Mathematician review: Haisch-Bernard (Step 2)

Reviewer role: PhD mathematician. I check derivations, notation, units, counting arguments and hidden assumptions. I leave the verdict on mainstream physics to the physicist, but I run orders of magnitude where the mathematics needs them.

Source tags: [read] the project documents (including page images I rendered myself); [web] with URL; [memory] with confidence. Interested parties: Haisch and Moddel are the patent inventors, Jovion Corporation is the assignee, and Moddel states a Jovion stock holding in the 2019 paper [web: scholar.colorado.edu/downloads/6395w823g]. Their numbers are the applicants' own characterisation. Forward (1984) was funded by the US Air Force Rocket Propulsion Laboratory and Hughes [read: Forward1984 p.3].

All calculations below were run by me (Python, SymPy and NumPy from the project venv, plus plain NumPy constants). Nothing here is quoted from the sources unless it is in quotation marks.

No source gap: all three project documents have readable text. The patent and Forward are OCR, and I checked the load-bearing passages against page images.

Notation and plain-language setup

• Larmor radiation: an accelerated charge radiates power P = e²a²/(6π ε0 c³), where e is the charge, a the acceleration, ε0 the vacuum permittivity and c the speed of light.
• Zero-point field (ZPF): in stochastic electrodynamics (SED), a classical random radiation field with ½ħω of energy per mode at every frequency ω, where ħ is the reduced Planck constant. SED treats it as a real classical field.
• Casimir cavity: two conducting surfaces a distance d apart. An ideal parallel-plate gap has a lowest allowed wavelength cutoff of 2d (modes with longer wavelengths cannot fit).
• Bohr radius a0 = 0.529 Å; 1 Å = 0.1 nm.

Reading of the project, in one paragraph

The patent says: SED explains why the hydrogen electron does not spiral into the nucleus (radiation lost to Larmor emission is balanced by ZPF absorption). If a Casimir cavity removes the ZPF modes that sustain the orbit, the electron should drop to a lower state and release 1 to 10 eV as heat. When the atom leaves the cavity it absorbs the same energy back from the ambient ZPF. Pump gas through a mesh of such cavities and the heat is "extracted locally, replenished globally". Forward's 1984 paper is a separate, older proposal: a stack of charged foils, collapsing under the Casimir force, doing electrical work like a battery.

---

Claim 1: atoms in a Casimir cavity lose 1 to 10 eV per transit by a "Casimir-Lamb shift"

a. Is the mathematics internally consistent?

Part of it is, and part is not.

What checks out. - Orbit speed v = αc (α = 1/137.036, the fine-structure constant) gives orbit frequency f = v/(2π a0) = 6.58 × 10^15 Hz. The patent prints "6.6×10^15 s^-1" [read: page image, patent p.4 col.] and this is right. - The patent's wavelength "2π0.529137 = 455 Å (0.0455 microns)" [read: page image] equals c/f = 45.6 nm. The OCR text layer shows "22", but the page image shows "2π*". Done with 2×2 instead of 2π one would get 29 nm, so the OCR is an error and the patent is right. - The patent's angular momentum "m c r/137 = h/2π" is m v a0 = ħ. I get 1.0000 ħ. Correct.

What does not check out. 1. The direction of the mode suppression is backwards for the stated purpose. The patent says "suppression of zero-point radiation at this wavelength and shorter in a Casimir cavity will result in the decay" [read: page image, patent p.4]. A parallel-plate gap removes wavelengths longer than 2d, not shorter. At d = 0.1 µm the cutoff is 200 nm, so the 45.6 nm modes the orbit needs are not suppressed. To cut off 45.6 nm one needs d ≈ 22.8 nm. The patent then argues that the "stairway" of modes at 2d/n gives partial suppression at shorter wavelengths, and that the electron's probability "tail" extends to 5 Bohr radii so longer wavelengths (0.1 to 0.2 µm) matter [read: patent p.3-4]. Neither step is derived. The stairway modification of the mode density at λ < 2d oscillates around zero and its relative size falls as 1/n; the patent gives no calculation of what this does to an atom's absorption. 2. Scaling to noble gases is a crude assumption the patent itself labels as such. It assumes an outermost electron fully shielded, orbital velocity ∝ r^(-1/2), so λ ∝ r^(3/2) [read: patent p.4]. With the patent's own radii I get the supporting wavelength: He 155 nm, Ne 175 nm, Ar 239 nm, Kr 286 nm, Xe 347 nm. So the 0.1 µm cavity (cutoff 200 nm) can only match Ar, Kr and Xe if the scaling holds. Note that these atoms have 10 to 54 electrons, and the patent says the balance is "not yet proven theoretically" for any atom other than hydrogen [read: patent p.4-5]. 3. No derivation of "1 to 10 eV". The patent argues "Since the binding energy of the electron is 13.6 eV, we assume ... on the order of 1 to 10 eV" [read: page image, patent p.4]. This is an assumption stated as a fraction (7% to 74%) of the hydrogen binding energy. The same assumption appears in the 2012 experiment as "10% of its energy, approximately 1 eV" [read: research/files Physics Procedia 38 p.16, via Investigator]. 4. The Larmor collapse time is misquoted. The patent says a classical atom spirals in "on time scales less than one-trillionth of a second" [read: patent p.3]. The standard result t = a0³/(4 r_e² c) (r_e is the classical electron radius) gives 1.56 × 10^-11 s, about 16 times longer than 10^-12 s. This is minor and does not change the argument; it is a factual slip.

Worked comparison: what ordinary quantum electrodynamics (QED) says an atom gains or loses near a wall

An atom near a conducting wall has its ground-state energy shifted by the Casimir-Polder interaction (atom-surface van der Waals force, with retardation at larger distances). This is the mainstream version of a "cavity-induced level shift".

For argon, with static polarizability volume 1.64 × 10^-30 m³ [memory, medium confidence] and a characteristic energy ħω0 ≈ 15.8 eV (its ionization energy), I get: - Non-retarded estimate U ≈ α_vol ħω0 / (8 z³), at distance z from a perfect mirror: z = 25 nm gives 2.1 × 10^-7 eV; z = 50 nm gives 2.6 × 10^-8 eV; z = 100 nm gives 3.2 × 10^-9 eV. - Retarded formula U = 3ħcα/(32π² ε0 z⁴): z = 50 nm gives 6 × 10^-9 eV.

Even allowing a factor of a few for two plates, a mainstream shift in a 0.1 µm gap is around 10^-8 eV. The patent assumes 1 to 10 eV, which is 8 to 9 orders of magnitude larger. For scale, the hydrogen 1S Lamb shift itself is about 8.17 GHz, or 3.4 × 10^-5 eV [memory, high confidence]. The patent's "shift comparable to but larger than the Lamb shift" is, as stated numerically, about 10^5 to 10^6 times the real Lamb shift, not "larger" in the sense used for a cavity modification. This is a mainstream benchmark, not a refutation of the SED idea, since SED is a different theory; but the patent presents the Lamb shift as support for its size.

A further point on the metal: a real metal is a good mirror only for wavelengths well above its plasma wavelength (the wavelength where the metal becomes transparent). A search result summarises that silver has a plasma wavelength near 130 nm and that metals become transparent in the ultraviolet [web: https://arxiv.org/pdf/0706.1184 and search summary]. The patent's supporting wavelengths of 45 to 350 nm lie at or below this for platinum (the patent's coating, 40 nm [read: patent p.15 via deep read]); I did not find a platinum value [memory, low confidence that it is similar]. The cutoff of 45.6 nm cannot be imposed by a metal gap at all. The 2012 paper itself says the coating should be "highly reflective in the near-ultraviolet" and that near-UV equals twice the cavity spacing [read: Physics Procedia 38 p.10], which reads as a cutoff of 0.2 to 0.8 µm for its 0.1 to 0.4 µm pores, nowhere near 45 nm.

b. Consistent with established experiments?

Cavity-QED level shifts are measured, for example atom-wall (Casimir-Polder) shifts and cavity-modified Lamb shifts [memory, medium confidence]. They are of the tiny size computed above. No established experiment shows eV-scale cavity shifts.

c. Tested?

Yes, once, by the co-inventor's own lab (an interested party). Dmitriyeva and Moddel, Physics Procedia 38, 8-17 (2012) [web: https://www.colorado.edu/faculty/moddel/media/59; copy in research/files/]. A pyroelectric detector saw an infrared signal for He, Ar, N2 and Xe through 0.1 to 0.4 µm polycarbonate pores. The authors write: "Emission in the infrared was clearly observed" and "ZP energy extraction from the quantum vacuum remains a possible explanation". Facts that bear on the claim, all from that paper [read: research/files copy, which I checked]: - Predicted trend: strongest for Xe with gold coating. Observed: strongest for He with uncoated membrane. The authors write "The results we obtained directly contradict this." - Measured radiated power about 10^-7 W at the detector; electrical power in (flow times pressure drop) about 2 × 10^-5 W at 5 sccm and 4 torr. Their own words: "it appears that the power out was less than the power in and there is no need to invoke an additional power source". - Their ZPE estimate is "on the order of milliwatts, which is thousands times larger than the measured power". The size of the shortfall is therefore 3 to 4 orders of magnitude. - The Joule-Thomson section says "the signal from He was so small we were not able to obtain reliable measurements from this gas" while the results section reports He as the strongest emitter. I flag this as an unresolved tension in the paper as I read it. - Reasons the authors give for the shortfall: few pores satisfy the cavity condition; "Atoms may need to undergo a collision in order to radiate power"; emission may fall outside the detector band (0.6 to 5 µm). The collision requirement contradicts the patent's mechanism, in which atoms radiate on entering. - The 2019 paper (Atoms 7, 51) says results are "tantalizing but, given the lower than expected power output, inconclusive" [web: https://scholar.colorado.edu/downloads/6395w823g]. - Henriques (2014, Coimbra): independent SED simulation plus a nanopore experiment with UV/visible spectroscopy. The Investigator reports "some evidence of radiation" for Xe and nothing above noise with pre-excited Xe, with a detector judged too insensitive [web: https://estudogeral.uc.pt/handle/10316/28082]. I have not read the thesis. - Puthoff, Little and Ibison (2002) tested hydrogen gas through a 1 µm Casimir cavity "so far without success" [web: arXiv:1012.5264 via Investigator]. - No replication by an independent neutral group of the Colorado experiment was found, and the gas-flow work goes quiet after 2012 (see Absence of evidence below).

d. What would have to change?

Quantum electrodynamics would need atoms to have eV-scale cavity-dependent binding energies at 100 nm gaps, contradicting measured Casimir-Polder shifts by about 10^8. Within SED, the theory's current status as a full atomic theory would need to be extended to many-electron atoms (see Claim 3).

Verdict (split)

• Qualitative idea that cavities shift atomic levels: Established (as the tiny Casimir-Polder / cavity-QED shifts).
• Claimed eV-scale shift per transit at 0.1 µm: Refuted by calculation against standard QED shifts (10^-8 eV vs 1 to 10 eV), which however does not test SED's different premise. Within SED: the stated mode-suppression direction is inconsistent with the cavity geometry, so internally inconsistent as written.
• Experimental test: Tested, not supported (one test, by a co-inventor, with trends opposite to prediction, output 3 to 4 orders below prediction, and described by the authors as inconclusive; not a neutral test).

---

Claim 2: on leaving the cavity the atom re-absorbs the energy from the vacuum, so cycling delivers usable power "without violating thermodynamics"

a. Internal consistency

The energy bookkeeping is stated but not derived.

1. Conservative-potential argument. A ground-state energy that depends on position is a potential energy U(x), and the force on the atom is -∇U. Around any closed loop (the atom returns to its start), ∮ ∇U · dx = 0. If entering lowers the energy by ΔE, the atom gains kinetic energy ΔE on entering, and must lose ΔE climbing out. The 2019 paper acknowledges this as the "potential hill" objection and concludes "we cannot totally rule out this potential hill argument" [web: scholar.colorado.edu/downloads/6395w823g, p.14]. The patent's cycle needs the ΔE to leave as radiation (not as kinetic energy) on entry and to enter from the vacuum (not from the pump) on exit. Nothing in the patent proves this can happen.
2. Detailed balance. The 2019 paper: "If the gas enters the Casimir cavity on a time scale that is much slower than the equilibration time, then the entire process occurs in quasi-equilibrium and detailed balance applies ... there would be no excess energy to shed", and it concludes it "is not clear whether equilibrium considerations render it incapable of providing harvestable energy" [web, same, p.14-15]. This is the inventors' own statement of an unresolved mathematical gap, and I keep it as pertinent.
3. The transition-counting argument is inconsistent with the thermal motion. The patent counts one "transition" per boundary crossing and sets the rate by the drift speed (10 cm/s). I reproduced its arithmetic exactly: the gap of 0.1 µm × 10 cm width × 10 cm/s = 10^-3 cm³/s; at STP (standard temperature and pressure) there are 2.69 × 10^25 atoms/m³, giving 2.7 × 10^16 atoms/s; times 5000 strips gives 1.34 × 10^20 transitions/s, matching the printed 1.3 × 10^20. But gas atoms move thermally at around 400 m/s (argon mean speed 394 m/s at 293 K; helium 1245 m/s). The one-way kinetic flux through a plane is n v̄/4, compared with the net drift flux n u. For argon with u = 0.1 m/s this ratio is about 1000. If each boundary entry releases energy, the thermal bouncing alone would give about a thousand times more events than the flow, and would do so with no flow at all. If instead energy release needs net flow, the "release on entry, re-absorb on exit" picture is not the mechanism. The 2012 paper meets the same problem: it assumes "if the atom that went into the pore ... bounced back and reabsorbed the radiation it had just shed, we need to account only for the signal from the atoms that went all the way through" [read: research/files copy]. That is an additional unstated hypothesis the patent does not contain.
4. Second-law argument. The patent cites Cole and Puthoff (1993) and calls Forward's scheme "simple, but not practical" [read: patent p.3]. It claims the process uses "energy extracted locally and replenished globally". This is a statement, not a derivation. A cyclic engine that takes heat from one reservoir with no cold sink would violate the Kelvin-Planck form of the second law unless the ZPF is treated as a non-thermal reservoir distinct from the matter. Cole and Puthoff's own analysis, as I recall it, allows extraction only from a ZPF that is out of equilibrium with the matter [memory, medium confidence]; I did not read it.

b. Established experiments

No experiment shows net energy gain from a Casimir cavity cycle. The Casimir force is itself measured (for example Lamoreaux 1997 and later work) [memory, high confidence].

c. Tested

Same tests as Claim 1. The one measurement that bears on net energy found "power out was less than the power in" by the authors' own balance [read: research/files copy]. The Investigator also found that Haisch said in 2009 they had "not yet been able to prove anything in the laboratory" and that sporadic signals "can't be ruled out as experimental error" [web: https://www.nextbigfuture.com/2009/02/jovion-corporation-gets-patent-for-zero.html], and on calphysics.org Haisch calls the Casimir-ZPE attribution "debatable" and the SED harmonic-oscillator basis "a weakness" [web: http://www.calphysics.org/zpe.html]. Both are statements by an interested party, and are the inventor's own caveats; I keep them.

d. What would have to change

Either the vacuum field would have to be a depletable source that atoms can draw down, which makes it a non-equilibrium reservoir in conflict with the standard picture of the vacuum as a lowest-energy state; or the cycle would need a loss mechanism making ∮ ∇U · dx nonzero for a gas in contact with a thermal environment.

Verdict (split)

• Strong form (net usable power drawn from the vacuum by the cycle): Never properly tested in a neutral, controlled way; the one co-inventor test measured less out than in. Mathematically the cycle is not closed (potential hill, detailed balance, transition counting). I rate it Internally inconsistent as presented (transition counting) and unproven; no neutral experiment exists.
• Weak form (the inventors' own modest statement that the scheme "does not appear to violate" thermodynamics): Not testable as stated until the potential-hill and detailed-balance objections are resolved.

---

Claim 3: SED underlies the hydrogen ground state, and the same balance holds for all atoms

a. Internal consistency

1. The circular-orbit balance is a consistent calculation at one point. Using the Bohr orbit, the Larmor power is P = e²a²/(6π ε0 c³) with a = v²/a0, which I compute as 4.67 × 10^-8 W. Dividing the 13.6 eV binding energy by this gives 4.7 × 10^-11 s for energy loss, which is a classical decay time scale. The ZPF absorption side must equal this at the Bohr radius. Puthoff (1987) is cited for this [read: patent p.3]; I did not read it and recall that the balance reduces to an angular-momentum condition m ω r² ≈ ħ [memory, medium confidence]. With the Coulomb force, the condition L = ħ gives exactly r = a0/Z. I verified L = m v a0 = 1.0000 ħ.
2. Haisch's own caveats. He calls the harmonic-oscillator basis of the absorption calculation "a weakness" (calphysics.org) [web, as above]. The absorption of a ZPF by an orbiting, non-harmonic system is the exact point where numerical factors enter (the result L = ħ depends on a factor that is exact for a harmonic oscillator, not for a Coulomb orbit). This is why Cole and Zou's simulations (2003 and 2004) matter. The patent reports they reproduce the Schrödinger probability density [read: patent p.3]. I did not read them.
3. Angular momentum. The patent notes that the SED orbit has instantaneous angular momentum ħ while the quantum 1s state has zero. It then cites Nickisch simulations that the time-averaged angular momentum is zero because of ZPF perturbations on the orbital plane [read: patent p.3]. It is a statement about simulation, not a derivation.
4. Extension to all atoms. The patent says it "has not yet been proven theoretically" for any atom other than hydrogen [read: patent p.4-5]. The extension to many-electron atoms is therefore an assumption, and the claim's patent protection depends on it, since the patent uses noble gases to avoid dissociation.
5. A counter-lead from inside SED. Cetto and de la Peña analysed environmental effects on the Lamb shift according to SED (Phys. Rev. A 37, 1952 (1988), with a companion conference version in Physica Scripta T21, 27 (1988)) [web: https://journals.aps.org/pra/abstract/10.1103/PhysRevA.37.1952; https://www.osti.gov/etdeweb/biblio/7187101]. Moddel and Dmitriyeva cite it as showing that "under particular simplifying assumptions this shift is not predicted" [web: scholar.colorado.edu/downloads/6395w823g, p.9]. I did not read the full paper, only the abstract-level search results, which say it studies how conducting objects that alter the mode distribution change radiative shifts. So the SED literature itself contains an analysis that, under its assumptions, does not support the patent's central effect. I keep this as pertinent.

b. Established experiments

Quantum mechanics reproduces hydrogen's spectrum, fine structure and Lamb shift to high precision [memory, high confidence]. The SED hydrogen ground state is a simulation result, not a derived spectrum. I did not review SED's record on excited states or other atoms and leave it to the physicist.

c. Tested

SED's atoms have been simulated, not experimentally distinguished from quantum mechanics in this setting. The cavity test is Claim 1's test.

d. What would have to change

If the SED balance were the true origin of atomic stability, then atomic ground-state energies would depend on the ambient radiation environment in a way quantum mechanics does not predict. Quantum electrodynamics and quantum mechanics would need to be understood as effective descriptions of a classical-field theory.

Verdict (split)

• Hydrogen circular-orbit balance as a model: Plausible, untested (as a mathematical model it is consistent at that single point; it is an alternative reading of a result quantum mechanics already gets, and SED is contested).
• Extension to all atoms and the cavity-induced decay: Never properly tested (the patent itself says unproven; an SED paper from 1988 does not predict the shift under its assumptions).

---

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

a. Internal consistency: what I recomputed

The arithmetic from assumptions to the quoted watts is correct throughout. The power numbers are therefore only as good as the inputs: the 1 to 10 eV per transition and the counting rule in Claim 2 item 3.

The 1700 Pa versus 17 atmospheres inconsistency is real and the deep read is right. I tried to decide which figure is correct with an independent estimate. In free-molecular (Knudsen) flow in a long tube of radius r, the diffusivity is D_K = (2r/3) v̄ (v̄ is the mean molecular speed), and the pressure rise required for a throughput is Δn = n u L / D_K, where u is the in-pore flow speed at STP density, L the length. For the patent's geometry (200 nm pores, 1 cm thick, 1 cm³/s over 1 cm², porosity 0.25 so in-pore speed 4 cm/s) I get for argon 15.2 STP densities of excess, i.e. about 16 atm; helium about 5 atm; xenon about 30 atm. Ignoring porosity (as the patent does) argon would give about 4 atm. So 17 atm is plausible in size, and 1700 Pa (0.017 atm) is wrong by a factor near 1000. The printed 1.7 mW would then be about 1.7 W. A better estimate of work per second for isothermal compression from 1 to 17 atm is p0 Q ln(17) = 0.29 W. Either way the pumping power is about 0.3 to 1.7 W, not 1.7 mW. These are rough Knudsen-flow estimates (+- a factor of a few, since mean free path is comparable to pore size), not exact.

Consequence. The patent compares the pumping power with the 2.2 to 22 kW tunnel-block estimate, and that comparison survives the 1000-fold correction (1.7 W versus kilowatts). But the preferred embodiment, the 1 cm² bonded-wafer device, is rated "between 1 and 10 watts for an input pressure of 8 atmospheres" [read: page image, patent p.15 via deep read] and no flux or pressure drop is given for it. The corrected pumping power (0.3 to 1.7 W) would be 3% to 170% of that output. So for the device the patent actually describes, the net power gain is not established by its own numbers, and may be negative. The deep read reached the same observation.

For the tunnel block, the same Knudsen estimate for 100 nm tunnels, 1 cm long, at 10 cm/s gives about 82 atm for argon (26 atm for helium), so a pressure drop of about 80 atm; with the volume flow of 1 cm³/s (STP) that is about 8 W. That is still far below kilowatts. This follows the patent's own premise of 10 cm/s through 1 cm of 100 nm tunnels, which the patent's pumping check does not use (it uses 200 nm pores).

Other issues: - Roy et al. is cited as 1993 in the body and 2003 in the references [read: patent p.2 and p.15 via deep read]. Pore size in the Roy figure is not stated in the patent; pressure drop scales strongly with pore size (D_K ∝ r), so scaling a 60 µm result linearly to a 1 cm thickness without matching pore diameter is unjustified. - The 2012 paper's 'power in' is "~2 × 10^-5 W" for 5 sccm and 4 torr. The simple product 5 sccm × 4 torr is 4.4 × 10^-5 W. I cannot reproduce the factor of 2 from the paper's text. Minor. - The 2012 paper's "milliwatts" expectation: 5 sccm is 2.2 × 10^18 atoms/s; at 1 eV per atom that is 0.36 W, not milliwatts. The paper reaches milliwatts by an unstated reduction (modulated component of the flow, fraction of qualifying pores). I could not derive the milliwatt figure from the text. - Water-bath absorption: at the stated lower bound α = 10 cm^-1 over 1 mm, αx = 1 and absorption is 1 - e^-1 = 63%, not "nearly all" (the Tunnels wording); the bonded-wafer wording "a large proportion" fits [read: patent via deep read]. Liquid water absorbs much more strongly at most mid-infrared wavelengths [memory, medium confidence], so this is cosmetic.

b. Established experiments

None report net power output. The 2012 measurement found less output power than input.

c. Tested

See Claim 1. No device has been built to the patent's figures, as far as the Investigator found. The 2019 paper reuses 2010 to 2012 data. Moddel's later lab work moved to metal-insulator-metal (MIM) Casimir-cavity devices (Phys. Rev. Research 3, L022007 (2021); Symmetry 13, 517 (2021)), with a separate patent family (US 11,133,758 B2) and a projected, not measured, "up to 250 W" for a 10 cm × 10 cm array [web: per Investigator, sites.md]. That is a different mechanism and is not evidence for this patent's mechanism.

d. What would have to change

The same as Claims 1 and 2.

Verdict

• Arithmetic from stated assumptions to watts: Established (it reproduces).
• The pumping calculation as printed: Internally inconsistent (1700 Pa vs 17 atm, 10^3).
• Net power at the quoted levels for the 1 cm² device: Never properly tested; not supported by its own corrected pumping figure; output magnitudes rest on the unproven 1 to 10 eV per transition.

---

Claim 5: Forward (1984), a "vacuum-fluctuation battery" of charged foil leaves

a. Internal consistency

I checked the paper's equations against the deep read and recomputed them. - Eq. (1) u = -(π/1440) hc/a³ and Eq. (2) P = (π/480) hc/a⁴ with h (not ħ): π h/1440 equals π² ħ/720 exactly, the standard Casimir energy per area, and P = π² ħ c/(240 a⁴) is the standard Casimir pressure. SymPy gives -du/da = -π h c/(480 a⁴) with a minus sign for the attractive force; the printed Eq. (2) lacks the minus sign (sign slip, as the deep read said). The magnitude is right. - Eq. (3), the Lifshitz dielectric force, prints a coefficient of 480 (OCR gave "980", an OCR error). In the dilute limit the Lifshitz result 23ħc(ε-1)²/(640π²a⁴), divided by Forward's prefactor, gives 23 × 960/(640 π⁴) = 0.354, matching his f = 0.35. His phrase "when e < 4" is loose; 0.35 is the ε → 1 limit. - My table for ideal metals (energy per area |u| and pressure P) matches the deep read: 1 µm gives P = 1.3 × 10^-3 Pa; 100 nm gives 13 Pa; 20 nm gives 8.1 × 10^3 Pa; 10 nm gives 1.3 × 10^5 Pa (about 1 atm); 1 nm gives 1.3 × 10^9 Pa. Energy per area at 1 nm is 0.43 J/m².

What the argument does and does not claim. Forward says there is "no rigorous proof known that the vacuum-fluctuation field is a conservative field", but that it is "highly probable" or infinite energy could be drawn [read: Forward p.2]. His hydroelectric-dam analogy treats the device as a converter and store, with the energy put in when the foliated state is prepared. That is mathematically the same as a spring: a conservative force F(a) = -dU/da gives zero net work around any closed expand-and-collapse cycle. Recharging costs the same energy as extracted, ideally. He gives no net-energy figure, and no power or efficiency estimate [read: deep read].

Gaps a mathematician should note. - The energy available comes almost entirely from the last few nanometres (|u| ∝ a^-3). Forward's own Eq. (4) says the 1/a⁴ law holds only down to a_min = λ_min/2π, below which the weaker, material-dependent law of Eq. (5) applies. Using ideal-conductor values below that overstates the energy. For aluminium with a plasma wavelength of about 80 nm [memory, medium confidence], a_min is about 13 nm, so the 0.43 J/m² at 1 nm is an upper bound that the theory itself does not support. - At spacings of 1 to 10 nm the available Casimir energy (0.4 mJ/m² to 0.4 J/m²) is smaller than or comparable to the ordinary surface energy of a metal, roughly 1 J/m² per surface [memory, medium confidence]. When the foils actually cohere ("cohering into a solid block"), the dominant energy released is the chemical-scale adhesion energy, not the Casimir term. That is a one-shot gain from a prepared high-energy state (foil), as in his own caveat that someone has to prepare the foliated state. - For ultra-thin foil of 10 nm thickness (my assumption, not in the paper) with aluminium density 2700 kg/m³, 0.43 J/m² per gap corresponds to about 1.6 × 10^4 J/kg, a small number compared with chemical fuel (about 10^7 J/kg) [memory, high confidence on fuel; the foil case is my illustration]. - The electrostatic suspension is unstable (his own caveat), and requires feedback.

b. Established experiments

Casimir forces between metals are measured at tens to hundreds of nm [memory, high confidence]. Forward cites the absence of good metal-plate data at 1 to 20 nm in 1984 (Arnold et al. deviated below 300 nm "probably indicating a lack of surface smoothness") [read: Forward p.1].

c. Tested

Not built, to my knowledge, and the paper itself reports no experiment. Whether anyone built it was not found by the Investigator.

d. What would have to change

Nothing in mainstream physics. A force that depends only on separation is conservative. The device is a store, like a charged capacitor or a raised weight.

Verdict

• As a conversion and storage device (energy comes from the prepared foil state): Plausible, untested.
• As a net source of energy from the vacuum: not what Forward claims; if read that way it is refuted by the conservative-force argument (a closed cycle gives zero net work).

---

Corrections to Step 1

1. claims.md claim 4: the inconsistency of "1700 Pa / 17 atmospheres" is correct, and the deep read inferred (without proof) that 17 atm is the correct figure. I confirm this independently with a Knudsen-flow estimate (about 16 atm for argon, 200 nm pores, porosity 0.25). I add: the pumping check is made against the tunnel block's output (2.2 to 22 kW), not against the 1-10 W bonded-wafer device the patent calls preferred, and for that device the corrected pumping power (0.3 to 1.7 W) is a large fraction of its estimated output. Step 1 reports the inconsistency but does not state this consequence.
2. claims.md claim 3 reads "the same balance is asserted, though not proven, for all atoms". Correct, and the patent adds in the scaling section that its own shielding assumption is "crude" [read: patent p.4]. Step 1 does not mention that the cavity needed to suppress the 455 Å modes is d about 23 nm, not the 0.1 µm of the claim, or that a parallel-plate gap suppresses longer wavelengths only. Added above.
3. claims.md claim 1 says the cavity "suppress[es] vacuum-field modes" and Step 1 treats this as the inventors' reasoning. The patent says "at this wavelength and shorter" [read: page image]. The mode-suppression direction needs correcting as above.
4. Digest/OCR: the patent text layer reads "220.529137"; the page image shows "2π0.529137 = 455 Å". Anyone checking the sentence from the OCR text alone would think the patent wrong by a factor of 1.57. It is correct.
5. Investigator leads: the Investigator reports "Measured power in (about 2x10^-5 W)". My simple product gives 4.4 × 10^-5 W (minor; the paper says ~2 × 10^-5 W). The Investigator's statement that the expected output is "milliwatts" is the paper's own, but 5 sccm at 1 eV per atom is 0.36 W; the milliwatt figure rests on an unstated reduction.
6. Deep read of the patent, section 3: "porosity of 0.25 is never used" is correct; including it raises my pressure estimate by about 4 times, which is how the 17 atm figure becomes plausible. The deep read suggests 1700 Pa was likely meant to be 1700 kPa; my estimate agrees.
7. Unflagged in Step 1: the patent's "less than one-trillionth of a second" is 16 times too short (1.56 × 10^-11 s); and the 2012 paper's Joule-Thomson section says He gave no reliable signal while the results section reports it strongest.

Investigator findings: kept in the record

Pertinent items stay here whatever my verdict. Each is a lead, not a verdict. - Inventor's own work: the 2012 experiment and 2010 SSE abstract (Claim 1); the 2019 Atoms review with its "potential hill" and detailed-balance admissions (Claim 2); Haisch's caveats on calphysics.org and in 2009 (Claims 2 and 3); Moddel's later MIM devices and patent family with a projected "up to 250 W" [web: per Investigator]. Patent assigned to the Regents of the University of Colorado on 2008-06-10 and priority 2004-09-27 [web: Google Patents, per Investigator]. The project's Haisch_abstractSSE10 is probably the SSE meeting of June 2010 (inference by the Investigator). - Support by others: Henriques (2014) "some evidence of radiation" for Xe; Puthoff, Little and Ibison (2002) tested a related idea and reported no success; Hathaway Research project with Haisch and Moddel, no results found. - Related work: Cole and Puthoff (1993) as the thermodynamics basis; third-party patents citing Haisch-Moddel; BlackLight Power 'hydrino' (a 2009 commenter's link, his characterisation); Kotov et al. (2025) on cavity modification of ground-state properties (position unknown). - Cetto and de la Peña (1988): a counter-lead, see Claim 3. - Absence of evidence. After 2012 there is no further gas-flow experiment by the Colorado group; Jovion's site has an empty "Progress" link; the patent status could not be confirmed. The Investigator lists four competing explanations (effect did not hold up and was dropped; funding ran out; trade secrecy; work moved to the MIM line). I add, from the mathematics: the 2012 data point to an unresolved gap of 3 to 4 orders of magnitude and opposite trends, and the 2019 paper lists detailed balance as unresolved, which fits explanation (a) or (d) without excluding the others. Records that would decide: the DARPA/SPAWAR final report for N66001-06-1-2026, USPTO maintenance-fee records, Hathaway's report, CU Technology Transfer files.

Summary table