Integrity review: "Perovskite computed tomography imager and three-dimensional reconstruction" (Nature Photonics, 2024)

Academic fraud report · Geng Detector

Summary

Verdict: No evidence of academic fraud or data fabrication was found in this paper. The Geng-style six-form audit was applied across image reuse, mathematical self-consistency, statistical behavior, equipment provenance, and timeline coherence, and the paper passed each check. The reported 91.4% absorption efficiency at 59.3 keV with a 980 μm absorber was independently verified using the authors' stated linear attenuation coefficient (25 cm⁻¹) and the Beer–Lambert relation 1 - e^(-\alpha d), yielding ~0.9138, i.e., 91.4%. Photocurrent mapping stability statistics (Mean = 0.67%, Median = 0.51%; Mean = 0.69%, Median = 0.52%) show a right-skewed distribution, consistent with genuine experimental noise rather than fabricated symmetric data. Equipment models (YXLON Y.TU/160 X-ray source, Hitachi SU8020 SEM, BMF S230 3D printer) and the submission-to-acceptance timeline (2023-07-26 to 2024-07-15) are internally consistent. One minor methodological imprecision is noted: using the 59.3 keV peak energy to approximate mean energy for sensitivity calculation, but it is explicitly disclosed in the Methods. Confidence is high within text-only analysis; pixel-level verification of original figures is not possible.

Verdict

✅ No fraud indicators detected. The paper appears methodologically and numerically sound at the textual level. One minor disclosed methodological simplification is noted but does not constitute misconduct.

Key findings

Mathematical self-consistency verified: Absorption efficiency 91.4% at 59.3 keV for d = 980 μm recomputes from the stated μ = 25 cm⁻¹ as 1 − e^(−25 × 0.098) ≈ 0.9138 (91.4%). No sign of fabricated constants or order-of-magnitude errors.
Plausible statistical signatures: Photocurrent mapping stability reports Mean > Median in both samples (0.67% vs 0.51%; 0.69% vs 0.52%), consistent with a right-skewed real-noise distribution rather than the suspiciously symmetric statistics typical of synthetic data.
Realistic equipment and reagent records: YXLON Y.TU/160, Hitachi SU8020, and BMF S230 (2 μm optical precision) are genuine models appropriate to the reported experiments. The 10 μm phantom-feature dimension is consistent with the printer's stated precision.
Timeline coherence: Submitted 2023-07-26, accepted 2024-07-15 (~12 months) — consistent with Nature Photonics' stringent review cycle. Cited 2023 Nat. Photonics references (Ref 26, 27) for GBAC composition and spray-coating are chronologically appropriate.
Textual contradiction scan (Figures 1–4): No mislabeled test conditions or control-group swaps detected at the caption level. Dark/0.63 nGy/2.31 nGy/6.72 nGy gradient in Figure 3g is internally consistent.
Disclosed methodological simplification: Peak energy (59.3 keV) used as a proxy for mean X-ray energy in theoretical sensitivity calculation; explicitly stated in Methods. Honest simplification, not misconduct.

Evidence highlights

DOI: 10.1038/s41566-024-01506-y
Journal: Nature Photonics, Volume 18, October 2024, pp. 1052–1058
Authors: Yuhong He, Jinmei Song, Mingbian Li, Kostiantyn Sakhatskyi, Weijun Li, Xiaopeng Feng, Bai Yang, Maksym Kovalenko & Haotong Wei
Verbatim numeric checks: 91.4% absorption at 59.3 keV; μ = 25 cm⁻¹; d = 980 μm → 0.9138
Stability stats: Mean = 0.67%, Median = 0.51%; Mean = 0.69%, Median = 0.52%

Notes

This is a text-only audit; pixel-level image-forensics (clone-stamp, splicing, noise inconsistencies) could not be performed without the original high-resolution figures.
The peak-vs-mean energy approximation may cause a small divergence between theoretical and measured sensitivity; this is a methodological caveat, not a fraud indicator, and is already disclosed.
DOI preserved exactly as provided.