Arbiter (Step 3): Larson-Tyson (Realm Industries)

Inputs weighed: physicist.md, mathematician.md, contrarian.md, facts.json, claims.md, deep/, sites.md, research/. Tags: [read] project documents; [web] with URL; [memory] with confidence; [calc] my own arithmetic (project venv).

Source gap: none. Both patent applications and the Ventura County Star article were fully read (graphs from page images by the deep readers; the contrarian re-rendered p.12 of US 2009/0090312). Every inventor-side statement is the applicants' own characterization (Realm Industries has a commercial stake). The regulators (EPA, Cal/OSHA), prosecutors and press are interested parties on the other side and are reported as such. No conclusion here rests on [memory] alone, but the relaxation-lifetime figure is a recalled order of magnitude (see Web checks).

1. What the reviewers agree on

All three reviewers agree on the following, and my recomputation confirms it: - The patents contain no composition analysis, no energy-in or energy-out figure, no flow, rpm, load or control run (plain DC, bottled H2/O2). [read] - No independent test, replication or witness measurement of the "bond-angle vapor" was found by the Investigator. This states what was found, not what exists. No role found a failed replication either. - The default reading of every reported observation is an electrolysis gas (H2 + O2, "HHO" or "Brown's gas"): bubbles on plates [read: [0035]-[0042]], 5 V across the gaps, engines running with air excluded (an H2:O2 premix carries its own oxidiser), a torch, detonating cylinders. - The inventors' own later PCT text concedes this. I fetched it: "the inventors acknowledge that the process described herein could result in the production of hydrogen gas (H2) and oxygen gas (O2)"; the product is "an energy carrier, which includes but is not limited to molecules with unnatural bond angles and H2 and O2 gas"; "it may be possible that covalent bonds are broken" [web: https://patents.google.com/patent/WO2010039283A2/en]. The same text still claims "less energy is consumed in making the product" than electrolysis, so the "advantage over electrolysis" sentence was kept; the "safer" sentence is not in my fetched passages (the Investigator and contrarian report it dropped; the fetch tool returns only excerpts, so I treat that as reported, not independently confirmed).

2. Numbers where reviewers differed (recomputed [calc])

3. Web checks (budget 8; used 4)

1. WO 2010/039283: quotes above verified [web: https://patents.google.com/patent/WO2010039283A2/en]. It names a "tri-fuel engine, a Hyundai GX 390" (note: the patents under review use Honda GX340/GX390; I did not resolve this naming difference, which might be a drafting slip or the fetch tool's paraphrase). The fetch did not return diesel-ratio "test experiments" wording; that detail stays as the Investigator's report.
2. US 7,793,621 B2: grant 14 Sep 2010, status "Expired - Fee Related", expiry shown 9 Nov 2027, i.e. lapsed early for non-payment [web: https://patents.google.com/patent/US7793621B2/en]. The fetch did not return the claim text, so "claim 1 requires metering air and mixing vapor, air and a fuel" remains the Investigator's report, repeated by physicist and contrarian. I could not independently confirm it.
3. Barrier to linearity: confirmed (table above).
4. Gas-phase relaxation time of excited H2O bend: search returned only that rate constants were measured (laser-induced fluorescence, 295-1020 K) and liquid-phase T1 of 170 +/- 30 fs [web: https://www.nature.com/articles/s41467-020-19759-w]; no gas-phase number retrieved. The "nanoseconds to microseconds" figure stays [memory, medium]. The verdict does not need it: months of storage versus any plausible collisional time (kinetic collision rate ~10^10 /s at 3 atm [calc, per contrarian]) is a gap of more than 10^15, and the condensation argument alone excludes pure H2O gas. Not rechecked (not verdict-bearing): IMR analyser CO range [memory, low]; tap-water resistivity [memory, medium]; double-layer capacitance [memory, medium]; Meyer court finding (not used in any verdict).

4. Corrected claim list (final verdicts)

Claim 1: kHz "RC resonating frequency" (f = 1/(RC), 1-22 kHz) combined with the supply frequency bends the H-O-H angle without breaking bonds

• Agreement: all three say the mechanism fails. Physicist: Refuted by calculation (strong form). Mathematician: Internally inconsistent; mechanism never tested. Contrarian: split, circuit time constant Established, bending mechanism Refuted by calculation and Internally inconsistent.
• Disagreement: only the label (Refuted vs Internally inconsistent vs Never tested). They are not in conflict; they describe different faults.
• Stronger argument: the mathematician's and contrarian's. (a) An RC network has no resonance, there is no 2 pi, and the stated tuning variables (area, gap, current) cancel out of eps/sigma or do not enter at all; (b) the polarity contradiction between [0019] and claim 2 [read]; (c) energy scale: a 22 kHz drive carries about 10^-10 eV against a 0.198 eV bend quantum; field-to-thermal ratio p E / kT for the stated 5 V over 4.5 mm (E ~ 1.1 kV/m, dipole 6.2e-30 C m) is ~1.7e-6 [physicist, calc]. Honest limit: the energy-scale argument rules out direct resonant excitation; it does not by itself rule out every nonlinear electrode effect, but the patent proposes no such mechanism and measures nothing.
• Final verdict (split): circuit has an RC time constant set by water conductivity: Established (trivial). "Resonance bends the bond angle without breaking bonds": Refuted by calculation (energy scale) and Internally inconsistent (terminology, units, tuning variables, polarity). The mechanism has also Never been properly tested; no angle measurement exists.
• What would settle it: infrared or Raman spectrum of the cell's product gas around 1595 cm^-1 (and the Raman bend band) with the "dual frequency" on and off, plus a measured impedance spectrum of the cell to see what its real RC is.

Claim 2: the pressurised "vapor" is a distinct substance (unnatural angles), stable for months at 1-100 psi

• Agreement: all three refute the strong form and call the identity of the gas never measured.
• Disagreement: none on verdict. The physicist rests on vapour pressure and no barrier; the mathematician on barrier and Boltzmann; the contrarian adds relaxation time and the 133 kJ/mol ceiling.
• Stronger argument: the vapour-pressure/condensation argument is strongest because it needs no recalled lifetime: at 25 C water's saturation pressure is 0.46 psi, so a gas at 1-100 psi that stays gas at room temperature is not H2O alone [calc]. Second: a molecule bent toward linear sits at up to 133 kJ/mol excess with thermal population ~7e-24, i.e. a hot molecule that sheds energy as heat in collisions.
• Evidence kept in the record: the company's TyLar MSDS "states that the material consists of compressed hydrogen gas", the invoice called it "compressed water vapor", shipping papers listed both an oxygen-hydrogen mixture and pure hydrogen; two of three cylinders detonated when shot [web: https://response.epa.gov/sites/7236/files/SSI_polrep_2.htm]. These are EPA's characterization (regulator) plus the company's own paperwork, each from one source. They match H2 + O2 and do not match a stable single-component H2O gas. The 2011 TyLar gas may not be the 2007-08 patent product; nobody has shown they are the same.
• Final verdict (split): strong form (a months-stable, bent-angle H2O gas, distinct from steam and from H2/O2): Refuted by calculation. Weak form (the cell makes a pressurised combustible gas; what it is): Never properly tested. It was dismissed by classification (Purdue's Raftery, EPA, Cal/OSHA), not by measurement, and the only physical sample was destroyed unsampled for safety reasons. The best-supported reading is H2 + O2.
• What would settle it: gas chromatography or mass spectrometry of gas drawn from the example generator, plus dew point and condensate at 35 psig. One run separates "H2 + O2" from "something else".

Claim 3: igniting the vapor releases usable energy, using less energy than electrolysis, no toxic gases, no atmospheric oxygen used, safer

• Agreement: no energy figures anywhere; burning releases energy trivially if the gas is H2/O2; the first law caps any over-unity reading. Physicist: split, angle-energy mechanism Refuted by calculation, accounting Never tested. Mathematician: split, "less than electrolysis" Never properly tested, any excess over input Refuted. Contrarian: "cheaper than electrolysis" Refuted by calculation if H2/O2 (it is electrolysis), "safer" Tested by events, not supported.
• Disagreement: whether "less energy than electrolysis" is refuted or untested. This turns on what the gas is. If H2 + O2, Faraday's law governs and a 5 V cell uses 965 kJ per mol H2 against 237 kJ minimum (25% LHV), i.e. worse than normal electrolysis. If the gas were a stored excited species, conservation does not forbid it as an energy store (at most 133 kJ/mol for 180 degrees), but it would be less than the energy needed to split water and would release as heat without ignition. Both branches are bounded by calculation; neither is measured.
• Final verdict (split): burning the gas releases energy and gives water: Established (if H2/O2). "Less energy than electrolysis": Never properly tested (no meter); the H2/O2 branch is Refuted by calculation and any output above electrical input would be Refuted by the first law (not claimed numerically). "No atmospheric oxygen consumed": see claim 4. "Safer than electrolysis hydrogen": Tested, not supported (uncontrolled incidents, not a designed test; label revised, see QE revision) (Dec 2008 tank failure, June 2010 fatality, Aug 2011 Sylmar explosion, Cal/OSHA's finding of H2 and O2 stored together; cause of the 2010 event "accidental but undetermined").
• What would settle it: a Joule (volt x amp x time) meter on the cell plus gas-volume measurement; compare energy in per standard litre with 2 F V per mole H2 at the operating voltage.

Claim 4: the vapor runs engines on its own (Honda GX340/GX390 with outside air excluded; Toyota FJ-40 with gasoline shut off; exhaust O2 ~20%, CO/CO2/HC near zero; Figs. 11-13)

• Reviewers: physicist "Tested, not supported"; mathematician "Plausible, untested" (as electrolysis gas); contrarian "Never properly tested".
• Which is stronger: the contrarian's. The scale's "Tested, not supported" requires controlled experiments that failed to support the claim. These were uncontrolled applicant runs with no rpm, load, flow, power, voltage, current, vapor quantity, witness or date; they neither support the new-substance claim nor test it. The physicist is right that the data fit lean hydrogen and that instrument readings are doubtful, but that is a reading of uncontrolled data, not a failed controlled test. The mathematician's "Plausible, untested" is right for "a plate-stack gas can run an engine" (hydrogen and oxyhydrogen engines are real [memory, high]) but understates two anomalies.
• Data problems all agree on, which I confirm: (i) CO2 of 18-19% exceeds the ~15.1-15.4% dry stoichiometric ceiling; CO sits flat at exactly 10.00% (probable analyser range limit); the oxygen balance fails in the late gasoline phase (table above); (ii) text says CO/CO2 "undetectable"/engine-off "unchanged", but the vapor-phase traces show CO2 ~0.2-0.6% (5-15 times ambient 0.04%), CO ~0.2-0.4% and an O2 step of 0.3-1.0 point at engine-off [read: deep reads; contrarian's p.12 render]; (iii) Fig. 13 text says gasoline = samples 1-36 and CO ">5%" but the graph shows ~51 and 10%; (iv) hydrocarbon plotted in %, only in Fig. 11; (v) the FJ-40 text does not say air was blocked (Fig. 14 shows air entering) and the Honda no-air runs have no data at all.
• Oxygen budget: the vapor-phase exhaust O2 of ~20.0-20.8% (0.3-1.0 point below ambient) corresponds to 1-2.5% H2 relative to air [calc], too little to power a 4.2 L engine; the physicist's order-of-magnitude (~4 kW load needs ~20% H2 in intake) and the contrarian's (35-55% of air O2 if air-fed) agree. So either the gas carried its own oxidiser (H2 + O2), the engine delivered almost no power, or residual gasoline did the work. The inventors' abstract claim of "no consumption of any atmospheric oxygen" is not demonstrated and the plotted O2 deficit is the opposite of zero consumption, albeit small.
• Strongest honest case for the inventors (contrarian), kept: oxyhydrogen engine operation is real, so "something derived from water ran an engine" is credible; the vapor-phase traces repeat across three runs; the inventors published readings that were not "zero" and a gasoline phase they could have trimmed; the authorities describe a working gas-production operation filling cylinders for sale.
• Final verdict (split): "A gas from the plate stack can fuel or supplement an engine": Established in general (hydrogen supplementation), Plausible, untested for this apparatus. "Vapor-only operation with air excluded, gasoline off, new substance, no oxygen consumed": Never properly tested (uncontrolled, indistinguishable from oxyhydrogen). The gasoline-phase exhaust data are internally unreliable (atom-balance failure). The claim was not tested-and-failed and not shown to be dismissed after a test.
• What would settle it: engine run with metered vapor flow (cylinder pressure decay), rpm, load, exhaust lambda and a calibrated multi-gas analyser (first verify atom-balance closure on gasoline), air path sealed or measured, and a control run with bottled H2/O2 at matched flow. A hydrogen sensor in the intake.

Claim 5: torch (5 psi, #00 tip) cut thin titanium and nickel-plated steel and melted tungsten; 60:40-70:30 vapor:diesel combusts on compression

• Reviewers: physicist Plausible, untested (tungsten possibly overstated); mathematician Plausible, untested / not testable for any bond-angle attribution; contrarian cutting Plausible, untested, tungsten bulk melting not supported as stated, diesel not testable as stated.
• Stronger argument: the contrarian's, because it quantifies: adiabatic oxyhydrogen flame ~3,075 K [web: https://etasr.com/index.php/ETASR/article/view/6132, via contrarian's search summary] against W melting 3,695 K, so bulk melting in a plain oxyhydrogen flame cannot occur; surface oxide formation and burn-back of a thin rod can. The report is qualitative (no rod size, time, photo). If bulk melting were genuine it would need a hotter-than-oxyhydrogen channel, which the patent does not argue. I keep it as the one observation mainstream expectation does not clearly cover, because the text says "melted" and the oxide explanation is mine, not the inventors'.
• Final verdict (split): cutting thin sheet: Plausible, untested. Tungsten "melted": Never properly tested; bulk melting is above the oxyhydrogen adiabatic limit by calculation (label revised, see QE revision). Diesel ratio: Not testable as stated (no data).
• What would settle it: a recorded torch run with pyrometry, a rod of known diameter, mass before/after, and a control oxyhydrogen torch from a commercial generator.

Claim 6: 17 June 2010 explosion killed one person (December 2008 earlier incident); authorities "were told" it was a water-based explosion involving extracting hydrogen from water (unnamed source; see QE revision)

• Agreement: all three: an event record, not a physics claim; it supports the H2/O2 reading but does not test the bond-angle claim.
• Record (kept as stated, interested parties named): the article does not name the victim [read]; the Simi Valley Acorn (25 Jun 2010) identifies co-inventor Tyson Larson, 28 [web: https://www.simivalleyacorn.com/articles/still-no-official-cause-of-explosion-on-easy-street/]; Cal/OSHA found three pressurised containers failed and issued $5,685 non-accident citations (appealed); the 9 Aug 2011 Sylmar explosion injured two; Cal/OSHA penalties of $540,890 (2012) included hydrogen and oxygen stored in the same tanks (AP/Daily News, via Investigator). Federal fraud case: Rossignol convicted 2014 (20 years); Stehl entered an Alford plea on 2 Sep 2014 [web: https://www.justice.gov/usao-ndny/pr/second-defendant-convicted-8-million-fraud-scheme-concerning-purported-alternative]. The prosecutor (interested party) said the case was "not alleging that this is some kind of fraudulent scientific process" (VC Star, via Investigator) and the charges concerned business-status misstatements. The legal record therefore neither validates nor refutes the physics. The fraud convictions are a separate reason to want independent data for the unwitnessed test claims; they do not refute any claim.
• Final verdict: Established as an event record; cause of the 2010 explosion Never determined; identity of the gas Never properly tested.

5. Corrections to Step 1 (claims.md and facts.json)

1. Claim 1: the "resonating frequency" is an RC time-constant frequency; unit unspecified (rad/s or Hz); "14 MHz" was 1.4e7 rad/s (2.2 MHz in Hz at sigma 0.01 S/m).
2. Claim 4: "CO, CO2 and hydrocarbons near zero" overstates the plotted vapor-phase traces (CO2 0.2-0.6%, CO 0.2-0.4%). "Honda ran with outside air excluded" is the patent text only, with no data; the three FJ-40 runs (Figs. 11-13) are the only graphs; claims.md merges them.
3. Claim 4: the CO2 problem is not only "above the usual maximum": gasoline-phase readings fail an oxygen balance (negative water budget at three points).
4. Claim 6: the victim was co-inventor Tyson Larson [web]; the "similar incident" 18 months earlier is the December 2008 tank failure.
5. The "no grant stated" note is outdated: the CIP issued as US 7,793,621 B2 on 14 Sep 2010 and lapsed for non-payment [web]; the parent 11/973,170 was abandoned 3 Sep 2013 after no reply to an office action (Investigator, Google Patents). The granted claim wording is unconfirmed by me.
6. Step 1 should note the claim 2 / [0019] polarity contradiction (anode at positive terminal vs negative terminal) as an internal inconsistency, not only a reader observation.
7. Deep-read figure differences of ~0.1-0.3 point in O2 baselines are immaterial to every verdict.

6. Where the record goes dark: absence-of-evidence analysis

What stops: after the Sylmar explosion (9 Aug 2011) and EPA's destruction of the cylinders (15 Sep 2011) there is no technical paper, test report, conference talk, independent analysis or product evidence. Realm Catalyst was dissolved in Nov 2011; EP 2209550 was deemed withdrawn 5 Jul 2013; the US parent was abandoned 3 Sep 2013; the CIP lapsed in 2018; the family had 15 members with a Chilean divisional filed Nov 2011 (Investigator). Larson died in 2010; Stehl was injured in 2011, indicted 2010, pleaded 2014 and was reportedly sentenced to 12 years (NCPR snippet, not opened). The only later technical echo is R2D2 LLC / H2Plus (2022-2025) reclaiming a "structurally altered gas molecule" (angle 94-104 degrees, below not above 104.5 degrees), which cites Realm's application as prior art [web: https://patents.google.com/patent/WO2023055615A1/en]; this is re-use of an idea, not confirmation, and no measurement data was found for it.

Competing explanations: 1. Ordinary electrolysis gas that failed commercially and legally. For: every computed number (5 V cell at 25% LHV, 1-2.5% H2 equivalent in exhaust, vapour pressure, MSDS "compressed hydrogen", detonating cylinders, H2 and O2 stored together, inventors' own PCT concession, Raftery's opinion) fits it. Against: nothing measured excludes more; Raftery worked from the application and news reports, not the apparatus. Decided by: gas analysis, which was never done. 2. Death, injuries, penalties and prosecution ended the work irrespective of merit. For: the timeline (death 2010, indictment 2010, Sylmar 2011, dissolution Nov 2011, patents unpaid 2013-2018, two inventors changed on the EP register 2010/2011). Against: a patent family filed in 15 jurisdictions could have been maintained by an assignee if the company believed in it; late fees were paid in 2011 and 2013 and then stopped, so someone was still paying in 2013. Decided by: assignee and payer records for the maintenance fees. 3. Work moved abroad or under another name (South Korea consignee; Strategic Sciences; Nevada). For: the August 2011 Korea export attempt and the company's pattern of renaming (Rainbow Technologies, BGX, Realm, Strategic Sciences, Rainbow of Hope, TyLar). Against: no evidence of later activity found; the Investigator did not reach the Nevada records. Decided by: the Korean consignee's records, Nevada/California Secretary of State filings, CD Cal exhibits. 4. Patent office technical rejection that the applicants could not or would not answer. For: abandonment "for failure to respond to an office action" (US); EP search opinion (31 May 2011) and examination report (27 Jan 2012) followed by withdrawal. Against: abandonment is also what happens when money, personnel or legal capacity run out; the objections are unread. Decided by: USPTO file wrapper (11/973,170) and EPO documents, document IDs ERR2FVMH8608FI4 (search opinion) and ESQUPCFQ1433FI4 (examination report) listed on the EPO register but not rendered [web: https://register.epo.org/application?number=EP08835437]. 5. Suppression. For: the Investigator saw Google's note that a DMCA notice removed 10 results on one "Sylmar/Tylar" query (filer and content unknown). Against: no evidence of who, what, or why; the most direct physical evidence loss (cylinders) has a documented, safety-based reason (AERO advised against opening them; EPA CID had asked for sampling). I record this as an unexplained item, not as evidence of suppression, and not as nothing. 6. Commercial secrecy. Weak: the company published full patents and shipped product; but the earlier Rainbow Technologies equipment sold to Timothy A. Larson in 2005 (Stehl's attorney, AP) and the items found in a Saranac Lake garage in 2015 (NCPR snippet) show an earlier process whose details are not in the record.

Judgement for the reader: explanation 1 (with 2 as the proximate cause of the end) is the best fit for the data on hand, but the record cannot exclude 3 or 4 because the decisive documents were not yet read. The loss of the cylinders is the single most consequential gap: it removed the one chance to identify the gas.

7. What would settle it (consolidated)

1. Compositional analysis (GC/MS, thermal conductivity, dew point) of the product gas from a rebuilt example generator: H2 : O2 : N2 : H2O.
2. Joule meter and gas-volume measurement: energy per standard litre versus 2 F V per mole H2.
3. Control runs: same cell on plain DC (no "dual frequency"); bottled oxyhydrogen at matched flow in the same engine and torch.
4. IR/Raman spectrum near 1595 cm^-1 of the product.
5. Engine runs with recorded rpm, load, vapor flow, sealed or measured air path and an analyser whose atom balance closes on gasoline first.
6. Documents: USPTO file wrapper 11/973,170 and 12/244,648 office actions; EPO search opinion and examination report; EPA POLREP 4 and CERCLA letters; Cal/OSHA inspection files and appeal decision; CD Cal and NDNY exhibits (any technical evaluations of Stehl's process); Strategic Sciences filings (Nevada); the Korean consignee.

8. Open questions

• Is the "Honda GX390" in the patents the same engine as the "Hyundai GX 390 tri-fuel engine" in the PCT text? (Fetch tool excerpt; unresolved.)
• What does granted claim 1 of US 7,793,621 actually say, and why was the CIP allowed while the parent was abandoned? (Claim text not confirmed by me.)
• Was the 2007-08 patent product the same as the 2011 TyLar gas?
• Did the Korean consignee receive earlier product and analyse it?
• What did the EP examiner write in January 2012?
• Did the plate-stack outputs match Faraday's law gas volume? Never measured.
• Is the unexplained vapor-phase CO2 (0.2-0.6%) residual gasoline or sampling carry-over, and why does the exhaust O2 drop below ambient? The data cannot tell.

9. Summary of final verdicts

10. QE revision (after qe.md, verdict MINOR; four further deep reads)

Inputs added: qe.md; deep/ reads of the Ventura County Star article (pp.1-3), US 2009/0092540 (added section 10: chamber wiring, waveform, "intrinsic frequency", and applications that presuppose a composition) and US 2009/0090312 (appended section 2: [0075]-[0096], claims 1-23). Tags as above. Web checks used: 5 of 8.

Net effect on verdicts: no substantive verdict changes. Two labels change to stay on the Step 2 scale (Claim 3 "safer", Claim 5 tungsten). Several statements are corrected or qualified. Sections 4 and 9 above are updated in place.

10.1 Claim 6: who said what (QE defect A6, confirmed by the deep read)

• The article says: "The cause of the blast is under investigation. Authorities were told it was a water-based explosion, and that the company's work involved extracting hydrogen from water to make fuel" [read: deep read, article p.2]. The informant is not named. Capt. Oatman is paraphrased on timing, the 2008 incident and the search only. No Realm employee, Larson or Stehl is quoted. The only mention of the owner's side is a "friend" and a "relative" in photo captions, neither quoted in the body. The victim is not named in the article.
• So my earlier wording "authorities described the work as extracting hydrogen from water" (section 4, claim 6 heading, and Corrections item 4's context) was wrong. The authorities relayed it. If the informant was company-side, it is closer to an inventor-side description; if a bystander, it is neither. Unknown.
• A different attribution exists in a later article: a pressurised container "being used to heat water in an apparent attempt to break hydrogen and oxygen apart" is the authorities' characterization [read: research/2026-10-04-news-2010.md, via Investigator]. I have not re-read that article. It is the authorities' own description and stays labelled so.
• Incident record to carry (all single-source each, stated as such): Dec 2008 tank "imploded due to mechanical valve failure", Larson burned (Simi Valley Acorn, via Investigator); the VC Star recorded early reports of a hydrogen cylinder top blowing off in 2008; EPA POLREP mentions "reports of an additional explosion in Simi Valley in 2008" [web: https://response.epa.gov/sites/7236/files/SSI_polrep_2.htm, via Investigator]; June 2010 fatality; 9 Aug 2011 Sylmar; Fire Engineering (2014), citing a Strategic Sciences/EPA report, "at least four explosions", two of them unreported in 2008 [read: research/2026-10-04-news-2011-2014.md]. Causes: 2008 stated as valve failure by one source; 2010 undetermined; Sylmar injured two. Not independently reconciled.

10.2 Claim 3: wiring and waveform of the 5 V

• The deep read settles the QE's open question only as far as the text allows: the examples say "four parallel chambers" ([0058], [0059]); [0049] separates parallel arrangement of chambers (water, space) from the electrical connection, which may be "in series or parallel as may suit the application". The electrical wiring of the example is therefore not stated. The waveform is not stated either (continuous DC, duty-cycled DC, PWM or AC all appear as options in [0024]-[0026], [0051], [0054]).
• Consequence for my numbers [calc]: sections 2 (rows "Electrolysis energy at 5 V" and "Example generator power") and Claim 3 assume 5 V across every gap. That holds if the chambers are parallel (the plates within a chamber are two interleaved sets, so every gap sees the supply). If the four chambers were in series, each sees about 1.25 V. At 1.25 V, 2 F V = 241.2 kJ per mol H2, which is 99.8% of the lower heating value (241.8 kJ/mol), i.e. a real cell would produce essentially no gas at the 25 C reversible voltage of 1.23 V plus overpotentials (practical cells need roughly 1.5-2 V, [memory, medium]). So the two readings give opposite pictures: parallel, a gas-producing but inefficient electrolyser (25% LHV); series, almost no electrolysis gas, which would make the 35 psi stored vapour and the engine runs harder to explain by electrolysis. The text does not choose. I do not treat either as established. The Claim 3 verdict is unchanged because it already rests on "no meter, no energy figure"; what changes is that the 25% figure is conditional on parallel wiring.
• Faraday's law (moles of gas per coulomb) governs either way; pulsing changes average current, not energy per mole. This was never measured.
• Label fix adopted from the QE: "Safer than electrolysis hydrogen": Tested, not supported, with the note that the tests were uncontrolled incidents, not a designed experiment. Caveat I add: there is no baseline accident rate for conventional electrolysis hydrogen in the record, and the stored H2 + O2 mixture is a hazard of the mixture, not evidence about bond angles. The verdict is weak in the same way the underlying evidence is.

10.3 Claim 1: "intrinsic frequency" and the radiation embodiment

• Correction to the physicist: the description does name oscillating sources: AC "from a standard outlet" ([0024]); duty-cycled DC, 0.005-0.5, e.g. 2 ms on / 98 ms off, which is 10 Hz, from a MOSFET array or SCR network, preferably battery powered ([0026]); PWM DC of "unequal pulse spacing" ([0051]). The physicist's statement is true of the examples, which give only "5 volts".
• Correction to the QE: the QE says the "intrinsic frequency" has a concrete meaning, the pulse rate. The deep read finds that the patent never gives it a value or says so. Claim 1(a) ties it to "a voltage source ... providing current"; for continuous DC it is undefined; the 10 Hz figure is only an illustration of duty cycle. The most plausible meaning is the switching or mains frequency, as an inference, not a statement. The patent also says the two frequencies "summate" to a "modifying frequency", which has no equation and, for a linear system, no defined meaning (a sum of two frequencies is not a frequency; products give sum and difference terms only in nonlinear systems).
• Second stated mechanism, kept: [0050]-[0051] radiation of 0.1-100 cm wavelength or 0-5 GHz (1.1 and 2.2 GHz named). Photon energy [calc]: 1.1 GHz 4.5e-6 eV, 2.2 GHz 9.1e-6 eV, against the 0.198 eV bend quantum, a gap of 4.4e4 and 2.2e4. At 10 Hz or 50 Hz the gap is about 5e12 and 1e12. The energy-scale argument therefore holds for every frequency named. Liquid water does absorb strongly in the GHz range (dielectric relaxation near 20 GHz, [memory, medium], not rechecked), but that is heating of liquid, not bending a gas molecule. The Claim 1 verdict is unchanged.

10.4 Claim 2: the inventors' own applications list points two ways

• The deep read (0092540 section 10) gives the inventors' own wording. Entries that read as free O2: [0145] Aircraft Oxygen Systems, [0146] Firefighter Breathing Apparatus, [0147] Portable Oxygen Apparatus, [0155] Tankless Underwater Breathing Apparatus, [0169] oxygen systems from water or ice, [0191] Atmospheric Oxygen Enrichment. Free H2 or reductant: [0216] Down Hole Hydrogenation, [0115] High Purity Gases, [0118] Reductive Purification of Metals without Carbon, [0193] nitrogen fixation. Redox: [0119]. Cutting and welding ([0150], [0179]) parallel the torch example.
• Against that, the same list says the product is a fuel that is not burned: [0170] "Rocket Fuel - Non-explosive", [0171] "Vehicle Fuel - Non-combustive", and many entries present a water-vapour product (water treatment, refrigeration). The inventors distinguish their product from electrolysis in [0021] (O-H bonds stay intact). The deep reader notes the list is not reconciled. I record both: the QE called the free-O2 entries the inventors' implicit characterization; that is fair but one-sided. The list is applicants' text (interested party), the grouping is a reader's inference, and it states no composition. It is weak evidence beside the PCT concession and Cal/OSHA's finding, not a substitute for analysis.
• Claim 2 verdict unchanged: strong form Refuted by calculation; identity Never properly tested.

10.5 Claim 4 and Claim 5: what the CIP claims and what it does not

• Of the 23 claims of US 2009/0090312, only independent claim 1 recites air (metered by a second mass flow controller, mixed with the vapor and "a fuel"). Independent claims 2, 12, 13 and 21 are silent on air and on conventional fuel; claims 9-11, 15 and 16 require a fuel. No claim excludes air or says "without outside air" [read: deep read, appended section 2, B]. The air-free Honda mode is described in [0071] and the abstract only, with no data. The inventors describe air-fed carburetor and injector configurations in [0078]-[0085].
• No numerical engine data appear in pp.28-32: no power, rpm, fuel flow, vapor consumption, energy balance or independent test. The diesel 60:40 to 70:30 sentence ([0079]) is bare text.
• If the Investigator's report of granted claim 1 of US 7,793,621 (metered air and fuel) is right, it matches CIP claim 1: the grant, if so, covers an air-and-fuel engine method, not the air-excluded mode. I have not seen the granted text, so this remains unconfirmed. It does not change any verdict, but it separates what the headline abstract asserts from what the claims recite. Both stay in the record.
• Lean limit (QE): 1-2.5% H2 in air [calc, section 2] is below hydrogen's commonly cited lower flammability limit of 4% for upward propagation (higher, 7-9.5%, for horizontal or downward propagation) [web: https://www.aiche.org/sites/default/files/docs/pages/the_elemental_-_hydrogen_flammability.pdf, via search summary]. Retagged from [memory]. This strengthens the three-way reading of the oxygen budget (oxidiser-bearing gas, negligible power, or residual gasoline). The Claim 4 verdict is unchanged: Never properly tested.
• Claim 5 label fix adopted: "Tungsten melted": Never properly tested; bulk melting is above the oxyhydrogen adiabatic limit (3,075 K against 3,695 K) by calculation. The observation stays in the record as the one the arbiter said mainstream expectation does not clearly cover, with the oxide-burn-back explanation as my suggestion, not the inventors'.

10.6 Absence of evidence (section 6): corrections and added items

• Explanation 2, "Against": the Google Patents legal events show the 4th-year maintenance fee on US 7,793,621 paid 12 March 2014, and lapse effective 14 Sep 2018 [read: research/2026-10-04-google-patents.md]. So someone paid after the EP withdrawal (5 Jul 2013) and parent abandonment (3 Sep 2013). My earlier "still paying in 2013" understated this. Payer and assignee records would decide between a believer still funding the patent and a routine renewal.
• Origin, kept: Stehl's companies (BGX Technologies, Rainbow Technologies, Saranac Lake NY) raised money from 2001; Timothy A. Larson, Realm's founder and Tyson's father, was an investor, not accused of wrongdoing, and per the prosecutor (Dooley, an interested party) "believed in the science" and paid to ship BGX equipment to California after late 2005 [read: research/2026-10-04-news-2010.md]. This is the prosecutor's characterization of a person's belief; I do not treat it as evidence about the physics.
• Outside expert, kept: Dan Raftery (Purdue chemistry) said, from reports and the application and without seeing the apparatus, that the application did not appear viable and looked like electrolysis. Single source; one informed opinion, weighed in section 4 already.
• Added leads: hazardous and reportedly radioactive items found in 2015 on a Saranac Lake lot owned by Stehl (NCPR and nny360 snippets, not opened; a lead on the original process's materials); a Silver City, NM slag-pile application named in the DOJ release as a use of Stehl's energy source [read: research/2026-10-04-justice-gov.md; not opened by me]. Neither verifies any claim.

10.7 Corrections to section 5 (Step 1 files)

Add to the list in section 5: (8) claims.md and facts.json item for the article must read "Authorities were told ... (unnamed source)"; (9) record the series/parallel and waveform ambiguity and the intrinsic-frequency definition gap; (10) record the CIP claim structure (only claim 1 recites air); (11) the maintenance fee date above.

10.8 Still open after this round

• Granted claim 1 text of US 7,793,621; "Honda" versus "Hyundai" GX 390 in WO 2010/039283 (the patents under review say Honda). Both go to the next Investigator run, as the QE recommends.
• Who told the authorities the explosion was "water-based" (VC Star 18 June 2010 follow-up).
• Whether the example chambers were wired in series or parallel; only the inventors or the apparatus can answer.