Physicist review: Kennard, "On Unipolar Induction" (1917)

Reviewer: PhD physicist (Step 2). Source: E. H. Kennard, Phil. Mag. ser. 6, vol. 33, pp. 179-190 (1917; signed University of Minnesota, August 1916). Tags: [read] = in the document or a saved file; [calc] = my own recomputation (sympy/numpy, this session); [web] = web source with URL; [memory] = recalled, with confidence.

No source gap: the whole paper was read. Pages 1-9 were deep read and checked against page images; pp.10-12 (section 7, "Fate of the Moving-Line Theory", and section 8, the conclusion) I read myself in the cleaner Internet Archive OCR [read: research/files/Kennard-1917-PhilMag-33-179-IA-OCR.txt]. Step 1's caveat that the aether claims rested on the digest alone is therefore closed.

0. The physics in plain terms (for the reader profile)

A unipolar (homopolar) generator is the Faraday disk: a conductor spinning about the axis of a magnet whose field is symmetric around that axis. A charge q moving with velocity v in a magnetic induction B feels the Lorentz force F = q(v x B) (in Gaussian units there is a 1/c; Kennard writes the "electromotive intensity" as (1/c)[v x B]). Here v x B is the vector cross product. Dividing by q gives an effective electric field in the moving metal.

The old question (1900s-1910s) was: v relative to what? Three answers were on the table [read: p.1]: 1. Lorentz/Faraday: relative to the frame in which the field is described (for Lorentz, the stationary aether). If the magnet rotates about its own symmetry axis, the field pattern is unchanged, so rotating the magnet does nothing to the field. 2. "Moving-line" theory: the magnetic "lines of force" are material-like and carry the field around with the magnet, so only motion relative to the magnet matters. 3. Hertz: no charge in this condenser case.

Kennard's apparatus (fig. 1) is a cylindrical condenser (two coaxial metal cylinders) inside a long coaxial solenoid (a coil), either of which can be spun separately. The three predictions are different and measurable:

Result: both together charges the condenser; solenoid alone does not. This eliminates the moving-line theory and is what Maxwell-Lorentz electromagnetism predicts.

Worked numbers [read: p.7, checked by [calc]]: 25 A in 1150 turns over 22.5 cm gives B = 4 pi (1150/22.5)(25)/10 = 1606 gauss (printed 1600). Flux through the 3.32 cm radius outer cylinder = 1606 x pi(3.32)^2 = 55,600 maxwells (printed 55,400). At 33 rev/s the induced potential difference between axis and outer cylinder is V0 = 55,400 x 33 x 10^-8 V = 0.0183 V. The rim speed is 2 pi (33)(0.0332 m) = 6.9 m/s, so v/c is about 2 x 10^-8. The effect is first order in v/c and needs no relativity to compute. The charge, -6.22 V0 in electrostatic units, gives about 3.8 x 10^-4 e.s.u. on a total capacity of 138 e.s.u., i.e. 2.7 x 10^-6 e.s.u. = 0.82 mV, which at 18,000 mm/V is a 14.8 mm electrometer deflection (printed 14.9 mm) [calc].

1. Kennard's own claims, in the order of claims.md

Claim 1. A cylindrical condenser rotated inside a magnetised coaxial solenoid becomes charged as Lorentz's theory predicts, even though a metallic screen rotates with it.

a. Mathematics. Internally consistent. The core statement (a metal point at flux-encircled N and n rev/s sits at potential Nn, in e.m.u.) is just the Faraday-disk EMF, E = (1/c) v x B integrated along r, giving V(r) = omega B r^2 / 2 for uniform B [read: p.5; calc]. The electrostatic boundary-value problem is a standard one (Laplace's equation in a cylindrical shell with prescribed boundary potentials); only the outer-space solution is printed. I did not re-solve it. As an order-of-magnitude check, the infinite coaxial capacity for a 15.1 cm inner cylinder of radius 0.77 cm in an outer cylinder of radius 3.32 cm is 15.1 / (2 ln 4.31) = 5.2 e.s.u. [calc]. Charge on the inner cylinder is then roughly -5 V0 for an infinite coax, against Kennard's -6.22 V0 including guard rings, end plates and the shaft region. Same order, difference plausible from end effects; not independently reproduced. The two printed arithmetic slips (0.117 for 0.114; 0.140 for 1.40) are real [calc; confirmed Step 1] and do not touch the 0.231 ratio.

b. Experiment. Consistent with everything established about the Lorentz force and Faraday-disk behaviour. A metallic screen co-rotating with the condenser cannot cancel the effect: the screening is electrostatic and the v x B force acts on the charges in the metal itself, whichever way they are arranged; the net charge is fixed by the electrostatics of the boundary potentials. This is textbook (Griffiths-level treatments of the Faraday disk and "exceptions to the flux rule") [memory, high confidence]. A recent PRResearch paper states the Barnett-Kennard-Pegram result as "a standard part of the electromagnetics literature" [read: Chyba, Hand & Chyba 2025, Phys. Rev. Research 7, 013285, research/files; their wording, and they are interested parties for Earth-rotation power generation, see section 3].

c. Tested/replicated. Kennard (1917, this experiment); Pegram, Phys. Rev. 10, 591 (1917), independent, condenser and coil rotating together, "found to become charged in accordance with the theory" per Barnett [read: Barnett 1919, Trans. AIEE, research/files, p.1510]. Pegram's own paper was not obtained by the Investigator (access blocked), so Pegram's numbers are unread by me. Barnett's 1919 report names both Kennard and Pegram as confirming the corollary.

d. If true: nothing in mainstream physics changes. It is already mainstream.

Verdict: Established. (For the experimental charging; the interpretation is claim 6.)

Claim 2. Measured electrometer/galvanometer ratio with condenser and solenoid rotating together is 0.242 against calculated 0.231 (5%), and 0.228 at half speed (1.4% low), using a computed charge of -6.22 V0 (+/-0.5%).

a. Mathematics. Consistent. I recomputed the theoretical ratio from the printed equation (20.5/62.2)(0.1 V/0.1197 V0)(6.22 V0/7.42 V) = 0.2308 [calc]. Step 1 correctly caught that the OCR "26.5" is really 20.5 (with 26.5 the ratio would be 0.298) [calc, 0.2984]. Percentages: 0.242/0.231 = +4.8% (printed "5%"); 0.228/0.231 = -1.3% (printed "1.4%", small rounding difference; Kennard may have used unrounded 0.2308 and 0.228, which gives -1.2%; the point is immaterial). Mean of the four printed ratios 0.237, 0.237, 0.236, 0.258 is 0.2420 [calc].

b. Experiment. Consistent with Lorentz-Maxwell. The agreement is better stated than it is discriminating. Two things limit its precision, both Kennard's own: individual groups ranged from 0.195 to 0.271 (alternative method) and 0.211 to 0.263 (half speed) [read: p.9], i.e. scatter of order +/-15% on single sets, and the 13 A and 25 A groups gave 0.25 and 0.27, an 8% difference [calc] that Kennard does not comment on (he calls those values "not comparable with the previous values" without saying why). He concludes the results "remain somewhat rough" [read: p.9]. The claimed +/-0.5% theory accuracy is therefore far tighter than the data; the 5% offset is about 10 times the claimed theory error, but within the experimental scatter. The discriminating power is not 5%: the alternatives (moving-line, Hertz) predict zero, so the observed ratio of about 0.24 against a predicted 0.23 versus 0 is a decisive discrimination, with the 5% agreement a bonus.

c. Tested/replicated. Single run by Kennard. Pegram (1917) independently saw charging; his quantitative agreement is not known to me.

d. If true: nothing changes.

Verdict: Established for the qualitative result and the order-of-magnitude (about 20%) quantitative agreement; the stated 5% and 0.5% precision claims are not supported by the scatter in the paper (a precision caveat, not a refutation).

Claim 3. Rotating the solenoid alone produces no charge on the condenser (reversing average +1.2 mm against about 15 mm for the full effect), confirming Barnett.

a. Mathematics. Consistent. Under Maxwell-Lorentz the solenoid's B field is axisymmetric and steady, so rotating it about its axis leaves B(r, z) unchanged; the stationary condenser sees no change. Checked [calc]: sign-corrected row effects are -0.4, +0.6, +1.1, +3.5 mm, mean +1.2, standard deviation 1.7, standard error 0.8 mm, i.e. about 1.5 sigma from zero and 8% of the 15-16 mm full effect [read: p.8 table; Step 1's transcription from the page image, which I trusted]. An earlier group gave +0.2 mm [read: p.8].

b. Experiment. Consistent with Barnett's 1912 null result [read: Barnett 1912, 1919] and with the standard lore that the field of an axisymmetric magnet does not turn with the magnet.

Limitation I add, [my inference from the table]: in the solenoid-alone rows the galvanometer reads 0, because the speed monitor (a Faraday disk on the condenser shaft) is not rotating. Rotation speed in those rows was therefore not recorded by the apparatus; comparison to "about 15 mm" assumes a similar speed. Kennard says as much ("No means being at hand to record the instantaneous speed", p.4). This weakens the null only slightly, since 1.2 +/- 0.8 mm would remain small at any plausible speed.

c. Tested/replicated. Barnett 1912 (original), Fehrle 1913 (Kennard says it is "marred by several results which must certainly be wrong", p.2; this is the characterization of a competitor in the same dispute), Kennard 1917. Barnett and Kennard were rivals over the interpretation, which makes their agreement on this fact more, not less, informative.

d. If true: nothing changes.

Verdict: Established.

Claim 4. The condenser-alone case, not completed because the solenoid short-circuited, is inferred by superposition (both minus solenoid alone), with a partial 14 mm check.

a. Mathematics. Sound within linear Maxwell-Lorentz theory: the force depends on the charge's velocity and the field, and the field does not depend on the solenoid's rotation, so the effect depends only on the condenser's motion. "Condenser alone = both together - solenoid alone" holds. It also holds in the moving-line theory (for its own different numbers), so it is a bookkeeping step, not a hidden assumption.

b. Experiment. The inference agrees with the independent observation of Barnett, whose original experiments in 1912 were of condenser rotation; Barnett's 1919 report describes the combined result as "a corollary of my experiments" and "superposed" [read: Barnett 1919, p.1510], so this is not new physics either.

The partial check is an electrometer deflection of 14 mm "under about the same conditions" [read: p.9], against 14.9 mm predicted at 25 A and 33 rev/s [calc]. No galvanometer normalisation is given, and speed and current are loosely specified, so the 14 mm check is a rough consistency, not a measurement of the ratio.

c. Tested/replicated. Not completed by Kennard (the solenoid failed and "it was not thought worth while to rewind it" [read: p.9]). Barnett measured the condenser rotating while the field source was stationary (1912 series, per his 1919 account).

d. Nothing changes.

Verdict: Plausible, untested by Kennard directly (inference from two measured cases plus one rough 14 mm run); the physical content is Established through Barnett's earlier condenser-rotating results.

Claim 5. The "moving-line" (relative-rotation) theory is disproved: its electromotive intensity [v x B] does not satisfy Laplace's equation in free aether, and the open-circuit experiments contradict it.

a. Mathematics. The key calculation is correct and I reproduced it. For rigid rotation v = omega x r about z in uniform B = B z-hat, v x B = omega B r r-hat (a radial vector of magnitude proportional to r). Its divergence is div(v x B) = 2 omega B, which is not zero [calc, sympy]. In empty space (no charges) an electric field satisfies div E = 0, so a vacuum field cannot equal (v x B) alone. Kennard's argument: if the "lines" drive an EMF of that form in free space (as Barnett assumed together with div(KE) = 4 pi rho), the two assumptions contradict each other; so one must write div(KE) = 4 pi rho "with Lorentz" and the lines act only on material bodies [read: pp.188-189, IA OCR]. This is a valid internal-consistency criticism of the specific Barnett-style version of the moving-line theory.

It should be said what the argument does and does not show. It shows the version of the theory that gives the lines an effect on free space is inconsistent with Maxwell's equations. It does not by itself rule out a modified theory with different assumptions; the experiments do that.

b. Experiment. The corotation positive result and solenoid-alone null result are exactly what moving-line theory forbids, since it needs only relative motion [read: p.2 predictions]. Mainstream view: "the field of a rotating axially symmetric electromagnet does not rotate with the magnet" [read: Chyba et al. 2025, their characterisation]. A further point from standard physics [memory, high]: for a steady axisymmetric field, the idea "the lines rotate with the magnet" has no physical content, because there is nothing in a steady symmetric field to tell lines apart; an observer cannot tell whether the lines are turning. That is why modern texts call the field lines a visual aid with no velocity. This is a point where later physics improves on both Kennard and Barnett.

Barnett's 1912-1913 papers call Kennard's 1912 work "inconclusive" and "based on incorrect theory", and say "Mr. Kennard's fallacies are quite apparent" [read: Barnett 1912/1913, as quoted in the Investigator's lead 4]. That is the characterisation of a direct rival in the dispute. His technical objection (in 1912 only the iron core rotated and the magnetising coil stayed fixed, so the fraction of "lines" that moved is unknown) is answered by the 1917 design in which the solenoid itself rotates. Barnett's objection was reasonable for 1912; the 1917 experiment removes it. By 1919 Barnett concedes that the corotation experiments leave two options, "(1) the ether does not exist and the tubes of induction do not move, or (2) the ether exists" [read: Barnett 1919, p.1511], so the moving-line theory in Barnett's treatment is also dead by his own account, though not for the reason Kennard prefers.

c. Tested/replicated. Barnett 1912 (solenoid-alone null), Kennard 1917, Pegram 1917 (corotation). Positive tests.

d. Nothing changes; moving-line theory is what had to go.

Verdict: Refuted by calculation (the specific vacuum-EMF version) and Tested, not supported (the moving-line theory in general). Split verdict: strong form (line-carried EMF in free space) is internally inconsistent with Maxwell's equations (div(v x B) = 2 omega B); weak form (field simply co-rotates with its source) was tested in 1912-1917 and rejected.

Claim 6. Induction therefore depends partly on absolute rotation, and the radial force on electrons in a rotating condenser can only be explained by motion relative to a stationary Lorentz aether.

Split into three parts.

6a. "Induction depends partly on absolute rotation" (rotation relative to something not the apparatus). a. Mathematics: consistent. b. Experiment: consistent and uncontroversial. Rotation is absolute in mechanics and optics too (Newton's bucket; gyroscopes and the Foucault pendulum; the Sagnac effect; Michelson-Gale-Pearson 1925 measured the Earth's rotation optically) [memory, high confidence]. Kennard's gyrocompass analogy [read: p.189] is in line with this. Special relativity's relativity principle covers uniform translation only; Barnett makes the same point: the experiments are "not inconsistent with the principle of relativity of Einstein, which relates to translatory motion only" [read: Barnett 1919, p.1511]. Nothing here is a problem for relativity. c. Test: the same experiments. d. Nothing changes. Verdict: Established.

6b. "Relative to the earth is improbable" (page 189). Kennard writes that "Rotation relative to the earth might in some unknown way be responsible ... but this seems improbable." My view: this is the weak step. The lab frame (earth-fixed, nearly inertial: Earth's rotation contributes terms of order (v/c)^2 ~ 10^-12 [read: Chyba 2025 on frame accuracy]) is where the solenoid's field pattern is defined and where the electrometer and wires sit. The electrons in the spinning condenser move relative to that frame at speed omega r, and the force is q(v x B) with v taken in that frame. There is nothing "unknown" involved; this is the plain mainstream reading, not an alternative to it.

6c. "Can only be explained by motion relative to a stationary Lorentz aether." a. Mathematics: the elimination argument (not radial acceleration since that goes as r^2, not rotation of the electron on its axis since that is the same for all, not motion relative to other parts since there is none) is logically tidy, but it leaves out a candidate Kennard did not consider: motion relative to the electromagnetic field itself, as specified in an inertial frame. That is the content of the Lorentz-force law as used in Maxwell-Lorentz theory. The field is a dynamical entity with its own state; an axisymmetric steady field has the same state whether or not its sources rotate, so the reference frame is fixed by the field, not by the sources. That makes "the stationary aether" an interpretive label for an operationally identical statement. b. Experiment: Lorentz's aether theory and special relativity are empirically equivalent for inertial frames [memory, high confidence]; no experiment on unipolar induction distinguishes them. Kennard's own 1912 paper says the method "cannot separate Lorentz from Hertz" [read: Investigator lead 1; 1912 text]. The 1917 claim that the experiment "seems to lend definite support to the existence of an electromagnetic ether" and is "a stumbling-block ... for those ultra-relativitists who would abandon the conception of an ether altogether" [read: pp.189-190] was therefore an overreach even for 1916, though the paper was clear about the argument it was making. Barnett in 1919 explicitly leaves it open either way [read: p.1511]. The modern authors the Investigator found (Chyba et al.) read v as relative to the frame in which the charge and field are specified, with no aether [web: https://journals.aps.org, Investigator's note 2026-10-04-aps-journals.md; their characterisation; they have a stake in the Earth-rotation power proposal]. c. Test: no test of "aether versus field in an inertial frame" in this setup is conceivable, because both predict the same charges. d. If it were true as stated (a material aether with a unique rest frame detectable by such experiments) one would need a preferred frame detectable by experiments. Michelson-Morley-type searches, which gave null results, already argue against a detectable preferred translational frame [memory, high]. Unipolar induction is a rotation effect and would not by itself detect a translating preferred frame. So Kennard's claim, taken narrowly (rotation relative to an inertial frame matters), is already mainstream; taken broadly (a stationary material aether), it adds an unobservable.

Verdict: Split: 6a Established; 6b Plausible, untested for any alternative but the mainstream reading is simple; 6c Not testable as stated (aether-versus-inertial-frame interpretation with no measurable difference here). Historical flavour: Historically superseded but instructive (a reasonable position in 1916 when many accepted an aether, since replaced by field-in-inertial-frame language).

2. Mathematics and presentation points that bear on the physics (summary)

• Two printed arithmetic slips (0.117 for 0.114; 0.140 for 1.40) confirmed [calc].
• OCR traps to remember: "20.5" not "26.5" (the ratio hinges on it), "+1.2" not "-12" for the solenoid-alone average.
• The electrostatic solution is printed only for the outer space; the inner space is "similar". A reader cannot reproduce -6.22 V0 without re-solving a boundary-value problem. This is a reproducibility limit, not an error.
• The 13 A (0.25) against 25 A (0.27) ratios differ by 8% unexplained. An expected-constant ratio varying by 8% hints at systematic effects (speed monitor lag, stray fields, brush noise) of the same size as the 5% "agreement". Kennard reports frequent "sudden large disturbances" and four rejected sets plus about as many abandoned [read: p.8], which cannot be audited.

3. Investigator's findings, weighed

• Kennard 1912 (Phil. Mag. 23, 937) [read: research/files, IA OCR]: steel bar rotated in a stationary solenoid, outer cylinder screened; 0.0 +/- 0.24 divisions against predicted about 37 [per Investigator]. Not decisive on moving-line theory because the magnetising coil stayed fixed, as Barnett objected. Kennard's own words on the conclusion: "a rotating magnet does not 'carry its force lines' around with it"; and induction "depends on absolute motion--that is, on motion relative to the ether". Kept: shows his aether reading was present from 1912 and predates the 1917 experiment, so the 1917 results were not what led to it.
• 1917 pp.188-190 clean text: used above.
• Barnett 1919 AIEE report: independent acknowledgement of claims 1-3 by the critic; on claim 6 a two-way reading. Kept as the strongest neutral-ish support (Barnett was Kennard's opponent on the interpretation).
• Pegram 1917: independent corotation experiment; I have not read it. Its strength is that it confirms claim 1 independently; unread.
• Chyba, Hand & Chyba 2025 and the 2026 Comment/Reply (Wijngaarden): they state the Barnett-Kennard-Pegram result as standard. They use it to argue that Earth's axisymmetric field does not co-rotate and so a conductor co-rotating with the Earth could generate voltage; Veltkamp and Wijngaarden reported a null result [per Investigator, not independently checked]. This downstream use of the Kennard principle is contested, so it is not evidence for or against the 1917 experiment; kept as the live modern application.
• Su 2005 (arXiv physics/0512179): argues that a short solenoid's nonuniform field could induce an EMF on a stationary capacitor outside or inside it, bearing on Kennard's 1912 and Barnett's outside-capacitor arrangement. Su is the author of a "local-ether" unified theory (Quantum Electromagnetics) [read: Su 2005 reference list], a stake to note. In the 1917 set-up Kennard measured the actual flux distribution with a ballistic galvanometer (section 4) and used its full non-uniform profile [read: p.5], so Su's concern for a pure uniform-field idealisation is largely handled in 1917, though Su's own claim is untested by me.
• Baumgartel and Maher 2022 (Sci. Rep.): place Kennard, Bartlett and Barnett among experiments some authors say need Ampere's force, while others say classical explanations suffice, "albeit a full re-analysis ... with due regard to the influence of the closing wire might be necessary" [read]. Unresolved among homopolar-motor theorists; it does not touch the voltage-level charge measurement but is a legitimate open flag for torque-type interpretations.
• Weber/Ampere-style authors (Wesley, Assis, Guala-Valverde): closely related work, said by mainstream sources to be fringe; that label is theirs. They agree the open-circuit results are not trivially relativistic; I agree with the narrower point that the interpretation of v is not enforced by the experiment (see 6c) but see no experimental support for Weber replacing Maxwell-Lorentz here [my assessment].
• Millis 1996 NASA Breakthrough Propulsion paper: cites Kennard 1917 in a Mach's-principle context. A citation, not a test.
• Historical record: the Kennard unipolar line ends in 1917. After this he moved to the 1922 NRC moving-media bulletin, quantum mechanics (his name is on the 1927 Kennard uncertainty bound [memory, high]), kinetic theory and Navy hydrodynamics [per AIP finding aid, Investigator]. Competing explanations the Investigator set out: (a) issue considered settled (Barnett accepted the experimental result by 1919); (b) apparatus failure (short-circuit, no rewind) and his move from Minnesota; (c) the field turned to relativity and quantum theory; (d) a change of view. Silence here is an anomaly rather than evidence for any one; the decisive records are his 1922 NRC chapter and the 1926 Cornell lecture notes (AIP Box 4 F4), which I have not seen.

4. Comments on the dispute as a whole

Kennard and Barnett agreed on the facts and fought over interpretation, in print, with some heat (Kennard: Barnett "maintained this view with some heat"; Barnett: "Mr. Kennard's fallacies are quite apparent"). Kennard was a young instructor (Rhodes Scholar, Cornell Ph.D. 1913, Minnesota instructor) and Barnett was an established Ohio State physicist [memory, medium: Barnett's affiliation; AIP finding aid for Kennard]. The factual record supports Kennard on the experimental outcomes; the interpretive record is a draw, resolved in the following decades by the field-in-inertial-frame reading. There is no sign in the record I read of experiments blocked or data suppressed; the dispute was open in the journals.

5. Summary verdict table

Corrections to Step 1

1. claims.md Source-gap/caveat on pp.10-12. "pp.10-12 rest on the digest alone" is now closed: I read the Internet Archive text of pp.188-190 (the clean OCR in research/files). The Laplace argument and the aether conclusion are as Step 1 summarised them. Section 7's actual claim is about assumptions Barnett used (div(KE) = 4 pi rho together with the lines' EMF); Step 1's summary "[v x B] does not satisfy Laplace's equation in free aether" is accurate but compressed. I verified it independently: div(v x B) = 2 omega B [calc].
2. Claim 2 percentage. The deep read repeats Kennard's "1.4%" for the half-speed ratio. Recomputed: 0.228 against 0.2308 is -1.2% to -1.3%. Trivial, noted for accuracy. The "5%" is 4.8% [calc].
3. Claim 2 precision. claims.md states the computed charge as "-6.22 V0 (+/-0.5%)". That is Kennard's stated accuracy of the calculation, not of the result. The measurements scatter by about +/-15% on single sets (0.195-0.271) and 8% between current groups. A reader could mistake the 0.5% for an accuracy of the comparison; it is not.
4. Claim 3 baseline. The deep read notes that the solenoid-alone rows have G = 0. It does not draw the implication I add: the speed monitor is the condenser-shaft disk, so the speed in these rows is unrecorded, and the comparison with the "about 15 mm" full effect assumes similar speed. Not an error in Step 1; a missing caveat.
5. Investigator lead on Chyba et al. sites.md says Chyba et al. read v relative to the observer's frame, not an aether. I confirmed the frame-based wording in the paper (e.g. discussion of reference frames K and K' and "the velocity of the charge q in the frame" [read: Chyba 2025]); I did not find the word "aether" in it, so "not an aether" is the Investigator's inference from silence, to be labelled as such. Also the Investigator says Chyba credits corotation to Pegram only; the text says Barnett's null was "reproduced by Kennard" and that Pegram showed corotation of the capacitor and connecting wire with the solenoid charged it. Barnett 1919 credits Kennard as well as Pegram with the corotation result, so there is a mild discrepancy in the secondary literature about priority between Kennard and Pegram that I could not settle without Pegram's paper.
6. Step 1 facts.json statement of "the 14 mm check". Both Step 1 and Kennard describe it as "partial confirmation". I note that the 14 mm has no galvanometer normalisation and a loosely specified speed; it is a rough consistency check only.

Source notes

• All verdict-bearing facts are [read] from the 1917 paper (pages checked against page images in the deep read, and the IA text for pp.188-190), Barnett 1919 (research/files/AIEE-1919-report-electromagnetic-induction.pdf), Chyba et al. 2025 and Su 2005, or [calc] from my recomputations.
• [memory] items not verified on the web (no verdict turns on them): Griffiths-level Faraday-disk treatment; Lorentz ether theory and SR empirical equivalence; Michelson-Gale-Pearson 1925; Kennard uncertainty relation 1927; Barnett's affiliation. All high confidence except the Barnett affiliation (medium); none was needed for a verdict. I made no web calls this pass, relying on the documents the Investigator saved.
• I did not read Pegram 1917, Fehrle 1913, Kennard 1913 (Phys. Rev. 1, 355), the Phys. Zeits. replies, the 1922 NRC bulletin or Kholmetskii 2003; statements about them are second-hand and labelled so.

Files read: review/Kennard/{claims.md, facts.json, sites.md, deep/On-Unipolar-Induction-60158c.md, _text/On-Unipolar-Induction-60158c.txt, research/files/*}.