SilverTooth: Arbiter (Step 3)

Project: E. W. Silvertooth, "Experimental detection of the ether" (received May 1986). Inputs: facts.json, claims.md, physicist.md, mathematician.md, contrarian.md, deep/, sites.md, research/. Tags: [read], [web] with URL, [memory] with confidence, [calc] my own computation (SymPy/NumPy, .venv).

Source gap note

The project rests on ONE five-page paper (modern retyping, text fully readable, no raw data, error bars, run counts, dates or place). Everything about the experiment's outcome beyond that paper rests on Silvertooth's later statements (1989 article), other people's reports (Marett 2012, Marinov and Wesley abstracts via the Investigator), and the saved 1976 Applied Optics paper [read by reviewers]. I did not open the Marett page itself beyond a search summary; the details of his findings are the Investigator's and reviewers' second-hand reading. Sponsor reports (RADC, DARPA) are not in the record. Where a conclusion rests on [memory] alone it is marked.

Recomputed where reviewers differed

1. Arithmetic of the headline number: v/c = lambda/(2 Delta) = 0.63e-6/(2 x 0.25e-3) = 1.26e-3; v = 377.7 km/s (c = 299,792.458 km/s) or 378.0 (c = 3e5) [calc]. All three reviewers agree. Note 378 is rounded input times 3e5, not an independent three-figure measurement; an error of 0.01 mm in Delta moves v by about 15 km/s [calc, contrarian].
2. Claim 3, the disputed physics. Physicist and Mathematician find the signal cancels; the Contrarian did not do that calculation and rated it "Tested, not supported". I redid it [calc, SymPy]. With eq. (5) wavenumbers k1 = 2 pi (1+b)/lambda, k2 = 2 pi (1-b)/lambda (b = v/c) and the slide making one beam path longer by Delta and the other shorter by Delta, the relative phase is phi = k1(sA+Delta+x) - k2(sB-Delta+X-x). Then d(phi)/d(Delta) = d(phi)/dx = k1+k2 = 4 pi/lambda exactly, with b cancelling. The standing-wave pattern shifts by exactly Delta per slide travel and the round-trip reference D2 tracks it identically, so D1-D2 stays in phase for every b. This confirms the two reviewers under that geometry. Two refinements they did not make. (i) If instead both paths lengthen together, the phase difference is (k1-k2)Delta = 4 pi b Delta/lambda [calc]; a 180 degree reversal then occurs at Delta = lambda/(4b), not lambda/(2b), a factor of two from the paper's eq. (9), and the 0.25 mm run would give a phase of one full cycle, not a half cycle. So even with the geometry that gives a first-order signal, eq. (9) as printed does not follow. (ii) The paper gives only words and a figure for the derivation of delta = lambda b, so which geometry is intended is unsettled. Verdict impact: the claim as derived in the paper is not supported by the paper's own equations on the geometry described; the geometry caveat is kept.
3. Claim 5 magnitude: L a/c^2 with L = 1e25 m, a = 3e-10 m/s^2 gives 0.033 [calc, matches Physicist and Contrarian]. Mathematician's point: orbit period 9.4e15 s vs light travel time 3.3e16 s, so the linear-in-L acceleration term is not valid for the author's own rotating quasar example (acceleration direction turns about 3.5 times in flight). I checked 2 pi d/v = 2 pi x 3e22 cm / 2e7 cm/s = 9.4e15 s [calc]; correct.
4. Cosine misalignment needed to fake the effect: 1 - cos(theta) = 1.26e-3 gives theta = 2.9 degrees [calc]. Contrarian's "2 degrees for lambda/4" and Physicist's 2.9 degrees refer to different criteria (quarter wave vs half-cycle) and are both right for their stated criterion. A fixed tilt is zeroed out; only drift over a run matters (Mathematician and Contrarian agree).
5. Doppler first term of eq. (10): at b = 0.5 the standard factor is 0.577, the 1976 form 0.433, the 1986 printed form 0.775 [calc, Mathematician]; the three agree only to first order. Not a retyping slip alone, since the 1976 form is also non-standard at second order.

Web checks done (budget 8; used 2 new plus those the reviewers cited)

• Krisher et al. 1990, Phys. Rev. D 42, 731: two hydrogen masers 21 km apart on a fibre link, upper limit on one-way linear anisotropy 100 m/s, delta c/c < 2e-7 (linear) relative to Earth velocity w.r.t. the CMB [web: https://pubmed.ncbi.nlm.nih.gov/10012893]. Upgrades the Physicist's check. Silvertooth's 378 km/s would be about 3800 times that limit. Caveat as the Physicist states it: such a link measures a quantity that may be synchronisation-convention dependent, so its bearing on a one-way wavelength claim is itself debated by conventionality writers.
• Marett replication exists, with the "very elaborate thermometer" conclusion and the room-temperature/laser-alignment attribution [web: http://www.conspiracyoflight.com/Silvertooth/Silvertooth.html, search summary only]. Level of detail (12 h pattern, wrong sign, local-time pattern) remains the Investigator's reading.
• Already confirmed by reviewers and accepted: CMB dipole 369.82 +/- 0.11 km/s toward Leo [web: https://www.aanda.org/articles/aa/full_html/2020/12/aa38053-20/aa38053-20.html]; Muller, Sci. Am. 238(5) 64 (1978) exists [web: https://ui.adsabs.harvard.edu/abs/1978SciAm.238e..64M/abstract]; Gunn and Tinsley, Nature 257, 454 (1975) [web: https://www.nature.com/articles/257454a0].
• Not verified, stay [memory]: Muller's figure of about 390 +/- 60 km/s (medium; reported by a search summary only), the CMB dipole declination near -7 degrees (medium), Riis 1988 first-order tests (medium), Gagnon et al. 1988 result (title only), air refractivity 2.7e-4 (high), GPS Sagnac correction up to about 133 ns (high). No verdict depends on the medium items alone.

Corrected claim list (post-review)

1. Round-trip experiments cannot fix a one-way speed or wavelength of the form c/(1 +/- b); the null needs no isotropic c. (Weak form, true as a conventionality statement. Strong form, a physically real lab anisotropy needing no contraction, is not shown.)
2. A Sagnac-style segment shift (L v/c^2 per segment) is assigned a physical, translation-surviving meaning, giving lambda/(1 +/- b). The author himself wrote in 1976 that the closed-loop shift is zero in translation [read: 1976 p.1101]. Premise A1 (frequency cannot change) also reverses his 1976 statement that frequency and velocity change, wavelength being "a resultant effect" (QE revision).
3. A slide-mounted two-beam standing-wave interferometer reads v/c = lambda/(2 Delta). Under the paper's own premises the first-order term cancels (above); the opposed-direction geometry is confirmed against the Fig. 2 page image (QE revision); eq. (9) also differs by factor 2 from the same-sign path geometry.
4. One reported value, Delta = 0.25 mm (about 397 wavelengths), v about 378 km/s, east-west with Leo on the horizon; nulls N-S and with Leo 6 or 18 h away; no errors, runs, dates or place. The value agrees with the CMB dipole (370 km/s toward Leo) to about 2%. Silvertooth's 1989 statement says all values lay within +/-5% over several years [web: https://spirit-science.fr/ArchivesScientifiques/1989Silvertooth.pdf, via Investigator transcription]. Related results by the inventor and others, kept beside it: 1972 JOSA one-way test recorded as a null by Roberts and Schleif; Marett 2011 single-laser one-way standing-wave null; Whitney and Silvertooth 1992 positive nodal-spacing claim (QE revision). 378 reported again in 1989 is the same value, not a second measurement.
5. A velocity-dependent wavelength count modifies the Doppler shift, with an extra L x (acceleration)/c^2 term (v-bar in 1986 is very probably an acceleration, per the 1976 paper), said to explain the accelerating universe, quasar redshift anomalies and GPS errors. No numbers in the 1986 paper.

Final verdicts

Claim 1

Agreement: all three: weak form Established as conventionality; argument circular (A2 in = A2 out). Disagreement: none material (Contrarian puts the strong form as "Not testable as stated" and routes it through claims 3-4; Physicist adds that a Galilean ether would give a second-order effect, so the chosen c_i is built to avoid it). Stronger argument: Physicist and Mathematician, who show eq. (5) is the Reichenbach epsilon = b convention. Verdict (split): weak form Established; strong form Not testable as stated (and circular as a rebuttal of contraction). What would settle it: nothing experimental for the weak form. For the strong form, claims 3-4.

Claim 2

Agreement: algebra of the rotating loop correct; extension to translation is an assumption (A3); author's own 1976 text concedes closed-loop cancellation. Disagreement on label: Physicist "Tested, not supported" for measurability (own 1976 null, Krisher 1990); Mathematician "Internally inconsistent" (derivation); Contrarian "Tested, not supported". Stronger: the Mathematician for the derivation; the Physicist for the experimental half. They do not conflict: different parts. Verdict (split): derivation from Sagnac: Internally inconsistent (assumption contradicts the author's own zero closed-loop statement; segment share depends on chosen centre). Resulting formula: Established only as a synchronisation convention (it equals the standard L v/c^2 offset). Claim that it is physically measurable in translation: Tested, not supported (own 1976 null of 79 expected cycles; Krisher 1990 limit 100 m/s, with the conventionality caveat). Settle: a calibrated two-oscillator path test over an east-west baseline with independent synchronisation (Silvertooth's own 1976 Arizona proposal, 59 cycles expected), whose report was not found.

Claim 3

Disagreement: Physicist "Refuted by calculation"; Mathematician "Internally inconsistent (conditional on reconstruction)"; Contrarian "Tested, not supported". Recomputation supports Physicist and Mathematician, with the caveat that the paper gives no derivation and the apparatus geometry is read from words and a figure. The Contrarian's grounds (Marett replication, precision one-way tests) remain relevant but address an empirical question that is moot if the apparatus should not respond at all in the model; they also leave unexplained why a signal appeared for Marett. That is a point the calculation cannot explain and that stays open (see Claim 4). Verdict: Internally inconsistent as stated: within eqs. (1)-(5) and the described opposed-path geometry the first-order signal cancels (matching the author's own 1976 null), and eq. (9) differs by a factor 2 under the alternative geometry. Held conditional: if a correct derivation of delta = lambda b from the real geometry is produced, revert to Never properly tested. Settle: the author's derivation or a drawing with dimensions; or a calibration run where a known path or index change is imposed with the effect absent, checking the D1/D2 reading.

Claim 4

Agreement: arithmetic right; no data; the CMB agreement is striking but without error bars means little; Marett's signal-plus-thermal result; nobody ran a neutral, controlled, blinded test. Disagreement: Physicist "Never properly tested" (fallback Tested, not supported); Contrarian split. Stronger: the Contrarian's split is the most informative and I adopt it, because it keeps both halves of Marett's finding. Additional points kept: laser frequency drift is about 1e-6, a thousand times too small to fake 1.3e-3 [calc, memory high for HeNe bandwidth], which favours the inventor against one objection; pointing/twist/AC-coupling remain open (Sherwin, Phipps, Marett); geometry predicts E-W maxima at rise and set and zero at transit, matching Marett's peaks near 5 h and 17 h sidereal, but also predicts a N-S signal near transit (about 0.82 of full) that the paper does not time-stamp; the 180 degree criterion is sign-blind; phase judged by eye with Delta chosen by the experimenter [read: p.4]. Verdict (split): the original run: Never properly tested (single value, no data, not dismissed without testing: Marinov, Wesley (abstract only), Marett engaged with it). A sidereally modulated signal of about the stated size in this instrument type: reproduced once, with the replicator's own thermal/pointing attribution (the best-documented replication). Interpretation as 378 km/s absolute velocity: Tested, not supported (and the apparatus, by Claim 3, should not respond to it). Settle: Contrarian's protocol: lock-in readout, thermally controlled and pointing-monitored laser, rotating table separating sidereal from solar/thermal pattern, declared N-S run near transit, raw time series published. Plus the original data.

Claim 5

Agreement: 1986 eq. (10) undefined; v-bar most likely acceleration (1976); magnitude order z about 0.03 is tunable; accelerating-universe and GPS statements have no derivation. Disagreement: Mathematician "Refuted by calculation" for the quasar example; Physicist and Contrarian "Not testable as stated". Stronger: Mathematician's calculation is sound for the rotating example (acceleration direction reverses within light flight time) but the claim also has other forms (Physicist notes the L a/c^2 shift is real for co-accelerating observer and source), so I limit the refutation to the quasar rotation example. GPS: the first-order Sagnac correction is already applied in GPS [memory, high], so "GPS ignores it" would be a mischaracterisation; the 1990 paper was not read. Verdict (split): eq. (10) as printed in 1986: Internally inconsistent (undefined symbols; units fail on literal reading). Quasar example with L a/c^2 for a rotating source: Refuted by calculation (period shorter than flight time). Accelerating universe: Not testable as stated. GPS: Not testable as stated in the 1986 paper; Tested, not supported insofar as GPS performs at metre level with standard corrections (not read: the 1990 paper). Settle: the numerical GPS error budget from Galilean Electrodynamics 1990 against published GPS residuals; differential redshift across rotating galaxies (the predicted L omega/c amplification, not known to be observed [memory, low]).

What would settle it (overall)

1. Obtain the missing primary documents: Nature 322:590 (1986) letter, Wireless World Nov 1986 and Nov 1989, JOSA 1972, Physics Essays 1992 (Whitney and Silvertooth), Galilean Electrodynamics 1990, Sherwin 1989, and Marett's full data.
2. Run the blinded, instrumented replication of Claim 4.
3. Derive the claim-3 phase shift from a stated geometry (or run the calibration).

Open questions

• Why did independent replicators (Marett, Marinov, perhaps Wesley) see a signal at the CMB scale and aligned with sidereal time if the apparatus should be blind? Marett's thermal explanation fits his data but a 12 h pattern with CMB-sized amplitude and correct phase is a coincidence that deserves a direct test (rotating table).
• Is the geometry of the apparatus different from what the reviewers reconstructed? The result of Claim 3 hinges on it.
• Wesley's "378 +/- 8 km/s" equals Silvertooth's value and is tighter than his own +/-5%; is it a re-analysis? Unknown.
• The cosine-of-orientation and E-W/N-S timing: when was the N-S null run?

Absence of evidence: where the record goes dark

1. Sponsorship (RADC, DARPA, "in part", 1986): no report, DTIC entry or acknowledgement found; a 1976 Arizona OSC experiment "under way" with no outcome found. Competing explanations: (a) informal or in-kind support with no report; (b) reports unindexed or restricted; (c) overstated sponsorship; (d) an RADC physicist (Newburgh) criticised the 1976 paper in print, so support may have been limited. Evidence for/against: for (d) the published critique exists [web via Investigator]; for (b) the Natural Philosophy Wiki biography mentions Air Force think-tank work but gives no source (see QE revision). Nothing decides between them. Decisive records: direct DTIC search, AFRL Rome/DARPA archives, FOIA, Arizona OSC FY75 report, Silvertooth's correspondence.
2. Original data: none of his papers publishes data graphs (Marett). Explanations: unpublished lab notebooks; journal (Speculations in Science and Technology (journal named by Marett's page [web via Investigator]; no data graphs published)) not requiring them; data never kept; inability to reproduce. The Nature letter promised details that did not reach a refereed journal [web: Koberlein/Forbes via Investigator]. The 1989 sidebar's suppression charge is an unsigned sidebar writer's characterisation; I find no document of the experiment being blocked, and it was repeated by others, so "dismissed without testing" does not fit; "never tested under controlled, published conditions" does.
3. Silvertooth after 1992: the claim sets does not continue in mainstream literature; explanation of a thin record includes an outsider's lack of channels, mainstream disinterest given the SR conventionality argument, and the thermal-artefact finding. Decisive: his archives and later correspondence.

Retention note

Items kept beside the verdicts: his 1983 standing-wave sensor (20,000:1 node/antinode, real instrument), 1941 Academy Award and Caltech record (competent optical engineer), his printing of his own 1976 null, Navia 2006 and Marinov/De Witte/Cahill related work (characterised by others as not accepted; not examined), and the critiques (Sherwin, Phipps, Newburgh, Gron) with their own stakes.

Summary table

QE revision (post-QE, MINOR)

Source: qe.md (verdict MINOR). I checked the 1976 paper's text layer myself [read: research/files 1976 pdf, journal p.1101 and the Arizona sentence] and the Tyapkin citation in the project text [read: _text, p.2-3 and ref. 5]. No verdict changes. Changes are retention additions, clarifications and label fixes.

1. Claim 3, geometry caveat narrowed. The QE read the Fig. 2 page image: beam lambda1 runs leftward inside the slide to the fixed mirror, beam lambda2 runs rightward to a fixed mirror, so a slide move of Delta shortens one leg and lengthens the other and the phase changes by (k1+k2)Delta, as in my recomputation 2 [read: page image p.3, via qe.md; not independently re-rendered by me]. The caveat is now only that the paper gives no dimensions and no derivation of delta = lambda b. Verdict stays Internally inconsistent as stated; the revert condition (a correct derivation from the real geometry gives Never properly tested) stays.
2. Claims 1-2, the inventor reverses himself. 1986 premise A1 says frequency cannot change, so only the wavelength changes [read: 1986 p.1]. In 1976 he wrote that the phase shift "has been experimentally demonstrated by the writer to be a change in the frequency and the velocity, the change in wavelength being a resultant effect" [read: 1976 p.1101]. The 1986 paper does not mention the earlier experiment. This is an internal inconsistency across his papers. The 1976 null used a different setup (two free-running lasers, beat frequency held under 3 MHz, two photocathode detectors, oppositely directed beams) [read: 1976], so it is a related null, not the same apparatus as 1986; my earlier phrase "own 1976 null" should be read that way. Verdicts for Claims 1-2 unchanged.
3. Claim 4, middle part labelled and evidence added. - Label: the Marett reproduction of a sidereally modulated signal is an observation of one replicator, with his thermal/pointing attribution and no controlled test of the cause. Scale label for the cause: Never properly tested. For the absolute-velocity reading: Tested, not supported (unchanged). - Added: Marett's 2011 companion single-laser one-way standing-wave experiment, reported as a null [web: http://www.conspiracyoflight.com/Silvertooth/Silvertooth.html, via Investigator]. Marett is an interested party in the sense that he proposes the thermal explanation. - Added: Silvertooth's 1972 JOSA "Isotropy of C" one-way test with frequency-doubling crystals, recorded by Roberts and Schleif as a null consistent with special relativity [web: https://math.ucr.edu/home/baez/physics/Relativity/SR/experiments.html; the JOSA paper not read]. This is the inventor's own earlier result against the effect. - Added: Whitney and Silvertooth 1992, Physics Essays, a standing-wave nodal spacing inside one Michelson arm departing from lambda/2, fitted v0 about 378 km/s [web: naturalphilosophy wiki abstract, via Investigator; paper not read]. It is the inventor's own later positive claim. In the eq. (5) model the node spacing is exactly lambda/2 (Mathematician), so the 1992 claim conflicts with the 1986 model as well. Not given a verdict (not read).
4. Claims 2-3, dismissals and a mainstream test. The paper says Turner-Hill (Moessbauer) and Cole (pulsar) results are "defeated through the second order by the argument of Tyapkin" and gives no argument [read: 1986 p.2-3, ref. 5 Tyapkin, Lett. Nuovo Cimento 7, 760 (1973)]. Tyapkin's argument is the conventionality argument that Marett uses against a one-way wavelength measurement, so it cuts against Claim 3. Gagnon, Torr, Kolen and Chang, Phys. Rev. A 38, 1767 (1988), a guided-wave first-order one-way test, is cited by the Contrarian as the strongest mainstream evidence against Claim 3 [web via Contrarian; result not read by me, so I do not rely on it for a verdict].
5. Corrections. The sponsorship note's biography source is the Natural Philosophy Wiki, which gives no source for the Air Force think-tank statement (Aerospace Corp.), not a "Wikipedia-type biography". I removed my "fringe venue" characterisation; no source states it in those words. The Contrarian's statement that the 1976 Doppler factor equals the standard sqrt((1-b)/(1+b)) is superseded by recomputation 5 (0.433 against 0.577 at b = 0.5). The orbits-in-flight figure depends on parameters: about 3.5 for d = 10 kpc, v = 200 km/s (Mathematician); 5.3 for omega = 1e-15 rad/s (example.py). Both support the Claim 5 quasar refutation.
6. Open leads added (retention). "The Silvertooth effect: TETET-99" (1999, UNIAM, ISBN 978-0968308547; authorship unknown); the 1989 statement that groups in the USA, West Germany, UK, Italy, France and Austria worked on explaining the result; Wireless World Nov 1986 and Nov 1989 coverage. Next Investigator run: read Marett's page directly, save the 1989 article PDF (owner approval), obtain Nature 322:590, JOSA 1972, Physics Essays 1992, Galilean Electrodynamics 1990 and Sherwin 1989, and run a DTIC search for the RADC/DARPA trail.

Verdict changes

None. Claim 4's middle part now carries a label (Never properly tested for the cause); the Claim 3 geometry caveat is weaker; the table and Corrected claim list above were updated accordingly.