Thompson-JJ (J. J. Thomson): Mathematician review (Step 2)

Reviewer: PhD mathematician. Inputs read in full: claims.md, all eight files in deep/, the "Leads for review" and research notes of the Investigator (sites.md, research/2026-10-08-*.md). I did not re-read the digests except where noted; the deep reads carry the page-image readings of the equations. No downloaded documents exist in research/files/ (it is empty), so the Investigator's leads are weighed as reported. Tags: [read] = from a deep read of the source (page numbers as in those files), [calc] = my own computation this session (Python in a scratch directory, not a deliverable), [web] = URL given, [memory] = recalled, with confidence.

Source gap: none for the claims as a whole, but three limits matter. (1) The Corpuscular Theory deep read lost its Chapter I section in a concurrent overwrite (noted in that file), so the e/m and charge values of Chapter I rest on the digest. (2) The Conduction of Electricity through Gases chapters I-XX (about 660 pages), Electricity and Matter chapters I-IV and VI, and Vortex Rings Parts I and IV rest on digests only; I have not checked their mathematics. (3) The Investigator's later works (Nature letter 1913, 1922 Proc. A paper, Structure of Light, Beyond the Electron) were not obtained, so what is said of them below is as the Investigator reports it.

What I re-checked myself ([calc])

I re-ran five of the deep readers' load-bearing checks rather than trust them:

1. Vortex polygon spectrum. The deep reader's closed form lambda^2/U^2 = -k(n-k)[2(n-1) - k(n-k)], k = 1..n-1, gives for n = 3..9 the sets {-4}, {-9,-8}, {-16,-12}, {-25,-16,-9}, {-36,-20,0}, {-49,-24,+15,+32}, {-64,-28,+36,+80}. Seven is marginal (a zero root), eight is the first with a positive root. Confirmed.
2. Vortex-ring eq. (71): inserting Thomson's printed A' and C' (as read from the page image) into his own limit formula (3/2)(1/15)(8A'/c^6 + 2C'/(c^4 k^2))(2/k), the ratio to the printed (71) is exactly 3 (symbolically, and 3.000 numerically). The two time integrals used are the Beta-function values 16/15 and 4/15, which I checked by hand. Confirmed, with the proviso that A' and C' are as read by the deep reader.
3. Conduction eq. (1) for the field momentum R: the printed closed form evaluates to 12.5 at v/V = 0.01 and 2.52 at 0.05, where the correct slow limit (2/3)(v/V) is 0.0067 and 0.033; the deep reader's corrected closed form gives 0.006667 and 0.03337. So the printed eq. (1) fails its own slow limit. Confirmed.
4. The "alpha" velocity factor for a point charge outside radius a, alpha(b) = (3/4)(1-b^2)^2 Int_0^pi sin^3 t/(1-b^2 sin^2 t)^3 dt, with c = 3.0 x 10^10: 2.874, 2.152, 1.819, 1.644, 1.511 at v = 2.85, 2.72, 2.59, 2.48, 2.36 (x 10^10). With c = 2.998: 2.891, 2.159, 1.822, 1.647, 1.513. Lorentz-Einstein gamma at the same speeds: 3.203, 2.371, 1.982, 1.777, 1.62. Confirmed the deep readers' tables.
5. Barkla count (Corpuscular Theory ch. VII): Ne = 0.00025 x 3/(8 pi) / (e (e/m)^2) = 8.6 with Thomson's e = 1.2 x 10^-20 emu and e/m = 1.7 x 10^7; N/n = 21.5 (with ne = 0.4), not 25. With e = 1.602 x 10^-20 and e/m = 1.759 x 10^7: Ne = 6.02, N/n = 14.0. The Thomson cross-section times air molecules per cc times 14.4 electrons gives 2.56 x 10^-4 per cm, which reproduces Barkla's 0.00025. Confirmed.

Corrections to Step 1

These concern claims.md, facts.json and the deep reads (the Step 1 products). The orchestrator should apply them.

1. Kaufmann comparison: the two books label the two columns oppositely, and Step 1 did not notice. In Conduction (2nd ed., Table 2, [read: Conduction deep read section 4]) the column 3.1, 2.42, 2.0, 1.66, 1.5 is p' = "observed by Kaufmann" and 3.09, 2.43, 2.04, 1.83, 1.65 is p = Thomson's calculation. In The Corpuscular Theory of Matter (p.48, [read: Corpuscular deep read section 3]) the column 3.09, 2.43, 2.04, 1.83, 1.65 is headed "Kaufmann" and 3.1, 2.42, 2.0, 1.66, 1.5 is headed "alpha (Thomson's calculation)". The two columns are the same numbers in both books; the labels are swapped. Internal evidence decides it in favour of the Conduction labelling: Table 1 of Conduction (e/m = 0.63, 0.77, 0.975, 1.17, 1.31 x 10^7) times the column 3.1, 2.42, 2.0, 1.66, 1.5 gives a constant slow-speed e/m of 1.95, 1.86, 1.95, 1.94, 1.97 [read, Conduction deep read], whereas times the other column it gives 1.95, 1.87, 1.99, 2.14, 2.16 [calc]. So the observed ratios are most probably 3.1, 2.42, 2.0, 1.66, 1.5. Consequences: (a) Thomson's "calculated" column 3.09, 2.43, 2.04, 1.83, 1.65 is not the exact integral of his own point-charge model (2.87, 2.15, 1.82, 1.64, 1.51 [calc]) and sits within about 3% of the Lorentz-Einstein gamma (3.20, 2.37, 1.98, 1.78, 1.62), a function he could not yet have had in this form; (b) the observed lower-speed points agree with his point-charge model, the higher-speed ones lie 8-12% above it. How he got his calculated column is unknown and cannot be reconstructed from either book. This is an open point, not a verdict on the electron model. claims.md claim 2 ("rising with speed as Kaufmann's radium data are said to show") is fair, but the corpuscular-theory deep read's statement that "Thomson's alpha is 8-12% above" his own model at the top three speeds is true only if that column is the calculation; under the better-supported labelling it is the data.
2. claims.md claim 6 and the vortex deep read: "stability of up to 6 vortices" is a correct restatement of Thomson, but Thomson's "seven unstable" is a rounding artefact (already recorded by the deep reader, confirmed here [calc]). Add that the linear test is inconclusive at seven, that eight is the first linearly unstable case (Havelock's classification of the heptagon as neutral: [web] summary at https://rcd.ics.org.ru/S1560354710010028/ and the Montaldi-Tokieda/Cabral-Schmidt literature), and that nonlinear analysis has since shown the heptagon (the "Thomson heptagon") to be stable. [web: search result summaries only, 2026-10-09; Kurakin-Yudovich 2002 attribution not confirmed against the paper, Cabral and Schmidt attribution reported in a secondary source.] So Thomson's conclusion "six is the greatest number" is wrong for seven (stable) though right in spirit (eight and more are unstable).
3. claims.md claim 2 wording "m = (2/3) e^2/a" is correct only in the slow limit and only for the field momentum. Add: the classical electromagnetic mass has two definitions that disagree (momentum gives (2/3) e^2/a; the energy gives m c^2 = U with U = (1/2) e^2 c^2/a, so U = (3/4) m c^2 in Thomson's own numbers, the "4/3 problem"); Thomson does not remark on it [read: Corpuscular deep read section 5; Conduction deep read section 6].
4. facts.json / claims.md claim 4: the table of rings has more problems than the two printing errors listed (13,8,3 twice; no N = 23). Add: the greedy rule does not give the lowest-energy planar arrangements for N = 9, 13, 16, 17, 19, 21, and the outer-ring boundaries at 58/59 and 67/68 on which the valency argument rests do not survive in the full planar model [read: Corpuscular deep read section 5]; Thomson himself calls the agreement of "sum of valencies = 8" with chemistry "quite accidental" [read: p.133]. Step 1's "supported by three counting methods" needs the qualifications in claim 4 below (the methods agree only if one rounds as Thomson did).
5. facts.json should carry the Barkla arithmetic: Thomson's Ne = 10 (from 8.6) and N = 25n (from 21.5) are rounded toward the atomic weight of nitrogen; with his own inputs the count is 21.5 per air molecule, with modern e it is 14 (atomic number, not atomic weight) [calc].
6. Small ones: the Conduction deep read gives the top Kaufmann speed as 2.83 in Table 1 and 2.85 in Table 2 (as printed, harmless); the Positive Rays book gives the helium content of air as 3 x 10^-6 on one page and 4 x 10^-6 on another (modern about 5.2 x 10^-6 [memory, high]); neither changes an argument.

Claim 1: the corpuscle (electron) is universal, has e/m about 10^7 emu/g, is a constituent of all atoms, and emits Zeeman light

a. Mathematically consistent? Yes, as far as I can check. The Lorentz Zeeman derivation is right: m omega^2 - H e omega - mu = 0 gives omega = (He/2m) + sqrt(...), with a first-order shift of He/2m and an exact splitting He/m; Thomson calls omega "frequency" though it is angular frequency [read: Corpuscular deep read section 8]. My worked number in that file (H = 10^4 gauss gives 0.012 nm at 500 nm) is the right size. The argument that an ion (e/m about 10^4) would give a shift about 1700 times smaller is a plain ratio. The e values are inconsistent inside the book: Chapter I gives about 1.03 x 10^-20 emu, Chapter VII uses 1.2 x 10^-20 (modern 1.602 x 10^-20) [read]; the misprint "e = 10^20" for 10^-20 is noted. The corpuscle-radius number a = (2/3) e (e/m) is circular (a is whatever makes (2/3) e^2/a equal the measured mass) [read].

b. Consistent with experiment? Yes: e/m near 1.76 x 10^7 emu/g, same for cathode rays, photoelectrons and thermionic electrons, and Zeeman splittings give the same order and sign [memory, high]. The "Zeeman light comes from corpuscles" argument is sound for normal triplets; anomalous patterns need spin (Lande 1921; Uhlenbeck-Goudsmit 1925) [memory, high], which Thomson's multi-corpuscle normal-mode models do not supply.

c. Tested? Thomson's 1897 e/m measurement, Zeeman 1896-97 and many later ones; replicated extensively [memory, high; the 1897 papers are not in the folder and the Chapter I deep read is lost].

d. If true: nothing in mainstream physics changes; this is mainstream.

Verdict: Established. Split note: "constituent of all atoms" is established; Thomson's picture of how (uniform positive sphere) is claim 4.

Claim 2: all of the corpuscle's inertia is electromagnetic, m = (2/3) e^2/a, rising with speed as Kaufmann's data show; mass is "bound ether" on Faraday tubes

a. Internal consistency. The slow-limit field momentum R = (2/3) mu e^2 w/a is correct (my integral of D x B over r > a; also the energy route (1/3) e^2 v^2/a gives the same mass) [read: both deep reads]. The effective-mass argument m = m' + R/w is correct and measures the transverse mass, the right quantity for Kaufmann's deflection experiment [read]. Problems: - The printed general formula (1) of Conduction does not reduce to the slow limit (confirmed [calc]); the correct closed form is given in the deep read. A misprint or a slip carried from Recent Researches (not checked, book not in the folder). - Thomson's calculated ratios p cannot be reproduced from his stated model (item 1 of the Corrections); the exact integral is 8-14% lower at the higher speeds. - The "bound ether" reading is arithmetically the same as D x B (momentum density 4 pi D^2 w sin(theta)), no separate prediction [read]. - The 4/3 problem: field momentum and field energy give different masses (U = (3/4) m c^2 in his own numbers). A classical point-charge model needs non-electromagnetic stresses (Poincare 1905-06) to be consistent; Thomson's model has none and he does not discuss the point [memory, high; read for the 3/4 relation]. - The inference "p matches alpha, therefore all mass is electrical" is valid given the model (it equals alpha at every speed only if the mechanical mass is zero, [calc in the deep read]) but the model fixes the charge distribution by assumption ("the part of the field outside a sphere of radius a"), and Thomson says so himself. - Thomson himself limits the claim in the same book: the corpuscles cannot carry the atom's mass (Arts. 161, 192; preface) [read]. - Misprint: pulse magnetic force formula lacks a factor e (Corpuscular p.62) [read].

b. Experiment. Kaufmann's five points (beta 0.79-0.95, scatter about 10%) show a steep rise of m/e with speed. They cannot separate Thomson's point-charge law, Abraham's rigid-sphere law and the Lorentz-Einstein gamma: at those speeds gamma fits the printed ratios within 2-4% as well as Thomson's own column does (see Correction 1) [calc]. Kaufmann in 1905 read his data as against Lorentz-Einstein and for Abraham; Bucherer's 1908 experiments favoured Lorentz-Einstein, but a historian's summary reports that the decisive accuracy was reached only much later [web: https://en.wikipedia.org/wiki/Alfred_Bucherer, search summary 2026-10-09; https://arxiv.org/pdf/1205.5951 listed]. Once relativity holds, a mass that rises as gamma says nothing about whether mass is electrical: the strong inference (velocity dependence implies all mass electrical) no longer follows [read: Corpuscular deep read item 4].

c. Tested by Kaufmann (1901-03, Gottingen), then Bucherer, Bestelmeyer and others 1905-09, and precise later work (the 1938 assessment and the 1940s confirmation are as reported in the Bucherer article, [web] same URL). The specific Thomson/Abraham velocity laws were not supported by the later, better data; the gamma law was. Not replicated in favour of the electromagnetic-only mass.

d. If true as stated (all mass electrical): the positive sphere's and every atom's inertia would be field inertia, so E = m c^2 would be a pure electrodynamic identity with the 4/3 factor repaired; mainstream relativity says mass-energy equivalence holds for all energy, so the "mass is electromagnetic" statement is a special case in the opposite direction. The Investigator's lead (Wikipedia, electromagnetic mass; "4/3 problem and Poincare stresses") is the same record [web, Investigator's note]. The Investigator's lead on the Structure of Light/Beyond the Electron (1924-28: electron with lines of force plus guiding waves) shows Thomson kept the Faraday-tube picture but I could not check its mathematics (not obtained).

Verdict (split): - Weak form (a moving charge's field carries momentum that adds (2/3) e^2/a to its inertia): Established. - Specific velocity law fitted to Kaufmann with the point-charge cut-off: Tested, not supported (later data favour gamma; his own calculated column is not reproducible). - Strong form (the whole mass of the corpuscle is electromagnetic): Historically superseded but instructive. It was a reasonable position in 1901-06; the 4/3 problem, relativity and quantum theory replaced it, and the radius a is not independently measurable. - Extension to atomic mass: Refuted by Thomson's own counting in the same book (Art. 355).

Claim 3: Roentgen rays are thin pulses emitted when a charge is stopped; matter conducts because the rays split molecules

a. Consistency. The kink argument is correct: tangential field T = e f sin(theta)/(4 pi V^2 r) (in his units), the Poynting flux V^2 T'H', the Larmor power (2/3) e^2 f^2/V, and the pulse energy (2 mu/3) e^2 w^2/delta [read: Conduction deep read sections 7-8; Corpuscular section 7]. The Art. 363 energy balance is solved correctly (table recomputed to rounding) but is an instantaneous balance, not the Abraham-Lorentz equation: it drops the Schott term, and its "zero initial acceleration" is a feature of that approximation, not a fact [read]. The ratio 2a/delta fails for delta < 2a (more energy than the particle had), and he does not address that regime. The Art. 361 sign of cos(theta) is a convention to confirm. Misprint: magnetic force formula in the Corpuscular Theory lacks e. In the Stokes lecture, the diffraction estimate is correct arithmetic: beta^2 = 2 r r' lambda/(r + r') gives beta = 0.002 inch for r = r' = 4 in and lambda = 10^-6 in, and the quoted 1/250 inch is 2 beta [read: Roentgen Rays deep read section 1.7]. But lambda = 10^-6 inch (about 25 nm) is an assumed thickness, not derived [read]. Thomson's ion-absorption equations in the 1898 reprint are printed with m missing from a denominator (conclusion that absorption fraction is proportional to d is unaffected for fixed ion); the mathematics of the original paper were omitted by its editor [read]. In the 1896 note there is no mathematics: all results are qualitative [read].

b. Experiment. Continuous bremsstrahlung from decelerated charges is exactly the classical result and is accepted [memory, high]. What does not survive: a purely pulse (non-periodic) character of X-rays. Diffraction of X-rays by crystals (von Laue 1912) shows wavelengths near 0.01-1 nm, with the Duane-Hunt short-wavelength cutoff (1915) and characteristic lines, which need quantum theory [memory, high]. The 1896 observations (insulators made conductive for the duration of the irradiation; discharge of both signs of charge; 3-4 V with no lower limit found) are experimental facts consistent with ionisation [read: 1896 deep read]; the "splitting up of molecules" is Thomson's tentative mechanism. Rayleigh's account of the ionisation fewer than one molecule in 10^9 (Investigator) is an early localisation argument [web, via Investigator]; I could not check it.

c. Tested: the 1896 conductivity result was immediately supported by a St Petersburg telegram read at the same meeting [read]; ionisation by X-rays is thoroughly replicated. The pulse theory was tested by diffraction work; the wavelength measurements of 1912 supersede the claim that there is no periodic structure. C. N. Wall 1924 extended the pulse theory (Investigator; not read).

d. If true as the whole story: X-rays would have no wavelength or diffraction pattern, which contradicts crystal diffraction. Nothing in mainstream would change for the bremsstrahlung part.

Verdict (split): radiation by a suddenly stopped charge, and conduction by X-ray ionisation: Established. "X-rays are only non-periodic pulses with no wave character": Historically superseded but instructive.

Claim 4: the atom is a uniform positive sphere with as many corpuscles as the atomic weight, arranged in rings; ring changes give the periodic law, valency and spectral series

a. Consistency. The derived parts check: equilibria for N = 1-4 (Q = 9/10, 21/10, 36/10, 54/10 e^2/a, nearest spacing a, per-corpuscle ratio 6:7:8:9), octahedron stable for 6, cube unstable for 8, hollow planar ring stable only up to 5 (6 weakly unstable), the n^3 growth of the required inner number [calc in the deep read]; three corpuscles at 1/sqrt(3) = 0.577 R, Thomson's ".57" truncated [read]. Weak parts: - The table of inner numbers i(n) for n = 30, 40 is not reproduced by the deep reader's model (about 141 and 345 against 101 and 232), so the 1904 paper's method matters; unknown without the paper. - The ring table has printing errors (13,8,3 twice; no 23) and is a greedy construction not guaranteed to give ground states (N = 9, 13, 16, 17, 19, 21 differ from the minimum-energy planar structures; his 13 = 10,3 relaxes to 9,4) [read: deep read section 5]. - The valency table: the text swaps the valencies of the 62 group; 59 is called electropositive then shown to have valency +0; the model gives period-9 only for the outer-20 series, while the next series (outer 21, N = 68-77) has ten groups and a valency sum of 9, so the model does not give a repeating eight [read: section 8-9]; Thomson says the sum 8 is "quite accidental" [read, p.133]. - The radiation table (Electricity and Matter p.123) is labelled "for each corpuscle" but values match the group total over one corpuscle (off by factor n), and the n = 5 row is ten times too large; the underlying scaling beta^(2(n-1)) is right [read: Electricity and Matter deep read section 5]. The qualitative conclusion that a symmetric ring radiates far less than a single orbiting charge is correct classical electrodynamics. - The energy estimate N (ne)^2/a = 1.02 x 10^19 erg per gram of hydrogen is arithmetically right, but the premise n = 1000 is false (revised by Thomson in 1906); with one electron and a Bohr radius, the same method gives about 2.6 MJ/mol, only twice the ionisation energy (virial factor) [read]. - The "three counting methods": Method I (Barkla scattering): arithmetic rounded as in Correction 5; Method II (cathode-ray opacity): the formula gives P/w = 2.2-3.6 (coefficient 70.6 rather than the printed 67), not "the same order"; Crowther's value 150 or beta-ray values 5-10 would give 0.27 or 0.01-0.02, which Thomson sets aside by arguing they measure total transmission rather than forward-beam loss; Method III (dispersion): hydrogen dispersion gives n near 1 [calc: 0.8-1.6], with a diatomic gas where the formula assumes monatomic; Drude's counts are a few per atom, reconciled by dividing corpuscles into loose valency ones and tight ones [read: Corpuscular ch. VII deep read]. These methods agree to "same order" only after the choices described; they do not independently pin the count at the atomic weight. - Atom-size table (Cuthbertson-Metcalfe): printed ratios right except arsenic (20 vs 20.7) and zinc (30 vs 31.7); the ratio for light elements is irregular, as Thomson says [read].

b. Experiment. Rutherford's nuclear atom (1911, from alpha scattering) superseded the uniform positive sphere; the electron count Z is about A/2, which is what Barkla's own datum gives with modern e [calc, 14 per air molecule]; Z is later fixed by Moseley (1913) [memory, high]. Thomson's central idea, that valency and periodicity come from a limit on stable ring populations and the ease of removing or adding a corpuscle, survives as shell and ionisation-energy ideas, but the numbers do not reproduce the periods 2, 8, 8, 18, 18, 32 [memory, high]. The Investigator's lead (LaFave 2013, J. Electrostatics 71:1029; arXiv:1403.2591: energy patterns of N charges on a sphere correlate with shell filling and ionisation energies) is support by another author for a limited, modern form of the idea; it is a numerical correspondence for the Thomson problem on a sphere, not Thomson's 1904 planar ring model, and I have only the abstract [web: https://arxiv.org/abs/1403.2591]. The 2-D ground states of charges in a parabolic trap (Bedanov-Peeters 1994) are what the deep reader's numerics reproduce; the attribution is [memory, medium], not checked.

c. Tested: Thomson's own magnet experiments (Mayer 1878; floating magnets reproduce small-N polygons and the 6 = 5+1 pattern) test the planar model only at small N and do not report observed vs predicted for N = 9, 13 [read]. Mayer's source for the stability result is tied to Kelvin's one-way analogy (stable magnets imply stable vortices), correctly stated by Thomson [read]. Note the book states that like poles repel "inversely as the distance"; the correct law is inverse square [read]. The ring-atom model as a whole was tested by Geiger-Marsden scattering (1909) and replaced.

d. If true: no nucleus; alpha scattering at large angles would not occur (single scattering by a diffuse positive sphere gives only small deflections [memory, high]). Mainstream physics already rejects it.

Verdict (split): the mathematics of the model (equilibria, ring instability beyond 5, need for inner corpuscles) is Established as mathematics, but the atom model as physics: Historically superseded but instructive. The claim that periodic law, valencies and spectral series follow from it: Not testable as stated in its quantitative form (Thomson's fits are admittedly accidental and use a greedy rule; the claimed spectral ratio laws are stated as expectations with no comparison). The counting claim (corpuscles equal atomic weight): Tested, not supported (modern count about A/2).

Claim 5: positive-ray analysis identifies masses; the 1913 line at 22; "X3" (H3); helium and neon from bombarded matter

a. Consistency. The parabola law e/m = y^2 A/(x B^2) and the ratio form (e/m)1/(e/m)2 = (y1^2/x1)/(y2^2/x2) are correct, and at fixed x a particle of mass M deflects as y/y_H = 1/sqrt(M): so mass 22 sits 4.9% from neon 20 in magnetic deflection [read, calc by deep reader; arithmetic trivial]. The multiple-charge handling (prolongations to 1/2, 1/3, 1/4 of the normal distance for charge 2, 3, 4) is consistent. The residue arithmetic is right: K (39) minus He (4) = 35 (chlorine); Na minus He = 19 (fluorine); Be minus He = 5, C minus He = 8 have no likely impurity [read]. Sensitivity numbers: the book gives the helium in 1 cc of air as 3 x 10^-6 cc on p.130 and 4 x 10^-6 on p.148; the tube gas is about 0.01 cc at atmospheric pressure equivalent; I cannot reconcile "1% of the gas" detection with these without the tube's working volume, which the passage does not give (so I do not check it). The one genuine mathematical weakness is the evidential logic, not the algebra: Thomson's own arithmetic shows a single cc of air entering the apparatus gives a helium line stronger than most he reports [read: p.148-149], so the experiments live below the air-leak level, which makes the controls (below) carry the whole weight.

b. Experiment. The mass-3 ion is today H3+ formed in the discharge (H2+ + H2 gives H3+ + H); it is stable and astronomically important, while the neutral H3 he also argued for was abandoned by the mid-1930s [web: Kragh 2011 Centaurus, https://onlinelibrary.wiley.com/doi/abs/10.1111/j.1600-0498.2011.00237.x, abstract; Kragh 2012 Phil. Trans. A 370:5225, abstract; Investigator's notes]. Thomson's chemical tests (storage for weeks, destruction by light with oxygen, palladium diffusion) were done on a gas that was mostly hydrogen with X3 present in small amounts; they are observations of what the mixture does, and how they map onto H3+ chemistry is not settled by the text [read; deep reader's note]. The 22 line is neon-22, about 9.25% of neon [memory, high]; Thomson's own later hedge, that the two gases may be "indistinguishable in chemical and spectroscopic properties" despite different atomic weights, anticipates isotopes (the word is not in these pages) [read]. This supports claim 5's first and second parts. For helium and neon from bombardment: contemporaries disagreed. Collie, Patterson and Masson (1913-14) and Baly supported disintegration; Strutt, Merton, Egerton and others reported failure to reproduce, and later authors say the claims "failed to be satisfactorily confirmed" [web: Redgrove 1922, https://www.gutenberg.org/files/43240/43240-h/43240-h.htm, via Investigator]. The alternative-science author Nelson writes that "The matter never has been resolved" [web: https://www.alchemywebsite.com/nelson2_6.html, via Investigator; Nelson's characterisation]. My own search for Strutt's 1914 paper returned only secondary summaries, so I cannot quote its result [web: search, 2026-10-09, https://makingscience.royalsociety.org/items/rr_21_53 is the referee file for Collie-Patterson-Masson, where the referee calls the results "currently unexplained" and names Strutt]; the claim that the episode ended with contamination and occlusion explanations is [memory, medium] and remains unverified by me beyond that. Wendt and Irion 1922 (exploding tungsten wires) are cited by the low-energy nuclear reaction community; Thomson's 1922 deflagrated-metals paper is the likely test but I have not read it (Investigator lead; result unknown).

Thomson's controls, recorded as given [read, Positive Rays pp.148-154]: (1) the helium signal is "not appreciable unless the bombardment is prolonged for an hour or so and increases in amount with the length of the bombardment", which an air leak would not do; (2) salts dissolved and evaporated 10-12 times still yield helium, with a half-aluminium blank (electrodes that had sat two days in helium at 3-4 mm Hg gave only slightly more than the unexposed half) arguing that solution removes absorbed gas; (3) the yield depends on the salt and on whether the cathode rays strike it, so the cathode is not the source; (4) KI (no hydrogen) gives more helium than KCl; (5) the rate generally declines with prolonged bombardment, which favours a limited reservoir, and he reports experiments in progress; (6) he says transmutation will not be accepted "until every other explanation has been shown to be untenable" and names the decisive test: detect the residue (Be minus He = 5, C minus He = 8). He gives the "loosened alpha particle" hypothesis as a bridge and does not claim proof. The Investigator's finding that the 1921 second edition still gives both explanations and the same test is consistent with this.

c. Tested: positive-ray analysis was taken up widely and developed by Aston into the mass spectrograph (1919), confirming neon-22 [memory, high]. The helium/neon-from-matter claim was tested by Strutt, Merton, Egerton, Harkins and Wilson, Allison and Harkins and others (Investigator's list; titles and volumes as given by Nelson) with negative outcomes as reported by the sources named; Collie, Patterson, Masson and Baly reported positive results.

d. If true (helium produced from other elements by cathode-ray bombardment at tens of keV): the energy threshold for removing an alpha particle from a nucleus is millions of eV, while a cathode ray of tens of kV carries tens of keV, so the process would need a mechanism unknown to nuclear physics [memory, high; order-of-magnitude, not computed here]. Thomson's own "loosening" is the same story in other words: some atoms already half-decayed. If the H3 claim were true, a stable neutral triatomic hydrogen molecule would be unexplained by valence theory (he says so himself [read: p.146]).

Verdict (split): - Mass analysis by parabolas, and the mass-22 line as a form of neon: Established. - Mass-3 line as the H3+ ion: Established. As neutral H3 (his X3 as "tri-atomic hydrogen"): Tested, not supported. - Helium and neon liberated from bombarded matter by disintegration of the atoms: Tested, not supported as a disintegration claim (replication failures in the published record, with contamination and occlusion the favoured explanation; I have not read the replications myself), while Thomson's controls (time-dependence, the aluminium blank, repeated solution and evaporation) are on record and did not by themselves exclude a small residual helium source. The reviewers cannot settle this from the folder; a controlled test with ultra-clean apparatus (Allison and Harkins 1924 is the Investigator's pointer) is the decisive record.

Claim 6: vortex rings in a perfect fluid interact by computable laws; a gas of vortex atoms gives kinetic-theory pressure and might explain chemistry

a. Consistency. This is the part of the project that I can judge most as mathematics, and it is mostly sound. Method: far-field (dipole) approximation of each ring, first-order perturbation along the undisturbed straight paths, retaining the largest term of mixed-order coefficients [read: Part II deep read]. Results that follow exactly from the displayed equations: in-plane bending (69), the sin 3 phi radius rule (74, with k^4), the 60 degree and 30 degree boundaries, the Class I/II summary, the sphere results of sections 37-38, the 1/c^3 law for deflection, and the 1/c^6 law for the sphere [calc by deep reader; my check of the time integrals]. Errors: eq. (71) is three times too small (confirmed [calc]); this reverses the comparison on p.75: corrected, rings whose paths just fail to cross (c = g) are deflected at least as much as rings whose paths cross; (74) and the g = 0 line on p.77 print k^3 where k^4 is needed (dimensionally a length² per time, not a length); the sphere's total angle lost the factor 1/u; eq. (75) and the section 32 impulses have inconsistent coefficients and c^2 where c^3 is needed; the "repel a ring passing by" remark rests on an averaged cos 3 phi that he does not show and which assumes uniform spread in phi; his n = 2 vibration uses log(2b/e') against log(8b/e') above [read]. Linked rings: the quick/slow frequency formulas (94-99), the unequal-ring constraint m log(d/e) = m' log(d/e') and the moment of momentum (105) check [calc]. The n-vortex circle analysis: his linearised equations (107), (111) match a numerically computed Jacobian to rounding, and the circulant factorisation is right; three real defects: seven vortices are marginal at linear order, not unstable (his +.002 comes from rounding beta to -2.130; full precision gives exactly 0 [calc to 50 digits in the deep read; my closed form confirms the zero]); for n = 5 and 6 the first listed period is a spurious k = 0 root excluded by his own constraints (harmless to stability); misprints (b' for n = 5 printed with r instead of r^4; the "-gamma" shorthand in (123); a cos 8 pi/7 for 4 pi/7 in the explicit line) [read]. The Part III summary "small compared with sqrt(I^3/m)" drops the density [read]. The linear criterion is correct in the closed form: stable only if k(n-k) <= 2(n-1) for all k, which holds for n <= 7 and fails from 8 [calc]. The vortex-sheet argument for a largest stable number is sound qualitatively (Kelvin-Helmholtz), and the exact spectrum's growth rate rising like n^2 agrees [calc in the deep read].

b. Experiment/physical consistency. The hydrodynamics (ideal fluid, vortex ring momentum and speed, the Kirchhoff-like n = 2 rotation, the attraction of a ring to a solid) is standard and correct in its regime [memory, high]. As atomic theory the vortex-atom idea (Kelvin 1867) was abandoned: atoms are not rings in an ideal fluid, and in the real world viscosity and the absence of such a medium matter [memory, high]. The mathematical descendants are real: the ring-of-vortices polygon problem (Thomson heptagon) and Mayer's magnets; the Investigator's Wikipedia pointers (Vortex theory of the atom; the Thomson problem as a descendant of the 1904 ring atom; Smale's 7th problem) apply here [web, via Investigator]. The chemistry and kinetic-theory claims (a gas of vortex atoms and chemical combination) are in Parts I and IV, which I have not seen except by digest; I do not assess them.

c. Tested: not as atomic theory. As fluid mechanics, smoke-ring collision experiments (Tait, Kelvin's lecture demonstrations) are the historical analogues [memory, medium; not checked]; no controlled test of the quantitative far-field laws is on record in the folder. The heptagon question was settled later by nonlinear analysis [web: see Correction 2].

d. If true as atomic theory: the physical vacuum would need to be an ideal fluid in which atoms are permanent vortex rings; this conflicts with quantum mechanics and spectroscopy. Mainstream already rejects it.

Verdict (split): the fluid-mechanical computations (with the errors listed): Plausible, untested as stated for physical rings, and Established as mathematics once corrected; the claim that seven vortices are unstable: Internally inconsistent with his own exact equations (rounding artefact); the claim that vortex atoms give matter, chemistry, kinetic pressure: Historically superseded but instructive.

Hidden assumptions and unit notes, collected

• Units: Thomson works in electromagnetic units (e_emu = e_esu / c); readers must convert e^2 -> e^2 / c^2 for Gaussian. Electricity and Matter p.126 says "measured in electrostatic units" while p.111 and p.122 use emu forms: a unit-system switch inside one book [read].
• The 1906-07 counting methods use e = 1.2 x 10^-20 emu, 25% below the modern value; N/n scales as 1/e [calc].
• All stability statements are linear (Vortex Rings; ring atom), and the one place where linear theory cannot decide (seven vortices) is the one where a rounding error produced a verdict.
• "Dragged ether" is an interpretation, not an extra prediction.
• The Darwin periodic-orbit argument for sharp spectral lines is a qualitative analogy; no orbit is calculated for an atom [read].

Investigator's findings carried into this review

• Positive (own work): Nature 90:645 (13 Feb 1913) reply to Ramsay, Collie and Patterson; 1921 second edition of Rays of Positive Electricity; Proc. A 101:290 (1922); Structure of Light (1925), Beyond the Electron (1928) with the electron's "dual structure" (lines of force plus guiding waves). These bear on claims 2, 3 and 5; none was available to me beyond the Investigator's description.
• Supporting by others: Kragh on H3 and H3+; LaFave on shell filling; C. N. Wall 1924 on the Stokes-Thomson pulse theory.
• Related work: the helium/neon controversy 1913-1924 (both sides, kept in claim 5), Wendt and Irion 1922 and the LENR literature, the Thomson problem, the 4/3 problem.
• Historical record: CUL MS Add. 7654 and Trinity THMJ/THMF (notebooks that would hold the raw positive-ray data); Morrisson's account of the 6 Feb 1913 meeting and Frankland's remark. Absence of evidence worth noting: I found no later paper by Thomson reporting residue atoms from carbon or beryllium bombardment (the test he named), and his 1922 deflagrated-metals result is unread; the record on how the helium work ended is a thin, secondary one and should be obtained (CUL notebooks; Proc. A 91:180 for Egerton; Allison and Harkins 1924). Competing explanations for the silence: the effect was contamination (the work was dropped as understood), or the effort moved to the mass spectrograph (Aston, 1919), or the work was not published; the record that would decide is the notebooks and the 1922 paper.
• Not reached by the Investigator: Kaufmann 1901-06 and Bucherer 1908 original papers (claim 2), Barkla and the 1906 corpuscle counts (claim 4), Thomson 1881 vortex-ring paper. I used the deep reads and short web checks for these.

Web checks done (budget about 6)

1. Thomson heptagon: summaries of nonlinear stability (Kurakin-Yudovich; Cabral-Schmidt; Montaldi-Tokieda) and of Havelock 1931: https://rcd.ics.org.ru/S1560354710010028/ and https://arxiv.org/pdf/1210.5662 (listed in results, not opened); retag [memory medium] to [web, search summary].
2. Kaufmann, Bucherer and the status of the early mass-velocity data: https://en.wikipedia.org/wiki/Alfred_Bucherer (search summary).
3. Strutt's 1914 negative paper: not retrieved; only a Royal Society referee file for Collie-Patterson-Masson (https://makingscience.royalsociety.org/items/rr_21_53) and secondary summaries. The check partly failed; the statement that Strutt, Merton and Egerton failed to reproduce rests on the Investigator's notes (Redgrove 1922, Nelson 2006) and my [memory, medium]. Not checked on the web: Bedanov-Peeters 1994 for planar ground states ([memory, medium]), Mayer 1878 ([memory, high]).

Summary table