Forensic Review of "Three-dimensional monolithic micro-LED display driven by atomically thin transistor matrix" (Nature Nanotechnology, 2021)

Academic fraud report · Geng Detector

Summary

This review evaluates the 2021 Nature Nanotechnology paper by Wanqing Meng, Feifan Xu, Zhihao Yu, Tao Tao and colleagues (Nanjing University, Xiamen University, and collaborators) on a monolithic 3D micro-LED display driven by an MoS2 transistor matrix. Based exclusively on the supplied text, the verdict is that no evidence of fraud or data manipulation was detected. Statistical figures, such as a Raman peak difference of 19.67 ± 0.47 cm⁻¹, photoluminescence peak of 1.865 eV (s.d. 1 meV), and FWHM 72 meV (s.d. 0.7 meV), are internally consistent and reflect realistic instrumental precision. Reported mobility distributions (median 17.9, FWHM 14.2 for the conventional process; median 54, FWHM 6.6 for the new process) and a 95% batch yield are plausibly imperfect. Percentage claims (475%, 60%) verify arithmetically. Limitations: only a text source was available, so pixel-level image reuse (Figure 1–4) and splicing checks (Figures 1d, 1e, 2a) could not be performed.

Verdict

Cleared based on the supplied textual record. No indicators of fraud, image manipulation, statistical fabrication, or methodological inconsistency were identified. The paper is judged credible within the limits of a text-only review.

Key findings

Internal statistical values are realistic and non-integer, consistent with genuine measurement noise (e.g., 19.67 ± 0.47 cm⁻¹ Raman shift; 1.865 eV PL peak with s.d. = 1 meV).
Reported device uniformity metrics show realistic scatter rather than idealized distributions; the conventional process exhibits large dispersion (median 17.9, FWHM 14.2) that approximates the median itself.
Quoted percentage improvements (475% and 60%) match the underlying numerical claims: (11.5 − 2)/2 = 4.75 and (775 − 310)/775 ≈ 0.60.
Sample sizes are large (200 devices per group, 400 total), Gaussian-fitted, and accompanied by dual-sweep hysteresis checks.
A 95% single-batch yield is reported rather than 99%/100%, consistent with realistic micro-fabrication outcomes.
References include a sister publication by the same team (DOI ending 963-8) in the same 2021 Nature Nanotechnology issue on MoS2 single-crystal growth on sapphire, indicating sustained systematic work.
Methods detail real, commercially available reagents and instruments (e.g., Futurrex IC1-1000 SOG, Transense TFA Au etchant, Asylum Cypher S AFM, FEI Helios 600i FIB, FEI Titan 80-300 TEM, Keithley 4200, Xilinx Spartan-6 FPGA), all available before the 2021 submission.
Authors openly discuss Au incompatibility with CMOS FEOL and propose a Cu/Au bilayer workaround in the supplement, a transparency pattern atypical of fabricated work.

Evidence highlights

Raman peak separation: 19.67 ± 0.47 cm⁻¹
PL statistics: peak 1.865 eV (s.d. 1 meV); FWHM 72 meV (s.d. 0.7 meV)
Conventional process mobility distribution: median 17.9, FWHM 14.2
New process mobility distribution: median 54, FWHM 6.6
Reported yield improvement from 2 to 11.5 (475%); contrast ratio shift from 775 to 310 (60%)
Sample size: 200 devices per group
Single-batch yield: 95%
DOI: 10.1038/s41565-021-00966-5
Sister-paper DOI ending: 963-8 (same 2021 issue, same team)

Notes

Scope limitation: the review relied solely on the supplied text; high-resolution figures (Figures 1–4) and specific panels (1d, 1e, 2a) were not accessible for pixel-level reuse or splicing analysis. Findings for image-related categories should be treated as inconclusive rather than exonerating.
No contact with authors was made, and no PubPeer posting is recommended on the basis of this text-only review.
The original Chinese reviewer concurs that this paper is suitable as a methodological reference for 2D-material and micro-LED research.
Confidence: moderate-to-high for text-derived checks; low-to-none for image-derived checks.
DOI preserved exactly as supplied: 10.1038/s41565-021-00966-5.