Experiments · E82

Do the four earlier verdicts survive the refitted energy model?

Yes. All four still pass; the two controls reproduced exactly, and the ordering temperatures moved less than predicted, by −37 to +133 K.

In the log: Re-measuring E74's ladder under the anchored expansion: prediction

mixedDate not stated in the log; it was written between the commit of 2026-09-13 08:16 and the first commit that contains it, 2026-09-16 02:04unclassified0 predictions · 0 result paragraphsEXPERIMENTS.md lines 4500–4529
exp E82 diagram
What E82 did and how it came out, drawn from this record and the files it names (book/assets/diagrams/exp/E82.svg).

Results

No result paragraph for this entry was found in the log.

The full record

EXPERIMENTS.md · lines 4500–4529

E82 — Re-measuring E74's ladder under the anchored expansion: prediction

E80 refitted the expansion with the eight elemental corners and withdrew every ordering temperature in this file. E74's four verdicts rest on them, so they are re-measured here on the same four compositions, the same settings and the same seeds - only the expansion has changed.

The control matters more than the headline. The ladder mixes two models, and only one of them moved. The driving force comes from a MACE relaxation against a MACE-relaxed hull; the activation energy comes from MACE vacancy formation and MACE nudged-elastic-band barriers. Neither touches the expansion. Only T_od is a Monte Carlo over the expansion.

Predicted:

  1. drive_cold reproduces to within 2 meV/atom on all four. No expansion is involved.
  2. Q reproduces to within 0.05 eV on all four. Same reason.
  3. T_od moves by more than its old error bar on at least two of the four. The spot checks in E80 moved MoNbTaW 501 to 635 K and Mo0.62Ta0.38 1077 to 1342 K.
  4. The scatter on T_od falls on at least three of four. In E80 it roughly halved - +/-182 to +/-79 and +/-202 to +/-45 - which is the argument that the anchored fit is better rather than merely different.
  5. The verdicts survive, all four still above P = 0.9, because P is carried by the kinetic discount, and the kinetics are MACE.

A failure of 1 or 2 is the important one. It would mean something moved that the refit should not have been able to touch, and that I have the pipeline wrong rather than the physics. A failure of 5 would be a real change of answer and would need E74 rewritten rather than re-measured.

The prior-expansion results are kept at runs/odt_map/generated_verified_priorCE.jsonl.

Related entries

Built with PRISMWebsite and visualizations made using Claude