TernionAG: Science Contrarian review

Role: Science Contrarian per prompt.txt. I judge by experiment, separate "tested and failed" from "never properly tested" from "dismissed without testing", build the strongest honest case, and apply the same rigour to the inventor and to the essay in the project folder. Tags: [read] source document, [calc] my arithmetic, [web] URL, [memory] recalled (confidence).

Source gap (repeated from Step 1)

Every Kassner/Ternion scientific claim rests on the patentee's own text. The folders hold no lab notebook, experimental report, independent test, replication or biography of Kassner, and no Ternion technical record (the Glarus register papers are corporate only). The patents are page scans (OCR); equations and tables were checked against page images only in the deep reads. Not read in depth: GB483062 pp.8-12, drawing sheets, GB448418 (not in folder), German priority applications, the Jacobsen Paperclip book and NARA guide (digest only). The three Morray-TH notes are secondary AI analyses, not evidence. Web verification: the session's web-search budget was exhausted (200 of 200) when I tried the three checks my verdicts lean on (Grundler/Keilmann replication, water Debye relaxation time, the 1945 Goodrich/Firestone purchase). One WebFetch (https://en.wikipedia.org/wiki/Dielectric_heating) returned only that water's optimum absorption is "around 10 GHz" and ovens run at 2.45 GHz, and nothing on non-thermal effects. All other [memory] items below are therefore unverified and marked so. Any conclusion on biological, preservation or colloid effects rests on the patents' own assertions, my arithmetic and [memory].

What I read myself

GB417501 deep read (pp.3-6, page images, with US2089966 and GB sibling comparison), US2089966 deep read (all 10 pp.), the Voss 1987 column deep read (including the cited passage, US2089966 printed p.6 col.1 l.61-71), the three 1945 newspaper clippings (page-image deep read), GB483062 deep read (Bessel tables, 40 cm disc), GB448419 tables, and the Morray-TH note 14 audit. I re-did the key arithmetic below. I also read the superseded _merged/Kassner-EEW-Ternion/contrarian.md as input to check; where I reuse its numbers I recomputed them.

Corrections to Step 1

1. Claim 4 / claims.md: the "factor of about 3" between the 1932 and 1938 tuning is correct for the 39-40 cm disc (GB417501: 39 cm, 1 cm gap, 116 cm; GB483062: 40 cm, 1 cm gap, 32.7 cm, a factor of about 3.5) [read: GB483062 deep read]. But the claim text says "lambda about 3 times the diameter ... disagree with it ... by a factor of about 3" without saying the 1932 numbers are a pure scaling rule (lambda/D = 3.00, 2.97 and 3.0 across three plate sizes at constant 1 cm gap) rather than Bessel theory [calc, deep read]. Step 1 should state that the 1932 figures match no Bessel root in the patentee's own tables (smallest listed root 1.84) and that dielectric loading cannot repair it: a 1 mm water layer in a 1 cm gap gives series epsilon about 1.11 [calc: 1/(0.9 + 0.1/80)], raising lambda about 5 percent, not 3-fold.
2. Claim 13: "the patent does name temperature and pressure as tuning handles" is thin. In US2089966, pressure appears only in claim 14, a measurement claim; the tuning handles in the body are time lag (temperature, concentration) [read: US2089966 deep read, section 5 table]. Temperature is a genuine handle; pressure is barely one.
3. Claim 3: the 28-day milk, 2-week veal and 4-month apple-juice figures are in GB417501's complete specification only. The US2089966 text is qualitative ("weeks versus days") and gives no powers or quantities [read]. The claim should say which document each number sits in. Also, the temperature rises the patent reports (2-8 C) are consistent with its own powers: milk 574 W x 120 s = 68.9 kJ into about 2.06 kg gives at most 8.6 K, reported about 2-3 K [calc; patent figures from the deep read of the superseded review, not re-read by me from the image]. Step 1 does not record that the numbers pass this check, which matters for judging whether the data look fabricated (they do not).
4. Claim 7: "bought basic electronic-vulcanisation patents" is right, but the clippings say only that the patents "originally were granted on discoveries made by" Dufour, Le Duc and Kassner; no seller, price, patent number or frequency [read: page image]. Step 1 correctly hedges, but the claim's framing "commercially significant" quietly equates a patent purchase with validation of Kassner's resonance mechanism. The clippings describe ordinary heating ("high-frequency oscillations shake the molecules ... creating uniform heat", cure in one-eighth of the time for sponge rubber [read: page image p.2, p.1]). Which Kassner patents were sold is unknown; they need not be the Ternion patents.
5. Claim 8: claims.md says "deleted from the register in August 1952". The deep read of note 14 notes the essay says Ternion "dissolved with no reasons provided" in 1952, while the register shows liquidation by resolution of 20 Dec 1937 for lack of funds. Step 1 already folds this in; I confirm it and add that the Dipole A.-G. Arbon loan account (balance sheet) is a lead nobody has followed: the research may have been done by a separate Arbon company, not Ternion itself [read: note-14 deep read item 3].
6. Claim 9: Step 1 is right that Bauch's IWM interviews never mention Ternion. I add that the sole source (a 1996 memoriam sentence, author unnamed in the folder) is secondhand and fifty years after the event; "the Ternion project at GEC" could equally be a courtesy bench job. Treat as one uncorroborated sentence, not a documented collaboration.
7. Claim 2: correct exponent (10^-9, not 10^-8), confirmed by me against the deep read of page image p.6. Add that GB417501 admits electrets were already known; US2089966 omits that sentence [read]. The 10^-9 C/cm^2 charge is modest compared with typical wax electrets (10^-8 to 10^-7, [memory, low, unverified]).
8. Missing from Step 1: Voss's rebuke is hedged ("sketched" in quotes, "perhaps that is the key word"), and he states the patent's object is not heating while diathermy is cited as prior art by Kassner himself [read: Voss deep read, GB417501 definitions]. This should limit claim 7.

Claim-by-claim

Claim 1: sharp natural frequencies of dipolar substances (electrolyte bands about 2% wide, serum about 90 cm, water 1.28-1.5 cm)

a. Mathematics. The Debye part is internally consistent. Debye's relaxation time tau = 4 pi eta a^3 / kT [read: GB417501 complete spec]. With eta = 1.0 mPa s, a = 1.4 A, T = 293 K: tau = 4 pi (1.0e-3)(1.4e-10)^3 / (1.381e-23 x 293) = 8.5 ps [calc]; loss peak at 1/(2 pi tau) = 18.7 GHz, lambda = 1.6 cm, matching the patent's "about 1.5 cm". The weakness is conceptual: Debye relaxation is overdamped (no Q, no resonance), yet the patent also calls the band a "natural frequency" at which the substance is "in resonance" and tunes it by temperature and concentration. The patent's own split into friction dispersion (broad) and "resonance dispersion" in conducting liquids (2% wide) avoids the contradiction only by asserting the second kind with no equation. The three "characteristic periods" are announced in US2089966 "explained more fully below" and never explained there [read: US2089966 deep read, section 1]. b. Experiment. Water's main dielectric relaxation is in the 10-20 GHz range at room temperature (the only web result I got, https://en.wikipedia.org/wiki/Dielectric_heating, says optimum absorption by water is "around 10 GHz"; it does not give the peak or the loss shape) [web partial]. That the 1.5 cm figure is right is therefore consistent, [memory, high; value 8-9 ps not web-checked]. Sharp (under 5% half-width) bands at 0.3-1 m in serum or gold-sol solutions are not what standard dielectric spectroscopy of electrolytes shows (conduction loss, broad protein and bound-water dispersions) [memory, medium, not web-verified]. Counter-evidence a contrarian must put on the table: claims of narrow microwave resonances in living cells (Webb and Booth on E. coli; Grundler and Keilmann on yeast, about 42 GHz) exist in the 1970s-80s literature [memory, medium; replication status unknown to me, web check failed]. They are in a different band (tens of GHz) and not Kassner's. c. Tests. None on record. Kassner reports detecting serum resonance by polarised-light changes; no spectra, no subjects, no controls. d. If true: electrolytes and serum would show narrow dielectric absorption lines at hundreds of MHz-GHz and a new ionic-resonance mechanism would be needed; Debye theory stays intact. Verdict (split): water's Debye band near 1.5 cm = Established. Sharp "resonance dispersion" in electrolytes/serum = Never properly tested by a neutral party (not "dismissed without testing": no document records anyone being asked to test it). Standard dielectric data as I recall them argue against, but I could not verify that.

Claim 2: a resonant field permanently changes "energy content" (orientation, double refraction, frozen-in "richer in energy" electrets)

a. "Energy content" change is defined by exclusion (anything not merely temperature rise) [read: GB417501 p.6], which is unfalsifiable as written; the "10-20 times smaller" electrostatic field assist is asserted without calculation. The electret numbers are self-consistent: 2e-9 C/cm^2 = 2e-5 C/m^2 on a 1 mm wax slab of relative permittivity 2.5 stores about 9e-3 J/m^2, about 0.9 microjoule/cm^2 [calc, from deep read]. That is the entire "extra energy"; it is ordinary electret physics and contradicts nothing. b. Field-induced orientation, Kerr birefringence and electrets (Eguchi 1919-25 [memory, medium]) are real. The 12 kV DC field used in the electret recipe does the poling; no experiment isolates the RF component. c. Tests: none; no run without the RF field is reported. d. If true: a non-thermal storage channel at photon energies about 7e-6 eV (18.6 cm: E = hc/lambda = 1.240e-6 eV m / 0.186 m = 6.7e-6 eV [calc], about 3,900 times below kT = 0.0259 eV at 300 K). Only a coherent collective mechanism (Froehlich-type [memory, medium]) could do this; not established. Verdict (split): "energy content" as defined = Not testable as stated. Narrow form (resonance-specific orientation/birefringence with controls) = Never properly tested. Electret storage of about a microjoule per cm^2 = Established (conventional).

Claim 3: practical effects (preservation, colloids, catalyst-free reactions, seeds, "profound effects in men and animals")

a. Energy budget [calc]: milk, veal, juice show 2-5 K rises against 4-9 K maximum from stated powers and times, so the reported heating is arithmetically believable. That shows the data are not obviously invented. It also shows exposures at 2-5 K above ambient are far below pasteurisation (about 63 C for 30 min [memory, high]); any preservation is therefore not bulk thermal pasteurisation. Hot-spot heating in a 1 cm plate gap carrying hundreds of watts is not excluded by a bulk temperature, and no field map or probe is described. Gold sol: 0.1 W x 3600 s = 360 J in a layer of 6.2 cm diameter x 1 mm (3 cm^3) is up to 120 J/cm^3 if absorbed, about 29 K if all went into 3 g of water [calc]; temperature control is not described, so a thermal cause of red-to-orange is not excluded. The "preservation by reversing colloid ageing" mechanism does not address bacterial growth or enzymes. b. Dielectric heating curing polymers, lacquers and rubber is established and thermal [memory, high]. Non-thermal RF preservation at these powers is not an established effect [memory, medium, unverified]. The 1945 purchase by Goodrich and Firestone shows industry paid for RF vulcanisation patents; it supports heating, not resonance. c. Tests: none on record. Each example is one run: no sample counts, spoilage criteria, storage temperatures or off-resonant, heat-matched and blinded controls. Bean seeds "twice as much" after four weeks has no n. The patent pre-excuses variation ("beneficial results ... must to some extent vary with the initial conditions"), a hedge against failed replication. d. If true: non-thermal bioeffects at 10^-5 eV photon energy would overturn the working premise behind RF exposure limits (thermal-only below heating thresholds) [memory, medium]. Verdict (split): RF/dielectric curing, vulcanisation, lacquer drying = Established (thermal). Resonance-specific non-thermal preservation, growth promotion, catalyst-free chemistry, "profound effects in men and animals" = Never properly tested. Neither tested-and-failed nor suppressed is documented. The experiment is cheap, so the absence of a test says something about the programme's fate (see claim 8), not about physics.

Claim 4: plate-pair resonator whose modes follow Bessel zeros (10 cm disc at 26.2 cm); 1932 tuning figures

a. GB483062 is correct radial-line theory: lambda = 2 pi R / x. For R = 10 cm, x = 2.405: 26.1 cm [calc], matching the quoted 26.2. The inconsistency lies in the 1932 numbers. 6.2 cm / 18.6 cm, 39 / 116 cm and 10 / 30 cm all give lambda = 3.0 D at a 1 cm gap. The lowest Bessel root in the patent's own tables (1.84) gives lambda about 3.4 R = 1.7 D for an open disc in its lowest mode [calc: 2 pi R/1.841 = 3.41 R], so 3 D (x about 1.05) matches no root; reaching it needs epsilon about 3 across the whole gap, whereas a 1 mm water layer in a 1 cm gap gives about 1.11 [calc]. GB483062's 40 cm disc at 1 cm gap resonates at 32.7 cm (x = 3.83), 3.5 times shorter than GB417501's 116 cm for the 39 cm disc; both were said to be calculated or confirmed by experiment [read]. And the 0.5 cm gap result (20 cm disc, 116 cm) implies lambda roughly scales with d/gap, not the square-root capacitor law (factor 1.93 versus 1.41) [calc]. b. Nothing in measured physics prohibits plate resonators; the problem is internal. Possible reconciliations, none documented: the 1932 oscillator is a tube-loaded tank (the plates are part of a longer feeder), the "wavelength" refers to the driven circuit, or the diameters are different quantities. Any of these would make "resonance of the sample" and "resonance of the apparatus" indistinguishable, which is the contrarian's real complaint: the sample's claimed resonance was only observed through the plate size chosen. c. Tests: none independent. GB483062 was "void / not accepted", an administrative outcome. d. Nothing in mainstream physics would change. Verdict: Internally inconsistent for the 1932 tuning figures versus the 1938 Bessel theory as stated (neither document reconciles them); the 1938 radial-line theory itself = Established. The sample-resonance interpretation = Never properly tested.

Claim 5: valves, coaxial resonators, wavemeter, polarimeter generate stable powerful quasi-optical waves; "about 40 percent" efficiency, 0.30-4.9 m

a. Tables in GB448419 are the patentee's; lambda/l ratios (1.5-1.9, falling to 7.9-12.9 for the loaded line) show lines are loaded, not simple quarter-wave lines [calc]. Anode is held negative against a positive grid (Barkhausen-Kurz regime) [read: deep read; interpretation memory, medium]. Efficiency 40 percent for space-charge oscillation tubes is large for 1930s decimetre sources [memory, low]; no measurement method (calorimetry, anode dissipation) is shown. b. The patent itself cites Barkhausen, Kurz and Hollmann; decimetre generation was real by 1934 [read, memory medium]. c. Independent evidence: one memoriam sentence that Bauch worked a Ternion UHF/decimetre project at GEC in 1934-35 [read]; no result, no performance figure. d. Nothing changes. Verdict: general feasibility = Established; specific power, stability and 40% efficiency = Plausible, untested.

Claim 6: Tesla-discharge applicator (from 0.5 mm) and later lamps raise efficiency or match absorption bands

a. Not re-read by me in depth beyond Step 1/deep reads. A spark gap is broadband and damped; "waves from 0.5 mm" is not in conflict with that; a narrow line at 0.5 mm is. No efficiency, spectrum or dose is given. b. Plasma and space-charge oscillations in discharges are real [memory, medium]; no data tie them to luminous efficacy. c. Tests: none in folder. US2104697 is assigned to Universal Inventions Ltd (London), not Vibranic [read: Step 1 correction]; a 1951 follow-up (US2558568) shows continued interest but not performance. d. Nothing changes. Verdict: radiation = Plausible, untested; therapeutic or efficacy gains = Never properly tested.

Claim 7: significance (Goodrich/Firestone purchase 1945; Voss 1987)

a/b. The clippings are consistent with the heating process and give one number: sponge rubber cured in about one-eighth of conventional time, "company engineers said" [read: page image p.1]. Voss's column is an editor's remark; he found the passage (US2089966 printed p.6, left col. l.61-71) "akin to the summary of a 1986 IMPI Conference session" [read]. Kassner does disclose multi-frequency exposure (claim 7 of GB417501), so "dual-frequency applicator" is a fair description of the disclosure [read]. But Kassner states his object is not heating and cites diathermy as prior art [read], so "foresaw the microwave oven" is an overreach. c. No test needed; a priority-of-concept question. Prior-art status is untested: did examiners in the 1980s know of the patent? Voss asks "Or do they?". d. The history of RF heating gets a footnote. Verdict (split): the 1945 purchase happened as reported = Established (secondary press, three clippings; seller and patent numbers unknown); the claim that it validates Kassner's resonance mechanism = Not testable as stated; Voss's "foresaw" = Historically superseded but instructive (early disclosure of multi-frequency applicators, not a working microwave oven).

Claim 8: Ternion AG company record (1932 founding, CHF 1.12 M research account, 20 Dec 1937 liquidation, 1952 deletion)

a. Accounts: research account CHF 1,120,380 against share capital CHF 100,000, loans of 1,016,535 and 80,000 [read: note-14 audit quoting translated register papers]. Internally consistent: a holding company financing research by loans eventually insolvent. b/c. Primary register copies certified 2017; the liquidation text blames lack of funds, lapsed patents unpaid, and "impossibility of industrial utilization" of the rest [read]. That is documented, not suppression. Only US1951731, GB417501 and GB417564 tie Kassner's patents to Ternion; US2089966 names no assignee, and Kassner is never named in the company papers [read]. The 1945 sale suggests some patents were later worth something, so the claim "no industrial use" is only true for what Ternion held. d. Nothing. Verdict: Established (documentary history), with open items: who owned and sold the patents and what happened to Dipole A.-G. Arbon.

Claim 9: Bauch did a Ternion UHF project at GEC, 1934

Documentary: one sentence in the 1996 memoriam [read]; the GEC secondment itself is corroborated by his IWM interview, his wife's interview and Swiss Sound 1983, none naming Ternion, Kassner or UHF [read]. Not testable by experiment; evidential only. Verdict: secondment = Established; Ternion project = Not testable as stated (uncorroborated single sentence); the essay's "employee of GEC ... collaborated jointly" overstates it.

Claim 10: engineered fields raise/lower covalent and ionic bond strength

a. Not in any Ternion text, whose terms are "energy content" and orientation [read]. Photon energy 6.7e-6 eV at 18.6 cm versus bond energies of 1-5 eV: five or six orders of magnitude, and versus kT about 3,900 times lower [calc]. No equation in the essay links a field to a bond energy. b. A strong static field can distort bonds (Stark, field ionisation) at about 1e9-1e10 V/m [memory, medium]; the plates carry fields many orders lower (230 W into a 1 cm gap gives volts per cm to kV/cm [calc, rough]). c. Never tested; no experiment even proposed. d. Would require coherent energy accumulation into one bond mode, with no known path. Verdict: Refuted by calculation in the single-photon form; the coherent/collective form is Not testable as stated (no mechanism or quantity given).

Claim 11: electrolysis-and-burning cycle or "chemical heat pump" yields net energy

a. Closed-cycle arithmetic: dG = 237.1 kJ/mol, dH = 285.8 kJ/mol are state functions of H2O -> H2 + 1/2 O2 [memory, high]; electrolysis then combustion returns exactly the enthalpy used, so any claimed net gain from a closed cycle must come from outside the cycle or violate energy conservation. If the fields lowered the energy needed for electrolysis, the same field lowering would lower the energy released on burning (detailed balance), so the net is not gained [calc]. A heat pump drawing heat from the environment could give COP > 1 but only by consuming work; no mechanism is offered. Never claimed by Ternion; the essay itself says the naive cycle "makes no sense" in mainstream science [read: note-14 audit item 34]. c. No test. Verdict: closed-cycle net energy = Refuted by calculation; "chemical heat pump" with COP > 1 = Not testable as stated (no claimed device or measurement).

Claim 12: weaponised binding energies; defence interests suppressed Ternion

a. n/a mathematically. b. Evidence offered is absence. The record shows the plainest explanation: liquidation for lack of funds in 1937, 45 years before the Marconi deaths (1982-91); none of the Marconi articles mentions Ternion or Bauch [read]; Bauch left GEC about 1936 [read]. Sharif's death in the project article is a ligature, Sands crashed into a cafe: details in note 14 are wrong [read]. BAE is not owned by American General Electric (GEC plc is British; GEC defence merged with British Aerospace in 1999 per the GEC Wikipedia page in the project) [read]. c. Not testable by experiment; archival questions only (GEC/Marconi corporate records, TNA). d. Nothing in physics; would require documents. How the claim was treated: no document in the folders shows blocked publication, reanalysis, omitted data or withheld funding. The commercial failure is documented. Verdict: Not testable as stated (no documented link); the specific facts offered (BAE owned by US GE) are Refuted. I make no claim either way about suppression beyond saying the record shows none.

Claim 13: effects hard to replicate because natural frequencies shift with temperature and pressure

a. Debye tau depends on viscosity and temperature, so a Debye band does shift (about 3% per degree for water, [memory, medium]); the patent's own tuning handles agree. Pressure appears only in a measurement claim (correction 2). b/c. No replication attempt of any kind is on record, so "hard to replicate" is not a finding. The explanation is unfalsifiable until a tuning protocol is stated; note the patents already pre-excuse variation. d. Nothing. Verdict: Not testable as stated; a falsifiable version exists (measure the dielectric spectrum of each sample at its temperature first).

Strongest honest case for Kassner

1. The theory starts from correct science: Debye 1929, water's band near 1.5 cm, tuning by viscosity and temperature, and the correct statement that electrostatic orientation needs 100-200 kV [read, calc].
2. The patents' reported heating (2-5 K) matches stated power and time to within a factor of 2-3, and he reports heating alongside the claimed effect rather than hiding it. Fabrication would be easier than reporting a lukewarm rise.
3. The "frozen-in orientation, released as radiation" idea is a testable and, for electrets, modestly true statement; he flags that electrets are already known (GB).
4. His multi-frequency applicator disclosure and RF vulcanisation list predate the field by decades and were reasonably identified by Voss's journal 50 years later; industry bought the basic vulcanisation patents in 1945.
5. The decisive experiment is cheap and was never recorded as done. That is the signature of "never properly tested", not "tested and failed".
6. The idea of frequency-selective biological interaction was pursued later by respected scientists (Froehlich theory; Webb, Grundler, Keilmann) [memory, medium, not web-verified]; replication unknown to me.

Where the researcher and the essay are weak (same rigour)

1. "Energy content" and "natural frequency" are defined so that resonance is true by definition (a natural frequency is "the range of anomalous dispersion").
2. The 1932 tuning figures contradict the 1938 theory by about a factor 3 and are never reconciled; the sample's resonance is only detected via the apparatus's own size.
3. No controls, sample sizes, statistics, storage temperatures or spoilage metrics. "Optimum" wavelengths quoted to three figures without tolerance.
4. The US text is weaker evidence than GB417501 (it drops quantities and the electret-prior-art admission).
5. The essay: bond-strength engineering has no equation; the closed-water cycle violates the first law; "BAE is owned by American GE" is wrong; the 1952 "no reasons" misses the documented 1937 liquidation minutes; the "suppression" case is built from absence.
6. Note 14 (secondary): "47-year-old Johnson" is the street number of 47 Lincoln's Inn Fields; Bauch did not go to Switzerland per any source; suppression-by-channeling contradicts the insolvency record.

Single experiment to settle claims 1-3

Identical, temperature-logged 2 L aliquots of raw milk, apple juice and gold sol, in a field-mapped plate oscillator: (A) resonant exposure per the patent, (B) off-resonant with equal absorbed power, (C) heat-only matched to A's measured temperature history, (D) untreated. Blind scoring (bacterial counts, pH, UV-vis) at fixed times over 28 days, and a broadband dielectric spectrum (0.1-5 GHz) of each liquid beforehand to see whether any narrow line exists at the claimed wavelength.

Verdict summary

• Established: water Debye band near 1.5 cm; RF/dielectric curing and vulcanisation (thermal); electret storage at about microjoule/cm^2; plate-resonator Bessel theory (GB483062); the Ternion company record; the 1945 patent purchase as a reported event.
• Never properly tested: sharp electrolyte/serum resonance bands, resonant non-thermal preservation, growth, colloid and catalyst-free effects; therapeutic lamp and Tesla-applicator claims; specific tube power/efficiency (Plausible, untested).
• Internally inconsistent: the 1932 plate-oscillator tuning versus the 1938 Bessel theory.
• Refuted by calculation: single-photon bond-strength change; the closed electrolysis-and-burning net-energy cycle; the BAE-owned-by-US-GE statement.
• Not testable as stated: "energy content", the suppression thesis, the replication-difficulty excuse, Bauch/Ternion link.

Investigator findings (2026-10-04)

Scope: I read every "Leads for review" entry in sites.md (leads 1-29), every research/ note (2026-10-03 and 2026-10-04 files) and the saved files bearing on the claims (OCR of US 2,094,602, 2,096,459, 2,415,025, 2,495,429; the Heller patent text; the Arnold and Heller note extracts). Nothing in the review above weighed any of this: it was written before the Investigator ran. Tags: [web] = Investigator's URL, [web, mine] = checked by me, [read] / [calc] / [memory] as before. Findings are leads, not verdicts; I keep every pertinent one, including those that cut against Kassner and those that cut against the essay's reading. Findings I did not open myself (Gos 1997, Gandhi 1980, Mittler 1976, Navy survey) are cited from the Investigator's notes; I did not re-read the full papers.

Web check I could now do

Water's principal Debye relaxation time: Google results give tau = 8.27 ps at 25 C and a loss peak "centered at about 20 GHz" (Colosi et al. 2013, RSC; Academia.edu "origin of the Debye relaxation in liquid water"); a second result gives 8.3 ps for the fastest Debye mode [web, mine: https://www.google.com/search?q=water+dielectric+relaxation+time+8.3+ps+25+C+Debye+Kaatze, run in Chrome in a new tab, closed]. My own 8.5 ps (Stokes-Einstein-Debye) and 1.6 cm are therefore confirmed to within a few percent, and claim 1's "Established" for water's band stands on a web source now. Still not web-checked: serum/electrolyte band widths and electret charge ranges (kept [memory], unverified).

Weighing, by priority

1. The inventor's own work

• FR 786,294 and FR 788,428 (1934) tie "natural frequencies" to Debye relaxation (lead 21) [web: research/2026-10-04-depatisnet.md, page images]. FR 786,294 p.1: the natural frequencies are "respectivement les fréquences de relaxation dites bandes de dispersion anomales", "en conformité avec la théorie des dipôles de Debye". FR 788,428 p.1: the first natural frequency is "donnée par le temps de relaxation pour l'orientation des dipôles". Bears on claim 1(a). This is the most important new inventor-side item. It supports the part of my split that was already "Established" (Debye relaxation of water near 1.5 cm) and weakens the part I called "never properly tested": in the French texts there is no separate narrow resonance, only relaxation, which is broad (water's peak spans about six decades [web, mine]). It sharpens my earlier internal-consistency complaint: GB417501 says friction dispersion is broad and a "hitherto unknown" resonance dispersion is 2% wide; the French members, filed 1934, say the frequencies are relaxation frequencies. The same inventor, same year, gives two accounts of what a natural frequency is. I do not conclude the sharp-band claim is false; I conclude that it is not what the family consistently asserts, and that the 2%-wide bands carry no spectrum anywhere. Only FR 786,294 pp.1-2, 15, 17 and FR 788,428 pp.1 and 6 were read, so the other 8-9 pages of FR 788,428 might say more.
• FR 788,428 repeats the GB417501 examples (lead 20): gold sol, agar, milk, veal "plus de 15 jours" (GB says "2 weeks"), apple juice four months, serum near 90 cm. Same single runs, same absence of controls. No new data; no change to claim 3. A third language family member does not make the examples three experiments.
• DE 1,128,930 and Kassner's death (lead 19) [web: DEPATISnet]. "Ernst E. W. Kassner †, München" named inventor on a 1960 filing, applicant Wilma Hencke; the dagger convention is [memory, high]. Bears on claim 8 and the "record goes dark" question: the record ends with his death around 1960-62, not in 1937. The patent promises UHF rotating fields for particle acceleration, plasma reactions and "Kernfusionsforschung"; no data. A fusion mention in a 1960 patent is a statement of intended use, not evidence, and I do not treat it as bearing on claims 1-3 (it only shows he was still working on circular plate resonators, claim 4).
• GB 435,643 = divisional of GB 417,501; plate-resonator light detector (lead 22). A plate resonator whose frequency shifts when a layer's dielectric constant changes is ordinary physics (loading a capacitor-like resonator changes its frequency [calc basis: f falls as the square root of the effective permittivity]). This sharpens my claim-4 point that a plate resonator reacts to what is in its gap: it is a fair design for a sensor, and it also means "sample resonance" and "apparatus resonance" are entangled in the same device. It does not reconcile the 1932 lambda = 3 D figures. Abstract only.
• GB 448,418 / US 2,094,602 (leads 4, 22). Ellipsoidal reflector, flow-through glass sphere, thermostat, optional pre-ionisation: a design for treating a liquid. The OCR of US 2,094,602 (my grep of the Investigator's file is not repeated; I rely on the Investigator's grep) reports no frequency, power, efficiency or result. One more small internal slip: "a diameter of 1 cm or a radius of 2 cm" in the same sentence [web: research/2026-10-03-uspto-ppubs.md]. US 2,096,459 repeats "about 40%" with no method [read: Investigator's OCR lines]. No change to claim 5.

2. Support by others

• Arnold (US 2,522,082; In re Arnold 1950; Arnold v. Watson 1955, 237 F.2d 584 (1956)) (lead 3, 24) [web: research/2026-10-03-caselaw-access-project.md; 2026-10-04-arnold.md]. A trained dielectric chemist (JACS 1936, J. Electrochem. Soc. 1938) independently proposed a "critical frequency" at the centre of the anomalous-dispersion range, selective surface activation "without abnormally activating" the interior, "destructive heating is not required", with temperature, concentration and voltage shifting the frequency and needing tracking. The CCPA summarised Kassner as teaching that "excitation is especially strong and energy absorption especially great under conditions of resonance" and held that "Kassner provides a sufficient disclosure of the balance of the claim". Arnold's broad claims were refused as "too broad and indefinite" (1956); the courts ruled on patentability only. The 1955 judgment was reached on "evidence" and says Arnold "has exposed the error of certain assumptions and conclusions by predecessors"; the record of that evidence (NARA RG 21 / RG 276) has not been seen. My weighing: (i) this is a second inventor, with measurement experience, using the same identification, which counts for claim 1's relaxation-frequency half and for claim 13 (Arnold states the temperature shift explicitly and builds in compensation); (ii) the sole result is one unquantified sentence ("excellent results ... alkyd resin monomers"), so it is another untested claim, not a test; (iii) this is "never properly tested" for the sub-thermal selectivity, and the legal outcome is not "tested and failed" nor "dismissed without testing": a patent-law refusal on indefiniteness. Note the cost to Kassner's side too: Arnold's own account is the ordinary Debye S-curve, broad and tracked, which is the weaker claim.
• Heller 1959 (Nature 183:905; US 3,095,359) and the 1960 Navy/MRI survey (lead 13, 23) [web: research/2026-10-04-heller-pearl-chain.md, files/run3/]. Heller's one-page letter reports garlic root-tip chromosome aberrations from a 27 Mc pulsed field with "No temperature increase in the water was noted" and a frequency-, pulse-, power-, axis-dependence; no cell counts, controls or statistics are given. The Heller patent text I read (files/google-patents/US3095359A.desc.txt) lists frequency-specific outcomes at named megacycle values (sex ratio at 25 Mc, eye character at 28 Mc, histidine reversion at 31 Mc, tumour regression in rats at 30 Mc versus none at 3 Mc), again with no n or controls, and claims it as "critical frequency and voltage" per organism, citing Kassner. The MRI report (Navy contract, interested party: radiation-hazard standard-setting) called it "frequency and species specific", "A new specific nonthermal effect has apparently been found", and, for the 1-20 m band, growth enhancement "not because of heating"; it records pearl-chain formation as real, with Maxwell-Wagner polarisation as the candidate mechanism, at fields of 300-1,200 V/cm. Weigh: this is a second, independent group claiming frequency-specific, non-heating effects in the same wavelength band as Kassner's examples (1-20 m), published in a major journal and reported by a government contractor. It is not replication of Kassner and not controlled data. The mechanism the report itself offers for the orientation effects (interfacial polarisation of particles, a broad frequency response at high pulsed field) is conventional and is not Kassner's molecular natural frequency. I record both: the finding raises the standing of "frequency-specific non-thermal effects deserve a controlled test"; it does not support the 2%-wide molecular band.
• Botanical Review 1944 cluster (lead 7, 25). Benedetti (yeast fermentation up or down with frequency), Liebesny and Wertheim (GB 417,863, 1.8 mm-120 m chosen per organism), Ternion GB 417,501: reviewed as claims, "not weighed" by the reviewer. Contemporary and independent of Kassner (Austrian priority 1932). For claim 3: a small cluster of 1930s claims of frequency-dependent effects, none with data in the sources I have; no test of any is recorded.
• Rubber World, April 1945 (lead 8): the excellent results "may be connected in some way with electronic power itself". Unquantified trade-press remark. I keep it beside the Goodrich staff's own 1950 account (below). It is weak evidence of a non-thermal effect and weak evidence only because it has no data, not because the source is unimportant.

3. Related work, including work mainstream calls false

• Sharp-resonance claims and their tests (lead 27) [web: Wiley abstract pages, per the Investigator]. For: Webb and Booth (Nature 1969); Grundler and Keilmann (PRL 51:1214, 1983, narrow windows near 42 GHz, yeast). Against: Gandhi et al. (Bioelectromagnetics 1980): swept 26.5-90 GHz, "Sharp spectra reported by other workers were not observed", also none in frozen samples; Gos et al. (1997): concurrent exposed/control chambers, 41.682-41.710 GHz in 2 MHz steps, "no consistently significant differences". This answers the web check that failed in my first pass: the claims exist, and the one controlled repeat I now know of did not reproduce Grundler and Keilmann, so my earlier "replication unknown to me" becomes "at least one designed repeat found no effect (abstract read by the Investigator only; I have not seen whether Grundler and Keilmann responded)". Weigh for Kassner: his 90 cm serum band (about 0.33 GHz) is two orders of magnitude lower in frequency than these tests, so they do not test his band. They do test the class of claim (sharp, narrow spectral features in biological liquids and cells) and they are against it. A Gandhi-type swept dielectric spectrum of serum at 0.1-5 GHz (my proposed experiment) would be the direct test, and I have found no record of one.
• 1930s German "specific effect" dispute (lead 26) [web: Google snippets of Kowarschik 1936 and a Yumpu text; the Investigator saw only snippets]. Texts of the period state that no specific short-wave effect on bacteria had been proven, and that experiments claiming "athermic" action did not prove it (Schliephake and Compère 1933 are the proponents named). The claim was being argued, not ignored, while Ternion was funded. Single snippets, sources and dates unopened; treat as pointers, not as the dispute's conclusion.
• Non-thermal microwave effects in chemistry (Kappe, Pieber and Dallinger, Angew. Chem. 2013; a published reply): the debate is alive in the 2010s; bears on catalyst-free curing (claim 3) as an open question, not an answer.
• Mittler 1976 (Environ. Res. 11:326): title "Failure of 2- and 10-meter radio waves to induce genetic damage in Drosophila". Different organism from Heller's garlic root tips; abstract not available; full text blocked. A negative in the Heller wavelength band, not a replication.

4. The historical record

• Goodrich 1950 (Miller and Wittig, Rubber World, Oct 1950) and the 1945 sale (lead 1, 14) [web: archive.org Rubber World 1950-10-01]. The Goodrich portfolio names "The Kassner patents ... 2,089,966; 2,109,843", saying they "touch on this problem somewhat" and "most of the work he reported specified frequencies about 150 million cycles per second" (2 m [calc]). This documents which Kassner patents were in the buyer's portfolio (answering my correction 4 that the patents sold were unknown), and the buyer's reading is heating-only. That is an interested party's characterization (licensor of the rubber field; its stake favours a heating account), and it sits beside Kassner's own statement in GB417501 that his object is not a temperature rise. Both are stated, neither is neutral. Note also: GB417501's examples run at 0.2-4.9 m, so "about 150 MHz" describes part of the range only, and the Goodrich paper does not say what it tested. Also: US agent Grell held the 1941 option on Dufour/Leduc (Vesting Order 679); Kassner is not named in that order; the August 1945 licensing company, Industry Inventions, Inc., is described as licensing "Leduc and Dufour" patents only, though the January release also named Kassner. Why Kassner drops out of the licensing description is unexplained, and that is worth recording as an anomaly, not a conclusion. Vibranic (Stoffel, Nov 1945, Fr. 200,000) bought Grell's foreign heating rights (FR 926,351, CH 250,508 share Grell's 1943/45 priority dates); whether any Kassner patent was in that package remains unknown.
• Patent-office and legal treatment as prior art (leads 3, 12). Spencer's microwave-cooking patent (US 2,495,429, filed Oct 1945) lists "2,094,602 Kassner" in References cited (I can see the OCR line in the Investigator's file; whether the examiner or applicant cited it is not stated); Raytheon's 1954 juice patent (Copson, US 2,833,657) cites 2,089,966; 32 citations of US 2,109,843 are mainstream UHF-tube patents; the DOE Emitron patent (GB 2,120,008, 1983) cites GB483062. Voss's 1987 question "Or do they?" can thus be partly answered: US examiners and applicants did know of the patents from the 1940s on. This supports Voss's historical point (priority of apparatus and multi-frequency exposure) and does not support the resonance mechanism: Spencer and Copson frame their inventions as heating. Kassner's patents were also abstracted in Chemical Abstracts 29:1178 (1935) and cited in Thorpe, Weissberger and Ellis, always as claims.
• Dipol A.-G. closed in 1933 (lead 15) [web: e-newspaperarchives, three reprints Dec 1933-Jan 1934]. "Die Aufhebung der Dipol A.G." named among reasons for Arbon's 1933 tax fall; Dipol was formally dissolved only on 6 Apr 1938. So the operating laboratory (factory premises on Philosophenweg, Fr. 25,000 capital, chemist Wiegner, trademark "Elektrische Apparate") ran about 18 months. Wiegner (lead 28; ETH dissertation on carbides and silicides in alloys; diploma on dyes) was an analytical/dye chemist, not a dielectrics specialist. For the reader: this supports "the research was real but short", and it sharpens an unanswered accounting question that I should state plainly: the Ternion research account reached CHF 1.12 million (financed by loans) by end-1936, while the lab that apparently did the work closed in 1933 and had Fr. 25,000 of capital, and Dipol's loan account on Ternion's 1936 balance sheet is only Fr. 51,155 [read: note-14 deep read]. Where the remainder was spent (patent costs, London work 1934-36, an unnamed valuation at the 17 Jul 1936 minutes) is not documented in any record I have. I do not assert fraud or misvaluation; I note that the capitalised figure is the company's valuation, not an audited expenditure, and that the minutes justifying it are the single document that would settle the question.
• The record does not go dark; it continues in other archives (leads 5, 10, 11, 19) [web: SHAB, archives-online, BAR, Google Patents]. 1929 Elektro-Physikalische Gesellschaft (identity likely, not proven); 1937 "Phono-Vision, Ernst E.W. Kassner" folder of "Fotos und vertrauliche Firmenunterlagen" (KB Appenzell Ausserrhoden, digitised); 5 Nov 1938 visiting card to Hans Kayser (SLA); 1943 General Infra Company A.G., Chur, board member "deutscher Staatsangehöriger", shares for rights from "Suisse Infra-Compagnie" which were cancelled against return of the contributed rights in Jan 1944; 1951 Montana-Vermala (US 2,558,568 to Vibranic); dissolution of General Infra 1953; death by May 1962. Three Swiss Federal Archives items are open and unread: a 1946-48 "Schwarze Liste" file titled "General Infra Co AG, Chur" (title only shows the foreign ministry kept a file in that category, not that the firm was listed), and police dossiers on "Ernst Kassner, 20.11.1900" (1939; 1946-47; identity with E. E. W. Kassner unconfirmed). NARA also has "Ernst Kassner" in captured German records and "Giodvad" in State Department and Army lists; contexts unread. Competing explanations for the missing scientific publications: (1) commercial failure: supported by the 1937 liquidation text, Dipol's 1933 closure, the 1944 return of General Infra's contributed rights, and Goodrich's 1950 judgment; (2) trade secrecy: supported by the "vertrauliche Firmenunterlagen" folder (one folder) and the nature of patent-only disclosure, nothing about the Ternion work itself; (3) wartime/security context: a German national on a board of a Swiss firm whose file sits in a blacklist series; plausible background, contents unread; (4) absorption under a new name: the chain Ternion/Dipol (1932-38) to Vibranic (1945-52) to General Infra (1943-53) shows continuity of people (Stoffel, Kassner) and subject; Vibranic's foreign-rights purchase is documented to be Grell's heating patents, not shown to be Kassner's; (5) suppression: no document found shows blocked publication, withheld funds or deleted data; the patents were openly litigated, abstracted and licensed. The records that would decide between (1)-(4): the Vibranic contract of 15 Nov 1945 (Landesarchiv Glarus), the Ternion minutes of 17 Jul 1936, the BAR files above, the Industry Inventions licence schedule, and the Arnold v. Watson trial exhibits.
• Closed leads: Kassner is not the pharmacologist E. W. Kassner (Coward and Kassner); the 1941 SHAB hit is a false match; Bauch's 1953 Wireless World biography confirms GEC and gives his Hertz Institute work as "supersonics", with no mention of Ternion or UHF (lead 6). That is silence in a short author note, not contradiction; the memoriam sentence remains documented in one published source, no second source found.

Questions a-d, as changed by the findings

Verdict changes

Changed: - Claim 1, second half (sharp "resonance dispersion" in electrolytes and serum). Was: "Never properly tested". Now split. (a) Kassner's own 2%-wide bands at 0.3-1 m: still Never properly tested (no test at his frequencies found); not "dismissed without testing". (b) The wider class, sharp narrow spectral features in biological liquids and cells (Webb/Booth, Grundler/Keilmann): Tested, not supported by the Gandhi (26.5-90 GHz) and Gos (41.7 GHz) work, which are at different frequencies and use different samples, so this is evidence against the class, not a test of Kassner. Why: the Investigator's web sources show the two tests; the claim-1 weakening is reinforced by Kassner's own 1934 French texts equating his frequencies with broad Debye relaxation. The water-band "Established" is now [web]-supported. - Claim 7, "foresaw the microwave oven". Kept at Historically superseded but instructive, but the supporting text is revised: the patents were treated as prior art by US examiners, applicants and courts from 1945 to 1955, and the buyer's own paper names US 2,089,966 and 2,109,843. This is a documentation upgrade, not a verdict change. - Claim 8 note. Verdict stays Established, with the "no later trace" open item replaced by documented later traces and the accounting question above. - Claim 12 wording only. Verdict stays Not testable as stated (no documented link; Kassner's patents were openly litigated, licensed and abstracted, which cuts against concealment). I withdraw the implication in my first pass that the record was complete enough to say it "shows none": the BAR files, NARA items and Landesarchiv files are unread, and they are the records that would decide the wartime and security hypotheses.

Unchanged: claims 2, 3, 4, 5, 6, 9, 10, 11, 13 keep their earlier verdicts. Claim 3 in particular remains Never properly tested for Kassner's own examples; the new material (Heller, Benedetti, Liebesny, MRI, Mittler) shows other claimants and one negative in another organism, not a test of milk, juice, gold sol or seeds. Claim 13 is strengthened as to physics (Arnold) but remains unfalsifiable without a tuning protocol, and Arnold's refusal for indefiniteness is the pattern that such a clause invites.

Where the Investigator's findings and my own review are weak (same rigour)

1. I have read none of the cited full texts (Gos, Gandhi, Mittler, Kowarschik, Grundler/Keilmann, the Arnold exhibits); several rest on abstract snippets, and my use of them is bounded by that. 2. Heller's and Arnold's results lack n, controls and statistics, as Kassner's do; "independent claimant" is not "independent confirmation". 3. The accounting question about CHF 1.12 million is an inference from a balance sheet and press items; the minutes of 17 Jul 1936 might answer it fully. 4. Kassner's identity with the 1921-26 "Ernst Kassner" (adding machines), the 1929 Elektro-Physikalische Gesellschaft Kassner and the 1939/1946 "Ernst Kassner b. 1900" police dossiers is unproven, and I have not relied on any of those links. 5. The Investigator's FR 788,428 reading covered 2 of 10 pages; a full comparison with GB417501 may change the claim-1 point above.

Investigator findings (2026-10-08)

Scope: the review above last weighed leads 1-29 (run 3, 2026-10-04). Since then the Investigator's STEP F follow-ups added leads 30-35 (Q03, 2026-10-06) and 36-40 (Q05, 2026-10-07) to sites.md, with notes research/2026-10-06-followup-Q03.md and research/2026-10-07-followup-Q05.md. No other research note or file in research/files/ is newer than my review. I read both notes in full and sites.md leads 1-40. I re-did the arithmetic that matters [calc] and made one web check [web, mine]. Findings I did not open (Niether 2018, Serra 2020, Roy/Rao/Kanzius, Hundt 2014, Saitta 2012, Bui 2021, the ACS 2026 review, NASA Morehouse) are cited from the Investigator's notes, mostly from abstracts or search snippets. I keep every pertinent finding, including those that cut against the essay's reading and those that cut for it. Findings are leads, not verdicts.

Web check I could do

Water vapour's 22.235 GHz line: a search found the Becker and Autler 1946 Physical Review 70:300 paper ("Water Vapor Absorption of Electromagnetic Radiation in the Centimeter Wave-Length Range") and Townes and Merritt 1946 (70:558) on the water spectrum near 1 cm, and Liebe et al. 1969 giving the transition as 22.235 GHz [web, mine: https://www.its.ntia.gov/publications/download/Liebe.Journ.Quant.Spectrosc.Radiat.Transf.Vol.9.1969.pdf]. I found no source saying these were the first observations, so priority is unverified. It does show that the measurements I can find date from 1946, 14 years after GB417501 (1932).

Weighing, by priority

1. The inventor's own work

• Lead 30: Kassner wrote nothing on electrolysis or bond energy; his water band is the Debye band; his chemistry claim is a rate claim (claims 10, 11). The Investigator's full-text search of _text/ found no "electrolysis", "dissociation of water", "binding energy" or "bond". The nearest text is GB417501 p.4, "Purely chemical reactions can also be enforced which otherwise only proceed under the influence of heat or of catalysts", with examples on p.5 (isoprene polymer, vulcanisation, Bakelite, varnish drying, acetylene plus water to acetaldehyde) [read via Investigator's quotation; I did not re-read the page]. Weighing: this confirms what I wrote in claim 10 (not in any Ternion text) and gives the extension to water its best textual footing: water is dipolar, Kassner's principle is stated for dipolar substances generally, and he names water's band. I accept that. Two cautions. The chemistry examples carry no yield, energy balance or control, so they are claims, not tests. And his own frame is energy going in: the object is a substance "richer in energy" and "not to produce a rise of temperature" (p.6), which is the opposite direction from a bond that is cheaper to break.
• Lead 36: "The excitation is especially strong and the absorption of energy therefore especially great in the case of resonance" (US2089966 p.3) (claims 1, 11). This is Kassner's own statement that load on the source is greatest at the natural frequency. It bears on the Q05 "lightly loaded tank" premise (see Q05 below): by Kassner's account, resonance means maximum draw. The same page defines the "standing field" (conductors set into natural oscillation so that standing waves fill the space holding the substance; "a pair of concentric tubes or a plate-resonator") and the "radiating field" (reflector-focused). So the 2026-10-07 description of the method as displacement current through the sample as the resonator's dielectric is Kassner's first method in his own words, and I revise my earlier claim-4 point accordingly: the entanglement of "sample resonance" and "apparatus resonance" is not an accident of one design, it is how the standing-field method is defined.
• Lead 31: 1.28 cm may be the water-vapour rotational line (claim 1). 22.235 GHz is 1.349 cm [calc], 5.0 percent from Kassner's 1.28 cm; the liquid Debye peak is 1.5-1.56 cm. The Investigator offers this as a possible explanation of the two water values in GB417501 (1.28 and 1.5). My check above makes it weak: I can find vapour-line measurements only from 1946, and a 1932 patentee would have had no measured value to copy [web, mine; priority unverified]. Alternatives I can state without a source: 1.28 cm comes from some other calculation, or from a measurement of his own that the patent does not describe. I record it as an unexplained internal discrepancy, as before. It does not change claim 1.

2. Support by others

• Lead 32: RF and microwave fields do speed electrolysis, at the catalyst (claim 11, rate part). Niether et al., Nature Energy 3:476 (2018): high-frequency magnetic heating of FeC-Ni core-shell nanoparticles cut the oxygen-evolution overpotential by 200 mV and hydrogen-evolution by 100 mV at 20 mA/cm2; the gain "is equivalent to a rise of the cell temperature to ~200 C, but in practice it increased by 5 C only" [web per Investigator: https://ideas.repec.org/a/nat/natene/v3y2018i6d10.1038_s41560-018-0132-1.html]. Serra et al., Nature Energy 5:910 (2020): microwaves reduce Gd-doped ceria, which then splits water below 250 C; authors' institutions have a patent application (interested party) [web per Investigator: https://www.nature.com/articles/s41560-020-00720-6]. Weighing: this is the strongest new item for the strongest honest case. It is peer-reviewed experimental evidence, with a measured quantity, that a high-frequency field can lower the voltage cost of splitting water without raising the bulk temperature. It is also an anomaly-flavoured result (local heating at the catalyst, bulk 5 C) that a purely bulk-thermal picture would not predict, and it is the "selective, non-bulk heating" idea in modern form. What it does not do: it does not act through water's bonds or its Debye band (the field acts on a nanoparticle or a ceria solid), it lowers overpotential, not the 237.1 kJ/mol free-energy floor, and the abstracts do not say whether field power was counted in an overall efficiency. A claim-11 third part ("RF-assisted electrolysis as a rate effect") therefore moves from untested to tested-and-positive at catalysts, and stays untested at water's own band in a standing field.
• Lead 33: Roy, Rao and Kanzius, RF "catalysis of dissociation" of salt water (claims 10, 11). Materials Research Innovations 12:3-6 (2008); Current Science 98:1500 (2010). 13.56 MHz, about 300 W; the gas above the solution burned (flame about 1,800 C) only while the field was on; Raman spectra differ in the O-H stretch region; the claimants propose "specific resonant coupling" and "never claimed" a net energy gain (Philip Ball, Chemistry World, snippet; ChEnected, snippet) [web per Investigator: https://www.i-sis.org.uk/canWaterBurn.php]. Mae-Wan Ho (Institute of Science in Society, an advocacy-leaning outlet) wrote "The phenomenon is real, but the mechanism is entirely unknown"; an anonymous reader comment on the same page says the spectra "look similar, apart from a minor difference in peak intensities". No independent replication was found, in favour or against; the Investigator also could not open the 2010 paper. Weighing: this is the only direct claim I know of that RF acts on water itself, and it is two papers by the claimants and press coverage. The record is thin, not refuted. Note that a salt-water flame under 13.56 MHz is consistent with conventional explanations (dielectric and ionic heating, ionisation of sodium in the field), which I cannot exclude or confirm without the papers. 13.56 MHz is 22 m, about 1,700 times longer than Kassner's water band, so it is not a Kassner-type resonance. Never properly tested by a neutral party: no replication attempt on record; not "dismissed without testing".
• Lead 37: second Wien effect and bipolar membranes (claims 2, 11). A measured, mainstream, static-field effect: strong fields (about 10^8 V/m [memory, medium]) orient water and speed its dissociation into H+ and OH- (Onsager 1934; Mafé et al. 1998; Bui et al., J. Phys. Chem. C 2021; Yamamoto et al. 2025); Stovall et al. 2026 question the second-Wien model (snippets only) [web per Investigator]. It is a rate effect and DC, not a resonance. Weighing: this is the best mainstream support for the idea behind Kassner's claim 2, that field orientation changes what a dipolar liquid does, and it argues that "orientation by a field alters the chemistry of water" is not outlandish. It does not support "energy content" or the 2%-wide bands.
• Lead 38: field-assisted electrolysis energy balances (claim 11). ACS Appl. Energy Mater. 9(16):10362 (2026): snippet says AC coil losses "aggravate this negative energy balance even further" and elsewhere "a net energy saving of 215.6 kWh was calculated"; context unseen, paywalled [web per Investigator]. I record both halves: at least one field method is judged a net loss once coil power is counted, and at least one (method unnamed in the snippet) a net saving. I give the second no weight until the context is read, and the first no more than its snippet carries. Both concern overpotential, not the floor.

3. Related work, including work mainstream calls false

• Lead 34: resonant bond weakening of water is established, in the infra-red, at a surface (claim 10). Hundt et al., Science 344:504 (2014): laser excitation of the D2O stretch strongly promotes dissociation on Ni(111); O-H versus O-D selectivity in HOD (J. Chem. Phys. 2018) [web per Investigator, snippets]. This is real, bond-selective, resonant activation of water. The Investigator gives the photon as about 35,000 times a 1.28 cm photon; from its own numbers I get about 3,400-4,700 times (0.33-0.46 eV against 9.7e-5 eV) [calc], and I list this under corrections below. The qualitative point stands. This weakens my claim-10 verdict "Refuted by calculation (single-photon form)" not at all, since the single-photon form at 1.28 cm is still five-plus orders below a bond; but it shows the correct statement is "microwave photons cannot do it, infra-red photons can, on a surface", which is more informative than the flat refutation and should be the wording a reader gets.
• Lead 35: later resonance-electrolysis claims (claim 11). Puharich US 4,394,230 (1983), "complex electrical wave form" water splitting (not read); Stanley Meyer's "water fuel cell", on which an Ohio court in 1996 found "gross and egregious fraud" after expert testimony that it used conventional electrolysis (the court's characterization in an adversarial proceeding, from a Wikipedia snippet); the non-thermal microwave debate in chemistry (Kappe, Pieber and Dallinger, Angew. Chem. 2013, and a published Correspondence) [web per Investigator]. I record all three. Puharich is unread, so no weight. Meyer is a court finding of fraud on specific investors and tells us nothing about whether Kassner's principle works. Kappe's thermal reading and its dispute remain the live question for catalyst-free curing in claim 3.
• Lead 39: thermally regenerative electrochemical cycles (claim 11). The mainstream form of "split water where it is cheap, burn it where it pays": NASA (Morehouse 1986, electrolyzer at 1,300 K), Chum's review. Carnot-bounded: for 373 K and 298 K, 1 - 298/373 = 0.20 [calc]. It shows what Q05's "environment pays" would have to be: a second reservoir at a different temperature.
• Lead 40: water's Debye loss and the dissociation field (claims 1, 4, 10, 11). Relaxation near 19 GHz, tau about 8.3 ps (Kaatze); Saitta et al., PRL 108:207801 (2012): dissociation beyond about 0.35 V/A [web per Investigator: https://arxiv.org/abs/1204.1120, abstract read]. I recomputed the single-Debye numbers (eps_s 78.4, eps_inf 5.2, tau 8.27 ps): loss tangent 0.87 at 19.2 GHz and 1.03 at 23.4 GHz, 0.12 at 2.45 GHz, 0.05 at 1 GHz [calc]; so a water-filled resonator at the band has a material Q of about 1.0-1.1. This agrees with the Investigator and also corroborates my claim-1 web check (tau 8.3 ps). The dissociation-field scale (3.5e9 V/m) is also consistent with the stored-energy estimate below.

Q05 (the "energy asymmetry" hypothesis), weighed

The Investigator's reasoning is, to my reading, correct on the points that decide it, and I checked its numbers. 1. Energy books [calc, memory high]: dG = 237.1 kJ/mol, dH = 285.8 kJ/mol, T dS = 48.7 kJ/mol, dH/dG = 1.205. Q05's "somewhat like air conditioning" is right for the established part: a reversible cell at 1.229 V takes 48.7 kJ/mol of room heat, so heat out over work in is up to 1.2. The asymmetry is in heat, not in work. Returning more than 237.1 kJ of work from the flame needs a second reservoir. 2. Lowering the floor: any field that loosens water first pays the difference itself. Linear dielectric storage, 0.5 eps0 (eps_r - 1) E^2 times 18.07 cm3/mol: 0.006 J/mol at 1e6 V/m, 0.6 J/mol at 1e7, 62 J/mol at 1e8 [calc]; the linear formula at 2e9 V/m gives 25 kJ/mol but it is out of range there (dielectric saturation, dissociation), so it is only an order-of-magnitude marker that a 10 percent saving needs fields near the Saitta threshold. 3. "Lightly loaded" fails at the band: Q about 1 from the water alone, and Kassner himself says absorption is greatest at resonance (lead 36). A tank can be lightly loaded only with a small water filling factor or a weak field, which means a small effect. 4. A lowered per-molecule cost must show as real power drawn from the tank, at least equal to the saving (6.6 W for 24 kJ/mol at 1 mol H2/hour [calc: 24,000/3600]). A saving with no such load would be the asymmetry and would violate the second law. 5. Environment as source: ambient radio of 0.1-10 microwatts/cm2 (an AI-aggregator snippet; the primary Piñuela 2013 survey unread) would need about 660 m2 of perfect collector for 6.6 W at 1 microwatt/cm2 [calc: 6.6 / 1e-6 W/cm2 = 6.6e6 cm2 = 660 m2], and a 19-23 GHz tank does not collect broadcast bands. 6. An essay draft in the folder says the machine "cools the local environment around the machine" and that "the equipment generating the natural frequencies must consume less power than required to electrolyze water" [read: per the Investigator's quotation of old-unexplored-technology-14 and -8]. The cooling prediction is a testable difference from the conventional account (a heat-pump effect of 48.7 kJ/mol is expected anyway at 1.23 V, so cooling alone does not discriminate; only total work per mole does). Not found by the Investigator: any measurement of total work (field plus electrical) per mole below 237.1 kJ/mol. That is absence in a 2-minute search, and as the quantity thermodynamics says cannot go below the floor, its absence is expected; I do not treat it as evidence in either direction beyond that.

Questions a-d, as changed by the findings

Verdict changes

Changed: - Claim 11, rate part (RF- or field-assisted electrolysis). Was: Plausible, untested (my first review did not address it separately; the arbiter had it so). Now split: (a) the claim that a high-frequency field lowers the voltage cost of electrolysis at a catalyst rests on measured overpotential reductions (Niether 2018; Serra 2020 for a ceria route), I give it Established (conventional measured result) for catalyst-mediated rate effects, with the caveat that the field power versus saving is not settled in the abstracts I have; (b) the same effect by a standing field at water's own Debye band, which is what Q05 proposes, stays Never properly tested; (c) net work below 237.1 kJ/mol including the field's real power stays Refuted by calculation; (d) the heat-pump ratio up to 1.2 stays Established. Why: two peer-reviewed measured results the earlier review had not seen. - Claim 10, wording. Verdict stays Refuted by calculation in the single-photon, microwave form, but the statement now says that resonant, bond-selective weakening of water is established in the infra-red at a surface (Hundt 2014) and that Roy, Rao and Kanzius claim an RF effect on water with no replication on record (Never properly tested by a neutral party). Why: so the verdict is not read as "no resonant weakening exists". - Claim 1, note only. Verdict unchanged. The water Debye band stays Established, now also by Kaatze's 19 GHz / 8.3 ps via the Investigator. The vapour-line explanation for 1.28 cm is recorded and weakened by the 1946 dating; the two water values remain an unreconciled internal discrepancy. - Claim 4, note only. Verdict unchanged. Lead 36 shows the sample/apparatus entanglement is built into the standing-field method.

Unchanged: claims 2, 3, 5, 6, 7, 8, 9, 12, 13. The new items touch claim 2 (second Wien effect) and 3 (Kappe) only as context.

Corrections to earlier documents

• Investigator's note research/2026-10-06-followup-Q03.md, section D, and sites.md lead 34: "roughly 35,000 times a 1.28 cm photon" does not match its own numbers. 0.33-0.46 eV against a 1.28 cm photon of 9.7e-5 eV gives about 3,400-4,700 [calc]. The "35,000" figure should read "roughly 4,000" (the O-H bond energy, 463 kJ/mol, is about 50,000 times the photon, which may be the source of the mix-up). The conclusion does not depend on it.
• sites.md lead 31 / Q03 lead 4 suggests the vapour line as Kassner's source; no measurement before 1946 found by me [web, mine], so it should be marked as unlikely rather than as an inference to check.
• My own earlier claim-10 wording ("Refuted by calculation" with no further statement) is superseded by the fuller wording above.

Where these findings and my weighing are weak (same rigour)

1. I have read none of the papers (Niether, Serra, Roy/Rao/Kanzius, Hundt, Saitta, Bui, the ACS review, Morehouse); I rely on abstracts and snippets, one of them (the 215.6 kWh sentence) with its context unseen. 2. The Niether and Serra results are the work of groups with a stake in their methods (Serra's patent application; Niether is a materials-chemistry paper); I have no reason to doubt the measurements but the energy accounting is not shown. 3. Kanzius evidence is the claimants plus an advocacy outlet; the anonymous reader's comment is also unsourced. 4. My Q05 numbers use a single-Debye model with remembered parameters (eps_s 78.4, eps_inf 5.2); real water has a second faster process that matters above about 100 GHz, not at 19-23 GHz. 5. The linear storage estimate is order-of-magnitude only. 6. The 1.28 versus 1.349 cm discrepancy is a 5 percent mismatch, not a match; I give it no explanatory weight.