Is ignoring magnetism a small error for chromium but large for iron, nickel, cobalt?

E216–E219 — pre-registered before any of them runs (2026-09-21 16:2x)

Queued as one pw.x-at-a-time chain behind the fcc references (runs/e216_e219_chain.sh), cheapest first. Each is the first number of its kind in this project.

E216 — the magnetic energy the frame throws away, per element. Cr (bcc, antiferro- magnetic: two sublattices with opposite starting moments), Fe (bcc, ferromagnetic), Ni and Co (fcc, ferromagnetic): nspin = 1 and nspin = 2, three volumes each, parabola minima, ΔE_mag = E_min(spin) − E_min(non-spin), 60/720 Ry, MV 0.02, same pseudopotentials. This turns "the ladder is spin-blind" from a sentence into four numbers. Predictions (PBE literature): Fe −400 to −600 meV/atom (the non-magnetic bcc Fe frame is grossly wrong); Co (fcc) −150 to −250; Ni −30 to −80; Cr (AFM) −5 to −40 (PBE overstabilises AFM Cr, moment ≈ 1 μB). Bars: Fe < −300; Co −100 to −300; Ni −20 to −100; Cr −3 to −60. Any element outside its bar is a finding about the pseudopotential or the cell, not about magnetism.

E216 RESULT — the magnetic blind spot is four numbers now (2026-09-21 18:0x real time)

All four predictions confirmed. The spin-polarised lattice constants land on experiment to 0.3–1 %; the non-spin ones for Fe and Co collapse so far (3.9 % and 2.7 %) that their minima fell below the 0.97–1.03 brackets and are extrapolated — a lower-bracket rerun is queued so those two E_min are bracketed too (the spin minima, which carry the physics, are inside). What it means for the ladder: the non-spin frame is off by 12 meV/atom on chromium, which SOB20 also judged small and neglected; it is off by 60–460 meV/atom on Ni, Co and Fe, which is not a correction but a different metal. Any fcc port containing those three needs a spin-polarised reference set from the first cell, and spec.py now carries these numbers where it used to say "spin-blind".