Integrity review report: 'Cation-π interactions enabled water-stable perovskite X-ray flat mini-panel imager' (Nature Communications, DOI: 10.1038/s41467-023-44644-7)

Academic fraud report · Geng Detector

Summary

This AI-assisted integrity review evaluates a 2024 Nature Communications paper describing a water-stable perovskite X-ray detector using cation-π interactions. The overall verdict is SUSPECT (yellow), reflecting methodological and physics-consistency concerns rather than confirmed fraud. Four substantive issues were flagged: (1) the reported μτ product (1.8×10⁻⁴ cm² V⁻¹) combined with 800 V mm⁻¹ (8000 V cm⁻¹) yields an average drift length of 1.44 cm—over 10× the ~1.25 mm device thickness, which, while mathematically tidy, raises questions about whether stated gains and dark-current stability are physically realistic; (2) the photolithography methods describe an 'aqueous solution' positive photoresist and 0.5 wt% NaOH as a developer/stripper, contradicting standard semiconductor chemistry; (3) a 17.2 lp mm⁻¹ resolution is extrapolated via MTF computation from a test card limited to 10.0 lp mm⁻¹, an over-extrapolation prone to model bias; (4) image-level analyses could not be performed from text alone. No timeline, instrumentation, or CCDC anomalies were found. Confidence is moderate; original data and code are required for a definitive judgement.

Verdict

🟡 Suspect — The paper contains several methodologically and physically questionable claims. No definitive evidence of data fabrication was identified, but the cumulative plausibility issues warrant author clarification and independent verification.

Key findings

Suspiciously perfect arithmetic on drift length — Reported μτ of 1.8×10⁻⁴ cm² V⁻¹ under an applied field of 800 V mm⁻¹ (= 8000 V cm⁻¹) yields a drift length of 1.44 cm, which is approximately 11.5× the stated device thickness (~1.25 mm). Combined with the claimed ultra-low dark current and ~500× photoconductive gain at 1000 V bias, this strains typical perovskite-device physics.
Photolithography protocol inconsistent with standard chemistry — The Methods describe a positive photoresist cast from an 'aqueous solution,' developed, then stripped using 0.5 wt% NaOH ('Wet RIE'). Standard positive photoresists (e.g., AZ, S1800) are solvent-borne and developed in TMAH-based developers; the described sequence is chemically implausible as written.
Resolution claim exceeds the test target's physical limit — Figure 5h–k and the corresponding text report 17.2 lp mm⁻¹ by computational MTF extrapolation from a line-pair card whose nominal maximum is 10.0 lp mm⁻¹, which is methodologically unreliable and sensitive to noise-filter and ESF-fit choices.
No anomalies in timeline, equipment, or crystallographic data — Submission 2023-05-15, acceptance 2023-12-27; instruments cited (Rigaku SmartLab III, Hitachi Regulus8100, Bruker AVANCE III 600 MHz) were commercially available; CCDC 2224700 follows a standard 2022 registration format.
Image-level integrity not assessable from text alone — SEM (Figure 3e), XRD (Figure 1b/4a), and long-term stability curves (Figure 4h) require original files for splice/Crop/contrast-stretch analysis.

Evidence highlights

μτ × E calculation: 1.8×10⁻⁴ cm² V⁻¹ × 8000 V cm�¹ = 1.44 cm; device thickness ≈ 0.125 cm (≈1.25 mm); reported sensitivity 2.5×10⁶ μC Gy_air⁻¹ cm�² at 1000 V bias.
Methods wording: 'spin-coating positive photoresist (aqueous solution) ... UV exposure 12 s ... develop 8 s ... immerse in 0.5 wt% NaOH (aq) for 8 s to remove exposed photoresist (Wet RIE).'
Resolution: 10.0 lp mm⁻¹ test-card upper bound vs. 17.2 lp mm⁻¹ claimed via computational MTF.
DOI: 10.1038/s41467-023-44644-7; submission/acceptance dates 2023-05-15 / 2023-12-27.

Notes

Confidence: moderate. No direct image manipulation or duplicated data was detected; concerns are inferential and methodological.
Recommended actions: (1) request the photoresist product name and SDS, plus the actual developer composition; (2) request MTF source code, raw slanted-edge images, and ESF/LSF fits; (3) request dark-current stability data at 1000 V bias over the 17.2 lp mm⁻¹ exposure window; (4) consider a PubPeer post summarizing the methodological concerns.
Limitations: this report is text-based; pixel-level image forensics were not performed.
Disclaimer: AI-assisted; not an official finding of misconduct.