Parameter Derivations

Sessions #64-65

Sensitivity ordering, measured 2026-09-03 (explorer, L2 field equation on 153 SPARC galaxies): this page audits hardest the parameters the model is least sensitive to. Moving γ by 40× and A by 10⁹× at a working floor changes χ²/N by only 1.07–2.7×; moving the floor(item 8) at fixed (γ, A) changes it by up to 6.35×10⁴. The Fisher correlation of (ln γ, ln A) is +1.000000 (κ = 3.6×10⁸): the two are one parameter in the galaxy sector, and neither is identifiable there. The CLT point below stands, but the practical verdict on γ is not “wrong sign” so much as unmeasurable where it was fitted. Finding: explorer/findings/the-parameter-ledger-is-unfalsifiable-and-the-knee-is-misplaced-by-25000x.md.

The 1/√N_corr scaling borrows from central-limit-theorem (CLT) statistics. The factor of 2 is motivated by phase-space dimensionality arguments (6D to 3 effective) but is not rigorously derived — integrating out momenta introduces temperature- and mass-dependent factors, not a clean factor of 2. Best understood as a physically motivated ansatz.

Internal inconsistency (2026-05-20): The CLT's 1/√N scaling governs the standard error of the mean for iid (independent, identically distributed) variables. But N_corr is by construction the count of correlated degrees of freedom — exactly the regime where the iid hypothesis fails and 1/√N does not apply. Invoking CLT for correlated N_corr is self-contradictory. The scaling is borrowed by analogy, not derived from the CLT. It is an ansatz with a fitted prefactor.

Sign inversion (2026-06-06 / 2026-06-29): Beyond being underived, the N_corr-dependence is sign-inverted relative to the mean-field physics it borrows. In any mean-field model (Ising, Curie–Weiss), increasing the number of correlated units makes the transition sharper: in Curie–Weiss the finite-size rounding width shrinks as N^−1/2, so a mean-field sharpness parameter would grow as √N, the reciprocal of the ansatz. That holds when N is the total size of an all-to-all model; N_corr is closer to the count in one coherence volume, where the Ginzburg criterion applies and the scaling is steeper (for BCS the fluctuation window goes roughly as (T_c/E_F)⁴; textbook, recalled, not re-derived), and that sharpness is in temperature while C's axis is density. The direction of the audit survives; “exact” did not. The deeper defect is the family: no γ makes this curve sharp. Here, γ = 2/√N_corr decreases with N_corr, giving the most-correlated systems (BCS, BEC, N_corr ≈ 10⁶–10⁹, depending on material and on what is counted —range widened 2026-09-10 and again 2026-10-16 to match the calculator: this read “≈ 10⁶” while the γ Calculator uses 10⁷ for BCS in the same argument, a factor of 10 on the same named system, caught by a graduate-physics reader. Both are defensible — a BCS pair's coherence volume holds ~10⁶–10⁹ electrons depending on whether you count ξ³n or pairs — which is itself the point this page makes: N_corr is a counting convention, not a measured invariant, so a factor of 10 in it is not an error so much as evidence the quantity is underdetermined) the flattest transitions. The γ-calculator (try γ = 10⁻³) shows this directly: the "sharpest" transition (γ = 2) belongs to the single uncorrelated particle, while BCS superconductors — among the most collectively ordered systems known — sit near C ≈ 0, the flattest possible curve. “Underived ansatz” understates the problem: the relation is anti-correlated with the physical intuition that motivates it. Any first-principles derivation must recover γ that increases with N_corrto match mean-field expectations — making the formula a falsified placeholder, not a gap to fill.

Session #66

tanh is a phenomenological choice from the sigmoid/compander family (μ-law, Hill, logistic, erf). Other sigmoids share the same qualitative properties and would produce indistinguishable physics near γ ≈ 1. The fractal coherence bridge failure (0/7 boundaries on 36 tests) is consistent with tanh being a generic sigmoid here, not a uniquely derived form. Not Landau: a saturating compander with argument ≥ 0 has no critical point, no diverging correlation length, and no critical exponents — nothing to put it in a universality class. The Landau framing was retired at the landing page (compander family); this page now matches.

The log argument, ln(1 + ρ/ρ_crit): chosen, no derivation. The “+1” keeps C finite and zero at ρ = 0. With it, tanh(γ ln(1+x)) = ((1+x)^2γ − 1)/((1+x)^2γ + 1) exactly, so the log turns the tanh into a power-law compander: C ≈ γx at small x, and 1 − C ≈ 2(1+x)^−2γat large x. The power law and its index come from this choice, not from a principle (MOND derives the corresponding deep-limit slope from scale invariance; this form inherits it from the regulator). No archive document motivates the log itself.

Note on the Ising analogy: tanh arises in mean-field Ising models as m = tanh(βJzm) — but that tanh comes from the self-consistency equationm = tanh(βJz·m), where m feeds back into itself. C(ρ) has no such self-consistency loop: it is evaluated directly at the input ρ with no fixed-point iteration. The Ising tanh is derived; this tanh is chosen. These are structurally different justifications.

Closed form: with x = ρ/ρ_crit, tanh(γ ln(1+x)) = [(1+x)^2γ − 1] / [(1+x)^2γ + 1] exactly (checked to 10⁻¹²). Limits: C ≈ γx for small x, and 1 − C ≈ 2x^−2γ for large x. At γ = ½ it is x/(x+2), MOND's simple μ with the scale doubled. It is a smooth sigmoid in (1+x)^2γ, analytic for every x > −1: a crossover (a compander), with no critical point anywhere.

Sessions #53, #66 — decisive test run 2026-06-07; badge split 2026-09-05

Lead with the large error, not the small one (explorer 2026-09-03). The 600× formula-vs-calibrated mismatch below is real but is the smaller discrepancy, and it moves ρ_crit toward the data. Pure density arithmetic, no fit: the calibrated A = 0.029 puts ρ_crit at 6.5×10² M_☉/pc³ (4.4×10⁻²⁰ g cm⁻³), while the SPARC median midplane density is 2.6×10⁻² M_☉/pc³ (1.8×10⁻²⁴) — the knee sits 2.5×10⁴ above every density the data sample. (Refracted Gravity, whose field equation this sector coincides with, puts its knee within 1.8× of the data.) Units, which this card never stated: A is in M_☉ pc⁻³ (km/s)⁻²; the “600×” reproduces only in those units (a simulated-visitor grad-student persona recomputed 635×). Why the normalization row is unrunnable, not merely negative: the 600× is repairable for free with N_corr = 6.7×10⁶, inside the 1–10⁷ range the γ Calculator admits — the third mutually inconsistent N_corr this site's own relations produce (1, 16.7, 6.7×10⁶). A row that any N_corr can rescue has no refutation criterion. The scaling (ρ_crit ∝ V², excluded at ~11σ, 2026-08-27) keeps its Audited-Negative.

β_J = λ_Jeans / R_half is the dimensionless Jeans-length-to-galaxy-size ratio (Session 53). Empirically β_J ≈ 1.1 ± 0.2 across SPARC galaxies.

Symbol note (2026-04-24 correction): β_J is the Jeans ratio λ_Jeans/R_half— an O(1) structural ratio, not the electromagnetic fine-structure constant α_em ≈ 1/137. The formula was previously written with α, which invited that misread. With α_em² ≈ 5×10⁻⁵, the formula yields A ≈ 550 (km/s)⁻² — 20,000× too large. The formula only closes at 5% with β_J = O(1). No electromagnetic coupling is implied.

The tightest constraint on this knee is the solar neighbourhood, and until 2026-09-18 this page did not mention it. A researcher visitor persona noted that the canonical ρ_crit page — the one that tabulates every sector's knee — named neither the Oort limit nor the local stellar density, while the single sharpest zero-cost exclusion on the site sat inside a card marked “unrun”. The exclusion, propagated here from TEST-02: at this row's published calibration ρ_crit = 0.029·V_flat² (= 1.52×10³ M_☉/pc³ for the Milky Way at V_flat = 229 km/s), the local density sits ~10⁴ below the knee, so C ≈ γx ≪ 1 and the division wiring gives a boost of about 3.5×10⁴ in g in the solar neighbourhood — excluded by the Oort limit and planetary ephemerides by orders of magnitude. The knee window that would instead reproduce the 0.05–0.4% wide-binary signal is ρ_crit ∈ [3.8×10⁻⁵, 3.2×10⁻⁴] M_☉/pc³ at γ = 0.489 — seven orders below this row's value and disjoint from every knee the galaxy sector uses.
Scope, unchanged: this is not a seventh refutation and the count stays 6. It is the same statement Key Claims already publishes as “rotation curves under this reading do not fail to flatten — they blow up”, evaluated where ρ is well measured. What was missing was not the result but its presence on the page a reader consults about this parameter. The visitor's recommendation to promote it to an executed refutation is declined for that reason, and the reason is recorded here rather than left implicit.

Sessions #87-88

The MOND acceleration scale is posited as a₀ = cH₀/(2π) by dimensional analysis: cH₀ is the only acceleration cosmology supplies. It does not follow from the coherence function, which contains no a₀. Base rate: a 43-member grammar of simple prefactors on cH₀ matches 89% of targets in [0.1, 0.3]×cH₀ to ±10%, and the archive itself used five such routes (cH₀/2π, /6, /2, Ω_m-power forms) and three H₀ values, so a match at this level cannot be evidence (explorer, 2026-10-14). The testable part is the slope of a₀(z), not the level. Milgrom's observed: 1.20×10⁻¹⁰ m/s² — the formula sits 13% below it. H₀ bookkeeping corrected 2026-07-22 (caught by a visitor physics persona): this card previously quoted a₀ ≈ 1.08×10⁻¹⁰(“within 10%”), which is the H₀ = 70 value — algebraically incompatible with item 5's Σ₀ = 119 at H₀ = 67.4, since Σ₀ = a₀/(2πG) exactly (one number, not two). At the site-standard H₀ = 67.4: a₀ = 1.04×10⁻¹⁰(13% below Milgrom) with Σ₀ = 119 — the consistent pair, now quoted site-wide. Convention caveat: at γ = ½ the compander collapses to x/(x+2) = μ_simple(x/2) — MOND's simple function with a₀ → 2a₀. Because ρ_crit (or a₀) is free in every fit this factor is silently absorbed, but it means the framework's a₀ is only defined up to a convention-dependent factor of 2 (the archive records a₀ = g^†/2 as exactly this “factor-2 anomaly,” 2026-08-24). A 13% comparison to Milgrom is not meaningful at that level; treat it as “same order, same dimensional origin,” not as agreement or disagreement.

This dimensional relation a₀ ∼ cH₀ has been noted since Milgrom (1983). Other authors reach the same dimensional origin with different prefactors, not a shared one: McCulloch 2007 has 0.99 cH, Verlinde 2016 cH₀/6, Smolin 2017 a_Λ/8.3 (table). The quantities c and H₀ are dimensionally sufficient to produce an acceleration — cH₀ is not a Synchronism-specific derivation. Classified as dimensional analysis / reparametrization on the honest assessment page.

The epoch fork is closed; branch (A) is forced, and every high-z test of it so far is non-discriminating (current state 2026-09-19). a₀(z) = cH(z)/2π has been checked against direct RAR intercepts (estimator-dominated; the anchor does not enter the slope; below), high-z TFR zero points, and 41 published f_DM(R_e) values; the f_DM route splits by fitting method (Price MCMC vs Genzel least squares), so the level is method-dominated. The level-free test is the trend ratio k(z_hi)/k(z_lo). Run 2026-10-06 on RC100 (z 0.6–2.5, pre-registered before the table was read): non-discriminating, with ratio 1.11 (galaxy-bootstrap error ±0.36 on ln of the ratio) against branch (A)'s 1.74 and constant-a₀ MOND's 1; all three are plain ratios. Post-hoc only: fitting a₀ ∝ E(z)ⁿ gives n = 0.0, with branch (A) at the ~2σ edge of the galaxy-bootstrap 95% interval (±1.0; the formal ±0.2 understates it). Deciding the row needs N ≳ 170 galaxies and a measured method floor. Del Popolo & Chan 2024 (arXiv:2405.01841) used the same table, with a cruder estimator, and also saw no growth. So the three direct datasets now read three ways: Ciocan rising, Del Popolo flat-to-falling, RC100 flat (explorer finding 2026-10-06). TFR zero points cannot constrain a₀(z) at all (Milgrom 2017 §4): the modelled branch-(A) lever is −0.125 dex at z ≈ 0.9 and −0.243 dex at z ≈ 2.3, against a 0.25–0.35 dex spread between two surveys measuring the same relation at z ≈ 0.9. Revision trail: this block opened 2026-08-01 as “tested and disfavored” after a 2026-07-30 explorer execution; revised to non-discriminating 2026-08-04 (anchor dependence, below) and re-confirmed 2026-09-19 on the TFR and f_DM handles. This relation is written in H₀ (today's expansion rate); the site previously said nothing about H(z), and two prior citations were doing the wrong job. Citation correction: arXiv:1703.06110 is not Milgrom proposing a₀ ~ cH/2π — it is the paper that constrains a₀(z) from six high-z discs (≈ 4a₀ at z ~ 2 is in tension; H(z) scaling, 3.0–3.5×, is said to be something the data “may help constrain,” not claimed excluded); on 41 discs the same test splits by fitting method (explorer 2026-09-19). Milgrom's actual proposals are arXiv:0801.3133 and ApJ 698, 1630 (2009), both citedinside 1703.06110 as references. Branch (B) (“present-epoch coincidence, no H(z) statement”) is not actually available: it requires a₀ to be fundamental, which is the MOND position this framework explicitly rejects elsewhere (a₀ is described as an emergent scale on this page and on Relation to MOND). An emergent a₀ tied to the coherence transition forces branch (A): a₀(z) = cH(z)/2π.

Branch (A), tested against direct RAR data: Ciocan et al. 2026 (MUSE-DARK III, A&A 709, L16; arXiv:2604.22613) fit the RAR directly in 79 star-forming galaxies, 0.33 < z < 1.44, and report a₀(z) = a₀(0) + a₁z with a₁ = 1.59 (+0.11/−0.10, a 95% CI; 1σ ±0.05) ×10⁻¹⁰, stating plainly:“our measured a₀(z) is faster than that of H(z).” Branch (A) is a zero-parameter prediction, a₀(z)/a₀(0) = E(z).

Correction (2026-08-04): the level deviation depends on the anchor, and the sign flips with the anchor. Branch (A) is a ratio prediction — converting it to a number at z~1 requires picking a₀(0), and four published values disagree by 69%. Revised 2026-10-12: the 2026-08-04 wording named the anchor as the row's limiting systematic. That is true of this level table only; the prediction's own statistic is the slope (table after this one), and the anchor does not enter the slope. Ciocan's z~1 value is 2.38 (+0.12/−0.10), a 95% CI (1σ ≈ 0.06):

The 2026-08-01 shipped figure (2.3–5.9σ low) used only the three low-side anchors; the fourth — the same McGaugh+2016 value the site uses as canonical a₀ elsewhere — is one sentence away in Ciocan's own paper and reverses the verdict to consistent. Signal (79% predicted growth to z~1) vs. systematic (69% spread among a₀(0) determinations) is signal/systematic ≈ 1.15 — this places the level comparison in the “untestable with foreseeable data” category, not “disfavoured.” (Corrected 2026-10-12: Ciocan's errors are 95% CIs; earlier σ's divided them as 1σ. That was true of the sentences below; it was not true of this table, which the explorer's script had already computed with 1σ = 0.051 (low side) / 0.061 (high side). The 10-12 edit doubled the first two rows to “~19σ” and “~18σ”, converting twice; restored 2026-10-19 to 9.4σ (anchor ±0.02 carried) and 9.8σ, after a simulated-visitor graduate persona recomputed them. The rows' branch (A) column de-evolves each anchor from its own sample redshift before multiplying by E(1) = 1.79, which is why the MIGHTEE-HI row reads 2.91, not 1.69 × 1.79 = 3.03. The anchor-carried rows (+0.5σ, −2.3σ) barely move. The signal/systematic sentence is a level statement; on the slope, below, the limit is the paper's own estimator spread, a methods question per-galaxy data would settle.)

The slope, level-free (added 2026-10-12; explorer 2026-10-10, pre-registered): fit a₀ ∝ E(z)ⁿ with the level free. Branch (A) is n = 1; constant a₀ is n = 0. The anchor does not enter this statistic (adding the SPARC anchor moves n by 0.05).

With the normalisation free, constant a₀ (n = 0) is excluded by the bins at about 5σ (bars pixel-read from Ciocan's Fig. 3, which prints no CI level, and read as 1σ; about 10σ if they are 95%). That is Ciocan's own headline, an apparent a₀ rising with z. It is not an in-house test of MOND, and it shares the degeneracy with a redshift drift in stellar M/L, gas fraction or pressure-support correction. n = 1 lies between the paper's two estimators, which disagree with each other at 3σ (the bins exclude n = 1 at 2.1σ; the global line sits above it). ΛCDM+baryons sits at n ≈ 1 too, so nothing here selects the framework. The row is estimator-dominated; the anchor does not enter the slope. RC100's slope error (±0.5) is too wide to test either; it is 1.4σ from Ciocan's bins. One degeneracy sits under every number on this row: at high z, “a₀(z) rises” and “the stellar M/L (Υ), gas fraction or pressure-support correction drifts with z” predict the same shift in the RAR intercept. “Faster than H(z)” is a statement about the global line's extrapolated z=0 intercept; within the data the line is n = 1.15, and the bins are slower than H(z).

ΛCDM predicts the same evolution, and the functional form is prior art. Mayer, Teklu, Dolag & Remus 2023 (Magneticum simulations, ΛCDM+baryons, no MOND, no fundamental a₀; MNRAS 518, 257, arXiv:2206.04333) fit a₀ growing by a factor ≈3 from z=0 to z=2 — branch (A) predicts E(2) = 3.03; in the slope convention above that factor is n ≈ 0.99, so on the slope the degeneracy is exact, not approximate. Their eq. (13) is branch (A), written down and tested inside a ΛCDM paper in 2022, where it is reported to fail to describe the simulated trend precisely. No outcome of the Ciocan measurement selects Synchronism over ΛCDM+baryons: both predict the RAR-fitted a₀ grows with epoch, by a similar factor, for unrelated reasons (galaxy-assembly physics vs. an emergent coherence scale).

Milgrom's coincidence has a second, unstated face: 2πa₀ ≈ cH₀ ≈ c²(Λ/3)^1/2 — the site has cited only the H₀ half. The Λ half predicts zero epoch evolution, since Λ is constant, and fares worse against Ciocan than branch (A): at the Milgrom-local normalization, a constant a₀ = 1.20 against the observed 2.38 (+0.12/−0.10, 95% CI) is roughly 21σ away on the measurement error alone (1σ ≈ 0.06; the paper's own figure is 19σ), and 4.4σ carrying the anchor's ±0.26 as the table above does for branch (A), against branch (A)'s ~4σ and 0.5σ under the same two conventions. Like for like, the evolving a₀ is the closer of the two on this sample (stated plainly 2026-10-10, with the RC100 counter-sample on Relation to MOND), and the level-free slope says the same thing without an anchor: constant a₀ is n = 0, excluded by Ciocan's bins at 5σ. If the paper's global slope (n = 1.15) holds, an a₀ evolving faster than H(z) embarrasses both halves of the 1983 coincidence, not this framework specifically.

Verdict — non-discriminating (revised 2026-08-04), and the count stays at 6: this row joins every other row in the ledger, reached the same way — a priori, from already-published numbers, with no new data or computation. It is not a live tension with a date on it; it is a forced, outstanding, prospective commitment (a rival with a fundamental a₀ only makes this bet optionally) whose power to discriminate was removed by the ΛCDM+baryons degeneracy above (exact on the slope, n ≈ 0.99) — the anchor dependence cited here since 2026-08-04 limits only the level comparison, and the slope does not see it (revised 2026-10-12) — not evidence against Synchronism specifically, and not evidence for it either. Consistency with a non-discriminating prediction (the McGaugh-anchor 0.5σ row) is not evidence for the framework. This does not join the refutation tally on Honest Assessment. One further note on the 13% figure two cards above: Ciocan's fitted z=0 intercept is 1.00 ± 0.02 (1σ) = ± 0.04 (95% CI), about 2σ from this framework's 1.04 — but that intercept is the extrapolated endpoint of a linear fit over 0.33 < z < 1.44 that the authors themselves call “phenomenological…rather than physically motivated,” not a local measurement, so this is a reason to stop treating the 13% gap as settled, not a claim of agreement.

Source: Ciocan et al. 2026, MUSE-DARK III, A&A 709, L16 (arXiv:2604.22613). See also Honest Assessment for the general pattern this row is an instance of: stated conclusions get read, buried ones get re-derived wrong — two prior simulated-visitor personas (2026-07-29, 2026-07-30) independently re-derived the pre-Ciocan version of this fork before it was closed.

Session #89

Arithmetic corrected 2026-07-09 (independently caught by two visitor personas the same day): using the a₀ ≈ 1.08×10⁻¹⁰ m/s² item 4 quoted at the time (the H₀ = 70 value; corrected to 1.04 at H₀ = 67.4 on 2026-07-22, since Σ₀ = a₀/(2πG) algebraically), Σ₀ ≈ 123.3 M☉/pc² — a 4% miss at the site-standard H₀ = 67.4 (119.0 vs Freeman's observed 124 M☉/pc², Freeman 1970; the often-quoted “0.5% match” needs H₀ ≈ 70, and the target itself is uncertain by ~2× through M/L, so no percentage is meaningful), not the previously stated “≈110, 12% error” (that 110 required H₀ ≈ 62 km/s/Mpc, inconsistent with the a₀ row directly above it). At the site-standard H₀ = 67.4 km/s/Mpc (adopted 2026-07-17, resolution note below) the same formula gives Σ₀ ≈ 119.0.

The combination cH₀/G has dimensions of surface density, so this is dimensional bookkeeping — expressing Freeman's empirical value via cosmological constants, not deriving it from physics. Σ₀ is not independent evidence from a₀: since Σ₀ = a₀/(2πG) exactly, this is the a₀ row propagated through a linear relation, not a second derivation. Which Σ_M: Milgrom's critical surface density is often defined as a₀/G, which is ≈ 861 M☉/pc² for a₀ = 1.2×10⁻¹⁰ m/s²; this page uses the a₀/(2πG) normalization throughout, which is 2π smaller. Under that convention Milgrom's Σ_M = a₀_,obs/(2πG) ≈ 137 M☉/pc² sits 10% above Freeman's 124, and this page's a₀ (1.04 at H₀ = 67.4) sits 13% below Milgrom's observed a₀ — equivalently, Σ₀ = 119 sits the same 13% below Milgrom's Σ_M = 137: one gap, one number, not two rows of independent support. The comparison to Freeman's 124 is itself convention-laden: that figure is a B-band central surface brightness converted with an assumed mass-to-light ratio Υ, so a different band or Υ moves it (see the caveat below). With the a₀/G normalization the same comparison would be off by 2π. (Definition note added 2026-10-06, graduate-physics simulated-visitor persona.) Re-badged from “Validated” (2026-04-28): a tight numeric match is not sufficient to claim derivation of what is, in origin, an observational law.

H₀ inconsistency RESOLVED (2026-07-17, root cause identified by a visitor physics persona 2026-07-17): the 119-vs-123.3 split across pages was exactly an undisclosed Hubble-constant switch — Σ₀ = cH₀/(4π²G) gives 119.0 M☉/pc² at H₀ = 67.4 km/s/Mpc (Planck 2018) and 123.6 at H₀ = 70. Site standard is now H₀ = 67.4 km/s/Mpc (matching /freemans-law and Honest Assessment), giving Σ₀ ≈ 119. Given the caveat below, no sub-percent “match” claim survives either H₀ choice, so the choice changes bookkeeping, not verdicts. (2) Neither error bar is meaningful on its own terms: Freeman (1970) reports a central surface brightness(21.65±0.30 B-mag/arcsec²), and converting to a surface density requires a mass-to-light ratio M/L_B that for disk galaxies spans roughly 1–3 — the target itself is uncertain by a factor of ∼2. Quoting agreement to 0.5% (or 4%) against a quantity known to a factor of two is precision theater; the honest statement is “consistent with Freeman's law within the M/L systematic,” not a specific percentage.

Attribution corrected 2026-08-08 — the badge said the wrong prior art. This row was labelled “Freeman's Law Re-expressed,” but a₀/(2πG) is not a re-expression of Freeman — it is Milgrom's critical surface density Σ_M, a standard MOND quantity, and “Freeman's law follows from Σ_M” is Milgrom's own 1983 result (the observation that disc galaxies cluster near the MOND transition surface density). So this row re-derives Σ_M and rediscovers Milgrom's explanation of Freeman's law — same demotion, correct lineage. The a₀ row above already credits McCulloch/Verlinde/Smolin for exactly this reason; the same standard now applies one row down. Caught by a visitor physics persona, 2026-08-08.

Session #91

Characteristic radius from velocity and acceleration. The natural MOND transition radius is V²/a₀; the factor 3 is fitted, not derived, and the archive's “97% accuracy” was quoted against a comparison set the session does not state — so the 3% is a one-parameter fit's residual, not a validation of anything. (Label rewritten 2026-09-06: it still carried the word “Validated” inside a parenthetical, and a graduate-physics persona read it as current.)

Sessions #44–45, #170, #185–186, #218–219 — provenance audit executed 2026-07-17

This page's bottom line is “zero parameters with first-principles derivations,” and φ is its strongest exhibit — previously missing from this page entirely. The golden ratio entered as an empirical fit (S44: virial exponent B = 1.62, noted “≈ φ”), was adjudicated the next day by the archive's own audit (S45: “intriguing coincidence but not significant — don't claim φ is fundamental”; 0/8 comparator scalings have φ exponents), then reappeared fully formed inside C(ρ) at S170 with no derivation and no citation of the S45 ruling.

The claimed derivations do not survive their own text: S186's “information conservation x + x² = 1” is the golden ratio's defining identity postulated as physics (circular); S219's “scale recursion theorem” inserts λ = 1 + 1/λ without anything forcing it; S218 concedes its own Boltzmann route “gives exponent 1, not 1/φ.” The three retro-justifications do not cite each other — derivations that get re-invented rather than accumulate are the signature of fitted-then-named. And at every measured slot the data preferred a different constant: the anchor fit was 0.66 (1.0% from 2/3, 6.8% from 1/φ); the a₀ slot prefers 3/2 (S217); Gaia gives 0.688 ± 0.10 (2/3 at center). Consequence for the ledger: TEST-09's BTFR kill gets stronger — the honest accounting is one derived ingredient (Ω_m) plus one free exponent dressed as a constant, and the parameter scan shows no exponent value rescues the slope. Full chain: explorer/findings/2026-07-17-phi-exponent-provenance-fitted-then-named.md.

Re-badged 2026-09-05 on the explorer's 2026-09-03 sensitivity scan. This row carried “Speculative — asserted, not derived” while items 1 and 3 carried Audited-Negative, which inverts the measured ordering: the floor C_min = Ω_m (equivalently B_max = 1/Ω_m) is the only parameter the galaxy sector is sensitive to (χ²/N moves up to 6.35×10⁴ with the floor, 1.07–2.7× with γ and A together), it is measurable (best-fit ε₀ = 0.220, B_max = 4.55, 2026-08-30), and its asserted value is excluded (TEST-09/TEST-10; RG's own fits demand 4–22; the weak-lensing RAR of Brouwer+2021 demands ν = 110–347 against any ceiling ≤ 18). So TEST-09/10 refute the chain's load-bearing parameter, not an optional closure. The 08-26 derivation-from-disc-geometry (Gauss averaging gives 1/C_min as the L2/L3 maximum) explains where the ceiling comes from; it does not make the value right.

The floor is not what fails in globular clusters; the knee is (explorer 2026-09-07, added 2026-09-08). On 42 Galactic globular clusters (Baumgardt & Hilker database) the outer velocity-dispersion slope is insensitive to the floor value — substituting this framework's Ω_m = 0.315 into Refracted Gravity's permittivity at RG's own knee (0.0083 M_☉/pc³, RG's elliptical-galaxy calibration from Cesare et al. 2022; its disc knees are lower) changes nothing and passes — while every knee this framework uses (0.11–1,400 M_☉/pc³) is excluded or marginal at γ = 0.489; at the registered γ = 2 the excluded band narrows to 0.5–100 and the framework's measured 0.161 becomes marginal rather than excluded (the fork). The one placement that escapes the shape test (host V = 220 km/s ⇒ knee at ~1 pc, whole cluster at the floor) is a pure G → G/Ω_m rescaling, which divides every cluster's stellar M/L_V by 3.175 and pushes 96% of 167 clusters below M/L_V = 1.2, beneath any 12 Gyr metal-poor population. So this row's parameter is refuted on SPARC by its value and on clusters by its mass budget; item 3's A is refuted on clusters by its location. Full result: Honest Assessment.

And freeing this parameter does not repair the galaxy sector — it makes it worse (measured 2026-09-08, read 2026-09-09). The obvious repair, given that 77–88 % of SPARC discs demand a boost above 1/Ω_m = 3.17, is to lower the floor and raise the ceiling. Run on 153 SPARC discs at Refracted Gravity's E0 floor f = 0.089 (ceiling 11.2), the diagnosis is confirmed and the cure fails: the fraction of discs that cannot be lifted falls from ~80 % to 10–23 %, and χ²/N rises from 69–85 to 195–2700 — 3 to 17× worse (computed by solving the field equation in disc geometry, not by the algebraic g_bar/C shortcut, and with this framework's switch at RG's floor — Refracted Gravity at its own published parameters was run separately and is reported on For Researchers). The extra boost arrives in the inner disc, where SPARC wants none. So this row's number is refuted, but replacing it with a better number does not rescue anything: the failure is that a switch keyed to local density turns on inside the baryonic disc. The ceiling is a symptom; the argument of C is the disease. See Honest Assessment for the table.

Added 2026-07-28 — flagged missing from this page by a simulated-visitor persona (grad student + researcher)

Honest Assessment calls the bounded acceleration boost B ≤ 1/Ω_m ≈ 3.17 “the framework's only structural difference from MOND,” and both executed discriminating tests (TEST-09's BTFR slope, TEST-10's dwarf DM fractions) are corollaries of this one number. It carries the site's entire discriminating weight and, until today, appeared nowhere in this chain — a parameter-inventory gap, not a derivation.

No archive document derives 1/Ω_m from the compander or from anything else — it is asserted. A live, unexecuted proposal (Research/proposals/boost_ceiling_provenance_and_class_exclusion.md) flags that the cosmic ratio a dynamical-to-baryonic boost should reference is arguably Ω_m/Ω_b ≈ 6.40 (the baryon budget), not 1/Ω_m(a matter-to-critical-density ratio). Under that convention the f_DM ceiling is 0.844, against 0.685 under 1/Ω_m. TEST-10's reported median f_DM is 0.755, so the median passes the 0.844 ceiling. Separately, the “69% of SPARC exceeds the ceiling” headline counts all 153 galaxies against the 0.685 ceiling (106/153), so it holds only under the 1/Ω_m convention. The kill does not depend on the convention choice at the tail: SPARC's maximum observed f_DM = 0.927 requires B ≥ 13.7, which no ratio built from Ω_m supplies (1/Ω_b = 20.3 does, and is excluded by the weak-lensing RAR one decade lower, 2026-09-25) — so the robust, convention-free statement is a class exclusion (B_max ≲ 14 is excluded by SPARC dwarfs), not the median-based percentage.

The sweep is now executed — TEST-09's half, run 2026-09-18, and it comes out the same way as TEST-10's. The paragraph above said this sweep “has not yet been executed against the full per-galaxy dataset.” That was true from 2026-07-27 to today for the BTFR-slope limb (TEST-10's exceedance limb was run 2026-07-29). The ceiling enters TEST-09 as the floor of C(a) = C_min + (1−C_min)·x/(1+x), so the sweep is a one-parameter change with everything else — sample cuts, V_flatestimator, bootstrap — held at TEST-09's own values. The identity control reproduces the published n = 3.35 ± 0.07 exactly before any new number is read.
• B_max = 3.175 (1/Ω_m): n = 3.347, |Δn| = 0.406 — fires
• B_max = 5.389 ((Ω_m−Ω_b)/Ω_b): n = 3.456, |Δn| = 0.297 — does not fire (the registered threshold is a strict > 0.3; this misses it by 0.003, about 0.03σ of the slope error, so the verdict is decided at the third decimal)
• B_max = 6.389 (Ω_m/Ω_b): n = 3.492, |Δn| = 0.260 — does not fire (data slope n_obs = 3.753 ± 0.100 throughout; the |Δn| = 0.300 edge falls at B_max ≈ 5.31, re-run 2026-10-30)
What each ceiling presupposes (added 2026-10-27): 1/Ω_m is the framework as written (C floored at the cosmic C₀ = Ω_m, which contains cold dark matter); Ω_m/Ω_b is a galaxy sector that replaces dark matter, a premise the dark-energy sector cannot carry (2026-09-22). The non-firing branch is a variant whose cosmology fails.
Pre-fixed rule: the kill stands iff it fires under every candidate. It does not. Both discriminating galaxy-sector kills now rest on the same undefended choice of cosmic ratio, and the convention-free form of TEST-09 is the class exclusion B_max ≲ 5.3 is excluded by the SPARC BTFR slope — the slope analogue of TEST-10's B_max ≲ 14. A free scan shows why the sweep bites at all: the predicted slope moves monotonically from 3.26 at B_max = 2 to 3.89 at B_max = 100, a range of 0.62, so the ceiling is a real lever on this observable rather than a labelling choice. Count unchanged; the recount gates on the project lead (proposal open question 4). Pre-registered at site commit 89e0467; maintainer/scripts/which_C_carries_the_floor.py (+ _PREREG.md, _output.txt).

And the floor is not pinned: the C this parameter floors is not the C the plotter draws (measured 2026-09-18). A simulated-visitor researcher pass read the plotter's “max C on this disk: 0.001” against this floor of 0.315 and concluded the floor binds at every radius, so the galaxy sector applies the constant 3.17 and TEST-09/TEST-10 are algebra. The premise conflates two functions. The plotter's “max C” readout reports the density-keyed C_ρ (unfloored; since 2026-10-08 its solid curve then applies the Ω_m floor by division, its dashed curve adds C_ρ in quadrature; its per-disc maximum over 123 real SPARC discs has median 1.2×10⁻³ at γ = 2, and 0 of 123 reach Ω_m — so the plotter's number does generalise). TEST-09/TEST-10 evaluate the acceleration-keyed C_a, whose floor is built into the functional form: over the same 2,856 radii it runs 0.329–0.954 (median 0.515), with 0.00% within 1% of the floor and a boost of 1.05–3.04× that never reaches the ceiling. The reading is refuted, and so are its two corollaries (predicted f_DM has s.d. 0.062 about a median of 0.585, not a delta at 0.685; the slope is not ceiling-independent). Recorded here at the same prominence a confirmed finding would get.

At which epoch? The evolving ceiling was tested on published high-z disc kinematics and failed as a reading (explorer 2026-09-11; corrected here 2026-09-14). Until then this box said the fork closed without data. An expert simulated-visitor pass noticed that this page evolves a₀ to a₀(z) = cH(z)/2π but freezes Ω_m. It proposed an evolving ceiling, f_DM,max(z) = 1 − Ω_m(z), which is ≈ 0.21 at z = 1 and ≈ 0.075 at z = 2. (This box said “≈ 0.05” for z = 2, but that is the value at z ≈ 2.4.) Branch (i), the Ω_m/Ω_b reading, still stands. Baryons and total matter are both dust, so Ω_b(z)/Ω_m(z) = 0.1565 at every z. The ceiling is 6.39 at all epochs, and there is nothing to measure. Branch (ii), the 1/Ω_m(z) reading. Under ΛCDM, B_max(z) = E(z)²/[Ω_m,0(1+z)³] falls from 3.17 to 1.27 at z = 1 and 1.08 at z = 2. Under the framework's own dark-energy sector the z = 2 value depends on γ: it runs from 1.002 to 1.201 across γ ∈ [0.3, 2], and is ≈ 1.08 near γ = ½. A single z = 2 value that holds at every γ does not exist. The 2026-08-08 closure was void. It called branch (ii) an internal contradiction, because the ceiling falls while a₀(z) rises 3.03×, and it concluded that the fork could be settled without data. There are three problems with that argument:
• It needs the evolving a₀(z) that this page's a₀ item marked disfavored on 2026-08-01.
• A density-keyed floor has no a₀ to collide with.
• The same ceiling at z = 0 is TEST-09/10, which the site treats as a test.
What was run instead. Session 100's closure, ρ_DE = ρ_m(1−C)/C, makes C at the mean matter density identically the model's own Ω_m(z) (maximum deviation 1.1×10⁻¹⁶). That is the literal reading of the floor as “the cosmic average coherence”. The rule was pre-registered before any table was fetched. It covers z ≥ 1.5 and refutes if at least max(5, 10%) of the discs sit more than 2σ above the cap, or if the median exceeds the median cap by more than 3 s.e. It was run across γ_DE ∈ [0.428, 0.560], the DESI DR2 wCDM 2σ envelope, mapped crudely at single pivot redshifts. On the N = 21 published discs:
• Price+2021 MCMC: 10–13 discs above the cap at > 2σ, median excess 6.7–7.4 s.e. Refuted.
• Genzel+2020 least-squares, same galaxies: 6 above (5 needed), 2.3–2.9 s.e. Refuted at every γ in the band.
• RC100 shape (N = 67 at z = 1.2–2.5): the data give f_DM ∝ (1+z)^−b with b = 1.07 ± 0.33. The cap needs b = 2.23–2.91, which is 3.5–5.6σ away.
A Salpeter IMF stacked on the least-squares fits removes the refutation (4 of 21 above the cap for Price, 3 for Genzel). But it leaves 2 (Price) or 5 (Genzel) of the 21 discs super-maximal at > 2σ, with baryons alone exceeding the measured rotation. The reading does get the sign and today's level right: RC100's z = 0 normalization is 0.75, against a cap of 0.685. It fails on steepness. No epoch or memory variant of the floor can loosen the z = 0 ceiling, because C_DE(ρ̄_m(z)) ≥ Ω_m on the past light cone. Branch (ii) is eliminated as a reading, and the floor is left as a fit constant set equal to Ω_m. It was never a registered prediction, so the refutation count stays 6. This candidate died on a measurement, not on a definitional choice. Script: explorer/findings/scripts/evolving_floor_highz_and_ambient_z0.py (+ _output.txt). The void closure is in Research/proposals/boost_ceiling_epoch_fork_closes_the_last_candidate_discriminator_20260808.md.

Which force law this ceiling is a statement about (2026-08-04): B_max = 3.17 is a statement about which of the site's live galaxy-sector formalisms is meant. Under g_obs = g_bar/C(ρ) — the reading the f_DM = 1−C identity implies — the boost B = 1/C is unbounded by construction and exceeds 3.17 by 2–5 orders of magnitude at every radius, including the disk centre, on the site's own five plotter galaxies. The ceiling is consistent only with the quadrature amplitude law (the Galaxy Curve Plotter's dashed curve since 2026-10-08; its solid curve is floored division) v² = v_b² + (V_flat·C)², where the extra term is bounded by V_flat irrespective of C. No physical density floor rescues the other reading: capping B at 3.17 there requires ρ ≥ 0.177 ρ_crit, a floor 10²–10⁴× the densest point of the model disk. See the C-convention note on Galaxy Rotation.