisation. Forward (1984) was funded by the US Air Force Rocket Propulsion Laboratory and Hughes [read: Forward p.3]; no stake in this patent.

The idea, briefly

Noble-gas atoms pushed through gaps about 0.1 micron wide (a "Casimir cavity") are said to drop 1-10 eV each on entry (heat), and to be re-energised from the zero-point field (ZPF: the vacuum's fluctuating electromagnetic field) on exit, so a gas flow becomes a heat engine fed by the vacuum. The basis is stochastic electrodynamics (SED), a classical theory in which the hydrogen electron is held in orbit by a balance of radiated and ZPF-absorbed power.

Recomputed numbers where reviewers differed or I relied on them

Web checks (budget 8; used 5)

• Nieuwenhuizen and Liska: abstract read; relativistic corrections "have little effect; the self-ionisation that occurs without them remains present" in SED hydrogen [web: https://arxiv.org/abs/1506.06787]. Retagged [web] (was medium).
• Cetto and de la Peña 1988 (PRA 37, 1952): abstract says vacuum "can be modified ... by the introduction of conducting objects that alter the distribution of modes" and that their SED calculations "coincide with previous quantum calculations" [web: https://journals.aps.org/pra/abstract/10.1103/PhysRevA.37.1952]. This is more informative than the reviewers' "not predicted under simplifying assumptions": SED itself, applied to environmental shifts, reproduces the ordinary QED-sized shifts. It is the inventors' own cited SED source, and it supports the ordinary small shift, not an eV shift. I read the abstract only.
• Polarizabilities of He/Ar/Xe: verified above (was medium).
• Plasma frequencies: Al about 15 eV, Au/Ag about 8-9 eV; above these metals become transparent [web search summary; e.g. https://arxiv.org/pdf/0908.1126]. The patent's 455 A (27 eV) modes lie above both, so a metal wall cannot impose a cutoff there. Retagged [web] (was memory, high/medium); the contrarian's point stands.
• Patent legal status: Google Patents (via fetch summary) lists US 7,379,286 B2 as "Active", expiring 2026-11-30, application US 11/236,142, inventors Haisch and Moddel, published 2008-05-27 [web: https://patents.google.com/patent/US7379286B2/en]. The summary named the assignee "University of Colorado Boulder and JOVIAN Corp" (probably a garble of Jovion; unverified). This conflicts with the reviewers' open question of a 2011-12 lapse: the fetched summary shows no lapse, but a summarizer may omit fee events. Status: unresolved, but "active, nominal term ends 2026-11-30" is the best current source. A patent kept in force ten-plus years through maintenance fees is a (weak) sign that someone has paid for it; it says nothing about whether it works.
• Not checked and still [memory]: Cole and Puthoff 1993 content (known only from Haisch's page via the Investigator); Lamoreaux 1997 / Mohideen-Roy 1998 (high); Henriques thesis (fetch failed for the contrarian; abstract only).

Corrected claim list

Claim 1: atoms lose 1-10 eV per transit of a ~0.1 micron cavity ("Casimir-Lamb shift")

Reviewers: Physicist: Refuted by calculation (strong form) and Tested, not supported. Mathematician: split: Established for tiny shifts, Refuted by calculation vs QED for eV scale, Internally inconsistent as written. Contrarian: Tested, not supported (strong); Never properly tested (weak, sub-eV). Agreement: all three agree that 1-10 eV is an assumption ("we assume", patent), not a derivation; that real cavity shifts exist but are 1e-9 to 1e-7 eV; that the 2012 signal was 1e-3 or lower of prediction with trends opposite to prediction. Disagreement: how to label it. The contrarian notes that QED does not refute SED, a different theory, and that the tests were weak instruments. Stronger argument: the physicist and mathematician, supported by two points the contrarian's framing underweights. (i) Geometry (hydrogen only; softened in the QE revision): for ideal mirrors a gap d removes modes longer than 2d, so the 45.6 nm hydrogen modes need d below ~22.8 nm; the patent uses a looser cutoff of lambda = d and pairs hydrogen with d = 250 A. The "stairway" and "tail to five Bohr radii" arguments are not computed. (ii) The inventors' own SED source (Cetto and de la Peña 1988) finds SED environmental shifts coincide with QED ones. So the claim is not rescued by SED as published. The contrarian is right that no neutral spectroscopic search exists, and that the 2012 experiment cannot cleanly separate a cavity effect from heating. Final verdict (split): - Strong form (1-10 eV per atom per transit at 0.1 micron): Refuted by calculation against accepted cavity QED (1e-9 to 1e-7 eV), and Tested, not supported (2012, 2019, Henriques 2014, Puthoff-Little-Ibison 2002; all non-neutral or weak instruments; none positive for the claimed size). The earlier label "internally inconsistent as written (mode-suppression direction)" is withdrawn; see QE revision, item 2 (the patent's wording is loose, its numbers are not shown inconsistent). - Weak form (a sub-eV or eV-fraction shift in some cavity geometry, or flow-driven IR emission): Never properly tested. Not dismissed without testing; tested by its proponents, who stopped. Corrections to Step 1: the 1-10 eV is an assumption; for hydrogen the patent itself specifies d = 250 A, close to the ~228 A the ideal 2d cutoff needs (QE revision, item 2); the 2012 paper itself says "power out was less than the power in" [read: 2012 p.13]; [WITHDRAWN: the helium "strongest signal vs unmeasurable" contradiction was a misreading, see QE revision item 1]; the paper's atom-must-collide-to-radiate hypothesis contradicts the patent's emit-on-entry mechanism.

Claim 2: re-absorption on exit; continuous net power "extracted locally, replenished globally"

Reviewers: Physicist: Refuted by calculation (conservative potential plus detailed balance); weak form Never properly tested. Mathematician: Internally inconsistent in counting and unproven; weak form Not testable as stated. Contrarian: Never properly tested; explicitly declines "Refuted by calculation" because the inventors treat it as open. Agreement: no net-power measurement exists; the only power balance (2012) has out < in; the inventors' own 2019 paper lists the "potential hill" and detailed-balance objections as unresolved ("we cannot totally rule out this potential hill argument"; "not clear whether equilibrium considerations render it incapable of providing harvestable energy" [read: Atoms 2019]). Both Haisch (2009: "not yet been able to prove anything in the laboratory") and Haisch's calphysics page ("no practical application" of Forward's scheme because it costs as much to re-separate plates) are inventor-side admissions kept in the record. Which is stronger: the physicist on logic, the contrarian on procedure. A closed loop in a position-dependent energy U(z) returns zero net work (the loop integral of the gradient vanishes), and thermal speeds are 540-3100 times the drift speed, so atoms cross and re-cross the stripe far faster than they flow; the gas equilibrates with the walls in nanoseconds [calc]. The mathematician's counting point adds that, if every boundary crossing released energy, thermal bouncing alone would give ~1000x the events with no flow at all. These are real calculations, but they assume the exit cost is paid by the same potential, which is exactly the inventors' denied step. The contrarian is right that the inventors' framework asserts a non-thermal ZPF reservoir (Cole and Puthoff), which I could not check. I side with the physicist for the strong form because it requires the extra hypothesis that exit re-excitation draws on the ZPF without any pump cost, with no derivation and with SED's own 1988 result (above) pointing the other way. But the labelling "Refuted by calculation" holds only conditional on claim 1's premise failing or on standard QED ground-state physics; I flag it as moderate confidence. Final verdict (split): - Strong form (net usable power from room-temperature gas cycling through a static cavity): Refuted by calculation (conservative potential, detailed balance, thermal-flux counting), moderate confidence; and the one power balance on record is Tested, not supported (out < in). - Weak form (a non-equilibrium flow effect that evades detailed balance): Never properly tested. Absence of evidence: see section below.

Claim 3: SED explains atomic stability; the hydrogen balance holds for all atoms

Reviewers: Physicist: Historically superseded but instructive; extension Not testable as stated. Mathematician: H model Plausible, untested; extension Never properly tested. Contrarian: H Tested, not supported (long simulations ionise); extension Never properly tested. Agreement: Larmor power at the Bohr orbit is 4.67e-8 W; the SED harmonic-oscillator derivation is a "weakness" per Haisch; the patent admits the all-atoms step is "not yet proven theoretically". Stronger argument: the physicist and contrarian, now backed by a web check: Nieuwenhuizen and Liska find self-ionisation persists with relativistic corrections [web: https://arxiv.org/abs/1506.06787], while Cole and Zou (2003-04) found a ground-state-like distribution for a limited time [read via patent; abstract at https://www.sciencedirect.com/science/article/abs/pii/S0375960103012696 per contrarian]. "Plausible, untested" understates it, because the simulations are tests of the model. "Tested, not supported" is too strong for SED as a whole: it is a numerical result about one model of hydrogen with a point electron, and the authors suggest the point-charge idealisation as the cause. Final verdict (split): - SED hydrogen ground state: Historically superseded but instructive (a legitimate classical programme that gets some real results but has a long-time stability problem even in hydrogen). - Extension to all atoms and to cavity-induced decay: Never properly tested (no derivation exists; the only indirect test is claim 1's experiment).

Claim 4: power estimates (21-210 W, 2150-21500 W, 1-10 W) and 1.7 mW pumping

Agreement: all three reproduce the arithmetic from the assumed 1-10 eV; all agree the pumping figure is inconsistent by ~1000. Resolution [calc and two independent reviewer derivations]: 17 atm is the plausible figure and "1700 Pa" is the error; the pumping power is ~0.3-1.7 W, not 1.7 mW. The comparison against the 2.2-22 kW cube still holds; against the 1-10 W bonded-wafer device (the preferred embodiment, with no worked calculation in the patent) the corrected pumping power is 3% to 170% of output, so net gain is not established by the patent's own numbers. Additional points: the cube estimate (10% open area, 0.1 micron) and the pumping check (0.25 porosity, 0.2 micron) are not the same device; Roy et al. dated 1993 in text, 2003 in references; water-bath "nearly all" absorbed is 63% at the stated lower coefficient; scaling pressure drop linearly in thickness ignores transitional flow. Final verdict: Pumping figure: Internally inconsistent. Output figures: arithmetic Established as arithmetic, but conditional on claim 1; as predictions of real output: Refuted by calculation to the extent claim 1 is (a 1e3-1e4 shortfall in the 2012 data would reduce the cube from kW to W or below); never built or measured.

Claim 5: Forward (1984) foil-stack "vacuum-fluctuation battery"

Agreement: all three: equations right (h, not hbar; coefficients 1440 and 480; Eq. 2 sign dropped; Eq. 3 prints 480 not OCR "980"); the Casimir force is established; the device is a store, not a source; never built. Notes kept: below a_min (lambda_min/2 pi, ~13 nm for aluminium with a plasma wavelength ~80 nm [memory, medium]) the ideal 1/a^4 law fails, so energies at 1-10 nm are overstated; real adhesion/surface energy (~1 J/m^2 [memory, medium]) may dominate; electrostatic suspension is unstable (Forward's own caveat). The patent's "simple, but not practical" description of Forward is fair only if read as "returns what is put in". Final verdict (split): Casimir force itself: Established (metals above ~100 nm; Lamoreaux 1997, Mohideen-Roy 1998 [memory, high]). Device as energy storage/conversion: Plausible, untested. As a net vacuum energy source over a cycle: Refuted by calculation (conservative force, zero net work) but Forward does not claim it. Values at 1-20 nm for metals: Never properly tested (as of 1984; unknown since).

What would settle it

1. Pore-size scan (0.05 to 10 micron) at fixed flow and pressure drop, with matched-impedance non-cavity controls, metered pump power, closed calorimetry, and a detector tested for pressure/acoustic pickup at the chopping frequency (settles claims 1 and 2: a cavity effect switches off for pores much larger than the supporting wavelength, a heating artifact does not). Contrarian and physicist both propose this.
2. UV-to-soft-X-ray spectroscopy of Xe or Ar in a precisely defined 25-250 nm slit made of a material reflecting at the relevant wavelengths: an eV shift would be glaring, the QED shift is ~1e-8 eV.
3. A derivation (not an assumption) of the energy an atom gains or loses entering a slit, including the exit cost, in SED and in QED.
4. Redo pumping power with slip/transitional flow and publish the corrected figure (resolves claim 4).
5. Read Cetto and de la Peña 1988 in full and state its assumptions against the patent's (claim 3).

Open questions

• [Withdrawn after QE: the "so small" helium sentence in the 2012 paper refers to the Mylar single-hole Joule-Thomson control, not to the nanopore membranes; see QE revision item 1.]
• Which polycarbonate pores, if any, are "cavities" (track-etched pores are non-uniform).
• Whether the 0.3% "qualifying fraction" implied by the milliwatt prediction has any physical basis.
• What Cole and Puthoff (1993) actually derive about extracting energy from a non-thermal ZPF (unread by all reviewers).
• Whether Kotov et al. (2025) and Moddel's 2021 MIM-device papers (Phys. Rev. Research 3, L022007; patent family US 11,133,758) support or contradict this patent (unread).
• Henriques's thesis: abstract says "some evidence" for Xe; Moddel and Dmitriyeva summarise it more negatively; the PDF has not been read by anyone.

Absence of evidence: the gas-flow line goes quiet after 2012

Facts: the Colorado group's last gas-flow experiment is 2012; the 2019 paper reuses 2010-12 data plus one 0.1-micron helium run; Jovion's site shows an empty "Progress" link (c) 2023; Hathaway Research International project page has no results; DARPA/SPAWAR grant N66001-06-1-2026, CU proof-of-concept funds, ~$200k private by 2009, $10M sought with no visible award, HUB Lab contract 2010-12 [web: Investigator notes in sites.md/research]. The patent appears still in force [web: Google Patents above]. Competing explanations: - (a) Effect did not hold up and was dropped. For: out < in; opposite trends; Haisch 2009 admission; 2019 closing sentence that fundamental constraints may preclude extraction; physics objections above. Against: nothing direct. Decide by: an internal lab report or the DARPA final report. - (b) Funding ran out. For: unmet $10M ask; modest conditional grants. Against: patent fees apparently paid. Decide by: DTIC final report, CU tech-transfer files. - (c) Trade secrecy. For: Jovion is a private company with a quiet website. Against: the 2012 and 2019 papers are public and candid about weak results; Moddel states he plays no active role in Jovion. No positive evidence beyond silence. Decide by: Jovion corporate filings. - (d) Work moved to the metal-insulator-metal (MIM) Casimir-cavity line (2019-21 papers, new patents, a projected 250 W per 10x10 cm array that is an estimate, not a measurement). For: documented in published papers and patents. Against: it is a different mechanism, so it does not validate this one. - (e) Ambiguous pyroelectric data (heating artefacts) made a cleaner follow-up impossible with the available budget. Plausible alongside (a) and (b). Not mutually exclusive; (a)+(b)+(d) together fit the record best, and no evidence of suppression or blocking of funding or publication was found. Records that would decide: USPTO maintenance-fee and assignment history, DTIC/DARPA final report, HRI report, Jovion filings, CU technology-transfer files.

Investigator findings kept in the record (whatever the verdict)

2012 and 2019 experiments with their numbers and candid caveats; Haisch's calphysics.org and 2009 interview caveats; Henriques (Coimbra 2014) "some evidence of radiation" for Xe and null with pre-excited Xe; Puthoff-Little-Ibison 2002 "so far without success"; Cetto and de la Peña 1988; Cole and Puthoff 1993; Moddel's MIM line with its projected 250 W; third-party patents citing Haisch-Moddel (US20190207537A1, US20180059704A1, US20230208321A1); a commenter's link to BlackLight Power "hydrino" claims (commenter's characterisation, not the inventors'); Kotov et al. 2025 (unread). Some of these are favourable to the inventors' candour and to the idea that cavity-modified ground states are a real research topic; the physics is just 7-10 orders smaller than claimed.

Headline verdicts

1. 1-10 eV per transit: Refuted by calculation (vs cavity QED) and Tested, not supported; weak sub-eV form Never properly tested.
2. Net power from gas cycling: Refuted by calculation (moderate confidence); weak form Never properly tested.
3. SED atom: Historically superseded but instructive; all-atoms extension Never properly tested.
4. Power estimates: Internally inconsistent (pumping, 1000x; 17 atm is right); outputs conditional on claim 1.
5. Forward battery: force Established; device Plausible, untested; net source Refuted by calculation (not claimed by Forward).

Changes after QE: see the table in "QE revision" below. Claim 1 loses the "internally inconsistent" sub-label; Claim 2 gains separate verdicts for the moving-wall and FIG. 8 variants; headline labels otherwise unchanged.

---

QE revision (2026-10-04)

Inputs: qe.md (verdict MINOR); the new deep reads in deep/Patent7379286-c0ceee.md (pp.10-12 cutoff convention, hydrogen d, Maclay, related patents; pp.19-21 claims and references; pp.17-19 later embodiments); the 2012 and 2019 papers in research/files/, which I checked directly. Tags: [read] deep read of the patent (page images); [read: paper] my own pdftotext check of the Investigator's copies; [calc] my arithmetic; [memory] with confidence. No new web checks (budget not needed; unchecked [memory] items are marked).

What changed, in one table

1. Withdrawal: the 2012 helium "contradiction"

I checked the Investigator's copy of the 2012 paper [read: paper, Dmitriyeva and Moddel, Physics Procedia 38, Joule-Thomson section]. The sentence "the signal from He was so small we were not able to obtain reliable measurements from this gas" belongs to a control experiment in which the filter was replaced by a 12.5 micron Mylar film with a single 3 mm hole; the lock-in phase there was "positive and 180 degrees off from the phase measured in the experiment with the nanopore filter, indicating cooling" [read: paper]. A small helium signal in a Joule-Thomson control is expected, because helium has the smallest Joule-Thomson coefficient (the paper quotes -0.0625 C/atm at 200 atm). So there is no internal contradiction. It is also mild evidence that the strong membrane signal is not Joule-Thomson cooling. The mathematician, the contrarian and my first pass were wrong to list it as a defect. The same paper's other controls stay in the record: power was linear in flow where simple friction predicts quadratic (the authors say this "does not automatically disqualify" friction because the chamber exchanged heat with the environment); the largest signal came from helium and the least from xenon, opposite to the friction ranking, which the authors say might be explained by gas heat conduction and diffusivity; their conclusion is that none of the explanations "clearly fit the data". That is a fair statement of an unresolved result, not a positive finding for a cavity effect.

2. Claim 1: what the patent actually says about the cutoff (deep read pp.10-12)

Quoted in the deep read [read: page images pp.10-11]: - The patent states modes "whose wavelengths are longer than the separation of the plates" are excluded, and that suppression "will be strongest for wavelengths of d or greater, but will begin to occur as well for wavelengths falling in between the 'stairway' d/n". So its cutoff is lambda = d. The textbook ideal-mirror cutoff is lambda = 2d [memory, high]. The patent does not say it is using a different convention and derives nothing. - For hydrogen it writes "a Casimir cavity of d=250 A or thereabouts", for 1-10 eV "we assume". - For noble gases it scales the SED wavelength as r^(3/2) (a "crude assumption") and says 0.1 micron "would have an effect" on the outermost s electrons.

My recomputation [calc], using the patent's own r values (He 1.2, Ne 1.3, Ar 1.6, Kr 1.8, Xe 2.05 A) and 455 A for r = 0.529 A: the relevant SED wavelengths come to about 155 nm (He), 175 nm (Ne), 240 nm (Ar), 286 nm (Kr), 347 nm (Xe). A 0.1 micron slit has the textbook cutoff at 200 nm and the patent's at 100 nm. So under the patent's convention all five gases lie in the suppressed range; under the textbook convention Ar, Kr and Xe do and He and Ne do not. For hydrogen, d = 250 A gives a textbook cutoff of 500 A, which is longer than 455 A, so hydrogen would need d below ~228 A; the patent's 250 A is within 10% of that.

Consequences, which correct what I wrote above: - My earlier geometry statement ("the 45.6 nm modes ... are not removed; a gap of ~23 nm would be needed") is true for hydrogen under the textbook convention, but it was presented as an objection to the patent when the patent had chosen d = 250 A for hydrogen. The "backwards" charge rested on the phrase "suppression ... at this wavelength and shorter", which is loosely worded; the patent's numbers remove the longer wavelengths. I withdraw "internally inconsistent" for this point. - My metal-transparency objection (plasma frequencies of Al about 15 eV, Au/Ag 8-9 eV, so 27 eV modes pass through) applies to hydrogen's 45.6 nm. It does not apply in the same way to the noble-gas wavelengths of ~150-350 nm (3.5-8 eV). There aluminium is a good reflector and the 40 nm platinum stripes of the preferred embodiment are partial reflectors [memory, medium; not web-checked]. So the objection is weaker for the actual working gases than I stated. - The verdict does not move, because the decisive point is size, not geometry. The patent never computes the shift. The ordinary atom-wall (Casimir-Polder) energy at 50 nm is 1e-9 to 3e-8 eV [calc, above], seven to ten orders below the assumed 1-10 eV, and the 1988 SED paper the inventors cite found SED environmental shifts coinciding with the quantum ones (abstract only; still an inference). The rest of the 7-10 order gap stays. - The patent's appeal to Jordan Maclay (a long cylindrical cavity gives an inward force, "this implies that a cylindrical cavity of diameter 0.1 micron would yield the desired decay") has no citation and is not in the 12-entry reference list [read: pp.11, 19]. It is the applicants' characterisation of a named theorist's work; what Maclay actually calculated is an open lead. The step from "inward force" to "electron decay" is the applicants' own. - The patent's "slow process" premise (atoms entering rapidly are still released gradually, so look for heat, not spectral lines [read: p.11]) explains why the 2012 experiment used a heat detector. It is also an untested assumption, and it is why a spectroscopic search (settling item 2 below) was never run by the inventors. - Related patents: the patent dismisses Shoulders (5,018,180), Mead and Nachamkin (5,590,031) and Pinto (6,477,028; 6,593,566; 6,665,167) with the single reason "does not deal with energy release from atoms in Casimir cavities". That is the applicants' characterisation, with a stake via Jovion; the Pinto patents propose to make a Casimir system "non-conservative", which is a different route to the same goal and is kept in the record.

3. Claim 2: the patent's own answer to the exit objection (FIG. 8, FIGS. 5-7)

What the patent says [read: pp.17-19]: - A general statement of the cycle: gas near a cavity has its orbital energy reduced, "the excess energy is emitted and absorbed by the apparatus"; on emerging, "the energy of the electron orbitals ... is again allowed to rise ... The compensation for the energy deficit is provided from the ambient electromagnetic modes." - A concession in the inventors' own words: if release and re-supply are "both local" or "both global", "no net energy is provided. To avoid these possibilities, we provide an asymmetry". No distance, time or power is attached to "local" or "global". - FIG. 8: an absorbing glass layer (84) lines the channel just upstream of the conducting cavity (86), on sapphire; nothing absorbs downstream (87). The absorber is placed "within roughly one emission wavelength" of the emitting atoms. The conducting length "is chosen so that the emerging atoms do not have substantial access to radiation emitted from the absorbing regions". No length, wavelength, absorber thickness, gas, flow rate or output is given; dimensions are "approximately the same" as FIGS. 4A-4D (200 nm gap, 2 micron stripes). The text calls the absorber "87" once by mistake (it is 84). - FIGS. 6-7: the gas does not move; the walls are stretched (piezo, photonic-crystal polymer with 200 nm gas gaps, silicon pillars) or rocked (electrostatically driven polysilicon torsion plate; "very little energy is required to maintain the oscillation"). No exit-specific mechanism; release and re-supply occur in the same place at alternate phases, which is the symmetric case the patent itself says may give no net energy. No dimensions for FIG. 7. No measurement for any of FIGS. 6-8.

My weighing: - Moving walls (FIGS. 5-7). The detailed-balance argument for flowing gas does not apply word for word, but a simpler one does. If an atom's energy E depends on wall spacing d, the wall feels F = -dE/dz and the actuator must pay on re-opening exactly what was released on closing; over a closed cycle of a single-valued E(d) the net work is zero. At the patent's own assumption (1 eV per atom, 1 atm gas, 293 K) the force term is the energy density n x E = 2.5e25 m^-3 x 1.6e-19 J = 4.0e6 Pa, about 40 atm, on the walls [calc; the same figure the deep reader obtained]. At the ordinary Casimir-Polder size (1e-8 eV) it is about 0.04 Pa. Either the shift is far smaller than assumed, or the walls feel a large extra pressure that the drive must overcome; "very little energy" is incompatible with the first. Verdict: Refuted by calculation (conservative force over a closed cycle); confidence higher than for the gas-flow case. This assumes the ground state follows the wall adiabatically; the inventors' premise of an extra ZPF reservoir is the same unresolved premise as before. - FIG. 8 asymmetric absorber. This is a better-engineered answer than the symmetric cases, and I record it as the inventors' strongest design response. It is the right idea for an energy-selective trap: put the sink where the atoms emit and none where they refill. It still does not escape the thermodynamic problem: (a) the absorber is a thermal emitter in the same infrared band it absorbs (Kirchhoff) [memory, high]; the conducting-length clause is the only safeguard and has no number; (b) the stated emission wavelengths do not agree: the inventors' 1-10 eV corresponds to 124 nm to 1.24 micron [calc: 1240 nm eV / E], where sapphire and silica glass are both largely transparent above ~0.2 micron [memory, medium], while the absorber choice rests on far-infrared glass absorption; (c) a 2 micron conductor stripe is much shorter than a far-infrared wavelength, so exiting atoms would sit within a wavelength of the absorber they just passed, contradicting "no substantial access" at the scale of FIGS. 4A-4D; (d) if the "global" supply is ordinary 300 K thermal radiation from the surroundings, the device is a loop ambient -> atoms -> absorber -> ambient with a single temperature, and work extractable from the absorber's small temperature rise is limited by the Carnot factor [calc by argument: dT/T of order 1e-3 or less]. So FIG. 8 gets Never properly tested (no FIG. 8 device was built by anyone on record; the 2012 experiment used track-etched polycarbonate filters, not this design) and, under thermal-equilibrium physics, Refuted by calculation. It survives only as the hypothesis that the ZPF is a non-thermal reservoir, which rests on Cole and Puthoff 1993, which no reviewer has read. - Net gain is not claimed. Reading the independent claims word for word [read: pp.19-21], no claim contains "net", "gain" or an output figure. Claims 1 and 19 require a fluid that releases energy on entering a Casimir cavity, a collector, and re-intake of energy on leaving; claims 15, 16, 18 and 20 add "relative movement between the cavity and fluid". Claim 20 is the broadest: it replaces the cavity with any "mechanism" that changes "the atoms and molecules" so as to release energy. The asymmetry of FIG. 8 appears in no claim. Claim 15 refers to "said means" and "said absorber" with no antecedent (printed that way). So the legal scope is "release then re-intake", which is weaker than the title and summary language ("from replacements for small batteries to power plant sized generators"). The distinction matters for the reader: the physics verdicts above apply to the title claim of a continuing net supply, not to the narrower wording that was granted. - The all-atoms and molecular-bond extensions are labelled by the patent itself as "not yet proven theoretically" (atoms) and "believed" (bonds), with liquids (ethylene glycol up to ~100 C, sodium) offered as working fluids [read: pp.12, 17]. They stay at Never properly tested.

Final verdict on claim 2 after revision: - Strong form, flowing gas (symmetric or FIG. 8): Refuted by calculation under thermal-equilibrium physics (moderate confidence, unchanged); the one power balance on record is Tested, not supported (out < in); FIG. 8 itself Never properly tested. - Moving-wall forms (FIGS. 5-7): Refuted by calculation (conservative force; higher confidence). - Weak form (a non-equilibrium flow effect evading detailed balance via a non-thermal ZPF): Never properly tested (unchanged).

4. Other corrections

• Cetto and de la Peña 1988: the claim that "SED itself reproduces ordinary QED-sized shifts" is an inference from the abstract ("coincide with previous quantum calculations"). Moddel and Dmitriyeva's own wording is narrower ("under particular simplifying assumptions this shift is not predicted"). Both statements stay in the record; the paper has not been read.
• Roy et al.: the reference list gives J. Appl. Phys. 93(8), 15 Apr 2003, pp.4870-4879; the body says 1993 [read: pp.15, 19]. The 2003 date is right (volume 93 [memory, high]).
• Reference list: 12 entries; Maclay and Chan et al. (Bell Labs MEMS Casimir, 2001) are cited in the body but not in the list [read: p.19, p.17].
• Digest errors recorded by QE (r^(1/2) for r^(3/2); 1.3e10 for 1.3e20; "Jovian") were corrected by the reviewers and the deep reads; no verdict used them.

5. Added to the kept record (qe.md retention defects)

• 2019 Fig. 9 [read: paper, Atoms 7, 51]: for helium through 0.1, 0.2 and 0.4 micron pores "the highest signal was registered from the 0.1 micron filter". This is the one trend in the direction the cavity hypothesis predicts. The authors give a second reading (more frictional heating at the larger pressure drop through the narrowest pores). Neither reading was tested with a matched-impedance control; item 1 of "What would settle it" does exactly that.
• 2012 controls (above): Mylar-hole control gave cooling (opposite phase); linear-in-flow power; helium largest and xenon smallest, opposite to the friction ranking; the authors' "none ... clearly fit the data".
• Moddel et al. 2021 (Phys. Rev. Research 3, L022007, via the Investigator [web: https://journals.aps.org/prresearch/abstract/10.1103/PhysRevResearch.3.L022007]): observed conductance increases with thinner Casimir cavities and "anomalous offsets in the current and voltage" in metal-insulator-metal devices. Different mechanism from this patent; a measured anomaly by the co-inventor, kept in the record. I have not re-read it.
• Historical record items: the 2009 LANL "New Frontiers in Casimir Force Control" workshop talk (abstract book is in research/files/, image-only, unread); patent priority 2004-09-27; the STAIF-2006 review and the Davis-Puthoff AIAA chapter (Haisch co-author); Dmitriyeva's parallel LENR thesis work and Coolescence funding.

6. Still open after QE

• Cole and Puthoff 1993 (unread; the claim-2 non-thermal-reservoir premise turns on it); Cetto and de la Peña 1988 full text; Henriques 2014 thesis; what Maclay calculated for cylindrical cavities; USPTO maintenance-fee and assignment history; the DTIC final report for N66001-06-1-2026.
• Metal reflectance at 150-350 nm for the actual stripe materials (Pt, Al) against the noble-gas wavelengths computed above [memory, medium; unchecked].
• An experiment on FIG. 8 itself. Nobody on record built it.