back to table · edit · history · where entries came from · files · download
3874 bytes, as of the version from 2026-10-25 06:53 (current). Recorded here, not run.
# T476 build notes
This is a measured transcription of the eight values in the “here” row of Hasenbusch's Table III (arXiv:1004.4983v3), not a fresh Monte Carlo computation. The selected range is that complete, recognizable set. All centres and uncertainties are preserved. No 100-digit precision claim is made.
Conventions fixed before checking: three dimensions, short-range isotropic bulk Ising criticality, a pure ordered phase, and high-temperature numerator for each cross-phase ratio. The second-moment and exponential correlation lengths remain distinct. Q-plus and Q-minus are alpha times the specific-heat amplitude times the cube of the second-moment length amplitude, not cube roots. The reduced free-energy derivative fixes the heat-capacity metric when the thermal coordinate changes. The ratio parameter enumerates the components of the universal bulk critical-amplitude data requested for this table; no Cartesian product of separate observables and classes is introduced.
Files:
- generate.py reproduces the entries and verifies T476; --fill writes only the existing table's entries and attachments.
- checks.py runs the transcription, interval, identity and literature checks. --stored repeats them on the API document.
- check.py is an unchanged copy of the campaign's shared checks.
- source-data.json holds the separate parenthetical transcription and independent comparison data.
- source-manifest.json records exact source versions, download addresses and SHA-256 digests.
- numerical-checks.json records the checks on the locally generated values.
The primary Table III and the comparison Table 1 were inspected visually in their PDFs. The primary formula definitions were inspected as well. All eight transcribed uncertainties agree exactly with the print. The two amplitude identities admit one exact rational point in all eight quoted intervals simultaneously. Positive controls, deliberately wrong identity controls, and an independent thermal/field/length rescaling control pass. The table holds eight entries; the longest number is 19 characters and the entries including comments occupy about 1.1 KB.
The independent comparison is not a validation of the paper's Monte Carlo uncertainty estimates. Six pairs of quoted intervals are disjoint: U2, Uxi, Q-plus, Q-minus and Qc against Campostrini et al. (2002), and u-star against Caselle and Hasenbusch (1997). All separations are less than three root-sum-squared quoted uncertainties. These measured uncertainties include systematic assessments; no simultaneous confidence level is inferred. The cause of the shifts between separate studies is not determined here. No contradictory result was silently discarded, averaged, or given a larger uncertainty.
The default command verifies all entries without writing. While the table is private, verification requires an authorized API key. The build does not publish or request review. T472, linked for critical exponents, is also a draft at build time, so it is a publication dependency for a later human review.
Final read-back: all eight entries matched the generator, with zero differing, missing or extra entries. All independent checks passed on the stored values, with the same six literature discrepancies recorded. The authenticated site audit returned clean=true and no findings. stored-checks.json and audit.json contain these results.
Implementation notes: the table API normalises Numbers into a mapping for this one-parameter table, even when submitted as flat records. The checks accept both shapes. The shared stored() helper requires Django, so these checks read the authenticated API instead. All consistency arithmetic uses Python Fraction exactly; this avoids an uninitialised Sage coercion path encountered when mixing Sage rationals and Python integers. No measured value changed during these local checker fixes.