QE (Step 4b): Frollov-Alexander (Alexander V. Frolov)

Inputs checked: _text/manifest.json, all 3 digests (digest/FreeEnergy-db73c1.01.md, .02.md, digest/single-wire-power-transmission-a19820.01.md), facts.json, claims.md, physicist.md, mathematician.md, contrarian.md, arbiter.md, primer.md, example.py. Spot checks were made against the extracted text in _text/ (FreeEnergy-db73c1.txt, 1,138 lines; single-wire-power-transmission-a19820.txt).

Page convention used below: "PDF p.N" means the === page N === marker in the extracted text, which is the convention the digests use. The 2003 article's own magazine page numbers are PDF page + 10 (PDF p.1 = magazine p.11, ..., PDF p.18 = magazine p.28). The reviews and primer cite magazine pages without saying so (see Process).

Verdict: MINOR

The reviews, arbiter and primer are careful and accurate about what they cover. The headline verdicts (Claims 1-5) are unlikely to change. But Step 1 never covered the second half of the 2003 article, PDF pp.10-16 (magazine pp.20-26). That half holds two experimental claims Frolov makes about his own work and about 20 further device or suppression claims, and none of them reaches facts.json, claims.md or the primer. The fix is to add sub-claims and per-device verdicts. That is a MINOR remediation (update facts, add an arbiter QE revision, update the primer), not a redo of Steps 2-4.

1. Coverage

Every manifest document was digested, and nothing was skipped. Both documents have a full text layer, so there is no source gap for the folder's contents. The real gap is the one claims.md already states: no primary documents for the devices Frolov cites.

2. Shortcuts

1. Defect (main finding): FreeEnergy.pdf was in effect sampled. facts.json still records "how_read": "text layer sampled", and its summary says the body was read "apart from the long patent lists". The arbiter accepted a correction to "body read in full", but facts.json was never updated. The content of facts.json tells a different story. Its summary and key claims stop at Zaev (PDF p.9-10) and never mention anything from PDF pp.10-16 except the two lists at the end. The reviewers and the arbiter read the text layer themselves, and the contrarian and arbiter picked up one item from PDF p.13 (the arc test). Every other item in the second half appears in no review file. Counts across facts.json, claims.md, the three reviews, the arbiter and the primer are zero for: Johnson, Adams, Hide, Echlin, ferrite (Frolov's own experiment), back-EMF rotor acceleration (Frolov's own), Swenson, Moray, Koler, De Palma, Tewari, MITI, Lambertson, Chernetsky, Latchinov, Goriatchev, Kanarev, Serogodsky, Buynov, Methernitha, Filimonenko, TOPAZ, Patterson/ENECO, AECL, electrolysis and teleportation. That is half of a key document, sampled away.
2. Acceptable: single-wire note read in full; arithmetic re-derived independently four times (facts, reviewers, arbiter, example.py).
3. Acceptable with caveat: none of the figures in FreeEnergy.pdf (Figs 1-12) was rendered as an image. The primer rendered Fig.1 of the 1994 note. The 2003 figures are schematics (tube circuit, pulse timing, Frolov's patent Fig.11), and claim 1 and the self-cooling or back-EMF claims depend on them, so they should be viewed in the deep read below.

3. Missed material (from my own reading of the digests and the text)

Frolov's own experimental claims (the most important omissions, because they are his, not second-hand):

• Ferrite self-cooling alternator [read, PDF p.11 / mag. p.21]: "The experiments to investigate the alternator principles were organized also by the author of this article and it was demonstrated that ferrite core of the generator coil is self-cooling." Setup: a motor rotates an iron plate through the gap between a magnet and a coil, with flux "decrease[d] ... in the rapprochement half-cycle and increase[d] for the moving off half-cycle". No temperatures or powers are given. If the claim is cooling while delivering electrical work from a single heat reservoir, it is a second-kind claim (Frolov's own classification). Needs a verdict.
• Rotor accelerated by back-EMF [read, PDF p.11-12]: "In this case the rotor is accelerated by the secondary field (back-EMF). From 1994 to 2003 several experiments were produced and main principles were claimed in the patent description, Fig.11." This is a quantitative-in-principle claim about his own patented device. It is not in facts.json.
• Arc / "negative resistance" test, stronger than recorded [read, PDF p.13]: the arbiter's sub-claim 1a says consumed power "falls". The text says more: "consumption current is decreasing until zero and then it can change its direction i.e. this system begins to generate the power". Frolov says "negative resistance ... was confirmed" and "One of the effects was demonstrated in 1996 during the conferences 'New Ideas in Natural Science', St.-Petersburg". He also says a bigger spark gap gives more surplus. The 1996 public demonstration and the current-reversal claim belong in facts.json and in the arbiter's 1a. He frames it as following Chernetsky's self-generated discharge (SGD, 1971 Moscow Aviation Institute transformer destroyed by a "reversed current" impulse "10 times" the consumed power), and the files never mention Chernetsky.
• Electrolysis as "field does work for free" [read, PDF p.14]: "ions in electrolyte are moving due to electric field only, i.e. work ... [is] produced by the potential field and expenses of the primary power are not required." This is Claim 1 applied to electrochemistry, and a clean teaching case: the cell's energy comes from the source's EMF (electromotive force) times the charge passed. It is not assessed anywhere.

Other people's claims, with numbers, that are missing from Claim 5's bundle [read, PDF pp.10-15]: Methernitha "Swiss machine" (750 kW total; 20 cm disk 200 W; 3 m disk 30 kW); Hide, US 4,897,592 (1990; screen over the decelerating arc, the same mechanism as Claim 1); Reed motor (1991); Johnson, US 4,151,431 (5 kW claimed; "electron spin" energy); Adams motor (10 kW planned, with Aspden); Echlin, US 4,567,407; Swenson (7.5 Hz field pulsations, 375 kHz, 10 m antenna); Moray (>50 kW, tested by Yakovlev in Nov 1929); RQM/Crane (25 and 200 kW); Koler (60 kW, 1925-45); De Palma unipolar (1991), Tewari, Gray, MITI 40 kW superconducting unipolar report (1994); Lambertson (1,600 W per unit); Correa PAGD (pulsed abnormal glow discharge), US 5,416,391 and 5,449,989 (483% efficiency claimed); Latchinov high-pressure electrolysis (1888); Goriatchev 0.2 V pulsed electrolysis (1500%); Kanarev plasma electrolysis; Serogodsky and Sheffer, German patent 4,244,016 (benzene-water retro-condensation at 154 °C, a second-kind claim); Buynov's "mono-thermal" N2O4 (dinitrogen tetroxide) cycle (1992, 1995); Filimonenko "warm nuclear synthesis" (1957), TOPAZ, Patterson cell, ENECO, AECL 1995 review (Brown's gas, "scalar (torsion) fields"); Chernetsky's group on biologically active radiation from SGD devices; Kozyrev gyroscope data put forward as the basis for "teleportation systems".

Suppression allegations not examined. The arbiter concludes that the "refused to test" criticism "does not apply to anything in these two documents". That holds for Frolov's own claims. But the article makes explicit suppression allegations that no reviewer addressed: Filimonenko's work supported by Kurchatov, Korolev and Zhukov in 1960, "secret resolution", then "suppressed by people of nuclear power plant team", and a 1994 Moscow City Council commission recommending renewal [read, PDF p.15]; Johnson's patent delayed 1973 to 1979 because "it is 'perpetuum mobile'" [read, PDF p.11]; Russian institutes blocking new ideas [read, PDF p.15-16]. The contrarian's brief requires looking at how claims were rejected, and the Curator weighs suppression stories, so each allegation needs at least a sourced or "unsourced in the document" status.

Smaller items: Frolov dates his "conception" to 1995 [read, PDF p.1]; Tesla's 1934 fuel-less car appears in facts.json but in no review (fine as a historical aside).

4. Accuracy spot checks (against _text/)

5. Process

• Source tags: present and honest throughout. The reviewers and the arbiter say plainly when a [web] tag rests on a search summary rather than an opened page (ARL/Bahder-Fazi, VIESH, Tajmar, amasci), and the arbiter flags the unresolved ARL vs Wikipedia tension. Good practice. Weak point: almost no primary web source was actually opened. That is acceptable here because no headline verdict turns on those details.
• Web checks: done for the medium/low-confidence facts that verdicts depend on (Brown in vacuum, Kozyrev 1960 commission, Tesla 1892, VIESH, Sweet). Within budget.
• Verdict scale: applied consistently. The arbiter correctly rejected the mathematician's off-scale "Refuted by his own data" and the misuse of "Historically superseded" for Kozyrev, and corrected "Not testable as stated" to "Never properly tested" for the quantitative Godin-Roshin claims.
• facts.json not corrected: the how_read for FreeEnergy.pdf is still "text layer sampled". All three reviewers and the arbiter flagged this, and no one applied it. Under the new process it should read "digest + deep read (pages)".
• Page citations: the reviews and primer cite magazine pages (11-28) and never say so; the digests use PDF pages (1-18). State the convention once in the primer's source note.
• Primer: follows all 9 sections of STEP 4 in order. Every symbol is defined on first use (ε₀, μ₀, ω, φ, Q, S, J). Textbook chapters are tagged [memory, high] and are correct as far as I know. The worked example uses real numbers and has a control-measurement step. The answers check out (Q2: X_C = 1.59 × 10⁷ Ω, I ≈ 0.13 mA). It suits the reader profile well. What it lacks is any mention of Frolov's own ferrite and back-EMF claims and of the electrolysis argument, which would be a natural second teaching case for "who pays".
• example.py: runs with the venv in about 10 s. Its output matches the primer: 0.0200 J/spark, 0.100 W, 3.3%; C_s = 1.01 pF symbolic (0.80 pF reviewers' estimate); no-diode source work ≈ 10⁻¹⁹ J; with the plug, source work 0.1000 J = 4 sparks × 0.020 J + 0.020 J stored, residual 4.7 × 10⁻⁸ J.

6. Remediation actions

Deep reads (from the extracted text, plus page images for the figures):

1. FreeEnergy-db73c1, pp.10-16, Sonnet. List every device or experiment claim in order, with its numbers and the evidence given (measured? by whom? instruments?). Separate Frolov's own experiments (ferrite self-cooling alternator, back-EMF rotor acceleration and the patent of Fig.11, arc/SGD negative-resistance test with current reversal and the 1996 St Petersburg demonstration, the electrolysis "field does the work" argument) from second-hand reports. Record each suppression allegation (Filimonenko and the 1960 resolution, the 1994 Moscow commission, the Johnson patent delay, institutes) with whatever source the article gives. Render Figs 8-12 as page images.
2. FreeEnergy-db73c1, pp.1-5, Sonnet. Render Figs 1-3 (tube with grid under cathode, pulsed-anode timing, single-wire diode circuit) as page images. State exactly how Frolov's closed-trajectory argument (Claim 1) is set up, including any equation. Give the details of his own 1995 flat-spiral-coil "asymmetrical mutual induction" experiment (what was measured). Give the context of the "20-100 kW" lines statement.

facts.json: for FreeEnergy.pdf, change how_read to "digest + deep read (pp.1-5, 10-16); appendices digest only". Add to key_claims: the ferrite self-cooling alternator (own); rotor acceleration by back-EMF, 1994-2003, patented (own); arc negative resistance, input current to zero then reversed, 1996 demonstration (own); electrolysis needing no primary power because the field moves the ions; the Filimonenko suppression allegation. Add the second-half devices with their numbers to the summary.

claims.md: add the ferrite and back-EMF experiments and the arc reversal as Frolov's own experimental sub-claims under Claim 1 (or as a new Claim 6, "Frolov's own bench experiments"). Extend Claim 5's list with the second-half devices, grouped as: magnet motors; unipolar; electrolysis and cold fusion; second-kind heat engines.

arbiter.md, add a "QE revision" section with verdicts for: - the ferrite self-cooling alternator (Never properly tested; as single-reservoir heat to work, Refuted by calculation, second law); - back-EMF acceleration (Never properly tested; Refuted by calculation if it means net output with no accounted input; show the energy-ledger argument for a generator whose back-torque aids rotation); - arc reversal plus the 1996 demonstration (revise 1a; probably still Never properly tested; record that a public demonstration is claimed with no data); - electrolysis "free" (Refuted by calculation: the energy delivered is EMF × charge, and the minimum water-splitting voltage is 1.23 V; Goriatchev's 0.2 V would need the source to supply the rest some other way. Web-check the 1.23 V figure or tag it [memory, high]); - Goriatchev, Kanarev, Correa 483%, Methernitha, Johnson, Adams, Moray, Koler, Lambertson (Never properly tested, or Tested, not supported where a web check finds one; budget about 4 web calls, with Methernitha/Testatika and Correa PAGD first); - Serogodsky/Sheffer and Buynov second-kind machines (Refuted by calculation, Kelvin-Planck); - the Filimonenko suppression allegation (state whether any source beyond Frolov exists, and whether the claim was ever tested).

Then revise the "Overall" and "Contrarian's distinction" paragraphs: Frolov does make suppression claims, though not about his own work.

primer.md: - §1/§4: add a sentence on Frolov's own bench claims. - §6 evidence table: add rows for the ferrite self-cooling, back-EMF, arc current reversal, electrolysis, and the grouped second-half devices. - §7 "Genuinely open": add the ferrite self-cooling test (thermocouple on the core with and without electrical load, as a cheap reader experiment) and the arc-reversal test with the 1996 demonstration. - Optional short second worked check: the electrolysis energy ledger (1.23 V × 2 × 96,485 C/mol ≈ 237 kJ per mole of H₂ (the Gibbs free energy, ΔG) vs the H₂ combustion enthalpy, 286 kJ/mol, which shows why "burning the gases gives more than the input" is a bookkeeping slip involving TΔS (temperature times entropy change) and heat from the surroundings). - Fix the page cite for "efficiency can be more than 100%" (mag. p.14). - State in the source note that 2003 page numbers are magazine pages.

qe.md: a Sonnet recheck appends a "Recheck" section after remediation.

Recheck

Rechecked against current facts.json, claims.md, arbiter.md, primer.md, example.py and deep/FreeEnergy-db73c1.md.

Defects 1. FreeEnergy.pdf sampled, second half (PDF pp.10-16) never covered: FIXED. A deep read now exists (25-item device table, Figs 8-12 rendered and viewed, suppression allegations recorded). 2. facts.json how_read "text layer sampled": FIXED. It now reads "digest + deep read (PDF pp.10-16, Figs 8-12 viewed) ...; appendices digest only" and states plainly that pp.1-5 got no dedicated deep read. 3. Frolov's own bench claims missing (ferrite self-cooling, back-EMF rotor, arc current reversal and 1996 demonstration, electrolysis argument): FIXED. They are in facts.json key_claims, in claims.md as new Claim 6, in arbiter.md QE revision items 1-4 (arc verdict 1a revised), and in primer rows 6a-6d. 4. Second-half devices and suppression allegations (Filimonenko, 1994 Moscow commission, Johnson delay, institutes): FIXED. Added to claims.md (Claim 5 extended, new Claim 7), arbiter items 5-7 and QE revision item 6, and the primer table and Section 7. 5. Page-citation defects (magazine vs PDF pages; "efficiency more than 100%" cited p.15 instead of p.14): FIXED. The primer source note states the convention and the cite now reads p.14. 6. Chernetsky/SGD context and the arbiter's Claim 1a wording: FIXED (arbiter QE revision item 1).

Remediation actions - Deep read 1 (pp.10-16, Sonnet): DONE. - Deep read 2 (pp.1-5, Sonnet; Figs 1-3 as images, the Claim 1 closed-trajectory setup, Frolov's 1995 flat-spiral-coil experiment, the context of the "20 to 100 kW" lines): REMAINS. Page images of Figs 1-3 were rendered (deep/img/FreeEnergy-p2-figs1-2-02.png, FreeEnergy-p5-fig3-05.png), but no deep-read text covers them. The earlier pp.1-5 deep read was overwritten, and the deep file's header and the arbiter both admit this. Claim 1, the central claim, therefore still rests on text-layer reads by the reviewers plus the digest. That is the central-claim-without-deep-read case the QE rules treat as a defect. The 1995 coil experiment appears in no deep read. - facts.json, claims.md, arbiter.md QE revision, primer.md updates: DONE. - Primer electrolysis ledger and extra worked check: DONE. example.py ran and its output includes the electrolysis ledger (237.1 kJ/mol, 285.8 kJ/mol, 16% at 0.2 V), matching the primer.

The headline verdicts are not expected to change, but one remediation item remains.

Final: OPEN (1 issues)

Recheck 2

Rechecked against current facts.json, claims.md, arbiter.md, primer.md and deep/FreeEnergy-db73c1.md.

Remaining issue from Recheck: deep read 2 (PDF pp.1-5; Figs 1-3 as images; Claim 1 closed-trajectory setup; 1995 flat-spiral-coil experiment; "20 to 100 kW" context). - Deep read exists: FIXED. deep/FreeEnergy-db73c1.md now has a pp.1-5 section (sections A-F, from line 127): the closed-trajectory argument (no equation; stated in words), Figs 1-3 described from page images (p.2 and p.5 viewed), the 1995 coil experiment quoted in full (section C: what was measured is not stated), the Strebkov sentence in context (section D), other pp.1-5 items (E) and a bottom line for Claim 1 (F). - Downstream files updated: FIXED. facts.json how_read and summary record the pp.1-5 deep read and its content; claims.md coverage line, arbiter.md Round 2 (items on pp.1-5, including the coil experiment and the unsourced Tesla and efficiency assertions) and primer.md source note all reflect it. - Residual, not blocking: PDF pp.6-9 (Figs 4-7) have no deep read or images, but every file says so openly, and none of Claims 1-5 turns on them. The deep file's header still carries the earlier overwrite note.

Final: PASS