Run the ampoule in real photon transport: budget, gate, synapse, boundary and dose

[MaxFreedomPollard]

Closes #14.

What this adds

A new directory, ampoule/, that runs the sealed machine of transistor/SEALED.md and transistor/EMBODIMENT.md in real photon transport: the vessel shell by shell in OpenMC 0.16 with the official ENDF/B-VIII.0 photoatomic library, driven by the ⁹⁰Sr and ⁹⁰Y beta spectra of its own core through OpenMC's thick target bremsstrahlung treatment, calibrated first against references that are not OpenMC (NIST XCOM attenuation to 0.4 percent, slab transmission to 1.0 percent of the exponential, thick target bremsstrahlung within the textbook rule's tolerance, beta spectra within 2 percent of the ICRP 107 mean energies).

What it measures

• The source. 43.9 M/s of bremsstrahlung leave the capsule, 0.0569 per decay at a mean energy of 252 keV; the vessel deposits 183 µW against the ledger's 181 µW per GBq.
• The compute shell. 32.9 k/s of interactions across the 64 sites as 2 mm plastic cells, 0.033× the proposal budget of SEALED.md Section 8; the same cells in CsI see 1.02 M/s, 31× more, which meets the budget.
• The GATE. The tungsten collar's transfer curve from closed to open: ON/OFF contrast 6.34 with plastic cells, 11.94 with CsI.
• The synapse. The 64 × 64 Green's function between sites (Open Problem 4): nearest neighbour weight 3×10⁻⁵ per photon, total fan out 0.1%, 3×10⁴ of attenuation per hop; the lead septum leaves the coupling it sits across at 0.49 of its open value against 1.00 on the unshielded side.
• The boundary. 3.25 M/s detected, 0.33× the note's 10⁷ per second, from 7% of core photons; 50 eight bit reads per second by the precision law, not 153.
• The dose. 67.8 µSv/h at the titanium wall, 0.261 µSv/h at one metre (ICRP 116), from 3.4 M/s escaping the shield, 7.7% of what left the core.

What it changes in the notes

Three of the design's numbers move and one holds. The power ledger holds. The cells must be CsI: in plastic the shell is 30× short of its proposal budget. The mouth is a factor 3.0 narrower than it was priced, so the 8 bit read rate is 50 per second rather than 153 at the present collection. And the two mechanical controls, the aperture and the septum, are analog knobs rather than switches, because the bremsstrahlung is hard and 2 mm of tungsten or 1 mm of lead is a filter for it: the same finding as the cadmium sheet of the neutron gate (#11), for the same reason. The dose figure the build note deferred is now on the table. SEALED.md and EMBODIMENT.md are amended in place, the README gains a section, and Open Problems 4 and 7 are updated with what is now measured.

Files

• ampoule/beta.py, ampoule/calibrate.py, ampoule/model.py, ampoule/report.py
• ampoule/tallies.json, ampoule/calibration.json: the committed run (7.6 minutes of wall clock)
• ampoule/results.md, ampoule/README.md, figures/fig17_ampoule.{svg,png}
• neutron/data.py: the fetcher now also takes the photoatomic files and the natural isotopes of the vessel's elements from the official archive
• README.md, transistor/SEALED.md, transistor/EMBODIMENT.md

Not in this PR: the CI line

python ampoule/report.py belongs beside the neutron and photon report steps in .github/workflows/build.yml; it needs no OpenMC. The token available to me still lacks workflow scope, so it joins the pending lines from #11 and #13.

Provenance

OpenMC 0.16.0 (conda-forge); ENDF/B-VIII.0 official OpenMC HDF5 distribution, photoatomic data EPICS2014 based, streamed from the 3.4 GB archive; NIST XCOM (Hubbell and Seltzer) for the attenuation references, transcribed; ICRP 116 anterior posterior photon coefficients as shipped with OpenMC; ICRP 107 mean beta energies; the thick target bremsstrahlung treatment validated by Lund and Romano (2018).

🤖 Generated with Claude Code