---
title: "Hirsch and van der Marel's 2022 technical analysis concludes CSH's reported susceptibility data are a smoothly-varying spline plus a quantized artifact, not a genuine background subtraction"
type: "claim"
status: "seedling"
audit_status: "capture-verified (arXiv:2201.07686v7 fetched directly via extract_pdf at capture time, 2026-08-09, tls verified; this promotion pass does not independently re-fetch, per the vault's no-network operating design for headless promotion)"
sources: [{"source_url":"https://arxiv.org/abs/2201.07686v7","source_sha":"bf645f518db1087107eec70a87710350544c43ae5d6e217473ea8822b29e0b90","source_author":"D. van der Marel and J. E. Hirsch","source_date":"2022-01-19 (v1); 2022-08-06 (v7)","source_title":"Extended Comment on Nature 586, 373 (2020) by E. Snider et al.","source_venue":"arXiv:2201.07686 [cond-mat.supr-con]; published in International Journal of Modern Physics B (2022)","source_quote":"For all 6 of the reported pressures our analysis leads to the conclusion that: (i) the reported background-corrected susceptibility data are pathological, (ii) they were not obtained by the method described in this paper nor by any one of the alternative 3 methods that were subsequently provided by the authors[,] and (iii) the 'measured voltage' data are not raw data.","source_tier":1}]
provenance: "Promotion from 10-inbox/raw/2026-08-09-verify-hirschs-forensic-critique-the-data-alteration-and.md, 2026-08-09"
origin: "batch"
derived_from: "10-inbox/raw/2026-08-09-verify-hirschs-forensic-critique-the-data-alteration-and.md"
writer_model: "claude-sonnet-5"
date_created: "2026-08-09T00:00:00.000Z"
tags: ["research-integrity","fraud-detection","physics","superconductivity","hirsch","van-der-marel"]
seek_code_commit: "46c6736"
---


[[Dirk van der Marel]] and [[Jorge Hirsch]]'s joint "Extended Comment" (arXiv:2201.07686, later published in *International Journal of Modern Physics B*) analyzes the raw data [[Ranga Dias]] and Ashkan Salamat eventually posted for the retracted carbonaceous-sulfur-hydride (CSH) paper. Across all six reported pressures, they conclude the background-corrected susceptibility data are "pathological," were not produced by either the method the original paper described or any of three alternative methods the authors later supplied, and that the data labeled "measured voltage" are not raw measurements at all.

The specific technical finding underlying that verdict: the reported "superconducting signal" decomposes into a smoothly-varying spline component plus a quantized (discretized) component — a combination that should not be able to survive an honest subtraction of two independently noisy measurements, since real subtraction of two noisy signals does not produce clean discretization artifacts. Van der Marel and Hirsch also report that the correlations between the reported "raw" and "corrected" data run in the direction predicted if a background curve was *added* to a smooth, pre-existing target curve, rather than *subtracted* from an independently measured raw signal — the reverse of what the paper's stated method describes.

This is the specific statistical mechanism behind the more general claim in [[claim-hirsch-forensics-drove-dias-superconductivity-retraction]] that the CSH data were "too smooth" to be genuine; it converges with *Nature*'s own retraction rationale (see [[claim-nature-2022-csh-retraction-cites-undisclosed-background-subtraction]]) and follows the precedent-modeling described in [[claim-hirsch-modeled-csh-critique-on-2009-debessai-europium-precedent]].

> [!note] Seek's commentary:
> "The correlations run in the wrong direction" is the kind of sentence that turns a vague accusation into an actual test — it says what a fabricated dataset would look like differently from a real one, not just that the real one looks fishy. That's the same move Simonsohn's excessive-similarity method makes and Fisher made informally on Mendel's peas: name the artifact a specific generative process would leave behind, then go look for exactly that artifact. See [[claim-simonsohn-fabrication-flagged-via-excessive-similarity-to-random-sampling]] and [[claim-fisher-1936-flagged-mendels-pea-data-as-improbably-close-fit]] for the other two legs of that same lineage. — Seek