Su-CC: Arbiter (Step 3)

Project: Ching-Chuan Su (C.-C. Su), National Tsing Hua University (NTHU), Hsinchu, Taiwan: the "local-ether" model and "Quantum Electromagnetics". Date: 2026-10-06. Inputs weighed: facts.json, claims.md, physicist.md, mathematician.md, contrarian.md, deep/, sites.md, research/. Tags: [read] from a project document or a review file that read it, [calc] my own calculation (venv Python, mpmath), [web] with URL, [memory] with confidence. Su is the originator and an interested party; his accounts of relativity, GPS practice and other people's experiments are his characterisations. The Investigator's sources Marsen and Gift are advocates (stake: they promote the model).

Source gap (repeated from claims.md)

The papers in the folder are all readable. Not in the folder: the 2005/2008 book (outline and web summary only), Su's papers on radar echo time, explicit field definitions, Sagnac (JEWA 15, 945), the railgun paper, and every experimental paper Su cites (Brillet and Hall, Staudenmann et al., Krisher et al., Mandelberg and Witten, Kundig, Champeney et al., Kennard and Barnett). Comparisons with those experiments rest on Su's citation or on [memory] unless tagged [web]. Two verdicts below rest partly on [memory]: "no dedicated satellite Michelson-Morley test has flown" (see claim 5) and the Compton coincidence experiments (claim 3).

1. Where the reviewers differed, and how I settled it

Most of the review is agreement: the algebra is correct, and nearly every number Su matches to data (GPS, Hafele-Keating, Ives-Stilwell, Sagnac, Shapiro delay, deflection) is also the special plus general relativity (SR/GR) number computed in an Earth-centred inertial (ECI) frame. The disagreements were five.

D1. Is the 62 Hz cavity prediction tested? (Physicist: never properly tested. Contrarian: never properly tested. Mathematician: provisionally tested, not supported, pending a check of Herrmann et al. 2009.) I did the check. I downloaded arXiv:1002.1284 and read its text [web: https://arxiv.org/pdf/1002.1284]. Herrmann et al. (Berlin, colatitude 37 deg, cavities near 282 THz, turntable period 45 s, data May 2007 to June 2008) fit the constant, laboratory-fixed 2 x turntable-rate amplitudes S0 and C0 separately from the sidereal ones. Their year-mean values are C0 = -0.2 +/- 11.7 and S0 = -10.2 +/- 14.4, in units of 1e-18 [web: Fig. 4 caption, same URL]. They say individual points deviate by several standard errors, "vary in magnitude and phase and thus average out", and that these terms are the ones most affected by residual systematics. Su's effect has the same signature: a term at twice the turntable rate, fixed to the laboratory because it is tied to geographic east. Its size is (v/c)^2/2 with v = 465 m/s x cos(latitude): at Berlin (52.5 deg) v = 283 m/s and the fractional shift is 4.5e-13 [calc]; Su's own Brillet-Hall figure (40 deg) is 7e-13 = 62 Hz at 3.39 micron [calc, agrees with all three reviewers]. The measured year mean is about 1e-17, a factor of roughly 4e4 below the prediction. A constant lab-fixed term would not average out over a year; that is exactly what distinguishes it from the varying systematics. Caveats I keep: (i) I did not re-derive Su's prediction for a pair of orthogonal cavities on a turntable (for two orthogonal arms the difference signal should keep the full amplitude, but this is my inference); (ii) the experiment was designed to test sidereal modulation, not Su, so it is an incidental test, not a dedicated one; (iii) I read the caption and text, not the raw data. Result: the Mathematician's provisional reading is right, and the Physicist's and Contrarian's "never properly tested" is out of date for the cavity quadrupole. This also reframes the 1979 Brillet-Hall residual (about 34 Hz peak to peak, 3.8e-13, per Su; their paper not read by anyone in the pipeline, and a web search returned only the abstract [web: https://vivo.colorado.edu/display/pubid_108412]): a lab-fixed term of that size in 1979 and under 1e-17 in 2008 means the 1979 term was most probably a systematic. That is an inference; it was one old measurement, and Marsen's reading of it as a possible positive signal (an advocate's reading) is recorded here and is not supported by the later data. The Contrarian's recommended clean test (re-orient the apparatus, change latitude) has in effect been run year-long by a rotating table whose lab-fixed amplitude stayed null.

D2. The one-way turntable link (Su eq. 33). (Physicist: internally inconsistent for atomic oscillators. Mathematician and Contrarian: untested, hinges on clock transport.) I checked the Physicist's derivation. Numerically, with light-time iteration (c = 299792458 m/s, V = 355 m/s, 10 m link on a turntable), the extra propagation delay is -V.d/c^2 per unit V, equal to -d cos(alpha), alpha the table angle, to 12 digits [calc]. If both oscillators obey Su's own f = f0 sqrt(1 - v^2/c^2) with v = V + u (u the table velocity of each end), their rate difference is -V.(u_B - u_A)/c^2. Integrated in time that is exactly -V.(change of the link vector)/c^2, with the opposite sign to the propagation term (u_B - u_A is the time derivative of the link vector). The two cancel up to a constant [calc, analytic; I did not integrate the clock term numerically]. So the Physicist is right: for atomic-clock endpoints Su's model, with its own eq. 23, predicts what SR predicts, and eq. 33 omits a term his own formula implies. The effect survives only for oscillators whose frequency does not scale that way, which Su says is "not precluded" [read: f7 p.4].

D3. The transverse radar Doppler "-2v^2/c^2". (Physicist: refuted as a difference from SR. Mathematician and Contrarian: listed as a prediction, no test.) I recomputed. For a stationary transceiver and a target on a straight line (v = 7.5 km/s, closest approach 500 km), the exact round-trip frequency ratio is dt_emit/dt_receive, a purely kinematic quantity that any theory with light speed c in the transceiver's frame must give. It equals the standard (1 - u')/(1 + u') with u' the radial speed at the reflection event, to 40 digits, at four points along the pass. The expression 1 - 2u - 2v^2 + 4u^2 (u at emission, as quoted by the Physicist from Su's eq. 34) differs from it by about 2e-19, which is third order [calc]. The "-2v^2/c^2" appears only because u is taken at emission instead of reflection. The Physicist is right: this is a convention difference, not a prediction that differs from relativity. The arithmetic of "a few Hz to 12 Hz at 10 GHz" is fine; the same shift is in the standard formula.

D4. Quasar claim: Internally inconsistent (Mathematician, Contrarian) or Refuted by calculation (Physicist). The numbers all reproduce: with R = 4.7e18 m and density 6e-11 kg/m^3 I get M = 2.6e46 kg (1.3e16 solar masses), Phi/c^2 = 4.1, escape speed 2.9 c, free-fall time 272 years, Su's formula z = sqrt(1 + 2 Phi/c^2) - 1 = 2.04 (Su rounds to M = 2.7e46, Phi/c^2 = 4.2, z = 2.07 [read: f5c]); the GR form 1/sqrt(1 - 2GM/Rc^2) is undefined because the Schwarzschild radius exceeds the radius [calc]. I side with "Internally inconsistent" as the verdict: it is demonstrable from Su's own premises (the index and frequency law were derived for Phi/c^2 much less than 1 [read: f5a p.3] and are used at 4.1). "Refuted by calculation" would need accepted laws applied correctly, and at escape speed above c the Newtonian free-fall time is itself outside its range. The stability objection (no pressure or rotation support, 1e16 solar masses in about 1000 light-years) stands as a serious unanswered problem that I keep next to the verdict, not as a separate refutation.

D5. Neutron Bragg-angle asymmetry. (Physicist: tested, not supported, weakly. Mathematician: never properly tested. Contrarian: ambiguous.) The one comparison is Su's own reading of two numbers from Staudenmann et al. 1980, which he calls "not good quantitatively"; with his stated zero the sign is opposite, with the other zero it matches, the zero-independent separation (140.49 deg) favours the standard law, and the wavelength may have been derived from the same readings [read: deep/f8a, via Contrarian and Mathematician]. I did not obtain the paper. That is not a controlled test. Verdict: Never properly tested, with the single comparison leaning against Su's size of effect. The Physicist's "tested, not supported" overstates a comparison that has an unknown zero offset.

Minor disagreements: (a) the Mathematician's "Internally inconsistent" for the full force law's dependence on the zero of the scalar potential. Su's law is built on potentials referenced to infinity, with a neutralisation assumption, and he never claims gauge freedom; it is a departure from standard electrodynamics and a defect (acceleration-dependent inertia m0 - q Phi/c^2, 20% for a 100 kV electron [read: deep/f2a]), not a self-contradiction. I record it as a defect and class the extra terms as Never properly tested. (b) Compton as wave scattering: the Physicist says the wave-only reading is tested and not supported by coincidence and antibunching experiments (his tag: memory, medium, not web-checked); the Mathematician says not testable as stated. Both agree the derivation re-uses Compton's energy and momentum balance. I split it below and keep the Physicist's point with its [memory, medium] tag.

2. Corrected claim list (what the documents claim, as corrected by the reviews)

1. A local ether carries light. Light travels at c relative to a "local ether" at rest in the ECI frame near Earth (heliocentric frame between planets). Correction: first-order Sagnac and "no Earth-orbit term" are also the SR/GR results in a locally inertial frame, and Su cites Ashby; the numbers match, so the ether is an interpretation at first order. The distinguishing content is a lab-fixed second-order anisotropy. Su admits one open problem, annual stellar aberration (the interface between two local ethers is undefined).
2. Gravity is an optical index n_g = 1 + 2GM/(c^2 r). The factor 2 is postulated, equals the GR weak-field value with gamma = 1, so deflection (1.75 arcsec), GPS gravitational delay (42.2 to 62.2 ps recomputed, Su says 43-60), and Earth-Venus delay (232.5 microseconds) are GR's numbers by construction. The "19.5 microsecond disagreement" with a textbook is probably a coordinate convention term [memory, medium]. (QE revision) Su also says the redshift "can be different in some other interaction" and lists "transition-dependence" as a prediction; he gives no formula or number for it, and his own wave equation, expanded, gives a non-universal coefficient for electronic transitions (QE revision, item Q1).
3. Matter waves obey the same preferred-frame equation; mass is a frequency. A Klein-Gordon-form equation; m0 = hbar omega0/c^2 is a calibration of the potential coefficient, not a derivation; f = f0 sqrt(1 - v^2/c^2) is shown for electronic states and assumed for hyperfine and nuclear ones; the equation is Lorentz invariant in form, the frame enters by interpretation and a Galilean step, and atoms do not contract in the model. Compton scattering is re-derived with Compton's conservation relations built in; its one new item is an unsized weak line at about lambda_C/(gamma - 1).
4. Magnetism can be Galilean. An augmented-potential (Weber/Riemann-type) force law reduces to the Lorentz force for neutralised currents at low speed (the textbook magnetic limit is already Galilean covariant [memory, medium]); the Galilean invariance is by labelling every quantity with a named frame. Fresnel drag is reproduced at first order. New content: extra terms F1 (electrostatic factor 1 + v^2/2c^2, 4.6% at 25 kV), F2 (about 7e-8 V/m at 1 cm from a 10 A wire), F3 (inertia shift), no drag in a purely magnetic moving medium, Barnett induction outside or with a short rotating solenoid. (QE revision) F2 is the one term with a related experimental record: Edwards, Kenyon and Lemon (1976) reported a coil-to-shield potential quadratic in current. Su's F2 has the same form but is 24 to 320 times smaller than their signal unless an unmeasured current-crowding factor alpha is that large (item Q2).
5. Testable predictions. Corrected list: the cavity quadrupole (now tested, below); the turntable one-way link (cancels for atomic oscillators within Su's own model); the transverse radar Doppler (same as SR once evaluated at the same instant, so not a difference); -11.5 Hz at 2.3 GHz (no data can test it, Su says; the fitted oscillator bias absorbs a constant; the prediction is ill defined without an interface rule); Bragg asymmetry (one ambiguous comparison); third-order rotor anisotropy (about 1e-18, beyond any rotor); no orbital Sagnac term (SR/GR also gives none); no magnetic-medium drag; Barnett short-solenoid. (QE revision) The outline (p.8) lists 11 predictions; four had not been reviewed and are now: transition-dependent gravitational redshift (no formula; Q1), star-mass-dependent stellar downshift (no formula; Q3), yoke/electrode relative motion in magnetic deflection (same as SR at first order; Q4), stronger second-order atom-radiation effect (no equation; physics/0208083 still unread; Q5). Also added: Sagnac index-dependences (matches SR; Q6) and the CMBR second-order test (no number given; Q7).
6. Quasar redshift as gravitational redshift. A fit with an assumed density, a Newtonian potential at Phi/c^2 = 4.1, a seed value z0 = 1.956 taken from Karlsson, the cloud's potential not uniform (central z = 2.69 against 2.07 at the surface under Su's formula [calc, Physicist]), and about 8 arcsec (not 0.7 arcsec) between quasar and galaxy nucleus (0.7 arcsec is Su's own size estimate [read: deep/f5c]). (QE revision) Su's observational support is a set of ten cases quoted at second hand from other authors (item Q8); the abrupt-redshift-boundary argument is a postulate of his model, not a test.

3. Verdict per claim

Scale names as in Step 2. Parts are listed separately; "Established" for a part means the numbers match experiment and accepted theory, which for much of this project means they are SR/GR's.

Claim 1. Local ether

• ECI-frame Sagnac correction for GPS and links; no Earth-orbit term in earthbound propagation: Established (as SR/GR practice; it is not evidence for an ether).
• The ether as the explanation at first order, against SR/GR in an ECI frame: Not testable as stated (identical numbers).
• Lab-fixed second-order cavity anisotropy, about 5e-13 to 7e-13: Tested, not supported. Herrmann et al. 2009 year-mean lab-fixed amplitudes are about 1e-17 with errors of about 1e-17 (D1). One caveat: an incidental test by a sidereal-search experiment, equivalence of geometry argued, not derived by me. It was not dismissed without testing; Su's own comparison (Brillet-Hall, per Su) was one residual of half the size, which he called "too early" for a conclusion, and which is now best read as a systematic.
• Annual stellar aberration with an Earth-comoving ether: Tested, not supported (aberration is observed; Su concedes his model does not yet reproduce it [read: EPJC p.12, p.15]).

Claim 2. Gravity as optical index

• Numerical results (deflection, Shapiro delay, GPS delay, first-order redshift): Established, as GR values, reproduced by construction (Cassini: gamma = 1 + (2.1 +/- 2.3)e-5 [web: https://ideas.repec.org/a/nat/nature/v425y2003i6956d10.1038_nature01997.html]).
• "No curvature needed": Not testable as stated at the orders tested. A scalar index with no field equation covers light only; perihelion precession, frame dragging and gravitational waves are not addressed. The second-order redshift difference (f0/sqrt(1 + 2 Phi/c^2) against GR) and heliocentric-versus-barycentric frame effects are unsized: Never properly tested.

Claim 3. Matter waves, mass as a frequency

• Algebra to de Broglie, Lorentz-factor relations; matter-wave Sagnac and COW phases; Hafele-Keating, GPS clock rates, Ives-Stilwell, maser, rotor at second order (all SR numbers in ECI): Established.
• Preferred-frame reading of the same formulas at those orders: Not testable as stated.
• m0 = hbar omega0/c^2 as "origin of inertia" and the inertial-gravitational identity: Not testable as stated (calibration and postulate; circular).
• f = f0 sqrt(1 - v^2/c^2) for hyperfine and nuclear transitions: Plausible, untested as a derivation (Su assumes it; SR gives the same).
• Third-order rotor anisotropy, v0^2/c^2 terms, Bragg asymmetry: Never properly tested.
• (QE revision) Hughes-Drever isotropy: Established as a null result, not discriminating: Su's agreement comes from the symmetry of the ECI speed under Earth's rotation, "whatever the dependence of nuclear magnetic interaction on mass and speed" [read: deep/f6 p.8], not from a calculation.
• (QE revision) Gravitational redshift is the same for every transition (Su, eq. 25, f6) but not by his own wave equation: Internally inconsistent for electronic transitions (item Q1).
• Compton as wave scattering: reproduces Compton by construction; the "direct radiation" prediction is unsized, Never properly tested. The claim that wave scattering replaces photon quantisation is Tested, not supported by the electron-photon coincidence experiments (Bothe-Geiger 1925: [web: https://en.wikipedia.org/wiki/Bothe%E2%80%93Geiger_coincidence_experiment]; Compton-Simon 1925 [memory, medium; not web-checked]) and photon antibunching (Kimble, Dagenais, Mandel, Phys. Rev. Lett. 39, 691, 1977 [web: https://resolver.caltech.edu/CaltechAUTHORS:KIMprl77]; the abstract-level result is that antibunching is understood only with a quantised field), which Su does not discuss. Caveat: antibunching in atomic fluorescence tests the light field, not Compton scattering by free electrons; the Compton-specific evidence is Bothe-Geiger and Compton-Simon.

Claim 4. Galilean magnetism

• Low-speed neutral-matrix reduction to F = q(E + v x B); Fresnel drag at first order: Established (same predictions as the standard law).
• "The Lorentz law holds only under hidden restrictions": Not testable as stated (true of any approximation).
• Extra terms F1-F3, dependence on the zero of Phi, Barnett outside/short solenoid, no drag in a magnetic medium: Never properly tested. Su's high-speed form is "tentative" [read: f2a]; Bertozzi 1964 (kinetic energy of electrons up to about 15 MeV agrees with qV and relativistic mechanics [web: https://en.wikipedia.org/wiki/Tests_of_relativistic_energy_and_momentum, as quoted by Mathematician and Contrarian]) would be in tension with a literal high-speed F1, but nobody in the pipeline, myself included, has worked the comparison through under Su's full law. I do not upgrade the verdict. Kennard (1912, 1917), as Su describes it, is single-source (Su) and ambiguous; unread. (QE revision) F2 specifically: Never properly tested, retained after weighing Edwards, Kenyon and Lemon 1976 (item Q2): it is neither a neutral test nor a refutation.

Claim 5. Predictions that differ from relativity

• Lab-fixed cavity quadrupole: Tested, not supported (claim 1).
• One-way turntable link with atomic oscillators: Internally inconsistent (Su's own eq. 23 cancels the effect eq. 33 predicts; D2). With non-atomic oscillators: Never properly tested.
• Transverse radar Doppler as a difference from relativity: Refuted by calculation (it is the standard formula; D3). As an observed quantity it is tested every day as standard Doppler tracking.
• -11.5 Hz interplanetary offset: Never properly tested (existing data cannot test it, by Su's account; the SR/GR barycentric-to-geocentric scaling of about 1.48e-8 contains it [memory, high]; ill defined without an interface rule).
• Bragg asymmetry about 1 degree: Never properly tested (D5).
• Third-order rotor, Barnett short solenoid, no drag in magnetic medium: Never properly tested (the rotor is beyond reach).
• No orbital Sagnac term: Established, not discriminating.
• (QE revision) Sagnac index-dependences (n^2 loop-rotation term, n-independent Earth-rotation term): Established, not discriminating (Post 1967 and first-order SR in the ECI frame give the same coefficients [read: deep/f3a; memory, high]).
• (QE revision) Transition-dependent gravitational redshift: as stated, Not testable as stated; read from Su's own equations, Internally inconsistent for electronic transitions (Q1). Stellar downshift by mass: Not testable as stated (Q3). Yoke/electrode relative motion: Not testable as stated as a difference from SR (Q4). Atom-radiation second-order effect: Not testable as stated (Q5). CMBR second-order test: Not testable as stated as written (Q7).
• Marsen's low-Earth-orbit and solar-orbit Michelson-Morley proposal (advocate): the prediction scales as (v/c)^2/2, 3e-10 at 7.6 km/s [calc]; it is the same physics as the ground cavity term, which is now bounded 4e4 times lower, so for Su's model as written it adds little. I found OPTIS (2001) only as a proposal [web: https://arxiv.org/pdf/gr-qc/0104067]; whether any satellite Michelson-Morley test flew was not confirmed by this check [memory, medium: none].

Claim 6. Quasar redshifts

• Weak-field gravitational redshift: Established.
• Su's redshift formula applied at Phi/c^2 = 4.1, and the cloud as stated: Internally inconsistent (outside its own weak-field domain; D4). Unanswered: free-fall time 272 years, escape speed 2.9 c, 1.3e16 solar masses, central redshift 2.69 against 2.07 at the surface, no support from pressure or rotation.
• The 2.07 against 2.114: a fit, not a prediction (free density; "slight adjustment" in size or density [read: f5c]).
• Karlsson periodicity, the premise: Tested, not supported by large surveys (Hawkins, Maddox, Merrifield on 1647 galaxy-quasar pairs [web: https://arxiv.org/pdf/astro-ph/0208117]; Tang and Zhang 2005 on 2dF and SDSS samples [web: https://en.wikipedia.org/wiki/Redshift_quantization]). Burbidge and Napier (2001) report the opposite on other samples [web: same page]; the proponents attack the selection of the large samples; contested.
• Physical association and ejection (Arp, Burbidge): not settled in this review; interest is recorded, no verdict. (QE revision) The cases Su cites are listed in Q8 with sources; the abrupt-boundary argument is Not testable as stated (Q8).
• Su's halving series giving 0.22, 0.14, 0.09: Never properly tested (no cited data).

4. What would settle it

5. Absence of evidence: where the record goes dark

What is dark. (i) Su's publications stop in 2006 (Crossref, arXiv, Wayback [read: sites.md]). (ii) One advocate source (Marsen 2021) says Su "passed away in approximately 2007" and that his website was removed; the Natural Philosophy Wiki repeats it, citing Marsen; no obituary or university notice was found. (iii) The book appeared from NTHU Press in March 2008 (announced in NTHU alumni e-news, 9 May 2008, which was not read), with an ISBN registered in 2005. (iv) No published Comment or Reply to the EPJC or EPL papers was found (one Google query; Scholar's cited-by lists blocked). (v) The live site now redirects to the department home page, yet Wayback holds captures of it in 2017 [read: research/wayback-machine.md], so the site was served for at least ten years after the stated death date.

Competing explanations for the end of the record. - Death around 2007 (Marsen). For: a clean stop in 2006, a book announced in 2002 and appearing in 2008 with an ISBN of 2005 (consistent with a posthumous edition), Marsen's "late Prof." and Beckmann also described as "late". Against: single advocate source; I know of no primary record; an NTHU site live in 2017 is not what one expects from an abandoned address, though departments do keep legacy pages up. Decides: NTHU or alumni notice, obituaries. - Illness or retirement. For: the same stop. Against: none specific; no evidence either way. Decides: NTHU personnel records. - Withdrawal after limited reception. For: no Comment or Reply found, so no engagement to answer. Against: Su kept writing through 2006 and compiled a book; the cavity result later showing his main prediction null cannot have caused this, because it appeared in 2009 and later. Decides: referee correspondence (unobtainable), his later activity. - Suppression. For: nothing in the record. Against: the core papers appeared in refereed journals (Eur. Phys. J. C 21, 701; Europhys. Lett. 56, 170; Eur. Phys. J. B 24, 231; J. Electromagn. Waves Applic.; Found. Phys. Lett. 2006 [read: facts.json; web: Crossref per research note]) and he held a faculty post from 1985. The "blocked from publication" story does not apply to what is in the record. My search for an obituary used the guess 蘇清泉 as his Chinese name (the Investigator never established the name) and returned nothing relevant; that check is inconclusive and the name is unverified [web: search, 2026-10-06].

Silence about rebuttal. Absent Comments and Replies are consistent with three explanations, none excluded: (1) the papers reproduce SR/GR numbers at the orders checked (Su's own text says so) so a referee or reader had nothing to rebut; (2) the end of the author's activity; (3) low engagement with the venues. Silence is not evidence of validity. Explanation (1) is supported by the reviews in this file; (2) and (3) have no evidence either way. Cited-by lists (46 and 32 on Scholar, about 20 and 10 on OpenAlex) are the deciding record and were not read.

Silence about the Brillet-Hall residual. If a lab-fixed term of 3.9e-13 had been a real Earth-rotation signal, later turntable cavity experiments would have seen it; the 2009 year-mean is about 1e-17. The absence of a Brillet-Hall follow-up on that term is explained by the term being a systematic, not by an omission by the community. Brillet and Hall's own account of why they called it spurious remains unread.

6. Open questions

1. Is the Herrmann bound equivalent to Su's prediction once the two-cavity geometry is derived? (If yes, the ground-fixed strong form is closed.)
2. Does any version of a nested-ether model reproduce annual aberration (an interface rule)? Nobody has written one in the documents reviewed.
3. Do the accelerator and electron-energy data (Bertozzi and later) rule out Su's F1 and F3 at high speed, and what does Su's full law predict there?
4. Do the Karlsson peaks survive in a sample selected without proximity to bright galaxies, and are quasars physically associated with low-redshift galaxies? (Contested; outside what Su's mechanism can decide.)
5. What happened to Su after 2006, and who saw the 2008 book through publication?
6. Beckmann's 1986 priority: did Su know the work? Su's EPJC text does not cite Beckmann [read: grep]. Beckmann's treatment of aberration is unchecked [memory, low].
7. Whether the weak ether-free parts (ECI Sagnac, Fresnel drag in polarisation-current form) are a useful teaching route to SR; the reviewers agree the algebra is correct.

7. Web checks done (arbiter budget 8, used 4)

• Herrmann et al. 2009: https://arxiv.org/pdf/1002.1284 (read locally after download; year-mean C0, S0 about 1e-17; retagged from Mathematician's "provisional").
• Brillet and Hall 1979: https://vivo.colorado.edu/display/pubid_108412 (abstract only: (1.5 +/- 2.5)e-15; no data on the 2-omega term obtainable; check failed for that item).
• NTHU obituary search (Chinese): no relevant result; inconclusive.
• OPTIS: https://arxiv.org/pdf/gr-qc/0104067 (proposal only; no flight or cancellation status found). Not checked (kept as [memory]): the Compton coincidence experiments, Bertozzi as an original, Le Bellac-Levy-Leblond, Mandelberg-Witten precision, Beckmann on aberration.

8. Final verdict summary

1. Local ether: ECI Sagnac and no orbital term, Established (not evidence for an ether); ether as explanation, Not testable as stated; lab-fixed 2nd-order cavity anisotropy, Tested, not supported (Herrmann 2009); annual aberration, Tested, not supported.
2. Gravity as index: numbers Established (GR's own); replacement of curvature, Not testable as stated.
3. Matter waves: standard relations Established; Hughes-Drever Established, not discriminating; gravitational redshift universality across transitions Internally inconsistent (QE revision); mass-as-frequency Not testable as stated; hyperfine extension Plausible, untested; Compton wave-only reading Tested, not supported (memory, medium); third-order effects and Bragg asymmetry Never properly tested.
4. Galilean magnetism: low-speed reduction Established; extra terms, Barnett, no-drag Never properly tested (tension with high-energy electron data suspected, not computed); F2 stays Never properly tested after Edwards et al. 1976 (QE revision).
5. Differing predictions: cavity Tested, not supported; one-way link Internally inconsistent (atomic clocks); radar Doppler Refuted by calculation as a difference; the rest Never properly tested; orbital-Sagnac null Established, not discriminating.
6. Quasars: weak-field redshift Established; the formula at Phi/c^2 = 4.1 Internally inconsistent; periodicity Tested, not supported (contested); halving series Never properly tested.

QE revision (2026-10-06, after qe.md: verdict MINOR)

Inputs added: six deep reads (outline pp.4-9; f6 pp.4-6; f5a pp.4-7; f2a pp.8-11 with Edwards, Kenyon and Lemon 1976; f5c pp.5-7; EPJC pp.11-12) and three web checks. Not done in this pass: physics/0208083 and f2d (Barnett) were not read in full, and Wu and Tsao 2019 and Persson 2021 were not weighed here (the primer should name them; neither bears on a verdict I could check). No headline verdict changes. One new verdict (Q1), several unreviewed predictions now classed, and the F2 verdict was reconsidered and kept.

Q1. Transition-dependent gravitational redshift (outline p.5 and p.8). On p.5 Su calls it a "discrepancy" with GR: the shift "originates from a quantum nature of the matter wave bound in the involved atom"; on p.8 the same point is listed as a prediction [read: deep/outline]. His text gives one formula for every clock, f = f0 sqrt(1 - v^2/c^2)(1 - Phi_g/c^2) (f6 eq. 25), derived only for spin-flip (hyperfine) transitions with identical spatial states, with the spin interaction assumed to act like q Phi, which he admits his equation cannot treat. He says other dependences "are not precluded" and gives no form and no number [read: deep/f6]. The paper cites no clock-comparison test of local position invariance. As stated, then: Not testable as stated. The new point is that his own equations do not give universality. His f5a eq. 3 (n_g = 1 + 2 Phi_g/c^2) expands to omega = omega0 n^(-1/2)(1 + P) - (c^2/2 omega0) n^(-3/2): the potential-energy part is redshifted by (1 - g) and the kinetic part by (1 - 3g), g = Phi_g/c^2 [calc: I re-did this two-line expansion; it agrees with the f6 deep reader's sympy result]. For a Coulomb-bound electronic transition the virial theorem (delta V = 2 delta E, delta T = -delta E [memory, high]) gives 2(1 - g) - (1 - 3g) = 1 + g, a blueshift factor against the hyperfine (1 - g): a difference of about 2 g = 1.4e-9 at Earth's surface (g = 6.96e-10). f5c eq. 6 instead puts one common factor n^(-1/2) on all terms, which would give universality; the two papers disagree and Su does not reconcile them. Verdict for the claim that the redshift is (1 - Phi_g/c^2) for every transition: Internally inconsistent (his eq. 3 against his eq. 25 and f5c eq. 7), for electronic transitions; for hyperfine clocks (Cs, H maser, as used in GPS and Hafele-Keating) the formula is the GR value and Established. Experimental bearing: Ashby et al. compared H masers with Cs fountains over a year of changing solar potential and bound |alpha_H - alpha_Cs| < 2.1e-5 [web: https://www.ncbi.nlm.nih.gov/pubmed/17359010, Phys. Rev. Lett. 98, 070802, as summarised by the search result]. Both are hyperfine clocks, so that bound does not test electronic against hyperfine. Optical-clock redshift comparisons would; their limits are [memory, medium; not web-checked, budget]. I do not give "Tested, not supported", because Su never committed to a species dependence and nobody has run his coefficient. Deciding calculation: derive the redshift of an electronic transition from f5a eq. 3 without dropping the 1/n^2 on the Laplacian, then compare with an optical-clock height or solar-potential comparison.

Q2. F2 against Edwards, Kenyon and Lemon (Utah State), Phys. Rev. D 14, 922 (1976). [read: deep/f2a section 13, from the scanned paper: web https://isidore.co/misc/Physics%20papers%20and%20books/Zotero/storage/C6B7H3S5/PhysRevD.14.html]. They reported a potential between a superconducting Nb-Ti coil and a shield, quadratic in current (fit exponent n = 2.02 +/- 0.05), of electron sign, surviving a bifilar winding and a Faraday cage, signals 3 to 13 mV in earlier runs and tens of mV (74 and 83 mV in single runs) later; they conclude "The signal appears to be a real field effect and is as yet unexplained" and exclude self-Hall, thermoelectric, flux-flow and other explanations they considered. Su's F2 (eq. 28), evaluated for their coil with uniform current: about 35 microvolts at 10 A and 89 microvolts at 16 A [calc, deep/f2a], the same form and the same sign as what they saw, but 24 to 320 times smaller unless the current-crowding factor alpha is that large; alpha was not measured. Which carriers (all conduction electrons or only superconducting ones) enter the formula is also open. f2a does not mention the experiment. Later record: Lemon, Edwards and Kenyon, Phys. Lett. A 162, 105 (1992) is cited as reporting an observed potential difference [web: abstract of arXiv:1009.0474, via the deep read]; a null test by Bartlett and Maglic (Rev. Sci. Instrum. 61, 1990) is recalled [memory, low-medium] and my search did not find it, so I cannot state that a null replication exists. Result: the strongest honest reading is that a published, unexplained anomalous signal has Su's F2 form but not his size, and no neutral, controlled test of F2 exists. Verdict: Never properly tested (kept). This is single-experiment evidence from one group, stated as such; it is not dismissed, and it is not support for the F2 coefficient. Settling it: measure alpha (current distribution) in a superconducting coil, repeat with an ordinary copper coil at Su's predicted 0.1 microvolt level, and read the 1992 and 1990 papers.

Q3. Star-mass-dependent stellar downshift (outline p.8). One clause, no formula, star or number [read: deep/outline]. In the weak field the only formula is Phi_g/c^2 as in GR (white-dwarf redshifts are the GR surface value; [memory, medium]); the strong-field form 1 + z = sqrt(1 + 2 Phi/c^2) applies only where Q8's quasar problem (D4) applies. Not testable as stated; numerically equal to GR's weak-field value.

Q4. Yoke/electrode relative motion in magnetic deflection (outline p.8). f2a does not size it. With u the electron speed relative to the electrodes and w the electrode speed relative to the yoke, Su's law gives a force q(u + w) x B, a first-order change w/u that special relativity also gives (the yoke field transforms) [calc and memory, high, deep/f2a section 13a]. The two differ only by the velocity-addition rule, about u w/c^2: 3e-8 for w = 30 m/s up to 4.7e-7 at 465 m/s for a 25 kV tube, against a spot-position resolution near 1e-3 [memory, medium]. Not testable as stated as a difference from SR. The separate F1 effect (about 5% on deflection at 25 kV, 0.317 against 0.302 for beta in the constant-mass estimate) is unchanged: Never properly tested, with the Bertozzi tension flagged in claim 4.

Q5. Stronger second-order atom-radiation effect (outline p.8). No equation, size or experiment in the outline; the candidate source physics/0208083 (outline ref. 8) is unread [read: deep/outline]. Not testable as stated; source gap.

Q6 and Q7. Sagnac index-dependences: Established, not discriminating (claim 5 list). CMBR (EPJC pp.11-12): Su treats the CMBR sources as an enclosing local ether, concedes the dipole (about 300 km/s, v/c about 1e-3) cannot discriminate his model, and says "a crucial test" needs the second-order anisotropy, giving no amplitude or method. Order of magnitude (v/c)^2 x 2.73 K = about 3 microkelvin [calc]. SR also has a quadrupole from the Sun's motion, and the paper does not separate the two [memory, medium]. Not testable as stated. His cited CMBR dipole measurements are 1977 to 1985 [read: deep/EPJC addendum].

Q8. Quasar cases and the abrupt boundary (retention). Su cites, at second hand [read: deep/f5c, reference details not web-verified]: NGC 7319 quasar 8 arcsec from the nucleus with a V-shaped filament (Galianni et al. 2005); Arp 220 (z = 0.018) with companions at z about 0.09 about 2 arcmin away and an H I stream ending on them (Arp 2001); NGC 3628 (z = 0.0028) with a filament to quasars at z = 0.995 and 2.15 (Arp et al. 2002); the NEQ3 filament and three compact objects at z = 0.1935, 0.1939, 0.2229 (Gutierrez and Lopez-Corredoira 2004); clusters and four quasars aligned across M82 (Arp and Russell 2001); the NGC 383 group (multi-direction ejection, which Su notes sits uneasily with axis ejection); and two abrupt boundaries: MCG 7-25-46 (z = 0.003 against 0.098, H-alpha emission ending exactly where the other begins, Lopez-Corredoira and Gutierrez 2005) and NGC 7320 (Gutierrez et al. 2002), where Su records that the authors read it "instead as evidence against the connection ... and for the projection effect". Stakes: Arp, Burbidge, Lopez-Corredoira and Gutierrez are proponents of association; Su adopts their reports for his own model. Su did no data analysis of his own on these objects. His argument that a sharp boundary shows a single exclusive potential is a postulate of the model; a sharp boundary is also what unrelated galaxies at different distances would show, and the paper offers no test that separates the two. Not testable as stated. Whether the objects are physically associated is not assessed here and stays open (claim 6).

Verdict changes. New or changed: Q1 gravitational-redshift universality for electronic transitions, Internally inconsistent (hyperfine clocks, Established); Hughes-Drever and Sagnac index-dependences added as Established, not discriminating; Q3, Q4, Q5, Q7 classed Not testable as stated; Q8 classed Not testable as stated. Unchanged: F2 Never properly tested; Compton wave-only reading Tested, not supported (now partly web-checked; Compton-Simon unchecked).

Web checks used in this revision (3; total 7 of 8): Kimble et al. and Bothe-Geiger (found, tagged above); local position invariance clock limits (Ashby et al. H-Cs bound; optical-clock limits not found in the summary); Bartlett and Maglic 1990 (not found).