Jorge Hirsch, the h-index's inventor, used data-pattern forensics to help force the retraction of Ranga Dias's room-temperature superconductivity claim

research-integrityfraud-detectionphysicscondensed-mattersuperconductivityhirschretraction

In 2020 physicist Ranga Dias and collaborators reported room-temperature superconductivity in carbonaceous sulfur hydride (CSH) under high pressure, a celebrated result published in Nature. Jorge Hirsch — the condensed-matter physicist who a generation earlier had invented the h-index and whose own field is superconductivity theory — became a leading critic, joined by Dirk van der Marel of the University of Geneva. Their argument was forensic rather than experimental: the published magnetic-susceptibility data fit an implausibly smooth curve for what should be noisy laboratory measurements, and per Chemistry World, "the duo stated" it was not possible to produce the team's results "without using procedures that can only be characterised as data alteration and manipulation" — a joint finding, not Hirsch's alone. Nature retracted the paper in September 2022, citing its own undisclosed background-subtraction procedure as the specific mechanism (see claim-nature-2022-csh-retraction-cites-undisclosed-background-subtraction); Hirsch and van der Marel's own technical papers ground the same mechanism independently (see claim-hirsch-modeled-csh-critique-on-2009-debessai-europium-precedent and claim-hirsch-van-der-marel-2022-csh-data-pathological-spline-quantized).

The critique carried a social cost that outran its technical vindication: arXiv administrators withdrew Hirsch's own CSH-fraud preprint in December 2021, and he was separately, temporarily banned from posting to arXiv in February 2022 for the inflammatory tone of his accusations, even as his substantive analysis held up — see claim-hirsch-arxiv-sanctions-two-distinct-events-dec2021-feb2022, which replaces this note's earlier, undated "~6-month ban" drawn only from Tier-4 Wikipedia sourcing.

This is the same "does the pattern of evidence make sense" instinct pointed in opposite directions — first to reward scientists with a citation number, then to unmask one by showing the numbers were too clean to be real. That "too clean to be real" test is not new to physics: it is the statistical descendant of a rule that runs from Fisher's 1936 verdict on Mendel back to Talmudic law, and it is the same principle by which the verbatim-but-non-independent Gersick quote gets graded down — developed in 2026-07-11-hop-suspicious-perfection. It is a data-forensics cousin of the "try to disprove, not prove" discipline in claim-ach-step-5-instructs-analysts-to-disprove-not-prove, and it belongs to the vault's small cluster on how scientific error does or does not get corrected — compare the propagation-failure case of claim-yellow-rain-bee-feces-refutation-never-formally-retracted, where a decisive refutation never reached the official record. Here, by contrast, the correction did land: the paper was withdrawn.