Geng Academic Fraud Detection Report: "A Valuable and Low-Budget Process Scheme of Equivalized 1 nm Technology Node Based on 2D Materials" (Nano-Micro Letters, DOI: 10.1007/s40820-025-01702-7)

Academic fraud report · Geng Detector

Summary

This AI-assisted integrity screening report rates the paper "A Valuable and Low-Budget Process Scheme of Equivalized 1 nm Technology Node Based on 2D Materials" (Nano-Micro Letters 17:191, 2025; DOI: 10.1007/s40820-025-01702-7) as HIGHLY SUSPICIOUS overall, though the automated Bayesian synthesis yields only 49.3% posterior probability (95% CI [15.8%, 79.6%], BF = 18.49) and a low-risk composite rating. The strongest concern is Figure 1d optical microscope images showing near-identical device units with matching reflection points and background texture; forensic analysis detected 36 copy-move block pairs at offset (96,0) in img-000.jpg and 22 pairs at (0,32) in img-003.jpg, though the periodic device array and claimed uniformity provide a partial benign explanation requiring original files. Additional confirmed issues include: internal contradictions between text and captions (Section 2 claims Figure 2d shows a STEM cross-section while its actual caption describes VTC curves), mislabeling of circuit-level VDD as device-level "VDS = 0.65 V" in Figure 4c,d, and a logically impossible self-comparison ("3 nm(2D+) offers a 28% improvement over 3 nm(2D+)"). The "1 nm node" claims rest entirely on TCAD/BSIM-CMG simulations from ~10 μm channel devices, constituting potential overstatement rather than data fraud. Benford/digit tests, noise-variance flags, and AI-text signals were excluded as unreliable. Final determination requires institutional investigation and original data.

Verdict

HIGHLY SUSPICIOUS (🟠) — Manual review flags this paper as highly suspicious, primarily due to potentially duplicated optical microscope imagery in Figure 1d and systematic text–figure mismatches. However, the automated Bayesian evidence synthesis yields a low-risk composite rating with a posterior probability of 49.3% (95% CI [15.8%, 79.6%], BF = 18.49, strong). No finding currently constitutes a basis for formal misconduct allegations; the paper is not suitable for institutional reporting at this stage, though author clarification and journal inquiry are warranted.

Key findings

Finding 1 (🟠, needs verification): Figure 1d optical microscope images of MoS₂ NSFET device arrays show nearly identical units in morphology, light reflection points, and background texture. Machine forensics detected 36 copy-move matched block pairs at offset (96,0) in img-000.jpg and 22 pairs at (0,32) in img-003.jpg. The periodic nature of device arrays and the paper's stated "uniformity and scalability" claim offer a partial benign explanation, but do not account for identical reflection points and background texture. Original image files required.
Finding 2 (🟡, insufficient basis): Internal contradiction — Section 2 states "Figure 2d shows a cross section of a MoS2 NSFET device under high-resolution transmission electron microscope (STEM)", while the Figure 1d caption mentions the cross-sectional STEM view. The Methods section omits cross-section sample preparation (e.g., FIB) and instrument details entirely.
Finding 3 (🟡, insufficient basis): All 1 nm node / 6 nm LG performance claims (36% / 28% / 48% frequency improvements; ION of 1.291 and 1.382 mA μm⁻¹) derive from TCAD + BSIM-CMG simulations; the fabricated device has an actual channel length of ~10 μm. The abstract's claim that the work "verifies the feasibility" is arguably overstated, though the Conclusion acknowledges this limitation. Not evidence of data fraud.
Finding 4 (🟡, confirmed): Section 3 states "the 3 nm(2D+) offers a 28% improvement over 3 nm(2D+)" — a logically impossible self-comparison; the abstract and Highlights use "36%" and "28%" in easily confused ways. Likely a revision slip, but indicates quality-control problems on a core selling point.
Finding 5 (🟡, confirmed): Figure 4c,d caption uses device-level "VDS = 0.65 V" where circuit-level VDD is appropriate. More seriously, the Section 2 text reference to Figure 2d (STEM cross-section) completely mismatches the actual Figure 2d caption ("Voltage transfer characteristics (VTC) of the three different devices based inverter") — possible systematic figure–text misalignment from revision.
Excluded clues: Benford/last-digit tests inapplicable to simulation/grid data; chart-element copy-move matches [I4][I6] and single weak noise-variance flags [I1][I3][I5][I7][I9] not individually probative; AI-text score 0.35/1.0 too weak to credit.

Evidence highlights

[I2] img-000.jpg: copy-move detection, 36 matched block pairs at offset (96,0) (logLR = 1.46; 51% of evidence contribution)
[I8] img-003.jpg: copy-move detection, 22 matched block pairs at offset (0,32) (logLR = 1.46; 49% of evidence contribution)
Verbatim text confirmed by manual comparison: "Figure 1d shows optical microscope images of MoS2 NSFETs and device arrays, confirming the high uniformity and scalability of the process"; "the 3 nm(2D+) offers a 28% improvement over 3 nm(2D+)"; "Figure 2d shows a cross section of a MoS2 NSFET device under high-resolution transmission electron microscope (STEM)"; "the actual channel length is about 10 μm"; "at VDS = 0.65 V"; Figure 2d caption "Voltage transfer characteristics (VTC) of the three different devices based inverter".

Notes

DOI: 10.1007/s40820-025-01702-7; Nano-Micro Letters (2025) 17:191; Received 18 Nov 2024, Accepted 18 Feb 2025, published online 18 Mar 2025. Authors: Yang Shen, Zhejia Zhang, Zhujun Yao, Mengge Jin, Jintian Gao, Yuhan Zhao, Wenzhong Bao, Yabin Sun, He Tian.
Recommended actions: request original Figure 1d image files and STEM cross-section raw data/preparation records from authors; ask authors to clarify the Figure 2d text/caption mismatch; correct the self-comparison and 28%/36% inconsistencies; consider raising questions on PubPeer depending on author response; not currently a basis for institutional reporting.
Report generated with AI assistance for academic discussion only; final misconduct determinations require professional institutional investigation. False positives and negatives are possible; the Bayesian synthesis assumes conditional independence of evidence and may overstate joint strength (prior fixed at 5%, LR version geng_lr_v3, n=11).