Physicist review (Step 2): Larson-Tyson (Realm Industries)

Reviewer: PhD physicist role. Sources: facts.json, claims.md, deep/ reads of both patent applications (US 2009/0092540 A1 and US 2009/0090312 A1), sites.md "Leads for review", research/ notes (EPA, Google Patents). I re-read the deep-read transcriptions of the patent passages (Examples, Figs. 11-13 tables) and re-did the arithmetic below. I did not re-render the page images myself; figure values are taken from the deep reader's page-image readings [read: page image, via deep read].

Tag key: [read] from the project documents; [web] with URL; [memory] with confidence. Interested parties: the inventors and Realm Industries (commercial stake; all descriptions of the "vapor" are theirs); the EPA, Cal/OSHA and prosecutors (regulatory/legal stake; their description of "TyLar gas" is their characterization). Web verification: two of my budgeted checks (barrier to linearity of water; maximum CO2 of gasoline exhaust) returned no usable number, so the first remains [memory, medium] and the second I replaced with my own stoichiometry calculation (below). No replication or independent test of the bond-angle vapor was found by anyone; I state none.

1. The mainstream view, stated fairly

Water molecule. Free H2O has an equilibrium H-O-H angle of 104.5 degrees and O-H length 0.96 angstrom [memory, high]. The angle is not a stored "state" held in place by anything: it is the minimum of a smooth potential energy surface for the bending motion. The molecule vibrates in the bend at ~1595 cm^-1 (frequency ~4.8e13 Hz, period ~21 fs) [web: bend ~1590 cm^-1 gas phase, https://www.nature.com/articles/s41467-020-19759-w]. Even at room temperature the zero-point bending motion already spans many degrees about 104.5. A molecule with an angle "greater than 104.5" is simply a molecule caught at some moment in its bending vibration, or a vibrationally excited molecule. Pushing it to linear (180 degrees) is a saddle point costing roughly 1.3-1.4 eV (~11,000 cm^-1, ~130 kJ/mol) [memory, medium; web check did not return the number]. There is no barrier on the way back: an excited bend relaxes by collisions in picoseconds to nanoseconds in a gas at 1-100 psi [memory, medium]. It cannot be "stored for several months" in a steel vessel.

Liquid water and electrodes. Passing current between immersed metal plates is electrolysis: at ~1.23 V (reversible) and in practice more (overpotentials), water is split to H2 at the cathode and O2 at the anode [memory, high]. Gas leaves as bubbles. The patent's own text describes "gas bubbles forming on electrode surfaces and rising" and a "pressure vessel" collecting the gas [read: US20090092540 [0035]-[0042], deep read].

What a "resonating frequency" means. A resonance needs an oscillator (energy swapping between two stores, e.g. L and C). A resistor in parallel with a capacitor does not resonate; it has a time constant RC and low-pass behaviour. The patent's f = 1/(RC) is a corner/relaxation frequency, not a resonance [read: [0021]].

Energy. Splitting 1 mol of water needs at least 237 kJ (Gibbs) or 286 kJ (thermoneutral, 1.48 V) [memory, high; computed 2 F V]. Burning the H2 back returns at most 286 kJ (HHV). Any scheme that returns a water-based "fuel" to ordinary water cannot release more than was put in, and with any losses releases less.

2. Claim-by-claim

Claim 1: current through interleaved plates creates an "RC resonating frequency" (1-22 kHz) that, combined with the supply frequency, bends the H2O bond angle without breaking O-H bonds

a. Mathematics consistent? Mostly undefined rather than contradicted. Observations: (i) f = 1/(RC) has no 2pi and no stated unit (rad/s or Hz) [read]. (ii) For a parallel-plate cell the bulk R = d/(sigma A) and C = eps0eps_rA/d, so RC = eps_reps0/sigma: the geometry cancels. My calculation: eps_r = 80, f = sigma/(2 pi eps0 eps_r): sigma = 5.5e-6 S/m (very pure water) gives 1.2 kHz; 1e-4 S/m gives 22 kHz; tap-water-like 0.05 S/m gives ~11 MHz. So the patent's 1-22 kHz window corresponds to nearly deionised water, yet it says purity-dependent R and area/gap-dependent C [read: [0021]] and says water need not be purified. At realistic electrolysis conductivities the bulk relaxation frequency is MHz, not kHz; with double-layer capacitance (tens of microfarads per cm^2) the electrode RC would be far lower (sub-kHz to ~Hz). The statement that C depends on "applied current" has no basis. (iii) A DC (5 V) supply drives no oscillation in an RC; no oscillating source is identified, and "dual frequency... better than the sum" is asserted with no data [read: [0023]]. (iv) Claim 2 and [0019] contradict each other on which terminal is anode [read, deep read p.22 vs p.15]. (v) The EM-radiation alternative mixes "0 to 5 GHz" and 0.1-100 cm wavelengths (0.3-300 GHz) [read]. So: not a derivation, a list of assertions, with a mislabelled resonance.

b. Consistent with experiment? No. Scale test: the applied electric field in the example cell is 5 V over a 0.1765 in (4.5 mm) gap, E ~ 1.1 kV/m. The water dipole (6.2e-30 C m) times E divided by kT at 300 K is ~1.7e-6 [computed], so the field cannot bend molecules measurably. A kHz drive is ~10^9 times slower than the bend vibration (~48 THz), so there is no resonant coupling to the bend; bending is excited by IR light near 1595 cm^-1 (6.3 um), which is established spectroscopy [memory, high]. No spectroscopy of the product was reported by the inventors. Also "without breaking bonds" is directly undercut by the inventors' own later filing (below).

c. Tested? By the inventors only, and only indirectly (engine exhaust). No composition measurement, spectroscopy, calorimetry or control run (steam alone, or electrolysis gas from the same cell) exists in the documents [read]. No independent test or replication found by the Investigator.

d. If true? Water would have a long-lived metastable geometry isomer with no barrier on the standard potential energy surface; Born-Oppenheimer-level quantum chemistry and the whole measured H2O vibrational spectroscopy would be wrong. Collisional relaxation rates would have to be vastly slower than measured.

Verdict: Refuted by calculation (strong form: a kHz-driven, months-stable bent-angle molecule); Not testable as stated in its weak form (the word "unnatural angle" attached to a gas with no measured property). The weak form "the gas is H2/O2 made at the electrodes" is Established physics (see Claim 2).

Claim 2: the pressurised vapor is a distinct substance from steam or electrolysis H2/O2 and can be stored 1-100 psi for months

a. Math: no quantitative statement. The distinction rests on asserted "electrical and magnetic properties", never measured [read: [0021]].

b. Experiment: Ordinary electrolysis gas from a plate stack is the default explanation for everything reported. Inventors' own later patent WO 2010/039283 (PCT filed 3 Apr 2009) concedes the process "could result in the production of hydrogen gas (H2) and oxygen gas (O2)", that "covalent bonds may break", and renames the product "energy carriers... molecules with unnatural bond angles and H2 and O2 gas"; it drops the "safer than electrolysis hydrogen" sentence [web: https://patents.google.com/patent/WO2010039283A2/en, via Investigator]. This is the inventors' own hedge in a commercial filing; it is not proof either way, but it is the inventors' own record and I keep it. Government record: TyLar MSDS "states that the material consists of compressed hydrogen gas"; invoice said "compressed water vapor"; shipping papers listed an oxygen-hydrogen mixture and pure hydrogen; EPA/AERO treated it as detonable, "similar to Brown's gas"; two of three cylinders detonated when shot [web: EPA POLREP #2, https://response.epa.gov/sites/7236/files/SSI_polrep_2.htm]. Cal/OSHA found H2 and O2 stored in the same tanks [web, via Investigator: Daily News 2012]. These are regulators' and the company's own paper descriptions (interested parties on both sides); they match H2+O2, and "compressed water vapor" is not a property of an H2O gas at 100 psi and room temperature (see below).

Additional point I can check: pure water vapor at 1-100 psi gauge and ambient temperature cannot exist as a gas (saturation vapor pressure of water at 25 C is ~3.2 kPa ~0.46 psi [memory, high]); at higher pressure it condenses. A "vapor" of H2O molecules stored at 35 psi for months at room temperature would be liquid. So the stored gas is either not H2O, or not only H2O. Only a non-condensable gas (H2, O2) fits the stated pressures.

c. Tested: The one physical sample that could have settled it was never analysed: three cylinders destroyed unsampled on 15 Sep 2011 for safety reasons (EPA CID had asked for sampling; AERO advised against) [web: EPA POLREP #2]. Absence of evidence worth recording: this is a safety-driven loss, not evidence of concealment as far as the record shows; nobody else (university lab, Korean consignee) is on record as analysing it.

d. If true: gas-phase H2O would need a new isomer with weeks-months lifetime; molecular spectroscopy would have found it.

Verdict: Refuted by calculation (as H2O at 1-100 psi gauge at room temperature, vapor pressure argument) for "stable bent-angle water vapor"; Never properly tested in the narrow sense that the identity of the gas was never measured by anyone neutral. The simplest reading (electrolysis gas, H2/O2) is consistent with all documents and records.

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

a. Math: no energy-in or energy-out figure at all [read]. Reader's arithmetic from the deep read: the example generator has 4 x 20 plates x 4 ft^2 = 320 ft^2 (my calc of one-side plate area for 80 plates: 80 x 0.372 m^2); at stated 10-100 W/ft^2 that is 3.2-32 kW, at 5 V 640-6,400 A, but the patent states no current. Electrolysis at 5 V costs 965 kJ per mol H2 (2 F V) versus 286 kJ recoverable [computed]; so unless a very small fraction of the current goes to gas, a 5 V plate stack is less efficient than conventional electrolysis, not more.

b. Experiment: Combustion of H2 with O2 does release energy and makes water (no CO2): established. Whether the system as a whole is energy-positive: no data; the first law forbids over-unity, and the patent does not claim over-unity explicitly, only "less than electrolysis". The "reversion of angle" energy source is capped by the ~130 kJ/mol that a fully linearised molecule would hold (and would first have had to be supplied) [memory, medium].

c. Tested: no calorimetry, no power measurements.

d. If true (as the angle-energy mechanism): the energy released would be vibrational energy dissipated, i.e. just heat, not a fuel; it would not need ignition.

Verdict: split. (i) "Burning the gas releases energy and gives water": Established if the gas is H2/O2. (ii) "Energy from reverting bond angle; less energy than electrolysis": Never properly tested (no energy accounting reported); the mechanism part is Refuted by calculation as in Claim 1. (iii) "No consumption of atmospheric oxygen": see Claim 4.

Claim 4: the vapor runs engines on its own (Honda GX340/GX390 with no outside air; Toyota FJ-40 with gasoline shut off; exhaust O2 near 20%, CO/CO2/HC near zero; Figs. 11-13)

a. Internally consistent? Mostly, with errors noted by the deep reader: text says Fig. 13 gasoline phase is points 1-36 and CO "exceeding 5%", the graph shows to ~51 and ~10%; text says engine-off readings are "unchanged", graphs show an O2 step of ~0.3-1.0 points to ~21% [read: deep read]; Fig. 11 hydrocarbons plotted in % (normally ppm); CO flat at exactly 10.00% for ~8 min (probably analyzer range ceiling; [memory, low]).

b. Consistent with experiment? Check each observation against ordinary hydrogen (or HHO) fuel: - Gasoline-phase CO2 of 18-19%. Stoichiometric gasoline (taken as CH1.85) burning in air gives dry CO2 = 1 / (1 + 1.4625 x 3.76) = 15.4% [computed]. Readings of 18-19% (Fig. 12, Fig. 13 spike) exceed this and signal analyzer cross-sensitivity/calibration error or extra oxygen. The instrument's reliability is therefore in doubt in the very data used to prove the claim (the inventors' data, not mine, show this). - Vapor-phase O2 of 20.0-20.8%. If an engine burns H2 in intake air, each mol of H2 removes half a mole of O2. For x percent H2 in the intake, dry exhaust O2 = (21 - x/2)/(100 - x/2) [computed]: 1% H2 gives 20.6% O2, 2% gives 20.2%, 4% gives 19.4%. The graphs' 20.0-20.8% (0.3-1.0 below ambient 21.0%) thus match ordinary hydrogen burning very lean (~1-3% H2 in the air stream) and indicate that outside air was consumed. The abstract's "no consumption of any atmospheric oxygen" is not supported by the inventors' own plotted O2 deficit (this deficit is the deep reader's observation; the H2 quantification is mine). Note that for the FJ-40 nothing says the air intake was blocked; the vapor was injected "below the carburetor" [read]. - Residual CO2 0.2-0.6% (5-15x ambient 0.04%) and CO 0.2-0.4% in vapor phase. Consistent with leftover gasoline in the float bowl/manifold, oil burning, or sampling carry-over; an H2O-only fuel has no carbon. The switchover takes ~70 s, and a carburetor float bowl holds minutes of fuel at idle, so a ~1 minute "switch" is within expected residual-fuel behaviour [memory, medium]. - Honda engines with no outside air (Example 2). A stoichiometric H2/O2 premix (Brown's gas, 2:1) carries its own oxidiser, so an engine can run with the air inlet closed. That is exactly what the claim describes, and it is the expected behaviour of H2+O2, not evidence of an unusual substance. If the gas were only H2O molecules with bent bonds, there would be no oxidiser and no fuel-energy source except the bending energy. - Engine power balance. An FJ-40 2F engine (4.2 L) at idle consumes roughly 3-5 kW of friction/accessory load [memory, low] and its intake flow is of order 6 L/s at throttled idle [computed rough: 4.2 L x 350 cycles/min x ~0.25 volumetric]. Running that on H2 at ~25% efficiency needs ~4 kW / 0.25 / 241 kJ/mol = ~0.07 mol/s, ~1.5 L/s H2 at STP, i.e. ~20% of the intake, far more than the 1-3% that the O2 readings imply. So the exhaust O2 readings are only consistent with a very low power output (or with another, unreported, fuel source such as residual gasoline, or a gauge/instrument problem). This is a rough order-of-magnitude check, low confidence, not a refutation. - No rpm, load, flow, power, voltage or current logged. The statement "The engine was still running" rests on the text alone [read]. The graphs cannot distinguish a running engine from air pumped through a non-firing engine (engine-off readings differ only by a 0.3-1.0 point O2 step) [read: deep read]. - The allowed claim. Granted US 7,793,621 B2 (14 Sep 2010) claim 1 requires metering air with a second mass flow controller and mixing vapor, air and a fuel [web: https://patents.google.com/patent/US7793621B2/en, via Investigator], i.e. the allowed form is a hydrogen-style dual-fuel use with air. The patent office's reading of the "no air" claim is unknown; the file wrapper (office actions leading to abandonment of the parent 2013-09-03 [web, via Investigator]) was not read. I do not infer that the office rejected the physics.

c. Tested? By the applicants, three times, no named witnesses, dates, places or independent verification [read]. No replication by others found. The data are consistent with an ordinary H2-containing gas, so the experiments as reported do not discriminate.

d. If true: a self-sustaining engine on a gas made from 5 V electrolysis-like input would have to produce more mechanical work than the electrical input, contradicting thermodynamics. Not claimed numerically, so it is not strictly a claim of over-unity.

Verdict: Tested, not supported (as evidence of a new substance or of oxygen-free, air-free, fuel-free operation): the applicants' own tests are uncontrolled, with an internally implausible instrument reading (CO2 above 15.4%), and their data are consistent with lean H2 combustion. The weaker statement "a plate-stack gas can partly fuel an engine" is Established (hydrogen supplementation is well known). The no-external-air Honda result is Plausible, untested for H2+O2 premix, with no data given.

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

a. Math: none. b. Oxyhydrogen flame temperature is ~2,800 C in oxygen [memory, medium]; tungsten melts at 3,422 C [memory, high]. A #00 tip at 5 psi gives a very small flame. An H2/O2 torch cutting thin sheet is ordinary and routine for HHO torches (a Brown's gas product, used for welding thin material); melting a tungsten rod with a small oxyhydrogen flame at 2,800 C is not obvious: tungsten melting exceeds the flame temperature, so the claim, if accurate, is surprising for plain H2/O2 and no flame temperature, rod diameter or time is given. Possible: a thin rod tip melting locally in an oxygen-rich flame, with oxidation (WO3 volatilises) mimicking melting; the report is qualitative. I flag this as the one reported observation that mainstream expectation does not clearly cover, and I keep it in the record. c. No replication or measurement. d. If true for H2/O2: nothing. If it needs more than ~2,800 C: a hotter-than-H2/O2 flame would need an extra energy source. The diesel statement ("in test experiments... still allowed the diesel to combust upon compression") has no data. Dual-fuel hydrogen/diesel operation with large hydrogen fractions is known in engine research [memory, medium]; it is consistent with ordinary H2.

Verdict: Plausible, untested (cutting thin metal with an H2/O2 flame is ordinary physics; tungsten melting is unmeasured and possibly overstated); diesel ratio: Not testable as stated (no data).

Claim 6: explosion on 17 June 2010 killed a person; Dec 2008 hydrogen cylinder incident; authorities described the work as extracting hydrogen from water (independent record)

Not a physics claim; I weigh it as evidence about what the product was. Documented: the victim was co-inventor Tyson Larson, 28 (Simi Valley Acorn 25 Jun 2010 [web, via Investigator]); Cal/OSHA found three containers used to generate and store pressurised fuel failed and issued $5,685 citations (appealed); later Sylmar explosion on 9 Aug 2011 injured two; Cal/OSHA penalties $540,890 in 2012, including that hydrogen and oxygen were stored in the same tanks [web, via Investigator: AP/Daily News; EPA POLREP #2]. Characterisations come from regulators, prosecutors and press; the company's own descriptions (stehl attorney: patents bought in 2005) are the other side. The injuries and fatality are consistent with stored premixed H2/O2 (detonable, static sensitive) and are not consistent with the patent's "safer than electrolysis hydrogen" claim: a gas that detonates in storage is behaving as H2/O2 would. The cause was recorded as "accidental but undetermined" [web, via Investigator]; I do not claim the cause.

Federal fraud case (NDNY, 2010-2015): Rossignol convicted 2014, 20 years; Stehl entered an Alford plea 2 Sep 2014 [web: DOJ press release linked in sites.md]. Prosecutor said the case was "not alleging that this is some kind of fraudulent scientific process" (VC Star 21 Jun 2010, via Investigator), so the legal record does not itself judge the science; and the charges concerned earlier Stehl ventures. Both facts are kept.

Verdict: Established as an event record (per press, Cal/OSHA, EPA); as evidence about the gas, it supports the "H2/O2 mixture" reading and does not test the bond-angle claim.

3. The Investigator's findings, weighed

1. Granted CIP US 7,793,621 B2 (lapsed 2018): a grant for a method of powering a vehicle with the vapor + metered air + fuel. Patent grant is not a verification of the physics; USPTO examiners do not require demonstrations of operability beyond enablement standards [memory, medium]. Keep.
2. WO 2010/039283 "energy carriers": inventors' own hedge conceding H2 and O2 may be the product. Strongest single piece of evidence in the record on Claim 2, and from the inventors themselves. Also read fairly: it is consistent with a patent lawyer's belt-and-braces drafting; it does not state the inventors abandoned their belief.
3. Family abandoned/withdrawn (US parent abandoned 2013, EP deemed withdrawn 2013, CN/RU/JP etc.): the examiners' technical objections are unread. Absence of evidence worth analysing; competing explanations in sites.md (commercial failure, injuries/prosecution, relocation) remain open. I cannot say which.
4. R2D2 LLC WO 2023/055615 and H2Plus (angle 94-104 degrees, "stable above 300 psig", made with crossed E and B fields; examiner cited Realm). Same concept re-claimed 14 years later by others; this is independent re-use of an idea, not independent confirmation, and I found no measurement data for it. Same physics objections apply (no barrier-stabilised isomer, vapour pressure argument). A bond angle below 104.5 for a stable gas-phase H2O is equally unsupported.
5. EPA cylinders destroyed unsampled; MSDS says compressed hydrogen; invoice says compressed water vapor. Kept under Claim 2.
6. Korea export / consignee, Nevada filings, CD Cal exhibits, Cal/OSHA files, USPTO and EPO file wrappers, PESN article: not yet researched. These are the records that could decide between "ordinary electrolysis gas" and anything else; I flag them as unread.
7. Stan's Legacy (Meyer archive) cites Realm as prior art; Brown's gas/HHO classification by EPA/Cal/OSHA: related work category. Stanley Meyer's "water fuel cell" claims were found in 1996 by an Ohio court to be "gross and egregious fraud" [memory, medium; not re-verified]; relevant as the same genre (electrolysis gas with claimed special properties), not as a judgement on Realm.

No replication, independent test, witness measurement or favourable independent theory of the bond-angle vapor was found; this is a statement about what was found, not about what exists.

4. Corrections to Step 1

1. claims.md Claim 4 states the Honda engines "ran with outside air excluded" and that the FJ-40 "kept running with gasoline shut off... in three runs". Reading: the Honda no-air result has no exhaust data at all (Examples are text only), and the three FJ-40 runs are Figs. 11-13; claims.md blends them. Clarified.
2. claims.md says the patent "reports no measurement of the vapor's composition" and facts say the same: confirmed. But the Step 1 claim implies engine-off readings equal vapor-phase readings; the graphs show a 0.3-1.0 point O2 step (deep read already notes it) and the deep read for 0090312 and 0092540 give slightly different numeric readings (e.g. 21.0 vs 20.9, Fig. 12 hydrocarbon omitted); I treat both as approximate (+/-0.3).
3. claims.md Claim 4 and sites.md say "CO plateaus at 10% (possible analyser limit)". Confirmed from deep reads; I add that Fig. 13 also hits 10.0% for samples 45-49, and the CO2 of 18-19% is above the 15.4% stoichiometric dry maximum I computed, not just "usual".
4. facts/claims Claim 1 says "f = 1/(RC), 1-22 kHz": I verified that this equals ε/σ scaling for pure water (1.2-22 kHz needs conductivity of 5.5e-6 to 1e-4 S/m) and is geometry independent; Step 1 did not note this.
5. claims.md note that the patent is "an application; no grant stated" is now outdated for the CIP (granted 2010, per Investigator).
6. The deep reader's reader-arithmetic "320 ft^2 total plate area" counts one face per plate; the plates are interleaved so electrode area facing each other is roughly 80 plates x 4 ft^2 = 320 ft^2 only if each plate is one-sided. Minor; not load-bearing since no current is stated.
7. Nothing found contradicting Step 1's quotations of the examples.

5. Summary of verdicts