SilverTooth: Physicist review (Step 2)

Project: E. W. Silvertooth, "Experimental detection of the ether" (received May 1986; Speculations in Science and Technology 10(1) 3, journal per Investigator [web]). Reviewer: PhD physicist. Tags: [read], [web], [memory]. No source gap: the 1986 paper is fully readable, but it is one paper with no data table. I also read the 1976 Applied Optics paper saved by the Investigator [read: research/files/Silvertooth-1976-AppliedOptics-15-1100...pdf, p.1100-1102].

Corrections to Step 1

1. "v-bar" in eq. (10) is almost certainly an acceleration. The deep read left it undefined, and the Investigator's lead is right. The 1976 paper says explicitly "the radial acceleration of the source" and gives the quasar example, radial acceleration = domega^2 [read: 1976 p.1102]. The units objection in the deep read (Lv/c^2 has units of time) falls away if it is an acceleration. Eq. (10) has v^2/c^2 inside both roots, but the 1976 text writes the factor as (1/gamma)(1 +/- v_r/c), so the 1986 form is probably a retyping slip. This is my reading of a garbled passage, not certain.
2. The sign swap of lambda1/lambda2 between eq. (5) and Fig. 2 is a real inconsistency, but it does not change magnitudes. I confirm it. The 1989 article uses lambda(1 +/- v/c), which agrees with eq. (5) only to first order [web: Investigator note, spirit-science.fr].
3. claims.md claim 2 says "straight segment in pure translation" is a Sagnac shift. The 1976 paper shows Silvertooth knew the closed-loop result: "no interferometer arrangement employing a single source can yield a fringe shift in this circumstance", because the shift along ab "is simply cancelled in the closed path" by the other two sides [read: 1976 p.1101]. Step 1 omitted this; it is central (see claim 2).
4. The deep read's arithmetic (397 wavelengths, 378 km/s, v/c = 1.26e-3) is correct. I reproduced it. The 1989 article (2v/c = lambda/Delta) is the same relation [web].
5. The Investigator's lead list is accurate as far as I could check. The Wesley "378 +/- 8 km/s" figure comes only from a wiki abstract, so I treat it as a report whose data I have not seen.

Mainstream view, stated fairly

Special relativity (SR) with Einstein's clock synchronisation makes the one-way speed of light c in every inertial frame. The round-trip speed is c in any theory that explains Michelson-Morley. It is well established that the one-way speed alone cannot be separated from the convention for synchronising distant clocks (Reichenbach 1928, Winnie 1970, Tyapkin 1973; all cited by Marett and the paper) [memory, high]. A Lorentz ether theory, in which the ether frame has a preferred simultaneity, is empirically equivalent to SR: it predicts no lab-detectable ether wind, because lengths contract and clocks slow in just the right way. So an honest critic does not say "the ether is impossible". The critic says "nothing in the lab can measure motion through it". Silvertooth claims the opposite, and this is the whole dispute.

Claim-by-claim (claims.md numbering)

Claim 1: Michelson-Morley cannot detect an ether, since a direction-dependent c shows up as a direction-dependent wavelength with frequency fixed; no isotropic c or Lorentz contraction is needed.

a. Mathematics. Eqs. (1)-(3), (5)-(7) are algebraically fine: with c_i = nulambda_i and a round trip fixed at 2L/c, the equations 1/c1 + 1/c2 = 2/c follow. I checked eq. (5) [read: p.1-2]. But the round-trip constraint (eq. 3) is an input (assumption A2), and "n and t independent of v" is just that input returned. Physical content: this is the Reichenbach synchronisation freedom, c1 = c/(1+beta), c2 = c/(1-beta) with beta = v/c, which is exactly the epsilon = beta convention. The "no Lorentz contraction needed" statement fails in an actual ether picture. A Galilean ether with c +/- v gives round trip (2L/c)/(1-beta^2), which is not 2L/c. Contraction or the special c_i of eq. (5) is needed to save the null, and eq. (5) is chosen precisely to do that. The transverse arm is not treated at all. b. Experiment. Consistent with all Michelson-Morley-type results, including modern ones to about 1e-17 [memory, medium]. The assumed null is built in. c. Tested. Not an experimental claim in itself. The conventionality of the one-way speed is a standard result that SR textbooks accept. d. If true:* nothing, in the weak form. The strong form (a physically measurable ether wind in the lab) would require relativity of simultaneity to fail operationally.

Verdict (split). Weak form (MM cannot fix the one-way speed): Established (as conventionality). Strong form (no contraction needed; the anisotropic c_i is physical rather than a synchronisation choice): Internally inconsistent in its use of the constraint (circular; A2 assumed). I would give it Not testable as stated for the part that remains a convention.

Claim 2: Sagnac shift on a straight segment in "pure translation" gives one-way wavelengths lambda/(1 +/- v/c).

a. Mathematics. Fig. 1 algebra is right: delta = 4Aomeganu/c^2 with A = Lr/2, omega = v/r gives 2Lv/(clambda) [read: p.2; checked]. In the rotating loop, the radial sides contribute zero because v is perpendicular to them, so the entire shift sits on the far side L. That is true. But the Sagnac phase is a closed-loop line integral of v.dl/c^2. For pure translation, v is constant and the integral around a closed loop is zero. Silvertooth himself says so in 1976 [read: p.1101]. His step is to assign the segment term alone a physical meaning. In SR the term Lv/c^2 is the relativity-of-simultaneity offset in t' = gamma(t - vx/c^2): it is the synchronisation gap between the ends of a moving rod. The 1976 paper derives the same phase "by invoking only the Lorentz transforms" [read: p.1102]. So the term is already in SR, and there it is a bookkeeping offset that cancels around any closed light path, so it is not an observable ether wind. The "discontinuity at zero curvature" argument (Fig. 2 of 1976) assumes the segment term is observable on its own; for a straight loop it is not observable, and for a rotating loop the closed integral is what is observed. b. Experiment. Sagnac, Michelson-Gale and ring-laser gyroscopes confirm the closed-loop effect (rotation) [memory, high]. No experiment shows a first-order closed-loop translation effect (Kennedy-Thorndike, Michelson-Morley, Ives-Stilwell). A direct one-way test of the kind Silvertooth proposed in 1976 (stable oscillators at both ends of an E-W path, watch the phase for a diurnal 59-cycle variation) was performed at JPL: Krisher et al. 1990 used two hydrogen masers on a 21 km fibre link and set an upper limit on one-way linear anisotropy of about 100 m/s [web: https://pubmed.ncbi.nlm.nih.gov/10012893 via search summary]. Silvertooth's 378 km/s would be about 3800 times larger. (Caveat: whether a two-clock link is sensitive to a synchronisation-convention effect is itself debated; the result is stated here as the experiment's own report.) c. Tested. Silvertooth's own 1975-76 two-laser experiment gave no phase shift [read: 1976 p.1101]. The Arizona OSC experiment he said was "under way" has no report in the record I could find (Investigator). d. If true:* simultaneity would have to be absolute operationally; SR's second postulate and the Lorentz invariance of QED would fail at first order. That would also break GPS-type clock-synchronisation, which works with Einstein synchronisation plus a Sagnac correction.

Verdict. The first-order algebra is Established as a clock-synchronisation convention (it reproduces the standard Lv/c^2 offset). The physical claim that it is measurable in translation is Tested, not supported (his own 1976 null; Krisher 1990) and conflicts with the closed-loop cancellation he acknowledged.

Claim 3: A slide-mounted two-beam standing-wave interferometer measures the one-way wavelength: v/c = lambda/(2*Delta).

a. Mathematics. The paper states the shift delta = lambda(v/c) but derives it only in words [read: p.3-4]. I tested it with his own premises (frequency nu common to both beams, k1 = 2pi(1+beta)/lambda, k2 = 2pi(1-beta)/lambda from eq. 5). Two counter-propagating beams of equal frequency form a standing wave with node spacing 2pi/(k1+k2) = lambda/2, which is exactly the isotropic value because 1/lambda1 + 1/lambda2 = 2/lambda. When the slide moves by Delta, one path grows by Delta and the other shrinks by Delta. The pattern's antinode position shifts by (k1Delta + k2Delta)/(k1+k2) = Delta, exactly, with beta cancelling (SymPy check, run in this review). The reference interferometer D2 (a round trip) tracks the same Delta. So the D1-D2 phase difference should be zero for any beta. The apparatus as drawn is blind to the effect in the model that is supposed to produce it. His step "waves advance by 1+v/c and retard by 1-v/c" gives a relative phase change of 2 waves (not 2v/c) when the signs are carried through. This is a calculation under his own eq. (5). It matches the independent arguments of Tyapkin and Marett that a one-way wavelength test is as impossible as a one-way speed test [web: Investigator, conspiracyoflight.com]. I am not claiming certainty about every detail of the beam geometry in Fig. 2, which I read from the deep read's description rather than a drawing of my own. b. Experiment. See claim 4. c. Tested. Replicated in some form: Marinov 1987 (different arrangement), Marett 2012, Wesley 1987. See claim 4. d. If true: a preferred frame measurable by pure optics at first order in v/c.

Verdict. Refuted by calculation (the standing-wave phase shift is independent of v under the paper's own premises), with the caveat above. If the calculation is wrong, the fallback is Tested, not supported (claim 4).

Claim 4: East-west, Leo on the horizon: Delta = 0.25 mm, so v = 378 km/s; no effect N-S or at Leo 6 or 18 h away; agrees with Muller's U2 result.

a. Mathematics. v/c = 0.63e-6/(20.25e-3) = 1.26e-3, v = 377.7 km/s with c = 299,792 km/s (378 with 3e5) [checked]. Consistent. b. Experimental consistency. The direction: Earth's CMB velocity is 369.82 +/- 0.11 km/s toward Leo, (l,b) = (264.0, 48.3) [web: https://www.aanda.org/articles/aa/full_html/2020/12/aa38053-20/aa38053-20.html, via search]. So 378 km/s toward Leo is a real number that matches the CMB dipole, to about 2%. This is striking and I do not dismiss it. Two cautions. (i) The paper states no uncertainty, so the 2% agreement means little; the 1989 article says values were "within +/-5%" over years [web: Investigator]. (ii) A velocity relative to the CMB radiation is not a velocity relative to an ether that SR forbids detecting; the CMB gives a preferred radiation frame, which is compatible with SR. Muller's 1978 value is, per my memory, about 390 +/- 60 km/s [memory, medium; my search did not confirm it]; the paper itself gives no number. Size of the effect: Delta = 0.25 mm over 397 wavelengths with a half-wavelength phase reversal means a scale mismatch of 1.26e-3 between D1 and D2. A cosine misalignment of about 2.9 degrees (1 - cos(theta) = 1.26e-3) between a beam and the slide axis gives the same mismatch, so ordinary alignment, tilt, wavefront or thermal-drift errors in a visual oscilloscope comparison can reach this size [calculation in this review]. Sherwin (1989) and Phipps argue a lambda/4 twist or AC-coupling artefacts could mimic the reversal [web: Investigator]. The paper gives no runs, errors, dates or location [read: deep read section 9]. c. Tested. (1) Marett (2012), the only replicator with detailed data I can see: reproduced a ~12-hour diurnal signal peaking near 5 h and 17 h sidereal, with magnitude about the CMB velocity after normalising. He also found a nonzero offset when perpendicular, large day-to-day scatter, the opposite sign to Silvertooth's prediction, the same pattern in local solar time, and correlation with room temperature and laser lock voltage. His conclusion: "a very elaborate thermometer" (beam-angle drift) [web: conspiracyoflight.com, Investigator]. Both halves stay in the record: the signal appeared; the replicator attributed it to temperature. His companion single-laser test gave a null (not read). (2) Marinov 1987: 386 +/- 38 km/s, RA 12.5 +/- 0.5 h, a different arrangement; Marinov was himself a claimant of absolute-motion detection (interested party). (3) Wesley 1987: 378 +/- 8 km/s, RA 11 +/- 2 h, from an abstract only; data unseen. (4) Silvertooth's 1989 repeatability statement is his own, with no data. (5) Nanning, Aspden, Wilczynski, Whitney 1992 papers not read. (6) Silvertooth's own 1972 JOSA one-way test is listed by Roberts and Schleif as a null consistent with SR [web, via Investigator; not read]. So: replicated as a signal by three people with unreported or weak controls; the one with the controls attributes it to an instrument artefact. No neutral controlled test (shuttered beams, thermal chamber, rotating table) is on record. Not dismissed without testing; also no mainstream group has tested it (the Investigator found no institutional replication). d. If true:* a first-order one-way anisotropy of 1.3e-3 would contradict Krisher 1990 (100 m/s limit), Ives-Stilwell-type and Riis 1988 first-order tests [memory, medium], and GPS, which uses one-way signals over 2e7 m ranges: 1.26e-3 x 2e7 m is about 25 km, while GPS works at metre level [calculation + memory, high; again subject to the synchronisation-convention caveat].

Verdict. Never properly tested as a controlled neutral measurement, with the replication evidence mixed (signal reproduced, instrumental explanation given by the best-documented replicator). If forced to one: Tested, not supported as evidence of an ether wind, because the one detailed replication traced it to temperature and found the wrong sign.

Claim 5: The same effect (velocity-dependent wavelength count altering the Doppler shift, with an extra L*v-dot/c^2 term) explains the accelerating universe, anomalous quasar redshifts and GPS errors.

a. Mathematics. Eq. (10) is undefined in the paper. With the 1976 meaning, nu_o = (nu_s/gamma)(1 +/- v_r/c)(1 +/- La_r/c^2) with a_r the radial acceleration [read: 1976 p.1102]. It is dimensionless. The term comes from the same Lv/c^2 segment phase differentiated in time, so it inherits the claim 2 problem. For comparison, the La/c^2 shift is real in physics, but for an observer and source accelerating together (equivalence principle: redshift gL/c^2 between heights). For a source accelerating relative to an inertial observer, the SR Doppler depends on the source velocity at emission and has no L-dependent term. Worked number: L = 1e27 cm = 1e25 m, a = domega^2 with omega about 1e-15 rad/s and d about 3e20 m gives a = 3e-10 m/s^2 and La/c^2 = 0.033. That is an effect of order z = 0.03 whose sign depends on rotation phase. Possible in principle; an assertion with numbers unspecified in the 1986 paper. b. Experiment. The acceleration of the universe is supported by Type Ia supernovae (1998) and later data; Gunn and Tinsley 1975 is a different, earlier argument [memory, high]. Quasar redshift differences within one object are not accepted evidence for rotation-induced shifts; mainstream attributes quasar redshifts to cosmological expansion [memory, high]. GPS performs to about a metre with the standard relativistic and Sagnac corrections; no residual of the claimed kind is documented. His 1990 Galilean Electrodynamics paper on "position error" is abstract-only in the record. c. Tested. No tests; no numbers in the paper. d. If true: a whole alternative explanation for supernova dimming, quasars and GPS; the equivalence of inertial frames would have to fail.

Verdict. Not testable as stated (no definitions or numbers in the 1986 paper). The GPS part is Tested, not supported by the fact that GPS works without the proposed correction.

Investigator findings kept in the record (whatever the verdict)

• 1976 paper: the null two-laser experiment (79 cycles expected for dv/c = 1e-4, none seen) is Silvertooth's own result against the wavelength-shift idea in that setup; he reads it as invariance of relative phase [read: 1976 p.1101]. Newcomb's 1881 letter to Michelson quoted [read]. The Arizona OSC experiment mentioned has no report found.
• 1983 Applied Optics standing-wave sensor (Silvertooth and Jacobs): node/antinode about 20,000:1, effective thickness < lambda/100; the D1 detector is a real instrument. A genuine contribution to optics, independent of the ether claim.
• 1941 Academy Award (Paramount relay condenser), Caltech degree and a long optics record show a competent optical engineer, not a naive claimant [web: Investigator].
• RADC/DARPA sponsorship: an absence-of-evidence anomaly. Competing explanations: informal or in-kind support with no report; unindexed or restricted reports; overstated sponsorship. An RADC physicist (Newburgh) criticised the 1976 paper in print. DTIC direct search, FOIA and the OSC FY75 report would decide. I take no position.
• The 1989 sidebar's complaint of suppression is the unsigned sidebar writer's characterisation; I cannot assess it from the record.
• Navia 2006 (laser diffraction anisotropy aligned with the CMB, a = -0.4106 +/- 0.0225 vs -0.5) and Marinov/De Witte/Cahill work share the claim; characterised by others as not accepted. Not independently examined here.
• Critiques (Sherwin, Phipps, Marett, Grøn, Newburgh) are retained beside the support.

Web checks done

CMB dipole 369.82 km/s toward Leo: confirmed [web: Planck/A&A]. Krisher 1990 limit ~100 m/s: confirmed from abstract summary [web: PubMed 10012893]. Muller's 390 +/- 60 km/s and Riis 1988: not verified [memory, medium]. Sagnac conventionality sources [memory, high].

Summary table