QE (Step 4b): Haisch-Bernard

QE run: 2026-10-04 (Opus). Inputs read: _text/manifest.json, _textprep.log, all four digests, both deep reads, facts.json, claims.md, sites.md, all ten research/ notes, physicist.md, mathematician.md, contrarian.md, arbiter.md, primer.md and example.py. Spot checks were made against the extracted text (_text/Patent7379286-c0ceee.txt), a page image of patent p.11 that I rendered (pdftoppm, 200 dpi, /tmp/qe-haisch/), and pdftotext of two Investigator downloads (research/files/Dmitriyeva-Moddel-2012-PhysicsProcedia38.pdf and Moddel-Dmitriyeva-2019-Atoms7-51.pdf).

Verdict: MINOR

The central verdicts hold. The work product is thorough and honestly tagged, and the arithmetic reproduces. There are four kinds of fixable defects:

• (1) One misreading of the 2012 paper went into three files.
• (2) The patent's hydrogen cavity size was described unfairly.
• (3) Pages 17-21 of the patent rest on the digest only, and they hold the inventors' own answer to the exit/detailed-balance objection.
• (4) Some favourable experimental details from the Investigator's findings were left out of the arbiter and the primer.

None of these is likely to move a headline verdict. Defect 3 could add nuance to claim 2.

---

1. Coverage

Every manifest document was digested. Nothing was skipped. _textprep.log shows no skipped files.

2. Shortcuts

• Patent pp.17-21 rest on the digest alone. These pages hold three embodiments that no reviewer, the arbiter or the primer examined:
• (a) flexible photonic-crystal polymer cavities switched by stretching (FIGS. 6A-6B);
• (b) MEMS (micro-electro-mechanical systems) modulation of the cavity wall spacing (FIGS. 7A-7B; it cites the 2001 Bell Labs Casimir MEMS demonstration, Chan et al.);
• (c) "Asymmetric Casimir cavity entry and exits including absorbing means" (FIG. 8). The pages also hold claims 1-20, which define the legal scope.

Embodiment (c) matters most. I read p.19 in the text layer myself. There the inventors put an absorbing glass layer on the entry side, and choose the conducting-layer length "so that the emerging atoms do not have substantial access to radiation emitted from the absorbing regions" [read: Patent7379286 p.19]. This is the inventors' own engineering answer to the "what you release going in you get back coming out" objection, and so to the claim 2 verdict (Refuted by calculation, moderate confidence). The arbiter's claim 2 reasoning does not engage with it. Embodiments (a) and (b) move the walls instead of the gas. For those, the exit cost would be paid by the actuator working against the Casimir/Casimir-Polder force, and the gas-flow detailed-balance argument does not apply word for word. This is a defect: a central claim's verdict was reached without these pages. - The deep read of patent pp.11-15 used page images where the numbers mattered. That is good. - Reviewers relied on the Investigator for Henriques (2014), Cetto and de la Peña (1988), Cole and Puthoff (1993) and the 2009 LANL abstract book. Each reviewer says so, which is honest. These are not manifest documents, so QE cannot order deep reads of them. They stay as open leads (see 6).

3. Accuracy spot checks

The other numbers I recomputed or saw reproduced all agree: Casimir pressure 13 Pa at 100 nm; the Casimir-Polder shifts; 21.5 W and 2152 W per eV; 16.7 atm; Larmor power 4.67e-8 W; spiral-in time 1.56e-11 s. example.py reproduces them.

4. Missed material and fairness points

1. The hydrogen cavity size was misdescribed. The mathematician and arbiter say that cutting off the 45.6 nm hydrogen modes needs d ≈ 23 nm, "not 0.1 micron". Primer section 4, row 3, and section 5 Step B set hydrogen against a 0.1 µm gap. But the patent itself specifies d = 250 Å (25 nm) for hydrogen [read: p.11, col. 4]. With the ideal 2d cutoff, that is within about 10% of the 22.8 nm the reviewers compute. The patent also states its own convention: suppression is "strongest for wavelengths of d or greater" [read: p.11, col. 3]. That is a cutoff at λ ≈ d, a factor of 2 looser than the ideal 2d. On that convention, d = 250 Å does remove 455 Å.

So the "mode-suppression direction is backwards / internally inconsistent" finding rests mainly on the phrase "at this wavelength and shorter". The patent's numbers are consistent with removing longer wavelengths. The fair statement is narrower: - (i) the patent's wording is loose, and its λ ≈ d convention is a factor of 2 more generous than the ideal-mirror 2d cutoff; - (ii) the strong objections remain: no metal reflects at 27 eV, nothing is computed, and the shift size is 10^7-10^10 too small.

The arbiter's split verdict on claim 1 ("Inside SED ... internally inconsistent as written") should be softened to match. This does not change the headline verdict on claim 1, which rests on the Casimir-Polder calculation and the experiments. 2. The asymmetric absorbing-layer embodiment (FIG. 8, p.19) and the moving-wall embodiments (pp.17-18) are absent from all reviews, the arbiter and the primer (see 2). They are the inventors' own design responses on claim 2. 3. The Jordan Maclay citation [read: p.11, col. 4]: "According to a Jordan Maclay, who has done theoretical Casimir cavity calculations, a long cylindrical cavity results in an inward force ... this implies that a cylindrical cavity of diameter 0.1 micron would yield the desired decay". This is the patent's appeal to a named Casimir theorist. It appears nowhere after the digest. It should be recorded as the inventors' characterisation of Maclay's work. It is also an Investigator lead: what did Maclay actually calculate? 4. Extension to molecules and molecular bonds [read: p.12; p.13 Summary; p.19 variation v] and to liquid working fluids (ethylene glycol, sodium). These are in the digest and facts.json only. They are peripheral, but they are claimed scope and should get one line in the primer's account of the patent. 5. The patent's "slow process" assumption [read: p.11]: "no matter how quickly the atom is injected ... the process will be a slow one as experienced by the orbiting electron. We therefore assume ... gradual release of energy in the form of heat, rather than a sudden optical radiation signature." The contrarian uses this well, against non-adiabatic entry. The arbiter and primer leave it out. Yet it is the patent's own premise for looking for heat rather than spectral lines, and it bears directly on the 2012 choice of an IR pyroelectric detector. It is worth one sentence in the primer. 6. Cetto and de la Peña (1988). The arbiter turns the abstract's "coincide with previous quantum calculations" into "SED itself ... reproduces the ordinary QED-sized shifts, not an eV shift". Primer section 3.4 restates this as "SED done carefully gives the same tiny QED-sized shifts". That is a reasonable inference from an abstract, but it is still an inference, and nobody read the paper. The primer should label it "[web: abstract only; inference]" and keep Moddel and Dmitriyeva's narrower wording ("under particular simplifying assumptions this shift is not predicted") beside it. 7. The physicist's "4.7e-11 s ... matching the patent's 'less than one-trillionth of a second'" is wrong, and the arbiter corrected it. No further action. 8. Digest errors that were caught downstream, recorded for completeness: - The digest has "1.3×10^10 transitions/s" for the channels. The page has 1.3×10^20. - The digest has 1.3×10^20 for the cube. The deep read gives 1.3×10^22. - The digest has r^(1/2). - The digest spells the assignee "Jovian". facts.json notes "Jovion".

5. Retention check

Rule: pertinent material from the sources or the Investigator may not be dropped from the arbiter or primer because a reviewer judged it negatively. The arbiter and primer keep most leads well: Henriques's "some evidence" next to the more negative summary in the 2019 paper; Haisch's caveats; Cetto and de la Peña; Cole and Puthoff; the MIM line; the hydrino commenter as the commenter's own characterisation; third-party patents; Kotov et al.; the funding chain; and a full absence-of-evidence analysis with five explanations. Defects:

1. 2019 Fig. 9 result dropped. The arbiter and primer mention "one 0.1-micron helium run" but not its result. Among the 0.1, 0.2 and 0.4 µm uncoated filters, "the highest signal was registered from the 0.1 µm filter" [web: Atoms 2019 p.11, via Investigator]. The authors give two readings: a better mode match, or more frictional heating from the larger pressure drop. This is the one trend in the direction the cavity hypothesis predicts. It is pertinent favourable evidence and belongs in the primer's Step E and evidence table, together with both readings.
2. The 2012 controls are missing from the primer. The primer says "Ordinary heating is the leading candidate" without the inventors' controls. Those were: the Mylar 3 mm hole control, which gave the opposite lock-in phase (cooling), so Joule-Thomson was ruled out; the finding that power was linear in flow, not quadratic as friction predicts; the finding that the signal did not follow membrane surface area, against adsorption; and the Reynolds-number check, against turbulence. The authors' conclusion was "None of these clearly fit the data, but it is possible that a combination of effects could" [read: 2012 paper]. The contrarian and physicist record these. The primer, which is the reader's document, should state them next to the heating hypothesis. The emissivity explanation of why uncoated beat gold (0.95 against 0.01-0.1) should also go in.
3. Moddel et al. 2021 observed result. The primer mentions the MIM line only through its projected 250 W. It leaves out that Phys. Rev. Research 3, L022007 (2021), a mainstream APS journal, reports observed conductance increases with thinner Casimir cavities and "anomalous offsets in the current and voltage" [web: https://journals.aps.org/prresearch/abstract/10.1103/PhysRevResearch.3.L022007, via Investigator]. This is a different mechanism, but it is a measured anomaly by the co-inventor and must stay in the record.
4. Historical record items dropped: - the 2009 presentation at the LANL CNLS "Workshop on New Frontiers in Casimir Force Control", Santa Fe (the abstract is unread, image-only PDF); - the patent priority date 2004-09-27 (provisional application); - the Davis-Puthoff AIAA chapter and the STAIF-2006 experimental-concepts review that lists Haisch as a co-author; - Dmitriyeva's parallel LENR (low-energy nuclear reactions) thesis work and the Coolescence funding.

Each is a short line in the primer's "Other material kept in the record".

6. Process

• Source tags: present and honest throughout. Reviewers say clearly what they did not read (Henriques thesis, Cetto and de la Peña, Cole and Puthoff, Cole and Zou). The interested-party framing (Haisch, Moddel, Jovion, Calphysics) is consistent in every file.
• Web checks: physicist 2 of about 6 (the remaining [memory] items, polarizabilities and Casimir-Polder scale, were later verified by the arbiter); mathematician about 3; contrarian 3 plus 1 failed (Henriques PDF, "too many redirects"); arbiter 5 of 8. That is acceptable, because every medium/low [memory] item that a verdict depends on was web-checked by the arbiter. Still [memory]: Lamoreaux 1997 and Mohideen-Roy 1998 (high confidence, acceptable), and the content of Cole and Puthoff 1993. That last item is a gap, because the claim 2 thermodynamic argument turns on it.
• Verdict scale: used consistently, with split verdicts stated part by part. One concern: claim 1 "Refuted by calculation" uses QED to refute a claim made within SED. The arbiter acknowledges this, and the scale allows it ("accepted laws"). The claim 2 "Refuted by calculation (moderate confidence)" is honest about its dependence, but it has not weighed the FIG. 8 asymmetric design (see 2).
• Primer: follows all nine Step 4 sections in order, with answers at the end. It defines every symbol and acronym (OCR, QED, SED, CP, MIM). The worked example uses real numbers. Its tone is fair to the inventors ("What the inventors get right"). It fits the reader profile.
• example.py: runs cleanly with the project venv (exit 0). Its output matches the primer's numbers (455.6 Å, 6.58e15 Hz, 4.67e-8 W, 1.56e-11 s, CP shifts 1.5e-9 / 1.2e-8 / 3.0e-8 eV, 21.5 W/eV, 2152 W/eV, 16.7 atm, 1.69 W, 0.29 W, thermal/drift ratios 550-3149, cycle work 0, 0.36 W, 4.4e-5 W). Small fix: section D prints the pumping fraction from hard-coded 0.28 and 1.69 instead of the computed variables. The output is correct, so this is cosmetic.
• Investigator: stayed within the 10-minute cap (about 6 minutes) and listed unreached sites with "Recheck? yes". Queries are recorded. Blocked sites name their cause (extension permission or HTTP 403).

---

Remediation (one round)

Deep reads (most important first; slugs from manifest.json)

1. Patent7379286-c0ceee, pp.17-19, Opus. Question: How do the three later embodiments deal with the exit/re-energisation step? These are the flexible photonic-crystal polymer (FIGS. 6A-6B), MEMS wall-spacing modulation (FIGS. 7A-7B), and especially the asymmetric entry/exit with absorbing layers (FIG. 8). Quote the exact wording on where the energy is absorbed, why the emerging atoms "do not have substantial access" to the emitted radiation, and how the design is meant to prevent the energy released on entry from being reabsorbed on exit. Give any dimensions, materials or numbers. State what does the work, and against which force, when the walls move instead of the gas. Check figures and units against page images.
2. Patent7379286-c0ceee, pp.19-21, Sonnet. Question: Quote the independent claims (claim 1, claim 15, claims 18-20) word for word from page images. Say which embodiments each covers. Say whether any claim needs net energy gain, or only the release and intake of radiation. List the full reference list as printed, checking whether Maclay appears and the Roy et al. year.
3. Patent7379286-c0ceee, pp.10-12, Sonnet. Question: From page images, quote exactly: - the "stairway" and "strongest for wavelengths of d or greater" passage; - the hydrogen d = 250 Å statement; - the Jordan Maclay cylindrical-cavity statement; - the molecular-bond extension; - the paragraph that dismisses the related patents (Shoulders 5,018,180; Mead and Nachamkin 5,590,031; Pinto 6,477,028, 6,593,566, 6,665,167).

Report the patent's own convention for the cutoff wavelength in terms of d.

facts.json and claims.md

• facts.json, Patent7379286 entry:
• fix how_read to say the deep read covers pp.1-2 and 11-16 of 21, and that pp.17-21 were digest only until remediation;
• add a key claim for the asymmetric absorbing-layer design;
• add the Maclay citation as the inventors' characterisation;
• add the molecular and liquid working-fluid extensions;
• add the patent's convention that suppression is strongest for λ ≥ d;
• add hydrogen d = 250 Å.
• claims.md, claim 1: add that the patent pairs hydrogen with d ≈ 250 Å and noble gases with 0.1 µm. Claim 2: add that the patent proposes an asymmetric entry/exit absorbing design (FIG. 8) to stop reabsorption, and moving-wall variants.

arbiter.md: add a "QE revision" section

• Withdraw the "helium is both strongest and unmeasurable" contradiction (Claim 1 corrections; Open questions bullet 1). The "so small" sentence is about the Mylar single-hole Joule-Thomson control (see 3, row 6).
• Claim 1: soften "internally inconsistent as written (mode-suppression direction)". The patent's numbers (d = 250 Å for hydrogen; λ ≥ d convention) are consistent with removing longer wavelengths. The real issues are the factor-of-2 cutoff convention, metal transparency at 27 eV, the absence of any computation, and a size gap of 7-10 orders of magnitude. Strike "the cavity needed would be ~23 nm, not 0.1 micron" for hydrogen, or rephrase it to say the patent's 250 Å is close to the needed 228 Å.
• Claim 2: weigh the FIG. 8 asymmetric design and the moving-wall variants (from deep read 1). State whether the conservative-potential/detailed-balance argument still applies, and why. Note that Cole and Puthoff 1993 is still unread.
• Label the Cetto and de la Peña reading as an inference from the abstract.
• Add to "kept in the record": the 2019 Fig. 9 result (0.1 µm highest He signal, with both readings); Moddel 2021's observed anomalous offsets; the 2009 LANL Santa Fe workshop talk; the STAIF-2006 review; the Maclay citation.

primer.md

• Section 4, row 3, and section 5 Step B: state that the patent pairs hydrogen with d = 250 Å (near the 228 Å the ideal 2d cutoff needs) and uses a λ ≈ d convention. Present 0.1 µm as the noble-gas and "tail" figure. Change the "Wrong, and why it is instructive: Direction of mode suppression" bullet to a fairer point: the cutoff is 2d, not d, and walls must actually reflect at that wavelength.
• Section 3.4 / Step C: mark the Cetto and de la Peña sentence as an inference from the abstract.
• Section 4: add one row or note on the FIG. 8 asymmetric absorber and the moving-wall embodiments, with the arbiter's revised weighing. Add the "slow process, heat not spectral lines" premise and the Maclay citation (as the inventors' characterisation). Mention the molecular and liquid extensions.
• Step E and evidence table: add the 2012 controls (Mylar hole opposite phase, linear-in-flow power, no surface-area dependence, Reynolds number, emissivity reading of plastic over gold) and the 2019 Fig. 9 result with both readings. Do not add the helium contradiction.
• "Other material kept in the record": add Moddel et al. 2021's observed conductance changes and anomalous offsets; the 2009 LANL CNLS workshop talk (abstract unread); patent priority 2004-09-27; the STAIF-2006 review and the Davis-Puthoff AIAA chapter on jovion.com; the Dmitriyeva LENR thesis and Coolescence funding.
• Optional: in example.py section D, compute the pumping fraction from the variables, not from literals.

mathematician.md and contrarian.md

No rewrite is needed. The arbiter's QE revision records the helium correction. The Recheck can note it.

Open leads for the next Investigator run (outside QE's deep-read budget)

• Henriques 2014 thesis PDF;
• Cetto and de la Peña 1988 full text;
• Cole and Puthoff 1993;
• OCR of the LANL 2009 abstract book;
• what Maclay calculated for cylindrical cavities;
• USPTO maintenance-fee record;
• DTIC final report for N66001-06-1-2026;
• Hathaway Research International results.

---

Recheck

Recheck run: 2026-10-04. I re-read the arbiter's "QE revision", deep/Patent7379286-c0ceee.md, facts.json, claims.md, primer.md, example.py, mathematician.md and contrarian.md, and ran example.py with the project venv (runs to the end, no traceback).

Remaining items are open leads only (Henriques 2014 thesis, Cetto and de la Peña full text, Cole and Puthoff 1993, LANL 2009 abstract book OCR, Maclay's actual calculation, USPTO maintenance record, DTIC report, Hathaway results). They are outside the QE deep-read budget and are already marked unread in the arbiter and primer.

Final: PASS