Mendeleev-D: Quality engineering (Step 4b)

QE agent (Opus), 2026-10-06. Inputs read in full: _text/manifest.json, _textprep.log, all ten digests, deep/A-Chemical-Conception-Of-The-Ether-37bf6d.md, facts.json, claims.md, sites.md, all nine research/ notes, physicist.md, mathematician.md, contrarian.md, arbiter.md, primer.md, example.py. Checked against the extracted text in _text/ and two page images I rendered to /tmp/qe-mendeleev/ (essay PDF p.34; Stewart PDF p.5). Source folders were not touched.

Verdict: MINOR. Coverage, deep reading and the core verdicts hold up. Every load-bearing number I checked matches the text. example.py runs. The defects are: Step 1 errors that the reviewers and arbiter found but that are still in facts.json and claims.md; a few pertinent items from the digests and the Investigator's notes that are missing from the primer (and partly from the arbiter); and some citation slips. None of them would change a central verdict.

---

1. Coverage

• PDF pp.61-100 are a Longmans publisher's catalogue and pp.8, 60 and 101-104 are blank. Digest-only coverage is right for those pages.
• No files were skipped. The 8 OCR pages per copy are front matter, tables and blank pages. The tables were read from page images.
• No defect found.

2. Shortcuts

• The essay (claims 1-5): the whole essay was deep-read, with page images for the tables and garbled equations. No sampling. OK.
• Stewart (2019) (claim 6): facts.json says "digest only". The digest is the only Step 1 coverage, but the contrarian and the arbiter quote Stewart's extracted text directly ([read: _text Stewart pp.3-5], note 12). So claim 6 does not in practice rest on the digest alone. It is a secondary source and claim 6 is not central.
• Defect (minor): facts.json how_read does not record that the text was later read directly, and several Stewart facts in facts.json are still wrong (section 6).
• One Sonnet deep read is requested so that claim 6 and the corrections rest on a file of their own (D1).
• Investigator downloads that no reviewer read directly. None of these is a manifest document. They are listed here because they bear on the "removed from the table" question and on the inventor's own later text:
• the Principles of Chemistry 3rd English ed. (1905) Appendix III, saved as djvu text: nobody checked whether its text or table differs from 1904 (for example, does it print "newtonium", x and y);
• Prometheus Nr 735/737/738 OCR and the Naturwiss. Rundschau excerpt: the "parabola" phrase is known only as Lente quotes it;
• Lente (2019): the PDF is saved, but the reviewers used the Investigator's note. These are remediation reads (R6), not defects in coverage.

3. Missed material (from my own reading of the digests and research notes)

Each item is pertinent and is missing from the primer. Where it is also missing from the arbiter, that is noted.

1. Werner's 1905 table also left spaces below hydrogen. Stewart: Werner "included spaces for two elements lighter than hydrogen and two elements between hydrogen and helium, but this just shows that such ideas were around at the time" [read: _text/mendeleev-predictions-bdcaef.txt, PDF p.6]. So a contemporary chemist, who later won the 1913 Nobel Prize, also entertained sub-hydrogen elements. This is context the reader needs to judge how eccentric the zero series was in 1902-05. - Arbiter: appears only inside open question 6. Primer: absent.
2. Dewar's Belfast presidential address (1902), which Mendeleev cites as close to his idea: "the highest regions of the atmosphere ... must be considered to be the province of hydrogen and of the argon analogues", "only a few steps" from a lighter gas filling space [read: PDF p.51]. This is a contemporary leading physical chemist's statement that Mendeleev offered as kindred support. - Absent from the arbiter and the primer. The primer mentions Dewar only for phosphorescence.
3. Mendeleev's ether programme of 1872-1881, and the spiritualism commission. Stewart says that after December 1871 Mendeleev turned to gas elasticity under a new Commission for Research in the Elasticity of Gases (from 1872), aimed at the nature of the "luminiferous aether". He adds that "his interest in the aether led naturally to an involvement with spiritualism": in May 1875 Mendeleev set up the Russian Physical Society's commission on mediumistic phenomena [read: Stewart PDF pp.2-3, digest p.4-5]. - The primer says only that the 1874 Boyle's-law work "was aimed at the ether question". - The ten-year funded programme is pertinent history for claims 1 and 3. Stewart's link to spiritualism is his own characterization and should be reported as his.
4. 1906 Nobel nomination, lost 5 to 4 to Moissan (Stewart) [read: digest p.6]. Pertinent to how the scientific world rated Mendeleev in the years of the essay. Absent from the arbiter and the primer.
5. Related "element zero" and sub-hydrogen proposals (Investigator priority 3; sites.md, research/2026-10-06-google.md and habr.md): - Antropoff's "neutronium" (1926), described as an element of about 0.1 of hydrogen's weight that hardly combines and penetrates everything, and the June 2026 ZAAC paper "100 Years of Element Zero"; - nebulium (1898) and "protofluorine". The physicist and the contrarian weigh these. The arbiter lists them only under "smaller dark spots". The primer omits them, including from Further reading. That is a retention defect (section 5).
6. Mendeleev's matter / force / mind (spirit) framing. It appears in the English (PDF pp.40-41: "a distinction must be made between matter, force, and mind") and, as the "trinity" opening, in the Russian and German texts (Investigator). It is the inventor's own stated reason for rejecting unity of matter (claim 5). It is absent from the arbiter and the primer.
7. Crookes's "fourth state". Mendeleev says his gas-ether "does not require the recognition of a peculiar fourth state" (Crookes) [read: PDF p.22]. A small point, useful context for claim 1. Optional.
8. Contemporary warnings in the essay itself. Stoney's and Rogovsky's argument that H and He escape Earth's atmosphere, and Mendeleev's own footnote that both were "very soon" found there. The contrarian has this; the primer could use it as a worked caution about the mean-speed escape criterion. Optional.

Digest error noted, not propagated: digest 37bf6d.02 p.49-50 gives the limits of x as "0.000000096 to 0.00000000000053". The printed text is 0.000,000,96 and 0.000,000,000,053 [read: _text line 1719-1720], that is 9.6e-7 and 5.3e-11. The deep read, facts.json, all reviews and the primer use the correct values.

4. Accuracy spot checks (against _text and page images)

Verdict on accuracy: no load-bearing error downstream.

5. Retention check

The test: was pertinent material left out of the arbiter or primer because a reviewer judged it negatively?

• Kept, with fair statement:
• the advocates' "removed / falsified table" claims (Atsyukovsky, Rodionov), with their stake, in the arbiter section 4 and primer section 6;
• the absence-of-evidence analysis, with four competing explanations and the deciding records;
• the Russian additions: newtonium, ether as a mixture, the light halogen, solar energy from ether, doubt of absolute zero;
• the weak form of claim 1, kept open;
• the contrarian's credit point: emanation is a zero-group gas;
• Yarkovsky 1901; Errera and Ramsay 1900; Bensaude-Vincent 1982; Nature 1904 and Scientific American (titles only);
• the Polytechnic Museum brochure;
• the neutrino and electron mass-scale coincidence, stated and weighed rather than dropped.
• Defect (minor): the primer omits the Investigator's priority-3 related work (Antropoff's 1926 neutronium and the 2026 ZAAC paper, nebulium, protofluorine) and Werner's 1905 sub-hydrogen spaces. The arbiter mentions them only in passing. These items are not negative in themselves. They show that the idea of elements below hydrogen recurred among mainstream chemists (Werner 1905, Antropoff 1926). Leaving them out makes the zero series look more isolated than the record shows.
• Defect (minor): Dewar's Belfast address (contemporary kindred view) is in neither the arbiter nor the primer.
• No case was found where material was removed because a reviewer judged it false.

6. Process

• Source tags. Present throughout and generally honest. [calc] is used consistently for the reviewers' own arithmetic. Points to fix:
• Contrarian section 3 item 1: "his 1874 work on departures from Boyle's law at low pressure is confirmed by later work [read: p.11]". That is Mendeleev's own statement ("confirmed by Ramsay and others (although still denied by some investigators)"). It should be attributed to him, not stated as fact under a [read] tag. The primer attributes it correctly ("which he says Ramsay confirmed").
• Arbiter citation slips:
  • "a series of thoughts" and "a hypothesis" are cited as "PDF pp.47, 58-59"; the hypothesis passage is PDF p.39 and nothing relevant is on p.47;
  • Emmens and Fittica are cited as "page image PDF p.9, near printed p.1"; they are on PDF p.17 (_text line 456);
  • Stewart is cited with journal page numbers (p.8, note 12) in some places and with PDF page numbers in others.
• Mathematician section 2.3 item 4 tags Kelvin's estimate "[memory, low]" next to the printed value. Better: "[read] as Mendeleev quotes it; Kelvin's original not found".
• Web checks. The physicist used 4 of 6 calls, the mathematician 2, the contrarian 3 of 6 and the arbiter 6 of 8. The medium-confidence items a verdict could rest on were checked: coronium/Fe XIV, emanation dates, Sirius B, the Oort density, and Fittica/Emmens (secondary summaries, flagged as such). Two items remain [memory, medium], and no final verdict rests on them alone:
• Dewar's phosphorescence mechanism;
• the identification of the Curie flask glow with radon. The arbiter says so for both. Acceptable.
• Verdict scale. Mostly consistent, with three uses that stretch the scale:
• "Established (as consistent arithmetic)" for claim 3a: the scale defines Established as "supported by experiment and accepted theory". Better to state "the arithmetic is correct" as a finding, not a scale verdict, or keep the label with that qualifier explicit (as now).
• Claim 6c, "Not testable as stated" for ether and coronium "as predictions": what is meant is "not a prediction in Stewart's sense". Acceptable with the qualifier, but the primer table should carry it.
• Emmens is left without a scale verdict: justified and explained in the arbiter.
• Primer against STEP 4 and the reader profile.
• All nine sections are present and in order.
• Every symbol is defined: k, T, m, G, gamma, G_s, Z, u. Acronyms are expanded on first use (rms, CMB).
• Textbook chapters are named only where safe: Schroeder ch.1 and ch.6; Griffiths ch.9.
• The worked example uses real numbers and matches example.py.
• Check-yourself questions have answers at the end.
• The primer adds no new verdicts.
• Wording slip: section 1 says the zero series is "below the hydrogen row". In the table, row 0 is above row 1 (hydrogen) [read: page image PDF p.34]. Change it to "a row before (above) hydrogen", or "lighter than hydrogen".
• example.py. Ran with the venv Python, exit code 0, "All checks ran". Parts A-G reproduce every [calc] number in primer section 5. The primer's fourth extrapolation ("quadratic in logarithms gives 135") is not computed in example.py; the primer correctly credits it to the mathematician.

7. Remediation actions

Deep reads (most important first; at most 8)

No further deep reads of manifest documents are needed: the essay was read in full.

Other remediation actions (non-manifest reads and file changes)

• R1. facts.json (Stewart entry):
• replace "identified as ionised iron (by Gruenwald)" with: "Stewart says the line was 'eventually identified as due to a highly ionised form of iron and named "coronium" by Gruenwald' (note 12 is Gruenwald 1887 on water-vapour spectra); the Fe XIV identification is Grotrian and Edlén, 1939-42";
• change "eleven briefly listed predictions" to "ten named in the body (the abstract says eleven)";
• mark "six more blanks between hydrogen and helium" as Stewart's statement, contradicted by his own Fig. 2 (the essay's table);
• update how_read to "digest + direct text reading by reviewers (+ D1)".
• R2. facts.json (essay entries):
• "mean molecular velocity" should read "root-mean-square speed (the mean speed is 1694 m/s)";
• mainstream_conflict: remove "(by Gruenwald)";
• add the coronal line's modern wavelength, 530.3 nm (531.7 nm as printed).
• R3. claims.md, applying arbiter section 7 items 4-6:
• claim 3: the adopted x is the upper edge (9.6e-7) of a range that runs down to 5.3e-11; it depends on the one assumed temperature; it does not use the 0.17 bound; the mean speed is 3.7 times the Sun's escape speed;
• claim 2: x <= 0.17 is a three-point quadratic extrapolation; x and y are printed without weights in rows 0 and 1;
• name forms: Kirpichev, Rogovsky, Nasini/Andreoli/Salvadori 1898, with the English spellings noted.
• R4. arbiter.md, "QE revision" section (no verdict changes expected):
• add Werner 1905 (sub-hydrogen spaces) and Dewar's Belfast address to the claim 2 and claim 1 records as pertinent context, stated as what the sources say;
• add Antropoff 1926 to section 4D with the description from habr.md, not title only;
• fix the citation slips listed in section 6 (pp.47 to p.39; p.9 to p.17; Stewart page numbering);
• note under claim 6 that "Not testable as stated" means "not a table-based prediction".
• R5. primer.md:
• add a short paragraph, in section 4 or in the notes under section 6, on contemporaries who entertained the same idea: Werner's 1905 table with spaces for elements lighter than hydrogen (Stewart), and Dewar's 1902 Belfast address (as Mendeleev cites it);
• add a "Related later proposals" entry to section 9 or the section 6 notes: Antropoff's 1926 neutronium and the 2026 ZAAC "100 Years of Element Zero"; nebulium (1898), later explained as ionised known elements; protofluorine. Mark these as titles or leads where they were not read;
• add to section 4's opening paragraph Mendeleev's funded gas-elasticity programme of 1872-1881 aimed at the ether, and, as Stewart's characterization, its link to the 1875 spiritualism commission (which was set up "to establish science as the source of truth", per the digest);
• add his matter/force/mind framing (PDF pp.40-41) as the inventor's own reason under claim 5;
• add the 1906 Nobel nomination (lost 5 to 4) to section 1 or 4 as context;
• fix "below the hydrogen row" in section 1;
• in the evidence table row 6c, write "Not testable as stated (not a table-based prediction)".
• R6. Investigator files to read before the next review update (Investigator or reviewer task; not manifest deep reads):
• research/files/Principles-of-Chemistry-3rd-English-ed-1905-vol1-2-djvu.txt, Appendix III and the author's preface: does the 1905 English text name newtonium, print x and y in its table, or differ from 1904?
• the 1905 Russian scan's p.25 table and newtonium footnote. The contrarian rendered p.25 but did not save page references for the other reviewers. Record the result in a deep-style note.
• the Lente (2019) PDF: confirm the "aus einer Parabel zweiter Ordnung" quotation and the reconstruction tables, which the mathematician used second-hand.
• Prometheus Nr 737 OCR (the coronium passage) and the Rundschau excerpt: confirm "Newtonium" in the 1904 German.
• R7. Investigator next run (already listed in sites.md; repeated because these records decide the open question): the 8th edition of Osnovy khimii (1906) and the 9th (1927) table pages; Kedrov's 1958 commentary (djvu-books.narod.ru); Bensaude-Vincent 1982 (BJHS); Nature 69:558 (1904); a recheck of rusneb.ru.

What each later file must change (summary)

• facts.json and claims.md: R1-R3.
• arbiter.md: a "QE revision" section (R4). No verdict changes.
• primer.md: R5.
• example.py: no change needed.
• sites.md: no change by QE. The Investigator's next run takes R6 and R7.

Recheck

QE recheck (Step 4b), 2026-10-06. Re-read the current facts.json, claims.md, arbiter.md and primer.md and the two deep notes, and grep-checked each remediation item against them. Only qe.md was written.

The two remaining reviewer-file slips are minor, were not routed to a downstream file, and do not touch any verdict. R6 and R7 are known record gaps, already disclosed in the arbiter and the primer. They are not defects in this pass.

Final: OPEN (2 issues)