Geng Report: Sand-Based Thermal Storage System for Human-Powered Energy Generation: A Review — Suspected AI/Contract Fabrication

Academic fraud report · Geng Detector

Summary

This report examines a 2025 review paper published in Energies (MDPI; DOI: 10.3390/en18225869) by Qirui Ding, Lili Zeng, Ying Zeng, Changhui Song, Liang Lei, and Weicheng Cui. The Geng detector verdict is 'confirmed misconduct' (🔴). The core evidence is a cross-domain image mismatch: Figure 2's caption describes TDTR spectroscopy measurements of h-GST (Ge-Sb-Te) across amorphous, cubic, and hexagonal phases — a Phase Change Memory (PCM) microelectronics material — embedded inside a review of Phase Change Materials (PCM) for thermal storage such as paraffin and hydrated salts. The shared acronym PCM almost certainly caused an AI hallucination or copy-paste error. Supporting findings include systematic layout failures (multiple missing 'Figure' numbering tokens throughout the manuscript), a draft instruction residue ('Data available: [38–41,82,83,96]') in Figure 6's caption, and an abruptly truncated Conclusion that ends mid-sentence. Two further claims (logical contradiction in efficiency calculations; truncation of the conclusion) are flagged as 'insufficient basis'. Detection confidence is high for the image-misuse and formatting findings but the report itself acknowledges possible false positives.

Verdict

🔴 Confirmed (实锤) — The paper shows strong indicators of AI-assisted or contract-generated fabrication, most clearly demonstrated by cross-domain image reuse between microelectronics Phase Change Memory and thermal-storage Phase Change Materials.

Key findings

Cross-domain image misuse (AI-hallucination/copy-paste signature). Figure 2 and its caption refer to TDTR spectroscopy of h-GST (Ge-Sb-Te) across amorphous (a-GST), cubic (c-GST), and hexagonal (h-GST) phases. GST is a Phase Change Memory (PCM) material used in non-volatile memory devices; the paper's subject is Phase Change Materials for thermal storage (paraffin, hydrated salts, etc.). The two fields share only the English acronym "PCM," suggesting automated splicing rather than expert authorship. Severity: 🔴. Re-verified: ✅ valid.
Systematic typesetting collapse. Multiple cross-references in the body text are missing figure numbers: "Figurea illustrates", "(Figure)", "(Figurea)", and other unrendered tokens appear across Pages 4, 7, 8, 11, 17, and 20. Such placeholders are characteristic of unedited automated/AI-generated drafts exported directly to a submission template. Severity: 🟠. Re-verified: ✅ valid.
Draft-instruction residue in Figure 6 caption. The caption ends with the raw string "Data available: [38–41,82,83,96].", which resembles a working note used during literature compilation rather than a legitimate data-availability statement. Severity: 🟠. Re-verified: ✅ valid.
Self-contradictory efficiency reasoning (Section 6.2.1, Equations 6–8). The model yields ~6.3% overall efficiency (mechanical 270 kJ, friction 90%, sand storage 82%, thermoelectric 10%), yet the text claims it falls within an "8–12% range when accounting for optimized thermal management." A 10% thermoelectric ceiling cannot be exceeded by upstream thermal losses, so the conclusion is inconsistent with the stated equations. Severity: 🟠. Re-verified: ⚠️ insufficient basis.
Abrupt truncation of the Conclusion (Section 7, Page 33). The final sentence ends mid-phrase: "In ma-terials, surface, and interface engineering—based on carbon atom vibration". A published paper terminating without sentence closure is a strong indicator of unfinished or spliced drafting. Severity: 🔴. Re-verified: ⚠️ insufficient basis.

Evidence highlights

Figure 2 caption references "h-GST" with "5 nm tungsten contact measured by TDTR spectroscopy" across a-GST/c-GST/h-GST phases — terminology specific to Phase Change Memory chips, not thermal-storage sand batteries.
DOI: 10.3390/en18225869 (Energies, MDPI, 2025).
Manuscript file referenced: energies-18-05869.pdf.
Affliation flagged for institutional notification: 西湖大学 (Westlake University).
Numeric example (Section 6.2.1): 270 kJ mechanical input × 0.90 (friction) × 0.82 (sand storage) × 0.10 (thermoelectric) ≈ 6.3% overall, vs. claimed 8–12%.

Notes

Two findings (efficiency self-contradiction; truncated conclusion) are marked ⚠️ insufficient basis by the original detector and should not be treated as independently conclusive.
The high-confidence finding is the GST/PCM cross-domain image swap, which is verifiable directly from the figure caption without specialized tooling.
Recommended actions listed in the source report include PubPeer post, contact with the journal editorial office (MDPI), and institutional notification to Westlake University.
This translation is provided for academic discussion; final determination of misconduct requires investigation by the journal and/or the authors' institution.