Four 3DSC_MP entries derive from retracted papers (MRh2Sn family, Eu)

[dl-eigenart]

While checking 3DSC_MP entries against their primary sources I found four rows whose
only or decisive SuperCon source has been retracted. This is separate from #3 (averaging
and the 5 K dispersion filter) and from PR #4; the fix is different.

The four entries

3DSC formula	3DSC Tc (K)	MP id	SuperCon source	retraction
Rh2Sn1Y1	4.1	mp-865518	Klimczuk et al., PRB 82, 134520 (2010)	PRB 83, 019901 (4 Jan 2011), 10.1103/PhysRevB.83.019901
Lu1Rh2Sn1	2.9	mp-865378	same	same
Rh2Sc1Sn1	2.0	mp-862623	same	same
Eu1	1.375	mp-1057315	Debessai et al., PRL 102, 197002 (2009), 2.75 K under pressure; averaged with a 1960 non-observation (floor 1.3 K, stored as 0)	PRL 127, 269902 (24 Dec 2021), 10.1103/PhysRevLett.127.269902

For the Heusler rows the retraction notice says: "Due to the mislabeling of a rebottled
chemical-starting material, the superconductors originally reported in this paper as Rh
intermetallics are now known to be Pd intermetallics." The rhodium compounds do not
exist; the measured samples were YPd2Sn, LuPd2Sn and ScPd2Sn, which 3DSC already carries
under their own formulas (Pd2Sn1Y1 4.81 K, Lu1Pd2Sn1 2.98 K, Pd2Sc1Sn1 2.12 K).

For Eu the 3DSC value is exactly (0 + 2.75)/2: the SuperCon non-observation "not
superconducting above 1.3 K" (1960) encoded as zero, averaged with the retracted
high-pressure value. Elemental Eu is not a superconductor at ambient pressure.

How I found them

I matched all 7,247 reference identifiers in MDR SuperCon Ver.240322 against the
Retraction Watch database (distributed by Crossref, 72,336 records as of 3 Sep 2026)
and confirmed each hit against the Crossref updated-by field of the original DOI.
Nine retracted papers with 15 rows are in SuperCon; the other seven (Schön's C60/C70
papers, the 1991 Rb/Tl-C60 report, the 2020 and 2023 room-temperature hydride papers,
one KOs2O6 report) do not reach 3DSC_MP, as far as I can see because the formulas have
no MP match or the compound has enough other sources. Full list, script and details are
in the next version of 10.5281/zenodo.22235399
(file S5, match_retractions.py).

Suggested fix

Not a patch to the cleaning script; the information is upstream of it. Two options:

1. In _2_0_clean_SuperCon.py, before aggregation, drop rows whose reference resolves to
   a DOI with a Crossref updated-by entry of type retraction. SuperCon has no DOI
   column, so this needs a one-off mapping of the ~7,000 citation strings to DOIs (title
   search against Crossref works for the ~6,600 with titles); I can contribute that
   mapping if wanted.
2. Short term: remove the four rows above and note them in the changelog.

Happy to open a PR for either, but I wanted the maintainers' view first, since option 1
changes what "SuperCon" means as an input.

S5_retracted_sources_in_mdr.txt

[dl-eigenart]

Follow-up: the status list this rests on is now published, https://doi.org/10.5281/zenodo.22771913. It gives a DOI for 5,814 of the 7,254 reference identifiers in MDR SuperCon Ver.240322 and carries the Retraction Watch status of the source where one exists. The nine retracted papers behind it, fifteen records in all, are listed in S5 of https://doi.org/10.5281/zenodo.22772787.