Search ENSDF for the photon keystone: level restoring pairs and the veto gateway

[MaxFreedomPollard]

Closes #12.

What this adds

A new directory, photon/, that runs the search the theory's Section 2 and Open Problem 1 asked for, and the second search the neutron gate (#11) made necessary, over the adopted level schemes of every nuclide that carries an isomer holding a bit for at least a second: 415 isomers from NUBASE2020 matched to ENSDF levels, fetched from the IAEA Live Chart and packed into a committed file so the census reproduces offline in seconds with numpy and matplotlib alone.

What it measures

• Gateways. Every level above every isomer that a photon or a captured electron could lift it into, in two classes: observed (ENSDF lists the line) and allowed (spin and parity permit E1, M1 or E2 and the line is unobserved, the class the 4.85 keV gateway of ⁹³ᵐMo belongs to). Each gateway's release cascade is followed through the adopted gammas with conversion coefficients: 2451 observed, 2747 allowed, 1802 of them release. The famous case reads back at leverage 501 against the theory's 500.
• Level restoring pairs. Every release line against every releasing gateway absorption, with ENSDF's own recoil convention (a gamma energy is the photon's; the level difference carries the recoil), both thermal Doppler widths, a rotor of up to 1 km/s (Moon, 1951), and the quoted uncertainties of the three energies: 12470 coincidences compatible within one standard deviation, 12402 heterogeneous, 3775 closed loops, and 0 that the data are precise enough to call resonant, because ENSDF quotes most energies to hundreds of electronvolts and a rotor closes a few (the median candidate is 475 times too coarsely known).
• The areal density wall. For every candidate, the inverted areal density the amplification condition demands from the gateway's integrated cross section, with measured widths kept apart from Weisskopf estimates. The least demanding candidate with a measured width (¹²⁷Ba feeding ⁵²Mn) needs 1×10²⁴ inverted nuclei per cm², 23 cm of a solid made entirely of the isomer.
• Inhibition. 214 isomers with a signal gateway and a veto gateway, two releasing paths whose cascades differ in whether they emit the signal line: the operation the neutron sector could not perform, native to isomers.
• The NEEC class. 34 releasing gateways within 30 keV of their isomer, ranked: the Phase B1 target list.

What it changes in the theory

Section 2's convertible class is now a table instead of a conjecture, and every entry fails Section 1.1's areal density wall by orders of magnitude whether or not it is resonant, so the photon amplifier must be engineered rather than found, which is what Phase B1's experiments have been for. Inhibition, which the neutron result added to the keystone's requirements, is available in the photon sector without new physics and comes off the list of things the keystone must be discovered to have. And the first concrete experimental request the search produces is not a cross section but three energies to an electronvolt.

Files

• photon/ensdf.py, photon/census.py: the fetcher and the census
• photon/levels.json.gz: the packed level schemes (committed; the raw cache is ignored)
• photon/gateways.csv, photon/pairs.csv, photon/inhibitors.csv, photon/results.md, photon/README.md, figures/fig16_photon_keystone.{svg,png}
• README.md, theory/THEORY.md: a new section, Phase B1 and the deliverables table updated, Sections 1.1 and 2 of the supplement corrected in place

Not in this PR: the CI line

python photon/census.py belongs in .github/workflows/build.yml beside the neutron report step; it needs no network and no OpenMC. The token available to me still carries repo but not workflow scope, so that line and the two workflow files from #11 remain pending on gh auth refresh -s workflow.

Provenance

IAEA Live Chart of Nuclides (ENSDF adopted levels and gammas), NUBASE2020 (already cached in gates/data), Weisskopf single particle estimates only where ENSDF gives no width and labelled as such; Moon's 1951 rotor for the compensation reach; Fröhner is not involved here.

🤖 Generated with Claude Code