Becquerel-Henri: Quality Engineering (Step 4b)

QE agent: Opus. Date: 2026-10-03. Inputs read: _text/manifest.json, all three digests, facts.json, claims.md, physicist.md, mathematician.md, contrarian.md, arbiter.md, primer.md, example.py. Spot checks were made against the extracted text in _text/ and, for Nature p.398 (PDF page 3), against a page image I rendered at 300 dpi into the scratchpad (nothing was written in the source folder).

Verdict: MINOR. Coverage is complete and no central verdict changes. There are four accuracy defects: two misread source statements in claims.md and facts.json, a misquotation repeated by three later files, and an overstated "10^5" energy-gap figure. All four can be fixed by editing facts.json, claims.md, arbiter.md and primer.md.

---

1. Coverage

• Every manifest document has a digest and a facts.json entry. No document was skipped.
• There is no deep/ folder. This is not a defect here: all three documents are short (3, 6 and 12 pages), and Step 1 and all three reviewers read them in full (Nature from page images, the others from the text layer). The digest was not the only reading of any central claim.
• One stale entry: facts.json says the Nature file has text_available: none and that only about 99 characters were extracted. Step 0 has since run OCR: the manifest shows 3 OCR pages and 23,028 characters. The OCR prose is good, but it mangles exponents ("1.6 x 10"", "4.107"", "1077"), so the page-image reading is still the right source for the numbers. facts.json should record both.

2. Shortcuts

• No sampling. Each document was read in full, so no central claim, experiment or derivation rests on a digest alone.
• The digests themselves contain errors. None of these errors reached the later files, but no later step should rely on the digests.
• The Nature digest gives the OCR-garbled exponents: velocity "1.6×10⁸ cm/s", m/e "10⁻⁷" and current "4×10⁻⁷ CGS". The page image reads 1.6 × 10^10, 10^-7 and 4 × 10^-13.
• The Harper's digest says Pierre Curie carried radium in his pocket. The text says Curie applied a tube to his arm for an hour and a half, and Becquerel carried radium in his pocket (_text lines 306 and 317).
• The Harper's digest says "fewer than 3 grams" of radium chloride existed. The text says "less than a gram exists" (line 272).
• The Harper's digest lists "magnetic deflection" among the properties of undeviable rays.

3. Missed material

These are items in the sources that facts.json, claims.md, the reviews, the arbiter or the primer missed or got wrong.

1. Becquerel says the non-deviable rays are LESS penetrating, not more [read: Nature p.397, _text line 329: "The non-deviable rays are much less penetrating than the deviable rays; the polonium radiation is limited in air to a kind of sheath of some centimetres"]. He then reports Villard's separate discovery of "very penetrating rays which are not capable of deviation" (line 333). - In other words, by 1901 Becquerel had in effect three components: deviable beta rays, weakly penetrating non-deviable alpha rays, and Villard's penetrating non-deviable gamma rays. - claims.md claim 3 and facts.json (Nature key_claim 3) both say "a non-deviable, more penetrating part", and the arbiter's heading for claim 3 repeats it. None of the four reviews catches this. - The verdict does not change, but the claim misstates the primary source. It also hides an instructive point: Becquerel was nearer the alpha/beta/gamma picture than the 1902-03 popularisers who said "non-deviable = X-rays".
2. Radium salts "increase spontaneously in activity for some time after their preparation" (the Curies and Giesel) [read: Nature p.397, line 265]. This is radon ingrowth observed before anyone understood it. - Primer Check-yourself Q2 asks the reader to work out exactly this rise, but the primer never says that the 1901 source already records it. The contrarian mentions the observation in passing. It would make a good link for the reader.
3. Uranium regained its activity after twelve months [read: Thomson, line 195]. After Becquerel moved uranium's activity to barium, the inactive uranium solution was active again about twelve months later. This answers Becquerel's own 1901 suggestion that uranium's activity might be "in great part, if not altogether" due to actinium: the activity regrows, so uranium produces it. - The contrarian cites the twelve-month regain. The primer's "Is uranium itself the source?" row does not use it, though it is the [read] evidence for that row's conclusion.
4. Thomson's hint about helium [read: Thomson, lines 261-263]. He says helium always occurs in radioactive minerals and suggests helium may be a product of radioactivity. facts.json notes this "link to helium". The primer, which explains alpha = helium only from [memory], does not mention that a 1903 source in the folder already pointed that way.
5. Two radioactivity findings from Elster and Geitel and Thomson [read: Thomson, lines 295-302]. Freshly fallen rain is radioactive for about an hour, and the Earth's own charge is enough to make points (lightning rods) radioactive. Both are direct evidence of airborne radon progeny, and they support the 5a/5b resolution. Not used.
6. Giesel's "polonium" was deviable and the Curies' was not, so they were different substances [read: Nature p.397]. This is an early sign that chemical separations of 1899-1900 were mixtures. It is a minor point that no file mentions.
7. Radiation burns. Harper's [read] reports Pierre Curie's deliberate 1.5-hour arm exposure, which left a sore lasting more than three months, and Becquerel's pocket burn. facts.json records both. The primer omits them. This is optional history, not a defect.

4. Accuracy spot checks

Load-bearing number that is overstated

The reviews, the arbiter (claim 4c) and primer §5 Step 6 all state that Becquerel's "no contradiction" argument undercounts radium's output by "about 10^5". That ratio divides the Curie-Laborde heat of pure radium (0.12 W/g, 1903) by a screen current that came from a 1900 preparation of unstated purity. The same article says the Curies' products of that period were only "several thousand times" as active as uranium; pure radium is about 3 × 10^6 times as active per gram (3.7 × 10^10 versus about 1.2 × 10^4 decays per second per gram; [memory: high]).

So the 10^5 is an upper-end figure. The verdict "Refuted by calculation" for 4c still stands on a sounder argument:

• in equilibrium, alpha decays release about 24 MeV per radium decay, against well under 2 MeV of beta energy, part of which goes to neutrinos [memory: high];
• most of the alpha energy is absorbed in the source, where the screen cannot see it;
• Becquerel's reasoning from mass to energy is mistaken whatever the numbers are.

The arbiter and the primer should present the gap as "several orders of magnitude (up to about 10^5 if the sample were pure radium)" and rest the verdict on the alpha-energy argument.

5. Process

• Source tags. Present and honest throughout. Page-image readings are tagged [read: page image p.N]. [memory] items carry confidence levels. The physicist correctly says that the chemlin figure came "via search summary".
• Web checks.
• Done: Curie-Laborde heat, field-emission threshold, Rutherford-Soddy, Kaufmann, Elster-Geitel.
• Still [memory: medium] and unchecked: atmospheric ion density, fair-weather field of about 100 V/m, and decay-rate independence underground. The arbiter states that verdict 5b does not depend on any of them being exact (it has 10^3 or more of margin), which is acceptable.
• The Comptes Rendus volume 122 citation in the primer is tagged medium; acceptable.
• Verdict scale.
• Applied consistently, and split verdicts are used where appropriate.
• The arbiter sensibly resolved the spread on 4b (space energy): Historically superseded, never directly tested.
• "Refuted by calculation" for 5b comes with the calculation shown, as the rule requires.
• 4c is a sub-claim the arbiter introduced, not one listed in claims.md. That is acceptable, but claims.md should be updated so the numbering matches the primer's evidence table (2a/2b, 3a/3b, 4a/4b/4c, 5a/5b).
• Arbiter corrections not carried back. The arbiter's "Corrected claim list" (claims 2, 3 and 4, and the 100,000 attribution) was never applied to claims.md or facts.json. Those files still say "independent of chemical form", "no noticeable decay", "more penetrating" and "100,000 (Nature)".
• Primer.
• Follows Step 4: all nine sections are present and in order, with answers at the end.
• Written for the reader profile: emu units, rigidity, gamma, the decay ODE and Q = Δm c^2 are each introduced before use, and every symbol is defined.
• The worked example uses real numbers and matches example.py.
• The textbook chapter references are given without chapter numbers, which complies with the "only if certain" rule.
• example.py. Runs cleanly with .venv/bin/python (exit 0). The output matches primer §5 (v = 0.534 c; β = 0.471 / 0.684; 1.26 mg; 5.12 erg/s/cm^2; 528 m; 0.029 and 0.143 W/g; ratio 2.3e5).

6. Remediation actions (one round)

Deep reads. Optional confirmatory reads; QE has already verified the key passages from page images and text.

1. Nature-Feb-1901-184ae1, pages 1-3, Sonnet. Write deep/Nature-Feb-1901-184ae1.md from page images (the OCR exponents are unreliable). Record: - each quantitative statement, with exact wording; - the penetration ordering of the deviable rays, the non-deviable rays and Villard's rays; - the exact "if not altogether" sentence; - radium's rise in activity after preparation and polonium's decay; - Rutherford's acid-removal observation.
2. J-J-Thomson-Becquerel-Rays-2f40a7, pages 3-5, Sonnet. Write deep/J-J-Thomson-Becquerel-Rays-2f40a7.md covering: - the twelve-month regrowth of uranium activity after barium separation; - the Rutherford-Soddy thorium-X test; - the helium remark; - the rain and lightning-rod findings of Elster and Geitel; - Thomson's exact wording of the ion-layer hypothesis and any data he gives.

facts.json

• Nature entry:
• set text_available to "partial (OCR text; exponents garbled, numbers read from page images)";
• change key_claim 3 to "a less-penetrating non-deviable component (later alpha), plus Villard's very penetrating non-deviable rays (later gamma)";
• change key_claim 5 to "several thousand times uranium (Nature); 100,000 is the Harper's figure";
• add the "if not altogether" actinium hedge, with its exact wording, and the observation that radium salts increase in activity after preparation.

claims.md

• Apply the arbiter's Corrected claim list, and renumber the sub-claims to match the primer (2a/2b, 3a/3b, 4a/4b/4c, 5a/5b).
• Claim 3: replace "more penetrating" with Becquerel's actual ordering, and attribute "like X-rays" to Duncan and Thomson.
• Claim 2: quote "if not altogether".

arbiter.md. Add a "QE revision" section that:

• corrects the misquote ("if not altogether"), noting that it strengthens Becquerel's 1901 doubt that uranium was the source, and that Thomson's twelve-month regrowth resolved that doubt;
• corrects the claim 3 heading on penetration and notes that Becquerel in effect already separated three components;
• restates 4c as "several orders of magnitude (up to about 10^5 for pure radium)", with the verdict resting on alpha energy absorbed in the source. The verdict stays Refuted by calculation;
• records that no verdict changes.

primer.md

• §4 table, "Non-deviable rays" row: add Becquerel's "much less penetrating" and Villard's penetrating rays.
• §4 "Is uranium itself the source?" row: cite Thomson's twelve-month regrowth [read].
• §5 Step 6 and §7: soften "five orders of magnitude" to "several orders (up to about 10^5 if pure radium)", and add the 24 MeV alpha versus under 2 MeV beta argument.
• Check-yourself answer 2: note that the Curies and Giesel saw this rise [read: Nature p.397].
• Optionally, §3: note Thomson's 1903 helium hint.
• Remove the parenthetical "Earlier review files quote this line as 'though not altogether'..." once the arbiter has been corrected, or keep it as a pointer to the QE revision.

example.py. No change is required. Optionally, print the Step 6 ratio with a note that it assumes pure radium.

No redo of Steps 2-4 is needed.

---

Recheck

Recheck agent: Sonnet. Date: 2026-10-03. Checked the current claims.md, facts.json, arbiter.md, primer.md, example.py and the new deep/ files against the defects and remediation actions above.

Defects

Remediation actions

Remaining optional (not defects): the digests still contain the errors noted in section 2; the Elster-Geitel rain and lightning-rod points are not confirmed in the primer; Step 2 review files keep the old misquote.

Final: PASS