Mills-Randal: Physicist review (Step 2)

Reviewer: PhD physicist. Subject: R. L. Mills, The Grand Unified Theory of Classical Physics (GUT-CP), project copy = "December 10, 2018 Edition" PDF, 1,876 pages. Date of review 2026-10-09. Tags: [read] = from the project documents, deep reads or downloaded files; [web] = Internet source with URL; [memory] = recalled, with confidence; [calc] = recomputed by a deep reader or me from printed numbers. Interested party: Mills founded and runs BlackLight Power / Brilliant Light Power (BLP), which sells the hydrino technology. The book, the company reports and most cited experimental papers come from that side. That is reported as a fact about the sources, not as a judgement of their data.

0. Source gap and what this review rests on

• The deep reads cover about 40 percent of the 1,876 pages (claims.md). Claims 1, 2, 4 and 5 rest on deep reads. Claim 6 rests on deep reads of Ch.32 (cosmology half), Ch.36-39 and Appendices. Claim 7 (Ch.24-30) rests on digests only; my verdict there is provisional. Ch.11 (the H2 and H2(1/p) molecule, source of the 4.478 eV and 31.68 eV numbers), Ch.3-4 (free electron, photon), Ch.9, Ch.17-21 and the first half of Ch.32 (Schwarzschild metric, perihelion, light deflection) were not deep-read by anyone; I do not judge them.
• Mills's mathematics was checked mainly by arithmetic reproduction. Equations are image-only in the text layer, so every number in the deep reads came from page images.
• The November 1995 edition (the alleged prediction of cosmic acceleration and the top quark) is not in the folder and I did not find it online. Both "predictions" are therefore unverified as predictions (see Claim 6).
• The Investigator's newer material (2025 edition, p <= 126 paper, Hagen-Mills 2022 and 2024/26 papers, NASA 1996, EarthTech 1997, Thermacore, USPTO and UK records, IIT Kanpur 2026) is weighed below. I read the Hagen-Mills 2022 paper myself.

1. The mainstream view, stated fairly (once, for all claims)

1. Electron and atom. In quantum mechanics (QM) the electron in hydrogen is described by a wave function psi with a stationary-state probability density |psi|^2 and charge density e|psi|^2 that do not depend on time. A static charge density and a steady probability current do not radiate under Maxwell's equations. The discrete spectrum comes from requiring psi to be square-integrable (finite everywhere, decaying at infinity); that requirement gives n = n_r + l + 1 with integers only, and the bound-state spectrum of H is bounded below at -13.6 eV [memory, high]. Spin 1/2 and g = 2 come from the Dirac equation; the 0.116 percent correction (g/2 - 1 = 0.00115965218...) comes from quantum electrodynamics (QED), calculated to five loops and measured to about 1e-13 [memory, high]. The measurement g/2 = 1.00115965218059(13) is Fan, Myers, Sukra and Gabrielse, PRL 130 (2023) [web: value confirmed in a search summary of the paper, https://cfp.physics.northwestern.edu/documents/2023MeasurementElectronMangeticMoment.pdf]. QM is a stunningly well-tested theory but does not explain why nature is quantum; interpretation (Born rule, measurement) is genuinely debated, and the Schrodinger "charge cloud" is not taken literally by experts. Mills's complaint about interpretation is partly fair; his conclusion that QM is wrong is not supported.
2. Molecules. Bond energies, lengths and angles of small molecules are computed from the many-electron Schrodinger equation with ab initio methods (coupled cluster, configuration interaction) to about 0.04 eV (1 kcal/mol) for atomisation energies of small molecules, and with density functional theory (DFT) to a few kcal/mol for typical organic molecules [memory, high]. For large organic molecules, group-additivity schemes (Benson) predict heats of formation to about 0.01-0.04 eV by fitted group values [memory, medium; as quoted in the Ch.15 deep read].
3. No state below n = 1. For hydrogen the electron is a bound state of the Coulomb potential. The ground state is the lowest eigenvalue; a nonradiative collision moves an atom between eigenstates, so cannot place an electron into a state that is not in the spectrum. The Klein-Gordon "hydrino" solution (Naudts 2005) and its Dirac analogues are extra mathematical solutions with unphysical behaviour at r = 0; Dombey (2006) shows they disappear if the nucleus has an arbitrarily small non-zero radius [read: files/arXiv-Dombey-2006 abstract; web: arXiv physics/0608095]. Precision hydrogen spectroscopy (1S-2S to parts in 1e15, Lamb shift, proton-radius work) agrees with standard theory [memory, high]; no spectral line of hydrogen is unexplained that would need sub-ground states.
4. Gravity, cosmology. General relativity plus the standard flat Lambda-CDM model fit the CMB (Planck: H0 = 67.4 +/- 0.5 km/s/Mpc, T_CMB = 2.7255 K, tensor-to-scalar ratio r < 0.036), supernovae, baryon acoustic oscillations and BBN; dark matter is non-baryonic by the CMB peak heights and BBN deuterium [memory, high]. There is a Hubble tension (SH0ES 73 versus Planck 67), but nothing near 78.5.
5. Bell tests. Local deterministic theories obey Bell/CHSH inequalities; experiments violate them, and since 2015 loophole-free tests (Delft, Vienna, NIST) close detection and locality loopholes together [web: Hensen et al., Nature 526, 682 (2015), https://arxiv.org/abs/1508.05949; S = 2.42 +/- 0.20, 245 trials].

2. Claim 1: classical electron shell, closed-form atoms

Plain statement. A zero-thickness spherical shell of charge and current obeying Maxwell and Newton, not radiating, gives spin, the Bohr magneton, g, Lamb shift, fine and hyperfine structure and ionization energies of 1-20-electron atoms, in closed form with integers and constants only.

a. Mathematics internally consistent? Partly, not fully. - The arithmetic reproduces: Bohr radius and energy (reduced-mass), Table 1.5 (equals the Dirac point-nucleus formula with a reduced-mass factor), Tables 7.1, 10.x row for row [read: deep reads Ch.1, 7, 10; calc]. - Inputs are postulated, not derived: the angular momentum hbar ("inalienable hbar", Eq. 1.35; Eq. 1.37 "derives" hbar by integrating the velocity built from hbar, a circular step); spin hbar/2 arises from a hemisphere-only great-circle construction chosen to give that target (Mills 0.495 hbar numerically); the full Bohr magneton needs an extra "trapped resonant Larmor photon" postulate; N = 3/2 is inserted by hand; the printed field energy is (4 pi)^2 too large in SI (mixed Gaussian/SI) [read: Ch.1 deep read]. - The nonradiation proof requires lambda_n = r_n (wavelength set equal to radius, in conflict with 2 pi r_n = lambda_n in Eq. 1.15). For l >= 1 the correct radial factor is spherical Bessel j_l(k r_n), not zero at k r_n = 2 pi (j_1 = -0.159). The Poynting-vector proof drops a charge term that the book's own continuity equation makes non-zero. The l = 0 state does not radiate, but only because it is time-independent, which is also true of QM's stationary states [read: Appendix I-II deep read]. - The self-force (Poincare stress) is removed by postulate; with it included, Mills's own force balance would give r = 2 a0 [read: Appendix II]. - g factor: Eq. 1.228, g/2 = 1 + alpha/2pi + (2/3)alpha^2(alpha/2pi) - (4/3)(alpha/2pi)^2, is single-parameter (alpha). Its leading term is Schwinger's. Several scale-setting steps (2 pi radius factor; v = c in one term and v = alpha c in another; 2/3 internal-field fraction; sign of the dissipative term) are modelling choices, none checked independently [read: Ch.1 deep read 2.4]. - Muon g (Appendix III): Eq. 26 uses beta^2 = 2/g_mu from the measured g_mu itself, so it returns the input within 7e-9 for any input value; it is circular, not a test [read].

b. Consistent with experiment? - g factor: with the 1987-era alpha, 1.0011596521215 vs the measurement of its day (the book's "eleven-figure agreement"). That test had resolution only about 7e-9 in g/2 because alpha was known to 1e-8 then. With today's independently measured alpha (Rb recoil 137.035999206(11), Morel et al. 2020 [web: https://www.cnrs.fr/en/french-team-has-improved-measurement-fundamental-physical-constant]; Cs recoil 137.035999046(27), Parker 2018 [memory, high]) Eq. 1.228 gives 1.00115965246, versus the measured 1.00115965218059(13): off by 2.9e-10, about 2,000 times the experimental uncertainty [calc, deep read]. The comparison also holds with the Cs/Rb discrepancy (5 sigma between them, [web same CNRS source]) because both differ from Mills's implied alpha^-1 = 137.036033 by about 3e-5, thousands of sigma. Mills's second and third coefficients are -1/3 and +pi^2/3 (in powers of alpha/pi) against QED's -0.3285 and +1.1812; the errors nearly cancel [calc]. The book's 2020 text does not report this comparison. - Atoms: He 24.58750 vs 24.58741 eV uses a helium-only formula whose bracket is borrowed from the muonium hyperfine section; the ion formula would give 23.68 eV [read: Ch.7]. Multi-electron ions show errors up to 4.6 percent (Ar3+, Ca3+, Cl3+) with low-Z ions known to six figures wrong by 2-4 percent; per-atom integers are chosen "to give an energy minimum" without minimisation shown; for neutral 2p atoms Mills's own equations would give lower energy for other assignments [read: Ch.10]. Ab initio atomic structure (relativistic multi-configuration Dirac-Hartree-Fock, coupled cluster) reproduces these ionization energies to much better than 1 percent [memory, high]. - Hydrogenic energies (Bohr, Dirac) reproduce because the force balance is the Bohr model; this is a re-description of correct results, not new content. Lamb shift, 21-cm line, muonic-hydrogen Lamb shift are derived with Mills's ad hoc factors (deep read Ch.2; summary table 14) and do not outperform QED. - Fair note: H- affinity 0.754175 eV vs Lykke 0.754195 +/- 0.000019 (about 1 sigma) [read: Ch.5, Ch.7; calc]; but this uses a special reduced mass and ab initio QM also reproduces H- to the same precision [memory, medium-high].

c. Tested? By whom? Replicated? The numbers are checkable arithmetic. Randy Booker (UNC Asheville) reports reproducing Tables 1.1-1.5, 7.1-7.2 and Eqs. 32.1-32.114 "to a very high degree of accuracy" in the company-hosted Theory-Reviews.zip (2017-2022); he also wrote company power-validation reports; compensation not stated in the pages read [read: research/2026-10-09-brilliantlightpower.md]. That confirms the numbers follow from the equations, which is what our deep readers also found; it does not test the postulates. D. Selke (2025, Eur. J. Appl. Phys., not read) extends the force-balance method to heavier ions [read: google-scholar note]. No outside test of the postulates (nonradiation, hbar/2 construction) is on record other than the critiques (Rathke 2005, New J. Phys. 7, 127, who writes that "severe inconsistencies in the deterministic model are pointed out" [read: files/arXiv-Rathke-2005 abstract]; Mills answered in a 165-page reply, 2008, not read).

d. If true, what would change? QM, Dirac theory and QED would be wrong or approximations of a classical electrodynamics; yet QED's other predictions (Lamb shift to 1e-6, g to 1e-13, hyperfine of muonium) would need to emerge too. The book gives no derivation of the multi-loop QED terms that match to 1e-12.

Verdict (split). - Strong form (a Maxwell-plus-Newton shell electron derives spin, g and all atoms, making QM unnecessary): Tested, not supported for the g factor against modern alpha (2,000 sigma off), and Internally inconsistent at the postulate level (circular hbar; lambda_n = r_n versus 2 pi r_n; mixed Gaussian/SI; Poincare stress by postulate; muon g circular). - Weak form (the shell with chosen integers reproduces one-electron energies and some atomic series to 0.1-4 percent): Established as arithmetic, not as new physics (equivalent to Bohr/Dirac for 1 electron; worse than standard methods for multi-electron ions).

3. Claim 2: hydrinos and the BlackLight process

Plain statement. Hydrogen has stable states n = 1/p (radius a_H/p, energy about -p^2 x 13.6 eV, p to 137). A catalyst absorbing m x 27.2 eV nonradiatively lets H drop to H(1/(p+m)), releasing [(p+m)^2 - p^2 - 2m] x 13.6 eV in a continuum. Evidence offered: EUV continua, hydride ion binding energy and NMR, molecular hydrino H2(1/p), the EPR signal (g = 2.00445) in SunCell product measured by W. R. Hagen and "replicated by Bruker", calorimetric power.

a. Mathematics. - The energy-hole algebra reproduces: p -> p+m releases [(p+m)^2 - p^2] x 13.6 eV; for p = 1, m = 1, 2, 3 the edges are 13.6, 54.4, 122.4 eV (91.2, 22.8, 10.1 nm); H -> H(1/4) releases 204 eV vs 1.48 eV per H atom for combustion of hydrogen [read: Ch.5 deep read, calc]. - But the force balance (Eq. 5.28) is satisfied identically for any real Z_eff = p + m: it does not itself select integers. The integer comes from Mills's separate "trapped photon" condition (Eq. 5.27, a Laplace-equation solution) and the 27.2 eV unit from a multipole-counting assumption (Eqs. 5.44-5.46) [read: Ch.5 deep read 2b, 2d]. The cap p <= 137 is justified (in other writing) by v = p alpha c reaching c [memory, medium]; the 2025 Mills-Hagen FeOOH paper now says p <= 126 "whereafter the proton and electron decay" [read: files/BLP-FeOOH_Hydrino_Paper.pdf p.1]. The limit changed between editions with a different stated reason. - Hydride formula (Eq. 5.11) mixes a_H and a_0 and uses a special reduced mass m_e/sqrt(3/4); the NMR formula is pi/2 larger than the standard Lamb shielding formula for the same shell [read: Ch.7 deep read]. H2(1/4) dimer rotation printed 44.30 cm^-1 does not follow from its own formula (26.43) [read: Ch.16 deep read 5.3]. Table 5.1 mixes 27.2114 and 27.2 in different rows (rounding-level). - Textual slips: 10.1/22.8/91.2 nm paired with m = 1/2/3 in reversed order on p.263; Lykke value printed as 0.75418 eV where 0.754195 is correct [read].

b. Consistent with established experiments? This is the crux. - Against. Precision hydrogen spectroscopy and the stability of the 1s state: no known transition requires sub-ground states. The standard-theory reason for no lower state is not a "definition" (Mills's word, Intro p.44) but square-integrability; for the relativistic route, the extra solutions vanish with a finite-size nucleus (Dombey 2006). If H(1/4) were produced in bulk at the claimed 100-kW scale, hydrino atoms (neutral, with small cross-section) would be a very large mass of ordinary-looking gas with no independent isotope or mass-spectrometry signature in the record I have; the Hagen-Mills paper itself says the sample is "contaminated" with regular H2 and the molecule "difficult to store" [read: Hagen-Mills 2022 p.3]. - Mixed. The 10.1 nm / 22.8 nm continua in pinch discharges were reproduced "at the Harvard Center for Astrophysics" per the book (refs 61-66, Ref 66 hosted on blacklightpower.com) [read: Ch.32 deep read 8]. I could not read the CfA report. Mainstream reanalyses by Phelps and Clementson (2012), Lawler and Goebel (2012), Loureiro and Amorim (2010) are cited in Hagen-Mills 2022 refs 71-84 [read: research note] but I did not read them; the mainstream explanation of fast-H Balmer broadening and EUV continua in hydrogen plasmas by ordinary processes is [memory, low-medium] and I do not rely on it. - Hydride and Lykke. The 0.75418 eV agreement is real but H- is also ordinary physics; it is not evidence for sub-ground states.

c. Tested? Replicated? Evidence classes and who produced them [read: Ch.5 deep read 6]: 1. All experimental citations for hydrino evidence on pp.248-264 (refs 1-9, 28) have Mills as author; ref 8 is "submitted". The p.252 sentence listing ten evidence classes carries no citation at all. 2. EPR (strongest recent evidence, read by me). Hagen and Mills, IJHE 47 (2022) 23751: an outside academic EPR specialist (Hagen, TU Delft) measured a solid made by BLP from SunCell Ga2O3, purified in KOH (Ga(O)OH fibres). At 9.8 GHz there is a multi-line pattern centred at g = 2.0046 [read: paper p.5-8]. Strengths: empty-tube baseline, 11 samples from 11 reactor runs (stated), frequency-dependent study showing features that stay at fixed field spacing while the main line stays at fixed g (consistent with frequency-independent hyperfine splitting around a single g), alternative "two radicals plus 14N" fitted but judged unlikely [read: Fig. 6, "Search for alternative interpretation"], ToF-SIMS finding "no nitrogen" and "no potentially paramagnetic transition ions" [read: p.3]. Hagen has no financial relationship with BLP per the declaration [read]. Honest limits, in the paper's own words: "The GUTCP fit... is semi-quantitative with an average error of 0.004 +/- 0.550 gauss over the 11 lines"; "fluxon separation... does not quantitatively fit the predicted value; in particular the separation is not a constant"; "the number and relative intensities of fluxon lines for a given satellite line are presently not understood"; "A broad signal underlies... origin and nature unknown" [read: p.8]. The book says the predicted spectrum "was confirmed" with "average error 0.097 G" [read: book p.1232]; the paper's own wording is weaker (0.550 gauss scatter). The sample was prepared by BLP; no sample was prepared by an outside lab; I found no blank run from a SunCell without hydrogen or with different gas treated identically (the book names control Ga(O)OH and control solutions only). The nuclear spin of 69Ga and 71Ga (I = 3/2, both) is not discussed in the text I searched, and a fixed-g hyperfine pattern from Ga or protons in a defect is exactly what an ordinary paramagnetic centre could give [memory, medium]; the authors' counter-argument (fixed g, no anisotropy of the powder pattern) is real but not a proof. The Bruker "replication": book gives one sentence and no reference; the 2022 paper adds "reproduced completely in duplicate on two different spectrometers in a lab of the spectrometer's manufacturer (Bruker, Billerica, MA, USA) [13]" where [13] is Mills's own submitted company PDF [read: Investigator note; paper p.5]. Single source on both counts, both co-authored by Mills; no Bruker-authored record found. Not suppressed here: it is evidence to be weighed and a Bruker or third-party EPR facility could settle it quickly. 3. Calorimetry. Company-hosted validation reports: Nansteel 2021 (169-210 kW plasma power, local gain about 4.2, a single 16 March 2021 test at BLP Cranbury); Booker 2020 (194-218 kW); Tse 2020 (Rutgers, hired observer); Glumac 2014 DSC (UIUC; 101-148 J/g exotherms "far more exothermic than the predicted thermochemistry"); 2016 summary claiming "1 MW+"; Jansson 194-page redacted report, image-only [read: research/2026-10-09-brilliantlightpower.md]. Hosted and selected by the interested party; several authors were hired or co-authors. They are measurements by named scientists, to be weighed, but none was published peer-reviewed as a neutral test and the full data were not read by me. Chemical alternatives for DSC exotherms and for steam-based calorimetry in a hot gallium/water cell (water reactions with gallium, hydrogen recombination, losses) are not excluded by anything I read. 4. Earlier electrolytic form. NASA Lewis TM-107167 (Niedra et al., Feb 1996): excess up to 11 W for 60 W in, gains 1.06-1.68, in a Hydrocatalysis-loaned cell; authors: recombination of dissolved H2/O2 "is an ordinary explanation" they could not exclude; further study "warranted" [read]. EarthTech (Austin, 1997): 164 h Ni-H2O-K2CO3 with Pd recombiner, "no detectable sign of excess heat", a 10 percent excess would have shown [web: https://earthtech.org/cold-fusion/mills/]. Thermacore SBIR (USAF Wright Lab, 1993-94): 25 W anomalous heat; Lehigh ESCA reported the predicted 55 eV peak [read: Investigator note]. The company lists 13 laboratories as confirmations; their own reports (AECL Chalk River, Westinghouse, MIT Lincoln, INEL, Hokkaido, Brookhaven) were not found and are not weighed as results. 5. Related and later. IIT Kanpur (Pala and Sharma, ECS 2026 abstract): first academic group announcing a replication programme for hydrino compounds; no results yet [web: https://iopscience.iop.org/article/10.1149/MA2026-01281445mtgabs]. NIAC Phase I (Rowan, 2002) took the anomalies as worth engineering study "regardless of the theoretical explanation" [read]. Hagen-Mills 2024/26 (broad-line EPR in other reactors, anti-Curie behaviour, transition near 107 K) is new data from the same pairing, not yet peer-reviewed in anything I saw.

Historical record (Investigator). USPTO allowed application 09/009,294 after the examiner reviewed experimental results and samples, then withdrew it after an "outside inquiry" because the Group Director's "main concern was... the electron going to a lower orbital in a fashion that I knew was contrary to the known laws of physics and chemistry"; "no one involved in the withdrawal had reviewed the application before" [read: research/2026-10-09-findlaw.md; web: https://caselaw.findlaw.com/court/us-federal-circuit/1362587.html]. UK IPO BL O/114/08 refused two BLP filings as contrary to accepted physical law; the High Court appeal [2008] EWHC 2763 (Pat), Floyd J, was allowed and remitted to the IPO [corrected 2026-10-09 after QE: was "was dismissed"] and is reported for the standard of proof (balance of probabilities) [web: https://caselaw.nationalarchives.gov.uk/ewhc/pat/2008/2763 via search summary; judgment text not read]. These are institutional dismissals on theory grounds, not neutral experimental tests. The absence-of-evidence item: 35 years, over $150 million claimed [read: BLP summary 2025, company's characterization], Form D raises $18.3M sold by 2018 [read: EDGAR note], yet no product sale found and only one non-company group (IIT Kanpur) announcing replication. Competing explanations: commercial delay or secrecy; effect not real; institutional refusal to test after 2000 and 2008. From the physics alone I weigh the second above the others, because a real 100 kW source of hydrino gas would be easy for outsiders to authenticate with a mass spectrometer and calorimeter loaned for a week, and because BLP itself invites it ("urgent invitation to academia at large to repeat" [read: paper p.8]). The first and third cannot be excluded from the record.

d. If true? The ground state of hydrogen would not be the lowest, QM's spectral theorem for Coulomb systems would fail, and the 13.6 eV per atom of combustion would be dwarfed by up to 200+ eV per atom: a new energy source, plus dark matter in the form of hydrinos. Thermodynamically nothing forbids a lower state of the H atom; the objection is that QM says there is none.

Verdict (split). - Existence of sub-n=1 states (theory): Tested, not supported in the sense that the standard-QM theorem excludes them, hydrogen spectroscopy shows no need for them and the relativistic "hydrino" solutions are shown unphysical (Dombey); the Mills version is Internally inconsistent at the nonradiation step (Claim 1). But note: one outside-lab measurement (Hagen EPR) is on record that Mills reads as support, and its ordinary explanations are not excluded by an independent blank. - Energy release at scale (kW levels): Never properly tested (company-hosted, interested-party calorimetry; no neutral controlled test; early electrolytic form: NASA ambiguous with recombination unexcluded, EarthTech negative = Tested, not supported for that earlier form). Not dismissed without testing at the claim's own level: it has been dismissed by agencies without testing, and tested only by affiliates. - EPR evidence for molecular hydrino: Never properly tested (single-sample-source, interested-party sample, partial independent replication claimed in two company-linked documents; no neutral blank).

4. Claim 3: hydrinos as dark matter and cosmic emission

a. Mathematics. The disproportionation cascades (Eqs. 5.65-5.80) use the same energy-hole formulas (arithmetically fine). The 3.48 keV assignment ("E = 3481.6 eV") is a chosen transition H(1/4)+H(1/4) -> H(1/17) type [read: Ch.5, Ch.32 deep reads]. With p up to 137 the set of available transition energies is a dense table, so a match to a single line is weak evidence unless the line list was fixed beforehand (not shown). b. Experiment. The original detections: Bulbul 2014, 3.55-3.57 keV; Boyarsky 2014, 3.52 keV [memory, medium]; the book prints 3.48 for the observed value. Hitomi's Perseus spectrum finds no line at the reported flux (excluded at 99-99.7 percent) but the data are limited by the spacecraft failure [web: Hitomi paper, https://www.repository.cam.ac.uk/handle/1810/276029 via search summary]. Dessert, Rodd and Safdi (Science 367, 1465, 2020) found no line from blank-sky XMM data inconsistent with decaying dark matter [web: https://arxiv.org/abs/1812.06976]; Abazajian argues the constraint is relaxed by a factor of about 20 and others disagree [web, same search]. So the line itself is contested; the book cites neither null result. The mainstream dark-matter case rests on CMB peak heights, BBN and the Bullet-Cluster-type separation of mass from gas; hydrinos are baryons by the book's own account, which conflicts with the BBN/CMB baryon budget; Mills's rejection of the CMB reading (his own peak formula, Claim 6) is why the two disagree. Whether neutral hydrino gas would separate from mass in the Bullet Cluster is a standard objection [memory, medium] I did not check. c. Tested. The 10.1 nm continuum in pinch discharges "reproduced at the Harvard CfA" (book; company-hosted report) and the Labov and Bowyer 1991 63.5 nm diffuse EUV line (their own favoured assignment was O V 63.0 nm) [read: Ch.32 deep read 8]. The 511 keV line assignments are Mills's. Hagen and Mills also have a cosmic-neutrino-background EPR paper (not read). d. If true. Dark matter would be ordinary baryonic hydrogen in a hidden state; the CMB-based cosmological parameters would be wrong. Verdict. Dark-matter identity: Never properly tested (conditional on Claim 2; no direct test possible until hydrinos are independently produced). 3.48 keV assignment specifically: Tested, not supported (line not confirmed by Hitomi, contested by blank-sky; assignment by selection from a dense table). Cascade arithmetic: reproducible.

5. Claim 4: one method for all molecules and matter

a. Mathematics. The equations reproduce arithmetically (hundreds of numbers reproduced by the deep readers, e.g. 0.91771 a0 carbon sp3 radius, Table 15.6 alkanes to five digits, Table 14.1 [read]). Internal slips: benzene has three different hybridisation numbers (0.85252, 0.79597, 0.91771) with separate justifications; CO2 bond energy 5.49557 vs 5.49553; NO2 angle error 3.1 degrees described as "excellent"; the muonium uncertainty printed 12 ppm for 12 ppb; Table 15 relative errors of nine rows do not follow (exp - calc)/exp [read]. Also "no fitted parameters" is over-stated: there is no least-squares fit, but (i) vibrational energies of the molecule being predicted enter as measured data (e.g. 993 cm^-1 for C-C; the size of this input, 0.06 eV per bond in ethane, is comparable to the 0.04-0.07 eV discrepancies claimed); (ii) heteroatom ionization energies are taken from tables; (iii) at least eight E_T values, seven c2 recipes and per-group integers are selected group by group "with the experimental totals in view"; (iv) a new functional group is defined whenever standard values would not match (CCl4, CBr4, CH2I2, CHI3) [read: Ch.14, 15 deep reads]. Whether these choices were made before or after seeing the data cannot be told from the text. b. Consistent with experiment? Mean absolute relative error of 431 summary rows 0.17 percent (median 0.056 percent); 317 within 0.1 percent; but cytosine +1.7, trimethylphosphine +2.3, D-ribose +2.4 and 13 of 20 amino acids +1.0 to 3.0 percent (mean 2.3), noble-gas van der Waals, DNA bases also 1-3 percent; the "experimental" values for 12 of those amino acids are crystalline-phase values compared with gas-phase calculated energies [read: Ch.15 deep read 0]. For n-alkanes the agreement is partly structural, since the 11 rows reduce to an intercept and a slope (additive by construction) [read]. Compared with mainstream, 0.17 percent of a ~10 eV total bond energy is ~0.02 eV, which is as good as or better than DFT and comparable to group additivity (Benson), but those use fitted parameters or large computing. Mills's novelty (if there is no fitting) would be remarkable, but the group-selection menu makes the independence of the 431 comparisons unestablished. Hydrino-molecule items in this claim: Raman lines of H2(1/4) "matched" to Diffuse Interstellar Bands (BLP-hosted PDF, ref 136; Hobbs et al. 2008 catalogue is independent) with no chance-coincidence estimate, and predicted lines dense at the tens-of-cm^-1 scale [read: Ch.16 deep read 5.2]. c. Tested. By Mills; Booker's checks in the company-hosted zip reproduce numbers from equations. No blinded test: the clean test (fix all rules, predict molecules whose groups were not used) is not on record. Not tested by others. Selke's extension is by an author outside the book but within the method. d. If true. Chemistry would reduce to a closed-form classical calculation; quantum chemistry's many-body methods would be unnecessary, yet they succeed on systems the book does not touch (e.g. transition-state structure, spectroscopy to wavenumber precision). Verdict (split). Weak form (a rule-based group-additivity scheme reproducing many bond energies to ~0.2 percent): Established as arithmetic, comparable to standard additivity schemes. Strong form (first-principles closed form, no fitted choices, with the 0.1 percent accuracy as prediction): Never properly tested (no pre-registered blind test; selection of discrete options per group; worse results 1-3 percent in several classes). Ch.11 (H2 itself, which several parts depend on) was not deep-read; this verdict is provisional there.

6. Claim 5: quantum mechanics is wrong or unnecessary

a. Mathematics. Mills's "Schrodinger charge must radiate" argument (Ch.40): he Fourier-transforms a radial function f(r) = r e^(-r/a0) (not the 1s density, which is e^(-2r/a0); it is not |psi|^2 of any single state), multiplies by a time-harmonic factor delta(omega - omega_n), and notes the transform is non-zero at |k| = omega/c. The closed form is correct up to a factor 2 [calc, deep read]; any smooth density has such a component, so the result is generic. The decisive premise is that the density oscillates at omega_n; for a stationary state psi_n e^(-iE_n t/hbar) the density |psi_n|^2 and the probability current are time independent and a static source does not radiate [memory, high]. Radiation arises in QM from superpositions of states (a transition). So the premise is the error, not the Fourier algebra. Elsewhere: Jackson's equation labelling is reversed in the book's use (17.70 is the quantum single-photon value, 17.71 the classical) [read: Ch.42 deep read 4.6]. Mills's Aspect "classical" curve R(phi)/R0 = 0.2490 + 0.2178 cos 2phi (Eq. 42.29) comes from adopting Horne's Eq. 5.17 (derived for the entangled two-photon cascade state) with the projection factor set to 1, so the "classical" number is the formula used for the QM prediction in Aspect's paper (the deep reader flags this as medium-confidence memory; the Aspect paper's abstract confirms results "in excellent agreement with the quantum mechanical predictions" but I could not confirm the exact numbers 5.8 versus 5.72 online [web: https://dpnc.unige.ch/~bravar/LaboIV/correlos/papers/Aspect_PRL47_81_460.pdf]). A genuine local model with Malus-law transmission and a shared random polarisation gives half the modulation (0.2490 + 0.1089 cos 2phi) and satisfies the inequality (delta = -0.096) [calc, deep reader]. Bell's theorem (Eq. 42.30) itself uses deterministic outcome functions A(a, lambda), so claiming it "does not apply" to deterministic classical physics is contrary to the derivation printed. Hess-Philipp (2001), cited favourably, was answered by Gill, Weihs, Zeilinger and Zukowski, PNAS 2002 [memory, high]; the book does not mention the reply. b. Experiments. Loophole-free Bell tests (Delft 2015: S = 2.42 +/- 0.20 over 245 trials, closing locality and detection loopholes; Vienna and NIST photon tests) are not addressed anywhere in the 1,876 pages (full-text search, deep reader) [web: https://arxiv.org/abs/1508.05949]. The book's newest Bell-related citation is 2007. The mechanism offered for the calcium cascade (photon "extended particle", pairs share a polarisation, "inefficient detectors") does not reach electron-spin tests. Stern-Gerlach, electron diffraction and superfluid-helium electron mobilities: Mills reproduces numbers with fitted-by-choice quantum numbers (about 69 candidate values for 15 peaks; the SQUID energy 0.14 K reproduces only with the book's own alpha factor, otherwise about 120 K [read: Ch.42 deep read 7-8]). Muon-ring g measurements and electron g to 1e-13 follow QED. c. Tested? The main question has been tested directly: Bell inequality violation has been replicated by many independent groups, with loophole-free versions since 2015. These support QM and refute local deterministic hidden-variable models (and by implication Mills's mechanism if it is such a model). d. If true. All of quantum information, laser physics, semiconductor physics and QED would need re-founding. Verdict (split). "Schrodinger charge density must radiate": Refuted by calculation (the conclusion depends on a time-harmonic density premise contradicted by stationarity; algebra otherwise fine). "Aspect/Bell results are classical": Tested, not supported (Bell tests, incl. loophole-free), plus Internally inconsistent (Jackson labelling; Bell's deterministic functions; the 'classical' curve is the QM formula). "Uncertainty principle and Born rule are refuted": Not testable as stated for the interpretation (no measurable difference proposed) and Tested, not supported where Mills offers observable differences. Mills's critique of interpretation (no physical picture of the electron) is a fair philosophical complaint but not a refutation.

7. Claim 6: spacetime, gravity, cosmology, particle masses

a. Mathematics. Reproducible arithmetic with internal slips: "1 Gyr = 3.1358e16 s" (correct 3.156e16); the "sec" fitted to 0.9975 s to close a 0.11 percent gap in the electron mass (Eq. 36.4); box 32.3 "ratio 0.8627" mislabelled; sign of the cos term in the digest wrong (digest error, not the book's) [read: claims.md; deep reads]. The field-correction factors that bring m_mu/m_e to 1.6e-8 are justified in words and "whether they were fixed before the comparison is open" [read: Ch.36 deep read]. Neutron mass 1838.674 m_e vs 1838.68366 is 5.2e-6 off, far outside measurement error [read]. A formula with several rational factors and alpha can match a handful of ratios by numerology; this is not a statistical test without a count of alternatives, which the Mathematician should provide. b. Experiment [read: Ch.32 and Ch.36-39 deep reads; calc; memory for observed values, high]:

The alleged 1995 prediction of acceleration: the book cites its own Nov. 1995 edition, Ch.22, "to the astonishment of cosmologists" [read: book p.1636, deep read 7]; the model predicts acceleration from t = 0 throughout the first quarter-cycle, with no earlier decelerating epoch, whereas the standard observational picture has deceleration then acceleration after z ~ 0.7 [memory, high]. The supernova references cited for the discovery are news items, not the discovery papers. The model's age (10 Gyr) conflicts with its own Hubble Ultra Deep Field figure caption "The 'Big Bang' is NOT observed". I could not verify the 1995 text; Investigator's worldCat route (LCCN etc.) not completed. c. Tested/replicated. Neither the "top quark" nor the "acceleration" prediction can be assessed without the 1995 edition: the file's dated 2018 text compares to 2002-era values and a CDF single channel. Booker reports reproducing Eqs. 32.1-32.114 from the equations (arithmetic only). d. If true. GR (Einstein's equations) would be wrong in the weak and strong field regimes, and Lambda-CDM would be replaced by an oscillating universe with period 9.83e11 yr; lepton masses derived from alpha would indicate a deeper law. The first-half of Ch.32 (Schwarzschild metric, perihelion, light deflection) was not read. Verdict (split). Cosmology numbers (H0, T_CMB, B-modes, CMB peaks, density): Tested, not supported. Particle masses (tops/Higgs): Tested, not supported (8-9 percent top mass; Higgs 2.8 percent); lepton and nucleon ratios agreeing to 1e-6 to 1e-8: Never properly tested (possible numerology; no count of alternatives). The claim to have predicted acceleration and the top mass in 1995: Never properly tested (source edition not examined). GR and Schwarzschild derivation: not assessed (not read).

8. Claim 7: collective phenomena are classical

Source note. Digest only (Ch.24-30). This is a verdict on the digests, not on deep reads. a. Mathematics. Ch.24 re-derives Maxwell-Boltzmann, Bose-Einstein and Fermi-Dirac distributions, the Planck law, Wien, Stefan-Boltzmann, Debye specific heat in textbook form, using h and photon counting in cavity modes with "Einstein-Bose statistics" [read: digest p.1536-1543]. These are standard results, not classical-only: the Planck law's derivation needs quantised energies hnu and occupancy f = 1/(e^{hnu/kT} - 1). Ch.25: superconductivity Tc from a "band-pass filter" of a dipole array: calculated Tc 30.8 K for a 3D metal (density 1e29 per m^3) versus 23.2 K for Nb3Ge, and for oxides 7 K (3D), 22 K (2D), 180 K (1D) [read: digest p.1554]. Offering three dimensionalities for the same material class leaves a free choice per comparison. Ch.26: R_H = h/(n e^2) derived with flux quantum Phi0 = h/2e and integers n inserted by linked wells (digest p.1558-1561). Ch.27: Aharonov-Bohm treated as a magnetoresistance effect [digest p.1562-1564]. b. Experiment [memory]: the BCS isotope effect, the energy gap, flux quantisation in units h/2e, Josephson effects, and the quantum Hall plateaus quantised to ~1e-10 and used as the resistance standard are established; Aharonov-Bohm interference with fully shielded flux has been shown (Tonomura) [memory, medium]. The book's classical derivation reproduces a number (h/e^2) only by supplying h and the integers. Whether it accounts for the isotope effect and the gap spectrum is not in the digests. c. Tested? No test beyond Mills's own comparison; in particular the Tc predictions have no stated predetermined dimensionality. d. Condensed-matter theory would need replacement. Verdict. Never properly tested (provisional, digest-based; not read in depth). The quantum Hall derivation as printed supplies h and the integers as input, which weakens it, but I have not confirmed this on the page images.

9. The Investigator's findings, kept in the record

• Newer edition (February 2025, 1,888 pp, modified 2026-04-28) served at the same file name; p limit now 126. Changes between editions are evidence of where the theory is still moving.
• Hagen-Mills 2022 (read) and 2024/26: independent-lab EPR with the limits above; strongest outside-lab datum on record.
• Company-hosted calorimetry 2014-2021; NASA 1996, Thermacore 1994 (Lehigh ESCA 55 eV), EarthTech 1997 negative; Booker's chapter-by-chapter checks (arithmetic confirmation).
• Related mainstream-formalism work: Naudts 2005 (Klein-Gordon), Khelashvili-Nadareishvili 2015 (Dirac), Giri 2008 (graphene with AB flux) get tight-bound states with singular solutions; Dombey 2006 and de Castro 2007 say these are unphysical. This is the nearest mainstream relative of Claim 2; it supports the point that "states below 1s" is a legitimate mathematical question, and is answered "no physical" by the finite-nucleus argument [read: files/arXiv-Dombey-2006; web arXiv physics/0608095].
• Rathke 2005 (New J. Phys. 7, 127) is the main published theory critique; Mills's 165-page reply (2008) is not weighed because I did not read it.
• USPTO/UK refusals: "contrary to known law" decisions by officials without experimental review; I give them no weight as physics, only as history.
• Absence of evidence: see section 3.
• Unread leads: Jansson 1997 thesis, Phelps-Clementson and Lawler-Goebel critiques, Hagen-Mills EPR supplementary file (Fig. S6, controls), Harvard CfA report on the continuum, IIT Kanpur results. Any of these could shift the Claim 2 verdict.

10. Corrections to Step 1 (earlier documents)

1. claims.md and deep read (pdf pp.1228-1232) on the Bruker replication. Already corrected by the Investigator: the 2022 paper says "two different spectrometers in a lab of the spectrometer's manufacturer (Bruker, Billerica)" citing Mills's company PDF. I confirm [read: paper p.5]. Add: the book's phrase "average error 0.097 G" differs from the paper's own "0.004 +/- 0.550 gauss over the 11 lines" [read: paper p.8]; the paper also says the fluxon separation "is not a constant" and "the number and relative intensities of fluxon lines... are presently not understood", a weaker result than the book's "confirmed" and "exactly maintained". The orchestrator should reflect both in claims.md.
2. claims.md claim 2 lists "p up to 137". Out of date for the inventor's current position (2025 paper: p <= 126). Not an error for the project copy.
3. Deep read / digest g-factor "eleven-figure agreement". claims.md writes g/2 = 1.0011596521 "eleven-figure agreement" as given by Mills and flags it. I add the reason: the 1987-era alpha has uncertainty producing ~7e-9 in g/2, so the book's comparison is not an eleven-figure test; with today's alpha it is 2.9e-10 off (2,000 sigma). Both in the deep read; keep as "Tested, not supported".
4. Aspect point. claims.md says "the 'classical' curve (42.23) is the QM prediction formula of the Aspect paper (medium-confidence memory)". My web check found the paper's abstract says its results agree with QM predictions, but I could not confirm the numerical values 5.8 and 5.72 on the page. Keep as medium confidence; the conclusion (Bell test supports QM) does not depend on it.
5. Deep-read file truncation. The Ch.16 (pdf 1214-1252) deep read's "Corrections" bullet is cut off by a later writer (stated in claims.md). I have not redone it. The 1.68-times mismatch for [H2(1/4)]2 dimer rotation (44.30 vs 26.43 cm^-1) is intact in that file.
6. Possible new correction. Deep read Ch.5 states the Lykke value as 0.754195 and the book prints 0.75418; the book's "agreement" is therefore at 1 sigma, not zero. Verified in the deep read's page images; no correction needed to our files.

11. Web checks done (budget about 6)

1. Perseus 3.5 keV line: Hitomi and Dessert-Rodd-Safdi results and disagreement (Abazajian) [web: https://www.repository.cam.ac.uk/handle/1810/276029; https://arxiv.org/abs/1812.06976]. Retagged from [memory, medium].
2. Floyd J, [2008] EWHC 2763 (Pat): appeal allowed and remitted [corrected 2026-10-09 after QE: was "confirmed as dismissed appeal"]; reported for burden/standard of proof; judgment text not read [web: https://caselaw.nationalarchives.gov.uk/ewhc/pat/2008/2763].
3. Hensen et al. 2015 loophole-free Bell: confirmed details, S = 2.42 +/- 0.20, 245 trials [web: https://arxiv.org/abs/1508.05949].
4. Fermilab final muon g-2 2025: a_mu = 0.001165920705(114) [web: https://www.advancedsciencenews.com/fermilab-confirms-muon-g-2-measurements/].
5. Fine-structure constant: Rb 137.035999206(11), Cs-Rb discrepancy about 5 sigma, g/2 = 1.00115965218059(13) [web: https://www.cnrs.fr/en/french-team-has-improved-measurement-fundamental-physical-constant].
6. Aspect 1981: abstract confirmed; specific numbers not found online (stays medium [memory]). Not checked (remain [memory], marked): Planck H0 and r < 0.036; Bullet Cluster argument; Benson group-additivity precision; Phelps/Lawler critiques; Tonomura AB; ab initio atomic ionization energies; BICEP2 withdrawal.

12. Summary table