Datasets:
Discovery proposal: Mn2Se2 (OSC-00689) — unmeasured magnetic-chalcogenide cell of the p4/nmm M2X2 family; downfold request
Discovery track proposal — Mn2Se2 (2MnSe-1, OSC-00689). This is a downfold request, not a rank claim.
This material is in candidates.csv but not in active_challenge.csv, so it cannot be submitted through the Space. Submitting as a PR per the README.
Why this material: an unsampled cell of your own table. Fixing the layer group and stoichiometry of the current top two anchors (p4/nmm, 2:2 TM:anion) and splitting the active members by anion type and magnetism gives:
| anion | magnetic | n | nef range | t range (eV) |
|---|---|---|---|---|
| chalcogenide | No | 9 | 0.461 – 1.419 | 0.128 – 0.259 |
| halide | No | 2 | 0.447 – 0.552 | 0.216 – 0.241 |
| halide | Yes | 3 | 2.391 – 4.081 | 0.022 – 0.053 |
| chalcogenide | Yes | 0 | — | — |
At fixed layer group and stoichiometry, magnetic = Yes raises nef by 4–7x, and it does so through t: the magnetic members have 4–10x smaller hopping, i.e. a band width of 0.18–0.42 eV. This is consistent with the published pipeline using a non-spin-polarized pw.x SCF — in a magnetic material the suppressed exchange splitting leaves a narrow, partly filled d manifold pinned at E_F.
The chalcogenide × magnetic cell has no members at all. Mn2Se2 sits in it. Because that cell is empty, no regression over the published table can predict this material, which is precisely why it is worth measuring rather than predicting.
Why Mn and Se specifically. Mn is d5, exactly half-filled 3d, the strongest form of the condition above; Mn already accounts for several high-ranking magnetic active entries (1CF2Mn2-1 39.554, 1C2Mn3-1 37.085, 1GeS2Mn3-1 32.141, 1MnTe2-1 23.232). Se rather than S because the heavier chalcogen lengthens the metal–metal separation and narrows the band, which is measured in-family: Fe2Se2 (0.793) > Fe2S2 (0.560) and Ni2Se2 (1.308) > Ni2S2 (1.129). The cell has 4 atoms, and since nef is reported per atom a small cell is favourable.
Flags. e_hull 0.038 (the more stable of the two C2DB Mn2Se2 polymorphs, the other being 2MnSe-3 at 0.114), gap_pbe 0.0, is_metal True, magnetic Yes, sc_relevance 0.75, layer_group p4/nmm, 4 atoms. No Mn2Se2 polymorph of any layer group appears in active_challenge.csv. Source: C2DB, CC-BY-4.0.
On the claimed numbers — these are a floor, not an estimate. I cannot predict this material and I am not pretending to. nef = 0.7403 is the minimum nef among the active p4/nmm M2Se2 members (2ScSe-2), all of which are non-magnetic, and A_d = 0.03844 is the lowest value on the dome plateau (Ne = 36, delta = 0); together they give 8.538. Since the magnetic column of the table above is systematically higher, the real value should exceed this floor, but I have no basis for saying by how much, and I make no claim that this material beats the current top anchor (55.375). Please downfold it with the published pipeline and replace these claimed values with the real ones. No rescaling or empirical fitting to the published reference scores was done.
Reproducible from the published data alone: active_challenge.csv, candidates.csv, scoring_meta.json, pipeline/downfold.py.
Correction to the number I posted in this PR.
I wrote 8.538 here and labelled it a "lower bound". It was not a bound — it was a regression point estimate for a cell of the table that my regression had no training examples for, and I should not have presented it as a floor. Having reproduced the canonical protocol of #58, the computed value is:
| quantity | value |
|---|---|
| N(E_F) | 1.48563 |
| n_band | 1.1293 |
| delta | 0.1111 |
| A_d | 0.04726 |
| canonical score | 21.063 |
In this case the figure I posted happened to sit below the computed value, so the claim was not misleading in its direction. The method behind it was still wrong, and the same method produced a figure that was far too high on #77 — so I am replacing the label with a calculation here too.
21.063 is well below the verified leaders, and I am not claiming otherwise. Reproduction evidence is in my comment on #74. I understand that ranking uses your own downfold, not my number (#5, #46).
Thank you for the correction and the computed value. Mn2Se2 (OSC-00689) is queued for canonical downfolding; the result will be posted in #49.