Rankings Primer Arbiter Physicist Mathematician Contrarian QE Sites log Claims Sources Research notes

Kennard: Science Contrarian review (Step 2)

Document reviewed: E. H. Kennard, "On Unipolar Induction: Another Experiment and its Significance as Evidence for the Existence of the Aether", Phil. Mag. (6) 33, 179-190 (1917; signed Univ. of Minnesota, August 1916). Reviewer role: judge by experimental evidence, build the strongest honest case, examine how the claims were received, and hold the inventor to the same standard as the critics. Source tags: [read] document or Investigator file; [web] URL; [memory] with confidence. Web checks: no verdict below depends on a medium- or low-confidence [memory] fact, so no extra web calls were spent. The Investigator's [web] items were re-read in the saved PDFs where noted.

Plain-language background (for the reader)

Unipolar (homopolar) induction is the voltage generated by a conductor that moves relative to a magnet whose field does not change at any point in space, for example a copper disk spun between the poles of a symmetric magnet (the Faraday disk). The open question around 1912 was this: if you spin the magnet itself, do its "lines of force" spin with it? Two views:

Lorentz view: the force on a charge q moving with velocity v in a magnetic field B is F = q v x B / c (Gaussian units, c = speed of light; "x" is the vector cross product). Here v is measured in the laboratory (or aether) frame, and the magnetic field of an axially symmetric magnet is simply there, whether or not the magnet turns. Spinning the magnet alone does nothing.
Moving-line view: the lines of force are attached to the magnet, so v is measured relative to the magnet. Spinning the magnet alone would act like spinning the conductor in the other direction, and spinning both together would give no effect.

Kennard's apparatus is a cylindrical condenser (two coaxial metal cylinders separated by insulation, so it can hold charge) inside a long coil (solenoid) that makes B along the axis. Either can be spun, alone or together. An electrometer (a sensitive voltmeter that draws no current) reads the charge on the inner cylinder. Worked scale: B = 4 pi n I / 10 = 4 pi (1150/22.5 cm)(25 A)/10 = 1606 gauss; flux through the outer cylinder = B pi r^2 = 1606 x pi x 3.32^2 = 5.56e4 maxwells (Kennard prints 55,400); at 33 rev/s the e.m.f. is 55,400 x 33 x 1e-8 = 0.0183 V. I reproduced these [read: deep read section 5; recomputed].

Corrections to Step 1

The deep read's "pages 10-12 rest on the digest alone" (claims.md header) is now closed: I read the extracted text of pp.9-12 and the Investigator's cleaner Internet Archive text [read: _text p.9-12; research/files/Kennard-1917-PhilMag-33-179-IA-OCR.txt]. Claims 5 and 6 are checked against the text, not the digest.
claims.md claim 5 says Kennard's argument is that [v x B] "does not satisfy Laplace's equation". That is Kennard's wording [read: p.10], but the precise point is that the field E = [v x B]/c is not divergence-free (not solenoidal) in free aether. For rigid rotation v = omega x r, div(v x B) = B.(curl v) - v.(curl B) = 2 omega.B inside the solenoid (curl B = 0 there). That is nonzero, so his objection is mathematically correct as stated [recomputed; standard vector identity, [memory] high]. The claims.md paraphrase is acceptable but should say "not solenoidal".
claims.md claim 3 says the reversing average of +1.2 mm is "against about 15 mm". The four raw solenoid-alone electrometer readings are +0.4, +0.6, -1.1 and +3.5 mm so the scatter (about 2 mm) is larger than the mean; statistically the mean is consistent with zero but also with a residual of a few percent of the full effect. The earlier group gave +0.2 mm [read: p.8]. No error is stated by Kennard. This is a nuance, not an error in Step 1.
claims.md claim 4 says "partial 14 mm check". The 14 mm is an electrometer deflection only, with no galvanometer reading and thus no speed/field normalisation; it was "under about the same conditions" [read: p.9]. Its comparison with the 14.9 mm absolute prediction is only rough (speed uncertain).
Step 1 correctly caught the printed misprints (0.117 vs 0.114; 0.140 vs 1.40) and the OCR error "26.5" for 20.5. I verified: (20.5/62.2)(0.1/0.1197)(6.22/7.42) = 0.2308; 0.114 x 18000/138 = 14.87 mm (0.117 gives 15.26) [recomputed]. 0.242/0.231 = +4.8%; 0.228/0.231 = -1.3% (Kennard prints 1.4%; trivial rounding).
Step 1 does not note a problem that I find: Kennard's four-row "Both" table has ratios .237, .237, .236, .258, but the mean of 16.1 mm electrometer deflection is 8% above the 14.9 mm absolute prediction [recomputed]. Because speed was not recorded, the ratio method was right to cancel it, but it means the agreement quoted (5%) is for the ratio only.
Facts.json states the journal is inferred from the page layout; the Investigator confirmed Phil. Mag. 33, 179-190 [web: https://archive.org/details/londonedinburg6331917lond; Crossref]. Fine.

What the Investigator found that bears on the claims (kept whatever my verdict)

Barnett, "A Report on Electromagnetic Induction", Trans. AIEE 1919 (the critic): "This corollary of my experiments has been confirmed in each of two recent investigations; one by Kennard ... and one by Pegram ... The condenser was found to become charged in accordance with the theory." [read: research/files/AIEE-1919-report-electromagnetic-induction.pdf]. Barnett is an interested party, a rival in the 1912-13 dispute; this is an acknowledgement against his own earlier position, which makes it more notable. Barnett also says the experiments "prove either (1) that the ether does not exist and that the tubes of induction do not move, or (2) that the ether exists" [as quoted by the Investigator; I saw the Kennard passage, p.737-745 of the extracted text, not the full quote location].
G. B. Pegram, Phys. Rev. 10, 591 (1917): an independent corotation experiment, charging seen (as reported by Barnett 1919 and Chyba 2025). PDF not obtained, so I cannot check his precision. No quantitative replication of Kennard's 0.231 ratio is known to me or the Investigator.
Chyba, Hand and Chyba 2025, Phys. Rev. Research 7, 013285: "These results proved that the field of a rotating axially symmetric electromagnet does not 'rotate with the magnet,' a conclusion that is now a standard part of the electromagnetics literature." They say v in v x B is the velocity of the charge "relative to a frame in which the magnetic flux density is B ... not a velocity relative to the translating or rotating magnetic field." [read: research/files/Chyba-Hand-Chyba-2025-PRResearch-7-013285.pdf, lines near 216-248]. Their characterization supports claims 1-3 and 5, and replaces Kennard's aether with an inertial frame in claim 6.
C. C. Su 2005 (arXiv physics/0512179): argues that a short solenoid's non-uniform field could induce an e.m.f. on a stationary capacitor, and that Kennard's 1912 outside-capacitor arrangement "actually observed" a non-vanishing electromotance at magnetisation reversal "attributed to the induction" [read: Su, lines 36-40 and 568-577]. Su concludes this experiment "cannot discriminate" between induction force and magnetostatic force. This is a modern re-analysis (Su is a single author; no independent confirmation found). It bears on claim 3 (finite-length edge effects) and on the 1912 background; it does not address the 1917 co-rotating result.
Baumgartel and Maher 2022 (Sci. Rep.) note that some authors think Ampere-force electrodynamics is needed to explain "Kennard's, Bartlett's and Barnett's experiments", while others say they can be explained classically, "albeit a full re-analysis of the experiments with due regard to the influence of the closing wire might be necessary" [read: lines 862-867]. This is an acknowledged open point in the modern literature, not a refutation.
Weber/Ampere-school treatments (Wesley, Assis) that cite Kennard 1917 and accept the open-circuit results as not trivially relativistic; mainstream sources often call these fringe. Kept as related work (the label is the mainstream's characterization). Also Millis 1996 (NASA) cites Kennard for absolute rotation / Mach's principle. I did not open these papers and make no claim about their content beyond the Investigator's note.
Historical record: Kennard's unipolar line ends in 1917; he went on to the 1922 NRC "Electrodynamics of Moving Media" report, quantum mechanics (1927), Cornell, and the Navy (David Taylor Model Basin) [Investigator, AIP finding aid]. The Investigator lists four explanations for the silence; see claim 6 below.
Not obtained: Kennard 1913 Phys. Rev. 1, 355 and Phys. Zeits. 13, 1155; 14, 256 (his replies to Barnett), Barnett 1922 Phil. Mag., the 1922 NRC bulletin, AIP papers. These are gaps; I did not fill them from memory.

The four questions, per claim

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

a. Mathematics: Consistent. The condenser, wire and electrometer obey Q = Q1 + CV; the potential difference is Nn (flux N times rev/s n) in e.m.u., a standard result for a point on a rotating conductor. The Bessel-function solution for the potential in the shell spaces gives a charge -6.22 V0; the ±0.5% claim is Kennard's, supported by a stated residual-error analysis. I cannot re-solve the boundary-value problem here; I checked every downstream number (see above) [read: pp.5-8; recomputed]. Two printed misprints (0.117 for 0.114; 0.140 for 1.40) do not change the result. b. Consistency with experiment: Yes. The result is what the Lorentz force plus a rotating conductor give; the same physics runs every homopolar generator (Faraday disk). The screen does not shield because the force is on the electrons in the metal itself. c. Tests: Kennard 1917 (four-row table, ratio 0.242; further groups). Independently Pegram, Phys. Rev. 10, 591 (1917), reported by Barnett 1919 and Chyba 2025 as seeing charging. Barnett 1919 states the confirmation in print. Pegram's PDF not read, so replication is documented by secondary reports only. No modern replication of this exact apparatus is known to me. The 1926+ literature treats it as accepted. d. If true: nothing changes. It is standard electromagnetism. Evidence quality (inventor held to the same standard): four sets rejected "for very obvious good reason" and "perhaps as many more" abandoned because of "an obvious large disturbance" [read: p.8]; the 10,000 ohm resistance "accidentally left in place" [read: p.3]; "sudden large disturbances were rather frequent" [read: p.3]. None is quantified. A critic could say that discarding disturbed runs after seeing them is a researcher degree of freedom; the counterweight is that the effect is large (16 mm against about 1 mm noise in the nil case) and that the sign reversals with both current and rotation direction in every row make a spurious cause unlikely. Verdict: Established.

Claim 2. The measured electrometer/galvanometer ratio (0.242) agrees with the calculated 0.231 to 5%, and 0.228 at half speed (1.4% low).

a. Mathematics: The ratio formula is (E/G)(0.1V/0.1197V0)(6.22V0/7.42V), recomputed to 0.2308. Consistent. The computed charge uses a rather elaborate correction chain (wire, amber rings, end fields) that Kennard says "just about cancels" (+0.03 V0 against -0.02 V0). That cancelation is asserted, not independently shown. b. Consistency: Yes. The agreement is within the scatter. c. Tests: One apparatus, one observer. The individual ratios range .211 to .263 (half speed), and .195 to .271 (alternative method), so the spread is roughly ±10%; a 5% mean offset is under one standard error of the mean for four sets (about 4-5%, my estimate from the quoted range). Kennard himself says the results "remain somewhat rough". Unexplained: ratio .25 at 13 A and .27 at 25 A (8% apart) with "not comparable with the previous values" and no explanation [read: p.9]; if the ratio is current-independent as theory says, this is a systematic of at least that size or noise. The quantity is therefore a few-percent-level agreement, not a precision test. No quantitative replication known. d. If true: nothing. Verdict: Established qualitatively (charging of the predicted sign and size); the quantitative claim of 5% agreement is Plausible but of limited precision. Split verdict: effect Established; "5% agreement" correctly reported but within the noise, not a precision test.

Claim 3. Rotating the solenoid alone produces no charge on the condenser, confirming Barnett.

a. Mathematics: Consistent; the Lorentz view predicts zero because the condenser is at rest in the lab frame. b. Consistency: Yes, and it matches Barnett 1912 and Fehrle 1913 (whom Kennard says has results "which must certainly be wrong"; this is Kennard's characterization of a rival and I have not checked Fehrle). c. Tests: Kennard 1917, Barnett 1912, Kennard 1912 (with a rotating iron bar), Fehrle. Barnett 1912/13 and Kennard 1912 are in the Investigator's files [Barnett 1912 PDF; Kennard 1912 OCR]. Chyba 2025 states the conclusion is standard. Honest limits: Kennard's mean +1.2 mm (8% of the full effect) with scatter -1.1 to +3.5 mm is a consistency-with-zero result, not a bound better than a few percent; the previous group gave +0.2 mm. Su 2005 suggests a short or nonuniform solenoid field could leave a small residual; Kennard's solenoid is 22.5 cm long with a 6.64 cm diameter condenser inside it, so end effects are present but the central region is the main contributor (my estimate, not computed). The 1912 experiment had a weakness Kennard himself conceded: a second small screen "might conceivably" cut off the effect [read: p.2]. The 1917 design addresses Barnett's objection about what rotates, but the screen concern was not eliminated for the solenoid-alone case; claim 1 covers it for the co-rotating case. d. If true: nothing. Verdict: Established (to a few per cent).

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

a. Mathematics: Consistent. Condenser alone = (both) - (solenoid alone), valid if the effects add (linear in the motion). b. Consistency: Yes, with the same caveat that the linearity assumption is a statement of the Lorentz theory, which is what is being tested. So the inference tests nothing independent: if the moving-line theory were true it would also be linear, but with a different split. In fact the three cases (solenoid alone, both, condenser alone) are exactly what separates the two theories, and only two were measured with controls. The 14 mm "partial" run lacks the normalising galvanometer. c. Tests: Not completed by Kennard. Barnett 1919 states a corollary that "condenser and disks alone produces a charge" from his own work [read: AIEE 1919 p.735-736]; I take that as a statement in the critic's text, not a measurement. No dedicated condenser-alone experiment under controlled conditions was found by the Investigator. d. If true: nothing. Verdict: Plausible, untested (directly); partial single-group check of 14 mm. The explicit statement "it was not thought worth while to rewind" [read: p.9] is a candid admission, to Kennard's credit.

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

a. Mathematics: The central argument [read: pp.10-11]: Barnett supposes the rotating lines set up a "[v x B]" electromotive intensity in the aether and then writes both div(K f) = 4 pi rho and f = e + [v x B]/c; but in free aether div E = 0 (no charge), so div f = -div[v x B]/c, which is not zero in general. I checked: for rigid rotation about the axis, div(v x B) = 2 omega.B, nonzero [recomputed]. Therefore the combination cannot hold in vacuum, and Kennard says the lines must act on matter only. The logic is sound as a reductio of Barnett's specific supplementary assumption (aether acting like a dielectric). It does not, by itself, rule out all relative-rotation theories; "disproved" is stronger than the algebra delivers. The mathematician should check the claim that Barnett's two assumptions are as Kennard states. b. Consistency: The experiments (nil effect for solenoid alone; charge for co-rotation) contradict the strict relative-rotation hypothesis, which would require the opposite results in both cases. Consistent with Kennard. c. Tests: As claims 1-3; the 1912 experiments (Barnett and Kennard) and Fehrle, Pegram. Rejection history: Barnett (1912, 1913) called Kennard's 1912 work "inconclusive" and "based on incorrect theory" and Kennard's "fallacies ... quite apparent" [Investigator, Barnett 1912/13 PDFs, not read in full by me]; Kennard says Barnett maintained his view "with some heat" [read: p.9]. This was a direct rival dispute between two experimentalists, resolved by Barnett's own 1919 acknowledgement that the condenser charges with co-rotation, and the later literature treats the lines-do-not-rotate result as standard (Chyba 2025). I find no evidence of suppression: the paper appeared in the Phil. Mag. and was abstracted in German journals in 1917 [Investigator, Google Books snippets]. The disagreement was settled by experiment, not dismissed. d. If true (the strict moving-line theory): mainstream electromagnetism would need frame-dependent field assignment. It is not what happened. Verdict: Split. (i) The strict moving-line theory (lines rotate with the magnet): Tested, not supported (experiments, at least by Barnett, Kennard and Pegram); Kennard's divergence argument adds a theoretical objection. (ii) That the experiments leave no room at all, as his "disproved" claims, is slightly stronger than the evidence: modern authors say the question is ill-posed since a uniform field cannot be said to move (Chyba 2025).

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

a. Mathematics: Kennard's argument of elimination [read: pp.11-12] says the force on each electron grows in proportion to its distance from the axis, and rules out centripetal acceleration, electron spin and motion relative to other parts of the apparatus. (His statement that the radial acceleration is "proportional to the square of the distance" is loosely worded: centripetal acceleration omega^2 r is also proportional to r; I take him to mean it scales as speed squared, so it cannot be linear in v. This is my reading.) The force q(v x B)/c with v = omega r and a fixed B is exactly linear in r. The solenoid co-rotates, so relative motion to the apparatus is correctly excluded, leaving the laboratory frame in which the solenoid's field is static. A modern reader names that frame the inertial frame, not the aether. His step "the only remaining alternative is ... something which is not matter" does not consider the electromagnetic field or inertial frames as the reference, so a stationary aether is one choice, not the unique one. b. Consistency: The observation (charging on co-rotation) is established. The inference "aether" is a matter of interpretation. Special relativity with rotation relative to an inertial frame explains it without an aether; Lorentz's stationary-aether theory is empirically indistinguishable from SR for such effects [memory, high]. So the experiment does not discriminate between the aether and relativity-with-inertial-frames; it discriminates only against "complete relativity of rotation" (a Machian reading). Millis 1996 and the Weber school (Investigator) treat it as bearing on absolute rotation / Mach's principle; I did not check their arguments. c. Tests: No experiment tested the aether interpretation as such. Barnett 1919 offered two options: either the aether does not exist and the tubes of induction do not move, or the aether exists [read: via Investigator, AIEE 1919]. Kennard himself allowed that "Rotation relative to the earth might in some unknown way be responsible ... but this seems improbable" [read: p.11]; that alternative (Mach-type) was not tested. Not tested directly by Michelson-Morley-type experiments either. The Baumgartel and Maher 2022 remark that a full re-analysis with the closing wire may be needed [read] keeps a small question open for Ampere-force models, not for the aether. Absence of evidence (Kennard's line ends in 1917): the Investigator's explanations (a) settled question, (b) apparatus damaged and he left Minnesota, (c) the field moved on to relativity and quantum theory, (d) he changed his view. The evidence: (a) Barnett's 1919 acknowledgement is for the experiment, not for the aether; (b) documented in the paper (solenoid short-circuited); (c) he co-wrote the 1922 NRC "Electrodynamics of Moving Media" report and did quantum work in 1927 [Investigator, AIP]; (d) unknown; deciding records are the 1922 NRC chapter, the 1926 Cornell lecture notes (AIP Box 4 F4) and the memorial statement. I can't say between them. It is not a case of suppression on the evidence found; it is also not shown that he held the view later. d. If true (a stationary Lorentz aether that is the preferred frame for rotation): special relativity's equivalence of inertial frames would need an extra preferred-frame element, and the experiment would need to be explained by something other than the lab frame. No other experiment supports a preferred frame (Michelson-Morley, Kennedy-Thorndike) [memory, high]. Verdict: Split. "Rotation relative to an inertial frame matters, not relative rotation of the magnet": Established. "Only a stationary Lorentz aether can explain it": Historically superseded but instructive (a reasonable reading in 1916; an inertial frame does the same work without the extra entity; Kennard's own caution about the earth-relative alternative is preserved).

Strongest honest case for Kennard

He answered Barnett's specific 1912 objection by building an apparatus where solenoid and condenser could each be rotated, and he got both positive and negative results with the same apparatus, which is the right design.
The result that the critic Barnett accepted in print in 1919, and that Pegram independently reproduced in 1917, is now standard.
He reported his own difficulties (disturbances, rejected sets, the unfinished run) and said "somewhat rough"; he did not hide the shortfalls.
His divergence argument against the moving-line theory is correct mathematics.
Related work (Weber/Ampere school, Wesley, Assis, Millis) uses his result in support of non-Maxwellian views; those authors are kept in the record as related work.

Strongest honest case against

The central quantitative claim (5% agreement) rests on four sets scattered ±10% with an unexplained 8% discrepancy in the proportionality test, and on discarded sets whose selection rules are not given.
The decisive condenser-alone run was never done; the inference is theory-dependent.
The aether conclusion is an argument from elimination that omits inertial frames; the experiment cannot distinguish Lorentz's aether from relativity. The heading of the paper oversells.
The absence of any error bar, any test of the amber rings, or an independent recheck of the Bessel solution leaves the exact theoretical number (0.231) only as strong as the stated ±0.5%.

Summary of verdicts

Condenser charges as Lorentz predicts: Established.
Ratio 0.242 vs 0.231: effect Established; 5% agreement Plausible, limited precision (scatter about ±10%, unexplained 8% current discrepancy).
Solenoid alone gives no charge: Established (to a few per cent).
Condenser-alone by superposition: Plausible, untested directly (partial 14 mm check).
Moving-line theory disproved: strict form Tested, not supported (and theoretically objectionable); "no room left" overstated.
Absolute rotation / stationary aether: absolute-rotation part Established (inertial frame); aether-only part Historically superseded but instructive.