Fedderson-Daedalus: PhD physicist review

Source gap (repeated from Step 1). Neither PDF has a text layer. Feddersen's own 1857-1862 papers are not in the folder. The file "Daedalus-Fedderson-109.pdf" is the 1890 Proceedings of the American Academy of Arts and Sciences vol. 25, not Feddersen and not the journal Daedalus. Claims 1 and 2 therefore rest partly on [memory]/[web] secondary sources, not on Feddersen's text. What I read myself, from rendered page images: Trowbridge and Sabine pp. 122-123 (conclusions), Trowbridge pp. 193-194, and Rich 1917 pp. 142-155 in full (including the middle pages Step 1 skipped). I did NOT read Trowbridge-Sabine pp. 109-121 (apparatus and data), so claim 2's evidence base is only their stated conclusions.

Source tags: [read] = page image read by me; [web] = URL given; [memory] = recalled, with confidence.

Corrections to Step 1

1. Cu-Zn numbers are swapped in claims/facts. Step 1 wrote "Cu+Zn- 1.27 vs Cu-Zn+ 4.77". The table (Rich p.154) heads the columns Cu-Zn+ = 1.27 (14 trains) and Cu+Zn- = 4.77 (143 trains) [read: Rich p.154]. Also the 1.27 divides by only the 11 half-cycles that had any spark: 21 of 32 half-cycles in that column produced no train, so per half-cycle it is 14/32 = 0.44. The other tables divide by all half-cycles, so the Cu-Zn averages are computed on a different basis.
2. Polarity convention (Step 1 flagged it as ambiguous). Resolved as far as the page allows. All tables agree with each other: the favoured column is the one where the first-named metal is negative in Fe-Bi+, Fe-Zn+, Cu+Bi-, Cu+Zn-, Cu+Fe-. So iron, bismuth and zinc start sparks more readily as the negative electrode (cathode), each against the metal it is paired with, and copper least readily. But Rich's prose repeatedly calls the favoured electrode the "anode" (e.g. "half-cycles in which iron was initially the anode", "from the bismuth anode", "zinc anode", "when iron was anode", pp. 150-153), which contradicts his table headings. His summary then says electrons are "emitted from iron more easily", which matches cathode. The paper is internally inconsistent in its wording; the tables are the reliable part [read: Rich pp.150-155].
3. Pages 13-19 of the scan (Rich pp. 144-151) have now been read. They contain: the three predictions (trains per half-cycle, sparks per train, odd/even count), Rich's own admission that primary-current flux changes could favour one polarity "whether the electrodes were of the same material or not" (p.145), the larger film-camera apparatus (25 mH coil, 0.125-0.5 microfarad condenser, film on a 40 cm wheel at 2000 rpm, 42 m/s, gap 0.1-4 mm, two years of work, p.146), the Cu-Fe table (Cu+Fe- 205 trains/30 half-cycles = 6.2; Cu-Fe+ 101/32 = 3.16, p.148), the Cu-Cu control (31 half-cycles: 63 trains = 2.03; 33: 74 = 2.24, symmetric, p.149), and a DC polarity test (221 V DC make/break with a kick-of-voltmeter determination of polarity, electrodes reversed on the same film, pp.150-151). Step 1's facts.json lists these as unread; they are the paper's best controls.
4. Step 1's key claim "number of sparks per train and odd/even are all asymmetric in the predicted direction" overstates. Odd/even percentages are not consistent across pairs (Cu-Bi: 88% odd vs 97% odd; Fe-Zn: 80% vs 95%; Cu-Zn: 86% vs 45%), and Rich himself writes "Not much reliance can be placed on the relative number of even and odd spark trains" and "no reliance can be placed" [read: Rich pp.151, 154].
5. Minor: Trowbridge's "1/100 of a wave length" (p.193: "a movement of 1/100 of a wave length") is literally what the page says, but it cannot mean 1/100 of an optical wavelength as the resolution of the instrument in any useful sense (see claim 3). The sentence is ambiguous; I treat the numerical sensitivity as unknown.
6. I checked the arithmetic of every Rich table (train counts times sparks-per-train sum to the printed totals; averages recompute). All internally add up; the only printed slip is Fe+Bi- sparks per train 2.21 (184/83 = 2.22), trivial.

Everything else in facts.json that I checked (title page identity, Trowbridge-Sabine conclusions 1-4, Trowbridge periods 3e-7 s and 2.4e-6 s, 6 mm distance, 20,000 lines/inch grating) matches the page images.

---

Claim 1. Condenser discharge through a spark gap is oscillatory, period 2 pi sqrt(LC), seen in a rotating mirror (Feddersen)

• a. Mathematics. Consistent. A series L, C, R circuit obeys L q'' + R q' + q/C = 0, with underdamped solution of angular frequency sqrt(1/LC - R^2/4L^2); the period is 2 pi sqrt(LC) when R is small. Worked number: L = 25 mH, C = 0.125 microfarad (Rich's circuit, p.146) gives T = 2 pi sqrt(0.025 x 1.25e-7) = 0.35 ms; for C = 0.5 microfarad, 0.70 ms (computed). Note these are low-frequency L values; Rich notes the high-frequency inductance differs, so this is only an order of magnitude.
• b. Experiment. Agrees with all later work: Hertz's spark-gap radio experiments (1887-88) and every spark-transmitter in the Marconi era relied on damped LC oscillation. Kelvin predicted it in 1853 and Feddersen confirmed it with a rotating mirror [web: https://en.wikipedia.org/wiki/Berend_Wilhelm_Feddersen via search result, years quoted 1857-1859; I could not open that page directly (404), the search summary gave 1857-59, so date confidence is medium; Step 1's 1857-1862 span is [memory] high-medium].
• c. Tested? Yes, by Feddersen himself, then by Trowbridge and Sabine 1890 with a revolving mirror, and by Rich 1917 (his Fig. 9 and tables show trains of 2-8 oscillation sparks separated in time) [read]. Widely replicated, including every spark-gap transmitter.
• d. If true, what changes? Nothing; it is mainstream.
• Verdict: Established.

Claim 2. Measured periods deviate from 2 pi sqrt(LC); attributed to "dielectric viscosity" and energy transformation in the air

• a. Mathematics. The formula is for a lossless, linear circuit with lumped constants. A discrepancy is expected: the spark gap is a nonlinear, time-varying resistance, and the circuit has stray capacitance and an inductance that depends on frequency (skin effect, iron wire). The authors' analogy to Laplace's correction of the speed of sound (isothermal vs adiabatic) is rhetorical; no quantitative model is given in the conclusions [read: pp.122-123]. A "hysteresis of the air dielectric" is not given equations in the part I read.
• b. Experiment. Air at these fields is a linear, essentially lossless dielectric to very good accuracy before breakdown (permittivity near 1.0006; negligible polarization loss at kHz-MHz). Real period shifts in spark circuits are well explained by spark-channel resistance and circuit stray elements [memory: high confidence, textbook spark-transmitter engineering]. "Dielectric viscosity" for glass is real in a loose sense (dielectric absorption, Rich p.146 notes his window-glass condenser had "very bad" absorption) but for air it is not supported.
• c. Tested? Only by the authors' own measurements. No replication of the "dielectric viscosity of air" interpretation that I could find; I did not search extensively [web: https://zenodo.org/records/2386894 lists the paper "Electrical oscillations in air" by Trowbridge and Sabine; no content]. A contemporary criticism of Trowbridge exists (a Phil. Mag. note, https://zenodo.org/records/1833238, title only seen).
• d. If true: Air would need frequency-dependent polarization lag at fields below breakdown, which would show in capacitor loss-tangent measurements; none are seen.
• Verdict: split. The observation that real oscillations differ from the ideal formula is Established (and the mainstream explanation is spark resistance and stray elements). The specific "dielectric viscosity of air / hysteresis" explanation is Historically superseded but instructive.

Claim 3. Metal atoms are not carried to and fro by the oscillating spark; iron lines show no Doppler shift (Trowbridge 1890)

• a. Mathematics. The Doppler formula quoted, lambda1 = lambda (V/(V+delta)), is fine [read: p.193]. Worked check: if atoms crossed the 6 mm gap in a half period (T/2 = 1.5e-7 s for T = 3e-7 s), v = 4e4 m/s, v/c = 1.3e-4, shift 0.53 angstrom at 4000 angstrom ("a whole tenth meter" in his units presumably of the plate scale). For T = 2.4e-6 s, v = 5000 m/s, 0.067 angstrom. A 56-amu iron atom at 4e4 m/s would need about 460 eV of kinetic energy, unrealistic for neutral or ion motion driven by a MHz field. Atoms are heavy and the field reverses faster than they can respond, so the null is expected by simple mechanics.
• b. Experiment. Consistent: modern spark spectroscopy attributes line broadening to Stark and thermal Doppler effects (thermal Fe speed at 1e4 K about 1.7 km/s, Doppler width about 0.04 angstrom, computed), not to to-and-fro motion at the oscillation frequency [memory: medium-high].
• c. Tested? Once, by Trowbridge, with a Rowland concave grating (20,000 lines/inch), photographs with the hollow electrode along and across the slit on the same plate showed no displacement [read: pp.193-194]. No quantitative limit with error bars is given. Not replicated by anyone I found.
• d. If true: Nothing. This is the expected null.
• Verdict: Established as a null result (consistent with simple mechanics and modern spark spectroscopy), though the 1890 experiment alone is weak (sensitivity not clearly stated) and a different question from whether electrode vapour moves at all.

Claim 4. Unlike-metal spark gap starts more readily on one polarity, so it partly rectifies (Rich 1917)

• a. Mathematics. No derivation to check; the argument is a counting scheme. The logic of Rich's three predictions (pp.144-145) is reasonable but not rigorous: a higher striking voltage on one half-cycle gives fewer trains per half-cycle because the condenser must charge further. His own caveat (primary-current flux asymmetry, p.145) is a confound he does not remove except by the Cu-Cu control.
• b. Experiment. Compatible with accepted physics. Rectification by asymmetric electrodes is real (point-plane corona and spark polarity effects; cathode-spot formation in arcs; the mercury-arc and Cu-oxide rectifiers) [memory: high]. Breakdown in a Townsend discharge depends on the cathode through the secondary-emission coefficient gamma, "strongly dependent on the cathode material, surface roughness, impurities" [web: https://en.wikipedia.org/wiki/Paschen%27s_law]; at larger gaps streamer breakdown makes the dependence weak, which fits the earlier null results at ordinary gaps. Rich's gaps were 0.1-4 mm [read: p.148], intermediate.
• c. Tested? By Rich only (single lab, 1915-1917, a PhD dissertation under Guthe). Controls present: Cu-Cu symmetric (2.03 vs 2.24 trains per half-cycle), DC make/break polarity check, many specimens over two years [read]. Statistics, my Poisson estimate of the difference in train rate in standard errors (computed): Fe-Bi 4.4, Cu-Bi 3.4, Fe-Zn 4.4, Cu-Zn 10, Cu-Fe 6.4, Cu-Cu -0.6. These are not trivially noise, but half-cycles within a film are not independent samples, one film per pair is shown as "typical", and there is no blinding, no oxide/surface control, no field-geometry control and no humidity or temperature record. I know of no replication [no source found; absence of evidence, not evidence of failure].
• d. If true: Nothing mainstream changes; the effect is the known cathode dependence of breakdown, here at the sub-4 mm regime. It would only contradict the old statement (Righi, Peace, Carr) that sparking is independent of electrode metal, and Rich himself notes those were at different gap lengths.
• Verdict: split. That some asymmetry exists with unlike metals: Plausible, untested (consistent with known physics, one uncontrolled-by-modern-standards study, not replicated on record). Rich's claim that the odd/even pattern confirms the mechanism: Not testable as stated in practice (he disclaims it himself).

Claim 5. Ordering Fe, Bi, Zn, Cu for ease of spark initiation, interpreted as electron emission; electrode material is not negligible

• a. Mathematics/consistency. Tables are arithmetically consistent (Step 1 correction 6). The interpretation is internally inconsistent in wording: prose says favoured electrode is the "anode", tables say it is the negative electrode, and the summary says electrons are emitted more easily from iron, which requires cathode (correction 2). Further, a rectification series is only meaningful if it is transitive: I saw Fe-Bi, Fe-Zn, Cu-Fe, Cu-Bi, Cu-Zn tables, but no Bi-Zn table in the pages I read, though the summary states Bi above Zn "in all possible combinations" [read: p.151, 155]. The Cu-Zn averaging basis differs (correction 1).
• b. Experiment. If the cause were simple electron emission, the order should follow work function. Values [web: https://en.wikipedia.org/wiki/Work_function, ranges by method]: Fe 4.67-4.81, Bi 4.31, Zn 3.63-4.9, Cu 4.53-5.10 eV. Fe has among the HIGHEST work functions here yet is ranked easiest; Bi (4.31) is below Fe. So the ordering does not track work function. More plausible [memory: medium, not web-checked]: after the first discharge, cathode-spot formation and metal-vapour supply, which depend on melting/boiling point, thermal conductivity and surface oxide, control restrike; Zn and Bi are volatile, Cu has high thermal conductivity. This is my suggestion, not a tested result.
• c. Tested? Only by Rich; Guthe's 1901 coherer series (Bi, Fe, Zn, Cu) is a different device, from [read: Rich p.154-155]; Rich notes it differs only in Bi/Fe positions.
• d. If true: Nothing in mainstream physics, only a statement about cathode phenomena in air.
• Verdict: split. The empirical series, as measured: Never properly tested (no neutral replication on record; not dismissed without testing, rather simply uncited later in what I found). The electron-emission-by-work-function interpretation: Tested, not supported is too strong (no one tested it), so I class it Refuted by calculation only in the weak sense that the ordering does not follow tabulated work functions (the table above); the mechanism remains open. The "anode/cathode" wording is Internally inconsistent.

---

Summary table

Confidence note: web checks done (4): work function values, Paschen/gamma, Feddersen dates (secondary search summary only), Trowbridge-Sabine paper existence. Not web-verified: Rich replications (none found, search not exhaustive), cathode-spot mechanism, spark-channel explanation of period shifts.